From 000725d56a196e72dc22328324c5ec5506265736 Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Thu, 8 Mar 2007 09:57:33 -0800 Subject: ioatdma: Push pending transactions to hardware more frequently Every 20 descriptors turns out to be to few append commands with newer/faster CPUs. Pushing every 4 still cuts down on MMIO writes to an acceptable level without letting the DMA engine run out of work. Signed-off-by: Chris Leech --- drivers/dma/ioatdma.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index 8500141..4e6afd7 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -310,7 +310,7 @@ static dma_cookie_t do_ioat_dma_memcpy(struct ioat_dma_chan *ioat_chan, list_splice_init(&new_chain, ioat_chan->used_desc.prev); ioat_chan->pending += desc_count; - if (ioat_chan->pending >= 20) { + if (ioat_chan->pending >= 4) { append = 1; ioat_chan->pending = 0; } @@ -818,7 +818,7 @@ static void __devexit ioat_remove(struct pci_dev *pdev) } /* MODULE API */ -MODULE_VERSION("1.7"); +MODULE_VERSION("1.9"); MODULE_LICENSE("GPL"); MODULE_AUTHOR("Intel Corporation"); -- cgit v1.1 From ff487fb773749124550a5ad2b7fbfe0376af6f0d Mon Sep 17 00:00:00 2001 From: Jeff Garzik Date: Thu, 8 Mar 2007 09:57:34 -0800 Subject: drivers/dma: handle sysfs errors From: Jeff Garzik Signed-off-by: Jeff Garzik Signed-off-by: Chris Leech --- drivers/dma/dmaengine.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c index 322ee29..828310d 100644 --- a/drivers/dma/dmaengine.c +++ b/drivers/dma/dmaengine.c @@ -316,7 +316,7 @@ EXPORT_SYMBOL(dma_async_client_chan_request); int dma_async_device_register(struct dma_device *device) { static int id; - int chancnt = 0; + int chancnt = 0, rc; struct dma_chan* chan; if (!device) @@ -338,8 +338,15 @@ int dma_async_device_register(struct dma_device *device) snprintf(chan->class_dev.class_id, BUS_ID_SIZE, "dma%dchan%d", device->dev_id, chan->chan_id); + rc = class_device_register(&chan->class_dev); + if (rc) { + chancnt--; + free_percpu(chan->local); + chan->local = NULL; + goto err_out; + } + kref_get(&device->refcount); - class_device_register(&chan->class_dev); } mutex_lock(&dma_list_mutex); @@ -349,6 +356,17 @@ int dma_async_device_register(struct dma_device *device) dma_chans_rebalance(); return 0; + +err_out: + list_for_each_entry(chan, &device->channels, device_node) { + if (chan->local == NULL) + continue; + kref_put(&device->refcount, dma_async_device_cleanup); + class_device_unregister(&chan->class_dev); + chancnt--; + free_percpu(chan->local); + } + return rc; } EXPORT_SYMBOL(dma_async_device_register); -- cgit v1.1 From e38288117c50fe22ed1693c2d8397245bb7e1a53 Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Thu, 8 Mar 2007 09:57:35 -0800 Subject: ioatdma: Remove the wrappers around read(bwl)/write(bwl) in ioatdma Signed-off-by: Chris Leech --- drivers/dma/ioatdma.c | 60 +++++++++++------------- drivers/dma/ioatdma_io.h | 118 ----------------------------------------------- 2 files changed, 28 insertions(+), 150 deletions(-) delete mode 100644 drivers/dma/ioatdma_io.h diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index 4e6afd7..2757070 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -32,7 +32,6 @@ #include #include #include "ioatdma.h" -#include "ioatdma_io.h" #include "ioatdma_registers.h" #include "ioatdma_hw.h" @@ -51,8 +50,8 @@ static int enumerate_dma_channels(struct ioat_device *device) int i; struct ioat_dma_chan *ioat_chan; - device->common.chancnt = ioatdma_read8(device, IOAT_CHANCNT_OFFSET); - xfercap_scale = ioatdma_read8(device, IOAT_XFERCAP_OFFSET); + device->common.chancnt = readb(device->reg_base + IOAT_CHANCNT_OFFSET); + xfercap_scale = readb(device->reg_base + IOAT_XFERCAP_OFFSET); xfercap = (xfercap_scale == 0 ? -1 : (1UL << xfercap_scale)); for (i = 0; i < device->common.chancnt; i++) { @@ -123,7 +122,7 @@ static int ioat_dma_alloc_chan_resources(struct dma_chan *chan) * In-use bit automatically set by reading chanctrl * If 0, we got it, if 1, someone else did */ - chanctrl = ioatdma_chan_read16(ioat_chan, IOAT_CHANCTRL_OFFSET); + chanctrl = readw(ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET); if (chanctrl & IOAT_CHANCTRL_CHANNEL_IN_USE) return -EBUSY; @@ -132,12 +131,12 @@ static int ioat_dma_alloc_chan_resources(struct dma_chan *chan) IOAT_CHANCTRL_ERR_INT_EN | IOAT_CHANCTRL_ANY_ERR_ABORT_EN | IOAT_CHANCTRL_ERR_COMPLETION_EN; - ioatdma_chan_write16(ioat_chan, IOAT_CHANCTRL_OFFSET, chanctrl); + writew(chanctrl, ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET); - chanerr = ioatdma_chan_read32(ioat_chan, IOAT_CHANERR_OFFSET); + chanerr = readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET); if (chanerr) { printk("IOAT: CHANERR = %x, clearing\n", chanerr); - ioatdma_chan_write32(ioat_chan, IOAT_CHANERR_OFFSET, chanerr); + writel(chanerr, ioat_chan->reg_base + IOAT_CHANERR_OFFSET); } /* Allocate descriptors */ @@ -161,10 +160,10 @@ static int ioat_dma_alloc_chan_resources(struct dma_chan *chan) &ioat_chan->completion_addr); memset(ioat_chan->completion_virt, 0, sizeof(*ioat_chan->completion_virt)); - ioatdma_chan_write32(ioat_chan, IOAT_CHANCMP_OFFSET_LOW, - ((u64) ioat_chan->completion_addr) & 0x00000000FFFFFFFF); - ioatdma_chan_write32(ioat_chan, IOAT_CHANCMP_OFFSET_HIGH, - ((u64) ioat_chan->completion_addr) >> 32); + writel(((u64) ioat_chan->completion_addr) & 0x00000000FFFFFFFF, + ioat_chan->reg_base + IOAT_CHANCMP_OFFSET_LOW); + writel(((u64) ioat_chan->completion_addr) >> 32, + ioat_chan->reg_base + IOAT_CHANCMP_OFFSET_HIGH); ioat_start_null_desc(ioat_chan); return i; @@ -182,7 +181,7 @@ static void ioat_dma_free_chan_resources(struct dma_chan *chan) ioat_dma_memcpy_cleanup(ioat_chan); - ioatdma_chan_write8(ioat_chan, IOAT_CHANCMD_OFFSET, IOAT_CHANCMD_RESET); + writeb(IOAT_CHANCMD_RESET, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); spin_lock_bh(&ioat_chan->desc_lock); list_for_each_entry_safe(desc, _desc, &ioat_chan->used_desc, node) { @@ -210,9 +209,9 @@ static void ioat_dma_free_chan_resources(struct dma_chan *chan) ioat_chan->last_completion = ioat_chan->completion_addr = 0; /* Tell hw the chan is free */ - chanctrl = ioatdma_chan_read16(ioat_chan, IOAT_CHANCTRL_OFFSET); + chanctrl = readw(ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET); chanctrl &= ~IOAT_CHANCTRL_CHANNEL_IN_USE; - ioatdma_chan_write16(ioat_chan, IOAT_CHANCTRL_OFFSET, chanctrl); + writew(chanctrl, ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET); } /** @@ -318,9 +317,8 @@ static dma_cookie_t do_ioat_dma_memcpy(struct ioat_dma_chan *ioat_chan, spin_unlock_bh(&ioat_chan->desc_lock); if (append) - ioatdma_chan_write8(ioat_chan, - IOAT_CHANCMD_OFFSET, - IOAT_CHANCMD_APPEND); + writeb(IOAT_CHANCMD_APPEND, + ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); return cookie; } @@ -417,9 +415,8 @@ static void ioat_dma_memcpy_issue_pending(struct dma_chan *chan) if (ioat_chan->pending != 0) { ioat_chan->pending = 0; - ioatdma_chan_write8(ioat_chan, - IOAT_CHANCMD_OFFSET, - IOAT_CHANCMD_APPEND); + writeb(IOAT_CHANCMD_APPEND, + ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); } } @@ -449,7 +446,7 @@ static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *chan) if ((chan->completion_virt->full & IOAT_CHANSTS_DMA_TRANSFER_STATUS) == IOAT_CHANSTS_DMA_TRANSFER_STATUS_HALTED) { printk("IOAT: Channel halted, chanerr = %x\n", - ioatdma_chan_read32(chan, IOAT_CHANERR_OFFSET)); + readl(chan->reg_base + IOAT_CHANERR_OFFSET)); /* TODO do something to salvage the situation */ } @@ -569,21 +566,21 @@ static irqreturn_t ioat_do_interrupt(int irq, void *data) unsigned long attnstatus; u8 intrctrl; - intrctrl = ioatdma_read8(instance, IOAT_INTRCTRL_OFFSET); + intrctrl = readb(instance->reg_base + IOAT_INTRCTRL_OFFSET); if (!(intrctrl & IOAT_INTRCTRL_MASTER_INT_EN)) return IRQ_NONE; if (!(intrctrl & IOAT_INTRCTRL_INT_STATUS)) { - ioatdma_write8(instance, IOAT_INTRCTRL_OFFSET, intrctrl); + writeb(intrctrl, instance->reg_base + IOAT_INTRCTRL_OFFSET); return IRQ_NONE; } - attnstatus = ioatdma_read32(instance, IOAT_ATTNSTATUS_OFFSET); + attnstatus = readl(instance->reg_base + IOAT_ATTNSTATUS_OFFSET); printk(KERN_ERR "ioatdma error: interrupt! status %lx\n", attnstatus); - ioatdma_write8(instance, IOAT_INTRCTRL_OFFSET, intrctrl); + writeb(intrctrl, instance->reg_base + IOAT_INTRCTRL_OFFSET); return IRQ_HANDLED; } @@ -612,14 +609,13 @@ static void ioat_start_null_desc(struct ioat_dma_chan *ioat_chan) spin_unlock_bh(&ioat_chan->desc_lock); #if (BITS_PER_LONG == 64) - ioatdma_chan_write64(ioat_chan, IOAT_CHAINADDR_OFFSET, desc->phys); + writeq(desc->phys, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET); #else - ioatdma_chan_write32(ioat_chan, - IOAT_CHAINADDR_OFFSET_LOW, - (u32) desc->phys); - ioatdma_chan_write32(ioat_chan, IOAT_CHAINADDR_OFFSET_HIGH, 0); + writel((u32) desc->phys, + ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_LOW); + writel(0, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_HIGH); #endif - ioatdma_chan_write8(ioat_chan, IOAT_CHANCMD_OFFSET, IOAT_CHANCMD_START); + writeb(IOAT_CHANCMD_START, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); } /* @@ -748,7 +744,7 @@ static int __devinit ioat_probe(struct pci_dev *pdev, device->reg_base = reg_base; - ioatdma_write8(device, IOAT_INTRCTRL_OFFSET, IOAT_INTRCTRL_MASTER_INT_EN); + writeb(IOAT_INTRCTRL_MASTER_INT_EN, device->reg_base + IOAT_INTRCTRL_OFFSET); pci_set_master(pdev); INIT_LIST_HEAD(&device->common.channels); diff --git a/drivers/dma/ioatdma_io.h b/drivers/dma/ioatdma_io.h deleted file mode 100644 index c0b4bf6..0000000 --- a/drivers/dma/ioatdma_io.h +++ /dev/null @@ -1,118 +0,0 @@ -/* - * Copyright(c) 2004 - 2006 Intel Corporation. All rights reserved. - * - * This program is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License as published by the Free - * Software Foundation; either version 2 of the License, or (at your option) - * any later version. - * - * This program is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for - * more details. - * - * You should have received a copy of the GNU General Public License along with - * this program; if not, write to the Free Software Foundation, Inc., 59 - * Temple Place - Suite 330, Boston, MA 02111-1307, USA. - * - * The full GNU General Public License is included in this distribution in the - * file called COPYING. - */ -#ifndef IOATDMA_IO_H -#define IOATDMA_IO_H - -#include - -/* - * device and per-channel MMIO register read and write functions - * this is a lot of anoying inline functions, but it's typesafe - */ - -static inline u8 ioatdma_read8(struct ioat_device *device, - unsigned int offset) -{ - return readb(device->reg_base + offset); -} - -static inline u16 ioatdma_read16(struct ioat_device *device, - unsigned int offset) -{ - return readw(device->reg_base + offset); -} - -static inline u32 ioatdma_read32(struct ioat_device *device, - unsigned int offset) -{ - return readl(device->reg_base + offset); -} - -static inline void ioatdma_write8(struct ioat_device *device, - unsigned int offset, u8 value) -{ - writeb(value, device->reg_base + offset); -} - -static inline void ioatdma_write16(struct ioat_device *device, - unsigned int offset, u16 value) -{ - writew(value, device->reg_base + offset); -} - -static inline void ioatdma_write32(struct ioat_device *device, - unsigned int offset, u32 value) -{ - writel(value, device->reg_base + offset); -} - -static inline u8 ioatdma_chan_read8(struct ioat_dma_chan *chan, - unsigned int offset) -{ - return readb(chan->reg_base + offset); -} - -static inline u16 ioatdma_chan_read16(struct ioat_dma_chan *chan, - unsigned int offset) -{ - return readw(chan->reg_base + offset); -} - -static inline u32 ioatdma_chan_read32(struct ioat_dma_chan *chan, - unsigned int offset) -{ - return readl(chan->reg_base + offset); -} - -static inline void ioatdma_chan_write8(struct ioat_dma_chan *chan, - unsigned int offset, u8 value) -{ - writeb(value, chan->reg_base + offset); -} - -static inline void ioatdma_chan_write16(struct ioat_dma_chan *chan, - unsigned int offset, u16 value) -{ - writew(value, chan->reg_base + offset); -} - -static inline void ioatdma_chan_write32(struct ioat_dma_chan *chan, - unsigned int offset, u32 value) -{ - writel(value, chan->reg_base + offset); -} - -#if (BITS_PER_LONG == 64) -static inline u64 ioatdma_chan_read64(struct ioat_dma_chan *chan, - unsigned int offset) -{ - return readq(chan->reg_base + offset); -} - -static inline void ioatdma_chan_write64(struct ioat_dma_chan *chan, - unsigned int offset, u64 value) -{ - writeq(value, chan->reg_base + offset); -} -#endif - -#endif /* IOATDMA_IO_H */ - -- cgit v1.1 From 70774b47392171faae0e98f795aa1507f9246af1 Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Thu, 8 Mar 2007 09:57:35 -0800 Subject: ioatdma: Remove the use of writeq from the ioatdma driver There's only one now anyway, and it's not in a performance path, so make it behave the same on 32-bit and 64-bit CPUs. Signed-off-by: Chris Leech --- drivers/dma/ioatdma.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index 2757070..b3fca12 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -608,13 +608,11 @@ static void ioat_start_null_desc(struct ioat_dma_chan *ioat_chan) list_add_tail(&desc->node, &ioat_chan->used_desc); spin_unlock_bh(&ioat_chan->desc_lock); -#if (BITS_PER_LONG == 64) - writeq(desc->phys, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET); -#else - writel((u32) desc->phys, + writel(((u64) desc->phys) & 0x00000000FFFFFFFF, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_LOW); - writel(0, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_HIGH); -#endif + writel(((u64) desc->phys) >> 32, + ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_HIGH); + writeb(IOAT_CHANCMD_START, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); } -- cgit v1.1 From 72d0b7a81d60f5e64ee7197bc190b9b3265f99dd Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Thu, 8 Mar 2007 09:57:35 -0800 Subject: I/OAT: Add documentation for the tcp_dma_copybreak sysctl Signed-off-by: Chris Leech --- Documentation/networking/ip-sysctl.txt | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt index af6a63a..8f6067e 100644 --- a/Documentation/networking/ip-sysctl.txt +++ b/Documentation/networking/ip-sysctl.txt @@ -433,6 +433,12 @@ tcp_workaround_signed_windows - BOOLEAN not receive a window scaling option from them. Default: 0 +tcp_dma_copybreak - INTEGER + Lower limit, in bytes, of the size of socket reads that will be + offloaded to a DMA copy engine, if one is present in the system + and CONFIG_NET_DMA is enabled. + Default: 4096 + CIPSOv4 Variables: cipso_cache_enable - BOOLEAN -- cgit v1.1 From 2b1244a43be97f504494b557a7f7a65fe0d00dba Mon Sep 17 00:00:00 2001 From: Chris Leech Date: Thu, 8 Mar 2007 09:57:36 -0800 Subject: I/OAT: Only offload copies for TCP when there will be a context switch The performance wins come with having the DMA copy engine doing the copies in parallel with the context switch. If there is enough data ready on the socket at recv time just use a regular copy. Signed-off-by: Chris Leech --- net/ipv4/tcp.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 450f44b..0eb7cf1 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -1116,6 +1116,8 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, long timeo; struct task_struct *user_recv = NULL; int copied_early = 0; + int available = 0; + struct sk_buff *skb; lock_sock(sk); @@ -1142,7 +1144,11 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, #ifdef CONFIG_NET_DMA tp->ucopy.dma_chan = NULL; preempt_disable(); - if ((len > sysctl_tcp_dma_copybreak) && !(flags & MSG_PEEK) && + skb = skb_peek_tail(&sk->sk_receive_queue); + if (skb) + available = TCP_SKB_CB(skb)->seq + skb->len - (*seq); + if ((available < target) && + (len > sysctl_tcp_dma_copybreak) && !(flags & MSG_PEEK) && !sysctl_tcp_low_latency && __get_cpu_var(softnet_data).net_dma) { preempt_enable_no_resched(); tp->ucopy.pinned_list = dma_pin_iovec_pages(msg->msg_iov, len); @@ -1151,7 +1157,6 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, #endif do { - struct sk_buff *skb; u32 offset; /* Are we at urgent data? Stop if we have read anything or have SIGURG pending. */ @@ -1439,7 +1444,6 @@ skip_copy: #ifdef CONFIG_NET_DMA if (tp->ucopy.dma_chan) { - struct sk_buff *skb; dma_cookie_t done, used; dma_async_memcpy_issue_pending(tp->ucopy.dma_chan); -- cgit v1.1 From e00c5d8b4d800b95b72b3f072e1d55d7c7034702 Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Thu, 8 Mar 2007 09:57:36 -0800 Subject: I/OAT: warning fix net/ipv4/tcp.c: In function 'tcp_recvmsg': net/ipv4/tcp.c:1111: warning: unused variable 'available' Signed-off-by: Andrew Morton Signed-off-by: Chris Leech --- net/ipv4/tcp.c | 26 ++++++++++++++++---------- 1 file changed, 16 insertions(+), 10 deletions(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 0eb7cf1..987b944 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -1116,7 +1116,6 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, long timeo; struct task_struct *user_recv = NULL; int copied_early = 0; - int available = 0; struct sk_buff *skb; lock_sock(sk); @@ -1145,15 +1144,22 @@ int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, tp->ucopy.dma_chan = NULL; preempt_disable(); skb = skb_peek_tail(&sk->sk_receive_queue); - if (skb) - available = TCP_SKB_CB(skb)->seq + skb->len - (*seq); - if ((available < target) && - (len > sysctl_tcp_dma_copybreak) && !(flags & MSG_PEEK) && - !sysctl_tcp_low_latency && __get_cpu_var(softnet_data).net_dma) { - preempt_enable_no_resched(); - tp->ucopy.pinned_list = dma_pin_iovec_pages(msg->msg_iov, len); - } else - preempt_enable_no_resched(); + { + int available = 0; + + if (skb) + available = TCP_SKB_CB(skb)->seq + skb->len - (*seq); + if ((available < target) && + (len > sysctl_tcp_dma_copybreak) && !(flags & MSG_PEEK) && + !sysctl_tcp_low_latency && + __get_cpu_var(softnet_data).net_dma) { + preempt_enable_no_resched(); + tp->ucopy.pinned_list = + dma_pin_iovec_pages(msg->msg_iov, len); + } else { + preempt_enable_no_resched(); + } + } #endif do { -- cgit v1.1 From 428ed6024fa74a271142f3257966e9b5e1cb37a1 Mon Sep 17 00:00:00 2001 From: Dan Aloni Date: Thu, 8 Mar 2007 09:57:36 -0800 Subject: I/OAT: fix I/OAT for kexec Under kexec, I/OAT initialization breaks over busy resources because the previous kernel did not release them. I'm not sure this fix can be considered a complete one but it works for me. I guess something similar to the *_remove method should occur there.. Signed-off-by: Dan Aloni Signed-off-by: Chris Leech Signed-off-by: Andrew Morton --- drivers/dma/ioatdma.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index b3fca12..c4209af 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -41,6 +41,7 @@ /* internal functions */ static int __devinit ioat_probe(struct pci_dev *pdev, const struct pci_device_id *ent); +static void ioat_shutdown(struct pci_dev *pdev); static void __devexit ioat_remove(struct pci_dev *pdev); static int enumerate_dma_channels(struct ioat_device *device) @@ -557,6 +558,7 @@ static struct pci_driver ioat_pci_driver = { .name = "ioatdma", .id_table = ioat_pci_tbl, .probe = ioat_probe, + .shutdown = ioat_shutdown, .remove = __devexit_p(ioat_remove), }; @@ -781,9 +783,20 @@ err_request_regions: err_set_dma_mask: pci_disable_device(pdev); err_enable_device: + + printk(KERN_ERR "Intel(R) I/OAT DMA Engine initialization failed\n"); + return err; } +static void ioat_shutdown(struct pci_dev *pdev) +{ + struct ioat_device *device; + device = pci_get_drvdata(pdev); + + dma_async_device_unregister(&device->common); +} + static void __devexit ioat_remove(struct pci_dev *pdev) { struct ioat_device *device; -- cgit v1.1 From 7405f74badf46b5d023c5d2b670b4471525f6c91 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 11:10:43 -0700 Subject: dmaengine: refactor dmaengine around dma_async_tx_descriptor The current dmaengine interface defines mutliple routines per operation, i.e. dma_async_memcpy_buf_to_buf, dma_async_memcpy_buf_to_page etc. Adding more operation types (xor, crc, etc) to this model would result in an unmanageable number of method permutations. Are we really going to add a set of hooks for each DMA engine whizbang feature? - Jeff Garzik The descriptor creation process is refactored using the new common dma_async_tx_descriptor structure. Instead of per driver do___to_ methods, drivers integrate dma_async_tx_descriptor into their private software descriptor and then define a 'prep' routine per operation. The prep routine allocates a descriptor and ensures that the tx_set_src, tx_set_dest, tx_submit routines are valid. Descriptor creation and submission becomes: struct dma_device *dev; struct dma_chan *chan; struct dma_async_tx_descriptor *tx; tx = dev->device_prep_dma_(chan, len, int_flag) tx->tx_set_src(dma_addr_t, tx, index /* for multi-source ops */) tx->tx_set_dest(dma_addr_t, tx, index) tx->tx_submit(tx) In addition to the refactoring, dma_async_tx_descriptor also lays the groundwork for definining cross-channel-operation dependencies, and a callback facility for asynchronous notification of operation completion. Changelog: * drop dma mapping methods, suggested by Chris Leech * fix ioat_dma_dependency_added, also caught by Andrew Morton * fix dma_sync_wait, change from Andrew Morton * uninline large functions, change from Andrew Morton * add tx->callback = NULL to dmaengine calls to interoperate with async_tx calls * hookup ioat_tx_submit * convert channel capabilities to a 'cpumask_t like' bitmap * removed DMA_TX_ARRAY_INIT, no longer needed * checkpatch.pl fixes * make set_src, set_dest, and tx_submit descriptor specific methods * fixup git-ioat merge * move group_list and phys to dma_async_tx_descriptor Cc: Jeff Garzik Cc: Chris Leech Signed-off-by: Shannon Nelson Signed-off-by: Dan Williams Acked-by: David S. Miller --- drivers/dma/dmaengine.c | 182 ++++++++++++++++++++++++++++ drivers/dma/ioatdma.c | 295 +++++++++++++++++++++------------------------- drivers/dma/ioatdma.h | 13 +- include/linux/dmaengine.h | 237 +++++++++++++++++++++++-------------- 4 files changed, 474 insertions(+), 253 deletions(-) diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c index 828310d..404cc7b 100644 --- a/drivers/dma/dmaengine.c +++ b/drivers/dma/dmaengine.c @@ -59,6 +59,7 @@ #include #include +#include #include #include #include @@ -66,6 +67,7 @@ #include #include #include +#include static DEFINE_MUTEX(dma_list_mutex); static LIST_HEAD(dma_device_list); @@ -165,6 +167,24 @@ static struct dma_chan *dma_client_chan_alloc(struct dma_client *client) return NULL; } +enum dma_status dma_sync_wait(struct dma_chan *chan, dma_cookie_t cookie) +{ + enum dma_status status; + unsigned long dma_sync_wait_timeout = jiffies + msecs_to_jiffies(5000); + + dma_async_issue_pending(chan); + do { + status = dma_async_is_tx_complete(chan, cookie, NULL, NULL); + if (time_after_eq(jiffies, dma_sync_wait_timeout)) { + printk(KERN_ERR "dma_sync_wait_timeout!\n"); + return DMA_ERROR; + } + } while (status == DMA_IN_PROGRESS); + + return status; +} +EXPORT_SYMBOL(dma_sync_wait); + /** * dma_chan_cleanup - release a DMA channel's resources * @kref: kernel reference structure that contains the DMA channel device @@ -322,6 +342,25 @@ int dma_async_device_register(struct dma_device *device) if (!device) return -ENODEV; + /* validate device routines */ + BUG_ON(dma_has_cap(DMA_MEMCPY, device->cap_mask) && + !device->device_prep_dma_memcpy); + BUG_ON(dma_has_cap(DMA_XOR, device->cap_mask) && + !device->device_prep_dma_xor); + BUG_ON(dma_has_cap(DMA_ZERO_SUM, device->cap_mask) && + !device->device_prep_dma_zero_sum); + BUG_ON(dma_has_cap(DMA_MEMSET, device->cap_mask) && + !device->device_prep_dma_memset); + BUG_ON(dma_has_cap(DMA_ZERO_SUM, device->cap_mask) && + !device->device_prep_dma_interrupt); + + BUG_ON(!device->device_alloc_chan_resources); + BUG_ON(!device->device_free_chan_resources); + BUG_ON(!device->device_dependency_added); + BUG_ON(!device->device_is_tx_complete); + BUG_ON(!device->device_issue_pending); + BUG_ON(!device->dev); + init_completion(&device->done); kref_init(&device->refcount); device->dev_id = id++; @@ -415,6 +454,149 @@ void dma_async_device_unregister(struct dma_device *device) } EXPORT_SYMBOL(dma_async_device_unregister); +/** + * dma_async_memcpy_buf_to_buf - offloaded copy between virtual addresses + * @chan: DMA channel to offload copy to + * @dest: destination address (virtual) + * @src: source address (virtual) + * @len: length + * + * Both @dest and @src must be mappable to a bus address according to the + * DMA mapping API rules for streaming mappings. + * Both @dest and @src must stay memory resident (kernel memory or locked + * user space pages). + */ +dma_cookie_t +dma_async_memcpy_buf_to_buf(struct dma_chan *chan, void *dest, + void *src, size_t len) +{ + struct dma_device *dev = chan->device; + struct dma_async_tx_descriptor *tx; + dma_addr_t addr; + dma_cookie_t cookie; + int cpu; + + tx = dev->device_prep_dma_memcpy(chan, len, 0); + if (!tx) + return -ENOMEM; + + tx->ack = 1; + tx->callback = NULL; + addr = dma_map_single(dev->dev, src, len, DMA_TO_DEVICE); + tx->tx_set_src(addr, tx, 0); + addr = dma_map_single(dev->dev, dest, len, DMA_FROM_DEVICE); + tx->tx_set_dest(addr, tx, 0); + cookie = tx->tx_submit(tx); + + cpu = get_cpu(); + per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; + per_cpu_ptr(chan->local, cpu)->memcpy_count++; + put_cpu(); + + return cookie; +} +EXPORT_SYMBOL(dma_async_memcpy_buf_to_buf); + +/** + * dma_async_memcpy_buf_to_pg - offloaded copy from address to page + * @chan: DMA channel to offload copy to + * @page: destination page + * @offset: offset in page to copy to + * @kdata: source address (virtual) + * @len: length + * + * Both @page/@offset and @kdata must be mappable to a bus address according + * to the DMA mapping API rules for streaming mappings. + * Both @page/@offset and @kdata must stay memory resident (kernel memory or + * locked user space pages) + */ +dma_cookie_t +dma_async_memcpy_buf_to_pg(struct dma_chan *chan, struct page *page, + unsigned int offset, void *kdata, size_t len) +{ + struct dma_device *dev = chan->device; + struct dma_async_tx_descriptor *tx; + dma_addr_t addr; + dma_cookie_t cookie; + int cpu; + + tx = dev->device_prep_dma_memcpy(chan, len, 0); + if (!tx) + return -ENOMEM; + + tx->ack = 1; + tx->callback = NULL; + addr = dma_map_single(dev->dev, kdata, len, DMA_TO_DEVICE); + tx->tx_set_src(addr, tx, 0); + addr = dma_map_page(dev->dev, page, offset, len, DMA_FROM_DEVICE); + tx->tx_set_dest(addr, tx, 0); + cookie = tx->tx_submit(tx); + + cpu = get_cpu(); + per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; + per_cpu_ptr(chan->local, cpu)->memcpy_count++; + put_cpu(); + + return cookie; +} +EXPORT_SYMBOL(dma_async_memcpy_buf_to_pg); + +/** + * dma_async_memcpy_pg_to_pg - offloaded copy from page to page + * @chan: DMA channel to offload copy to + * @dest_pg: destination page + * @dest_off: offset in page to copy to + * @src_pg: source page + * @src_off: offset in page to copy from + * @len: length + * + * Both @dest_page/@dest_off and @src_page/@src_off must be mappable to a bus + * address according to the DMA mapping API rules for streaming mappings. + * Both @dest_page/@dest_off and @src_page/@src_off must stay memory resident + * (kernel memory or locked user space pages). + */ +dma_cookie_t +dma_async_memcpy_pg_to_pg(struct dma_chan *chan, struct page *dest_pg, + unsigned int dest_off, struct page *src_pg, unsigned int src_off, + size_t len) +{ + struct dma_device *dev = chan->device; + struct dma_async_tx_descriptor *tx; + dma_addr_t addr; + dma_cookie_t cookie; + int cpu; + + tx = dev->device_prep_dma_memcpy(chan, len, 0); + if (!tx) + return -ENOMEM; + + tx->ack = 1; + tx->callback = NULL; + addr = dma_map_page(dev->dev, src_pg, src_off, len, DMA_TO_DEVICE); + tx->tx_set_src(addr, tx, 0); + addr = dma_map_page(dev->dev, dest_pg, dest_off, len, DMA_FROM_DEVICE); + tx->tx_set_dest(addr, tx, 0); + cookie = tx->tx_submit(tx); + + cpu = get_cpu(); + per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; + per_cpu_ptr(chan->local, cpu)->memcpy_count++; + put_cpu(); + + return cookie; +} +EXPORT_SYMBOL(dma_async_memcpy_pg_to_pg); + +void dma_async_tx_descriptor_init(struct dma_async_tx_descriptor *tx, + struct dma_chan *chan) +{ + tx->chan = chan; + spin_lock_init(&tx->lock); + INIT_LIST_HEAD(&tx->depend_node); + INIT_LIST_HEAD(&tx->depend_list); +} +EXPORT_SYMBOL(dma_async_tx_descriptor_init); + static int __init dma_bus_init(void) { mutex_init(&dma_list_mutex); diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index c4209af..1710449 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -38,6 +38,7 @@ #define to_ioat_chan(chan) container_of(chan, struct ioat_dma_chan, common) #define to_ioat_device(dev) container_of(dev, struct ioat_device, common) #define to_ioat_desc(lh) container_of(lh, struct ioat_desc_sw, node) +#define tx_to_ioat_desc(tx) container_of(tx, struct ioat_desc_sw, async_tx) /* internal functions */ static int __devinit ioat_probe(struct pci_dev *pdev, const struct pci_device_id *ent); @@ -78,6 +79,73 @@ static int enumerate_dma_channels(struct ioat_device *device) return device->common.chancnt; } +static void +ioat_set_src(dma_addr_t addr, struct dma_async_tx_descriptor *tx, int index) +{ + struct ioat_desc_sw *iter, *desc = tx_to_ioat_desc(tx); + struct ioat_dma_chan *ioat_chan = to_ioat_chan(tx->chan); + + pci_unmap_addr_set(desc, src, addr); + + list_for_each_entry(iter, &desc->async_tx.tx_list, node) { + iter->hw->src_addr = addr; + addr += ioat_chan->xfercap; + } + +} + +static void +ioat_set_dest(dma_addr_t addr, struct dma_async_tx_descriptor *tx, int index) +{ + struct ioat_desc_sw *iter, *desc = tx_to_ioat_desc(tx); + struct ioat_dma_chan *ioat_chan = to_ioat_chan(tx->chan); + + pci_unmap_addr_set(desc, dst, addr); + + list_for_each_entry(iter, &desc->async_tx.tx_list, node) { + iter->hw->dst_addr = addr; + addr += ioat_chan->xfercap; + } +} + +static dma_cookie_t +ioat_tx_submit(struct dma_async_tx_descriptor *tx) +{ + struct ioat_dma_chan *ioat_chan = to_ioat_chan(tx->chan); + struct ioat_desc_sw *desc = tx_to_ioat_desc(tx); + int append = 0; + dma_cookie_t cookie; + struct ioat_desc_sw *group_start; + + group_start = list_entry(desc->async_tx.tx_list.next, + struct ioat_desc_sw, node); + spin_lock_bh(&ioat_chan->desc_lock); + /* cookie incr and addition to used_list must be atomic */ + cookie = ioat_chan->common.cookie; + cookie++; + if (cookie < 0) + cookie = 1; + ioat_chan->common.cookie = desc->async_tx.cookie = cookie; + + /* write address into NextDescriptor field of last desc in chain */ + to_ioat_desc(ioat_chan->used_desc.prev)->hw->next = + group_start->async_tx.phys; + list_splice_init(&desc->async_tx.tx_list, ioat_chan->used_desc.prev); + + ioat_chan->pending += desc->tx_cnt; + if (ioat_chan->pending >= 4) { + append = 1; + ioat_chan->pending = 0; + } + spin_unlock_bh(&ioat_chan->desc_lock); + + if (append) + writeb(IOAT_CHANCMD_APPEND, + ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); + + return cookie; +} + static struct ioat_desc_sw *ioat_dma_alloc_descriptor( struct ioat_dma_chan *ioat_chan, gfp_t flags) @@ -99,8 +167,13 @@ static struct ioat_desc_sw *ioat_dma_alloc_descriptor( } memset(desc, 0, sizeof(*desc)); + dma_async_tx_descriptor_init(&desc_sw->async_tx, &ioat_chan->common); + desc_sw->async_tx.tx_set_src = ioat_set_src; + desc_sw->async_tx.tx_set_dest = ioat_set_dest; + desc_sw->async_tx.tx_submit = ioat_tx_submit; + INIT_LIST_HEAD(&desc_sw->async_tx.tx_list); desc_sw->hw = desc; - desc_sw->phys = phys; + desc_sw->async_tx.phys = phys; return desc_sw; } @@ -188,12 +261,14 @@ static void ioat_dma_free_chan_resources(struct dma_chan *chan) list_for_each_entry_safe(desc, _desc, &ioat_chan->used_desc, node) { in_use_descs++; list_del(&desc->node); - pci_pool_free(ioat_device->dma_pool, desc->hw, desc->phys); + pci_pool_free(ioat_device->dma_pool, desc->hw, + desc->async_tx.phys); kfree(desc); } list_for_each_entry_safe(desc, _desc, &ioat_chan->free_desc, node) { list_del(&desc->node); - pci_pool_free(ioat_device->dma_pool, desc->hw, desc->phys); + pci_pool_free(ioat_device->dma_pool, desc->hw, + desc->async_tx.phys); kfree(desc); } spin_unlock_bh(&ioat_chan->desc_lock); @@ -215,45 +290,25 @@ static void ioat_dma_free_chan_resources(struct dma_chan *chan) writew(chanctrl, ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET); } -/** - * do_ioat_dma_memcpy - actual function that initiates a IOAT DMA transaction - * @ioat_chan: IOAT DMA channel handle - * @dest: DMA destination address - * @src: DMA source address - * @len: transaction length in bytes - */ - -static dma_cookie_t do_ioat_dma_memcpy(struct ioat_dma_chan *ioat_chan, - dma_addr_t dest, - dma_addr_t src, - size_t len) +static struct dma_async_tx_descriptor * +ioat_dma_prep_memcpy(struct dma_chan *chan, size_t len, int int_en) { - struct ioat_desc_sw *first; - struct ioat_desc_sw *prev; - struct ioat_desc_sw *new; - dma_cookie_t cookie; + struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan); + struct ioat_desc_sw *first, *prev, *new; LIST_HEAD(new_chain); u32 copy; size_t orig_len; - dma_addr_t orig_src, orig_dst; - unsigned int desc_count = 0; - unsigned int append = 0; - - if (!ioat_chan || !dest || !src) - return -EFAULT; + int desc_count = 0; if (!len) - return ioat_chan->common.cookie; + return NULL; orig_len = len; - orig_src = src; - orig_dst = dest; first = NULL; prev = NULL; spin_lock_bh(&ioat_chan->desc_lock); - while (len) { if (!list_empty(&ioat_chan->free_desc)) { new = to_ioat_desc(ioat_chan->free_desc.next); @@ -270,140 +325,36 @@ static dma_cookie_t do_ioat_dma_memcpy(struct ioat_dma_chan *ioat_chan, new->hw->size = copy; new->hw->ctl = 0; - new->hw->src_addr = src; - new->hw->dst_addr = dest; - new->cookie = 0; + new->async_tx.cookie = 0; + new->async_tx.ack = 1; /* chain together the physical address list for the HW */ if (!first) first = new; else - prev->hw->next = (u64) new->phys; + prev->hw->next = (u64) new->async_tx.phys; prev = new; - len -= copy; - dest += copy; - src += copy; - list_add_tail(&new->node, &new_chain); desc_count++; } - new->hw->ctl = IOAT_DMA_DESCRIPTOR_CTL_CP_STS; - new->hw->next = 0; - /* cookie incr and addition to used_list must be atomic */ + list_splice(&new_chain, &new->async_tx.tx_list); - cookie = ioat_chan->common.cookie; - cookie++; - if (cookie < 0) - cookie = 1; - ioat_chan->common.cookie = new->cookie = cookie; + new->hw->ctl = IOAT_DMA_DESCRIPTOR_CTL_CP_STS; + new->hw->next = 0; + new->tx_cnt = desc_count; + new->async_tx.ack = 0; /* client is in control of this ack */ + new->async_tx.cookie = -EBUSY; - pci_unmap_addr_set(new, src, orig_src); - pci_unmap_addr_set(new, dst, orig_dst); pci_unmap_len_set(new, src_len, orig_len); pci_unmap_len_set(new, dst_len, orig_len); - - /* write address into NextDescriptor field of last desc in chain */ - to_ioat_desc(ioat_chan->used_desc.prev)->hw->next = first->phys; - list_splice_init(&new_chain, ioat_chan->used_desc.prev); - - ioat_chan->pending += desc_count; - if (ioat_chan->pending >= 4) { - append = 1; - ioat_chan->pending = 0; - } - spin_unlock_bh(&ioat_chan->desc_lock); - if (append) - writeb(IOAT_CHANCMD_APPEND, - ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); - return cookie; + return new ? &new->async_tx : NULL; } -/** - * ioat_dma_memcpy_buf_to_buf - wrapper that takes src & dest bufs - * @chan: IOAT DMA channel handle - * @dest: DMA destination address - * @src: DMA source address - * @len: transaction length in bytes - */ - -static dma_cookie_t ioat_dma_memcpy_buf_to_buf(struct dma_chan *chan, - void *dest, - void *src, - size_t len) -{ - dma_addr_t dest_addr; - dma_addr_t src_addr; - struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan); - - dest_addr = pci_map_single(ioat_chan->device->pdev, - dest, len, PCI_DMA_FROMDEVICE); - src_addr = pci_map_single(ioat_chan->device->pdev, - src, len, PCI_DMA_TODEVICE); - - return do_ioat_dma_memcpy(ioat_chan, dest_addr, src_addr, len); -} - -/** - * ioat_dma_memcpy_buf_to_pg - wrapper, copying from a buf to a page - * @chan: IOAT DMA channel handle - * @page: pointer to the page to copy to - * @offset: offset into that page - * @src: DMA source address - * @len: transaction length in bytes - */ - -static dma_cookie_t ioat_dma_memcpy_buf_to_pg(struct dma_chan *chan, - struct page *page, - unsigned int offset, - void *src, - size_t len) -{ - dma_addr_t dest_addr; - dma_addr_t src_addr; - struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan); - - dest_addr = pci_map_page(ioat_chan->device->pdev, - page, offset, len, PCI_DMA_FROMDEVICE); - src_addr = pci_map_single(ioat_chan->device->pdev, - src, len, PCI_DMA_TODEVICE); - - return do_ioat_dma_memcpy(ioat_chan, dest_addr, src_addr, len); -} - -/** - * ioat_dma_memcpy_pg_to_pg - wrapper, copying between two pages - * @chan: IOAT DMA channel handle - * @dest_pg: pointer to the page to copy to - * @dest_off: offset into that page - * @src_pg: pointer to the page to copy from - * @src_off: offset into that page - * @len: transaction length in bytes. This is guaranteed not to make a copy - * across a page boundary. - */ - -static dma_cookie_t ioat_dma_memcpy_pg_to_pg(struct dma_chan *chan, - struct page *dest_pg, - unsigned int dest_off, - struct page *src_pg, - unsigned int src_off, - size_t len) -{ - dma_addr_t dest_addr; - dma_addr_t src_addr; - struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan); - - dest_addr = pci_map_page(ioat_chan->device->pdev, - dest_pg, dest_off, len, PCI_DMA_FROMDEVICE); - src_addr = pci_map_page(ioat_chan->device->pdev, - src_pg, src_off, len, PCI_DMA_TODEVICE); - - return do_ioat_dma_memcpy(ioat_chan, dest_addr, src_addr, len); -} /** * ioat_dma_memcpy_issue_pending - push potentially unrecognized appended descriptors to hw @@ -465,8 +416,8 @@ static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *chan) * exceeding xfercap, perhaps. If so, only the last one will * have a cookie, and require unmapping. */ - if (desc->cookie) { - cookie = desc->cookie; + if (desc->async_tx.cookie) { + cookie = desc->async_tx.cookie; /* yes we are unmapping both _page and _single alloc'd regions with unmap_page. Is this *really* that bad? @@ -481,14 +432,19 @@ static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *chan) PCI_DMA_TODEVICE); } - if (desc->phys != phys_complete) { - /* a completed entry, but not the last, so cleanup */ - list_del(&desc->node); - list_add_tail(&desc->node, &chan->free_desc); + if (desc->async_tx.phys != phys_complete) { + /* a completed entry, but not the last, so cleanup + * if the client is done with the descriptor + */ + if (desc->async_tx.ack) { + list_del(&desc->node); + list_add_tail(&desc->node, &chan->free_desc); + } else + desc->async_tx.cookie = 0; } else { /* last used desc. Do not remove, so we can append from it, but don't look at it next time, either */ - desc->cookie = 0; + desc->async_tx.cookie = 0; /* TODO check status bits? */ break; @@ -504,6 +460,17 @@ static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *chan) spin_unlock(&chan->cleanup_lock); } +static void ioat_dma_dependency_added(struct dma_chan *chan) +{ + struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan); + spin_lock_bh(&ioat_chan->desc_lock); + if (ioat_chan->pending == 0) { + spin_unlock_bh(&ioat_chan->desc_lock); + ioat_dma_memcpy_cleanup(ioat_chan); + } else + spin_unlock_bh(&ioat_chan->desc_lock); +} + /** * ioat_dma_is_complete - poll the status of a IOAT DMA transaction * @chan: IOAT DMA channel handle @@ -606,13 +573,14 @@ static void ioat_start_null_desc(struct ioat_dma_chan *ioat_chan) desc->hw->ctl = IOAT_DMA_DESCRIPTOR_NUL; desc->hw->next = 0; + desc->async_tx.ack = 1; list_add_tail(&desc->node, &ioat_chan->used_desc); spin_unlock_bh(&ioat_chan->desc_lock); - writel(((u64) desc->phys) & 0x00000000FFFFFFFF, + writel(((u64) desc->async_tx.phys) & 0x00000000FFFFFFFF, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_LOW); - writel(((u64) desc->phys) >> 32, + writel(((u64) desc->async_tx.phys) >> 32, ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_HIGH); writeb(IOAT_CHANCMD_START, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET); @@ -629,6 +597,8 @@ static int ioat_self_test(struct ioat_device *device) u8 *src; u8 *dest; struct dma_chan *dma_chan; + struct dma_async_tx_descriptor *tx; + dma_addr_t addr; dma_cookie_t cookie; int err = 0; @@ -654,7 +624,15 @@ static int ioat_self_test(struct ioat_device *device) goto out; } - cookie = ioat_dma_memcpy_buf_to_buf(dma_chan, dest, src, IOAT_TEST_SIZE); + tx = ioat_dma_prep_memcpy(dma_chan, IOAT_TEST_SIZE, 0); + async_tx_ack(tx); + addr = dma_map_single(dma_chan->device->dev, src, IOAT_TEST_SIZE, + DMA_TO_DEVICE); + ioat_set_src(addr, tx, 0); + addr = dma_map_single(dma_chan->device->dev, dest, IOAT_TEST_SIZE, + DMA_FROM_DEVICE); + ioat_set_dest(addr, tx, 0); + cookie = ioat_tx_submit(tx); ioat_dma_memcpy_issue_pending(dma_chan); msleep(1); @@ -750,13 +728,14 @@ static int __devinit ioat_probe(struct pci_dev *pdev, INIT_LIST_HEAD(&device->common.channels); enumerate_dma_channels(device); + dma_cap_set(DMA_MEMCPY, device->common.cap_mask); device->common.device_alloc_chan_resources = ioat_dma_alloc_chan_resources; device->common.device_free_chan_resources = ioat_dma_free_chan_resources; - device->common.device_memcpy_buf_to_buf = ioat_dma_memcpy_buf_to_buf; - device->common.device_memcpy_buf_to_pg = ioat_dma_memcpy_buf_to_pg; - device->common.device_memcpy_pg_to_pg = ioat_dma_memcpy_pg_to_pg; - device->common.device_memcpy_complete = ioat_dma_is_complete; - device->common.device_memcpy_issue_pending = ioat_dma_memcpy_issue_pending; + device->common.device_prep_dma_memcpy = ioat_dma_prep_memcpy; + device->common.device_is_tx_complete = ioat_dma_is_complete; + device->common.device_issue_pending = ioat_dma_memcpy_issue_pending; + device->common.device_dependency_added = ioat_dma_dependency_added; + device->common.dev = &pdev->dev; printk(KERN_INFO "Intel(R) I/OAT DMA Engine found, %d channels\n", device->common.chancnt); diff --git a/drivers/dma/ioatdma.h b/drivers/dma/ioatdma.h index 62b26a9..d3f69bb 100644 --- a/drivers/dma/ioatdma.h +++ b/drivers/dma/ioatdma.h @@ -105,21 +105,20 @@ struct ioat_dma_chan { /** * struct ioat_desc_sw - wrapper around hardware descriptor * @hw: hardware DMA descriptor - * @node: - * @cookie: - * @phys: + * @node: this descriptor will either be on the free list, + * or attached to a transaction list (async_tx.tx_list) + * @tx_cnt: number of descriptors required to complete the transaction + * @async_tx: the generic software descriptor for all engines */ - struct ioat_desc_sw { struct ioat_dma_descriptor *hw; struct list_head node; - dma_cookie_t cookie; - dma_addr_t phys; + int tx_cnt; DECLARE_PCI_UNMAP_ADDR(src) DECLARE_PCI_UNMAP_LEN(src_len) DECLARE_PCI_UNMAP_ADDR(dst) DECLARE_PCI_UNMAP_LEN(dst_len) + struct dma_async_tx_descriptor async_tx; }; #endif /* IOATDMA_H */ - diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index c94d8f1..3de1cf7 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -21,13 +21,12 @@ #ifndef DMAENGINE_H #define DMAENGINE_H -#ifdef CONFIG_DMA_ENGINE - #include #include #include #include #include +#include /** * enum dma_event - resource PNP/power managment events @@ -65,6 +64,31 @@ enum dma_status { }; /** + * enum dma_transaction_type - DMA transaction types/indexes + */ +enum dma_transaction_type { + DMA_MEMCPY, + DMA_XOR, + DMA_PQ_XOR, + DMA_DUAL_XOR, + DMA_PQ_UPDATE, + DMA_ZERO_SUM, + DMA_PQ_ZERO_SUM, + DMA_MEMSET, + DMA_MEMCPY_CRC32C, + DMA_INTERRUPT, +}; + +/* last transaction type for creation of the capabilities mask */ +#define DMA_TX_TYPE_END (DMA_INTERRUPT + 1) + +/** + * dma_cap_mask_t - capabilities bitmap modeled after cpumask_t. + * See linux/cpumask.h + */ +typedef struct { DECLARE_BITMAP(bits, DMA_TX_TYPE_END); } dma_cap_mask_t; + +/** * struct dma_chan_percpu - the per-CPU part of struct dma_chan * @refcount: local_t used for open-coded "bigref" counting * @memcpy_count: transaction counter @@ -157,48 +181,106 @@ struct dma_client { struct list_head global_node; }; +typedef void (*dma_async_tx_callback)(void *dma_async_param); +/** + * struct dma_async_tx_descriptor - async transaction descriptor + * ---dma generic offload fields--- + * @cookie: tracking cookie for this transaction, set to -EBUSY if + * this tx is sitting on a dependency list + * @ack: the descriptor can not be reused until the client acknowledges + * receipt, i.e. has has a chance to establish any dependency chains + * @phys: physical address of the descriptor + * @tx_list: driver common field for operations that require multiple + * descriptors + * @chan: target channel for this operation + * @tx_submit: set the prepared descriptor(s) to be executed by the engine + * @tx_set_dest: set a destination address in a hardware descriptor + * @tx_set_src: set a source address in a hardware descriptor + * @callback: routine to call after this operation is complete + * @callback_param: general parameter to pass to the callback routine + * ---async_tx api specific fields--- + * @depend_list: at completion this list of transactions are submitted + * @depend_node: allow this transaction to be executed after another + * transaction has completed, possibly on another channel + * @parent: pointer to the next level up in the dependency chain + * @lock: protect the dependency list + */ +struct dma_async_tx_descriptor { + dma_cookie_t cookie; + int ack; + dma_addr_t phys; + struct list_head tx_list; + struct dma_chan *chan; + dma_cookie_t (*tx_submit)(struct dma_async_tx_descriptor *tx); + void (*tx_set_dest)(dma_addr_t addr, + struct dma_async_tx_descriptor *tx, int index); + void (*tx_set_src)(dma_addr_t addr, + struct dma_async_tx_descriptor *tx, int index); + dma_async_tx_callback callback; + void *callback_param; + struct list_head depend_list; + struct list_head depend_node; + struct dma_async_tx_descriptor *parent; + spinlock_t lock; +}; + /** * struct dma_device - info on the entity supplying DMA services * @chancnt: how many DMA channels are supported * @channels: the list of struct dma_chan * @global_node: list_head for global dma_device_list + * @cap_mask: one or more dma_capability flags + * @max_xor: maximum number of xor sources, 0 if no capability * @refcount: reference count * @done: IO completion struct * @dev_id: unique device ID + * @dev: struct device reference for dma mapping api * @device_alloc_chan_resources: allocate resources and return the * number of allocated descriptors * @device_free_chan_resources: release DMA channel's resources - * @device_memcpy_buf_to_buf: memcpy buf pointer to buf pointer - * @device_memcpy_buf_to_pg: memcpy buf pointer to struct page - * @device_memcpy_pg_to_pg: memcpy struct page/offset to struct page/offset - * @device_memcpy_complete: poll the status of an IOAT DMA transaction - * @device_memcpy_issue_pending: push appended descriptors to hardware + * @device_prep_dma_memcpy: prepares a memcpy operation + * @device_prep_dma_xor: prepares a xor operation + * @device_prep_dma_zero_sum: prepares a zero_sum operation + * @device_prep_dma_memset: prepares a memset operation + * @device_prep_dma_interrupt: prepares an end of chain interrupt operation + * @device_dependency_added: async_tx notifies the channel about new deps + * @device_issue_pending: push pending transactions to hardware */ struct dma_device { unsigned int chancnt; struct list_head channels; struct list_head global_node; + dma_cap_mask_t cap_mask; + int max_xor; struct kref refcount; struct completion done; int dev_id; + struct device *dev; int (*device_alloc_chan_resources)(struct dma_chan *chan); void (*device_free_chan_resources)(struct dma_chan *chan); - dma_cookie_t (*device_memcpy_buf_to_buf)(struct dma_chan *chan, - void *dest, void *src, size_t len); - dma_cookie_t (*device_memcpy_buf_to_pg)(struct dma_chan *chan, - struct page *page, unsigned int offset, void *kdata, - size_t len); - dma_cookie_t (*device_memcpy_pg_to_pg)(struct dma_chan *chan, - struct page *dest_pg, unsigned int dest_off, - struct page *src_pg, unsigned int src_off, size_t len); - enum dma_status (*device_memcpy_complete)(struct dma_chan *chan, + + struct dma_async_tx_descriptor *(*device_prep_dma_memcpy)( + struct dma_chan *chan, size_t len, int int_en); + struct dma_async_tx_descriptor *(*device_prep_dma_xor)( + struct dma_chan *chan, unsigned int src_cnt, size_t len, + int int_en); + struct dma_async_tx_descriptor *(*device_prep_dma_zero_sum)( + struct dma_chan *chan, unsigned int src_cnt, size_t len, + u32 *result, int int_en); + struct dma_async_tx_descriptor *(*device_prep_dma_memset)( + struct dma_chan *chan, int value, size_t len, int int_en); + struct dma_async_tx_descriptor *(*device_prep_dma_interrupt)( + struct dma_chan *chan); + + void (*device_dependency_added)(struct dma_chan *chan); + enum dma_status (*device_is_tx_complete)(struct dma_chan *chan, dma_cookie_t cookie, dma_cookie_t *last, dma_cookie_t *used); - void (*device_memcpy_issue_pending)(struct dma_chan *chan); + void (*device_issue_pending)(struct dma_chan *chan); }; /* --- public DMA engine API --- */ @@ -207,96 +289,72 @@ struct dma_client *dma_async_client_register(dma_event_callback event_callback); void dma_async_client_unregister(struct dma_client *client); void dma_async_client_chan_request(struct dma_client *client, unsigned int number); +dma_cookie_t dma_async_memcpy_buf_to_buf(struct dma_chan *chan, + void *dest, void *src, size_t len); +dma_cookie_t dma_async_memcpy_buf_to_pg(struct dma_chan *chan, + struct page *page, unsigned int offset, void *kdata, size_t len); +dma_cookie_t dma_async_memcpy_pg_to_pg(struct dma_chan *chan, + struct page *dest_pg, unsigned int dest_off, struct page *src_pg, + unsigned int src_off, size_t len); +void dma_async_tx_descriptor_init(struct dma_async_tx_descriptor *tx, + struct dma_chan *chan); -/** - * dma_async_memcpy_buf_to_buf - offloaded copy between virtual addresses - * @chan: DMA channel to offload copy to - * @dest: destination address (virtual) - * @src: source address (virtual) - * @len: length - * - * Both @dest and @src must be mappable to a bus address according to the - * DMA mapping API rules for streaming mappings. - * Both @dest and @src must stay memory resident (kernel memory or locked - * user space pages). - */ -static inline dma_cookie_t dma_async_memcpy_buf_to_buf(struct dma_chan *chan, - void *dest, void *src, size_t len) -{ - int cpu = get_cpu(); - per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; - per_cpu_ptr(chan->local, cpu)->memcpy_count++; - put_cpu(); - return chan->device->device_memcpy_buf_to_buf(chan, dest, src, len); +static inline void +async_tx_ack(struct dma_async_tx_descriptor *tx) +{ + tx->ack = 1; } -/** - * dma_async_memcpy_buf_to_pg - offloaded copy from address to page - * @chan: DMA channel to offload copy to - * @page: destination page - * @offset: offset in page to copy to - * @kdata: source address (virtual) - * @len: length - * - * Both @page/@offset and @kdata must be mappable to a bus address according - * to the DMA mapping API rules for streaming mappings. - * Both @page/@offset and @kdata must stay memory resident (kernel memory or - * locked user space pages) - */ -static inline dma_cookie_t dma_async_memcpy_buf_to_pg(struct dma_chan *chan, - struct page *page, unsigned int offset, void *kdata, size_t len) +#define first_dma_cap(mask) __first_dma_cap(&(mask)) +static inline int __first_dma_cap(const dma_cap_mask_t *srcp) { - int cpu = get_cpu(); - per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; - per_cpu_ptr(chan->local, cpu)->memcpy_count++; - put_cpu(); + return min_t(int, DMA_TX_TYPE_END, + find_first_bit(srcp->bits, DMA_TX_TYPE_END)); +} - return chan->device->device_memcpy_buf_to_pg(chan, page, offset, - kdata, len); +#define next_dma_cap(n, mask) __next_dma_cap((n), &(mask)) +static inline int __next_dma_cap(int n, const dma_cap_mask_t *srcp) +{ + return min_t(int, DMA_TX_TYPE_END, + find_next_bit(srcp->bits, DMA_TX_TYPE_END, n+1)); } -/** - * dma_async_memcpy_pg_to_pg - offloaded copy from page to page - * @chan: DMA channel to offload copy to - * @dest_pg: destination page - * @dest_off: offset in page to copy to - * @src_pg: source page - * @src_off: offset in page to copy from - * @len: length - * - * Both @dest_page/@dest_off and @src_page/@src_off must be mappable to a bus - * address according to the DMA mapping API rules for streaming mappings. - * Both @dest_page/@dest_off and @src_page/@src_off must stay memory resident - * (kernel memory or locked user space pages). - */ -static inline dma_cookie_t dma_async_memcpy_pg_to_pg(struct dma_chan *chan, - struct page *dest_pg, unsigned int dest_off, struct page *src_pg, - unsigned int src_off, size_t len) +#define dma_cap_set(tx, mask) __dma_cap_set((tx), &(mask)) +static inline void +__dma_cap_set(enum dma_transaction_type tx_type, dma_cap_mask_t *dstp) { - int cpu = get_cpu(); - per_cpu_ptr(chan->local, cpu)->bytes_transferred += len; - per_cpu_ptr(chan->local, cpu)->memcpy_count++; - put_cpu(); + set_bit(tx_type, dstp->bits); +} - return chan->device->device_memcpy_pg_to_pg(chan, dest_pg, dest_off, - src_pg, src_off, len); +#define dma_has_cap(tx, mask) __dma_has_cap((tx), &(mask)) +static inline int +__dma_has_cap(enum dma_transaction_type tx_type, dma_cap_mask_t *srcp) +{ + return test_bit(tx_type, srcp->bits); } +#define for_each_dma_cap_mask(cap, mask) \ + for ((cap) = first_dma_cap(mask); \ + (cap) < DMA_TX_TYPE_END; \ + (cap) = next_dma_cap((cap), (mask))) + /** - * dma_async_memcpy_issue_pending - flush pending copies to HW + * dma_async_issue_pending - flush pending transactions to HW * @chan: target DMA channel * * This allows drivers to push copies to HW in batches, * reducing MMIO writes where possible. */ -static inline void dma_async_memcpy_issue_pending(struct dma_chan *chan) +static inline void dma_async_issue_pending(struct dma_chan *chan) { - return chan->device->device_memcpy_issue_pending(chan); + return chan->device->device_issue_pending(chan); } +#define dma_async_memcpy_issue_pending(chan) dma_async_issue_pending(chan) + /** - * dma_async_memcpy_complete - poll for transaction completion + * dma_async_is_tx_complete - poll for transaction completion * @chan: DMA channel * @cookie: transaction identifier to check status of * @last: returns last completed cookie, can be NULL @@ -306,12 +364,15 @@ static inline void dma_async_memcpy_issue_pending(struct dma_chan *chan) * internal state and can be used with dma_async_is_complete() to check * the status of multiple cookies without re-checking hardware state. */ -static inline enum dma_status dma_async_memcpy_complete(struct dma_chan *chan, +static inline enum dma_status dma_async_is_tx_complete(struct dma_chan *chan, dma_cookie_t cookie, dma_cookie_t *last, dma_cookie_t *used) { - return chan->device->device_memcpy_complete(chan, cookie, last, used); + return chan->device->device_is_tx_complete(chan, cookie, last, used); } +#define dma_async_memcpy_complete(chan, cookie, last, used)\ + dma_async_is_tx_complete(chan, cookie, last, used) + /** * dma_async_is_complete - test a cookie against chan state * @cookie: transaction identifier to test status of @@ -334,6 +395,7 @@ static inline enum dma_status dma_async_is_complete(dma_cookie_t cookie, return DMA_IN_PROGRESS; } +enum dma_status dma_sync_wait(struct dma_chan *chan, dma_cookie_t cookie); /* --- DMA device --- */ @@ -362,5 +424,4 @@ dma_cookie_t dma_memcpy_pg_to_iovec(struct dma_chan *chan, struct iovec *iov, struct dma_pinned_list *pinned_list, struct page *page, unsigned int offset, size_t len); -#endif /* CONFIG_DMA_ENGINE */ #endif /* DMAENGINE_H */ -- cgit v1.1 From d379b01e9087a582d58f4b678208a4f8d8376fe7 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Mon, 9 Jul 2007 11:56:42 -0700 Subject: dmaengine: make clients responsible for managing channels The current implementation assumes that a channel will only be used by one client at a time. In order to enable channel sharing the dmaengine core is changed to a model where clients subscribe to channel-available-events. Instead of tracking how many channels a client wants and how many it has received the core just broadcasts the available channels and lets the clients optionally take a reference. The core learns about the clients' needs at dma_event_callback time. In support of multiple operation types, clients can specify a capability mask to only be notified of channels that satisfy a certain set of capabilities. Changelog: * removed DMA_TX_ARRAY_INIT, no longer needed * dma_client_chan_free -> dma_chan_release: switch to global reference counting only at device unregistration time, before it was also happening at client unregistration time * clients now return dma_state_client to dmaengine (ack, dup, nak) * checkpatch.pl fixes * fixup merge with git-ioat Cc: Chris Leech Signed-off-by: Shannon Nelson Signed-off-by: Dan Williams Acked-by: David S. Miller --- drivers/dma/dmaengine.c | 217 ++++++++++++++++++++++++---------------------- drivers/dma/ioatdma.c | 1 - drivers/dma/ioatdma.h | 3 - include/linux/dmaengine.h | 58 ++++++++----- net/core/dev.c | 112 ++++++++++++++++-------- 5 files changed, 224 insertions(+), 167 deletions(-) diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c index 404cc7b..8248992 100644 --- a/drivers/dma/dmaengine.c +++ b/drivers/dma/dmaengine.c @@ -37,11 +37,11 @@ * Each device has a channels list, which runs unlocked but is never modified * once the device is registered, it's just setup by the driver. * - * Each client has a channels list, it's only modified under the client->lock - * and in an RCU callback, so it's safe to read under rcu_read_lock(). + * Each client is responsible for keeping track of the channels it uses. See + * the definition of dma_event_callback in dmaengine.h. * * Each device has a kref, which is initialized to 1 when the device is - * registered. A kref_put is done for each class_device registered. When the + * registered. A kref_get is done for each class_device registered. When the * class_device is released, the coresponding kref_put is done in the release * method. Every time one of the device's channels is allocated to a client, * a kref_get occurs. When the channel is freed, the coresponding kref_put @@ -51,10 +51,12 @@ * references to finish. * * Each channel has an open-coded implementation of Rusty Russell's "bigref," - * with a kref and a per_cpu local_t. A single reference is set when on an - * ADDED event, and removed with a REMOVE event. Net DMA client takes an - * extra reference per outstanding transaction. The relase function does a - * kref_put on the device. -ChrisL + * with a kref and a per_cpu local_t. A dma_chan_get is called when a client + * signals that it wants to use a channel, and dma_chan_put is called when + * a channel is removed or a client using it is unregesitered. A client can + * take extra references per outstanding transaction, as is the case with + * the NET DMA client. The release function does a kref_put on the device. + * -ChrisL, DanW */ #include @@ -102,8 +104,19 @@ static ssize_t show_bytes_transferred(struct class_device *cd, char *buf) static ssize_t show_in_use(struct class_device *cd, char *buf) { struct dma_chan *chan = container_of(cd, struct dma_chan, class_dev); + int in_use = 0; + + if (unlikely(chan->slow_ref) && + atomic_read(&chan->refcount.refcount) > 1) + in_use = 1; + else { + if (local_read(&(per_cpu_ptr(chan->local, + get_cpu())->refcount)) > 0) + in_use = 1; + put_cpu(); + } - return sprintf(buf, "%d\n", (chan->client ? 1 : 0)); + return sprintf(buf, "%d\n", in_use); } static struct class_device_attribute dma_class_attrs[] = { @@ -129,42 +142,53 @@ static struct class dma_devclass = { /* --- client and device registration --- */ +#define dma_chan_satisfies_mask(chan, mask) \ + __dma_chan_satisfies_mask((chan), &(mask)) +static int +__dma_chan_satisfies_mask(struct dma_chan *chan, dma_cap_mask_t *want) +{ + dma_cap_mask_t has; + + bitmap_and(has.bits, want->bits, chan->device->cap_mask.bits, + DMA_TX_TYPE_END); + return bitmap_equal(want->bits, has.bits, DMA_TX_TYPE_END); +} + /** - * dma_client_chan_alloc - try to allocate a channel to a client + * dma_client_chan_alloc - try to allocate channels to a client * @client: &dma_client * * Called with dma_list_mutex held. */ -static struct dma_chan *dma_client_chan_alloc(struct dma_client *client) +static void dma_client_chan_alloc(struct dma_client *client) { struct dma_device *device; struct dma_chan *chan; - unsigned long flags; int desc; /* allocated descriptor count */ + enum dma_state_client ack; - /* Find a channel, any DMA engine will do */ - list_for_each_entry(device, &dma_device_list, global_node) { + /* Find a channel */ + list_for_each_entry(device, &dma_device_list, global_node) list_for_each_entry(chan, &device->channels, device_node) { - if (chan->client) + if (!dma_chan_satisfies_mask(chan, client->cap_mask)) continue; desc = chan->device->device_alloc_chan_resources(chan); if (desc >= 0) { - kref_get(&device->refcount); - kref_init(&chan->refcount); - chan->slow_ref = 0; - INIT_RCU_HEAD(&chan->rcu); - chan->client = client; - spin_lock_irqsave(&client->lock, flags); - list_add_tail_rcu(&chan->client_node, - &client->channels); - spin_unlock_irqrestore(&client->lock, flags); - return chan; + ack = client->event_callback(client, + chan, + DMA_RESOURCE_AVAILABLE); + + /* we are done once this client rejects + * an available resource + */ + if (ack == DMA_ACK) { + dma_chan_get(chan); + kref_get(&device->refcount); + } else if (ack == DMA_NAK) + return; } } - } - - return NULL; } enum dma_status dma_sync_wait(struct dma_chan *chan, dma_cookie_t cookie) @@ -193,7 +217,6 @@ void dma_chan_cleanup(struct kref *kref) { struct dma_chan *chan = container_of(kref, struct dma_chan, refcount); chan->device->device_free_chan_resources(chan); - chan->client = NULL; kref_put(&chan->device->refcount, dma_async_device_cleanup); } EXPORT_SYMBOL(dma_chan_cleanup); @@ -209,7 +232,7 @@ static void dma_chan_free_rcu(struct rcu_head *rcu) kref_put(&chan->refcount, dma_chan_cleanup); } -static void dma_client_chan_free(struct dma_chan *chan) +static void dma_chan_release(struct dma_chan *chan) { atomic_add(0x7FFFFFFF, &chan->refcount.refcount); chan->slow_ref = 1; @@ -217,70 +240,57 @@ static void dma_client_chan_free(struct dma_chan *chan) } /** - * dma_chans_rebalance - reallocate channels to clients - * - * When the number of DMA channel in the system changes, - * channels need to be rebalanced among clients. + * dma_chans_notify_available - broadcast available channels to the clients */ -static void dma_chans_rebalance(void) +static void dma_clients_notify_available(void) { struct dma_client *client; - struct dma_chan *chan; - unsigned long flags; mutex_lock(&dma_list_mutex); - list_for_each_entry(client, &dma_client_list, global_node) { - while (client->chans_desired > client->chan_count) { - chan = dma_client_chan_alloc(client); - if (!chan) - break; - client->chan_count++; - client->event_callback(client, - chan, - DMA_RESOURCE_ADDED); - } - while (client->chans_desired < client->chan_count) { - spin_lock_irqsave(&client->lock, flags); - chan = list_entry(client->channels.next, - struct dma_chan, - client_node); - list_del_rcu(&chan->client_node); - spin_unlock_irqrestore(&client->lock, flags); - client->chan_count--; - client->event_callback(client, - chan, - DMA_RESOURCE_REMOVED); - dma_client_chan_free(chan); - } - } + list_for_each_entry(client, &dma_client_list, global_node) + dma_client_chan_alloc(client); mutex_unlock(&dma_list_mutex); } /** - * dma_async_client_register - allocate and register a &dma_client - * @event_callback: callback for notification of channel addition/removal + * dma_chans_notify_available - tell the clients that a channel is going away + * @chan: channel on its way out */ -struct dma_client *dma_async_client_register(dma_event_callback event_callback) +static void dma_clients_notify_removed(struct dma_chan *chan) { struct dma_client *client; + enum dma_state_client ack; - client = kzalloc(sizeof(*client), GFP_KERNEL); - if (!client) - return NULL; + mutex_lock(&dma_list_mutex); + + list_for_each_entry(client, &dma_client_list, global_node) { + ack = client->event_callback(client, chan, + DMA_RESOURCE_REMOVED); + + /* client was holding resources for this channel so + * free it + */ + if (ack == DMA_ACK) { + dma_chan_put(chan); + kref_put(&chan->device->refcount, + dma_async_device_cleanup); + } + } - INIT_LIST_HEAD(&client->channels); - spin_lock_init(&client->lock); - client->chans_desired = 0; - client->chan_count = 0; - client->event_callback = event_callback; + mutex_unlock(&dma_list_mutex); +} +/** + * dma_async_client_register - register a &dma_client + * @client: ptr to a client structure with valid 'event_callback' and 'cap_mask' + */ +void dma_async_client_register(struct dma_client *client) +{ mutex_lock(&dma_list_mutex); list_add_tail(&client->global_node, &dma_client_list); mutex_unlock(&dma_list_mutex); - - return client; } EXPORT_SYMBOL(dma_async_client_register); @@ -292,40 +302,42 @@ EXPORT_SYMBOL(dma_async_client_register); */ void dma_async_client_unregister(struct dma_client *client) { + struct dma_device *device; struct dma_chan *chan; + enum dma_state_client ack; if (!client) return; - rcu_read_lock(); - list_for_each_entry_rcu(chan, &client->channels, client_node) - dma_client_chan_free(chan); - rcu_read_unlock(); - mutex_lock(&dma_list_mutex); + /* free all channels the client is holding */ + list_for_each_entry(device, &dma_device_list, global_node) + list_for_each_entry(chan, &device->channels, device_node) { + ack = client->event_callback(client, chan, + DMA_RESOURCE_REMOVED); + + if (ack == DMA_ACK) { + dma_chan_put(chan); + kref_put(&chan->device->refcount, + dma_async_device_cleanup); + } + } + list_del(&client->global_node); mutex_unlock(&dma_list_mutex); - - kfree(client); - dma_chans_rebalance(); } EXPORT_SYMBOL(dma_async_client_unregister); /** - * dma_async_client_chan_request - request DMA channels - * @client: &dma_client - * @number: count of DMA channels requested - * - * Clients call dma_async_client_chan_request() to specify how many - * DMA channels they need, 0 to free all currently allocated. - * The resulting allocations/frees are indicated to the client via the - * event callback. + * dma_async_client_chan_request - send all available channels to the + * client that satisfy the capability mask + * @client - requester */ -void dma_async_client_chan_request(struct dma_client *client, - unsigned int number) +void dma_async_client_chan_request(struct dma_client *client) { - client->chans_desired = number; - dma_chans_rebalance(); + mutex_lock(&dma_list_mutex); + dma_client_chan_alloc(client); + mutex_unlock(&dma_list_mutex); } EXPORT_SYMBOL(dma_async_client_chan_request); @@ -386,13 +398,16 @@ int dma_async_device_register(struct dma_device *device) } kref_get(&device->refcount); + kref_init(&chan->refcount); + chan->slow_ref = 0; + INIT_RCU_HEAD(&chan->rcu); } mutex_lock(&dma_list_mutex); list_add_tail(&device->global_node, &dma_device_list); mutex_unlock(&dma_list_mutex); - dma_chans_rebalance(); + dma_clients_notify_available(); return 0; @@ -428,26 +443,16 @@ static void dma_async_device_cleanup(struct kref *kref) void dma_async_device_unregister(struct dma_device *device) { struct dma_chan *chan; - unsigned long flags; mutex_lock(&dma_list_mutex); list_del(&device->global_node); mutex_unlock(&dma_list_mutex); list_for_each_entry(chan, &device->channels, device_node) { - if (chan->client) { - spin_lock_irqsave(&chan->client->lock, flags); - list_del(&chan->client_node); - chan->client->chan_count--; - spin_unlock_irqrestore(&chan->client->lock, flags); - chan->client->event_callback(chan->client, - chan, - DMA_RESOURCE_REMOVED); - dma_client_chan_free(chan); - } + dma_clients_notify_removed(chan); class_device_unregister(&chan->class_dev); + dma_chan_release(chan); } - dma_chans_rebalance(); kref_put(&device->refcount, dma_async_device_cleanup); wait_for_completion(&device->done); diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index 1710449..81810b3 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -72,7 +72,6 @@ static int enumerate_dma_channels(struct ioat_device *device) INIT_LIST_HEAD(&ioat_chan->used_desc); /* This should be made common somewhere in dmaengine.c */ ioat_chan->common.device = &device->common; - ioat_chan->common.client = NULL; list_add_tail(&ioat_chan->common.device_node, &device->common.channels); } diff --git a/drivers/dma/ioatdma.h b/drivers/dma/ioatdma.h index d3f69bb..d372647 100644 --- a/drivers/dma/ioatdma.h +++ b/drivers/dma/ioatdma.h @@ -30,9 +30,6 @@ #define IOAT_LOW_COMPLETION_MASK 0xffffffc0 -extern struct list_head dma_device_list; -extern struct list_head dma_client_list; - /** * struct ioat_device - internal representation of a IOAT device * @pdev: PCI-Express device diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 3de1cf7..a3b6035 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -29,20 +29,32 @@ #include /** - * enum dma_event - resource PNP/power managment events + * enum dma_state - resource PNP/power managment state * @DMA_RESOURCE_SUSPEND: DMA device going into low power state * @DMA_RESOURCE_RESUME: DMA device returning to full power - * @DMA_RESOURCE_ADDED: DMA device added to the system + * @DMA_RESOURCE_AVAILABLE: DMA device available to the system * @DMA_RESOURCE_REMOVED: DMA device removed from the system */ -enum dma_event { +enum dma_state { DMA_RESOURCE_SUSPEND, DMA_RESOURCE_RESUME, - DMA_RESOURCE_ADDED, + DMA_RESOURCE_AVAILABLE, DMA_RESOURCE_REMOVED, }; /** + * enum dma_state_client - state of the channel in the client + * @DMA_ACK: client would like to use, or was using this channel + * @DMA_DUP: client has already seen this channel, or is not using this channel + * @DMA_NAK: client does not want to see any more channels + */ +enum dma_state_client { + DMA_ACK, + DMA_DUP, + DMA_NAK, +}; + +/** * typedef dma_cookie_t - an opaque DMA cookie * * if dma_cookie_t is >0 it's a DMA request cookie, <0 it's an error code @@ -104,7 +116,6 @@ struct dma_chan_percpu { /** * struct dma_chan - devices supply DMA channels, clients use them - * @client: ptr to the client user of this chan, will be %NULL when unused * @device: ptr to the dma device who supplies this channel, always !%NULL * @cookie: last cookie value returned to client * @chan_id: channel ID for sysfs @@ -112,12 +123,10 @@ struct dma_chan_percpu { * @refcount: kref, used in "bigref" slow-mode * @slow_ref: indicates that the DMA channel is free * @rcu: the DMA channel's RCU head - * @client_node: used to add this to the client chan list * @device_node: used to add this to the device chan list * @local: per-cpu pointer to a struct dma_chan_percpu */ struct dma_chan { - struct dma_client *client; struct dma_device *device; dma_cookie_t cookie; @@ -129,11 +138,11 @@ struct dma_chan { int slow_ref; struct rcu_head rcu; - struct list_head client_node; struct list_head device_node; struct dma_chan_percpu *local; }; + void dma_chan_cleanup(struct kref *kref); static inline void dma_chan_get(struct dma_chan *chan) @@ -158,26 +167,31 @@ static inline void dma_chan_put(struct dma_chan *chan) /* * typedef dma_event_callback - function pointer to a DMA event callback + * For each channel added to the system this routine is called for each client. + * If the client would like to use the channel it returns '1' to signal (ack) + * the dmaengine core to take out a reference on the channel and its + * corresponding device. A client must not 'ack' an available channel more + * than once. When a channel is removed all clients are notified. If a client + * is using the channel it must 'ack' the removal. A client must not 'ack' a + * removed channel more than once. + * @client - 'this' pointer for the client context + * @chan - channel to be acted upon + * @state - available or removed */ -typedef void (*dma_event_callback) (struct dma_client *client, - struct dma_chan *chan, enum dma_event event); +struct dma_client; +typedef enum dma_state_client (*dma_event_callback) (struct dma_client *client, + struct dma_chan *chan, enum dma_state state); /** * struct dma_client - info on the entity making use of DMA services * @event_callback: func ptr to call when something happens - * @chan_count: number of chans allocated - * @chans_desired: number of chans requested. Can be +/- chan_count - * @lock: protects access to the channels list - * @channels: the list of DMA channels allocated + * @cap_mask: only return channels that satisfy the requested capabilities + * a value of zero corresponds to any capability * @global_node: list_head for global dma_client_list */ struct dma_client { dma_event_callback event_callback; - unsigned int chan_count; - unsigned int chans_desired; - - spinlock_t lock; - struct list_head channels; + dma_cap_mask_t cap_mask; struct list_head global_node; }; @@ -285,10 +299,9 @@ struct dma_device { /* --- public DMA engine API --- */ -struct dma_client *dma_async_client_register(dma_event_callback event_callback); +void dma_async_client_register(struct dma_client *client); void dma_async_client_unregister(struct dma_client *client); -void dma_async_client_chan_request(struct dma_client *client, - unsigned int number); +void dma_async_client_chan_request(struct dma_client *client); dma_cookie_t dma_async_memcpy_buf_to_buf(struct dma_chan *chan, void *dest, void *src, size_t len); dma_cookie_t dma_async_memcpy_buf_to_pg(struct dma_chan *chan, @@ -299,7 +312,6 @@ dma_cookie_t dma_async_memcpy_pg_to_pg(struct dma_chan *chan, void dma_async_tx_descriptor_init(struct dma_async_tx_descriptor *tx, struct dma_chan *chan); - static inline void async_tx_ack(struct dma_async_tx_descriptor *tx) { diff --git a/net/core/dev.c b/net/core/dev.c index ee051bb..835202f 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -151,9 +151,22 @@ static struct list_head ptype_base[16] __read_mostly; /* 16 way hashed list */ static struct list_head ptype_all __read_mostly; /* Taps */ #ifdef CONFIG_NET_DMA -static struct dma_client *net_dma_client; -static unsigned int net_dma_count; -static spinlock_t net_dma_event_lock; +struct net_dma { + struct dma_client client; + spinlock_t lock; + cpumask_t channel_mask; + struct dma_chan *channels[NR_CPUS]; +}; + +static enum dma_state_client +netdev_dma_event(struct dma_client *client, struct dma_chan *chan, + enum dma_state state); + +static struct net_dma net_dma = { + .client = { + .event_callback = netdev_dma_event, + }, +}; #endif /* @@ -2015,12 +2028,13 @@ out: * There may not be any more sk_buffs coming right now, so push * any pending DMA copies to hardware */ - if (net_dma_client) { - struct dma_chan *chan; - rcu_read_lock(); - list_for_each_entry_rcu(chan, &net_dma_client->channels, client_node) - dma_async_memcpy_issue_pending(chan); - rcu_read_unlock(); + if (!cpus_empty(net_dma.channel_mask)) { + int chan_idx; + for_each_cpu_mask(chan_idx, net_dma.channel_mask) { + struct dma_chan *chan = net_dma.channels[chan_idx]; + if (chan) + dma_async_memcpy_issue_pending(chan); + } } #endif return; @@ -3563,12 +3577,13 @@ static int dev_cpu_callback(struct notifier_block *nfb, * This is called when the number of channels allocated to the net_dma_client * changes. The net_dma_client tries to have one DMA channel per CPU. */ -static void net_dma_rebalance(void) + +static void net_dma_rebalance(struct net_dma *net_dma) { - unsigned int cpu, i, n; + unsigned int cpu, i, n, chan_idx; struct dma_chan *chan; - if (net_dma_count == 0) { + if (cpus_empty(net_dma->channel_mask)) { for_each_online_cpu(cpu) rcu_assign_pointer(per_cpu(softnet_data, cpu).net_dma, NULL); return; @@ -3577,10 +3592,12 @@ static void net_dma_rebalance(void) i = 0; cpu = first_cpu(cpu_online_map); - rcu_read_lock(); - list_for_each_entry(chan, &net_dma_client->channels, client_node) { - n = ((num_online_cpus() / net_dma_count) - + (i < (num_online_cpus() % net_dma_count) ? 1 : 0)); + for_each_cpu_mask(chan_idx, net_dma->channel_mask) { + chan = net_dma->channels[chan_idx]; + + n = ((num_online_cpus() / cpus_weight(net_dma->channel_mask)) + + (i < (num_online_cpus() % + cpus_weight(net_dma->channel_mask)) ? 1 : 0)); while(n) { per_cpu(softnet_data, cpu).net_dma = chan; @@ -3589,7 +3606,6 @@ static void net_dma_rebalance(void) } i++; } - rcu_read_unlock(); } /** @@ -3598,23 +3614,53 @@ static void net_dma_rebalance(void) * @chan: DMA channel for the event * @event: event type */ -static void netdev_dma_event(struct dma_client *client, struct dma_chan *chan, - enum dma_event event) -{ - spin_lock(&net_dma_event_lock); - switch (event) { - case DMA_RESOURCE_ADDED: - net_dma_count++; - net_dma_rebalance(); +static enum dma_state_client +netdev_dma_event(struct dma_client *client, struct dma_chan *chan, + enum dma_state state) +{ + int i, found = 0, pos = -1; + struct net_dma *net_dma = + container_of(client, struct net_dma, client); + enum dma_state_client ack = DMA_DUP; /* default: take no action */ + + spin_lock(&net_dma->lock); + switch (state) { + case DMA_RESOURCE_AVAILABLE: + for (i = 0; i < NR_CPUS; i++) + if (net_dma->channels[i] == chan) { + found = 1; + break; + } else if (net_dma->channels[i] == NULL && pos < 0) + pos = i; + + if (!found && pos >= 0) { + ack = DMA_ACK; + net_dma->channels[pos] = chan; + cpu_set(pos, net_dma->channel_mask); + net_dma_rebalance(net_dma); + } break; case DMA_RESOURCE_REMOVED: - net_dma_count--; - net_dma_rebalance(); + for (i = 0; i < NR_CPUS; i++) + if (net_dma->channels[i] == chan) { + found = 1; + pos = i; + break; + } + + if (found) { + ack = DMA_ACK; + cpu_clear(pos, net_dma->channel_mask); + net_dma->channels[i] = NULL; + net_dma_rebalance(net_dma); + } break; default: break; } - spin_unlock(&net_dma_event_lock); + spin_unlock(&net_dma->lock); + + return ack; } /** @@ -3622,12 +3668,10 @@ static void netdev_dma_event(struct dma_client *client, struct dma_chan *chan, */ static int __init netdev_dma_register(void) { - spin_lock_init(&net_dma_event_lock); - net_dma_client = dma_async_client_register(netdev_dma_event); - if (net_dma_client == NULL) - return -ENOMEM; - - dma_async_client_chan_request(net_dma_client, num_online_cpus()); + spin_lock_init(&net_dma.lock); + dma_cap_set(DMA_MEMCPY, net_dma.client.cap_mask); + dma_async_client_register(&net_dma.client); + dma_async_client_chan_request(&net_dma.client); return 0; } -- cgit v1.1 From 685784aaf3cd0e3ff5e36c7ecf6f441cdbf57f73 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Mon, 9 Jul 2007 11:56:42 -0700 Subject: xor: make 'xor_blocks' a library routine for use with async_tx The async_tx api tries to use a dma engine for an operation, but will fall back to an optimized software routine otherwise. Xor support is implemented using the raid5 xor routines. For organizational purposes this routine is moved to a common area. The following fixes are also made: * rename xor_block => xor_blocks, suggested by Adrian Bunk * ensure that xor.o initializes before md.o in the built-in case * checkpatch.pl fixes * mark calibrate_xor_blocks __init, Adrian Bunk Cc: Adrian Bunk Cc: NeilBrown Cc: Herbert Xu Signed-off-by: Dan Williams --- crypto/Kconfig | 6 ++ crypto/Makefile | 6 ++ crypto/xor.c | 156 +++++++++++++++++++++++++++++++++++++++++++++++ drivers/md/Kconfig | 1 + drivers/md/Makefile | 4 +- drivers/md/md.c | 2 +- drivers/md/raid5.c | 10 +-- drivers/md/xor.c | 154 ---------------------------------------------- include/linux/raid/xor.h | 2 +- 9 files changed, 178 insertions(+), 163 deletions(-) create mode 100644 crypto/xor.c delete mode 100644 drivers/md/xor.c diff --git a/crypto/Kconfig b/crypto/Kconfig index 4ca0ab3..b749a1a 100644 --- a/crypto/Kconfig +++ b/crypto/Kconfig @@ -1,4 +1,10 @@ # +# Generic algorithms support +# +config XOR_BLOCKS + tristate + +# # Cryptographic API Configuration # diff --git a/crypto/Makefile b/crypto/Makefile index cce46a1..68e934b 100644 --- a/crypto/Makefile +++ b/crypto/Makefile @@ -50,3 +50,9 @@ obj-$(CONFIG_CRYPTO_MICHAEL_MIC) += michael_mic.o obj-$(CONFIG_CRYPTO_CRC32C) += crc32c.o obj-$(CONFIG_CRYPTO_TEST) += tcrypt.o + +# +# generic algorithms and the async_tx api +# +obj-$(CONFIG_XOR_BLOCKS) += xor.o + diff --git a/crypto/xor.c b/crypto/xor.c new file mode 100644 index 0000000..8281ac5 --- /dev/null +++ b/crypto/xor.c @@ -0,0 +1,156 @@ +/* + * xor.c : Multiple Devices driver for Linux + * + * Copyright (C) 1996, 1997, 1998, 1999, 2000, + * Ingo Molnar, Matti Aarnio, Jakub Jelinek, Richard Henderson. + * + * Dispatch optimized RAID-5 checksumming functions. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2, or (at your option) + * any later version. + * + * You should have received a copy of the GNU General Public License + * (for example /usr/src/linux/COPYING); if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#define BH_TRACE 0 +#include +#include +#include +#include + +/* The xor routines to use. */ +static struct xor_block_template *active_template; + +void +xor_blocks(unsigned int count, unsigned int bytes, void **ptr) +{ + unsigned long *p0, *p1, *p2, *p3, *p4; + + p0 = (unsigned long *) ptr[0]; + p1 = (unsigned long *) ptr[1]; + if (count == 2) { + active_template->do_2(bytes, p0, p1); + return; + } + + p2 = (unsigned long *) ptr[2]; + if (count == 3) { + active_template->do_3(bytes, p0, p1, p2); + return; + } + + p3 = (unsigned long *) ptr[3]; + if (count == 4) { + active_template->do_4(bytes, p0, p1, p2, p3); + return; + } + + p4 = (unsigned long *) ptr[4]; + active_template->do_5(bytes, p0, p1, p2, p3, p4); +} +EXPORT_SYMBOL(xor_blocks); + +/* Set of all registered templates. */ +static struct xor_block_template *template_list; + +#define BENCH_SIZE (PAGE_SIZE) + +static void +do_xor_speed(struct xor_block_template *tmpl, void *b1, void *b2) +{ + int speed; + unsigned long now; + int i, count, max; + + tmpl->next = template_list; + template_list = tmpl; + + /* + * Count the number of XORs done during a whole jiffy, and use + * this to calculate the speed of checksumming. We use a 2-page + * allocation to have guaranteed color L1-cache layout. + */ + max = 0; + for (i = 0; i < 5; i++) { + now = jiffies; + count = 0; + while (jiffies == now) { + mb(); /* prevent loop optimzation */ + tmpl->do_2(BENCH_SIZE, b1, b2); + mb(); + count++; + mb(); + } + if (count > max) + max = count; + } + + speed = max * (HZ * BENCH_SIZE / 1024); + tmpl->speed = speed; + + printk(KERN_INFO " %-10s: %5d.%03d MB/sec\n", tmpl->name, + speed / 1000, speed % 1000); +} + +static int __init +calibrate_xor_blocks(void) +{ + void *b1, *b2; + struct xor_block_template *f, *fastest; + + b1 = (void *) __get_free_pages(GFP_KERNEL, 2); + if (!b1) { + printk(KERN_WARNING "xor: Yikes! No memory available.\n"); + return -ENOMEM; + } + b2 = b1 + 2*PAGE_SIZE + BENCH_SIZE; + + /* + * If this arch/cpu has a short-circuited selection, don't loop through + * all the possible functions, just test the best one + */ + + fastest = NULL; + +#ifdef XOR_SELECT_TEMPLATE + fastest = XOR_SELECT_TEMPLATE(fastest); +#endif + +#define xor_speed(templ) do_xor_speed((templ), b1, b2) + + if (fastest) { + printk(KERN_INFO "xor: automatically using best " + "checksumming function: %s\n", + fastest->name); + xor_speed(fastest); + } else { + printk(KERN_INFO "xor: measuring checksumming speed\n"); + XOR_TRY_TEMPLATES; + fastest = template_list; + for (f = fastest; f; f = f->next) + if (f->speed > fastest->speed) + fastest = f; + } + + printk(KERN_INFO "xor: using function: %s (%d.%03d MB/sec)\n", + fastest->name, fastest->speed / 1000, fastest->speed % 1000); + +#undef xor_speed + + free_pages((unsigned long)b1, 2); + + active_template = fastest; + return 0; +} + +static __exit void xor_exit(void) { } + +MODULE_LICENSE("GPL"); + +/* when built-in xor.o must initialize before drivers/md/md.o */ +core_initcall(calibrate_xor_blocks); +module_exit(xor_exit); diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig index 7df934d..24d93d0 100644 --- a/drivers/md/Kconfig +++ b/drivers/md/Kconfig @@ -109,6 +109,7 @@ config MD_RAID10 config MD_RAID456 tristate "RAID-4/RAID-5/RAID-6 mode" depends on BLK_DEV_MD + select XOR_BLOCKS ---help--- A RAID-5 set of N drives with a capacity of C MB per drive provides the capacity of C * (N - 1) MB, and protects against a failure diff --git a/drivers/md/Makefile b/drivers/md/Makefile index 3875408..71eb45f 100644 --- a/drivers/md/Makefile +++ b/drivers/md/Makefile @@ -17,7 +17,7 @@ raid456-objs := raid5.o raid6algos.o raid6recov.o raid6tables.o \ hostprogs-y := mktables # Note: link order is important. All raid personalities -# and xor.o must come before md.o, as they each initialise +# and must come before md.o, as they each initialise # themselves, and md.o may use the personalities when it # auto-initialised. @@ -25,7 +25,7 @@ obj-$(CONFIG_MD_LINEAR) += linear.o obj-$(CONFIG_MD_RAID0) += raid0.o obj-$(CONFIG_MD_RAID1) += raid1.o obj-$(CONFIG_MD_RAID10) += raid10.o -obj-$(CONFIG_MD_RAID456) += raid456.o xor.o +obj-$(CONFIG_MD_RAID456) += raid456.o obj-$(CONFIG_MD_MULTIPATH) += multipath.o obj-$(CONFIG_MD_FAULTY) += faulty.o obj-$(CONFIG_BLK_DEV_MD) += md-mod.o diff --git a/drivers/md/md.c b/drivers/md/md.c index 1c54f3c..33beaa7 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5814,7 +5814,7 @@ static __exit void md_exit(void) } } -module_init(md_init) +subsys_initcall(md_init); module_exit(md_exit) static int get_ro(char *buffer, struct kernel_param *kp) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 061375e..5adbe0b 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -918,7 +918,7 @@ static void copy_data(int frombio, struct bio *bio, #define check_xor() do { \ if (count == MAX_XOR_BLOCKS) { \ - xor_block(count, STRIPE_SIZE, ptr); \ + xor_blocks(count, STRIPE_SIZE, ptr); \ count = 1; \ } \ } while(0) @@ -949,7 +949,7 @@ static void compute_block(struct stripe_head *sh, int dd_idx) check_xor(); } if (count != 1) - xor_block(count, STRIPE_SIZE, ptr); + xor_blocks(count, STRIPE_SIZE, ptr); set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); } @@ -1004,7 +1004,7 @@ static void compute_parity5(struct stripe_head *sh, int method) break; } if (count>1) { - xor_block(count, STRIPE_SIZE, ptr); + xor_blocks(count, STRIPE_SIZE, ptr); count = 1; } @@ -1038,7 +1038,7 @@ static void compute_parity5(struct stripe_head *sh, int method) } } if (count != 1) - xor_block(count, STRIPE_SIZE, ptr); + xor_blocks(count, STRIPE_SIZE, ptr); if (method != CHECK_PARITY) { set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); @@ -1160,7 +1160,7 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) check_xor(); } if (count != 1) - xor_block(count, STRIPE_SIZE, ptr); + xor_blocks(count, STRIPE_SIZE, ptr); if (!nozero) set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); else clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); } diff --git a/drivers/md/xor.c b/drivers/md/xor.c deleted file mode 100644 index 324897c..0000000 --- a/drivers/md/xor.c +++ /dev/null @@ -1,154 +0,0 @@ -/* - * xor.c : Multiple Devices driver for Linux - * - * Copyright (C) 1996, 1997, 1998, 1999, 2000, - * Ingo Molnar, Matti Aarnio, Jakub Jelinek, Richard Henderson. - * - * Dispatch optimized RAID-5 checksumming functions. - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2, or (at your option) - * any later version. - * - * You should have received a copy of the GNU General Public License - * (for example /usr/src/linux/COPYING); if not, write to the Free - * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. - */ - -#define BH_TRACE 0 -#include -#include -#include -#include - -/* The xor routines to use. */ -static struct xor_block_template *active_template; - -void -xor_block(unsigned int count, unsigned int bytes, void **ptr) -{ - unsigned long *p0, *p1, *p2, *p3, *p4; - - p0 = (unsigned long *) ptr[0]; - p1 = (unsigned long *) ptr[1]; - if (count == 2) { - active_template->do_2(bytes, p0, p1); - return; - } - - p2 = (unsigned long *) ptr[2]; - if (count == 3) { - active_template->do_3(bytes, p0, p1, p2); - return; - } - - p3 = (unsigned long *) ptr[3]; - if (count == 4) { - active_template->do_4(bytes, p0, p1, p2, p3); - return; - } - - p4 = (unsigned long *) ptr[4]; - active_template->do_5(bytes, p0, p1, p2, p3, p4); -} - -/* Set of all registered templates. */ -static struct xor_block_template *template_list; - -#define BENCH_SIZE (PAGE_SIZE) - -static void -do_xor_speed(struct xor_block_template *tmpl, void *b1, void *b2) -{ - int speed; - unsigned long now; - int i, count, max; - - tmpl->next = template_list; - template_list = tmpl; - - /* - * Count the number of XORs done during a whole jiffy, and use - * this to calculate the speed of checksumming. We use a 2-page - * allocation to have guaranteed color L1-cache layout. - */ - max = 0; - for (i = 0; i < 5; i++) { - now = jiffies; - count = 0; - while (jiffies == now) { - mb(); - tmpl->do_2(BENCH_SIZE, b1, b2); - mb(); - count++; - mb(); - } - if (count > max) - max = count; - } - - speed = max * (HZ * BENCH_SIZE / 1024); - tmpl->speed = speed; - - printk(" %-10s: %5d.%03d MB/sec\n", tmpl->name, - speed / 1000, speed % 1000); -} - -static int -calibrate_xor_block(void) -{ - void *b1, *b2; - struct xor_block_template *f, *fastest; - - b1 = (void *) __get_free_pages(GFP_KERNEL, 2); - if (! b1) { - printk("raid5: Yikes! No memory available.\n"); - return -ENOMEM; - } - b2 = b1 + 2*PAGE_SIZE + BENCH_SIZE; - - /* - * If this arch/cpu has a short-circuited selection, don't loop through all - * the possible functions, just test the best one - */ - - fastest = NULL; - -#ifdef XOR_SELECT_TEMPLATE - fastest = XOR_SELECT_TEMPLATE(fastest); -#endif - -#define xor_speed(templ) do_xor_speed((templ), b1, b2) - - if (fastest) { - printk(KERN_INFO "raid5: automatically using best checksumming function: %s\n", - fastest->name); - xor_speed(fastest); - } else { - printk(KERN_INFO "raid5: measuring checksumming speed\n"); - XOR_TRY_TEMPLATES; - fastest = template_list; - for (f = fastest; f; f = f->next) - if (f->speed > fastest->speed) - fastest = f; - } - - printk("raid5: using function: %s (%d.%03d MB/sec)\n", - fastest->name, fastest->speed / 1000, fastest->speed % 1000); - -#undef xor_speed - - free_pages((unsigned long)b1, 2); - - active_template = fastest; - return 0; -} - -static __exit void xor_exit(void) { } - -EXPORT_SYMBOL(xor_block); -MODULE_LICENSE("GPL"); - -module_init(calibrate_xor_block); -module_exit(xor_exit); diff --git a/include/linux/raid/xor.h b/include/linux/raid/xor.h index f0d67cb..7d6c20b 100644 --- a/include/linux/raid/xor.h +++ b/include/linux/raid/xor.h @@ -5,7 +5,7 @@ #define MAX_XOR_BLOCKS 5 -extern void xor_block(unsigned int count, unsigned int bytes, void **ptr); +extern void xor_blocks(unsigned int count, unsigned int bytes, void **ptr); struct xor_block_template { struct xor_block_template *next; -- cgit v1.1 From 9bc89cd82d6f88fb0ca39b30445c329a430fd66b Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 11:10:44 -0700 Subject: async_tx: add the async_tx api The async_tx api provides methods for describing a chain of asynchronous bulk memory transfers/transforms with support for inter-transactional dependencies. It is implemented as a dmaengine client that smooths over the details of different hardware offload engine implementations. Code that is written to the api can optimize for asynchronous operation and the api will fit the chain of operations to the available offload resources. I imagine that any piece of ADMA hardware would register with the 'async_*' subsystem, and a call to async_X would be routed as appropriate, or be run in-line. - Neil Brown async_tx exploits the capabilities of struct dma_async_tx_descriptor to provide an api of the following general format: struct dma_async_tx_descriptor * async_(..., struct dma_async_tx_descriptor *depend_tx, dma_async_tx_callback cb_fn, void *cb_param) { struct dma_chan *chan = async_tx_find_channel(depend_tx, ); struct dma_device *device = chan ? chan->device : NULL; int int_en = cb_fn ? 1 : 0; struct dma_async_tx_descriptor *tx = device ? device->device_prep_dma_(chan, len, int_en) : NULL; if (tx) { /* run asynchronously */ ... tx->tx_set_dest(addr, tx, index); ... tx->tx_set_src(addr, tx, index); ... async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); } else { /* run synchronously */ ... ... async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); } return tx; } async_tx_find_channel() returns a capable channel from its pool. The channel pool is organized as a per-cpu array of channel pointers. The async_tx_rebalance() routine is tasked with managing these arrays. In the uniprocessor case async_tx_rebalance() tries to spread responsibility evenly over channels of similar capabilities. For example if there are two copy+xor channels, one will handle copy operations and the other will handle xor. In the SMP case async_tx_rebalance() attempts to spread the operations evenly over the cpus, e.g. cpu0 gets copy channel0 and xor channel0 while cpu1 gets copy channel 1 and xor channel 1. When a dependency is specified async_tx_find_channel defaults to keeping the operation on the same channel. A xor->copy->xor chain will stay on one channel if it supports both operation types, otherwise the transaction will transition between a copy and a xor resource. Currently the raid5 implementation in the MD raid456 driver has been converted to the async_tx api. A driver for the offload engines on the Intel Xscale series of I/O processors, iop-adma, is provided in a later commit. With the iop-adma driver and async_tx, raid456 is able to offload copy, xor, and xor-zero-sum operations to hardware engines. On iop342 tiobench showed higher throughput for sequential writes (20 - 30% improvement) and sequential reads to a degraded array (40 - 55% improvement). For the other cases performance was roughly equal, +/- a few percentage points. On a x86-smp platform the performance of the async_tx implementation (in synchronous mode) was also +/- a few percentage points of the original implementation. According to 'top' on iop342 CPU utilization drops from ~50% to ~15% during a 'resync' while the speed according to /proc/mdstat doubles from ~25 MB/s to ~50 MB/s. The tiobench command line used for testing was: tiobench --size 2048 --block 4096 --block 131072 --dir /mnt/raid --numruns 5 * iop342 had 1GB of memory available Details: * if CONFIG_DMA_ENGINE=n the asynchronous path is compiled away by making async_tx_find_channel a static inline routine that always returns NULL * when a callback is specified for a given transaction an interrupt will fire at operation completion time and the callback will occur in a tasklet. if the the channel does not support interrupts then a live polling wait will be performed * the api is written as a dmaengine client that requests all available channels * In support of dependencies the api implicitly schedules channel-switch interrupts. The interrupt triggers the cleanup tasklet which causes pending operations to be scheduled on the next channel * Xor engines treat an xor destination address differently than a software xor routine. To the software routine the destination address is an implied source, whereas engines treat it as a write-only destination. This patch modifies the xor_blocks routine to take a an explicit destination address to mirror the hardware. Changelog: * fixed a leftover debug print * don't allow callbacks in async_interrupt_cond * fixed xor_block changes * fixed usage of ASYNC_TX_XOR_DROP_DEST * drop dma mapping methods, suggested by Chris Leech * printk warning fixups from Andrew Morton * don't use inline in C files, Adrian Bunk * select the API when MD is enabled * BUG_ON xor source counts <= 1 * implicitly handle hardware concerns like channel switching and interrupts, Neil Brown * remove the per operation type list, and distribute operation capabilities evenly amongst the available channels * simplify async_tx_find_channel to optimize the fast path * introduce the channel_table_initialized flag to prevent early calls to the api * reorganize the code to mimic crypto * include mm.h as not all archs include it in dma-mapping.h * make the Kconfig options non-user visible, Adrian Bunk * move async_tx under crypto since it is meant as 'core' functionality, and the two may share algorithms in the future * move large inline functions into c files * checkpatch.pl fixes * gpl v2 only correction Cc: Herbert Xu Signed-off-by: Dan Williams Acked-By: NeilBrown --- crypto/Kconfig | 6 +- crypto/Makefile | 2 +- crypto/async_tx/Kconfig | 16 ++ crypto/async_tx/Makefile | 4 + crypto/async_tx/async_memcpy.c | 131 +++++++++++ crypto/async_tx/async_memset.c | 109 +++++++++ crypto/async_tx/async_tx.c | 497 +++++++++++++++++++++++++++++++++++++++++ crypto/async_tx/async_xor.c | 327 +++++++++++++++++++++++++++ crypto/xor.c | 29 ++- drivers/dma/Kconfig | 5 +- drivers/md/Kconfig | 3 +- drivers/md/raid5.c | 54 ++--- include/linux/async_tx.h | 156 +++++++++++++ include/linux/raid/xor.h | 5 +- 14 files changed, 1294 insertions(+), 50 deletions(-) create mode 100644 crypto/async_tx/Kconfig create mode 100644 crypto/async_tx/Makefile create mode 100644 crypto/async_tx/async_memcpy.c create mode 100644 crypto/async_tx/async_memset.c create mode 100644 crypto/async_tx/async_tx.c create mode 100644 crypto/async_tx/async_xor.c create mode 100644 include/linux/async_tx.h diff --git a/crypto/Kconfig b/crypto/Kconfig index b749a1a..07090e9 100644 --- a/crypto/Kconfig +++ b/crypto/Kconfig @@ -5,9 +5,13 @@ config XOR_BLOCKS tristate # -# Cryptographic API Configuration +# async_tx api: hardware offloaded memory transfer/transform support # +source "crypto/async_tx/Kconfig" +# +# Cryptographic API Configuration +# menu "Cryptographic options" config CRYPTO diff --git a/crypto/Makefile b/crypto/Makefile index 68e934b..0cf17f1 100644 --- a/crypto/Makefile +++ b/crypto/Makefile @@ -55,4 +55,4 @@ obj-$(CONFIG_CRYPTO_TEST) += tcrypt.o # generic algorithms and the async_tx api # obj-$(CONFIG_XOR_BLOCKS) += xor.o - +obj-$(CONFIG_ASYNC_CORE) += async_tx/ diff --git a/crypto/async_tx/Kconfig b/crypto/async_tx/Kconfig new file mode 100644 index 0000000..d8fb3914 --- /dev/null +++ b/crypto/async_tx/Kconfig @@ -0,0 +1,16 @@ +config ASYNC_CORE + tristate + +config ASYNC_MEMCPY + tristate + select ASYNC_CORE + +config ASYNC_XOR + tristate + select ASYNC_CORE + select XOR_BLOCKS + +config ASYNC_MEMSET + tristate + select ASYNC_CORE + diff --git a/crypto/async_tx/Makefile b/crypto/async_tx/Makefile new file mode 100644 index 0000000..27baa7d --- /dev/null +++ b/crypto/async_tx/Makefile @@ -0,0 +1,4 @@ +obj-$(CONFIG_ASYNC_CORE) += async_tx.o +obj-$(CONFIG_ASYNC_MEMCPY) += async_memcpy.o +obj-$(CONFIG_ASYNC_MEMSET) += async_memset.o +obj-$(CONFIG_ASYNC_XOR) += async_xor.o diff --git a/crypto/async_tx/async_memcpy.c b/crypto/async_tx/async_memcpy.c new file mode 100644 index 0000000..a973f4e --- /dev/null +++ b/crypto/async_tx/async_memcpy.c @@ -0,0 +1,131 @@ +/* + * copy offload engine support + * + * Copyright © 2006, Intel Corporation. + * + * Dan Williams + * + * with architecture considerations by: + * Neil Brown + * Jeff Garzik + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#include +#include +#include +#include +#include + +/** + * async_memcpy - attempt to copy memory with a dma engine. + * @dest: destination page + * @src: src page + * @offset: offset in pages to start transaction + * @len: length in bytes + * @flags: ASYNC_TX_ASSUME_COHERENT, ASYNC_TX_ACK, ASYNC_TX_DEP_ACK, + * ASYNC_TX_KMAP_SRC, ASYNC_TX_KMAP_DST + * @depend_tx: memcpy depends on the result of this transaction + * @cb_fn: function to call when the memcpy completes + * @cb_param: parameter to pass to the callback routine + */ +struct dma_async_tx_descriptor * +async_memcpy(struct page *dest, struct page *src, unsigned int dest_offset, + unsigned int src_offset, size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_MEMCPY); + struct dma_device *device = chan ? chan->device : NULL; + int int_en = cb_fn ? 1 : 0; + struct dma_async_tx_descriptor *tx = device ? + device->device_prep_dma_memcpy(chan, len, + int_en) : NULL; + + if (tx) { /* run the memcpy asynchronously */ + dma_addr_t addr; + enum dma_data_direction dir; + + pr_debug("%s: (async) len: %zu\n", __FUNCTION__, len); + + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_FROM_DEVICE; + + addr = dma_map_page(device->dev, dest, dest_offset, len, dir); + tx->tx_set_dest(addr, tx, 0); + + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_TO_DEVICE; + + addr = dma_map_page(device->dev, src, src_offset, len, dir); + tx->tx_set_src(addr, tx, 0); + + async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); + } else { /* run the memcpy synchronously */ + void *dest_buf, *src_buf; + pr_debug("%s: (sync) len: %zu\n", __FUNCTION__, len); + + /* wait for any prerequisite operations */ + if (depend_tx) { + /* if ack is already set then we cannot be sure + * we are referring to the correct operation + */ + BUG_ON(depend_tx->ack); + if (dma_wait_for_async_tx(depend_tx) == DMA_ERROR) + panic("%s: DMA_ERROR waiting for depend_tx\n", + __FUNCTION__); + } + + if (flags & ASYNC_TX_KMAP_DST) + dest_buf = kmap_atomic(dest, KM_USER0) + dest_offset; + else + dest_buf = page_address(dest) + dest_offset; + + if (flags & ASYNC_TX_KMAP_SRC) + src_buf = kmap_atomic(src, KM_USER0) + src_offset; + else + src_buf = page_address(src) + src_offset; + + memcpy(dest_buf, src_buf, len); + + if (flags & ASYNC_TX_KMAP_DST) + kunmap_atomic(dest_buf, KM_USER0); + + if (flags & ASYNC_TX_KMAP_SRC) + kunmap_atomic(src_buf, KM_USER0); + + async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); + } + + return tx; +} +EXPORT_SYMBOL_GPL(async_memcpy); + +static int __init async_memcpy_init(void) +{ + return 0; +} + +static void __exit async_memcpy_exit(void) +{ + do { } while (0); +} + +module_init(async_memcpy_init); +module_exit(async_memcpy_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("asynchronous memcpy api"); +MODULE_LICENSE("GPL"); diff --git a/crypto/async_tx/async_memset.c b/crypto/async_tx/async_memset.c new file mode 100644 index 0000000..66ef635 --- /dev/null +++ b/crypto/async_tx/async_memset.c @@ -0,0 +1,109 @@ +/* + * memory fill offload engine support + * + * Copyright © 2006, Intel Corporation. + * + * Dan Williams + * + * with architecture considerations by: + * Neil Brown + * Jeff Garzik + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#include +#include +#include +#include +#include + +/** + * async_memset - attempt to fill memory with a dma engine. + * @dest: destination page + * @val: fill value + * @offset: offset in pages to start transaction + * @len: length in bytes + * @flags: ASYNC_TX_ASSUME_COHERENT, ASYNC_TX_ACK, ASYNC_TX_DEP_ACK + * @depend_tx: memset depends on the result of this transaction + * @cb_fn: function to call when the memcpy completes + * @cb_param: parameter to pass to the callback routine + */ +struct dma_async_tx_descriptor * +async_memset(struct page *dest, int val, unsigned int offset, + size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_MEMSET); + struct dma_device *device = chan ? chan->device : NULL; + int int_en = cb_fn ? 1 : 0; + struct dma_async_tx_descriptor *tx = device ? + device->device_prep_dma_memset(chan, val, len, + int_en) : NULL; + + if (tx) { /* run the memset asynchronously */ + dma_addr_t dma_addr; + enum dma_data_direction dir; + + pr_debug("%s: (async) len: %zu\n", __FUNCTION__, len); + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_FROM_DEVICE; + + dma_addr = dma_map_page(device->dev, dest, offset, len, dir); + tx->tx_set_dest(dma_addr, tx, 0); + + async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); + } else { /* run the memset synchronously */ + void *dest_buf; + pr_debug("%s: (sync) len: %zu\n", __FUNCTION__, len); + + dest_buf = (void *) (((char *) page_address(dest)) + offset); + + /* wait for any prerequisite operations */ + if (depend_tx) { + /* if ack is already set then we cannot be sure + * we are referring to the correct operation + */ + BUG_ON(depend_tx->ack); + if (dma_wait_for_async_tx(depend_tx) == DMA_ERROR) + panic("%s: DMA_ERROR waiting for depend_tx\n", + __FUNCTION__); + } + + memset(dest_buf, val, len); + + async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); + } + + return tx; +} +EXPORT_SYMBOL_GPL(async_memset); + +static int __init async_memset_init(void) +{ + return 0; +} + +static void __exit async_memset_exit(void) +{ + do { } while (0); +} + +module_init(async_memset_init); +module_exit(async_memset_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("asynchronous memset api"); +MODULE_LICENSE("GPL"); diff --git a/crypto/async_tx/async_tx.c b/crypto/async_tx/async_tx.c new file mode 100644 index 0000000..0350071 --- /dev/null +++ b/crypto/async_tx/async_tx.c @@ -0,0 +1,497 @@ +/* + * core routines for the asynchronous memory transfer/transform api + * + * Copyright © 2006, Intel Corporation. + * + * Dan Williams + * + * with architecture considerations by: + * Neil Brown + * Jeff Garzik + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#include +#include + +#ifdef CONFIG_DMA_ENGINE +static enum dma_state_client +dma_channel_add_remove(struct dma_client *client, + struct dma_chan *chan, enum dma_state state); + +static struct dma_client async_tx_dma = { + .event_callback = dma_channel_add_remove, + /* .cap_mask == 0 defaults to all channels */ +}; + +/** + * dma_cap_mask_all - enable iteration over all operation types + */ +static dma_cap_mask_t dma_cap_mask_all; + +/** + * chan_ref_percpu - tracks channel allocations per core/opertion + */ +struct chan_ref_percpu { + struct dma_chan_ref *ref; +}; + +static int channel_table_initialized; +static struct chan_ref_percpu *channel_table[DMA_TX_TYPE_END]; + +/** + * async_tx_lock - protect modification of async_tx_master_list and serialize + * rebalance operations + */ +static spinlock_t async_tx_lock; + +static struct list_head +async_tx_master_list = LIST_HEAD_INIT(async_tx_master_list); + +/* async_tx_issue_pending_all - start all transactions on all channels */ +void async_tx_issue_pending_all(void) +{ + struct dma_chan_ref *ref; + + rcu_read_lock(); + list_for_each_entry_rcu(ref, &async_tx_master_list, node) + ref->chan->device->device_issue_pending(ref->chan); + rcu_read_unlock(); +} +EXPORT_SYMBOL_GPL(async_tx_issue_pending_all); + +/* dma_wait_for_async_tx - spin wait for a transcation to complete + * @tx: transaction to wait on + */ +enum dma_status +dma_wait_for_async_tx(struct dma_async_tx_descriptor *tx) +{ + enum dma_status status; + struct dma_async_tx_descriptor *iter; + + if (!tx) + return DMA_SUCCESS; + + /* poll through the dependency chain, return when tx is complete */ + do { + iter = tx; + while (iter->cookie == -EBUSY) + iter = iter->parent; + + status = dma_sync_wait(iter->chan, iter->cookie); + } while (status == DMA_IN_PROGRESS || (iter != tx)); + + return status; +} +EXPORT_SYMBOL_GPL(dma_wait_for_async_tx); + +/* async_tx_run_dependencies - helper routine for dma drivers to process + * (start) dependent operations on their target channel + * @tx: transaction with dependencies + */ +void +async_tx_run_dependencies(struct dma_async_tx_descriptor *tx) +{ + struct dma_async_tx_descriptor *dep_tx, *_dep_tx; + struct dma_device *dev; + struct dma_chan *chan; + + list_for_each_entry_safe(dep_tx, _dep_tx, &tx->depend_list, + depend_node) { + chan = dep_tx->chan; + dev = chan->device; + /* we can't depend on ourselves */ + BUG_ON(chan == tx->chan); + list_del(&dep_tx->depend_node); + tx->tx_submit(dep_tx); + + /* we need to poke the engine as client code does not + * know about dependency submission events + */ + dev->device_issue_pending(chan); + } +} +EXPORT_SYMBOL_GPL(async_tx_run_dependencies); + +static void +free_dma_chan_ref(struct rcu_head *rcu) +{ + struct dma_chan_ref *ref; + ref = container_of(rcu, struct dma_chan_ref, rcu); + kfree(ref); +} + +static void +init_dma_chan_ref(struct dma_chan_ref *ref, struct dma_chan *chan) +{ + INIT_LIST_HEAD(&ref->node); + INIT_RCU_HEAD(&ref->rcu); + ref->chan = chan; + atomic_set(&ref->count, 0); +} + +/** + * get_chan_ref_by_cap - returns the nth channel of the given capability + * defaults to returning the channel with the desired capability and the + * lowest reference count if the index can not be satisfied + * @cap: capability to match + * @index: nth channel desired, passing -1 has the effect of forcing the + * default return value + */ +static struct dma_chan_ref * +get_chan_ref_by_cap(enum dma_transaction_type cap, int index) +{ + struct dma_chan_ref *ret_ref = NULL, *min_ref = NULL, *ref; + + rcu_read_lock(); + list_for_each_entry_rcu(ref, &async_tx_master_list, node) + if (dma_has_cap(cap, ref->chan->device->cap_mask)) { + if (!min_ref) + min_ref = ref; + else if (atomic_read(&ref->count) < + atomic_read(&min_ref->count)) + min_ref = ref; + + if (index-- == 0) { + ret_ref = ref; + break; + } + } + rcu_read_unlock(); + + if (!ret_ref) + ret_ref = min_ref; + + if (ret_ref) + atomic_inc(&ret_ref->count); + + return ret_ref; +} + +/** + * async_tx_rebalance - redistribute the available channels, optimize + * for cpu isolation in the SMP case, and opertaion isolation in the + * uniprocessor case + */ +static void async_tx_rebalance(void) +{ + int cpu, cap, cpu_idx = 0; + unsigned long flags; + + if (!channel_table_initialized) + return; + + spin_lock_irqsave(&async_tx_lock, flags); + + /* undo the last distribution */ + for_each_dma_cap_mask(cap, dma_cap_mask_all) + for_each_possible_cpu(cpu) { + struct dma_chan_ref *ref = + per_cpu_ptr(channel_table[cap], cpu)->ref; + if (ref) { + atomic_set(&ref->count, 0); + per_cpu_ptr(channel_table[cap], cpu)->ref = + NULL; + } + } + + for_each_dma_cap_mask(cap, dma_cap_mask_all) + for_each_online_cpu(cpu) { + struct dma_chan_ref *new; + if (NR_CPUS > 1) + new = get_chan_ref_by_cap(cap, cpu_idx++); + else + new = get_chan_ref_by_cap(cap, -1); + + per_cpu_ptr(channel_table[cap], cpu)->ref = new; + } + + spin_unlock_irqrestore(&async_tx_lock, flags); +} + +static enum dma_state_client +dma_channel_add_remove(struct dma_client *client, + struct dma_chan *chan, enum dma_state state) +{ + unsigned long found, flags; + struct dma_chan_ref *master_ref, *ref; + enum dma_state_client ack = DMA_DUP; /* default: take no action */ + + switch (state) { + case DMA_RESOURCE_AVAILABLE: + found = 0; + rcu_read_lock(); + list_for_each_entry_rcu(ref, &async_tx_master_list, node) + if (ref->chan == chan) { + found = 1; + break; + } + rcu_read_unlock(); + + pr_debug("async_tx: dma resource available [%s]\n", + found ? "old" : "new"); + + if (!found) + ack = DMA_ACK; + else + break; + + /* add the channel to the generic management list */ + master_ref = kmalloc(sizeof(*master_ref), GFP_KERNEL); + if (master_ref) { + /* keep a reference until async_tx is unloaded */ + dma_chan_get(chan); + init_dma_chan_ref(master_ref, chan); + spin_lock_irqsave(&async_tx_lock, flags); + list_add_tail_rcu(&master_ref->node, + &async_tx_master_list); + spin_unlock_irqrestore(&async_tx_lock, + flags); + } else { + printk(KERN_WARNING "async_tx: unable to create" + " new master entry in response to" + " a DMA_RESOURCE_ADDED event" + " (-ENOMEM)\n"); + return 0; + } + + async_tx_rebalance(); + break; + case DMA_RESOURCE_REMOVED: + found = 0; + spin_lock_irqsave(&async_tx_lock, flags); + list_for_each_entry_rcu(ref, &async_tx_master_list, node) + if (ref->chan == chan) { + /* permit backing devices to go away */ + dma_chan_put(ref->chan); + list_del_rcu(&ref->node); + call_rcu(&ref->rcu, free_dma_chan_ref); + found = 1; + break; + } + spin_unlock_irqrestore(&async_tx_lock, flags); + + pr_debug("async_tx: dma resource removed [%s]\n", + found ? "ours" : "not ours"); + + if (found) + ack = DMA_ACK; + else + break; + + async_tx_rebalance(); + break; + case DMA_RESOURCE_SUSPEND: + case DMA_RESOURCE_RESUME: + printk(KERN_WARNING "async_tx: does not support dma channel" + " suspend/resume\n"); + break; + default: + BUG(); + } + + return ack; +} + +static int __init +async_tx_init(void) +{ + enum dma_transaction_type cap; + + spin_lock_init(&async_tx_lock); + bitmap_fill(dma_cap_mask_all.bits, DMA_TX_TYPE_END); + + /* an interrupt will never be an explicit operation type. + * clearing this bit prevents allocation to a slot in 'channel_table' + */ + clear_bit(DMA_INTERRUPT, dma_cap_mask_all.bits); + + for_each_dma_cap_mask(cap, dma_cap_mask_all) { + channel_table[cap] = alloc_percpu(struct chan_ref_percpu); + if (!channel_table[cap]) + goto err; + } + + channel_table_initialized = 1; + dma_async_client_register(&async_tx_dma); + dma_async_client_chan_request(&async_tx_dma); + + printk(KERN_INFO "async_tx: api initialized (async)\n"); + + return 0; +err: + printk(KERN_ERR "async_tx: initialization failure\n"); + + while (--cap >= 0) + free_percpu(channel_table[cap]); + + return 1; +} + +static void __exit async_tx_exit(void) +{ + enum dma_transaction_type cap; + + channel_table_initialized = 0; + + for_each_dma_cap_mask(cap, dma_cap_mask_all) + if (channel_table[cap]) + free_percpu(channel_table[cap]); + + dma_async_client_unregister(&async_tx_dma); +} + +/** + * async_tx_find_channel - find a channel to carry out the operation or let + * the transaction execute synchronously + * @depend_tx: transaction dependency + * @tx_type: transaction type + */ +struct dma_chan * +async_tx_find_channel(struct dma_async_tx_descriptor *depend_tx, + enum dma_transaction_type tx_type) +{ + /* see if we can keep the chain on one channel */ + if (depend_tx && + dma_has_cap(tx_type, depend_tx->chan->device->cap_mask)) + return depend_tx->chan; + else if (likely(channel_table_initialized)) { + struct dma_chan_ref *ref; + int cpu = get_cpu(); + ref = per_cpu_ptr(channel_table[tx_type], cpu)->ref; + put_cpu(); + return ref ? ref->chan : NULL; + } else + return NULL; +} +EXPORT_SYMBOL_GPL(async_tx_find_channel); +#else +static int __init async_tx_init(void) +{ + printk(KERN_INFO "async_tx: api initialized (sync-only)\n"); + return 0; +} + +static void __exit async_tx_exit(void) +{ + do { } while (0); +} +#endif + +void +async_tx_submit(struct dma_chan *chan, struct dma_async_tx_descriptor *tx, + enum async_tx_flags flags, struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + tx->callback = cb_fn; + tx->callback_param = cb_param; + + /* set this new tx to run after depend_tx if: + * 1/ a dependency exists (depend_tx is !NULL) + * 2/ the tx can not be submitted to the current channel + */ + if (depend_tx && depend_tx->chan != chan) { + /* if ack is already set then we cannot be sure + * we are referring to the correct operation + */ + BUG_ON(depend_tx->ack); + + tx->parent = depend_tx; + spin_lock_bh(&depend_tx->lock); + list_add_tail(&tx->depend_node, &depend_tx->depend_list); + if (depend_tx->cookie == 0) { + struct dma_chan *dep_chan = depend_tx->chan; + struct dma_device *dep_dev = dep_chan->device; + dep_dev->device_dependency_added(dep_chan); + } + spin_unlock_bh(&depend_tx->lock); + + /* schedule an interrupt to trigger the channel switch */ + async_trigger_callback(ASYNC_TX_ACK, depend_tx, NULL, NULL); + } else { + tx->parent = NULL; + tx->tx_submit(tx); + } + + if (flags & ASYNC_TX_ACK) + async_tx_ack(tx); + + if (depend_tx && (flags & ASYNC_TX_DEP_ACK)) + async_tx_ack(depend_tx); +} +EXPORT_SYMBOL_GPL(async_tx_submit); + +/** + * async_trigger_callback - schedules the callback function to be run after + * any dependent operations have been completed. + * @flags: ASYNC_TX_ACK, ASYNC_TX_DEP_ACK + * @depend_tx: 'callback' requires the completion of this transaction + * @cb_fn: function to call after depend_tx completes + * @cb_param: parameter to pass to the callback routine + */ +struct dma_async_tx_descriptor * +async_trigger_callback(enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + struct dma_chan *chan; + struct dma_device *device; + struct dma_async_tx_descriptor *tx; + + if (depend_tx) { + chan = depend_tx->chan; + device = chan->device; + + /* see if we can schedule an interrupt + * otherwise poll for completion + */ + if (device && !dma_has_cap(DMA_INTERRUPT, device->cap_mask)) + device = NULL; + + tx = device ? device->device_prep_dma_interrupt(chan) : NULL; + } else + tx = NULL; + + if (tx) { + pr_debug("%s: (async)\n", __FUNCTION__); + + async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); + } else { + pr_debug("%s: (sync)\n", __FUNCTION__); + + /* wait for any prerequisite operations */ + if (depend_tx) { + /* if ack is already set then we cannot be sure + * we are referring to the correct operation + */ + BUG_ON(depend_tx->ack); + if (dma_wait_for_async_tx(depend_tx) == DMA_ERROR) + panic("%s: DMA_ERROR waiting for depend_tx\n", + __FUNCTION__); + } + + async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); + } + + return tx; +} +EXPORT_SYMBOL_GPL(async_trigger_callback); + +module_init(async_tx_init); +module_exit(async_tx_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("Asynchronous Bulk Memory Transactions API"); +MODULE_LICENSE("GPL"); diff --git a/crypto/async_tx/async_xor.c b/crypto/async_tx/async_xor.c new file mode 100644 index 0000000..2575f67 --- /dev/null +++ b/crypto/async_tx/async_xor.c @@ -0,0 +1,327 @@ +/* + * xor offload engine api + * + * Copyright © 2006, Intel Corporation. + * + * Dan Williams + * + * with architecture considerations by: + * Neil Brown + * Jeff Garzik + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#include +#include +#include +#include +#include +#include + +static void +do_async_xor(struct dma_async_tx_descriptor *tx, struct dma_device *device, + struct dma_chan *chan, struct page *dest, struct page **src_list, + unsigned int offset, unsigned int src_cnt, size_t len, + enum async_tx_flags flags, struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + dma_addr_t dma_addr; + enum dma_data_direction dir; + int i; + + pr_debug("%s: len: %zu\n", __FUNCTION__, len); + + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_FROM_DEVICE; + + dma_addr = dma_map_page(device->dev, dest, offset, len, dir); + tx->tx_set_dest(dma_addr, tx, 0); + + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_TO_DEVICE; + + for (i = 0; i < src_cnt; i++) { + dma_addr = dma_map_page(device->dev, src_list[i], + offset, len, dir); + tx->tx_set_src(dma_addr, tx, i); + } + + async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); +} + +static void +do_sync_xor(struct page *dest, struct page **src_list, unsigned int offset, + unsigned int src_cnt, size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + void *_dest; + int i; + + pr_debug("%s: len: %zu\n", __FUNCTION__, len); + + /* reuse the 'src_list' array to convert to buffer pointers */ + for (i = 0; i < src_cnt; i++) + src_list[i] = (struct page *) + (page_address(src_list[i]) + offset); + + /* set destination address */ + _dest = page_address(dest) + offset; + + if (flags & ASYNC_TX_XOR_ZERO_DST) + memset(_dest, 0, len); + + xor_blocks(src_cnt, len, _dest, + (void **) src_list); + + async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); +} + +/** + * async_xor - attempt to xor a set of blocks with a dma engine. + * xor_blocks always uses the dest as a source so the ASYNC_TX_XOR_ZERO_DST + * flag must be set to not include dest data in the calculation. The + * assumption with dma eninges is that they only use the destination + * buffer as a source when it is explicity specified in the source list. + * @dest: destination page + * @src_list: array of source pages (if the dest is also a source it must be + * at index zero). The contents of this array may be overwritten. + * @offset: offset in pages to start transaction + * @src_cnt: number of source pages + * @len: length in bytes + * @flags: ASYNC_TX_XOR_ZERO_DST, ASYNC_TX_XOR_DROP_DEST, + * ASYNC_TX_ASSUME_COHERENT, ASYNC_TX_ACK, ASYNC_TX_DEP_ACK + * @depend_tx: xor depends on the result of this transaction. + * @cb_fn: function to call when the xor completes + * @cb_param: parameter to pass to the callback routine + */ +struct dma_async_tx_descriptor * +async_xor(struct page *dest, struct page **src_list, unsigned int offset, + int src_cnt, size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_XOR); + struct dma_device *device = chan ? chan->device : NULL; + struct dma_async_tx_descriptor *tx = NULL; + dma_async_tx_callback _cb_fn; + void *_cb_param; + unsigned long local_flags; + int xor_src_cnt; + int i = 0, src_off = 0, int_en; + + BUG_ON(src_cnt <= 1); + + while (src_cnt) { + local_flags = flags; + if (device) { /* run the xor asynchronously */ + xor_src_cnt = min(src_cnt, device->max_xor); + /* if we are submitting additional xors + * only set the callback on the last transaction + */ + if (src_cnt > xor_src_cnt) { + local_flags &= ~ASYNC_TX_ACK; + _cb_fn = NULL; + _cb_param = NULL; + } else { + _cb_fn = cb_fn; + _cb_param = cb_param; + } + + int_en = _cb_fn ? 1 : 0; + + tx = device->device_prep_dma_xor( + chan, xor_src_cnt, len, int_en); + + if (tx) { + do_async_xor(tx, device, chan, dest, + &src_list[src_off], offset, xor_src_cnt, len, + local_flags, depend_tx, _cb_fn, + _cb_param); + } else /* fall through */ + goto xor_sync; + } else { /* run the xor synchronously */ +xor_sync: + /* in the sync case the dest is an implied source + * (assumes the dest is at the src_off index) + */ + if (flags & ASYNC_TX_XOR_DROP_DST) { + src_cnt--; + src_off++; + } + + /* process up to 'MAX_XOR_BLOCKS' sources */ + xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS); + + /* if we are submitting additional xors + * only set the callback on the last transaction + */ + if (src_cnt > xor_src_cnt) { + local_flags &= ~ASYNC_TX_ACK; + _cb_fn = NULL; + _cb_param = NULL; + } else { + _cb_fn = cb_fn; + _cb_param = cb_param; + } + + /* wait for any prerequisite operations */ + if (depend_tx) { + /* if ack is already set then we cannot be sure + * we are referring to the correct operation + */ + BUG_ON(depend_tx->ack); + if (dma_wait_for_async_tx(depend_tx) == + DMA_ERROR) + panic("%s: DMA_ERROR waiting for " + "depend_tx\n", + __FUNCTION__); + } + + do_sync_xor(dest, &src_list[src_off], offset, + xor_src_cnt, len, local_flags, depend_tx, + _cb_fn, _cb_param); + } + + /* the previous tx is hidden from the client, + * so ack it + */ + if (i && depend_tx) + async_tx_ack(depend_tx); + + depend_tx = tx; + + if (src_cnt > xor_src_cnt) { + /* drop completed sources */ + src_cnt -= xor_src_cnt; + src_off += xor_src_cnt; + + /* unconditionally preserve the destination */ + flags &= ~ASYNC_TX_XOR_ZERO_DST; + + /* use the intermediate result a source, but remember + * it's dropped, because it's implied, in the sync case + */ + src_list[--src_off] = dest; + src_cnt++; + flags |= ASYNC_TX_XOR_DROP_DST; + } else + src_cnt = 0; + i++; + } + + return tx; +} +EXPORT_SYMBOL_GPL(async_xor); + +static int page_is_zero(struct page *p, unsigned int offset, size_t len) +{ + char *a = page_address(p) + offset; + return ((*(u32 *) a) == 0 && + memcmp(a, a + 4, len - 4) == 0); +} + +/** + * async_xor_zero_sum - attempt a xor parity check with a dma engine. + * @dest: destination page used if the xor is performed synchronously + * @src_list: array of source pages. The dest page must be listed as a source + * at index zero. The contents of this array may be overwritten. + * @offset: offset in pages to start transaction + * @src_cnt: number of source pages + * @len: length in bytes + * @result: 0 if sum == 0 else non-zero + * @flags: ASYNC_TX_ASSUME_COHERENT, ASYNC_TX_ACK, ASYNC_TX_DEP_ACK + * @depend_tx: xor depends on the result of this transaction. + * @cb_fn: function to call when the xor completes + * @cb_param: parameter to pass to the callback routine + */ +struct dma_async_tx_descriptor * +async_xor_zero_sum(struct page *dest, struct page **src_list, + unsigned int offset, int src_cnt, size_t len, + u32 *result, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_param) +{ + struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_ZERO_SUM); + struct dma_device *device = chan ? chan->device : NULL; + int int_en = cb_fn ? 1 : 0; + struct dma_async_tx_descriptor *tx = device ? + device->device_prep_dma_zero_sum(chan, src_cnt, len, result, + int_en) : NULL; + int i; + + BUG_ON(src_cnt <= 1); + + if (tx) { + dma_addr_t dma_addr; + enum dma_data_direction dir; + + pr_debug("%s: (async) len: %zu\n", __FUNCTION__, len); + + dir = (flags & ASYNC_TX_ASSUME_COHERENT) ? + DMA_NONE : DMA_TO_DEVICE; + + for (i = 0; i < src_cnt; i++) { + dma_addr = dma_map_page(device->dev, src_list[i], + offset, len, dir); + tx->tx_set_src(dma_addr, tx, i); + } + + async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param); + } else { + unsigned long xor_flags = flags; + + pr_debug("%s: (sync) len: %zu\n", __FUNCTION__, len); + + xor_flags |= ASYNC_TX_XOR_DROP_DST; + xor_flags &= ~ASYNC_TX_ACK; + + tx = async_xor(dest, src_list, offset, src_cnt, len, xor_flags, + depend_tx, NULL, NULL); + + if (tx) { + if (dma_wait_for_async_tx(tx) == DMA_ERROR) + panic("%s: DMA_ERROR waiting for tx\n", + __FUNCTION__); + async_tx_ack(tx); + } + + *result = page_is_zero(dest, offset, len) ? 0 : 1; + + tx = NULL; + + async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param); + } + + return tx; +} +EXPORT_SYMBOL_GPL(async_xor_zero_sum); + +static int __init async_xor_init(void) +{ + return 0; +} + +static void __exit async_xor_exit(void) +{ + do { } while (0); +} + +module_init(async_xor_init); +module_exit(async_xor_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("asynchronous xor/xor-zero-sum api"); +MODULE_LICENSE("GPL"); diff --git a/crypto/xor.c b/crypto/xor.c index 8281ac5..b2e6db0 100644 --- a/crypto/xor.c +++ b/crypto/xor.c @@ -26,31 +26,30 @@ static struct xor_block_template *active_template; void -xor_blocks(unsigned int count, unsigned int bytes, void **ptr) +xor_blocks(unsigned int src_count, unsigned int bytes, void *dest, void **srcs) { - unsigned long *p0, *p1, *p2, *p3, *p4; + unsigned long *p1, *p2, *p3, *p4; - p0 = (unsigned long *) ptr[0]; - p1 = (unsigned long *) ptr[1]; - if (count == 2) { - active_template->do_2(bytes, p0, p1); + p1 = (unsigned long *) srcs[0]; + if (src_count == 1) { + active_template->do_2(bytes, dest, p1); return; } - p2 = (unsigned long *) ptr[2]; - if (count == 3) { - active_template->do_3(bytes, p0, p1, p2); + p2 = (unsigned long *) srcs[1]; + if (src_count == 2) { + active_template->do_3(bytes, dest, p1, p2); return; } - p3 = (unsigned long *) ptr[3]; - if (count == 4) { - active_template->do_4(bytes, p0, p1, p2, p3); + p3 = (unsigned long *) srcs[2]; + if (src_count == 3) { + active_template->do_4(bytes, dest, p1, p2, p3); return; } - p4 = (unsigned long *) ptr[4]; - active_template->do_5(bytes, p0, p1, p2, p3, p4); + p4 = (unsigned long *) srcs[3]; + active_template->do_5(bytes, dest, p1, p2, p3, p4); } EXPORT_SYMBOL(xor_blocks); @@ -128,7 +127,7 @@ calibrate_xor_blocks(void) fastest->name); xor_speed(fastest); } else { - printk(KERN_INFO "xor: measuring checksumming speed\n"); + printk(KERN_INFO "xor: measuring software checksum speed\n"); XOR_TRY_TEMPLATES; fastest = template_list; for (f = fastest; f; f = f->next) diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index 72be6c6..492aa08 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -8,8 +8,8 @@ menu "DMA Engine support" config DMA_ENGINE bool "Support for DMA engines" ---help--- - DMA engines offload copy operations from the CPU to dedicated - hardware, allowing the copies to happen asynchronously. + DMA engines offload bulk memory operations from the CPU to dedicated + hardware, allowing the operations to happen asynchronously. comment "DMA Clients" @@ -31,5 +31,4 @@ config INTEL_IOATDMA default m ---help--- Enable support for the Intel(R) I/OAT DMA engine. - endmenu diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig index 24d93d0..bfd9b9c 100644 --- a/drivers/md/Kconfig +++ b/drivers/md/Kconfig @@ -109,7 +109,8 @@ config MD_RAID10 config MD_RAID456 tristate "RAID-4/RAID-5/RAID-6 mode" depends on BLK_DEV_MD - select XOR_BLOCKS + select ASYNC_MEMCPY + select ASYNC_XOR ---help--- A RAID-5 set of N drives with a capacity of C MB per drive provides the capacity of C * (N - 1) MB, and protects against a failure diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 5adbe0b..4f51dfa 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -916,25 +916,25 @@ static void copy_data(int frombio, struct bio *bio, } } -#define check_xor() do { \ - if (count == MAX_XOR_BLOCKS) { \ - xor_blocks(count, STRIPE_SIZE, ptr); \ - count = 1; \ - } \ +#define check_xor() do { \ + if (count == MAX_XOR_BLOCKS) { \ + xor_blocks(count, STRIPE_SIZE, dest, ptr);\ + count = 0; \ + } \ } while(0) static void compute_block(struct stripe_head *sh, int dd_idx) { int i, count, disks = sh->disks; - void *ptr[MAX_XOR_BLOCKS], *p; + void *ptr[MAX_XOR_BLOCKS], *dest, *p; PRINTK("compute_block, stripe %llu, idx %d\n", (unsigned long long)sh->sector, dd_idx); - ptr[0] = page_address(sh->dev[dd_idx].page); - memset(ptr[0], 0, STRIPE_SIZE); - count = 1; + dest = page_address(sh->dev[dd_idx].page); + memset(dest, 0, STRIPE_SIZE); + count = 0; for (i = disks ; i--; ) { if (i == dd_idx) continue; @@ -948,8 +948,8 @@ static void compute_block(struct stripe_head *sh, int dd_idx) check_xor(); } - if (count != 1) - xor_blocks(count, STRIPE_SIZE, ptr); + if (count) + xor_blocks(count, STRIPE_SIZE, dest, ptr); set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); } @@ -957,14 +957,14 @@ static void compute_parity5(struct stripe_head *sh, int method) { raid5_conf_t *conf = sh->raid_conf; int i, pd_idx = sh->pd_idx, disks = sh->disks, count; - void *ptr[MAX_XOR_BLOCKS]; + void *ptr[MAX_XOR_BLOCKS], *dest; struct bio *chosen; PRINTK("compute_parity5, stripe %llu, method %d\n", (unsigned long long)sh->sector, method); - count = 1; - ptr[0] = page_address(sh->dev[pd_idx].page); + count = 0; + dest = page_address(sh->dev[pd_idx].page); switch(method) { case READ_MODIFY_WRITE: BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags)); @@ -987,7 +987,7 @@ static void compute_parity5(struct stripe_head *sh, int method) } break; case RECONSTRUCT_WRITE: - memset(ptr[0], 0, STRIPE_SIZE); + memset(dest, 0, STRIPE_SIZE); for (i= disks; i-- ;) if (i!=pd_idx && sh->dev[i].towrite) { chosen = sh->dev[i].towrite; @@ -1003,9 +1003,9 @@ static void compute_parity5(struct stripe_head *sh, int method) case CHECK_PARITY: break; } - if (count>1) { - xor_blocks(count, STRIPE_SIZE, ptr); - count = 1; + if (count) { + xor_blocks(count, STRIPE_SIZE, dest, ptr); + count = 0; } for (i = disks; i--;) @@ -1037,9 +1037,9 @@ static void compute_parity5(struct stripe_head *sh, int method) check_xor(); } } - if (count != 1) - xor_blocks(count, STRIPE_SIZE, ptr); - + if (count) + xor_blocks(count, STRIPE_SIZE, dest, ptr); + if (method != CHECK_PARITY) { set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); set_bit(R5_LOCKED, &sh->dev[pd_idx].flags); @@ -1132,7 +1132,7 @@ static void compute_parity6(struct stripe_head *sh, int method) static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) { int i, count, disks = sh->disks; - void *ptr[MAX_XOR_BLOCKS], *p; + void *ptr[MAX_XOR_BLOCKS], *dest, *p; int pd_idx = sh->pd_idx; int qd_idx = raid6_next_disk(pd_idx, disks); @@ -1143,9 +1143,9 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) /* We're actually computing the Q drive */ compute_parity6(sh, UPDATE_PARITY); } else { - ptr[0] = page_address(sh->dev[dd_idx].page); - if (!nozero) memset(ptr[0], 0, STRIPE_SIZE); - count = 1; + dest = page_address(sh->dev[dd_idx].page); + if (!nozero) memset(dest, 0, STRIPE_SIZE); + count = 0; for (i = disks ; i--; ) { if (i == dd_idx || i == qd_idx) continue; @@ -1159,8 +1159,8 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) check_xor(); } - if (count != 1) - xor_blocks(count, STRIPE_SIZE, ptr); + if (count) + xor_blocks(count, STRIPE_SIZE, dest, ptr); if (!nozero) set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); else clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); } diff --git a/include/linux/async_tx.h b/include/linux/async_tx.h new file mode 100644 index 0000000..ff12550 --- /dev/null +++ b/include/linux/async_tx.h @@ -0,0 +1,156 @@ +/* + * Copyright © 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#ifndef _ASYNC_TX_H_ +#define _ASYNC_TX_H_ +#include +#include +#include + +/** + * dma_chan_ref - object used to manage dma channels received from the + * dmaengine core. + * @chan - the channel being tracked + * @node - node for the channel to be placed on async_tx_master_list + * @rcu - for list_del_rcu + * @count - number of times this channel is listed in the pool + * (for channels with multiple capabiities) + */ +struct dma_chan_ref { + struct dma_chan *chan; + struct list_head node; + struct rcu_head rcu; + atomic_t count; +}; + +/** + * async_tx_flags - modifiers for the async_* calls + * @ASYNC_TX_XOR_ZERO_DST: this flag must be used for xor operations where the + * the destination address is not a source. The asynchronous case handles this + * implicitly, the synchronous case needs to zero the destination block. + * @ASYNC_TX_XOR_DROP_DST: this flag must be used if the destination address is + * also one of the source addresses. In the synchronous case the destination + * address is an implied source, whereas the asynchronous case it must be listed + * as a source. The destination address must be the first address in the source + * array. + * @ASYNC_TX_ASSUME_COHERENT: skip cache maintenance operations + * @ASYNC_TX_ACK: immediately ack the descriptor, precludes setting up a + * dependency chain + * @ASYNC_TX_DEP_ACK: ack the dependency descriptor. Useful for chaining. + * @ASYNC_TX_KMAP_SRC: if the transaction is to be performed synchronously + * take an atomic mapping (KM_USER0) on the source page(s) + * @ASYNC_TX_KMAP_DST: if the transaction is to be performed synchronously + * take an atomic mapping (KM_USER0) on the dest page(s) + */ +enum async_tx_flags { + ASYNC_TX_XOR_ZERO_DST = (1 << 0), + ASYNC_TX_XOR_DROP_DST = (1 << 1), + ASYNC_TX_ASSUME_COHERENT = (1 << 2), + ASYNC_TX_ACK = (1 << 3), + ASYNC_TX_DEP_ACK = (1 << 4), + ASYNC_TX_KMAP_SRC = (1 << 5), + ASYNC_TX_KMAP_DST = (1 << 6), +}; + +#ifdef CONFIG_DMA_ENGINE +void async_tx_issue_pending_all(void); +enum dma_status dma_wait_for_async_tx(struct dma_async_tx_descriptor *tx); +void async_tx_run_dependencies(struct dma_async_tx_descriptor *tx); +struct dma_chan * +async_tx_find_channel(struct dma_async_tx_descriptor *depend_tx, + enum dma_transaction_type tx_type); +#else +static inline void async_tx_issue_pending_all(void) +{ + do { } while (0); +} + +static inline enum dma_status +dma_wait_for_async_tx(struct dma_async_tx_descriptor *tx) +{ + return DMA_SUCCESS; +} + +static inline void +async_tx_run_dependencies(struct dma_async_tx_descriptor *tx, + struct dma_chan *host_chan) +{ + do { } while (0); +} + +static inline struct dma_chan * +async_tx_find_channel(struct dma_async_tx_descriptor *depend_tx, + enum dma_transaction_type tx_type) +{ + return NULL; +} +#endif + +/** + * async_tx_sync_epilog - actions to take if an operation is run synchronously + * @flags: async_tx flags + * @depend_tx: transaction depends on depend_tx + * @cb_fn: function to call when the transaction completes + * @cb_fn_param: parameter to pass to the callback routine + */ +static inline void +async_tx_sync_epilog(unsigned long flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param) +{ + if (cb_fn) + cb_fn(cb_fn_param); + + if (depend_tx && (flags & ASYNC_TX_DEP_ACK)) + async_tx_ack(depend_tx); +} + +void +async_tx_submit(struct dma_chan *chan, struct dma_async_tx_descriptor *tx, + enum async_tx_flags flags, struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); + +struct dma_async_tx_descriptor * +async_xor(struct page *dest, struct page **src_list, unsigned int offset, + int src_cnt, size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); + +struct dma_async_tx_descriptor * +async_xor_zero_sum(struct page *dest, struct page **src_list, + unsigned int offset, int src_cnt, size_t len, + u32 *result, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); + +struct dma_async_tx_descriptor * +async_memcpy(struct page *dest, struct page *src, unsigned int dest_offset, + unsigned int src_offset, size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); + +struct dma_async_tx_descriptor * +async_memset(struct page *dest, int val, unsigned int offset, + size_t len, enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); + +struct dma_async_tx_descriptor * +async_trigger_callback(enum async_tx_flags flags, + struct dma_async_tx_descriptor *depend_tx, + dma_async_tx_callback cb_fn, void *cb_fn_param); +#endif /* _ASYNC_TX_H_ */ diff --git a/include/linux/raid/xor.h b/include/linux/raid/xor.h index 7d6c20b..3e12058 100644 --- a/include/linux/raid/xor.h +++ b/include/linux/raid/xor.h @@ -3,9 +3,10 @@ #include -#define MAX_XOR_BLOCKS 5 +#define MAX_XOR_BLOCKS 4 -extern void xor_blocks(unsigned int count, unsigned int bytes, void **ptr); +extern void xor_blocks(unsigned int count, unsigned int bytes, + void *dest, void **srcs); struct xor_block_template { struct xor_block_template *next; -- cgit v1.1 From a445685647e825c713175d180ffc8dd54d90589b Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Mon, 9 Jul 2007 11:56:43 -0700 Subject: raid5: refactor handle_stripe5 and handle_stripe6 (v3) handle_stripe5 and handle_stripe6 have very deep logic paths handling the various states of a stripe_head. By introducing the 'stripe_head_state' and 'r6_state' objects, large portions of the logic can be moved to sub-routines. 'struct stripe_head_state' consumes all of the automatic variables that previously stood alone in handle_stripe5,6. 'struct r6_state' contains the handle_stripe6 specific variables like p_failed and q_failed. One of the nice side effects of the 'stripe_head_state' change is that it allows for further reductions in code duplication between raid5 and raid6. The following new routines are shared between raid5 and raid6: handle_completed_write_requests handle_requests_to_failed_array handle_stripe_expansion Changes: * v2: fixed 'conf->raid_disk-1' for the raid6 'handle_stripe_expansion' path * v3: removed the unused 'dirty' field from struct stripe_head_state * v3: coalesced open coded bi_end_io routines into return_io() Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 1526 +++++++++++++++++++++----------------------- include/linux/raid/raid5.h | 16 + 2 files changed, 756 insertions(+), 786 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 4f51dfa..38232fa 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -104,6 +104,23 @@ static inline int raid6_next_disk(int disk, int raid_disks) disk++; return (disk < raid_disks) ? disk : 0; } + +static void return_io(struct bio *return_bi) +{ + struct bio *bi = return_bi; + while (bi) { + int bytes = bi->bi_size; + + return_bi = bi->bi_next; + bi->bi_next = NULL; + bi->bi_size = 0; + bi->bi_end_io(bi, bytes, + test_bit(BIO_UPTODATE, &bi->bi_flags) + ? 0 : -EIO); + bi = return_bi; + } +} + static void print_raid5_conf (raid5_conf_t *conf); static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh) @@ -1326,6 +1343,608 @@ static int stripe_to_pdidx(sector_t stripe, raid5_conf_t *conf, int disks) return pd_idx; } +static void +handle_requests_to_failed_array(raid5_conf_t *conf, struct stripe_head *sh, + struct stripe_head_state *s, int disks, + struct bio **return_bi) +{ + int i; + for (i = disks; i--; ) { + struct bio *bi; + int bitmap_end = 0; + + if (test_bit(R5_ReadError, &sh->dev[i].flags)) { + mdk_rdev_t *rdev; + rcu_read_lock(); + rdev = rcu_dereference(conf->disks[i].rdev); + if (rdev && test_bit(In_sync, &rdev->flags)) + /* multiple read failures in one stripe */ + md_error(conf->mddev, rdev); + rcu_read_unlock(); + } + spin_lock_irq(&conf->device_lock); + /* fail all writes first */ + bi = sh->dev[i].towrite; + sh->dev[i].towrite = NULL; + if (bi) { + s->to_write--; + bitmap_end = 1; + } + + if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) + wake_up(&conf->wait_for_overlap); + + while (bi && bi->bi_sector < + sh->dev[i].sector + STRIPE_SECTORS) { + struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector); + clear_bit(BIO_UPTODATE, &bi->bi_flags); + if (--bi->bi_phys_segments == 0) { + md_write_end(conf->mddev); + bi->bi_next = *return_bi; + *return_bi = bi; + } + bi = nextbi; + } + /* and fail all 'written' */ + bi = sh->dev[i].written; + sh->dev[i].written = NULL; + if (bi) bitmap_end = 1; + while (bi && bi->bi_sector < + sh->dev[i].sector + STRIPE_SECTORS) { + struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector); + clear_bit(BIO_UPTODATE, &bi->bi_flags); + if (--bi->bi_phys_segments == 0) { + md_write_end(conf->mddev); + bi->bi_next = *return_bi; + *return_bi = bi; + } + bi = bi2; + } + + /* fail any reads if this device is non-operational */ + if (!test_bit(R5_Insync, &sh->dev[i].flags) || + test_bit(R5_ReadError, &sh->dev[i].flags)) { + bi = sh->dev[i].toread; + sh->dev[i].toread = NULL; + if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) + wake_up(&conf->wait_for_overlap); + if (bi) s->to_read--; + while (bi && bi->bi_sector < + sh->dev[i].sector + STRIPE_SECTORS) { + struct bio *nextbi = + r5_next_bio(bi, sh->dev[i].sector); + clear_bit(BIO_UPTODATE, &bi->bi_flags); + if (--bi->bi_phys_segments == 0) { + bi->bi_next = *return_bi; + *return_bi = bi; + } + bi = nextbi; + } + } + spin_unlock_irq(&conf->device_lock); + if (bitmap_end) + bitmap_endwrite(conf->mddev->bitmap, sh->sector, + STRIPE_SECTORS, 0, 0); + } + +} + +static void handle_issuing_new_read_requests5(struct stripe_head *sh, + struct stripe_head_state *s, int disks) +{ + int i; + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (!test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && + (dev->toread || + (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) || + s->syncing || s->expanding || + (s->failed && (sh->dev[s->failed_num].toread || + (sh->dev[s->failed_num].towrite && + !test_bit(R5_OVERWRITE, &sh->dev[s->failed_num].flags)) + )))) { + /* we would like to get this block, possibly + * by computing it, but we might not be able to + */ + if (s->uptodate == disks-1) { + PRINTK("Computing block %d\n", i); + compute_block(sh, i); + s->uptodate++; + } else if (test_bit(R5_Insync, &dev->flags)) { + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + s->locked++; + PRINTK("Reading block %d (sync=%d)\n", + i, s->syncing); + } + } + } + set_bit(STRIPE_HANDLE, &sh->state); +} + +static void handle_issuing_new_read_requests6(struct stripe_head *sh, + struct stripe_head_state *s, struct r6_state *r6s, + int disks) +{ + int i; + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (!test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && + (dev->toread || (dev->towrite && + !test_bit(R5_OVERWRITE, &dev->flags)) || + s->syncing || s->expanding || + (s->failed >= 1 && + (sh->dev[r6s->failed_num[0]].toread || + s->to_write)) || + (s->failed >= 2 && + (sh->dev[r6s->failed_num[1]].toread || + s->to_write)))) { + /* we would like to get this block, possibly + * by computing it, but we might not be able to + */ + if (s->uptodate == disks-1) { + PRINTK("Computing stripe %llu block %d\n", + (unsigned long long)sh->sector, i); + compute_block_1(sh, i, 0); + s->uptodate++; + } else if ( s->uptodate == disks-2 && s->failed >= 2 ) { + /* Computing 2-failure is *very* expensive; only + * do it if failed >= 2 + */ + int other; + for (other = disks; other--; ) { + if (other == i) + continue; + if (!test_bit(R5_UPTODATE, + &sh->dev[other].flags)) + break; + } + BUG_ON(other < 0); + PRINTK("Computing stripe %llu blocks %d,%d\n", + (unsigned long long)sh->sector, + i, other); + compute_block_2(sh, i, other); + s->uptodate += 2; + } else if (test_bit(R5_Insync, &dev->flags)) { + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + s->locked++; + PRINTK("Reading block %d (sync=%d)\n", + i, s->syncing); + } + } + } + set_bit(STRIPE_HANDLE, &sh->state); +} + + +/* handle_completed_write_requests + * any written block on an uptodate or failed drive can be returned. + * Note that if we 'wrote' to a failed drive, it will be UPTODATE, but + * never LOCKED, so we don't need to test 'failed' directly. + */ +static void handle_completed_write_requests(raid5_conf_t *conf, + struct stripe_head *sh, int disks, struct bio **return_bi) +{ + int i; + struct r5dev *dev; + + for (i = disks; i--; ) + if (sh->dev[i].written) { + dev = &sh->dev[i]; + if (!test_bit(R5_LOCKED, &dev->flags) && + test_bit(R5_UPTODATE, &dev->flags)) { + /* We can return any write requests */ + struct bio *wbi, *wbi2; + int bitmap_end = 0; + PRINTK("Return write for disc %d\n", i); + spin_lock_irq(&conf->device_lock); + wbi = dev->written; + dev->written = NULL; + while (wbi && wbi->bi_sector < + dev->sector + STRIPE_SECTORS) { + wbi2 = r5_next_bio(wbi, dev->sector); + if (--wbi->bi_phys_segments == 0) { + md_write_end(conf->mddev); + wbi->bi_next = *return_bi; + *return_bi = wbi; + } + wbi = wbi2; + } + if (dev->towrite == NULL) + bitmap_end = 1; + spin_unlock_irq(&conf->device_lock); + if (bitmap_end) + bitmap_endwrite(conf->mddev->bitmap, + sh->sector, + STRIPE_SECTORS, + !test_bit(STRIPE_DEGRADED, &sh->state), + 0); + } + } +} + +static void handle_issuing_new_write_requests5(raid5_conf_t *conf, + struct stripe_head *sh, struct stripe_head_state *s, int disks) +{ + int rmw = 0, rcw = 0, i; + for (i = disks; i--; ) { + /* would I have to read this buffer for read_modify_write */ + struct r5dev *dev = &sh->dev[i]; + if ((dev->towrite || i == sh->pd_idx) && + !test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags)) { + if (test_bit(R5_Insync, &dev->flags)) + rmw++; + else + rmw += 2*disks; /* cannot read it */ + } + /* Would I have to read this buffer for reconstruct_write */ + if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx && + !test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags)) { + if (test_bit(R5_Insync, &dev->flags)) + rcw++; + else + rcw += 2*disks; + } + } + PRINTK("for sector %llu, rmw=%d rcw=%d\n", + (unsigned long long)sh->sector, rmw, rcw); + set_bit(STRIPE_HANDLE, &sh->state); + if (rmw < rcw && rmw > 0) + /* prefer read-modify-write, but need to get some data */ + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if ((dev->towrite || i == sh->pd_idx) && + !test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && + test_bit(R5_Insync, &dev->flags)) { + if ( + test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + PRINTK("Read_old block " + "%d for r-m-w\n", i); + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + s->locked++; + } else { + set_bit(STRIPE_DELAYED, &sh->state); + set_bit(STRIPE_HANDLE, &sh->state); + } + } + } + if (rcw <= rmw && rcw > 0) + /* want reconstruct write, but need to get some data */ + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (!test_bit(R5_OVERWRITE, &dev->flags) && + i != sh->pd_idx && + !test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && + test_bit(R5_Insync, &dev->flags)) { + if ( + test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + PRINTK("Read_old block " + "%d for Reconstruct\n", i); + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + s->locked++; + } else { + set_bit(STRIPE_DELAYED, &sh->state); + set_bit(STRIPE_HANDLE, &sh->state); + } + } + } + /* now if nothing is locked, and if we have enough data, + * we can start a write request + */ + if (s->locked == 0 && (rcw == 0 || rmw == 0) && + !test_bit(STRIPE_BIT_DELAY, &sh->state)) { + PRINTK("Computing parity...\n"); + compute_parity5(sh, rcw == 0 ? + RECONSTRUCT_WRITE : READ_MODIFY_WRITE); + /* now every locked buffer is ready to be written */ + for (i = disks; i--; ) + if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { + PRINTK("Writing block %d\n", i); + s->locked++; + set_bit(R5_Wantwrite, &sh->dev[i].flags); + if (!test_bit(R5_Insync, &sh->dev[i].flags) + || (i == sh->pd_idx && s->failed == 0)) + set_bit(STRIPE_INSYNC, &sh->state); + } + if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + atomic_dec(&conf->preread_active_stripes); + if (atomic_read(&conf->preread_active_stripes) < + IO_THRESHOLD) + md_wakeup_thread(conf->mddev->thread); + } + } +} + +static void handle_issuing_new_write_requests6(raid5_conf_t *conf, + struct stripe_head *sh, struct stripe_head_state *s, + struct r6_state *r6s, int disks) +{ + int rcw = 0, must_compute = 0, pd_idx = sh->pd_idx, i; + int qd_idx = r6s->qd_idx; + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + /* Would I have to read this buffer for reconstruct_write */ + if (!test_bit(R5_OVERWRITE, &dev->flags) + && i != pd_idx && i != qd_idx + && (!test_bit(R5_LOCKED, &dev->flags) + ) && + !test_bit(R5_UPTODATE, &dev->flags)) { + if (test_bit(R5_Insync, &dev->flags)) rcw++; + else { + PRINTK("raid6: must_compute: " + "disk %d flags=%#lx\n", i, dev->flags); + must_compute++; + } + } + } + PRINTK("for sector %llu, rcw=%d, must_compute=%d\n", + (unsigned long long)sh->sector, rcw, must_compute); + set_bit(STRIPE_HANDLE, &sh->state); + + if (rcw > 0) + /* want reconstruct write, but need to get some data */ + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (!test_bit(R5_OVERWRITE, &dev->flags) + && !(s->failed == 0 && (i == pd_idx || i == qd_idx)) + && !test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && + test_bit(R5_Insync, &dev->flags)) { + if ( + test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + PRINTK("Read_old stripe %llu " + "block %d for Reconstruct\n", + (unsigned long long)sh->sector, i); + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + s->locked++; + } else { + PRINTK("Request delayed stripe %llu " + "block %d for Reconstruct\n", + (unsigned long long)sh->sector, i); + set_bit(STRIPE_DELAYED, &sh->state); + set_bit(STRIPE_HANDLE, &sh->state); + } + } + } + /* now if nothing is locked, and if we have enough data, we can start a + * write request + */ + if (s->locked == 0 && rcw == 0 && + !test_bit(STRIPE_BIT_DELAY, &sh->state)) { + if (must_compute > 0) { + /* We have failed blocks and need to compute them */ + switch (s->failed) { + case 0: + BUG(); + case 1: + compute_block_1(sh, r6s->failed_num[0], 0); + break; + case 2: + compute_block_2(sh, r6s->failed_num[0], + r6s->failed_num[1]); + break; + default: /* This request should have been failed? */ + BUG(); + } + } + + PRINTK("Computing parity for stripe %llu\n", + (unsigned long long)sh->sector); + compute_parity6(sh, RECONSTRUCT_WRITE); + /* now every locked buffer is ready to be written */ + for (i = disks; i--; ) + if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { + PRINTK("Writing stripe %llu block %d\n", + (unsigned long long)sh->sector, i); + s->locked++; + set_bit(R5_Wantwrite, &sh->dev[i].flags); + } + /* after a RECONSTRUCT_WRITE, the stripe MUST be in-sync */ + set_bit(STRIPE_INSYNC, &sh->state); + + if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + atomic_dec(&conf->preread_active_stripes); + if (atomic_read(&conf->preread_active_stripes) < + IO_THRESHOLD) + md_wakeup_thread(conf->mddev->thread); + } + } +} + +static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh, + struct stripe_head_state *s, int disks) +{ + set_bit(STRIPE_HANDLE, &sh->state); + if (s->failed == 0) { + BUG_ON(s->uptodate != disks); + compute_parity5(sh, CHECK_PARITY); + s->uptodate--; + if (page_is_zero(sh->dev[sh->pd_idx].page)) { + /* parity is correct (on disc, not in buffer any more) + */ + set_bit(STRIPE_INSYNC, &sh->state); + } else { + conf->mddev->resync_mismatches += STRIPE_SECTORS; + if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) + /* don't try to repair!! */ + set_bit(STRIPE_INSYNC, &sh->state); + else { + compute_block(sh, sh->pd_idx); + s->uptodate++; + } + } + } + if (!test_bit(STRIPE_INSYNC, &sh->state)) { + struct r5dev *dev; + /* either failed parity check, or recovery is happening */ + if (s->failed == 0) + s->failed_num = sh->pd_idx; + dev = &sh->dev[s->failed_num]; + BUG_ON(!test_bit(R5_UPTODATE, &dev->flags)); + BUG_ON(s->uptodate != disks); + + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + clear_bit(STRIPE_DEGRADED, &sh->state); + s->locked++; + set_bit(STRIPE_INSYNC, &sh->state); + } +} + + +static void handle_parity_checks6(raid5_conf_t *conf, struct stripe_head *sh, + struct stripe_head_state *s, + struct r6_state *r6s, struct page *tmp_page, + int disks) +{ + int update_p = 0, update_q = 0; + struct r5dev *dev; + int pd_idx = sh->pd_idx; + int qd_idx = r6s->qd_idx; + + set_bit(STRIPE_HANDLE, &sh->state); + + BUG_ON(s->failed > 2); + BUG_ON(s->uptodate < disks); + /* Want to check and possibly repair P and Q. + * However there could be one 'failed' device, in which + * case we can only check one of them, possibly using the + * other to generate missing data + */ + + /* If !tmp_page, we cannot do the calculations, + * but as we have set STRIPE_HANDLE, we will soon be called + * by stripe_handle with a tmp_page - just wait until then. + */ + if (tmp_page) { + if (s->failed == r6s->q_failed) { + /* The only possible failed device holds 'Q', so it + * makes sense to check P (If anything else were failed, + * we would have used P to recreate it). + */ + compute_block_1(sh, pd_idx, 1); + if (!page_is_zero(sh->dev[pd_idx].page)) { + compute_block_1(sh, pd_idx, 0); + update_p = 1; + } + } + if (!r6s->q_failed && s->failed < 2) { + /* q is not failed, and we didn't use it to generate + * anything, so it makes sense to check it + */ + memcpy(page_address(tmp_page), + page_address(sh->dev[qd_idx].page), + STRIPE_SIZE); + compute_parity6(sh, UPDATE_PARITY); + if (memcmp(page_address(tmp_page), + page_address(sh->dev[qd_idx].page), + STRIPE_SIZE) != 0) { + clear_bit(STRIPE_INSYNC, &sh->state); + update_q = 1; + } + } + if (update_p || update_q) { + conf->mddev->resync_mismatches += STRIPE_SECTORS; + if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) + /* don't try to repair!! */ + update_p = update_q = 0; + } + + /* now write out any block on a failed drive, + * or P or Q if they need it + */ + + if (s->failed == 2) { + dev = &sh->dev[r6s->failed_num[1]]; + s->locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + } + if (s->failed >= 1) { + dev = &sh->dev[r6s->failed_num[0]]; + s->locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + } + + if (update_p) { + dev = &sh->dev[pd_idx]; + s->locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + } + if (update_q) { + dev = &sh->dev[qd_idx]; + s->locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + } + clear_bit(STRIPE_DEGRADED, &sh->state); + + set_bit(STRIPE_INSYNC, &sh->state); + } +} + +static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh, + struct r6_state *r6s) +{ + int i; + + /* We have read all the blocks in this stripe and now we need to + * copy some of them into a target stripe for expand. + */ + clear_bit(STRIPE_EXPAND_SOURCE, &sh->state); + for (i = 0; i < sh->disks; i++) + if (i != sh->pd_idx && (r6s && i != r6s->qd_idx)) { + int dd_idx, pd_idx, j; + struct stripe_head *sh2; + + sector_t bn = compute_blocknr(sh, i); + sector_t s = raid5_compute_sector(bn, conf->raid_disks, + conf->raid_disks - + conf->max_degraded, &dd_idx, + &pd_idx, conf); + sh2 = get_active_stripe(conf, s, conf->raid_disks, + pd_idx, 1); + if (sh2 == NULL) + /* so far only the early blocks of this stripe + * have been requested. When later blocks + * get requested, we will try again + */ + continue; + if (!test_bit(STRIPE_EXPANDING, &sh2->state) || + test_bit(R5_Expanded, &sh2->dev[dd_idx].flags)) { + /* must have already done this block */ + release_stripe(sh2); + continue; + } + memcpy(page_address(sh2->dev[dd_idx].page), + page_address(sh->dev[i].page), + STRIPE_SIZE); + set_bit(R5_Expanded, &sh2->dev[dd_idx].flags); + set_bit(R5_UPTODATE, &sh2->dev[dd_idx].flags); + for (j = 0; j < conf->raid_disks; j++) + if (j != sh2->pd_idx && + (r6s && j != r6s->qd_idx) && + !test_bit(R5_Expanded, &sh2->dev[j].flags)) + break; + if (j == conf->raid_disks) { + set_bit(STRIPE_EXPAND_READY, &sh2->state); + set_bit(STRIPE_HANDLE, &sh2->state); + } + release_stripe(sh2); + } +} /* * handle_stripe - do things to a stripe. @@ -1344,20 +1963,16 @@ static int stripe_to_pdidx(sector_t stripe, raid5_conf_t *conf, int disks) * get BH_Lock set before the stripe lock is released. * */ - + static void handle_stripe5(struct stripe_head *sh) { raid5_conf_t *conf = sh->raid_conf; - int disks = sh->disks; - struct bio *return_bi= NULL; - struct bio *bi; - int i; - int syncing, expanding, expanded; - int locked=0, uptodate=0, to_read=0, to_write=0, failed=0, written=0; - int non_overwrite = 0; - int failed_num=0; + int disks = sh->disks, i; + struct bio *return_bi = NULL; + struct stripe_head_state s; struct r5dev *dev; + memset(&s, 0, sizeof(s)); PRINTK("handling stripe %llu, cnt=%d, pd_idx=%d\n", (unsigned long long)sh->sector, atomic_read(&sh->count), sh->pd_idx); @@ -1366,15 +1981,15 @@ static void handle_stripe5(struct stripe_head *sh) clear_bit(STRIPE_HANDLE, &sh->state); clear_bit(STRIPE_DELAYED, &sh->state); - syncing = test_bit(STRIPE_SYNCING, &sh->state); - expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state); - expanded = test_bit(STRIPE_EXPAND_READY, &sh->state); + s.syncing = test_bit(STRIPE_SYNCING, &sh->state); + s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state); + s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state); /* Now to look around and see what can be done */ rcu_read_lock(); for (i=disks; i--; ) { mdk_rdev_t *rdev; - dev = &sh->dev[i]; + struct r5dev *dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); PRINTK("check %d: state 0x%lx read %p write %p written %p\n", @@ -1403,17 +2018,18 @@ static void handle_stripe5(struct stripe_head *sh) } /* now count some things */ - if (test_bit(R5_LOCKED, &dev->flags)) locked++; - if (test_bit(R5_UPTODATE, &dev->flags)) uptodate++; + if (test_bit(R5_LOCKED, &dev->flags)) s.locked++; + if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++; - - if (dev->toread) to_read++; + if (dev->toread) + s.to_read++; if (dev->towrite) { - to_write++; + s.to_write++; if (!test_bit(R5_OVERWRITE, &dev->flags)) - non_overwrite++; + s.non_overwrite++; } - if (dev->written) written++; + if (dev->written) + s.written++; rdev = rcu_dereference(conf->disks[i].rdev); if (!rdev || !test_bit(In_sync, &rdev->flags)) { /* The ReadError flag will just be confusing now */ @@ -1422,306 +2038,59 @@ static void handle_stripe5(struct stripe_head *sh) } if (!rdev || !test_bit(In_sync, &rdev->flags) || test_bit(R5_ReadError, &dev->flags)) { - failed++; - failed_num = i; + s.failed++; + s.failed_num = i; } else set_bit(R5_Insync, &dev->flags); } rcu_read_unlock(); PRINTK("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d\n", - locked, uptodate, to_read, to_write, failed, failed_num); + s.locked, s.uptodate, s.to_read, s.to_write, + s.failed, s.failed_num); /* check if the array has lost two devices and, if so, some requests might * need to be failed */ - if (failed > 1 && to_read+to_write+written) { - for (i=disks; i--; ) { - int bitmap_end = 0; - - if (test_bit(R5_ReadError, &sh->dev[i].flags)) { - mdk_rdev_t *rdev; - rcu_read_lock(); - rdev = rcu_dereference(conf->disks[i].rdev); - if (rdev && test_bit(In_sync, &rdev->flags)) - /* multiple read failures in one stripe */ - md_error(conf->mddev, rdev); - rcu_read_unlock(); - } - - spin_lock_irq(&conf->device_lock); - /* fail all writes first */ - bi = sh->dev[i].towrite; - sh->dev[i].towrite = NULL; - if (bi) { to_write--; bitmap_end = 1; } - - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS){ - struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - bi->bi_next = return_bi; - return_bi = bi; - } - bi = nextbi; - } - /* and fail all 'written' */ - bi = sh->dev[i].written; - sh->dev[i].written = NULL; - if (bi) bitmap_end = 1; - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS) { - struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - bi->bi_next = return_bi; - return_bi = bi; - } - bi = bi2; - } - - /* fail any reads if this device is non-operational */ - if (!test_bit(R5_Insync, &sh->dev[i].flags) || - test_bit(R5_ReadError, &sh->dev[i].flags)) { - bi = sh->dev[i].toread; - sh->dev[i].toread = NULL; - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - if (bi) to_read--; - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS){ - struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - bi->bi_next = return_bi; - return_bi = bi; - } - bi = nextbi; - } - } - spin_unlock_irq(&conf->device_lock); - if (bitmap_end) - bitmap_endwrite(conf->mddev->bitmap, sh->sector, - STRIPE_SECTORS, 0, 0); - } - } - if (failed > 1 && syncing) { + if (s.failed > 1 && s.to_read+s.to_write+s.written) + handle_requests_to_failed_array(conf, sh, &s, disks, + &return_bi); + if (s.failed > 1 && s.syncing) { md_done_sync(conf->mddev, STRIPE_SECTORS,0); clear_bit(STRIPE_SYNCING, &sh->state); - syncing = 0; + s.syncing = 0; } /* might be able to return some write requests if the parity block * is safe, or on a failed drive */ dev = &sh->dev[sh->pd_idx]; - if ( written && - ( (test_bit(R5_Insync, &dev->flags) && !test_bit(R5_LOCKED, &dev->flags) && - test_bit(R5_UPTODATE, &dev->flags)) - || (failed == 1 && failed_num == sh->pd_idx)) - ) { - /* any written block on an uptodate or failed drive can be returned. - * Note that if we 'wrote' to a failed drive, it will be UPTODATE, but - * never LOCKED, so we don't need to test 'failed' directly. - */ - for (i=disks; i--; ) - if (sh->dev[i].written) { - dev = &sh->dev[i]; - if (!test_bit(R5_LOCKED, &dev->flags) && - test_bit(R5_UPTODATE, &dev->flags) ) { - /* We can return any write requests */ - struct bio *wbi, *wbi2; - int bitmap_end = 0; - PRINTK("Return write for disc %d\n", i); - spin_lock_irq(&conf->device_lock); - wbi = dev->written; - dev->written = NULL; - while (wbi && wbi->bi_sector < dev->sector + STRIPE_SECTORS) { - wbi2 = r5_next_bio(wbi, dev->sector); - if (--wbi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - wbi->bi_next = return_bi; - return_bi = wbi; - } - wbi = wbi2; - } - if (dev->towrite == NULL) - bitmap_end = 1; - spin_unlock_irq(&conf->device_lock); - if (bitmap_end) - bitmap_endwrite(conf->mddev->bitmap, sh->sector, - STRIPE_SECTORS, - !test_bit(STRIPE_DEGRADED, &sh->state), 0); - } - } - } + if ( s.written && + ((test_bit(R5_Insync, &dev->flags) && + !test_bit(R5_LOCKED, &dev->flags) && + test_bit(R5_UPTODATE, &dev->flags)) || + (s.failed == 1 && s.failed_num == sh->pd_idx))) + handle_completed_write_requests(conf, sh, disks, &return_bi); /* Now we might consider reading some blocks, either to check/generate * parity, or to satisfy requests * or to load a block that is being partially written. */ - if (to_read || non_overwrite || (syncing && (uptodate < disks)) || expanding) { - for (i=disks; i--;) { - dev = &sh->dev[i]; - if (!test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) && - (dev->toread || - (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) || - syncing || - expanding || - (failed && (sh->dev[failed_num].toread || - (sh->dev[failed_num].towrite && !test_bit(R5_OVERWRITE, &sh->dev[failed_num].flags)))) - ) - ) { - /* we would like to get this block, possibly - * by computing it, but we might not be able to - */ - if (uptodate == disks-1) { - PRINTK("Computing block %d\n", i); - compute_block(sh, i); - uptodate++; - } else if (test_bit(R5_Insync, &dev->flags)) { - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - locked++; - PRINTK("Reading block %d (sync=%d)\n", - i, syncing); - } - } - } - set_bit(STRIPE_HANDLE, &sh->state); - } + if (s.to_read || s.non_overwrite || + (s.syncing && (s.uptodate < disks)) || s.expanding) + handle_issuing_new_read_requests5(sh, &s, disks); /* now to consider writing and what else, if anything should be read */ - if (to_write) { - int rmw=0, rcw=0; - for (i=disks ; i--;) { - /* would I have to read this buffer for read_modify_write */ - dev = &sh->dev[i]; - if ((dev->towrite || i == sh->pd_idx) && - (!test_bit(R5_LOCKED, &dev->flags) - ) && - !test_bit(R5_UPTODATE, &dev->flags)) { - if (test_bit(R5_Insync, &dev->flags) -/* && !(!mddev->insync && i == sh->pd_idx) */ - ) - rmw++; - else rmw += 2*disks; /* cannot read it */ - } - /* Would I have to read this buffer for reconstruct_write */ - if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx && - (!test_bit(R5_LOCKED, &dev->flags) - ) && - !test_bit(R5_UPTODATE, &dev->flags)) { - if (test_bit(R5_Insync, &dev->flags)) rcw++; - else rcw += 2*disks; - } - } - PRINTK("for sector %llu, rmw=%d rcw=%d\n", - (unsigned long long)sh->sector, rmw, rcw); - set_bit(STRIPE_HANDLE, &sh->state); - if (rmw < rcw && rmw > 0) - /* prefer read-modify-write, but need to get some data */ - for (i=disks; i--;) { - dev = &sh->dev[i]; - if ((dev->towrite || i == sh->pd_idx) && - !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) && - test_bit(R5_Insync, &dev->flags)) { - if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) - { - PRINTK("Read_old block %d for r-m-w\n", i); - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - locked++; - } else { - set_bit(STRIPE_DELAYED, &sh->state); - set_bit(STRIPE_HANDLE, &sh->state); - } - } - } - if (rcw <= rmw && rcw > 0) - /* want reconstruct write, but need to get some data */ - for (i=disks; i--;) { - dev = &sh->dev[i]; - if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx && - !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) && - test_bit(R5_Insync, &dev->flags)) { - if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) - { - PRINTK("Read_old block %d for Reconstruct\n", i); - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - locked++; - } else { - set_bit(STRIPE_DELAYED, &sh->state); - set_bit(STRIPE_HANDLE, &sh->state); - } - } - } - /* now if nothing is locked, and if we have enough data, we can start a write request */ - if (locked == 0 && (rcw == 0 ||rmw == 0) && - !test_bit(STRIPE_BIT_DELAY, &sh->state)) { - PRINTK("Computing parity...\n"); - compute_parity5(sh, rcw==0 ? RECONSTRUCT_WRITE : READ_MODIFY_WRITE); - /* now every locked buffer is ready to be written */ - for (i=disks; i--;) - if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { - PRINTK("Writing block %d\n", i); - locked++; - set_bit(R5_Wantwrite, &sh->dev[i].flags); - if (!test_bit(R5_Insync, &sh->dev[i].flags) - || (i==sh->pd_idx && failed == 0)) - set_bit(STRIPE_INSYNC, &sh->state); - } - if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - atomic_dec(&conf->preread_active_stripes); - if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) - md_wakeup_thread(conf->mddev->thread); - } - } - } + if (s.to_write) + handle_issuing_new_write_requests5(conf, sh, &s, disks); /* maybe we need to check and possibly fix the parity for this stripe * Any reads will already have been scheduled, so we just see if enough data * is available */ - if (syncing && locked == 0 && - !test_bit(STRIPE_INSYNC, &sh->state)) { - set_bit(STRIPE_HANDLE, &sh->state); - if (failed == 0) { - BUG_ON(uptodate != disks); - compute_parity5(sh, CHECK_PARITY); - uptodate--; - if (page_is_zero(sh->dev[sh->pd_idx].page)) { - /* parity is correct (on disc, not in buffer any more) */ - set_bit(STRIPE_INSYNC, &sh->state); - } else { - conf->mddev->resync_mismatches += STRIPE_SECTORS; - if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) - /* don't try to repair!! */ - set_bit(STRIPE_INSYNC, &sh->state); - else { - compute_block(sh, sh->pd_idx); - uptodate++; - } - } - } - if (!test_bit(STRIPE_INSYNC, &sh->state)) { - /* either failed parity check, or recovery is happening */ - if (failed==0) - failed_num = sh->pd_idx; - dev = &sh->dev[failed_num]; - BUG_ON(!test_bit(R5_UPTODATE, &dev->flags)); - BUG_ON(uptodate != disks); - - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantwrite, &dev->flags); - clear_bit(STRIPE_DEGRADED, &sh->state); - locked++; - set_bit(STRIPE_INSYNC, &sh->state); - } - } - if (syncing && locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { + if (s.syncing && s.locked == 0 && + !test_bit(STRIPE_INSYNC, &sh->state)) + handle_parity_checks5(conf, sh, &s, disks); + if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { md_done_sync(conf->mddev, STRIPE_SECTORS,1); clear_bit(STRIPE_SYNCING, &sh->state); } @@ -1729,99 +2098,50 @@ static void handle_stripe5(struct stripe_head *sh) /* If the failed drive is just a ReadError, then we might need to progress * the repair/check process */ - if (failed == 1 && ! conf->mddev->ro && - test_bit(R5_ReadError, &sh->dev[failed_num].flags) - && !test_bit(R5_LOCKED, &sh->dev[failed_num].flags) - && test_bit(R5_UPTODATE, &sh->dev[failed_num].flags) + if (s.failed == 1 && !conf->mddev->ro && + test_bit(R5_ReadError, &sh->dev[s.failed_num].flags) + && !test_bit(R5_LOCKED, &sh->dev[s.failed_num].flags) + && test_bit(R5_UPTODATE, &sh->dev[s.failed_num].flags) ) { - dev = &sh->dev[failed_num]; + dev = &sh->dev[s.failed_num]; if (!test_bit(R5_ReWrite, &dev->flags)) { set_bit(R5_Wantwrite, &dev->flags); set_bit(R5_ReWrite, &dev->flags); set_bit(R5_LOCKED, &dev->flags); - locked++; + s.locked++; } else { /* let's read it back */ set_bit(R5_Wantread, &dev->flags); set_bit(R5_LOCKED, &dev->flags); - locked++; + s.locked++; } } - if (expanded && test_bit(STRIPE_EXPANDING, &sh->state)) { + if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state)) { /* Need to write out all blocks after computing parity */ sh->disks = conf->raid_disks; sh->pd_idx = stripe_to_pdidx(sh->sector, conf, conf->raid_disks); compute_parity5(sh, RECONSTRUCT_WRITE); - for (i= conf->raid_disks; i--;) { + for (i = conf->raid_disks; i--; ) { set_bit(R5_LOCKED, &sh->dev[i].flags); - locked++; + s.locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); } clear_bit(STRIPE_EXPANDING, &sh->state); - } else if (expanded) { + } else if (s.expanded) { clear_bit(STRIPE_EXPAND_READY, &sh->state); atomic_dec(&conf->reshape_stripes); wake_up(&conf->wait_for_overlap); md_done_sync(conf->mddev, STRIPE_SECTORS, 1); } - if (expanding && locked == 0) { - /* We have read all the blocks in this stripe and now we need to - * copy some of them into a target stripe for expand. - */ - clear_bit(STRIPE_EXPAND_SOURCE, &sh->state); - for (i=0; i< sh->disks; i++) - if (i != sh->pd_idx) { - int dd_idx, pd_idx, j; - struct stripe_head *sh2; - - sector_t bn = compute_blocknr(sh, i); - sector_t s = raid5_compute_sector(bn, conf->raid_disks, - conf->raid_disks-1, - &dd_idx, &pd_idx, conf); - sh2 = get_active_stripe(conf, s, conf->raid_disks, pd_idx, 1); - if (sh2 == NULL) - /* so far only the early blocks of this stripe - * have been requested. When later blocks - * get requested, we will try again - */ - continue; - if(!test_bit(STRIPE_EXPANDING, &sh2->state) || - test_bit(R5_Expanded, &sh2->dev[dd_idx].flags)) { - /* must have already done this block */ - release_stripe(sh2); - continue; - } - memcpy(page_address(sh2->dev[dd_idx].page), - page_address(sh->dev[i].page), - STRIPE_SIZE); - set_bit(R5_Expanded, &sh2->dev[dd_idx].flags); - set_bit(R5_UPTODATE, &sh2->dev[dd_idx].flags); - for (j=0; jraid_disks; j++) - if (j != sh2->pd_idx && - !test_bit(R5_Expanded, &sh2->dev[j].flags)) - break; - if (j == conf->raid_disks) { - set_bit(STRIPE_EXPAND_READY, &sh2->state); - set_bit(STRIPE_HANDLE, &sh2->state); - } - release_stripe(sh2); - } - } + if (s.expanding && s.locked == 0) + handle_stripe_expansion(conf, sh, NULL); spin_unlock(&sh->lock); - while ((bi=return_bi)) { - int bytes = bi->bi_size; + return_io(return_bi); - return_bi = bi->bi_next; - bi->bi_next = NULL; - bi->bi_size = 0; - bi->bi_end_io(bi, bytes, - test_bit(BIO_UPTODATE, &bi->bi_flags) - ? 0 : -EIO); - } for (i=disks; i-- ;) { int rw; struct bio *bi; @@ -1850,7 +2170,7 @@ static void handle_stripe5(struct stripe_head *sh) rcu_read_unlock(); if (rdev) { - if (syncing || expanding || expanded) + if (s.syncing || s.expanding || s.expanded) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; @@ -1886,29 +2206,26 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) { raid6_conf_t *conf = sh->raid_conf; int disks = sh->disks; - struct bio *return_bi= NULL; - struct bio *bi; - int i; - int syncing, expanding, expanded; - int locked=0, uptodate=0, to_read=0, to_write=0, failed=0, written=0; - int non_overwrite = 0; - int failed_num[2] = {0, 0}; + struct bio *return_bi = NULL; + int i, pd_idx = sh->pd_idx; + struct stripe_head_state s; + struct r6_state r6s; struct r5dev *dev, *pdev, *qdev; - int pd_idx = sh->pd_idx; - int qd_idx = raid6_next_disk(pd_idx, disks); - int p_failed, q_failed; - PRINTK("handling stripe %llu, state=%#lx cnt=%d, pd_idx=%d, qd_idx=%d\n", - (unsigned long long)sh->sector, sh->state, atomic_read(&sh->count), - pd_idx, qd_idx); + r6s.qd_idx = raid6_next_disk(pd_idx, disks); + PRINTK("handling stripe %llu, state=%#lx cnt=%d, " + "pd_idx=%d, qd_idx=%d\n", + (unsigned long long)sh->sector, sh->state, + atomic_read(&sh->count), pd_idx, r6s.qd_idx); + memset(&s, 0, sizeof(s)); spin_lock(&sh->lock); clear_bit(STRIPE_HANDLE, &sh->state); clear_bit(STRIPE_DELAYED, &sh->state); - syncing = test_bit(STRIPE_SYNCING, &sh->state); - expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state); - expanded = test_bit(STRIPE_EXPAND_READY, &sh->state); + s.syncing = test_bit(STRIPE_SYNCING, &sh->state); + s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state); + s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state); /* Now to look around and see what can be done */ rcu_read_lock(); @@ -1943,17 +2260,19 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) } /* now count some things */ - if (test_bit(R5_LOCKED, &dev->flags)) locked++; - if (test_bit(R5_UPTODATE, &dev->flags)) uptodate++; + if (test_bit(R5_LOCKED, &dev->flags)) s.locked++; + if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++; - if (dev->toread) to_read++; + if (dev->toread) + s.to_read++; if (dev->towrite) { - to_write++; + s.to_write++; if (!test_bit(R5_OVERWRITE, &dev->flags)) - non_overwrite++; + s.non_overwrite++; } - if (dev->written) written++; + if (dev->written) + s.written++; rdev = rcu_dereference(conf->disks[i].rdev); if (!rdev || !test_bit(In_sync, &rdev->flags)) { /* The ReadError flag will just be confusing now */ @@ -1962,96 +2281,27 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) } if (!rdev || !test_bit(In_sync, &rdev->flags) || test_bit(R5_ReadError, &dev->flags)) { - if ( failed < 2 ) - failed_num[failed] = i; - failed++; + if (s.failed < 2) + r6s.failed_num[s.failed] = i; + s.failed++; } else set_bit(R5_Insync, &dev->flags); } rcu_read_unlock(); PRINTK("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d,%d\n", - locked, uptodate, to_read, to_write, failed, - failed_num[0], failed_num[1]); - /* check if the array has lost >2 devices and, if so, some requests might - * need to be failed + s.locked, s.uptodate, s.to_read, s.to_write, s.failed, + r6s.failed_num[0], r6s.failed_num[1]); + /* check if the array has lost >2 devices and, if so, some requests + * might need to be failed */ - if (failed > 2 && to_read+to_write+written) { - for (i=disks; i--; ) { - int bitmap_end = 0; - - if (test_bit(R5_ReadError, &sh->dev[i].flags)) { - mdk_rdev_t *rdev; - rcu_read_lock(); - rdev = rcu_dereference(conf->disks[i].rdev); - if (rdev && test_bit(In_sync, &rdev->flags)) - /* multiple read failures in one stripe */ - md_error(conf->mddev, rdev); - rcu_read_unlock(); - } - - spin_lock_irq(&conf->device_lock); - /* fail all writes first */ - bi = sh->dev[i].towrite; - sh->dev[i].towrite = NULL; - if (bi) { to_write--; bitmap_end = 1; } - - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS){ - struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - bi->bi_next = return_bi; - return_bi = bi; - } - bi = nextbi; - } - /* and fail all 'written' */ - bi = sh->dev[i].written; - sh->dev[i].written = NULL; - if (bi) bitmap_end = 1; - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS) { - struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - bi->bi_next = return_bi; - return_bi = bi; - } - bi = bi2; - } - - /* fail any reads if this device is non-operational */ - if (!test_bit(R5_Insync, &sh->dev[i].flags) || - test_bit(R5_ReadError, &sh->dev[i].flags)) { - bi = sh->dev[i].toread; - sh->dev[i].toread = NULL; - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - if (bi) to_read--; - while (bi && bi->bi_sector < sh->dev[i].sector + STRIPE_SECTORS){ - struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector); - clear_bit(BIO_UPTODATE, &bi->bi_flags); - if (--bi->bi_phys_segments == 0) { - bi->bi_next = return_bi; - return_bi = bi; - } - bi = nextbi; - } - } - spin_unlock_irq(&conf->device_lock); - if (bitmap_end) - bitmap_endwrite(conf->mddev->bitmap, sh->sector, - STRIPE_SECTORS, 0, 0); - } - } - if (failed > 2 && syncing) { + if (s.failed > 2 && s.to_read+s.to_write+s.written) + handle_requests_to_failed_array(conf, sh, &s, disks, + &return_bi); + if (s.failed > 2 && s.syncing) { md_done_sync(conf->mddev, STRIPE_SECTORS,0); clear_bit(STRIPE_SYNCING, &sh->state); - syncing = 0; + s.syncing = 0; } /* @@ -2059,279 +2309,41 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) * are safe, or on a failed drive */ pdev = &sh->dev[pd_idx]; - p_failed = (failed >= 1 && failed_num[0] == pd_idx) - || (failed >= 2 && failed_num[1] == pd_idx); - qdev = &sh->dev[qd_idx]; - q_failed = (failed >= 1 && failed_num[0] == qd_idx) - || (failed >= 2 && failed_num[1] == qd_idx); - - if ( written && - ( p_failed || ((test_bit(R5_Insync, &pdev->flags) + r6s.p_failed = (s.failed >= 1 && r6s.failed_num[0] == pd_idx) + || (s.failed >= 2 && r6s.failed_num[1] == pd_idx); + qdev = &sh->dev[r6s.qd_idx]; + r6s.q_failed = (s.failed >= 1 && r6s.failed_num[0] == r6s.qd_idx) + || (s.failed >= 2 && r6s.failed_num[1] == r6s.qd_idx); + + if ( s.written && + ( r6s.p_failed || ((test_bit(R5_Insync, &pdev->flags) && !test_bit(R5_LOCKED, &pdev->flags) - && test_bit(R5_UPTODATE, &pdev->flags))) ) && - ( q_failed || ((test_bit(R5_Insync, &qdev->flags) + && test_bit(R5_UPTODATE, &pdev->flags)))) && + ( r6s.q_failed || ((test_bit(R5_Insync, &qdev->flags) && !test_bit(R5_LOCKED, &qdev->flags) - && test_bit(R5_UPTODATE, &qdev->flags))) ) ) { - /* any written block on an uptodate or failed drive can be - * returned. Note that if we 'wrote' to a failed drive, - * it will be UPTODATE, but never LOCKED, so we don't need - * to test 'failed' directly. - */ - for (i=disks; i--; ) - if (sh->dev[i].written) { - dev = &sh->dev[i]; - if (!test_bit(R5_LOCKED, &dev->flags) && - test_bit(R5_UPTODATE, &dev->flags) ) { - /* We can return any write requests */ - int bitmap_end = 0; - struct bio *wbi, *wbi2; - PRINTK("Return write for stripe %llu disc %d\n", - (unsigned long long)sh->sector, i); - spin_lock_irq(&conf->device_lock); - wbi = dev->written; - dev->written = NULL; - while (wbi && wbi->bi_sector < dev->sector + STRIPE_SECTORS) { - wbi2 = r5_next_bio(wbi, dev->sector); - if (--wbi->bi_phys_segments == 0) { - md_write_end(conf->mddev); - wbi->bi_next = return_bi; - return_bi = wbi; - } - wbi = wbi2; - } - if (dev->towrite == NULL) - bitmap_end = 1; - spin_unlock_irq(&conf->device_lock); - if (bitmap_end) - bitmap_endwrite(conf->mddev->bitmap, sh->sector, - STRIPE_SECTORS, - !test_bit(STRIPE_DEGRADED, &sh->state), 0); - } - } - } + && test_bit(R5_UPTODATE, &qdev->flags))))) + handle_completed_write_requests(conf, sh, disks, &return_bi); /* Now we might consider reading some blocks, either to check/generate * parity, or to satisfy requests * or to load a block that is being partially written. */ - if (to_read || non_overwrite || (to_write && failed) || - (syncing && (uptodate < disks)) || expanding) { - for (i=disks; i--;) { - dev = &sh->dev[i]; - if (!test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) && - (dev->toread || - (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) || - syncing || - expanding || - (failed >= 1 && (sh->dev[failed_num[0]].toread || to_write)) || - (failed >= 2 && (sh->dev[failed_num[1]].toread || to_write)) - ) - ) { - /* we would like to get this block, possibly - * by computing it, but we might not be able to - */ - if (uptodate == disks-1) { - PRINTK("Computing stripe %llu block %d\n", - (unsigned long long)sh->sector, i); - compute_block_1(sh, i, 0); - uptodate++; - } else if ( uptodate == disks-2 && failed >= 2 ) { - /* Computing 2-failure is *very* expensive; only do it if failed >= 2 */ - int other; - for (other=disks; other--;) { - if ( other == i ) - continue; - if ( !test_bit(R5_UPTODATE, &sh->dev[other].flags) ) - break; - } - BUG_ON(other < 0); - PRINTK("Computing stripe %llu blocks %d,%d\n", - (unsigned long long)sh->sector, i, other); - compute_block_2(sh, i, other); - uptodate += 2; - } else if (test_bit(R5_Insync, &dev->flags)) { - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - locked++; - PRINTK("Reading block %d (sync=%d)\n", - i, syncing); - } - } - } - set_bit(STRIPE_HANDLE, &sh->state); - } + if (s.to_read || s.non_overwrite || (s.to_write && s.failed) || + (s.syncing && (s.uptodate < disks)) || s.expanding) + handle_issuing_new_read_requests6(sh, &s, &r6s, disks); /* now to consider writing and what else, if anything should be read */ - if (to_write) { - int rcw=0, must_compute=0; - for (i=disks ; i--;) { - dev = &sh->dev[i]; - /* Would I have to read this buffer for reconstruct_write */ - if (!test_bit(R5_OVERWRITE, &dev->flags) - && i != pd_idx && i != qd_idx - && (!test_bit(R5_LOCKED, &dev->flags) - ) && - !test_bit(R5_UPTODATE, &dev->flags)) { - if (test_bit(R5_Insync, &dev->flags)) rcw++; - else { - PRINTK("raid6: must_compute: disk %d flags=%#lx\n", i, dev->flags); - must_compute++; - } - } - } - PRINTK("for sector %llu, rcw=%d, must_compute=%d\n", - (unsigned long long)sh->sector, rcw, must_compute); - set_bit(STRIPE_HANDLE, &sh->state); - - if (rcw > 0) - /* want reconstruct write, but need to get some data */ - for (i=disks; i--;) { - dev = &sh->dev[i]; - if (!test_bit(R5_OVERWRITE, &dev->flags) - && !(failed == 0 && (i == pd_idx || i == qd_idx)) - && !test_bit(R5_LOCKED, &dev->flags) && !test_bit(R5_UPTODATE, &dev->flags) && - test_bit(R5_Insync, &dev->flags)) { - if (test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) - { - PRINTK("Read_old stripe %llu block %d for Reconstruct\n", - (unsigned long long)sh->sector, i); - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - locked++; - } else { - PRINTK("Request delayed stripe %llu block %d for Reconstruct\n", - (unsigned long long)sh->sector, i); - set_bit(STRIPE_DELAYED, &sh->state); - set_bit(STRIPE_HANDLE, &sh->state); - } - } - } - /* now if nothing is locked, and if we have enough data, we can start a write request */ - if (locked == 0 && rcw == 0 && - !test_bit(STRIPE_BIT_DELAY, &sh->state)) { - if ( must_compute > 0 ) { - /* We have failed blocks and need to compute them */ - switch ( failed ) { - case 0: BUG(); - case 1: compute_block_1(sh, failed_num[0], 0); break; - case 2: compute_block_2(sh, failed_num[0], failed_num[1]); break; - default: BUG(); /* This request should have been failed? */ - } - } - - PRINTK("Computing parity for stripe %llu\n", (unsigned long long)sh->sector); - compute_parity6(sh, RECONSTRUCT_WRITE); - /* now every locked buffer is ready to be written */ - for (i=disks; i--;) - if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { - PRINTK("Writing stripe %llu block %d\n", - (unsigned long long)sh->sector, i); - locked++; - set_bit(R5_Wantwrite, &sh->dev[i].flags); - } - /* after a RECONSTRUCT_WRITE, the stripe MUST be in-sync */ - set_bit(STRIPE_INSYNC, &sh->state); - - if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - atomic_dec(&conf->preread_active_stripes); - if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) - md_wakeup_thread(conf->mddev->thread); - } - } - } + if (s.to_write) + handle_issuing_new_write_requests6(conf, sh, &s, &r6s, disks); /* maybe we need to check and possibly fix the parity for this stripe - * Any reads will already have been scheduled, so we just see if enough data - * is available + * Any reads will already have been scheduled, so we just see if enough + * data is available */ - if (syncing && locked == 0 && !test_bit(STRIPE_INSYNC, &sh->state)) { - int update_p = 0, update_q = 0; - struct r5dev *dev; - - set_bit(STRIPE_HANDLE, &sh->state); + if (s.syncing && s.locked == 0 && !test_bit(STRIPE_INSYNC, &sh->state)) + handle_parity_checks6(conf, sh, &s, &r6s, tmp_page, disks); - BUG_ON(failed>2); - BUG_ON(uptodate < disks); - /* Want to check and possibly repair P and Q. - * However there could be one 'failed' device, in which - * case we can only check one of them, possibly using the - * other to generate missing data - */ - - /* If !tmp_page, we cannot do the calculations, - * but as we have set STRIPE_HANDLE, we will soon be called - * by stripe_handle with a tmp_page - just wait until then. - */ - if (tmp_page) { - if (failed == q_failed) { - /* The only possible failed device holds 'Q', so it makes - * sense to check P (If anything else were failed, we would - * have used P to recreate it). - */ - compute_block_1(sh, pd_idx, 1); - if (!page_is_zero(sh->dev[pd_idx].page)) { - compute_block_1(sh,pd_idx,0); - update_p = 1; - } - } - if (!q_failed && failed < 2) { - /* q is not failed, and we didn't use it to generate - * anything, so it makes sense to check it - */ - memcpy(page_address(tmp_page), - page_address(sh->dev[qd_idx].page), - STRIPE_SIZE); - compute_parity6(sh, UPDATE_PARITY); - if (memcmp(page_address(tmp_page), - page_address(sh->dev[qd_idx].page), - STRIPE_SIZE)!= 0) { - clear_bit(STRIPE_INSYNC, &sh->state); - update_q = 1; - } - } - if (update_p || update_q) { - conf->mddev->resync_mismatches += STRIPE_SECTORS; - if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) - /* don't try to repair!! */ - update_p = update_q = 0; - } - - /* now write out any block on a failed drive, - * or P or Q if they need it - */ - - if (failed == 2) { - dev = &sh->dev[failed_num[1]]; - locked++; - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantwrite, &dev->flags); - } - if (failed >= 1) { - dev = &sh->dev[failed_num[0]]; - locked++; - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantwrite, &dev->flags); - } - - if (update_p) { - dev = &sh->dev[pd_idx]; - locked ++; - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantwrite, &dev->flags); - } - if (update_q) { - dev = &sh->dev[qd_idx]; - locked++; - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantwrite, &dev->flags); - } - clear_bit(STRIPE_DEGRADED, &sh->state); - - set_bit(STRIPE_INSYNC, &sh->state); - } - } - - if (syncing && locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { + if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { md_done_sync(conf->mddev, STRIPE_SECTORS,1); clear_bit(STRIPE_SYNCING, &sh->state); } @@ -2339,9 +2351,9 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) /* If the failed drives are just a ReadError, then we might need * to progress the repair/check process */ - if (failed <= 2 && ! conf->mddev->ro) - for (i=0; idev[failed_num[i]]; + if (s.failed <= 2 && !conf->mddev->ro) + for (i = 0; i < s.failed; i++) { + dev = &sh->dev[r6s.failed_num[i]]; if (test_bit(R5_ReadError, &dev->flags) && !test_bit(R5_LOCKED, &dev->flags) && test_bit(R5_UPTODATE, &dev->flags) @@ -2358,7 +2370,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) } } - if (expanded && test_bit(STRIPE_EXPANDING, &sh->state)) { + if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state)) { /* Need to write out all blocks after computing P&Q */ sh->disks = conf->raid_disks; sh->pd_idx = stripe_to_pdidx(sh->sector, conf, @@ -2366,82 +2378,24 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) compute_parity6(sh, RECONSTRUCT_WRITE); for (i = conf->raid_disks ; i-- ; ) { set_bit(R5_LOCKED, &sh->dev[i].flags); - locked++; + s.locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); } clear_bit(STRIPE_EXPANDING, &sh->state); - } else if (expanded) { + } else if (s.expanded) { clear_bit(STRIPE_EXPAND_READY, &sh->state); atomic_dec(&conf->reshape_stripes); wake_up(&conf->wait_for_overlap); md_done_sync(conf->mddev, STRIPE_SECTORS, 1); } - if (expanding && locked == 0) { - /* We have read all the blocks in this stripe and now we need to - * copy some of them into a target stripe for expand. - */ - clear_bit(STRIPE_EXPAND_SOURCE, &sh->state); - for (i = 0; i < sh->disks ; i++) - if (i != pd_idx && i != qd_idx) { - int dd_idx2, pd_idx2, j; - struct stripe_head *sh2; - - sector_t bn = compute_blocknr(sh, i); - sector_t s = raid5_compute_sector( - bn, conf->raid_disks, - conf->raid_disks - conf->max_degraded, - &dd_idx2, &pd_idx2, conf); - sh2 = get_active_stripe(conf, s, - conf->raid_disks, - pd_idx2, 1); - if (sh2 == NULL) - /* so for only the early blocks of - * this stripe have been requests. - * When later blocks get requests, we - * will try again - */ - continue; - if (!test_bit(STRIPE_EXPANDING, &sh2->state) || - test_bit(R5_Expanded, - &sh2->dev[dd_idx2].flags)) { - /* must have already done this block */ - release_stripe(sh2); - continue; - } - memcpy(page_address(sh2->dev[dd_idx2].page), - page_address(sh->dev[i].page), - STRIPE_SIZE); - set_bit(R5_Expanded, &sh2->dev[dd_idx2].flags); - set_bit(R5_UPTODATE, &sh2->dev[dd_idx2].flags); - for (j = 0 ; j < conf->raid_disks ; j++) - if (j != sh2->pd_idx && - j != raid6_next_disk(sh2->pd_idx, - sh2->disks) && - !test_bit(R5_Expanded, - &sh2->dev[j].flags)) - break; - if (j == conf->raid_disks) { - set_bit(STRIPE_EXPAND_READY, - &sh2->state); - set_bit(STRIPE_HANDLE, &sh2->state); - } - release_stripe(sh2); - } - } + if (s.expanding && s.locked == 0) + handle_stripe_expansion(conf, sh, &r6s); spin_unlock(&sh->lock); - while ((bi=return_bi)) { - int bytes = bi->bi_size; + return_io(return_bi); - return_bi = bi->bi_next; - bi->bi_next = NULL; - bi->bi_size = 0; - bi->bi_end_io(bi, bytes, - test_bit(BIO_UPTODATE, &bi->bi_flags) - ? 0 : -EIO); - } for (i=disks; i-- ;) { int rw; struct bio *bi; @@ -2470,7 +2424,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) rcu_read_unlock(); if (rdev) { - if (syncing || expanding || expanded) + if (s.syncing || s.expanding || s.expanded) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index d8286db..b99d354 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -145,6 +145,22 @@ struct stripe_head { unsigned long flags; } dev[1]; /* allocated with extra space depending of RAID geometry */ }; + +/* stripe_head_state - collects and tracks the dynamic state of a stripe_head + * for handle_stripe. It is only valid under spin_lock(sh->lock); + */ +struct stripe_head_state { + int syncing, expanding, expanded; + int locked, uptodate, to_read, to_write, failed, written; + int non_overwrite; + int failed_num; +}; + +/* r6_state - extra state data only relevant to r6 */ +struct r6_state { + int p_failed, q_failed, qd_idx, failed_num[2]; +}; + /* Flags */ #define R5_UPTODATE 0 /* page contains current data */ #define R5_LOCKED 1 /* IO has been submitted on "req" */ -- cgit v1.1 From 45b4233caac05da0118b608a9fc2a40a9fc580cd Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Mon, 9 Jul 2007 11:56:43 -0700 Subject: raid5: replace custom debug PRINTKs with standard pr_debug Replaces PRINTK with pr_debug, and kills the RAID5_DEBUG definition in favor of the global DEBUG definition. To get local debug messages just add '#define DEBUG' to the top of the file. Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 116 ++++++++++++++++++++++++++--------------------------- 1 file changed, 58 insertions(+), 58 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 38232fa..e372e57 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -80,7 +80,6 @@ /* * The following can be used to debug the driver */ -#define RAID5_DEBUG 0 #define RAID5_PARANOIA 1 #if RAID5_PARANOIA && defined(CONFIG_SMP) # define CHECK_DEVLOCK() assert_spin_locked(&conf->device_lock) @@ -88,8 +87,7 @@ # define CHECK_DEVLOCK() #endif -#define PRINTK(x...) ((void)(RAID5_DEBUG && printk(x))) -#if RAID5_DEBUG +#ifdef DEBUG #define inline #define __inline__ #endif @@ -169,7 +167,8 @@ static void release_stripe(struct stripe_head *sh) static inline void remove_hash(struct stripe_head *sh) { - PRINTK("remove_hash(), stripe %llu\n", (unsigned long long)sh->sector); + pr_debug("remove_hash(), stripe %llu\n", + (unsigned long long)sh->sector); hlist_del_init(&sh->hash); } @@ -178,7 +177,8 @@ static inline void insert_hash(raid5_conf_t *conf, struct stripe_head *sh) { struct hlist_head *hp = stripe_hash(conf, sh->sector); - PRINTK("insert_hash(), stripe %llu\n", (unsigned long long)sh->sector); + pr_debug("insert_hash(), stripe %llu\n", + (unsigned long long)sh->sector); CHECK_DEVLOCK(); hlist_add_head(&sh->hash, hp); @@ -243,7 +243,7 @@ static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int BUG_ON(test_bit(STRIPE_HANDLE, &sh->state)); CHECK_DEVLOCK(); - PRINTK("init_stripe called, stripe %llu\n", + pr_debug("init_stripe called, stripe %llu\n", (unsigned long long)sh->sector); remove_hash(sh); @@ -277,11 +277,11 @@ static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector, in struct hlist_node *hn; CHECK_DEVLOCK(); - PRINTK("__find_stripe, sector %llu\n", (unsigned long long)sector); + pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector); hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash) if (sh->sector == sector && sh->disks == disks) return sh; - PRINTK("__stripe %llu not in cache\n", (unsigned long long)sector); + pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector); return NULL; } @@ -293,7 +293,7 @@ static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector { struct stripe_head *sh; - PRINTK("get_stripe, sector %llu\n", (unsigned long long)sector); + pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector); spin_lock_irq(&conf->device_lock); @@ -554,8 +554,8 @@ static int raid5_end_read_request(struct bio * bi, unsigned int bytes_done, if (bi == &sh->dev[i].req) break; - PRINTK("end_read_request %llu/%d, count: %d, uptodate %d.\n", - (unsigned long long)sh->sector, i, atomic_read(&sh->count), + pr_debug("end_read_request %llu/%d, count: %d, uptodate %d.\n", + (unsigned long long)sh->sector, i, atomic_read(&sh->count), uptodate); if (i == disks) { BUG(); @@ -630,7 +630,7 @@ static int raid5_end_write_request (struct bio *bi, unsigned int bytes_done, if (bi == &sh->dev[i].req) break; - PRINTK("end_write_request %llu/%d, count %d, uptodate: %d.\n", + pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n", (unsigned long long)sh->sector, i, atomic_read(&sh->count), uptodate); if (i == disks) { @@ -675,7 +675,7 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev) { char b[BDEVNAME_SIZE]; raid5_conf_t *conf = (raid5_conf_t *) mddev->private; - PRINTK("raid5: error called\n"); + pr_debug("raid5: error called\n"); if (!test_bit(Faulty, &rdev->flags)) { set_bit(MD_CHANGE_DEVS, &mddev->flags); @@ -946,7 +946,7 @@ static void compute_block(struct stripe_head *sh, int dd_idx) int i, count, disks = sh->disks; void *ptr[MAX_XOR_BLOCKS], *dest, *p; - PRINTK("compute_block, stripe %llu, idx %d\n", + pr_debug("compute_block, stripe %llu, idx %d\n", (unsigned long long)sh->sector, dd_idx); dest = page_address(sh->dev[dd_idx].page); @@ -977,7 +977,7 @@ static void compute_parity5(struct stripe_head *sh, int method) void *ptr[MAX_XOR_BLOCKS], *dest; struct bio *chosen; - PRINTK("compute_parity5, stripe %llu, method %d\n", + pr_debug("compute_parity5, stripe %llu, method %d\n", (unsigned long long)sh->sector, method); count = 0; @@ -1075,7 +1075,7 @@ static void compute_parity6(struct stripe_head *sh, int method) qd_idx = raid6_next_disk(pd_idx, disks); d0_idx = raid6_next_disk(qd_idx, disks); - PRINTK("compute_parity, stripe %llu, method %d\n", + pr_debug("compute_parity, stripe %llu, method %d\n", (unsigned long long)sh->sector, method); switch(method) { @@ -1153,7 +1153,7 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) int pd_idx = sh->pd_idx; int qd_idx = raid6_next_disk(pd_idx, disks); - PRINTK("compute_block_1, stripe %llu, idx %d\n", + pr_debug("compute_block_1, stripe %llu, idx %d\n", (unsigned long long)sh->sector, dd_idx); if ( dd_idx == qd_idx ) { @@ -1200,7 +1200,7 @@ static void compute_block_2(struct stripe_head *sh, int dd_idx1, int dd_idx2) BUG_ON(faila == failb); if ( failb < faila ) { int tmp = faila; faila = failb; failb = tmp; } - PRINTK("compute_block_2, stripe %llu, idx %d,%d (%d,%d)\n", + pr_debug("compute_block_2, stripe %llu, idx %d,%d (%d,%d)\n", (unsigned long long)sh->sector, dd_idx1, dd_idx2, faila, failb); if ( failb == disks-1 ) { @@ -1259,7 +1259,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in raid5_conf_t *conf = sh->raid_conf; int firstwrite=0; - PRINTK("adding bh b#%llu to stripe s#%llu\n", + pr_debug("adding bh b#%llu to stripe s#%llu\n", (unsigned long long)bi->bi_sector, (unsigned long long)sh->sector); @@ -1288,7 +1288,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in spin_unlock_irq(&conf->device_lock); spin_unlock(&sh->lock); - PRINTK("added bi b#%llu to stripe s#%llu, disk %d.\n", + pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n", (unsigned long long)bi->bi_sector, (unsigned long long)sh->sector, dd_idx); @@ -1448,14 +1448,14 @@ static void handle_issuing_new_read_requests5(struct stripe_head *sh, * by computing it, but we might not be able to */ if (s->uptodate == disks-1) { - PRINTK("Computing block %d\n", i); + pr_debug("Computing block %d\n", i); compute_block(sh, i); s->uptodate++; } else if (test_bit(R5_Insync, &dev->flags)) { set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); s->locked++; - PRINTK("Reading block %d (sync=%d)\n", + pr_debug("Reading block %d (sync=%d)\n", i, s->syncing); } } @@ -1485,7 +1485,7 @@ static void handle_issuing_new_read_requests6(struct stripe_head *sh, * by computing it, but we might not be able to */ if (s->uptodate == disks-1) { - PRINTK("Computing stripe %llu block %d\n", + pr_debug("Computing stripe %llu block %d\n", (unsigned long long)sh->sector, i); compute_block_1(sh, i, 0); s->uptodate++; @@ -1502,7 +1502,7 @@ static void handle_issuing_new_read_requests6(struct stripe_head *sh, break; } BUG_ON(other < 0); - PRINTK("Computing stripe %llu blocks %d,%d\n", + pr_debug("Computing stripe %llu blocks %d,%d\n", (unsigned long long)sh->sector, i, other); compute_block_2(sh, i, other); @@ -1511,7 +1511,7 @@ static void handle_issuing_new_read_requests6(struct stripe_head *sh, set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); s->locked++; - PRINTK("Reading block %d (sync=%d)\n", + pr_debug("Reading block %d (sync=%d)\n", i, s->syncing); } } @@ -1539,7 +1539,7 @@ static void handle_completed_write_requests(raid5_conf_t *conf, /* We can return any write requests */ struct bio *wbi, *wbi2; int bitmap_end = 0; - PRINTK("Return write for disc %d\n", i); + pr_debug("Return write for disc %d\n", i); spin_lock_irq(&conf->device_lock); wbi = dev->written; dev->written = NULL; @@ -1591,7 +1591,7 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, rcw += 2*disks; } } - PRINTK("for sector %llu, rmw=%d rcw=%d\n", + pr_debug("for sector %llu, rmw=%d rcw=%d\n", (unsigned long long)sh->sector, rmw, rcw); set_bit(STRIPE_HANDLE, &sh->state); if (rmw < rcw && rmw > 0) @@ -1604,7 +1604,7 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, test_bit(R5_Insync, &dev->flags)) { if ( test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - PRINTK("Read_old block " + pr_debug("Read_old block " "%d for r-m-w\n", i); set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); @@ -1626,7 +1626,7 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, test_bit(R5_Insync, &dev->flags)) { if ( test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - PRINTK("Read_old block " + pr_debug("Read_old block " "%d for Reconstruct\n", i); set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); @@ -1642,13 +1642,13 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, */ if (s->locked == 0 && (rcw == 0 || rmw == 0) && !test_bit(STRIPE_BIT_DELAY, &sh->state)) { - PRINTK("Computing parity...\n"); + pr_debug("Computing parity...\n"); compute_parity5(sh, rcw == 0 ? RECONSTRUCT_WRITE : READ_MODIFY_WRITE); /* now every locked buffer is ready to be written */ for (i = disks; i--; ) if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { - PRINTK("Writing block %d\n", i); + pr_debug("Writing block %d\n", i); s->locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); if (!test_bit(R5_Insync, &sh->dev[i].flags) @@ -1680,13 +1680,13 @@ static void handle_issuing_new_write_requests6(raid5_conf_t *conf, !test_bit(R5_UPTODATE, &dev->flags)) { if (test_bit(R5_Insync, &dev->flags)) rcw++; else { - PRINTK("raid6: must_compute: " + pr_debug("raid6: must_compute: " "disk %d flags=%#lx\n", i, dev->flags); must_compute++; } } } - PRINTK("for sector %llu, rcw=%d, must_compute=%d\n", + pr_debug("for sector %llu, rcw=%d, must_compute=%d\n", (unsigned long long)sh->sector, rcw, must_compute); set_bit(STRIPE_HANDLE, &sh->state); @@ -1701,14 +1701,14 @@ static void handle_issuing_new_write_requests6(raid5_conf_t *conf, test_bit(R5_Insync, &dev->flags)) { if ( test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - PRINTK("Read_old stripe %llu " + pr_debug("Read_old stripe %llu " "block %d for Reconstruct\n", (unsigned long long)sh->sector, i); set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); s->locked++; } else { - PRINTK("Request delayed stripe %llu " + pr_debug("Request delayed stripe %llu " "block %d for Reconstruct\n", (unsigned long long)sh->sector, i); set_bit(STRIPE_DELAYED, &sh->state); @@ -1738,13 +1738,13 @@ static void handle_issuing_new_write_requests6(raid5_conf_t *conf, } } - PRINTK("Computing parity for stripe %llu\n", + pr_debug("Computing parity for stripe %llu\n", (unsigned long long)sh->sector); compute_parity6(sh, RECONSTRUCT_WRITE); /* now every locked buffer is ready to be written */ for (i = disks; i--; ) if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { - PRINTK("Writing stripe %llu block %d\n", + pr_debug("Writing stripe %llu block %d\n", (unsigned long long)sh->sector, i); s->locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); @@ -1973,7 +1973,7 @@ static void handle_stripe5(struct stripe_head *sh) struct r5dev *dev; memset(&s, 0, sizeof(s)); - PRINTK("handling stripe %llu, cnt=%d, pd_idx=%d\n", + pr_debug("handling stripe %llu, cnt=%d, pd_idx=%d\n", (unsigned long long)sh->sector, atomic_read(&sh->count), sh->pd_idx); @@ -1992,12 +1992,12 @@ static void handle_stripe5(struct stripe_head *sh) struct r5dev *dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); - PRINTK("check %d: state 0x%lx read %p write %p written %p\n", + pr_debug("check %d: state 0x%lx read %p write %p written %p\n", i, dev->flags, dev->toread, dev->towrite, dev->written); /* maybe we can reply to a read */ if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread) { struct bio *rbi, *rbi2; - PRINTK("Return read for disc %d\n", i); + pr_debug("Return read for disc %d\n", i); spin_lock_irq(&conf->device_lock); rbi = dev->toread; dev->toread = NULL; @@ -2044,7 +2044,7 @@ static void handle_stripe5(struct stripe_head *sh) set_bit(R5_Insync, &dev->flags); } rcu_read_unlock(); - PRINTK("locked=%d uptodate=%d to_read=%d" + pr_debug("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d\n", s.locked, s.uptodate, s.to_read, s.to_write, s.failed, s.failed_num); @@ -2174,7 +2174,7 @@ static void handle_stripe5(struct stripe_head *sh) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; - PRINTK("for %llu schedule op %ld on disc %d\n", + pr_debug("for %llu schedule op %ld on disc %d\n", (unsigned long long)sh->sector, bi->bi_rw, i); atomic_inc(&sh->count); bi->bi_sector = sh->sector + rdev->data_offset; @@ -2194,7 +2194,7 @@ static void handle_stripe5(struct stripe_head *sh) } else { if (rw == WRITE) set_bit(STRIPE_DEGRADED, &sh->state); - PRINTK("skip op %ld on disc %d for sector %llu\n", + pr_debug("skip op %ld on disc %d for sector %llu\n", bi->bi_rw, i, (unsigned long long)sh->sector); clear_bit(R5_LOCKED, &sh->dev[i].flags); set_bit(STRIPE_HANDLE, &sh->state); @@ -2213,7 +2213,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) struct r5dev *dev, *pdev, *qdev; r6s.qd_idx = raid6_next_disk(pd_idx, disks); - PRINTK("handling stripe %llu, state=%#lx cnt=%d, " + pr_debug("handling stripe %llu, state=%#lx cnt=%d, " "pd_idx=%d, qd_idx=%d\n", (unsigned long long)sh->sector, sh->state, atomic_read(&sh->count), pd_idx, r6s.qd_idx); @@ -2234,12 +2234,12 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); - PRINTK("check %d: state 0x%lx read %p write %p written %p\n", + pr_debug("check %d: state 0x%lx read %p write %p written %p\n", i, dev->flags, dev->toread, dev->towrite, dev->written); /* maybe we can reply to a read */ if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread) { struct bio *rbi, *rbi2; - PRINTK("Return read for disc %d\n", i); + pr_debug("Return read for disc %d\n", i); spin_lock_irq(&conf->device_lock); rbi = dev->toread; dev->toread = NULL; @@ -2288,7 +2288,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) set_bit(R5_Insync, &dev->flags); } rcu_read_unlock(); - PRINTK("locked=%d uptodate=%d to_read=%d" + pr_debug("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d,%d\n", s.locked, s.uptodate, s.to_read, s.to_write, s.failed, r6s.failed_num[0], r6s.failed_num[1]); @@ -2428,7 +2428,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; - PRINTK("for %llu schedule op %ld on disc %d\n", + pr_debug("for %llu schedule op %ld on disc %d\n", (unsigned long long)sh->sector, bi->bi_rw, i); atomic_inc(&sh->count); bi->bi_sector = sh->sector + rdev->data_offset; @@ -2448,7 +2448,7 @@ static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) } else { if (rw == WRITE) set_bit(STRIPE_DEGRADED, &sh->state); - PRINTK("skip op %ld on disc %d for sector %llu\n", + pr_debug("skip op %ld on disc %d for sector %llu\n", bi->bi_rw, i, (unsigned long long)sh->sector); clear_bit(R5_LOCKED, &sh->dev[i].flags); set_bit(STRIPE_HANDLE, &sh->state); @@ -2692,7 +2692,7 @@ static int raid5_align_endio(struct bio *bi, unsigned int bytes, int error) } - PRINTK("raid5_align_endio : io error...handing IO for a retry\n"); + pr_debug("raid5_align_endio : io error...handing IO for a retry\n"); add_bio_to_retry(raid_bi, conf); return 0; @@ -2730,7 +2730,7 @@ static int chunk_aligned_read(request_queue_t *q, struct bio * raid_bio) mdk_rdev_t *rdev; if (!in_chunk_boundary(mddev, raid_bio)) { - PRINTK("chunk_aligned_read : non aligned\n"); + pr_debug("chunk_aligned_read : non aligned\n"); return 0; } /* @@ -2854,7 +2854,7 @@ static int make_request(request_queue_t *q, struct bio * bi) new_sector = raid5_compute_sector(logical_sector, disks, data_disks, &dd_idx, &pd_idx, conf); - PRINTK("raid5: make_request, sector %llu logical %llu\n", + pr_debug("raid5: make_request, sector %llu logical %llu\n", (unsigned long long)new_sector, (unsigned long long)logical_sector); @@ -3227,7 +3227,7 @@ static void raid5d (mddev_t *mddev) raid5_conf_t *conf = mddev_to_conf(mddev); int handled; - PRINTK("+++ raid5d active\n"); + pr_debug("+++ raid5d active\n"); md_check_recovery(mddev); @@ -3279,13 +3279,13 @@ static void raid5d (mddev_t *mddev) spin_lock_irq(&conf->device_lock); } - PRINTK("%d stripes handled\n", handled); + pr_debug("%d stripes handled\n", handled); spin_unlock_irq(&conf->device_lock); unplug_slaves(mddev); - PRINTK("--- raid5d inactive\n"); + pr_debug("--- raid5d inactive\n"); } static ssize_t @@ -3461,7 +3461,7 @@ static int run(mddev_t *mddev) atomic_set(&conf->preread_active_stripes, 0); atomic_set(&conf->active_aligned_reads, 0); - PRINTK("raid5: run(%s) called.\n", mdname(mddev)); + pr_debug("raid5: run(%s) called.\n", mdname(mddev)); ITERATE_RDEV(mddev,rdev,tmp) { raid_disk = rdev->raid_disk; @@ -3644,7 +3644,7 @@ static int stop(mddev_t *mddev) return 0; } -#if RAID5_DEBUG +#ifdef DEBUG static void print_sh (struct seq_file *seq, struct stripe_head *sh) { int i; @@ -3691,7 +3691,7 @@ static void status (struct seq_file *seq, mddev_t *mddev) conf->disks[i].rdev && test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_"); seq_printf (seq, "]"); -#if RAID5_DEBUG +#ifdef DEBUG seq_printf (seq, "\n"); printall(seq, conf); #endif -- cgit v1.1 From 91c00924846a0034020451c280c76baa4299f9dc Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:30 -0700 Subject: md: raid5_run_ops - run stripe operations outside sh->lock When the raid acceleration work was proposed, Neil laid out the following attack plan: 1/ move the xor and copy operations outside spin_lock(&sh->lock) 2/ find/implement an asynchronous offload api The raid5_run_ops routine uses the asynchronous offload api (async_tx) and the stripe_operations member of a stripe_head to carry out xor+copy operations asynchronously, outside the lock. To perform operations outside the lock a new set of state flags is needed to track new requests, in-flight requests, and completed requests. In this new model handle_stripe is tasked with scanning the stripe_head for work, updating the stripe_operations structure, and finally dropping the lock and calling raid5_run_ops for processing. The following flags outline the requests that handle_stripe can make of raid5_run_ops: STRIPE_OP_BIOFILL - copy data into request buffers to satisfy a read request STRIPE_OP_COMPUTE_BLK - generate a missing block in the cache from the other blocks STRIPE_OP_PREXOR - subtract existing data as part of the read-modify-write process STRIPE_OP_BIODRAIN - copy data out of request buffers to satisfy a write request STRIPE_OP_POSTXOR - recalculate parity for new data that has entered the cache STRIPE_OP_CHECK - verify that the parity is correct STRIPE_OP_IO - submit i/o to the member disks (note this was already performed outside the stripe lock, but it made sense to add it as an operation type The flow is: 1/ handle_stripe sets STRIPE_OP_* in sh->ops.pending 2/ raid5_run_ops reads sh->ops.pending, sets sh->ops.ack, and submits the operation to the async_tx api 3/ async_tx triggers the completion callback routine to set sh->ops.complete and release the stripe 4/ handle_stripe runs again to finish the operation and optionally submit new operations that were previously blocked Note this patch just defines raid5_run_ops, subsequent commits (one per major operation type) modify handle_stripe to take advantage of this routine. Changelog: * removed ops_complete_biodrain in favor of ops_complete_postxor and ops_complete_write. * removed the raid5_run_ops workqueue * call bi_end_io for reads in ops_complete_biofill, saves a call to handle_stripe * explicitly handle the 2-disk raid5 case (xor becomes memcpy), Neil Brown * fix race between async engines and bi_end_io call for reads, Neil Brown * remove unnecessary spin_lock from ops_complete_biofill * remove test_and_set/test_and_clear BUG_ONs, Neil Brown * remove explicit interrupt handling for channel switching, this feature was absorbed (i.e. it is now implicit) by the async_tx api * use return_io in ops_complete_biofill Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 536 +++++++++++++++++++++++++++++++++++++++++++++ include/linux/raid/raid5.h | 81 ++++++- 2 files changed, 614 insertions(+), 3 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index e372e57..0b70024 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -52,6 +52,7 @@ #include "raid6.h" #include +#include /* * Stripe cache @@ -341,6 +342,541 @@ static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector return sh; } +static int +raid5_end_read_request(struct bio *bi, unsigned int bytes_done, int error); +static int +raid5_end_write_request (struct bio *bi, unsigned int bytes_done, int error); + +static void ops_run_io(struct stripe_head *sh) +{ + raid5_conf_t *conf = sh->raid_conf; + int i, disks = sh->disks; + + might_sleep(); + + for (i = disks; i--; ) { + int rw; + struct bio *bi; + mdk_rdev_t *rdev; + if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) + rw = WRITE; + else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags)) + rw = READ; + else + continue; + + bi = &sh->dev[i].req; + + bi->bi_rw = rw; + if (rw == WRITE) + bi->bi_end_io = raid5_end_write_request; + else + bi->bi_end_io = raid5_end_read_request; + + rcu_read_lock(); + rdev = rcu_dereference(conf->disks[i].rdev); + if (rdev && test_bit(Faulty, &rdev->flags)) + rdev = NULL; + if (rdev) + atomic_inc(&rdev->nr_pending); + rcu_read_unlock(); + + if (rdev) { + if (test_bit(STRIPE_SYNCING, &sh->state) || + test_bit(STRIPE_EXPAND_SOURCE, &sh->state) || + test_bit(STRIPE_EXPAND_READY, &sh->state)) + md_sync_acct(rdev->bdev, STRIPE_SECTORS); + + bi->bi_bdev = rdev->bdev; + pr_debug("%s: for %llu schedule op %ld on disc %d\n", + __FUNCTION__, (unsigned long long)sh->sector, + bi->bi_rw, i); + atomic_inc(&sh->count); + bi->bi_sector = sh->sector + rdev->data_offset; + bi->bi_flags = 1 << BIO_UPTODATE; + bi->bi_vcnt = 1; + bi->bi_max_vecs = 1; + bi->bi_idx = 0; + bi->bi_io_vec = &sh->dev[i].vec; + bi->bi_io_vec[0].bv_len = STRIPE_SIZE; + bi->bi_io_vec[0].bv_offset = 0; + bi->bi_size = STRIPE_SIZE; + bi->bi_next = NULL; + if (rw == WRITE && + test_bit(R5_ReWrite, &sh->dev[i].flags)) + atomic_add(STRIPE_SECTORS, + &rdev->corrected_errors); + generic_make_request(bi); + } else { + if (rw == WRITE) + set_bit(STRIPE_DEGRADED, &sh->state); + pr_debug("skip op %ld on disc %d for sector %llu\n", + bi->bi_rw, i, (unsigned long long)sh->sector); + clear_bit(R5_LOCKED, &sh->dev[i].flags); + set_bit(STRIPE_HANDLE, &sh->state); + } + } +} + +static struct dma_async_tx_descriptor * +async_copy_data(int frombio, struct bio *bio, struct page *page, + sector_t sector, struct dma_async_tx_descriptor *tx) +{ + struct bio_vec *bvl; + struct page *bio_page; + int i; + int page_offset; + + if (bio->bi_sector >= sector) + page_offset = (signed)(bio->bi_sector - sector) * 512; + else + page_offset = (signed)(sector - bio->bi_sector) * -512; + bio_for_each_segment(bvl, bio, i) { + int len = bio_iovec_idx(bio, i)->bv_len; + int clen; + int b_offset = 0; + + if (page_offset < 0) { + b_offset = -page_offset; + page_offset += b_offset; + len -= b_offset; + } + + if (len > 0 && page_offset + len > STRIPE_SIZE) + clen = STRIPE_SIZE - page_offset; + else + clen = len; + + if (clen > 0) { + b_offset += bio_iovec_idx(bio, i)->bv_offset; + bio_page = bio_iovec_idx(bio, i)->bv_page; + if (frombio) + tx = async_memcpy(page, bio_page, page_offset, + b_offset, clen, + ASYNC_TX_DEP_ACK | ASYNC_TX_KMAP_SRC, + tx, NULL, NULL); + else + tx = async_memcpy(bio_page, page, b_offset, + page_offset, clen, + ASYNC_TX_DEP_ACK | ASYNC_TX_KMAP_DST, + tx, NULL, NULL); + } + if (clen < len) /* hit end of page */ + break; + page_offset += len; + } + + return tx; +} + +static void ops_complete_biofill(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + struct bio *return_bi = NULL; + raid5_conf_t *conf = sh->raid_conf; + int i, more_to_read = 0; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + /* clear completed biofills */ + for (i = sh->disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + /* check if this stripe has new incoming reads */ + if (dev->toread) + more_to_read++; + + /* acknowledge completion of a biofill operation */ + /* and check if we need to reply to a read request + */ + if (test_bit(R5_Wantfill, &dev->flags) && !dev->toread) { + struct bio *rbi, *rbi2; + clear_bit(R5_Wantfill, &dev->flags); + + /* The access to dev->read is outside of the + * spin_lock_irq(&conf->device_lock), but is protected + * by the STRIPE_OP_BIOFILL pending bit + */ + BUG_ON(!dev->read); + rbi = dev->read; + dev->read = NULL; + while (rbi && rbi->bi_sector < + dev->sector + STRIPE_SECTORS) { + rbi2 = r5_next_bio(rbi, dev->sector); + spin_lock_irq(&conf->device_lock); + if (--rbi->bi_phys_segments == 0) { + rbi->bi_next = return_bi; + return_bi = rbi; + } + spin_unlock_irq(&conf->device_lock); + rbi = rbi2; + } + } + } + clear_bit(STRIPE_OP_BIOFILL, &sh->ops.ack); + clear_bit(STRIPE_OP_BIOFILL, &sh->ops.pending); + + return_io(return_bi); + + if (more_to_read) + set_bit(STRIPE_HANDLE, &sh->state); + release_stripe(sh); +} + +static void ops_run_biofill(struct stripe_head *sh) +{ + struct dma_async_tx_descriptor *tx = NULL; + raid5_conf_t *conf = sh->raid_conf; + int i; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + for (i = sh->disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (test_bit(R5_Wantfill, &dev->flags)) { + struct bio *rbi; + spin_lock_irq(&conf->device_lock); + dev->read = rbi = dev->toread; + dev->toread = NULL; + spin_unlock_irq(&conf->device_lock); + while (rbi && rbi->bi_sector < + dev->sector + STRIPE_SECTORS) { + tx = async_copy_data(0, rbi, dev->page, + dev->sector, tx); + rbi = r5_next_bio(rbi, dev->sector); + } + } + } + + atomic_inc(&sh->count); + async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx, + ops_complete_biofill, sh); +} + +static void ops_complete_compute5(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + int target = sh->ops.target; + struct r5dev *tgt = &sh->dev[target]; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + set_bit(R5_UPTODATE, &tgt->flags); + BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags)); + clear_bit(R5_Wantcompute, &tgt->flags); + set_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete); + set_bit(STRIPE_HANDLE, &sh->state); + release_stripe(sh); +} + +static struct dma_async_tx_descriptor * +ops_run_compute5(struct stripe_head *sh, unsigned long pending) +{ + /* kernel stack size limits the total number of disks */ + int disks = sh->disks; + struct page *xor_srcs[disks]; + int target = sh->ops.target; + struct r5dev *tgt = &sh->dev[target]; + struct page *xor_dest = tgt->page; + int count = 0; + struct dma_async_tx_descriptor *tx; + int i; + + pr_debug("%s: stripe %llu block: %d\n", + __FUNCTION__, (unsigned long long)sh->sector, target); + BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags)); + + for (i = disks; i--; ) + if (i != target) + xor_srcs[count++] = sh->dev[i].page; + + atomic_inc(&sh->count); + + if (unlikely(count == 1)) + tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, + 0, NULL, ops_complete_compute5, sh); + else + tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, + ASYNC_TX_XOR_ZERO_DST, NULL, + ops_complete_compute5, sh); + + /* ack now if postxor is not set to be run */ + if (tx && !test_bit(STRIPE_OP_POSTXOR, &pending)) + async_tx_ack(tx); + + return tx; +} + +static void ops_complete_prexor(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + set_bit(STRIPE_OP_PREXOR, &sh->ops.complete); +} + +static struct dma_async_tx_descriptor * +ops_run_prexor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx) +{ + /* kernel stack size limits the total number of disks */ + int disks = sh->disks; + struct page *xor_srcs[disks]; + int count = 0, pd_idx = sh->pd_idx, i; + + /* existing parity data subtracted */ + struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + /* Only process blocks that are known to be uptodate */ + if (dev->towrite && test_bit(R5_Wantprexor, &dev->flags)) + xor_srcs[count++] = dev->page; + } + + tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, + ASYNC_TX_DEP_ACK | ASYNC_TX_XOR_DROP_DST, tx, + ops_complete_prexor, sh); + + return tx; +} + +static struct dma_async_tx_descriptor * +ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx) +{ + int disks = sh->disks; + int pd_idx = sh->pd_idx, i; + + /* check if prexor is active which means only process blocks + * that are part of a read-modify-write (Wantprexor) + */ + int prexor = test_bit(STRIPE_OP_PREXOR, &sh->ops.pending); + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + struct bio *chosen; + int towrite; + + towrite = 0; + if (prexor) { /* rmw */ + if (dev->towrite && + test_bit(R5_Wantprexor, &dev->flags)) + towrite = 1; + } else { /* rcw */ + if (i != pd_idx && dev->towrite && + test_bit(R5_LOCKED, &dev->flags)) + towrite = 1; + } + + if (towrite) { + struct bio *wbi; + + spin_lock(&sh->lock); + chosen = dev->towrite; + dev->towrite = NULL; + BUG_ON(dev->written); + wbi = dev->written = chosen; + spin_unlock(&sh->lock); + + while (wbi && wbi->bi_sector < + dev->sector + STRIPE_SECTORS) { + tx = async_copy_data(1, wbi, dev->page, + dev->sector, tx); + wbi = r5_next_bio(wbi, dev->sector); + } + } + } + + return tx; +} + +static void ops_complete_postxor(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + set_bit(STRIPE_OP_POSTXOR, &sh->ops.complete); + set_bit(STRIPE_HANDLE, &sh->state); + release_stripe(sh); +} + +static void ops_complete_write(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + int disks = sh->disks, i, pd_idx = sh->pd_idx; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (dev->written || i == pd_idx) + set_bit(R5_UPTODATE, &dev->flags); + } + + set_bit(STRIPE_OP_BIODRAIN, &sh->ops.complete); + set_bit(STRIPE_OP_POSTXOR, &sh->ops.complete); + + set_bit(STRIPE_HANDLE, &sh->state); + release_stripe(sh); +} + +static void +ops_run_postxor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx) +{ + /* kernel stack size limits the total number of disks */ + int disks = sh->disks; + struct page *xor_srcs[disks]; + + int count = 0, pd_idx = sh->pd_idx, i; + struct page *xor_dest; + int prexor = test_bit(STRIPE_OP_PREXOR, &sh->ops.pending); + unsigned long flags; + dma_async_tx_callback callback; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + /* check if prexor is active which means only process blocks + * that are part of a read-modify-write (written) + */ + if (prexor) { + xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page; + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (dev->written) + xor_srcs[count++] = dev->page; + } + } else { + xor_dest = sh->dev[pd_idx].page; + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (i != pd_idx) + xor_srcs[count++] = dev->page; + } + } + + /* check whether this postxor is part of a write */ + callback = test_bit(STRIPE_OP_BIODRAIN, &sh->ops.pending) ? + ops_complete_write : ops_complete_postxor; + + /* 1/ if we prexor'd then the dest is reused as a source + * 2/ if we did not prexor then we are redoing the parity + * set ASYNC_TX_XOR_DROP_DST and ASYNC_TX_XOR_ZERO_DST + * for the synchronous xor case + */ + flags = ASYNC_TX_DEP_ACK | ASYNC_TX_ACK | + (prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST); + + atomic_inc(&sh->count); + + if (unlikely(count == 1)) { + flags &= ~(ASYNC_TX_XOR_DROP_DST | ASYNC_TX_XOR_ZERO_DST); + tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, + flags, tx, callback, sh); + } else + tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, + flags, tx, callback, sh); +} + +static void ops_complete_check(void *stripe_head_ref) +{ + struct stripe_head *sh = stripe_head_ref; + int pd_idx = sh->pd_idx; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + if (test_and_clear_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending) && + sh->ops.zero_sum_result == 0) + set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); + + set_bit(STRIPE_OP_CHECK, &sh->ops.complete); + set_bit(STRIPE_HANDLE, &sh->state); + release_stripe(sh); +} + +static void ops_run_check(struct stripe_head *sh) +{ + /* kernel stack size limits the total number of disks */ + int disks = sh->disks; + struct page *xor_srcs[disks]; + struct dma_async_tx_descriptor *tx; + + int count = 0, pd_idx = sh->pd_idx, i; + struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page; + + pr_debug("%s: stripe %llu\n", __FUNCTION__, + (unsigned long long)sh->sector); + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (i != pd_idx) + xor_srcs[count++] = dev->page; + } + + tx = async_xor_zero_sum(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, + &sh->ops.zero_sum_result, 0, NULL, NULL, NULL); + + if (tx) + set_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending); + else + clear_bit(STRIPE_OP_MOD_DMA_CHECK, &sh->ops.pending); + + atomic_inc(&sh->count); + tx = async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx, + ops_complete_check, sh); +} + +static void raid5_run_ops(struct stripe_head *sh, unsigned long pending) +{ + int overlap_clear = 0, i, disks = sh->disks; + struct dma_async_tx_descriptor *tx = NULL; + + if (test_bit(STRIPE_OP_BIOFILL, &pending)) { + ops_run_biofill(sh); + overlap_clear++; + } + + if (test_bit(STRIPE_OP_COMPUTE_BLK, &pending)) + tx = ops_run_compute5(sh, pending); + + if (test_bit(STRIPE_OP_PREXOR, &pending)) + tx = ops_run_prexor(sh, tx); + + if (test_bit(STRIPE_OP_BIODRAIN, &pending)) { + tx = ops_run_biodrain(sh, tx); + overlap_clear++; + } + + if (test_bit(STRIPE_OP_POSTXOR, &pending)) + ops_run_postxor(sh, tx); + + if (test_bit(STRIPE_OP_CHECK, &pending)) + ops_run_check(sh); + + if (test_bit(STRIPE_OP_IO, &pending)) + ops_run_io(sh); + + if (overlap_clear) + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (test_and_clear_bit(R5_Overlap, &dev->flags)) + wake_up(&sh->raid_conf->wait_for_overlap); + } +} + static int grow_one_stripe(raid5_conf_t *conf) { struct stripe_head *sh; diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index b99d354..6fb9d94 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -116,13 +116,46 @@ * attach a request to an active stripe (add_stripe_bh()) * lockdev attach-buffer unlockdev * handle a stripe (handle_stripe()) - * lockstripe clrSTRIPE_HANDLE ... (lockdev check-buffers unlockdev) .. change-state .. record io needed unlockstripe schedule io + * lockstripe clrSTRIPE_HANDLE ... + * (lockdev check-buffers unlockdev) .. + * change-state .. + * record io/ops needed unlockstripe schedule io/ops * release an active stripe (release_stripe()) * lockdev if (!--cnt) { if STRIPE_HANDLE, add to handle_list else add to inactive-list } unlockdev * * The refcount counts each thread that have activated the stripe, * plus raid5d if it is handling it, plus one for each active request - * on a cached buffer. + * on a cached buffer, and plus one if the stripe is undergoing stripe + * operations. + * + * Stripe operations are performed outside the stripe lock, + * the stripe operations are: + * -copying data between the stripe cache and user application buffers + * -computing blocks to save a disk access, or to recover a missing block + * -updating the parity on a write operation (reconstruct write and + * read-modify-write) + * -checking parity correctness + * -running i/o to disk + * These operations are carried out by raid5_run_ops which uses the async_tx + * api to (optionally) offload operations to dedicated hardware engines. + * When requesting an operation handle_stripe sets the pending bit for the + * operation and increments the count. raid5_run_ops is then run whenever + * the count is non-zero. + * There are some critical dependencies between the operations that prevent some + * from being requested while another is in flight. + * 1/ Parity check operations destroy the in cache version of the parity block, + * so we prevent parity dependent operations like writes and compute_blocks + * from starting while a check is in progress. Some dma engines can perform + * the check without damaging the parity block, in these cases the parity + * block is re-marked up to date (assuming the check was successful) and is + * not re-read from disk. + * 2/ When a write operation is requested we immediately lock the affected + * blocks, and mark them as not up to date. This causes new read requests + * to be held off, as well as parity checks and compute block operations. + * 3/ Once a compute block operation has been requested handle_stripe treats + * that block as if it is up to date. raid5_run_ops guaruntees that any + * operation that is dependent on the compute block result is initiated after + * the compute block completes. */ struct stripe_head { @@ -136,11 +169,26 @@ struct stripe_head { spinlock_t lock; int bm_seq; /* sequence number for bitmap flushes */ int disks; /* disks in stripe */ + /* stripe_operations + * @pending - pending ops flags (set for request->issue->complete) + * @ack - submitted ops flags (set for issue->complete) + * @complete - completed ops flags (set for complete) + * @target - STRIPE_OP_COMPUTE_BLK target + * @count - raid5_runs_ops is set to run when this is non-zero + */ + struct stripe_operations { + unsigned long pending; + unsigned long ack; + unsigned long complete; + int target; + int count; + u32 zero_sum_result; + } ops; struct r5dev { struct bio req; struct bio_vec vec; struct page *page; - struct bio *toread, *towrite, *written; + struct bio *toread, *read, *towrite, *written; sector_t sector; /* sector of this page */ unsigned long flags; } dev[1]; /* allocated with extra space depending of RAID geometry */ @@ -174,6 +222,15 @@ struct r6_state { #define R5_ReWrite 9 /* have tried to over-write the readerror */ #define R5_Expanded 10 /* This block now has post-expand data */ +#define R5_Wantcompute 11 /* compute_block in progress treat as + * uptodate + */ +#define R5_Wantfill 12 /* dev->toread contains a bio that needs + * filling + */ +#define R5_Wantprexor 13 /* distinguish blocks ready for rmw from + * other "towrites" + */ /* * Write method */ @@ -196,6 +253,24 @@ struct r6_state { #define STRIPE_EXPAND_SOURCE 10 #define STRIPE_EXPAND_READY 11 /* + * Operations flags (in issue order) + */ +#define STRIPE_OP_BIOFILL 0 +#define STRIPE_OP_COMPUTE_BLK 1 +#define STRIPE_OP_PREXOR 2 +#define STRIPE_OP_BIODRAIN 3 +#define STRIPE_OP_POSTXOR 4 +#define STRIPE_OP_CHECK 5 +#define STRIPE_OP_IO 6 + +/* modifiers to the base operations + * STRIPE_OP_MOD_REPAIR_PD - compute the parity block and write it back + * STRIPE_OP_MOD_DMA_CHECK - parity is not corrupted by the check + */ +#define STRIPE_OP_MOD_REPAIR_PD 7 +#define STRIPE_OP_MOD_DMA_CHECK 8 + +/* * Plugging: * * To improve write throughput, we need to delay the handling of some -- cgit v1.1 From d84e0f10d38393f617227f0c831a99c69294651f Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:30 -0700 Subject: md: common infrastructure for running operations with raid5_run_ops All the handle_stripe operations that are to be transitioned to use raid5_run_ops need a method to coherently gather work under the stripe-lock and hand that work off to raid5_run_ops. The 'get_stripe_work' routine runs under the lock to read all the bits in sh->ops.pending that do not have the corresponding bit set in sh->ops.ack. This modified 'pending' bitmap is then passed to raid5_run_ops for processing. The transition from 'ack' to 'completion' does not need similar protection as the existing release_stripe infrastructure will guarantee that handle_stripe will run again after a completion bit is set, and handle_stripe can tolerate a sh->ops.completed bit being set while the lock is held. A call to async_tx_issue_pending_all() is added to raid5d to kick the offload engines once all pending stripe operations work has been submitted. This enables batching of the submission and completion of operations. Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 67 ++++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 58 insertions(+), 9 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 0b70024..d89a25e 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -141,6 +141,7 @@ static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh) } md_wakeup_thread(conf->mddev->thread); } else { + BUG_ON(sh->ops.pending); if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { atomic_dec(&conf->preread_active_stripes); if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) @@ -242,7 +243,8 @@ static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int BUG_ON(atomic_read(&sh->count) != 0); BUG_ON(test_bit(STRIPE_HANDLE, &sh->state)); - + BUG_ON(sh->ops.pending || sh->ops.ack || sh->ops.complete); + CHECK_DEVLOCK(); pr_debug("init_stripe called, stripe %llu\n", (unsigned long long)sh->sector); @@ -258,11 +260,11 @@ static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int for (i = sh->disks; i--; ) { struct r5dev *dev = &sh->dev[i]; - if (dev->toread || dev->towrite || dev->written || + if (dev->toread || dev->read || dev->towrite || dev->written || test_bit(R5_LOCKED, &dev->flags)) { - printk("sector=%llx i=%d %p %p %p %d\n", + printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n", (unsigned long long)sh->sector, i, dev->toread, - dev->towrite, dev->written, + dev->read, dev->towrite, dev->written, test_bit(R5_LOCKED, &dev->flags)); BUG(); } @@ -342,6 +344,44 @@ static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector return sh; } +/* test_and_ack_op() ensures that we only dequeue an operation once */ +#define test_and_ack_op(op, pend) \ +do { \ + if (test_bit(op, &sh->ops.pending) && \ + !test_bit(op, &sh->ops.complete)) { \ + if (test_and_set_bit(op, &sh->ops.ack)) \ + clear_bit(op, &pend); \ + else \ + ack++; \ + } else \ + clear_bit(op, &pend); \ +} while (0) + +/* find new work to run, do not resubmit work that is already + * in flight + */ +static unsigned long get_stripe_work(struct stripe_head *sh) +{ + unsigned long pending; + int ack = 0; + + pending = sh->ops.pending; + + test_and_ack_op(STRIPE_OP_BIOFILL, pending); + test_and_ack_op(STRIPE_OP_COMPUTE_BLK, pending); + test_and_ack_op(STRIPE_OP_PREXOR, pending); + test_and_ack_op(STRIPE_OP_BIODRAIN, pending); + test_and_ack_op(STRIPE_OP_POSTXOR, pending); + test_and_ack_op(STRIPE_OP_CHECK, pending); + if (test_and_clear_bit(STRIPE_OP_IO, &sh->ops.pending)) + ack++; + + sh->ops.count -= ack; + BUG_ON(sh->ops.count < 0); + + return pending; +} + static int raid5_end_read_request(struct bio *bi, unsigned int bytes_done, int error); static int @@ -2494,7 +2534,6 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh, * schedule a write of some buffers * return confirmation of parity correctness * - * Parity calculations are done inside the stripe lock * buffers are taken off read_list or write_list, and bh_cache buffers * get BH_Lock set before the stripe lock is released. * @@ -2507,11 +2546,13 @@ static void handle_stripe5(struct stripe_head *sh) struct bio *return_bi = NULL; struct stripe_head_state s; struct r5dev *dev; + unsigned long pending = 0; memset(&s, 0, sizeof(s)); - pr_debug("handling stripe %llu, cnt=%d, pd_idx=%d\n", - (unsigned long long)sh->sector, atomic_read(&sh->count), - sh->pd_idx); + pr_debug("handling stripe %llu, state=%#lx cnt=%d, pd_idx=%d " + "ops=%lx:%lx:%lx\n", (unsigned long long)sh->sector, sh->state, + atomic_read(&sh->count), sh->pd_idx, + sh->ops.pending, sh->ops.ack, sh->ops.complete); spin_lock(&sh->lock); clear_bit(STRIPE_HANDLE, &sh->state); @@ -2674,8 +2715,14 @@ static void handle_stripe5(struct stripe_head *sh) if (s.expanding && s.locked == 0) handle_stripe_expansion(conf, sh, NULL); + if (sh->ops.count) + pending = get_stripe_work(sh); + spin_unlock(&sh->lock); + if (pending) + raid5_run_ops(sh, pending); + return_io(return_bi); for (i=disks; i-- ;) { @@ -3798,8 +3845,10 @@ static void raid5d (mddev_t *mddev) handled++; } - if (list_empty(&conf->handle_list)) + if (list_empty(&conf->handle_list)) { + async_tx_issue_pending_all(); break; + } first = conf->handle_list.next; sh = list_entry(first, struct stripe_head, lru); -- cgit v1.1 From e33129d84130459dbb764a1a52a4bfceab3da978 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:30 -0700 Subject: md: handle_stripe5 - add request/completion logic for async write ops After handle_stripe5 decides whether it wants to perform a read-modify-write, or a reconstruct write it calls handle_write_operations5. A read-modify-write operation will perform an xor subtraction of the blocks marked with the R5_Wantprexor flag, copy the new data into the stripe (biodrain) and perform a postxor operation across all up-to-date blocks to generate the new parity. A reconstruct write is run when all blocks are already up-to-date in the cache so all that is needed is a biodrain and postxor. On the completion path STRIPE_OP_PREXOR will be set if the operation was a read-modify-write. The STRIPE_OP_BIODRAIN flag is used in the completion path to differentiate write-initiated postxor operations versus expansion-initiated postxor operations. Completion of a write triggers i/o to the drives. Changelog: * make the 'rcw' parameter to handle_write_operations5 a simple flag, Neil Brown * remove test_and_set/test_and_clear BUG_ONs, Neil Brown Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 161 +++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 138 insertions(+), 23 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index d89a25e..d9521aa 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -1822,7 +1822,79 @@ static void compute_block_2(struct stripe_head *sh, int dd_idx1, int dd_idx2) } } +static int +handle_write_operations5(struct stripe_head *sh, int rcw, int expand) +{ + int i, pd_idx = sh->pd_idx, disks = sh->disks; + int locked = 0; + + if (rcw) { + /* if we are not expanding this is a proper write request, and + * there will be bios with new data to be drained into the + * stripe cache + */ + if (!expand) { + set_bit(STRIPE_OP_BIODRAIN, &sh->ops.pending); + sh->ops.count++; + } + + set_bit(STRIPE_OP_POSTXOR, &sh->ops.pending); + sh->ops.count++; + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + + if (dev->towrite) { + set_bit(R5_LOCKED, &dev->flags); + if (!expand) + clear_bit(R5_UPTODATE, &dev->flags); + locked++; + } + } + } else { + BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) || + test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags))); + + set_bit(STRIPE_OP_PREXOR, &sh->ops.pending); + set_bit(STRIPE_OP_BIODRAIN, &sh->ops.pending); + set_bit(STRIPE_OP_POSTXOR, &sh->ops.pending); + + sh->ops.count += 3; + + for (i = disks; i--; ) { + struct r5dev *dev = &sh->dev[i]; + if (i == pd_idx) + continue; + + /* For a read-modify write there may be blocks that are + * locked for reading while others are ready to be + * written so we distinguish these blocks by the + * R5_Wantprexor bit + */ + if (dev->towrite && + (test_bit(R5_UPTODATE, &dev->flags) || + test_bit(R5_Wantcompute, &dev->flags))) { + set_bit(R5_Wantprexor, &dev->flags); + set_bit(R5_LOCKED, &dev->flags); + clear_bit(R5_UPTODATE, &dev->flags); + locked++; + } + } + } + + /* keep the parity disk locked while asynchronous operations + * are in flight + */ + set_bit(R5_LOCKED, &sh->dev[pd_idx].flags); + clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); + locked++; + pr_debug("%s: stripe %llu locked: %d pending: %lx\n", + __FUNCTION__, (unsigned long long)sh->sector, + locked, sh->ops.pending); + + return locked; +} /* * Each stripe/dev can have one or more bion attached. @@ -2217,27 +2289,8 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, * we can start a write request */ if (s->locked == 0 && (rcw == 0 || rmw == 0) && - !test_bit(STRIPE_BIT_DELAY, &sh->state)) { - pr_debug("Computing parity...\n"); - compute_parity5(sh, rcw == 0 ? - RECONSTRUCT_WRITE : READ_MODIFY_WRITE); - /* now every locked buffer is ready to be written */ - for (i = disks; i--; ) - if (test_bit(R5_LOCKED, &sh->dev[i].flags)) { - pr_debug("Writing block %d\n", i); - s->locked++; - set_bit(R5_Wantwrite, &sh->dev[i].flags); - if (!test_bit(R5_Insync, &sh->dev[i].flags) - || (i == sh->pd_idx && s->failed == 0)) - set_bit(STRIPE_INSYNC, &sh->state); - } - if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { - atomic_dec(&conf->preread_active_stripes); - if (atomic_read(&conf->preread_active_stripes) < - IO_THRESHOLD) - md_wakeup_thread(conf->mddev->thread); - } - } + !test_bit(STRIPE_BIT_DELAY, &sh->state)) + s->locked += handle_write_operations5(sh, rcw == 0, 0); } static void handle_issuing_new_write_requests6(raid5_conf_t *conf, @@ -2656,8 +2709,70 @@ static void handle_stripe5(struct stripe_head *sh) (s.syncing && (s.uptodate < disks)) || s.expanding) handle_issuing_new_read_requests5(sh, &s, disks); - /* now to consider writing and what else, if anything should be read */ - if (s.to_write) + /* Now we check to see if any write operations have recently + * completed + */ + + /* leave prexor set until postxor is done, allows us to distinguish + * a rmw from a rcw during biodrain + */ + if (test_bit(STRIPE_OP_PREXOR, &sh->ops.complete) && + test_bit(STRIPE_OP_POSTXOR, &sh->ops.complete)) { + + clear_bit(STRIPE_OP_PREXOR, &sh->ops.complete); + clear_bit(STRIPE_OP_PREXOR, &sh->ops.ack); + clear_bit(STRIPE_OP_PREXOR, &sh->ops.pending); + + for (i = disks; i--; ) + clear_bit(R5_Wantprexor, &sh->dev[i].flags); + } + + /* if only POSTXOR is set then this is an 'expand' postxor */ + if (test_bit(STRIPE_OP_BIODRAIN, &sh->ops.complete) && + test_bit(STRIPE_OP_POSTXOR, &sh->ops.complete)) { + + clear_bit(STRIPE_OP_BIODRAIN, &sh->ops.complete); + clear_bit(STRIPE_OP_BIODRAIN, &sh->ops.ack); + clear_bit(STRIPE_OP_BIODRAIN, &sh->ops.pending); + + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.complete); + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.ack); + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.pending); + + /* All the 'written' buffers and the parity block are ready to + * be written back to disk + */ + BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags)); + for (i = disks; i--; ) { + dev = &sh->dev[i]; + if (test_bit(R5_LOCKED, &dev->flags) && + (i == sh->pd_idx || dev->written)) { + pr_debug("Writing block %d\n", i); + set_bit(R5_Wantwrite, &dev->flags); + if (!test_and_set_bit( + STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; + if (!test_bit(R5_Insync, &dev->flags) || + (i == sh->pd_idx && s.failed == 0)) + set_bit(STRIPE_INSYNC, &sh->state); + } + } + if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { + atomic_dec(&conf->preread_active_stripes); + if (atomic_read(&conf->preread_active_stripes) < + IO_THRESHOLD) + md_wakeup_thread(conf->mddev->thread); + } + } + + /* Now to consider new write requests and what else, if anything + * should be read. We do not handle new writes when: + * 1/ A 'write' operation (copy+xor) is already in flight. + * 2/ A 'check' operation is in flight, as it may clobber the parity + * block. + */ + if (s.to_write && !test_bit(STRIPE_OP_POSTXOR, &sh->ops.pending) && + !test_bit(STRIPE_OP_CHECK, &sh->ops.pending)) handle_issuing_new_write_requests5(conf, sh, &s, disks); /* maybe we need to check and possibly fix the parity for this stripe -- cgit v1.1 From f38e12199a94ca458e4f03c5a2c984fb80adadc5 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:30 -0700 Subject: md: handle_stripe5 - add request/completion logic for async compute ops handle_stripe will compute a block when a backing disk has failed, or when it determines it can save a disk read by computing the block from all the other up-to-date blocks. Previously a block would be computed under the lock and subsequent logic in handle_stripe could use the newly up-to-date block. With the raid5_run_ops implementation the compute operation is carried out a later time outside the lock. To preserve the old functionality we take advantage of the dependency chain feature of async_tx to flag the block as R5_Wantcompute and then let other parts of handle_stripe operate on the block as if it were up-to-date. raid5_run_ops guarantees that the block will be ready before it is used in another operation. However, this only works in cases where the compute and the dependent operation are scheduled at the same time. If a previous call to handle_stripe sets the R5_Wantcompute flag there is no facility to pass the async_tx dependency chain across successive calls to raid5_run_ops. The req_compute variable protects against this case. Changelog: * remove the req_compute BUG_ON Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 149 ++++++++++++++++++++++++++++++++++----------- include/linux/raid/raid5.h | 2 +- 2 files changed, 115 insertions(+), 36 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index d9521aa..42439a4 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2077,36 +2077,101 @@ handle_requests_to_failed_array(raid5_conf_t *conf, struct stripe_head *sh, } +/* __handle_issuing_new_read_requests5 - returns 0 if there are no more disks + * to process + */ +static int __handle_issuing_new_read_requests5(struct stripe_head *sh, + struct stripe_head_state *s, int disk_idx, int disks) +{ + struct r5dev *dev = &sh->dev[disk_idx]; + struct r5dev *failed_dev = &sh->dev[s->failed_num]; + + /* don't schedule compute operations or reads on the parity block while + * a check is in flight + */ + if ((disk_idx == sh->pd_idx) && + test_bit(STRIPE_OP_CHECK, &sh->ops.pending)) + return ~0; + + /* is the data in this block needed, and can we get it? */ + if (!test_bit(R5_LOCKED, &dev->flags) && + !test_bit(R5_UPTODATE, &dev->flags) && (dev->toread || + (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) || + s->syncing || s->expanding || (s->failed && + (failed_dev->toread || (failed_dev->towrite && + !test_bit(R5_OVERWRITE, &failed_dev->flags) + ))))) { + /* 1/ We would like to get this block, possibly by computing it, + * but we might not be able to. + * + * 2/ Since parity check operations potentially make the parity + * block !uptodate it will need to be refreshed before any + * compute operations on data disks are scheduled. + * + * 3/ We hold off parity block re-reads until check operations + * have quiesced. + */ + if ((s->uptodate == disks - 1) && + !test_bit(STRIPE_OP_CHECK, &sh->ops.pending)) { + set_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending); + set_bit(R5_Wantcompute, &dev->flags); + sh->ops.target = disk_idx; + s->req_compute = 1; + sh->ops.count++; + /* Careful: from this point on 'uptodate' is in the eye + * of raid5_run_ops which services 'compute' operations + * before writes. R5_Wantcompute flags a block that will + * be R5_UPTODATE by the time it is needed for a + * subsequent operation. + */ + s->uptodate++; + return 0; /* uptodate + compute == disks */ + } else if ((s->uptodate < disks - 1) && + test_bit(R5_Insync, &dev->flags)) { + /* Note: we hold off compute operations while checks are + * in flight, but we still prefer 'compute' over 'read' + * hence we only read if (uptodate < * disks-1) + */ + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantread, &dev->flags); + if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; + s->locked++; + pr_debug("Reading block %d (sync=%d)\n", disk_idx, + s->syncing); + } + } + + return ~0; +} + static void handle_issuing_new_read_requests5(struct stripe_head *sh, struct stripe_head_state *s, int disks) { int i; - for (i = disks; i--; ) { - struct r5dev *dev = &sh->dev[i]; - if (!test_bit(R5_LOCKED, &dev->flags) && - !test_bit(R5_UPTODATE, &dev->flags) && - (dev->toread || - (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) || - s->syncing || s->expanding || - (s->failed && (sh->dev[s->failed_num].toread || - (sh->dev[s->failed_num].towrite && - !test_bit(R5_OVERWRITE, &sh->dev[s->failed_num].flags)) - )))) { - /* we would like to get this block, possibly - * by computing it, but we might not be able to - */ - if (s->uptodate == disks-1) { - pr_debug("Computing block %d\n", i); - compute_block(sh, i); - s->uptodate++; - } else if (test_bit(R5_Insync, &dev->flags)) { - set_bit(R5_LOCKED, &dev->flags); - set_bit(R5_Wantread, &dev->flags); - s->locked++; - pr_debug("Reading block %d (sync=%d)\n", - i, s->syncing); - } - } + + /* Clear completed compute operations. Parity recovery + * (STRIPE_OP_MOD_REPAIR_PD) implies a write-back which is handled + * later on in this routine + */ + if (test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete) && + !test_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending)) { + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete); + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.ack); + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending); + } + + /* look for blocks to read/compute, skip this if a compute + * is already in flight, or if the stripe contents are in the + * midst of changing due to a write + */ + if (!test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending) && + !test_bit(STRIPE_OP_PREXOR, &sh->ops.pending) && + !test_bit(STRIPE_OP_POSTXOR, &sh->ops.pending)) { + for (i = disks; i--; ) + if (__handle_issuing_new_read_requests5( + sh, s, i, disks) == 0) + break; } set_bit(STRIPE_HANDLE, &sh->state); } @@ -2223,7 +2288,8 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, struct r5dev *dev = &sh->dev[i]; if ((dev->towrite || i == sh->pd_idx) && !test_bit(R5_LOCKED, &dev->flags) && - !test_bit(R5_UPTODATE, &dev->flags)) { + !(test_bit(R5_UPTODATE, &dev->flags) || + test_bit(R5_Wantcompute, &dev->flags))) { if (test_bit(R5_Insync, &dev->flags)) rmw++; else @@ -2232,9 +2298,9 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, /* Would I have to read this buffer for reconstruct_write */ if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx && !test_bit(R5_LOCKED, &dev->flags) && - !test_bit(R5_UPTODATE, &dev->flags)) { - if (test_bit(R5_Insync, &dev->flags)) - rcw++; + !(test_bit(R5_UPTODATE, &dev->flags) || + test_bit(R5_Wantcompute, &dev->flags))) { + if (test_bit(R5_Insync, &dev->flags)) rcw++; else rcw += 2*disks; } @@ -2248,7 +2314,8 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, struct r5dev *dev = &sh->dev[i]; if ((dev->towrite || i == sh->pd_idx) && !test_bit(R5_LOCKED, &dev->flags) && - !test_bit(R5_UPTODATE, &dev->flags) && + !(test_bit(R5_UPTODATE, &dev->flags) || + test_bit(R5_Wantcompute, &dev->flags)) && test_bit(R5_Insync, &dev->flags)) { if ( test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { @@ -2270,7 +2337,8 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx && !test_bit(R5_LOCKED, &dev->flags) && - !test_bit(R5_UPTODATE, &dev->flags) && + !(test_bit(R5_UPTODATE, &dev->flags) || + test_bit(R5_Wantcompute, &dev->flags)) && test_bit(R5_Insync, &dev->flags)) { if ( test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { @@ -2288,8 +2356,17 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, /* now if nothing is locked, and if we have enough data, * we can start a write request */ - if (s->locked == 0 && (rcw == 0 || rmw == 0) && - !test_bit(STRIPE_BIT_DELAY, &sh->state)) + /* since handle_stripe can be called at any time we need to handle the + * case where a compute block operation has been submitted and then a + * subsequent call wants to start a write request. raid5_run_ops only + * handles the case where compute block and postxor are requested + * simultaneously. If this is not the case then new writes need to be + * held off until the compute completes. + */ + if ((s->req_compute || + !test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending)) && + (s->locked == 0 && (rcw == 0 || rmw == 0) && + !test_bit(STRIPE_BIT_DELAY, &sh->state))) s->locked += handle_write_operations5(sh, rcw == 0, 0); } @@ -2650,6 +2727,7 @@ static void handle_stripe5(struct stripe_head *sh) /* now count some things */ if (test_bit(R5_LOCKED, &dev->flags)) s.locked++; if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++; + if (test_bit(R5_Wantcompute, &dev->flags)) s.compute++; if (dev->toread) s.to_read++; @@ -2706,7 +2784,8 @@ static void handle_stripe5(struct stripe_head *sh) * or to load a block that is being partially written. */ if (s.to_read || s.non_overwrite || - (s.syncing && (s.uptodate < disks)) || s.expanding) + (s.syncing && (s.uptodate + s.compute < disks)) || s.expanding || + test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending)) handle_issuing_new_read_requests5(sh, &s, disks); /* Now we check to see if any write operations have recently diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index 6fb9d94..2293015 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -200,7 +200,7 @@ struct stripe_head { struct stripe_head_state { int syncing, expanding, expanded; int locked, uptodate, to_read, to_write, failed, written; - int non_overwrite; + int compute, req_compute, non_overwrite; int failed_num; }; -- cgit v1.1 From e89f89629b5de76e504d1be75c82c4a6b2419583 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: md: handle_stripe5 - add request/completion logic for async check ops Check operations are scheduled when the array is being resynced or an explicit 'check/repair' command was sent to the array. Previously check operations would destroy the parity block in the cache such that even if parity turned out to be correct the parity block would be marked !R5_UPTODATE at the completion of the check. When the operation can be carried out by a dma engine the assumption is that it can check parity as a read-only operation. If raid5_run_ops notices that the check was handled by hardware it will preserve the R5_UPTODATE status of the parity disk. When a check operation determines that the parity needs to be repaired we reuse the existing compute block infrastructure to carry out the operation. Repair operations imply an immediate write back of the data, so to differentiate a repair from a normal compute operation the STRIPE_OP_MOD_REPAIR_PD flag is added. Changelog: * remove test_and_set/test_and_clear BUG_ONs, Neil Brown Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 84 ++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 65 insertions(+), 19 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 42439a4..810cf83 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2471,26 +2471,67 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh, struct stripe_head_state *s, int disks) { set_bit(STRIPE_HANDLE, &sh->state); - if (s->failed == 0) { - BUG_ON(s->uptodate != disks); - compute_parity5(sh, CHECK_PARITY); - s->uptodate--; - if (page_is_zero(sh->dev[sh->pd_idx].page)) { - /* parity is correct (on disc, not in buffer any more) - */ - set_bit(STRIPE_INSYNC, &sh->state); - } else { - conf->mddev->resync_mismatches += STRIPE_SECTORS; - if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) - /* don't try to repair!! */ + /* Take one of the following actions: + * 1/ start a check parity operation if (uptodate == disks) + * 2/ finish a check parity operation and act on the result + * 3/ skip to the writeback section if we previously + * initiated a recovery operation + */ + if (s->failed == 0 && + !test_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending)) { + if (!test_and_set_bit(STRIPE_OP_CHECK, &sh->ops.pending)) { + BUG_ON(s->uptodate != disks); + clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags); + sh->ops.count++; + s->uptodate--; + } else if ( + test_and_clear_bit(STRIPE_OP_CHECK, &sh->ops.complete)) { + clear_bit(STRIPE_OP_CHECK, &sh->ops.ack); + clear_bit(STRIPE_OP_CHECK, &sh->ops.pending); + + if (sh->ops.zero_sum_result == 0) + /* parity is correct (on disc, + * not in buffer any more) + */ set_bit(STRIPE_INSYNC, &sh->state); else { - compute_block(sh, sh->pd_idx); - s->uptodate++; + conf->mddev->resync_mismatches += + STRIPE_SECTORS; + if (test_bit( + MD_RECOVERY_CHECK, &conf->mddev->recovery)) + /* don't try to repair!! */ + set_bit(STRIPE_INSYNC, &sh->state); + else { + set_bit(STRIPE_OP_COMPUTE_BLK, + &sh->ops.pending); + set_bit(STRIPE_OP_MOD_REPAIR_PD, + &sh->ops.pending); + set_bit(R5_Wantcompute, + &sh->dev[sh->pd_idx].flags); + sh->ops.target = sh->pd_idx; + sh->ops.count++; + s->uptodate++; + } } } } - if (!test_bit(STRIPE_INSYNC, &sh->state)) { + + /* check if we can clear a parity disk reconstruct */ + if (test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete) && + test_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending)) { + + clear_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending); + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.complete); + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.ack); + clear_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending); + } + + /* Wait for check parity and compute block operations to complete + * before write-back + */ + if (!test_bit(STRIPE_INSYNC, &sh->state) && + !test_bit(STRIPE_OP_CHECK, &sh->ops.pending) && + !test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending)) { struct r5dev *dev; /* either failed parity check, or recovery is happening */ if (s->failed == 0) @@ -2855,12 +2896,17 @@ static void handle_stripe5(struct stripe_head *sh) handle_issuing_new_write_requests5(conf, sh, &s, disks); /* maybe we need to check and possibly fix the parity for this stripe - * Any reads will already have been scheduled, so we just see if enough data - * is available + * Any reads will already have been scheduled, so we just see if enough + * data is available. The parity check is held off while parity + * dependent operations are in flight. */ - if (s.syncing && s.locked == 0 && - !test_bit(STRIPE_INSYNC, &sh->state)) + if ((s.syncing && s.locked == 0 && + !test_bit(STRIPE_OP_COMPUTE_BLK, &sh->ops.pending) && + !test_bit(STRIPE_INSYNC, &sh->state)) || + test_bit(STRIPE_OP_CHECK, &sh->ops.pending) || + test_bit(STRIPE_OP_MOD_REPAIR_PD, &sh->ops.pending)) handle_parity_checks5(conf, sh, &s, disks); + if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { md_done_sync(conf->mddev, STRIPE_SECTORS,1); clear_bit(STRIPE_SYNCING, &sh->state); -- cgit v1.1 From b5e98d65d34a1c11a2135ea8a9b2619dbc7216c8 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: md: handle_stripe5 - add request/completion logic for async read ops When a read bio is attached to the stripe and the corresponding block is marked R5_UPTODATE, then a read (biofill) operation is scheduled to copy the data from the stripe cache to the bio buffer. handle_stripe flags the blocks to be operated on with the R5_Wantfill flag. If new read requests arrive while raid5_run_ops is running they will not be handled until handle_stripe is scheduled to run again. Changelog: * cleanup to_read and to_fill accounting * do not fail reads that have reached the cache Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 53 ++++++++++++++++++++++------------------------ include/linux/raid/raid5.h | 2 +- 2 files changed, 26 insertions(+), 29 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 810cf83..a33dac7 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2049,9 +2049,12 @@ handle_requests_to_failed_array(raid5_conf_t *conf, struct stripe_head *sh, bi = bi2; } - /* fail any reads if this device is non-operational */ - if (!test_bit(R5_Insync, &sh->dev[i].flags) || - test_bit(R5_ReadError, &sh->dev[i].flags)) { + /* fail any reads if this device is non-operational and + * the data has not reached the cache yet. + */ + if (!test_bit(R5_Wantfill, &sh->dev[i].flags) && + (!test_bit(R5_Insync, &sh->dev[i].flags) || + test_bit(R5_ReadError, &sh->dev[i].flags))) { bi = sh->dev[i].toread; sh->dev[i].toread = NULL; if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) @@ -2740,37 +2743,27 @@ static void handle_stripe5(struct stripe_head *sh) struct r5dev *dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); - pr_debug("check %d: state 0x%lx read %p write %p written %p\n", - i, dev->flags, dev->toread, dev->towrite, dev->written); - /* maybe we can reply to a read */ - if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread) { - struct bio *rbi, *rbi2; - pr_debug("Return read for disc %d\n", i); - spin_lock_irq(&conf->device_lock); - rbi = dev->toread; - dev->toread = NULL; - if (test_and_clear_bit(R5_Overlap, &dev->flags)) - wake_up(&conf->wait_for_overlap); - spin_unlock_irq(&conf->device_lock); - while (rbi && rbi->bi_sector < dev->sector + STRIPE_SECTORS) { - copy_data(0, rbi, dev->page, dev->sector); - rbi2 = r5_next_bio(rbi, dev->sector); - spin_lock_irq(&conf->device_lock); - if (--rbi->bi_phys_segments == 0) { - rbi->bi_next = return_bi; - return_bi = rbi; - } - spin_unlock_irq(&conf->device_lock); - rbi = rbi2; - } - } + pr_debug("check %d: state 0x%lx toread %p read %p write %p " + "written %p\n", i, dev->flags, dev->toread, dev->read, + dev->towrite, dev->written); + + /* maybe we can request a biofill operation + * + * new wantfill requests are only permitted while + * STRIPE_OP_BIOFILL is clear + */ + if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread && + !test_bit(STRIPE_OP_BIOFILL, &sh->ops.pending)) + set_bit(R5_Wantfill, &dev->flags); /* now count some things */ if (test_bit(R5_LOCKED, &dev->flags)) s.locked++; if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++; if (test_bit(R5_Wantcompute, &dev->flags)) s.compute++; - if (dev->toread) + if (test_bit(R5_Wantfill, &dev->flags)) + s.to_fill++; + else if (dev->toread) s.to_read++; if (dev->towrite) { s.to_write++; @@ -2793,6 +2786,10 @@ static void handle_stripe5(struct stripe_head *sh) set_bit(R5_Insync, &dev->flags); } rcu_read_unlock(); + + if (s.to_fill && !test_and_set_bit(STRIPE_OP_BIOFILL, &sh->ops.pending)) + sh->ops.count++; + pr_debug("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d\n", s.locked, s.uptodate, s.to_read, s.to_write, diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index 2293015..93678f5 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -200,7 +200,7 @@ struct stripe_head { struct stripe_head_state { int syncing, expanding, expanded; int locked, uptodate, to_read, to_write, failed, written; - int compute, req_compute, non_overwrite; + int to_fill, compute, req_compute, non_overwrite; int failed_num; }; -- cgit v1.1 From f0a50d3754c7f1b7f05f45b1c0b35d20445316b5 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: md: handle_stripe5 - add request/completion logic for async expand ops When a stripe is being expanded bulk copying takes place to move the data from the old stripe to the new. Since raid5_run_ops only operates on one stripe at a time these bulk copies are handled in-line under the stripe lock. In the dma offload case we poll for the completion of the operation. After the data has been copied into the new stripe the parity needs to be recalculated across the new disks. We reuse the existing postxor functionality to carry out this calculation. By setting STRIPE_OP_POSTXOR without setting STRIPE_OP_BIODRAIN the completion path in handle stripe can differentiate expand operations from normal write operations. Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 50 ++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 38 insertions(+), 12 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index a33dac7..c6e0e2b 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2653,6 +2653,7 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh, /* We have read all the blocks in this stripe and now we need to * copy some of them into a target stripe for expand. */ + struct dma_async_tx_descriptor *tx = NULL; clear_bit(STRIPE_EXPAND_SOURCE, &sh->state); for (i = 0; i < sh->disks; i++) if (i != sh->pd_idx && (r6s && i != r6s->qd_idx)) { @@ -2678,9 +2679,12 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh, release_stripe(sh2); continue; } - memcpy(page_address(sh2->dev[dd_idx].page), - page_address(sh->dev[i].page), - STRIPE_SIZE); + + /* place all the copies on one channel */ + tx = async_memcpy(sh2->dev[dd_idx].page, + sh->dev[i].page, 0, 0, STRIPE_SIZE, + ASYNC_TX_DEP_ACK, tx, NULL, NULL); + set_bit(R5_Expanded, &sh2->dev[dd_idx].flags); set_bit(R5_UPTODATE, &sh2->dev[dd_idx].flags); for (j = 0; j < conf->raid_disks; j++) @@ -2693,6 +2697,12 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh, set_bit(STRIPE_HANDLE, &sh2->state); } release_stripe(sh2); + + /* done submitting copies, wait for them to complete */ + if (i + 1 >= sh->disks) { + async_tx_ack(tx); + dma_wait_for_async_tx(tx); + } } } @@ -2931,18 +2941,34 @@ static void handle_stripe5(struct stripe_head *sh) } } - if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state)) { - /* Need to write out all blocks after computing parity */ - sh->disks = conf->raid_disks; - sh->pd_idx = stripe_to_pdidx(sh->sector, conf, conf->raid_disks); - compute_parity5(sh, RECONSTRUCT_WRITE); + /* Finish postxor operations initiated by the expansion + * process + */ + if (test_bit(STRIPE_OP_POSTXOR, &sh->ops.complete) && + !test_bit(STRIPE_OP_BIODRAIN, &sh->ops.pending)) { + + clear_bit(STRIPE_EXPANDING, &sh->state); + + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.pending); + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.ack); + clear_bit(STRIPE_OP_POSTXOR, &sh->ops.complete); + for (i = conf->raid_disks; i--; ) { - set_bit(R5_LOCKED, &sh->dev[i].flags); - s.locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); + if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; } - clear_bit(STRIPE_EXPANDING, &sh->state); - } else if (s.expanded) { + } + + if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state) && + !test_bit(STRIPE_OP_POSTXOR, &sh->ops.pending)) { + /* Need to write out all blocks after computing parity */ + sh->disks = conf->raid_disks; + sh->pd_idx = stripe_to_pdidx(sh->sector, conf, + conf->raid_disks); + s.locked += handle_write_operations5(sh, 0, 1); + } else if (s.expanded && + !test_bit(STRIPE_OP_POSTXOR, &sh->ops.pending)) { clear_bit(STRIPE_EXPAND_READY, &sh->state); atomic_dec(&conf->reshape_stripes); wake_up(&conf->wait_for_overlap); -- cgit v1.1 From 830ea01673a397798d1281d2022615559f5001bb Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: md: handle_stripe5 - request io processing in raid5_run_ops I/O submission requests were already handled outside of the stripe lock in handle_stripe. Now that handle_stripe is only tasked with finding work, this logic belongs in raid5_run_ops. Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 71 ++++++++++-------------------------------------------- 1 file changed, 13 insertions(+), 58 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index c6e0e2b..7e1cc07 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2326,6 +2326,9 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, "%d for r-m-w\n", i); set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); + if (!test_and_set_bit( + STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; s->locked++; } else { set_bit(STRIPE_DELAYED, &sh->state); @@ -2349,6 +2352,9 @@ static void handle_issuing_new_write_requests5(raid5_conf_t *conf, "%d for Reconstruct\n", i); set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantread, &dev->flags); + if (!test_and_set_bit( + STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; s->locked++; } else { set_bit(STRIPE_DELAYED, &sh->state); @@ -2545,6 +2551,9 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh, set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantwrite, &dev->flags); + if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; + clear_bit(STRIPE_DEGRADED, &sh->state); s->locked++; set_bit(STRIPE_INSYNC, &sh->state); @@ -2930,12 +2939,16 @@ static void handle_stripe5(struct stripe_head *sh) dev = &sh->dev[s.failed_num]; if (!test_bit(R5_ReWrite, &dev->flags)) { set_bit(R5_Wantwrite, &dev->flags); + if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; set_bit(R5_ReWrite, &dev->flags); set_bit(R5_LOCKED, &dev->flags); s.locked++; } else { /* let's read it back */ set_bit(R5_Wantread, &dev->flags); + if (!test_and_set_bit(STRIPE_OP_IO, &sh->ops.pending)) + sh->ops.count++; set_bit(R5_LOCKED, &dev->flags); s.locked++; } @@ -2988,64 +3001,6 @@ static void handle_stripe5(struct stripe_head *sh) return_io(return_bi); - for (i=disks; i-- ;) { - int rw; - struct bio *bi; - mdk_rdev_t *rdev; - if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) - rw = WRITE; - else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags)) - rw = READ; - else - continue; - - bi = &sh->dev[i].req; - - bi->bi_rw = rw; - if (rw == WRITE) - bi->bi_end_io = raid5_end_write_request; - else - bi->bi_end_io = raid5_end_read_request; - - rcu_read_lock(); - rdev = rcu_dereference(conf->disks[i].rdev); - if (rdev && test_bit(Faulty, &rdev->flags)) - rdev = NULL; - if (rdev) - atomic_inc(&rdev->nr_pending); - rcu_read_unlock(); - - if (rdev) { - if (s.syncing || s.expanding || s.expanded) - md_sync_acct(rdev->bdev, STRIPE_SECTORS); - - bi->bi_bdev = rdev->bdev; - pr_debug("for %llu schedule op %ld on disc %d\n", - (unsigned long long)sh->sector, bi->bi_rw, i); - atomic_inc(&sh->count); - bi->bi_sector = sh->sector + rdev->data_offset; - bi->bi_flags = 1 << BIO_UPTODATE; - bi->bi_vcnt = 1; - bi->bi_max_vecs = 1; - bi->bi_idx = 0; - bi->bi_io_vec = &sh->dev[i].vec; - bi->bi_io_vec[0].bv_len = STRIPE_SIZE; - bi->bi_io_vec[0].bv_offset = 0; - bi->bi_size = STRIPE_SIZE; - bi->bi_next = NULL; - if (rw == WRITE && - test_bit(R5_ReWrite, &sh->dev[i].flags)) - atomic_add(STRIPE_SECTORS, &rdev->corrected_errors); - generic_make_request(bi); - } else { - if (rw == WRITE) - set_bit(STRIPE_DEGRADED, &sh->state); - pr_debug("skip op %ld on disc %d for sector %llu\n", - bi->bi_rw, i, (unsigned long long)sh->sector); - clear_bit(R5_LOCKED, &sh->dev[i].flags); - set_bit(STRIPE_HANDLE, &sh->state); - } - } } static void handle_stripe6(struct stripe_head *sh, struct page *tmp_page) -- cgit v1.1 From f6dff381af01006ffae3c23cd2e07e30584de0ec Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: md: remove raid5 compute_block and compute_parity5 replaced by raid5_run_ops Signed-off-by: Dan Williams Acked-By: NeilBrown --- drivers/md/raid5.c | 124 ----------------------------------------------------- 1 file changed, 124 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 7e1cc07..0b66afe 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -1516,130 +1516,6 @@ static void copy_data(int frombio, struct bio *bio, } \ } while(0) - -static void compute_block(struct stripe_head *sh, int dd_idx) -{ - int i, count, disks = sh->disks; - void *ptr[MAX_XOR_BLOCKS], *dest, *p; - - pr_debug("compute_block, stripe %llu, idx %d\n", - (unsigned long long)sh->sector, dd_idx); - - dest = page_address(sh->dev[dd_idx].page); - memset(dest, 0, STRIPE_SIZE); - count = 0; - for (i = disks ; i--; ) { - if (i == dd_idx) - continue; - p = page_address(sh->dev[i].page); - if (test_bit(R5_UPTODATE, &sh->dev[i].flags)) - ptr[count++] = p; - else - printk(KERN_ERR "compute_block() %d, stripe %llu, %d" - " not present\n", dd_idx, - (unsigned long long)sh->sector, i); - - check_xor(); - } - if (count) - xor_blocks(count, STRIPE_SIZE, dest, ptr); - set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); -} - -static void compute_parity5(struct stripe_head *sh, int method) -{ - raid5_conf_t *conf = sh->raid_conf; - int i, pd_idx = sh->pd_idx, disks = sh->disks, count; - void *ptr[MAX_XOR_BLOCKS], *dest; - struct bio *chosen; - - pr_debug("compute_parity5, stripe %llu, method %d\n", - (unsigned long long)sh->sector, method); - - count = 0; - dest = page_address(sh->dev[pd_idx].page); - switch(method) { - case READ_MODIFY_WRITE: - BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags)); - for (i=disks ; i-- ;) { - if (i==pd_idx) - continue; - if (sh->dev[i].towrite && - test_bit(R5_UPTODATE, &sh->dev[i].flags)) { - ptr[count++] = page_address(sh->dev[i].page); - chosen = sh->dev[i].towrite; - sh->dev[i].towrite = NULL; - - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - - BUG_ON(sh->dev[i].written); - sh->dev[i].written = chosen; - check_xor(); - } - } - break; - case RECONSTRUCT_WRITE: - memset(dest, 0, STRIPE_SIZE); - for (i= disks; i-- ;) - if (i!=pd_idx && sh->dev[i].towrite) { - chosen = sh->dev[i].towrite; - sh->dev[i].towrite = NULL; - - if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags)) - wake_up(&conf->wait_for_overlap); - - BUG_ON(sh->dev[i].written); - sh->dev[i].written = chosen; - } - break; - case CHECK_PARITY: - break; - } - if (count) { - xor_blocks(count, STRIPE_SIZE, dest, ptr); - count = 0; - } - - for (i = disks; i--;) - if (sh->dev[i].written) { - sector_t sector = sh->dev[i].sector; - struct bio *wbi = sh->dev[i].written; - while (wbi && wbi->bi_sector < sector + STRIPE_SECTORS) { - copy_data(1, wbi, sh->dev[i].page, sector); - wbi = r5_next_bio(wbi, sector); - } - - set_bit(R5_LOCKED, &sh->dev[i].flags); - set_bit(R5_UPTODATE, &sh->dev[i].flags); - } - - switch(method) { - case RECONSTRUCT_WRITE: - case CHECK_PARITY: - for (i=disks; i--;) - if (i != pd_idx) { - ptr[count++] = page_address(sh->dev[i].page); - check_xor(); - } - break; - case READ_MODIFY_WRITE: - for (i = disks; i--;) - if (sh->dev[i].written) { - ptr[count++] = page_address(sh->dev[i].page); - check_xor(); - } - } - if (count) - xor_blocks(count, STRIPE_SIZE, dest, ptr); - - if (method != CHECK_PARITY) { - set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); - set_bit(R5_LOCKED, &sh->dev[pd_idx].flags); - } else - clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags); -} - static void compute_parity6(struct stripe_head *sh, int method) { raid6_conf_t *conf = sh->raid_conf; -- cgit v1.1 From c211092313b90f898dec61f35207fc282d1eadc3 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:26 -0700 Subject: dmaengine: driver for the iop32x, iop33x, and iop13xx raid engines The Intel(R) IOP series of i/o processors integrate an Xscale core with raid acceleration engines. The capabilities per platform are: iop219: (2) copy engines iop321: (2) copy engines (1) xor and block fill engine iop33x: (2) copy and crc32c engines (1) xor, xor zero sum, pq, pq zero sum, and block fill engine iop34x (iop13xx): (2) copy, crc32c, xor, xor zero sum, and block fill engines (1) copy, crc32c, xor, xor zero sum, pq, pq zero sum, and block fill engine The driver supports the features of the async_tx api: * asynchronous notification of operation completion * implicit (interupt triggered) handling of inter-channel transaction dependencies The driver adapts to the platform it is running by two methods. 1/ #include which defines the hardware specific iop_chan_* and iop_desc_* routines as a series of static inline functions 2/ The private platform data attached to the platform_device defines the capabilities of the channels 20070626: Callbacks are run in a tasklet. Given the recent discussion on LKML about killing tasklets in favor of workqueues I did a quick conversion of the driver. Raid5 resync performance dropped from 50MB/s to 30MB/s, so the tasklet implementation remains until a generic softirq interface is available. Changelog: * fixed a slot allocation bug in do_iop13xx_adma_xor that caused too few slots to be requested eventually leading to data corruption * enabled the slot allocation routine to attempt to free slots before returning -ENOMEM * switched the cleanup routine to solely use the software chain and the status register to determine if a descriptor is complete. This is necessary to support other IOP engines that do not have status writeback capability * make the driver iop generic * modified the allocation routines to understand allocating a group of slots for a single operation * added a null xor initialization operation for the xor only channel on iop3xx * support xor operations on buffers larger than the hardware maximum * split the do_* routines into separate prep, src/dest set, submit stages * added async_tx support (dependent operations initiation at cleanup time) * simplified group handling * added interrupt support (callbacks via tasklets) * brought the pending depth inline with ioat (i.e. 4 descriptors) * drop dma mapping methods, suggested by Chris Leech * don't use inline in C files, Adrian Bunk * remove static tasklet declarations * make iop_adma_alloc_slots easier to read and remove chances for a corrupted descriptor chain * fix locking bug in iop_adma_alloc_chan_resources, Benjamin Herrenschmidt * convert capabilities over to dma_cap_mask_t * fixup sparse warnings * add descriptor flush before iop_chan_enable * checkpatch.pl fixes * gpl v2 only correction * move set_src, set_dest, submit to async_tx methods * move group_list and phys to async_tx Cc: Russell King Signed-off-by: Dan Williams --- drivers/dma/Kconfig | 9 + drivers/dma/Makefile | 1 + drivers/dma/iop-adma.c | 1467 +++++++++++++++++++++++++++++++++++ include/asm-arm/hardware/iop_adma.h | 118 +++ 4 files changed, 1595 insertions(+) create mode 100644 drivers/dma/iop-adma.c create mode 100644 include/asm-arm/hardware/iop_adma.h diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index 492aa08..b31756d 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -31,4 +31,13 @@ config INTEL_IOATDMA default m ---help--- Enable support for the Intel(R) I/OAT DMA engine. + +config INTEL_IOP_ADMA + tristate "Intel IOP ADMA support" + depends on DMA_ENGINE && (ARCH_IOP32X || ARCH_IOP33X || ARCH_IOP13XX) + select ASYNC_CORE + default m + ---help--- + Enable support for the Intel(R) IOP Series RAID engines. + endmenu diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile index bdcfdbd..b3839b6 100644 --- a/drivers/dma/Makefile +++ b/drivers/dma/Makefile @@ -1,3 +1,4 @@ obj-$(CONFIG_DMA_ENGINE) += dmaengine.o obj-$(CONFIG_NET_DMA) += iovlock.o obj-$(CONFIG_INTEL_IOATDMA) += ioatdma.o +obj-$(CONFIG_INTEL_IOP_ADMA) += iop-adma.o diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c new file mode 100644 index 0000000..5a1d426 --- /dev/null +++ b/drivers/dma/iop-adma.c @@ -0,0 +1,1467 @@ +/* + * offload engine driver for the Intel Xscale series of i/o processors + * Copyright © 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ + +/* + * This driver supports the asynchrounous DMA copy and RAID engines available + * on the Intel Xscale(R) family of I/O Processors (IOP 32x, 33x, 134x) + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define to_iop_adma_chan(chan) container_of(chan, struct iop_adma_chan, common) +#define to_iop_adma_device(dev) \ + container_of(dev, struct iop_adma_device, common) +#define tx_to_iop_adma_slot(tx) \ + container_of(tx, struct iop_adma_desc_slot, async_tx) + +/** + * iop_adma_free_slots - flags descriptor slots for reuse + * @slot: Slot to free + * Caller must hold &iop_chan->lock while calling this function + */ +static void iop_adma_free_slots(struct iop_adma_desc_slot *slot) +{ + int stride = slot->slots_per_op; + + while (stride--) { + slot->slots_per_op = 0; + slot = list_entry(slot->slot_node.next, + struct iop_adma_desc_slot, + slot_node); + } +} + +static dma_cookie_t +iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *iop_chan, dma_cookie_t cookie) +{ + BUG_ON(desc->async_tx.cookie < 0); + spin_lock_bh(&desc->async_tx.lock); + if (desc->async_tx.cookie > 0) { + cookie = desc->async_tx.cookie; + desc->async_tx.cookie = 0; + + /* call the callback (must not sleep or submit new + * operations to this channel) + */ + if (desc->async_tx.callback) + desc->async_tx.callback( + desc->async_tx.callback_param); + + /* unmap dma addresses + * (unmap_single vs unmap_page?) + */ + if (desc->group_head && desc->unmap_len) { + struct iop_adma_desc_slot *unmap = desc->group_head; + struct device *dev = + &iop_chan->device->pdev->dev; + u32 len = unmap->unmap_len; + u32 src_cnt = unmap->unmap_src_cnt; + dma_addr_t addr = iop_desc_get_dest_addr(unmap, + iop_chan); + + dma_unmap_page(dev, addr, len, DMA_FROM_DEVICE); + while (src_cnt--) { + addr = iop_desc_get_src_addr(unmap, + iop_chan, + src_cnt); + dma_unmap_page(dev, addr, len, + DMA_TO_DEVICE); + } + desc->group_head = NULL; + } + } + + /* run dependent operations */ + async_tx_run_dependencies(&desc->async_tx); + spin_unlock_bh(&desc->async_tx.lock); + + return cookie; +} + +static int +iop_adma_clean_slot(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *iop_chan) +{ + /* the client is allowed to attach dependent operations + * until 'ack' is set + */ + if (!desc->async_tx.ack) + return 0; + + /* leave the last descriptor in the chain + * so we can append to it + */ + if (desc->chain_node.next == &iop_chan->chain) + return 1; + + dev_dbg(iop_chan->device->common.dev, + "\tfree slot: %d slots_per_op: %d\n", + desc->idx, desc->slots_per_op); + + list_del(&desc->chain_node); + iop_adma_free_slots(desc); + + return 0; +} + +static void __iop_adma_slot_cleanup(struct iop_adma_chan *iop_chan) +{ + struct iop_adma_desc_slot *iter, *_iter, *grp_start = NULL; + dma_cookie_t cookie = 0; + u32 current_desc = iop_chan_get_current_descriptor(iop_chan); + int busy = iop_chan_is_busy(iop_chan); + int seen_current = 0, slot_cnt = 0, slots_per_op = 0; + + dev_dbg(iop_chan->device->common.dev, "%s\n", __FUNCTION__); + /* free completed slots from the chain starting with + * the oldest descriptor + */ + list_for_each_entry_safe(iter, _iter, &iop_chan->chain, + chain_node) { + pr_debug("\tcookie: %d slot: %d busy: %d " + "this_desc: %#x next_desc: %#x ack: %d\n", + iter->async_tx.cookie, iter->idx, busy, + iter->async_tx.phys, iop_desc_get_next_desc(iter), + iter->async_tx.ack); + prefetch(_iter); + prefetch(&_iter->async_tx); + + /* do not advance past the current descriptor loaded into the + * hardware channel, subsequent descriptors are either in + * process or have not been submitted + */ + if (seen_current) + break; + + /* stop the search if we reach the current descriptor and the + * channel is busy, or if it appears that the current descriptor + * needs to be re-read (i.e. has been appended to) + */ + if (iter->async_tx.phys == current_desc) { + BUG_ON(seen_current++); + if (busy || iop_desc_get_next_desc(iter)) + break; + } + + /* detect the start of a group transaction */ + if (!slot_cnt && !slots_per_op) { + slot_cnt = iter->slot_cnt; + slots_per_op = iter->slots_per_op; + if (slot_cnt <= slots_per_op) { + slot_cnt = 0; + slots_per_op = 0; + } + } + + if (slot_cnt) { + pr_debug("\tgroup++\n"); + if (!grp_start) + grp_start = iter; + slot_cnt -= slots_per_op; + } + + /* all the members of a group are complete */ + if (slots_per_op != 0 && slot_cnt == 0) { + struct iop_adma_desc_slot *grp_iter, *_grp_iter; + int end_of_chain = 0; + pr_debug("\tgroup end\n"); + + /* collect the total results */ + if (grp_start->xor_check_result) { + u32 zero_sum_result = 0; + slot_cnt = grp_start->slot_cnt; + grp_iter = grp_start; + + list_for_each_entry_from(grp_iter, + &iop_chan->chain, chain_node) { + zero_sum_result |= + iop_desc_get_zero_result(grp_iter); + pr_debug("\titer%d result: %d\n", + grp_iter->idx, zero_sum_result); + slot_cnt -= slots_per_op; + if (slot_cnt == 0) + break; + } + pr_debug("\tgrp_start->xor_check_result: %p\n", + grp_start->xor_check_result); + *grp_start->xor_check_result = zero_sum_result; + } + + /* clean up the group */ + slot_cnt = grp_start->slot_cnt; + grp_iter = grp_start; + list_for_each_entry_safe_from(grp_iter, _grp_iter, + &iop_chan->chain, chain_node) { + cookie = iop_adma_run_tx_complete_actions( + grp_iter, iop_chan, cookie); + + slot_cnt -= slots_per_op; + end_of_chain = iop_adma_clean_slot(grp_iter, + iop_chan); + + if (slot_cnt == 0 || end_of_chain) + break; + } + + /* the group should be complete at this point */ + BUG_ON(slot_cnt); + + slots_per_op = 0; + grp_start = NULL; + if (end_of_chain) + break; + else + continue; + } else if (slots_per_op) /* wait for group completion */ + continue; + + /* write back zero sum results (single descriptor case) */ + if (iter->xor_check_result && iter->async_tx.cookie) + *iter->xor_check_result = + iop_desc_get_zero_result(iter); + + cookie = iop_adma_run_tx_complete_actions( + iter, iop_chan, cookie); + + if (iop_adma_clean_slot(iter, iop_chan)) + break; + } + + BUG_ON(!seen_current); + + iop_chan_idle(busy, iop_chan); + + if (cookie > 0) { + iop_chan->completed_cookie = cookie; + pr_debug("\tcompleted cookie %d\n", cookie); + } +} + +static void +iop_adma_slot_cleanup(struct iop_adma_chan *iop_chan) +{ + spin_lock_bh(&iop_chan->lock); + __iop_adma_slot_cleanup(iop_chan); + spin_unlock_bh(&iop_chan->lock); +} + +static void iop_adma_tasklet(unsigned long data) +{ + struct iop_adma_chan *chan = (struct iop_adma_chan *) data; + __iop_adma_slot_cleanup(chan); +} + +static struct iop_adma_desc_slot * +iop_adma_alloc_slots(struct iop_adma_chan *iop_chan, int num_slots, + int slots_per_op) +{ + struct iop_adma_desc_slot *iter, *_iter, *alloc_start = NULL; + struct list_head chain = LIST_HEAD_INIT(chain); + int slots_found, retry = 0; + + /* start search from the last allocated descrtiptor + * if a contiguous allocation can not be found start searching + * from the beginning of the list + */ +retry: + slots_found = 0; + if (retry == 0) + iter = iop_chan->last_used; + else + iter = list_entry(&iop_chan->all_slots, + struct iop_adma_desc_slot, + slot_node); + + list_for_each_entry_safe_continue( + iter, _iter, &iop_chan->all_slots, slot_node) { + prefetch(_iter); + prefetch(&_iter->async_tx); + if (iter->slots_per_op) { + /* give up after finding the first busy slot + * on the second pass through the list + */ + if (retry) + break; + + slots_found = 0; + continue; + } + + /* start the allocation if the slot is correctly aligned */ + if (!slots_found++) { + if (iop_desc_is_aligned(iter, slots_per_op)) + alloc_start = iter; + else { + slots_found = 0; + continue; + } + } + + if (slots_found == num_slots) { + struct iop_adma_desc_slot *alloc_tail = NULL; + struct iop_adma_desc_slot *last_used = NULL; + iter = alloc_start; + while (num_slots) { + int i; + dev_dbg(iop_chan->device->common.dev, + "allocated slot: %d " + "(desc %p phys: %#x) slots_per_op %d\n", + iter->idx, iter->hw_desc, + iter->async_tx.phys, slots_per_op); + + /* pre-ack all but the last descriptor */ + if (num_slots != slots_per_op) + iter->async_tx.ack = 1; + else + iter->async_tx.ack = 0; + + list_add_tail(&iter->chain_node, &chain); + alloc_tail = iter; + iter->async_tx.cookie = 0; + iter->slot_cnt = num_slots; + iter->xor_check_result = NULL; + for (i = 0; i < slots_per_op; i++) { + iter->slots_per_op = slots_per_op - i; + last_used = iter; + iter = list_entry(iter->slot_node.next, + struct iop_adma_desc_slot, + slot_node); + } + num_slots -= slots_per_op; + } + alloc_tail->group_head = alloc_start; + alloc_tail->async_tx.cookie = -EBUSY; + list_splice(&chain, &alloc_tail->async_tx.tx_list); + iop_chan->last_used = last_used; + iop_desc_clear_next_desc(alloc_start); + iop_desc_clear_next_desc(alloc_tail); + return alloc_tail; + } + } + if (!retry++) + goto retry; + + /* try to free some slots if the allocation fails */ + tasklet_schedule(&iop_chan->irq_tasklet); + + return NULL; +} + +static dma_cookie_t +iop_desc_assign_cookie(struct iop_adma_chan *iop_chan, + struct iop_adma_desc_slot *desc) +{ + dma_cookie_t cookie = iop_chan->common.cookie; + cookie++; + if (cookie < 0) + cookie = 1; + iop_chan->common.cookie = desc->async_tx.cookie = cookie; + return cookie; +} + +static void iop_adma_check_threshold(struct iop_adma_chan *iop_chan) +{ + dev_dbg(iop_chan->device->common.dev, "pending: %d\n", + iop_chan->pending); + + if (iop_chan->pending >= IOP_ADMA_THRESHOLD) { + iop_chan->pending = 0; + iop_chan_append(iop_chan); + } +} + +static dma_cookie_t +iop_adma_tx_submit(struct dma_async_tx_descriptor *tx) +{ + struct iop_adma_desc_slot *sw_desc = tx_to_iop_adma_slot(tx); + struct iop_adma_chan *iop_chan = to_iop_adma_chan(tx->chan); + struct iop_adma_desc_slot *grp_start, *old_chain_tail; + int slot_cnt; + int slots_per_op; + dma_cookie_t cookie; + + grp_start = sw_desc->group_head; + slot_cnt = grp_start->slot_cnt; + slots_per_op = grp_start->slots_per_op; + + spin_lock_bh(&iop_chan->lock); + cookie = iop_desc_assign_cookie(iop_chan, sw_desc); + + old_chain_tail = list_entry(iop_chan->chain.prev, + struct iop_adma_desc_slot, chain_node); + list_splice_init(&sw_desc->async_tx.tx_list, + &old_chain_tail->chain_node); + + /* fix up the hardware chain */ + iop_desc_set_next_desc(old_chain_tail, grp_start->async_tx.phys); + + /* 1/ don't add pre-chained descriptors + * 2/ dummy read to flush next_desc write + */ + BUG_ON(iop_desc_get_next_desc(sw_desc)); + + /* increment the pending count by the number of slots + * memcpy operations have a 1:1 (slot:operation) relation + * other operations are heavier and will pop the threshold + * more often. + */ + iop_chan->pending += slot_cnt; + iop_adma_check_threshold(iop_chan); + spin_unlock_bh(&iop_chan->lock); + + dev_dbg(iop_chan->device->common.dev, "%s cookie: %d slot: %d\n", + __FUNCTION__, sw_desc->async_tx.cookie, sw_desc->idx); + + return cookie; +} + +static void +iop_adma_set_dest(dma_addr_t addr, struct dma_async_tx_descriptor *tx, + int index) +{ + struct iop_adma_desc_slot *sw_desc = tx_to_iop_adma_slot(tx); + struct iop_adma_chan *iop_chan = to_iop_adma_chan(tx->chan); + + /* to do: support transfers lengths > IOP_ADMA_MAX_BYTE_COUNT */ + iop_desc_set_dest_addr(sw_desc->group_head, iop_chan, addr); +} + +static void iop_chan_start_null_memcpy(struct iop_adma_chan *iop_chan); +static void iop_chan_start_null_xor(struct iop_adma_chan *iop_chan); + +/* returns the number of allocated descriptors */ +static int iop_adma_alloc_chan_resources(struct dma_chan *chan) +{ + char *hw_desc; + int idx; + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *slot = NULL; + int init = iop_chan->slots_allocated ? 0 : 1; + struct iop_adma_platform_data *plat_data = + iop_chan->device->pdev->dev.platform_data; + int num_descs_in_pool = plat_data->pool_size/IOP_ADMA_SLOT_SIZE; + + /* Allocate descriptor slots */ + do { + idx = iop_chan->slots_allocated; + if (idx == num_descs_in_pool) + break; + + slot = kzalloc(sizeof(*slot), GFP_KERNEL); + if (!slot) { + printk(KERN_INFO "IOP ADMA Channel only initialized" + " %d descriptor slots", idx); + break; + } + hw_desc = (char *) iop_chan->device->dma_desc_pool_virt; + slot->hw_desc = (void *) &hw_desc[idx * IOP_ADMA_SLOT_SIZE]; + + dma_async_tx_descriptor_init(&slot->async_tx, chan); + slot->async_tx.tx_submit = iop_adma_tx_submit; + slot->async_tx.tx_set_dest = iop_adma_set_dest; + INIT_LIST_HEAD(&slot->chain_node); + INIT_LIST_HEAD(&slot->slot_node); + INIT_LIST_HEAD(&slot->async_tx.tx_list); + hw_desc = (char *) iop_chan->device->dma_desc_pool; + slot->async_tx.phys = + (dma_addr_t) &hw_desc[idx * IOP_ADMA_SLOT_SIZE]; + slot->idx = idx; + + spin_lock_bh(&iop_chan->lock); + iop_chan->slots_allocated++; + list_add_tail(&slot->slot_node, &iop_chan->all_slots); + spin_unlock_bh(&iop_chan->lock); + } while (iop_chan->slots_allocated < num_descs_in_pool); + + if (idx && !iop_chan->last_used) + iop_chan->last_used = list_entry(iop_chan->all_slots.next, + struct iop_adma_desc_slot, + slot_node); + + dev_dbg(iop_chan->device->common.dev, + "allocated %d descriptor slots last_used: %p\n", + iop_chan->slots_allocated, iop_chan->last_used); + + /* initialize the channel and the chain with a null operation */ + if (init) { + if (dma_has_cap(DMA_MEMCPY, + iop_chan->device->common.cap_mask)) + iop_chan_start_null_memcpy(iop_chan); + else if (dma_has_cap(DMA_XOR, + iop_chan->device->common.cap_mask)) + iop_chan_start_null_xor(iop_chan); + else + BUG(); + } + + return (idx > 0) ? idx : -ENOMEM; +} + +static struct dma_async_tx_descriptor * +iop_adma_prep_dma_interrupt(struct dma_chan *chan) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *sw_desc, *grp_start; + int slot_cnt, slots_per_op; + + dev_dbg(iop_chan->device->common.dev, "%s\n", __FUNCTION__); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_interrupt_slot_count(&slots_per_op, iop_chan); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + iop_desc_init_interrupt(grp_start, iop_chan); + grp_start->unmap_len = 0; + } + spin_unlock_bh(&iop_chan->lock); + + return sw_desc ? &sw_desc->async_tx : NULL; +} + +static void +iop_adma_memcpy_set_src(dma_addr_t addr, struct dma_async_tx_descriptor *tx, + int index) +{ + struct iop_adma_desc_slot *sw_desc = tx_to_iop_adma_slot(tx); + struct iop_adma_desc_slot *grp_start = sw_desc->group_head; + + iop_desc_set_memcpy_src_addr(grp_start, addr); +} + +static struct dma_async_tx_descriptor * +iop_adma_prep_dma_memcpy(struct dma_chan *chan, size_t len, int int_en) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *sw_desc, *grp_start; + int slot_cnt, slots_per_op; + + if (unlikely(!len)) + return NULL; + BUG_ON(unlikely(len > IOP_ADMA_MAX_BYTE_COUNT)); + + dev_dbg(iop_chan->device->common.dev, "%s len: %u\n", + __FUNCTION__, len); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_memcpy_slot_count(len, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + iop_desc_init_memcpy(grp_start, int_en); + iop_desc_set_byte_count(grp_start, iop_chan, len); + sw_desc->unmap_src_cnt = 1; + sw_desc->unmap_len = len; + sw_desc->async_tx.tx_set_src = iop_adma_memcpy_set_src; + } + spin_unlock_bh(&iop_chan->lock); + + return sw_desc ? &sw_desc->async_tx : NULL; +} + +static struct dma_async_tx_descriptor * +iop_adma_prep_dma_memset(struct dma_chan *chan, int value, size_t len, + int int_en) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *sw_desc, *grp_start; + int slot_cnt, slots_per_op; + + if (unlikely(!len)) + return NULL; + BUG_ON(unlikely(len > IOP_ADMA_MAX_BYTE_COUNT)); + + dev_dbg(iop_chan->device->common.dev, "%s len: %u\n", + __FUNCTION__, len); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_memset_slot_count(len, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + iop_desc_init_memset(grp_start, int_en); + iop_desc_set_byte_count(grp_start, iop_chan, len); + iop_desc_set_block_fill_val(grp_start, value); + sw_desc->unmap_src_cnt = 1; + sw_desc->unmap_len = len; + } + spin_unlock_bh(&iop_chan->lock); + + return sw_desc ? &sw_desc->async_tx : NULL; +} + +static void +iop_adma_xor_set_src(dma_addr_t addr, struct dma_async_tx_descriptor *tx, + int index) +{ + struct iop_adma_desc_slot *sw_desc = tx_to_iop_adma_slot(tx); + struct iop_adma_desc_slot *grp_start = sw_desc->group_head; + + iop_desc_set_xor_src_addr(grp_start, index, addr); +} + +static struct dma_async_tx_descriptor * +iop_adma_prep_dma_xor(struct dma_chan *chan, unsigned int src_cnt, size_t len, + int int_en) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *sw_desc, *grp_start; + int slot_cnt, slots_per_op; + + if (unlikely(!len)) + return NULL; + BUG_ON(unlikely(len > IOP_ADMA_XOR_MAX_BYTE_COUNT)); + + dev_dbg(iop_chan->device->common.dev, + "%s src_cnt: %d len: %u int_en: %d\n", + __FUNCTION__, src_cnt, len, int_en); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_xor_slot_count(len, src_cnt, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + iop_desc_init_xor(grp_start, src_cnt, int_en); + iop_desc_set_byte_count(grp_start, iop_chan, len); + sw_desc->unmap_src_cnt = src_cnt; + sw_desc->unmap_len = len; + sw_desc->async_tx.tx_set_src = iop_adma_xor_set_src; + } + spin_unlock_bh(&iop_chan->lock); + + return sw_desc ? &sw_desc->async_tx : NULL; +} + +static void +iop_adma_xor_zero_sum_set_src(dma_addr_t addr, + struct dma_async_tx_descriptor *tx, + int index) +{ + struct iop_adma_desc_slot *sw_desc = tx_to_iop_adma_slot(tx); + struct iop_adma_desc_slot *grp_start = sw_desc->group_head; + + iop_desc_set_zero_sum_src_addr(grp_start, index, addr); +} + +static struct dma_async_tx_descriptor * +iop_adma_prep_dma_zero_sum(struct dma_chan *chan, unsigned int src_cnt, + size_t len, u32 *result, int int_en) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *sw_desc, *grp_start; + int slot_cnt, slots_per_op; + + if (unlikely(!len)) + return NULL; + + dev_dbg(iop_chan->device->common.dev, "%s src_cnt: %d len: %u\n", + __FUNCTION__, src_cnt, len); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_zero_sum_slot_count(len, src_cnt, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + iop_desc_init_zero_sum(grp_start, src_cnt, int_en); + iop_desc_set_zero_sum_byte_count(grp_start, len); + grp_start->xor_check_result = result; + pr_debug("\t%s: grp_start->xor_check_result: %p\n", + __FUNCTION__, grp_start->xor_check_result); + sw_desc->unmap_src_cnt = src_cnt; + sw_desc->unmap_len = len; + sw_desc->async_tx.tx_set_src = iop_adma_xor_zero_sum_set_src; + } + spin_unlock_bh(&iop_chan->lock); + + return sw_desc ? &sw_desc->async_tx : NULL; +} + +static void iop_adma_dependency_added(struct dma_chan *chan) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + tasklet_schedule(&iop_chan->irq_tasklet); +} + +static void iop_adma_free_chan_resources(struct dma_chan *chan) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + struct iop_adma_desc_slot *iter, *_iter; + int in_use_descs = 0; + + iop_adma_slot_cleanup(iop_chan); + + spin_lock_bh(&iop_chan->lock); + list_for_each_entry_safe(iter, _iter, &iop_chan->chain, + chain_node) { + in_use_descs++; + list_del(&iter->chain_node); + } + list_for_each_entry_safe_reverse( + iter, _iter, &iop_chan->all_slots, slot_node) { + list_del(&iter->slot_node); + kfree(iter); + iop_chan->slots_allocated--; + } + iop_chan->last_used = NULL; + + dev_dbg(iop_chan->device->common.dev, "%s slots_allocated %d\n", + __FUNCTION__, iop_chan->slots_allocated); + spin_unlock_bh(&iop_chan->lock); + + /* one is ok since we left it on there on purpose */ + if (in_use_descs > 1) + printk(KERN_ERR "IOP: Freeing %d in use descriptors!\n", + in_use_descs - 1); +} + +/** + * iop_adma_is_complete - poll the status of an ADMA transaction + * @chan: ADMA channel handle + * @cookie: ADMA transaction identifier + */ +static enum dma_status iop_adma_is_complete(struct dma_chan *chan, + dma_cookie_t cookie, + dma_cookie_t *done, + dma_cookie_t *used) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + dma_cookie_t last_used; + dma_cookie_t last_complete; + enum dma_status ret; + + last_used = chan->cookie; + last_complete = iop_chan->completed_cookie; + + if (done) + *done = last_complete; + if (used) + *used = last_used; + + ret = dma_async_is_complete(cookie, last_complete, last_used); + if (ret == DMA_SUCCESS) + return ret; + + iop_adma_slot_cleanup(iop_chan); + + last_used = chan->cookie; + last_complete = iop_chan->completed_cookie; + + if (done) + *done = last_complete; + if (used) + *used = last_used; + + return dma_async_is_complete(cookie, last_complete, last_used); +} + +static irqreturn_t iop_adma_eot_handler(int irq, void *data) +{ + struct iop_adma_chan *chan = data; + + dev_dbg(chan->device->common.dev, "%s\n", __FUNCTION__); + + tasklet_schedule(&chan->irq_tasklet); + + iop_adma_device_clear_eot_status(chan); + + return IRQ_HANDLED; +} + +static irqreturn_t iop_adma_eoc_handler(int irq, void *data) +{ + struct iop_adma_chan *chan = data; + + dev_dbg(chan->device->common.dev, "%s\n", __FUNCTION__); + + tasklet_schedule(&chan->irq_tasklet); + + iop_adma_device_clear_eoc_status(chan); + + return IRQ_HANDLED; +} + +static irqreturn_t iop_adma_err_handler(int irq, void *data) +{ + struct iop_adma_chan *chan = data; + unsigned long status = iop_chan_get_status(chan); + + dev_printk(KERN_ERR, chan->device->common.dev, + "error ( %s%s%s%s%s%s%s)\n", + iop_is_err_int_parity(status, chan) ? "int_parity " : "", + iop_is_err_mcu_abort(status, chan) ? "mcu_abort " : "", + iop_is_err_int_tabort(status, chan) ? "int_tabort " : "", + iop_is_err_int_mabort(status, chan) ? "int_mabort " : "", + iop_is_err_pci_tabort(status, chan) ? "pci_tabort " : "", + iop_is_err_pci_mabort(status, chan) ? "pci_mabort " : "", + iop_is_err_split_tx(status, chan) ? "split_tx " : ""); + + iop_adma_device_clear_err_status(chan); + + BUG(); + + return IRQ_HANDLED; +} + +static void iop_adma_issue_pending(struct dma_chan *chan) +{ + struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan); + + if (iop_chan->pending) { + iop_chan->pending = 0; + iop_chan_append(iop_chan); + } +} + +/* + * Perform a transaction to verify the HW works. + */ +#define IOP_ADMA_TEST_SIZE 2000 + +static int __devinit iop_adma_memcpy_self_test(struct iop_adma_device *device) +{ + int i; + void *src, *dest; + dma_addr_t src_dma, dest_dma; + struct dma_chan *dma_chan; + dma_cookie_t cookie; + struct dma_async_tx_descriptor *tx; + int err = 0; + struct iop_adma_chan *iop_chan; + + dev_dbg(device->common.dev, "%s\n", __FUNCTION__); + + src = kzalloc(sizeof(u8) * IOP_ADMA_TEST_SIZE, GFP_KERNEL); + if (!src) + return -ENOMEM; + dest = kzalloc(sizeof(u8) * IOP_ADMA_TEST_SIZE, GFP_KERNEL); + if (!dest) { + kfree(src); + return -ENOMEM; + } + + /* Fill in src buffer */ + for (i = 0; i < IOP_ADMA_TEST_SIZE; i++) + ((u8 *) src)[i] = (u8)i; + + memset(dest, 0, IOP_ADMA_TEST_SIZE); + + /* Start copy, using first DMA channel */ + dma_chan = container_of(device->common.channels.next, + struct dma_chan, + device_node); + if (iop_adma_alloc_chan_resources(dma_chan) < 1) { + err = -ENODEV; + goto out; + } + + tx = iop_adma_prep_dma_memcpy(dma_chan, IOP_ADMA_TEST_SIZE, 1); + dest_dma = dma_map_single(dma_chan->device->dev, dest, + IOP_ADMA_TEST_SIZE, DMA_FROM_DEVICE); + iop_adma_set_dest(dest_dma, tx, 0); + src_dma = dma_map_single(dma_chan->device->dev, src, + IOP_ADMA_TEST_SIZE, DMA_TO_DEVICE); + iop_adma_memcpy_set_src(src_dma, tx, 0); + + cookie = iop_adma_tx_submit(tx); + iop_adma_issue_pending(dma_chan); + async_tx_ack(tx); + msleep(1); + + if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != + DMA_SUCCESS) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test copy timed out, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + iop_chan = to_iop_adma_chan(dma_chan); + dma_sync_single_for_cpu(&iop_chan->device->pdev->dev, dest_dma, + IOP_ADMA_TEST_SIZE, DMA_FROM_DEVICE); + if (memcmp(src, dest, IOP_ADMA_TEST_SIZE)) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test copy failed compare, disabling\n"); + err = -ENODEV; + goto free_resources; + } + +free_resources: + iop_adma_free_chan_resources(dma_chan); +out: + kfree(src); + kfree(dest); + return err; +} + +#define IOP_ADMA_NUM_SRC_TEST 4 /* must be <= 15 */ +static int __devinit +iop_adma_xor_zero_sum_self_test(struct iop_adma_device *device) +{ + int i, src_idx; + struct page *dest; + struct page *xor_srcs[IOP_ADMA_NUM_SRC_TEST]; + struct page *zero_sum_srcs[IOP_ADMA_NUM_SRC_TEST + 1]; + dma_addr_t dma_addr, dest_dma; + struct dma_async_tx_descriptor *tx; + struct dma_chan *dma_chan; + dma_cookie_t cookie; + u8 cmp_byte = 0; + u32 cmp_word; + u32 zero_sum_result; + int err = 0; + struct iop_adma_chan *iop_chan; + + dev_dbg(device->common.dev, "%s\n", __FUNCTION__); + + for (src_idx = 0; src_idx < IOP_ADMA_NUM_SRC_TEST; src_idx++) { + xor_srcs[src_idx] = alloc_page(GFP_KERNEL); + if (!xor_srcs[src_idx]) + while (src_idx--) { + __free_page(xor_srcs[src_idx]); + return -ENOMEM; + } + } + + dest = alloc_page(GFP_KERNEL); + if (!dest) + while (src_idx--) { + __free_page(xor_srcs[src_idx]); + return -ENOMEM; + } + + /* Fill in src buffers */ + for (src_idx = 0; src_idx < IOP_ADMA_NUM_SRC_TEST; src_idx++) { + u8 *ptr = page_address(xor_srcs[src_idx]); + for (i = 0; i < PAGE_SIZE; i++) + ptr[i] = (1 << src_idx); + } + + for (src_idx = 0; src_idx < IOP_ADMA_NUM_SRC_TEST; src_idx++) + cmp_byte ^= (u8) (1 << src_idx); + + cmp_word = (cmp_byte << 24) | (cmp_byte << 16) | + (cmp_byte << 8) | cmp_byte; + + memset(page_address(dest), 0, PAGE_SIZE); + + dma_chan = container_of(device->common.channels.next, + struct dma_chan, + device_node); + if (iop_adma_alloc_chan_resources(dma_chan) < 1) { + err = -ENODEV; + goto out; + } + + /* test xor */ + tx = iop_adma_prep_dma_xor(dma_chan, IOP_ADMA_NUM_SRC_TEST, + PAGE_SIZE, 1); + dest_dma = dma_map_page(dma_chan->device->dev, dest, 0, + PAGE_SIZE, DMA_FROM_DEVICE); + iop_adma_set_dest(dest_dma, tx, 0); + + for (i = 0; i < IOP_ADMA_NUM_SRC_TEST; i++) { + dma_addr = dma_map_page(dma_chan->device->dev, xor_srcs[i], 0, + PAGE_SIZE, DMA_TO_DEVICE); + iop_adma_xor_set_src(dma_addr, tx, i); + } + + cookie = iop_adma_tx_submit(tx); + iop_adma_issue_pending(dma_chan); + async_tx_ack(tx); + msleep(8); + + if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != + DMA_SUCCESS) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test xor timed out, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + iop_chan = to_iop_adma_chan(dma_chan); + dma_sync_single_for_cpu(&iop_chan->device->pdev->dev, dest_dma, + PAGE_SIZE, DMA_FROM_DEVICE); + for (i = 0; i < (PAGE_SIZE / sizeof(u32)); i++) { + u32 *ptr = page_address(dest); + if (ptr[i] != cmp_word) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test xor failed compare, disabling\n"); + err = -ENODEV; + goto free_resources; + } + } + dma_sync_single_for_device(&iop_chan->device->pdev->dev, dest_dma, + PAGE_SIZE, DMA_TO_DEVICE); + + /* skip zero sum if the capability is not present */ + if (!dma_has_cap(DMA_ZERO_SUM, dma_chan->device->cap_mask)) + goto free_resources; + + /* zero sum the sources with the destintation page */ + for (i = 0; i < IOP_ADMA_NUM_SRC_TEST; i++) + zero_sum_srcs[i] = xor_srcs[i]; + zero_sum_srcs[i] = dest; + + zero_sum_result = 1; + + tx = iop_adma_prep_dma_zero_sum(dma_chan, IOP_ADMA_NUM_SRC_TEST + 1, + PAGE_SIZE, &zero_sum_result, 1); + for (i = 0; i < IOP_ADMA_NUM_SRC_TEST + 1; i++) { + dma_addr = dma_map_page(dma_chan->device->dev, zero_sum_srcs[i], + 0, PAGE_SIZE, DMA_TO_DEVICE); + iop_adma_xor_zero_sum_set_src(dma_addr, tx, i); + } + + cookie = iop_adma_tx_submit(tx); + iop_adma_issue_pending(dma_chan); + async_tx_ack(tx); + msleep(8); + + if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test zero sum timed out, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + if (zero_sum_result != 0) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test zero sum failed compare, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + /* test memset */ + tx = iop_adma_prep_dma_memset(dma_chan, 0, PAGE_SIZE, 1); + dma_addr = dma_map_page(dma_chan->device->dev, dest, 0, + PAGE_SIZE, DMA_FROM_DEVICE); + iop_adma_set_dest(dma_addr, tx, 0); + + cookie = iop_adma_tx_submit(tx); + iop_adma_issue_pending(dma_chan); + async_tx_ack(tx); + msleep(8); + + if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test memset timed out, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + for (i = 0; i < PAGE_SIZE/sizeof(u32); i++) { + u32 *ptr = page_address(dest); + if (ptr[i]) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test memset failed compare, disabling\n"); + err = -ENODEV; + goto free_resources; + } + } + + /* test for non-zero parity sum */ + zero_sum_result = 0; + tx = iop_adma_prep_dma_zero_sum(dma_chan, IOP_ADMA_NUM_SRC_TEST + 1, + PAGE_SIZE, &zero_sum_result, 1); + for (i = 0; i < IOP_ADMA_NUM_SRC_TEST + 1; i++) { + dma_addr = dma_map_page(dma_chan->device->dev, zero_sum_srcs[i], + 0, PAGE_SIZE, DMA_TO_DEVICE); + iop_adma_xor_zero_sum_set_src(dma_addr, tx, i); + } + + cookie = iop_adma_tx_submit(tx); + iop_adma_issue_pending(dma_chan); + async_tx_ack(tx); + msleep(8); + + if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test non-zero sum timed out, disabling\n"); + err = -ENODEV; + goto free_resources; + } + + if (zero_sum_result != 1) { + dev_printk(KERN_ERR, dma_chan->device->dev, + "Self-test non-zero sum failed compare, disabling\n"); + err = -ENODEV; + goto free_resources; + } + +free_resources: + iop_adma_free_chan_resources(dma_chan); +out: + src_idx = IOP_ADMA_NUM_SRC_TEST; + while (src_idx--) + __free_page(xor_srcs[src_idx]); + __free_page(dest); + return err; +} + +static int __devexit iop_adma_remove(struct platform_device *dev) +{ + struct iop_adma_device *device = platform_get_drvdata(dev); + struct dma_chan *chan, *_chan; + struct iop_adma_chan *iop_chan; + int i; + struct iop_adma_platform_data *plat_data = dev->dev.platform_data; + + dma_async_device_unregister(&device->common); + + for (i = 0; i < 3; i++) { + unsigned int irq; + irq = platform_get_irq(dev, i); + free_irq(irq, device); + } + + dma_free_coherent(&dev->dev, plat_data->pool_size, + device->dma_desc_pool_virt, device->dma_desc_pool); + + do { + struct resource *res; + res = platform_get_resource(dev, IORESOURCE_MEM, 0); + release_mem_region(res->start, res->end - res->start); + } while (0); + + list_for_each_entry_safe(chan, _chan, &device->common.channels, + device_node) { + iop_chan = to_iop_adma_chan(chan); + list_del(&chan->device_node); + kfree(iop_chan); + } + kfree(device); + + return 0; +} + +static int __devinit iop_adma_probe(struct platform_device *pdev) +{ + struct resource *res; + int ret = 0, i; + struct iop_adma_device *adev; + struct iop_adma_chan *iop_chan; + struct dma_device *dma_dev; + struct iop_adma_platform_data *plat_data = pdev->dev.platform_data; + + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (!res) + return -ENODEV; + + if (!devm_request_mem_region(&pdev->dev, res->start, + res->end - res->start, pdev->name)) + return -EBUSY; + + adev = kzalloc(sizeof(*adev), GFP_KERNEL); + if (!adev) + return -ENOMEM; + dma_dev = &adev->common; + + /* allocate coherent memory for hardware descriptors + * note: writecombine gives slightly better performance, but + * requires that we explicitly flush the writes + */ + if ((adev->dma_desc_pool_virt = dma_alloc_writecombine(&pdev->dev, + plat_data->pool_size, + &adev->dma_desc_pool, + GFP_KERNEL)) == NULL) { + ret = -ENOMEM; + goto err_free_adev; + } + + dev_dbg(&pdev->dev, "%s: allocted descriptor pool virt %p phys %p\n", + __FUNCTION__, adev->dma_desc_pool_virt, + (void *) adev->dma_desc_pool); + + adev->id = plat_data->hw_id; + + /* discover transaction capabilites from the platform data */ + dma_dev->cap_mask = plat_data->cap_mask; + + adev->pdev = pdev; + platform_set_drvdata(pdev, adev); + + INIT_LIST_HEAD(&dma_dev->channels); + + /* set base routines */ + dma_dev->device_alloc_chan_resources = iop_adma_alloc_chan_resources; + dma_dev->device_free_chan_resources = iop_adma_free_chan_resources; + dma_dev->device_is_tx_complete = iop_adma_is_complete; + dma_dev->device_issue_pending = iop_adma_issue_pending; + dma_dev->device_dependency_added = iop_adma_dependency_added; + dma_dev->dev = &pdev->dev; + + /* set prep routines based on capability */ + if (dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask)) + dma_dev->device_prep_dma_memcpy = iop_adma_prep_dma_memcpy; + if (dma_has_cap(DMA_MEMSET, dma_dev->cap_mask)) + dma_dev->device_prep_dma_memset = iop_adma_prep_dma_memset; + if (dma_has_cap(DMA_XOR, dma_dev->cap_mask)) { + dma_dev->max_xor = iop_adma_get_max_xor(); + dma_dev->device_prep_dma_xor = iop_adma_prep_dma_xor; + } + if (dma_has_cap(DMA_ZERO_SUM, dma_dev->cap_mask)) + dma_dev->device_prep_dma_zero_sum = + iop_adma_prep_dma_zero_sum; + if (dma_has_cap(DMA_INTERRUPT, dma_dev->cap_mask)) + dma_dev->device_prep_dma_interrupt = + iop_adma_prep_dma_interrupt; + + iop_chan = kzalloc(sizeof(*iop_chan), GFP_KERNEL); + if (!iop_chan) { + ret = -ENOMEM; + goto err_free_dma; + } + iop_chan->device = adev; + + iop_chan->mmr_base = devm_ioremap(&pdev->dev, res->start, + res->end - res->start); + if (!iop_chan->mmr_base) { + ret = -ENOMEM; + goto err_free_iop_chan; + } + tasklet_init(&iop_chan->irq_tasklet, iop_adma_tasklet, (unsigned long) + iop_chan); + + /* clear errors before enabling interrupts */ + iop_adma_device_clear_err_status(iop_chan); + + for (i = 0; i < 3; i++) { + irq_handler_t handler[] = { iop_adma_eot_handler, + iop_adma_eoc_handler, + iop_adma_err_handler }; + int irq = platform_get_irq(pdev, i); + if (irq < 0) { + ret = -ENXIO; + goto err_free_iop_chan; + } else { + ret = devm_request_irq(&pdev->dev, irq, + handler[i], 0, pdev->name, iop_chan); + if (ret) + goto err_free_iop_chan; + } + } + + spin_lock_init(&iop_chan->lock); + init_timer(&iop_chan->cleanup_watchdog); + iop_chan->cleanup_watchdog.data = (unsigned long) iop_chan; + iop_chan->cleanup_watchdog.function = iop_adma_tasklet; + INIT_LIST_HEAD(&iop_chan->chain); + INIT_LIST_HEAD(&iop_chan->all_slots); + INIT_RCU_HEAD(&iop_chan->common.rcu); + iop_chan->common.device = dma_dev; + list_add_tail(&iop_chan->common.device_node, &dma_dev->channels); + + if (dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask)) { + ret = iop_adma_memcpy_self_test(adev); + dev_dbg(&pdev->dev, "memcpy self test returned %d\n", ret); + if (ret) + goto err_free_iop_chan; + } + + if (dma_has_cap(DMA_XOR, dma_dev->cap_mask) || + dma_has_cap(DMA_MEMSET, dma_dev->cap_mask)) { + ret = iop_adma_xor_zero_sum_self_test(adev); + dev_dbg(&pdev->dev, "xor self test returned %d\n", ret); + if (ret) + goto err_free_iop_chan; + } + + dev_printk(KERN_INFO, &pdev->dev, "Intel(R) IOP: " + "( %s%s%s%s%s%s%s%s%s%s)\n", + dma_has_cap(DMA_PQ_XOR, dma_dev->cap_mask) ? "pq_xor " : "", + dma_has_cap(DMA_PQ_UPDATE, dma_dev->cap_mask) ? "pq_update " : "", + dma_has_cap(DMA_PQ_ZERO_SUM, dma_dev->cap_mask) ? "pq_zero_sum " : "", + dma_has_cap(DMA_XOR, dma_dev->cap_mask) ? "xor " : "", + dma_has_cap(DMA_DUAL_XOR, dma_dev->cap_mask) ? "dual_xor " : "", + dma_has_cap(DMA_ZERO_SUM, dma_dev->cap_mask) ? "xor_zero_sum " : "", + dma_has_cap(DMA_MEMSET, dma_dev->cap_mask) ? "fill " : "", + dma_has_cap(DMA_MEMCPY_CRC32C, dma_dev->cap_mask) ? "cpy+crc " : "", + dma_has_cap(DMA_MEMCPY, dma_dev->cap_mask) ? "cpy " : "", + dma_has_cap(DMA_INTERRUPT, dma_dev->cap_mask) ? "intr " : ""); + + dma_async_device_register(dma_dev); + goto out; + + err_free_iop_chan: + kfree(iop_chan); + err_free_dma: + dma_free_coherent(&adev->pdev->dev, plat_data->pool_size, + adev->dma_desc_pool_virt, adev->dma_desc_pool); + err_free_adev: + kfree(adev); + out: + return ret; +} + +static void iop_chan_start_null_memcpy(struct iop_adma_chan *iop_chan) +{ + struct iop_adma_desc_slot *sw_desc, *grp_start; + dma_cookie_t cookie; + int slot_cnt, slots_per_op; + + dev_dbg(iop_chan->device->common.dev, "%s\n", __FUNCTION__); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_memcpy_slot_count(0, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + + list_splice_init(&sw_desc->async_tx.tx_list, &iop_chan->chain); + sw_desc->async_tx.ack = 1; + iop_desc_init_memcpy(grp_start, 0); + iop_desc_set_byte_count(grp_start, iop_chan, 0); + iop_desc_set_dest_addr(grp_start, iop_chan, 0); + iop_desc_set_memcpy_src_addr(grp_start, 0); + + cookie = iop_chan->common.cookie; + cookie++; + if (cookie <= 1) + cookie = 2; + + /* initialize the completed cookie to be less than + * the most recently used cookie + */ + iop_chan->completed_cookie = cookie - 1; + iop_chan->common.cookie = sw_desc->async_tx.cookie = cookie; + + /* channel should not be busy */ + BUG_ON(iop_chan_is_busy(iop_chan)); + + /* clear any prior error-status bits */ + iop_adma_device_clear_err_status(iop_chan); + + /* disable operation */ + iop_chan_disable(iop_chan); + + /* set the descriptor address */ + iop_chan_set_next_descriptor(iop_chan, sw_desc->async_tx.phys); + + /* 1/ don't add pre-chained descriptors + * 2/ dummy read to flush next_desc write + */ + BUG_ON(iop_desc_get_next_desc(sw_desc)); + + /* run the descriptor */ + iop_chan_enable(iop_chan); + } else + dev_printk(KERN_ERR, iop_chan->device->common.dev, + "failed to allocate null descriptor\n"); + spin_unlock_bh(&iop_chan->lock); +} + +static void iop_chan_start_null_xor(struct iop_adma_chan *iop_chan) +{ + struct iop_adma_desc_slot *sw_desc, *grp_start; + dma_cookie_t cookie; + int slot_cnt, slots_per_op; + + dev_dbg(iop_chan->device->common.dev, "%s\n", __FUNCTION__); + + spin_lock_bh(&iop_chan->lock); + slot_cnt = iop_chan_xor_slot_count(0, 2, &slots_per_op); + sw_desc = iop_adma_alloc_slots(iop_chan, slot_cnt, slots_per_op); + if (sw_desc) { + grp_start = sw_desc->group_head; + list_splice_init(&sw_desc->async_tx.tx_list, &iop_chan->chain); + sw_desc->async_tx.ack = 1; + iop_desc_init_null_xor(grp_start, 2, 0); + iop_desc_set_byte_count(grp_start, iop_chan, 0); + iop_desc_set_dest_addr(grp_start, iop_chan, 0); + iop_desc_set_xor_src_addr(grp_start, 0, 0); + iop_desc_set_xor_src_addr(grp_start, 1, 0); + + cookie = iop_chan->common.cookie; + cookie++; + if (cookie <= 1) + cookie = 2; + + /* initialize the completed cookie to be less than + * the most recently used cookie + */ + iop_chan->completed_cookie = cookie - 1; + iop_chan->common.cookie = sw_desc->async_tx.cookie = cookie; + + /* channel should not be busy */ + BUG_ON(iop_chan_is_busy(iop_chan)); + + /* clear any prior error-status bits */ + iop_adma_device_clear_err_status(iop_chan); + + /* disable operation */ + iop_chan_disable(iop_chan); + + /* set the descriptor address */ + iop_chan_set_next_descriptor(iop_chan, sw_desc->async_tx.phys); + + /* 1/ don't add pre-chained descriptors + * 2/ dummy read to flush next_desc write + */ + BUG_ON(iop_desc_get_next_desc(sw_desc)); + + /* run the descriptor */ + iop_chan_enable(iop_chan); + } else + dev_printk(KERN_ERR, iop_chan->device->common.dev, + "failed to allocate null descriptor\n"); + spin_unlock_bh(&iop_chan->lock); +} + +static struct platform_driver iop_adma_driver = { + .probe = iop_adma_probe, + .remove = iop_adma_remove, + .driver = { + .owner = THIS_MODULE, + .name = "iop-adma", + }, +}; + +static int __init iop_adma_init (void) +{ + /* it's currently unsafe to unload this module */ + /* if forced, worst case is that rmmod hangs */ + __unsafe(THIS_MODULE); + + return platform_driver_register(&iop_adma_driver); +} + +static void __exit iop_adma_exit (void) +{ + platform_driver_unregister(&iop_adma_driver); + return; +} + +module_init(iop_adma_init); +module_exit(iop_adma_exit); + +MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("IOP ADMA Engine Driver"); +MODULE_LICENSE("GPL"); diff --git a/include/asm-arm/hardware/iop_adma.h b/include/asm-arm/hardware/iop_adma.h new file mode 100644 index 0000000..ca8e71f --- /dev/null +++ b/include/asm-arm/hardware/iop_adma.h @@ -0,0 +1,118 @@ +/* + * Copyright © 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#ifndef IOP_ADMA_H +#define IOP_ADMA_H +#include +#include +#include + +#define IOP_ADMA_SLOT_SIZE 32 +#define IOP_ADMA_THRESHOLD 4 + +/** + * struct iop_adma_device - internal representation of an ADMA device + * @pdev: Platform device + * @id: HW ADMA Device selector + * @dma_desc_pool: base of DMA descriptor region (DMA address) + * @dma_desc_pool_virt: base of DMA descriptor region (CPU address) + * @common: embedded struct dma_device + */ +struct iop_adma_device { + struct platform_device *pdev; + int id; + dma_addr_t dma_desc_pool; + void *dma_desc_pool_virt; + struct dma_device common; +}; + +/** + * struct iop_adma_chan - internal representation of an ADMA device + * @pending: allows batching of hardware operations + * @completed_cookie: identifier for the most recently completed operation + * @lock: serializes enqueue/dequeue operations to the slot pool + * @mmr_base: memory mapped register base + * @chain: device chain view of the descriptors + * @device: parent device + * @common: common dmaengine channel object members + * @last_used: place holder for allocation to continue from where it left off + * @all_slots: complete domain of slots usable by the channel + * @cleanup_watchdog: workaround missed interrupts on iop3xx + * @slots_allocated: records the actual size of the descriptor slot pool + * @irq_tasklet: bottom half where iop_adma_slot_cleanup runs + */ +struct iop_adma_chan { + int pending; + dma_cookie_t completed_cookie; + spinlock_t lock; /* protects the descriptor slot pool */ + void __iomem *mmr_base; + struct list_head chain; + struct iop_adma_device *device; + struct dma_chan common; + struct iop_adma_desc_slot *last_used; + struct list_head all_slots; + struct timer_list cleanup_watchdog; + int slots_allocated; + struct tasklet_struct irq_tasklet; +}; + +/** + * struct iop_adma_desc_slot - IOP-ADMA software descriptor + * @slot_node: node on the iop_adma_chan.all_slots list + * @chain_node: node on the op_adma_chan.chain list + * @hw_desc: virtual address of the hardware descriptor chain + * @phys: hardware address of the hardware descriptor chain + * @group_head: first operation in a transaction + * @slot_cnt: total slots used in an transaction (group of operations) + * @slots_per_op: number of slots per operation + * @idx: pool index + * @unmap_src_cnt: number of xor sources + * @unmap_len: transaction bytecount + * @async_tx: support for the async_tx api + * @group_list: list of slots that make up a multi-descriptor transaction + * for example transfer lengths larger than the supported hw max + * @xor_check_result: result of zero sum + * @crc32_result: result crc calculation + */ +struct iop_adma_desc_slot { + struct list_head slot_node; + struct list_head chain_node; + void *hw_desc; + struct iop_adma_desc_slot *group_head; + u16 slot_cnt; + u16 slots_per_op; + u16 idx; + u16 unmap_src_cnt; + size_t unmap_len; + struct dma_async_tx_descriptor async_tx; + union { + u32 *xor_check_result; + u32 *crc32_result; + }; +}; + +struct iop_adma_platform_data { + int hw_id; + dma_cap_mask_t cap_mask; + size_t pool_size; +}; + +#define to_iop_sw_desc(addr_hw_desc) \ + container_of(addr_hw_desc, struct iop_adma_desc_slot, hw_desc) +#define iop_hw_desc_slot_idx(hw_desc, idx) \ + ( (void *) (((unsigned long) hw_desc) + ((idx) << 5)) ) +#endif -- cgit v1.1 From 39a8d7d13c113e4a98bfdfc45c7233188e4d715f Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: iop13xx: surface the iop13xx adma units to the iop-adma driver Adds the platform device definitions and the architecture specific support routines (i.e. register initialization and descriptor formats) for the iop-adma driver. Changelog: * added 'descriptor pool size' to the platform data * add base support for buffer sizes larger than 16MB (hw max) * build error fix from Kirill A. Shutemov * rebase for async_tx changes * add interrupt support * do not call platform register macros in driver code * remove unnecessary ARM assembly statement * checkpatch.pl fixes * gpl v2 only correction Cc: Russell King Signed-off-by: Dan Williams --- arch/arm/mach-iop13xx/setup.c | 217 ++++++++++++- include/asm-arm/arch-iop13xx/adma.h | 544 +++++++++++++++++++++++++++++++++ include/asm-arm/arch-iop13xx/iop13xx.h | 38 +-- 3 files changed, 774 insertions(+), 25 deletions(-) create mode 100644 include/asm-arm/arch-iop13xx/adma.h diff --git a/arch/arm/mach-iop13xx/setup.c b/arch/arm/mach-iop13xx/setup.c index bc48715..bfe0c87 100644 --- a/arch/arm/mach-iop13xx/setup.c +++ b/arch/arm/mach-iop13xx/setup.c @@ -25,6 +25,7 @@ #include #include #include +#include #define IOP13XX_UART_XTAL 33334000 #define IOP13XX_SETUP_DEBUG 0 @@ -236,19 +237,143 @@ static unsigned long iq8134x_probe_flash_size(void) } #endif +/* ADMA Channels */ +static struct resource iop13xx_adma_0_resources[] = { + [0] = { + .start = IOP13XX_ADMA_PHYS_BASE(0), + .end = IOP13XX_ADMA_UPPER_PA(0), + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_IOP13XX_ADMA0_EOT, + .end = IRQ_IOP13XX_ADMA0_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_IOP13XX_ADMA0_EOC, + .end = IRQ_IOP13XX_ADMA0_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_IOP13XX_ADMA0_ERR, + .end = IRQ_IOP13XX_ADMA0_ERR, + .flags = IORESOURCE_IRQ + } +}; + +static struct resource iop13xx_adma_1_resources[] = { + [0] = { + .start = IOP13XX_ADMA_PHYS_BASE(1), + .end = IOP13XX_ADMA_UPPER_PA(1), + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_IOP13XX_ADMA1_EOT, + .end = IRQ_IOP13XX_ADMA1_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_IOP13XX_ADMA1_EOC, + .end = IRQ_IOP13XX_ADMA1_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_IOP13XX_ADMA1_ERR, + .end = IRQ_IOP13XX_ADMA1_ERR, + .flags = IORESOURCE_IRQ + } +}; + +static struct resource iop13xx_adma_2_resources[] = { + [0] = { + .start = IOP13XX_ADMA_PHYS_BASE(2), + .end = IOP13XX_ADMA_UPPER_PA(2), + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_IOP13XX_ADMA2_EOT, + .end = IRQ_IOP13XX_ADMA2_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_IOP13XX_ADMA2_EOC, + .end = IRQ_IOP13XX_ADMA2_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_IOP13XX_ADMA2_ERR, + .end = IRQ_IOP13XX_ADMA2_ERR, + .flags = IORESOURCE_IRQ + } +}; + +static u64 iop13xx_adma_dmamask = DMA_64BIT_MASK; +static struct iop_adma_platform_data iop13xx_adma_0_data = { + .hw_id = 0, + .pool_size = PAGE_SIZE, +}; + +static struct iop_adma_platform_data iop13xx_adma_1_data = { + .hw_id = 1, + .pool_size = PAGE_SIZE, +}; + +static struct iop_adma_platform_data iop13xx_adma_2_data = { + .hw_id = 2, + .pool_size = PAGE_SIZE, +}; + +/* The ids are fixed up later in iop13xx_platform_init */ +static struct platform_device iop13xx_adma_0_channel = { + .name = "iop-adma", + .id = 0, + .num_resources = 4, + .resource = iop13xx_adma_0_resources, + .dev = { + .dma_mask = &iop13xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop13xx_adma_0_data, + }, +}; + +static struct platform_device iop13xx_adma_1_channel = { + .name = "iop-adma", + .id = 0, + .num_resources = 4, + .resource = iop13xx_adma_1_resources, + .dev = { + .dma_mask = &iop13xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop13xx_adma_1_data, + }, +}; + +static struct platform_device iop13xx_adma_2_channel = { + .name = "iop-adma", + .id = 0, + .num_resources = 4, + .resource = iop13xx_adma_2_resources, + .dev = { + .dma_mask = &iop13xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop13xx_adma_2_data, + }, +}; + void __init iop13xx_map_io(void) { /* Initialize the Static Page Table maps */ iotable_init(iop13xx_std_desc, ARRAY_SIZE(iop13xx_std_desc)); } -static int init_uart = 0; -static int init_i2c = 0; +static int init_uart; +static int init_i2c; +static int init_adma; void __init iop13xx_platform_init(void) { int i; - u32 uart_idx, i2c_idx, plat_idx; + u32 uart_idx, i2c_idx, adma_idx, plat_idx; struct platform_device *iop13xx_devices[IQ81340_MAX_PLAT_DEVICES]; /* set the bases so we can read the device id */ @@ -294,6 +419,12 @@ void __init iop13xx_platform_init(void) } } + if (init_adma == IOP13XX_INIT_ADMA_DEFAULT) { + init_adma |= IOP13XX_INIT_ADMA_0; + init_adma |= IOP13XX_INIT_ADMA_1; + init_adma |= IOP13XX_INIT_ADMA_2; + } + plat_idx = 0; uart_idx = 0; i2c_idx = 0; @@ -332,6 +463,56 @@ void __init iop13xx_platform_init(void) } } + /* initialize adma channel ids and capabilities */ + adma_idx = 0; + for (i = 0; i < IQ81340_NUM_ADMA; i++) { + struct iop_adma_platform_data *plat_data; + if ((init_adma & (1 << i)) && IOP13XX_SETUP_DEBUG) + printk(KERN_INFO + "Adding adma%d to platform device list\n", i); + switch (init_adma & (1 << i)) { + case IOP13XX_INIT_ADMA_0: + iop13xx_adma_0_channel.id = adma_idx++; + iop13xx_devices[plat_idx++] = &iop13xx_adma_0_channel; + plat_data = &iop13xx_adma_0_data; + dma_cap_set(DMA_MEMCPY, plat_data->cap_mask); + dma_cap_set(DMA_XOR, plat_data->cap_mask); + dma_cap_set(DMA_DUAL_XOR, plat_data->cap_mask); + dma_cap_set(DMA_ZERO_SUM, plat_data->cap_mask); + dma_cap_set(DMA_MEMSET, plat_data->cap_mask); + dma_cap_set(DMA_MEMCPY_CRC32C, plat_data->cap_mask); + dma_cap_set(DMA_INTERRUPT, plat_data->cap_mask); + break; + case IOP13XX_INIT_ADMA_1: + iop13xx_adma_1_channel.id = adma_idx++; + iop13xx_devices[plat_idx++] = &iop13xx_adma_1_channel; + plat_data = &iop13xx_adma_1_data; + dma_cap_set(DMA_MEMCPY, plat_data->cap_mask); + dma_cap_set(DMA_XOR, plat_data->cap_mask); + dma_cap_set(DMA_DUAL_XOR, plat_data->cap_mask); + dma_cap_set(DMA_ZERO_SUM, plat_data->cap_mask); + dma_cap_set(DMA_MEMSET, plat_data->cap_mask); + dma_cap_set(DMA_MEMCPY_CRC32C, plat_data->cap_mask); + dma_cap_set(DMA_INTERRUPT, plat_data->cap_mask); + break; + case IOP13XX_INIT_ADMA_2: + iop13xx_adma_2_channel.id = adma_idx++; + iop13xx_devices[plat_idx++] = &iop13xx_adma_2_channel; + plat_data = &iop13xx_adma_2_data; + dma_cap_set(DMA_MEMCPY, plat_data->cap_mask); + dma_cap_set(DMA_XOR, plat_data->cap_mask); + dma_cap_set(DMA_DUAL_XOR, plat_data->cap_mask); + dma_cap_set(DMA_ZERO_SUM, plat_data->cap_mask); + dma_cap_set(DMA_MEMSET, plat_data->cap_mask); + dma_cap_set(DMA_MEMCPY_CRC32C, plat_data->cap_mask); + dma_cap_set(DMA_INTERRUPT, plat_data->cap_mask); + dma_cap_set(DMA_PQ_XOR, plat_data->cap_mask); + dma_cap_set(DMA_PQ_UPDATE, plat_data->cap_mask); + dma_cap_set(DMA_PQ_ZERO_SUM, plat_data->cap_mask); + break; + } + } + #ifdef CONFIG_MTD_PHYSMAP iq8134x_flash_resource.end = iq8134x_flash_resource.start + iq8134x_probe_flash_size() - 1; @@ -399,5 +580,35 @@ static int __init iop13xx_init_i2c_setup(char *str) return 1; } +static int __init iop13xx_init_adma_setup(char *str) +{ + if (str) { + while (*str != '\0') { + switch (*str) { + case '0': + init_adma |= IOP13XX_INIT_ADMA_0; + break; + case '1': + init_adma |= IOP13XX_INIT_ADMA_1; + break; + case '2': + init_adma |= IOP13XX_INIT_ADMA_2; + break; + case ',': + case '=': + break; + default: + PRINTK("\"iop13xx_init_adma\" malformed" + " at character: \'%c\'", *str); + *(str + 1) = '\0'; + init_adma = IOP13XX_INIT_ADMA_DEFAULT; + } + str++; + } + } + return 1; +} + +__setup("iop13xx_init_adma", iop13xx_init_adma_setup); __setup("iop13xx_init_uart", iop13xx_init_uart_setup); __setup("iop13xx_init_i2c", iop13xx_init_i2c_setup); diff --git a/include/asm-arm/arch-iop13xx/adma.h b/include/asm-arm/arch-iop13xx/adma.h new file mode 100644 index 0000000..04006c1 --- /dev/null +++ b/include/asm-arm/arch-iop13xx/adma.h @@ -0,0 +1,544 @@ +/* + * Copyright(c) 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#ifndef _ADMA_H +#define _ADMA_H +#include +#include +#include +#include + +#define ADMA_ACCR(chan) (chan->mmr_base + 0x0) +#define ADMA_ACSR(chan) (chan->mmr_base + 0x4) +#define ADMA_ADAR(chan) (chan->mmr_base + 0x8) +#define ADMA_IIPCR(chan) (chan->mmr_base + 0x18) +#define ADMA_IIPAR(chan) (chan->mmr_base + 0x1c) +#define ADMA_IIPUAR(chan) (chan->mmr_base + 0x20) +#define ADMA_ANDAR(chan) (chan->mmr_base + 0x24) +#define ADMA_ADCR(chan) (chan->mmr_base + 0x28) +#define ADMA_CARMD(chan) (chan->mmr_base + 0x2c) +#define ADMA_ABCR(chan) (chan->mmr_base + 0x30) +#define ADMA_DLADR(chan) (chan->mmr_base + 0x34) +#define ADMA_DUADR(chan) (chan->mmr_base + 0x38) +#define ADMA_SLAR(src, chan) (chan->mmr_base + (0x3c + (src << 3))) +#define ADMA_SUAR(src, chan) (chan->mmr_base + (0x40 + (src << 3))) + +struct iop13xx_adma_src { + u32 src_addr; + union { + u32 upper_src_addr; + struct { + unsigned int pq_upper_src_addr:24; + unsigned int pq_dmlt:8; + }; + }; +}; + +struct iop13xx_adma_desc_ctrl { + unsigned int int_en:1; + unsigned int xfer_dir:2; + unsigned int src_select:4; + unsigned int zero_result:1; + unsigned int block_fill_en:1; + unsigned int crc_gen_en:1; + unsigned int crc_xfer_dis:1; + unsigned int crc_seed_fetch_dis:1; + unsigned int status_write_back_en:1; + unsigned int endian_swap_en:1; + unsigned int reserved0:2; + unsigned int pq_update_xfer_en:1; + unsigned int dual_xor_en:1; + unsigned int pq_xfer_en:1; + unsigned int p_xfer_dis:1; + unsigned int reserved1:10; + unsigned int relax_order_en:1; + unsigned int no_snoop_en:1; +}; + +struct iop13xx_adma_byte_count { + unsigned int byte_count:24; + unsigned int host_if:3; + unsigned int reserved:2; + unsigned int zero_result_err_q:1; + unsigned int zero_result_err:1; + unsigned int tx_complete:1; +}; + +struct iop13xx_adma_desc_hw { + u32 next_desc; + union { + u32 desc_ctrl; + struct iop13xx_adma_desc_ctrl desc_ctrl_field; + }; + union { + u32 crc_addr; + u32 block_fill_data; + u32 q_dest_addr; + }; + union { + u32 byte_count; + struct iop13xx_adma_byte_count byte_count_field; + }; + union { + u32 dest_addr; + u32 p_dest_addr; + }; + union { + u32 upper_dest_addr; + u32 pq_upper_dest_addr; + }; + struct iop13xx_adma_src src[1]; +}; + +struct iop13xx_adma_desc_dual_xor { + u32 next_desc; + u32 desc_ctrl; + u32 reserved; + u32 byte_count; + u32 h_dest_addr; + u32 h_upper_dest_addr; + u32 src0_addr; + u32 upper_src0_addr; + u32 src1_addr; + u32 upper_src1_addr; + u32 h_src_addr; + u32 h_upper_src_addr; + u32 d_src_addr; + u32 d_upper_src_addr; + u32 d_dest_addr; + u32 d_upper_dest_addr; +}; + +struct iop13xx_adma_desc_pq_update { + u32 next_desc; + u32 desc_ctrl; + u32 reserved; + u32 byte_count; + u32 p_dest_addr; + u32 p_upper_dest_addr; + u32 src0_addr; + u32 upper_src0_addr; + u32 src1_addr; + u32 upper_src1_addr; + u32 p_src_addr; + u32 p_upper_src_addr; + u32 q_src_addr; + struct { + unsigned int q_upper_src_addr:24; + unsigned int q_dmlt:8; + }; + u32 q_dest_addr; + u32 q_upper_dest_addr; +}; + +static inline int iop_adma_get_max_xor(void) +{ + return 16; +} + +static inline u32 iop_chan_get_current_descriptor(struct iop_adma_chan *chan) +{ + return __raw_readl(ADMA_ADAR(chan)); +} + +static inline void iop_chan_set_next_descriptor(struct iop_adma_chan *chan, + u32 next_desc_addr) +{ + __raw_writel(next_desc_addr, ADMA_ANDAR(chan)); +} + +#define ADMA_STATUS_BUSY (1 << 13) + +static inline char iop_chan_is_busy(struct iop_adma_chan *chan) +{ + if (__raw_readl(ADMA_ACSR(chan)) & + ADMA_STATUS_BUSY) + return 1; + else + return 0; +} + +static inline int +iop_chan_get_desc_align(struct iop_adma_chan *chan, int num_slots) +{ + return 1; +} +#define iop_desc_is_aligned(x, y) 1 + +static inline int +iop_chan_memcpy_slot_count(size_t len, int *slots_per_op) +{ + *slots_per_op = 1; + return 1; +} + +#define iop_chan_interrupt_slot_count(s, c) iop_chan_memcpy_slot_count(0, s) + +static inline int +iop_chan_memset_slot_count(size_t len, int *slots_per_op) +{ + *slots_per_op = 1; + return 1; +} + +static inline int +iop_chan_xor_slot_count(size_t len, int src_cnt, int *slots_per_op) +{ + int num_slots; + /* slots_to_find = 1 for basic descriptor + 1 per 4 sources above 1 + * (1 source => 8 bytes) (1 slot => 32 bytes) + */ + num_slots = 1 + (((src_cnt - 1) << 3) >> 5); + if (((src_cnt - 1) << 3) & 0x1f) + num_slots++; + + *slots_per_op = num_slots; + + return num_slots; +} + +#define ADMA_MAX_BYTE_COUNT (16 * 1024 * 1024) +#define IOP_ADMA_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT +#define IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT +#define IOP_ADMA_XOR_MAX_BYTE_COUNT ADMA_MAX_BYTE_COUNT +#define iop_chan_zero_sum_slot_count(l, s, o) iop_chan_xor_slot_count(l, s, o) + +static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + return hw_desc->dest_addr; +} + +static inline u32 iop_desc_get_byte_count(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + return hw_desc->byte_count_field.byte_count; +} + +static inline u32 iop_desc_get_src_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + int src_idx) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + return hw_desc->src[src_idx].src_addr; +} + +static inline u32 iop_desc_get_src_count(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + return hw_desc->desc_ctrl_field.src_select + 1; +} + +static inline void +iop_desc_init_memcpy(struct iop_adma_desc_slot *desc, int int_en) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop13xx_adma_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */ + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + hw_desc->crc_addr = 0; +} + +static inline void +iop_desc_init_memset(struct iop_adma_desc_slot *desc, int int_en) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop13xx_adma_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */ + u_desc_ctrl.field.block_fill_en = 1; + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + hw_desc->crc_addr = 0; +} + +/* to do: support buffers larger than ADMA_MAX_BYTE_COUNT */ +static inline void +iop_desc_init_xor(struct iop_adma_desc_slot *desc, int src_cnt, int int_en) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop13xx_adma_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.src_select = src_cnt - 1; + u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */ + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + hw_desc->crc_addr = 0; + +} +#define iop_desc_init_null_xor(d, s, i) iop_desc_init_xor(d, s, i) + +/* to do: support buffers larger than ADMA_MAX_BYTE_COUNT */ +static inline int +iop_desc_init_zero_sum(struct iop_adma_desc_slot *desc, int src_cnt, int int_en) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop13xx_adma_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.src_select = src_cnt - 1; + u_desc_ctrl.field.xfer_dir = 3; /* local to internal bus */ + u_desc_ctrl.field.zero_result = 1; + u_desc_ctrl.field.status_write_back_en = 1; + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + hw_desc->crc_addr = 0; + + return 1; +} + +static inline void iop_desc_set_byte_count(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + u32 byte_count) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + hw_desc->byte_count = byte_count; +} + +static inline void +iop_desc_set_zero_sum_byte_count(struct iop_adma_desc_slot *desc, u32 len) +{ + int slots_per_op = desc->slots_per_op; + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc, *iter; + int i = 0; + + if (len <= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT) { + hw_desc->byte_count = len; + } else { + do { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iter->byte_count = IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + len -= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + i += slots_per_op; + } while (len > IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT); + + if (len) { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iter->byte_count = len; + } + } +} + + +static inline void iop_desc_set_dest_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + dma_addr_t addr) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + hw_desc->dest_addr = addr; + hw_desc->upper_dest_addr = 0; +} + +static inline void iop_desc_set_memcpy_src_addr(struct iop_adma_desc_slot *desc, + dma_addr_t addr) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + hw_desc->src[0].src_addr = addr; + hw_desc->src[0].upper_src_addr = 0; +} + +static inline void iop_desc_set_xor_src_addr(struct iop_adma_desc_slot *desc, + int src_idx, dma_addr_t addr) +{ + int slot_cnt = desc->slot_cnt, slots_per_op = desc->slots_per_op; + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc, *iter; + int i = 0; + + do { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iter->src[src_idx].src_addr = addr; + iter->src[src_idx].upper_src_addr = 0; + slot_cnt -= slots_per_op; + if (slot_cnt) { + i += slots_per_op; + addr += IOP_ADMA_XOR_MAX_BYTE_COUNT; + } + } while (slot_cnt); +} + +static inline void +iop_desc_init_interrupt(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + iop_desc_init_memcpy(desc, 1); + iop_desc_set_byte_count(desc, chan, 0); + iop_desc_set_dest_addr(desc, chan, 0); + iop_desc_set_memcpy_src_addr(desc, 0); +} + +#define iop_desc_set_zero_sum_src_addr iop_desc_set_xor_src_addr + +static inline void iop_desc_set_next_desc(struct iop_adma_desc_slot *desc, + u32 next_desc_addr) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + BUG_ON(hw_desc->next_desc); + hw_desc->next_desc = next_desc_addr; +} + +static inline u32 iop_desc_get_next_desc(struct iop_adma_desc_slot *desc) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + return hw_desc->next_desc; +} + +static inline void iop_desc_clear_next_desc(struct iop_adma_desc_slot *desc) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + hw_desc->next_desc = 0; +} + +static inline void iop_desc_set_block_fill_val(struct iop_adma_desc_slot *desc, + u32 val) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + hw_desc->block_fill_data = val; +} + +static inline int iop_desc_get_zero_result(struct iop_adma_desc_slot *desc) +{ + struct iop13xx_adma_desc_hw *hw_desc = desc->hw_desc; + struct iop13xx_adma_desc_ctrl desc_ctrl = hw_desc->desc_ctrl_field; + struct iop13xx_adma_byte_count byte_count = hw_desc->byte_count_field; + + BUG_ON(!(byte_count.tx_complete && desc_ctrl.zero_result)); + + if (desc_ctrl.pq_xfer_en) + return byte_count.zero_result_err_q; + else + return byte_count.zero_result_err; +} + +static inline void iop_chan_append(struct iop_adma_chan *chan) +{ + u32 adma_accr; + + adma_accr = __raw_readl(ADMA_ACCR(chan)); + adma_accr |= 0x2; + __raw_writel(adma_accr, ADMA_ACCR(chan)); +} + +static inline void iop_chan_idle(int busy, struct iop_adma_chan *chan) +{ + do { } while (0); +} + +static inline u32 iop_chan_get_status(struct iop_adma_chan *chan) +{ + return __raw_readl(ADMA_ACSR(chan)); +} + +static inline void iop_chan_disable(struct iop_adma_chan *chan) +{ + u32 adma_chan_ctrl = __raw_readl(ADMA_ACCR(chan)); + adma_chan_ctrl &= ~0x1; + __raw_writel(adma_chan_ctrl, ADMA_ACCR(chan)); +} + +static inline void iop_chan_enable(struct iop_adma_chan *chan) +{ + u32 adma_chan_ctrl; + + adma_chan_ctrl = __raw_readl(ADMA_ACCR(chan)); + adma_chan_ctrl |= 0x1; + __raw_writel(adma_chan_ctrl, ADMA_ACCR(chan)); +} + +static inline void iop_adma_device_clear_eot_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(ADMA_ACSR(chan)); + status &= (1 << 12); + __raw_writel(status, ADMA_ACSR(chan)); +} + +static inline void iop_adma_device_clear_eoc_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(ADMA_ACSR(chan)); + status &= (1 << 11); + __raw_writel(status, ADMA_ACSR(chan)); +} + +static inline void iop_adma_device_clear_err_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(ADMA_ACSR(chan)); + status &= (1 << 9) | (1 << 5) | (1 << 4) | (1 << 3); + __raw_writel(status, ADMA_ACSR(chan)); +} + +static inline int +iop_is_err_int_parity(unsigned long status, struct iop_adma_chan *chan) +{ + return test_bit(9, &status); +} + +static inline int +iop_is_err_mcu_abort(unsigned long status, struct iop_adma_chan *chan) +{ + return test_bit(5, &status); +} + +static inline int +iop_is_err_int_tabort(unsigned long status, struct iop_adma_chan *chan) +{ + return test_bit(4, &status); +} + +static inline int +iop_is_err_int_mabort(unsigned long status, struct iop_adma_chan *chan) +{ + return test_bit(3, &status); +} + +static inline int +iop_is_err_pci_tabort(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +static inline int +iop_is_err_pci_mabort(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +static inline int +iop_is_err_split_tx(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +#endif /* _ADMA_H */ diff --git a/include/asm-arm/arch-iop13xx/iop13xx.h b/include/asm-arm/arch-iop13xx/iop13xx.h index e6736c3..d4e4f82 100644 --- a/include/asm-arm/arch-iop13xx/iop13xx.h +++ b/include/asm-arm/arch-iop13xx/iop13xx.h @@ -166,12 +166,22 @@ static inline int iop13xx_cpu_id(void) #define IOP13XX_INIT_I2C_1 (1 << 1) #define IOP13XX_INIT_I2C_2 (1 << 2) -#define IQ81340_NUM_UART 2 -#define IQ81340_NUM_I2C 3 -#define IQ81340_NUM_PHYS_MAP_FLASH 1 -#define IQ81340_MAX_PLAT_DEVICES (IQ81340_NUM_UART +\ - IQ81340_NUM_I2C +\ - IQ81340_NUM_PHYS_MAP_FLASH) +/* ADMA selection flags */ +/* INIT_ADMA_DEFAULT = Rely on CONFIG_IOP13XX_ADMA* */ +#define IOP13XX_INIT_ADMA_DEFAULT (0) +#define IOP13XX_INIT_ADMA_0 (1 << 0) +#define IOP13XX_INIT_ADMA_1 (1 << 1) +#define IOP13XX_INIT_ADMA_2 (1 << 2) + +/* Platform devices */ +#define IQ81340_NUM_UART 2 +#define IQ81340_NUM_I2C 3 +#define IQ81340_NUM_PHYS_MAP_FLASH 1 +#define IQ81340_NUM_ADMA 3 +#define IQ81340_MAX_PLAT_DEVICES (IQ81340_NUM_UART + \ + IQ81340_NUM_I2C + \ + IQ81340_NUM_PHYS_MAP_FLASH + \ + IQ81340_NUM_ADMA) /*========================== PMMR offsets for key registers ============*/ #define IOP13XX_ATU0_PMMR_OFFSET 0x00048000 @@ -444,22 +454,6 @@ static inline int iop13xx_cpu_id(void) /*==============================ADMA UNITS===============================*/ #define IOP13XX_ADMA_PHYS_BASE(chan) IOP13XX_REG_ADDR32_PHYS((chan << 9)) #define IOP13XX_ADMA_UPPER_PA(chan) (IOP13XX_ADMA_PHYS_BASE(chan) + 0xc0) -#define IOP13XX_ADMA_OFFSET(chan, ofs) IOP13XX_REG_ADDR32((chan << 9) + (ofs)) - -#define IOP13XX_ADMA_ACCR(chan) IOP13XX_ADMA_OFFSET(chan, 0x0) -#define IOP13XX_ADMA_ACSR(chan) IOP13XX_ADMA_OFFSET(chan, 0x4) -#define IOP13XX_ADMA_ADAR(chan) IOP13XX_ADMA_OFFSET(chan, 0x8) -#define IOP13XX_ADMA_IIPCR(chan) IOP13XX_ADMA_OFFSET(chan, 0x18) -#define IOP13XX_ADMA_IIPAR(chan) IOP13XX_ADMA_OFFSET(chan, 0x1c) -#define IOP13XX_ADMA_IIPUAR(chan) IOP13XX_ADMA_OFFSET(chan, 0x20) -#define IOP13XX_ADMA_ANDAR(chan) IOP13XX_ADMA_OFFSET(chan, 0x24) -#define IOP13XX_ADMA_ADCR(chan) IOP13XX_ADMA_OFFSET(chan, 0x28) -#define IOP13XX_ADMA_CARMD(chan) IOP13XX_ADMA_OFFSET(chan, 0x2c) -#define IOP13XX_ADMA_ABCR(chan) IOP13XX_ADMA_OFFSET(chan, 0x30) -#define IOP13XX_ADMA_DLADR(chan) IOP13XX_ADMA_OFFSET(chan, 0x34) -#define IOP13XX_ADMA_DUADR(chan) IOP13XX_ADMA_OFFSET(chan, 0x38) -#define IOP13XX_ADMA_SLAR(src, chan) IOP13XX_ADMA_OFFSET(chan, 0x3c + (src <<3)) -#define IOP13XX_ADMA_SUAR(src, chan) IOP13XX_ADMA_OFFSET(chan, 0x40 + (src <<3)) /*==============================XSI BRIDGE===============================*/ #define IOP13XX_XBG_BECSR IOP13XX_REG_ADDR32(0x178c) -- cgit v1.1 From 2492c845189a961a92d8537a44d233e8e1e45c6d Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 13:52:31 -0700 Subject: iop3xx: surface the iop3xx DMA and AAU units to the iop-adma driver Adds the platform device definitions and the architecture specific support routines (i.e. register initialization and descriptor formats) for the iop-adma driver. Changelog: * add support for > 1k zero sum buffer sizes * added dma/aau platform devices to iq80321 and iq80332 setup * fixed the calculation in iop_desc_is_aligned * support xor buffer sizes larger than 16MB * fix places where software descriptors are assumed to be contiguous, only hardware descriptors are contiguous for up to a PAGE_SIZE buffer size * convert to async_tx * add interrupt support * add platform devices for 80219 boards * do not call platform register macros in driver code * remove switch() statements for compatible register offsets/layouts * change over to bitmap based capabilities * remove unnecessary ARM assembly statement * checkpatch.pl fixes * gpl v2 only correction * phys move to dma_async_tx_descriptor Cc: Russell King Signed-off-by: Dan Williams --- arch/arm/mach-iop32x/glantank.c | 2 + arch/arm/mach-iop32x/iq31244.c | 5 + arch/arm/mach-iop32x/iq80321.c | 3 + arch/arm/mach-iop32x/n2100.c | 2 + arch/arm/mach-iop33x/iq80331.c | 3 + arch/arm/mach-iop33x/iq80332.c | 3 + arch/arm/plat-iop/Makefile | 2 + arch/arm/plat-iop/adma.c | 209 ++++++++ include/asm-arm/arch-iop32x/adma.h | 5 + include/asm-arm/arch-iop33x/adma.h | 5 + include/asm-arm/hardware/iop3xx-adma.h | 892 +++++++++++++++++++++++++++++++++ include/asm-arm/hardware/iop3xx.h | 68 +-- 12 files changed, 1139 insertions(+), 60 deletions(-) create mode 100644 arch/arm/plat-iop/adma.c create mode 100644 include/asm-arm/arch-iop32x/adma.h create mode 100644 include/asm-arm/arch-iop33x/adma.h create mode 100644 include/asm-arm/hardware/iop3xx-adma.h diff --git a/arch/arm/mach-iop32x/glantank.c b/arch/arm/mach-iop32x/glantank.c index 5776fd8..2b086ab 100644 --- a/arch/arm/mach-iop32x/glantank.c +++ b/arch/arm/mach-iop32x/glantank.c @@ -180,6 +180,8 @@ static void __init glantank_init_machine(void) platform_device_register(&iop3xx_i2c1_device); platform_device_register(&glantank_flash_device); platform_device_register(&glantank_serial_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); pm_power_off = glantank_power_off; } diff --git a/arch/arm/mach-iop32x/iq31244.c b/arch/arm/mach-iop32x/iq31244.c index d4eefbe..98cfa1c 100644 --- a/arch/arm/mach-iop32x/iq31244.c +++ b/arch/arm/mach-iop32x/iq31244.c @@ -298,9 +298,14 @@ static void __init iq31244_init_machine(void) platform_device_register(&iop3xx_i2c1_device); platform_device_register(&iq31244_flash_device); platform_device_register(&iq31244_serial_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); if (is_ep80219()) pm_power_off = ep80219_power_off; + + if (!is_80219()) + platform_device_register(&iop3xx_aau_channel); } static int __init force_ep80219_setup(char *str) diff --git a/arch/arm/mach-iop32x/iq80321.c b/arch/arm/mach-iop32x/iq80321.c index 8d9f4916..18ad29f 100644 --- a/arch/arm/mach-iop32x/iq80321.c +++ b/arch/arm/mach-iop32x/iq80321.c @@ -181,6 +181,9 @@ static void __init iq80321_init_machine(void) platform_device_register(&iop3xx_i2c1_device); platform_device_register(&iq80321_flash_device); platform_device_register(&iq80321_serial_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); + platform_device_register(&iop3xx_aau_channel); } MACHINE_START(IQ80321, "Intel IQ80321") diff --git a/arch/arm/mach-iop32x/n2100.c b/arch/arm/mach-iop32x/n2100.c index d55005d..390a97d 100644 --- a/arch/arm/mach-iop32x/n2100.c +++ b/arch/arm/mach-iop32x/n2100.c @@ -245,6 +245,8 @@ static void __init n2100_init_machine(void) platform_device_register(&iop3xx_i2c0_device); platform_device_register(&n2100_flash_device); platform_device_register(&n2100_serial_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); pm_power_off = n2100_power_off; diff --git a/arch/arm/mach-iop33x/iq80331.c b/arch/arm/mach-iop33x/iq80331.c index 2b06318..433188e 100644 --- a/arch/arm/mach-iop33x/iq80331.c +++ b/arch/arm/mach-iop33x/iq80331.c @@ -136,6 +136,9 @@ static void __init iq80331_init_machine(void) platform_device_register(&iop33x_uart0_device); platform_device_register(&iop33x_uart1_device); platform_device_register(&iq80331_flash_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); + platform_device_register(&iop3xx_aau_channel); } MACHINE_START(IQ80331, "Intel IQ80331") diff --git a/arch/arm/mach-iop33x/iq80332.c b/arch/arm/mach-iop33x/iq80332.c index 7889ce3..416c095 100644 --- a/arch/arm/mach-iop33x/iq80332.c +++ b/arch/arm/mach-iop33x/iq80332.c @@ -136,6 +136,9 @@ static void __init iq80332_init_machine(void) platform_device_register(&iop33x_uart0_device); platform_device_register(&iop33x_uart1_device); platform_device_register(&iq80332_flash_device); + platform_device_register(&iop3xx_dma_0_channel); + platform_device_register(&iop3xx_dma_1_channel); + platform_device_register(&iop3xx_aau_channel); } MACHINE_START(IQ80332, "Intel IQ80332") diff --git a/arch/arm/plat-iop/Makefile b/arch/arm/plat-iop/Makefile index 4d2b1da..36bff03 100644 --- a/arch/arm/plat-iop/Makefile +++ b/arch/arm/plat-iop/Makefile @@ -12,6 +12,7 @@ obj-$(CONFIG_ARCH_IOP32X) += setup.o obj-$(CONFIG_ARCH_IOP32X) += time.o obj-$(CONFIG_ARCH_IOP32X) += io.o obj-$(CONFIG_ARCH_IOP32X) += cp6.o +obj-$(CONFIG_ARCH_IOP32X) += adma.o # IOP33X obj-$(CONFIG_ARCH_IOP33X) += gpio.o @@ -21,6 +22,7 @@ obj-$(CONFIG_ARCH_IOP33X) += setup.o obj-$(CONFIG_ARCH_IOP33X) += time.o obj-$(CONFIG_ARCH_IOP33X) += io.o obj-$(CONFIG_ARCH_IOP33X) += cp6.o +obj-$(CONFIG_ARCH_IOP33X) += adma.o # IOP13XX obj-$(CONFIG_ARCH_IOP13XX) += cp6.o diff --git a/arch/arm/plat-iop/adma.c b/arch/arm/plat-iop/adma.c new file mode 100644 index 0000000..53c5e9a --- /dev/null +++ b/arch/arm/plat-iop/adma.c @@ -0,0 +1,209 @@ +/* + * platform device definitions for the iop3xx dma/xor engines + * Copyright © 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#include +#include +#include +#include +#include + +#ifdef CONFIG_ARCH_IOP32X +#define IRQ_DMA0_EOT IRQ_IOP32X_DMA0_EOT +#define IRQ_DMA0_EOC IRQ_IOP32X_DMA0_EOC +#define IRQ_DMA0_ERR IRQ_IOP32X_DMA0_ERR + +#define IRQ_DMA1_EOT IRQ_IOP32X_DMA1_EOT +#define IRQ_DMA1_EOC IRQ_IOP32X_DMA1_EOC +#define IRQ_DMA1_ERR IRQ_IOP32X_DMA1_ERR + +#define IRQ_AA_EOT IRQ_IOP32X_AA_EOT +#define IRQ_AA_EOC IRQ_IOP32X_AA_EOC +#define IRQ_AA_ERR IRQ_IOP32X_AA_ERR +#endif +#ifdef CONFIG_ARCH_IOP33X +#define IRQ_DMA0_EOT IRQ_IOP33X_DMA0_EOT +#define IRQ_DMA0_EOC IRQ_IOP33X_DMA0_EOC +#define IRQ_DMA0_ERR IRQ_IOP33X_DMA0_ERR + +#define IRQ_DMA1_EOT IRQ_IOP33X_DMA1_EOT +#define IRQ_DMA1_EOC IRQ_IOP33X_DMA1_EOC +#define IRQ_DMA1_ERR IRQ_IOP33X_DMA1_ERR + +#define IRQ_AA_EOT IRQ_IOP33X_AA_EOT +#define IRQ_AA_EOC IRQ_IOP33X_AA_EOC +#define IRQ_AA_ERR IRQ_IOP33X_AA_ERR +#endif +/* AAU and DMA Channels */ +static struct resource iop3xx_dma_0_resources[] = { + [0] = { + .start = IOP3XX_DMA_PHYS_BASE(0), + .end = IOP3XX_DMA_UPPER_PA(0), + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_DMA0_EOT, + .end = IRQ_DMA0_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_DMA0_EOC, + .end = IRQ_DMA0_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_DMA0_ERR, + .end = IRQ_DMA0_ERR, + .flags = IORESOURCE_IRQ + } +}; + +static struct resource iop3xx_dma_1_resources[] = { + [0] = { + .start = IOP3XX_DMA_PHYS_BASE(1), + .end = IOP3XX_DMA_UPPER_PA(1), + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_DMA1_EOT, + .end = IRQ_DMA1_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_DMA1_EOC, + .end = IRQ_DMA1_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_DMA1_ERR, + .end = IRQ_DMA1_ERR, + .flags = IORESOURCE_IRQ + } +}; + + +static struct resource iop3xx_aau_resources[] = { + [0] = { + .start = IOP3XX_AAU_PHYS_BASE, + .end = IOP3XX_AAU_UPPER_PA, + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = IRQ_AA_EOT, + .end = IRQ_AA_EOT, + .flags = IORESOURCE_IRQ + }, + [2] = { + .start = IRQ_AA_EOC, + .end = IRQ_AA_EOC, + .flags = IORESOURCE_IRQ + }, + [3] = { + .start = IRQ_AA_ERR, + .end = IRQ_AA_ERR, + .flags = IORESOURCE_IRQ + } +}; + +static u64 iop3xx_adma_dmamask = DMA_32BIT_MASK; + +static struct iop_adma_platform_data iop3xx_dma_0_data = { + .hw_id = DMA0_ID, + .pool_size = PAGE_SIZE, +}; + +static struct iop_adma_platform_data iop3xx_dma_1_data = { + .hw_id = DMA1_ID, + .pool_size = PAGE_SIZE, +}; + +static struct iop_adma_platform_data iop3xx_aau_data = { + .hw_id = AAU_ID, + .pool_size = 3 * PAGE_SIZE, +}; + +struct platform_device iop3xx_dma_0_channel = { + .name = "iop-adma", + .id = 0, + .num_resources = 4, + .resource = iop3xx_dma_0_resources, + .dev = { + .dma_mask = &iop3xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop3xx_dma_0_data, + }, +}; + +struct platform_device iop3xx_dma_1_channel = { + .name = "iop-adma", + .id = 1, + .num_resources = 4, + .resource = iop3xx_dma_1_resources, + .dev = { + .dma_mask = &iop3xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop3xx_dma_1_data, + }, +}; + +struct platform_device iop3xx_aau_channel = { + .name = "iop-adma", + .id = 2, + .num_resources = 4, + .resource = iop3xx_aau_resources, + .dev = { + .dma_mask = &iop3xx_adma_dmamask, + .coherent_dma_mask = DMA_64BIT_MASK, + .platform_data = (void *) &iop3xx_aau_data, + }, +}; + +static int __init iop3xx_adma_cap_init(void) +{ + #ifdef CONFIG_ARCH_IOP32X /* the 32x DMA does not perform CRC32C */ + dma_cap_set(DMA_MEMCPY, iop3xx_dma_0_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_dma_0_data.cap_mask); + #else + dma_cap_set(DMA_MEMCPY, iop3xx_dma_0_data.cap_mask); + dma_cap_set(DMA_MEMCPY_CRC32C, iop3xx_dma_0_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_dma_0_data.cap_mask); + #endif + + #ifdef CONFIG_ARCH_IOP32X /* the 32x DMA does not perform CRC32C */ + dma_cap_set(DMA_MEMCPY, iop3xx_dma_1_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_dma_1_data.cap_mask); + #else + dma_cap_set(DMA_MEMCPY, iop3xx_dma_1_data.cap_mask); + dma_cap_set(DMA_MEMCPY_CRC32C, iop3xx_dma_1_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_dma_1_data.cap_mask); + #endif + + #ifdef CONFIG_ARCH_IOP32X /* the 32x AAU does not perform zero sum */ + dma_cap_set(DMA_XOR, iop3xx_aau_data.cap_mask); + dma_cap_set(DMA_MEMSET, iop3xx_aau_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_aau_data.cap_mask); + #else + dma_cap_set(DMA_XOR, iop3xx_aau_data.cap_mask); + dma_cap_set(DMA_ZERO_SUM, iop3xx_aau_data.cap_mask); + dma_cap_set(DMA_MEMSET, iop3xx_aau_data.cap_mask); + dma_cap_set(DMA_INTERRUPT, iop3xx_aau_data.cap_mask); + #endif + + return 0; +} + +arch_initcall(iop3xx_adma_cap_init); diff --git a/include/asm-arm/arch-iop32x/adma.h b/include/asm-arm/arch-iop32x/adma.h new file mode 100644 index 0000000..5ed9203 --- /dev/null +++ b/include/asm-arm/arch-iop32x/adma.h @@ -0,0 +1,5 @@ +#ifndef IOP32X_ADMA_H +#define IOP32X_ADMA_H +#include +#endif + diff --git a/include/asm-arm/arch-iop33x/adma.h b/include/asm-arm/arch-iop33x/adma.h new file mode 100644 index 0000000..4b92f79 --- /dev/null +++ b/include/asm-arm/arch-iop33x/adma.h @@ -0,0 +1,5 @@ +#ifndef IOP33X_ADMA_H +#define IOP33X_ADMA_H +#include +#endif + diff --git a/include/asm-arm/hardware/iop3xx-adma.h b/include/asm-arm/hardware/iop3xx-adma.h new file mode 100644 index 0000000..10834b5 --- /dev/null +++ b/include/asm-arm/hardware/iop3xx-adma.h @@ -0,0 +1,892 @@ +/* + * Copyright © 2006, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. + * + */ +#ifndef _ADMA_H +#define _ADMA_H +#include +#include +#include +#include + +/* Memory copy units */ +#define DMA_CCR(chan) (chan->mmr_base + 0x0) +#define DMA_CSR(chan) (chan->mmr_base + 0x4) +#define DMA_DAR(chan) (chan->mmr_base + 0xc) +#define DMA_NDAR(chan) (chan->mmr_base + 0x10) +#define DMA_PADR(chan) (chan->mmr_base + 0x14) +#define DMA_PUADR(chan) (chan->mmr_base + 0x18) +#define DMA_LADR(chan) (chan->mmr_base + 0x1c) +#define DMA_BCR(chan) (chan->mmr_base + 0x20) +#define DMA_DCR(chan) (chan->mmr_base + 0x24) + +/* Application accelerator unit */ +#define AAU_ACR(chan) (chan->mmr_base + 0x0) +#define AAU_ASR(chan) (chan->mmr_base + 0x4) +#define AAU_ADAR(chan) (chan->mmr_base + 0x8) +#define AAU_ANDAR(chan) (chan->mmr_base + 0xc) +#define AAU_SAR(src, chan) (chan->mmr_base + (0x10 + ((src) << 2))) +#define AAU_DAR(chan) (chan->mmr_base + 0x20) +#define AAU_ABCR(chan) (chan->mmr_base + 0x24) +#define AAU_ADCR(chan) (chan->mmr_base + 0x28) +#define AAU_SAR_EDCR(src_edc) (chan->mmr_base + (0x02c + ((src_edc-4) << 2))) +#define AAU_EDCR0_IDX 8 +#define AAU_EDCR1_IDX 17 +#define AAU_EDCR2_IDX 26 + +#define DMA0_ID 0 +#define DMA1_ID 1 +#define AAU_ID 2 + +struct iop3xx_aau_desc_ctrl { + unsigned int int_en:1; + unsigned int blk1_cmd_ctrl:3; + unsigned int blk2_cmd_ctrl:3; + unsigned int blk3_cmd_ctrl:3; + unsigned int blk4_cmd_ctrl:3; + unsigned int blk5_cmd_ctrl:3; + unsigned int blk6_cmd_ctrl:3; + unsigned int blk7_cmd_ctrl:3; + unsigned int blk8_cmd_ctrl:3; + unsigned int blk_ctrl:2; + unsigned int dual_xor_en:1; + unsigned int tx_complete:1; + unsigned int zero_result_err:1; + unsigned int zero_result_en:1; + unsigned int dest_write_en:1; +}; + +struct iop3xx_aau_e_desc_ctrl { + unsigned int reserved:1; + unsigned int blk1_cmd_ctrl:3; + unsigned int blk2_cmd_ctrl:3; + unsigned int blk3_cmd_ctrl:3; + unsigned int blk4_cmd_ctrl:3; + unsigned int blk5_cmd_ctrl:3; + unsigned int blk6_cmd_ctrl:3; + unsigned int blk7_cmd_ctrl:3; + unsigned int blk8_cmd_ctrl:3; + unsigned int reserved2:7; +}; + +struct iop3xx_dma_desc_ctrl { + unsigned int pci_transaction:4; + unsigned int int_en:1; + unsigned int dac_cycle_en:1; + unsigned int mem_to_mem_en:1; + unsigned int crc_data_tx_en:1; + unsigned int crc_gen_en:1; + unsigned int crc_seed_dis:1; + unsigned int reserved:21; + unsigned int crc_tx_complete:1; +}; + +struct iop3xx_desc_dma { + u32 next_desc; + union { + u32 pci_src_addr; + u32 pci_dest_addr; + u32 src_addr; + }; + union { + u32 upper_pci_src_addr; + u32 upper_pci_dest_addr; + }; + union { + u32 local_pci_src_addr; + u32 local_pci_dest_addr; + u32 dest_addr; + }; + u32 byte_count; + union { + u32 desc_ctrl; + struct iop3xx_dma_desc_ctrl desc_ctrl_field; + }; + u32 crc_addr; +}; + +struct iop3xx_desc_aau { + u32 next_desc; + u32 src[4]; + u32 dest_addr; + u32 byte_count; + union { + u32 desc_ctrl; + struct iop3xx_aau_desc_ctrl desc_ctrl_field; + }; + union { + u32 src_addr; + u32 e_desc_ctrl; + struct iop3xx_aau_e_desc_ctrl e_desc_ctrl_field; + } src_edc[31]; +}; + +struct iop3xx_aau_gfmr { + unsigned int gfmr1:8; + unsigned int gfmr2:8; + unsigned int gfmr3:8; + unsigned int gfmr4:8; +}; + +struct iop3xx_desc_pq_xor { + u32 next_desc; + u32 src[3]; + union { + u32 data_mult1; + struct iop3xx_aau_gfmr data_mult1_field; + }; + u32 dest_addr; + u32 byte_count; + union { + u32 desc_ctrl; + struct iop3xx_aau_desc_ctrl desc_ctrl_field; + }; + union { + u32 src_addr; + u32 e_desc_ctrl; + struct iop3xx_aau_e_desc_ctrl e_desc_ctrl_field; + u32 data_multiplier; + struct iop3xx_aau_gfmr data_mult_field; + u32 reserved; + } src_edc_gfmr[19]; +}; + +struct iop3xx_desc_dual_xor { + u32 next_desc; + u32 src0_addr; + u32 src1_addr; + u32 h_src_addr; + u32 d_src_addr; + u32 h_dest_addr; + u32 byte_count; + union { + u32 desc_ctrl; + struct iop3xx_aau_desc_ctrl desc_ctrl_field; + }; + u32 d_dest_addr; +}; + +union iop3xx_desc { + struct iop3xx_desc_aau *aau; + struct iop3xx_desc_dma *dma; + struct iop3xx_desc_pq_xor *pq_xor; + struct iop3xx_desc_dual_xor *dual_xor; + void *ptr; +}; + +static inline int iop_adma_get_max_xor(void) +{ + return 32; +} + +static inline u32 iop_chan_get_current_descriptor(struct iop_adma_chan *chan) +{ + int id = chan->device->id; + + switch (id) { + case DMA0_ID: + case DMA1_ID: + return __raw_readl(DMA_DAR(chan)); + case AAU_ID: + return __raw_readl(AAU_ADAR(chan)); + default: + BUG(); + } + return 0; +} + +static inline void iop_chan_set_next_descriptor(struct iop_adma_chan *chan, + u32 next_desc_addr) +{ + int id = chan->device->id; + + switch (id) { + case DMA0_ID: + case DMA1_ID: + __raw_writel(next_desc_addr, DMA_NDAR(chan)); + break; + case AAU_ID: + __raw_writel(next_desc_addr, AAU_ANDAR(chan)); + break; + } + +} + +#define IOP_ADMA_STATUS_BUSY (1 << 10) +#define IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT (1024) +#define IOP_ADMA_XOR_MAX_BYTE_COUNT (16 * 1024 * 1024) +#define IOP_ADMA_MAX_BYTE_COUNT (16 * 1024 * 1024) + +static inline int iop_chan_is_busy(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(DMA_CSR(chan)); + return (status & IOP_ADMA_STATUS_BUSY) ? 1 : 0; +} + +static inline int iop_desc_is_aligned(struct iop_adma_desc_slot *desc, + int num_slots) +{ + /* num_slots will only ever be 1, 2, 4, or 8 */ + return (desc->idx & (num_slots - 1)) ? 0 : 1; +} + +/* to do: support large (i.e. > hw max) buffer sizes */ +static inline int iop_chan_memcpy_slot_count(size_t len, int *slots_per_op) +{ + *slots_per_op = 1; + return 1; +} + +/* to do: support large (i.e. > hw max) buffer sizes */ +static inline int iop_chan_memset_slot_count(size_t len, int *slots_per_op) +{ + *slots_per_op = 1; + return 1; +} + +static inline int iop3xx_aau_xor_slot_count(size_t len, int src_cnt, + int *slots_per_op) +{ + const static int slot_count_table[] = { 0, + 1, 1, 1, 1, /* 01 - 04 */ + 2, 2, 2, 2, /* 05 - 08 */ + 4, 4, 4, 4, /* 09 - 12 */ + 4, 4, 4, 4, /* 13 - 16 */ + 8, 8, 8, 8, /* 17 - 20 */ + 8, 8, 8, 8, /* 21 - 24 */ + 8, 8, 8, 8, /* 25 - 28 */ + 8, 8, 8, 8, /* 29 - 32 */ + }; + *slots_per_op = slot_count_table[src_cnt]; + return *slots_per_op; +} + +static inline int +iop_chan_interrupt_slot_count(int *slots_per_op, struct iop_adma_chan *chan) +{ + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return iop_chan_memcpy_slot_count(0, slots_per_op); + case AAU_ID: + return iop3xx_aau_xor_slot_count(0, 2, slots_per_op); + default: + BUG(); + } + return 0; +} + +static inline int iop_chan_xor_slot_count(size_t len, int src_cnt, + int *slots_per_op) +{ + int slot_cnt = iop3xx_aau_xor_slot_count(len, src_cnt, slots_per_op); + + if (len <= IOP_ADMA_XOR_MAX_BYTE_COUNT) + return slot_cnt; + + len -= IOP_ADMA_XOR_MAX_BYTE_COUNT; + while (len > IOP_ADMA_XOR_MAX_BYTE_COUNT) { + len -= IOP_ADMA_XOR_MAX_BYTE_COUNT; + slot_cnt += *slots_per_op; + } + + if (len) + slot_cnt += *slots_per_op; + + return slot_cnt; +} + +/* zero sum on iop3xx is limited to 1k at a time so it requires multiple + * descriptors + */ +static inline int iop_chan_zero_sum_slot_count(size_t len, int src_cnt, + int *slots_per_op) +{ + int slot_cnt = iop3xx_aau_xor_slot_count(len, src_cnt, slots_per_op); + + if (len <= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT) + return slot_cnt; + + len -= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + while (len > IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT) { + len -= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + slot_cnt += *slots_per_op; + } + + if (len) + slot_cnt += *slots_per_op; + + return slot_cnt; +} + +static inline u32 iop_desc_get_dest_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return hw_desc.dma->dest_addr; + case AAU_ID: + return hw_desc.aau->dest_addr; + default: + BUG(); + } + return 0; +} + +static inline u32 iop_desc_get_byte_count(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return hw_desc.dma->byte_count; + case AAU_ID: + return hw_desc.aau->byte_count; + default: + BUG(); + } + return 0; +} + +/* translate the src_idx to a descriptor word index */ +static inline int __desc_idx(int src_idx) +{ + const static int desc_idx_table[] = { 0, 0, 0, 0, + 0, 1, 2, 3, + 5, 6, 7, 8, + 9, 10, 11, 12, + 14, 15, 16, 17, + 18, 19, 20, 21, + 23, 24, 25, 26, + 27, 28, 29, 30, + }; + + return desc_idx_table[src_idx]; +} + +static inline u32 iop_desc_get_src_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + int src_idx) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return hw_desc.dma->src_addr; + case AAU_ID: + break; + default: + BUG(); + } + + if (src_idx < 4) + return hw_desc.aau->src[src_idx]; + else + return hw_desc.aau->src_edc[__desc_idx(src_idx)].src_addr; +} + +static inline void iop3xx_aau_desc_set_src_addr(struct iop3xx_desc_aau *hw_desc, + int src_idx, dma_addr_t addr) +{ + if (src_idx < 4) + hw_desc->src[src_idx] = addr; + else + hw_desc->src_edc[__desc_idx(src_idx)].src_addr = addr; +} + +static inline void +iop_desc_init_memcpy(struct iop_adma_desc_slot *desc, int int_en) +{ + struct iop3xx_desc_dma *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop3xx_dma_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.mem_to_mem_en = 1; + u_desc_ctrl.field.pci_transaction = 0xe; /* memory read block */ + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + hw_desc->upper_pci_src_addr = 0; + hw_desc->crc_addr = 0; +} + +static inline void +iop_desc_init_memset(struct iop_adma_desc_slot *desc, int int_en) +{ + struct iop3xx_desc_aau *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop3xx_aau_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + u_desc_ctrl.field.blk1_cmd_ctrl = 0x2; /* memory block fill */ + u_desc_ctrl.field.dest_write_en = 1; + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; +} + +static inline u32 +iop3xx_desc_init_xor(struct iop3xx_desc_aau *hw_desc, int src_cnt, int int_en) +{ + int i, shift; + u32 edcr; + union { + u32 value; + struct iop3xx_aau_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + switch (src_cnt) { + case 25 ... 32: + u_desc_ctrl.field.blk_ctrl = 0x3; /* use EDCR[2:0] */ + edcr = 0; + shift = 1; + for (i = 24; i < src_cnt; i++) { + edcr |= (1 << shift); + shift += 3; + } + hw_desc->src_edc[AAU_EDCR2_IDX].e_desc_ctrl = edcr; + src_cnt = 24; + /* fall through */ + case 17 ... 24: + if (!u_desc_ctrl.field.blk_ctrl) { + hw_desc->src_edc[AAU_EDCR2_IDX].e_desc_ctrl = 0; + u_desc_ctrl.field.blk_ctrl = 0x3; /* use EDCR[2:0] */ + } + edcr = 0; + shift = 1; + for (i = 16; i < src_cnt; i++) { + edcr |= (1 << shift); + shift += 3; + } + hw_desc->src_edc[AAU_EDCR1_IDX].e_desc_ctrl = edcr; + src_cnt = 16; + /* fall through */ + case 9 ... 16: + if (!u_desc_ctrl.field.blk_ctrl) + u_desc_ctrl.field.blk_ctrl = 0x2; /* use EDCR0 */ + edcr = 0; + shift = 1; + for (i = 8; i < src_cnt; i++) { + edcr |= (1 << shift); + shift += 3; + } + hw_desc->src_edc[AAU_EDCR0_IDX].e_desc_ctrl = edcr; + src_cnt = 8; + /* fall through */ + case 2 ... 8: + shift = 1; + for (i = 0; i < src_cnt; i++) { + u_desc_ctrl.value |= (1 << shift); + shift += 3; + } + + if (!u_desc_ctrl.field.blk_ctrl && src_cnt > 4) + u_desc_ctrl.field.blk_ctrl = 0x1; /* use mini-desc */ + } + + u_desc_ctrl.field.dest_write_en = 1; + u_desc_ctrl.field.blk1_cmd_ctrl = 0x7; /* direct fill */ + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; + + return u_desc_ctrl.value; +} + +static inline void +iop_desc_init_xor(struct iop_adma_desc_slot *desc, int src_cnt, int int_en) +{ + iop3xx_desc_init_xor(desc->hw_desc, src_cnt, int_en); +} + +/* return the number of operations */ +static inline int +iop_desc_init_zero_sum(struct iop_adma_desc_slot *desc, int src_cnt, int int_en) +{ + int slot_cnt = desc->slot_cnt, slots_per_op = desc->slots_per_op; + struct iop3xx_desc_aau *hw_desc, *prev_hw_desc, *iter; + union { + u32 value; + struct iop3xx_aau_desc_ctrl field; + } u_desc_ctrl; + int i, j; + + hw_desc = desc->hw_desc; + + for (i = 0, j = 0; (slot_cnt -= slots_per_op) >= 0; + i += slots_per_op, j++) { + iter = iop_hw_desc_slot_idx(hw_desc, i); + u_desc_ctrl.value = iop3xx_desc_init_xor(iter, src_cnt, int_en); + u_desc_ctrl.field.dest_write_en = 0; + u_desc_ctrl.field.zero_result_en = 1; + u_desc_ctrl.field.int_en = int_en; + iter->desc_ctrl = u_desc_ctrl.value; + + /* for the subsequent descriptors preserve the store queue + * and chain them together + */ + if (i) { + prev_hw_desc = + iop_hw_desc_slot_idx(hw_desc, i - slots_per_op); + prev_hw_desc->next_desc = + (u32) (desc->async_tx.phys + (i << 5)); + } + } + + return j; +} + +static inline void +iop_desc_init_null_xor(struct iop_adma_desc_slot *desc, int src_cnt, int int_en) +{ + struct iop3xx_desc_aau *hw_desc = desc->hw_desc; + union { + u32 value; + struct iop3xx_aau_desc_ctrl field; + } u_desc_ctrl; + + u_desc_ctrl.value = 0; + switch (src_cnt) { + case 25 ... 32: + u_desc_ctrl.field.blk_ctrl = 0x3; /* use EDCR[2:0] */ + hw_desc->src_edc[AAU_EDCR2_IDX].e_desc_ctrl = 0; + /* fall through */ + case 17 ... 24: + if (!u_desc_ctrl.field.blk_ctrl) { + hw_desc->src_edc[AAU_EDCR2_IDX].e_desc_ctrl = 0; + u_desc_ctrl.field.blk_ctrl = 0x3; /* use EDCR[2:0] */ + } + hw_desc->src_edc[AAU_EDCR1_IDX].e_desc_ctrl = 0; + /* fall through */ + case 9 ... 16: + if (!u_desc_ctrl.field.blk_ctrl) + u_desc_ctrl.field.blk_ctrl = 0x2; /* use EDCR0 */ + hw_desc->src_edc[AAU_EDCR0_IDX].e_desc_ctrl = 0; + /* fall through */ + case 1 ... 8: + if (!u_desc_ctrl.field.blk_ctrl && src_cnt > 4) + u_desc_ctrl.field.blk_ctrl = 0x1; /* use mini-desc */ + } + + u_desc_ctrl.field.dest_write_en = 0; + u_desc_ctrl.field.int_en = int_en; + hw_desc->desc_ctrl = u_desc_ctrl.value; +} + +static inline void iop_desc_set_byte_count(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + u32 byte_count) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + hw_desc.dma->byte_count = byte_count; + break; + case AAU_ID: + hw_desc.aau->byte_count = byte_count; + break; + default: + BUG(); + } +} + +static inline void +iop_desc_init_interrupt(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + iop_desc_init_memcpy(desc, 1); + hw_desc.dma->byte_count = 0; + hw_desc.dma->dest_addr = 0; + hw_desc.dma->src_addr = 0; + break; + case AAU_ID: + iop_desc_init_null_xor(desc, 2, 1); + hw_desc.aau->byte_count = 0; + hw_desc.aau->dest_addr = 0; + hw_desc.aau->src[0] = 0; + hw_desc.aau->src[1] = 0; + break; + default: + BUG(); + } +} + +static inline void +iop_desc_set_zero_sum_byte_count(struct iop_adma_desc_slot *desc, u32 len) +{ + int slots_per_op = desc->slots_per_op; + struct iop3xx_desc_aau *hw_desc = desc->hw_desc, *iter; + int i = 0; + + if (len <= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT) { + hw_desc->byte_count = len; + } else { + do { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iter->byte_count = IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + len -= IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT; + i += slots_per_op; + } while (len > IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT); + + if (len) { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iter->byte_count = len; + } + } +} + +static inline void iop_desc_set_dest_addr(struct iop_adma_desc_slot *desc, + struct iop_adma_chan *chan, + dma_addr_t addr) +{ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + hw_desc.dma->dest_addr = addr; + break; + case AAU_ID: + hw_desc.aau->dest_addr = addr; + break; + default: + BUG(); + } +} + +static inline void iop_desc_set_memcpy_src_addr(struct iop_adma_desc_slot *desc, + dma_addr_t addr) +{ + struct iop3xx_desc_dma *hw_desc = desc->hw_desc; + hw_desc->src_addr = addr; +} + +static inline void +iop_desc_set_zero_sum_src_addr(struct iop_adma_desc_slot *desc, int src_idx, + dma_addr_t addr) +{ + + struct iop3xx_desc_aau *hw_desc = desc->hw_desc, *iter; + int slot_cnt = desc->slot_cnt, slots_per_op = desc->slots_per_op; + int i; + + for (i = 0; (slot_cnt -= slots_per_op) >= 0; + i += slots_per_op, addr += IOP_ADMA_ZERO_SUM_MAX_BYTE_COUNT) { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iop3xx_aau_desc_set_src_addr(iter, src_idx, addr); + } +} + +static inline void iop_desc_set_xor_src_addr(struct iop_adma_desc_slot *desc, + int src_idx, dma_addr_t addr) +{ + + struct iop3xx_desc_aau *hw_desc = desc->hw_desc, *iter; + int slot_cnt = desc->slot_cnt, slots_per_op = desc->slots_per_op; + int i; + + for (i = 0; (slot_cnt -= slots_per_op) >= 0; + i += slots_per_op, addr += IOP_ADMA_XOR_MAX_BYTE_COUNT) { + iter = iop_hw_desc_slot_idx(hw_desc, i); + iop3xx_aau_desc_set_src_addr(iter, src_idx, addr); + } +} + +static inline void iop_desc_set_next_desc(struct iop_adma_desc_slot *desc, + u32 next_desc_addr) +{ + /* hw_desc->next_desc is the same location for all channels */ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + BUG_ON(hw_desc.dma->next_desc); + hw_desc.dma->next_desc = next_desc_addr; +} + +static inline u32 iop_desc_get_next_desc(struct iop_adma_desc_slot *desc) +{ + /* hw_desc->next_desc is the same location for all channels */ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + return hw_desc.dma->next_desc; +} + +static inline void iop_desc_clear_next_desc(struct iop_adma_desc_slot *desc) +{ + /* hw_desc->next_desc is the same location for all channels */ + union iop3xx_desc hw_desc = { .ptr = desc->hw_desc, }; + hw_desc.dma->next_desc = 0; +} + +static inline void iop_desc_set_block_fill_val(struct iop_adma_desc_slot *desc, + u32 val) +{ + struct iop3xx_desc_aau *hw_desc = desc->hw_desc; + hw_desc->src[0] = val; +} + +static inline int iop_desc_get_zero_result(struct iop_adma_desc_slot *desc) +{ + struct iop3xx_desc_aau *hw_desc = desc->hw_desc; + struct iop3xx_aau_desc_ctrl desc_ctrl = hw_desc->desc_ctrl_field; + + BUG_ON(!(desc_ctrl.tx_complete && desc_ctrl.zero_result_en)); + return desc_ctrl.zero_result_err; +} + +static inline void iop_chan_append(struct iop_adma_chan *chan) +{ + u32 dma_chan_ctrl; + /* workaround dropped interrupts on 3xx */ + mod_timer(&chan->cleanup_watchdog, jiffies + msecs_to_jiffies(3)); + + dma_chan_ctrl = __raw_readl(DMA_CCR(chan)); + dma_chan_ctrl |= 0x2; + __raw_writel(dma_chan_ctrl, DMA_CCR(chan)); +} + +static inline void iop_chan_idle(int busy, struct iop_adma_chan *chan) +{ + if (!busy) + del_timer(&chan->cleanup_watchdog); +} + +static inline u32 iop_chan_get_status(struct iop_adma_chan *chan) +{ + return __raw_readl(DMA_CSR(chan)); +} + +static inline void iop_chan_disable(struct iop_adma_chan *chan) +{ + u32 dma_chan_ctrl = __raw_readl(DMA_CCR(chan)); + dma_chan_ctrl &= ~1; + __raw_writel(dma_chan_ctrl, DMA_CCR(chan)); +} + +static inline void iop_chan_enable(struct iop_adma_chan *chan) +{ + u32 dma_chan_ctrl = __raw_readl(DMA_CCR(chan)); + + dma_chan_ctrl |= 1; + __raw_writel(dma_chan_ctrl, DMA_CCR(chan)); +} + +static inline void iop_adma_device_clear_eot_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(DMA_CSR(chan)); + status &= (1 << 9); + __raw_writel(status, DMA_CSR(chan)); +} + +static inline void iop_adma_device_clear_eoc_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(DMA_CSR(chan)); + status &= (1 << 8); + __raw_writel(status, DMA_CSR(chan)); +} + +static inline void iop_adma_device_clear_err_status(struct iop_adma_chan *chan) +{ + u32 status = __raw_readl(DMA_CSR(chan)); + + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + status &= (1 << 5) | (1 << 3) | (1 << 2) | (1 << 1); + break; + case AAU_ID: + status &= (1 << 5); + break; + default: + BUG(); + } + + __raw_writel(status, DMA_CSR(chan)); +} + +static inline int +iop_is_err_int_parity(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +static inline int +iop_is_err_mcu_abort(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +static inline int +iop_is_err_int_tabort(unsigned long status, struct iop_adma_chan *chan) +{ + return 0; +} + +static inline int +iop_is_err_int_mabort(unsigned long status, struct iop_adma_chan *chan) +{ + return test_bit(5, &status); +} + +static inline int +iop_is_err_pci_tabort(unsigned long status, struct iop_adma_chan *chan) +{ + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return test_bit(2, &status); + default: + return 0; + } +} + +static inline int +iop_is_err_pci_mabort(unsigned long status, struct iop_adma_chan *chan) +{ + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return test_bit(3, &status); + default: + return 0; + } +} + +static inline int +iop_is_err_split_tx(unsigned long status, struct iop_adma_chan *chan) +{ + switch (chan->device->id) { + case DMA0_ID: + case DMA1_ID: + return test_bit(1, &status); + default: + return 0; + } +} +#endif /* _ADMA_H */ diff --git a/include/asm-arm/hardware/iop3xx.h b/include/asm-arm/hardware/iop3xx.h index 63feceb..81ca5d3 100644 --- a/include/asm-arm/hardware/iop3xx.h +++ b/include/asm-arm/hardware/iop3xx.h @@ -144,24 +144,9 @@ extern int init_atu; #define IOP3XX_IAR (volatile u32 *)IOP3XX_REG_ADDR(0x0380) /* DMA Controller */ -#define IOP3XX_DMA0_CCR (volatile u32 *)IOP3XX_REG_ADDR(0x0400) -#define IOP3XX_DMA0_CSR (volatile u32 *)IOP3XX_REG_ADDR(0x0404) -#define IOP3XX_DMA0_DAR (volatile u32 *)IOP3XX_REG_ADDR(0x040c) -#define IOP3XX_DMA0_NDAR (volatile u32 *)IOP3XX_REG_ADDR(0x0410) -#define IOP3XX_DMA0_PADR (volatile u32 *)IOP3XX_REG_ADDR(0x0414) -#define IOP3XX_DMA0_PUADR (volatile u32 *)IOP3XX_REG_ADDR(0x0418) -#define IOP3XX_DMA0_LADR (volatile u32 *)IOP3XX_REG_ADDR(0x041c) -#define IOP3XX_DMA0_BCR (volatile u32 *)IOP3XX_REG_ADDR(0x0420) -#define IOP3XX_DMA0_DCR (volatile u32 *)IOP3XX_REG_ADDR(0x0424) -#define IOP3XX_DMA1_CCR (volatile u32 *)IOP3XX_REG_ADDR(0x0440) -#define IOP3XX_DMA1_CSR (volatile u32 *)IOP3XX_REG_ADDR(0x0444) -#define IOP3XX_DMA1_DAR (volatile u32 *)IOP3XX_REG_ADDR(0x044c) -#define IOP3XX_DMA1_NDAR (volatile u32 *)IOP3XX_REG_ADDR(0x0450) -#define IOP3XX_DMA1_PADR (volatile u32 *)IOP3XX_REG_ADDR(0x0454) -#define IOP3XX_DMA1_PUADR (volatile u32 *)IOP3XX_REG_ADDR(0x0458) -#define IOP3XX_DMA1_LADR (volatile u32 *)IOP3XX_REG_ADDR(0x045c) -#define IOP3XX_DMA1_BCR (volatile u32 *)IOP3XX_REG_ADDR(0x0460) -#define IOP3XX_DMA1_DCR (volatile u32 *)IOP3XX_REG_ADDR(0x0464) +#define IOP3XX_DMA_PHYS_BASE(chan) (IOP3XX_PERIPHERAL_PHYS_BASE + \ + (0x400 + (chan << 6))) +#define IOP3XX_DMA_UPPER_PA(chan) (IOP3XX_DMA_PHYS_BASE(chan) + 0x27) /* Peripheral bus interface */ #define IOP3XX_PBCR (volatile u32 *)IOP3XX_REG_ADDR(0x0680) @@ -210,48 +195,8 @@ extern int init_atu; #define IOP_TMR_RATIO_1_1 0x00 /* Application accelerator unit */ -#define IOP3XX_AAU_ACR (volatile u32 *)IOP3XX_REG_ADDR(0x0800) -#define IOP3XX_AAU_ASR (volatile u32 *)IOP3XX_REG_ADDR(0x0804) -#define IOP3XX_AAU_ADAR (volatile u32 *)IOP3XX_REG_ADDR(0x0808) -#define IOP3XX_AAU_ANDAR (volatile u32 *)IOP3XX_REG_ADDR(0x080c) -#define IOP3XX_AAU_SAR1 (volatile u32 *)IOP3XX_REG_ADDR(0x0810) -#define IOP3XX_AAU_SAR2 (volatile u32 *)IOP3XX_REG_ADDR(0x0814) -#define IOP3XX_AAU_SAR3 (volatile u32 *)IOP3XX_REG_ADDR(0x0818) -#define IOP3XX_AAU_SAR4 (volatile u32 *)IOP3XX_REG_ADDR(0x081c) -#define IOP3XX_AAU_DAR (volatile u32 *)IOP3XX_REG_ADDR(0x0820) -#define IOP3XX_AAU_ABCR (volatile u32 *)IOP3XX_REG_ADDR(0x0824) -#define IOP3XX_AAU_ADCR (volatile u32 *)IOP3XX_REG_ADDR(0x0828) -#define IOP3XX_AAU_SAR5 (volatile u32 *)IOP3XX_REG_ADDR(0x082c) -#define IOP3XX_AAU_SAR6 (volatile u32 *)IOP3XX_REG_ADDR(0x0830) -#define IOP3XX_AAU_SAR7 (volatile u32 *)IOP3XX_REG_ADDR(0x0834) -#define IOP3XX_AAU_SAR8 (volatile u32 *)IOP3XX_REG_ADDR(0x0838) -#define IOP3XX_AAU_EDCR0 (volatile u32 *)IOP3XX_REG_ADDR(0x083c) -#define IOP3XX_AAU_SAR9 (volatile u32 *)IOP3XX_REG_ADDR(0x0840) -#define IOP3XX_AAU_SAR10 (volatile u32 *)IOP3XX_REG_ADDR(0x0844) -#define IOP3XX_AAU_SAR11 (volatile u32 *)IOP3XX_REG_ADDR(0x0848) -#define IOP3XX_AAU_SAR12 (volatile u32 *)IOP3XX_REG_ADDR(0x084c) -#define IOP3XX_AAU_SAR13 (volatile u32 *)IOP3XX_REG_ADDR(0x0850) -#define IOP3XX_AAU_SAR14 (volatile u32 *)IOP3XX_REG_ADDR(0x0854) -#define IOP3XX_AAU_SAR15 (volatile u32 *)IOP3XX_REG_ADDR(0x0858) -#define IOP3XX_AAU_SAR16 (volatile u32 *)IOP3XX_REG_ADDR(0x085c) -#define IOP3XX_AAU_EDCR1 (volatile u32 *)IOP3XX_REG_ADDR(0x0860) -#define IOP3XX_AAU_SAR17 (volatile u32 *)IOP3XX_REG_ADDR(0x0864) -#define IOP3XX_AAU_SAR18 (volatile u32 *)IOP3XX_REG_ADDR(0x0868) -#define IOP3XX_AAU_SAR19 (volatile u32 *)IOP3XX_REG_ADDR(0x086c) -#define IOP3XX_AAU_SAR20 (volatile u32 *)IOP3XX_REG_ADDR(0x0870) -#define IOP3XX_AAU_SAR21 (volatile u32 *)IOP3XX_REG_ADDR(0x0874) -#define IOP3XX_AAU_SAR22 (volatile u32 *)IOP3XX_REG_ADDR(0x0878) -#define IOP3XX_AAU_SAR23 (volatile u32 *)IOP3XX_REG_ADDR(0x087c) -#define IOP3XX_AAU_SAR24 (volatile u32 *)IOP3XX_REG_ADDR(0x0880) -#define IOP3XX_AAU_EDCR2 (volatile u32 *)IOP3XX_REG_ADDR(0x0884) -#define IOP3XX_AAU_SAR25 (volatile u32 *)IOP3XX_REG_ADDR(0x0888) -#define IOP3XX_AAU_SAR26 (volatile u32 *)IOP3XX_REG_ADDR(0x088c) -#define IOP3XX_AAU_SAR27 (volatile u32 *)IOP3XX_REG_ADDR(0x0890) -#define IOP3XX_AAU_SAR28 (volatile u32 *)IOP3XX_REG_ADDR(0x0894) -#define IOP3XX_AAU_SAR29 (volatile u32 *)IOP3XX_REG_ADDR(0x0898) -#define IOP3XX_AAU_SAR30 (volatile u32 *)IOP3XX_REG_ADDR(0x089c) -#define IOP3XX_AAU_SAR31 (volatile u32 *)IOP3XX_REG_ADDR(0x08a0) -#define IOP3XX_AAU_SAR32 (volatile u32 *)IOP3XX_REG_ADDR(0x08a4) +#define IOP3XX_AAU_PHYS_BASE (IOP3XX_PERIPHERAL_PHYS_BASE + 0x800) +#define IOP3XX_AAU_UPPER_PA (IOP3XX_AAU_PHYS_BASE + 0xa7) /* I2C bus interface unit */ #define IOP3XX_ICR0 (volatile u32 *)IOP3XX_REG_ADDR(0x1680) @@ -329,6 +274,9 @@ static inline void write_tisr(u32 val) asm volatile("mcr p6, 0, %0, c6, c1, 0" : : "r" (val)); } +extern struct platform_device iop3xx_dma_0_channel; +extern struct platform_device iop3xx_dma_1_channel; +extern struct platform_device iop3xx_aau_channel; extern struct platform_device iop3xx_i2c0_device; extern struct platform_device iop3xx_i2c1_device; -- cgit v1.1 From 5816815f7850509ed51ab94eb4f644e405ccb865 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 2 Jan 2007 11:10:43 -0700 Subject: ARM: Add drivers/dma to arch/arm/Kconfig Cc: Russell King Signed-off-by: Dan Williams --- arch/arm/Kconfig | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 50d9f3e..0cb2d4f 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -1034,6 +1034,8 @@ source "drivers/mmc/Kconfig" source "drivers/rtc/Kconfig" +source "drivers/dma/Kconfig" + endmenu source "fs/Kconfig" -- cgit v1.1 From 3039f0735a280b54c7364fbfe6a9287f7f0b510a Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 13 Jul 2007 08:06:19 -0700 Subject: ioatdma: add the unisys "i/oat" pci vendor/device id Cc: John Magolan Signed-off-by: Shannon Nelson Signed-off-by: Dan Williams --- drivers/dma/ioatdma.c | 2 ++ include/linux/pci_ids.h | 3 +++ 2 files changed, 5 insertions(+) diff --git a/drivers/dma/ioatdma.c b/drivers/dma/ioatdma.c index 81810b3..5fbe56b 100644 --- a/drivers/dma/ioatdma.c +++ b/drivers/dma/ioatdma.c @@ -517,6 +517,8 @@ static enum dma_status ioat_dma_is_complete(struct dma_chan *chan, static struct pci_device_id ioat_pci_tbl[] = { { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IOAT) }, + { PCI_DEVICE(PCI_VENDOR_ID_UNISYS, + PCI_DEVICE_ID_UNISYS_DMA_DIRECTOR) }, { 0, } }; diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h index 5b1c999..0275f69 100644 --- a/include/linux/pci_ids.h +++ b/include/linux/pci_ids.h @@ -475,6 +475,9 @@ #define PCI_DEVICE_ID_IBM_ICOM_V2_ONE_PORT_RVX_ONE_PORT_MDM_PCIE 0x0361 #define PCI_DEVICE_ID_IBM_ICOM_FOUR_PORT_MODEL 0x252 +#define PCI_VENDOR_ID_UNISYS 0x1018 +#define PCI_DEVICE_ID_UNISYS_DMA_DIRECTOR 0x001C + #define PCI_VENDOR_ID_COMPEX2 0x101a /* pci.ids says "AT&T GIS (NCR)" */ #define PCI_DEVICE_ID_COMPEX2_100VG 0x0005 -- cgit v1.1