diff options
Diffstat (limited to 'arch')
50 files changed, 1094 insertions, 457 deletions
diff --git a/arch/arm/mach-realview/realview_eb.c b/arch/arm/mach-realview/realview_eb.c index 3dba666..ecec2f8 100644 --- a/arch/arm/mach-realview/realview_eb.c +++ b/arch/arm/mach-realview/realview_eb.c @@ -165,7 +165,7 @@ static void __init gic_init_irq(void) #endif gic_dist_init(0, __io_address(REALVIEW_GIC_DIST_BASE), 29); gic_cpu_init(0, __io_address(REALVIEW_GIC_CPU_BASE)); -#ifdef CONFIG_REALVIEW_MPCORE +#if defined(CONFIG_REALVIEW_MPCORE) && !defined(CONFIG_REALVIEW_MPCORE_REVB) gic_dist_init(1, __io_address(REALVIEW_GIC1_DIST_BASE), 64); gic_cpu_init(1, __io_address(REALVIEW_GIC1_CPU_BASE)); gic_cascade_irq(1, IRQ_EB_IRQ1); diff --git a/arch/arm/vfp/vfpmodule.c b/arch/arm/vfp/vfpmodule.c index 04ddab2..eea3f50 100644 --- a/arch/arm/vfp/vfpmodule.c +++ b/arch/arm/vfp/vfpmodule.c @@ -323,6 +323,7 @@ static int __init vfp_init(void) * we just need to read the VFPSID register. */ vfp_vector = vfp_testing_entry; + barrier(); vfpsid = fmrx(FPSID); barrier(); vfp_vector = vfp_null_entry; diff --git a/arch/blackfin/mach-common/pm.c b/arch/blackfin/mach-common/pm.c index 1772d8d..b103027 100644 --- a/arch/blackfin/mach-common/pm.c +++ b/arch/blackfin/mach-common/pm.c @@ -158,10 +158,16 @@ static int bfin_pm_finish(suspend_state_t state) return 0; } +static int bfin_pm_valid(suspend_state_t state) +{ + return (state == PM_SUSPEND_STANDBY); +} + struct pm_ops bfin_pm_ops = { .prepare = bfin_pm_prepare, .enter = bfin_pm_enter, .finish = bfin_pm_finish, + .valid = bfin_pm_valid, }; static int __init bfin_pm_init(void) diff --git a/arch/h8300/Kconfig.debug b/arch/h8300/Kconfig.debug index 554efe6..996d97e 100644 --- a/arch/h8300/Kconfig.debug +++ b/arch/h8300/Kconfig.debug @@ -59,7 +59,7 @@ config BLKDEV_RESERVE help Reserved BLKDEV area. -config CONFIG_BLKDEV_RESERVE_ADDRESS +config BLKDEV_RESERVE_ADDRESS hex 'start address' depends on BLKDEV_RESERVE help diff --git a/arch/i386/kernel/cpu/intel_cacheinfo.c b/arch/i386/kernel/cpu/intel_cacheinfo.c index d5a456d..db6c25a 100644 --- a/arch/i386/kernel/cpu/intel_cacheinfo.c +++ b/arch/i386/kernel/cpu/intel_cacheinfo.c @@ -515,7 +515,7 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu) cpuid4_info[cpu] = kzalloc( sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL); - if (unlikely(cpuid4_info[cpu] == NULL)) + if (cpuid4_info[cpu] == NULL) return -ENOMEM; oldmask = current->cpus_allowed; @@ -748,6 +748,8 @@ static void __cpuinit cache_remove_dev(struct sys_device * sys_dev) unsigned int cpu = sys_dev->id; unsigned long i; + if (cpuid4_info[cpu] == NULL) + return; for (i = 0; i < num_cache_leaves; i++) { cache_remove_shared_cpu_map(cpu, i); kobject_unregister(&(INDEX_KOBJECT_PTR(cpu,i)->kobj)); diff --git a/arch/i386/mach-generic/Makefile b/arch/i386/mach-generic/Makefile index 77fbc9f..6914485 100644 --- a/arch/i386/mach-generic/Makefile +++ b/arch/i386/mach-generic/Makefile @@ -2,6 +2,6 @@ # Makefile for the generic architecture # -EXTRA_CFLAGS += -I../kernel +EXTRA_CFLAGS := -Iarch/i386/kernel obj-y := probe.o summit.o bigsmp.o es7000.o default.o ../mach-es7000/ diff --git a/arch/i386/mach-voyager/Makefile b/arch/i386/mach-voyager/Makefile index f24d296..33b74cf 100644 --- a/arch/i386/mach-voyager/Makefile +++ b/arch/i386/mach-voyager/Makefile @@ -2,7 +2,7 @@ # Makefile for the linux kernel. # -EXTRA_CFLAGS += -I../kernel +EXTRA_CFLAGS := -Iarch/i386/kernel obj-y := setup.o voyager_basic.o voyager_thread.o obj-$(CONFIG_SMP) += voyager_smp.o voyager_cat.o diff --git a/arch/i386/mm/highmem.c b/arch/i386/mm/highmem.c index ad8d86c..1c3bf95 100644 --- a/arch/i386/mm/highmem.c +++ b/arch/i386/mm/highmem.c @@ -34,17 +34,16 @@ void *kmap_atomic_prot(struct page *page, enum km_type type, pgprot_t prot) /* even !CONFIG_PREEMPT needs this, for in_atomic in do_page_fault */ pagefault_disable(); - idx = type + KM_TYPE_NR*smp_processor_id(); - BUG_ON(!pte_none(*(kmap_pte-idx))); - if (!PageHighMem(page)) return page_address(page); + idx = type + KM_TYPE_NR*smp_processor_id(); vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx); + BUG_ON(!pte_none(*(kmap_pte-idx))); set_pte(kmap_pte-idx, mk_pte(page, prot)); arch_flush_lazy_mmu_mode(); - return (void*) vaddr; + return (void *)vaddr; } void *kmap_atomic(struct page *page, enum km_type type) diff --git a/arch/i386/pci/irq.c b/arch/i386/pci/irq.c index 665db06..8434f23 100644 --- a/arch/i386/pci/irq.c +++ b/arch/i386/pci/irq.c @@ -550,6 +550,7 @@ static __init int intel_router_probe(struct irq_router *r, struct pci_dev *route case PCI_DEVICE_ID_INTEL_ICH9_3: case PCI_DEVICE_ID_INTEL_ICH9_4: case PCI_DEVICE_ID_INTEL_ICH9_5: + case PCI_DEVICE_ID_INTEL_TOLAPAI_0: r->name = "PIIX/ICH"; r->get = pirq_piix_get; r->set = pirq_piix_set; diff --git a/arch/i386/pci/pcbios.c b/arch/i386/pci/pcbios.c index 5f51934..10ac8c3 100644 --- a/arch/i386/pci/pcbios.c +++ b/arch/i386/pci/pcbios.c @@ -412,7 +412,7 @@ struct irq_routing_options { u16 segment; } __attribute__((packed)); -struct irq_routing_table * __devinit pcibios_get_irq_routing_table(void) +struct irq_routing_table * pcibios_get_irq_routing_table(void) { struct irq_routing_options opt; struct irq_routing_table *rt = NULL; diff --git a/arch/m68k/kernel/entry.S b/arch/m68k/kernel/entry.S index e162ee68..918f5db 100644 --- a/arch/m68k/kernel/entry.S +++ b/arch/m68k/kernel/entry.S @@ -735,4 +735,14 @@ sys_call_table: .long sys_tee .long sys_vmsplice .long sys_move_pages /* 310 */ + .long sys_sched_setaffinity + .long sys_sched_getaffinity + .long sys_kexec_load + .long sys_getcpu + .long sys_epoll_pwait /* 315 */ + .long sys_utimensat + .long sys_signalfd + .long sys_timerfd + .long sys_eventfd + .long sys_fallocate /* 320 */ diff --git a/arch/m68knommu/kernel/syscalltable.S b/arch/m68knommu/kernel/syscalltable.S index 4603f4f..9620093 100644 --- a/arch/m68knommu/kernel/syscalltable.S +++ b/arch/m68knommu/kernel/syscalltable.S @@ -329,6 +329,16 @@ ENTRY(sys_call_table) .long sys_tee .long sys_vmsplice .long sys_move_pages /* 310 */ + .long sys_sched_setaffinity + .long sys_sched_getaffinity + .long sys_kexec_load + .long sys_getcpu + .long sys_epoll_pwait /* 315 */ + .long sys_utimensat + .long sys_signalfd + .long sys_timerfd + .long sys_eventfd + .long sys_fallocate /* 320 */ .rept NR_syscalls-(.-sys_call_table)/4 .long sys_ni_syscall diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c index 06448a9..3e00416 100644 --- a/arch/mips/kernel/cpu-probe.c +++ b/arch/mips/kernel/cpu-probe.c @@ -199,7 +199,14 @@ static inline void check_wait(void) if ((c->processor_id & 0xff) <= 0x64) break; - cpu_wait = r4k_wait; + /* + * Another rev is incremeting c0_count at a reduced clock + * rate while in WAIT mode. So we basically have the choice + * between using the cp0 timer as clocksource or avoiding + * the WAIT instruction. Until more details are known, + * disable the use of WAIT for 20Kc entirely. + cpu_wait = r4k_wait; + */ break; case CPU_RM9000: if ((c->processor_id & 0x00ff) >= 0x40) diff --git a/arch/mips/kernel/rtlx.c b/arch/mips/kernel/rtlx.c index aab89e9..1ba00c1 100644 --- a/arch/mips/kernel/rtlx.c +++ b/arch/mips/kernel/rtlx.c @@ -56,8 +56,6 @@ static struct chan_waitqueues { struct mutex mutex; } channel_wqs[RTLX_CHANNELS]; -static struct irqaction irq; -static int irq_num; static struct vpe_notifications notify; static int sp_stopping = 0; @@ -111,7 +109,7 @@ static void __used dump_rtlx(void) static int rtlx_init(struct rtlx_info *rtlxi) { if (rtlxi->id != RTLX_ID) { - printk(KERN_ERR "no valid RTLX id at 0x%p 0x%x\n", rtlxi, rtlxi->id); + printk(KERN_ERR "no valid RTLX id at 0x%p 0x%lx\n", rtlxi, rtlxi->id); return -ENOEXEC; } diff --git a/arch/mips/kernel/scall64-n32.S b/arch/mips/kernel/scall64-n32.S index 53d7a97..118be24 100644 --- a/arch/mips/kernel/scall64-n32.S +++ b/arch/mips/kernel/scall64-n32.S @@ -375,7 +375,7 @@ EXPORT(sysn32_call_table) PTR sys_mkdirat PTR sys_mknodat PTR sys_fchownat - PTR sys_futimesat /* 6255 */ + PTR compat_sys_futimesat /* 6255 */ PTR sys_newfstatat PTR sys_unlinkat PTR sys_renameat diff --git a/arch/mips/kernel/time.c b/arch/mips/kernel/time.c index d48d1d5..9a5596b 100644 --- a/arch/mips/kernel/time.c +++ b/arch/mips/kernel/time.c @@ -16,6 +16,7 @@ #include <linux/init.h> #include <linux/sched.h> #include <linux/param.h> +#include <linux/profile.h> #include <linux/time.h> #include <linux/timex.h> #include <linux/smp.h> diff --git a/arch/mips/mips-boards/malta/malta_int.c b/arch/mips/mips-boards/malta/malta_int.c index 97aeb8c..b73f218 100644 --- a/arch/mips/mips-boards/malta/malta_int.c +++ b/arch/mips/mips-boards/malta/malta_int.c @@ -256,7 +256,7 @@ asmlinkage void plat_irq_dispatch(void) if (irq == MIPSCPU_INT_I8259A) malta_hw0_irqdispatch(); - else if (irq > 0) + else if (irq >= 0) do_IRQ(MIPS_CPU_IRQ_BASE + irq); else spurious_interrupt(); diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c index 09d9150..5240432 100644 --- a/arch/mips/mm/init.c +++ b/arch/mips/mm/init.c @@ -8,6 +8,7 @@ * Kevin D. Kissell, kevink@mips.com and Carsten Langgaard, carstenl@mips.com * Copyright (C) 2000 MIPS Technologies, Inc. All rights reserved. */ +#include <linux/bug.h> #include <linux/init.h> #include <linux/module.h> #include <linux/signal.h> @@ -132,6 +133,8 @@ void *kmap_coherent(struct page *page, unsigned long addr) pte_t pte; int tlbidx; + BUG_ON(Page_dcache_dirty(page)); + inc_preempt_count(); idx = (addr >> PAGE_SHIFT) & (FIX_N_COLOURS - 1); #ifdef CONFIG_MIPS_MT_SMTC @@ -208,7 +211,7 @@ void copy_user_highpage(struct page *to, struct page *from, void *vfrom, *vto; vto = kmap_atomic(to, KM_USER1); - if (cpu_has_dc_aliases) { + if (cpu_has_dc_aliases && !Page_dcache_dirty(from)) { vfrom = kmap_coherent(from, vaddr); copy_page(vto, vfrom); kunmap_coherent(); diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c index 9cb3964..6c425b0 100644 --- a/arch/mips/mm/tlbex.c +++ b/arch/mips/mm/tlbex.c @@ -58,6 +58,21 @@ static __init int __maybe_unused r10000_llsc_war(void) } /* + * Found by experiment: At least some revisions of the 4kc throw under + * some circumstances a machine check exception, triggered by invalid + * values in the index register. Delaying the tlbp instruction until + * after the next branch, plus adding an additional nop in front of + * tlbwi/tlbwr avoids the invalid index register values. Nobody knows + * why; it's not an issue caused by the core RTL. + * + */ +static __init int __attribute__((unused)) m4kc_tlbp_war(void) +{ + return (current_cpu_data.processor_id & 0xffff00) == + (PRID_COMP_MIPS | PRID_IMP_4KC); +} + +/* * A little micro-assembler, intended for TLB refill handler * synthesizing. It is intentionally kept simple, does only support * a subset of instructions, and does not try to hide pipeline effects @@ -894,6 +909,8 @@ static __init void build_tlb_write_entry(u32 **p, struct label **l, case CPU_20KC: case CPU_25KF: case CPU_LOONGSON2: + if (m4kc_tlbp_war()) + i_nop(p); tlbw(p); break; @@ -1705,7 +1722,8 @@ build_r4000_tlbchange_handler_head(u32 **p, struct label **l, l_smp_pgtable_change(l, *p); # endif iPTE_LW(p, l, pte, ptr); /* get even pte */ - build_tlb_probe_entry(p); + if (!m4kc_tlbp_war()) + build_tlb_probe_entry(p); } static void __init @@ -1747,6 +1765,8 @@ static void __init build_r4000_tlb_load_handler(void) build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); build_pte_present(&p, &l, &r, K0, K1, label_nopage_tlbl); + if (m4kc_tlbp_war()) + build_tlb_probe_entry(&p); build_make_valid(&p, &r, K0, K1); build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); @@ -1781,6 +1801,8 @@ static void __init build_r4000_tlb_store_handler(void) build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); build_pte_writable(&p, &l, &r, K0, K1, label_nopage_tlbs); + if (m4kc_tlbp_war()) + build_tlb_probe_entry(&p); build_make_write(&p, &r, K0, K1); build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); @@ -1815,6 +1837,8 @@ static void __init build_r4000_tlb_modify_handler(void) build_r4000_tlbchange_handler_head(&p, &l, &r, K0, K1); build_pte_modifiable(&p, &l, &r, K0, K1, label_nopage_tlbm); + if (m4kc_tlbp_war()) + build_tlb_probe_entry(&p); /* Present and writable bits set, set accessed and dirty bits. */ build_make_write(&p, &r, K0, K1); build_r4000_tlbchange_handler_tail(&p, &l, &r, K0, K1); diff --git a/arch/mips/sgi-ip22/ip22-platform.c b/arch/mips/sgi-ip22/ip22-platform.c index 78b608d2..28ffec8 100644 --- a/arch/mips/sgi-ip22/ip22-platform.c +++ b/arch/mips/sgi-ip22/ip22-platform.c @@ -150,8 +150,8 @@ static int __init sgiseeq_devinit(void) return res; /* Second HPC is missing? */ - if (ip22_is_fullhouse() || - !get_dbe(tmp, (unsigned int *)&hpc3c1->pbdma[1])) + if (!ip22_is_fullhouse() || + get_dbe(tmp, (unsigned int *)&hpc3c1->pbdma[1])) return 0; sgimc->giopar |= SGIMC_GIOPAR_MASTEREXP1 | SGIMC_GIOPAR_EXP164 | diff --git a/arch/mips/sgi-ip22/ip22-time.c b/arch/mips/sgi-ip22/ip22-time.c index 8e88a44..de3d018 100644 --- a/arch/mips/sgi-ip22/ip22-time.c +++ b/arch/mips/sgi-ip22/ip22-time.c @@ -114,8 +114,8 @@ static unsigned long dosample(void) } while (msb); /* Stop the counter. */ - writeb(sgint->tcword, (SGINT_TCWORD_CNT2 | SGINT_TCWORD_CALL | - SGINT_TCWORD_MSWST)); + writeb(SGINT_TCWORD_CNT2 | SGINT_TCWORD_CALL | SGINT_TCWORD_MSWST, + &sgint->tcword); /* * Return the difference, this is how far the r4k counter increments * for every 1/HZ seconds. We round off the nearest 1 MHz of master diff --git a/arch/powerpc/kernel/legacy_serial.c b/arch/powerpc/kernel/legacy_serial.c index cea8045..90fa11c 100644 --- a/arch/powerpc/kernel/legacy_serial.c +++ b/arch/powerpc/kernel/legacy_serial.c @@ -493,7 +493,7 @@ static int __init serial_dev_init(void) return platform_device_register(&serial_device); } -arch_initcall(serial_dev_init); +device_initcall(serial_dev_init); /* diff --git a/arch/powerpc/platforms/85xx/mpc8544_ds.c b/arch/powerpc/platforms/85xx/mpc8544_ds.c index 0f834d8..48983bc 100644 --- a/arch/powerpc/platforms/85xx/mpc8544_ds.c +++ b/arch/powerpc/platforms/85xx/mpc8544_ds.c @@ -178,7 +178,9 @@ define_machine(mpc8544_ds) { .probe = mpc8544_ds_probe, .setup_arch = mpc8544_ds_setup_arch, .init_IRQ = mpc8544_ds_pic_init, +#ifdef CONFIG_PCI .pcibios_fixup_bus = fsl_pcibios_fixup_bus, +#endif .get_irq = mpic_get_irq, .restart = mpc85xx_restart, .calibrate_decr = generic_calibrate_decr, diff --git a/arch/powerpc/platforms/85xx/mpc85xx_cds.c b/arch/powerpc/platforms/85xx/mpc85xx_cds.c index 6a171e9..2d4cb78 100644 --- a/arch/powerpc/platforms/85xx/mpc85xx_cds.c +++ b/arch/powerpc/platforms/85xx/mpc85xx_cds.c @@ -351,10 +351,10 @@ define_machine(mpc85xx_cds) { .get_irq = mpic_get_irq, #ifdef CONFIG_PCI .restart = mpc85xx_cds_restart, + .pcibios_fixup_bus = fsl_pcibios_fixup_bus, #else .restart = mpc85xx_restart, #endif .calibrate_decr = generic_calibrate_decr, .progress = udbg_progress, - .pcibios_fixup_bus = fsl_pcibios_fixup_bus, }; diff --git a/arch/powerpc/platforms/85xx/mpc85xx_mds.c b/arch/powerpc/platforms/85xx/mpc85xx_mds.c index be25ecd..7ca7e67 100644 --- a/arch/powerpc/platforms/85xx/mpc85xx_mds.c +++ b/arch/powerpc/platforms/85xx/mpc85xx_mds.c @@ -207,5 +207,7 @@ define_machine(mpc85xx_mds) { .restart = mpc85xx_restart, .calibrate_decr = generic_calibrate_decr, .progress = udbg_progress, +#ifdef CONFIG_PCI .pcibios_fixup_bus = fsl_pcibios_fixup_bus, +#endif }; diff --git a/arch/powerpc/platforms/86xx/mpc86xx_hpcn.c b/arch/powerpc/platforms/86xx/mpc86xx_hpcn.c index 56b27ca..47aafa7 100644 --- a/arch/powerpc/platforms/86xx/mpc86xx_hpcn.c +++ b/arch/powerpc/platforms/86xx/mpc86xx_hpcn.c @@ -250,5 +250,7 @@ define_machine(mpc86xx_hpcn) { .time_init = mpc86xx_time_init, .calibrate_decr = generic_calibrate_decr, .progress = udbg_progress, +#ifdef CONFIG_PCI .pcibios_fixup_bus = fsl_pcibios_fixup_bus, +#endif }; diff --git a/arch/powerpc/platforms/Kconfig b/arch/powerpc/platforms/Kconfig index cfc2497..19d4628 100644 --- a/arch/powerpc/platforms/Kconfig +++ b/arch/powerpc/platforms/Kconfig @@ -285,6 +285,7 @@ config AXON_RAM config FSL_ULI1575 bool default n + select GENERIC_ISA_DMA help Supports for the ULI1575 PCIe south bridge that exists on some Freescale reference boards. The boards all use the ULI in pretty diff --git a/arch/powerpc/sysdev/cpm2_common.c b/arch/powerpc/sysdev/cpm2_common.c index 9244129..c827715 100644 --- a/arch/powerpc/sysdev/cpm2_common.c +++ b/arch/powerpc/sysdev/cpm2_common.c @@ -102,7 +102,7 @@ cpm_setbrg(uint brg, uint rate) brg -= 4; } bp += brg; - *bp = ((BRG_UART_CLK / rate) << 1) | CPM_BRG_EN; + out_be32(bp, (((BRG_UART_CLK / rate) - 1) << 1) | CPM_BRG_EN); cpm2_unmap(bp); } diff --git a/arch/ppc/kernel/head_8xx.S b/arch/ppc/kernel/head_8xx.S index 944c35c..eb8d26f 100644 --- a/arch/ppc/kernel/head_8xx.S +++ b/arch/ppc/kernel/head_8xx.S @@ -495,9 +495,7 @@ LoadLargeDTLB: lwz r11, 4(r0) lwz r12, 16(r0) -#ifdef CONFIG_8xx_CPU6 lwz r3, 8(r0) -#endif rfi /* This is the data TLB error on the MPC8xx. This could be due to diff --git a/arch/sparc64/kernel/pci.c b/arch/sparc64/kernel/pci.c index 139b4cf..e8dac81 100644 --- a/arch/sparc64/kernel/pci.c +++ b/arch/sparc64/kernel/pci.c @@ -744,7 +744,7 @@ static void __devinit pci_of_scan_bus(struct pci_pbm_info *pbm, { struct device_node *child; const u32 *reg; - int reglen, devfn; + int reglen, devfn, prev_devfn; struct pci_dev *dev; if (ofpci_verbose) @@ -752,14 +752,25 @@ static void __devinit pci_of_scan_bus(struct pci_pbm_info *pbm, node->full_name, bus->number); child = NULL; + prev_devfn = -1; while ((child = of_get_next_child(node, child)) != NULL) { if (ofpci_verbose) printk(" * %s\n", child->full_name); reg = of_get_property(child, "reg", ®len); if (reg == NULL || reglen < 20) continue; + devfn = (reg[0] >> 8) & 0xff; + /* This is a workaround for some device trees + * which list PCI devices twice. On the V100 + * for example, device number 3 is listed twice. + * Once as "pm" and once again as "lomp". + */ + if (devfn == prev_devfn) + continue; + prev_devfn = devfn; + /* create a new pci_dev for this device */ dev = of_create_pci_dev(pbm, child, bus, devfn, 0); if (!dev) diff --git a/arch/x86_64/vdso/vclock_gettime.c b/arch/x86_64/vdso/vclock_gettime.c index 17f6a00..5b54cdf 100644 --- a/arch/x86_64/vdso/vclock_gettime.c +++ b/arch/x86_64/vdso/vclock_gettime.c @@ -34,10 +34,11 @@ static long vdso_fallback_gettime(long clock, struct timespec *ts) static inline long vgetns(void) { + long v; cycles_t (*vread)(void); vread = gtod->clock.vread; - return ((vread() - gtod->clock.cycle_last) * gtod->clock.mult) >> - gtod->clock.shift; + v = (vread() - gtod->clock.cycle_last) & gtod->clock.mask; + return (v * gtod->clock.mult) >> gtod->clock.shift; } static noinline int do_realtime(struct timespec *ts) diff --git a/arch/xtensa/Makefile b/arch/xtensa/Makefile index 95f836d..acf05be 100644 --- a/arch/xtensa/Makefile +++ b/arch/xtensa/Makefile @@ -27,7 +27,12 @@ platform-$(CONFIG_XTENSA_PLATFORM_ISS) := iss PLATFORM = $(platform-y) export PLATFORM -CFLAGS += -pipe -mlongcalls +# temporarily until string.h is fixed +cflags-y += -ffreestanding + +cflags-y += -pipe -mlongcalls + +CFLAGS += $(cflags-y) KBUILD_DEFCONFIG := iss_defconfig diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile index 71f733c..f582d6a 100644 --- a/arch/xtensa/kernel/Makefile +++ b/arch/xtensa/kernel/Makefile @@ -7,7 +7,7 @@ extra-y := head.o vmlinux.lds obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o semaphore.o \ setup.o signal.o syscall.o time.o traps.o vectors.o platform.o \ - pci-dma.o + pci-dma.o init_task.o io.o ## windowspill.o diff --git a/arch/xtensa/kernel/asm-offsets.c b/arch/xtensa/kernel/asm-offsets.c index d0323cd..d5ffe7b 100644 --- a/arch/xtensa/kernel/asm-offsets.c +++ b/arch/xtensa/kernel/asm-offsets.c @@ -18,12 +18,13 @@ #include <linux/stddef.h> #include <linux/thread_info.h> #include <linux/ptrace.h> +#include <linux/mm.h> + #include <asm/ptrace.h> #include <asm/processor.h> #include <asm/uaccess.h> #define DEFINE(sym, val) asm volatile("\n->" #sym " %0 " #val : : "i" (val)) -#define BLANK() asm volatile("\n->" : : ) int main(void) { @@ -63,7 +64,6 @@ int main(void) DEFINE(PT_SIZE, sizeof(struct pt_regs)); DEFINE(PT_AREG_END, offsetof (struct pt_regs, areg[XCHAL_NUM_AREGS])); DEFINE(PT_USER_SIZE, offsetof(struct pt_regs, areg[XCHAL_NUM_AREGS])); - BLANK(); /* struct task_struct */ DEFINE(TASK_PTRACE, offsetof (struct task_struct, ptrace)); @@ -73,27 +73,26 @@ int main(void) DEFINE(TASK_THREAD, offsetof (struct task_struct, thread)); DEFINE(TASK_THREAD_INFO, offsetof (struct task_struct, stack)); DEFINE(TASK_STRUCT_SIZE, sizeof (struct task_struct)); - BLANK(); /* struct thread_info (offset from start_struct) */ DEFINE(THREAD_RA, offsetof (struct task_struct, thread.ra)); DEFINE(THREAD_SP, offsetof (struct task_struct, thread.sp)); DEFINE(THREAD_CP_SAVE, offsetof (struct task_struct, thread.cp_save)); DEFINE(THREAD_CURRENT_DS, offsetof (struct task_struct, thread.current_ds)); - BLANK(); /* struct mm_struct */ DEFINE(MM_USERS, offsetof(struct mm_struct, mm_users)); DEFINE(MM_PGD, offsetof (struct mm_struct, pgd)); DEFINE(MM_CONTEXT, offsetof (struct mm_struct, context)); - BLANK(); - DEFINE(PT_SINGLESTEP_BIT, PT_SINGLESTEP_BIT); + + /* struct page */ + DEFINE(PAGE_FLAGS, offsetof(struct page, flags)); /* constants */ DEFINE(_CLONE_VM, CLONE_VM); DEFINE(_CLONE_UNTRACED, CLONE_UNTRACED); + DEFINE(PG_ARCH_1, PG_arch_1); return 0; } - diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S index 8dc7a2c..91a689e 100644 --- a/arch/xtensa/kernel/entry.S +++ b/arch/xtensa/kernel/entry.S @@ -7,7 +7,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2004-2005 by Tensilica Inc. + * Copyright (C) 2004-2007 by Tensilica Inc. * * Chris Zankel <chris@zankel.net> * @@ -169,7 +169,7 @@ _user_exception: * We have to save all registers up to the first '1' from * the right, except the current frame (bit 0). * Assume a2 is: 001001000110001 - * All regiser frames starting from the top fiel to the marked '1' + * All register frames starting from the top field to the marked '1' * must be saved. */ @@ -1572,10 +1572,12 @@ ENTRY(fast_second_level_miss) l32i a0, a1, TASK_MM # tsk->mm beqz a0, 9f -8: rsr a1, EXCVADDR # fault address - _PGD_OFFSET(a0, a1, a1) + + /* We deliberately destroy a3 that holds the exception table. */ + +8: rsr a3, EXCVADDR # fault address + _PGD_OFFSET(a0, a3, a1) l32i a0, a0, 0 # read pmdval - //beqi a0, _PAGE_USER, 2f beqz a0, 2f /* Read ptevaddr and convert to top of page-table page. @@ -1588,7 +1590,7 @@ ENTRY(fast_second_level_miss) * The messy computation for 'pteval' above really simplifies * into the following: * - * pteval = ((pmdval - PAGE_OFFSET) & PAGE_MASK) | PAGE_KERNEL + * pteval = ((pmdval - PAGE_OFFSET) & PAGE_MASK) | PAGE_DIRECTORY */ movi a1, -PAGE_OFFSET @@ -1596,20 +1598,34 @@ ENTRY(fast_second_level_miss) extui a1, a0, 0, PAGE_SHIFT # ... & PAGE_MASK xor a0, a0, a1 - - movi a1, PAGE_DIRECTORY + movi a1, _PAGE_DIRECTORY or a0, a0, a1 # ... | PAGE_DIRECTORY + /* + * We utilize all three wired-ways (7-9) to hold pmd translations. + * Memory regions are mapped to the DTLBs according to bits 28 and 29. + * This allows to map the three most common regions to three different + * DTLBs: + * 0,1 -> way 7 program (0040.0000) and virtual (c000.0000) + * 2 -> way 8 shared libaries (2000.0000) + * 3 -> way 0 stack (3000.0000) + */ + + extui a3, a3, 28, 2 # addr. bit 28 and 29 0,1,2,3 rsr a1, PTEVADDR + addx2 a3, a3, a3 # -> 0,3,6,9 srli a1, a1, PAGE_SHIFT + extui a3, a3, 2, 2 # -> 0,0,1,2 slli a1, a1, PAGE_SHIFT # ptevaddr & PAGE_MASK - addi a1, a1, DTLB_WAY_PGD # ... + way_number + addi a3, a3, DTLB_WAY_PGD + add a1, a1, a3 # ... + way_number - wdtlb a0, a1 +3: wdtlb a0, a1 dsync /* Exit critical section. */ +4: movi a3, exc_table # restore a3 movi a0, 0 s32i a0, a3, EXC_TABLE_FIXUP @@ -1636,8 +1652,76 @@ ENTRY(fast_second_level_miss) 9: l32i a0, a1, TASK_ACTIVE_MM # unlikely case mm == 0 j 8b +#if (DCACHE_WAY_SIZE > PAGE_SIZE) + +2: /* Special case for cache aliasing. + * We (should) only get here if a clear_user_page, copy_user_page + * or the aliased cache flush functions got preemptively interrupted + * by another task. Re-establish temporary mapping to the + * TLBTEMP_BASE areas. + */ + + /* We shouldn't be in a double exception */ + + l32i a0, a2, PT_DEPC + bgeui a0, VALID_DOUBLE_EXCEPTION_ADDRESS, 2f + + /* Make sure the exception originated in the special functions */ + + movi a0, __tlbtemp_mapping_start + rsr a3, EPC_1 + bltu a3, a0, 2f + movi a0, __tlbtemp_mapping_end + bgeu a3, a0, 2f + + /* Check if excvaddr was in one of the TLBTEMP_BASE areas. */ + + movi a3, TLBTEMP_BASE_1 + rsr a0, EXCVADDR + bltu a0, a3, 2f + + addi a1, a0, -(2 << (DCACHE_ALIAS_ORDER + PAGE_SHIFT)) + bgeu a1, a3, 2f + + /* Check if we have to restore an ITLB mapping. */ + + movi a1, __tlbtemp_mapping_itlb + rsr a3, EPC_1 + sub a3, a3, a1 + + /* Calculate VPN */ + + movi a1, PAGE_MASK + and a1, a1, a0 + + /* Jump for ITLB entry */ + + bgez a3, 1f + + /* We can use up to two TLBTEMP areas, one for src and one for dst. */ + + extui a3, a0, PAGE_SHIFT + DCACHE_ALIAS_ORDER, 1 + add a1, a3, a1 + + /* PPN is in a6 for the first TLBTEMP area and in a7 for the second. */ + + mov a0, a6 + movnez a0, a7, a3 + j 3b + + /* ITLB entry. We only use dst in a6. */ + +1: witlb a6, a1 + isync + j 4b + + +#endif // DCACHE_WAY_SIZE > PAGE_SIZE + + 2: /* Invalid PGD, default exception handling */ + movi a3, exc_table rsr a1, DEPC xsr a3, EXCSAVE_1 s32i a1, a2, PT_AREG2 @@ -1682,15 +1766,15 @@ ENTRY(fast_store_prohibited) 8: rsr a1, EXCVADDR # fault address _PGD_OFFSET(a0, a1, a4) l32i a0, a0, 0 - //beqi a0, _PAGE_USER, 2f # FIXME use _PAGE_INVALID beqz a0, 2f + /* Note that we assume _PAGE_WRITABLE_BIT is only set if pte is valid.*/ + _PTE_OFFSET(a0, a1, a4) l32i a4, a0, 0 # read pteval - movi a1, _PAGE_VALID | _PAGE_RW - bnall a4, a1, 2f + bbci.l a4, _PAGE_WRITABLE_BIT, 2f - movi a1, _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_WRENABLE + movi a1, _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HW_WRITE or a4, a4, a1 rsr a1, EXCVADDR s32i a4, a0, 0 @@ -1700,10 +1784,7 @@ ENTRY(fast_store_prohibited) dhwb a0, 0 #endif pdtlb a0, a1 - beqz a0, 1f - idtlb a0 // FIXME do we need this? wdtlb a4, a0 -1: /* Exit critical section. */ diff --git a/arch/xtensa/kernel/init_task.c b/arch/xtensa/kernel/init_task.c new file mode 100644 index 0000000..021b4f4 --- /dev/null +++ b/arch/xtensa/kernel/init_task.c @@ -0,0 +1,38 @@ +/* + * arch/xtensa/kernel/init_task.c + * + * Xtensa Processor version. + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2007 Tensilica Inc. + * + * Chris Zankel <chris@zankel.net> + */ + +#include <linux/mm.h> +#include <linux/fs.h> +#include <linux/init.h> +#include <linux/init_task.h> +#include <linux/module.h> +#include <linux/mqueue.h> + +#include <asm/uaccess.h> + +static struct fs_struct init_fs = INIT_FS; +static struct files_struct init_files = INIT_FILES; +static struct signal_struct init_signals = INIT_SIGNALS(init_signals); +static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand); +struct mm_struct init_mm = INIT_MM(init_mm); + +EXPORT_SYMBOL(init_mm); + +union thread_union init_thread_union + __attribute__((__section__(".data.init_task"))) = +{ INIT_THREAD_INFO(init_task) }; + +struct task_struct init_task = INIT_TASK(init_task); + +EXPORT_SYMBOL(init_task); diff --git a/arch/xtensa/kernel/io.c b/arch/xtensa/kernel/io.c new file mode 100644 index 0000000..5b65269 --- /dev/null +++ b/arch/xtensa/kernel/io.c @@ -0,0 +1,75 @@ +/* + * arch/xtensa/io.c + * + * IO primitives + * + * 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. + * + * Copied from sparc. + * + * Chris Zankel <chris@zankel.net> + * + */ + +#include <asm/io.h> +#include <asm/byteorder.h> + +void outsb(unsigned long addr, const void *src, unsigned long count) { + while (count) { + count -= 1; + writeb(*(const char *)src, addr); + src += 1; + addr += 1; + } +} + +void outsw(unsigned long addr, const void *src, unsigned long count) { + while (count) { + count -= 2; + writew(*(const short *)src, addr); + src += 2; + addr += 2; + } +} + +void outsl(unsigned long addr, const void *src, unsigned long count) { + while (count) { + count -= 4; + writel(*(const long *)src, addr); + src += 4; + addr += 4; + } +} + +void insb(unsigned long addr, void *dst, unsigned long count) { + while (count) { + count -= 1; + *(unsigned char *)dst = readb(addr); + dst += 1; + addr += 1; + } +} + +void insw(unsigned long addr, void *dst, unsigned long count) { + while (count) { + count -= 2; + *(unsigned short *)dst = readw(addr); + dst += 2; + addr += 2; + } +} + +void insl(unsigned long addr, void *dst, unsigned long count) { + while (count) { + count -= 4; + /* + * XXX I am sure we are in for an unaligned trap here. + */ + *(unsigned long *)dst = readl(addr); + dst += 4; + addr += 4; + } +} diff --git a/arch/xtensa/kernel/module.c b/arch/xtensa/kernel/module.c index 2ea1755..ddf14dc 100644 --- a/arch/xtensa/kernel/module.c +++ b/arch/xtensa/kernel/module.c @@ -7,7 +7,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2001 - 2005 Tensilica Inc. + * Copyright (C) 2001 - 2006 Tensilica Inc. * * Chris Zankel <chris@zankel.net> * @@ -22,57 +22,216 @@ #include <linux/kernel.h> #include <linux/cache.h> -LIST_HEAD(module_buf_list); +#undef DEBUG_RELOCATE void *module_alloc(unsigned long size) { - panic("module_alloc not implemented"); + if (size == 0) + return NULL; + return vmalloc(size); } void module_free(struct module *mod, void *module_region) { - panic("module_free not implemented"); + vfree(module_region); + /* FIXME: If module_region == mod->init_region, trim exception + table entries. */ } int module_frob_arch_sections(Elf32_Ehdr *hdr, Elf32_Shdr *sechdrs, char *secstrings, - struct module *me) + struct module *mod) { - panic("module_frob_arch_sections not implemented"); + return 0; +} + +static int +decode_calln_opcode (unsigned char *location) +{ +#ifdef __XTENSA_EB__ + return (location[0] & 0xf0) == 0x50; +#endif +#ifdef __XTENSA_EL__ + return (location[0] & 0xf) == 0x5; +#endif +} + +static int +decode_l32r_opcode (unsigned char *location) +{ +#ifdef __XTENSA_EB__ + return (location[0] & 0xf0) == 0x10; +#endif +#ifdef __XTENSA_EL__ + return (location[0] & 0xf) == 0x1; +#endif } int apply_relocate(Elf32_Shdr *sechdrs, const char *strtab, unsigned int symindex, unsigned int relsec, - struct module *module) + struct module *mod) { - panic ("apply_relocate not implemented"); + printk(KERN_ERR "module %s: REL RELOCATION unsupported\n", + mod->name); + return -ENOEXEC; + } int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab, unsigned int symindex, unsigned int relsec, - struct module *module) + struct module *mod) { - panic("apply_relocate_add not implemented"); + unsigned int i; + Elf32_Rela *rela = (void *)sechdrs[relsec].sh_addr; + Elf32_Sym *sym; + unsigned char *location; + uint32_t value; + +#ifdef DEBUG_RELOCATE + printk("Applying relocate section %u to %u\n", relsec, + sechdrs[relsec].sh_info); +#endif + for (i = 0; i < sechdrs[relsec].sh_size / sizeof(*rela); i++) { + location = (char *)sechdrs[sechdrs[relsec].sh_info].sh_addr + + rela[i].r_offset; + sym = (Elf32_Sym *)sechdrs[symindex].sh_addr + + ELF32_R_SYM(rela[i].r_info); + value = sym->st_value + rela[i].r_addend; + + switch (ELF32_R_TYPE(rela[i].r_info)) { + case R_XTENSA_NONE: + case R_XTENSA_DIFF8: + case R_XTENSA_DIFF16: + case R_XTENSA_DIFF32: + case R_XTENSA_ASM_EXPAND: + break; + + case R_XTENSA_32: + case R_XTENSA_PLT: + *(uint32_t *)location += value; + break; + + case R_XTENSA_SLOT0_OP: + if (decode_calln_opcode(location)) { + value -= ((unsigned long)location & -4) + 4; + if ((value & 3) != 0 || + ((value + (1 << 19)) >> 20) != 0) { + printk("%s: relocation out of range, " + "section %d reloc %d " + "sym '%s'\n", + mod->name, relsec, i, + strtab + sym->st_name); + return -ENOEXEC; + } + value = (signed int)value >> 2; +#ifdef __XTENSA_EB__ + location[0] = ((location[0] & ~0x3) | + ((value >> 16) & 0x3)); + location[1] = (value >> 8) & 0xff; + location[2] = value & 0xff; +#endif +#ifdef __XTENSA_EL__ + location[0] = ((location[0] & ~0xc0) | + ((value << 6) & 0xc0)); + location[1] = (value >> 2) & 0xff; + location[2] = (value >> 10) & 0xff; +#endif + } else if (decode_l32r_opcode(location)) { + value -= (((unsigned long)location + 3) & -4); + if ((value & 3) != 0 || + (signed int)value >> 18 != -1) { + printk("%s: relocation out of range, " + "section %d reloc %d " + "sym '%s'\n", + mod->name, relsec, i, + strtab + sym->st_name); + return -ENOEXEC; + } + value = (signed int)value >> 2; + +#ifdef __XTENSA_EB__ + location[1] = (value >> 8) & 0xff; + location[2] = value & 0xff; +#endif +#ifdef __XTENSA_EL__ + location[1] = value & 0xff; + location[2] = (value >> 8) & 0xff; +#endif + } + /* FIXME: Ignore any other opcodes. The Xtensa + assembler currently assumes that the linker will + always do relaxation and so all PC-relative + operands need relocations. (The assembler also + writes out the tentative PC-relative values, + assuming no link-time relaxation, so it is usually + safe to ignore the relocations.) If the + assembler's "--no-link-relax" flag can be made to + work, and if all kernel modules can be assembled + with that flag, then unexpected relocations could + be detected here. */ + break; + + case R_XTENSA_SLOT1_OP: + case R_XTENSA_SLOT2_OP: + case R_XTENSA_SLOT3_OP: + case R_XTENSA_SLOT4_OP: + case R_XTENSA_SLOT5_OP: + case R_XTENSA_SLOT6_OP: + case R_XTENSA_SLOT7_OP: + case R_XTENSA_SLOT8_OP: + case R_XTENSA_SLOT9_OP: + case R_XTENSA_SLOT10_OP: + case R_XTENSA_SLOT11_OP: + case R_XTENSA_SLOT12_OP: + case R_XTENSA_SLOT13_OP: + case R_XTENSA_SLOT14_OP: + printk("%s: unexpected FLIX relocation: %u\n", + mod->name, + ELF32_R_TYPE(rela[i].r_info)); + return -ENOEXEC; + + case R_XTENSA_SLOT0_ALT: + case R_XTENSA_SLOT1_ALT: + case R_XTENSA_SLOT2_ALT: + case R_XTENSA_SLOT3_ALT: + case R_XTENSA_SLOT4_ALT: + case R_XTENSA_SLOT5_ALT: + case R_XTENSA_SLOT6_ALT: + case R_XTENSA_SLOT7_ALT: + case R_XTENSA_SLOT8_ALT: + case R_XTENSA_SLOT9_ALT: + case R_XTENSA_SLOT10_ALT: + case R_XTENSA_SLOT11_ALT: + case R_XTENSA_SLOT12_ALT: + case R_XTENSA_SLOT13_ALT: + case R_XTENSA_SLOT14_ALT: + printk("%s: unexpected ALT relocation: %u\n", + mod->name, + ELF32_R_TYPE(rela[i].r_info)); + return -ENOEXEC; + + default: + printk("%s: unexpected relocation: %u\n", + mod->name, + ELF32_R_TYPE(rela[i].r_info)); + return -ENOEXEC; + } + } + return 0; } int module_finalize(const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs, - struct module *me) + struct module *mod) { - panic ("module_finalize not implemented"); + return 0; } void module_arch_cleanup(struct module *mod) { - panic("module_arch_cleanup not implemented"); -} - -struct bug_entry *module_find_bug(unsigned long bugaddr) -{ - panic("module_find_bug not implemented"); } diff --git a/arch/xtensa/kernel/pci.c b/arch/xtensa/kernel/pci.c index 77deae5..b7c0734 100644 --- a/arch/xtensa/kernel/pci.c +++ b/arch/xtensa/kernel/pci.c @@ -394,72 +394,3 @@ int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma, return ret; } - -/* - * This probably belongs here rather than ioport.c because - * we do not want this crud linked into SBus kernels. - * Also, think for a moment about likes of floppy.c that - * include architecture specific parts. They may want to redefine ins/outs. - * - * We do not use horrible macros here because we want to - * advance pointer by sizeof(size). - */ -void outsb(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 1; - writeb(*(const char *)src, addr); - src += 1; - addr += 1; - } -} - -void outsw(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 2; - writew(*(const short *)src, addr); - src += 2; - addr += 2; - } -} - -void outsl(unsigned long addr, const void *src, unsigned long count) { - while (count) { - count -= 4; - writel(*(const long *)src, addr); - src += 4; - addr += 4; - } -} - -void insb(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 1; - *(unsigned char *)dst = readb(addr); - dst += 1; - addr += 1; - } -} - -void insw(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 2; - *(unsigned short *)dst = readw(addr); - dst += 2; - addr += 2; - } -} - -void insl(unsigned long addr, void *dst, unsigned long count) { - while (count) { - count -= 4; - /* - * XXX I am sure we are in for an unaligned trap here. - */ - *(unsigned long *)dst = readl(addr); - dst += 4; - addr += 4; - } -} - - - diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c index dd498f1..f53d7bd 100644 --- a/arch/xtensa/kernel/process.c +++ b/arch/xtensa/kernel/process.c @@ -46,20 +46,6 @@ extern void ret_from_fork(void); -static struct fs_struct init_fs = INIT_FS; -static struct files_struct init_files = INIT_FILES; -static struct signal_struct init_signals = INIT_SIGNALS(init_signals); -static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand); -struct mm_struct init_mm = INIT_MM(init_mm); -EXPORT_SYMBOL(init_mm); - -union thread_union init_thread_union - __attribute__((__section__(".data.init_task"))) = -{ INIT_THREAD_INFO(init_task) }; - -struct task_struct init_task = INIT_TASK(init_task); -EXPORT_SYMBOL(init_task); - struct task_struct *current_set[NR_CPUS] = {&init_task, }; void (*pm_power_off)(void) = NULL; diff --git a/arch/xtensa/kernel/semaphore.c b/arch/xtensa/kernel/semaphore.c index d40f4b1..995c641 100644 --- a/arch/xtensa/kernel/semaphore.c +++ b/arch/xtensa/kernel/semaphore.c @@ -100,7 +100,7 @@ static __inline__ int waking_non_zero_trylock(struct semaphore *sem) return ret; } -spinlock_t semaphore_wake_lock; +DEFINE_SPINLOCK(semaphore_wake_lock); /* * Semaphores are implemented using a two-way counter: diff --git a/arch/xtensa/kernel/syscall.c b/arch/xtensa/kernel/syscall.c index fe3834b..f3e16ef 100644 --- a/arch/xtensa/kernel/syscall.c +++ b/arch/xtensa/kernel/syscall.c @@ -93,3 +93,8 @@ asmlinkage long xtensa_shmat(int shmid, char __user *shmaddr, int shmflg) return (long)ret; } +asmlinkage long xtensa_fadvise64_64(int fd, int advice, unsigned long long offset, unsigned long long len) +{ + return sys_fadvise64_64(fd, offset, len, advice); +} + diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c index 22949be..60d29fe 100644 --- a/arch/xtensa/kernel/time.c +++ b/arch/xtensa/kernel/time.c @@ -32,12 +32,20 @@ EXPORT_SYMBOL(rtc_lock); #ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT unsigned long ccount_per_jiffy; /* per 1/HZ */ -unsigned long ccount_nsec; /* nsec per ccount increment */ +unsigned long nsec_per_ccount; /* nsec per ccount increment */ #endif -unsigned int last_ccount_stamp; static long last_rtc_update = 0; +/* + * Scheduler clock - returns current tim in nanosec units. + */ + +unsigned long long sched_clock(void) +{ + return (unsigned long long)jiffies * (1000000000 / HZ); +} + static irqreturn_t timer_interrupt(int irq, void *dev_id); static struct irqaction timer_irqaction = { .handler = timer_interrupt, @@ -69,7 +77,6 @@ void __init time_init(void) xtime.tv_nsec = 0; last_rtc_update = xtime.tv_sec = sec_n; - last_ccount_stamp = get_ccount(); set_normalized_timespec(&wall_to_monotonic, -xtime.tv_sec, -xtime.tv_nsec); @@ -85,7 +92,7 @@ int do_settimeofday(struct timespec *tv) { time_t wtm_sec, sec = tv->tv_sec; long wtm_nsec, nsec = tv->tv_nsec; - unsigned long ccount; + unsigned long delta; if ((unsigned long)tv->tv_nsec >= NSEC_PER_SEC) return -EINVAL; @@ -97,8 +104,10 @@ int do_settimeofday(struct timespec *tv) * wall time. Discover what correction gettimeofday() would have * made, and then undo it! */ - ccount = get_ccount(); - nsec -= (ccount - last_ccount_stamp) * CCOUNT_NSEC; + + delta = CCOUNT_PER_JIFFY; + delta += get_ccount() - get_linux_timer(); + nsec -= delta * NSEC_PER_CCOUNT; wtm_sec = wall_to_monotonic.tv_sec + (xtime.tv_sec - sec); wtm_nsec = wall_to_monotonic.tv_nsec + (xtime.tv_nsec - nsec); @@ -117,17 +126,21 @@ EXPORT_SYMBOL(do_settimeofday); void do_gettimeofday(struct timeval *tv) { unsigned long flags; - unsigned long sec, usec, delta, seq; + unsigned long volatile sec, usec, delta, seq; do { seq = read_seqbegin_irqsave(&xtime_lock, flags); - delta = get_ccount() - last_ccount_stamp; sec = xtime.tv_sec; usec = (xtime.tv_nsec / NSEC_PER_USEC); + + delta = get_linux_timer() - get_ccount(); + } while (read_seqretry_irqrestore(&xtime_lock, seq, flags)); - usec += (delta * CCOUNT_NSEC) / NSEC_PER_USEC; + usec += (((unsigned long) CCOUNT_PER_JIFFY - delta) + * (unsigned long) NSEC_PER_CCOUNT) / NSEC_PER_USEC; + for (; usec >= 1000000; sec++, usec -= 1000000) ; @@ -158,9 +171,12 @@ again: write_seqlock(&xtime_lock); - last_ccount_stamp = next; + do_timer(1); /* Linux handler in kernel/timer.c */ + + /* Note that writing CCOMPARE clears the interrupt. */ + next += CCOUNT_PER_JIFFY; - do_timer (1); /* Linux handler in kernel/timer.c */ + set_linux_timer(next); if (ntp_synced() && xtime.tv_sec - last_rtc_update >= 659 && @@ -175,19 +191,15 @@ again: write_sequnlock(&xtime_lock); } - /* NOTE: writing CCOMPAREn clears the interrupt. */ + /* Allow platform to do something useful (Wdog). */ - set_linux_timer (next); + platform_heartbeat(); /* Make sure we didn't miss any tick... */ if ((signed long)(get_ccount() - next) > 0) goto again; - /* Allow platform to do something useful (Wdog). */ - - platform_heartbeat(); - return IRQ_HANDLED; } diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c index c5e62f9..8be99c7 100644 --- a/arch/xtensa/kernel/traps.c +++ b/arch/xtensa/kernel/traps.c @@ -83,7 +83,7 @@ typedef struct { void* handler; } dispatch_init_table_t; -dispatch_init_table_t __init dispatch_init_table[] = { +static dispatch_init_table_t __initdata dispatch_init_table[] = { { EXCCAUSE_ILLEGAL_INSTRUCTION, 0, do_illegal_instruction}, { EXCCAUSE_SYSTEM_CALL, KRNL, fast_syscall_kernel }, @@ -305,7 +305,7 @@ do_debug(struct pt_regs *regs) #define set_handler(idx,handler) (exc_table[idx] = (unsigned long) (handler)) -void trap_init(void) +void __init trap_init(void) { int i; diff --git a/arch/xtensa/mm/Makefile b/arch/xtensa/mm/Makefile index a5aed59..10aec22 100644 --- a/arch/xtensa/mm/Makefile +++ b/arch/xtensa/mm/Makefile @@ -5,9 +5,5 @@ # removes any old dependencies. DON'T put your own dependencies here # unless it's something special (ie not a .c file). # -# Note 2! The CFLAGS definition is now in the main makefile... -obj-y := init.o fault.o tlb.o misc.o -obj-m := -obj-n := -obj- := +obj-y := init.o fault.o tlb.o misc.o cache.o diff --git a/arch/xtensa/mm/cache.c b/arch/xtensa/mm/cache.c new file mode 100644 index 0000000..9a1fa94 --- /dev/null +++ b/arch/xtensa/mm/cache.c @@ -0,0 +1,256 @@ +/* + * arch/xtensa/mm/cache.c + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2001-2006 Tensilica Inc. + * + * Chris Zankel <chris@zankel.net> + * Joe Taylor + * Marc Gauthier + * + */ + +#include <linux/init.h> +#include <linux/signal.h> +#include <linux/sched.h> +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/string.h> +#include <linux/types.h> +#include <linux/ptrace.h> +#include <linux/bootmem.h> +#include <linux/swap.h> +#include <linux/pagemap.h> + +#include <asm/pgtable.h> +#include <asm/bootparam.h> +#include <asm/mmu_context.h> +#include <asm/tlb.h> +#include <asm/tlbflush.h> +#include <asm/page.h> +#include <asm/pgalloc.h> +#include <asm/pgtable.h> + +//#define printd(x...) printk(x) +#define printd(x...) do { } while(0) + +/* + * Note: + * The kernel provides one architecture bit PG_arch_1 in the page flags that + * can be used for cache coherency. + * + * I$-D$ coherency. + * + * The Xtensa architecture doesn't keep the instruction cache coherent with + * the data cache. We use the architecture bit to indicate if the caches + * are coherent. The kernel clears this bit whenever a page is added to the + * page cache. At that time, the caches might not be in sync. We, therefore, + * define this flag as 'clean' if set. + * + * D-cache aliasing. + * + * With cache aliasing, we have to always flush the cache when pages are + * unmapped (see tlb_start_vma(). So, we use this flag to indicate a dirty + * page. + * + * + * + */ + +#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK + +/* + * Any time the kernel writes to a user page cache page, or it is about to + * read from a page cache page this routine is called. + * + */ + +void flush_dcache_page(struct page *page) +{ + struct address_space *mapping = page_mapping(page); + + /* + * If we have a mapping but the page is not mapped to user-space + * yet, we simply mark this page dirty and defer flushing the + * caches until update_mmu(). + */ + + if (mapping && !mapping_mapped(mapping)) { + if (!test_bit(PG_arch_1, &page->flags)) + set_bit(PG_arch_1, &page->flags); + return; + + } else { + + unsigned long phys = page_to_phys(page); + unsigned long temp = page->index << PAGE_SHIFT; + unsigned long alias = !(DCACHE_ALIAS_EQ(temp, phys)); + unsigned long virt; + + /* + * Flush the page in kernel space and user space. + * Note that we can omit that step if aliasing is not + * an issue, but we do have to synchronize I$ and D$ + * if we have a mapping. + */ + + if (!alias && !mapping) + return; + + __flush_invalidate_dcache_page((long)page_address(page)); + + virt = TLBTEMP_BASE_1 + (temp & DCACHE_ALIAS_MASK); + + if (alias) + __flush_invalidate_dcache_page_alias(virt, phys); + + if (mapping) + __invalidate_icache_page_alias(virt, phys); + } + + /* There shouldn't be an entry in the cache for this page anymore. */ +} + + +/* + * For now, flush the whole cache. FIXME?? + */ + +void flush_cache_range(struct vm_area_struct* vma, + unsigned long start, unsigned long end) +{ + __flush_invalidate_dcache_all(); + __invalidate_icache_all(); +} + +/* + * Remove any entry in the cache for this page. + * + * Note that this function is only called for user pages, so use the + * alias versions of the cache flush functions. + */ + +void flush_cache_page(struct vm_area_struct* vma, unsigned long address, + unsigned long pfn) +{ + /* Note that we have to use the 'alias' address to avoid multi-hit */ + + unsigned long phys = page_to_phys(pfn_to_page(pfn)); + unsigned long virt = TLBTEMP_BASE_1 + (address & DCACHE_ALIAS_MASK); + + __flush_invalidate_dcache_page_alias(virt, phys); + __invalidate_icache_page_alias(virt, phys); +} + +#endif + +void +update_mmu_cache(struct vm_area_struct * vma, unsigned long addr, pte_t pte) +{ + unsigned long pfn = pte_pfn(pte); + struct page *page; + + if (!pfn_valid(pfn)) + return; + + page = pfn_to_page(pfn); + + /* Invalidate old entry in TLBs */ + + invalidate_itlb_mapping(addr); + invalidate_dtlb_mapping(addr); + +#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK + + if (!PageReserved(page) && test_bit(PG_arch_1, &page->flags)) { + + unsigned long vaddr = TLBTEMP_BASE_1 + (addr & DCACHE_ALIAS_MASK); + unsigned long paddr = (unsigned long) page_address(page); + unsigned long phys = page_to_phys(page); + + __flush_invalidate_dcache_page(paddr); + + __flush_invalidate_dcache_page_alias(vaddr, phys); + __invalidate_icache_page_alias(vaddr, phys); + + clear_bit(PG_arch_1, &page->flags); + } +#else + if (!PageReserved(page) && !test_bit(PG_arch_1, &page->flags) + && (vma->vm_flags & VM_EXEC) != 0) { + unsigned long vaddr = addr & PAGE_MASK; + __flush_dcache_page(vaddr); + __invalidate_icache_page(vaddr); + set_bit(PG_arch_1, &page->flags); + } +#endif +} + +/* + * access_process_vm() has called get_user_pages(), which has done a + * flush_dcache_page() on the page. + */ + +#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK + +void copy_to_user_page(struct vm_area_struct *vma, struct page *page, + unsigned long vaddr, void *dst, const void *src, + unsigned long len) +{ + unsigned long phys = page_to_phys(page); + unsigned long alias = !(DCACHE_ALIAS_EQ(vaddr, phys)); + + /* Flush and invalidate user page if aliased. */ + + if (alias) { + unsigned long temp = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK); + __flush_invalidate_dcache_page_alias(temp, phys); + } + + /* Copy data */ + + memcpy(dst, src, len); + + /* + * Flush and invalidate kernel page if aliased and synchronize + * data and instruction caches for executable pages. + */ + + if (alias) { + unsigned long temp = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK); + + __flush_invalidate_dcache_range((unsigned long) dst, len); + if ((vma->vm_flags & VM_EXEC) != 0) { + __invalidate_icache_page_alias(temp, phys); + } + + } else if ((vma->vm_flags & VM_EXEC) != 0) { + __flush_dcache_range((unsigned long)dst,len); + __invalidate_icache_range((unsigned long) dst, len); + } +} + +extern void copy_from_user_page(struct vm_area_struct *vma, struct page *page, + unsigned long vaddr, void *dst, const void *src, + unsigned long len) +{ + unsigned long phys = page_to_phys(page); + unsigned long alias = !(DCACHE_ALIAS_EQ(vaddr, phys)); + + /* + * Flush user page if aliased. + * (Note: a simply flush would be sufficient) + */ + + if (alias) { + unsigned long temp = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK); + __flush_invalidate_dcache_page_alias(temp, phys); + } + + memcpy(dst, src, len); +} + +#endif diff --git a/arch/xtensa/mm/fault.c b/arch/xtensa/mm/fault.c index 1600406..45d28f2 100644 --- a/arch/xtensa/mm/fault.c +++ b/arch/xtensa/mm/fault.c @@ -24,6 +24,8 @@ unsigned long asid_cache = ASID_USER_FIRST; void bad_page_fault(struct pt_regs*, unsigned long, int); +#undef DEBUG_PAGE_FAULT + /* * This routine handles page faults. It determines the address, * and the problem, and then passes it off to one of the appropriate @@ -64,7 +66,7 @@ void do_page_fault(struct pt_regs *regs) exccause == EXCCAUSE_ITLB_MISS || exccause == EXCCAUSE_FETCH_CACHE_ATTRIBUTE) ? 1 : 0; -#if 0 +#ifdef DEBUG_PAGE_FAULT printk("[%s:%d:%08x:%d:%08x:%s%s]\n", current->comm, current->pid, address, exccause, regs->pc, is_write? "w":"", is_exec? "x":""); #endif @@ -219,7 +221,7 @@ bad_page_fault(struct pt_regs *regs, unsigned long address, int sig) /* Are we prepared to handle this kernel fault? */ if ((entry = search_exception_tables(regs->pc)) != NULL) { -#if 1 +#ifdef DEBUG_PAGE_FAULT printk(KERN_DEBUG "%s: Exception at pc=%#010lx (%lx)\n", current->comm, regs->pc, entry->fixup); #endif diff --git a/arch/xtensa/mm/init.c b/arch/xtensa/mm/init.c index 8415c76..b3086f3 100644 --- a/arch/xtensa/mm/init.c +++ b/arch/xtensa/mm/init.c @@ -15,40 +15,24 @@ * Kevin Chea */ -#include <linux/init.h> -#include <linux/signal.h> -#include <linux/sched.h> #include <linux/kernel.h> #include <linux/errno.h> -#include <linux/string.h> -#include <linux/types.h> -#include <linux/ptrace.h> #include <linux/bootmem.h> #include <linux/swap.h> +#include <linux/mman.h> +#include <linux/nodemask.h> +#include <linux/mm.h> +#include <linux/slab.h> #include <asm/pgtable.h> #include <asm/bootparam.h> #include <asm/mmu_context.h> #include <asm/tlb.h> -#include <asm/tlbflush.h> #include <asm/page.h> #include <asm/pgalloc.h> -#include <asm/pgtable.h> - -#define DEBUG 0 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); -//static DEFINE_SPINLOCK(tlb_lock); - -/* - * This flag is used to indicate that the page was mapped and modified in - * kernel space, so the cache is probably dirty at that address. - * If cache aliasing is enabled and the page color mismatches, update_mmu_cache - * synchronizes the caches if this bit is set. - */ - -#define PG_cache_clean PG_arch_1 /* References to section boundaries */ @@ -323,228 +307,22 @@ void show_mem(void) printk("%d free pages\n", free); } -/* ------------------------------------------------------------------------- */ - -#if (DCACHE_WAY_SIZE > PAGE_SIZE) - -/* - * With cache aliasing, the page color of the page in kernel space and user - * space might mismatch. We temporarily map the page to a different virtual - * address with the same color and clear the page there. - */ - -void clear_user_page(void *kaddr, unsigned long vaddr, struct page* page) -{ - - /* There shouldn't be any entries for this page. */ - - __flush_invalidate_dcache_page_phys(__pa(page_address(page))); - - if (!PAGE_COLOR_EQ(vaddr, kaddr)) { - unsigned long v, p; - - /* Temporarily map page to DTLB_WAY_DCACHE_ALIAS0. */ - - spin_lock(&tlb_lock); - - p = (unsigned long)pte_val((mk_pte(page,PAGE_KERNEL))); - kaddr = (void*)PAGE_COLOR_MAP0(vaddr); - v = (unsigned long)kaddr | DTLB_WAY_DCACHE_ALIAS0; - __asm__ __volatile__("wdtlb %0,%1; dsync" : :"a" (p), "a" (v)); - - clear_page(kaddr); - - spin_unlock(&tlb_lock); - } else { - clear_page(kaddr); - } - - /* We need to make sure that i$ and d$ are coherent. */ - - clear_bit(PG_cache_clean, &page->flags); -} - -/* - * With cache aliasing, we have to make sure that the page color of the page - * in kernel space matches that of the virtual user address before we read - * the page. If the page color differ, we create a temporary DTLB entry with - * the corrent page color and use this 'temporary' address as the source. - * We then use the same approach as in clear_user_page and copy the data - * to the kernel space and clear the PG_cache_clean bit to synchronize caches - * later. - * - * Note: - * Instead of using another 'way' for the temporary DTLB entry, we could - * probably use the same entry that points to the kernel address (after - * saving the original value and restoring it when we are done). - */ +struct kmem_cache *pgtable_cache __read_mostly; -void copy_user_page(void* to, void* from, unsigned long vaddr, - struct page* to_page) +static void pgd_ctor(void *addr, struct kmem_cache *cache, unsigned long flags) { - /* There shouldn't be any entries for the new page. */ - - __flush_invalidate_dcache_page_phys(__pa(page_address(to_page))); - - spin_lock(&tlb_lock); - - if (!PAGE_COLOR_EQ(vaddr, from)) { - unsigned long v, p, t; - - __asm__ __volatile__ ("pdtlb %1,%2; rdtlb1 %0,%1" - : "=a"(p), "=a"(t) : "a"(from)); - from = (void*)PAGE_COLOR_MAP0(vaddr); - v = (unsigned long)from | DTLB_WAY_DCACHE_ALIAS0; - __asm__ __volatile__ ("wdtlb %0,%1; dsync" ::"a" (p), "a" (v)); - } - - if (!PAGE_COLOR_EQ(vaddr, to)) { - unsigned long v, p; - - p = (unsigned long)pte_val((mk_pte(to_page,PAGE_KERNEL))); - to = (void*)PAGE_COLOR_MAP1(vaddr); - v = (unsigned long)to | DTLB_WAY_DCACHE_ALIAS1; - __asm__ __volatile__ ("wdtlb %0,%1; dsync" ::"a" (p), "a" (v)); - } - copy_page(to, from); - - spin_unlock(&tlb_lock); - - /* We need to make sure that i$ and d$ are coherent. */ - - clear_bit(PG_cache_clean, &to_page->flags); -} - - - -/* - * Any time the kernel writes to a user page cache page, or it is about to - * read from a page cache page this routine is called. - * - * Note: - * The kernel currently only provides one architecture bit in the page - * flags that we use for I$/D$ coherency. Maybe, in future, we can - * use a sepearte bit for deferred dcache aliasing: - * If the page is not mapped yet, we only need to set a flag, - * if mapped, we need to invalidate the page. - */ -// FIXME: we probably need this for WB caches not only for Page Coloring.. - -void flush_dcache_page(struct page *page) -{ - unsigned long addr = __pa(page_address(page)); - struct address_space *mapping = page_mapping(page); - - __flush_invalidate_dcache_page_phys(addr); - - if (!test_bit(PG_cache_clean, &page->flags)) - return; - - /* If this page hasn't been mapped, yet, handle I$/D$ coherency later.*/ -#if 0 - if (mapping && !mapping_mapped(mapping)) - clear_bit(PG_cache_clean, &page->flags); - else -#endif - __invalidate_icache_page_phys(addr); -} - -void flush_cache_range(struct vm_area_struct* vma, unsigned long s, - unsigned long e) -{ - __flush_invalidate_cache_all(); -} - -void flush_cache_page(struct vm_area_struct* vma, unsigned long address, - unsigned long pfn) -{ - struct page *page = pfn_to_page(pfn); - - /* Remove any entry for the old mapping. */ - - if (current->active_mm == vma->vm_mm) { - unsigned long addr = __pa(page_address(page)); - __flush_invalidate_dcache_page_phys(addr); - if ((vma->vm_flags & VM_EXEC) != 0) - __invalidate_icache_page_phys(addr); - } else { - BUG(); - } -} - -#endif /* (DCACHE_WAY_SIZE > PAGE_SIZE) */ - - -pte_t* pte_alloc_one_kernel (struct mm_struct* mm, unsigned long addr) -{ - pte_t* pte = (pte_t*)__get_free_pages(GFP_KERNEL|__GFP_REPEAT, 0); - if (likely(pte)) { - pte_t* ptep = (pte_t*)(pte_val(*pte) + PAGE_OFFSET); - int i; - for (i = 0; i < 1024; i++, ptep++) - pte_clear(mm, addr, ptep); - } - return pte; -} - -struct page* pte_alloc_one(struct mm_struct *mm, unsigned long addr) -{ - struct page *page; - - page = alloc_pages(GFP_KERNEL | __GFP_REPEAT, 0); - - if (likely(page)) { - pte_t* ptep = kmap_atomic(page, KM_USER0); - int i; + pte_t* ptep = (pte_t*)addr; + int i; - for (i = 0; i < 1024; i++, ptep++) - pte_clear(mm, addr, ptep); + for (i = 0; i < 1024; i++, ptep++) + pte_clear(NULL, 0, ptep); - kunmap_atomic(ptep, KM_USER0); - } - return page; } - -/* - * Handle D$/I$ coherency. - * - * Note: - * We only have one architecture bit for the page flags, so we cannot handle - * cache aliasing, yet. - */ - -void -update_mmu_cache(struct vm_area_struct * vma, unsigned long addr, pte_t pte) +void __init pgtable_cache_init(void) { - unsigned long pfn = pte_pfn(pte); - struct page *page; - unsigned long vaddr = addr & PAGE_MASK; - - if (!pfn_valid(pfn)) - return; - - page = pfn_to_page(pfn); - - invalidate_itlb_mapping(addr); - invalidate_dtlb_mapping(addr); - - /* We have a new mapping. Use it. */ - - write_dtlb_entry(pte, dtlb_probe(addr)); - - /* If the processor can execute from this page, synchronize D$/I$. */ - - if ((vma->vm_flags & VM_EXEC) != 0) { - - write_itlb_entry(pte, itlb_probe(addr)); - - /* Synchronize caches, if not clean. */ - - if (!test_and_set_bit(PG_cache_clean, &page->flags)) { - __flush_dcache_page(vaddr); - __invalidate_icache_page(vaddr); - } - } + pgtable_cache = kmem_cache_create("pgd", + PAGE_SIZE, PAGE_SIZE, + SLAB_HWCACHE_ALIGN, + pgd_ctor); } - diff --git a/arch/xtensa/mm/misc.S b/arch/xtensa/mm/misc.S index ae08533..e1f8803 100644 --- a/arch/xtensa/mm/misc.S +++ b/arch/xtensa/mm/misc.S @@ -7,29 +7,33 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 2001 - 2005 Tensilica Inc. + * Copyright (C) 2001 - 2007 Tensilica Inc. * * Chris Zankel <chris@zankel.net> */ -/* Note: we might want to implement some of the loops as zero-overhead-loops, - * where applicable and if supported by the processor. - */ #include <linux/linkage.h> #include <asm/page.h> #include <asm/pgtable.h> #include <asm/asmmacro.h> #include <asm/cacheasm.h> +#include <asm/tlbflush.h> + -/* clear_page (page) */ +/* + * clear_page and clear_user_page are the same for non-cache-aliased configs. + * + * clear_page (unsigned long page) + * a2 + */ ENTRY(clear_page) entry a1, 16 - addi a4, a2, PAGE_SIZE - movi a3, 0 -1: s32i a3, a2, 0 + movi a3, 0 + __loopi a2, a7, PAGE_SIZE, 32 + s32i a3, a2, 0 s32i a3, a2, 4 s32i a3, a2, 8 s32i a3, a2, 12 @@ -37,42 +41,277 @@ ENTRY(clear_page) s32i a3, a2, 20 s32i a3, a2, 24 s32i a3, a2, 28 - addi a2, a2, 32 - blt a2, a4, 1b + __endla a2, a7, 32 retw /* + * copy_page and copy_user_page are the same for non-cache-aliased configs. + * * copy_page (void *to, void *from) - * a2 a3 + * a2 a3 */ ENTRY(copy_page) entry a1, 16 - addi a4, a2, PAGE_SIZE - -1: l32i a5, a3, 0 - l32i a6, a3, 4 - l32i a7, a3, 8 - s32i a5, a2, 0 - s32i a6, a2, 4 - s32i a7, a2, 8 - l32i a5, a3, 12 - l32i a6, a3, 16 - l32i a7, a3, 20 - s32i a5, a2, 12 - s32i a6, a2, 16 - s32i a7, a2, 20 - l32i a5, a3, 24 - l32i a6, a3, 28 - s32i a5, a2, 24 - s32i a6, a2, 28 - addi a2, a2, 32 - addi a3, a3, 32 - blt a2, a4, 1b + __loopi a2, a4, PAGE_SIZE, 32 + + l32i a8, a3, 0 + l32i a9, a3, 4 + s32i a8, a2, 0 + s32i a9, a2, 4 + + l32i a8, a3, 8 + l32i a9, a3, 12 + s32i a8, a2, 8 + s32i a9, a2, 12 + + l32i a8, a3, 16 + l32i a9, a3, 20 + s32i a8, a2, 16 + s32i a9, a2, 20 + + l32i a8, a3, 24 + l32i a9, a3, 28 + s32i a8, a2, 24 + s32i a9, a2, 28 + + addi a2, a2, 32 + addi a3, a3, 32 + + __endl a2, a4 + + retw + +/* + * If we have to deal with cache aliasing, we use temporary memory mappings + * to ensure that the source and destination pages have the same color as + * the virtual address. We use way 0 and 1 for temporary mappings in such cases. + * + * The temporary DTLB entries shouldn't be flushed by interrupts, but are + * flushed by preemptive task switches. Special code in the + * fast_second_level_miss handler re-established the temporary mapping. + * It requires that the PPNs for the destination and source addresses are + * in a6, and a7, respectively. + */ + +/* TLB miss exceptions are treated special in the following region */ + +ENTRY(__tlbtemp_mapping_start) + +#if (DCACHE_WAY_SIZE > PAGE_SIZE) + +/* + * clear_user_page (void *addr, unsigned long vaddr, struct page *page) + * a2 a3 a4 + */ + +ENTRY(clear_user_page) + entry a1, 32 + + /* Mark page dirty and determine alias. */ + + movi a7, (1 << PG_ARCH_1) + l32i a5, a4, PAGE_FLAGS + xor a6, a2, a3 + extui a3, a3, PAGE_SHIFT, DCACHE_ALIAS_ORDER + extui a6, a6, PAGE_SHIFT, DCACHE_ALIAS_ORDER + or a5, a5, a7 + slli a3, a3, PAGE_SHIFT + s32i a5, a4, PAGE_FLAGS + + /* Skip setting up a temporary DTLB if not aliased. */ + + beqz a6, 1f + + /* Invalidate kernel page. */ + + mov a10, a2 + call8 __invalidate_dcache_page + + /* Setup a temporary DTLB with the color of the VPN */ + + movi a4, -PAGE_OFFSET + (PAGE_KERNEL | _PAGE_HW_WRITE) + movi a5, TLBTEMP_BASE_1 # virt + add a6, a2, a4 # ppn + add a2, a5, a3 # add 'color' + + wdtlb a6, a2 + dsync + +1: movi a3, 0 + __loopi a2, a7, PAGE_SIZE, 32 + s32i a3, a2, 0 + s32i a3, a2, 4 + s32i a3, a2, 8 + s32i a3, a2, 12 + s32i a3, a2, 16 + s32i a3, a2, 20 + s32i a3, a2, 24 + s32i a3, a2, 28 + __endla a2, a7, 32 + + bnez a6, 1f + retw + + /* We need to invalidate the temporary idtlb entry, if any. */ + +1: addi a2, a2, -PAGE_SIZE + idtlb a2 + dsync + + retw + +/* + * copy_page_user (void *to, void *from, unsigned long vaddr, struct page *page) + * a2 a3 a4 a5 + */ + +ENTRY(copy_user_page) + + entry a1, 32 + + /* Mark page dirty and determine alias for destination. */ + + movi a8, (1 << PG_ARCH_1) + l32i a9, a5, PAGE_FLAGS + xor a6, a2, a4 + xor a7, a3, a4 + extui a4, a4, PAGE_SHIFT, DCACHE_ALIAS_ORDER + extui a6, a6, PAGE_SHIFT, DCACHE_ALIAS_ORDER + extui a7, a7, PAGE_SHIFT, DCACHE_ALIAS_ORDER + or a9, a9, a8 + slli a4, a4, PAGE_SHIFT + s32i a9, a5, PAGE_FLAGS + movi a5, -PAGE_OFFSET + (PAGE_KERNEL | _PAGE_HW_WRITE) + + beqz a6, 1f + + /* Invalidate dcache */ + + mov a10, a2 + call8 __invalidate_dcache_page + + /* Setup a temporary DTLB with a matching color. */ + + movi a8, TLBTEMP_BASE_1 # base + add a6, a2, a5 # ppn + add a2, a8, a4 # add 'color' + + wdtlb a6, a2 + dsync + + /* Skip setting up a temporary DTLB for destination if not aliased. */ + +1: beqz a7, 1f + + /* Setup a temporary DTLB with a matching color. */ + + movi a8, TLBTEMP_BASE_2 # base + add a7, a3, a5 # ppn + add a3, a8, a4 + addi a8, a3, 1 # way1 + + wdtlb a7, a8 + dsync + +1: __loopi a2, a4, PAGE_SIZE, 32 + + l32i a8, a3, 0 + l32i a9, a3, 4 + s32i a8, a2, 0 + s32i a9, a2, 4 + + l32i a8, a3, 8 + l32i a9, a3, 12 + s32i a8, a2, 8 + s32i a9, a2, 12 + + l32i a8, a3, 16 + l32i a9, a3, 20 + s32i a8, a2, 16 + s32i a9, a2, 20 + + l32i a8, a3, 24 + l32i a9, a3, 28 + s32i a8, a2, 24 + s32i a9, a2, 28 + + addi a2, a2, 32 + addi a3, a3, 32 + + __endl a2, a4 + + /* We need to invalidate any temporary mapping! */ + + bnez a6, 1f + bnez a7, 2f + retw + +1: addi a2, a2, -PAGE_SIZE + idtlb a2 + dsync + bnez a7, 2f + retw + +2: addi a3, a3, -PAGE_SIZE+1 + idtlb a3 + dsync + + retw + +#endif + +#if (DCACHE_WAY_SIZE > PAGE_SIZE) + +/* + * void __flush_invalidate_dcache_page_alias (addr, phys) + * a2 a3 + */ + +ENTRY(__flush_invalidate_dcache_page_alias) + entry sp, 16 + + movi a7, 0 # required for exception handler + addi a6, a3, (PAGE_KERNEL | _PAGE_HW_WRITE) + mov a4, a2 + wdtlb a6, a2 + dsync + + ___flush_invalidate_dcache_page a2 a3 + + idtlb a4 + dsync + + retw + +#endif + +ENTRY(__tlbtemp_mapping_itlb) + +#if (ICACHE_WAY_SIZE > PAGE_SIZE) + +ENTRY(__invalidate_icache_page_alias) + entry sp, 16 + + addi a6, a3, (PAGE_KERNEL | _PAGE_HW_WRITE) + mov a4, a2 + witlb a6, a2 + isync + + ___invalidate_icache_page a2 a3 + + iitlb a4 + isync retw +#endif + +/* End of special treatment in tlb miss exception */ + +ENTRY(__tlbtemp_mapping_end) + /* * void __invalidate_icache_page(ulong start) */ @@ -121,8 +360,6 @@ ENTRY(__flush_dcache_page) dsync retw - - /* * void __invalidate_icache_range(ulong start, ulong size) */ @@ -168,7 +405,6 @@ ENTRY(__invalidate_dcache_range) ___invalidate_dcache_range a2 a3 a4 - retw /* diff --git a/arch/xtensa/platform-iss/console.c b/arch/xtensa/platform-iss/console.c index 2f4f20f..854677d 100644 --- a/arch/xtensa/platform-iss/console.c +++ b/arch/xtensa/platform-iss/console.c @@ -20,7 +20,6 @@ #include <linux/param.h> #include <linux/serial.h> #include <linux/serialP.h> -#include <linux/console.h> #include <asm/uaccess.h> #include <asm/irq.h> |