summaryrefslogtreecommitdiffstats
path: root/arch/x86_64
diff options
context:
space:
mode:
authorJon Mason <jdmason@us.ibm.com>2006-07-29 21:42:43 +0200
committerLinus Torvalds <torvalds@g5.osdl.org>2006-07-29 20:59:55 -0700
commitd2105b10fe0f460c388fe4e09226313f519d8c00 (patch)
tree59ad2f99eeb124ecea6506801eb7f5c0a0a1395d /arch/x86_64
parent089bbbcb36979166131868a89ca5f4e695d6637d (diff)
downloadop-kernel-dev-d2105b10fe0f460c388fe4e09226313f519d8c00.zip
op-kernel-dev-d2105b10fe0f460c388fe4e09226313f519d8c00.tar.gz
[PATCH] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix
Calgary hits a NULL pointer dereference when booting in a multi-chassis NUMA system. See Redhat bugzilla number 198498, found by Konrad Rzeszutek (konradr@redhat.com). There are many issues that had to be resolved to fix this problem. Firstly when I originally wrote the code to handle NUMA systems, I had a large misunderstanding that was not corrected until now. That was that I thought the "number of nodes online" referred to number of physical systems connected. So that if NUMA was disabled, there would only be 1 node and it would only show that node's PCI bus. In reality if NUMA is disabled, the system displays all of the connected chassis as one node but is only ignorant of the delays in accessing main memory. Therefore, references to num_online_nodes() and MAX_NUMNODES are incorrect and need to be set to the maximum number of nodes that can be accessed (which are 8). I created a variable, MAX_NUM_CHASSIS, and set it to 8 to fix this. Secondly, when walking the PCI in detect_calgary, the code only checked the first "slot" when looking to see if a device is present. This will work for most cases, but unfortunately it isn't always the case. In the NUMA MXE drawers, there are USB devices present on the 3rd slot (with slot 1 being empty). So, to work around this, all slots (up to 8) are scanned to see if there are any devices present. Lastly, the bus is being enumerated on large systems in a different way the we originally thought. This throws the ugly logic we had out the window. To more elegantly handle this, I reorganized the kva array to be sparse (which removed the need to have any bus number to kva slot logic in tce.c) and created a secondary space array to contain the bus number to phb mapping. With these changes Calgary boots on an x460 with 4 nodes with and without NUMA enabled. Signed-off-by: Jon Mason <jdmason@us.ibm.com> Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com> Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'arch/x86_64')
-rw-r--r--arch/x86_64/kernel/pci-calgary.c76
-rw-r--r--arch/x86_64/kernel/tce.c4
2 files changed, 45 insertions, 35 deletions
diff --git a/arch/x86_64/kernel/pci-calgary.c b/arch/x86_64/kernel/pci-calgary.c
index 92744ab..146924b 100644
--- a/arch/x86_64/kernel/pci-calgary.c
+++ b/arch/x86_64/kernel/pci-calgary.c
@@ -85,7 +85,8 @@
#define CSR_AGENT_MASK 0xffe0ffff
#define MAX_NUM_OF_PHBS 8 /* how many PHBs in total? */
-#define MAX_PHB_BUS_NUM (MAX_NUM_OF_PHBS * 2) /* max dev->bus->number */
+#define MAX_NUM_CHASSIS 8 /* max number of chassis */
+#define MAX_PHB_BUS_NUM (MAX_NUM_OF_PHBS * MAX_NUM_CHASSIS * 2) /* max dev->bus->number */
#define PHBS_PER_CALGARY 4
/* register offsets in Calgary's internal register space */
@@ -110,7 +111,8 @@ static const unsigned long phb_offsets[] = {
0xB000 /* PHB3 */
};
-void* tce_table_kva[MAX_NUM_OF_PHBS * MAX_NUMNODES];
+static char bus_to_phb[MAX_PHB_BUS_NUM];
+void* tce_table_kva[MAX_PHB_BUS_NUM];
unsigned int specified_table_size = TCE_TABLE_SIZE_UNSPECIFIED;
static int translate_empty_slots __read_mostly = 0;
static int calgary_detected __read_mostly = 0;
@@ -119,7 +121,7 @@ static int calgary_detected __read_mostly = 0;
* the bitmap of PHBs the user requested that we disable
* translation on.
*/
-static DECLARE_BITMAP(translation_disabled, MAX_NUMNODES * MAX_PHB_BUS_NUM);
+static DECLARE_BITMAP(translation_disabled, MAX_PHB_BUS_NUM);
static void tce_cache_blast(struct iommu_table *tbl);
@@ -452,7 +454,7 @@ static struct dma_mapping_ops calgary_dma_ops = {
static inline int busno_to_phbid(unsigned char num)
{
- return bus_to_phb(num) % PHBS_PER_CALGARY;
+ return bus_to_phb[num];
}
static inline unsigned long split_queue_offset(unsigned char num)
@@ -812,7 +814,7 @@ static int __init calgary_init(void)
int i, ret = -ENODEV;
struct pci_dev *dev = NULL;
- for (i = 0; i < num_online_nodes() * MAX_NUM_OF_PHBS; i++) {
+ for (i = 0; i < MAX_PHB_BUS_NUM; i++) {
dev = pci_get_device(PCI_VENDOR_ID_IBM,
PCI_DEVICE_ID_IBM_CALGARY,
dev);
@@ -822,7 +824,7 @@ static int __init calgary_init(void)
calgary_init_one_nontraslated(dev);
continue;
}
- if (!tce_table_kva[i] && !translate_empty_slots) {
+ if (!tce_table_kva[dev->bus->number] && !translate_empty_slots) {
pci_dev_put(dev);
continue;
}
@@ -842,7 +844,7 @@ error:
pci_dev_put(dev);
continue;
}
- if (!tce_table_kva[i] && !translate_empty_slots)
+ if (!tce_table_kva[dev->bus->number] && !translate_empty_slots)
continue;
calgary_disable_translation(dev);
calgary_free_tar(dev);
@@ -876,9 +878,10 @@ static inline int __init determine_tce_table_size(u64 ram)
void __init detect_calgary(void)
{
u32 val;
- int bus, table_idx;
+ int bus;
void *tbl;
- int detected = 0;
+ int calgary_found = 0;
+ int phb = -1;
/*
* if the user specified iommu=off or iommu=soft or we found
@@ -889,37 +892,46 @@ void __init detect_calgary(void)
specified_table_size = determine_tce_table_size(end_pfn * PAGE_SIZE);
- for (bus = 0, table_idx = 0;
- bus < num_online_nodes() * MAX_PHB_BUS_NUM;
- bus++) {
+ for (bus = 0; bus < MAX_PHB_BUS_NUM; bus++) {
+ int dev;
+
+ tce_table_kva[bus] = NULL;
+ bus_to_phb[bus] = -1;
+
if (read_pci_config(bus, 0, 0, 0) != PCI_VENDOR_DEVICE_ID_CALGARY)
continue;
+
+ /*
+ * There are 4 PHBs per Calgary chip. Set phb to which phb (0-3)
+ * it is connected to releative to the clagary chip.
+ */
+ phb = (phb + 1) % PHBS_PER_CALGARY;
+
if (test_bit(bus, translation_disabled)) {
printk(KERN_INFO "Calgary: translation is disabled for "
"PHB 0x%x\n", bus);
/* skip this phb, don't allocate a tbl for it */
- tce_table_kva[table_idx] = NULL;
- table_idx++;
continue;
}
/*
- * scan the first slot of the PCI bus to see if there
- * are any devices present
+ * Scan the slots of the PCI bus to see if there is a device present.
+ * The parent bus will be the zero-ith device, so start at 1.
*/
- val = read_pci_config(bus, 1, 0, 0);
- if (val != 0xffffffff || translate_empty_slots) {
- tbl = alloc_tce_table();
- if (!tbl)
- goto cleanup;
- detected = 1;
- } else
- tbl = NULL;
-
- tce_table_kva[table_idx] = tbl;
- table_idx++;
+ for (dev = 1; dev < 8; dev++) {
+ val = read_pci_config(bus, dev, 0, 0);
+ if (val != 0xffffffff || translate_empty_slots) {
+ tbl = alloc_tce_table();
+ if (!tbl)
+ goto cleanup;
+ tce_table_kva[bus] = tbl;
+ bus_to_phb[bus] = phb;
+ calgary_found = 1;
+ break;
+ }
+ }
}
- if (detected) {
+ if (calgary_found) {
iommu_detected = 1;
calgary_detected = 1;
printk(KERN_INFO "PCI-DMA: Calgary IOMMU detected. "
@@ -928,9 +940,9 @@ void __init detect_calgary(void)
return;
cleanup:
- for (--table_idx; table_idx >= 0; --table_idx)
- if (tce_table_kva[table_idx])
- free_tce_table(tce_table_kva[table_idx]);
+ for (--bus; bus >= 0; --bus)
+ if (tce_table_kva[bus])
+ free_tce_table(tce_table_kva[bus]);
}
int __init calgary_iommu_init(void)
@@ -1001,7 +1013,7 @@ static int __init calgary_parse_options(char *p)
if (p == endp)
break;
- if (bridge < (num_online_nodes() * MAX_PHB_BUS_NUM)) {
+ if (bridge < MAX_PHB_BUS_NUM) {
printk(KERN_INFO "Calgary: disabling "
"translation for PHB 0x%x\n", bridge);
set_bit(bridge, translation_disabled);
diff --git a/arch/x86_64/kernel/tce.c b/arch/x86_64/kernel/tce.c
index d3a9e79..5530dda 100644
--- a/arch/x86_64/kernel/tce.c
+++ b/arch/x86_64/kernel/tce.c
@@ -96,7 +96,6 @@ static inline unsigned int table_size_to_number_of_entries(unsigned char size)
static int tce_table_setparms(struct pci_dev *dev, struct iommu_table *tbl)
{
unsigned int bitmapsz;
- unsigned int tce_table_index;
unsigned long bmppages;
int ret;
@@ -105,8 +104,7 @@ static int tce_table_setparms(struct pci_dev *dev, struct iommu_table *tbl)
/* set the tce table size - measured in entries */
tbl->it_size = table_size_to_number_of_entries(specified_table_size);
- tce_table_index = bus_to_phb(tbl->it_busno);
- tbl->it_base = (unsigned long)tce_table_kva[tce_table_index];
+ tbl->it_base = (unsigned long)tce_table_kva[dev->bus->number];
if (!tbl->it_base) {
printk(KERN_ERR "Calgary: iommu_table_setparms: "
"no table allocated?!\n");
OpenPOWER on IntegriCloud