public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix
@ 2006-07-17 23:18 Jon Mason
  2006-07-18 13:50 ` [discuss] " Andi Kleen
  0 siblings, 1 reply; 4+ messages in thread
From: Jon Mason @ 2006-07-17 23:18 UTC (permalink / raw)
  To: ak; +Cc: muli, discuss, linux-kernel, konradr

Hey Andi,

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.

Please consider for inclusion in 2.6.18

Thanks,
Jon

Signed-Off-By: Jon Mason <jdmason@us.ibm.com>
Signed-Off-By: Muli Ben-Yehuda <muli@il.ibm.com>

diff -r c59d0a0b1975 arch/x86_64/kernel/pci-calgary.c
--- a/arch/x86_64/kernel/pci-calgary.c	Thu Jul 13 08:50:58 2006 -0500
+++ b/arch/x86_64/kernel/pci-calgary.c	Mon Jul 17 11:43:40 2006 -0500
@@ -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_mostl
  * 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_dm
 
 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 @@ void __init detect_calgary(void)
 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++;
-	}
-
-	if (detected) {
+		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 (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(
 			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 -r c59d0a0b1975 arch/x86_64/kernel/tce.c
--- a/arch/x86_64/kernel/tce.c	Thu Jul 13 08:50:58 2006 -0500
+++ b/arch/x86_64/kernel/tce.c	Mon Jul 17 11:43:40 2006 -0500
@@ -96,7 +96,6 @@ static int tce_table_setparms(struct pci
 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
 	/* 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");
diff -r c59d0a0b1975 include/asm-x86_64/calgary.h
--- a/include/asm-x86_64/calgary.h	Thu Jul 13 08:50:58 2006 -0500
+++ b/include/asm-x86_64/calgary.h	Mon Jul 17 11:43:40 2006 -0500
@@ -60,9 +60,4 @@ static inline void detect_calgary(void) 
 static inline void detect_calgary(void) { return; }
 #endif
 
-static inline unsigned int bus_to_phb(unsigned char busno)
-{
-	return ((busno % 15 == 0) ? 0 : busno / 2 + 1);
-}
-
 #endif /* _ASM_X86_64_CALGARY_H */

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [discuss] [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix
  2006-07-17 23:18 [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix Jon Mason
@ 2006-07-18 13:50 ` Andi Kleen
  2006-07-18 14:09   ` Muli Ben-Yehuda
  0 siblings, 1 reply; 4+ messages in thread
From: Andi Kleen @ 2006-07-18 13:50 UTC (permalink / raw)
  To: discuss; +Cc: Jon Mason, muli, linux-kernel, konradr

On Tuesday 18 July 2006 01:18, Jon Mason wrote:
> Hey Andi,
>
> 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).

The patch doesn't apply at all to rc2.

patching file arch/x86_64/kernel/pci-calgary.c
Hunk #5 FAILED at 814.
Hunk #9 FAILED at 892.
Hunk #10 succeeded at 941 (offset 1 line).
Hunk #11 FAILED at 1014.
3 out of 11 hunks FAILED -- rejects in file arch/x86_64/kernel/pci-calgary.c

Also where is 2/2 ? I only see 1/2 

-Andi

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [discuss] [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix
  2006-07-18 13:50 ` [discuss] " Andi Kleen
@ 2006-07-18 14:09   ` Muli Ben-Yehuda
  2006-07-18 14:17     ` Andi Kleen
  0 siblings, 1 reply; 4+ messages in thread
From: Muli Ben-Yehuda @ 2006-07-18 14:09 UTC (permalink / raw)
  To: Andi Kleen; +Cc: discuss, Jon Mason, linux-kernel, konradr

On Tue, Jul 18, 2006 at 03:50:57PM +0200, Andi Kleen wrote:
> On Tuesday 18 July 2006 01:18, Jon Mason wrote:
> > Hey Andi,
> >
> > 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).
> 
> The patch doesn't apply at all to rc2.
>
> Also where is 2/2 ? I only see 1/2 

Hmpf, there were two Calgary patches, both mistakenly labeled
1/2. This is actually the second one. When applied in order, both
apply to today's head.

Sorry about the confusion, please let us know if you want a resend. I
guess Jon got a smidgen too excited about OLS :-)

Cheers,
Muli

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [discuss] [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix
  2006-07-18 14:09   ` Muli Ben-Yehuda
@ 2006-07-18 14:17     ` Andi Kleen
  0 siblings, 0 replies; 4+ messages in thread
From: Andi Kleen @ 2006-07-18 14:17 UTC (permalink / raw)
  To: Muli Ben-Yehuda; +Cc: discuss, Jon Mason, linux-kernel, konradr


>
> Sorry about the confusion, please let us know if you want a resend. 

Please resend.

-Andi

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2006-07-18 14:16 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-07-17 23:18 [PATCH 1/2] x86_64: Calgary IOMMU - Multi-Node NULL pointer dereference fix Jon Mason
2006-07-18 13:50 ` [discuss] " Andi Kleen
2006-07-18 14:09   ` Muli Ben-Yehuda
2006-07-18 14:17     ` Andi Kleen

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox