Skip to contentSkip to conversation

iommu/vt-d: use dedicated bitmap to track remapping entry allocation status

PATCH Yijing Wang 1 participant 6 messages Dec 2, 2013 7 messages not in the archive PCI

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

#2 lore raw
Tested-and-reviewed-by Yijing Wang <[email protected]>
Quoted 125 lines · Jiang Liu

Currently Intel interrupt remapping drivers uses the "present" flag bit in remapping entry to track whether an entry is allocated or not. It works as follow:
1) allocate a remapping entry and set its "present" flag bit to 1
2) compose other fields for the entry
3) update the remapping entry with the composed value

The remapping hardware may access the entry between step 1 and step 3, which then obervers an entry with the "present" flag set but random values in all other fields.

This patch introduces a dedicated bitmap to track remapping entry allocation status instead of sharing the "present" flag with hardware, thus eliminate the race window. It also simplifies the implementation.

---
 drivers/iommu/intel_irq_remapping.c |   51 +++++++++++++++++------------------
 include/linux/intel-iommu.h         |    1 +
 2 files changed, 25 insertions(+), 27 deletions(-)
drivers/iommu/intel_irq_remapping.c +24 -27
index bab10b1..282d392 100644
@@ -72,7 +72,6 @@ static int alloc_irte(struct intel_iommu *iommu, int irq, u16 count)
 	u16 index, start_index;
 	unsigned int mask = 0;
 	unsigned long flags;
-	int i;
 
 	if (!count || !irq_iommu)
 		return -1;
@@ -96,32 +95,17 @@ static int alloc_irte(struct intel_iommu *iommu, int irq, u16 count)
 	}
 
 	raw_spin_lock_irqsave(&irq_2_ir_lock, flags);
-	do {
-		for (i = index; i < index + count; i++)
-			if  (table->base[i].present)
-				break;
-		/* empty index found */
-		if (i == index + count)
-			break;
-
-		index = (index + count) % INTR_REMAP_TABLE_ENTRIES;
-
-		if (index == start_index) {
-			raw_spin_unlock_irqrestore(&irq_2_ir_lock, flags);
-			printk(KERN_ERR "can't allocate an IRTE\n");
-			return -1;
-		}
-	} while (1);
-
-	for (i = index; i < index + count; i++)
-		table->base[i].present = 1;
-
-	cfg->remapped = 1;
-	irq_iommu->iommu = iommu;
-	irq_iommu->irte_index =  index;
-	irq_iommu->sub_handle = 0;
-	irq_iommu->irte_mask = mask;
-
+	index = bitmap_find_free_region(table->bitmap,
+					INTR_REMAP_TABLE_ENTRIES, mask);
+	if (index < 0) {
+		printk(KERN_ERR "can't allocate an IRTE\n");
+	} else {
+		cfg->remapped = 1;
+		irq_iommu->iommu = iommu;
+		irq_iommu->irte_index =  index;
+		irq_iommu->sub_handle = 0;
+		irq_iommu->irte_mask = mask;
+	}
 	raw_spin_unlock_irqrestore(&irq_2_ir_lock, flags);
 
 	return index;
@@ -254,6 +238,8 @@ static int clear_entries(struct irq_2_iommu *irq_iommu)
 		set_64bit(&entry->low, 0);
 		set_64bit(&entry->high, 0);
 	}
+	bitmap_release_region(iommu->ir_table->bitmap, index,
+			      irq_iommu->irte_mask);
 
 	return qi_flush_iec(iommu, index, irq_iommu->irte_mask);
 }
@@ -453,6 +439,7 @@ static int intel_setup_irq_remapping(struct intel_iommu *iommu, int mode)
 {
 	struct ir_table *ir_table;
 	struct page *pages;
+	unsigned long *bitmap;
 
 	ir_table = iommu->ir_table = kzalloc(sizeof(struct ir_table),
 					     GFP_ATOMIC);
@@ -470,7 +457,17 @@ static int intel_setup_irq_remapping(struct intel_iommu *iommu, int mode)
 		return -ENOMEM;
 	}
 
+	bitmap = kcalloc(BITS_TO_LONGS(INTR_REMAP_TABLE_ENTRIES),
+			 sizeof(long), GFP_ATOMIC);
+	if (bitmap == NULL) {
+		printk(KERN_ERR "failed to allocate bitmap\n");
+		__free_pages(pages, INTR_REMAP_PAGE_ORDER);
+		kfree(ir_table);
+		return -ENOMEM;
+	}
+
 	ir_table->base = page_address(pages);
+	ir_table->bitmap = bitmap;
 
 	iommu_set_irq_remapping(iommu, mode);
 	return 0;
include/linux/intel-iommu.h +1 -0
index d380c5e..de1e5e9 100644
@@ -288,6 +288,7 @@ struct q_inval {
 
 struct ir_table {
 	struct irte *base;
+	unsigned long *bitmap;
 };
 #endif
 
Signature · 2 lines
Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): [email protected]

author Reviewed-by
#4 lore raw

Re: [Patch Part1 V2 02/17] iommu/vt-d: fix PCI device reference leakage on error recovery path

Quoted 73 lines · Jiang Liu

Function dmar_parse_dev_scope() should release the PCI device reference count gained in function dmar_parse_one_dev_scope() on error recovery, otherwise will cause PCI device object leakage.

This patch also introduces dmar_free_dev_scope(), which will be used to support DMAR device hotplug.

---
 drivers/iommu/dmar.c |   15 +++++++++++++--
 include/linux/dmar.h |    1 +
 2 files changed, 14 insertions(+), 2 deletions(-)
drivers/iommu/dmar.c +13 -2
index 8b452c9..f3043a2 100644
@@ -72,6 +72,7 @@ static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
 	struct acpi_dmar_pci_path *path;
 	int count;
 
+	*dev = NULL;
 	bus = pci_find_bus(segment, scope->bus);
 	path = (struct acpi_dmar_pci_path *)(scope + 1);
 	count = (scope->length - sizeof(struct acpi_dmar_device_scope))
@@ -100,7 +101,6 @@ static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
 	if (!pdev) {
 		pr_warn("Device scope device [%04x:%02x:%02x.%02x] not found\n",
 			segment, scope->bus, path->device, path->function);
-		*dev = NULL;
 		return 0;
 	}
 	if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT && \
@@ -151,7 +151,7 @@ int __init dmar_parse_dev_scope(void *start, void *end, int *cnt,
 			ret = dmar_parse_one_dev_scope(scope,
 				&(*devices)[index], segment);
 			if (ret) {
-				kfree(*devices);
+				dmar_free_dev_scope(devices, cnt);
 				return ret;
 			}
 			index ++;
@@ -162,6 +162,17 @@ int __init dmar_parse_dev_scope(void *start, void *end, int *cnt,
 	return 0;
 }
 
+void dmar_free_dev_scope(struct pci_dev ***devices, int *cnt)
+{
+	if (*devices && *cnt) {
+		while (--*cnt >= 0)
+			pci_dev_put((*devices)[*cnt]);
+		kfree(*devices);
+		*devices = NULL;
+		*cnt = 0;
+	}
+}
+
 /**
  * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition
  * structure which uniquely represent one DMA remapping hardware unit
include/linux/dmar.h +1 -0
index b029d1a..8adfce0 100644
@@ -159,6 +159,7 @@ extern int dmar_parse_one_rmrr(struct acpi_dmar_header *header);
 extern int dmar_parse_one_atsr(struct acpi_dmar_header *header);
 extern int dmar_parse_dev_scope(void *start, void *end, int *cnt,
 				struct pci_dev ***devices, u16 segment);
+extern void dmar_free_dev_scope(struct pci_dev ***devices, int *cnt);
 extern int intel_iommu_init(void);
 #else /* !CONFIG_INTEL_IOMMU: */
 static inline int intel_iommu_init(void) { return -ENODEV; }
Signature · 2 lines
Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): [email protected]

author Acked-by
#6 lore raw

Re: [Patch Part1 V2 04/17] iommu/vt-d: fix resource leakage on error recovery path in iommu_init_domains()

Quoted 14 lines · Jiang Liu

Release allocated resources on error recovery path in function iommu_init_domains().

---
 drivers/iommu/intel-iommu.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/iommu/intel-iommu.c +0 -0
index b8e3b48..2398876 100644

drivers/iommu/intel-iommu.c static int iommu_init_domains(struct intel_iommu *iommu) · line 1280

Quoted 8 lines
drivers/iommu/intel-iommu.c +2 -0
@@ -1273,6 +1273,8 @@ static int iommu_init_domains(struct intel_iommu *iommu)
 			GFP_KERNEL);
 	if (!iommu->domains) {
 		printk(KERN_ERR "Allocating domain array failed\n");
+		kfree(iommu->domain_ids);
+		iommu->domain_ids = NULL;
 		return -ENOMEM;
 	}
 
Signature · 2 lines
Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): [email protected]

#8 lore raw

Re: [Patch Part1 V2 07/17] iommu/vt-d. trivial: check suitable flag in function detect_intel_iommu()

This patch is the same as the last.:)

Quoted 24 lines · Jiang Liu

Flag irq_remapping_enabled is only set by intel_enable_irq_remapping(), which is called after detect_intel_iommu(). So we should check flag disable_irq_remap instead of irq_remapping_enabled in function detect_intel_iommu().

---
 drivers/iommu/dmar.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/dmar.c +1 -1
index f3043a2..77a066b 100644
@@ -561,7 +561,7 @@ int __init detect_intel_iommu(void)
 
 		dmar = (struct acpi_table_dmar *) dmar_tbl;
 
-		if (ret && irq_remapping_enabled && cpu_has_x2apic &&
+		if (ret && !disable_irq_remap && cpu_has_x2apic &&
 		    dmar->flags & 0x1)
 			pr_info("Queued invalidation will be enabled to support x2apic and Intr-remapping.\n");
 
Signature · 2 lines
Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): [email protected]

author Reviewed-by
#10 lore raw

Re: [Patch Part1 V2 07/17] iommu/vt-d, trivial: check suitable flag in function detect_intel_iommu()

Quoted 16 lines · Jiang Liu

Flag irq_remapping_enabled is only set by intel_enable_irq_remapping(), which is called after detect_intel_iommu(). So we should check flag disable_irq_remap instead of irq_remapping_enabled in function detect_intel_iommu().

---
 drivers/iommu/dmar.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/dmar.c +0 -0
index f3043a2..77a066b 100644

drivers/iommu/dmar.c int __init detect_intel_iommu(void) · line 567

Quoted 8 lines
drivers/iommu/dmar.c +1 -1
@@ -561,7 +561,7 @@ int __init detect_intel_iommu(void)
 
 		dmar = (struct acpi_table_dmar *) dmar_tbl;
 
-		if (ret && irq_remapping_enabled && cpu_has_x2apic &&
+		if (ret && !disable_irq_remap && cpu_has_x2apic &&
 		    dmar->flags & 0x1)
 			pr_info("Queued invalidation will be enabled to support x2apic and Intr-remapping.\n");
 
Signature · 2 lines
Thanks!
Yijing

A message that is not in the archive (sent off-list or not yet received): [email protected]

author Reviewed-by
#12 lore raw

Re: [Patch Part1 V2 12/17] iommu/vt-d: fix invalid memory access when freeing DMAR irq

Quoted 82 lines · Jiang Liu

In function free_dmar_iommu(), it sets IRQ handler data to NULL before calling free_irq(), which will cause invalid memory access because free_irq() will access IRQ handler data when calling function dmar_msi_mask(). So only set IRQ handler data to NULL after calling free_irq().

Sample stack dump:

[   13.094010] BUG: unable to handle kernel NULL pointer dereference at 0000000000000048
[   13.103215] IP: [<ffffffff810a97cd>] __lock_acquire+0x4d/0x12a0
[   13.110104] PGD 0
[   13.112614] Oops: 0000 [#1] SMP
[   13.116585] Modules linked in:
[   13.120260] CPU: 60 PID: 1 Comm: swapper/0 Tainted: G        W    3.13.0-rc1-gerry+ #9
[   13.129367] Hardware name: Intel Corporation LH Pass ........../SVRBD-ROW_T, BIOS SE5C600.86B.99.99.x059.091020121352 09/10/2012
[   13.142555] task: ffff88042dd38010 ti: ffff88042dd32000 task.ti: ffff88042dd32000
[   13.151179] RIP: 0010:[<ffffffff810a97cd>]  [<ffffffff810a97cd>] __lock_acquire+0x4d/0x12a0
[   13.160867] RSP: 0000:ffff88042dd33b78  EFLAGS: 00010046
[   13.166969] RAX: 0000000000000046 RBX: 0000000000000002 RCX: 0000000000000000
[   13.175122] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000048
[   13.183274] RBP: ffff88042dd33bd8 R08: 0000000000000002 R09: 0000000000000001
[   13.191417] R10: 0000000000000000 R11: 0000000000000001 R12: ffff88042dd38010
[   13.199571] R13: 0000000000000000 R14: 0000000000000048 R15: 0000000000000000
[   13.207725] FS:  0000000000000000(0000) GS:ffff88103f200000(0000) knlGS:0000000000000000
[   13.217014] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   13.223596] CR2: 0000000000000048 CR3: 0000000001a0b000 CR4: 00000000000407e0
[   13.231747] Stack:
[   13.234160]  0000000000000004 0000000000000046 ffff88042dd33b98 ffffffff810a567d
[   13.243059]  ffff88042dd33c08 ffffffff810bb14c ffffffff828995a0 0000000000000046
[   13.251969]  0000000000000000 0000000000000000 0000000000000002 0000000000000000
[   13.260862] Call Trace:
[   13.263775]  [<ffffffff810a567d>] ? trace_hardirqs_off+0xd/0x10
[   13.270571]  [<ffffffff810bb14c>] ? vprintk_emit+0x23c/0x570
[   13.277058]  [<ffffffff810ab1e3>] lock_acquire+0x93/0x120
[   13.283269]  [<ffffffff814623f7>] ? dmar_msi_mask+0x47/0x70
[   13.289677]  [<ffffffff8156b449>] _raw_spin_lock_irqsave+0x49/0x90
[   13.296748]  [<ffffffff814623f7>] ? dmar_msi_mask+0x47/0x70
[   13.303153]  [<ffffffff814623f7>] dmar_msi_mask+0x47/0x70
[   13.309354]  [<ffffffff810c0d93>] irq_shutdown+0x53/0x60
[   13.315467]  [<ffffffff810bdd9d>] __free_irq+0x26d/0x280
[   13.321580]  [<ffffffff810be920>] free_irq+0xf0/0x180
[   13.327395]  [<ffffffff81466591>] free_dmar_iommu+0x271/0x2b0
[   13.333996]  [<ffffffff810a947d>] ? trace_hardirqs_on+0xd/0x10
[   13.340696]  [<ffffffff81461a17>] free_iommu+0x17/0x50
[   13.346597]  [<ffffffff81dc75a5>] init_dmars+0x691/0x77a
[   13.352711]  [<ffffffff81dc7afd>] intel_iommu_init+0x351/0x438
[   13.359400]  [<ffffffff81d8a711>] ? iommu_setup+0x27d/0x27d
[   13.365806]  [<ffffffff81d8a739>] pci_iommu_init+0x28/0x52
[   13.372114]  [<ffffffff81000342>] do_one_initcall+0x122/0x180
[   13.378707]  [<ffffffff81077738>] ? parse_args+0x1e8/0x320
[   13.385016]  [<ffffffff81d850e8>] kernel_init_freeable+0x1e1/0x26c
[   13.392100]  [<ffffffff81d84833>] ? do_early_param+0x88/0x88
[   13.398596]  [<ffffffff8154f8b0>] ? rest_init+0xd0/0xd0
[   13.404614]  [<ffffffff8154f8be>] kernel_init+0xe/0x130
[   13.410626]  [<ffffffff81574d6c>] ret_from_fork+0x7c/0xb0
[   13.416829]  [<ffffffff8154f8b0>] ? rest_init+0xd0/0xd0
[   13.422842] Code: ec 99 00 85 c0 8b 05 53 05 a5 00 41 0f 45 d8 85 c0 0f 84 ff 00 00 00 8b 05 99 f9 7e 01 49 89 fe 41 89 f7 85 c0 0f 84 03 01 00 00 <49> 8b 06 be 01 00 00 00 48 3d c0 0e 01 82 0f 44 de 41 83 ff 01
[   13.450191] RIP  [<ffffffff810a97cd>] __lock_acquire+0x4d/0x12a0
[   13.458598]  RSP <ffff88042dd33b78>
[   13.462671] CR2: 0000000000000048
[   13.466551] ---[ end trace c5bd26a37c81d760 ]---
---
 drivers/iommu/intel-iommu.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/intel-iommu.c +1 -1
index 0ec49da..426095e 100644
@@ -1289,9 +1289,9 @@ void free_dmar_iommu(struct intel_iommu *iommu)
 		iommu_disable_translation(iommu);
 
 	if (iommu->irq) {
-		irq_set_handler_data(iommu->irq, NULL);
 		/* This will mask the irq */
 		free_irq(iommu->irq, iommu);
+		irq_set_handler_data(iommu->irq, NULL);
 		destroy_irq(iommu->irq);
 	}
 
Signature · 2 lines
Thanks!
Yijing