Skip to contentSkip to conversation

IOMMU updates for 2.6.33-rc1

GIT PULL In review Joerg Roedel 2 participants 5 messages Dec 21, 2009 – Dec 28, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi Ingo,

The following changes since commit 55639353a0035052d9ea6cfe4dde0ac7fcbb2c9f:
Linus Torvalds (1):

        Linux 2.6.33-rc1

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu.git iommu/fixes

Joerg Roedel (2):
x86/amd-iommu: Fix initialization failure panic
Merge branches 'dma-debug/fixes' and 'amd-iommu/fixes' into iommu/fixes

Shaun Ruffell (1):
dma-debug: Do not add notifier when dma debugging is disabled.

 arch/x86/kernel/amd_iommu_init.c |   10 ++++++++++
 lib/dma-debug.c                  |    5 +++++

These two patches fix possible kernel null-pointer dereferences in the initialization path of AMD IOMMU and the device notifier path of dma-debug. Please pull.

Joerg

 2 files changed, 15 insertions(+), 0 deletions(-)
#1 lore raw

PATCH 1/2 dma-debug: Do not add notifier when dma debugging is disabled.

From: Shaun Ruffell <[email protected]>

If CONFIG_HAVE_DMA_API_DEBUG is defined and "dma_debug=off" is specified on the kernel command line, when you detach a driver from a device you can cause the following NULL pointer dereference:

BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<c0580d35>] dma_debug_device_change+0x5d/0x117

The problem is that the dma_debug_device_change notifier function is added to the bus notifier chain even though the dma_entry_hash array was never initialized. If dma debugging is disabled, this patch both prevents dma_debug_device_change notifiers from being added to the chain, and additionally ensures that the dma_debug_device_change notifier function is a no-op.

---
 lib/dma-debug.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)
lib/dma-debug.c +5 -0
index d9b08e0..7399744 100644
@@ -676,6 +676,8 @@ static int dma_debug_device_change(struct notifier_block *nb,
 	struct device *dev = data;
 	int count;
 
+	if (global_disable)
+		return;
 
 	switch (action) {
 	case BUS_NOTIFY_UNBOUND_DRIVER:
@@ -697,6 +699,9 @@ void dma_debug_add_bus(struct bus_type *bus)
 {
 	struct notifier_block *nb;
 
+	if (global_disable)
+		return;
+
 	nb = kzalloc(sizeof(struct notifier_block), GFP_KERNEL);
 	if (nb == NULL) {
 		pr_err("dma_debug_add_bus: out of memory\n");
Signature · 1 line
1.6.5.4
#2 lore raw

PATCH 2/2 x86/amd-iommu: Fix initialization failure panic

In reply to Joerg Roedel

The assumption that acpi_table_parse passes the return value of the hanlder function to the caller proved wrong recently. The return value of the handler function is totally ignored. This makes the initialization code for AMD IOMMU buggy in a way that could cause a kernel panic on initialization. This patch fixes the issue in the AMD IOMMU driver.

---
 arch/x86/kernel/amd_iommu_init.c |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)
arch/x86/kernel/amd_iommu_init.c +10 -0
index 1dca9c3..fb490ce 100644
@@ -138,6 +138,11 @@ int amd_iommus_present;
 bool amd_iommu_np_cache __read_mostly;
 
 /*
+ * Set to true if ACPI table parsing and hardware intialization went properly
+ */
+static bool amd_iommu_initialized;
+
+/*
  * List of protection domains - used during resume
  */
 LIST_HEAD(amd_iommu_pd_list);
@@ -929,6 +934,8 @@ static int __init init_iommu_all(struct acpi_table_header *table)
 	}
 	WARN_ON(p != end);
 
+	amd_iommu_initialized = true;
+
 	return 0;
 }
 
@@ -1263,6 +1270,9 @@ static int __init amd_iommu_init(void)
 	if (acpi_table_parse("IVRS", init_iommu_all) != 0)
 		goto free;
 
+	if (!amd_iommu_initialized)
+		goto free;
+
 	if (acpi_table_parse("IVRS", init_memory_definitions) != 0)
 		goto free;
 
Signature · 1 line
1.6.5.4
#3 lore raw

Re: [PATCH 2/2] x86/amd-iommu: Fix initialization failure panic

Quoted 19 lines · Joerg Roedel

The assumption that acpi_table_parse passes the return value of the hanlder function to the caller proved wrong recently. The return value of the handler function is totally ignored. This makes the initialization code for AMD IOMMU buggy in a way that could cause a kernel panic on initialization. This patch fixes the issue in the AMD IOMMU driver.

---
 arch/x86/kernel/amd_iommu_init.c |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)
arch/x86/kernel/amd_iommu_init.c +0 -0
index 1dca9c3..fb490ce 100644

arch/x86/kernel/amd_iommu_init.c int amd_iommus_present; · line 143

Quoted 6 lines
arch/x86/kernel/amd_iommu_init.c +3 -0
@@ -138,6 +138,11 @@ int amd_iommus_present;
 bool amd_iommu_np_cache __read_mostly;
 
 /*
+ * Set to true if ACPI table parsing and hardware intialization went properly
+ */
+static bool amd_iommu_initialized;

Btw., could this new variable be __init?

Ingo

#4 lore raw
In reply to Joerg Roedel
Quoted 21 lines · Joerg Roedel

Hi Ingo,

The following changes since commit 55639353a0035052d9ea6cfe4dde0ac7fcbb2c9f:
Linus Torvalds (1):

        Linux 2.6.33-rc1

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu.git iommu/fixes

Joerg Roedel (2):
x86/amd-iommu: Fix initialization failure panic
Merge branches 'dma-debug/fixes' and 'amd-iommu/fixes' into iommu/fixes

Shaun Ruffell (1):
dma-debug: Do not add notifier when dma debugging is disabled.

 arch/x86/kernel/amd_iommu_init.c |   10 ++++++++++
 lib/dma-debug.c                  |    5 +++++
Quoted 7 lines

These two patches fix possible kernel null-pointer dereferences in the initialization path of AMD IOMMU and the device notifier path of dma-debug. Please pull.

Joerg

 2 files changed, 15 insertions(+), 0 deletions(-)

Pulled, thanks a lot Joerg!

Ingo