Skip to contentSkip to conversation

iommu/vt-d: Fix WARN_ON in iommu probe path

PATCH In review Lu Baolu 4 participants 10 messages Apr 7, 2024 – Nov 18

Subsystems (3)
lore
Share & tools
lore

Lu Baolu

author maintainer
#0 lore raw

PATCH 1/1 iommu/vt-d: Fix WARN_ON in iommu probe path

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on some platforms below kernel splat will be displayed and the iommu probe results in failure.

 WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158 intel_iommu_probe_device+0x319/0xd90
 Call Trace:
  <TASK>
  ? __warn+0x7e/0x180
  ? intel_iommu_probe_device+0x319/0xd90
  ? report_bug+0x1f8/0x200
  ? handle_bug+0x3c/0x70
  ? exc_invalid_op+0x18/0x70
  ? asm_exc_invalid_op+0x1a/0x20
  ? intel_iommu_probe_device+0x319/0xd90
  ? debug_mutex_init+0x37/0x50
  __iommu_probe_device+0xf2/0x4f0
  iommu_probe_device+0x22/0x70
  iommu_bus_notifier+0x1e/0x40
  notifier_call_chain+0x46/0x150
  blocking_notifier_call_chain+0x42/0x60
  bus_notify+0x2f/0x50
  device_add+0x5ed/0x7e0
  platform_device_add+0xf5/0x240
  mfd_add_devices+0x3f9/0x500
  ? preempt_count_add+0x4c/0xa0
  ? up_write+0xa2/0x1b0
  ? __debugfs_create_file+0xe3/0x150
  intel_lpss_probe+0x49f/0x5b0
  ? pci_conf1_write+0xa3/0xf0
  intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
  pci_device_probe+0x95/0x120
  really_probe+0xd9/0x370
  ? __pfx___driver_attach+0x10/0x10
  __driver_probe_device+0x73/0x150
  driver_probe_device+0x19/0xa0
  __driver_attach+0xb6/0x180
  ? __pfx___driver_attach+0x10/0x10
  bus_for_each_dev+0x77/0xd0
  bus_add_driver+0x114/0x210
  driver_register+0x5b/0x110
  ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
  do_one_initcall+0x57/0x2b0
  ? kmalloc_trace+0x21e/0x280
  ? do_init_module+0x1e/0x210
  do_init_module+0x5f/0x210
  load_module+0x1d37/0x1fc0
  ? init_module_from_file+0x86/0xd0
  init_module_from_file+0x86/0xd0
  idempotent_init_module+0x17c/0x230
  __x64_sys_finit_module+0x56/0xb0
  do_syscall_64+0x6e/0x140
  entry_SYSCALL_64_after_hwframe+0x71/0x79
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices")
---
 drivers/iommu/intel/iommu.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iommu/intel/iommu.c +7 -4
index 50eb9aed47cc..a7ecd90303dc 100644
@@ -4299,9 +4299,11 @@ static struct iommu_device *intel_iommu_probe_device(struct device *dev)
 	}
 
 	dev_iommu_priv_set(dev, info);
-	ret = device_rbtree_insert(iommu, info);
-	if (ret)
-		goto free;
+	if (pdev && pci_ats_supported(pdev)) {
+		ret = device_rbtree_insert(iommu, info);
+		if (ret)
+			goto free;
+	}
 
 	if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
 		ret = intel_pasid_alloc_table(dev);
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct device *dev)
 	struct intel_iommu *iommu = info->iommu;
 
 	mutex_lock(&iommu->iopf_lock);
-	device_rbtree_remove(info);
+	if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+		device_rbtree_remove(info);
 	mutex_unlock(&iommu->iopf_lock);
 
 	if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev) &&
Signature · 1 line
2.34.1
#1 lore raw
Quoted 9 lines

From: Lu Baolu <[email protected]>
Sent: Sunday, April 7, 2024 9:14 AM

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

Quoted 8 lines

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on some platforms below kernel splat will be displayed and the iommu probe results in failure.

Just be curious. What about two ATS capable devices putting behind a PCI-to-PCIe bridge?

Quoted 65 lines
 WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158
intel_iommu_probe_device+0x319/0xd90
 Call Trace:
  <TASK>
  ? __warn+0x7e/0x180
  ? intel_iommu_probe_device+0x319/0xd90
  ? report_bug+0x1f8/0x200
  ? handle_bug+0x3c/0x70
  ? exc_invalid_op+0x18/0x70
  ? asm_exc_invalid_op+0x1a/0x20
  ? intel_iommu_probe_device+0x319/0xd90
  ? debug_mutex_init+0x37/0x50
  __iommu_probe_device+0xf2/0x4f0
  iommu_probe_device+0x22/0x70
  iommu_bus_notifier+0x1e/0x40
  notifier_call_chain+0x46/0x150
  blocking_notifier_call_chain+0x42/0x60
  bus_notify+0x2f/0x50
  device_add+0x5ed/0x7e0
  platform_device_add+0xf5/0x240
  mfd_add_devices+0x3f9/0x500
  ? preempt_count_add+0x4c/0xa0
  ? up_write+0xa2/0x1b0
  ? __debugfs_create_file+0xe3/0x150
  intel_lpss_probe+0x49f/0x5b0
  ? pci_conf1_write+0xa3/0xf0
  intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
  pci_device_probe+0x95/0x120
  really_probe+0xd9/0x370
  ? __pfx___driver_attach+0x10/0x10
  __driver_probe_device+0x73/0x150
  driver_probe_device+0x19/0xa0
  __driver_attach+0xb6/0x180
  ? __pfx___driver_attach+0x10/0x10
  bus_for_each_dev+0x77/0xd0
  bus_add_driver+0x114/0x210
  driver_register+0x5b/0x110
  ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
  do_one_initcall+0x57/0x2b0
  ? kmalloc_trace+0x21e/0x280
  ? do_init_module+0x1e/0x210
  do_init_module+0x5f/0x210
  load_module+0x1d37/0x1fc0
  ? init_module_from_file+0x86/0xd0
  init_module_from_file+0x86/0xd0
  idempotent_init_module+0x17c/0x230
  __x64_sys_finit_module+0x56/0xb0
  do_syscall_64+0x6e/0x140
  entry_SYSCALL_64_after_hwframe+0x71/0x79
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed

devices")

---
 drivers/iommu/intel/iommu.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iommu/intel/iommu.c +0 -0
index 50eb9aed47cc..a7ecd90303dc 100644
@@ -4299,9 +4299,11 @@ static struct iommu_device
 *intel_iommu_probe_device(struct device *dev)
 	}

drivers/iommu/intel/iommu.c static struct iommu_device · line 4307

Quoted 10 lines
drivers/iommu/intel/iommu.c +5 -3
 
 	dev_iommu_priv_set(dev, info);
-	ret = device_rbtree_insert(iommu, info);
-	if (ret)
-		goto free;
+	if (pdev && pci_ats_supported(pdev)) {
+		ret = device_rbtree_insert(iommu, info);
+		if (ret)
+			goto free;
+	}

probably replace device_rbtree with ats_rbtree?

Quoted 16 lines
drivers/iommu/intel/iommu.c +2 -1
 	if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
 		ret = intel_pasid_alloc_table(dev);
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct
 device *dev)
 	struct intel_iommu *iommu = info->iommu;
 
 	mutex_lock(&iommu->iopf_lock);
-	device_rbtree_remove(info);
+	if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+		device_rbtree_remove(info);
 	mutex_unlock(&iommu->iopf_lock);
 

if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev) &&
--
2.34.1

author maintainer
#2 lore raw
Quoted 13 lines · Kevin
Quoted 13 lines

From: Lu Baolu<[email protected]>
Sent: Sunday, April 7, 2024 9:14 AM

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Quoted 6 lines
Quoted 4 lines

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on some platforms below kernel splat will be displayed and the iommu probe results in failure.

Just be curious. What about two ATS capable devices putting behind a PCI-to-PCIe bridge?

I don't think ATS capable device putting behind a PCI-to-PCIe bridge is a right configuration for the ATS service. The PCIe spec requires this in section 10.1.1, that

"
ATS requires the following:
- ATS capable components must interoperate with [PCIe-1.1] compliant components.
...
"

My understanding is that PCI-to-PCIe bridge is not a PCIe compliant device.

Quoted 66 lines
Quoted 66 lines
  WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158
intel_iommu_probe_device+0x319/0xd90
  Call Trace:
   <TASK>
   ? __warn+0x7e/0x180
   ? intel_iommu_probe_device+0x319/0xd90
   ? report_bug+0x1f8/0x200
   ? handle_bug+0x3c/0x70
   ? exc_invalid_op+0x18/0x70
   ? asm_exc_invalid_op+0x1a/0x20
   ? intel_iommu_probe_device+0x319/0xd90
   ? debug_mutex_init+0x37/0x50
   __iommu_probe_device+0xf2/0x4f0
   iommu_probe_device+0x22/0x70
   iommu_bus_notifier+0x1e/0x40
   notifier_call_chain+0x46/0x150
   blocking_notifier_call_chain+0x42/0x60
   bus_notify+0x2f/0x50
   device_add+0x5ed/0x7e0
   platform_device_add+0xf5/0x240
   mfd_add_devices+0x3f9/0x500
   ? preempt_count_add+0x4c/0xa0
   ? up_write+0xa2/0x1b0
   ? __debugfs_create_file+0xe3/0x150
   intel_lpss_probe+0x49f/0x5b0
   ? pci_conf1_write+0xa3/0xf0
   intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
   pci_device_probe+0x95/0x120
   really_probe+0xd9/0x370
   ? __pfx___driver_attach+0x10/0x10
   __driver_probe_device+0x73/0x150
   driver_probe_device+0x19/0xa0
   __driver_attach+0xb6/0x180
   ? __pfx___driver_attach+0x10/0x10
   bus_for_each_dev+0x77/0xd0
   bus_add_driver+0x114/0x210
   driver_register+0x5b/0x110
   ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
   do_one_initcall+0x57/0x2b0
   ? kmalloc_trace+0x21e/0x280
   ? do_init_module+0x1e/0x210
   do_init_module+0x5f/0x210
   load_module+0x1d37/0x1fc0
   ? init_module_from_file+0x86/0xd0
   init_module_from_file+0x86/0xd0
   idempotent_init_module+0x17c/0x230
   __x64_sys_finit_module+0x56/0xb0
   do_syscall_64+0x6e/0x140
   entry_SYSCALL_64_after_hwframe+0x71/0x79
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed

devices")

---
  drivers/iommu/intel/iommu.c | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iommu/intel/iommu.c +0 -0
index 50eb9aed47cc..a7ecd90303dc 100644
@@ -4299,9 +4299,11 @@ static struct iommu_device
 *intel_iommu_probe_device(struct device *dev)
  	}
 
Quoted 10 lines

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4306

Quoted 9 lines
drivers/iommu/intel/iommu.c +5 -3
  	dev_iommu_priv_set(dev, info);
-	ret = device_rbtree_insert(iommu, info);
-	if (ret)
-		goto free;
+	if (pdev && pci_ats_supported(pdev)) {
+		ret = device_rbtree_insert(iommu, info);
+		if (ret)
+			goto free;
+	}

probably replace device_rbtree with ats_rbtree?

It makes sense. Probably I need a follow-up patch to do such cleanup.

Best regards,
baolu

#3 lore raw

+Bjorn

Quoted 37 lines

From: Baolu Lu <[email protected]>
Sent: Monday, April 8, 2024 3:15 PM

Quoted 9 lines · Kevin
Quoted 9 lines

From: Lu Baolu<[email protected]>
Sent: Sunday, April 7, 2024 9:14 AM

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

Quoted 6 lines
Quoted 6 lines

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on

some

Quoted 1 line
Quoted 1 line

platforms below kernel splat will be displayed and the iommu probe

results

Quoted 3 lines
Quoted 1 line

in failure.

Just be curious. What about two ATS capable devices putting behind a PCI-to-PCIe bridge?

I don't think ATS capable device putting behind a PCI-to-PCIe bridge is a right configuration for the ATS service. The PCIe spec requires this in section 10.1.1, that

"
ATS requires the following:
- ATS capable components must interoperate with [PCIe-1.1] compliant components.
...
"

Quoted 3 lines

My understanding is that PCI-to-PCIe bridge is not a PCIe compliant device.

what is the pci core policy on such device? pci_enable_ats() doesn't prevent such device and I saw various places in pci core do handle the PCI_EXP_TYPE_PCIE_BRIDGE type...

author maintainer
#5 lore raw
In reply to Lu Baolu
Quoted 64 lines · Lu Baolu

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on some platforms below kernel splat will be displayed and the iommu probe results in failure.

  WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158 intel_iommu_probe_device+0x319/0xd90
  Call Trace:
   <TASK>
   ? __warn+0x7e/0x180
   ? intel_iommu_probe_device+0x319/0xd90
   ? report_bug+0x1f8/0x200
   ? handle_bug+0x3c/0x70
   ? exc_invalid_op+0x18/0x70
   ? asm_exc_invalid_op+0x1a/0x20
   ? intel_iommu_probe_device+0x319/0xd90
   ? debug_mutex_init+0x37/0x50
   __iommu_probe_device+0xf2/0x4f0
   iommu_probe_device+0x22/0x70
   iommu_bus_notifier+0x1e/0x40
   notifier_call_chain+0x46/0x150
   blocking_notifier_call_chain+0x42/0x60
   bus_notify+0x2f/0x50
   device_add+0x5ed/0x7e0
   platform_device_add+0xf5/0x240
   mfd_add_devices+0x3f9/0x500
   ? preempt_count_add+0x4c/0xa0
   ? up_write+0xa2/0x1b0
   ? __debugfs_create_file+0xe3/0x150
   intel_lpss_probe+0x49f/0x5b0
   ? pci_conf1_write+0xa3/0xf0
   intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
   pci_device_probe+0x95/0x120
   really_probe+0xd9/0x370
   ? __pfx___driver_attach+0x10/0x10
   __driver_probe_device+0x73/0x150
   driver_probe_device+0x19/0xa0
   __driver_attach+0xb6/0x180
   ? __pfx___driver_attach+0x10/0x10
   bus_for_each_dev+0x77/0xd0
   bus_add_driver+0x114/0x210
   driver_register+0x5b/0x110
   ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
   do_one_initcall+0x57/0x2b0
   ? kmalloc_trace+0x21e/0x280
   ? do_init_module+0x1e/0x210
   do_init_module+0x5f/0x210
   load_module+0x1d37/0x1fc0
   ? init_module_from_file+0x86/0xd0
   init_module_from_file+0x86/0xd0
   idempotent_init_module+0x17c/0x230
   __x64_sys_finit_module+0x56/0xb0
   do_syscall_64+0x6e/0x140
   entry_SYSCALL_64_after_hwframe+0x71/0x79
Quoted 5 lines
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices")
---
  drivers/iommu/intel/iommu.c | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)

This patch closes a boot bug described here.

I will queue it for v6.9-rc if no objection.

Best regards,
baolu

Reviewed-by
#6 lore raw
Quoted 77 lines

From: Baolu Lu <[email protected]>
Sent: Thursday, April 11, 2024 10:48 AM

Quoted 6 lines · Lu Baolu

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

Quoted 6 lines

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on

some

Quoted 4 lines

platforms below kernel splat will be displayed and the iommu probe results in failure.

  WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158

intel_iommu_probe_device+0x319/0xd90

Quoted 48 lines
  Call Trace:
   <TASK>
   ? __warn+0x7e/0x180
   ? intel_iommu_probe_device+0x319/0xd90
   ? report_bug+0x1f8/0x200
   ? handle_bug+0x3c/0x70
   ? exc_invalid_op+0x18/0x70
   ? asm_exc_invalid_op+0x1a/0x20
   ? intel_iommu_probe_device+0x319/0xd90
   ? debug_mutex_init+0x37/0x50
   __iommu_probe_device+0xf2/0x4f0
   iommu_probe_device+0x22/0x70
   iommu_bus_notifier+0x1e/0x40
   notifier_call_chain+0x46/0x150
   blocking_notifier_call_chain+0x42/0x60
   bus_notify+0x2f/0x50
   device_add+0x5ed/0x7e0
   platform_device_add+0xf5/0x240
   mfd_add_devices+0x3f9/0x500
   ? preempt_count_add+0x4c/0xa0
   ? up_write+0xa2/0x1b0
   ? __debugfs_create_file+0xe3/0x150
   intel_lpss_probe+0x49f/0x5b0
   ? pci_conf1_write+0xa3/0xf0
   intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
   pci_device_probe+0x95/0x120
   really_probe+0xd9/0x370
   ? __pfx___driver_attach+0x10/0x10
   __driver_probe_device+0x73/0x150
   driver_probe_device+0x19/0xa0
   __driver_attach+0xb6/0x180
   ? __pfx___driver_attach+0x10/0x10
   bus_for_each_dev+0x77/0xd0
   bus_add_driver+0x114/0x210
   driver_register+0x5b/0x110
   ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
   do_one_initcall+0x57/0x2b0
   ? kmalloc_trace+0x21e/0x280
   ? do_init_module+0x1e/0x210
   do_init_module+0x5f/0x210
   load_module+0x1d37/0x1fc0
   ? init_module_from_file+0x86/0xd0
   init_module_from_file+0x86/0xd0
   idempotent_init_module+0x17c/0x230
   __x64_sys_finit_module+0x56/0xb0
   do_syscall_64+0x6e/0x140
   entry_SYSCALL_64_after_hwframe+0x71/0x79
Quoted 1 line
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed

devices")

Quoted 4 lines
---
  drivers/iommu/intel/iommu.c | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)
Quoted 6 lines

This patch closes a boot bug described here.

I will queue it for v6.9-rc if no objection.

let's do it. the issue which I raised is rare and can be tackled when it arises.

#4 lore raw
In reply to Tian, Kevin
Quoted 13 lines · Kevin
Quoted 13 lines

From: Lu Baolu <[email protected]>
Sent: Sunday, April 7, 2024 9:14 AM

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.

The reason for using a rbtree to track devices is to look up the device with PCI bus and devfunc in the paths of handling ATS invalidation time out error and the PRI I/O page faults. Both are PCI ATS feature related.

Quoted 6 lines
Quoted 4 lines

Only track the devices that have PCI ATS capabilities in the rbtree to avoid unnecessary WARN_ON in the iommu probe path. Otherwise, on some platforms below kernel splat will be displayed and the iommu probe results in failure.

Just be curious. What about two ATS capable devices putting behind a PCI-to-PCIe bridge?

No PCI-to-PCIe bridges such device AFAIK, ATS TLP couldn't be routed via PCIe-to-PCI bridge unmodified. ATS is based on PCIe.

But the device-rbtree is possible holding mutiple device nodes behind PCIe-to-PCI bridge indexed by the same requester-ID.

Thanks,
Ethan

Quoted 92 lines
Quoted 64 lines
  WARNING: CPU: 3 PID: 166 at drivers/iommu/intel/iommu.c:158
intel_iommu_probe_device+0x319/0xd90
  Call Trace:
   <TASK>
   ? __warn+0x7e/0x180
   ? intel_iommu_probe_device+0x319/0xd90
   ? report_bug+0x1f8/0x200
   ? handle_bug+0x3c/0x70
   ? exc_invalid_op+0x18/0x70
   ? asm_exc_invalid_op+0x1a/0x20
   ? intel_iommu_probe_device+0x319/0xd90
   ? debug_mutex_init+0x37/0x50
   __iommu_probe_device+0xf2/0x4f0
   iommu_probe_device+0x22/0x70
   iommu_bus_notifier+0x1e/0x40
   notifier_call_chain+0x46/0x150
   blocking_notifier_call_chain+0x42/0x60
   bus_notify+0x2f/0x50
   device_add+0x5ed/0x7e0
   platform_device_add+0xf5/0x240
   mfd_add_devices+0x3f9/0x500
   ? preempt_count_add+0x4c/0xa0
   ? up_write+0xa2/0x1b0
   ? __debugfs_create_file+0xe3/0x150
   intel_lpss_probe+0x49f/0x5b0
   ? pci_conf1_write+0xa3/0xf0
   intel_lpss_pci_probe+0xcf/0x110 [intel_lpss_pci]
   pci_device_probe+0x95/0x120
   really_probe+0xd9/0x370
   ? __pfx___driver_attach+0x10/0x10
   __driver_probe_device+0x73/0x150
   driver_probe_device+0x19/0xa0
   __driver_attach+0xb6/0x180
   ? __pfx___driver_attach+0x10/0x10
   bus_for_each_dev+0x77/0xd0
   bus_add_driver+0x114/0x210
   driver_register+0x5b/0x110
   ? __pfx_intel_lpss_pci_driver_init+0x10/0x10 [intel_lpss_pci]
   do_one_initcall+0x57/0x2b0
   ? kmalloc_trace+0x21e/0x280
   ? do_init_module+0x1e/0x210
   do_init_module+0x5f/0x210
   load_module+0x1d37/0x1fc0
   ? init_module_from_file+0x86/0xd0
   init_module_from_file+0x86/0xd0
   idempotent_init_module+0x17c/0x230
   __x64_sys_finit_module+0x56/0xb0
   do_syscall_64+0x6e/0x140
   entry_SYSCALL_64_after_hwframe+0x71/0x79
Fixes 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed

devices")

---
  drivers/iommu/intel/iommu.c | 11 +++++++----
  1 file changed, 7 insertions(+), 4 deletions(-)
drivers/iommu/intel/iommu.c +0 -0
index 50eb9aed47cc..a7ecd90303dc 100644
@@ -4299,9 +4299,11 @@ static struct iommu_device
 *intel_iommu_probe_device(struct device *dev)
  	}

drivers/iommu/intel/iommu.c static struct iommu_device · line 4307

Quoted 10 lines
drivers/iommu/intel/iommu.c +5 -3
 
  	dev_iommu_priv_set(dev, info);
-	ret = device_rbtree_insert(iommu, info);
-	if (ret)
-		goto free;
+	if (pdev && pci_ats_supported(pdev)) {
+		ret = device_rbtree_insert(iommu, info);
+		if (ret)
+			goto free;
+	}

probably replace device_rbtree with ats_rbtree?

Quoted 15 lines
drivers/iommu/intel/iommu.c +2 -1
  	if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
  		ret = intel_pasid_alloc_table(dev);
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct
 device *dev)
  	struct intel_iommu *iommu = info->iommu;
 
  	mutex_lock(&iommu->iopf_lock);
-	device_rbtree_remove(info);
+	if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+		device_rbtree_remove(info);
  	mutex_unlock(&iommu->iopf_lock);
 

if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev) &&
--
2.34.1

#7 lore raw
In reply to Lu Baolu

[+cc Danny and others because this is related to ongoing work at
https://lore.kernel.org/all/[email protected]/]

Quoted 10 lines · Lu Baolu

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.
...

+++ b/drivers/iommu/intel/iommu.c

drivers/iommu/intel/iommu.c +0 -0
@@ -4299,9 +4299,11 @@ static struct iommu_device *intel_iommu_probe_device(struct device *dev)
      }

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4306

Quoted 10 lines
drivers/iommu/intel/iommu.c +5 -3
 
      dev_iommu_priv_set(dev, info);
-     ret = device_rbtree_insert(iommu, info);
-     if (ret)
-             goto free;
+     if (pdev && pci_ats_supported(pdev)) {
+             ret = device_rbtree_insert(iommu, info);
+             if (ret)
+                     goto free;
+     }

This change is upstream as 89436f4f5412 ("iommu/vt-d: Fix WARN_ON in iommu probe path"), but I think there's a problem here. Previously the rbtree insertion was unconditional, so the unconditional remove in the error path was correct.

After 89436f4f5412, the rbtree insertion is conditional. If intel_pasid_alloc_table() fails, we try to remove it even though it may never have been inserted:

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4309

Quoted 2 lines
drivers/iommu/intel/iommu.c +0 -0
      if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
              ret = intel_pasid_alloc_table(dev);
                if (ret) {
                        dev_err(dev, "PASID table allocation failed\n");
                        goto clear_rbtree;
                }
    ...
    clear_rbtree:
        device_rbtree_remove(info);
    free:
        kfree(info);

drivers/iommu/intel/iommu.c static void intel_iommu_release_device(struct device *dev) · line 4342

Quoted 7 lines
drivers/iommu/intel/iommu.c +2 -1
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct device *dev)
      struct intel_iommu *iommu = info->iommu;
 
      mutex_lock(&iommu->iopf_lock);
-     device_rbtree_remove(info);
+     if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+             device_rbtree_remove(info);

This is also somewhat problematic because it relies on the assumption that pci_ats_supported(X) always returns the same value it returned in intel_iommu_probe_device(). That may be true today, but I don't think you should rely on it.

I think both issues could be solved if intel/iommu.c kept track of the rbtree insertion itself, e.g., with a bit in struct device_domain_info.

Quoted 6 lines
drivers/iommu/intel/iommu.c +0 -0
      mutex_unlock(&iommu->iopf_lock);
 
      if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev) &&
Signature · 1 line
2.34.1
author maintainer
#8 lore raw
Quoted 30 lines · Bjorn Helgaas

[+cc Danny and others because this is related to ongoing work at
https://lore.kernel.org/all/[email protected]/]

Quoted 10 lines · Lu Baolu

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.
...

+++ b/drivers/iommu/intel/iommu.c

drivers/iommu/intel/iommu.c +0 -0
@@ -4299,9 +4299,11 @@ static struct iommu_device *intel_iommu_probe_device(struct device *dev)
       }

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4306

Quoted 10 lines
drivers/iommu/intel/iommu.c +5 -3
 
       dev_iommu_priv_set(dev, info);
-     ret = device_rbtree_insert(iommu, info);
-     if (ret)
-             goto free;
+     if (pdev && pci_ats_supported(pdev)) {
+             ret = device_rbtree_insert(iommu, info);
+             if (ret)
+                     goto free;
+     }

This change is upstream as 89436f4f5412 ("iommu/vt-d: Fix WARN_ON in iommu probe path"), but I think there's a problem here. Previously the rbtree insertion was unconditional, so the unconditional remove in the error path was correct.

Quoted 3 lines

After 89436f4f5412, the rbtree insertion is conditional. If intel_pasid_alloc_table() fails, we try to remove it even though it may never have been inserted:

info->node is initialized with RB_CLEAR_NODE() before the insertion, and the remove helper checks RB_EMPTY_NODE() before erasing. If insertion was skipped, removal is a no-op, so this error path does not need another state check.

Quoted 16 lines

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4309

Quoted 3 lines
drivers/iommu/intel/iommu.c +0 -0
       if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
               ret = intel_pasid_alloc_table(dev);
                 if (ret) {
                         dev_err(dev, "PASID table allocation failed\n");
                         goto clear_rbtree;
                 }
     ...
     clear_rbtree:
         device_rbtree_remove(info);
     free:
         kfree(info);
Quoted 3 lines
drivers/iommu/intel/iommu.c +0 -0
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct device *dev)
       struct intel_iommu *iommu = info->iommu;
 
Quoted 9 lines

drivers/iommu/intel/iommu.c static void intel_iommu_release_device(struct device *dev) · line 4342

Quoted 4 lines
drivers/iommu/intel/iommu.c +2 -1
       mutex_lock(&iommu->iopf_lock);
-     device_rbtree_remove(info);
+     if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+             device_rbtree_remove(info);

This is also somewhat problematic because it relies on the assumption that pci_ats_supported(X) always returns the same value it returned in intel_iommu_probe_device(). That may be true today, but I don't think you should rely on it.

Agreed. The release path should not re-evaluate pci_ats_supported().

Quoted 3 lines

I think both issues could be solved if intel/iommu.c kept track of the rbtree insertion itself, e.g., with a bit in struct device_domain_info.

Sure. I will address this with a follow-up patch.

Quoted 7 lines
Quoted 7 lines
drivers/iommu/intel/iommu.c +0 -0
       mutex_unlock(&iommu->iopf_lock);
 
       if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev) &&
Signature · 1 line
2.34.1

Thanks,
baolu

#9 lore raw
Quoted 30 lines · Baolu Lu
Quoted 29 lines · Bjorn Helgaas

[+cc Danny and others because this is related to ongoing work at
https://lore.kernel.org/all/[email protected]/]

Quoted 9 lines · Lu Baolu

Commit 1a75cc710b95 ("iommu/vt-d: Use rbtree to track iommu probed devices") adds all devices probed by the iommu driver in a rbtree indexed by the source ID of each device. It assumes that each device has a unique source ID. This assumption is incorrect and the VT-d spec doesn't state this requirement either.
...

+++ b/drivers/iommu/intel/iommu.c

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4305

Quoted 10 lines
drivers/iommu/intel/iommu.c +5 -3
@@ -4299,9 +4299,11 @@ static struct iommu_device *intel_iommu_probe_device(struct device *dev)
       }
       dev_iommu_priv_set(dev, info);
-     ret = device_rbtree_insert(iommu, info);
-     if (ret)
-             goto free;
+     if (pdev && pci_ats_supported(pdev)) {
+             ret = device_rbtree_insert(iommu, info);
+             if (ret)
+                     goto free;
+     }

This change is upstream as 89436f4f5412 ("iommu/vt-d: Fix WARN_ON in iommu probe path"), but I think there's a problem here. Previously the rbtree insertion was unconditional, so the unconditional remove in the error path was correct.

Quoted 8 lines
Quoted 3 lines

After 89436f4f5412, the rbtree insertion is conditional. If intel_pasid_alloc_table() fails, we try to remove it even though it may never have been inserted:

info->node is initialized with RB_CLEAR_NODE() before the insertion, and the remove helper checks RB_EMPTY_NODE() before erasing. If insertion was skipped, removal is a no-op, so this error path does not need another state check.

Seems like RB_CLEAR_NODE() and RB_EMPTY_NODE() are lacking
documentation: RB_CLEAR_NODE() isn't used at all by lib/rbtree.c and Documentation/core-api/rbtree.rst doesn't mention it.

But I agree, this does look safe as-is, though requiring a little inside knowledge of rbtree internals.

Quoted 24 lines
Quoted 12 lines

drivers/iommu/intel/iommu.c static struct iommu_device *intel_iommu_probe_device(struct device *dev) · line 4309

Quoted 2 lines
drivers/iommu/intel/iommu.c +0 -0
       if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
               ret = intel_pasid_alloc_table(dev);
                 if (ret) {
                         dev_err(dev, "PASID table allocation failed\n");
                         goto clear_rbtree;
                 }
     ...
     clear_rbtree:
         device_rbtree_remove(info);
     free:
         kfree(info);
Quoted 11 lines

drivers/iommu/intel/iommu.c static void intel_iommu_release_device(struct device *dev) · line 4341

Quoted 6 lines
drivers/iommu/intel/iommu.c +2 -1
@@ -4336,7 +4338,8 @@ static void intel_iommu_release_device(struct device *dev)
       struct intel_iommu *iommu = info->iommu;
       mutex_lock(&iommu->iopf_lock);
-     device_rbtree_remove(info);
+     if (dev_is_pci(dev) && pci_ats_supported(to_pci_dev(dev)))
+             device_rbtree_remove(info);

This is also somewhat problematic because it relies on the assumption that pci_ats_supported(X) always returns the same value it returned in intel_iommu_probe_device(). That may be true today, but I don't think you should rely on it.

Quoted 7 lines

Agreed. The release path should not re-evaluate pci_ats_supported().

Quoted 3 lines

I think both issues could be solved if intel/iommu.c kept track of the rbtree insertion itself, e.g., with a bit in struct device_domain_info.

Sure. I will address this with a follow-up patch.

Great, thank you!

Bjorn