Skip to contentSkip to conversation

vfio: Fix ksize arg while copying user struct in vfio_df_ioctl_bind_iommufd()

PATCH Superseded Raghavendra Rao Ananta 3 participants 6 messages Oct 30, 2025 – Nov 3, 2025

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 vfio: Fix ksize arg while copying user struct in vfio_df_ioctl_bind_iommufd()

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")
---
 drivers/vfio/device_cdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/vfio/device_cdev.c +1 -1
index 480cac3a0c274..8ceca24ac136c 100644
@@ -99,7 +99,7 @@ long vfio_df_ioctl_bind_iommufd(struct vfio_device_file *df,
 		return ret;
 	if (user_size < minsz)
 		return -EINVAL;
-	ret = copy_struct_from_user(&bind, minsz, arg, user_size);
+	ret = copy_struct_from_user(&bind, sizeof(bind), arg, user_size);
 	if (ret)
 		return ret;
 

base-commit: 3a8660878839faadb4f1a6dd72c3179c1df56787

Signature · 1 line
2.51.1.930.gacf6e81ea2-goog
Reviewed-by on v1 1/1
#1 lore raw
Quoted 11 lines · Raghavendra Rao Ananta

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Quoted 5 lines
Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")
---
 drivers/vfio/device_cdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Though I feel this was copied from some other spot in vfio so I wonder if we have a larger set of things that are a little off..

Jason

#2 lore raw
Quoted 19 lines · Jason Gunthorpe
Quoted 11 lines · Raghavendra Rao Ananta

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Quoted 5 lines
Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")
---
 drivers/vfio/device_cdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 6 lines

Though I feel this was copied from some other spot in vfio so I wonder if we have a larger set of things that are a little off..

I could only find vfio_df_ioctl_bind_iommufd() in vfio referencing copy_struct_from_user(). The other closest would be in drivers/iommu/iommufd/main.c::iommufd_fops_ioctl(), which seems to be doing the right thing.

Thank you.
Raghavendra

maintainer
#3 lore raw
In reply to Raghavendra Rao Ananta
Quoted 6 lines · Raghavendra Rao Ananta

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Quoted 6 lines

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")

Hi Ananta,

This patch also has another bug: in the hisi_acc_vfio_pci.c driver, It have two "struct vfio_device_ops" Only one of them, "hisi_acc_vfio_pci_ops" has match_token_uuid added,
while the other one, "hisi_acc_vfio_pci_migrn_ops", is missing it.
This will cause a QEMU crash (call trace) when QEMU tries to start the device.

Could you please help include this fix in your patchset as well?

drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +1 -0
@@ -1637,6 +1637,7 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
 	.mmap = hisi_acc_vfio_pci_mmap,
 	.request = vfio_pci_core_request,
 	.match = vfio_pci_core_match,
+	.match_token_uuid = vfio_pci_core_match_token_uuid,
 	.bind_iommufd = vfio_iommufd_physical_bind,
 	.unbind_iommufd = vfio_iommufd_physical_unbind,
 	.attach_ioas = vfio_iommufd_physical_attach_ioas,

Thanks.
Longfang.

Quoted 21 lines
---
 drivers/vfio/device_cdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/vfio/device_cdev.c +1 -1
index 480cac3a0c274..8ceca24ac136c 100644
@@ -99,7 +99,7 @@ long vfio_df_ioctl_bind_iommufd(struct vfio_device_file *df,
 		return ret;
 	if (user_size < minsz)
 		return -EINVAL;
-	ret = copy_struct_from_user(&bind, minsz, arg, user_size);
+	ret = copy_struct_from_user(&bind, sizeof(bind), arg, user_size);
 	if (ret)
 		return ret;
 

base-commit: 3a8660878839faadb4f1a6dd72c3179c1df56787

#4 lore raw
Quoted 28 lines · liulongfang
Quoted 6 lines · Raghavendra Rao Ananta

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Quoted 6 lines

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")

Hi Ananta,

This patch also has another bug: in the hisi_acc_vfio_pci.c driver, It have two "struct vfio_device_ops" Only one of them, "hisi_acc_vfio_pci_ops" has match_token_uuid added,
while the other one, "hisi_acc_vfio_pci_migrn_ops", is missing it.
This will cause a QEMU crash (call trace) when QEMU tries to start the device.

Could you please help include this fix in your patchset as well?

drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = { · line 1643

Quoted 7 lines
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +1 -0
@@ -1637,6 +1637,7 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
        .mmap = hisi_acc_vfio_pci_mmap,
        .request = vfio_pci_core_request,
        .match = vfio_pci_core_match,
+       .match_token_uuid = vfio_pci_core_match_token_uuid,
        .bind_iommufd = vfio_iommufd_physical_bind,
        .unbind_iommufd = vfio_iommufd_physical_unbind,
        .attach_ioas = vfio_iommufd_physical_attach_ioas,

Sent as a separate patch in v2:
https://lore.kernel.org/all/[email protected]/
(untested).

Thank you.
Raghavendra

maintainer
#5 lore raw
Quoted 25 lines · Raghavendra Rao Ananta
Quoted 24 lines · liulongfang
Quoted 6 lines · Raghavendra Rao Ananta

For the cases where user includes a non-zero value in 'token_uuid_ptr' field of 'struct vfio_device_bind_iommufd', the copy_struct_from_user() in vfio_df_ioctl_bind_iommufd() fails with -E2BIG. For the 'minsz' passed, copy_struct_from_user() expects the newly introduced field to be zero-ed, which would be incorrect in this case.

Quoted 6 lines

Fix this by passing the actual size of the kernel struct. If working with a newer userspace, copy_struct_from_user() would copy the 'token_uuid_ptr' field, and if working with an old userspace, it would zero out this field, thus still retaining backward compatibility.

Fixes 86624ba3b522 ("vfio/pci: Do vf_token checks for VFIO_DEVICE_BIND_IOMMUFD")

Hi Ananta,

This patch also has another bug: in the hisi_acc_vfio_pci.c driver, It have two "struct vfio_device_ops" Only one of them, "hisi_acc_vfio_pci_ops" has match_token_uuid added,
while the other one, "hisi_acc_vfio_pci_migrn_ops", is missing it.
This will cause a QEMU crash (call trace) when QEMU tries to start the device.

Could you please help include this fix in your patchset as well?

Quoted 15 lines

drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = { · line 1643

Quoted 11 lines
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +1 -0
@@ -1637,6 +1637,7 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
        .mmap = hisi_acc_vfio_pci_mmap,
        .request = vfio_pci_core_request,
        .match = vfio_pci_core_match,
+       .match_token_uuid = vfio_pci_core_match_token_uuid,
        .bind_iommufd = vfio_iommufd_physical_bind,
        .unbind_iommufd = vfio_iommufd_physical_unbind,
        .attach_ioas = vfio_iommufd_physical_attach_ioas,

Sent as a separate patch in v2:
https://lore.kernel.org/all/[email protected]/
(untested).

I've tested this patch locally, and after applying it, QEMU no longer fails to start and the functionality works as expected.

Thanks.
Longfang.

Quoted 4 lines

Thank you.
Raghavendra
.