drivers/vfio/device_cdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
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")
Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
---
drivers/vfio/device_cdev.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/vfio/device_cdev.c b/drivers/vfio/device_cdev.c
index 480cac3a0c274..8ceca24ac136c 100644
--- a/drivers/vfio/device_cdev.c
+++ b/drivers/vfio/device_cdev.c
@@ -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
--
2.51.1.930.gacf6e81ea2-goog
On 2025/10/31 1:12, Raghavendra Rao Ananta wrote:
> 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")
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?
--- a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
+++ b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
@@ -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.
> Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
> ---
> drivers/vfio/device_cdev.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/vfio/device_cdev.c b/drivers/vfio/device_cdev.c
> index 480cac3a0c274..8ceca24ac136c 100644
> --- a/drivers/vfio/device_cdev.c
> +++ b/drivers/vfio/device_cdev.c
> @@ -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
>
On Thu, Oct 30, 2025 at 6:34 PM liulongfang <liulongfang@huawei.com> wrote:
>
> On 2025/10/31 1:12, Raghavendra Rao Ananta wrote:
> > 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")
>
> 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?
>
> --- a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
> +++ b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
> @@ -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/20251031170603.2260022-3-rananta@google.com/
(untested).
Thank you.
Raghavendra
On 2025/11/1 1:09, Raghavendra Rao Ananta wrote:
> On Thu, Oct 30, 2025 at 6:34 PM liulongfang <liulongfang@huawei.com> wrote:
>>
>> On 2025/10/31 1:12, Raghavendra Rao Ananta wrote:
>>> 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")
>>
>> 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?
>>
>> --- a/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
>> +++ b/drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c
>> @@ -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/20251031170603.2260022-3-rananta@google.com/
> (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.
> Thank you.
> Raghavendra
> .
>
On Thu, Oct 30, 2025 at 05:12:38PM +0000, Raghavendra Rao Ananta wrote:
> 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")
> Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
> ---
> drivers/vfio/device_cdev.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
Cc: stable@vger.kernel.org
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
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
On Thu, Oct 30, 2025 at 11:31 AM Jason Gunthorpe <jgg@ziepe.ca> wrote:
>
> On Thu, Oct 30, 2025 at 05:12:38PM +0000, Raghavendra Rao Ananta wrote:
> > 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")
> > Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
> > ---
> > drivers/vfio/device_cdev.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
>
> Cc: stable@vger.kernel.org
> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
>
> 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
© 2016 - 2026 Red Hat, Inc.