2022-10-25 08:00:43

by Yang Li

Subject: [PATCH -next] drm/amdkfd: Fix NULL pointer dereference in svm_migrate_to_ram()

./drivers/gpu/drm/amd/amdkfd/kfd_migrate.c:985:58-62: ERROR: p is NULL but dereferenced.

Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=2549
Reported-by: Abaci Robot <[email protected]>
Signed-off-by: Yang Li <[email protected]>
---
drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
index cddf259875c0..405dd51521dc 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
@@ -981,7 +981,8 @@ static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
out_mmput:
mmput(mm);

- pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr);
+ if (p)
+ pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr);

return r ? VM_FAULT_SIGBUS : 0;
}
--
2.20.1.7.g153144c

2022-10-25 20:39:38

by Felix Kuehling

Subject: Re: [PATCH -next] drm/amdkfd: Fix NULL pointer dereference in svm_migrate_to_ram()

Am 2022-10-25 um 03:28 schrieb Yang Li:
> ./drivers/gpu/drm/amd/amdkfd/kfd_migrate.c:985:58-62: ERROR: p is NULL but dereferenced.
>
> Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=2549
> Reported-by: Abaci Robot <[email protected]>
> Signed-off-by: Yang Li <[email protected]>
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> index cddf259875c0..405dd51521dc 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> @@ -981,7 +981,8 @@ static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
> out_mmput:
> mmput(mm);
>
> - pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr);
> + if (p)
> + pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr);

Thank you for catching and reporting this problem. I think the correct
solution would be to move the pr_debug up before the kfd_unref_process
call. That way you're sure that the pointer is initialized and that it
represents a valid reference to the kfd_process structure.

Regards,
  Felix


>
> return r ? VM_FAULT_SIGBUS : 0;
> }