2019-09-18 17:38:44

by Allen

Subject: [PATCH] drm/amdkfd: fix a potential NULL pointer dereference

alloc_workqueue is not checked for errors and as a result,
a potential NULL dereference could occur.

Signed-off-by: Allen Pais <[email protected]>
---
drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
index c56ac47..caa82a8 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
@@ -62,6 +62,11 @@ int kfd_interrupt_init(struct kfd_dev *kfd)
}

kfd->ih_wq = alloc_workqueue("KFD IH", WQ_HIGHPRI, 1);
+ if (unlikely(!kfd->ih_wq)) {
+ fifo_free(&kfd->ih_fifo);
+ dev_err(kfd_chardev(), "Failed to allocate KFD IH workqueue\n");
+ return kfd->ih_wq;
+ }
spin_lock_init(&kfd->interrupt_lock);

INIT_WORK(&kfd->interrupt_work, interrupt_wq);
--
1.9.1

2019-09-19 22:20:18

by Felix Kuehling

Subject: Re: [PATCH] drm/amdkfd: fix a potential NULL pointer dereference

On 2019-09-18 12:30 p.m., Allen Pais wrote:
> alloc_workqueue is not checked for errors and as a result,
> a potential NULL dereference could occur.
>
> Signed-off-by: Allen Pais <[email protected]>
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
> index c56ac47..caa82a8 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
> @@ -62,6 +62,11 @@ int kfd_interrupt_init(struct kfd_dev *kfd)
> }
>
> kfd->ih_wq = alloc_workqueue("KFD IH", WQ_HIGHPRI, 1);
> + if (unlikely(!kfd->ih_wq)) {
> + fifo_free(&kfd->ih_fifo);

This does not compile. I think this should be kfifo_free.

> + dev_err(kfd_chardev(), "Failed to allocate KFD IH workqueue\n");
> + return kfd->ih_wq;

This throws a compiler warning "return makes integer from pointer
without a cast". What's worse, kfd->ih_wq is NULL here and
kfd_interrupt_init returns 0 to mean success, so returning a NULL
pointer is definitely not what you want here. This function should
return a negative error code on failure. I propose -ENOMEM.

I'm going to apply your patch with those fixes.

Regards,
  Felix


> + }
> spin_lock_init(&kfd->interrupt_lock);
>
> INIT_WORK(&kfd->interrupt_work, interrupt_wq);