2022-04-01 02:38:22

by Lee Jones

Subject: [PATCH v2 1/1] drm/amdkfd: Create file descriptor after client is added to smi_clients list

This ensures userspace cannot prematurely clean-up the client before
it is fully initialised which has been proven to cause issues in the
past.

Cc: Felix Kuehling <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: "Christian König" <[email protected]>
Cc: "Pan, Xinhui" <[email protected]>
Cc: David Airlie <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
---

CAVEAT: This patch is completely untested
I can't seem to find a configuration strategy to compile test this
allyesconfig and allmodconfig do not appear sufficient

v2: Also remove Client from RCU list in error path

drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 24 +++++++++++++--------
1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index e4beebb1c80a2..f2e1d506ba211 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -247,15 +247,6 @@ int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
return ret;
}

- ret = anon_inode_getfd(kfd_smi_name, &kfd_smi_ev_fops, (void *)client,
- O_RDWR);
- if (ret < 0) {
- kfifo_free(&client->fifo);
- kfree(client);
- return ret;
- }
- *fd = ret;
-
init_waitqueue_head(&client->wait_queue);
spin_lock_init(&client->lock);
client->events = 0;
@@ -265,5 +256,20 @@ int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
list_add_rcu(&client->list, &dev->smi_clients);
spin_unlock(&dev->smi_lock);

+ ret = anon_inode_getfd(kfd_smi_name, &kfd_smi_ev_fops, (void *)client,
+ O_RDWR);
+ if (ret < 0) {
+ spin_lock(&dev->smi_lock);
+ list_del_rcu(&client->list);
+ spin_unlock(&dev->smi_lock);
+
+ synchronize_rcu();
+
+ kfifo_free(&client->fifo);
+ kfree(client);
+ return ret;
+ }
+ *fd = ret;
+
return 0;
}
--
2.35.1.1021.g381101b075-goog

2022-04-01 13:01:36

by Felix Kuehling

Subject: Re: [PATCH v2 1/1] drm/amdkfd: Create file descriptor after client is added to smi_clients list

Am 2022-03-31 um 08:21 schrieb Lee Jones:
> This ensures userspace cannot prematurely clean-up the client before
> it is fully initialised which has been proven to cause issues in the
> past.
>
> Cc: Felix Kuehling <[email protected]>
> Cc: Alex Deucher <[email protected]>
> Cc: "Christian König" <[email protected]>
> Cc: "Pan, Xinhui" <[email protected]>
> Cc: David Airlie <[email protected]>
> Cc: Daniel Vetter <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Signed-off-by: Lee Jones <[email protected]>

Reviewed-by: Felix Kuehling <[email protected]>

I will apply to patch to amd-staging-drm-next.

> ---
>
> CAVEAT: This patch is completely untested
> I can't seem to find a configuration strategy to compile test this
> allyesconfig and allmodconfig do not appear sufficient

That's very strange. It should get compiled if CONFIG_HSA_AMD is
enabled. I will at least compile-test it before I submit.

Regards,
  Felix


>
> v2: Also remove Client from RCU list in error path
>
> drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 24 +++++++++++++--------
> 1 file changed, 15 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> index e4beebb1c80a2..f2e1d506ba211 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
> @@ -247,15 +247,6 @@ int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
> return ret;
> }
>
> - ret = anon_inode_getfd(kfd_smi_name, &kfd_smi_ev_fops, (void *)client,
> - O_RDWR);
> - if (ret < 0) {
> - kfifo_free(&client->fifo);
> - kfree(client);
> - return ret;
> - }
> - *fd = ret;
> -
> init_waitqueue_head(&client->wait_queue);
> spin_lock_init(&client->lock);
> client->events = 0;
> @@ -265,5 +256,20 @@ int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
> list_add_rcu(&client->list, &dev->smi_clients);
> spin_unlock(&dev->smi_lock);
>
> + ret = anon_inode_getfd(kfd_smi_name, &kfd_smi_ev_fops, (void *)client,
> + O_RDWR);
> + if (ret < 0) {
> + spin_lock(&dev->smi_lock);
> + list_del_rcu(&client->list);
> + spin_unlock(&dev->smi_lock);
> +
> + synchronize_rcu();
> +
> + kfifo_free(&client->fifo);
> + kfree(client);
> + return ret;
> + }
> + *fd = ret;
> +
> return 0;
> }

2022-04-01 15:46:13

by Lee Jones

Subject: Re: [PATCH v2 1/1] drm/amdkfd: Create file descriptor after client is added to smi_clients list

On Thu, 31 Mar 2022, Felix Kuehling wrote:

>
> Am 2022-03-31 um 08:21 schrieb Lee Jones:
> > This ensures userspace cannot prematurely clean-up the client before
> > it is fully initialised which has been proven to cause issues in the
> > past.
> >
> > Cc: Felix Kuehling <[email protected]>
> > Cc: Alex Deucher <[email protected]>
> > Cc: "Christian König" <[email protected]>
> > Cc: "Pan, Xinhui" <[email protected]>
> > Cc: David Airlie <[email protected]>
> > Cc: Daniel Vetter <[email protected]>
> > Cc: [email protected]
> > Cc: [email protected]
> > Signed-off-by: Lee Jones <[email protected]>
>
> Reviewed-by: Felix Kuehling <[email protected]>
>
> I will apply to patch to amd-staging-drm-next.
>
> > ---
> >
> > CAVEAT: This patch is completely untested
> > I can't seem to find a configuration strategy to compile test this
> > allyesconfig and allmodconfig do not appear sufficient
>
> That's very strange. It should get compiled if CONFIG_HSA_AMD is
> enabled.

Okay, I figured it out.

Apparently you can't give 'drivers/gpu/drm/amd' as a build param.

'drivers/gpu/drm' works as I expect however.

> I will at least compile-test it before I submit.

Great, thank you.

--
Lee Jones [李琼斯]
Principal Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog