2017-04-23 13:37:29

by Pan Bian

Subject: [PATCH 1/1] drm/radeon: check return value of radeon_ring_lock

From: Pan Bian <[email protected]>

Function radeon_ring_lock() returns an errno on failure, and its return
value should be validated. However, in functions r420_cp_errata_init()
and r420_cp_errata_fini(), its return value is not checked. This patch
adds the checks.

Signed-off-by: Pan Bian <[email protected]>
---
drivers/gpu/drm/radeon/r420.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c
index 2828605..a8c2b37 100644
--- a/drivers/gpu/drm/radeon/r420.c
+++ b/drivers/gpu/drm/radeon/r420.c
@@ -215,7 +215,8 @@ static void r420_cp_errata_init(struct radeon_device *rdev)
* of the CP init, apparently.
*/
radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
- radeon_ring_lock(rdev, ring, 8);
+ if (radeon_ring_lock(rdev, ring, 8))
+ return;
radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1));
radeon_ring_write(ring, rdev->config.r300.resync_scratch);
radeon_ring_write(ring, 0xDEADBEEF);
@@ -229,7 +230,8 @@ static void r420_cp_errata_fini(struct radeon_device *rdev)
/* Catch the RESYNC we dispatched all the way back,
* at the very beginning of the CP init.
*/
- radeon_ring_lock(rdev, ring, 8);
+ if (radeon_ring_lock(rdev, ring, 8))
+ return;
radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
radeon_ring_write(ring, R300_RB3D_DC_FINISH);
radeon_ring_unlock_commit(rdev, ring, false);
--
1.9.1

2017-04-24 07:47:48

by Christian König

Subject: Re: [PATCH 1/1] drm/radeon: check return value of radeon_ring_lock

Am 23.04.2017 um 15:36 schrieb Pan Bian:
> From: Pan Bian <[email protected]>
>
> Function radeon_ring_lock() returns an errno on failure, and its return
> value should be validated. However, in functions r420_cp_errata_init()
> and r420_cp_errata_fini(), its return value is not checked. This patch
> adds the checks.
>
> Signed-off-by: Pan Bian <[email protected]>
> ---
> drivers/gpu/drm/radeon/r420.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c
> index 2828605..a8c2b37 100644
> --- a/drivers/gpu/drm/radeon/r420.c
> +++ b/drivers/gpu/drm/radeon/r420.c
> @@ -215,7 +215,8 @@ static void r420_cp_errata_init(struct radeon_device *rdev)
> * of the CP init, apparently.
> */
> radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
> - radeon_ring_lock(rdev, ring, 8);
> + if (radeon_ring_lock(rdev, ring, 8))
> + return;

Nice that somebody wants to clean that up, but just returning here is
not a good idea.

Additional to that radeon_ring_lock() can only fail if we try to
allocate to many dw (impossible with only 8) or the hardware is crashed
and then it doesn't matter anyway.

I suggest to just add a WARN_ON() here.

Regards,
Christian.

> radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1));
> radeon_ring_write(ring, rdev->config.r300.resync_scratch);
> radeon_ring_write(ring, 0xDEADBEEF);
> @@ -229,7 +230,8 @@ static void r420_cp_errata_fini(struct radeon_device *rdev)
> /* Catch the RESYNC we dispatched all the way back,
> * at the very beginning of the CP init.
> */
> - radeon_ring_lock(rdev, ring, 8);
> + if (radeon_ring_lock(rdev, ring, 8))
> + return;
> radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
> radeon_ring_write(ring, R300_RB3D_DC_FINISH);
> radeon_ring_unlock_commit(rdev, ring, false);