Maybe I'm doing something wrong, but my cloned repo of drm-radeon-testing is giving build errors. What I'm seeing is
[...] CC drivers/gpu/drm/radeon/radeon_gem.o drivers/gpu/drm/radeon/radeon_gem.c: In function ‘radeon_gem_create_ioctl’: drivers/gpu/drm/radeon/radeon_gem.c:221:3: error: implicit declaration of function ‘radeon_mutex_lock’ [-Werror=implicit-function-declaration] drivers/gpu/drm/radeon/radeon_gem.c:221:26: error: ‘struct radeon_device’ has no member named ‘cs_mutex’ drivers/gpu/drm/radeon/radeon_gem.c:222:3: error: implicit declaration of function ‘radeon_mutex_unlock’ [-Werror=implicit-function-declaration] drivers/gpu/drm/radeon/radeon_gem.c:222:28: error: ‘struct radeon_device’ has no member named ‘cs_mutex’ drivers/gpu/drm/radeon/radeon_gem.c: In function ‘radeon_gem_set_domain_ioctl’: drivers/gpu/drm/radeon/radeon_gem.c:268:26: error: ‘struct radeon_device’ has no member named ‘cs_mutex’ drivers/gpu/drm/radeon/radeon_gem.c:269:28: error: ‘struct radeon_device’ has no member named ‘cs_mutex’ cc1: some warnings being treated as errors make[6]: *** [drivers/gpu/drm/radeon/radeon_gem.o] Error 1
Looking around a bit, I find this commit
commit 36ff39c4045ee71cd306f8af5f8c2a1c6e998eba Author: Christian König deathsimple@vodafone.de Date: Wed May 9 10:07:08 2012 +0200
drm/radeon: replace cs_mutex with vm_mutex v3
replaces radeon_mutex_lock()/radeon_mutex_unlock() with mutex_lock()/mutex_unlock() in most places, eliminating it in others. A more recent commit
commit 0a01063d5fd7417175512f0f916629008c0a842e Author: Jerome Glisse jglisse@redhat.com Date: Wed Jun 27 14:14:21 2012 -0400
drm/radeon: disable any GPU activity after unrecovered lockup v5
introduces this hunk:
diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index f28bd4b..74176c5 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -217,6 +217,22 @@ int radeon_gem_create_ioctl(struct drm_device *dev, void *data, uint32_t handle; int r;
+ /* if in middle of gpu reset wait on the mutex, if reset + * failed don't do anything, otherwise proceed. + */ + if (!rdev->accel_working) { + /* lockup situation wait for cs mutex to be droped */ + radeon_mutex_lock(&rdev->cs_mutex); + radeon_mutex_unlock(&rdev->cs_mutex); + /* we are after a gpu reset */ + if (!rdev->accel_working) { + /* gpu reset failed, don't create anymore GPU + * object + */ + return -EBUSY; + } + } +
If it's not just something stupid I've done on my local machine, then it looks like the 0a01063d patch was applied without noticing the changes made in 36ff39c4. (If the problem is really just me, sorry for the spam/ noise.)
Dave W.