From: Jerome Glisse jglisse@redhat.com

After unrecovered GPU lockup avoid any GPU activities to avoid things like kernel segfault and alike to happen in any of the path that assume hw is working.

The segfault is due to PCIE vram gart table being unmapped after suspend in the GPU reset path. To avoid segault to happen and to avoid further GPU activity if unsuccessful at reseting GPU we use the accel_working boolean to transform ttm activities into noop. It does not impact the module load path because in that path ttm have an empty schedule queue and accel_working will be set to true as soon as the gart table is in valid state. Because ttm might have work queued it is better to use the accel working then disabling radeon_bo ioctl.

To trigger the segfault launch a program that repeatly create bo in ttm and let it run in background, then trigger gpu lockup from another process.

This patch also for video mode restoring on r1xx,r2xx,r3xx,r4xx, r5xx,rs4xx,rs6xx GPU even if GPU reset fail. When GPU reset fails it is very likely (so far i never had it not working) that the modesetting part of the GPU is still alive. So we can have a chance to get kernel backtrace or other debugging informations on the screen if we always restore the video mode.

v2: fix spelling error and disable accel before suspend and reenable it after pcie gart initialization to be even more cautious about possible segfault. Improve commit message v3: Improve commit message to describe the video mode restoring no matter what. v4: Avoid issue after successfull GPU lockup recovery. Don't do noop ttm move, instead report error if move needs bind or unbind or fallback to memcpy. Don't restrict new bo domain instead refuse to create new bo if gpu reset failed. Disable accel working in gart vram table unpin thus we don't change the behavior of the suspend path.

cc: stable@vger.kernel.org Signed-off-by: Jerome Glisse jglisse@redhat.com --- drivers/gpu/drm/radeon/evergreen.c | 2 +- drivers/gpu/drm/radeon/ni.c | 2 +- drivers/gpu/drm/radeon/r300.c | 2 +- drivers/gpu/drm/radeon/r520.c | 2 +- drivers/gpu/drm/radeon/r600.c | 2 +- drivers/gpu/drm/radeon/radeon_device.c | 8 +++++--- drivers/gpu/drm/radeon/radeon_gart.c | 1 + drivers/gpu/drm/radeon/radeon_gem.c | 13 +++++++++++++ drivers/gpu/drm/radeon/radeon_ttm.c | 23 +++++++++++++++++++++++ drivers/gpu/drm/radeon/rs400.c | 2 +- drivers/gpu/drm/radeon/rs600.c | 2 +- drivers/gpu/drm/radeon/rs690.c | 2 +- drivers/gpu/drm/radeon/rv515.c | 2 +- drivers/gpu/drm/radeon/rv770.c | 2 +- drivers/gpu/drm/radeon/si.c | 2 +- 15 files changed, 53 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index c3073f7..5f154e3 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -3071,6 +3071,7 @@ static int evergreen_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true; evergreen_gpu_init(rdev);

r = evergreen_blit_init(rdev); @@ -3145,7 +3146,6 @@ int evergreen_resume(struct radeon_device *rdev) /* post card */ atom_asic_init(rdev->mode_info.atom_context);

- rdev->accel_working = true; r = evergreen_startup(rdev); if (r) { DRM_ERROR("evergreen startup failed on resume\n"); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index dc2e34d..486faa8 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1245,6 +1245,7 @@ static int cayman_startup(struct radeon_device *rdev) r = cayman_pcie_gart_enable(rdev); if (r) return r; + rdev->accel_working = true; cayman_gpu_init(rdev);

r = evergreen_blit_init(rdev); @@ -1337,7 +1338,6 @@ int cayman_resume(struct radeon_device *rdev) /* post card */ atom_asic_init(rdev->mode_info.atom_context);

- rdev->accel_working = true; r = cayman_startup(rdev); if (r) { DRM_ERROR("cayman startup failed on resume\n"); diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 97722a3..206ac1f 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -1358,6 +1358,7 @@ static int r300_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true;

if (rdev->family == CHIP_R300 || rdev->family == CHIP_R350 || @@ -1426,7 +1427,6 @@ int r300_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = r300_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/r520.c b/drivers/gpu/drm/radeon/r520.c index b5cf837..6409eb0 100644 --- a/drivers/gpu/drm/radeon/r520.c +++ b/drivers/gpu/drm/radeon/r520.c @@ -181,6 +181,7 @@ static int r520_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true;

/* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -236,7 +237,6 @@ int r520_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = r520_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index bb1958d..d022533 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2394,6 +2394,7 @@ int r600_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true; r600_gpu_init(rdev); r = r600_blit_init(rdev); if (r) { @@ -2477,7 +2478,6 @@ int r600_resume(struct radeon_device *rdev) /* post card */ atom_asic_init(rdev->mode_info.atom_context);

- rdev->accel_working = true; r = r600_startup(rdev); if (r) { DRM_ERROR("r600 startup failed on resume\n"); diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 066c98b..27e6e64 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -998,11 +998,13 @@ int radeon_gpu_reset(struct radeon_device *rdev) if (!r) { dev_info(rdev->dev, "GPU reset succeed\n"); radeon_resume(rdev); - radeon_restore_bios_scratch_regs(rdev); - drm_helper_resume_force_mode(rdev->ddev); - ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched); }

+ /* no matter what restore video mode */ + radeon_restore_bios_scratch_regs(rdev); + drm_helper_resume_force_mode(rdev->ddev); + ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched); + if (r) { /* bad news, how to tell it to userspace ? */ dev_info(rdev->dev, "GPU reset failed\n"); diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c index 59d4493..c343297 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -117,6 +117,7 @@ void radeon_gart_table_vram_unpin(struct radeon_device *rdev) if (rdev->gart.robj == NULL) { return; } + rdev->accel_working = false; r = radeon_bo_reserve(rdev->gart.robj, false); if (likely(r == 0)) { radeon_bo_kunmap(rdev->gart.robj); diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index f28bd4b..ac8ba62 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -217,6 +217,19 @@ int radeon_gem_create_ioctl(struct drm_device *dev, void *data, uint32_t handle; int r;

+ 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; + } + } + /* create a gem object to contain this object in */ args->size = roundup(args->size, PAGE_SIZE); r = radeon_gem_object_create(rdev, args->size, args->alignment, diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index c94a225..287c6a8 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -399,6 +399,15 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, radeon_move_null(bo, new_mem); return 0; } + if (!rdev->accel_working) { + if (new_mem->mem_type == TTM_PL_TT || + old_mem->mem_type == TTM_PL_TT) { + /* can't bind/unbind tt */ + return -EBUSY; + } + /* try memcpy */ + goto memcpy; + } if ((old_mem->mem_type == TTM_PL_TT && new_mem->mem_type == TTM_PL_SYSTEM) || (old_mem->mem_type == TTM_PL_SYSTEM && @@ -545,6 +554,13 @@ static int radeon_ttm_backend_bind(struct ttm_tt *ttm, WARN(1, "nothing to bind %lu pages for mreg %p back %p!\n", ttm->num_pages, bo_mem, ttm); } + if (!gtt->rdev->accel_working) { + /* when accel is not working GPU is in broken state just + * do nothing for any ttm operation to avoid making the + * situation worse than it is + */ + return 0; + } r = radeon_gart_bind(gtt->rdev, gtt->offset, ttm->num_pages, ttm->pages, gtt->ttm.dma_address); if (r) { @@ -559,6 +575,13 @@ static int radeon_ttm_backend_unbind(struct ttm_tt *ttm) { struct radeon_ttm_tt *gtt = (void *)ttm;

+ if (!gtt->rdev->accel_working) { + /* when accel is not working GPU is in broken state just + * do nothing for any ttm operation to avoid making the + * situation worse than it is + */ + return 0; + } radeon_gart_unbind(gtt->rdev, gtt->offset, ttm->num_pages); return 0; } diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c index a464eb5..39fac11 100644 --- a/drivers/gpu/drm/radeon/rs400.c +++ b/drivers/gpu/drm/radeon/rs400.c @@ -404,6 +404,7 @@ static int rs400_startup(struct radeon_device *rdev) r = rs400_gart_enable(rdev); if (r) return r; + rdev->accel_working = true;

/* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -460,7 +461,6 @@ int rs400_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = rs400_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c index e95c5e6..5633203 100644 --- a/drivers/gpu/drm/radeon/rs600.c +++ b/drivers/gpu/drm/radeon/rs600.c @@ -886,6 +886,7 @@ static int rs600_startup(struct radeon_device *rdev) r = rs600_gart_enable(rdev); if (r) return r; + rdev->accel_working = true;

/* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -946,7 +947,6 @@ int rs600_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = rs600_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c index 159b6a4..633220e 100644 --- a/drivers/gpu/drm/radeon/rs690.c +++ b/drivers/gpu/drm/radeon/rs690.c @@ -615,6 +615,7 @@ static int rs690_startup(struct radeon_device *rdev) r = rs400_gart_enable(rdev); if (r) return r; + rdev->accel_working = true;

/* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -675,7 +676,6 @@ int rs690_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = rs690_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c index 7f08ced..81af328 100644 --- a/drivers/gpu/drm/radeon/rv515.c +++ b/drivers/gpu/drm/radeon/rv515.c @@ -386,6 +386,7 @@ static int rv515_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true;

/* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -441,7 +442,6 @@ int rv515_resume(struct radeon_device *rdev) /* Initialize surface registers */ radeon_surface_init(rdev);

- rdev->accel_working = true; r = rv515_startup(rdev); if (r) { rdev->accel_working = false; diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index b4f51c5..95a594c 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -910,6 +910,7 @@ static int rv770_startup(struct radeon_device *rdev) if (r) return r; } + rdev->accel_working = true;

rv770_gpu_init(rdev); r = r600_blit_init(rdev); @@ -979,7 +980,6 @@ int rv770_resume(struct radeon_device *rdev) /* post card */ atom_asic_init(rdev->mode_info.atom_context);

- rdev->accel_working = true; r = rv770_startup(rdev); if (r) { DRM_ERROR("r600 startup failed on resume\n"); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index c7b61f1..813cc34 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -3675,6 +3675,7 @@ static int si_startup(struct radeon_device *rdev) r = si_pcie_gart_enable(rdev); if (r) return r; + rdev->accel_working = true; si_gpu_init(rdev);

#if 0 @@ -3795,7 +3796,6 @@ int si_resume(struct radeon_device *rdev) /* post card */ atom_asic_init(rdev->mode_info.atom_context);

- rdev->accel_working = true; r = si_startup(rdev); if (r) { DRM_ERROR("si startup failed on resume\n");