From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?ISO-8859-1?Q?Christian_K=F6nig?= Subject: Re: drm-next radeon patches Date: Thu, 21 Jun 2012 12:35:56 +0200 Message-ID: <4FE2F90C.1080900@vodafone.de> References: <1339681530-1078-1-git-send-email-deathsimple@vodafone.de> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="------------080300090405010500040404" Return-path: Received: from outgoing.email.vodafone.de (outgoing.email.vodafone.de [139.7.28.128]) by gabe.freedesktop.org (Postfix) with ESMTP id 4C6419E75E for ; Thu, 21 Jun 2012 03:42:14 -0700 (PDT) In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dri-devel-bounces+sf-dri-devel=m.gmane.org@lists.freedesktop.org Errors-To: dri-devel-bounces+sf-dri-devel=m.gmane.org@lists.freedesktop.org To: Dave Airlie Cc: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org This is a multi-part message in MIME format. --------------080300090405010500040404 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Hi Dave, On 20.06.2012 20:53, Dave Airlie wrote: > Hi Christian, > > the 3rd patch doesnt' apply cleanly to my -next tree, > > Can you rebase them onto my current drm-core-next? sure, re-based patches are attached. Additional to that you can find them under git://people.freedesktop.org/~deathsimple/linux branch next. Cheers, Christian. > > Dave. > --------------080300090405010500040404 Content-Type: text/x-patch; name="0001-drm-radeon-remove-radeon_fence_create.patch" Content-Disposition: attachment; filename="0001-drm-radeon-remove-radeon_fence_create.patch" Content-Transfer-Encoding: quoted-printable >>From 876dc9f32907e57e0298bcd0f1607cb7a2582f63 Mon Sep 17 00:00:00 2001 From: =3D?UTF-8?q?Christian=3D20K=3DC3=3DB6nig?=3D Date: Tue, 8 May 2012 14:24:01 +0200 Subject: [PATCH 1/9] drm/radeon: remove radeon_fence_create MIME-Version: 1.0 Content-Type: text/plain; charset=3DUTF-8 Content-Transfer-Encoding: 8bit It is completely unnecessary to create fences before they are emitted, so remove it and a bunch of checks if fences are emitted or not. Signed-off-by: Christian K=C3=B6nig Reviewed-by: Jerome Glisse --- drivers/gpu/drm/radeon/evergreen.c | 2 +- drivers/gpu/drm/radeon/ni.c | 2 +- drivers/gpu/drm/radeon/r100.c | 4 +- drivers/gpu/drm/radeon/r200.c | 4 +- drivers/gpu/drm/radeon/r600.c | 4 +- drivers/gpu/drm/radeon/r600_blit_kms.c | 6 +-- drivers/gpu/drm/radeon/radeon.h | 11 +++-- drivers/gpu/drm/radeon/radeon_asic.h | 8 ++-- drivers/gpu/drm/radeon/radeon_benchmark.c | 10 +---- drivers/gpu/drm/radeon/radeon_fence.c | 42 ++++++------------ drivers/gpu/drm/radeon/radeon_ring.c | 19 +++++---- drivers/gpu/drm/radeon/radeon_sa.c | 2 +- drivers/gpu/drm/radeon/radeon_test.c | 66 ++++++++++++-----------= ------ drivers/gpu/drm/radeon/radeon_ttm.c | 30 +++++-------- drivers/gpu/drm/radeon/si.c | 6 +-- 15 files changed, 86 insertions(+), 130 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/= evergreen.c index 7fb3d2e..8e40ba4 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -1371,7 +1371,7 @@ void evergreen_mc_program(struct radeon_device *rde= v) */ void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon= _ib *ib) { - struct radeon_ring *ring =3D &rdev->ring[ib->fence->ring]; + struct radeon_ring *ring =3D &rdev->ring[ib->ring]; =20 /* set to DX10/11 mode */ radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0)); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index b7bf18e..2366be3 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -850,7 +850,7 @@ void cayman_fence_ring_emit(struct radeon_device *rde= v, =20 void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib= *ib) { - struct radeon_ring *ring =3D &rdev->ring[ib->fence->ring]; + struct radeon_ring *ring =3D &rdev->ring[ib->ring]; =20 /* set to DX10/11 mode */ radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0)); diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.= c index fb44e7e..415b7d8 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -883,7 +883,7 @@ int r100_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence) + struct radeon_fence **fence) { struct radeon_ring *ring =3D &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; uint32_t cur_pages; @@ -947,7 +947,7 @@ int r100_copy_blit(struct radeon_device *rdev, RADEON_WAIT_HOST_IDLECLEAN | RADEON_WAIT_DMA_GUI_IDLE); if (fence) { - r =3D radeon_fence_emit(rdev, fence); + r =3D radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX); } radeon_ring_unlock_commit(rdev, ring); return r; diff --git a/drivers/gpu/drm/radeon/r200.c b/drivers/gpu/drm/radeon/r200.= c index a26144d..f088925 100644 --- a/drivers/gpu/drm/radeon/r200.c +++ b/drivers/gpu/drm/radeon/r200.c @@ -85,7 +85,7 @@ int r200_copy_dma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence) + struct radeon_fence **fence) { struct radeon_ring *ring =3D &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; uint32_t size; @@ -120,7 +120,7 @@ int r200_copy_dma(struct radeon_device *rdev, radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0)); radeon_ring_write(ring, RADEON_WAIT_DMA_GUI_IDLE); if (fence) { - r =3D radeon_fence_emit(rdev, fence); + r =3D radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX); } radeon_ring_unlock_commit(rdev, ring); return r; diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.= c index bff6272..a80e61e 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2309,7 +2309,7 @@ int r600_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence) + struct radeon_fence **fence) { struct radeon_sa_bo *vb =3D NULL; int r; @@ -2607,7 +2607,7 @@ void r600_fini(struct radeon_device *rdev) */ void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *= ib) { - struct radeon_ring *ring =3D &rdev->ring[ib->fence->ring]; + struct radeon_ring *ring =3D &rdev->ring[ib->ring]; =20 /* FIXME: implement */ radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/rad= eon/r600_blit_kms.c index 03b6e0d..02f4eeb 100644 --- a/drivers/gpu/drm/radeon/r600_blit_kms.c +++ b/drivers/gpu/drm/radeon/r600_blit_kms.c @@ -703,20 +703,20 @@ int r600_blit_prepare_copy(struct radeon_device *rd= ev, unsigned num_gpu_pages, return 0; } =20 -void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= *fence, +void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= **fence, struct radeon_sa_bo *vb) { struct radeon_ring *ring =3D &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; int r; =20 - r =3D radeon_fence_emit(rdev, fence); + r =3D radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX); if (r) { radeon_ring_unlock_undo(rdev, ring); return; } =20 radeon_ring_unlock_commit(rdev, ring); - radeon_sa_bo_free(rdev, &vb, fence); + radeon_sa_bo_free(rdev, &vb, *fence); } =20 void r600_kms_blit_copy(struct radeon_device *rdev, diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h index fefcca5..e2feddd 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -113,7 +113,6 @@ extern int radeon_lockup_timeout; =20 /* fence seq are set to this number when signaled */ #define RADEON_FENCE_SIGNALED_SEQ 0LL -#define RADEON_FENCE_NOTEMITED_SEQ (~0LL) =20 /* internal ring indices */ /* r1xx+ has gfx CP ring */ @@ -277,8 +276,7 @@ struct radeon_fence { int radeon_fence_driver_start_ring(struct radeon_device *rdev, int ring)= ; int radeon_fence_driver_init(struct radeon_device *rdev); void radeon_fence_driver_fini(struct radeon_device *rdev); -int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence = **fence, int ring); -int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence *f= ence); +int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence **= fence, int ring); void radeon_fence_process(struct radeon_device *rdev, int ring); bool radeon_fence_signaled(struct radeon_fence *fence); int radeon_fence_wait(struct radeon_fence *fence, bool interruptible); @@ -630,6 +628,7 @@ struct radeon_ib { uint32_t length_dw; uint64_t gpu_addr; uint32_t *ptr; + int ring; struct radeon_fence *fence; unsigned vm_id; bool is_const_ib; @@ -1192,20 +1191,20 @@ struct radeon_asic { uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence); + struct radeon_fence **fence); u32 blit_ring_index; int (*dma)(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence); + struct radeon_fence **fence); u32 dma_ring_index; /* method used for bo copy */ int (*copy)(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence); + struct radeon_fence **fence); /* ring used for bo copies */ u32 copy_ring_index; } copy; diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeo= n/radeon_asic.h index e76a941..8cdf075 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -79,7 +79,7 @@ int r100_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence); + struct radeon_fence **fence); int r100_set_surface_reg(struct radeon_device *rdev, int reg, uint32_t tiling_flags, uint32_t pitch, uint32_t offset, uint32_t obj_size); @@ -144,7 +144,7 @@ extern int r200_copy_dma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, - struct radeon_fence *fence); + struct radeon_fence **fence); void r200_set_safe_registers(struct radeon_device *rdev); =20 /* @@ -318,7 +318,7 @@ void r600_ring_ib_execute(struct radeon_device *rdev,= struct radeon_ib *ib); int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); int r600_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, struct radeon_fence *fence); + unsigned num_gpu_pages, struct radeon_fence **fence); void r600_hpd_init(struct radeon_device *rdev); void r600_hpd_fini(struct radeon_device *rdev); bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd); @@ -364,7 +364,7 @@ void r600_hdmi_update_audio_settings(struct drm_encod= er *encoder); /* r600 blit */ int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_= pages, struct radeon_sa_bo **vb); -void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= *fence, +void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= **fence, struct radeon_sa_bo *vb); void r600_kms_blit_copy(struct radeon_device *rdev, u64 src_gpu_addr, u64 dst_gpu_addr, diff --git a/drivers/gpu/drm/radeon/radeon_benchmark.c b/drivers/gpu/drm/= radeon/radeon_benchmark.c index 364f5b1..bedda9c 100644 --- a/drivers/gpu/drm/radeon/radeon_benchmark.c +++ b/drivers/gpu/drm/radeon/radeon_benchmark.c @@ -45,20 +45,14 @@ static int radeon_benchmark_do_move(struct radeon_dev= ice *rdev, unsigned size, for (i =3D 0; i < n; i++) { switch (flag) { case RADEON_BENCHMARK_COPY_DMA: - r =3D radeon_fence_create(rdev, &fence, radeon_copy_dma_ring_index(rd= ev)); - if (r) - return r; r =3D radeon_copy_dma(rdev, saddr, daddr, size / RADEON_GPU_PAGE_SIZE, - fence); + &fence); break; case RADEON_BENCHMARK_COPY_BLIT: - r =3D radeon_fence_create(rdev, &fence, radeon_copy_blit_ring_index(r= dev)); - if (r) - return r; r =3D radeon_copy_blit(rdev, saddr, daddr, size / RADEON_GPU_PAGE_SIZE, - fence); + &fence); break; default: DRM_ERROR("Unknown copy method\n"); diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/rade= on/radeon_fence.c index 11f5f40..401d346 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -61,15 +61,21 @@ static u32 radeon_fence_read(struct radeon_device *rd= ev, int ring) return seq; } =20 -int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence *f= ence) +int radeon_fence_emit(struct radeon_device *rdev, + struct radeon_fence **fence, + int ring) { /* we are protected by the ring emission mutex */ - if (fence->seq && fence->seq < RADEON_FENCE_NOTEMITED_SEQ) { - return 0; + *fence =3D kmalloc(sizeof(struct radeon_fence), GFP_KERNEL); + if ((*fence) =3D=3D NULL) { + return -ENOMEM; } - fence->seq =3D ++rdev->fence_drv[fence->ring].seq; - radeon_fence_ring_emit(rdev, fence->ring, fence); - trace_radeon_fence_emit(rdev->ddev, fence->seq); + kref_init(&((*fence)->kref)); + (*fence)->rdev =3D rdev; + (*fence)->seq =3D ++rdev->fence_drv[ring].seq; + (*fence)->ring =3D ring; + radeon_fence_ring_emit(rdev, ring, *fence); + trace_radeon_fence_emit(rdev->ddev, (*fence)->seq); return 0; } =20 @@ -138,25 +144,9 @@ static void radeon_fence_destroy(struct kref *kref) struct radeon_fence *fence; =20 fence =3D container_of(kref, struct radeon_fence, kref); - fence->seq =3D RADEON_FENCE_NOTEMITED_SEQ; kfree(fence); } =20 -int radeon_fence_create(struct radeon_device *rdev, - struct radeon_fence **fence, - int ring) -{ - *fence =3D kmalloc(sizeof(struct radeon_fence), GFP_KERNEL); - if ((*fence) =3D=3D NULL) { - return -ENOMEM; - } - kref_init(&((*fence)->kref)); - (*fence)->rdev =3D rdev; - (*fence)->seq =3D RADEON_FENCE_NOTEMITED_SEQ; - (*fence)->ring =3D ring; - return 0; -} - static bool radeon_fence_seq_signaled(struct radeon_device *rdev, u64 seq, unsigned ring) { @@ -176,10 +166,6 @@ bool radeon_fence_signaled(struct radeon_fence *fenc= e) if (!fence) { return true; } - if (fence->seq =3D=3D RADEON_FENCE_NOTEMITED_SEQ) { - WARN(1, "Querying an unemitted fence : %p !\n", fence); - return true; - } if (fence->seq =3D=3D RADEON_FENCE_SIGNALED_SEQ) { return true; } @@ -444,9 +430,7 @@ int radeon_fence_wait_any(struct radeon_device *rdev, return 0; } =20 - if (fences[i]->seq < RADEON_FENCE_NOTEMITED_SEQ) { - seq[i] =3D fences[i]->seq; - } + seq[i] =3D fences[i]->seq; } =20 r =3D radeon_fence_wait_any_seq(rdev, seq, intr); diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeo= n/radeon_ring.c index 983658c..dd506c2 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -49,13 +49,9 @@ int radeon_ib_get(struct radeon_device *rdev, int ring= , dev_err(rdev->dev, "failed to get a new IB (%d)\n", r); return r; } - r =3D radeon_fence_create(rdev, &ib->fence, ring); - if (r) { - dev_err(rdev->dev, "failed to create fence for new IB (%d)\n", r); - radeon_sa_bo_free(rdev, &ib->sa_bo, NULL); - return r; - } =20 + ib->ring =3D ring; + ib->fence =3D NULL; ib->ptr =3D radeon_sa_bo_cpu_addr(ib->sa_bo); ib->gpu_addr =3D radeon_sa_bo_gpu_addr(ib->sa_bo); ib->vm_id =3D 0; @@ -74,7 +70,7 @@ void radeon_ib_free(struct radeon_device *rdev, struct = radeon_ib *ib) =20 int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib) { - struct radeon_ring *ring =3D &rdev->ring[ib->fence->ring]; + struct radeon_ring *ring =3D &rdev->ring[ib->ring]; int r =3D 0; =20 if (!ib->length_dw || !ring->ready) { @@ -89,8 +85,13 @@ int radeon_ib_schedule(struct radeon_device *rdev, str= uct radeon_ib *ib) dev_err(rdev->dev, "scheduling IB failed (%d).\n", r); return r; } - radeon_ring_ib_execute(rdev, ib->fence->ring, ib); - radeon_fence_emit(rdev, ib->fence); + radeon_ring_ib_execute(rdev, ib->ring, ib); + r =3D radeon_fence_emit(rdev, &ib->fence, ib->ring); + if (r) { + dev_err(rdev->dev, "failed to emit fence for new IB (%d)\n", r); + radeon_ring_unlock_undo(rdev, ring); + return r; + } radeon_ring_unlock_commit(rdev, ring); return 0; } diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/= radeon_sa.c index 32059b7..81dbb5b 100644 --- a/drivers/gpu/drm/radeon/radeon_sa.c +++ b/drivers/gpu/drm/radeon/radeon_sa.c @@ -349,7 +349,7 @@ void radeon_sa_bo_free(struct radeon_device *rdev, st= ruct radeon_sa_bo **sa_bo, =20 sa_manager =3D (*sa_bo)->manager; spin_lock(&sa_manager->lock); - if (fence && fence->seq && fence->seq < RADEON_FENCE_NOTEMITED_SEQ) { + if (fence && !radeon_fence_signaled(fence)) { (*sa_bo)->fence =3D radeon_fence_ref(fence); list_add_tail(&(*sa_bo)->flist, &sa_manager->flist[fence->ring]); diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeo= n/radeon_test.c index efff929..47e1535 100644 --- a/drivers/gpu/drm/radeon/radeon_test.c +++ b/drivers/gpu/drm/radeon/radeon_test.c @@ -106,13 +106,7 @@ void radeon_test_moves(struct radeon_device *rdev) =20 radeon_bo_kunmap(gtt_obj[i]); =20 - r =3D radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX); - if (r) { - DRM_ERROR("Failed to create GTT->VRAM fence %d\n", i); - goto out_cleanup; - } - - r =3D radeon_copy(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SI= ZE, fence); + r =3D radeon_copy(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SI= ZE, &fence); if (r) { DRM_ERROR("Failed GTT->VRAM copy %d\n", i); goto out_cleanup; @@ -155,13 +149,7 @@ void radeon_test_moves(struct radeon_device *rdev) =20 radeon_bo_kunmap(vram_obj); =20 - r =3D radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX); - if (r) { - DRM_ERROR("Failed to create VRAM->GTT fence %d\n", i); - goto out_cleanup; - } - - r =3D radeon_copy(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SI= ZE, fence); + r =3D radeon_copy(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SI= ZE, &fence); if (r) { DRM_ERROR("Failed VRAM->GTT copy %d\n", i); goto out_cleanup; @@ -245,17 +233,6 @@ void radeon_test_ring_sync(struct radeon_device *rde= v, int ridxB =3D radeon_ring_index(rdev, ringB); int r; =20 - r =3D radeon_fence_create(rdev, &fence1, ridxA); - if (r) { - DRM_ERROR("Failed to create sync fence 1\n"); - goto out_cleanup; - } - r =3D radeon_fence_create(rdev, &fence2, ridxA); - if (r) { - DRM_ERROR("Failed to create sync fence 2\n"); - goto out_cleanup; - } - r =3D radeon_semaphore_create(rdev, &semaphore); if (r) { DRM_ERROR("Failed to create semaphore\n"); @@ -268,9 +245,19 @@ void radeon_test_ring_sync(struct radeon_device *rde= v, goto out_cleanup; } radeon_semaphore_emit_wait(rdev, ridxA, semaphore); - radeon_fence_emit(rdev, fence1); + r =3D radeon_fence_emit(rdev, &fence1, ridxA); + if (r) { + DRM_ERROR("Failed to emit fence 1\n"); + radeon_ring_unlock_undo(rdev, ringA); + goto out_cleanup; + } radeon_semaphore_emit_wait(rdev, ridxA, semaphore); - radeon_fence_emit(rdev, fence2); + r =3D radeon_fence_emit(rdev, &fence2, ridxA); + if (r) { + DRM_ERROR("Failed to emit fence 2\n"); + radeon_ring_unlock_undo(rdev, ringA); + goto out_cleanup; + } radeon_ring_unlock_commit(rdev, ringA); =20 mdelay(1000); @@ -342,17 +329,6 @@ void radeon_test_ring_sync2(struct radeon_device *rd= ev, bool sigA, sigB; int i, r; =20 - r =3D radeon_fence_create(rdev, &fenceA, ridxA); - if (r) { - DRM_ERROR("Failed to create sync fence 1\n"); - goto out_cleanup; - } - r =3D radeon_fence_create(rdev, &fenceB, ridxB); - if (r) { - DRM_ERROR("Failed to create sync fence 2\n"); - goto out_cleanup; - } - r =3D radeon_semaphore_create(rdev, &semaphore); if (r) { DRM_ERROR("Failed to create semaphore\n"); @@ -365,7 +341,12 @@ void radeon_test_ring_sync2(struct radeon_device *rd= ev, goto out_cleanup; } radeon_semaphore_emit_wait(rdev, ridxA, semaphore); - radeon_fence_emit(rdev, fenceA); + r =3D radeon_fence_emit(rdev, &fenceA, ridxA); + if (r) { + DRM_ERROR("Failed to emit sync fence 1\n"); + radeon_ring_unlock_undo(rdev, ringA); + goto out_cleanup; + } radeon_ring_unlock_commit(rdev, ringA); =20 r =3D radeon_ring_lock(rdev, ringB, 64); @@ -374,7 +355,12 @@ void radeon_test_ring_sync2(struct radeon_device *rd= ev, goto out_cleanup; } radeon_semaphore_emit_wait(rdev, ridxB, semaphore); - radeon_fence_emit(rdev, fenceB); + r =3D radeon_fence_emit(rdev, &fenceB, ridxB); + if (r) { + DRM_ERROR("Failed to create sync fence 2\n"); + radeon_ring_unlock_undo(rdev, ringB); + goto out_cleanup; + } radeon_ring_unlock_commit(rdev, ringB); =20 mdelay(1000); diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon= /radeon_ttm.c index c94a225..2d36bdd 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -222,15 +222,12 @@ static int radeon_move_blit(struct ttm_buffer_objec= t *bo, { struct radeon_device *rdev; uint64_t old_start, new_start; - struct radeon_fence *fence, *old_fence; + struct radeon_fence *fence; struct radeon_semaphore *sem =3D NULL; - int r; + int r, ridx; =20 rdev =3D radeon_get_rdev(bo->bdev); - r =3D radeon_fence_create(rdev, &fence, radeon_copy_ring_index(rdev)); - if (unlikely(r)) { - return r; - } + ridx =3D radeon_copy_ring_index(rdev); old_start =3D old_mem->start << PAGE_SHIFT; new_start =3D new_mem->start << PAGE_SHIFT; =20 @@ -243,7 +240,6 @@ static int radeon_move_blit(struct ttm_buffer_object = *bo, break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); - radeon_fence_unref(&fence); return -EINVAL; } switch (new_mem->mem_type) { @@ -255,42 +251,38 @@ static int radeon_move_blit(struct ttm_buffer_objec= t *bo, break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); - radeon_fence_unref(&fence); return -EINVAL; } - if (!rdev->ring[radeon_copy_ring_index(rdev)].ready) { + if (!rdev->ring[ridx].ready) { DRM_ERROR("Trying to move memory with ring turned off.\n"); - radeon_fence_unref(&fence); return -EINVAL; } =20 BUILD_BUG_ON((PAGE_SIZE % RADEON_GPU_PAGE_SIZE) !=3D 0); =20 /* sync other rings */ - old_fence =3D bo->sync_obj; - if (old_fence && old_fence->ring !=3D fence->ring - && !radeon_fence_signaled(old_fence)) { + fence =3D bo->sync_obj; + if (fence && fence->ring !=3D ridx + && !radeon_fence_signaled(fence)) { bool sync_to_ring[RADEON_NUM_RINGS] =3D { }; - sync_to_ring[old_fence->ring] =3D true; + sync_to_ring[fence->ring] =3D true; =20 r =3D radeon_semaphore_create(rdev, &sem); if (r) { - radeon_fence_unref(&fence); return r; } =20 - r =3D radeon_semaphore_sync_rings(rdev, sem, - sync_to_ring, fence->ring); + r =3D radeon_semaphore_sync_rings(rdev, sem, sync_to_ring, ridx); if (r) { radeon_semaphore_free(rdev, sem, NULL); - radeon_fence_unref(&fence); return r; } } =20 + fence =3D NULL; r =3D radeon_copy(rdev, old_start, new_start, new_mem->num_pages * (PAGE_SIZE / RADEON_GPU_PAGE_SIZE), /* GPU pages= */ - fence); + &fence); /* FIXME: handle copy error */ r =3D ttm_bo_move_accel_cleanup(bo, (void *)fence, NULL, evict, no_wait_reserve, no_wait_gpu, new_mem); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index c7b61f1..8868a1f 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -1762,7 +1762,7 @@ void si_fence_ring_emit(struct radeon_device *rdev, */ void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib= ) { - struct radeon_ring *ring =3D &rdev->ring[ib->fence->ring]; + struct radeon_ring *ring =3D &rdev->ring[ib->ring]; u32 header; =20 if (ib->is_const_ib) @@ -2702,7 +2702,7 @@ int si_ib_parse(struct radeon_device *rdev, struct = radeon_ib *ib) if (ib->is_const_ib) ret =3D si_vm_packet3_ce_check(rdev, ib->ptr, &pkt); else { - switch (ib->fence->ring) { + switch (ib->ring) { case RADEON_RING_TYPE_GFX_INDEX: ret =3D si_vm_packet3_gfx_check(rdev, ib->ptr, &pkt); break; @@ -2711,7 +2711,7 @@ int si_ib_parse(struct radeon_device *rdev, struct = radeon_ib *ib) ret =3D si_vm_packet3_compute_check(rdev, ib->ptr, &pkt); break; default: - dev_err(rdev->dev, "Non-PM4 ring %d !\n", ib->fence->ring); + dev_err(rdev->dev, "Non-PM4 ring %d !\n", ib->ring); ret =3D -EINVAL; break; } --=20 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0002-drm-radeon-add-infrastructure-for-advanced-ring-sync.patch" Content-Disposition: attachment; filename*0="0002-drm-radeon-add-infrastructure-for-advanced-ring-sync.pa"; filename*1="tch" Content-Transfer-Encoding: quoted-printable >>From 68e250b7c281dbb75ea2a892a7d4ca27f974fc91 Mon Sep 17 00:00:00 2001 From: =3D?UTF-8?q?Christian=3D20K=3DC3=3DB6nig?=3D Date: Thu, 10 May 2012 15:57:31 +0200 Subject: [PATCH 2/9] drm/radeon: add infrastructure for advanced ring synchronization v2 MIME-Version: 1.0 Content-Type: text/plain; charset=3DUTF-8 Content-Transfer-Encoding: 8bit v2: BUG_ON not matching rings. Signed-off-by: Christian K=C3=B6nig Reviewed-by: Jerome Glisse --- drivers/gpu/drm/radeon/radeon.h | 25 ++++++++++- drivers/gpu/drm/radeon/radeon_fence.c | 73 +++++++++++++++++++++++++++= ++---- 2 files changed, 87 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h index e2feddd..33a72dc 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -257,8 +257,8 @@ struct radeon_fence_driver { uint32_t scratch_reg; uint64_t gpu_addr; volatile uint32_t *cpu_addr; - /* seq is protected by ring emission lock */ - uint64_t seq; + /* sync_seq is protected by ring emission lock */ + uint64_t sync_seq[RADEON_NUM_RINGS]; atomic64_t last_seq; unsigned long last_activity; bool initialized; @@ -288,6 +288,27 @@ int radeon_fence_wait_any(struct radeon_device *rdev= , struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence); void radeon_fence_unref(struct radeon_fence **fence); unsigned radeon_fence_count_emitted(struct radeon_device *rdev, int ring= ); +bool radeon_fence_need_sync(struct radeon_fence *fence, int ring); +void radeon_fence_note_sync(struct radeon_fence *fence, int ring); +static inline struct radeon_fence *radeon_fence_later(struct radeon_fenc= e *a, + struct radeon_fence *b) +{ + if (!a) { + return b; + } + + if (!b) { + return a; + } + + BUG_ON(a->ring !=3D b->ring); + + if (a->seq > b->seq) { + return a; + } else { + return b; + } +} =20 /* * Tiling registers diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/rade= on/radeon_fence.c index 401d346..7b55625 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -72,7 +72,7 @@ int radeon_fence_emit(struct radeon_device *rdev, } kref_init(&((*fence)->kref)); (*fence)->rdev =3D rdev; - (*fence)->seq =3D ++rdev->fence_drv[ring].seq; + (*fence)->seq =3D ++rdev->fence_drv[ring].sync_seq[ring]; (*fence)->ring =3D ring; radeon_fence_ring_emit(rdev, ring, *fence); trace_radeon_fence_emit(rdev->ddev, (*fence)->seq); @@ -449,7 +449,7 @@ int radeon_fence_wait_next_locked(struct radeon_devic= e *rdev, int ring) * wait. */ seq =3D atomic64_read(&rdev->fence_drv[ring].last_seq) + 1ULL; - if (seq >=3D rdev->fence_drv[ring].seq) { + if (seq >=3D rdev->fence_drv[ring].sync_seq[ring]) { /* nothing to wait for, last_seq is already the last emited fence */ return -ENOENT; @@ -464,7 +464,7 @@ int radeon_fence_wait_empty_locked(struct radeon_devi= ce *rdev, int ring) * activity can be scheduled so there won't be concurrent access * to seq value. */ - return radeon_fence_wait_seq(rdev, rdev->fence_drv[ring].seq, + return radeon_fence_wait_seq(rdev, rdev->fence_drv[ring].sync_seq[ring]= , ring, false, false); } =20 @@ -492,7 +492,8 @@ unsigned radeon_fence_count_emitted(struct radeon_dev= ice *rdev, int ring) * but it's ok to report slightly wrong fence count here. */ radeon_fence_process(rdev, ring); - emitted =3D rdev->fence_drv[ring].seq - atomic64_read(&rdev->fence_drv[= ring].last_seq); + emitted =3D rdev->fence_drv[ring].sync_seq[ring] + - atomic64_read(&rdev->fence_drv[ring].last_seq); /* to avoid 32bits warp around */ if (emitted > 0x10000000) { emitted =3D 0x10000000; @@ -500,6 +501,51 @@ unsigned radeon_fence_count_emitted(struct radeon_de= vice *rdev, int ring) return (unsigned)emitted; } =20 +bool radeon_fence_need_sync(struct radeon_fence *fence, int dst_ring) +{ + struct radeon_fence_driver *fdrv; + + if (!fence) { + return false; + } + + if (fence->ring =3D=3D dst_ring) { + return false; + } + + /* we are protected by the ring mutex */ + fdrv =3D &fence->rdev->fence_drv[dst_ring]; + if (fence->seq <=3D fdrv->sync_seq[fence->ring]) { + return false; + } + + return true; +} + +void radeon_fence_note_sync(struct radeon_fence *fence, int dst_ring) +{ + struct radeon_fence_driver *dst, *src; + unsigned i; + + if (!fence) { + return; + } + + if (fence->ring =3D=3D dst_ring) { + return; + } + + /* we are protected by the ring mutex */ + src =3D &fence->rdev->fence_drv[fence->ring]; + dst =3D &fence->rdev->fence_drv[dst_ring]; + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { + if (i =3D=3D dst_ring) { + continue; + } + dst->sync_seq[i] =3D max(dst->sync_seq[i], src->sync_seq[i]); + } +} + int radeon_fence_driver_start_ring(struct radeon_device *rdev, int ring) { uint64_t index; @@ -521,7 +567,7 @@ int radeon_fence_driver_start_ring(struct radeon_devi= ce *rdev, int ring) } rdev->fence_drv[ring].cpu_addr =3D &rdev->wb.wb[index/4]; rdev->fence_drv[ring].gpu_addr =3D rdev->wb.gpu_addr + index; - radeon_fence_write(rdev, rdev->fence_drv[ring].seq, ring); + radeon_fence_write(rdev, rdev->fence_drv[ring].sync_seq[ring], ring); rdev->fence_drv[ring].initialized =3D true; dev_info(rdev->dev, "fence driver on ring %d use gpu addr 0x%016llx and= cpu addr 0x%p\n", ring, rdev->fence_drv[ring].gpu_addr, rdev->fence_drv[ring].cpu_addr)= ; @@ -530,10 +576,13 @@ int radeon_fence_driver_start_ring(struct radeon_de= vice *rdev, int ring) =20 static void radeon_fence_driver_init_ring(struct radeon_device *rdev, in= t ring) { + int i; + rdev->fence_drv[ring].scratch_reg =3D -1; rdev->fence_drv[ring].cpu_addr =3D NULL; rdev->fence_drv[ring].gpu_addr =3D 0; - rdev->fence_drv[ring].seq =3D 0; + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) + rdev->fence_drv[ring].sync_seq[i] =3D 0; atomic64_set(&rdev->fence_drv[ring].last_seq, 0); rdev->fence_drv[ring].last_activity =3D jiffies; rdev->fence_drv[ring].initialized =3D false; @@ -579,7 +628,7 @@ static int radeon_debugfs_fence_info(struct seq_file = *m, void *data) struct drm_info_node *node =3D (struct drm_info_node *)m->private; struct drm_device *dev =3D node->minor->dev; struct radeon_device *rdev =3D dev->dev_private; - int i; + int i, j; =20 for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { if (!rdev->fence_drv[i].initialized) @@ -588,8 +637,14 @@ static int radeon_debugfs_fence_info(struct seq_file= *m, void *data) seq_printf(m, "--- ring %d ---\n", i); seq_printf(m, "Last signaled fence 0x%016llx\n", (unsigned long long)atomic64_read(&rdev->fence_drv[i].last_seq)); - seq_printf(m, "Last emitted 0x%016llx\n", - rdev->fence_drv[i].seq); + seq_printf(m, "Last emitted 0x%016llx\n", + rdev->fence_drv[i].sync_seq[i]); + + for (j =3D 0; j < RADEON_NUM_RINGS; ++j) { + if (i !=3D j && rdev->fence_drv[j].initialized) + seq_printf(m, "Last sync to ring %d 0x%016llx\n", + j, rdev->fence_drv[i].sync_seq[j]); + } } return 0; } --=20 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0003-drm-radeon-rework-ring-syncing-code.patch" Content-Disposition: attachment; filename="0003-drm-radeon-rework-ring-syncing-code.patch" Content-Transfer-Encoding: quoted-printable >>From 220907d9835ce5181b9f782c862b1ee7a4d24c23 Mon Sep 17 00:00:00 2001 From: =3D?UTF-8?q?Christian=3D20K=3DC3=3DB6nig?=3D Date: Thu, 10 May 2012 16:46:43 +0200 Subject: [PATCH 3/9] drm/radeon: rework ring syncing code MIME-Version: 1.0 Content-Type: text/plain; charset=3DUTF-8 Content-Transfer-Encoding: 8bit Move inter ring syncing with semaphores into the existing ring allocations, with that we need to lock the ring mutex only once. Signed-off-by: Christian K=C3=B6nig Reviewed-by: Jerome Glisse --- drivers/gpu/drm/radeon/evergreen_blit_kms.c | 3 +- drivers/gpu/drm/radeon/r600.c | 5 +- drivers/gpu/drm/radeon/r600_blit_kms.c | 24 +++++++-- drivers/gpu/drm/radeon/radeon.h | 6 +-- drivers/gpu/drm/radeon/radeon_asic.h | 5 +- drivers/gpu/drm/radeon/radeon_cs.c | 38 +++----------- drivers/gpu/drm/radeon/radeon_ring.c | 30 +++++++++-- drivers/gpu/drm/radeon/radeon_semaphore.c | 71 ++++++++++-----------= ------ drivers/gpu/drm/radeon/radeon_test.c | 6 +-- drivers/gpu/drm/radeon/radeon_ttm.c | 20 -------- 10 files changed, 92 insertions(+), 116 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen_blit_kms.c b/drivers/gpu/dr= m/radeon/evergreen_blit_kms.c index 1e96bd4..e512560 100644 --- a/drivers/gpu/drm/radeon/evergreen_blit_kms.c +++ b/drivers/gpu/drm/radeon/evergreen_blit_kms.c @@ -622,7 +622,8 @@ int evergreen_blit_init(struct radeon_device *rdev) rdev->r600_blit.primitives.draw_auto =3D draw_auto; rdev->r600_blit.primitives.set_default_state =3D set_default_state; =20 - rdev->r600_blit.ring_size_common =3D 55; /* shaders + def state */ + rdev->r600_blit.ring_size_common =3D 8; /* sync semaphore */ + rdev->r600_blit.ring_size_common +=3D 55; /* shaders + def state */ rdev->r600_blit.ring_size_common +=3D 16; /* fence emit for VB IB */ rdev->r600_blit.ring_size_common +=3D 5; /* done copy */ rdev->r600_blit.ring_size_common +=3D 16; /* fence emit for done copy *= / diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.= c index a80e61e..aced97b 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2311,15 +2311,16 @@ int r600_copy_blit(struct radeon_device *rdev, unsigned num_gpu_pages, struct radeon_fence **fence) { + struct radeon_semaphore *sem =3D NULL; struct radeon_sa_bo *vb =3D NULL; int r; =20 - r =3D r600_blit_prepare_copy(rdev, num_gpu_pages, &vb); + r =3D r600_blit_prepare_copy(rdev, num_gpu_pages, fence, &vb, &sem); if (r) { return r; } r600_kms_blit_copy(rdev, src_offset, dst_offset, num_gpu_pages, vb); - r600_blit_done_copy(rdev, fence, vb); + r600_blit_done_copy(rdev, fence, vb, sem); return 0; } =20 diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/rad= eon/r600_blit_kms.c index 02f4eeb..2b8d641 100644 --- a/drivers/gpu/drm/radeon/r600_blit_kms.c +++ b/drivers/gpu/drm/radeon/r600_blit_kms.c @@ -512,7 +512,8 @@ int r600_blit_init(struct radeon_device *rdev) rdev->r600_blit.primitives.draw_auto =3D draw_auto; rdev->r600_blit.primitives.set_default_state =3D set_default_state; =20 - rdev->r600_blit.ring_size_common =3D 40; /* shaders + def state */ + rdev->r600_blit.ring_size_common =3D 8; /* sync semaphore */ + rdev->r600_blit.ring_size_common +=3D 40; /* shaders + def state */ rdev->r600_blit.ring_size_common +=3D 5; /* done copy */ rdev->r600_blit.ring_size_common +=3D 16; /* fence emit for done copy *= / =20 @@ -666,7 +667,8 @@ static unsigned r600_blit_create_rect(unsigned num_gp= u_pages, =20 =20 int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_= pages, - struct radeon_sa_bo **vb) + struct radeon_fence **fence, struct radeon_sa_bo **vb, + struct radeon_semaphore **sem) { struct radeon_ring *ring =3D &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; int r; @@ -689,22 +691,37 @@ int r600_blit_prepare_copy(struct radeon_device *rd= ev, unsigned num_gpu_pages, return r; } =20 + r =3D radeon_semaphore_create(rdev, sem); + if (r) { + radeon_sa_bo_free(rdev, vb, NULL); + return r; + } + /* calculate number of loops correctly */ ring_size =3D num_loops * dwords_per_loop; ring_size +=3D rdev->r600_blit.ring_size_common; r =3D radeon_ring_lock(rdev, ring, ring_size); if (r) { radeon_sa_bo_free(rdev, vb, NULL); + radeon_semaphore_free(rdev, sem, NULL); return r; } =20 + if (radeon_fence_need_sync(*fence, RADEON_RING_TYPE_GFX_INDEX)) { + radeon_semaphore_sync_rings(rdev, *sem, (*fence)->ring, + RADEON_RING_TYPE_GFX_INDEX); + radeon_fence_note_sync(*fence, RADEON_RING_TYPE_GFX_INDEX); + } else { + radeon_semaphore_free(rdev, sem, NULL); + } + rdev->r600_blit.primitives.set_default_state(rdev); rdev->r600_blit.primitives.set_shaders(rdev); return 0; } =20 void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= **fence, - struct radeon_sa_bo *vb) + struct radeon_sa_bo *vb, struct radeon_semaphore *sem) { struct radeon_ring *ring =3D &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; int r; @@ -717,6 +734,7 @@ void r600_blit_done_copy(struct radeon_device *rdev, = struct radeon_fence **fence =20 radeon_ring_unlock_commit(rdev, ring); radeon_sa_bo_free(rdev, &vb, *fence); + radeon_semaphore_free(rdev, &sem, *fence); } =20 void r600_kms_blit_copy(struct radeon_device *rdev, diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h index 33a72dc..4563e50 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -470,10 +470,9 @@ void radeon_semaphore_emit_wait(struct radeon_device= *rdev, int ring, struct radeon_semaphore *semaphore); int radeon_semaphore_sync_rings(struct radeon_device *rdev, struct radeon_semaphore *semaphore, - bool sync_to[RADEON_NUM_RINGS], - int dst_ring); + int signaler, int waiter); void radeon_semaphore_free(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, + struct radeon_semaphore **semaphore, struct radeon_fence *fence); =20 /* @@ -653,6 +652,7 @@ struct radeon_ib { struct radeon_fence *fence; unsigned vm_id; bool is_const_ib; + struct radeon_fence *sync_to[RADEON_NUM_RINGS]; struct radeon_semaphore *semaphore; }; =20 diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeo= n/radeon_asic.h index 8cdf075..94c427a 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -363,9 +363,10 @@ int r600_hdmi_buffer_status_changed(struct drm_encod= er *encoder); void r600_hdmi_update_audio_settings(struct drm_encoder *encoder); /* r600 blit */ int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_= pages, - struct radeon_sa_bo **vb); + struct radeon_fence **fence, struct radeon_sa_bo **vb, + struct radeon_semaphore **sem); void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence= **fence, - struct radeon_sa_bo *vb); + struct radeon_sa_bo *vb, struct radeon_semaphore *sem); void r600_kms_blit_copy(struct radeon_device *rdev, u64 src_gpu_addr, u64 dst_gpu_addr, unsigned num_gpu_pages, diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/= radeon_cs.c index 142f894..dd3e234 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -115,36 +115,20 @@ static int radeon_cs_get_ring(struct radeon_cs_pars= er *p, u32 ring, s32 priority return 0; } =20 -static int radeon_cs_sync_rings(struct radeon_cs_parser *p) +static void radeon_cs_sync_rings(struct radeon_cs_parser *p) { - bool sync_to_ring[RADEON_NUM_RINGS] =3D { }; - bool need_sync =3D false; - int i, r; + int i; =20 for (i =3D 0; i < p->nrelocs; i++) { - struct radeon_fence *fence; + struct radeon_fence *a, *b; =20 if (!p->relocs[i].robj || !p->relocs[i].robj->tbo.sync_obj) continue; =20 - fence =3D p->relocs[i].robj->tbo.sync_obj; - if (fence->ring !=3D p->ring && !radeon_fence_signaled(fence)) { - sync_to_ring[fence->ring] =3D true; - need_sync =3D true; - } - } - - if (!need_sync) { - return 0; - } - - r =3D radeon_semaphore_create(p->rdev, &p->ib.semaphore); - if (r) { - return r; + a =3D p->relocs[i].robj->tbo.sync_obj; + b =3D p->ib.sync_to[a->ring]; + p->ib.sync_to[a->ring] =3D radeon_fence_later(a, b); } - - return radeon_semaphore_sync_rings(p->rdev, p->ib.semaphore, - sync_to_ring, p->ring); } =20 /* XXX: note that this is called from the legacy UMS CS ioctl as well */ @@ -368,10 +352,7 @@ static int radeon_cs_ib_chunk(struct radeon_device *= rdev, DRM_ERROR("Invalid command stream !\n"); return r; } - r =3D radeon_cs_sync_rings(parser); - if (r) { - DRM_ERROR("Failed to synchronize rings !\n"); - } + radeon_cs_sync_rings(parser); parser->ib.vm_id =3D 0; r =3D radeon_ib_schedule(rdev, &parser->ib); if (r) { @@ -468,10 +449,7 @@ static int radeon_cs_ib_vm_chunk(struct radeon_devic= e *rdev, if (r) { goto out; } - r =3D radeon_cs_sync_rings(parser); - if (r) { - DRM_ERROR("Failed to synchronize rings !\n"); - } + radeon_cs_sync_rings(parser); =20 if ((rdev->family >=3D CHIP_TAHITI) && (parser->chunk_const_ib_idx !=3D -1)) { diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeo= n/radeon_ring.c index dd506c2..0826e77 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -42,7 +42,7 @@ int radeon_debugfs_sa_init(struct radeon_device *rdev); int radeon_ib_get(struct radeon_device *rdev, int ring, struct radeon_ib *ib, unsigned size) { - int r; + int i, r; =20 r =3D radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, &ib->sa_bo, size, 256,= true); if (r) { @@ -50,20 +50,26 @@ int radeon_ib_get(struct radeon_device *rdev, int rin= g, return r; } =20 + r =3D radeon_semaphore_create(rdev, &ib->semaphore); + if (r) { + return r; + } + ib->ring =3D ring; ib->fence =3D NULL; ib->ptr =3D radeon_sa_bo_cpu_addr(ib->sa_bo); ib->gpu_addr =3D radeon_sa_bo_gpu_addr(ib->sa_bo); ib->vm_id =3D 0; ib->is_const_ib =3D false; - ib->semaphore =3D NULL; + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) + ib->sync_to[i] =3D NULL; =20 return 0; } =20 void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib) { - radeon_semaphore_free(rdev, ib->semaphore, ib->fence); + radeon_semaphore_free(rdev, &ib->semaphore, ib->fence); radeon_sa_bo_free(rdev, &ib->sa_bo, ib->fence); radeon_fence_unref(&ib->fence); } @@ -71,7 +77,8 @@ void radeon_ib_free(struct radeon_device *rdev, struct = radeon_ib *ib) int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib) { struct radeon_ring *ring =3D &rdev->ring[ib->ring]; - int r =3D 0; + bool need_sync =3D false; + int i, r =3D 0; =20 if (!ib->length_dw || !ring->ready) { /* TODO: Nothings in the ib we should report. */ @@ -80,11 +87,24 @@ int radeon_ib_schedule(struct radeon_device *rdev, st= ruct radeon_ib *ib) } =20 /* 64 dwords should be enough for fence too */ - r =3D radeon_ring_lock(rdev, ring, 64); + r =3D radeon_ring_lock(rdev, ring, 64 + RADEON_NUM_RINGS * 8); if (r) { dev_err(rdev->dev, "scheduling IB failed (%d).\n", r); return r; } + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { + struct radeon_fence *fence =3D ib->sync_to[i]; + if (radeon_fence_need_sync(fence, ib->ring)) { + need_sync =3D true; + radeon_semaphore_sync_rings(rdev, ib->semaphore, + fence->ring, ib->ring); + radeon_fence_note_sync(fence, ib->ring); + } + } + /* immediately free semaphore when we don't need to sync */ + if (!need_sync) { + radeon_semaphore_free(rdev, &ib->semaphore, NULL); + } radeon_ring_ib_execute(rdev, ib->ring, ib); r =3D radeon_fence_emit(rdev, &ib->fence, ib->ring); if (r) { diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/= radeon/radeon_semaphore.c index e2ace5d..7cc78de 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -68,70 +68,49 @@ void radeon_semaphore_emit_wait(struct radeon_device = *rdev, int ring, radeon_semaphore_ring_emit(rdev, ring, &rdev->ring[ring], semaphore, tr= ue); } =20 +/* caller must hold ring lock */ int radeon_semaphore_sync_rings(struct radeon_device *rdev, struct radeon_semaphore *semaphore, - bool sync_to[RADEON_NUM_RINGS], - int dst_ring) + int signaler, int waiter) { - int i =3D 0, r; + int r; =20 - mutex_lock(&rdev->ring_lock); - r =3D radeon_ring_alloc(rdev, &rdev->ring[dst_ring], RADEON_NUM_RINGS *= 8); - if (r) { - goto error; + /* no need to signal and wait on the same ring */ + if (signaler =3D=3D waiter) { + return 0; } =20 - for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { - /* no need to sync to our own or unused rings */ - if (!sync_to[i] || i =3D=3D dst_ring) - continue; - - /* prevent GPU deadlocks */ - if (!rdev->ring[i].ready) { - dev_err(rdev->dev, "Trying to sync to a disabled ring!"); - r =3D -EINVAL; - goto error; - } - - r =3D radeon_ring_alloc(rdev, &rdev->ring[i], 8); - if (r) { - goto error; - } - - radeon_semaphore_emit_signal(rdev, i, semaphore); - radeon_semaphore_emit_wait(rdev, dst_ring, semaphore); + /* prevent GPU deadlocks */ + if (!rdev->ring[signaler].ready) { + dev_err(rdev->dev, "Trying to sync to a disabled ring!"); + return -EINVAL; + } =20 - radeon_ring_commit(rdev, &rdev->ring[i]); + r =3D radeon_ring_alloc(rdev, &rdev->ring[signaler], 8); + if (r) { + return r; } + radeon_semaphore_emit_signal(rdev, signaler, semaphore); + radeon_ring_commit(rdev, &rdev->ring[signaler]); =20 - radeon_ring_commit(rdev, &rdev->ring[dst_ring]); - mutex_unlock(&rdev->ring_lock); + /* we assume caller has already allocated space on waiters ring */ + radeon_semaphore_emit_wait(rdev, waiter, semaphore); =20 return 0; - -error: - /* unlock all locks taken so far */ - for (--i; i >=3D 0; --i) { - if (sync_to[i] || i =3D=3D dst_ring) { - radeon_ring_undo(&rdev->ring[i]); - } - } - radeon_ring_undo(&rdev->ring[dst_ring]); - mutex_unlock(&rdev->ring_lock); - return r; } =20 void radeon_semaphore_free(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, + struct radeon_semaphore **semaphore, struct radeon_fence *fence) { - if (semaphore =3D=3D NULL) { + if (semaphore =3D=3D NULL || *semaphore =3D=3D NULL) { return; } - if (semaphore->waiters > 0) { + if ((*semaphore)->waiters > 0) { dev_err(rdev->dev, "semaphore %p has more waiters than signalers," - " hardware lockup imminent!\n", semaphore); + " hardware lockup imminent!\n", *semaphore); } - radeon_sa_bo_free(rdev, &semaphore->sa_bo, fence); - kfree(semaphore); + radeon_sa_bo_free(rdev, &(*semaphore)->sa_bo, fence); + kfree(*semaphore); + *semaphore =3D NULL; } diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeo= n/radeon_test.c index 47e1535..a94f66f 100644 --- a/drivers/gpu/drm/radeon/radeon_test.c +++ b/drivers/gpu/drm/radeon/radeon_test.c @@ -303,8 +303,7 @@ void radeon_test_ring_sync(struct radeon_device *rdev= , } =20 out_cleanup: - if (semaphore) - radeon_semaphore_free(rdev, semaphore, NULL); + radeon_semaphore_free(rdev, &semaphore, NULL); =20 if (fence1) radeon_fence_unref(&fence1); @@ -422,8 +421,7 @@ void radeon_test_ring_sync2(struct radeon_device *rde= v, } =20 out_cleanup: - if (semaphore) - radeon_semaphore_free(rdev, semaphore, NULL); + radeon_semaphore_free(rdev, &semaphore, NULL); =20 if (fenceA) radeon_fence_unref(&fenceA); diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon= /radeon_ttm.c index 2d36bdd..c43035c 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -223,7 +223,6 @@ static int radeon_move_blit(struct ttm_buffer_object = *bo, struct radeon_device *rdev; uint64_t old_start, new_start; struct radeon_fence *fence; - struct radeon_semaphore *sem =3D NULL; int r, ridx; =20 rdev =3D radeon_get_rdev(bo->bdev); @@ -262,31 +261,12 @@ static int radeon_move_blit(struct ttm_buffer_objec= t *bo, =20 /* sync other rings */ fence =3D bo->sync_obj; - if (fence && fence->ring !=3D ridx - && !radeon_fence_signaled(fence)) { - bool sync_to_ring[RADEON_NUM_RINGS] =3D { }; - sync_to_ring[fence->ring] =3D true; - - r =3D radeon_semaphore_create(rdev, &sem); - if (r) { - return r; - } - - r =3D radeon_semaphore_sync_rings(rdev, sem, sync_to_ring, ridx); - if (r) { - radeon_semaphore_free(rdev, sem, NULL); - return r; - } - } - - fence =3D NULL; r =3D radeon_copy(rdev, old_start, new_start, new_mem->num_pages * (PAGE_SIZE / RADEON_GPU_PAGE_SIZE), /* GPU pages= */ &fence); /* FIXME: handle copy error */ r =3D ttm_bo_move_accel_cleanup(bo, (void *)fence, NULL, evict, no_wait_reserve, no_wait_gpu, new_mem); - radeon_semaphore_free(rdev, sem, fence); radeon_fence_unref(&fence); return r; } --=20 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0004-drm-radeon-replace-vmram_mutex-with-mclk_lock-v2.patch" Content-Disposition: attachment; filename*0="0004-drm-radeon-replace-vmram_mutex-with-mclk_lock-v2.patch" Content-Transfer-Encoding: quoted-printable >>From db7fce3983ad9b3deebda450121af4aaf6809ce2 Mon Sep 17 00:00:00 2001 From: =3D?UTF-8?q?Christian=3D20K=3DC3=3DB6nig?=3D Date: Fri, 11 May 2012 14:57:18 +0200 Subject: [PATCH 4/9] drm/radeon: replace vmram_mutex with mclk_lock v2 MIME-Version: 1.0 Content-Type: text/plain; charset=3DUTF-8 Content-Transfer-Encoding: 8bit It is a rw_semaphore now and only write locked while changing the clock. Also the lock is renamed to better reflect what it is protecting. v2: Keep the ttm_vm_ops on IGPs Signed-off-by: Christian K=C3=B6nig Reviewed-by: Jerome Glisse --- drivers/gpu/drm/radeon/radeon.h | 3 ++- drivers/gpu/drm/radeon/radeon_device.c | 2 +- drivers/gpu/drm/radeon/radeon_object.c | 8 ++++---- drivers/gpu/drm/radeon/radeon_pm.c | 4 ++-- drivers/gpu/drm/radeon/radeon_ttm.c | 4 ++-- 5 files changed, 11 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h index 4563e50..586f30e 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1059,6 +1059,8 @@ struct radeon_power_state { =20 struct radeon_pm { struct mutex mutex; + /* write locked while reprogramming mclk */ + struct rw_semaphore mclk_lock; u32 active_crtcs; int active_crtc_count; int req_vblank; @@ -1554,7 +1556,6 @@ struct radeon_device { struct work_struct audio_work; int num_crtc; /* number of crtcs */ struct mutex dc_hw_i2c_mutex; /* display controller hw i2c mutex */ - struct mutex vram_mutex; bool audio_enabled; struct r600_audio audio_status; /* audio stuff */ struct notifier_block acpi_nb; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/rad= eon/radeon_device.c index 066c98b..7667184 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -735,7 +735,7 @@ int radeon_device_init(struct radeon_device *rdev, spin_lock_init(&rdev->ih.lock); mutex_init(&rdev->gem.mutex); mutex_init(&rdev->pm.mutex); - mutex_init(&rdev->vram_mutex); + init_rwsem(&rdev->pm.mclk_lock); init_waitqueue_head(&rdev->irq.vblank_queue); init_waitqueue_head(&rdev->irq.idle_queue); r =3D radeon_gem_init(rdev); diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/rad= eon/radeon_object.c index 830f1a7..6ecb200 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -154,11 +154,11 @@ retry: INIT_LIST_HEAD(&bo->va); radeon_ttm_placement_from_domain(bo, domain); /* Kernel allocation are uninterruptible */ - mutex_lock(&rdev->vram_mutex); + down_read(&rdev->pm.mclk_lock); r =3D ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, 0, !kernel, NULL, acc_size, sg, &radeon_ttm_bo_destroy); - mutex_unlock(&rdev->vram_mutex); + up_read(&rdev->pm.mclk_lock); if (unlikely(r !=3D 0)) { if (r !=3D -ERESTARTSYS) { if (domain =3D=3D RADEON_GEM_DOMAIN_VRAM) { @@ -219,9 +219,9 @@ void radeon_bo_unref(struct radeon_bo **bo) return; rdev =3D (*bo)->rdev; tbo =3D &((*bo)->tbo); - mutex_lock(&rdev->vram_mutex); + down_read(&rdev->pm.mclk_lock); ttm_bo_unref(&tbo); - mutex_unlock(&rdev->vram_mutex); + up_read(&rdev->pm.mclk_lock); if (tbo =3D=3D NULL) *bo =3D NULL; } diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/= radeon_pm.c index 5b37e28..9dc0b54 100644 --- a/drivers/gpu/drm/radeon/radeon_pm.c +++ b/drivers/gpu/drm/radeon/radeon_pm.c @@ -251,7 +251,7 @@ static void radeon_pm_set_clocks(struct radeon_device= *rdev) return; =20 mutex_lock(&rdev->ddev->struct_mutex); - mutex_lock(&rdev->vram_mutex); + down_write(&rdev->pm.mclk_lock); mutex_lock(&rdev->ring_lock); =20 /* gui idle int has issues on older chips it seems */ @@ -303,7 +303,7 @@ static void radeon_pm_set_clocks(struct radeon_device= *rdev) rdev->pm.dynpm_planned_action =3D DYNPM_ACTION_NONE; =20 mutex_unlock(&rdev->ring_lock); - mutex_unlock(&rdev->vram_mutex); + up_write(&rdev->pm.mclk_lock); mutex_unlock(&rdev->ddev->struct_mutex); } =20 diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon= /radeon_ttm.c index c43035c..0881131 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -797,9 +797,9 @@ static int radeon_ttm_fault(struct vm_area_struct *vm= a, struct vm_fault *vmf) return VM_FAULT_NOPAGE; } rdev =3D radeon_get_rdev(bo->bdev); - mutex_lock(&rdev->vram_mutex); + down_read(&rdev->pm.mclk_lock); r =3D ttm_vm_ops->fault(vma, vmf); - mutex_unlock(&rdev->vram_mutex); + up_read(&rdev->pm.mclk_lock); return r; } =20 --=20 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0005-drm-radeon-remove-some-unneeded-structure-members.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename*0="0005-drm-radeon-remove-some-unneeded-structure-members.patch" >>From 6823d74003abedd688a3f535aefe6ce0e06444fd Mon Sep 17 00:00:00 2001 From: Christian Koenig Date: Wed, 16 May 2012 20:24:36 +0200 Subject: [PATCH 5/9] drm/radeon: remove some unneeded structure members Signed-off-by: Christian Koenig Reviewed-by: Jerome Glisse --- drivers/gpu/drm/radeon/radeon.h | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 586f30e..9d6c1de 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -738,9 +738,7 @@ struct r600_ih { struct radeon_bo *ring_obj; volatile uint32_t *ring; unsigned rptr; - unsigned rptr_offs; unsigned wptr; - unsigned wptr_old; unsigned ring_size; uint64_t gpu_addr; uint32_t ptr_mask; -- 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0006-drm-radeon-fix-improve-ih-ring-handling-v3.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename*0="0006-drm-radeon-fix-improve-ih-ring-handling-v3.patch" >>From c20dc3698dc7ecf053e2bf77299ae5982c0c2c45 Mon Sep 17 00:00:00 2001 From: Christian Koenig Date: Wed, 16 May 2012 21:45:24 +0200 Subject: [PATCH 6/9] drm/radeon: fix & improve ih ring handling v3 The spinlock was actually there to protect the rptr, but rptr was read outside of the locked area. Also we don't really need a spinlock here, an atomic should to quite fine since we only need to prevent it from being reentrant. v2: Keep the spinlock.... v3: Back to an atomic again after finding & fixing the real bug. Signed-off-by: Christian Koenig --- drivers/gpu/drm/radeon/evergreen.c | 26 +++++++++++++------------- drivers/gpu/drm/radeon/r600.c | 29 +++++++++++++---------------- drivers/gpu/drm/radeon/radeon.h | 3 +-- drivers/gpu/drm/radeon/radeon_device.c | 3 +-- drivers/gpu/drm/radeon/si.c | 27 +++++++++++++-------------- 5 files changed, 41 insertions(+), 47 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 8e40ba4..765bd1f 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2676,7 +2676,6 @@ int evergreen_irq_process(struct radeon_device *rdev) u32 rptr; u32 src_id, src_data; u32 ring_index; - unsigned long flags; bool queue_hotplug = false; bool queue_hdmi = false; @@ -2684,22 +2683,21 @@ int evergreen_irq_process(struct radeon_device *rdev) return IRQ_NONE; wptr = evergreen_get_ih_wptr(rdev); + +restart_ih: + /* is somebody else already processing irqs? */ + if (atomic_xchg(&rdev->ih.lock, 1)) + return IRQ_NONE; + rptr = rdev->ih.rptr; DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr); - spin_lock_irqsave(&rdev->ih.lock, flags); - if (rptr == wptr) { - spin_unlock_irqrestore(&rdev->ih.lock, flags); - return IRQ_NONE; - } -restart_ih: /* Order reading of wptr vs. reading of IH ring data */ rmb(); /* display interrupts */ evergreen_irq_ack(rdev); - rdev->ih.wptr = wptr; while (rptr != wptr) { /* wptr/rptr are in bytes! */ ring_index = rptr / 4; @@ -2998,17 +2996,19 @@ restart_ih: rptr += 16; rptr &= rdev->ih.ptr_mask; } - /* make sure wptr hasn't changed while processing */ - wptr = evergreen_get_ih_wptr(rdev); - if (wptr != rdev->ih.wptr) - goto restart_ih; if (queue_hotplug) schedule_work(&rdev->hotplug_work); if (queue_hdmi) schedule_work(&rdev->audio_work); rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); - spin_unlock_irqrestore(&rdev->ih.lock, flags); + atomic_set(&rdev->ih.lock, 0); + + /* make sure wptr hasn't changed while processing */ + wptr = evergreen_get_ih_wptr(rdev); + if (wptr != rptr) + goto restart_ih; + return IRQ_HANDLED; } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index aced97b..b487c69 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2858,7 +2858,6 @@ void r600_disable_interrupts(struct radeon_device *rdev) WREG32(IH_RB_RPTR, 0); WREG32(IH_RB_WPTR, 0); rdev->ih.enabled = false; - rdev->ih.wptr = 0; rdev->ih.rptr = 0; } @@ -3310,7 +3309,6 @@ int r600_irq_process(struct radeon_device *rdev) u32 rptr; u32 src_id, src_data; u32 ring_index; - unsigned long flags; bool queue_hotplug = false; bool queue_hdmi = false; @@ -3322,24 +3320,21 @@ int r600_irq_process(struct radeon_device *rdev) RREG32(IH_RB_WPTR); wptr = r600_get_ih_wptr(rdev); - rptr = rdev->ih.rptr; - DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr); - - spin_lock_irqsave(&rdev->ih.lock, flags); - if (rptr == wptr) { - spin_unlock_irqrestore(&rdev->ih.lock, flags); +restart_ih: + /* is somebody else already processing irqs? */ + if (atomic_xchg(&rdev->ih.lock, 1)) return IRQ_NONE; - } -restart_ih: + rptr = rdev->ih.rptr; + DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr); + /* Order reading of wptr vs. reading of IH ring data */ rmb(); /* display interrupts */ r600_irq_ack(rdev); - rdev->ih.wptr = wptr; while (rptr != wptr) { /* wptr/rptr are in bytes! */ ring_index = rptr / 4; @@ -3493,17 +3488,19 @@ restart_ih: rptr += 16; rptr &= rdev->ih.ptr_mask; } - /* make sure wptr hasn't changed while processing */ - wptr = r600_get_ih_wptr(rdev); - if (wptr != rdev->ih.wptr) - goto restart_ih; if (queue_hotplug) schedule_work(&rdev->hotplug_work); if (queue_hdmi) schedule_work(&rdev->audio_work); rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); - spin_unlock_irqrestore(&rdev->ih.lock, flags); + atomic_set(&rdev->ih.lock, 0); + + /* make sure wptr hasn't changed while processing */ + wptr = r600_get_ih_wptr(rdev); + if (wptr != rptr) + goto restart_ih; + return IRQ_HANDLED; } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 9d6c1de..5ddf895 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -738,11 +738,10 @@ struct r600_ih { struct radeon_bo *ring_obj; volatile uint32_t *ring; unsigned rptr; - unsigned wptr; unsigned ring_size; uint64_t gpu_addr; uint32_t ptr_mask; - spinlock_t lock; + atomic_t lock; bool enabled; }; diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 7667184..3c563d1 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -731,8 +731,7 @@ int radeon_device_init(struct radeon_device *rdev, radeon_mutex_init(&rdev->cs_mutex); mutex_init(&rdev->ring_lock); mutex_init(&rdev->dc_hw_i2c_mutex); - if (rdev->family >= CHIP_R600) - spin_lock_init(&rdev->ih.lock); + atomic_set(&rdev->ih.lock, 0); mutex_init(&rdev->gem.mutex); mutex_init(&rdev->pm.mutex); init_rwsem(&rdev->pm.mclk_lock); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 8868a1f..ecef972 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -2942,7 +2942,6 @@ static void si_disable_interrupts(struct radeon_device *rdev) WREG32(IH_RB_RPTR, 0); WREG32(IH_RB_WPTR, 0); rdev->ih.enabled = false; - rdev->ih.wptr = 0; rdev->ih.rptr = 0; } @@ -3359,29 +3358,27 @@ int si_irq_process(struct radeon_device *rdev) u32 rptr; u32 src_id, src_data, ring_id; u32 ring_index; - unsigned long flags; bool queue_hotplug = false; if (!rdev->ih.enabled || rdev->shutdown) return IRQ_NONE; wptr = si_get_ih_wptr(rdev); + +restart_ih: + /* is somebody else already processing irqs? */ + if (atomic_xchg(&rdev->ih.lock, 1)) + return IRQ_NONE; + rptr = rdev->ih.rptr; DRM_DEBUG("si_irq_process start: rptr %d, wptr %d\n", rptr, wptr); - spin_lock_irqsave(&rdev->ih.lock, flags); - if (rptr == wptr) { - spin_unlock_irqrestore(&rdev->ih.lock, flags); - return IRQ_NONE; - } -restart_ih: /* Order reading of wptr vs. reading of IH ring data */ rmb(); /* display interrupts */ si_irq_ack(rdev); - rdev->ih.wptr = wptr; while (rptr != wptr) { /* wptr/rptr are in bytes! */ ring_index = rptr / 4; @@ -3632,15 +3629,17 @@ restart_ih: rptr += 16; rptr &= rdev->ih.ptr_mask; } - /* make sure wptr hasn't changed while processing */ - wptr = si_get_ih_wptr(rdev); - if (wptr != rdev->ih.wptr) - goto restart_ih; if (queue_hotplug) schedule_work(&rdev->hotplug_work); rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); - spin_unlock_irqrestore(&rdev->ih.lock, flags); + atomic_set(&rdev->ih.lock, 0); + + /* make sure wptr hasn't changed while processing */ + wptr = si_get_ih_wptr(rdev); + if (wptr != rptr) + goto restart_ih; + return IRQ_HANDLED; } -- 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0007-drm-radeon-apply-Murphy-s-law-to-the-kms-irq-code-v3.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename*0="0007-drm-radeon-apply-Murphy-s-law-to-the-kms-irq-code-v3.pa"; filename*1="tch" >>From fb98257a9d9d2089972b18079d5bdd4412e107e2 Mon Sep 17 00:00:00 2001 From: Christian Koenig Date: Thu, 17 May 2012 01:33:30 +0200 Subject: [PATCH 7/9] drm/radeon: apply Murphy's law to the kms irq code v3 1. It is really dangerous to have more than one spinlock protecting the same information. 2. radeon_irq_set sometimes wasn't called with lock protection, so it can happen that more than one CPU would tamper with the irq regs at the same time. 3. The pm.gui_idle variable was assuming that the 3D engine wasn't becoming idle between testing the register and setting the variable. So just remove it and test the register directly. v2: Also handle the hpd irq code the same way. v3: Rename hpd parameter for clarification. Signed-off-by: Christian Koenig Reviewed-by: Alex Deucher --- drivers/gpu/drm/radeon/evergreen.c | 21 ++----- drivers/gpu/drm/radeon/r100.c | 29 ++-------- drivers/gpu/drm/radeon/r600.c | 38 ++++-------- drivers/gpu/drm/radeon/r600_hdmi.c | 6 +- drivers/gpu/drm/radeon/radeon.h | 35 +++++------ drivers/gpu/drm/radeon/radeon_irq_kms.c | 96 +++++++++++++++++++++++++++---- drivers/gpu/drm/radeon/radeon_kms.c | 12 +++- drivers/gpu/drm/radeon/radeon_pm.c | 12 +--- drivers/gpu/drm/radeon/rs600.c | 13 ++--- drivers/gpu/drm/radeon/si.c | 1 - 10 files changed, 144 insertions(+), 119 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 765bd1f..bdc1f30 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -428,6 +428,7 @@ void evergreen_hpd_init(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned enabled = 0; u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) | DC_HPDx_RX_INT_TIMER(0xfa) | DC_HPDx_EN; @@ -436,73 +437,64 @@ void evergreen_hpd_init(struct radeon_device *rdev) switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HPD1_CONTROL, tmp); - rdev->irq.hpd[0] = true; break; case RADEON_HPD_2: WREG32(DC_HPD2_CONTROL, tmp); - rdev->irq.hpd[1] = true; break; case RADEON_HPD_3: WREG32(DC_HPD3_CONTROL, tmp); - rdev->irq.hpd[2] = true; break; case RADEON_HPD_4: WREG32(DC_HPD4_CONTROL, tmp); - rdev->irq.hpd[3] = true; break; case RADEON_HPD_5: WREG32(DC_HPD5_CONTROL, tmp); - rdev->irq.hpd[4] = true; break; case RADEON_HPD_6: WREG32(DC_HPD6_CONTROL, tmp); - rdev->irq.hpd[5] = true; break; default: break; } radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd); + enabled |= 1 << radeon_connector->hpd.hpd; } - if (rdev->irq.installed) - evergreen_irq_set(rdev); + radeon_irq_kms_enable_hpd(rdev, enabled); } void evergreen_hpd_fini(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned disabled = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HPD1_CONTROL, 0); - rdev->irq.hpd[0] = false; break; case RADEON_HPD_2: WREG32(DC_HPD2_CONTROL, 0); - rdev->irq.hpd[1] = false; break; case RADEON_HPD_3: WREG32(DC_HPD3_CONTROL, 0); - rdev->irq.hpd[2] = false; break; case RADEON_HPD_4: WREG32(DC_HPD4_CONTROL, 0); - rdev->irq.hpd[3] = false; break; case RADEON_HPD_5: WREG32(DC_HPD5_CONTROL, 0); - rdev->irq.hpd[4] = false; break; case RADEON_HPD_6: WREG32(DC_HPD6_CONTROL, 0); - rdev->irq.hpd[5] = false; break; default: break; } + disabled |= 1 << radeon_connector->hpd.hpd; } + radeon_irq_kms_disable_hpd(rdev, disabled); } /* watermark setup */ @@ -2984,7 +2976,6 @@ restart_ih: break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); - rdev->pm.gui_idle = true; wake_up(&rdev->irq.idle_queue); break; default: diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index 415b7d8..e8fe4ae 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -567,43 +567,27 @@ void r100_hpd_init(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned enable = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); - switch (radeon_connector->hpd.hpd) { - case RADEON_HPD_1: - rdev->irq.hpd[0] = true; - break; - case RADEON_HPD_2: - rdev->irq.hpd[1] = true; - break; - default: - break; - } + enable |= 1 << radeon_connector->hpd.hpd; radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd); } - if (rdev->irq.installed) - r100_irq_set(rdev); + radeon_irq_kms_enable_hpd(rdev, enable); } void r100_hpd_fini(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned disable = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); - switch (radeon_connector->hpd.hpd) { - case RADEON_HPD_1: - rdev->irq.hpd[0] = false; - break; - case RADEON_HPD_2: - rdev->irq.hpd[1] = false; - break; - default: - break; - } + disable |= 1 << radeon_connector->hpd.hpd; } + radeon_irq_kms_disable_hpd(rdev, disable); } /* @@ -782,7 +766,6 @@ int r100_irq_process(struct radeon_device *rdev) /* gui idle interrupt */ if (status & RADEON_GUI_IDLE_STAT) { rdev->irq.gui_idle_acked = true; - rdev->pm.gui_idle = true; wake_up(&rdev->irq.idle_queue); } /* Vertical blank interrupts */ diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index b487c69..9cd7728 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -709,6 +709,7 @@ void r600_hpd_init(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned enable = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); @@ -729,28 +730,22 @@ void r600_hpd_init(struct radeon_device *rdev) switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HPD1_CONTROL, tmp); - rdev->irq.hpd[0] = true; break; case RADEON_HPD_2: WREG32(DC_HPD2_CONTROL, tmp); - rdev->irq.hpd[1] = true; break; case RADEON_HPD_3: WREG32(DC_HPD3_CONTROL, tmp); - rdev->irq.hpd[2] = true; break; case RADEON_HPD_4: WREG32(DC_HPD4_CONTROL, tmp); - rdev->irq.hpd[3] = true; break; /* DCE 3.2 */ case RADEON_HPD_5: WREG32(DC_HPD5_CONTROL, tmp); - rdev->irq.hpd[4] = true; break; case RADEON_HPD_6: WREG32(DC_HPD6_CONTROL, tmp); - rdev->irq.hpd[5] = true; break; default: break; @@ -759,85 +754,73 @@ void r600_hpd_init(struct radeon_device *rdev) switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN); - rdev->irq.hpd[0] = true; break; case RADEON_HPD_2: WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN); - rdev->irq.hpd[1] = true; break; case RADEON_HPD_3: WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN); - rdev->irq.hpd[2] = true; break; default: break; } } + enable |= 1 << radeon_connector->hpd.hpd; radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd); } - if (rdev->irq.installed) - r600_irq_set(rdev); + radeon_irq_kms_enable_hpd(rdev, enable); } void r600_hpd_fini(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned disable = 0; - if (ASIC_IS_DCE3(rdev)) { - list_for_each_entry(connector, &dev->mode_config.connector_list, head) { - struct radeon_connector *radeon_connector = to_radeon_connector(connector); + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + if (ASIC_IS_DCE3(rdev)) { switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HPD1_CONTROL, 0); - rdev->irq.hpd[0] = false; break; case RADEON_HPD_2: WREG32(DC_HPD2_CONTROL, 0); - rdev->irq.hpd[1] = false; break; case RADEON_HPD_3: WREG32(DC_HPD3_CONTROL, 0); - rdev->irq.hpd[2] = false; break; case RADEON_HPD_4: WREG32(DC_HPD4_CONTROL, 0); - rdev->irq.hpd[3] = false; break; /* DCE 3.2 */ case RADEON_HPD_5: WREG32(DC_HPD5_CONTROL, 0); - rdev->irq.hpd[4] = false; break; case RADEON_HPD_6: WREG32(DC_HPD6_CONTROL, 0); - rdev->irq.hpd[5] = false; break; default: break; } - } - } else { - list_for_each_entry(connector, &dev->mode_config.connector_list, head) { - struct radeon_connector *radeon_connector = to_radeon_connector(connector); + } else { switch (radeon_connector->hpd.hpd) { case RADEON_HPD_1: WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0); - rdev->irq.hpd[0] = false; break; case RADEON_HPD_2: WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0); - rdev->irq.hpd[1] = false; break; case RADEON_HPD_3: WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0); - rdev->irq.hpd[2] = false; break; default: break; } } + disable |= 1 << radeon_connector->hpd.hpd; } + radeon_irq_kms_disable_hpd(rdev, disable); } /* @@ -3476,7 +3459,6 @@ restart_ih: break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); - rdev->pm.gui_idle = true; wake_up(&rdev->irq.idle_queue); break; default: diff --git a/drivers/gpu/drm/radeon/r600_hdmi.c b/drivers/gpu/drm/radeon/r600_hdmi.c index 82a0a4c..e3558c3 100644 --- a/drivers/gpu/drm/radeon/r600_hdmi.c +++ b/drivers/gpu/drm/radeon/r600_hdmi.c @@ -519,8 +519,7 @@ void r600_hdmi_enable(struct drm_encoder *encoder) if (rdev->irq.installed) { /* if irq is available use it */ - rdev->irq.afmt[dig->afmt->id] = true; - radeon_irq_set(rdev); + radeon_irq_kms_enable_afmt(rdev, dig->afmt->id); } dig->afmt->enabled = true; @@ -556,8 +555,7 @@ void r600_hdmi_disable(struct drm_encoder *encoder) offset, radeon_encoder->encoder_id); /* disable irq */ - rdev->irq.afmt[dig->afmt->id] = false; - radeon_irq_set(rdev); + radeon_irq_kms_disable_afmt(rdev, dig->afmt->id); /* Older chipsets not handled by AtomBIOS */ if (rdev->family >= CHIP_R600 && !ASIC_IS_DCE3(rdev)) { diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 5ddf895..353a183 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -615,21 +615,20 @@ union radeon_irq_stat_regs { #define RADEON_MAX_AFMT_BLOCKS 6 struct radeon_irq { - bool installed; - bool sw_int[RADEON_NUM_RINGS]; - bool crtc_vblank_int[RADEON_MAX_CRTCS]; - bool pflip[RADEON_MAX_CRTCS]; - wait_queue_head_t vblank_queue; - bool hpd[RADEON_MAX_HPD_PINS]; - bool gui_idle; - bool gui_idle_acked; - wait_queue_head_t idle_queue; - bool afmt[RADEON_MAX_AFMT_BLOCKS]; - spinlock_t sw_lock; - int sw_refcount[RADEON_NUM_RINGS]; - union radeon_irq_stat_regs stat_regs; - spinlock_t pflip_lock[RADEON_MAX_CRTCS]; - int pflip_refcount[RADEON_MAX_CRTCS]; + bool installed; + spinlock_t lock; + bool sw_int[RADEON_NUM_RINGS]; + int sw_refcount[RADEON_NUM_RINGS]; + bool crtc_vblank_int[RADEON_MAX_CRTCS]; + bool pflip[RADEON_MAX_CRTCS]; + int pflip_refcount[RADEON_MAX_CRTCS]; + wait_queue_head_t vblank_queue; + bool hpd[RADEON_MAX_HPD_PINS]; + bool gui_idle; + bool gui_idle_acked; + wait_queue_head_t idle_queue; + bool afmt[RADEON_MAX_AFMT_BLOCKS]; + union radeon_irq_stat_regs stat_regs; }; int radeon_irq_kms_init(struct radeon_device *rdev); @@ -638,6 +637,11 @@ void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev, int ring); void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev, int ring); void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc); void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc); +void radeon_irq_kms_enable_afmt(struct radeon_device *rdev, int block); +void radeon_irq_kms_disable_afmt(struct radeon_device *rdev, int block); +void radeon_irq_kms_enable_hpd(struct radeon_device *rdev, unsigned hpd_mask); +void radeon_irq_kms_disable_hpd(struct radeon_device *rdev, unsigned hpd_mask); +int radeon_irq_kms_wait_gui_idle(struct radeon_device *rdev); /* * CP & rings. @@ -1062,7 +1066,6 @@ struct radeon_pm { int active_crtc_count; int req_vblank; bool vblank_sync; - bool gui_idle; fixed20_12 max_bandwidth; fixed20_12 igp_sideport_mclk; fixed20_12 igp_system_mclk; diff --git a/drivers/gpu/drm/radeon/radeon_irq_kms.c b/drivers/gpu/drm/radeon/radeon_irq_kms.c index 5df58d1..c5eb7a1 100644 --- a/drivers/gpu/drm/radeon/radeon_irq_kms.c +++ b/drivers/gpu/drm/radeon/radeon_irq_kms.c @@ -32,6 +32,8 @@ #include "radeon.h" #include "atom.h" +#define RADEON_WAIT_IDLE_TIMEOUT 200 + irqreturn_t radeon_driver_irq_handler_kms(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; @@ -62,8 +64,10 @@ static void radeon_hotplug_work_func(struct work_struct *work) void radeon_driver_irq_preinstall_kms(struct drm_device *dev) { struct radeon_device *rdev = dev->dev_private; + unsigned long irqflags; unsigned i; + spin_lock_irqsave(&rdev->irq.lock, irqflags); /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) rdev->irq.sw_int[i] = false; @@ -76,6 +80,7 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev) rdev->irq.afmt[i] = false; } radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); /* Clear bits */ radeon_irq_process(rdev); } @@ -83,23 +88,28 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev) int radeon_driver_irq_postinstall_kms(struct drm_device *dev) { struct radeon_device *rdev = dev->dev_private; + unsigned long irqflags; unsigned i; dev->max_vblank_count = 0x001fffff; + spin_lock_irqsave(&rdev->irq.lock, irqflags); for (i = 0; i < RADEON_NUM_RINGS; i++) rdev->irq.sw_int[i] = true; radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); return 0; } void radeon_driver_irq_uninstall_kms(struct drm_device *dev) { struct radeon_device *rdev = dev->dev_private; + unsigned long irqflags; unsigned i; if (rdev == NULL) { return; } + spin_lock_irqsave(&rdev->irq.lock, irqflags); /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) rdev->irq.sw_int[i] = false; @@ -112,6 +122,7 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev) rdev->irq.afmt[i] = false; } radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } static bool radeon_msi_ok(struct radeon_device *rdev) @@ -168,15 +179,12 @@ static bool radeon_msi_ok(struct radeon_device *rdev) int radeon_irq_kms_init(struct radeon_device *rdev) { - int i; int r = 0; INIT_WORK(&rdev->hotplug_work, radeon_hotplug_work_func); INIT_WORK(&rdev->audio_work, r600_audio_update_hdmi); - spin_lock_init(&rdev->irq.sw_lock); - for (i = 0; i < rdev->num_crtc; i++) - spin_lock_init(&rdev->irq.pflip_lock[i]); + spin_lock_init(&rdev->irq.lock); r = drm_vblank_init(rdev->ddev, rdev->num_crtc); if (r) { return r; @@ -217,25 +225,25 @@ void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev, int ring) { unsigned long irqflags; - spin_lock_irqsave(&rdev->irq.sw_lock, irqflags); + spin_lock_irqsave(&rdev->irq.lock, irqflags); if (rdev->ddev->irq_enabled && (++rdev->irq.sw_refcount[ring] == 1)) { rdev->irq.sw_int[ring] = true; radeon_irq_set(rdev); } - spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev, int ring) { unsigned long irqflags; - spin_lock_irqsave(&rdev->irq.sw_lock, irqflags); + spin_lock_irqsave(&rdev->irq.lock, irqflags); BUG_ON(rdev->ddev->irq_enabled && rdev->irq.sw_refcount[ring] <= 0); if (rdev->ddev->irq_enabled && (--rdev->irq.sw_refcount[ring] == 0)) { rdev->irq.sw_int[ring] = false; radeon_irq_set(rdev); } - spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc) @@ -245,12 +253,12 @@ void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc) if (crtc < 0 || crtc >= rdev->num_crtc) return; - spin_lock_irqsave(&rdev->irq.pflip_lock[crtc], irqflags); + spin_lock_irqsave(&rdev->irq.lock, irqflags); if (rdev->ddev->irq_enabled && (++rdev->irq.pflip_refcount[crtc] == 1)) { rdev->irq.pflip[crtc] = true; radeon_irq_set(rdev); } - spin_unlock_irqrestore(&rdev->irq.pflip_lock[crtc], irqflags); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc) @@ -260,12 +268,76 @@ void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc) if (crtc < 0 || crtc >= rdev->num_crtc) return; - spin_lock_irqsave(&rdev->irq.pflip_lock[crtc], irqflags); + spin_lock_irqsave(&rdev->irq.lock, irqflags); BUG_ON(rdev->ddev->irq_enabled && rdev->irq.pflip_refcount[crtc] <= 0); if (rdev->ddev->irq_enabled && (--rdev->irq.pflip_refcount[crtc] == 0)) { rdev->irq.pflip[crtc] = false; radeon_irq_set(rdev); } - spin_unlock_irqrestore(&rdev->irq.pflip_lock[crtc], irqflags); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } +void radeon_irq_kms_enable_afmt(struct radeon_device *rdev, int block) +{ + unsigned long irqflags; + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + rdev->irq.afmt[block] = true; + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); + +} + +void radeon_irq_kms_disable_afmt(struct radeon_device *rdev, int block) +{ + unsigned long irqflags; + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + rdev->irq.afmt[block] = false; + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); +} + +void radeon_irq_kms_enable_hpd(struct radeon_device *rdev, unsigned hpd_mask) +{ + unsigned long irqflags; + int i; + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + for (i = 0; i < RADEON_MAX_HPD_PINS; ++i) + rdev->irq.hpd[i] |= !!(hpd_mask & (1 << i)); + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); +} + +void radeon_irq_kms_disable_hpd(struct radeon_device *rdev, unsigned hpd_mask) +{ + unsigned long irqflags; + int i; + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + for (i = 0; i < RADEON_MAX_HPD_PINS; ++i) + rdev->irq.hpd[i] &= !(hpd_mask & (1 << i)); + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); +} + +int radeon_irq_kms_wait_gui_idle(struct radeon_device *rdev) +{ + unsigned long irqflags; + int r; + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + rdev->irq.gui_idle = true; + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); + + r = wait_event_timeout(rdev->irq.idle_queue, radeon_gui_idle(rdev), + msecs_to_jiffies(RADEON_WAIT_IDLE_TIMEOUT)); + + spin_lock_irqsave(&rdev->irq.lock, irqflags); + rdev->irq.gui_idle = false; + radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); + return r; +} diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 5c58d7d..18b81d6 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -382,29 +382,35 @@ u32 radeon_get_vblank_counter_kms(struct drm_device *dev, int crtc) int radeon_enable_vblank_kms(struct drm_device *dev, int crtc) { struct radeon_device *rdev = dev->dev_private; + unsigned long irqflags; + int r; if (crtc < 0 || crtc >= rdev->num_crtc) { DRM_ERROR("Invalid crtc %d\n", crtc); return -EINVAL; } + spin_lock_irqsave(&rdev->irq.lock, irqflags); rdev->irq.crtc_vblank_int[crtc] = true; - - return radeon_irq_set(rdev); + r = radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); + return r; } void radeon_disable_vblank_kms(struct drm_device *dev, int crtc) { struct radeon_device *rdev = dev->dev_private; + unsigned long irqflags; if (crtc < 0 || crtc >= rdev->num_crtc) { DRM_ERROR("Invalid crtc %d\n", crtc); return; } + spin_lock_irqsave(&rdev->irq.lock, irqflags); rdev->irq.crtc_vblank_int[crtc] = false; - radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } int radeon_get_vblank_timestamp_kms(struct drm_device *dev, int crtc, diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c index 9dc0b54..7ae6066 100644 --- a/drivers/gpu/drm/radeon/radeon_pm.c +++ b/drivers/gpu/drm/radeon/radeon_pm.c @@ -34,7 +34,6 @@ #define RADEON_IDLE_LOOP_MS 100 #define RADEON_RECLOCK_DELAY_MS 200 #define RADEON_WAIT_VBLANK_TIMEOUT 200 -#define RADEON_WAIT_IDLE_TIMEOUT 200 static const char *radeon_pm_state_type_name[5] = { "Default", @@ -257,15 +256,8 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev) /* gui idle int has issues on older chips it seems */ if (rdev->family >= CHIP_R600) { if (rdev->irq.installed) { - /* wait for GPU idle */ - rdev->pm.gui_idle = false; - rdev->irq.gui_idle = true; - radeon_irq_set(rdev); - wait_event_interruptible_timeout( - rdev->irq.idle_queue, rdev->pm.gui_idle, - msecs_to_jiffies(RADEON_WAIT_IDLE_TIMEOUT)); - rdev->irq.gui_idle = false; - radeon_irq_set(rdev); + /* wait for GPU to become idle */ + radeon_irq_kms_wait_gui_idle(rdev); } } else { struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c index e95c5e6..f9aee25 100644 --- a/drivers/gpu/drm/radeon/rs600.c +++ b/drivers/gpu/drm/radeon/rs600.c @@ -294,6 +294,7 @@ void rs600_hpd_init(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned enable = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); @@ -301,26 +302,25 @@ void rs600_hpd_init(struct radeon_device *rdev) case RADEON_HPD_1: WREG32(R_007D00_DC_HOT_PLUG_DETECT1_CONTROL, S_007D00_DC_HOT_PLUG_DETECT1_EN(1)); - rdev->irq.hpd[0] = true; break; case RADEON_HPD_2: WREG32(R_007D10_DC_HOT_PLUG_DETECT2_CONTROL, S_007D10_DC_HOT_PLUG_DETECT2_EN(1)); - rdev->irq.hpd[1] = true; break; default: break; } + enable |= 1 << radeon_connector->hpd.hpd; radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd); } - if (rdev->irq.installed) - rs600_irq_set(rdev); + radeon_irq_kms_enable_hpd(rdev, enable); } void rs600_hpd_fini(struct radeon_device *rdev) { struct drm_device *dev = rdev->ddev; struct drm_connector *connector; + unsigned disable = 0; list_for_each_entry(connector, &dev->mode_config.connector_list, head) { struct radeon_connector *radeon_connector = to_radeon_connector(connector); @@ -328,17 +328,17 @@ void rs600_hpd_fini(struct radeon_device *rdev) case RADEON_HPD_1: WREG32(R_007D00_DC_HOT_PLUG_DETECT1_CONTROL, S_007D00_DC_HOT_PLUG_DETECT1_EN(0)); - rdev->irq.hpd[0] = false; break; case RADEON_HPD_2: WREG32(R_007D10_DC_HOT_PLUG_DETECT2_CONTROL, S_007D10_DC_HOT_PLUG_DETECT2_EN(0)); - rdev->irq.hpd[1] = false; break; default: break; } + disable |= 1 << radeon_connector->hpd.hpd; } + radeon_irq_kms_disable_hpd(rdev, disable); } int rs600_asic_reset(struct radeon_device *rdev) @@ -686,7 +686,6 @@ int rs600_irq_process(struct radeon_device *rdev) /* GUI idle */ if (G_000040_GUI_IDLE(status)) { rdev->irq.gui_idle_acked = true; - rdev->pm.gui_idle = true; wake_up(&rdev->irq.idle_queue); } /* Vertical blank interrupts */ diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index ecef972..a65045b 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -3617,7 +3617,6 @@ restart_ih: break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); - rdev->pm.gui_idle = true; wake_up(&rdev->irq.idle_queue); break; default: -- 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0008-drm-radeon-replace-pflip-and-sw_int-counters-with-at.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename*0="0008-drm-radeon-replace-pflip-and-sw_int-counters-with-at.pa"; filename*1="tch" >>From 736fc37fd7c7634e939e9ec0c67765941913bb82 Mon Sep 17 00:00:00 2001 From: Christian Koenig Date: Thu, 17 May 2012 19:52:00 +0200 Subject: [PATCH 8/9] drm/radeon: replace pflip and sw_int counters with atomics So we can skip the locking. Also renames sw_int to ring_int, cause that better matches its purpose. Signed-off-by: Christian Koenig --- drivers/gpu/drm/radeon/evergreen.c | 32 ++++++++--------- drivers/gpu/drm/radeon/r100.c | 10 +++--- drivers/gpu/drm/radeon/r600.c | 10 +++--- drivers/gpu/drm/radeon/radeon.h | 6 ++-- drivers/gpu/drm/radeon/radeon_irq_kms.c | 59 +++++++++++++++---------------- drivers/gpu/drm/radeon/rs600.c | 10 +++--- drivers/gpu/drm/radeon/si.c | 30 ++++++++-------- 7 files changed, 76 insertions(+), 81 deletions(-) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index bdc1f30..f716e08 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -2340,20 +2340,20 @@ int evergreen_irq_set(struct radeon_device *rdev) if (rdev->family >= CHIP_CAYMAN) { /* enable CP interrupts on all rings */ - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("evergreen_irq_set: sw int gfx\n"); cp_int_cntl |= TIME_STAMP_INT_ENABLE; } - if (rdev->irq.sw_int[CAYMAN_RING_TYPE_CP1_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP1_INDEX])) { DRM_DEBUG("evergreen_irq_set: sw int cp1\n"); cp_int_cntl1 |= TIME_STAMP_INT_ENABLE; } - if (rdev->irq.sw_int[CAYMAN_RING_TYPE_CP2_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP2_INDEX])) { DRM_DEBUG("evergreen_irq_set: sw int cp2\n"); cp_int_cntl2 |= TIME_STAMP_INT_ENABLE; } } else { - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("evergreen_irq_set: sw int gfx\n"); cp_int_cntl |= RB_INT_ENABLE; cp_int_cntl |= TIME_STAMP_INT_ENABLE; @@ -2361,32 +2361,32 @@ int evergreen_irq_set(struct radeon_device *rdev) } if (rdev->irq.crtc_vblank_int[0] || - rdev->irq.pflip[0]) { + atomic_read(&rdev->irq.pflip[0])) { DRM_DEBUG("evergreen_irq_set: vblank 0\n"); crtc1 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[1] || - rdev->irq.pflip[1]) { + atomic_read(&rdev->irq.pflip[1])) { DRM_DEBUG("evergreen_irq_set: vblank 1\n"); crtc2 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[2] || - rdev->irq.pflip[2]) { + atomic_read(&rdev->irq.pflip[2])) { DRM_DEBUG("evergreen_irq_set: vblank 2\n"); crtc3 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[3] || - rdev->irq.pflip[3]) { + atomic_read(&rdev->irq.pflip[3])) { DRM_DEBUG("evergreen_irq_set: vblank 3\n"); crtc4 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[4] || - rdev->irq.pflip[4]) { + atomic_read(&rdev->irq.pflip[4])) { DRM_DEBUG("evergreen_irq_set: vblank 4\n"); crtc5 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[5] || - rdev->irq.pflip[5]) { + atomic_read(&rdev->irq.pflip[5])) { DRM_DEBUG("evergreen_irq_set: vblank 5\n"); crtc6 |= VBLANK_INT_MASK; } @@ -2706,7 +2706,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[0]) + if (atomic_read(&rdev->irq.pflip[0])) radeon_crtc_handle_flip(rdev, 0); rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VBLANK_INTERRUPT; DRM_DEBUG("IH: D1 vblank\n"); @@ -2732,7 +2732,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[1]) + if (atomic_read(&rdev->irq.pflip[1])) radeon_crtc_handle_flip(rdev, 1); rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT; DRM_DEBUG("IH: D2 vblank\n"); @@ -2758,7 +2758,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[2]) + if (atomic_read(&rdev->irq.pflip[2])) radeon_crtc_handle_flip(rdev, 2); rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT; DRM_DEBUG("IH: D3 vblank\n"); @@ -2784,7 +2784,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[3]) + if (atomic_read(&rdev->irq.pflip[3])) radeon_crtc_handle_flip(rdev, 3); rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT; DRM_DEBUG("IH: D4 vblank\n"); @@ -2810,7 +2810,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[4]) + if (atomic_read(&rdev->irq.pflip[4])) radeon_crtc_handle_flip(rdev, 4); rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT; DRM_DEBUG("IH: D5 vblank\n"); @@ -2836,7 +2836,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[5]) + if (atomic_read(&rdev->irq.pflip[5])) radeon_crtc_handle_flip(rdev, 5); rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT; DRM_DEBUG("IH: D6 vblank\n"); diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index e8fe4ae..35825bf 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -689,18 +689,18 @@ int r100_irq_set(struct radeon_device *rdev) WREG32(R_000040_GEN_INT_CNTL, 0); return -EINVAL; } - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { tmp |= RADEON_SW_INT_ENABLE; } if (rdev->irq.gui_idle) { tmp |= RADEON_GUI_IDLE_MASK; } if (rdev->irq.crtc_vblank_int[0] || - rdev->irq.pflip[0]) { + atomic_read(&rdev->irq.pflip[0])) { tmp |= RADEON_CRTC_VBLANK_MASK; } if (rdev->irq.crtc_vblank_int[1] || - rdev->irq.pflip[1]) { + atomic_read(&rdev->irq.pflip[1])) { tmp |= RADEON_CRTC2_VBLANK_MASK; } if (rdev->irq.hpd[0]) { @@ -775,7 +775,7 @@ int r100_irq_process(struct radeon_device *rdev) rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[0]) + if (atomic_read(&rdev->irq.pflip[0])) radeon_crtc_handle_flip(rdev, 0); } if (status & RADEON_CRTC2_VBLANK_STAT) { @@ -784,7 +784,7 @@ int r100_irq_process(struct radeon_device *rdev) rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[1]) + if (atomic_read(&rdev->irq.pflip[1])) radeon_crtc_handle_flip(rdev, 1); } if (status & RADEON_FP_DETECT_STAT) { diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 9cd7728..43d0c41 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3025,18 +3025,18 @@ int r600_irq_set(struct radeon_device *rdev) hdmi1 = RREG32(HDMI1_AUDIO_PACKET_CONTROL) & ~HDMI0_AZ_FORMAT_WTRIG_MASK; } - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("r600_irq_set: sw int\n"); cp_int_cntl |= RB_INT_ENABLE; cp_int_cntl |= TIME_STAMP_INT_ENABLE; } if (rdev->irq.crtc_vblank_int[0] || - rdev->irq.pflip[0]) { + atomic_read(&rdev->irq.pflip[0])) { DRM_DEBUG("r600_irq_set: vblank 0\n"); mode_int |= D1MODE_VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[1] || - rdev->irq.pflip[1]) { + atomic_read(&rdev->irq.pflip[1])) { DRM_DEBUG("r600_irq_set: vblank 1\n"); mode_int |= D2MODE_VBLANK_INT_MASK; } @@ -3334,7 +3334,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[0]) + if (atomic_read(&rdev->irq.pflip[0])) radeon_crtc_handle_flip(rdev, 0); rdev->irq.stat_regs.r600.disp_int &= ~LB_D1_VBLANK_INTERRUPT; DRM_DEBUG("IH: D1 vblank\n"); @@ -3360,7 +3360,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[1]) + if (atomic_read(&rdev->irq.pflip[1])) radeon_crtc_handle_flip(rdev, 1); rdev->irq.stat_regs.r600.disp_int &= ~LB_D2_VBLANK_INTERRUPT; DRM_DEBUG("IH: D2 vblank\n"); diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 353a183..126bac5 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -617,11 +617,9 @@ union radeon_irq_stat_regs { struct radeon_irq { bool installed; spinlock_t lock; - bool sw_int[RADEON_NUM_RINGS]; - int sw_refcount[RADEON_NUM_RINGS]; + atomic_t ring_int[RADEON_NUM_RINGS]; bool crtc_vblank_int[RADEON_MAX_CRTCS]; - bool pflip[RADEON_MAX_CRTCS]; - int pflip_refcount[RADEON_MAX_CRTCS]; + atomic_t pflip[RADEON_MAX_CRTCS]; wait_queue_head_t vblank_queue; bool hpd[RADEON_MAX_HPD_PINS]; bool gui_idle; diff --git a/drivers/gpu/drm/radeon/radeon_irq_kms.c b/drivers/gpu/drm/radeon/radeon_irq_kms.c index c5eb7a1..6664514 100644 --- a/drivers/gpu/drm/radeon/radeon_irq_kms.c +++ b/drivers/gpu/drm/radeon/radeon_irq_kms.c @@ -70,13 +70,13 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev) spin_lock_irqsave(&rdev->irq.lock, irqflags); /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) - rdev->irq.sw_int[i] = false; + atomic_set(&rdev->irq.ring_int[i], 0); rdev->irq.gui_idle = false; for (i = 0; i < RADEON_MAX_HPD_PINS; i++) rdev->irq.hpd[i] = false; for (i = 0; i < RADEON_MAX_CRTCS; i++) { rdev->irq.crtc_vblank_int[i] = false; - rdev->irq.pflip[i] = false; + atomic_set(&rdev->irq.pflip[i], 0); rdev->irq.afmt[i] = false; } radeon_irq_set(rdev); @@ -87,16 +87,7 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev) int radeon_driver_irq_postinstall_kms(struct drm_device *dev) { - struct radeon_device *rdev = dev->dev_private; - unsigned long irqflags; - unsigned i; - dev->max_vblank_count = 0x001fffff; - spin_lock_irqsave(&rdev->irq.lock, irqflags); - for (i = 0; i < RADEON_NUM_RINGS; i++) - rdev->irq.sw_int[i] = true; - radeon_irq_set(rdev); - spin_unlock_irqrestore(&rdev->irq.lock, irqflags); return 0; } @@ -112,13 +103,13 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev) spin_lock_irqsave(&rdev->irq.lock, irqflags); /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) - rdev->irq.sw_int[i] = false; + atomic_set(&rdev->irq.ring_int[i], 0); rdev->irq.gui_idle = false; for (i = 0; i < RADEON_MAX_HPD_PINS; i++) rdev->irq.hpd[i] = false; for (i = 0; i < RADEON_MAX_CRTCS; i++) { rdev->irq.crtc_vblank_int[i] = false; - rdev->irq.pflip[i] = false; + atomic_set(&rdev->irq.pflip[i], 0); rdev->irq.afmt[i] = false; } radeon_irq_set(rdev); @@ -225,25 +216,28 @@ void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev, int ring) { unsigned long irqflags; - spin_lock_irqsave(&rdev->irq.lock, irqflags); - if (rdev->ddev->irq_enabled && (++rdev->irq.sw_refcount[ring] == 1)) { - rdev->irq.sw_int[ring] = true; + if (!rdev->ddev->irq_enabled) + return; + + if (atomic_inc_return(&rdev->irq.ring_int[ring]) == 1) { + spin_lock_irqsave(&rdev->irq.lock, irqflags); radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } - spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev, int ring) { unsigned long irqflags; - spin_lock_irqsave(&rdev->irq.lock, irqflags); - BUG_ON(rdev->ddev->irq_enabled && rdev->irq.sw_refcount[ring] <= 0); - if (rdev->ddev->irq_enabled && (--rdev->irq.sw_refcount[ring] == 0)) { - rdev->irq.sw_int[ring] = false; + if (!rdev->ddev->irq_enabled) + return; + + if (atomic_dec_and_test(&rdev->irq.ring_int[ring])) { + spin_lock_irqsave(&rdev->irq.lock, irqflags); radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } - spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc) @@ -253,12 +247,14 @@ void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc) if (crtc < 0 || crtc >= rdev->num_crtc) return; - spin_lock_irqsave(&rdev->irq.lock, irqflags); - if (rdev->ddev->irq_enabled && (++rdev->irq.pflip_refcount[crtc] == 1)) { - rdev->irq.pflip[crtc] = true; + if (!rdev->ddev->irq_enabled) + return; + + if (atomic_inc_return(&rdev->irq.pflip[crtc]) == 1) { + spin_lock_irqsave(&rdev->irq.lock, irqflags); radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } - spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc) @@ -268,13 +264,14 @@ void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc) if (crtc < 0 || crtc >= rdev->num_crtc) return; - spin_lock_irqsave(&rdev->irq.lock, irqflags); - BUG_ON(rdev->ddev->irq_enabled && rdev->irq.pflip_refcount[crtc] <= 0); - if (rdev->ddev->irq_enabled && (--rdev->irq.pflip_refcount[crtc] == 0)) { - rdev->irq.pflip[crtc] = false; + if (!rdev->ddev->irq_enabled) + return; + + if (atomic_dec_and_test(&rdev->irq.pflip[crtc])) { + spin_lock_irqsave(&rdev->irq.lock, irqflags); radeon_irq_set(rdev); + spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } - spin_unlock_irqrestore(&rdev->irq.lock, irqflags); } void radeon_irq_kms_enable_afmt(struct radeon_device *rdev, int block) diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c index f9aee25..e11bc46 100644 --- a/drivers/gpu/drm/radeon/rs600.c +++ b/drivers/gpu/drm/radeon/rs600.c @@ -564,18 +564,18 @@ int rs600_irq_set(struct radeon_device *rdev) WREG32(R_000040_GEN_INT_CNTL, 0); return -EINVAL; } - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { tmp |= S_000040_SW_INT_EN(1); } if (rdev->irq.gui_idle) { tmp |= S_000040_GUI_IDLE(1); } if (rdev->irq.crtc_vblank_int[0] || - rdev->irq.pflip[0]) { + atomic_read(&rdev->irq.pflip[0])) { mode_int |= S_006540_D1MODE_VBLANK_INT_MASK(1); } if (rdev->irq.crtc_vblank_int[1] || - rdev->irq.pflip[1]) { + atomic_read(&rdev->irq.pflip[1])) { mode_int |= S_006540_D2MODE_VBLANK_INT_MASK(1); } if (rdev->irq.hpd[0]) { @@ -695,7 +695,7 @@ int rs600_irq_process(struct radeon_device *rdev) rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[0]) + if (atomic_read(&rdev->irq.pflip[0])) radeon_crtc_handle_flip(rdev, 0); } if (G_007EDC_LB_D2_VBLANK_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) { @@ -704,7 +704,7 @@ int rs600_irq_process(struct radeon_device *rdev) rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[1]) + if (atomic_read(&rdev->irq.pflip[1])) radeon_crtc_handle_flip(rdev, 1); } if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) { diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index a65045b..34603b3c8 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -3092,45 +3092,45 @@ int si_irq_set(struct radeon_device *rdev) hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN; /* enable CP interrupts on all rings */ - if (rdev->irq.sw_int[RADEON_RING_TYPE_GFX_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("si_irq_set: sw int gfx\n"); cp_int_cntl |= TIME_STAMP_INT_ENABLE; } - if (rdev->irq.sw_int[CAYMAN_RING_TYPE_CP1_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP1_INDEX])) { DRM_DEBUG("si_irq_set: sw int cp1\n"); cp_int_cntl1 |= TIME_STAMP_INT_ENABLE; } - if (rdev->irq.sw_int[CAYMAN_RING_TYPE_CP2_INDEX]) { + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP2_INDEX])) { DRM_DEBUG("si_irq_set: sw int cp2\n"); cp_int_cntl2 |= TIME_STAMP_INT_ENABLE; } if (rdev->irq.crtc_vblank_int[0] || - rdev->irq.pflip[0]) { + atomic_read(&rdev->irq.pflip[0])) { DRM_DEBUG("si_irq_set: vblank 0\n"); crtc1 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[1] || - rdev->irq.pflip[1]) { + atomic_read(&rdev->irq.pflip[1])) { DRM_DEBUG("si_irq_set: vblank 1\n"); crtc2 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[2] || - rdev->irq.pflip[2]) { + atomic_read(&rdev->irq.pflip[2])) { DRM_DEBUG("si_irq_set: vblank 2\n"); crtc3 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[3] || - rdev->irq.pflip[3]) { + atomic_read(&rdev->irq.pflip[3])) { DRM_DEBUG("si_irq_set: vblank 3\n"); crtc4 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[4] || - rdev->irq.pflip[4]) { + atomic_read(&rdev->irq.pflip[4])) { DRM_DEBUG("si_irq_set: vblank 4\n"); crtc5 |= VBLANK_INT_MASK; } if (rdev->irq.crtc_vblank_int[5] || - rdev->irq.pflip[5]) { + atomic_read(&rdev->irq.pflip[5])) { DRM_DEBUG("si_irq_set: vblank 5\n"); crtc6 |= VBLANK_INT_MASK; } @@ -3396,7 +3396,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[0]) + if (atomic_read(&rdev->irq.pflip[0])) radeon_crtc_handle_flip(rdev, 0); rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VBLANK_INTERRUPT; DRM_DEBUG("IH: D1 vblank\n"); @@ -3422,7 +3422,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[1]) + if (atomic_read(&rdev->irq.pflip[1])) radeon_crtc_handle_flip(rdev, 1); rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT; DRM_DEBUG("IH: D2 vblank\n"); @@ -3448,7 +3448,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[2]) + if (atomic_read(&rdev->irq.pflip[2])) radeon_crtc_handle_flip(rdev, 2); rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT; DRM_DEBUG("IH: D3 vblank\n"); @@ -3474,7 +3474,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[3]) + if (atomic_read(&rdev->irq.pflip[3])) radeon_crtc_handle_flip(rdev, 3); rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT; DRM_DEBUG("IH: D4 vblank\n"); @@ -3500,7 +3500,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[4]) + if (atomic_read(&rdev->irq.pflip[4])) radeon_crtc_handle_flip(rdev, 4); rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT; DRM_DEBUG("IH: D5 vblank\n"); @@ -3526,7 +3526,7 @@ restart_ih: rdev->pm.vblank_sync = true; wake_up(&rdev->irq.vblank_queue); } - if (rdev->irq.pflip[5]) + if (atomic_read(&rdev->irq.pflip[5])) radeon_crtc_handle_flip(rdev, 5); rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT; DRM_DEBUG("IH: D6 vblank\n"); -- 1.7.9.5 --------------080300090405010500040404 Content-Type: text/x-patch; name="0009-drm-radeon-replace-cs_mutex-with-vm_mutex-v3.patch" Content-Disposition: attachment; filename*0="0009-drm-radeon-replace-cs_mutex-with-vm_mutex-v3.patch" Content-Transfer-Encoding: quoted-printable >>From 36ff39c4045ee71cd306f8af5f8c2a1c6e998eba Mon Sep 17 00:00:00 2001 From: =3D?UTF-8?q?Christian=3D20K=3DC3=3DB6nig?=3D Date: Wed, 9 May 2012 10:07:08 +0200 Subject: [PATCH 9/9] drm/radeon: replace cs_mutex with vm_mutex v3 MIME-Version: 1.0 Content-Type: text/plain; charset=3DUTF-8 Content-Transfer-Encoding: 8bit Try to remove or replace the cs_mutex with a vm_mutex where it is still needed. v2: fix locking order v3: rebased on drm-next Signed-off-by: Christian K=C3=B6nig --- drivers/gpu/drm/radeon/radeon.h | 44 +-------------------------= ------ drivers/gpu/drm/radeon/radeon_cs.c | 9 +++---- drivers/gpu/drm/radeon/radeon_device.c | 2 +- drivers/gpu/drm/radeon/radeon_gart.c | 24 ++++++++--------- drivers/gpu/drm/radeon/radeon_gem.c | 2 -- 5 files changed, 17 insertions(+), 64 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h index 126bac5..77b4519b 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -159,48 +159,6 @@ static inline int radeon_atrm_get_bios_chunk(uint8_t= *bios, int offset, int len) #endif bool radeon_get_bios(struct radeon_device *rdev); =20 - -/* - * Mutex which allows recursive locking from the same process. - */ -struct radeon_mutex { - struct mutex mutex; - struct task_struct *owner; - int level; -}; - -static inline void radeon_mutex_init(struct radeon_mutex *mutex) -{ - mutex_init(&mutex->mutex); - mutex->owner =3D NULL; - mutex->level =3D 0; -} - -static inline void radeon_mutex_lock(struct radeon_mutex *mutex) -{ - if (mutex_trylock(&mutex->mutex)) { - /* The mutex was unlocked before, so it's ours now */ - mutex->owner =3D current; - } else if (mutex->owner !=3D current) { - /* Another process locked the mutex, take it */ - mutex_lock(&mutex->mutex); - mutex->owner =3D current; - } - /* Otherwise the mutex was already locked by this process */ - - mutex->level++; -} - -static inline void radeon_mutex_unlock(struct radeon_mutex *mutex) -{ - if (--mutex->level > 0) - return; - - mutex->owner =3D NULL; - mutex_unlock(&mutex->mutex); -} - - /* * Dummy page */ @@ -712,6 +670,7 @@ struct radeon_vm_funcs { }; =20 struct radeon_vm_manager { + struct mutex lock; struct list_head lru_vm; uint32_t use_bitmap; struct radeon_sa_manager sa_manager; @@ -1532,7 +1491,6 @@ struct radeon_device { struct radeon_gem gem; struct radeon_pm pm; uint32_t bios_scratch[RADEON_BIOS_NUM_SCRATCH]; - struct radeon_mutex cs_mutex; struct radeon_wb wb; struct radeon_dummy_page dummy_page; bool shutdown; diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/= radeon_cs.c index dd3e234..f1b7527 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -440,6 +440,7 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device= *rdev, return r; } =20 + mutex_lock(&rdev->vm_manager.lock); mutex_lock(&vm->mutex); r =3D radeon_vm_bind(rdev, vm); if (r) { @@ -477,7 +478,8 @@ out: } vm->fence =3D radeon_fence_ref(parser->ib.fence); } - mutex_unlock(&fpriv->vm.mutex); + mutex_unlock(&vm->mutex); + mutex_unlock(&rdev->vm_manager.lock); return r; } =20 @@ -497,9 +499,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *dat= a, struct drm_file *filp) struct radeon_cs_parser parser; int r; =20 - radeon_mutex_lock(&rdev->cs_mutex); if (!rdev->accel_working) { - radeon_mutex_unlock(&rdev->cs_mutex); return -EBUSY; } /* initialize parser */ @@ -513,7 +513,6 @@ int radeon_cs_ioctl(struct drm_device *dev, void *dat= a, struct drm_file *filp) DRM_ERROR("Failed to initialize parser !\n"); radeon_cs_parser_fini(&parser, r); r =3D radeon_cs_handle_lockup(rdev, r); - radeon_mutex_unlock(&rdev->cs_mutex); return r; } r =3D radeon_cs_parser_relocs(&parser); @@ -522,7 +521,6 @@ int radeon_cs_ioctl(struct drm_device *dev, void *dat= a, struct drm_file *filp) DRM_ERROR("Failed to parse relocation %d!\n", r); radeon_cs_parser_fini(&parser, r); r =3D radeon_cs_handle_lockup(rdev, r); - radeon_mutex_unlock(&rdev->cs_mutex); return r; } r =3D radeon_cs_ib_chunk(rdev, &parser); @@ -536,7 +534,6 @@ int radeon_cs_ioctl(struct drm_device *dev, void *dat= a, struct drm_file *filp) out: radeon_cs_parser_fini(&parser, r); r =3D radeon_cs_handle_lockup(rdev, r); - radeon_mutex_unlock(&rdev->cs_mutex); return r; } =20 diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/rad= eon/radeon_device.c index 3c563d1..f654ba8 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -728,7 +728,6 @@ int radeon_device_init(struct radeon_device *rdev, =20 /* mutex initialization are all done here so we * can recall function without having locking issues */ - radeon_mutex_init(&rdev->cs_mutex); mutex_init(&rdev->ring_lock); mutex_init(&rdev->dc_hw_i2c_mutex); atomic_set(&rdev->ih.lock, 0); @@ -741,6 +740,7 @@ int radeon_device_init(struct radeon_device *rdev, if (r) return r; /* initialize vm here */ + mutex_init(&rdev->vm_manager.lock); rdev->vm_manager.use_bitmap =3D 1; rdev->vm_manager.max_pfn =3D 1 << 20; INIT_LIST_HEAD(&rdev->vm_manager.lru_vm); diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeo= n/radeon_gart.c index 59d4493..2b34c1a 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -305,7 +305,7 @@ int radeon_vm_manager_init(struct radeon_device *rdev= ) return r; } =20 -/* cs mutex must be lock */ +/* global mutex must be lock */ static void radeon_vm_unbind_locked(struct radeon_device *rdev, struct radeon_vm *vm) { @@ -356,17 +356,17 @@ int radeon_vm_manager_suspend(struct radeon_device = *rdev) { struct radeon_vm *vm, *tmp; =20 - radeon_mutex_lock(&rdev->cs_mutex); + mutex_lock(&rdev->vm_manager.lock); /* unbind all active vm */ list_for_each_entry_safe(vm, tmp, &rdev->vm_manager.lru_vm, list) { radeon_vm_unbind_locked(rdev, vm); } rdev->vm_manager.funcs->fini(rdev); - radeon_mutex_unlock(&rdev->cs_mutex); + mutex_unlock(&rdev->vm_manager.lock); return radeon_sa_bo_manager_suspend(rdev, &rdev->vm_manager.sa_manager)= ; } =20 -/* cs mutex must be lock */ +/* global mutex must be locked */ void radeon_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm) { mutex_lock(&vm->mutex); @@ -374,7 +374,7 @@ void radeon_vm_unbind(struct radeon_device *rdev, str= uct radeon_vm *vm) mutex_unlock(&vm->mutex); } =20 -/* cs mutex must be lock & vm mutex must be lock */ +/* global and local mutex must be locked */ int radeon_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm) { struct radeon_vm *vm_evict; @@ -478,7 +478,7 @@ int radeon_vm_bo_add(struct radeon_device *rdev, if (last_pfn > vm->last_pfn) { /* release mutex and lock in right order */ mutex_unlock(&vm->mutex); - radeon_mutex_lock(&rdev->cs_mutex); + mutex_lock(&rdev->vm_manager.lock); mutex_lock(&vm->mutex); /* and check again */ if (last_pfn > vm->last_pfn) { @@ -487,7 +487,7 @@ int radeon_vm_bo_add(struct radeon_device *rdev, radeon_vm_unbind_locked(rdev, vm); vm->last_pfn =3D (last_pfn + align) & ~align; } - radeon_mutex_unlock(&rdev->cs_mutex); + mutex_unlock(&rdev->vm_manager.lock); } head =3D &vm->va; last_offset =3D 0; @@ -542,7 +542,7 @@ static u64 radeon_vm_get_addr(struct radeon_device *r= dev, return addr; } =20 -/* object have to be reserved & cs mutex took & vm mutex took */ +/* object have to be reserved & global and local mutex must be locked */ int radeon_vm_bo_update_pte(struct radeon_device *rdev, struct radeon_vm *vm, struct radeon_bo *bo, @@ -601,10 +601,10 @@ int radeon_vm_bo_rmv(struct radeon_device *rdev, if (bo_va =3D=3D NULL) return 0; =20 - radeon_mutex_lock(&rdev->cs_mutex); + mutex_lock(&rdev->vm_manager.lock); mutex_lock(&vm->mutex); radeon_vm_bo_update_pte(rdev, vm, bo, NULL); - radeon_mutex_unlock(&rdev->cs_mutex); + mutex_unlock(&rdev->vm_manager.lock); list_del(&bo_va->vm_list); mutex_unlock(&vm->mutex); list_del(&bo_va->bo_list); @@ -647,10 +647,10 @@ void radeon_vm_fini(struct radeon_device *rdev, str= uct radeon_vm *vm) struct radeon_bo_va *bo_va, *tmp; int r; =20 - radeon_mutex_lock(&rdev->cs_mutex); + mutex_lock(&rdev->vm_manager.lock); mutex_lock(&vm->mutex); radeon_vm_unbind_locked(rdev, vm); - radeon_mutex_unlock(&rdev->cs_mutex); + mutex_unlock(&rdev->vm_manager.lock); =20 /* remove all bo */ r =3D radeon_bo_reserve(rdev->ring_tmp_bo.bo, false); diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon= /radeon_gem.c index f28bd4b..d9b0809 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -159,11 +159,9 @@ void radeon_gem_object_close(struct drm_gem_object *= obj, static int radeon_gem_handle_lockup(struct radeon_device *rdev, int r) { if (r =3D=3D -EDEADLK) { - radeon_mutex_lock(&rdev->cs_mutex); r =3D radeon_gpu_reset(rdev); if (!r) r =3D -EAGAIN; - radeon_mutex_unlock(&rdev->cs_mutex); } return r; } --=20 1.7.9.5 --------------080300090405010500040404 Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel --------------080300090405010500040404--