From patchwork Fri Sep 12 11:52:33 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?Christian_K=C3=B6nig?= X-Patchwork-Id: 4894061 Return-Path: X-Original-To: patchwork-dri-devel@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork2.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.19.201]) by patchwork2.web.kernel.org (Postfix) with ESMTP id 581A4BEEA6 for ; Fri, 12 Sep 2014 11:53:10 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 7E127202C8 for ; Fri, 12 Sep 2014 11:53:07 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by mail.kernel.org (Postfix) with ESMTP id 001BF202E6 for ; Fri, 12 Sep 2014 11:53:04 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 3C9036E696; Fri, 12 Sep 2014 04:53:04 -0700 (PDT) X-Original-To: dri-devel@lists.freedesktop.org Delivered-To: dri-devel@lists.freedesktop.org Received: from pegasos-out.vodafone.de (pegasos-out.vodafone.de [80.84.1.38]) by gabe.freedesktop.org (Postfix) with ESMTP id 07EC66E68C for ; Fri, 12 Sep 2014 04:52:57 -0700 (PDT) Received: from localhost (localhost.localdomain [127.0.0.1]) by pegasos-out.vodafone.de (Rohrpostix1 Daemon) with ESMTP id 75437261406 for ; Fri, 12 Sep 2014 13:52:56 +0200 (CEST) X-Virus-Scanned: amavisd-new at vodafone.de X-Spam-Score: 0.201 X-Spam-Level: X-Spam-Status: No, score=-6.3 required=5.0 tests=BAYES_00,DKIM_SIGNED, RCVD_IN_DNSWL_MED,RP_MATCHES_RCVD,T_DKIM_INVALID,UNPARSEABLE_RELAY autolearn=ham version=3.3.1 Authentication-Results: rohrpostix1.prod.vfnet.de (amavisd-new); dkim=pass header.i=@vodafone.de Received: from pegasos-out.vodafone.de ([127.0.0.1]) by localhost (rohrpostix1.prod.vfnet.de [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id xvB66KzJxNgb for ; Fri, 12 Sep 2014 13:52:52 +0200 (CEST) Received: from smtp-01.vodafone.de (xsmail-dmz5.prod.vfnet.de [10.215.254.36]) by pegasos-out.vodafone.de (Rohrpostix1 Daemon) with ESMTP id CF6F4261432 for ; Fri, 12 Sep 2014 13:52:51 +0200 (CEST) X-DKIM: OpenDKIM Filter v2.6.8 pegasos-out.vodafone.de CF6F4261432 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=vodafone.de; s=mail; t=1410522771; bh=lcXYshLBpKOSbNK3kwFMaadFUU1Ib7YDiOAK51drzrE=; h=From:To:Subject:Date:In-Reply-To:References; b=a2sWwc0S/ew/M3yTX0RGtlGicl8BxQfjkL+/j1Dey10gAbu4zWgpxeUC0eENw3D2P eODVs3jJpsMqDqIbF4z+LPGaOSechGzQBEFK4VgQoFGqIEfpolRWVs4U0y6FTTbtge lv9hZCuAghav2O0iITJDgLl7MH3Ofz+oREYpMnlc= X-DKIM: OpenDKIM Filter v2.0.2 smtp-01.vodafone.de B07B1E5343 X-Virus-Scanned: amavisd-new at vodafone.de Received: from smtp-01.vodafone.de ([127.0.0.1]) by localhost (xsmail-dmz5.prod.vfnet.de [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id UdTTcKxmPvyj for ; Fri, 12 Sep 2014 13:52:43 +0200 (CEST) From: =?UTF-8?q?Christian=20K=C3=B6nig?= To: dri-devel@lists.freedesktop.org Subject: [PATCH 04/11] drm/radeon: split semaphore and sync object handling Date: Fri, 12 Sep 2014 13:52:33 +0200 Message-Id: <1410522760-26204-5-git-send-email-deathsimple@vodafone.de> X-Mailer: git-send-email 1.9.1 In-Reply-To: <1410522760-26204-1-git-send-email-deathsimple@vodafone.de> References: <1410522760-26204-1-git-send-email-deathsimple@vodafone.de> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP From: Christian König Previously we just allocated space for four hardware semaphores in each software semaphore object. Make software semaphore objects represent only one hardware semaphore address again by splitting the sync code into it's own object. Signed-off-by: Christian König --- drivers/gpu/drm/radeon/Makefile | 3 +- drivers/gpu/drm/radeon/cik.c | 18 ++- drivers/gpu/drm/radeon/cik_sdma.c | 18 ++- drivers/gpu/drm/radeon/evergreen_dma.c | 18 ++- drivers/gpu/drm/radeon/r600.c | 18 ++- drivers/gpu/drm/radeon/r600_dma.c | 18 ++- drivers/gpu/drm/radeon/radeon.h | 40 +++--- drivers/gpu/drm/radeon/radeon_cs.c | 7 +- drivers/gpu/drm/radeon/radeon_ib.c | 13 +- drivers/gpu/drm/radeon/radeon_semaphore.c | 139 +-------------------- drivers/gpu/drm/radeon/radeon_sync.c | 198 ++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/radeon_vm.c | 4 +- drivers/gpu/drm/radeon/rv770_dma.c | 18 ++- drivers/gpu/drm/radeon/si_dma.c | 18 ++- 14 files changed, 286 insertions(+), 244 deletions(-) create mode 100644 drivers/gpu/drm/radeon/radeon_sync.c diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index 7d7aed5..8cc9f68 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -80,7 +80,8 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \ r600_dpm.o rs780_dpm.o rv6xx_dpm.o rv770_dpm.o rv730_dpm.o rv740_dpm.o \ rv770_smc.o cypress_dpm.o btc_dpm.o sumo_dpm.o sumo_smc.o trinity_dpm.o \ trinity_smc.o ni_dpm.o si_smc.o si_dpm.o kv_smc.o kv_dpm.o ci_smc.o \ - ci_dpm.o dce6_afmt.o radeon_vm.o radeon_ucode.o radeon_ib.o radeon_mn.o + ci_dpm.o dce6_afmt.o radeon_vm.o radeon_ucode.o radeon_ib.o radeon_mn.o \ + radeon_sync.o # add async DMA block radeon-y += \ diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 16a861d..6dd55ea 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -3970,31 +3970,27 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.blit_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes, control; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 7 + 18); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -4018,12 +4014,12 @@ struct radeon_fence *cik_copy_cpdma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c index dd73246..06602e4 100644 --- a/drivers/gpu/drm/radeon/cik_sdma.c +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -548,31 +548,27 @@ struct radeon_fence *cik_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -593,12 +589,12 @@ struct radeon_fence *cik_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/evergreen_dma.c b/drivers/gpu/drm/radeon/evergreen_dma.c index 946f37d..2958fa8 100644 --- a/drivers/gpu/drm/radeon/evergreen_dma.c +++ b/drivers/gpu/drm/radeon/evergreen_dma.c @@ -110,31 +110,27 @@ struct radeon_fence *evergreen_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xfffff); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -153,12 +149,12 @@ struct radeon_fence *evergreen_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 5e9146b..5205070 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2899,31 +2899,27 @@ struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.blit_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes, tmp; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); r = radeon_ring_lock(rdev, ring, num_loops * 6 + 24); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); @@ -2952,12 +2948,12 @@ struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/r600_dma.c b/drivers/gpu/drm/radeon/r600_dma.c index fc54224..8c39659 100644 --- a/drivers/gpu/drm/radeon/r600_dma.c +++ b/drivers/gpu/drm/radeon/r600_dma.c @@ -447,31 +447,27 @@ struct radeon_fence *r600_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFE); r = radeon_ring_lock(rdev, ring, num_loops * 4 + 8); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -490,12 +486,12 @@ struct radeon_fence *r600_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index a93c9d9..c326fae 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -147,9 +147,6 @@ extern int radeon_bapm; /* number of hw syncs before falling back on blocking */ #define RADEON_NUM_SYNCS 4 -/* number of hw syncs before falling back on blocking */ -#define RADEON_NUM_SYNCS 4 - /* hardcode those limit for now */ #define RADEON_VA_IB_OFFSET (1 << 20) #define RADEON_VA_RESERVED_SIZE (8 << 20) @@ -573,10 +570,9 @@ int radeon_mode_dumb_mmap(struct drm_file *filp, * Semaphores. */ struct radeon_semaphore { - struct radeon_sa_bo *sa_bo; - signed waiters; - uint64_t gpu_addr; - struct radeon_fence *sync_to[RADEON_NUM_RINGS]; + struct radeon_sa_bo *sa_bo; + signed waiters; + uint64_t gpu_addr; }; int radeon_semaphore_create(struct radeon_device *rdev, @@ -585,19 +581,31 @@ bool radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore); bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring, struct radeon_semaphore *semaphore); -void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore, - struct radeon_fence *fence); -void radeon_semaphore_sync_resv(struct radeon_semaphore *semaphore, - struct reservation_object *resv, - bool shared); -int radeon_semaphore_sync_rings(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, - int waiting_ring); void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_semaphore **semaphore, struct radeon_fence *fence); /* + * Synchronization + */ +struct radeon_sync { + struct radeon_semaphore *semaphores[RADEON_NUM_SYNCS]; + struct radeon_fence *sync_to[RADEON_NUM_RINGS]; +}; + +void radeon_sync_create(struct radeon_sync *sync); +void radeon_sync_fence(struct radeon_sync *sync, + struct radeon_fence *fence); +void radeon_sync_resv(struct radeon_sync *sync, + struct reservation_object *resv, + bool shared); +int radeon_sync_rings(struct radeon_device *rdev, + struct radeon_sync *sync, + int waiting_ring); +void radeon_sync_free(struct radeon_device *rdev, struct radeon_sync *sync, + struct radeon_fence *fence); + +/* * GART structures, functions & helpers */ struct radeon_mc; @@ -810,7 +818,7 @@ struct radeon_ib { struct radeon_fence *fence; struct radeon_vm *vm; bool is_const_ib; - struct radeon_semaphore *semaphore; + struct radeon_sync sync; }; struct radeon_ring { diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index f662de4..0619c32 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -260,8 +260,7 @@ static void radeon_cs_sync_rings(struct radeon_cs_parser *p) continue; resv = p->relocs[i].robj->tbo.resv; - radeon_semaphore_sync_resv(p->ib.semaphore, resv, - p->relocs[i].tv.shared); + radeon_sync_resv(&p->ib.sync, resv, p->relocs[i].tv.shared); } } @@ -281,9 +280,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) INIT_LIST_HEAD(&p->validated); p->idx = 0; p->ib.sa_bo = NULL; - p->ib.semaphore = NULL; p->const_ib.sa_bo = NULL; - p->const_ib.semaphore = NULL; p->chunk_ib_idx = -1; p->chunk_relocs_idx = -1; p->chunk_flags_idx = -1; @@ -566,7 +563,7 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev, goto out; } radeon_cs_sync_rings(parser); - radeon_semaphore_sync_fence(parser->ib.semaphore, vm->fence); + radeon_sync_fence(&parser->ib.sync, vm->fence); if ((rdev->family >= CHIP_TAHITI) && (parser->chunk_const_ib_idx != -1)) { diff --git a/drivers/gpu/drm/radeon/radeon_ib.c b/drivers/gpu/drm/radeon/radeon_ib.c index 3f39fcc..56a1704 100644 --- a/drivers/gpu/drm/radeon/radeon_ib.c +++ b/drivers/gpu/drm/radeon/radeon_ib.c @@ -64,10 +64,7 @@ int radeon_ib_get(struct radeon_device *rdev, int ring, return r; } - r = radeon_semaphore_create(rdev, &ib->semaphore); - if (r) { - return r; - } + radeon_sync_create(&ib->sync); ib->ring = ring; ib->fence = NULL; @@ -96,7 +93,7 @@ int radeon_ib_get(struct radeon_device *rdev, int ring, */ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib) { - radeon_semaphore_free(rdev, &ib->semaphore, ib->fence); + radeon_sync_free(rdev, &ib->sync, ib->fence); radeon_sa_bo_free(rdev, &ib->sa_bo, ib->fence); radeon_fence_unref(&ib->fence); } @@ -145,11 +142,11 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib, if (ib->vm) { struct radeon_fence *vm_id_fence; vm_id_fence = radeon_vm_grab_id(rdev, ib->vm, ib->ring); - radeon_semaphore_sync_fence(ib->semaphore, vm_id_fence); + radeon_sync_fence(&ib->sync, vm_id_fence); } /* sync with other rings */ - r = radeon_semaphore_sync_rings(rdev, ib->semaphore, ib->ring); + r = radeon_sync_rings(rdev, &ib->sync, ib->ring); if (r) { dev_err(rdev->dev, "failed to sync rings (%d)\n", r); radeon_ring_unlock_undo(rdev, ring); @@ -161,7 +158,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib, if (const_ib) { radeon_ring_ib_execute(rdev, const_ib->ring, const_ib); - radeon_semaphore_free(rdev, &const_ib->semaphore, NULL); + radeon_sync_free(rdev, &const_ib->sync, NULL); } radeon_ring_ib_execute(rdev, ib->ring, ib); r = radeon_fence_emit(rdev, &ib->fence, ib->ring); diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c index 5bfbd83..e6ad54c 100644 --- a/drivers/gpu/drm/radeon/radeon_semaphore.c +++ b/drivers/gpu/drm/radeon/radeon_semaphore.c @@ -34,15 +34,14 @@ int radeon_semaphore_create(struct radeon_device *rdev, struct radeon_semaphore **semaphore) { - uint32_t *cpu_addr; - int i, r; + int r; *semaphore = kmalloc(sizeof(struct radeon_semaphore), GFP_KERNEL); if (*semaphore == NULL) { return -ENOMEM; } - r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, &(*semaphore)->sa_bo, - 8 * RADEON_NUM_SYNCS, 8); + r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, + &(*semaphore)->sa_bo, 8, 8); if (r) { kfree(*semaphore); *semaphore = NULL; @@ -51,12 +50,7 @@ int radeon_semaphore_create(struct radeon_device *rdev, (*semaphore)->waiters = 0; (*semaphore)->gpu_addr = radeon_sa_bo_gpu_addr((*semaphore)->sa_bo); - cpu_addr = radeon_sa_bo_cpu_addr((*semaphore)->sa_bo); - for (i = 0; i < RADEON_NUM_SYNCS; ++i) - cpu_addr[i] = 0; - - for (i = 0; i < RADEON_NUM_RINGS; ++i) - (*semaphore)->sync_to[i] = NULL; + *((uint64_t *)radeon_sa_bo_cpu_addr((*semaphore)->sa_bo)) = 0; return 0; } @@ -95,131 +89,6 @@ bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ridx, return false; } -/** - * radeon_semaphore_sync_fence - use the semaphore to sync to a fence - * - * @semaphore: semaphore object to add fence to - * @fence: fence to sync to - * - * Sync to the fence using this semaphore object - */ -void radeon_semaphore_sync_fence(struct radeon_semaphore *semaphore, - struct radeon_fence *fence) -{ - struct radeon_fence *other; - - if (!fence) - return; - - other = semaphore->sync_to[fence->ring]; - semaphore->sync_to[fence->ring] = radeon_fence_later(fence, other); -} - -/** - * radeon_semaphore_sync_to - use the semaphore to sync to a reservation object - * - * @sema: semaphore object to add fence from reservation object to - * @resv: reservation object with embedded fence - * @shared: true if we should onyl sync to the exclusive fence - * - * Sync to the fence using this semaphore object - */ -void radeon_semaphore_sync_resv(struct radeon_semaphore *sema, - struct reservation_object *resv, - bool shared) -{ - struct reservation_object_list *flist; - struct fence *f; - unsigned i; - - /* always sync to the exclusive fence */ - f = reservation_object_get_excl(resv); - radeon_semaphore_sync_fence(sema, (struct radeon_fence*)f); - - flist = reservation_object_get_list(resv); - if (shared || !flist) - return; - - for (i = 0; i < flist->shared_count; ++i) { - f = rcu_dereference_protected(flist->shared[i], - reservation_object_held(resv)); - radeon_semaphore_sync_fence(sema, (struct radeon_fence*)f); - } -} - -/** - * radeon_semaphore_sync_rings - sync ring to all registered fences - * - * @rdev: radeon_device pointer - * @semaphore: semaphore object to use for sync - * @ring: ring that needs sync - * - * Ensure that all registered fences are signaled before letting - * the ring continue. The caller must hold the ring lock. - */ -int radeon_semaphore_sync_rings(struct radeon_device *rdev, - struct radeon_semaphore *semaphore, - int ring) -{ - unsigned count = 0; - int i, r; - - for (i = 0; i < RADEON_NUM_RINGS; ++i) { - struct radeon_fence *fence = semaphore->sync_to[i]; - - /* check if we really need to sync */ - if (!radeon_fence_need_sync(fence, ring)) - continue; - - /* prevent GPU deadlocks */ - if (!rdev->ring[i].ready) { - dev_err(rdev->dev, "Syncing to a disabled ring!"); - return -EINVAL; - } - - if (++count > RADEON_NUM_SYNCS) { - /* not enough room, wait manually */ - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - /* allocate enough space for sync command */ - r = radeon_ring_alloc(rdev, &rdev->ring[i], 16); - if (r) { - return r; - } - - /* emit the signal semaphore */ - if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) { - /* signaling wasn't successful wait manually */ - radeon_ring_undo(&rdev->ring[i]); - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - /* we assume caller has already allocated space on waiters ring */ - if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) { - /* waiting wasn't successful wait manually */ - radeon_ring_undo(&rdev->ring[i]); - r = radeon_fence_wait(fence, false); - if (r) - return r; - continue; - } - - radeon_ring_commit(rdev, &rdev->ring[i], false); - radeon_fence_note_sync(fence, ring); - - semaphore->gpu_addr += 8; - } - - return 0; -} - void radeon_semaphore_free(struct radeon_device *rdev, struct radeon_semaphore **semaphore, struct radeon_fence *fence) diff --git a/drivers/gpu/drm/radeon/radeon_sync.c b/drivers/gpu/drm/radeon/radeon_sync.c new file mode 100644 index 0000000..77b7401 --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_sync.c @@ -0,0 +1,198 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + */ +/* + * Authors: + * Christian König + */ + +#include +#include "radeon.h" +#include "radeon_trace.h" + +/** + * radeon_sync_create - zero init sync object + * + * @sync: sync object to initialize + * + * Just clear the sync object for now. + */ +void radeon_sync_create(struct radeon_sync *sync) +{ + unsigned i; + + for (i = 0; i < RADEON_NUM_SYNCS; ++i) + sync->semaphores[i] = NULL; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) + sync->sync_to[i] = NULL; +} + +/** + * radeon_sync_fence - use the semaphore to sync to a fence + * + * @sync: sync object to add fence to + * @fence: fence to sync to + * + * Sync to the fence using the semaphore objects + */ +void radeon_sync_fence(struct radeon_sync *sync, + struct radeon_fence *fence) +{ + struct radeon_fence *other; + + if (!fence) + return; + + other = sync->sync_to[fence->ring]; + sync->sync_to[fence->ring] = radeon_fence_later(fence, other); +} + +/** + * radeon_sync_resv - use the semaphores to sync to a reservation object + * + * @sync: sync object to add fences from reservation object to + * @resv: reservation object with embedded fence + * @shared: true if we should onyl sync to the exclusive fence + * + * Sync to the fence using the semaphore objects + */ +void radeon_sync_resv(struct radeon_sync *sync, + struct reservation_object *resv, + bool shared) +{ + struct reservation_object_list *flist; + struct fence *f; + unsigned i; + + /* always sync to the exclusive fence */ + f = reservation_object_get_excl(resv); + radeon_sync_fence(sync, (struct radeon_fence*)f); + + flist = reservation_object_get_list(resv); + if (shared || !flist) + return; + + for (i = 0; i < flist->shared_count; ++i) { + f = rcu_dereference_protected(flist->shared[i], + reservation_object_held(resv)); + radeon_sync_fence(sync, (struct radeon_fence*)f); + } +} + +/** + * radeon_sync_rings - sync ring to all registered fences + * + * @rdev: radeon_device pointer + * @sync: sync object to use + * @ring: ring that needs sync + * + * Ensure that all registered fences are signaled before letting + * the ring continue. The caller must hold the ring lock. + */ +int radeon_sync_rings(struct radeon_device *rdev, + struct radeon_sync *sync, + int ring) +{ + unsigned count = 0; + int i, r; + + for (i = 0; i < RADEON_NUM_RINGS; ++i) { + struct radeon_fence *fence = sync->sync_to[i]; + struct radeon_semaphore *semaphore; + + /* check if we really need to sync */ + if (!radeon_fence_need_sync(fence, ring)) + continue; + + /* prevent GPU deadlocks */ + if (!rdev->ring[i].ready) { + dev_err(rdev->dev, "Syncing to a disabled ring!"); + return -EINVAL; + } + + if (count >= RADEON_NUM_SYNCS) { + /* not enough room, wait manually */ + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + r = radeon_semaphore_create(rdev, &semaphore); + if (r) + return r; + + sync->semaphores[count++] = semaphore; + + /* allocate enough space for sync command */ + r = radeon_ring_alloc(rdev, &rdev->ring[i], 16); + if (r) + return r; + + /* emit the signal semaphore */ + if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) { + /* signaling wasn't successful wait manually */ + radeon_ring_undo(&rdev->ring[i]); + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + + /* we assume caller has already allocated space on waiters ring */ + if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) { + /* waiting wasn't successful wait manually */ + radeon_ring_undo(&rdev->ring[i]); + r = radeon_fence_wait(fence, false); + if (r) + return r; + continue; + } + + radeon_ring_commit(rdev, &rdev->ring[i], false); + radeon_fence_note_sync(fence, ring); + } + + return 0; +} + +/** + * radeon_sync_free - free the sync object + * + * @rdev: radeon_device pointer + * @sync: sync object to use + * @fence: fence to use for the free + * + * Free the sync object by freeing all semaphores in it. + */ +void radeon_sync_free(struct radeon_device *rdev, + struct radeon_sync *sync, + struct radeon_fence *fence) +{ + unsigned i; + + for (i = 0; i < RADEON_NUM_SYNCS; ++i) + radeon_semaphore_free(rdev, &sync->semaphores[i], fence); +} diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index f684c18..411146a 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -699,7 +699,7 @@ int radeon_vm_update_page_directory(struct radeon_device *rdev, if (ib.length_dw != 0) { radeon_asic_vm_pad_ib(rdev, &ib); - radeon_semaphore_sync_resv(ib.semaphore, pd->tbo.resv, false); + radeon_sync_resv(&ib.sync, pd->tbo.resv, false); WARN_ON(ib.length_dw > ndw); r = radeon_ib_schedule(rdev, &ib, NULL, false); if (r) { @@ -825,7 +825,7 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev, unsigned nptes; uint64_t pte; - radeon_semaphore_sync_resv(ib->semaphore, pt->tbo.resv, false); + radeon_sync_resv(&ib->sync, pt->tbo.resv, false); if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; diff --git a/drivers/gpu/drm/radeon/rv770_dma.c b/drivers/gpu/drm/radeon/rv770_dma.c index c112764..564822c 100644 --- a/drivers/gpu/drm/radeon/rv770_dma.c +++ b/drivers/gpu/drm/radeon/rv770_dma.c @@ -44,31 +44,27 @@ struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_dw, cur_size_in_dw; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFF); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 8); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_dw = size_in_dw; @@ -87,12 +83,12 @@ struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; } diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c index 6573843..0b7d3c5 100644 --- a/drivers/gpu/drm/radeon/si_dma.c +++ b/drivers/gpu/drm/radeon/si_dma.c @@ -226,31 +226,27 @@ struct radeon_fence *si_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct reservation_object *resv) { - struct radeon_semaphore *sem = NULL; struct radeon_fence *fence; + struct radeon_sync sync; int ring_index = rdev->asic->copy.dma_ring_index; struct radeon_ring *ring = &rdev->ring[ring_index]; u32 size_in_bytes, cur_size_in_bytes; int i, num_loops; int r = 0; - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return ERR_PTR(r); - } + radeon_sync_create(&sync); size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); num_loops = DIV_ROUND_UP(size_in_bytes, 0xfffff); r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); if (r) { DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } - radeon_semaphore_sync_resv(sem, resv, false); - radeon_semaphore_sync_rings(rdev, sem, ring->idx); + radeon_sync_resv(&sync, resv, false); + radeon_sync_rings(rdev, &sync, ring->idx); for (i = 0; i < num_loops; i++) { cur_size_in_bytes = size_in_bytes; @@ -269,12 +265,12 @@ struct radeon_fence *si_copy_dma(struct radeon_device *rdev, r = radeon_fence_emit(rdev, &fence, ring->idx); if (r) { radeon_ring_unlock_undo(rdev, ring); - radeon_semaphore_free(rdev, &sem, NULL); + radeon_sync_free(rdev, &sync, NULL); return ERR_PTR(r); } radeon_ring_unlock_commit(rdev, ring, false); - radeon_semaphore_free(rdev, &sem, fence); + radeon_sync_free(rdev, &sync, fence); return fence; }