From patchwork Tue Dec 1 20:13:51 2009 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 64032 Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by demeter.kernel.org (8.14.2/8.14.2) with ESMTP id nB1KUqIb010405 for ; Tue, 1 Dec 2009 20:30:52 GMT Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 76FDB9F0DB; Tue, 1 Dec 2009 12:30:52 -0800 (PST) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org X-Greylist: delayed 1000 seconds by postgrey-1.30 at gabe; Tue, 01 Dec 2009 12:30:48 PST Received: from fireflyinternet.com (smtp.fireflyinternet.com [94.30.43.99]) by gabe.freedesktop.org (Postfix) with ESMTP id EC17C9E808 for ; Tue, 1 Dec 2009 12:30:48 -0800 (PST) Received: from localhost.localdomain (unverified [78.156.66.37]) by fireflyinternet.com (Firefly Internet SMTP) with ESMTP id 66536379-1948518 for multiple; Tue, 01 Dec 2009 20:14:00 +0000 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Tue, 1 Dec 2009 20:13:51 +0000 Message-Id: <1259698431-15149-1-git-send-email-chris@chris-wilson.co.uk> X-Mailer: git-send-email 1.6.3.3 In-Reply-To: <20090918154255.4e56ef1a@jbarnes-g45> References: <20090918154255.4e56ef1a@jbarnes-g45> X-Originating-IP: 78.156.66.37 Subject: [Intel-gfx] [PATCH] drm/i915: execbuf2 support X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.9 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Sender: intel-gfx-bounces@lists.freedesktop.org Errors-To: intel-gfx-bounces@lists.freedesktop.org diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index fe89d0c..663016f 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -813,9 +813,13 @@ static int i915_getparam(struct drm_device *dev, void *data, case I915_PARAM_HAS_PAGEFLIPPING: value = 1; break; + case I915_PARAM_HAS_EXECBUF2: + /* depends on GEM */ + value = dev_priv->has_gem; + break; default: DRM_DEBUG_DRIVER("Unknown parameter %d\n", - param->param); + param->param); return -EINVAL; } @@ -1652,6 +1656,7 @@ struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER2, i915_gem_execbuffer2, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH), diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index ca1ba42..c3922b5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -784,6 +784,8 @@ int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); int i915_gem_execbuffer(struct drm_device *dev, void *data, struct drm_file *file_priv); +int i915_gem_execbuffer2(struct drm_device *dev, void *data, + struct drm_file *file_priv); int i915_gem_pin_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); int i915_gem_unpin_ioctl(struct drm_device *dev, void *data, @@ -850,6 +852,9 @@ void i915_gem_shrinker_exit(void); void i915_gem_detect_bit_6_swizzle(struct drm_device *dev); void i915_gem_object_do_bit_17_swizzle(struct drm_gem_object *obj); void i915_gem_object_save_bit_17_swizzle(struct drm_gem_object *obj); +bool i915_tiling_ok(struct drm_device *dev, int stride, int size, + int tiling_mode); +bool i915_obj_fenceable(struct drm_device *dev, struct drm_gem_object *obj); /* i915_gem_debug.c */ void i915_gem_dump_object(struct drm_gem_object *obj, int len, diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 214fb18..d78e32d 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3197,7 +3197,7 @@ i915_gem_object_set_cpu_read_domain_range(struct drm_gem_object *obj, static int i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, struct drm_file *file_priv, - struct drm_i915_gem_exec_object *entry, + struct drm_i915_gem_exec_object2 *entry, struct drm_i915_gem_relocation_entry *relocs) { struct drm_device *dev = obj->dev; @@ -3205,12 +3205,35 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, struct drm_i915_gem_object *obj_priv = obj->driver_private; int i, ret; void __iomem *reloc_page; + bool need_fence; + + need_fence = entry->flags & EXEC_OBJECT_NEEDS_FENCE && + obj_priv->tiling_mode != I915_TILING_NONE; + + /* Check fence reg constraints and rebind if necessary */ + if (need_fence && !i915_obj_fenceable(dev, obj)) + i915_gem_object_unbind(obj); /* Choose the GTT offset for our buffer and put it there. */ ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment); if (ret) return ret; + /* + * Pre-965 chips need a fence register set up in order to + * properly handle blits to/from tiled surfaces. + */ + if (need_fence) { + ret = i915_gem_object_get_fence_reg(obj); + if (ret != 0) { + if (ret != -EBUSY && ret != -ERESTARTSYS) + DRM_ERROR("Failure to install fence: %d\n", + ret); + i915_gem_object_unpin(obj); + return ret; + } + } + entry->offset = obj_priv->gtt_offset; /* Apply the relocations, using the GTT aperture to avoid cache @@ -3372,7 +3395,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, */ static int i915_dispatch_gem_execbuffer(struct drm_device *dev, - struct drm_i915_gem_execbuffer *exec, + struct drm_i915_gem_execbuffer2 *exec, struct drm_clip_rect *cliprects, uint64_t exec_offset) { @@ -3462,7 +3485,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file_priv) } static int -i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list, +i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object2 *exec_list, uint32_t buffer_count, struct drm_i915_gem_relocation_entry **relocs) { @@ -3502,7 +3525,7 @@ i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list, } static int -i915_gem_put_relocs_to_user(struct drm_i915_gem_exec_object *exec_list, +i915_gem_put_relocs_to_user(struct drm_i915_gem_exec_object2 *exec_list, uint32_t buffer_count, struct drm_i915_gem_relocation_entry *relocs) { @@ -3535,7 +3558,7 @@ err: } static int -i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer *exec, +i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer2 *exec, uint64_t exec_offset) { uint32_t exec_start, exec_len; @@ -3588,18 +3611,18 @@ i915_gem_wait_for_pending_flip(struct drm_device *dev, } int -i915_gem_execbuffer(struct drm_device *dev, void *data, - struct drm_file *file_priv) +i915_gem_do_execbuffer(struct drm_device *dev, void *data, + struct drm_file *file_priv, + struct drm_i915_gem_execbuffer2 *args, + struct drm_i915_gem_exec_object2 *exec_list) { drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_i915_gem_execbuffer *args = data; - struct drm_i915_gem_exec_object *exec_list = NULL; struct drm_gem_object **object_list = NULL; struct drm_gem_object *batch_obj; struct drm_i915_gem_object *obj_priv; struct drm_clip_rect *cliprects = NULL; struct drm_i915_gem_relocation_entry *relocs; - int ret, ret2, i, pinned = 0; + int ret = 0, ret2, i, pinned = 0; uint64_t exec_offset; uint32_t seqno, flush_domains, reloc_index; int pin_tries, flips; @@ -3623,15 +3646,6 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, ret = -ENOMEM; goto pre_mutex_err; } - ret = copy_from_user(exec_list, - (struct drm_i915_relocation_entry __user *) - (uintptr_t) args->buffers_ptr, - sizeof(*exec_list) * args->buffer_count); - if (ret != 0) { - DRM_ERROR("copy %d exec entries failed %d\n", - args->buffer_count, ret); - goto pre_mutex_err; - } if (args->num_cliprects != 0) { cliprects = kcalloc(args->num_cliprects, sizeof(*cliprects), @@ -3885,20 +3899,6 @@ err: mutex_unlock(&dev->struct_mutex); - if (!ret) { - /* Copy the new buffer offsets back to the user's exec list. */ - ret = copy_to_user((struct drm_i915_relocation_entry __user *) - (uintptr_t) args->buffers_ptr, - exec_list, - sizeof(*exec_list) * args->buffer_count); - if (ret) { - ret = -EFAULT; - DRM_ERROR("failed to copy %d exec entries " - "back to user (%d)\n", - args->buffer_count, ret); - } - } - /* Copy the updated relocations out regardless of current error * state. Failure to update the relocs would mean that the next * time userland calls execbuf, it would do so with presumed offset @@ -3915,12 +3915,156 @@ err: pre_mutex_err: drm_free_large(object_list); - drm_free_large(exec_list); kfree(cliprects); return ret; } +/* + * Legacy execbuffer just creates an exec2 list from the original exec object + * list array and passes it to the real function. + */ +int +i915_gem_execbuffer(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_execbuffer *args = data; + struct drm_i915_gem_execbuffer2 exec2; + struct drm_i915_gem_exec_object *exec_list = NULL; + struct drm_i915_gem_exec_object2 *exec2_list = NULL; + int ret, i; + +#if WATCH_EXEC + DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n", + (int) args->buffers_ptr, args->buffer_count, args->batch_len); +#endif + + if (args->buffer_count < 1) { + DRM_ERROR("execbuf with %d buffers\n", args->buffer_count); + return -EINVAL; + } + + /* Copy in the exec list from userland */ + exec_list = drm_malloc_ab(sizeof(*exec_list), args->buffer_count); + exec2_list = drm_malloc_ab(sizeof(*exec2_list), args->buffer_count); + if (exec_list == NULL || exec2_list == NULL) { + DRM_ERROR("Failed to allocate exec list for %d buffers\n", + args->buffer_count); + drm_free_large(exec_list); + drm_free_large(exec2_list); + return -ENOMEM; + } + ret = copy_from_user(exec_list, + (struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + sizeof(*exec_list) * args->buffer_count); + if (ret != 0) { + DRM_ERROR("copy %d exec entries failed %d\n", + args->buffer_count, ret); + drm_free_large(exec_list); + drm_free_large(exec2_list); + return -EFAULT; + } + + for (i = 0; i < args->buffer_count; i++) { + exec2_list[i].handle = exec_list[i].handle; + exec2_list[i].relocation_count = exec_list[i].relocation_count; + exec2_list[i].relocs_ptr = exec_list[i].relocs_ptr; + exec2_list[i].alignment = exec_list[i].alignment; + exec2_list[i].offset = exec_list[i].offset; + if (!IS_I965G(dev)) + exec2_list[i].flags = EXEC_OBJECT_NEEDS_FENCE; + else + exec2_list[i].flags = 0; + } + + exec2.buffers_ptr = args->buffers_ptr; + exec2.buffer_count = args->buffer_count; + exec2.batch_start_offset = args->batch_start_offset; + exec2.batch_len = args->batch_len; + exec2.DR1 = args->DR1; + exec2.DR4 = args->DR4; + exec2.num_cliprects = args->num_cliprects; + exec2.cliprects_ptr = args->cliprects_ptr; + exec2.flags = 0; + + ret = i915_gem_do_execbuffer(dev, data, file_priv, &exec2, exec2_list); + if (!ret) { + /* Copy the new buffer offsets back to the user's exec list. */ + for (i = 0; i < args->buffer_count; i++) + exec_list[i].offset = exec2_list[i].offset; + /* ... and back out to userspace */ + ret = copy_to_user((struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + exec_list, + sizeof(*exec_list) * args->buffer_count); + if (ret) { + ret = -EFAULT; + DRM_ERROR("failed to copy %d exec entries " + "back to user (%d)\n", + args->buffer_count, ret); + } + } + + drm_free_large(exec_list); + drm_free_large(exec2_list); + return ret; +} + +int +i915_gem_execbuffer2(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_execbuffer2 *args = data; + struct drm_i915_gem_exec_object2 *exec2_list = NULL; + int ret; + +#if WATCH_EXEC + DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n", + (int) args->buffers_ptr, args->buffer_count, args->batch_len); +#endif + + if (args->buffer_count < 1) { + DRM_ERROR("execbuf2 with %d buffers\n", args->buffer_count); + return -EINVAL; + } + + exec2_list = drm_malloc_ab(sizeof(*exec2_list), args->buffer_count); + if (exec2_list == NULL) { + DRM_ERROR("Failed to allocate exec list for %d buffers\n", + args->buffer_count); + return -ENOMEM; + } + ret = copy_from_user(exec2_list, + (struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + sizeof(*exec2_list) * args->buffer_count); + if (ret != 0) { + DRM_ERROR("copy %d exec entries failed %d\n", + args->buffer_count, ret); + drm_free_large(exec2_list); + return -EFAULT; + } + + ret = i915_gem_do_execbuffer(dev, data, file_priv, args, exec2_list); + if (!ret) { + /* Copy the new buffer offsets back to the user's exec list. */ + ret = copy_to_user((struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + exec2_list, + sizeof(*exec2_list) * args->buffer_count); + if (ret) { + ret = -EFAULT; + DRM_ERROR("failed to copy %d exec entries " + "back to user (%d)\n", + args->buffer_count, ret); + } + } + + drm_free_large(exec2_list); + return ret; +} + int i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment) { @@ -3934,19 +4078,7 @@ i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment) if (ret) return ret; } - /* - * Pre-965 chips need a fence register set up in order to - * properly handle tiled surfaces. - */ - if (!IS_I965G(dev) && obj_priv->tiling_mode != I915_TILING_NONE) { - ret = i915_gem_object_get_fence_reg(obj); - if (ret != 0) { - if (ret != -EBUSY && ret != -ERESTARTSYS) - DRM_ERROR("Failure to install fence: %d\n", - ret); - return ret; - } - } + obj_priv->pin_count++; /* If the object is not active and not pending a flush, diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 0c8df96..50bb487 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -304,35 +304,39 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) /** - * Returns the size of the fence for a tiled object of the given size. + * Returns whether an object is currently fenceable. If not, it may need + * to be unbound and have its pitch adjusted. */ -static int -i915_get_fence_size(struct drm_device *dev, int size) +bool +i915_obj_fenceable(struct drm_device *dev, struct drm_gem_object *obj) { - int i; - int start; + struct drm_i915_gem_object *obj_priv = obj->driver_private; if (IS_I965G(dev)) { /* The 965 can have fences at any page boundary. */ - return ALIGN(size, 4096); + if (obj->size & 4095) + return false; + return true; + } else if (IS_I9XX(dev)) { + if (obj_priv->gtt_offset & ~I915_FENCE_START_MASK) + return false; } else { - /* Align the size to a power of two greater than the smallest - * fence size. - */ - if (IS_I9XX(dev)) - start = 1024 * 1024; - else - start = 512 * 1024; + if (obj_priv->gtt_offset & ~I830_FENCE_START_MASK) + return false; + } - for (i = start; i < size; i <<= 1) - ; + /* Power of two sized... */ + if (obj->size & (obj->size - 1)) + return false; - return i; - } + /* Objects must be size aligned as well */ + if (obj_priv->gtt_offset & (obj->size - 1)) + return false; + return true; } /* Check pitch constriants for all chips & tiling formats */ -static bool +bool i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode) { int tile_width; @@ -384,12 +388,6 @@ i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode) if (stride & (stride - 1)) return false; - /* We don't 0handle the aperture area covered by the fence being bigger - * than the object size. - */ - if (i915_get_fence_size(dev, size) != size) - return false; - return true; } diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index 02045ed..cb61ec6 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -188,6 +188,7 @@ typedef struct _drm_i915_sarea { #define DRM_I915_GEM_MADVISE 0x26 #define DRM_I915_OVERLAY_PUT_IMAGE 0x27 #define DRM_I915_OVERLAY_ATTRS 0x28 +#define DRM_I915_GEM_EXECBUFFER2 0x29 #define DRM_IOCTL_I915_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t) #define DRM_IOCTL_I915_FLUSH DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH) @@ -207,6 +208,7 @@ typedef struct _drm_i915_sarea { #define DRM_IOCTL_I915_VBLANK_SWAP DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_VBLANK_SWAP, drm_i915_vblank_swap_t) #define DRM_IOCTL_I915_GEM_INIT DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_INIT, struct drm_i915_gem_init) #define DRM_IOCTL_I915_GEM_EXECBUFFER DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER, struct drm_i915_gem_execbuffer) +#define DRM_IOCTL_I915_GEM_EXECBUFFER2 DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER2, struct drm_i915_gem_execbuffer2) #define DRM_IOCTL_I915_GEM_PIN DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_PIN, struct drm_i915_gem_pin) #define DRM_IOCTL_I915_GEM_UNPIN DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_UNPIN, struct drm_i915_gem_unpin) #define DRM_IOCTL_I915_GEM_BUSY DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_BUSY, struct drm_i915_gem_busy) @@ -272,6 +274,7 @@ typedef struct drm_i915_irq_wait { #define I915_PARAM_NUM_FENCES_AVAIL 6 #define I915_PARAM_HAS_OVERLAY 7 #define I915_PARAM_HAS_PAGEFLIPPING 8 +#define I915_PARAM_HAS_EXECBUF2 9 typedef struct drm_i915_getparam { int param; @@ -567,6 +570,57 @@ struct drm_i915_gem_execbuffer { __u64 cliprects_ptr; }; +struct drm_i915_gem_exec_object2 { + /** + * User's handle for a buffer to be bound into the GTT for this + * operation. + */ + __u32 handle; + + /** Number of relocations to be performed on this buffer */ + __u32 relocation_count; + /** + * Pointer to array of struct drm_i915_gem_relocation_entry containing + * the relocations to be performed in this buffer. + */ + __u64 relocs_ptr; + + /** Required alignment in graphics aperture */ + __u64 alignment; + + /** + * Returned value of the updated offset of the object, for future + * presumed_offset writes. + */ + __u64 offset; + +#define EXEC_OBJECT_NEEDS_FENCE (1<<0) + __u64 flags; + __u64 rsvd1; + __u64 rsvd2; +}; + +struct drm_i915_gem_execbuffer2 { + /** + * List of gem_exec_object2 structs + */ + __u64 buffers_ptr; + __u32 buffer_count; + + /** Offset in the batchbuffer to start execution from. */ + __u32 batch_start_offset; + /** Bytes used in batchbuffer from batch_start_offset */ + __u32 batch_len; + __u32 DR1; + __u32 DR4; + __u32 num_cliprects; + /** This is a struct drm_clip_rect *cliprects */ + __u64 cliprects_ptr; + __u64 flags; /* currently unused */ + __u64 rsvd1; + __u64 rsvd2; +}; + struct drm_i915_gem_pin { /** Handle of the buffer to be pinned. */ __u32 handle;