diff options
Diffstat (limited to 'drivers/gpu/drm/i915/i915_gem.c')
| -rw-r--r-- | drivers/gpu/drm/i915/i915_gem.c | 414 |
1 files changed, 332 insertions, 82 deletions
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index abfc27b0c2e..0c67924ca80 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1288,6 +1288,7 @@ i915_gem_create_mmap_offset(struct drm_gem_object *obj) list->hash.key = list->file_offset_node->start; if (drm_ht_insert_item(&mm->offset_hash, &list->hash)) { DRM_ERROR("failed to add to map hash\n"); + ret = -ENOMEM; goto out_free_mm; } @@ -1309,7 +1310,7 @@ out_free_list: * i915_gem_release_mmap - remove physical page mappings * @obj: obj in question * - * Preserve the reservation of the mmaping with the DRM core code, but + * Preserve the reservation of the mmapping with the DRM core code, but * relinquish ownership of the pages back to the system. * * It is vital that we remove the page mapping if we have mapped a tiled @@ -1583,7 +1584,7 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj) * * Returned sequence numbers are nonzero on success. */ -static uint32_t +uint32_t i915_add_request(struct drm_device *dev, struct drm_file *file_priv, uint32_t flush_domains) { @@ -1617,7 +1618,7 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv, OUT_RING(MI_USER_INTERRUPT); ADVANCE_LP_RING(); - DRM_DEBUG("%d\n", seqno); + DRM_DEBUG_DRIVER("%d\n", seqno); request->seqno = seqno; request->emitted_jiffies = jiffies; @@ -1820,12 +1821,8 @@ i915_gem_retire_work_handler(struct work_struct *work) mutex_unlock(&dev->struct_mutex); } -/** - * Waits for a sequence number to be signaled, and cleans up the - * request and object lists appropriately for that event. - */ -static int -i915_wait_request(struct drm_device *dev, uint32_t seqno) +int +i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible) { drm_i915_private_t *dev_priv = dev->dev_private; u32 ier; @@ -1837,7 +1834,7 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno) return -EIO; if (!i915_seqno_passed(i915_get_gem_seqno(dev), seqno)) { - if (IS_IGDNG(dev)) + if (IS_IRONLAKE(dev)) ier = I915_READ(DEIER) | I915_READ(GTIER); else ier = I915_READ(IER); @@ -1852,10 +1849,15 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno) dev_priv->mm.waiting_gem_seqno = seqno; i915_user_irq_get(dev); - ret = wait_event_interruptible(dev_priv->irq_queue, - i915_seqno_passed(i915_get_gem_seqno(dev), - seqno) || - atomic_read(&dev_priv->mm.wedged)); + if (interruptible) + ret = wait_event_interruptible(dev_priv->irq_queue, + i915_seqno_passed(i915_get_gem_seqno(dev), seqno) || + atomic_read(&dev_priv->mm.wedged)); + else + wait_event(dev_priv->irq_queue, + i915_seqno_passed(i915_get_gem_seqno(dev), seqno) || + atomic_read(&dev_priv->mm.wedged)); + i915_user_irq_put(dev); dev_priv->mm.waiting_gem_seqno = 0; @@ -1879,6 +1881,16 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno) return ret; } +/** + * Waits for a sequence number to be signaled, and cleans up the + * request and object lists appropriately for that event. + */ +static int +i915_wait_request(struct drm_device *dev, uint32_t seqno) +{ + return i915_do_wait_request(dev, seqno, 1); +} + static void i915_gem_flush(struct drm_device *dev, uint32_t invalidate_domains, @@ -1947,7 +1959,7 @@ i915_gem_flush(struct drm_device *dev, #endif BEGIN_LP_RING(2); OUT_RING(cmd); - OUT_RING(0); /* noop */ + OUT_RING(MI_NOOP); ADVANCE_LP_RING(); } } @@ -2009,9 +2021,6 @@ i915_gem_object_unbind(struct drm_gem_object *obj) /* blow away mappings if mapped through GTT */ i915_gem_release_mmap(obj); - if (obj_priv->fence_reg != I915_FENCE_REG_NONE) - i915_gem_clear_fence_reg(obj); - /* Move the object to the CPU domain to ensure that * any possible CPU writes while it's not in the GTT * are flushed when we go to remap it. This will @@ -2027,6 +2036,10 @@ i915_gem_object_unbind(struct drm_gem_object *obj) BUG_ON(obj_priv->active); + /* release the fence reg _after_ flushing */ + if (obj_priv->fence_reg != I915_FENCE_REG_NONE) + i915_gem_clear_fence_reg(obj); + if (obj_priv->agp_mem != NULL) { drm_unbind_agp(obj_priv->agp_mem); drm_free_agp(obj_priv->agp_mem, obj->size / PAGE_SIZE); @@ -2569,9 +2582,6 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment) bool retry_alloc = false; int ret; - if (dev_priv->mm.suspended) - return -EBUSY; - if (obj_priv->madv != I915_MADV_WILLNEED) { DRM_ERROR("Attempting to bind a purgeable object\n"); return -EINVAL; @@ -2760,6 +2770,22 @@ i915_gem_object_flush_cpu_write_domain(struct drm_gem_object *obj) old_write_domain); } +void +i915_gem_object_flush_write_domain(struct drm_gem_object *obj) +{ + switch (obj->write_domain) { + case I915_GEM_DOMAIN_GTT: + i915_gem_object_flush_gtt_write_domain(obj); + break; + case I915_GEM_DOMAIN_CPU: + i915_gem_object_flush_cpu_write_domain(obj); + break; + default: + i915_gem_object_flush_gpu_write_domain(obj); + break; + } +} + /** * Moves a single object to the GTT read, and possibly write domain. * @@ -2811,6 +2837,57 @@ i915_gem_object_set_to_gtt_domain(struct drm_gem_object *obj, int write) return 0; } +/* + * Prepare buffer for display plane. Use uninterruptible for possible flush + * wait, as in modesetting process we're not supposed to be interrupted. + */ +int +i915_gem_object_set_to_display_plane(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + uint32_t old_write_domain, old_read_domains; + int ret; + + /* Not valid to be called on unbound objects. */ + if (obj_priv->gtt_space == NULL) + return -EINVAL; + + i915_gem_object_flush_gpu_write_domain(obj); + + /* Wait on any GPU rendering and flushing to occur. */ + if (obj_priv->active) { +#if WATCH_BUF + DRM_INFO("%s: object %p wait for seqno %08x\n", + __func__, obj, obj_priv->last_rendering_seqno); +#endif + ret = i915_do_wait_request(dev, obj_priv->last_rendering_seqno, 0); + if (ret != 0) + return ret; + } + + old_write_domain = obj->write_domain; + old_read_domains = obj->read_domains; + + obj->read_domains &= I915_GEM_DOMAIN_GTT; + + i915_gem_object_flush_cpu_write_domain(obj); + + /* It should now be out of any other write domains, and we can update + * the domain values for our changes. + */ + BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_GTT) != 0); + obj->read_domains |= I915_GEM_DOMAIN_GTT; + obj->write_domain = I915_GEM_DOMAIN_GTT; + obj_priv->dirty = 1; + + trace_i915_gem_object_change_domain(obj, + old_read_domains, + old_write_domain); + + return 0; +} + /** * Moves a single object to the CPU read, and possibly write domain. * @@ -3170,7 +3247,7 @@ 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; @@ -3178,12 +3255,35 @@ 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 @@ -3345,7 +3445,7 @@ 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) { @@ -3435,7 +3535,7 @@ 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) { @@ -3450,8 +3550,10 @@ i915_gem_get_relocs_from_user(struct drm_i915_gem_exec_object *exec_list, } *relocs = drm_calloc_large(reloc_count, sizeof(**relocs)); - if (*relocs == NULL) + if (*relocs == NULL) { + DRM_ERROR("failed to alloc relocs, count %d\n", reloc_count); return -ENOMEM; + } for (i = 0; i < buffer_count; i++) { struct drm_i915_gem_relocation_entry __user *user_relocs; @@ -3475,7 +3577,7 @@ 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) { @@ -3508,7 +3610,7 @@ 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; @@ -3525,22 +3627,57 @@ i915_gem_check_execbuffer (struct drm_i915_gem_execbuffer *exec, return 0; } +static int +i915_gem_wait_for_pending_flip(struct drm_device *dev, + struct drm_gem_object **object_list, + int count) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + DEFINE_WAIT(wait); + int i, ret = 0; + + for (;;) { + prepare_to_wait(&dev_priv->pending_flip_queue, + &wait, TASK_INTERRUPTIBLE); + for (i = 0; i < count; i++) { + obj_priv = object_list[i]->driver_private; + if (atomic_read(&obj_priv->pending_flip) > 0) + break; + } + if (i == count) + break; + + if (!signal_pending(current)) { + mutex_unlock(&dev->struct_mutex); + schedule(); + mutex_lock(&dev->struct_mutex); + continue; + } + ret = -ERESTARTSYS; + break; + } + finish_wait(&dev_priv->pending_flip_queue, &wait); + + return ret; +} + 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; + int pin_tries, flips; #if WATCH_EXEC DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n", @@ -3551,25 +3688,13 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, DRM_ERROR("execbuf with %d buffers\n", args->buffer_count); return -EINVAL; } - /* Copy in the exec list from userland */ - exec_list = drm_calloc_large(sizeof(*exec_list), args->buffer_count); - object_list = drm_calloc_large(sizeof(*object_list), args->buffer_count); - if (exec_list == NULL || object_list == NULL) { - DRM_ERROR("Failed to allocate exec or object list " - "for %d buffers\n", + object_list = drm_malloc_ab(sizeof(*object_list), args->buffer_count); + if (object_list == NULL) { + DRM_ERROR("Failed to allocate object list for %d buffers\n", args->buffer_count); 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), @@ -3598,20 +3723,19 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, i915_verify_inactive(dev, __FILE__, __LINE__); if (atomic_read(&dev_priv->mm.wedged)) { - DRM_ERROR("Execbuf while wedged\n"); mutex_unlock(&dev->struct_mutex); ret = -EIO; goto pre_mutex_err; } if (dev_priv->mm.suspended) { - DRM_ERROR("Execbuf while VT-switched.\n"); mutex_unlock(&dev->struct_mutex); ret = -EBUSY; goto pre_mutex_err; } /* Look up object handles */ + flips = 0; for (i = 0; i < args->buffer_count; i++) { object_list[i] = drm_gem_object_lookup(dev, file_priv, exec_list[i].handle); @@ -3630,6 +3754,14 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, goto err; } obj_priv->in_execbuffer = true; + flips += atomic_read(&obj_priv->pending_flip); + } + + if (flips > 0) { + ret = i915_gem_wait_for_pending_flip(dev, object_list, + args->buffer_count); + if (ret) + goto err; } /* Pin and relocate */ @@ -3814,20 +3946,6 @@ 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 @@ -3844,12 +3962,156 @@ 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) { @@ -3863,19 +4125,7 @@ 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, @@ -4356,7 +4606,7 @@ i915_gem_init_hws(struct drm_device *dev) memset(dev_priv->hw_status_page, 0, PAGE_SIZE); I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); I915_READ(HWS_PGA); /* posting read */ - DRM_DEBUG("hws offset: 0x%08x\n", dev_priv->status_gfx_addr); + DRM_DEBUG_DRIVER("hws offset: 0x%08x\n", dev_priv->status_gfx_addr); return 0; } @@ -4614,8 +4864,8 @@ i915_gem_load(struct drm_device *dev) for (i = 0; i < 8; i++) I915_WRITE(FENCE_REG_945_8 + (i * 4), 0); } - i915_gem_detect_bit_6_swizzle(dev); + init_waitqueue_head(&dev_priv->pending_flip_queue); } /* @@ -4638,7 +4888,7 @@ int i915_gem_init_phys_object(struct drm_device *dev, phys_obj->id = id; - phys_obj->handle = drm_pci_alloc(dev, size, 0, 0xffffffff); + phys_obj->handle = drm_pci_alloc(dev, size, 0); if (!phys_obj->handle) { ret = -ENOMEM; goto kfree_obj; @@ -4790,7 +5040,7 @@ i915_gem_phys_pwrite(struct drm_device *dev, struct drm_gem_object *obj, user_data = (char __user *) (uintptr_t) args->data_ptr; obj_addr = obj_priv->phys_obj->handle->vaddr + args->offset; - DRM_DEBUG("obj_addr %p, %lld\n", obj_addr, args->size); + DRM_DEBUG_DRIVER("obj_addr %p, %lld\n", obj_addr, args->size); ret = copy_from_user(obj_addr, user_data, args->size); if (ret) return -EFAULT; |