diff options
Diffstat (limited to 'drivers/gpu/drm/i915/i915_gem.c')
| -rw-r--r-- | drivers/gpu/drm/i915/i915_gem.c | 559 |
1 files changed, 388 insertions, 171 deletions
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 3326770c9ed261..f36126383d2601 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -31,6 +31,7 @@ #include "i915_drv.h" #include "i915_trace.h" #include "intel_drv.h" +#include <linux/oom.h> #include <linux/shmem_fs.h> #include <linux/slab.h> #include <linux/swap.h> @@ -43,6 +44,8 @@ static void i915_gem_object_flush_cpu_write_domain(struct drm_i915_gem_object *o static __must_check int i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, bool readonly); +static void +i915_gem_object_retire(struct drm_i915_gem_object *obj); static void i915_gem_write_fence(struct drm_device *dev, int reg, struct drm_i915_gem_object *obj); @@ -50,14 +53,15 @@ static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj, struct drm_i915_fence_reg *fence, bool enable); -static unsigned long i915_gem_inactive_count(struct shrinker *shrinker, +static unsigned long i915_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc); -static unsigned long i915_gem_inactive_scan(struct shrinker *shrinker, +static unsigned long i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc); +static int i915_gem_shrinker_oom(struct notifier_block *nb, + unsigned long event, + void *ptr); static unsigned long i915_gem_purge(struct drm_i915_private *dev_priv, long target); static unsigned long i915_gem_shrink_all(struct drm_i915_private *dev_priv); -static void i915_gem_object_truncate(struct drm_i915_gem_object *obj); -static void i915_gem_retire_requests_ring(struct intel_ring_buffer *ring); static bool cpu_cache_is_coherent(struct drm_device *dev, enum i915_cache_level level) @@ -470,6 +474,8 @@ int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj, ret = i915_gem_object_wait_rendering(obj, true); if (ret) return ret; + + i915_gem_object_retire(obj); } ret = i915_gem_object_get_pages(obj); @@ -885,6 +891,8 @@ i915_gem_shmem_pwrite(struct drm_device *dev, ret = i915_gem_object_wait_rendering(obj, false); if (ret) return ret; + + i915_gem_object_retire(obj); } /* Same trick applies to invalidate partially written cachelines read * before writing. */ @@ -1088,7 +1096,7 @@ i915_gem_check_wedge(struct i915_gpu_error *error, * equal. */ static int -i915_gem_check_olr(struct intel_ring_buffer *ring, u32 seqno) +i915_gem_check_olr(struct intel_engine_cs *ring, u32 seqno) { int ret; @@ -1107,7 +1115,7 @@ static void fake_irq(unsigned long data) } static bool missed_irq(struct drm_i915_private *dev_priv, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { return test_bit(ring->id, &dev_priv->gpu_error.missed_irq_rings); } @@ -1138,7 +1146,7 @@ static bool can_wait_boost(struct drm_i915_file_private *file_priv) * Returns 0 if the seqno was found within the alloted time. Else returns the * errno with remaining time filled in timeout argument. */ -static int __wait_seqno(struct intel_ring_buffer *ring, u32 seqno, +static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno, unsigned reset_counter, bool interruptible, struct timespec *timeout, @@ -1245,7 +1253,7 @@ static int __wait_seqno(struct intel_ring_buffer *ring, u32 seqno, * request and object lists appropriately for that event. */ int -i915_wait_seqno(struct intel_ring_buffer *ring, uint32_t seqno) +i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) { struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -1270,9 +1278,10 @@ i915_wait_seqno(struct intel_ring_buffer *ring, uint32_t seqno) static int i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { - i915_gem_retire_requests_ring(ring); + if (!obj->active) + return 0; /* Manually manage the write flush as we may have not yet * retired the buffer. @@ -1282,7 +1291,6 @@ i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj, * we know we have passed the last write. */ obj->last_write_seqno = 0; - obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS; return 0; } @@ -1295,7 +1303,7 @@ static __must_check int i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, bool readonly) { - struct intel_ring_buffer *ring = obj->ring; + struct intel_engine_cs *ring = obj->ring; u32 seqno; int ret; @@ -1320,7 +1328,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, { struct drm_device *dev = obj->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring = obj->ring; + struct intel_engine_cs *ring = obj->ring; unsigned reset_counter; u32 seqno; int ret; @@ -1536,7 +1544,7 @@ int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf) /* Access to snoopable pages through the GTT is incoherent. */ if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(dev)) { - ret = -EINVAL; + ret = -EFAULT; goto unlock; } @@ -1803,12 +1811,16 @@ i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data, return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset); } +static inline int +i915_gem_object_is_purgeable(struct drm_i915_gem_object *obj) +{ + return obj->madv == I915_MADV_DONTNEED; +} + /* Immediately discard the backing storage */ static void i915_gem_object_truncate(struct drm_i915_gem_object *obj) { - struct inode *inode; - i915_gem_object_free_mmap_offset(obj); if (obj->base.filp == NULL) @@ -1819,16 +1831,28 @@ i915_gem_object_truncate(struct drm_i915_gem_object *obj) * To do this we must instruct the shmfs to drop all of its * backing pages, *now*. */ - inode = file_inode(obj->base.filp); - shmem_truncate_range(inode, 0, (loff_t)-1); - + shmem_truncate_range(file_inode(obj->base.filp), 0, (loff_t)-1); obj->madv = __I915_MADV_PURGED; } -static inline int -i915_gem_object_is_purgeable(struct drm_i915_gem_object *obj) +/* Try to discard unwanted pages */ +static void +i915_gem_object_invalidate(struct drm_i915_gem_object *obj) { - return obj->madv == I915_MADV_DONTNEED; + struct address_space *mapping; + + switch (obj->madv) { + case I915_MADV_DONTNEED: + i915_gem_object_truncate(obj); + case __I915_MADV_PURGED: + return; + } + + if (obj->base.filp == NULL) + return; + + mapping = file_inode(obj->base.filp)->i_mapping, + invalidate_mapping_pages(mapping, 0, (loff_t)-1); } static void @@ -1893,8 +1917,7 @@ i915_gem_object_put_pages(struct drm_i915_gem_object *obj) ops->put_pages(obj); obj->pages = NULL; - if (i915_gem_object_is_purgeable(obj)) - i915_gem_object_truncate(obj); + i915_gem_object_invalidate(obj); return 0; } @@ -1903,58 +1926,58 @@ static unsigned long __i915_gem_shrink(struct drm_i915_private *dev_priv, long target, bool purgeable_only) { - struct list_head still_bound_list; - struct drm_i915_gem_object *obj, *next; + struct list_head still_in_list; + struct drm_i915_gem_object *obj; unsigned long count = 0; - list_for_each_entry_safe(obj, next, - &dev_priv->mm.unbound_list, - global_list) { - if ((i915_gem_object_is_purgeable(obj) || !purgeable_only) && - i915_gem_object_put_pages(obj) == 0) { - count += obj->base.size >> PAGE_SHIFT; - if (count >= target) - return count; - } - } - /* - * As we may completely rewrite the bound list whilst unbinding + * As we may completely rewrite the (un)bound list whilst unbinding * (due to retiring requests) we have to strictly process only * one element of the list at the time, and recheck the list * on every iteration. + * + * In particular, we must hold a reference whilst removing the + * object as we may end up waiting for and/or retiring the objects. + * This might release the final reference (held by the active list) + * and result in the object being freed from under us. This is + * similar to the precautions the eviction code must take whilst + * removing objects. + * + * Also note that although these lists do not hold a reference to + * the object we can safely grab one here: The final object + * unreferencing and the bound_list are both protected by the + * dev->struct_mutex and so we won't ever be able to observe an + * object on the bound_list with a reference count equals 0. */ - INIT_LIST_HEAD(&still_bound_list); + INIT_LIST_HEAD(&still_in_list); + while (count < target && !list_empty(&dev_priv->mm.unbound_list)) { + obj = list_first_entry(&dev_priv->mm.unbound_list, + typeof(*obj), global_list); + list_move_tail(&obj->global_list, &still_in_list); + + if (!i915_gem_object_is_purgeable(obj) && purgeable_only) + continue; + + drm_gem_object_reference(&obj->base); + + if (i915_gem_object_put_pages(obj) == 0) + count += obj->base.size >> PAGE_SHIFT; + + drm_gem_object_unreference(&obj->base); + } + list_splice(&still_in_list, &dev_priv->mm.unbound_list); + + INIT_LIST_HEAD(&still_in_list); while (count < target && !list_empty(&dev_priv->mm.bound_list)) { struct i915_vma *vma, *v; obj = list_first_entry(&dev_priv->mm.bound_list, typeof(*obj), global_list); - list_move_tail(&obj->global_list, &still_bound_list); + list_move_tail(&obj->global_list, &still_in_list); if (!i915_gem_object_is_purgeable(obj) && purgeable_only) continue; - /* - * Hold a reference whilst we unbind this object, as we may - * end up waiting for and retiring requests. This might - * release the final reference (held by the active list) - * and result in the object being freed from under us. - * in this object being freed. - * - * Note 1: Shrinking the bound list is special since only active - * (and hence bound objects) can contain such limbo objects, so - * we don't need special tricks for shrinking the unbound list. - * The only other place where we have to be careful with active - * objects suddenly disappearing due to retiring requests is the - * eviction code. - * - * Note 2: Even though the bound list doesn't hold a reference - * to the object we can safely grab one here: The final object - * unreferencing and the bound_list are both protected by the - * dev->struct_mutex and so we won't ever be able to observe an - * object on the bound_list with a reference count equals 0. - */ drm_gem_object_reference(&obj->base); list_for_each_entry_safe(vma, v, &obj->vma_list, vma_link) @@ -1966,7 +1989,7 @@ __i915_gem_shrink(struct drm_i915_private *dev_priv, long target, drm_gem_object_unreference(&obj->base); } - list_splice(&still_bound_list, &dev_priv->mm.bound_list); + list_splice(&still_in_list, &dev_priv->mm.bound_list); return count; } @@ -1980,17 +2003,8 @@ i915_gem_purge(struct drm_i915_private *dev_priv, long target) static unsigned long i915_gem_shrink_all(struct drm_i915_private *dev_priv) { - struct drm_i915_gem_object *obj, *next; - long freed = 0; - i915_gem_evict_everything(dev_priv->dev); - - list_for_each_entry_safe(obj, next, &dev_priv->mm.unbound_list, - global_list) { - if (i915_gem_object_put_pages(obj) == 0) - freed += obj->base.size >> PAGE_SHIFT; - } - return freed; + return __i915_gem_shrink(dev_priv, LONG_MAX, false); } static int @@ -2094,7 +2108,19 @@ err_pages: page_cache_release(sg_page_iter_page(&sg_iter)); sg_free_table(st); kfree(st); - return PTR_ERR(page); + + /* shmemfs first checks if there is enough memory to allocate the page + * and reports ENOSPC should there be insufficient, along with the usual + * ENOMEM for a genuine allocation failure. + * + * We use ENOSPC in our driver to mean that we have run out of aperture + * space and so want to translate the error from shmemfs back to our + * usual understanding of ENOMEM. + */ + if (PTR_ERR(page) == -ENOSPC) + return -ENOMEM; + else + return PTR_ERR(page); } /* Ensure that the associated pages are gathered from the backing storage @@ -2131,7 +2157,7 @@ i915_gem_object_get_pages(struct drm_i915_gem_object *obj) static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { struct drm_device *dev = obj->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -2169,7 +2195,7 @@ i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, } void i915_vma_move_to_active(struct i915_vma *vma, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { list_move_tail(&vma->mm_list, &vma->vm->active_list); return i915_gem_object_move_to_active(vma->obj, ring); @@ -2207,11 +2233,24 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj) WARN_ON(i915_verify_lists(dev)); } +static void +i915_gem_object_retire(struct drm_i915_gem_object *obj) +{ + struct intel_engine_cs *ring = obj->ring; + + if (ring == NULL) + return; + + if (i915_seqno_passed(ring->get_seqno(ring, true), + obj->last_read_seqno)) + i915_gem_object_move_to_inactive(obj); +} + static int i915_gem_init_seqno(struct drm_device *dev, u32 seqno) { struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring; + struct intel_engine_cs *ring; int ret, i, j; /* Carefully retire all requests without writing to the rings */ @@ -2226,8 +2265,8 @@ i915_gem_init_seqno(struct drm_device *dev, u32 seqno) for_each_ring(ring, dev_priv, i) { intel_ring_init_seqno(ring, seqno); - for (j = 0; j < ARRAY_SIZE(ring->sync_seqno); j++) - ring->sync_seqno[j] = 0; + for (j = 0; j < ARRAY_SIZE(ring->semaphore.sync_seqno); j++) + ring->semaphore.sync_seqno[j] = 0; } return 0; @@ -2277,7 +2316,7 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno) return 0; } -int __i915_add_request(struct intel_ring_buffer *ring, +int __i915_add_request(struct intel_engine_cs *ring, struct drm_file *file, struct drm_i915_gem_object *obj, u32 *out_seqno) @@ -2382,7 +2421,7 @@ i915_gem_request_remove_from_client(struct drm_i915_gem_request *request) } static bool i915_context_is_banned(struct drm_i915_private *dev_priv, - const struct i915_hw_context *ctx) + const struct intel_context *ctx) { unsigned long elapsed; @@ -2395,8 +2434,9 @@ static bool i915_context_is_banned(struct drm_i915_private *dev_priv, if (!i915_gem_context_is_default(ctx)) { DRM_DEBUG("context hanging too fast, banning!\n"); return true; - } else if (dev_priv->gpu_error.stop_rings == 0) { - DRM_ERROR("gpu hanging too fast, banning!\n"); + } else if (i915_stop_ring_allow_ban(dev_priv)) { + if (i915_stop_ring_allow_warn(dev_priv)) + DRM_ERROR("gpu hanging too fast, banning!\n"); return true; } } @@ -2405,7 +2445,7 @@ static bool i915_context_is_banned(struct drm_i915_private *dev_priv, } static void i915_set_reset_status(struct drm_i915_private *dev_priv, - struct i915_hw_context *ctx, + struct intel_context *ctx, const bool guilty) { struct i915_ctx_hang_stats *hs; @@ -2436,7 +2476,7 @@ static void i915_gem_free_request(struct drm_i915_gem_request *request) } struct drm_i915_gem_request * -i915_gem_find_active_request(struct intel_ring_buffer *ring) +i915_gem_find_active_request(struct intel_engine_cs *ring) { struct drm_i915_gem_request *request; u32 completed_seqno; @@ -2454,7 +2494,7 @@ i915_gem_find_active_request(struct intel_ring_buffer *ring) } static void i915_gem_reset_ring_status(struct drm_i915_private *dev_priv, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { struct drm_i915_gem_request *request; bool ring_hung; @@ -2473,7 +2513,7 @@ static void i915_gem_reset_ring_status(struct drm_i915_private *dev_priv, } static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, - struct intel_ring_buffer *ring) + struct intel_engine_cs *ring) { while (!list_empty(&ring->active_list)) { struct drm_i915_gem_object *obj; @@ -2501,6 +2541,11 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, i915_gem_free_request(request); } + + /* These may not have been flush before the reset, do so now */ + kfree(ring->preallocated_lazy_request); + ring->preallocated_lazy_request = NULL; + ring->outstanding_lazy_seqno = 0; } void i915_gem_restore_fences(struct drm_device *dev) @@ -2527,7 +2572,7 @@ void i915_gem_restore_fences(struct drm_device *dev) void i915_gem_reset(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring; + struct intel_engine_cs *ring; int i; /* @@ -2541,8 +2586,6 @@ void i915_gem_reset(struct drm_device *dev) for_each_ring(ring, dev_priv, i) i915_gem_reset_ring_cleanup(dev_priv, ring); - i915_gem_cleanup_ringbuffer(dev); - i915_gem_context_reset(dev); i915_gem_restore_fences(dev); @@ -2551,8 +2594,8 @@ void i915_gem_reset(struct drm_device *dev) /** * This function clears the request list as sequence numbers are passed. */ -static void -i915_gem_retire_requests_ring(struct intel_ring_buffer *ring) +void +i915_gem_retire_requests_ring(struct intel_engine_cs *ring) { uint32_t seqno; @@ -2597,7 +2640,7 @@ i915_gem_retire_requests_ring(struct intel_ring_buffer *ring) * of tail of the request to update the last known position * of the GPU head. */ - ring->last_retired_head = request->tail; + ring->buffer->last_retired_head = request->tail; i915_gem_free_request(request); } @@ -2615,7 +2658,7 @@ bool i915_gem_retire_requests(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring; + struct intel_engine_cs *ring; bool idle = true; int i; @@ -2709,7 +2752,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_gem_wait *args = data; struct drm_i915_gem_object *obj; - struct intel_ring_buffer *ring = NULL; + struct intel_engine_cs *ring = NULL; struct timespec timeout_stack, *timeout = NULL; unsigned reset_counter; u32 seqno = 0; @@ -2780,9 +2823,9 @@ out: */ int i915_gem_object_sync(struct drm_i915_gem_object *obj, - struct intel_ring_buffer *to) + struct intel_engine_cs *to) { - struct intel_ring_buffer *from = obj->ring; + struct intel_engine_cs *from = obj->ring; u32 seqno; int ret, idx; @@ -2795,7 +2838,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, idx = intel_ring_sync_index(from, to); seqno = obj->last_read_seqno; - if (seqno <= from->sync_seqno[idx]) + if (seqno <= from->semaphore.sync_seqno[idx]) return 0; ret = i915_gem_check_olr(obj->ring, seqno); @@ -2803,13 +2846,13 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, return ret; trace_i915_gem_ring_sync_to(from, to, seqno); - ret = to->sync_to(to, from, seqno); + ret = to->semaphore.sync_to(to, from, seqno); if (!ret) /* We use last_read_seqno because sync_to() * might have just caused seqno wrap under * the radar. */ - from->sync_seqno[idx] = obj->last_read_seqno; + from->semaphore.sync_seqno[idx] = obj->last_read_seqno; return ret; } @@ -2865,12 +2908,14 @@ int i915_vma_unbind(struct i915_vma *vma) * cause memory corruption through use-after-free. */ - i915_gem_object_finish_gtt(obj); + if (i915_is_ggtt(vma->vm)) { + i915_gem_object_finish_gtt(obj); - /* release the fence reg _after_ flushing */ - ret = i915_gem_object_put_fence(obj); - if (ret) - return ret; + /* release the fence reg _after_ flushing */ + ret = i915_gem_object_put_fence(obj); + if (ret) + return ret; + } trace_i915_vma_unbind(vma); @@ -2903,7 +2948,7 @@ int i915_vma_unbind(struct i915_vma *vma) int i915_gpu_idle(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring; + struct intel_engine_cs *ring; int ret, i; /* Flush everything onto the inactive list. */ @@ -3144,6 +3189,9 @@ i915_gem_object_put_fence(struct drm_i915_gem_object *obj) fence = &dev_priv->fence_regs[obj->fence_reg]; + if (WARN_ON(fence->pin_count)) + return -EBUSY; + i915_gem_object_fence_lost(obj); i915_gem_object_update_fence(obj, fence, false); @@ -3548,6 +3596,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write) if (ret) return ret; + i915_gem_object_retire(obj); i915_gem_object_flush_cpu_write_domain(obj, false); /* Serialise direct access to this object with the barriers for @@ -3646,6 +3695,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj, * in obj->write_domain and have been skipping the clflushes. * Just set it to the CPU cache for now. */ + i915_gem_object_retire(obj); WARN_ON(obj->base.write_domain & ~I915_GEM_DOMAIN_CPU); old_read_domains = obj->base.read_domains; @@ -3743,6 +3793,15 @@ unlock: static bool is_pin_display(struct drm_i915_gem_object *obj) { + struct i915_vma *vma; + + if (list_empty(&obj->vma_list)) + return false; + + vma = i915_gem_obj_to_ggtt(obj); + if (!vma) + return false; + /* There are 3 sources that pin objects: * 1. The display engine (scanouts, sprites, cursors); * 2. Reservations for execbuffer; @@ -3754,7 +3813,7 @@ static bool is_pin_display(struct drm_i915_gem_object *obj) * subtracting the potential reference by the user, any pin_count * remains, it must be due to another use by the display engine. */ - return i915_gem_obj_to_ggtt(obj)->pin_count - !!obj->user_pin_count; + return vma->pin_count - !!obj->user_pin_count; } /* @@ -3765,9 +3824,10 @@ static bool is_pin_display(struct drm_i915_gem_object *obj) int i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, u32 alignment, - struct intel_ring_buffer *pipelined) + struct intel_engine_cs *pipelined) { u32 old_read_domains, old_write_domain; + bool was_pin_display; int ret; if (pipelined != obj->ring) { @@ -3779,6 +3839,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, /* Mark the pin_display early so that we account for the * display coherency whilst setting up the cache domains. */ + was_pin_display = obj->pin_display; obj->pin_display = true; /* The display engine is not coherent with the LLC cache on gen6. As @@ -3821,7 +3882,8 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, return 0; err_unpin_display: - obj->pin_display = is_pin_display(obj); + WARN_ON(was_pin_display != is_pin_display(obj)); + obj->pin_display = was_pin_display; return ret; } @@ -3868,6 +3930,7 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write) if (ret) return ret; + i915_gem_object_retire(obj); i915_gem_object_flush_gtt_write_domain(obj); old_write_domain = obj->base.write_domain; @@ -3917,7 +3980,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file) struct drm_i915_file_private *file_priv = file->driver_priv; unsigned long recent_enough = jiffies - msecs_to_jiffies(20); struct drm_i915_gem_request *request; - struct intel_ring_buffer *ring = NULL; + struct intel_engine_cs *ring = NULL; unsigned reset_counter; u32 seqno = 0; int ret; @@ -3976,9 +4039,13 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj, uint32_t alignment, uint64_t flags) { + struct drm_i915_private *dev_priv = obj->base.dev->dev_private; struct i915_vma *vma; int ret; + if (WARN_ON(vm == &dev_priv->mm.aliasing_ppgtt->base)) + return -ENODEV; + if (WARN_ON(flags & (PIN_GLOBAL | PIN_MAPPABLE) && !i915_is_ggtt(vm))) return -EINVAL; @@ -4032,6 +4099,32 @@ i915_gem_object_ggtt_unpin(struct drm_i915_gem_object *obj) obj->pin_mappable = false; } +bool +i915_gem_object_pin_fence(struct drm_i915_gem_object *obj) +{ + if (obj->fence_reg != I915_FENCE_REG_NONE) { + struct drm_i915_private *dev_priv = obj->base.dev->dev_private; + struct i915_vma *ggtt_vma = i915_gem_obj_to_ggtt(obj); + + WARN_ON(!ggtt_vma || + dev_priv->fence_regs[obj->fence_reg].pin_count > + ggtt_vma->pin_count); + dev_priv->fence_regs[obj->fence_reg].pin_count++; + return true; + } else + return false; +} + +void +i915_gem_object_unpin_fence(struct drm_i915_gem_object *obj) +{ + if (obj->fence_reg != I915_FENCE_REG_NONE) { + struct drm_i915_private *dev_priv = obj->base.dev->dev_private; + WARN_ON(dev_priv->fence_regs[obj->fence_reg].pin_count <= 0); + dev_priv->fence_regs[obj->fence_reg].pin_count--; + } +} + int i915_gem_pin_ioctl(struct drm_device *dev, void *data, struct drm_file *file) @@ -4292,6 +4385,30 @@ struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev, return obj; } +static bool discard_backing_storage(struct drm_i915_gem_object *obj) +{ + /* If we are the last user of the backing storage (be it shmemfs + * pages or stolen etc), we know that the pages are going to be + * immediately released. In this case, we can then skip copying + * back the contents from the GPU. + */ + + if (obj->madv != I915_MADV_WILLNEED) + return false; + + if (obj->base.filp == NULL) + return true; + + /* At first glance, this looks racy, but then again so would be + * userspace racing mmap against close. However, the first external + * reference to the filp can only be obtained through the + * i915_gem_mmap_ioctl() which safeguards us against the user + * acquiring such a reference whilst we are in the middle of + * freeing the object. + */ + return atomic_long_read(&obj->base.filp->f_count) == 1; +} + void i915_gem_free_object(struct drm_gem_object *gem_obj) { struct drm_i915_gem_object *obj = to_intel_bo(gem_obj); @@ -4329,6 +4446,8 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj) if (WARN_ON(obj->pages_pin_count)) obj->pages_pin_count = 0; + if (discard_backing_storage(obj)) + obj->madv = I915_MADV_DONTNEED; i915_gem_object_put_pages(obj); i915_gem_object_free_mmap_offset(obj); i915_gem_object_release_stolen(obj); @@ -4338,6 +4457,9 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj) if (obj->base.import_attach) drm_prime_gem_destroy(&obj->base, NULL); + if (obj->ops->release) + obj->ops->release(obj); + drm_gem_object_release(&obj->base); i915_gem_info_remove_obj(dev_priv, obj->base.size); @@ -4371,6 +4493,17 @@ void i915_gem_vma_destroy(struct i915_vma *vma) kfree(vma); } +static void +i915_gem_stop_ringbuffers(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_engine_cs *ring; + int i; + + for_each_ring(ring, dev_priv, i) + intel_stop_ring_buffer(ring); +} + int i915_gem_suspend(struct drm_device *dev) { @@ -4392,7 +4525,7 @@ i915_gem_suspend(struct drm_device *dev) i915_gem_evict_everything(dev); i915_kernel_lost_context(dev); - i915_gem_cleanup_ringbuffer(dev); + i915_gem_stop_ringbuffers(dev); /* Hack! Don't let anybody do execbuf while we don't control the chip. * We need to replace this with a semaphore, or something. @@ -4413,7 +4546,7 @@ err: return ret; } -int i915_gem_l3_remap(struct intel_ring_buffer *ring, int slice) +int i915_gem_l3_remap(struct intel_engine_cs *ring, int slice) { struct drm_device *dev = ring->dev; struct drm_i915_private *dev_priv = dev->dev_private; @@ -4512,13 +4645,20 @@ static int i915_gem_init_rings(struct drm_device *dev) goto cleanup_blt_ring; } + if (HAS_BSD2(dev)) { + ret = intel_init_bsd2_ring_buffer(dev); + if (ret) + goto cleanup_vebox_ring; + } ret = i915_gem_set_seqno(dev, ((u32)~0 - 0x1000)); if (ret) - goto cleanup_vebox_ring; + goto cleanup_bsd2_ring; return 0; +cleanup_bsd2_ring: + intel_cleanup_ring_buffer(&dev_priv->ring[VCS2]); cleanup_vebox_ring: intel_cleanup_ring_buffer(&dev_priv->ring[VECS]); cleanup_blt_ring: @@ -4576,15 +4716,11 @@ i915_gem_init_hw(struct drm_device *dev) * the do_switch), but before enabling PPGTT. So don't move this. */ ret = i915_gem_context_enable(dev_priv); - if (ret) { + if (ret && ret != -EIO) { DRM_ERROR("Context enable failed %d\n", ret); - goto err_out; + i915_gem_cleanup_ringbuffer(dev); } - return 0; - -err_out: - i915_gem_cleanup_ringbuffer(dev); return ret; } @@ -4597,11 +4733,13 @@ int i915_gem_init(struct drm_device *dev) if (IS_VALLEYVIEW(dev)) { /* VLVA0 (potential hack), BIOS isn't actually waking us */ - I915_WRITE(VLV_GTLC_WAKE_CTRL, 1); - if (wait_for((I915_READ(VLV_GTLC_PW_STATUS) & 1) == 1, 10)) + I915_WRITE(VLV_GTLC_WAKE_CTRL, VLV_GTLC_ALLOWWAKEREQ); + if (wait_for((I915_READ(VLV_GTLC_PW_STATUS) & + VLV_GTLC_ALLOWWAKEACK), 10)) DRM_DEBUG_DRIVER("allow wake ack timed out\n"); } + i915_gem_init_userptr(dev); i915_gem_init_global_gtt(dev); ret = i915_gem_context_init(dev); @@ -4611,25 +4749,28 @@ int i915_gem_init(struct drm_device *dev) } ret = i915_gem_init_hw(dev); - mutex_unlock(&dev->struct_mutex); - if (ret) { - WARN_ON(dev_priv->mm.aliasing_ppgtt); - i915_gem_context_fini(dev); - drm_mm_takedown(&dev_priv->gtt.base.mm); - return ret; + if (ret == -EIO) { + /* Allow ring initialisation to fail by marking the GPU as + * wedged. But we only want to do this where the GPU is angry, + * for all other failure, such as an allocation failure, bail. + */ + DRM_ERROR("Failed to initialize GPU, declaring it wedged\n"); + atomic_set_mask(I915_WEDGED, &dev_priv->gpu_error.reset_counter); + ret = 0; } + mutex_unlock(&dev->struct_mutex); /* Allow hardware batchbuffers unless told otherwise, but not for KMS. */ if (!drm_core_check_feature(dev, DRIVER_MODESET)) dev_priv->dri1.allow_batchbuffer = 1; - return 0; + return ret; } void i915_gem_cleanup_ringbuffer(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - struct intel_ring_buffer *ring; + struct intel_engine_cs *ring; int i; for_each_ring(ring, dev_priv, i) @@ -4661,16 +4802,15 @@ i915_gem_entervt_ioctl(struct drm_device *dev, void *data, } BUG_ON(!list_empty(&dev_priv->gtt.base.active_list)); - mutex_unlock(&dev->struct_mutex); - ret = drm_irq_install(dev); + ret = drm_irq_install(dev, dev->pdev->irq); if (ret) goto cleanup_ringbuffer; + mutex_unlock(&dev->struct_mutex); return 0; cleanup_ringbuffer: - mutex_lock(&dev->struct_mutex); i915_gem_cleanup_ringbuffer(dev); dev_priv->ums.mm_suspended = 1; mutex_unlock(&dev->struct_mutex); @@ -4685,7 +4825,9 @@ i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, if (drm_core_check_feature(dev, DRIVER_MODESET)) return 0; + mutex_lock(&dev->struct_mutex); drm_irq_uninstall(dev); + mutex_unlock(&dev->struct_mutex); return i915_gem_suspend(dev); } @@ -4704,7 +4846,7 @@ i915_gem_lastclose(struct drm_device *dev) } static void -init_ring_lists(struct intel_ring_buffer *ring) +init_ring_lists(struct intel_engine_cs *ring) { INIT_LIST_HEAD(&ring->active_list); INIT_LIST_HEAD(&ring->request_list); @@ -4752,7 +4894,7 @@ i915_gem_load(struct drm_device *dev) init_waitqueue_head(&dev_priv->gpu_error.reset_queue); /* On GEN3 we really need to make sure the ARB C3 LP bit is set */ - if (IS_GEN3(dev)) { + if (!drm_core_check_feature(dev, DRIVER_MODESET) && IS_GEN3(dev)) { I915_WRITE(MI_ARB_STATE, _MASKED_BIT_ENABLE(MI_ARB_C3_LP_WRITE_ENABLE)); } @@ -4779,10 +4921,13 @@ i915_gem_load(struct drm_device *dev) dev_priv->mm.interruptible = true; - dev_priv->mm.inactive_shrinker.scan_objects = i915_gem_inactive_scan; - dev_priv->mm.inactive_shrinker.count_objects = i915_gem_inactive_count; - dev_priv->mm.inactive_shrinker.seeks = DEFAULT_SEEKS; - register_shrinker(&dev_priv->mm.inactive_shrinker); + dev_priv->mm.shrinker.scan_objects = i915_gem_shrinker_scan; + dev_priv->mm.shrinker.count_objects = i915_gem_shrinker_count; + dev_priv->mm.shrinker.seeks = DEFAULT_SEEKS; + register_shrinker(&dev_priv->mm.shrinker); + + dev_priv->mm.oom_notifier.notifier_call = i915_gem_shrinker_oom; + register_oom_notifier(&dev_priv->mm.oom_notifier); } void i915_gem_release(struct drm_device *dev, struct drm_file *file) @@ -4857,27 +5002,46 @@ static bool mutex_is_locked_by(struct mutex *mutex, struct task_struct *task) #endif } +static bool i915_gem_shrinker_lock(struct drm_device *dev, bool *unlock) +{ + if (!mutex_trylock(&dev->struct_mutex)) { + if (!mutex_is_locked_by(&dev->struct_mutex, current)) + return false; + + if (to_i915(dev)->mm.shrinker_no_lock_stealing) + return false; + + *unlock = false; + } else + *unlock = true; + + return true; +} + +static int num_vma_bound(struct drm_i915_gem_object *obj) +{ + struct i915_vma *vma; + int count = 0; + + list_for_each_entry(vma, &obj->vma_list, vma_link) + if (drm_mm_node_allocated(&vma->node)) + count++; + + return count; +} + static unsigned long -i915_gem_inactive_count(struct shrinker *shrinker, struct shrink_control *sc) +i915_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc) { struct drm_i915_private *dev_priv = - container_of(shrinker, - struct drm_i915_private, - mm.inactive_shrinker); + container_of(shrinker, struct drm_i915_private, mm.shrinker); struct drm_device *dev = dev_priv->dev; struct drm_i915_gem_object *obj; - bool unlock = true; unsigned long count; + bool unlock; - if (!mutex_trylock(&dev->struct_mutex)) { - if (!mutex_is_locked_by(&dev->struct_mutex, current)) - return 0; - - if (dev_priv->mm.shrinker_no_lock_stealing) - return 0; - - unlock = false; - } + if (!i915_gem_shrinker_lock(dev, &unlock)) + return 0; count = 0; list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) @@ -4885,10 +5049,8 @@ i915_gem_inactive_count(struct shrinker *shrinker, struct shrink_control *sc) count += obj->base.size >> PAGE_SHIFT; list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) { - if (obj->active) - continue; - - if (!i915_gem_obj_is_pinned(obj) && obj->pages_pin_count == 0) + if (!i915_gem_obj_is_pinned(obj) && + obj->pages_pin_count == num_vma_bound(obj)) count += obj->base.size >> PAGE_SHIFT; } @@ -4961,44 +5123,99 @@ unsigned long i915_gem_obj_size(struct drm_i915_gem_object *o, } static unsigned long -i915_gem_inactive_scan(struct shrinker *shrinker, struct shrink_control *sc) +i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) { struct drm_i915_private *dev_priv = - container_of(shrinker, - struct drm_i915_private, - mm.inactive_shrinker); + container_of(shrinker, struct drm_i915_private, mm.shrinker); struct drm_device *dev = dev_priv->dev; unsigned long freed; - bool unlock = true; + bool unlock; - if (!mutex_trylock(&dev->struct_mutex)) { - if (!mutex_is_locked_by(&dev->struct_mutex, current)) - return SHRINK_STOP; - - if (dev_priv->mm.shrinker_no_lock_stealing) - return SHRINK_STOP; - - unlock = false; - } + if (!i915_gem_shrinker_lock(dev, &unlock)) + return SHRINK_STOP; freed = i915_gem_purge(dev_priv, sc->nr_to_scan); if (freed < sc->nr_to_scan) freed += __i915_gem_shrink(dev_priv, sc->nr_to_scan - freed, false); - if (freed < sc->nr_to_scan) - freed += i915_gem_shrink_all(dev_priv); - if (unlock) mutex_unlock(&dev->struct_mutex); return freed; } +static int +i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) +{ + struct drm_i915_private *dev_priv = + container_of(nb, struct drm_i915_private, mm.oom_notifier); + struct drm_device *dev = dev_priv->dev; + struct drm_i915_gem_object *obj; + unsigned long timeout = msecs_to_jiffies(5000) + 1; + unsigned long pinned, bound, unbound, freed; + bool was_interruptible; + bool unlock; + + while (!i915_gem_shrinker_lock(dev, &unlock) && --timeout) + schedule_timeout_killable(1); + if (timeout == 0) { + pr_err("Unable to purge GPU memory due lock contention.\n"); + return NOTIFY_DONE; + } + + was_interruptible = dev_priv->mm.interruptible; + dev_priv->mm.interruptible = false; + + freed = i915_gem_shrink_all(dev_priv); + + dev_priv->mm.interruptible = was_interruptible; + + /* Because we may be allocating inside our own driver, we cannot + * assert that there are no objects with pinned pages that are not + * being pointed to by hardware. + */ + unbound = bound = pinned = 0; + list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) { + if (!obj->base.filp) /* not backed by a freeable object */ + continue; + + if (obj->pages_pin_count) + pinned += obj->base.size; + else + unbound += obj->base.size; + } + list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) { + if (!obj->base.filp) + continue; + + if (obj->pages_pin_count) + pinned += obj->base.size; + else + bound += obj->base.size; + } + + if (unlock) + mutex_unlock(&dev->struct_mutex); + + pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n", + freed, pinned); + if (unbound || bound) + pr_err("%lu and %lu bytes still available in the " + "bound and unbound GPU page lists.\n", + bound, unbound); + + *(unsigned long *)ptr += freed; + return NOTIFY_DONE; +} + struct i915_vma *i915_gem_obj_to_ggtt(struct drm_i915_gem_object *obj) { struct i915_vma *vma; + /* This WARN has probably outlived its usefulness (callers already + * WARN if they don't find the GGTT vma they expect). When removing, + * remember to remove the pre-check in is_pin_display() as well */ if (WARN_ON(list_empty(&obj->vma_list))) return NULL; |