Replace seqno values with request structures
Subsystems (2)
cover Replace seqno values with request structures
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
There is a general feeling that it is better to move away from using a simple integer 'seqno' value to track batch buffer completion. Instead, the request structure should be used. That provides for much more flexibility going forwards. Especially which things like a GPU scheduler (which can re-order batch buffers and hence seqnos after submission to the hardware), Android sync points and other such features which potentially make seqno usage more and more complex.
The current set of patches do most of the seqno to request structure conversion. There are still a couple of direct seqno comparisons in the semaphore code. The final conversion of a seqno test into a 'completed' flag inside the request structure is still do to as well. Along with whatever changes are required to maintain such a flag.
The patches are being posted now to make sure that the various people involved agree that it is heading in the right direction.
[Patches against drm-intel-nightly tree fetched 11/09/2014]
John Harrison (21):
Bug: missing i915_seqno_passed() call?
drm/i915: Remove redundant parameter to i915_gem_object_wait_rendering__tail() drm/i915: Ensure OLS & PLR are always in sync
drm/i915: Add reference count to request structure
drm/i915: Add helper functions to aid seqno -> request transition drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
drm/i915: Ensure requests stick around during waits
drm/i915: Remove 'outstanding_lazy_seqno'
drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno drm/i915: Convert 'last_flip_req' to be a request not a seqno drm/i915: Convert i915_wait_seqno to i915_wait_request
drm/i915: Convert 'i915_add_request' to take a request not a seqno drm/i915: Convert mmio_flip::seqno to struct request
drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' drm/i915: Convert __wait_seqno() to __wait_request()
drm/i915: Convert trace functions from seqno to request
drm/i915: Convert 'trace_irq' to use requests rather than seqnos drm/i915: Convert semaphores to handle requests not seqnos
drm/i915: Convert 'ring_idle()' to use requests not seqnos
drm/i915: Remove 'obj->ring'
drivers/gpu/drm/i915/i915_debugfs.c | 23 +-- drivers/gpu/drm/i915/i915_drv.h | 64 +++++-- drivers/gpu/drm/i915/i915_gem.c | 265 ++++++++++++++-------------- drivers/gpu/drm/i915/i915_gem_context.c | 12 +- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 +- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 19 +- drivers/gpu/drm/i915/i915_irq.c | 12 +- drivers/gpu/drm/i915/i915_trace.h | 47 ++--- drivers/gpu/drm/i915/intel_display.c | 47 ++--- drivers/gpu/drm/i915/intel_drv.h | 4 +- drivers/gpu/drm/i915/intel_lrc.c | 49 ++--- drivers/gpu/drm/i915/intel_overlay.c | 17 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 88 +++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 24 ++- 16 files changed, 392 insertions(+), 293 deletions(-)
Signature · 1 line
1.7.9.5
RFC 1/21 Bug: missing i915_seqno_passed() call?
From: John Harrison <[email protected]>
--- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +1 -1
@@ -2886,7 +2886,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, seqno = obj->last_read_seqno;/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ -if (seqno <= from->semaphore.sync_seqno[idx]) +if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; ret =i915_gem_check_olr (obj->ring, seqno);
Signature · 1 line
1.7.9.5
RFC 2/21 drm/i915: Remove redundant parameter to i915_gem_object_wait_rendering__tail()
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +3 -4
@@ -1281,8 +1281,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) }static int -i915_gem_object_wait_rendering__tail (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring) +i915_gem_object_wait_rendering__tail (struct drm_i915_gem_object *obj) {if (!obj->active)return 0 ;
@@ -1319,7 +1318,7 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,if (ret)return ret; -return i915_gem_object_wait_rendering__tail (obj, ring); +return i915_gem_object_wait_rendering__tail (obj); }/* A nonblocking variant of the above wait. This is a highly dangerous routine
@@ -1359,7 +1358,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,if (ret)return ret; -return i915_gem_object_wait_rendering__tail (obj, ring); +return i915_gem_object_wait_rendering__tail (obj); }/**
Signature · 1 line
1.7.9.5
RFC 3/21 drm/i915: Ensure OLS & PLR are always in sync
From: John Harrison <[email protected]>
The new seqno alloction code pre-allocates a 'lazy' request structure and then tries to allocate the 'lazy' seqno. The seqno allocation can potential wrap around zero and when doing so, tries to idle the ring by waiting for all oustanding work to complete. With a scheduler in place, this can mean first submitting extra work to the ring. However, at this point in time, the lazy request is valid but the lazy seqno is not. Some existing code was getting confused by this state and Bad Things would happen.
The safest solution is to still allocate the lazy request in advance (to avoid having to roll back in an out of memory sitation) but to save the pointer in a local variable rather than immediately updating the lazy pointer. Only after a valid seqno has been acquired is the lazy request pointer actually updated.
This guarantees that both lazy values are either invalid or both valid. There can no longer be an inconsistent state.
For: VIZ-4377
--- drivers/gpu/drm/i915/intel_lrc.c | 42 ++++++++++++++++++++----------- drivers/gpu/drm/i915/intel_ringbuffer.c | 29 +++++++++++++++------ 2 files changed, 48 insertions(+), 23 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +27 -15
@@ -796,27 +796,39 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf)static int logical_ring_alloc_seqno (struct intel_engine_cs *ring,struct intel_context *ctx) { -if (ring->outstanding_lazy_seqno) -return 0 ; +struct drm_i915_gem_request *request; +int ret; -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request; +/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL ); +return 0 ; + } +BUG_ON (ring->preallocated_lazy_request !=NULL ); -/* Hold a reference to the context this request belongs to -* (we will need it when the time comes to emit/retire the -* request). -*/ - request->ctx = ctx; -i915_gem_context_reference (request->ctx); + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +/* Hold a reference to the context this request belongs to +* (we will need it when the time comes to emit/retire the +* request). +*/ + request->ctx = ctx; +i915_gem_context_reference (request->ctx); + + ring->preallocated_lazy_request = request; +return 0 ; } static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf,
drivers/gpu/drm/i915/intel_ringbuffer.c +21 -8
@@ -2000,20 +2000,33 @@ int intel_ring_idle(struct intel_engine_cs *ring)static int intel_ring_alloc_seqno (struct intel_engine_cs *ring) { -if (ring->outstanding_lazy_seqno) +int ret; +struct drm_i915_gem_request *request; + +/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ + +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL );return 0 ; + } -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request; +BUG_ON (ring->preallocated_lazy_request !=NULL ); - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ring->preallocated_lazy_request = request; +return 0 ; } static int __intel_ring_prepare(struct intel_engine_cs *ring,
Signature · 1 line
1.7.9.5
RFC 4/21 drm/i915: Add reference count to request structure
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 16 ++++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 20 ++++++++++++++++---- drivers/gpu/drm/i915/intel_lrc.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ 4 files changed, 36 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +16 -0
@@ -1913,6 +1913,8 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,* an emission time with seqnos for tracking how far ahead of the GPU we are. */ struct drm_i915_gem_request { +struct kref ref; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;
@@ -1942,6 +1944,20 @@ struct drm_i915_gem_request {struct list_head client_list; }; +void i915_gem_request_free (struct kref *req_ref); + +static inline void +i915_gem_request_reference (struct drm_i915_gem_request *req) +{ +kref_get (&req->ref); +} + +static inline void +i915_gem_request_unreference (struct drm_i915_gem_request *req) +{ +kref_put (&req->ref, i915_gem_request_free); +} +struct drm_i915_file_private {struct drm_i915_private *dev_priv;struct drm_file *file;
drivers/gpu/drm/i915/i915_gem.c +16 -4
@@ -2379,6 +2379,8 @@ int __i915_add_request(struct intel_engine_cs *ring,return ret; } +i915_gem_request_reference (request); + request->seqno =intel_ring_get_seqno (ring); request->ring = ring; request->head = request_start;
@@ -2495,12 +2497,22 @@ static void i915_set_reset_status(struct drm_i915_private *dev_priv,static void i915_gem_free_request (struct drm_i915_gem_request *request) {list_del (&request->list); -i915_gem_request_remove_from_client (request); -if (request->ctx) -i915_gem_context_unreference (request->ctx); +i915_gem_request_unreference (request); +} + +void i915_gem_request_free (struct kref *req_ref) +{ +struct drm_i915_gem_request *req =container_of (req_ref, +typeof (*req), ref); + +/*list_del(&req->list);*/ +i915_gem_request_remove_from_client (req); + +if (req->ctx) +i915_gem_context_unreference (req->ctx); -kfree (request); +kfree (req); } struct drm_i915_gem_request *
drivers/gpu/drm/i915/intel_lrc.c +2 -0
@@ -814,6 +814,8 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,if (request ==NULL )return -ENOMEM ; +kref_init (&request->ref); + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno);if (ret) {kfree (request);
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -0
@@ -2019,6 +2019,8 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)if (request ==NULL )return -ENOMEM ; +kref_init (&request->ref); + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno);if (ret) {kfree (request);
Signature · 1 line
1.7.9.5
RFC 5/21 drm/i915: Add helper functions to aid seqno -> request transition
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 12 ++++++++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 7 +++++++ 2 files changed, 19 insertions(+)
drivers/gpu/drm/i915/i915_drv.h +12 -0
@@ -1946,6 +1946,18 @@ struct drm_i915_gem_request {void i915_gem_request_free (struct kref *req_ref); +static inline uint32_t +i915_gem_request_get_seqno (struct drm_i915_gem_request *req) +{ +return req ? req->seqno :0 ; +} + +static inline struct intel_engine_cs * +i915_gem_request_get_ring (struct drm_i915_gem_request *req) +{ +return req ? req->ring :NULL ; +} +static inline void i915_gem_request_reference (struct drm_i915_gem_request *req) {
drivers/gpu/drm/i915/intel_ringbuffer.h +7 -0
@@ -435,6 +435,13 @@ static inline u32 intel_ring_get_seqno(struct intel_engine_cs *ring)return ring->outstanding_lazy_seqno; } +static inline struct drm_i915_gem_request * +intel_ring_get_request (struct intel_engine_cs *ring) +{ +BUG_ON (ring->preallocated_lazy_request ==NULL ); +return ring->preallocated_lazy_request; +} +static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno) {if (ring->trace_irq_seqno ==0 && ring->irq_get (ring))
Signature · 1 line
1.7.9.5
RFC 6/21 drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 6 +-- drivers/gpu/drm/i915/i915_drv.h | 6 +-- drivers/gpu/drm/i915/i915_gem.c | 66 ++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_execbuffer.c | 6 +-- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 4 +- drivers/gpu/drm/i915/intel_display.c | 10 ++--- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 +- 9 files changed, 58 insertions(+), 48 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +3 -3
@@ -133,9 +133,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) obj->base.size /1024 , obj->base.read_domains, obj->base.write_domain, - obj->last_read_seqno, - obj->last_write_seqno, - obj->last_fenced_seqno, +i915_gem_request_get_seqno (obj->last_read_req), +i915_gem_request_get_seqno (obj->last_write_req), +i915_gem_request_get_seqno (obj->last_fenced_req),i915_cache_level_str (to_i915 (obj->base.dev), obj->cache_level), obj->dirty ?" dirty" :"" , obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");
drivers/gpu/drm/i915/i915_drv.h +3 -3
@@ -1862,10 +1862,10 @@ struct drm_i915_gem_object {struct intel_engine_cs *ring;/** Breadcrumb of last rendering to the buffer. */ -uint32_t last_read_seqno; -uint32_t last_write_seqno; +struct drm_i915_gem_request *last_read_req; +struct drm_i915_gem_request *last_write_req;/** Breadcrumb of last fenced GPU access to the buffer. */ -uint32_t last_fenced_seqno; +struct drm_i915_gem_request *last_fenced_req;/** Current tiling stride for the object, if it's tiled. */ uint32_t stride;
drivers/gpu/drm/i915/i915_gem.c +38 -28
173 lines not on this page · raw message
drivers/gpu/drm/i915/i915_gem_execbuffer.c +3 -3
@@ -946,7 +946,7 @@ void i915_gem_execbuffer_move_to_active(struct list_head *vmas, struct intel_engine_cs *ring) { - u32 seqno = intel_ring_get_seqno(ring); + struct drm_i915_gem_request *req = intel_ring_get_request(ring);struct i915_vma *vma;list_for_each_entry (vma, vmas, exec_list) {
@@ -963,7 +963,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,i915_vma_move_to_active (vma, ring);if (obj->base.write_domain) { obj->dirty =1 ; - obj->last_write_seqno = seqno; + obj->last_write_req = req;intel_fb_obj_invalidate (obj, ring);
@@ -971,7 +971,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas, obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS ; }if (entry->flags &EXEC_OBJECT_NEEDS_FENCE ) { - obj->last_fenced_seqno = seqno; + obj->last_fenced_req = req;if (entry->flags & __EXEC_OBJECT_HAS_FENCE) {struct drm_i915_private *dev_priv =to_i915 (ring->dev); list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,
drivers/gpu/drm/i915/i915_gem_gtt.h +2 -2
@@ -178,7 +178,7 @@ struct i915_address_space {* List of objects currently involved in rendering. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list.
@@ -189,7 +189,7 @@ struct i915_address_space {* LRU list of objects which are not in the ringbuffer and * are ready to unbind, but are still in the GTT. * -* last_rendering_seqno is 0 while an object is in this list. +* last_rendering_req is NULL while an object is in this list. * * A reference is not held on the buffer while on this list, * as merely being GTT-bound shouldn't prevent its being
drivers/gpu/drm/i915/i915_gem_tiling.c +1 -1
@@ -383,7 +383,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,if (ret ==0 ) { obj->fence_dirty = - obj->last_fenced_seqno || + obj->last_fenced_req || obj->fence_reg !=I915_FENCE_REG_NONE ; obj->tiling_mode = args->tiling_mode;
drivers/gpu/drm/i915/i915_gpu_error.c +2 -2
@@ -666,8 +666,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->size = obj->base.size; err->name = obj->base.name; - err->rseqno = obj->last_read_seqno; - err->wseqno = obj->last_write_seqno; + err->rseqno =i915_gem_request_get_seqno (obj->last_read_req); + err->wseqno =i915_gem_request_get_seqno (obj->last_write_req); err->gtt_offset = vma->node.start; err->read_domains = obj->base.read_domains; err->write_domain = obj->base.write_domain;
drivers/gpu/drm/i915/intel_display.c +5 -5
@@ -9766,16 +9766,16 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)lockdep_assert_held (&obj->base.dev->struct_mutex); -if (!obj->last_write_seqno) +if (!obj->last_write_req)return 0 ; ring = obj->ring;if (i915_seqno_passed (ring->get_seqno (ring, true), - obj->last_write_seqno)) +i915_gem_request_get_seqno (obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring, obj->last_write_seqno); + ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req));if (ret)return ret;
@@ -9838,7 +9838,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev, }spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); - intel_crtc->mmio_flip.seqno = obj->last_write_seqno; + intel_crtc->mmio_flip.seqno =i915_gem_request_get_seqno (obj->last_write_req); intel_crtc->mmio_flip.ring_id = obj->ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);
@@ -10045,7 +10045,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno = obj->last_write_seqno; + work->flip_queued_seqno =i915_gem_request_get_seqno (obj->last_write_req); work->flip_queued_ring = obj->ring; }else { ret = dev_priv->display.queue_flip(dev, crtc, fb, obj, ring,
drivers/gpu/drm/i915/intel_ringbuffer.h +1 -1
@@ -249,7 +249,7 @@ struct intel_engine_cs {* ringbuffer. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list.
Signature · 1 line
1.7.9.5
RFC 7/21 drm/i915: Ensure requests stick around during waits
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 25 ++++++++++++++++--------- 1 file changed, 16 insertions(+), 9 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +16 -9
@@ -1360,9 +1360,11 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,return ret; reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter); +i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); ret =__wait_seqno (ring, seqno, reset_counter, true,NULL , file_priv);mutex_lock (&dev->struct_mutex); +i915_gem_request_unreference (req);if (ret)return ret;
@@ -2825,6 +2827,7 @@ 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 drm_i915_gem_request *req;struct intel_engine_cs *ring =NULL ;unsigned reset_counter;u32 seqno =0 ;
@@ -2845,14 +2848,13 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)if (ret) goto out; -if (obj->active) { -if (!obj->last_read_req) - goto out; +if (!obj->active || !obj->last_read_req) + goto out; - seqno =i915_gem_request_get_seqno (obj->last_read_req); -BUG_ON (seqno ==0 ); - ring = obj->ring; - } + req = obj->last_read_req; + seqno =i915_gem_request_get_seqno (req); +BUG_ON (seqno ==0 ); + ring = obj->ring;/* Do this after OLR check to make sure we make forward progress polling * on this IOCTL with a timeout <=0 (like busy ioctl)
@@ -2864,10 +2866,15 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)drm_gem_object_unreference (&obj->base); reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter); +i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); -return __wait_seqno (ring, seqno, reset_counter, true, &args->timeout_ns, - file->driver_priv); + ret =__wait_seqno (ring, seqno, reset_counter, true, &args->timeout_ns, + file->driver_priv); +mutex_lock (&dev->struct_mutex); +i915_gem_request_unreference (req); +mutex_unlock (&dev->struct_mutex); +return ret; out:drm_gem_object_unreference (&obj->base);
Signature · 1 line
1.7.9.5
RFC 8/21 drm/i915: Remove 'outstanding_lazy_seqno'
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 13 +++----- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 20 ++++-------- drivers/gpu/drm/i915/intel_ringbuffer.c | 48 +++++++++++++++------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 ++------ 6 files changed, 41 insertions(+), 57 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +5 -8
@@ -1108,7 +1108,7 @@ i915_gem_check_olr(struct intel_engine_cs *ring, u32 seqno)BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); ret =0 ; -if (seqno == ring->outstanding_lazy_seqno) +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) ret =i915_add_request (ring,NULL );return ret;
@@ -2344,7 +2344,7 @@ int __i915_add_request(struct intel_engine_cs *ring,u32 request_ring_position, request_start;int ret; - request = ring->preallocated_lazy_request; + request = ring->outstanding_lazy_request;if (WARN_ON (request ==NULL ))return -ENOMEM ;
@@ -2391,7 +2391,6 @@ int __i915_add_request(struct intel_engine_cs *ring,i915_gem_request_reference (request); - request->seqno =intel_ring_get_seqno (ring); request->ring = ring; request->head = request_start; request->tail = request_ring_position;
@@ -2428,8 +2427,7 @@ int __i915_add_request(struct intel_engine_cs *ring, }trace_i915_gem_request_add (ring, request->seqno); - ring->outstanding_lazy_seqno =0 ; - ring->preallocated_lazy_request =NULL ; + ring->outstanding_lazy_request =NULL ;if (!dev_priv->ums.mm_suspended) {i915_queue_hangcheck (ring->dev);
@@ -2605,9 +2603,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, }/* 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 ; +kfree (ring->outstanding_lazy_request); + ring->outstanding_lazy_request =NULL ; }void i915_gem_restore_fences (struct drm_device *dev)
drivers/gpu/drm/i915/i915_gem_execbuffer.c +1 -1
@@ -1170,7 +1170,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,return ret; } -trace_i915_gem_ring_dispatch (ring,intel_ring_get_seqno (ring), flags); +trace_i915_gem_ring_dispatch (ring,i915_gem_request_get_seqno (intel_ring_get_request (ring)), flags);i915_gem_execbuffer_move_to_active (vmas, ring);i915_gem_execbuffer_retire_commands (dev, file, ring, batch_obj);
drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -10053,7 +10053,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =intel_ring_get_seqno (ring); + work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); work->flip_queued_ring = ring; }
drivers/gpu/drm/i915/intel_lrc.c +6 -14
@@ -799,16 +799,8 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,struct drm_i915_gem_request *request;int ret; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */ - -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )
@@ -816,7 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret;
@@ -829,7 +821,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, request->ctx = ctx;i915_gem_context_reference (request->ctx); - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }
@@ -1212,7 +1204,8 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf) (ring->status_page.gfx_addr + (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT)));intel_logical_ring_emit (ringbuf,0 ); -intel_logical_ring_emit (ringbuf, ring->outstanding_lazy_seqno); +intel_logical_ring_emit (ringbuf, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_logical_ring_emit (ringbuf,MI_USER_INTERRUPT );intel_logical_ring_emit (ringbuf,MI_NOOP );intel_logical_ring_advance_and_submit (ringbuf);
@@ -1235,8 +1228,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *ring)intel_logical_ring_stop (ring);WARN_ON ((I915_READ_MODE (ring) &MODE_IDLE ) ==0 ); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);
drivers/gpu/drm/i915/intel_ringbuffer.c +25 -23
147 lines not on this page · raw message
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -10
@@ -265,8 +265,7 @@ struct intel_engine_cs {/** * Do we have some not yet emitted requests outstanding? */ -struct drm_i915_gem_request *preallocated_lazy_request; -u32 outstanding_lazy_seqno; +struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty;bool fbc_dirty;
@@ -429,17 +428,11 @@ static inline u32 intel_ring_get_tail(struct intel_ringbuffer *ringbuf)return ringbuf->tail; } -static inline u32 intel_ring_get_seqno (struct intel_engine_cs *ring) -{ -BUG_ON (ring->outstanding_lazy_seqno ==0 ); -return ring->outstanding_lazy_seqno; -} -static inline struct drm_i915_gem_request *intel_ring_get_request (struct intel_engine_cs *ring) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); -return ring->preallocated_lazy_request; +BUG_ON (ring->outstanding_lazy_request ==NULL ); +return ring->outstanding_lazy_request; }static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno)
Signature · 1 line
1.7.9.5
RFC 9/21 drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 17 ++++++++++++++++- drivers/gpu/drm/i915/i915_gem.c | 25 ++++++++++--------------- drivers/gpu/drm/i915/intel_display.c | 2 +- 3 files changed, 27 insertions(+), 17 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +16 -1
@@ -2483,7 +2483,7 @@ bool i915_gem_retire_requests(struct drm_device *dev);void i915_gem_retire_requests_ring (struct intel_engine_cs *ring);int __must_check i915_gem_check_wedge (struct i915_gpu_error *error,bool interruptible); -int __must_check i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno); +int __must_check i915_gem_check_olr (struct drm_i915_gem_request *req);static inline bool i915_reset_in_progress (struct i915_gpu_error *error) {
@@ -3020,4 +3020,19 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +/************************* Deprecated *************************/ +static inline int __must_check i915_gem_check_ols (struct intel_engine_cs *ring,u32 seqno) +{ +int ret; + +BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); + + ret =0 ; +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) + ret =i915_add_request (ring,NULL ); + +return ret; +} +/************************* Deprecated *************************/ +#endif
drivers/gpu/drm/i915/i915_gem.c +10 -15
@@ -1097,19 +1097,18 @@ i915_gem_check_wedge(struct i915_gpu_error *error, }/* -* Compare seqno against outstanding lazy request. Emit a request if they are -* equal. +* Compare arbitrary request against outstanding lazy request. Emit on match. */ int -i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno) +i915_gem_check_olr (struct drm_i915_gem_request *req) {int ret; -BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); +BUG_ON (!mutex_is_locked (&req->ring->dev->struct_mutex)); ret =0 ; -if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) - ret =i915_add_request (ring,NULL ); +if (req == req->ring->outstanding_lazy_request) + ret =i915_add_request (req->ring,NULL );return ret; }
@@ -1271,7 +1270,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno)if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_ols (ring, seqno);if (ret)return ret;
@@ -1338,7 +1337,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,struct drm_i915_private *dev_priv = dev->dev_private;struct intel_engine_cs *ring = obj->ring;unsigned reset_counter; -u32 seqno;int ret;BUG_ON (!mutex_is_locked (&dev->struct_mutex));
@@ -1348,21 +1346,18 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,if (!req)return 0 ; - seqno =i915_gem_request_get_seqno (req); -BUG_ON (seqno ==0 ); - ret =i915_gem_check_wedge (&dev_priv->gpu_error, true);if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_olr (req);if (ret)return ret; reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter);i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); - ret =__wait_seqno (ring, seqno, reset_counter, true,NULL , file_priv); + ret =__wait_seqno (ring,i915_gem_request_get_seqno (req), reset_counter, true,NULL , file_priv);mutex_lock (&dev->struct_mutex);i915_gem_request_unreference (req);if (ret)
@@ -2786,7 +2781,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)int ret;if (obj->active) { - ret =i915_gem_check_olr (obj->ring,i915_gem_request_get_seqno (obj->last_read_req)); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;
@@ -2913,7 +2908,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; - ret =i915_gem_check_olr (obj->ring, seqno); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;
drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -9775,7 +9775,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) i915_gem_request_get_seqno(obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req)); + ret =i915_gem_check_olr (obj->last_write_req);if (ret)return ret;
Signature · 1 line
1.7.9.5
RFC 10/21 drm/i915: Convert 'last_flip_req' to be a request not a seqno
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/intel_overlay.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/intel_overlay.c +8 -8
@@ -182,7 +182,7 @@ struct intel_overlay {u32 flip_addr;struct drm_i915_gem_object *reg_bo;/* flip handling */ -uint32_t last_flip_req; +struct drm_i915_gem_request *last_flip_req;void (*flip_tail)(struct intel_overlay *); };
@@ -217,17 +217,17 @@ static int intel_overlay_do_wait_request(struct intel_overlay *overlay,int ret;BUG_ON (overlay->last_flip_req); - ret =i915_add_request (ring, &overlay->last_flip_req); + ret =i915_add_request (ring, &overlay->last_flip_req->seqno);if (ret)return ret; overlay->flip_tail = tail; - ret =i915_wait_seqno (ring, overlay->last_flip_req); + ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req));if (ret)return ret;i915_gem_retire_requests (dev); - overlay->last_flip_req =0 ; + overlay->last_flip_req =NULL ;return 0 ; }
@@ -286,7 +286,7 @@ static int intel_overlay_continue(struct intel_overlay *overlay,intel_ring_emit (ring, flip_addr);intel_ring_advance (ring); -return i915_add_request (ring, &overlay->last_flip_req); +return i915_add_request (ring, &overlay->last_flip_req->seqno); }static void intel_overlay_release_old_vid_tail (struct intel_overlay *overlay)
@@ -366,10 +366,10 @@ static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)struct intel_engine_cs *ring = &dev_priv->ring[RCS ];int ret; -if (overlay->last_flip_req ==0 ) +if (overlay->last_flip_req ==NULL )return 0 ; - ret =i915_wait_seqno (ring, overlay->last_flip_req); + ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req));if (ret)return ret;i915_gem_retire_requests (dev);
@@ -377,7 +377,7 @@ static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)if (overlay->flip_tail) overlay->flip_tail (overlay); - overlay->last_flip_req =0 ; + overlay->last_flip_req =NULL ;return 0 ; }
Signature · 1 line
1.7.9.5
RFC 11/21 drm/i915: Convert i915_wait_seqno to i915_wait_request
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 18 +----------------- drivers/gpu/drm/i915/i915_gem.c | 30 +++++++++++++++--------------- drivers/gpu/drm/i915/intel_lrc.c | 6 ++---- drivers/gpu/drm/i915/intel_overlay.c | 9 +++------ drivers/gpu/drm/i915/intel_ringbuffer.c | 14 ++++++-------- 5 files changed, 27 insertions(+), 50 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -17
@@ -2530,8 +2530,7 @@ int __i915_add_request(struct intel_engine_cs *ring, u32 *seqno);#define i915_add_request (ring, seqno) \ __i915_add_request(ring, NULL, NULL, seqno) -int __must_check i915_wait_seqno (struct intel_engine_cs *ring, -uint32_t seqno); +int __must_check i915_wait_request (struct drm_i915_gem_request *req);int i915_gem_fault (struct vm_area_struct *vma,struct vm_fault *vmf); int __must_check i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj,
@@ -3020,19 +3019,4 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } -/************************* Deprecated *************************/ -static inline int __must_check i915_gem_check_ols (struct intel_engine_cs *ring,u32 seqno) -{ -int ret; - -BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); - - ret =0 ; -if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) - ret =i915_add_request (ring,NULL ); - -return ret; -} -/************************* Deprecated *************************/ -#endif
drivers/gpu/drm/i915/i915_gem.c +15 -15
@@ -1252,29 +1252,34 @@ static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno, }/** -* Waits for a sequence number to be signaled, and cleans up the +* Waits for a request to be signaled, and cleans up the * request and object lists appropriately for that event. */ int -i915_wait_seqno (struct intel_engine_cs *ring,uint32_t seqno) +i915_wait_request (struct drm_i915_gem_request *req) { -struct drm_device *dev = ring->dev; -struct drm_i915_private *dev_priv = dev->dev_private; -bool interruptible = dev_priv->mm.interruptible; +struct drm_device *dev; +struct drm_i915_private *dev_priv; +bool interruptible;int ret; +BUG_ON (req ==NULL ); + + dev = req->ring->dev; + dev_priv = dev->dev_private; + interruptible = dev_priv->mm.interruptible; +BUG_ON (!mutex_is_locked (&dev->struct_mutex)); -BUG_ON (seqno ==0 ); ret =i915_gem_check_wedge (&dev_priv->gpu_error, interruptible);if (ret)return ret; - ret =i915_gem_check_ols (ring, seqno); + ret =i915_gem_check_olr (req);if (ret)return ret; -return __wait_seqno (ring, seqno, +return __wait_seqno (req->ring,i915_gem_request_get_seqno (req),atomic_read (&dev_priv->gpu_error.reset_counter), interruptible,NULL ,NULL ); }
@@ -1306,18 +1311,13 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, bool readonly) {struct drm_i915_gem_request *req; -struct intel_engine_cs *ring = obj->ring; -u32 seqno;int ret; req = readonly ? obj->last_write_req : obj->last_read_req;if (!req)return 0 ; - seqno =i915_gem_request_get_seqno (req); -BUG_ON (seqno ==0 ); - - ret =i915_wait_seqno (ring, seqno); + ret =i915_wait_request (req);if (ret)return ret;
@@ -3234,7 +3234,7 @@ static int i915_gem_object_wait_fence(struct drm_i915_gem_object *obj) {if (obj->last_fenced_req) { -int ret =i915_wait_seqno (obj->ring,i915_gem_request_get_seqno (obj->last_fenced_req)); +int ret =i915_wait_request (obj->last_fenced_req);if (ret)return ret;
drivers/gpu/drm/i915/intel_lrc.c +2 -4
@@ -830,7 +830,6 @@ static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf, {struct intel_engine_cs *ring = ringbuf->ring;struct drm_i915_gem_request *request; -u32 seqno =0 ;int ret;if (ringbuf->last_retired_head !=-1 ) {
@@ -845,15 +844,14 @@ static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf,list_for_each_entry (request, &ring->request_list, list) {if (__intel_ring_space (request->tail, ringbuf->tail, ringbuf->size) >= bytes) { - seqno = request->seqno;break ; } } -if (seqno ==0 ) +if (&request->list == &ring->request_list)return -ENOSPC ; - ret =i915_wait_seqno (ring, seqno); + ret =i915_wait_request (request);if (ret)return ret;
drivers/gpu/drm/i915/intel_overlay.c +3 -6
@@ -222,7 +222,7 @@ static int intel_overlay_do_wait_request(struct intel_overlay *overlay,return ret; overlay->flip_tail = tail; - ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req)); + ret =i915_wait_request (overlay->last_flip_req);if (ret)return ret;i915_gem_retire_requests (dev);
@@ -361,18 +361,15 @@ static int intel_overlay_off(struct intel_overlay *overlay)* We have to be careful not to repeat work forever an make forward progess. */ static int intel_overlay_recover_from_interrupt (struct intel_overlay *overlay) { -struct drm_device *dev = overlay->dev; -struct drm_i915_private *dev_priv = dev->dev_private; -struct intel_engine_cs *ring = &dev_priv->ring[RCS ];int ret;if (overlay->last_flip_req ==NULL )return 0 ; - ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req)); + ret =i915_wait_request (overlay->last_flip_req);if (ret)return ret; -i915_gem_retire_requests (dev); +i915_gem_retire_requests (overlay->dev);if (overlay->flip_tail) overlay->flip_tail (overlay);
drivers/gpu/drm/i915/intel_ringbuffer.c +6 -8
@@ -1868,7 +1868,6 @@ static int intel_ring_wait_request(struct intel_engine_cs *ring, int n) {struct intel_ringbuffer *ringbuf = ring->buffer;struct drm_i915_gem_request *request; -u32 seqno =0 ;int ret;if (ringbuf->last_retired_head !=-1 ) {
@@ -1883,15 +1882,14 @@ static int intel_ring_wait_request(struct intel_engine_cs *ring, int n)list_for_each_entry (request, &ring->request_list, list) {if (__intel_ring_space (request->tail, ringbuf->tail, ringbuf->size) >= n) { - seqno = request->seqno;break ; } } -if (seqno ==0 ) +if (&request->list == &ring->request_list)return -ENOSPC ; - ret =i915_wait_seqno (ring, seqno); + ret =i915_wait_request (request);if (ret)return ret;
@@ -1987,7 +1985,7 @@ static int intel_wrap_ring_buffer(struct intel_engine_cs *ring)int intel_ring_idle (struct intel_engine_cs *ring) { -u32 seqno; +struct drm_i915_gem_request *req;int ret;/* We need to add any requests required to flush the objects and ring */
@@ -2001,11 +1999,11 @@ int intel_ring_idle(struct intel_engine_cs *ring)if (list_empty (&ring->request_list))return 0 ; - seqno =list_entry (ring->request_list.prev, + req =list_entry (ring->request_list.prev, struct drm_i915_gem_request, - list)->seqno; + list); -return i915_wait_seqno (ring, seqno); +return i915_wait_request (req); } static int
Signature · 1 line
1.7.9.5
RFC 12/21 drm/i915: Convert 'i915_add_request' to take a request not a seqno
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 6 +++--- drivers/gpu/drm/i915/i915_gem.c | 6 +++--- drivers/gpu/drm/i915/intel_overlay.c | 4 ++-- 3 files changed, 8 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +3 -3
@@ -2527,9 +2527,9 @@ int __must_check i915_gem_suspend(struct drm_device *dev);int __i915_add_request (struct intel_engine_cs *ring,struct drm_file *file,struct drm_i915_gem_object *batch_obj, -u32 *seqno); -#define i915_add_request (ring, seqno) \ - __i915_add_request(ring, NULL, NULL, seqno) +struct drm_i915_gem_request **out_req); +#define i915_add_request (ring, out_req) \ + __i915_add_request(ring, NULL, NULL, out_req)int __must_check i915_wait_request (struct drm_i915_gem_request *req);int i915_gem_fault (struct vm_area_struct *vma,struct vm_fault *vmf);int __must_check
drivers/gpu/drm/i915/i915_gem.c +3 -3
@@ -2331,7 +2331,7 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno)int __i915_add_request (struct intel_engine_cs *ring,struct drm_file *file,struct drm_i915_gem_object *obj, -u32 *out_seqno) +struct drm_i915_gem_request **out_req) {struct drm_i915_private *dev_priv = ring->dev->dev_private;struct drm_i915_gem_request *request;
@@ -2434,8 +2434,8 @@ int __i915_add_request(struct intel_engine_cs *ring,intel_mark_busy (dev_priv->dev); } -if (out_seqno) - *out_seqno = request->seqno; +if (out_req) + *out_req = request;return 0 ; }
drivers/gpu/drm/i915/intel_overlay.c +2 -2
@@ -217,7 +217,7 @@ static int intel_overlay_do_wait_request(struct intel_overlay *overlay,int ret;BUG_ON (overlay->last_flip_req); - ret =i915_add_request (ring, &overlay->last_flip_req->seqno); + ret =i915_add_request (ring, &overlay->last_flip_req);if (ret)return ret;
@@ -286,7 +286,7 @@ static int intel_overlay_continue(struct intel_overlay *overlay,intel_ring_emit (ring, flip_addr);intel_ring_advance (ring); -return i915_add_request (ring, &overlay->last_flip_req->seqno); +return i915_add_request (ring, &overlay->last_flip_req); }static void intel_overlay_release_old_vid_tail (struct intel_overlay *overlay)
Signature · 1 line
1.7.9.5
RFC 13/21 drm/i915: Convert mmio_flip::seqno to struct request
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/intel_display.c | 12 +++++++----- drivers/gpu/drm/i915/intel_drv.h | 2 +- 2 files changed, 8 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/intel_display.c +7 -5
@@ -9799,15 +9799,16 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct intel_mmio_flip *mmio_flip; mmio_flip = &intel_crtc->mmio_flip; -if (mmio_flip->seqno ==0 ) +if (mmio_flip->req ==NULL )continue ;if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno, mmio_flip->seqno)) { +if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) {intel_do_mmio_flip (intel_crtc); - mmio_flip->seqno =0 ; +i915_gem_request_unreference (mmio_flip->req); + mmio_flip->req =NULL ; ring->irq_put (ring); } }
@@ -9826,7 +9827,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev,unsigned long irq_flags;int ret; -if (WARN_ON (intel_crtc->mmio_flip.seqno)) +if (WARN_ON (intel_crtc->mmio_flip.req))return -EBUSY ; ret =intel_postpone_flip (obj);
@@ -9838,7 +9839,8 @@ static int intel_queue_mmio_flip(struct drm_device *dev, }spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); - intel_crtc->mmio_flip.seqno =i915_gem_request_get_seqno (obj->last_write_req); + intel_crtc->mmio_flip.req = obj->last_write_req; +i915_gem_request_reference (intel_crtc->mmio_flip.req); intel_crtc->mmio_flip.ring_id = obj->ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);
drivers/gpu/drm/i915/intel_drv.h +1 -1
@@ -395,7 +395,7 @@ struct intel_pipe_wm { };struct intel_mmio_flip { -u32 seqno; +struct drm_i915_gem_request *req;u32 ring_id; };
Signature · 1 line
1.7.9.5
RFC 14/21 drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request'
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 4 ++-- drivers/gpu/drm/i915/intel_display.c | 9 ++++++--- drivers/gpu/drm/i915/intel_drv.h | 2 +- 3 files changed, 9 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +2 -2
@@ -547,11 +547,11 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)if (work->flip_queued_ring) {seq_printf (m,"Flip queued on %s at seqno %u, next seqno %u [current breadcrumb %u], completed? %d\n" , work->flip_queued_ring->name, - work->flip_queued_seqno, +i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true),i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - work->flip_queued_seqno)); +i915_gem_request_get_seqno (work->flip_queued_req))); }else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",
drivers/gpu/drm/i915/intel_display.c +6 -3
@@ -9879,7 +9879,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - work->flip_queued_seqno)) +i915_gem_request_get_seqno (work->flip_queued_req)))return false; work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);
@@ -10047,7 +10047,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =i915_gem_request_get_seqno (obj->last_write_req); + work->flip_queued_req = obj->last_write_req; work->flip_queued_ring = obj->ring; }else { ret = dev_priv->display.queue_flip(dev, crtc, fb, obj, ring,
@@ -10055,10 +10055,13 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); + work->flip_queued_req =intel_ring_get_request (ring); work->flip_queued_ring = ring; } +if (work->flip_queued_req) +i915_gem_request_reference (work->flip_queued_req); + work->flip_queued_vblank =drm_vblank_count (dev, intel_crtc->pipe); work->enable_stall_check = true;
drivers/gpu/drm/i915/intel_drv.h +1 -1
@@ -687,7 +687,7 @@ struct intel_unpin_work {u32 flip_count;u32 gtt_offset;struct intel_engine_cs *flip_queued_ring; -u32 flip_queued_seqno; +struct drm_i915_gem_request *flip_queued_req;int flip_queued_vblank;int flip_ready_vblank;bool enable_stall_check;
Signature · 1 line
1.7.9.5
RFC 15/21 drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed'
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 3 +-- drivers/gpu/drm/i915/i915_drv.h | 17 +++++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 12 ++++-------- drivers/gpu/drm/i915/intel_display.c | 11 +++-------- drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 1 + 6 files changed, 27 insertions(+), 18 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -2
@@ -550,8 +550,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - i915_gem_request_get_seqno(work->flip_queued_req))); + i915_gem_request_completed(work->flip_queued_req, true)); }else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",
drivers/gpu/drm/i915/i915_drv.h +17 -0
@@ -1970,6 +1970,8 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } +/* ??? i915_gem_request_completed should be here ??? */ +struct drm_i915_file_private {struct drm_i915_private *dev_priv;struct drm_file *file;
@@ -3019,4 +3021,19 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +u32 seqno; + +BUG_ON (req ==NULL ); + +if (req->ring ==NULL ) +return false; + + seqno = req->ring->get_seqno (req->ring, lazy_coherency); + +return i915_seqno_passed (seqno, req->seqno); +} +#endif
drivers/gpu/drm/i915/i915_gem.c +4 -8
@@ -2253,8 +2253,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)if (ring ==NULL )return ; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_read_req))) +if (i915_gem_request_completed (obj->last_read_req, true))i915_gem_object_move_to_inactive (obj); }
@@ -2522,12 +2521,9 @@ struct drm_i915_gem_request * i915_gem_find_active_request(struct intel_engine_cs *ring) {struct drm_i915_gem_request *request; -u32 completed_seqno; - - completed_seqno = ring->get_seqno (ring, false);list_for_each_entry (request, &ring->request_list, list) { -if (i915_seqno_passed (completed_seqno, request->seqno)) +if (i915_gem_request_completed (request, false))continue ;return request;
@@ -2671,7 +2667,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req))) +if (!i915_gem_request_completed (obj->last_read_req, true))break ;i915_gem_object_move_to_inactive (obj);
@@ -2686,7 +2682,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_request, list); -if (!i915_seqno_passed (seqno, request->seqno)) +if (!i915_gem_request_completed (request, true))break ;trace_i915_gem_request_retire (ring, request->seqno);
drivers/gpu/drm/i915/intel_display.c +3 -8
@@ -9771,8 +9771,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) ring = obj->ring; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_write_req))) +if (i915_gem_request_completed (obj->last_write_req, true))return 0 ; ret =i915_gem_check_olr (obj->last_write_req);
@@ -9790,9 +9789,6 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct drm_i915_private *dev_priv =to_i915 (ring->dev);struct intel_crtc *intel_crtc;unsigned long irq_flags; -u32 seqno; - - seqno = ring->get_seqno (ring, false);spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags);for_each_intel_crtc (ring->dev, intel_crtc) {
@@ -9805,7 +9801,7 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) { +if (i915_gem_request_completed (mmio_flip->req, false)) {intel_do_mmio_flip (intel_crtc);i915_gem_request_unreference (mmio_flip->req); mmio_flip->req =NULL ;
@@ -9878,8 +9874,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && - !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_gem_request_get_seqno (work->flip_queued_req))) + !i915_gem_request_completed (work->flip_queued_req, true))return false; work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);
drivers/gpu/drm/i915/intel_lrc.c +1 -0
@@ -807,6 +807,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,return -ENOMEM ;kref_init (&request->ref); + request->ring =NULL ; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
drivers/gpu/drm/i915/intel_ringbuffer.c +1 -0
@@ -2020,6 +2020,7 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)return -ENOMEM ;kref_init (&request->ref); + request->ring =NULL ; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
Signature · 1 line
1.7.9.5
RFC 16/21 drm/i915: Convert __wait_seqno() to __wait_request()
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 57 +++++++++++++++++---------------------- 1 file changed, 24 insertions(+), 33 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +24 -33
@@ -1133,10 +1133,9 @@ static bool can_wait_boost(struct drm_i915_file_private *file_priv) }/** -* __wait_seqno - wait until execution of seqno has finished -* @ring: the ring expected to report seqno -* @seqno: duh! -* @reset_counter: reset sequence associated with the given seqno +* __wait_request - wait until execution of request has finished +* @req: duh! +* @reset_counter: reset sequence associated with the given request * @interruptible: do an interruptible wait (normally yes) * @timeout: in - how long to wait (NULL forever); out - how much time remaining *
@@ -1147,15 +1146,16 @@ static bool can_wait_boost(struct drm_i915_file_private *file_priv)* reset_counter _must_ be read before, and an appropriate smp_rmb must be * inserted. * -* Returns 0 if the seqno was found within the alloted time. Else returns the +* Returns 0 if the request was found within the alloted time. Else returns the * errno with remaining time filled in timeout argument. */ -static int __wait_seqno (struct intel_engine_cs *ring,u32 seqno, -unsigned reset_counter, -bool interruptible, -s64 *timeout, -struct drm_i915_file_private *file_priv) +static int __wait_request (struct drm_i915_gem_request *req, +unsigned reset_counter, +bool interruptible, +s64 *timeout, +struct drm_i915_file_private *file_priv) { +struct intel_engine_cs *ring =i915_gem_request_get_ring (req);struct drm_device *dev = ring->dev;struct drm_i915_private *dev_priv = dev->dev_private; const bool irq_test_in_progress =
@@ -1167,7 +1167,7 @@ static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno,WARN (!intel_irqs_enabled (dev_priv),"IRQs disabled" ); -if (i915_seqno_passed (ring->get_seqno (ring, true), seqno)) +if (i915_gem_request_completed (req, true))return 0 ; timeout_expire = timeout ? jiffies +nsecs_to_jiffies ((u64)*timeout) :0 ;
@@ -1184,7 +1184,7 @@ static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno,return -ENODEV ;/* Record current time in case interrupted by signal, or wedged */ -trace_i915_gem_request_wait_begin (ring, seqno); +trace_i915_gem_request_wait_begin (i915_gem_request_get_ring (req),i915_gem_request_get_seqno (req)); before =ktime_get_raw_ns ();for (;;) {struct timer_list timer;
@@ -1203,7 +1203,7 @@ static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno,break ; } -if (i915_seqno_passed (ring->get_seqno (ring, false), seqno)) { +if (i915_gem_request_completed (req, false)) { ret =0 ;break ; }
@@ -1235,7 +1235,7 @@ static int __wait_seqno(struct intel_engine_cs *ring, u32 seqno, } } now =ktime_get_raw_ns (); -trace_i915_gem_request_wait_end (ring, seqno); +trace_i915_gem_request_wait_end (i915_gem_request_get_ring (req),i915_gem_request_get_seqno (req));if (!irq_test_in_progress) ring->irq_put (ring);
@@ -1279,9 +1279,9 @@ i915_wait_request(struct drm_i915_gem_request *req)if (ret)return ret; -return __wait_seqno (req->ring,i915_gem_request_get_seqno (req), -atomic_read (&dev_priv->gpu_error.reset_counter), - interruptible,NULL ,NULL ); +return __wait_request (req, +atomic_read (&dev_priv->gpu_error.reset_counter), + interruptible,NULL ,NULL ); } static int
@@ -1335,7 +1335,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,struct drm_i915_gem_request *req;struct drm_device *dev = obj->base.dev;struct drm_i915_private *dev_priv = dev->dev_private; -struct intel_engine_cs *ring = obj->ring;unsigned reset_counter;int ret;
@@ -1357,7 +1356,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter);i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); - ret =__wait_seqno (ring,i915_gem_request_get_seqno (req), reset_counter, true,NULL , file_priv); + ret =__wait_request (req, reset_counter, true,NULL , file_priv);mutex_lock (&dev->struct_mutex);i915_gem_request_unreference (req);if (ret)
@@ -2816,9 +2815,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)struct drm_i915_gem_wait *args = data;struct drm_i915_gem_object *obj;struct drm_i915_gem_request *req; -struct intel_engine_cs *ring =NULL ;unsigned reset_counter; -u32 seqno =0 ;int ret =0 ; ret =i915_mutex_lock_interruptible (dev);
@@ -2840,9 +2837,6 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) goto out; req = obj->last_read_req; - seqno =i915_gem_request_get_seqno (req); -BUG_ON (seqno ==0 ); - ring = obj->ring;/* Do this after OLR check to make sure we make forward progress polling * on this IOCTL with a timeout <=0 (like busy ioctl)
@@ -2857,8 +2851,8 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); - ret =__wait_seqno (ring, seqno, reset_counter, true, &args->timeout_ns, - file->driver_priv); + ret =__wait_request (req, reset_counter, true, &args->timeout_ns, + file->driver_priv);mutex_lock (&dev->struct_mutex);i915_gem_request_unreference (req);mutex_unlock (&dev->struct_mutex);
@@ -4053,10 +4047,8 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)struct drm_i915_private *dev_priv = dev->dev_private;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_engine_cs *ring =NULL ; +struct drm_i915_gem_request *request, *target =NULL ;unsigned reset_counter; -u32 seqno =0 ;int ret; ret =i915_gem_wait_for_error (&dev_priv->gpu_error);
@@ -4072,16 +4064,15 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)if (time_after_eq (request->emitted_jiffies, recent_enough))break ; - ring = request->ring; - seqno = request->seqno; + target = request; } reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter);spin_unlock (&file_priv->mm.lock); -if (seqno ==0 ) +if (target ==NULL )return 0 ; - ret =__wait_seqno (ring, seqno, reset_counter, true,NULL ,NULL ); + ret =__wait_request (target, reset_counter, true,NULL ,NULL );if (ret ==0 )queue_delayed_work (dev_priv->wq, &dev_priv->mm.retire_work,0 );
Signature · 1 line
1.7.9.5
RFC 17/21 drm/i915: Convert trace functions from seqno to request
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 10 +++--- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/i915_trace.h | 47 ++++++++++++++++------------ 3 files changed, 33 insertions(+), 26 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +5 -5
@@ -1184,7 +1184,7 @@ static int __wait_request(struct drm_i915_gem_request *req,return -ENODEV ;/* Record current time in case interrupted by signal, or wedged */ -trace_i915_gem_request_wait_begin (i915_gem_request_get_ring (req),i915_gem_request_get_seqno (req)); +trace_i915_gem_request_wait_begin (req); before =ktime_get_raw_ns ();for (;;) {struct timer_list timer;
@@ -1235,7 +1235,7 @@ static int __wait_request(struct drm_i915_gem_request *req, } } now =ktime_get_raw_ns (); -trace_i915_gem_request_wait_end (i915_gem_request_get_ring (req),i915_gem_request_get_seqno (req)); +trace_i915_gem_request_wait_end (req);if (!irq_test_in_progress) ring->irq_put (ring);
@@ -2419,7 +2419,7 @@ int __i915_add_request(struct intel_engine_cs *ring,spin_unlock (&file_priv->mm.lock); } -trace_i915_gem_request_add (ring, request->seqno); +trace_i915_gem_request_add (request); ring->outstanding_lazy_request =NULL ;if (!dev_priv->ums.mm_suspended) {
@@ -2684,7 +2684,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring)if (!i915_gem_request_completed (request, true))break ; -trace_i915_gem_request_retire (ring, request->seqno); +trace_i915_gem_request_retire (request);/* This is one of the few common intersection points * between legacy ringbuffer submission and execlists:
@@ -2902,7 +2902,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,if (ret)return ret; -trace_i915_gem_ring_sync_to (from, to, seqno); +trace_i915_gem_ring_sync_to (from, to, obj->last_read_req); ret = to->semaphore.sync_to (to, from, seqno);if (!ret)/* We use last_read_req because sync_to()
drivers/gpu/drm/i915/i915_gem_execbuffer.c +1 -1
@@ -1170,7 +1170,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,return ret; } -trace_i915_gem_ring_dispatch (ring,i915_gem_request_get_seqno (intel_ring_get_request (ring)), flags); +trace_i915_gem_ring_dispatch (intel_ring_get_request (ring), flags);i915_gem_execbuffer_move_to_active (vmas, ring);i915_gem_execbuffer_retire_commands (dev, file, ring, batch_obj);
drivers/gpu/drm/i915/i915_trace.h +27 -20
117 lines not on this page · raw message
Signature · 1 line
1.7.9.5
RFC 18/21 drm/i915: Convert 'trace_irq' to use requests rather than seqnos
From: John Harrison <[email protected]>
Note: i915_trace_irq_get is no longer inline as it requires accessing the internals of the request structure. As i915_drv.h includes intel_ringbuffer.h, an inline within the former is never going to be able to use details from the latter.
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 10 +++------- drivers/gpu/drm/i915/i915_trace.h | 2 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 8 ++++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 8 ++------ 4 files changed, 14 insertions(+), 14 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +3 -7
@@ -2646,15 +2646,11 @@ void i915_gem_reset(struct drm_device *dev)void i915_gem_retire_requests_ring (struct intel_engine_cs *ring) { -uint32_t seqno; -if (list_empty (&ring->request_list))return ;WARN_ON (i915_verify_lists (ring->dev)); - seqno = ring->get_seqno (ring, true); -/* Move any buffers on the active list that are no longer referenced * by the ringbuffer to the flushing/inactive lists as appropriate, * before we free the context associated with the requests.
@@ -2707,10 +2703,10 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring)i915_gem_free_request (request); } -if (unlikely (ring->trace_irq_seqno && -i915_seqno_passed (seqno, ring->trace_irq_seqno))) { +if (unlikely (ring->trace_irq_req && +i915_gem_request_completed (ring->trace_irq_req, true))) { ring->irq_put (ring); - ring->trace_irq_seqno =0 ; + ring->trace_irq_req =NULL ; }WARN_ON (i915_verify_lists (ring->dev));
drivers/gpu/drm/i915/i915_trace.h +1 -1
@@ -369,7 +369,7 @@ TRACE_EVENT(i915_gem_ring_dispatch, __entry->ring = ring->id; __entry->seqno =i915_gem_request_get_seqno (req); __entry->flags = flags; -i915_trace_irq_get (ring, __entry->seqno); +i915_trace_irq_get (req); ), TP_printk("dev=%u, ring=%u, seqno=%u, flags=%x",
drivers/gpu/drm/i915/intel_ringbuffer.c +8 -0
@@ -2791,3 +2791,11 @@ intel_stop_ring_buffer(struct intel_engine_cs *ring)stop_ring (ring); } + +void i915_trace_irq_get (struct drm_i915_gem_request *req) +{ +struct intel_engine_cs *ring =i915_gem_request_get_ring (req); + +if (ring->trace_irq_req ==NULL && ring->irq_get (ring)) + ring->trace_irq_req = req; +}
drivers/gpu/drm/i915/intel_ringbuffer.h +2 -6
@@ -142,7 +142,7 @@ struct intel_engine_cs {unsigned irq_refcount;/* protected by dev_priv->irq_lock */ u32 irq_enable_mask;/* bitmask to enable ring interrupt */ -u32 trace_irq_seqno; +struct drm_i915_gem_request *trace_irq_req;bool __must_check (*irq_get)(struct intel_engine_cs *ring);void (*irq_put)(struct intel_engine_cs *ring);
@@ -435,11 +435,7 @@ intel_ring_get_request(struct intel_engine_cs *ring)return ring->outstanding_lazy_request; } -static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno) -{ -if (ring->trace_irq_seqno ==0 && ring->irq_get (ring)) - ring->trace_irq_seqno = seqno; -} +void i915_trace_irq_get (struct drm_i915_gem_request *req);/* DRI warts */ int intel_render_ring_init_dri (struct drm_device *dev,u64 start,u32 size);
Signature · 1 line
1.7.9.5
RFC 19/21 drm/i915: Convert semaphores to handle requests not seqnos
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 3 ++- drivers/gpu/drm/i915/i915_gem.c | 14 ++++++-------- drivers/gpu/drm/i915/i915_gpu_error.c | 12 ++++++++---- drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 4 ++-- 5 files changed, 24 insertions(+), 19 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +2 -1
@@ -2604,7 +2604,8 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)seq_puts (m,"\nSync seqno:\n" );for_each_ring (ring, dev_priv, i) {for (j =0 ; j < num_rings; j++) { -seq_printf (m," 0x%08x " , ring->semaphore.sync_seqno[j]); +seq_printf (m," 0x%08x " , +i915_gem_request_get_seqno (ring->semaphore.sync_req[j])); }seq_putc (m,'\n' ); }
drivers/gpu/drm/i915/i915_gem.c +6 -8
@@ -2275,8 +2275,8 @@ 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->semaphore.sync_seqno); j++) - ring->semaphore.sync_seqno[j] =0 ; +for (j =0 ; j <ARRAY_SIZE (ring->semaphore.sync_req); j++) + ring->semaphore.sync_req[j] =NULL ; }return 0 ;
@@ -2877,7 +2877,6 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, struct intel_engine_cs *to) {struct intel_engine_cs *from = obj->ring; -u32 seqno;int ret, idx;if (from ==NULL || to == from)
@@ -2888,10 +2887,10 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, idx =intel_ring_sync_index (from, to); - seqno =i915_gem_request_get_seqno (obj->last_read_req);/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ -if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ +if (i915_gem_request_get_seqno (obj->last_read_req) <= +i915_gem_request_get_seqno (from->semaphore.sync_req[idx]))/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; ret =i915_gem_check_olr (obj->last_read_req);
@@ -2899,14 +2898,13 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,return ret;trace_i915_gem_ring_sync_to (from, to, obj->last_read_req); - ret = to->semaphore.sync_to (to, from, seqno); + ret = to->semaphore.sync_to (to, from, obj->last_read_req);if (!ret)/* We use last_read_req because sync_to() * might have just caused seqno wrap under * the radar. */ - from->semaphore.sync_seqno[idx] = -i915_gem_request_get_seqno (obj->last_read_req); + from->semaphore.sync_req[idx] = obj->last_read_req;return ret; }
drivers/gpu/drm/i915/i915_gpu_error.c +8 -4
@@ -822,7 +822,8 @@ static void gen8_record_semaphore_state(struct drm_i915_private *dev_priv, idx =intel_ring_sync_index (ring, to); ering->semaphore_mboxes[idx] = tmp[signal_offset]; - ering->semaphore_seqno[idx] = ring->semaphore.sync_seqno[idx]; + ering->semaphore_seqno[idx] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[idx]); } }
@@ -832,13 +833,16 @@ static void gen6_record_semaphore_state(struct drm_i915_private *dev_priv, { ering->semaphore_mboxes[0 ] =I915_READ (RING_SYNC_0 (ring->mmio_base)); ering->semaphore_mboxes[1 ] =I915_READ (RING_SYNC_1 (ring->mmio_base)); - ering->semaphore_seqno[0 ] = ring->semaphore.sync_seqno[0 ]; - ering->semaphore_seqno[1 ] = ring->semaphore.sync_seqno[1 ]; + ering->semaphore_seqno[0 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[0 ]); + ering->semaphore_seqno[1 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[1 ]);if (HAS_VEBOX (dev_priv->dev)) { ering->semaphore_mboxes[2 ] =I915_READ (RING_SYNC_2 (ring->mmio_base)); - ering->semaphore_seqno[2 ] = ring->semaphore.sync_seqno[2 ]; + ering->semaphore_seqno[2 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[2 ]); } }
drivers/gpu/drm/i915/intel_ringbuffer.c +6 -4
@@ -1042,7 +1042,7 @@ static inline bool i915_gem_has_seqno_wrapped(struct drm_device *dev,static int gen8_ring_sync (struct intel_engine_cs *waiter,struct intel_engine_cs *signaller, -u32 seqno) +struct drm_i915_gem_request *req) {struct drm_i915_private *dev_priv = waiter->dev->dev_private;int ret;
@@ -1055,7 +1055,7 @@ gen8_ring_sync(struct intel_engine_cs *waiter,MI_SEMAPHORE_GLOBAL_GTT |MI_SEMAPHORE_POLL | MI_SEMAPHORE_SAD_GTE_SDD); -intel_ring_emit (waiter, seqno); +intel_ring_emit (waiter,i915_gem_request_get_seqno (req));intel_ring_emit (waiter,lower_32_bits (GEN8_WAIT_OFFSET (waiter, signaller->id))); intel_ring_emit(waiter,
@@ -1067,18 +1067,20 @@ gen8_ring_sync(struct intel_engine_cs *waiter,static int gen6_ring_sync (struct intel_engine_cs *waiter,struct intel_engine_cs *signaller, -u32 seqno) +struct drm_i915_gem_request *req) {u32 dw1 =MI_SEMAPHORE_MBOX |MI_SEMAPHORE_COMPARE |MI_SEMAPHORE_REGISTER ;u32 wait_mbox = signaller->semaphore.mbox.wait[waiter->id];int ret; +u32 seqno;/* Throughout all of the GEM code, seqno passed implies our current * seqno is >= the last seqno executed. However for hardware the * comparison is strictly greater than. */ + seqno =i915_gem_request_get_seqno (req); seqno -=1 ;WARN_ON (wait_mbox ==MI_SEMAPHORE_SYNC_INVALID );
@@ -1794,7 +1796,7 @@ static int intel_init_ring_buffer(struct drm_device *dev,INIT_LIST_HEAD (&ring->execlist_queue); ringbuf->size =32 *PAGE_SIZE ; ringbuf->ring = ring; -memset (ring->semaphore.sync_seqno,0 ,sizeof (ring->semaphore.sync_seqno)); +memset (ring->semaphore.sync_req,0 ,sizeof (ring->semaphore.sync_req));init_waitqueue_head (&ring->irq_queue);
drivers/gpu/drm/i915/intel_ringbuffer.h +2 -2
@@ -211,7 +211,7 @@ struct intel_engine_cs {* ie. transpose of f(x, y) */ struct { - u32 sync_seqno[I915_NUM_RINGS-1]; + struct drm_i915_gem_request *sync_req[I915_NUM_RINGS-1]; union { struct {
@@ -226,7 +226,7 @@ struct intel_engine_cs {/* AKA wait() */ int (*sync_to)(struct intel_engine_cs *ring,struct intel_engine_cs *to, -u32 seqno); +struct drm_i915_gem_request *req);int (*signal)(struct intel_engine_cs *signaller,/* num_dwords needed by caller */ unsigned int num_dwords);
Signature · 1 line
1.7.9.5
RFC 20/21 drm/i915: Convert 'ring_idle()' to use requests not seqnos
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_irq.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_irq.c +6 -6
@@ -3030,18 +3030,18 @@ static void gen8_disable_vblank(struct drm_device *dev, int pipe)spin_unlock_irqrestore (&dev_priv->irq_lock, irqflags); } -static u32 -ring_last_seqno (struct intel_engine_cs *ring) +static struct drm_i915_gem_request * +ring_last_request (struct intel_engine_cs *ring) {return list_entry (ring->request_list.prev, - struct drm_i915_gem_request, list)->seqno; + struct drm_i915_gem_request, list); }static bool -ring_idle (struct intel_engine_cs *ring,u32 seqno) +ring_idle (struct intel_engine_cs *ring) {return (list_empty (&ring->request_list) || -i915_seqno_passed (seqno,ring_last_seqno (ring))); +i915_gem_request_completed (ring_last_request (ring), false)); } static bool
@@ -3261,7 +3261,7 @@ static void i915_hangcheck_elapsed(unsigned long data) acthd =intel_ring_get_active_head (ring);if (ring->hangcheck.seqno == seqno) { -if (ring_idle (ring, seqno)) { +if (ring_idle (ring)) { ring->hangcheck.action =HANGCHECK_IDLE ;if (waitqueue_active (&ring->irq_queue)) {
Signature · 1 line
1.7.9.5
RFC 21/21 drm/i915: Remove 'obj->ring'
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 9 +++++---- drivers/gpu/drm/i915/i915_drv.h | 2 -- drivers/gpu/drm/i915/i915_gem.c | 32 +++++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_context.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gpu_error.c | 3 ++- drivers/gpu/drm/i915/intel_display.c | 15 ++++++++------- 6 files changed, 46 insertions(+), 27 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +5 -4
@@ -168,8 +168,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) *t ='\0' ;seq_printf (m," (%s mappable)" , s); } -if (obj->ring !=NULL ) -seq_printf (m," (%s)" , obj->ring->name); +if (obj->last_read_req !=NULL ) +seq_printf (m," (%s)" , +i915_gem_request_get_ring (obj->last_read_req)->name);if (obj->frontbuffer_bits)seq_printf (m," (frontbuffer: 0x%03x)" , obj->frontbuffer_bits); }
@@ -336,7 +337,7 @@ static int per_file_stats(int id, void *ptr, void *data)if (ppgtt->file_priv != stats->file_priv)continue ; -if (obj->ring)/* XXX per-vma statistic */ +if (obj->last_read_req)/* XXX per-vma statistic */ stats->active += obj->base.size;else stats->inactive += obj->base.size;
@@ -346,7 +347,7 @@ static int per_file_stats(int id, void *ptr, void *data) }else {if (i915_gem_obj_ggtt_bound (obj)) { stats->global += obj->base.size; -if (obj->ring) +if (obj->last_read_req) stats->active += obj->base.size;else stats->inactive += obj->base.size;
drivers/gpu/drm/i915/i915_drv.h +0 -2
@@ -1859,8 +1859,6 @@ struct drm_i915_gem_object {void *dma_buf_vmapping;int vmapping_count; -struct intel_engine_cs *ring; -/** Breadcrumb of last rendering to the buffer. */ struct drm_i915_gem_request *last_read_req;struct drm_i915_gem_request *last_write_req;
drivers/gpu/drm/i915/i915_gem.c +20 -12
@@ -2184,14 +2184,18 @@ static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring) { - struct drm_i915_gem_request *req = intel_ring_get_request(ring); + struct drm_i915_gem_request *req; +struct intel_engine_cs *old_ring;BUG_ON (ring ==NULL ); -if (obj->ring != ring && obj->last_write_req) { + + req =intel_ring_get_request (ring); + old_ring =i915_gem_request_get_ring (obj->last_read_req); + +if (old_ring != ring && obj->last_write_req) {/* Keep the request relative to the current ring */ obj->last_write_req = req; } - obj->ring = ring;/* Add a reference if we're newly entering the active list. */ if (!obj->active) {
@@ -2230,7 +2234,6 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)intel_fb_obj_flush (obj, true);list_del_init (&obj->ring_list); - obj->ring =NULL ; obj->last_read_req =NULL ; obj->last_write_req =NULL ;
@@ -2247,9 +2250,7 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)static void i915_gem_object_retire (struct drm_i915_gem_object *obj) { -struct intel_engine_cs *ring = obj->ring; - -if (ring ==NULL ) +if (obj->last_read_req ==NULL )return ;if (i915_gem_request_completed (obj->last_read_req, true))
@@ -2769,14 +2770,17 @@ i915_gem_idle_work_handler(struct work_struct *work)static int i915_gem_object_flush_active (struct drm_i915_gem_object *obj) { +struct intel_engine_cs *ring;int ret;if (obj->active) { + ring =i915_gem_request_get_ring (obj->last_read_req); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret; -i915_gem_retire_requests_ring (obj->ring); +i915_gem_retire_requests_ring (ring); }return 0 ;
@@ -2876,9 +2880,11 @@ int i915_gem_object_sync(struct drm_i915_gem_object *obj, struct intel_engine_cs *to) { - struct intel_engine_cs *from = obj->ring; + struct intel_engine_cs *from;int ret, idx; + from =i915_gem_request_get_ring (obj->last_read_req); +if (from ==NULL || to == from)return 0 ;
@@ -3889,7 +3895,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,bool was_pin_display;int ret; -if (pipelined != obj->ring) { +if (pipelined !=i915_gem_request_get_ring (obj->last_read_req)) { ret =i915_gem_object_sync (obj, pipelined);if (ret)return ret;
@@ -4303,9 +4309,11 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data, ret =i915_gem_object_flush_active (obj); args->busy = obj->active; -if (obj->ring) { +if (obj->last_read_req) { +struct intel_engine_cs *ring;BUILD_BUG_ON (I915_NUM_RINGS >16 ); - args->busy |=intel_ring_flag (obj->ring) <<16 ; + ring =i915_gem_request_get_ring (obj->last_read_req); + args->busy |=intel_ring_flag (ring) <<16 ; }drm_gem_object_unreference (&obj->base);
drivers/gpu/drm/i915/i915_gem_context.c +11 -1
@@ -613,7 +613,17 @@ static int do_switch(struct intel_engine_cs *ring,* swapped, but there is no way to do that yet. */ from->legacy_hw_ctx.rcs_state->dirty =1 ; -BUG_ON (from->legacy_hw_ctx.rcs_state->ring != ring); + +/* BUG_ON(i915_gem_request_get_ring( +from->legacy_hw_ctx.rcs_state->last_read_req) != ring); */ +/* NB: last_read_req has already been updated to the current +* request however, that request has not yet been submitted. +* Thus last_read_req->ring is guaranteed to be null! +* NB2: Doing the check before the update of last_read_req +* (which happens in i915_vma_move_to_active() just above), +* also fails because last_read_req is almost always null on +* entry! +*/ /* obj is kept alive until the next request by its active ref */ i915_gem_object_ggtt_unpin (from->legacy_hw_ctx.rcs_state);
drivers/gpu/drm/i915/i915_gpu_error.c +2 -1
@@ -681,7 +681,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->dirty = obj->dirty; err->purgeable = obj->madv !=I915_MADV_WILLNEED ; err->userptr = obj->userptr.mm !=NULL ; - err->ring = obj->ring ? obj->ring->id :-1 ; + err->ring = obj->last_read_req ? +i915_gem_request_get_ring (obj->last_read_req)->id :-1 ; err->cache_level = obj->cache_level; }
drivers/gpu/drm/i915/intel_display.c +8 -7
@@ -9728,7 +9728,7 @@ static bool use_mmio_flip(struct intel_engine_cs *ring,else if (i915.enable_execlists)return true;else -return ring != obj->ring; +return ring !=i915_gem_request_get_ring (obj->last_read_req); }static void intel_do_mmio_flip (struct intel_crtc *intel_crtc)
@@ -9769,8 +9769,6 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (!obj->last_write_req)return 0 ; - ring = obj->ring; -if (i915_gem_request_completed (obj->last_write_req, true))return 0 ;
@@ -9778,6 +9776,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (ret)return ret; + ring =i915_gem_request_get_ring (obj->last_write_req);if (WARN_ON (!ring->irq_get (ring)))return 0 ;
@@ -9837,14 +9836,15 @@ static int intel_queue_mmio_flip(struct drm_device *dev,spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); intel_crtc->mmio_flip.req = obj->last_write_req;i915_gem_request_reference (intel_crtc->mmio_flip.req); - intel_crtc->mmio_flip.ring_id = obj->ring->id; +BUG_ON (ring !=i915_gem_request_get_ring (intel_crtc->mmio_flip.req)); + intel_crtc->mmio_flip.ring_id = ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);/* * Double check to catch cases where irq fired before * mmio flip data was ready */ -intel_notify_mmio_flip (obj->ring); +intel_notify_mmio_flip (ring);return 0 ; }
@@ -10022,7 +10022,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, }else if (IS_IVYBRIDGE (dev)) { ring = &dev_priv->ring[BCS ]; }else if (INTEL_INFO (dev)->gen >=7 ) { - ring = obj->ring; + ring =i915_gem_request_get_ring (obj->last_read_req);if (ring ==NULL || ring->id !=RCS ) ring = &dev_priv->ring[BCS ]; }else {
@@ -10043,7 +10043,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, goto cleanup_unpin; work->flip_queued_req = obj->last_write_req; - work->flip_queued_ring = obj->ring; + work->flip_queued_ring = +i915_gem_request_get_ring (obj->last_write_req); }else { ret = dev_priv->display.queue_flip (dev, crtc, fb, obj, ring, page_flip_flags);
Signature · 1 line
1.7.9.5
Re: [RFC 21/21] drm/i915: Remove 'obj->ring'
Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:25PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
I think this should be split up into the different parts:
- s/obj->ring/obj->last_read_req->ring/ for all the cases that just want the current ring.
- s/obj->ring/obj->last_read_req/ I think in a bunch of places the code would actually be more readable if we'd check for obj->active instead.
- All the oddball special cases probably deserve their own commit + nice explanation in the commit message about why the change is correct.
Cheers, Daniel
Quoted 267 lines
--- drivers/gpu/drm/i915/i915_debugfs.c | 9 +++++---- drivers/gpu/drm/i915/i915_drv.h | 2 -- drivers/gpu/drm/i915/i915_gem.c | 32 +++++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_context.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gpu_error.c | 3 ++- drivers/gpu/drm/i915/intel_display.c | 15 ++++++++------- 6 files changed, 46 insertions(+), 27 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +5
-4@@ -168,8 +168,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) *t ='\0' ;seq_printf (m," (%s mappable)" , s); } -if (obj->ring !=NULL ) -seq_printf (m," (%s)" , obj->ring->name); +if (obj->last_read_req !=NULL ) +seq_printf (m," (%s)" , +i915_gem_request_get_ring (obj->last_read_req)->name);if (obj->frontbuffer_bits)seq_printf (m," (frontbuffer: 0x%03x)" , obj->frontbuffer_bits); }@@ -336,7 +337,7 @@ static int per_file_stats(int id, void *ptr, void *data)if (ppgtt->file_priv != stats->file_priv)continue ; -if (obj->ring)/* XXX per-vma statistic */ +if (obj->last_read_req)/* XXX per-vma statistic */ stats->active += obj->base.size;else stats->inactive += obj->base.size;@@ -346,7 +347,7 @@ static int per_file_stats(int id, void *ptr, void *data) }else {if (i915_gem_obj_ggtt_bound (obj)) { stats->global += obj->base.size; -if (obj->ring) +if (obj->last_read_req) stats->active += obj->base.size;else stats->inactive += obj->base.size;drivers/gpu/drm/i915/i915_drv.h +0
-2@@ -1859,8 +1859,6 @@ struct drm_i915_gem_object {void *dma_buf_vmapping;int vmapping_count; -struct intel_engine_cs *ring; -/** Breadcrumb of last rendering to the buffer. */ struct drm_i915_gem_request *last_read_req;struct drm_i915_gem_request *last_write_req;drivers/gpu/drm/i915/i915_gem.c +20
-12@@ -2184,14 +2184,18 @@ static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring) { - struct drm_i915_gem_request *req = intel_ring_get_request(ring); + struct drm_i915_gem_request *req; +struct intel_engine_cs *old_ring;BUG_ON (ring ==NULL ); -if (obj->ring != ring && obj->last_write_req) { + + req =intel_ring_get_request (ring); + old_ring =i915_gem_request_get_ring (obj->last_read_req); + +if (old_ring != ring && obj->last_write_req) {/* Keep the request relative to the current ring */ obj->last_write_req = req; } - obj->ring = ring;/* Add a reference if we're newly entering the active list. */ if (!obj->active) {@@ -2230,7 +2234,6 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)intel_fb_obj_flush (obj, true);list_del_init (&obj->ring_list); - obj->ring =NULL ; obj->last_read_req =NULL ; obj->last_write_req =NULL ;@@ -2247,9 +2250,7 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)static void i915_gem_object_retire (struct drm_i915_gem_object *obj) { -struct intel_engine_cs *ring = obj->ring; - -if (ring ==NULL ) +if (obj->last_read_req ==NULL )return ;if (i915_gem_request_completed (obj->last_read_req, true))@@ -2769,14 +2770,17 @@ i915_gem_idle_work_handler(struct work_struct *work)static int i915_gem_object_flush_active (struct drm_i915_gem_object *obj) { +struct intel_engine_cs *ring;int ret;if (obj->active) { + ring =i915_gem_request_get_ring (obj->last_read_req); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret; -i915_gem_retire_requests_ring (obj->ring); +i915_gem_retire_requests_ring (ring); }return 0 ;@@ -2876,9 +2880,11 @@ int i915_gem_object_sync(struct drm_i915_gem_object *obj, struct intel_engine_cs *to) { - struct intel_engine_cs *from = obj->ring; + struct intel_engine_cs *from;int ret, idx; + from =i915_gem_request_get_ring (obj->last_read_req); +if (from ==NULL || to == from)return 0 ;@@ -3889,7 +3895,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,bool was_pin_display;int ret; -if (pipelined != obj->ring) { +if (pipelined !=i915_gem_request_get_ring (obj->last_read_req)) { ret =i915_gem_object_sync (obj, pipelined);if (ret)return ret;@@ -4303,9 +4309,11 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data, ret =i915_gem_object_flush_active (obj); args->busy = obj->active; -if (obj->ring) { +if (obj->last_read_req) { +struct intel_engine_cs *ring;BUILD_BUG_ON (I915_NUM_RINGS >16 ); - args->busy |=intel_ring_flag (obj->ring) <<16 ; + ring =i915_gem_request_get_ring (obj->last_read_req); + args->busy |=intel_ring_flag (ring) <<16 ; }drm_gem_object_unreference (&obj->base);drivers/gpu/drm/i915/i915_gem_context.c +11
-1@@ -613,7 +613,17 @@ static int do_switch(struct intel_engine_cs *ring,* swapped, but there is no way to do that yet. */ from->legacy_hw_ctx.rcs_state->dirty =1 ; -BUG_ON (from->legacy_hw_ctx.rcs_state->ring != ring); + +/* BUG_ON(i915_gem_request_get_ring( +from->legacy_hw_ctx.rcs_state->last_read_req) != ring); */ +/* NB: last_read_req has already been updated to the current +* request however, that request has not yet been submitted. +* Thus last_read_req->ring is guaranteed to be null! +* NB2: Doing the check before the update of last_read_req +* (which happens in i915_vma_move_to_active() just above), +* also fails because last_read_req is almost always null on +* entry! +*/ /* obj is kept alive until the next request by its active ref */ i915_gem_object_ggtt_unpin (from->legacy_hw_ctx.rcs_state);drivers/gpu/drm/i915/i915_gpu_error.c +2
-1@@ -681,7 +681,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->dirty = obj->dirty; err->purgeable = obj->madv !=I915_MADV_WILLNEED ; err->userptr = obj->userptr.mm !=NULL ; - err->ring = obj->ring ? obj->ring->id :-1 ; + err->ring = obj->last_read_req ? +i915_gem_request_get_ring (obj->last_read_req)->id :-1 ; err->cache_level = obj->cache_level; }drivers/gpu/drm/i915/intel_display.c +8
-7@@ -9728,7 +9728,7 @@ static bool use_mmio_flip(struct intel_engine_cs *ring,else if (i915.enable_execlists)return true;else -return ring != obj->ring; +return ring !=i915_gem_request_get_ring (obj->last_read_req); }static void intel_do_mmio_flip (struct intel_crtc *intel_crtc)@@ -9769,8 +9769,6 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (!obj->last_write_req)return 0 ; - ring = obj->ring; -if (i915_gem_request_completed (obj->last_write_req, true))return 0 ;@@ -9778,6 +9776,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (ret)return ret; + ring =i915_gem_request_get_ring (obj->last_write_req);if (WARN_ON (!ring->irq_get (ring)))return 0 ;@@ -9837,14 +9836,15 @@ static int intel_queue_mmio_flip(struct drm_device *dev,spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); intel_crtc->mmio_flip.req = obj->last_write_req;i915_gem_request_reference (intel_crtc->mmio_flip.req); - intel_crtc->mmio_flip.ring_id = obj->ring->id; +BUG_ON (ring !=i915_gem_request_get_ring (intel_crtc->mmio_flip.req)); + intel_crtc->mmio_flip.ring_id = ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);/* * Double check to catch cases where irq fired before * mmio flip data was ready */ -intel_notify_mmio_flip (obj->ring); +intel_notify_mmio_flip (ring);return 0 ; }@@ -10022,7 +10022,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, }else if (IS_IVYBRIDGE (dev)) { ring = &dev_priv->ring[BCS ]; }else if (INTEL_INFO (dev)->gen >=7 ) { - ring = obj->ring; + ring =i915_gem_request_get_ring (obj->last_read_req);if (ring ==NULL || ring->id !=RCS ) ring = &dev_priv->ring[BCS ]; }else {@@ -10043,7 +10043,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, goto cleanup_unpin; work->flip_queued_req = obj->last_write_req; - work->flip_queued_ring = obj->ring; + work->flip_queued_ring = +i915_gem_request_get_ring (obj->last_write_req); }else { ret = dev_priv->display.queue_flip (dev, crtc, fb, obj, ring, page_flip_flags);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 21/21] drm/i915: Remove 'obj->ring'
Quoted 5 lines · Daniel Vetter
Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:25PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
Quoted 8 lines
I think this should be split up into the different parts:
- s/obj->ring/obj->last_read_req->ring/ for all the cases that just want the current ring.
- s/obj->ring/obj->last_read_req/ I think in a bunch of places the code would actually be more readable if we'd check forobj->activeinstead.
- All the oddball special cases probably deserve their own commit + nice explanation in the commit message about why the change is correct.
Can you explain which is which? As for why the change is correct, it is not a functional change. obj->ring was assigned with obj->last_read_seqno and cleared with obj->last_read_seqno. Thus querying obj->ring is equivalent to querying obj->last_read_req->ring in all cases. Given that the ring is now available from obj->lrr, it seemed redundant to also have it explicitly saved in obj->ring, hence the patch to remove it.
AFAICT, there are two examples in debugfs that should just be querying obj->active. The rest are pretty much wanting the currently in use ring and/or are about to use last_read_req anyway. It seems more sensible to say 'if(obj->lrr) { do_stuff(obj->lrr); }' than 'if(obj->active) { do_stuff(obj->lrr); }'. Nothing looks particularly 'oddball' to me!
Quoted 264 lines
Cheers, Daniel
Quoted 263 lines
--- drivers/gpu/drm/i915/i915_debugfs.c | 9 +++++---- drivers/gpu/drm/i915/i915_drv.h | 2 -- drivers/gpu/drm/i915/i915_gem.c | 32 +++++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_context.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gpu_error.c | 3 ++- drivers/gpu/drm/i915/intel_display.c | 15 ++++++++------- 6 files changed, 46 insertions(+), 27 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +5
-4@@ -168,8 +168,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) *t ='\0' ;seq_printf (m," (%s mappable)" , s); } -if (obj->ring !=NULL ) -seq_printf (m," (%s)" , obj->ring->name); +if (obj->last_read_req !=NULL ) +seq_printf (m," (%s)" , +i915_gem_request_get_ring (obj->last_read_req)->name);if (obj->frontbuffer_bits)seq_printf (m," (frontbuffer: 0x%03x)" , obj->frontbuffer_bits); }@@ -336,7 +337,7 @@ static int per_file_stats(int id, void *ptr, void *data)if (ppgtt->file_priv != stats->file_priv)continue ; -if (obj->ring)/* XXX per-vma statistic */ +if (obj->last_read_req)/* XXX per-vma statistic */ stats->active += obj->base.size;else stats->inactive += obj->base.size;@@ -346,7 +347,7 @@ static int per_file_stats(int id, void *ptr, void *data) }else {if (i915_gem_obj_ggtt_bound (obj)) { stats->global += obj->base.size; -if (obj->ring) +if (obj->last_read_req) stats->active += obj->base.size;else stats->inactive += obj->base.size;drivers/gpu/drm/i915/i915_drv.h +0
-2@@ -1859,8 +1859,6 @@ struct drm_i915_gem_object {void *dma_buf_vmapping;int vmapping_count; -struct intel_engine_cs *ring; -/** Breadcrumb of last rendering to the buffer. */ struct drm_i915_gem_request *last_read_req;struct drm_i915_gem_request *last_write_req;drivers/gpu/drm/i915/i915_gem.c +20
-12@@ -2184,14 +2184,18 @@ static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring) { - struct drm_i915_gem_request *req = intel_ring_get_request(ring); + struct drm_i915_gem_request *req; +struct intel_engine_cs *old_ring;BUG_ON (ring ==NULL ); -if (obj->ring != ring && obj->last_write_req) { + + req =intel_ring_get_request (ring); + old_ring =i915_gem_request_get_ring (obj->last_read_req); + +if (old_ring != ring && obj->last_write_req) {/* Keep the request relative to the current ring */ obj->last_write_req = req; } - obj->ring = ring;/* Add a reference if we're newly entering the active list. */ if (!obj->active) {@@ -2230,7 +2234,6 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)intel_fb_obj_flush (obj, true);list_del_init (&obj->ring_list); - obj->ring =NULL ; obj->last_read_req =NULL ; obj->last_write_req =NULL ;@@ -2247,9 +2250,7 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)static void i915_gem_object_retire (struct drm_i915_gem_object *obj) { -struct intel_engine_cs *ring = obj->ring; - -if (ring ==NULL ) +if (obj->last_read_req ==NULL )return ;if (i915_gem_request_completed (obj->last_read_req, true))@@ -2769,14 +2770,17 @@ i915_gem_idle_work_handler(struct work_struct *work)static int i915_gem_object_flush_active (struct drm_i915_gem_object *obj) { +struct intel_engine_cs *ring;int ret;if (obj->active) { + ring =i915_gem_request_get_ring (obj->last_read_req); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret; -i915_gem_retire_requests_ring (obj->ring); +i915_gem_retire_requests_ring (ring); }return 0 ;@@ -2876,9 +2880,11 @@ int i915_gem_object_sync(struct drm_i915_gem_object *obj, struct intel_engine_cs *to) { - struct intel_engine_cs *from = obj->ring; + struct intel_engine_cs *from;int ret, idx; + from =i915_gem_request_get_ring (obj->last_read_req); +if (from ==NULL || to == from)return 0 ;@@ -3889,7 +3895,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,bool was_pin_display;int ret; -if (pipelined != obj->ring) { +if (pipelined !=i915_gem_request_get_ring (obj->last_read_req)) { ret =i915_gem_object_sync (obj, pipelined);if (ret)return ret;@@ -4303,9 +4309,11 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data, ret =i915_gem_object_flush_active (obj); args->busy = obj->active; -if (obj->ring) { +if (obj->last_read_req) { +struct intel_engine_cs *ring;BUILD_BUG_ON (I915_NUM_RINGS >16 ); - args->busy |=intel_ring_flag (obj->ring) <<16 ; + ring =i915_gem_request_get_ring (obj->last_read_req); + args->busy |=intel_ring_flag (ring) <<16 ; }drm_gem_object_unreference (&obj->base);drivers/gpu/drm/i915/i915_gem_context.c +11
-1@@ -613,7 +613,17 @@ static int do_switch(struct intel_engine_cs *ring,* swapped, but there is no way to do that yet. */ from->legacy_hw_ctx.rcs_state->dirty =1 ; -BUG_ON (from->legacy_hw_ctx.rcs_state->ring != ring); + +/* BUG_ON(i915_gem_request_get_ring( +from->legacy_hw_ctx.rcs_state->last_read_req) != ring); */ +/* NB: last_read_req has already been updated to the current +* request however, that request has not yet been submitted. +* Thus last_read_req->ring is guaranteed to be null! +* NB2: Doing the check before the update of last_read_req +* (which happens in i915_vma_move_to_active() just above), +* also fails because last_read_req is almost always null on +* entry! +*/ /* obj is kept alive until the next request by its active ref */ i915_gem_object_ggtt_unpin (from->legacy_hw_ctx.rcs_state);drivers/gpu/drm/i915/i915_gpu_error.c +2
-1@@ -681,7 +681,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->dirty = obj->dirty; err->purgeable = obj->madv !=I915_MADV_WILLNEED ; err->userptr = obj->userptr.mm !=NULL ; - err->ring = obj->ring ? obj->ring->id :-1 ; + err->ring = obj->last_read_req ? +i915_gem_request_get_ring (obj->last_read_req)->id :-1 ; err->cache_level = obj->cache_level; }drivers/gpu/drm/i915/intel_display.c +8
-7@@ -9728,7 +9728,7 @@ static bool use_mmio_flip(struct intel_engine_cs *ring,else if (i915.enable_execlists)return true;else -return ring != obj->ring; +return ring !=i915_gem_request_get_ring (obj->last_read_req); }static void intel_do_mmio_flip (struct intel_crtc *intel_crtc)@@ -9769,8 +9769,6 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (!obj->last_write_req)return 0 ; - ring = obj->ring; -if (i915_gem_request_completed (obj->last_write_req, true))return 0 ;@@ -9778,6 +9776,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)if (ret)return ret; + ring =i915_gem_request_get_ring (obj->last_write_req);if (WARN_ON (!ring->irq_get (ring)))return 0 ;@@ -9837,14 +9836,15 @@ static int intel_queue_mmio_flip(struct drm_device *dev,spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); intel_crtc->mmio_flip.req = obj->last_write_req;i915_gem_request_reference (intel_crtc->mmio_flip.req); - intel_crtc->mmio_flip.ring_id = obj->ring->id; +BUG_ON (ring !=i915_gem_request_get_ring (intel_crtc->mmio_flip.req)); + intel_crtc->mmio_flip.ring_id = ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);/* * Double check to catch cases where irq fired before * mmio flip data was ready */ -intel_notify_mmio_flip (obj->ring); +intel_notify_mmio_flip (ring);return 0 ; }@@ -10022,7 +10022,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, }else if (IS_IVYBRIDGE (dev)) { ring = &dev_priv->ring[BCS ]; }else if (INTEL_INFO (dev)->gen >=7 ) { - ring = obj->ring; + ring =i915_gem_request_get_ring (obj->last_read_req);if (ring ==NULL || ring->id !=RCS ) ring = &dev_priv->ring[BCS ]; }else {@@ -10043,7 +10043,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, goto cleanup_unpin; work->flip_queued_req = obj->last_write_req; - work->flip_queued_ring = obj->ring; + work->flip_queued_ring = +i915_gem_request_get_ring (obj->last_write_req); }else { ret = dev_priv->display.queue_flip (dev, crtc, fb, obj, ring, page_flip_flags);Signature · 1 line
1.7.9.5
Quoted 4 lines
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 21/21] drm/i915: Remove 'obj->ring'
Quoted 27 lines · John Harrison
Quoted 5 lines · Daniel Vetter
Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:25PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]Quoted 8 lines
I think this should be split up into the different parts:
- s/obj->ring/obj->last_read_req->ring/ for all the cases that just want the current ring.
- s/obj->ring/obj->last_read_req/ I think in a bunch of places the code would actually be more readable if we'd check forobj->activeinstead.
- All the oddball special cases probably deserve their own commit + nice explanation in the commit message about why the change is correct.Can you explain which is which? As for why the change is correct, it is not a functional change.
obj->ringwas assigned withobj->last_read_seqnoand cleared withobj->last_read_seqno. Thus queryingobj->ringis equivalent to queryingobj->last_read_req->ringin all cases. Given that the ring is now available fromobj->lrr, it seemed redundant to also have it explicitly saved inobj->ring, hence the patch to remove it.AFAICT, there are two examples in debugfs that should just be querying
obj->active. The rest are pretty much wanting the currently in use ring and/or are about to uselast_read_reqanyway. It seems more sensible to say 'if(obj->lrr){do_stuff(obj->lrr); }' than 'if(obj->active){do_stuff(obj->lrr); }'. Nothing looks particularly 'oddball' to me!
[snip]
Quoted 2 lines
Quoted 2 lines
drivers/gpu/drm/i915/i915_debugfs.c static int per_file_stats(int id, void *ptr, void *data) · line 341
Quoted 2 lines
drivers/gpu/drm/i915/i915_debugfs.c +1
-0+if (obj->last_read_req)/* XXX per-vma statistic */ stats->active += obj->base.size;
Stuff like this here is what I've meant. Checking for obj->active here is better, and you might as well ditch the XXX comment too.
You're right that it's all semantically equivalent and your change here is the less risky since there's no way behaviour can change. But code is mostly written so that other humans can understand it (when debugging it), so conveying as much meaning as possible is important. And this looks like a good opportunity to review all the various places and make sure the code is sane. We have accumulated a bit of cruft from the per-vma rework and there's more cruft on the horizon with potentially multiple read fences, so I think this is worth it.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 20/21] drm/i915: Convert 'ring_idle()' to use requests not seqnos
In reply to [email protected]Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:24PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
We have places that shovel stuff onto the ring without an explicit add_request. Or at least we've had, so this needs a full audit, and that audit needs to be in the commit message.
-Daniel
Quoted 48 lines
--- drivers/gpu/drm/i915/i915_irq.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)drivers/gpu/drm/i915/i915_irq.c +6
-6@@ -3030,18 +3030,18 @@ static void gen8_disable_vblank(struct drm_device *dev, int pipe)spin_unlock_irqrestore (&dev_priv->irq_lock, irqflags); } -static u32 -ring_last_seqno (struct intel_engine_cs *ring) +static struct drm_i915_gem_request * +ring_last_request (struct intel_engine_cs *ring) {return list_entry (ring->request_list.prev, - struct drm_i915_gem_request, list)->seqno; + struct drm_i915_gem_request, list); }static bool -ring_idle (struct intel_engine_cs *ring,u32 seqno) +ring_idle (struct intel_engine_cs *ring) {return (list_empty (&ring->request_list) || -i915_seqno_passed (seqno,ring_last_seqno (ring))); +i915_gem_request_completed (ring_last_request (ring), false)); } static bool@@ -3261,7 +3261,7 @@ static void i915_hangcheck_elapsed(unsigned long data) acthd =intel_ring_get_active_head (ring);if (ring->hangcheck.seqno == seqno) { -if (ring_idle (ring, seqno)) { +if (ring_idle (ring)) { ring->hangcheck.action =HANGCHECK_IDLE ;if (waitqueue_active (&ring->irq_queue)) {Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 20/21] drm/i915: Convert 'ring_idle()' to use requests not seqnos
Quoted 9 lines · Daniel Vetter
Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:24PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]We have places that shovel stuff onto the ring without an explicit
add_request. Or at least we've had, so this needs a full audit, and that audit needs to be in the commit message.
-Daniel
Not sure what you mean. There is no functional change here. The old version pulled the seqno out of the last request entry in the list and compared that to the hardware seqno value to check for completion. The new version gets the same request entry and does the same completion test, just in request parlance. If commands are being dumped on the ring without an request being added to the request list then the old version would still not have been checking for them. It would still only look at the last piece of work that actually did call add_request and see if that has completed or not.
Quoted 40 lines
Quoted 40 lines
--- drivers/gpu/drm/i915/i915_irq.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)drivers/gpu/drm/i915/i915_irq.c +6
-6@@ -3030,18 +3030,18 @@ static void gen8_disable_vblank(struct drm_device *dev, int pipe)spin_unlock_irqrestore (&dev_priv->irq_lock, irqflags); } -static u32 -ring_last_seqno (struct intel_engine_cs *ring) +static struct drm_i915_gem_request * +ring_last_request (struct intel_engine_cs *ring) {return list_entry (ring->request_list.prev, - struct drm_i915_gem_request, list)->seqno; + struct drm_i915_gem_request, list); }static bool -ring_idle (struct intel_engine_cs *ring,u32 seqno) +ring_idle (struct intel_engine_cs *ring) {return (list_empty (&ring->request_list) || -i915_seqno_passed (seqno,ring_last_seqno (ring))); +i915_gem_request_completed (ring_last_request (ring), false)); } static bool@@ -3261,7 +3261,7 @@ static void i915_hangcheck_elapsed(unsigned long data) acthd =intel_ring_get_active_head (ring);if (ring->hangcheck.seqno == seqno) { -if (ring_idle (ring, seqno)) { +if (ring_idle (ring)) { ring->hangcheck.action =HANGCHECK_IDLE ;
Quoted 8 lines
Quoted 8 lines
drivers/gpu/drm/i915/i915_irq.c +0
-0if (waitqueue_active (&ring->irq_queue)) {Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 20/21] drm/i915: Convert 'ring_idle()' to use requests not seqnos
Quoted 4 lines · John Harrison
Quoted 3 lines · Daniel Vetter
We have places that shovel stuff onto the ring without an explicit
add_request. Or at least we've had, so this needs a full audit, and that audit needs to be in the commit message.
Quoted 8 lines
Not sure what you mean. There is no functional change here. The old version pulled the seqno out of the last request entry in the list and compared that to the hardware seqno value to check for completion. The new version gets the same request entry and does the same completion test, just in request parlance. If commands are being dumped on the ring without an request being added to the request list then the old version would still not have been checking for them. It would still only look at the last piece of work that actually did call
add_requestand see if that has completed or not.
Oh, I've pretty badly mixed stuff up ;-)
The issue I'm seeing here though is acessing the request list, from irq context, without any locks. Pre-existing, but still deserves a FIXME comment.
Wrt the change itself I'm not sure it's that useful, this is fairly low-level code.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 19/21] drm/i915: Convert semaphores to handle requests not seqnos
In reply to [email protected]Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:23PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
This smells funky on a quick look. I'd expect some reference counting in here, or at least an explanatation for why it's not needed.
But even without that I'm not sure we want to track requests at the semaphore level. After all hw semaphores work with seqno numbers and not requests, and depending upon how the scheduler materializes seqnos shoveling requests through these functions makes no sense at all.
I'm maybe missing something big here, but with the thin commit message I can't tell. So I vote to just drop this one for now.
-Daniel
Quoted 191 lines
--- drivers/gpu/drm/i915/i915_debugfs.c | 3 ++- drivers/gpu/drm/i915/i915_gem.c | 14 ++++++-------- drivers/gpu/drm/i915/i915_gpu_error.c | 12 ++++++++---- drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 4 ++-- 5 files changed, 24 insertions(+), 19 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +2
-1@@ -2604,7 +2604,8 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)seq_puts (m,"\nSync seqno:\n" );for_each_ring (ring, dev_priv, i) {for (j =0 ; j < num_rings; j++) { -seq_printf (m," 0x%08x " , ring->semaphore.sync_seqno[j]); +seq_printf (m," 0x%08x " , +i915_gem_request_get_seqno (ring->semaphore.sync_req[j])); }seq_putc (m,'\n' ); }drivers/gpu/drm/i915/i915_gem.c +6
-8@@ -2275,8 +2275,8 @@ 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->semaphore.sync_seqno); j++) - ring->semaphore.sync_seqno[j] =0 ; +for (j =0 ; j <ARRAY_SIZE (ring->semaphore.sync_req); j++) + ring->semaphore.sync_req[j] =NULL ; }return 0 ;@@ -2877,7 +2877,6 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, struct intel_engine_cs *to) {struct intel_engine_cs *from = obj->ring; -u32 seqno;int ret, idx;if (from ==NULL || to == from)@@ -2888,10 +2887,10 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, idx =intel_ring_sync_index (from, to); - seqno =i915_gem_request_get_seqno (obj->last_read_req);/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ -if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ +if (i915_gem_request_get_seqno (obj->last_read_req) <= +i915_gem_request_get_seqno (from->semaphore.sync_req[idx]))/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; ret =i915_gem_check_olr (obj->last_read_req);@@ -2899,14 +2898,13 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,return ret;trace_i915_gem_ring_sync_to (from, to, obj->last_read_req); - ret = to->semaphore.sync_to (to, from, seqno); + ret = to->semaphore.sync_to (to, from, obj->last_read_req);if (!ret)/* We use last_read_req because sync_to() * might have just caused seqno wrap under * the radar. */ - from->semaphore.sync_seqno[idx] = -i915_gem_request_get_seqno (obj->last_read_req); + from->semaphore.sync_req[idx] = obj->last_read_req;return ret; }drivers/gpu/drm/i915/i915_gpu_error.c +8
-4@@ -822,7 +822,8 @@ static void gen8_record_semaphore_state(struct drm_i915_private *dev_priv, idx =intel_ring_sync_index (ring, to); ering->semaphore_mboxes[idx] = tmp[signal_offset]; - ering->semaphore_seqno[idx] = ring->semaphore.sync_seqno[idx]; + ering->semaphore_seqno[idx] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[idx]); } }@@ -832,13 +833,16 @@ static void gen6_record_semaphore_state(struct drm_i915_private *dev_priv, { ering->semaphore_mboxes[0 ] =I915_READ (RING_SYNC_0 (ring->mmio_base)); ering->semaphore_mboxes[1 ] =I915_READ (RING_SYNC_1 (ring->mmio_base)); - ering->semaphore_seqno[0 ] = ring->semaphore.sync_seqno[0 ]; - ering->semaphore_seqno[1 ] = ring->semaphore.sync_seqno[1 ]; + ering->semaphore_seqno[0 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[0 ]); + ering->semaphore_seqno[1 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[1 ]);if (HAS_VEBOX (dev_priv->dev)) { ering->semaphore_mboxes[2 ] =I915_READ (RING_SYNC_2 (ring->mmio_base)); - ering->semaphore_seqno[2 ] = ring->semaphore.sync_seqno[2 ]; + ering->semaphore_seqno[2 ] = +i915_gem_request_get_seqno (ring->semaphore.sync_req[2 ]); } }drivers/gpu/drm/i915/intel_ringbuffer.c +6
-4@@ -1042,7 +1042,7 @@ static inline bool i915_gem_has_seqno_wrapped(struct drm_device *dev,static int gen8_ring_sync (struct intel_engine_cs *waiter,struct intel_engine_cs *signaller, -u32 seqno) +struct drm_i915_gem_request *req) {struct drm_i915_private *dev_priv = waiter->dev->dev_private;int ret;@@ -1055,7 +1055,7 @@ gen8_ring_sync(struct intel_engine_cs *waiter,MI_SEMAPHORE_GLOBAL_GTT |MI_SEMAPHORE_POLL | MI_SEMAPHORE_SAD_GTE_SDD); -intel_ring_emit (waiter, seqno); +intel_ring_emit (waiter,i915_gem_request_get_seqno (req));intel_ring_emit (waiter,lower_32_bits (GEN8_WAIT_OFFSET (waiter, signaller->id))); intel_ring_emit(waiter,@@ -1067,18 +1067,20 @@ gen8_ring_sync(struct intel_engine_cs *waiter,static int gen6_ring_sync (struct intel_engine_cs *waiter,struct intel_engine_cs *signaller, -u32 seqno) +struct drm_i915_gem_request *req) {u32 dw1 =MI_SEMAPHORE_MBOX |MI_SEMAPHORE_COMPARE |MI_SEMAPHORE_REGISTER ;u32 wait_mbox = signaller->semaphore.mbox.wait[waiter->id];int ret; +u32 seqno;/* Throughout all of the GEM code, seqno passed implies our current * seqno is >= the last seqno executed. However for hardware the * comparison is strictly greater than. */ + seqno =i915_gem_request_get_seqno (req); seqno -=1 ;WARN_ON (wait_mbox ==MI_SEMAPHORE_SYNC_INVALID );@@ -1794,7 +1796,7 @@ static int intel_init_ring_buffer(struct drm_device *dev,INIT_LIST_HEAD (&ring->execlist_queue); ringbuf->size =32 *PAGE_SIZE ; ringbuf->ring = ring; -memset (ring->semaphore.sync_seqno,0 ,sizeof (ring->semaphore.sync_seqno)); +memset (ring->semaphore.sync_req,0 ,sizeof (ring->semaphore.sync_req));init_waitqueue_head (&ring->irq_queue);drivers/gpu/drm/i915/intel_ringbuffer.h +2
-2@@ -211,7 +211,7 @@ struct intel_engine_cs {* ie. transpose of f(x, y) */ struct { - u32 sync_seqno[I915_NUM_RINGS-1]; + struct drm_i915_gem_request *sync_req[I915_NUM_RINGS-1]; union { struct {@@ -226,7 +226,7 @@ struct intel_engine_cs {/* AKA wait() */ int (*sync_to)(struct intel_engine_cs *ring,struct intel_engine_cs *to, -u32 seqno); +struct drm_i915_gem_request *req);int (*signal)(struct intel_engine_cs *signaller,/* num_dwords needed by caller */ unsigned int num_dwords);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
RFC 16/25 drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed'
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_debugfs.c | 3 +-- drivers/gpu/drm/i915/i915_drv.h | 18 ++++++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 12 ++++-------- drivers/gpu/drm/i915/intel_display.c | 11 +++-------- 4 files changed, 26 insertions(+), 18 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -2
@@ -550,8 +550,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - i915_gem_request_get_seqno(work->flip_queued_req))); + i915_gem_request_completed(work->flip_queued_req, true)); }else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",
drivers/gpu/drm/i915/i915_drv.h +18 -0
@@ -1970,6 +1970,12 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } +/* +* NB: i915_gem_request_completed should be here but currently needs the +* definition of i915_seqno_passed() which is below. It will be moved in +* a later patch when the call to i915_seqno_passed() is obsoleted... +*/ +struct drm_i915_file_private {struct drm_i915_private *dev_priv;struct drm_file *file;
@@ -3019,4 +3025,16 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +u32 seqno; + +BUG_ON (req ==NULL ); + + seqno = req->ring->get_seqno (req->ring, lazy_coherency); + +return i915_seqno_passed (seqno, req->seqno); +} +#endif
drivers/gpu/drm/i915/i915_gem.c +4 -8
@@ -2253,8 +2253,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)if (ring ==NULL )return ; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_read_req))) +if (i915_gem_request_completed (obj->last_read_req, true))i915_gem_object_move_to_inactive (obj); }
@@ -2521,12 +2520,9 @@ struct drm_i915_gem_request * i915_gem_find_active_request(struct intel_engine_cs *ring) {struct drm_i915_gem_request *request; -u32 completed_seqno; - - completed_seqno = ring->get_seqno (ring, false);list_for_each_entry (request, &ring->request_list, list) { -if (i915_seqno_passed (completed_seqno, request->seqno)) +if (i915_gem_request_completed (request, false))continue ;return request;
@@ -2670,7 +2666,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req))) +if (!i915_gem_request_completed (obj->last_read_req, true))break ;i915_gem_object_move_to_inactive (obj);
@@ -2685,7 +2681,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_request, list); -if (!i915_seqno_passed (seqno, request->seqno)) +if (!i915_gem_request_completed (request, true))break ;trace_i915_gem_request_retire (ring, request->seqno);
drivers/gpu/drm/i915/intel_display.c +3 -8
@@ -9771,8 +9771,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) ring = obj->ring; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_write_req))) +if (i915_gem_request_completed (obj->last_write_req, true))return 0 ; ret =i915_gem_check_olr (obj->last_write_req);
@@ -9790,9 +9789,6 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct drm_i915_private *dev_priv =to_i915 (ring->dev);struct intel_crtc *intel_crtc;unsigned long irq_flags; -u32 seqno; - - seqno = ring->get_seqno (ring, false);spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags);for_each_intel_crtc (ring->dev, intel_crtc) {
@@ -9805,7 +9801,7 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) { +if (i915_gem_request_completed (mmio_flip->req, false)) {intel_do_mmio_flip (intel_crtc);i915_gem_request_unreference (mmio_flip->req); mmio_flip->req =NULL ;
@@ -9878,8 +9874,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && - !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_gem_request_get_seqno (work->flip_queued_req))) + !i915_gem_request_completed (work->flip_queued_req, true))return false; work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);
Signature · 1 line
1.7.9.5
Re: [RFC 16/25] drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed'
Quoted 47 lines · On Fri, Oct 10, 2014 at 12:39:49PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_debugfs.c | 3 +-- drivers/gpu/drm/i915/i915_drv.h | 18 ++++++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 12 ++++-------- drivers/gpu/drm/i915/intel_display.c | 11 +++-------- 4 files changed, 26 insertions(+), 18 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +1
-2@@ -550,8 +550,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - i915_gem_request_get_seqno(work->flip_queued_req))); + i915_gem_request_completed(work->flip_queued_req, true)); }else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",drivers/gpu/drm/i915/i915_drv.h +7
-0@@ -1970,6 +1970,12 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } +/* +* NB: i915_gem_request_completed should be here but currently needs the +* definition of i915_seqno_passed() which is below. It will be moved in +* a later patch when the call to i915_seqno_passed() is obsoleted... +*/ +struct drm_i915_file_private {struct drm_i915_private *dev_priv;struct drm_file *file;@@ -3019,4 +3025,16 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,
drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3038
Quoted 10 lines
drivers/gpu/drm/i915/i915_drv.h +10
-0+ bool lazy_coherency) +{ +u32 seqno; + +BUG_ON (req ==NULL ); + + seqno = req->ring->get_seqno (req->ring, lazy_coherency); + +return i915_seqno_passed (seqno, req->seqno); +}
Tbh this thing looks a bit big for a static inline. Have you double-checked that inlining this doesn't increase binary size? If it does then we need performance data justifying the cost. Same probably applies to some of the other static inline functions you've added. -Daniel
Quoted 98 lines
drivers/gpu/drm/i915/i915_drv.h +1
-0+#endif drivers/gpu/drm/i915/i915_gem.c +4
-8@@ -2253,8 +2253,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)if (ring ==NULL )return ; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_read_req))) +if (i915_gem_request_completed (obj->last_read_req, true))i915_gem_object_move_to_inactive (obj); }@@ -2521,12 +2520,9 @@ struct drm_i915_gem_request * i915_gem_find_active_request(struct intel_engine_cs *ring) {struct drm_i915_gem_request *request; -u32 completed_seqno; - - completed_seqno = ring->get_seqno (ring, false);list_for_each_entry (request, &ring->request_list, list) { -if (i915_seqno_passed (completed_seqno, request->seqno)) +if (i915_gem_request_completed (request, false))continue ;return request;@@ -2670,7 +2666,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req))) +if (!i915_gem_request_completed (obj->last_read_req, true))break ;i915_gem_object_move_to_inactive (obj);@@ -2685,7 +2681,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_request, list); -if (!i915_seqno_passed (seqno, request->seqno)) +if (!i915_gem_request_completed (request, true))break ;trace_i915_gem_request_retire (ring, request->seqno);drivers/gpu/drm/i915/intel_display.c +3
-8@@ -9771,8 +9771,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) ring = obj->ring; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_write_req))) +if (i915_gem_request_completed (obj->last_write_req, true))return 0 ; ret =i915_gem_check_olr (obj->last_write_req);@@ -9790,9 +9789,6 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct drm_i915_private *dev_priv =to_i915 (ring->dev);struct intel_crtc *intel_crtc;unsigned long irq_flags; -u32 seqno; - - seqno = ring->get_seqno (ring, false);spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags);for_each_intel_crtc (ring->dev, intel_crtc) {@@ -9805,7 +9801,7 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) { +if (i915_gem_request_completed (mmio_flip->req, false)) {intel_do_mmio_flip (intel_crtc);i915_gem_request_unreference (mmio_flip->req); mmio_flip->req =NULL ;@@ -9878,8 +9874,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && - !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_gem_request_get_seqno (work->flip_queued_req))) + !i915_gem_request_completed (work->flip_queued_req, true))return false; work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 16/25] drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed'
Quoted 47 lines · Daniel Vetter
Quoted 46 lines · On Fri, Oct 10, 2014 at 12:39:49PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_debugfs.c | 3 +-- drivers/gpu/drm/i915/i915_drv.h | 18 ++++++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 12 ++++-------- drivers/gpu/drm/i915/intel_display.c | 11 +++-------- 4 files changed, 26 insertions(+), 18 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +1
-2@@ -550,8 +550,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - i915_gem_request_get_seqno(work->flip_queued_req))); + i915_gem_request_completed(work->flip_queued_req, true)); }else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",drivers/gpu/drm/i915/i915_drv.h +6
-0@@ -1970,6 +1970,12 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } +/* +* NB: i915_gem_request_completed should be here but currently needs the +* definition of i915_seqno_passed() which is below. It will be moved in +* a later patch when the call to i915_seqno_passed() is obsoleted... +*/ +struct drm_i915_file_private {struct drm_i915_private *dev_priv;struct drm_file *file;@@ -3019,4 +3025,16 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } }
Quoted 16 lines
drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3038
Quoted 11 lines
drivers/gpu/drm/i915/i915_drv.h +11
-0+static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +u32 seqno; + +BUG_ON (req ==NULL ); + + seqno = req->ring->get_seqno (req->ring, lazy_coherency); + +return i915_seqno_passed (seqno, req->seqno); +}Tbh this thing looks a bit big for a static inline. Have you double-checked that inlining this doesn't increase binary size? If it does then we need performance data justifying the cost. Same probably applies to some of the other static inline functions you've added. -Daniel
It is a replacement for i915_seqno_passed() which is a static inline so it seemed sensible that this should be too. Note that the ultimate aim is for this to be a simple 'return req->completed'. At which point it should certainly be worthy of inlining.
But no I haven't done any actual analysis of the benefit/cost yet. As has been mentioned, this RFC was a quick knock up to get something to discuss as soon as possible.
Quoted 90 lines
Quoted 90 lines
drivers/gpu/drm/i915/i915_drv.h +1
-0+#endif drivers/gpu/drm/i915/i915_gem.c +4
-8@@ -2253,8 +2253,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)if (ring ==NULL )return ; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_read_req))) +if (i915_gem_request_completed (obj->last_read_req, true))i915_gem_object_move_to_inactive (obj); }@@ -2521,12 +2520,9 @@ struct drm_i915_gem_request * i915_gem_find_active_request(struct intel_engine_cs *ring) {struct drm_i915_gem_request *request; -u32 completed_seqno; - - completed_seqno = ring->get_seqno (ring, false);list_for_each_entry (request, &ring->request_list, list) { -if (i915_seqno_passed (completed_seqno, request->seqno)) +if (i915_gem_request_completed (request, false))continue ;return request;@@ -2670,7 +2666,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req))) +if (!i915_gem_request_completed (obj->last_read_req, true))break ;i915_gem_object_move_to_inactive (obj);@@ -2685,7 +2681,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_request, list); -if (!i915_seqno_passed (seqno, request->seqno)) +if (!i915_gem_request_completed (request, true))break ;trace_i915_gem_request_retire (ring, request->seqno);drivers/gpu/drm/i915/intel_display.c +3
-8@@ -9771,8 +9771,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) ring = obj->ring; -if (i915_seqno_passed (ring->get_seqno (ring, true), -i915_gem_request_get_seqno (obj->last_write_req))) +if (i915_gem_request_completed (obj->last_write_req, true))return 0 ; ret =i915_gem_check_olr (obj->last_write_req);@@ -9790,9 +9789,6 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct drm_i915_private *dev_priv =to_i915 (ring->dev);struct intel_crtc *intel_crtc;unsigned long irq_flags; -u32 seqno; - - seqno = ring->get_seqno (ring, false);spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags);for_each_intel_crtc (ring->dev, intel_crtc) {@@ -9805,7 +9801,7 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) { +if (i915_gem_request_completed (mmio_flip->req, false)) {intel_do_mmio_flip (intel_crtc);i915_gem_request_unreference (mmio_flip->req); mmio_flip->req =NULL ;@@ -9878,8 +9874,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && - !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), -i915_gem_request_get_seqno (work->flip_queued_req))) + !i915_gem_request_completed (work->flip_queued_req, true))return false;
Quoted 8 lines
Quoted 8 lines
drivers/gpu/drm/i915/intel_display.c +0
-0work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 14/21] drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request'
In reply to [email protected]Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:18PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Again on the topic of thin commit message: Some additional words about why we can noodle around in request structures from hardirq context (those fields are invariant after add_request) would be good. Since with the scheduler a lot of that stuff suddenly is a lot less invariant (if we delay the seqno allocation until we emit requests, which I'm pretty sure we have to for a bunch of reasons).
Quoted 14 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_debugfs.c | 4 ++-- drivers/gpu/drm/i915/intel_display.c | 9 ++++++--- drivers/gpu/drm/i915/intel_drv.h | 2 +- 3 files changed, 9 insertions(+), 6 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +0
-0
drivers/gpu/drm/i915/i915_debugfs.c static int i915_gem_pageflip_info(struct seq_file *m, void *data) · line 554
Quoted 10 lines
drivers/gpu/drm/i915/i915_debugfs.c +2
-2@@ -547,11 +547,11 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)if (work->flip_queued_ring) {seq_printf (m,"Flip queued on %s at seqno %u, next seqno %u [current breadcrumb %u], completed? %d\n" , work->flip_queued_ring->name, - work->flip_queued_seqno, +i915_gem_request_get_seqno (work->flip_queued_req), dev_priv->next_seqno, work->flip_queued_ring->get_seqno (work->flip_queued_ring, true),i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - work->flip_queued_seqno)); +i915_gem_request_get_seqno (work->flip_queued_req)));
Yeah it's no your patch, but this stuff has defintely fallen over the edge. A prep patch to extract some helper functions to bring sanity to i915_gem_pageflip_info's indentation levels is needed here imo. -Daniel
Quoted 60 lines
drivers/gpu/drm/i915/i915_debugfs.c +0
-0}else seq_printf (m,"Flip not associated with any ring\n" ); seq_printf(m, "Flip queued on frame %d, (was ready on frame %d), now %d\n",drivers/gpu/drm/i915/intel_display.c +6
-3@@ -9879,7 +9879,7 @@ static bool __intel_pageflip_stall_check(struct drm_device *dev,if (work->flip_ready_vblank ==0 ) {if (work->flip_queued_ring && !i915_seqno_passed (work->flip_queued_ring->get_seqno (work->flip_queued_ring, true), - work->flip_queued_seqno)) +i915_gem_request_get_seqno (work->flip_queued_req)))return false; work->flip_ready_vblank =drm_vblank_count (dev, intel_crtc->pipe);@@ -10047,7 +10047,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =i915_gem_request_get_seqno (obj->last_write_req); + work->flip_queued_req = obj->last_write_req; work->flip_queued_ring = obj->ring; }else { ret = dev_priv->display.queue_flip(dev, crtc, fb, obj, ring,@@ -10055,10 +10055,13 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); + work->flip_queued_req =intel_ring_get_request (ring); work->flip_queued_ring = ring; } +if (work->flip_queued_req) +i915_gem_request_reference (work->flip_queued_req); + work->flip_queued_vblank =drm_vblank_count (dev, intel_crtc->pipe); work->enable_stall_check = true;drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -687,7 +687,7 @@ struct intel_unpin_work {u32 flip_count;u32 gtt_offset;struct intel_engine_cs *flip_queued_ring; -u32 flip_queued_seqno; +struct drm_i915_gem_request *flip_queued_req;int flip_queued_vblank;int flip_ready_vblank;bool enable_stall_check;Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 13/21] drm/i915: Convert mmio_flip::seqno to struct request
In reply to [email protected]Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:17PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Again on the topic of thin commit messages: Beyond the boilerplate blabla explaining why we do all the s/seqno/request/ stuff for the benefit of the future reader this definitely needs a mention of the refcounting needed for the async work.
-Daniel
Quoted 71 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/intel_display.c | 12 +++++++----- drivers/gpu/drm/i915/intel_drv.h | 2 +- 2 files changed, 8 insertions(+), 6 deletions(-)drivers/gpu/drm/i915/intel_display.c +7
-5@@ -9799,15 +9799,16 @@ void intel_notify_mmio_flip(struct intel_engine_cs *ring)struct intel_mmio_flip *mmio_flip; mmio_flip = &intel_crtc->mmio_flip; -if (mmio_flip->seqno ==0 ) +if (mmio_flip->req ==NULL )continue ;if (ring->id != mmio_flip->ring_id)continue ; -if (i915_seqno_passed (seqno, mmio_flip->seqno)) { +if (i915_seqno_passed (seqno,i915_gem_request_get_seqno (mmio_flip->req))) {intel_do_mmio_flip (intel_crtc); - mmio_flip->seqno =0 ; +i915_gem_request_unreference (mmio_flip->req); + mmio_flip->req =NULL ; ring->irq_put (ring); } }@@ -9826,7 +9827,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev,unsigned long irq_flags;int ret; -if (WARN_ON (intel_crtc->mmio_flip.seqno)) +if (WARN_ON (intel_crtc->mmio_flip.req))return -EBUSY ; ret =intel_postpone_flip (obj);@@ -9838,7 +9839,8 @@ static int intel_queue_mmio_flip(struct drm_device *dev, }spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); - intel_crtc->mmio_flip.seqno =i915_gem_request_get_seqno (obj->last_write_req); + intel_crtc->mmio_flip.req = obj->last_write_req; +i915_gem_request_reference (intel_crtc->mmio_flip.req); intel_crtc->mmio_flip.ring_id = obj->ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -395,7 +395,7 @@ struct intel_pipe_wm { };struct intel_mmio_flip { -u32 seqno; +struct drm_i915_gem_request *req;u32 ring_id; };Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 10/21] drm/i915: Convert 'last_flip_req' to be a request not a seqno
In reply to [email protected]Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:14PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
I know there's often not a lot to talk about for if you have a refactoring step that needs to be applied n times. But even then a small commit message to reiterate what is going on and why and a small note if there's anything funky goes a long way.
Since in a few months someone will digg out your patch here using git blame and git lock --pickaxe and will be totally lost without the context of the entire series. So each patch really needs to be able to be understood on its own.
-Daniel
Quoted 79 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/intel_overlay.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)drivers/gpu/drm/i915/intel_overlay.c +8
-8@@ -182,7 +182,7 @@ struct intel_overlay {u32 flip_addr;struct drm_i915_gem_object *reg_bo;/* flip handling */ -uint32_t last_flip_req; +struct drm_i915_gem_request *last_flip_req;void (*flip_tail)(struct intel_overlay *); };@@ -217,17 +217,17 @@ static int intel_overlay_do_wait_request(struct intel_overlay *overlay,int ret;BUG_ON (overlay->last_flip_req); - ret =i915_add_request (ring, &overlay->last_flip_req); + ret =i915_add_request (ring, &overlay->last_flip_req->seqno);if (ret)return ret; overlay->flip_tail = tail; - ret =i915_wait_seqno (ring, overlay->last_flip_req); + ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req));if (ret)return ret;i915_gem_retire_requests (dev); - overlay->last_flip_req =0 ; + overlay->last_flip_req =NULL ;return 0 ; }@@ -286,7 +286,7 @@ static int intel_overlay_continue(struct intel_overlay *overlay,intel_ring_emit (ring, flip_addr);intel_ring_advance (ring); -return i915_add_request (ring, &overlay->last_flip_req); +return i915_add_request (ring, &overlay->last_flip_req->seqno); }static void intel_overlay_release_old_vid_tail (struct intel_overlay *overlay)@@ -366,10 +366,10 @@ static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)struct intel_engine_cs *ring = &dev_priv->ring[RCS ];int ret; -if (overlay->last_flip_req ==0 ) +if (overlay->last_flip_req ==NULL )return 0 ; - ret =i915_wait_seqno (ring, overlay->last_flip_req); + ret =i915_wait_seqno (ring,i915_gem_request_get_seqno (overlay->last_flip_req));if (ret)return ret;i915_gem_retire_requests (dev);@@ -377,7 +377,7 @@ static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)if (overlay->flip_tail) overlay->flip_tail (overlay); - overlay->last_flip_req =0 ; + overlay->last_flip_req =NULL ;return 0 ; }Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 09/21] drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno
In reply to [email protected]Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:13PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
To thin commit message.
Also I wonder whethere we should track the olr state more explicitly in the request structure instead of jumping through all these hoops. And explicit olr state for a request might also help to clarify control flow and checks in a bunch of places.
Quoted 22 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_drv.h | 17 ++++++++++++++++- drivers/gpu/drm/i915/i915_gem.c | 25 ++++++++++--------------- drivers/gpu/drm/i915/intel_display.c | 2 +- 3 files changed, 27 insertions(+), 17 deletions(-)drivers/gpu/drm/i915/i915_drv.h +1
-1@@ -2483,7 +2483,7 @@ bool i915_gem_retire_requests(struct drm_device *dev);void i915_gem_retire_requests_ring (struct intel_engine_cs *ring);int __must_check i915_gem_check_wedge (struct i915_gpu_error *error,bool interruptible); -int __must_check i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno); +int __must_check i915_gem_check_olr (struct drm_i915_gem_request *req);static inline bool i915_reset_in_progress (struct i915_gpu_error *error) {
drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3023
Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +1
-0@@ -3020,4 +3020,19 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +/************************* Deprecated *************************/
There's a gcc flag for functions if you want to mark this as depracated, otherwise I'd just go with a "XXX: This temporary function will disappear at the end of the patch series" somewhere.
drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3028
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +5
-0+static inline int __must_check i915_gem_check_ols (struct intel_engine_cs *ring,u32 seqno) +{ +int ret; + +BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex));
WARN_ON instead of BUG_ON for locking checks.
Quoted 20 lines
drivers/gpu/drm/i915/i915_drv.h +9
-0+ + ret =0 ; +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) + ret =i915_add_request (ring,NULL ); + + return ret; +} +/************************* Deprecated *************************/ +#endif drivers/gpu/drm/i915/i915_gem.c +0
-2@@ -1097,19 +1097,18 @@ i915_gem_check_wedge(struct i915_gpu_error *error, }/* -* Compare seqno against outstanding lazy request. Emit a request if they are -* equal.
drivers/gpu/drm/i915/i915_gem.c i915_gem_check_wedge(struct i915_gpu_error *error, · line 1107
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +3
-2+* Compare arbitrary request against outstanding lazy request. Emit on match. */ int -i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno) +i915_gem_check_olr (struct drm_i915_gem_request *req) {int ret; -BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); +BUG_ON (!mutex_is_locked (&req->ring->dev->struct_mutex));
Same here. Plus please a stern warning in the commit message and cc: to whomever who managed to sneak this in (and failed to catch it in review).
Quoted 88 lines
drivers/gpu/drm/i915/i915_gem.c +7
-11ret =0 ; -if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) - ret =i915_add_request (ring,NULL ); +if (req == req->ring->outstanding_lazy_request) + ret =i915_add_request (req->ring,NULL );return ret; }@@ -1271,7 +1270,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno)if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_ols (ring, seqno);if (ret)return ret;@@ -1338,7 +1337,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,struct drm_i915_private *dev_priv = dev->dev_private;struct intel_engine_cs *ring = obj->ring;unsigned reset_counter; -u32 seqno;int ret;BUG_ON (!mutex_is_locked (&dev->struct_mutex));@@ -1348,21 +1346,18 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,if (!req)return 0 ; - seqno =i915_gem_request_get_seqno (req); -BUG_ON (seqno ==0 ); - ret =i915_gem_check_wedge (&dev_priv->gpu_error, true);if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_olr (req);if (ret)return ret; reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter);i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); - ret =__wait_seqno (ring, seqno, reset_counter, true,NULL , file_priv); + ret =__wait_seqno (ring,i915_gem_request_get_seqno (req), reset_counter, true,NULL , file_priv);mutex_lock (&dev->struct_mutex);i915_gem_request_unreference (req);if (ret)@@ -2786,7 +2781,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)int ret;if (obj->active) { - ret =i915_gem_check_olr (obj->ring,i915_gem_request_get_seqno (obj->last_read_req)); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;@@ -2913,7 +2908,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; - ret =i915_gem_check_olr (obj->ring, seqno); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;drivers/gpu/drm/i915/intel_display.c +1
-1@@ -9775,7 +9775,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) i915_gem_request_get_seqno(obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req)); + ret =i915_gem_check_olr (obj->last_write_req);if (ret)return ret;Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 09/21] drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno
Quoted 8 lines · Daniel Vetter
Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:13PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
To thin commit message.
Also I wonder whethere we should track the olr state more explicitly in the request structure instead of jumping through all these hoops. And explicit olr state for a request might also help to clarify control flow and checks in a bunch of places.
What would be nicer would be to get rid of the OLR completely. If code wants to write to the ring then it should request a request structure at the start, pass that request in to all the ring writing code and submit the same request structure at the end. The whole idea of some random request floating around in the ether scooping up any ring accesses that happen to occur with no idea about ownership, sequence, etc. just seems highly dodgy. Indeed it definitely causes issues with the scheduler with things like page flips being mixed up in completely unrelated batch buffer submissions! I believe that is one of the things Chris did in his re-write the universe patch.
Quoted 150 lines
Quoted 22 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_drv.h | 17 ++++++++++++++++- drivers/gpu/drm/i915/i915_gem.c | 25 ++++++++++--------------- drivers/gpu/drm/i915/intel_display.c | 2 +- 3 files changed, 27 insertions(+), 17 deletions(-)drivers/gpu/drm/i915/i915_drv.h +1
-1@@ -2483,7 +2483,7 @@ bool i915_gem_retire_requests(struct drm_device *dev);void i915_gem_retire_requests_ring (struct intel_engine_cs *ring);int __must_check i915_gem_check_wedge (struct i915_gpu_error *error,bool interruptible); -int __must_check i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno); +int __must_check i915_gem_check_olr (struct drm_i915_gem_request *req);static inline bool i915_reset_in_progress (struct i915_gpu_error *error) {drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3023
Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +1
-0@@ -3020,4 +3020,19 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } +/************************* Deprecated *************************/ There's a gcc flag for functions if you want to mark this as depracated, otherwise I'd just go with a "XXX: This temporary function will disappear at the end of the patch series" somewhere.
drivers/gpu/drm/i915/i915_drv.h wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) · line 3028
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +5
-0+static inline int __must_check i915_gem_check_ols (struct intel_engine_cs *ring,u32 seqno) +{ +int ret; + +BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex));
WARN_ONinstead ofBUG_ONfor locking checks.Quoted 20 lines
drivers/gpu/drm/i915/i915_drv.h +9
-0+ + ret =0 ; +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) + ret =i915_add_request (ring,NULL ); + + return ret; +} +/************************* Deprecated *************************/ +#endif drivers/gpu/drm/i915/i915_gem.c +0
-2@@ -1097,19 +1097,18 @@ i915_gem_check_wedge(struct i915_gpu_error *error, }/* -* Compare seqno against outstanding lazy request. Emit a request if they are -* equal. drivers/gpu/drm/i915/i915_gem.c i915_gem_check_wedge(struct i915_gpu_error *error, · line 1107
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +3
-2+* Compare arbitrary request against outstanding lazy request. Emit on match. */ int -i915_gem_check_olr (struct intel_engine_cs *ring,u32 seqno) +i915_gem_check_olr (struct drm_i915_gem_request *req) {int ret; -BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); +BUG_ON (!mutex_is_locked (&req->ring->dev->struct_mutex));Same here. Plus please a stern warning in the commit message and cc: to whomever who managed to sneak this in (and failed to catch it in review).
Quoted 81 lines
drivers/gpu/drm/i915/i915_gem.c +7
-11ret =0 ; -if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) - ret =i915_add_request (ring,NULL ); +if (req == req->ring->outstanding_lazy_request) + ret =i915_add_request (req->ring,NULL );return ret; }@@ -1271,7 +1270,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno)if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_ols (ring, seqno);if (ret)return ret;@@ -1338,7 +1337,6 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,struct drm_i915_private *dev_priv = dev->dev_private;struct intel_engine_cs *ring = obj->ring;unsigned reset_counter; -u32 seqno;int ret;BUG_ON (!mutex_is_locked (&dev->struct_mutex));@@ -1348,21 +1346,18 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,if (!req)return 0 ; - seqno =i915_gem_request_get_seqno (req); -BUG_ON (seqno ==0 ); - ret =i915_gem_check_wedge (&dev_priv->gpu_error, true);if (ret)return ret; - ret =i915_gem_check_olr (ring, seqno); + ret =i915_gem_check_olr (req);if (ret)return ret; reset_counter =atomic_read (&dev_priv->gpu_error.reset_counter);i915_gem_request_reference (req);mutex_unlock (&dev->struct_mutex); - ret =__wait_seqno (ring, seqno, reset_counter, true,NULL , file_priv); + ret =__wait_seqno (ring,i915_gem_request_get_seqno (req), reset_counter, true,NULL , file_priv);mutex_lock (&dev->struct_mutex);i915_gem_request_unreference (req);if (ret)@@ -2786,7 +2781,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)int ret;if (obj->active) { - ret =i915_gem_check_olr (obj->ring,i915_gem_request_get_seqno (obj->last_read_req)); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;@@ -2913,7 +2908,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ return 0 ; - ret =i915_gem_check_olr (obj->ring, seqno); + ret =i915_gem_check_olr (obj->last_read_req);if (ret)return ret;drivers/gpu/drm/i915/intel_display.c +1
-1@@ -9775,7 +9775,7 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj) i915_gem_request_get_seqno(obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req)); + ret =i915_gem_check_olr (obj->last_write_req);if (ret)return ret;
Quoted 7 lines
Quoted 7 lines
Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 09/21] drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno
Quoted 10 lines · John Harrison
Quoted 8 lines · Daniel Vetter
Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:13PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
To thin commit message.
Also I wonder whethere we should track the olr state more explicitly in the request structure instead of jumping through all these hoops. And explicit olr state for a request might also help to clarify control flow and checks in a bunch of places.
Quoted 9 lines
What would be nicer would be to get rid of the OLR completely. If code wants to write to the ring then it should request a request structure at the start, pass that request in to all the ring writing code and submit the same request structure at the end. The whole idea of some random request floating around in the ether scooping up any ring accesses that happen to occur with no idea about ownership, sequence, etc. just seems highly dodgy. Indeed it definitely causes issues with the scheduler with things like page flips being mixed up in completely unrelated batch buffer submissions! I believe that is one of the things Chris did in his re-write the universe patch.
Oh OLR is highly dodgy ;-)
It is unfortunately a bit of work to fix things up, especially without a rewrite-the-universe patch. Perhaps:
1. Explicitly preallocate the request, still store it at the same place.
2. Pass the explicit request to add_request, add any missing add_requests call where we currently relied on the olr to scoop up register writes.
3. Shovel the request pointer from the engine to the ring (since that's where we need it it for execlist), that means we need to allocate the request for a specific engine. Steps 1&3 might be better if switched, dunno.
4. Sprinkle asserts over all ring functions to make sure the request is there already.
5. Throw away olr logic.
6. Optional, refactoring over the next few years: Use requests as the primary cmd submission object everywhere.
Comments?
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
RFC 8/25 drm/i915: Remove 'outstanding_lazy_seqno'
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 13 +++---- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 26 +++++--------- drivers/gpu/drm/i915/intel_ringbuffer.c | 52 +++++++++++++++------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 ++----- 6 files changed, 46 insertions(+), 62 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +5 -8
@@ -1108,7 +1108,7 @@ i915_gem_check_olr(struct intel_engine_cs *ring, u32 seqno)BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); ret =0 ; -if (seqno == ring->outstanding_lazy_seqno) +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) ret =i915_add_request (ring,NULL );return ret;
@@ -2344,7 +2344,7 @@ int __i915_add_request(struct intel_engine_cs *ring,u32 request_ring_position, request_start;int ret; - request = ring->preallocated_lazy_request; + request = ring->outstanding_lazy_request;if (WARN_ON (request ==NULL ))return -ENOMEM ;
@@ -2391,7 +2391,6 @@ int __i915_add_request(struct intel_engine_cs *ring,i915_gem_request_reference (request); - request->seqno =intel_ring_get_seqno (ring); request->ring = ring; request->head = request_start; request->tail = request_ring_position;
@@ -2428,8 +2427,7 @@ int __i915_add_request(struct intel_engine_cs *ring, }trace_i915_gem_request_add (ring, request->seqno); - ring->outstanding_lazy_seqno =0 ; - ring->preallocated_lazy_request =NULL ; + ring->outstanding_lazy_request =NULL ;if (!dev_priv->ums.mm_suspended) {i915_queue_hangcheck (ring->dev);
@@ -2605,9 +2603,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, }/* 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 ; +kfree (ring->outstanding_lazy_request); + ring->outstanding_lazy_request =NULL ; }void i915_gem_restore_fences (struct drm_device *dev)
drivers/gpu/drm/i915/i915_gem_execbuffer.c +1 -1
@@ -1170,7 +1170,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,return ret; } -trace_i915_gem_ring_dispatch (ring,intel_ring_get_seqno (ring), flags); +trace_i915_gem_ring_dispatch (ring,i915_gem_request_get_seqno (intel_ring_get_request (ring)), flags);i915_gem_execbuffer_move_to_active (vmas, ring);i915_gem_execbuffer_retire_commands (dev, file, ring, batch_obj);
drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -10053,7 +10053,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =intel_ring_get_seqno (ring); + work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); work->flip_queued_ring = ring; }
drivers/gpu/drm/i915/intel_lrc.c +9 -17
@@ -793,22 +793,14 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf)execlists_context_queue (ring, ctx, ringbuf->tail); } -static int logical_ring_alloc_seqno (struct intel_engine_cs *ring, -struct intel_context *ctx) +static int logical_ring_alloc_request (struct intel_engine_cs *ring, +struct intel_context *ctx) {struct drm_i915_gem_request *request;int ret; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */ - -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )
@@ -816,7 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret;
@@ -829,7 +821,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, request->ctx = ctx;i915_gem_context_reference (request->ctx); - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }
@@ -997,7 +989,7 @@ int intel_logical_ring_begin(struct intel_ringbuffer *ringbuf, int num_dwords)return ret;/* Preallocate the olr before touching the ring */ - ret =logical_ring_alloc_seqno (ring, ringbuf->FIXME_lrc_ctx); + ret =logical_ring_alloc_request (ring, ringbuf->FIXME_lrc_ctx);if (ret)return ret;
@@ -1212,7 +1204,8 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf) (ring->status_page.gfx_addr + (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT)));intel_logical_ring_emit (ringbuf,0 ); -intel_logical_ring_emit (ringbuf, ring->outstanding_lazy_seqno); +intel_logical_ring_emit (ringbuf, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_logical_ring_emit (ringbuf,MI_USER_INTERRUPT );intel_logical_ring_emit (ringbuf,MI_NOOP );intel_logical_ring_advance_and_submit (ringbuf);
@@ -1235,8 +1228,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *ring)intel_logical_ring_stop (ring);WARN_ON ((I915_READ_MODE (ring) &MODE_IDLE ) ==0 ); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);
drivers/gpu/drm/i915/intel_ringbuffer.c +27 -25
161 lines not on this page · raw message
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -10
@@ -265,8 +265,7 @@ struct intel_engine_cs {/** * Do we have some not yet emitted requests outstanding? */ -struct drm_i915_gem_request *preallocated_lazy_request; -u32 outstanding_lazy_seqno; +struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty;bool fbc_dirty;
@@ -429,17 +428,11 @@ static inline u32 intel_ring_get_tail(struct intel_ringbuffer *ringbuf)return ringbuf->tail; } -static inline u32 intel_ring_get_seqno (struct intel_engine_cs *ring) -{ -BUG_ON (ring->outstanding_lazy_seqno ==0 ); -return ring->outstanding_lazy_seqno; -} -static inline struct drm_i915_gem_request *intel_ring_get_request (struct intel_engine_cs *ring) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); -return ring->preallocated_lazy_request; +BUG_ON (ring->outstanding_lazy_request ==NULL ); +return ring->outstanding_lazy_request; }static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno)
Signature · 1 line
1.7.9.5
Re: [RFC 08/25] drm/i915: Remove 'outstanding_lazy_seqno'
Quoted 4 lines · On Fri, Oct 10, 2014 at 12:38:22PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
This looks like a v2 of the previous patch 08, but the commit message fails to mention what exactly changed and why.
-Daniel
Quoted 383 lines
--- drivers/gpu/drm/i915/i915_gem.c | 13 +++---- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 26 +++++--------- drivers/gpu/drm/i915/intel_ringbuffer.c | 52 +++++++++++++++------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 ++----- 6 files changed, 46 insertions(+), 62 deletions(-)drivers/gpu/drm/i915/i915_gem.c +5
-8@@ -1108,7 +1108,7 @@ i915_gem_check_olr(struct intel_engine_cs *ring, u32 seqno)BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); ret =0 ; -if (seqno == ring->outstanding_lazy_seqno) +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) ret =i915_add_request (ring,NULL );return ret;@@ -2344,7 +2344,7 @@ int __i915_add_request(struct intel_engine_cs *ring,u32 request_ring_position, request_start;int ret; - request = ring->preallocated_lazy_request; + request = ring->outstanding_lazy_request;if (WARN_ON (request ==NULL ))return -ENOMEM ;@@ -2391,7 +2391,6 @@ int __i915_add_request(struct intel_engine_cs *ring,i915_gem_request_reference (request); - request->seqno =intel_ring_get_seqno (ring); request->ring = ring; request->head = request_start; request->tail = request_ring_position;@@ -2428,8 +2427,7 @@ int __i915_add_request(struct intel_engine_cs *ring, }trace_i915_gem_request_add (ring, request->seqno); - ring->outstanding_lazy_seqno =0 ; - ring->preallocated_lazy_request =NULL ; + ring->outstanding_lazy_request =NULL ;if (!dev_priv->ums.mm_suspended) {i915_queue_hangcheck (ring->dev);@@ -2605,9 +2603,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, }/* 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 ; +kfree (ring->outstanding_lazy_request); + ring->outstanding_lazy_request =NULL ; }void i915_gem_restore_fences (struct drm_device *dev)drivers/gpu/drm/i915/i915_gem_execbuffer.c +1
-1@@ -1170,7 +1170,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,return ret; } -trace_i915_gem_ring_dispatch (ring,intel_ring_get_seqno (ring), flags); +trace_i915_gem_ring_dispatch (ring,i915_gem_request_get_seqno (intel_ring_get_request (ring)), flags);i915_gem_execbuffer_move_to_active (vmas, ring);i915_gem_execbuffer_retire_commands (dev, file, ring, batch_obj);drivers/gpu/drm/i915/intel_display.c +1
-1@@ -10053,7 +10053,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =intel_ring_get_seqno (ring); + work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); work->flip_queued_ring = ring; }drivers/gpu/drm/i915/intel_lrc.c +9
-17@@ -793,22 +793,14 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf)execlists_context_queue (ring, ctx, ringbuf->tail); } -static int logical_ring_alloc_seqno (struct intel_engine_cs *ring, -struct intel_context *ctx) +static int logical_ring_alloc_request (struct intel_engine_cs *ring, +struct intel_context *ctx) {struct drm_i915_gem_request *request;int ret; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */ - -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )@@ -816,7 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret;@@ -829,7 +821,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, request->ctx = ctx;i915_gem_context_reference (request->ctx); - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }@@ -997,7 +989,7 @@ int intel_logical_ring_begin(struct intel_ringbuffer *ringbuf, int num_dwords)return ret;/* Preallocate the olr before touching the ring */ - ret =logical_ring_alloc_seqno (ring, ringbuf->FIXME_lrc_ctx); + ret =logical_ring_alloc_request (ring, ringbuf->FIXME_lrc_ctx);if (ret)return ret;@@ -1212,7 +1204,8 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf) (ring->status_page.gfx_addr + (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT)));intel_logical_ring_emit (ringbuf,0 ); -intel_logical_ring_emit (ringbuf, ring->outstanding_lazy_seqno); +intel_logical_ring_emit (ringbuf, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_logical_ring_emit (ringbuf,MI_USER_INTERRUPT );intel_logical_ring_emit (ringbuf,MI_NOOP );intel_logical_ring_advance_and_submit (ringbuf);@@ -1235,8 +1228,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *ring)intel_logical_ring_stop (ring);WARN_ON ((I915_READ_MODE (ring) &MODE_IDLE ) ==0 ); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);drivers/gpu/drm/i915/intel_ringbuffer.c +27
-25@@ -895,17 +895,20 @@ static int gen8_rcs_signal(struct intel_engine_cs *signaller,return ret;for_each_ring (waiter, dev_priv, i) { +u32 seqno;u64 gtt_offset = signaller->semaphore.signal_ggtt[i];if (gtt_offset ==MI_SEMAPHORE_SYNC_INVALID )continue ; + seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller,GFX_OP_PIPE_CONTROL (6 ));intel_ring_emit (signaller,PIPE_CONTROL_GLOBAL_GTT_IVB |PIPE_CONTROL_QW_WRITE |PIPE_CONTROL_FLUSH_ENABLE );intel_ring_emit (signaller,lower_32_bits (gtt_offset));intel_ring_emit (signaller,upper_32_bits (gtt_offset)); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno);intel_ring_emit (signaller,0 );intel_ring_emit (signaller,MI_SEMAPHORE_SIGNAL |MI_SEMAPHORE_TARGET (waiter->id));@@ -933,16 +936,19 @@ static int gen8_xcs_signal(struct intel_engine_cs *signaller,return ret;for_each_ring (waiter, dev_priv, i) { +u32 seqno;u64 gtt_offset = signaller->semaphore.signal_ggtt[i];if (gtt_offset ==MI_SEMAPHORE_SYNC_INVALID )continue ; + seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller, (MI_FLUSH_DW +1 ) |MI_FLUSH_DW_OP_STOREDW );intel_ring_emit (signaller,lower_32_bits (gtt_offset) |MI_FLUSH_DW_USE_GTT );intel_ring_emit (signaller,upper_32_bits (gtt_offset)); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno);intel_ring_emit (signaller,MI_SEMAPHORE_SIGNAL |MI_SEMAPHORE_TARGET (waiter->id));intel_ring_emit (signaller,0 );@@ -971,9 +977,11 @@ static int gen6_signal(struct intel_engine_cs *signaller,for_each_ring (useless, dev_priv, i) {u32 mbox_reg = signaller->semaphore.mbox.signal[i];if (mbox_reg !=GEN6_NOSYNC ) { +u32 seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller,MI_LOAD_REGISTER_IMM (1 ));intel_ring_emit (signaller, mbox_reg); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno); } }@@ -1008,7 +1016,8 @@ gen6_add_request(struct intel_engine_cs *ring)intel_ring_emit (ring,MI_STORE_DWORD_INDEX );intel_ring_emit (ring,I915_GEM_HWS_INDEX <<MI_STORE_DWORD_INDEX_SHIFT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,MI_USER_INTERRUPT );__intel_ring_advance (ring);@@ -1126,7 +1135,8 @@ pc_render_add_request(struct intel_engine_cs *ring)PIPE_CONTROL_WRITE_FLUSH | PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);intel_ring_emit (ring, ring->scratch.gtt_offset |PIPE_CONTROL_GLOBAL_GTT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,0 );PIPE_CONTROL_FLUSH (ring, scratch_addr); scratch_addr +=2 *CACHELINE_BYTES ;/* write to separate cachelines */ @@ -1145,7 +1155,8 @@ pc_render_add_request(struct intel_engine_cs *ring)PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE | PIPE_CONTROL_NOTIFY);intel_ring_emit (ring, ring->scratch.gtt_offset |PIPE_CONTROL_GLOBAL_GTT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,0 );__intel_ring_advance (ring);@@ -1385,7 +1396,8 @@ i9xx_add_request(struct intel_engine_cs *ring)intel_ring_emit (ring,MI_STORE_DWORD_INDEX );intel_ring_emit (ring,I915_GEM_HWS_INDEX <<MI_STORE_DWORD_INDEX_SHIFT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,MI_USER_INTERRUPT );__intel_ring_advance (ring);@@ -1839,8 +1851,7 @@ void intel_cleanup_ring_buffer(struct intel_engine_cs *ring)WARN_ON (!IS_GEN2 (ring->dev) && (I915_READ_MODE (ring) &MODE_IDLE ) ==0 );intel_destroy_ringbuffer_obj (ringbuf); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);@@ -1980,7 +1991,7 @@ int intel_ring_idle(struct intel_engine_cs *ring)int ret;/* We need to add any requests required to flush the objects and ring */ -if (ring->outstanding_lazy_seqno) { +if (ring->outstanding_lazy_request) { ret =i915_add_request (ring,NULL );if (ret)return ret;@@ -1998,22 +2009,13 @@ int intel_ring_idle(struct intel_engine_cs *ring) }static int -intel_ring_alloc_seqno (struct intel_engine_cs *ring) +intel_ring_alloc_request (struct intel_engine_cs *ring) {int ret;struct drm_i915_gem_request *request; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */ - -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } - -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )@@ -2021,13 +2023,13 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret; } - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }@@ -2068,7 +2070,7 @@ int intel_ring_begin(struct intel_engine_cs *ring,return ret;/* Preallocate the olr before touching the ring */ - ret =intel_ring_alloc_seqno (ring); + ret =intel_ring_alloc_request (ring);if (ret)return ret;@@ -2103,7 +2105,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)struct drm_device *dev = ring->dev;struct drm_i915_private *dev_priv = dev->dev_private; -BUG_ON (ring->outstanding_lazy_seqno); +BUG_ON (ring->outstanding_lazy_request);if (INTEL_INFO (dev)->gen ==6 ||INTEL_INFO (dev)->gen ==7 ) {I915_WRITE (RING_SYNC_0 (ring->mmio_base),0 );drivers/gpu/drm/i915/intel_ringbuffer.h +3
-10@@ -265,8 +265,7 @@ struct intel_engine_cs {/** * Do we have some not yet emitted requests outstanding? */ -struct drm_i915_gem_request *preallocated_lazy_request; -u32 outstanding_lazy_seqno; +struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty;bool fbc_dirty;@@ -429,17 +428,11 @@ static inline u32 intel_ring_get_tail(struct intel_ringbuffer *ringbuf)return ringbuf->tail; } -static inline u32 intel_ring_get_seqno (struct intel_engine_cs *ring) -{ -BUG_ON (ring->outstanding_lazy_seqno ==0 ); -return ring->outstanding_lazy_seqno; -} -static inline struct drm_i915_gem_request *intel_ring_get_request (struct intel_engine_cs *ring) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); -return ring->preallocated_lazy_request; +BUG_ON (ring->outstanding_lazy_request ==NULL ); +return ring->outstanding_lazy_request; }static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno)Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 08/21] drm/i915: Remove 'outstanding_lazy_seqno'
In reply to [email protected]Quoted 4 lines · On Mon, Oct 06, 2014 at 03:15:12PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
given the scary commit message for patch 3 I'm confused that we can just do this. Or maybe we should move patch 3 right before this one here and instead of mentioning fun in combination with the scheduler as the justification use this patch here. Since obviously if we dig out the ols from the plr then they obviously need to be allocated together.
In any case this patch here should at least mention why this is suddenly possible without fuzz. A few comments below.
-Daniel
Quoted 50 lines
--- drivers/gpu/drm/i915/i915_gem.c | 13 +++----- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 20 ++++-------- drivers/gpu/drm/i915/intel_ringbuffer.c | 48 +++++++++++++++------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 ++------ 6 files changed, 41 insertions(+), 57 deletions(-)drivers/gpu/drm/i915/i915_gem.c +3
-5@@ -1108,7 +1108,7 @@ i915_gem_check_olr(struct intel_engine_cs *ring, u32 seqno)BUG_ON (!mutex_is_locked (&ring->dev->struct_mutex)); ret =0 ; -if (seqno == ring->outstanding_lazy_seqno) +if (seqno ==i915_gem_request_get_seqno (ring->outstanding_lazy_request)) ret =i915_add_request (ring,NULL );return ret;@@ -2344,7 +2344,7 @@ int __i915_add_request(struct intel_engine_cs *ring,u32 request_ring_position, request_start;int ret; - request = ring->preallocated_lazy_request; + request = ring->outstanding_lazy_request;if (WARN_ON (request ==NULL ))return -ENOMEM ;@@ -2391,7 +2391,6 @@ int __i915_add_request(struct intel_engine_cs *ring,i915_gem_request_reference (request); - request->seqno =intel_ring_get_seqno (ring); request->ring = ring; request->head = request_start; request->tail = request_ring_position;@@ -2428,8 +2427,7 @@ int __i915_add_request(struct intel_engine_cs *ring, }trace_i915_gem_request_add (ring, request->seqno); - ring->outstanding_lazy_seqno =0 ; - ring->preallocated_lazy_request =NULL ; + ring->outstanding_lazy_request =NULL ;if (!dev_priv->ums.mm_suspended) {i915_queue_hangcheck (ring->dev);
drivers/gpu/drm/i915/i915_gem.c static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, · line 2607
Quoted 8 lines
drivers/gpu/drm/i915/i915_gem.c +2
-3@@ -2605,9 +2603,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, }/* 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 ; +kfree (ring->outstanding_lazy_request); + ring->outstanding_lazy_request =NULL ;
Somewhat unrelated, but shouldn't we move to refcount even the ring->plr? If we allocate it then the ring->plr pointer would hold the ref, then it gets transferred to the request list.
This isn't too terribly important yet, but sometimes (when we have per-engine reset) we need to fix up all the request access the hangcheck and error state capture code does. Which probably means we really need to refcount requests super-carefully everywhere (because the hangcheck can happen almost anywhere at all).
Quoted 34 lines
drivers/gpu/drm/i915/i915_gem.c +0
-0}void i915_gem_restore_fences (struct drm_device *dev)drivers/gpu/drm/i915/i915_gem_execbuffer.c +1
-1@@ -1170,7 +1170,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,return ret; } -trace_i915_gem_ring_dispatch (ring,intel_ring_get_seqno (ring), flags); +trace_i915_gem_ring_dispatch (ring,i915_gem_request_get_seqno (intel_ring_get_request (ring)), flags);i915_gem_execbuffer_move_to_active (vmas, ring);i915_gem_execbuffer_retire_commands (dev, file, ring, batch_obj);drivers/gpu/drm/i915/intel_display.c +1
-1@@ -10053,7 +10053,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno =intel_ring_get_seqno (ring); + work->flip_queued_seqno =i915_gem_request_get_seqno (intel_ring_get_request (ring)); work->flip_queued_ring = ring; }drivers/gpu/drm/i915/intel_lrc.c +0
-0
drivers/gpu/drm/i915/intel_lrc.c static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, · line 805
Quoted 7 lines
drivers/gpu/drm/i915/intel_lrc.c +0
-4@@ -799,16 +799,8 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,struct drm_i915_gem_request *request;int ret; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */
If you add a comment only to remove it later on in the series I usually smash an "XXX: This will get cleaned up further" or similar on top of it. it aids review if you can gauge this whitout peeking ahead.
Quoted 252 lines
drivers/gpu/drm/i915/intel_lrc.c +6
-10- -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )@@ -816,7 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret;@@ -829,7 +821,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring, request->ctx = ctx;i915_gem_context_reference (request->ctx); - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }@@ -1212,7 +1204,8 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf) (ring->status_page.gfx_addr + (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT)));intel_logical_ring_emit (ringbuf,0 ); -intel_logical_ring_emit (ringbuf, ring->outstanding_lazy_seqno); +intel_logical_ring_emit (ringbuf, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_logical_ring_emit (ringbuf,MI_USER_INTERRUPT );intel_logical_ring_emit (ringbuf,MI_NOOP );intel_logical_ring_advance_and_submit (ringbuf);@@ -1235,8 +1228,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *ring)intel_logical_ring_stop (ring);WARN_ON ((I915_READ_MODE (ring) &MODE_IDLE ) ==0 ); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);drivers/gpu/drm/i915/intel_ringbuffer.c +25
-23@@ -895,17 +895,20 @@ static int gen8_rcs_signal(struct intel_engine_cs *signaller,return ret;for_each_ring (waiter, dev_priv, i) { +u32 seqno;u64 gtt_offset = signaller->semaphore.signal_ggtt[i];if (gtt_offset ==MI_SEMAPHORE_SYNC_INVALID )continue ; + seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller,GFX_OP_PIPE_CONTROL (6 ));intel_ring_emit (signaller,PIPE_CONTROL_GLOBAL_GTT_IVB |PIPE_CONTROL_QW_WRITE |PIPE_CONTROL_FLUSH_ENABLE );intel_ring_emit (signaller,lower_32_bits (gtt_offset));intel_ring_emit (signaller,upper_32_bits (gtt_offset)); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno);intel_ring_emit (signaller,0 );intel_ring_emit (signaller,MI_SEMAPHORE_SIGNAL |MI_SEMAPHORE_TARGET (waiter->id));@@ -933,16 +936,19 @@ static int gen8_xcs_signal(struct intel_engine_cs *signaller,return ret;for_each_ring (waiter, dev_priv, i) { +u32 seqno;u64 gtt_offset = signaller->semaphore.signal_ggtt[i];if (gtt_offset ==MI_SEMAPHORE_SYNC_INVALID )continue ; + seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller, (MI_FLUSH_DW +1 ) |MI_FLUSH_DW_OP_STOREDW );intel_ring_emit (signaller,lower_32_bits (gtt_offset) |MI_FLUSH_DW_USE_GTT );intel_ring_emit (signaller,upper_32_bits (gtt_offset)); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno);intel_ring_emit (signaller,MI_SEMAPHORE_SIGNAL |MI_SEMAPHORE_TARGET (waiter->id));intel_ring_emit (signaller,0 );@@ -971,9 +977,11 @@ static int gen6_signal(struct intel_engine_cs *signaller,for_each_ring (useless, dev_priv, i) {u32 mbox_reg = signaller->semaphore.mbox.signal[i];if (mbox_reg !=GEN6_NOSYNC ) { +u32 seqno =i915_gem_request_get_seqno ( + signaller->outstanding_lazy_request);intel_ring_emit (signaller,MI_LOAD_REGISTER_IMM (1 ));intel_ring_emit (signaller, mbox_reg); -intel_ring_emit (signaller, signaller->outstanding_lazy_seqno); +intel_ring_emit (signaller, seqno); } }@@ -1008,7 +1016,8 @@ gen6_add_request(struct intel_engine_cs *ring)intel_ring_emit (ring,MI_STORE_DWORD_INDEX );intel_ring_emit (ring,I915_GEM_HWS_INDEX <<MI_STORE_DWORD_INDEX_SHIFT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,MI_USER_INTERRUPT );__intel_ring_advance (ring);@@ -1126,7 +1135,8 @@ pc_render_add_request(struct intel_engine_cs *ring)PIPE_CONTROL_WRITE_FLUSH | PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);intel_ring_emit (ring, ring->scratch.gtt_offset |PIPE_CONTROL_GLOBAL_GTT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,0 );PIPE_CONTROL_FLUSH (ring, scratch_addr); scratch_addr +=2 *CACHELINE_BYTES ;/* write to separate cachelines */ @@ -1145,7 +1155,8 @@ pc_render_add_request(struct intel_engine_cs *ring)PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE | PIPE_CONTROL_NOTIFY);intel_ring_emit (ring, ring->scratch.gtt_offset |PIPE_CONTROL_GLOBAL_GTT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,0 );__intel_ring_advance (ring);@@ -1385,7 +1396,8 @@ i9xx_add_request(struct intel_engine_cs *ring)intel_ring_emit (ring,MI_STORE_DWORD_INDEX );intel_ring_emit (ring,I915_GEM_HWS_INDEX <<MI_STORE_DWORD_INDEX_SHIFT ); -intel_ring_emit (ring, ring->outstanding_lazy_seqno); +intel_ring_emit (ring, +i915_gem_request_get_seqno (ring->outstanding_lazy_request));intel_ring_emit (ring,MI_USER_INTERRUPT );__intel_ring_advance (ring);@@ -1839,8 +1851,7 @@ void intel_cleanup_ring_buffer(struct intel_engine_cs *ring)WARN_ON (!IS_GEN2 (ring->dev) && (I915_READ_MODE (ring) &MODE_IDLE ) ==0 );intel_destroy_ringbuffer_obj (ringbuf); - ring->preallocated_lazy_request =NULL ; - ring->outstanding_lazy_seqno =0 ; + ring->outstanding_lazy_request =NULL ;if (ring->cleanup) ring->cleanup (ring);@@ -1980,7 +1991,7 @@ int intel_ring_idle(struct intel_engine_cs *ring)int ret;/* We need to add any requests required to flush the objects and ring */ -if (ring->outstanding_lazy_seqno) { +if (ring->outstanding_lazy_request) { ret =i915_add_request (ring,NULL );if (ret)return ret;@@ -2003,17 +2014,8 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)int ret;struct drm_i915_gem_request *request; -/* The aim is to replace seqno values with request structures. A step -* along the way is to switch to using the PLR in preference to the -* OLS. That requires the PLR to only be valid when the OLS is also -* valid. I.e., the two must be kept in step. */ - -if (ring->outstanding_lazy_seqno) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); +if (ring->outstanding_lazy_request)return 0 ; - } - -BUG_ON (ring->preallocated_lazy_request !=NULL ); request =kmalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )@@ -2021,13 +2023,13 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)kref_init (&request->ref); - ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {kfree (request);return ret; } - ring->preallocated_lazy_request = request; + ring->outstanding_lazy_request = request;return 0 ; }@@ -2103,7 +2105,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)struct drm_device *dev = ring->dev;struct drm_i915_private *dev_priv = dev->dev_private; -BUG_ON (ring->outstanding_lazy_seqno); +BUG_ON (ring->outstanding_lazy_request);if (INTEL_INFO (dev)->gen ==6 ||INTEL_INFO (dev)->gen ==7 ) {I915_WRITE (RING_SYNC_0 (ring->mmio_base),0 );drivers/gpu/drm/i915/intel_ringbuffer.h +3
-10@@ -265,8 +265,7 @@ struct intel_engine_cs {/** * Do we have some not yet emitted requests outstanding? */ -struct drm_i915_gem_request *preallocated_lazy_request; -u32 outstanding_lazy_seqno; +struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty;bool fbc_dirty;@@ -429,17 +428,11 @@ static inline u32 intel_ring_get_tail(struct intel_ringbuffer *ringbuf)return ringbuf->tail; } -static inline u32 intel_ring_get_seqno (struct intel_engine_cs *ring) -{ -BUG_ON (ring->outstanding_lazy_seqno ==0 ); -return ring->outstanding_lazy_seqno; -} -static inline struct drm_i915_gem_request *intel_ring_get_request (struct intel_engine_cs *ring) { -BUG_ON (ring->preallocated_lazy_request ==NULL ); -return ring->preallocated_lazy_request; +BUG_ON (ring->outstanding_lazy_request ==NULL ); +return ring->outstanding_lazy_request; }static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno)Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 08/21] drm/i915: Remove 'outstanding_lazy_seqno'
Quoted 13 lines · Daniel Vetter
drivers/gpu/drm/i915/i915_gem.c static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, · line 2607
Quoted 9 lines
drivers/gpu/drm/i915/i915_gem.c +2
-3@@ -2605,9 +2603,8 @@ static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv, }/* 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 ; +kfree (ring->outstanding_lazy_request); + ring->outstanding_lazy_request =NULL ;Somewhat unrelated, but shouldn't we move to refcount even the
ring->plr? If we allocate it then thering->plrpointer would hold the ref, then it gets transferred to the request list.
Actually this is 100% related to this patch, since it replaces the ol seqno with the ol request. Would be good to clarify this in the commit message, since this patch doesn't remove the logic around olr, but really just does the switch from seqno to request.
-Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 06/21] drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
In reply to [email protected]Quoted 77 lines · On Mon, Oct 06, 2014 at 03:15:10PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_debugfs.c | 6 +-- drivers/gpu/drm/i915/i915_drv.h | 6 +-- drivers/gpu/drm/i915/i915_gem.c | 66 ++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_execbuffer.c | 6 +-- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 4 +- drivers/gpu/drm/i915/intel_display.c | 10 ++--- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 +- 9 files changed, 58 insertions(+), 48 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +3
-3@@ -133,9 +133,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) obj->base.size /1024 , obj->base.read_domains, obj->base.write_domain, - obj->last_read_seqno, - obj->last_write_seqno, - obj->last_fenced_seqno, +i915_gem_request_get_seqno (obj->last_read_req), +i915_gem_request_get_seqno (obj->last_write_req), +i915_gem_request_get_seqno (obj->last_fenced_req),i915_cache_level_str (to_i915 (obj->base.dev), obj->cache_level), obj->dirty ?" dirty" :"" , obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");drivers/gpu/drm/i915/i915_drv.h +3
-3@@ -1862,10 +1862,10 @@ struct drm_i915_gem_object {struct intel_engine_cs *ring;/** Breadcrumb of last rendering to the buffer. */ -uint32_t last_read_seqno; -uint32_t last_write_seqno; +struct drm_i915_gem_request *last_read_req; +struct drm_i915_gem_request *last_write_req;/** Breadcrumb of last fenced GPU access to the buffer. */ -uint32_t last_fenced_seqno; +struct drm_i915_gem_request *last_fenced_req;/** Current tiling stride for the object, if it's tiled. */ uint32_t stride;drivers/gpu/drm/i915/i915_gem.c +4
-3@@ -1289,11 +1289,11 @@ i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj)/* Manually manage the write flush as we may have not yet * retired the buffer. * -* Note that the last_write_seqno is always the earlier of -* the two (read/write) seqno, so if we haved successfully waited, +* Note that the last_write_req is always the earlier of +* the two (read/write) requests, so if we haved successfully waited, * we know we have passed the last write. */ - obj->last_write_seqno =0 ; + obj->last_write_req =NULL ;return 0 ; }@@ -1306,14 +1306,18 @@ static __must_check int i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, bool readonly) { +struct drm_i915_gem_request *req;struct intel_engine_cs *ring = obj->ring;u32 seqno;
drivers/gpu/drm/i915/i915_gem.c static __must_check int · line 1319
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +4
-2int ret; - seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno; -if (seqno ==0 ) + req = readonly ? obj->last_write_req : obj->last_read_req; +if (!req)return 0 ; + seqno =i915_gem_request_get_seqno (req); +BUG_ON (seqno ==0 );
Again, you like BUG_ON a bit too much for my taste. If you want these checks imo a WARN_ON in i915_gem_request_get_seqno (iff req != NULL ofc) in the previous patch would be much better.
-Daniel
Quoted 298 lines
drivers/gpu/drm/i915/i915_gem.c +30
-23+ ret =i915_wait_seqno (ring, seqno);if (ret)return ret;@@ -1329,6 +1333,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, struct drm_i915_file_private *file_priv, bool readonly) { +struct drm_i915_gem_request *req;struct drm_device *dev = obj->base.dev;struct drm_i915_private *dev_priv = dev->dev_private;struct intel_engine_cs *ring = obj->ring;@@ -1339,10 +1344,13 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,BUG_ON (!mutex_is_locked (&dev->struct_mutex));BUG_ON (!dev_priv->mm.interruptible); - seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno; -if (seqno ==0 ) + req = readonly ? obj->last_write_req : obj->last_read_req; +if (!req)return 0 ; + seqno =i915_gem_request_get_seqno (req); +BUG_ON (seqno ==0 ); + ret =i915_gem_check_wedge (&dev_priv->gpu_error, true);if (ret)return ret;@@ -2180,12 +2188,12 @@ static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring) { - u32 seqno = intel_ring_get_seqno(ring); + struct drm_i915_gem_request *req = intel_ring_get_request(ring);BUG_ON (ring ==NULL ); -if (obj->ring != ring && obj->last_write_seqno) { -/* Keep the seqno relative to the current ring */ - obj->last_write_seqno = seqno; +if (obj->ring != ring && obj->last_write_req) { +/* Keep the request relative to the current ring */ + obj->last_write_req = req; } obj->ring = ring;@@ -2197,7 +2205,7 @@ i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,list_move_tail (&obj->ring_list, &ring->active_list); - obj->last_read_seqno = seqno; + obj->last_read_req = req; } void i915_vma_move_to_active(struct i915_vma *vma,@@ -2228,11 +2236,11 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)list_del_init (&obj->ring_list); obj->ring =NULL ; - obj->last_read_seqno =0 ; - obj->last_write_seqno =0 ; + obj->last_read_req =NULL ; + obj->last_write_req =NULL ; obj->base.write_domain =0 ; - obj->last_fenced_seqno =0 ; + obj->last_fenced_req =NULL ; obj->active =0 ;drm_gem_object_unreference (&obj->base);@@ -2249,7 +2257,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)return ;if (i915_seqno_passed (ring->get_seqno (ring, true), - obj->last_read_seqno)) +i915_gem_request_get_seqno (obj->last_read_req)))i915_gem_object_move_to_inactive (obj); }@@ -2669,7 +2677,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno, obj->last_read_seqno)) +if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req)))break ;i915_gem_object_move_to_inactive (obj);@@ -2779,7 +2787,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)int ret;if (obj->active) { - ret =i915_gem_check_olr (obj->ring, obj->last_read_seqno); + ret =i915_gem_check_olr (obj->ring,i915_gem_request_get_seqno (obj->last_read_req));if (ret)return ret;@@ -2838,13 +2846,14 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) goto out;if (obj->active) { - seqno = obj->last_read_seqno; +if (!obj->last_read_req) + goto out; + + seqno =i915_gem_request_get_seqno (obj->last_read_req); +BUG_ON (seqno ==0 ); ring = obj->ring; } -if (seqno ==0 ) - goto out; -/* Do this after OLR check to make sure we make forward progress polling * on this IOCTL with a timeout <=0 (like busy ioctl) */ @@ -2894,7 +2903,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, idx =intel_ring_sync_index (from, to); - seqno = obj->last_read_seqno; + seqno =i915_gem_request_get_seqno (obj->last_read_req);/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ @@ -2907,11 +2916,12 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,trace_i915_gem_ring_sync_to (from, to, seqno); ret = to->semaphore.sync_to (to, from, seqno);if (!ret) -/* We use last_read_seqno because sync_to() +/* We use last_read_req because sync_to() * might have just caused seqno wrap under * the radar. */ - from->semaphore.sync_seqno[idx] = obj->last_read_seqno; + from->semaphore.sync_seqno[idx] = +i915_gem_request_get_seqno (obj->last_read_req);return ret; }@@ -3224,12 +3234,12 @@ static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj,static int i915_gem_object_wait_fence (struct drm_i915_gem_object *obj) { -if (obj->last_fenced_seqno) { -int ret =i915_wait_seqno (obj->ring, obj->last_fenced_seqno); +if (obj->last_fenced_req) { +int ret =i915_wait_seqno (obj->ring,i915_gem_request_get_seqno (obj->last_fenced_req));if (ret)return ret; - obj->last_fenced_seqno =0 ; + obj->last_fenced_req =NULL ; }return 0 ;drivers/gpu/drm/i915/i915_gem_execbuffer.c +3
-3@@ -946,7 +946,7 @@ void i915_gem_execbuffer_move_to_active(struct list_head *vmas, struct intel_engine_cs *ring) { - u32 seqno = intel_ring_get_seqno(ring); + struct drm_i915_gem_request *req = intel_ring_get_request(ring);struct i915_vma *vma;list_for_each_entry (vma, vmas, exec_list) {@@ -963,7 +963,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,i915_vma_move_to_active (vma, ring);if (obj->base.write_domain) { obj->dirty =1 ; - obj->last_write_seqno = seqno; + obj->last_write_req = req;intel_fb_obj_invalidate (obj, ring);@@ -971,7 +971,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas, obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS ; }if (entry->flags &EXEC_OBJECT_NEEDS_FENCE ) { - obj->last_fenced_seqno = seqno; + obj->last_fenced_req = req;if (entry->flags & __EXEC_OBJECT_HAS_FENCE) {struct drm_i915_private *dev_priv =to_i915 (ring->dev); list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,drivers/gpu/drm/i915/i915_gem_gtt.h +2
-2@@ -178,7 +178,7 @@ struct i915_address_space {* List of objects currently involved in rendering. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list. @@ -189,7 +189,7 @@ struct i915_address_space {* LRU list of objects which are not in the ringbuffer and * are ready to unbind, but are still in the GTT. * -* last_rendering_seqno is 0 while an object is in this list. +* last_rendering_req is NULL while an object is in this list. * * A reference is not held on the buffer while on this list, * as merely being GTT-bound shouldn't prevent its being drivers/gpu/drm/i915/i915_gem_tiling.c +1
-1@@ -383,7 +383,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,if (ret ==0 ) { obj->fence_dirty = - obj->last_fenced_seqno || + obj->last_fenced_req || obj->fence_reg !=I915_FENCE_REG_NONE ; obj->tiling_mode = args->tiling_mode;drivers/gpu/drm/i915/i915_gpu_error.c +2
-2@@ -666,8 +666,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->size = obj->base.size; err->name = obj->base.name; - err->rseqno = obj->last_read_seqno; - err->wseqno = obj->last_write_seqno; + err->rseqno =i915_gem_request_get_seqno (obj->last_read_req); + err->wseqno =i915_gem_request_get_seqno (obj->last_write_req); err->gtt_offset = vma->node.start; err->read_domains = obj->base.read_domains; err->write_domain = obj->base.write_domain;drivers/gpu/drm/i915/intel_display.c +5
-5@@ -9766,16 +9766,16 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)lockdep_assert_held (&obj->base.dev->struct_mutex); -if (!obj->last_write_seqno) +if (!obj->last_write_req)return 0 ; ring = obj->ring;if (i915_seqno_passed (ring->get_seqno (ring, true), - obj->last_write_seqno)) +i915_gem_request_get_seqno (obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring, obj->last_write_seqno); + ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req));if (ret)return ret;@@ -9838,7 +9838,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev, }spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); - intel_crtc->mmio_flip.seqno = obj->last_write_seqno; + intel_crtc->mmio_flip.seqno =i915_gem_request_get_seqno (obj->last_write_req); intel_crtc->mmio_flip.ring_id = obj->ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);@@ -10045,7 +10045,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno = obj->last_write_seqno; + work->flip_queued_seqno =i915_gem_request_get_seqno (obj->last_write_req); work->flip_queued_ring = obj->ring; }else { ret = dev_priv->display.queue_flip(dev, crtc, fb, obj, ring,drivers/gpu/drm/i915/intel_ringbuffer.h +1
-1@@ -249,7 +249,7 @@ struct intel_engine_cs {* ringbuffer. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list. Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 06/21] drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
Quoted 86 lines · Daniel Vetter
Quoted 85 lines · On Mon, Oct 06, 2014 at 03:15:10PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_debugfs.c | 6 +-- drivers/gpu/drm/i915/i915_drv.h | 6 +-- drivers/gpu/drm/i915/i915_gem.c | 66 ++++++++++++++++------------ drivers/gpu/drm/i915/i915_gem_execbuffer.c | 6 +-- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 4 +- drivers/gpu/drm/i915/intel_display.c | 10 ++--- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 +- 9 files changed, 58 insertions(+), 48 deletions(-)drivers/gpu/drm/i915/i915_debugfs.c +3
-3@@ -133,9 +133,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj) obj->base.size /1024 , obj->base.read_domains, obj->base.write_domain, - obj->last_read_seqno, - obj->last_write_seqno, - obj->last_fenced_seqno, +i915_gem_request_get_seqno (obj->last_read_req), +i915_gem_request_get_seqno (obj->last_write_req), +i915_gem_request_get_seqno (obj->last_fenced_req),i915_cache_level_str (to_i915 (obj->base.dev), obj->cache_level), obj->dirty ?" dirty" :"" , obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");drivers/gpu/drm/i915/i915_drv.h +3
-3@@ -1862,10 +1862,10 @@ struct drm_i915_gem_object {struct intel_engine_cs *ring;/** Breadcrumb of last rendering to the buffer. */ -uint32_t last_read_seqno; -uint32_t last_write_seqno; +struct drm_i915_gem_request *last_read_req; +struct drm_i915_gem_request *last_write_req;/** Breadcrumb of last fenced GPU access to the buffer. */ -uint32_t last_fenced_seqno; +struct drm_i915_gem_request *last_fenced_req;/** Current tiling stride for the object, if it's tiled. */ uint32_t stride;drivers/gpu/drm/i915/i915_gem.c +6
-5@@ -1289,11 +1289,11 @@ i915_gem_object_wait_rendering__tail(struct drm_i915_gem_object *obj)/* Manually manage the write flush as we may have not yet * retired the buffer. * -* Note that the last_write_seqno is always the earlier of -* the two (read/write) seqno, so if we haved successfully waited, +* Note that the last_write_req is always the earlier of +* the two (read/write) requests, so if we haved successfully waited, * we know we have passed the last write. */ - obj->last_write_seqno =0 ; + obj->last_write_req =NULL ;return 0 ; }@@ -1306,14 +1306,18 @@ static __must_check int i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj, bool readonly) { +struct drm_i915_gem_request *req;struct intel_engine_cs *ring = obj->ring;u32 seqno;int ret; - seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno; -if (seqno ==0 ) + req = readonly ? obj->last_write_req : obj->last_read_req; +if (!req)return 0 ;
Quoted 6 lines
drivers/gpu/drm/i915/i915_gem.c static __must_check int · line 1319
Quoted 2 lines
drivers/gpu/drm/i915/i915_gem.c +2
-0+ seqno =i915_gem_request_get_seqno (req); +BUG_ON (seqno ==0 );Again, you like
BUG_ONa bit too much for my taste. If you want these checks imo aWARN_ONini915_gem_request_get_seqno(iff req != NULL ofc) in the previous patch would be much better.
-Daniel
Again a) this patch set was posted as a work in progress, I would like some quick feedback before attempting to finish it, polish it, flush out the corner cases, etc. And for internal development debug, BUG_ONs are a lot more use than WARN_ON. And b) this will disappear in a later patch when the seqno is removed completely and 'i915_wait_seqno(seqno)' becomes 'i915_wait_request(request)'.
Quoted 298 lines
Quoted 298 lines
drivers/gpu/drm/i915/i915_gem.c +30
-23+ ret =i915_wait_seqno (ring, seqno);if (ret)return ret;@@ -1329,6 +1333,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj, struct drm_i915_file_private *file_priv, bool readonly) { +struct drm_i915_gem_request *req;struct drm_device *dev = obj->base.dev;struct drm_i915_private *dev_priv = dev->dev_private;struct intel_engine_cs *ring = obj->ring;@@ -1339,10 +1344,13 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,BUG_ON (!mutex_is_locked (&dev->struct_mutex));BUG_ON (!dev_priv->mm.interruptible); - seqno = readonly ? obj->last_write_seqno : obj->last_read_seqno; -if (seqno ==0 ) + req = readonly ? obj->last_write_req : obj->last_read_req; +if (!req)return 0 ; + seqno =i915_gem_request_get_seqno (req); +BUG_ON (seqno ==0 ); + ret =i915_gem_check_wedge (&dev_priv->gpu_error, true);if (ret)return ret;@@ -2180,12 +2188,12 @@ static void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj, struct intel_engine_cs *ring) { - u32 seqno = intel_ring_get_seqno(ring); + struct drm_i915_gem_request *req = intel_ring_get_request(ring);BUG_ON (ring ==NULL ); -if (obj->ring != ring && obj->last_write_seqno) { -/* Keep the seqno relative to the current ring */ - obj->last_write_seqno = seqno; +if (obj->ring != ring && obj->last_write_req) { +/* Keep the request relative to the current ring */ + obj->last_write_req = req; } obj->ring = ring;@@ -2197,7 +2205,7 @@ i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,list_move_tail (&obj->ring_list, &ring->active_list); - obj->last_read_seqno = seqno; + obj->last_read_req = req; } void i915_vma_move_to_active(struct i915_vma *vma,@@ -2228,11 +2236,11 @@ i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)list_del_init (&obj->ring_list); obj->ring =NULL ; - obj->last_read_seqno =0 ; - obj->last_write_seqno =0 ; + obj->last_read_req =NULL ; + obj->last_write_req =NULL ; obj->base.write_domain =0 ; - obj->last_fenced_seqno =0 ; + obj->last_fenced_req =NULL ; obj->active =0 ;drm_gem_object_unreference (&obj->base);@@ -2249,7 +2257,7 @@ i915_gem_object_retire(struct drm_i915_gem_object *obj)return ;if (i915_seqno_passed (ring->get_seqno (ring, true), - obj->last_read_seqno)) +i915_gem_request_get_seqno (obj->last_read_req)))i915_gem_object_move_to_inactive (obj); }@@ -2669,7 +2677,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring) struct drm_i915_gem_object, ring_list); -if (!i915_seqno_passed (seqno, obj->last_read_seqno)) +if (!i915_seqno_passed (seqno,i915_gem_request_get_seqno (obj->last_read_req)))break ;i915_gem_object_move_to_inactive (obj);@@ -2779,7 +2787,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)int ret;if (obj->active) { - ret =i915_gem_check_olr (obj->ring, obj->last_read_seqno); + ret =i915_gem_check_olr (obj->ring,i915_gem_request_get_seqno (obj->last_read_req));if (ret)return ret;@@ -2838,13 +2846,14 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file) goto out;if (obj->active) { - seqno = obj->last_read_seqno; +if (!obj->last_read_req) + goto out; + + seqno =i915_gem_request_get_seqno (obj->last_read_req); +BUG_ON (seqno ==0 ); ring = obj->ring; } -if (seqno ==0 ) - goto out; -/* Do this after OLR check to make sure we make forward progress polling * on this IOCTL with a timeout <=0 (like busy ioctl) */ @@ -2894,7 +2903,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, idx =intel_ring_sync_index (from, to); - seqno = obj->last_read_seqno; + seqno =i915_gem_request_get_seqno (obj->last_read_req);/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ @@ -2907,11 +2916,12 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,trace_i915_gem_ring_sync_to (from, to, seqno); ret = to->semaphore.sync_to (to, from, seqno);if (!ret) -/* We use last_read_seqno because sync_to() +/* We use last_read_req because sync_to() * might have just caused seqno wrap under * the radar. */ - from->semaphore.sync_seqno[idx] = obj->last_read_seqno; + from->semaphore.sync_seqno[idx] = +i915_gem_request_get_seqno (obj->last_read_req);return ret; }@@ -3224,12 +3234,12 @@ static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj,static int i915_gem_object_wait_fence (struct drm_i915_gem_object *obj) { -if (obj->last_fenced_seqno) { -int ret =i915_wait_seqno (obj->ring, obj->last_fenced_seqno); +if (obj->last_fenced_req) { +int ret =i915_wait_seqno (obj->ring,i915_gem_request_get_seqno (obj->last_fenced_req));if (ret)return ret; - obj->last_fenced_seqno =0 ; + obj->last_fenced_req =NULL ; }return 0 ;drivers/gpu/drm/i915/i915_gem_execbuffer.c +3
-3@@ -946,7 +946,7 @@ void i915_gem_execbuffer_move_to_active(struct list_head *vmas, struct intel_engine_cs *ring) { - u32 seqno = intel_ring_get_seqno(ring); + struct drm_i915_gem_request *req = intel_ring_get_request(ring);struct i915_vma *vma;list_for_each_entry (vma, vmas, exec_list) {@@ -963,7 +963,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,i915_vma_move_to_active (vma, ring);if (obj->base.write_domain) { obj->dirty =1 ; - obj->last_write_seqno = seqno; + obj->last_write_req = req;intel_fb_obj_invalidate (obj, ring);@@ -971,7 +971,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas, obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS ; }if (entry->flags &EXEC_OBJECT_NEEDS_FENCE ) { - obj->last_fenced_seqno = seqno; + obj->last_fenced_req = req;if (entry->flags & __EXEC_OBJECT_HAS_FENCE) {struct drm_i915_private *dev_priv =to_i915 (ring->dev); list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,drivers/gpu/drm/i915/i915_gem_gtt.h +2
-2@@ -178,7 +178,7 @@ struct i915_address_space {* List of objects currently involved in rendering. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list. @@ -189,7 +189,7 @@ struct i915_address_space {* LRU list of objects which are not in the ringbuffer and * are ready to unbind, but are still in the GTT. * -* last_rendering_seqno is 0 while an object is in this list. +* last_rendering_req is NULL while an object is in this list. * * A reference is not held on the buffer while on this list, * as merely being GTT-bound shouldn't prevent its being drivers/gpu/drm/i915/i915_gem_tiling.c +1
-1@@ -383,7 +383,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,if (ret ==0 ) { obj->fence_dirty = - obj->last_fenced_seqno || + obj->last_fenced_req || obj->fence_reg !=I915_FENCE_REG_NONE ; obj->tiling_mode = args->tiling_mode;drivers/gpu/drm/i915/i915_gpu_error.c +2
-2@@ -666,8 +666,8 @@ static void capture_bo(struct drm_i915_error_buffer *err, err->size = obj->base.size; err->name = obj->base.name; - err->rseqno = obj->last_read_seqno; - err->wseqno = obj->last_write_seqno; + err->rseqno =i915_gem_request_get_seqno (obj->last_read_req); + err->wseqno =i915_gem_request_get_seqno (obj->last_write_req); err->gtt_offset = vma->node.start; err->read_domains = obj->base.read_domains; err->write_domain = obj->base.write_domain;drivers/gpu/drm/i915/intel_display.c +5
-5@@ -9766,16 +9766,16 @@ static int intel_postpone_flip(struct drm_i915_gem_object *obj)lockdep_assert_held (&obj->base.dev->struct_mutex); -if (!obj->last_write_seqno) +if (!obj->last_write_req)return 0 ; ring = obj->ring;if (i915_seqno_passed (ring->get_seqno (ring, true), - obj->last_write_seqno)) +i915_gem_request_get_seqno (obj->last_write_req)))return 0 ; - ret =i915_gem_check_olr (ring, obj->last_write_seqno); + ret =i915_gem_check_olr (ring,i915_gem_request_get_seqno (obj->last_write_req));if (ret)return ret;@@ -9838,7 +9838,7 @@ static int intel_queue_mmio_flip(struct drm_device *dev, }spin_lock_irqsave (&dev_priv->mmio_flip_lock, irq_flags); - intel_crtc->mmio_flip.seqno = obj->last_write_seqno; + intel_crtc->mmio_flip.seqno =i915_gem_request_get_seqno (obj->last_write_req); intel_crtc->mmio_flip.ring_id = obj->ring->id;spin_unlock_irqrestore (&dev_priv->mmio_flip_lock, irq_flags);@@ -10045,7 +10045,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,if (ret) goto cleanup_unpin; - work->flip_queued_seqno = obj->last_write_seqno; + work->flip_queued_seqno =i915_gem_request_get_seqno (obj->last_write_req); work->flip_queued_ring = obj->ring; }else { ret = dev_priv->display.queue_flip(dev, crtc, fb, obj, ring,drivers/gpu/drm/i915/intel_ringbuffer.h +1
-1@@ -249,7 +249,7 @@ struct intel_engine_cs {* ringbuffer. * * Includes buffers having the contents of their GPU caches -* flushed, not necessarily primitives. last_rendering_seqno +* flushed, not necessarily primitives. last_rendering_req * represents when the rendering involved will be completed. * * A reference is held on the buffer while on this list. Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 05/21] drm/i915: Add helper functions to aid seqno -> request transition
In reply to [email protected]Quoted 2 lines · On Mon, Oct 06, 2014 at 03:15:09PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
I think a few words about how these helpers help the transitions (and example maybe) would be great. Since without peeking ahead (which I didn't) and looking at the JIRA (which I've forgotten all about already right after writing it) this doesn't make a lot of sense ;-)
Quoted 35 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_drv.h | 12 ++++++++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 7 +++++++ 2 files changed, 19 insertions(+)drivers/gpu/drm/i915/i915_drv.h +12
-0@@ -1946,6 +1946,18 @@ struct drm_i915_gem_request {void i915_gem_request_free (struct kref *req_ref); +static inline uint32_t +i915_gem_request_get_seqno (struct drm_i915_gem_request *req) +{ +return req ? req->seqno :0 ; +} + +static inline struct intel_engine_cs * +i915_gem_request_get_ring (struct drm_i915_gem_request *req) +{ +return req ? req->ring :NULL ; +} +static inline void i915_gem_request_reference (struct drm_i915_gem_request *req) {drivers/gpu/drm/i915/intel_ringbuffer.h +0
-0
drivers/gpu/drm/i915/intel_ringbuffer.h static inline u32 intel_ring_get_seqno(struct intel_engine_cs *ring) · line 441
Quoted 7 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +4
-0@@ -435,6 +435,13 @@ static inline u32 intel_ring_get_seqno(struct intel_engine_cs *ring)return ring->outstanding_lazy_seqno; } +static inline struct drm_i915_gem_request * +intel_ring_get_request (struct intel_engine_cs *ring) +{ +BUG_ON (ring->preallocated_lazy_request ==NULL );
Again, I'm not a fan of BUG_ON. But if this is just a transitional thing and will disappear at the end of the series I'm ok with it. -Daniel
Quoted 13 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +3
-0+return ring->preallocated_lazy_request; +} +static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno) {if (ring->trace_irq_seqno ==0 && ring->irq_get (ring))Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 05/21] drm/i915: Add helper functions to aid seqno -> request transition
Quoted 46 lines · Daniel Vetter
Quoted 2 lines · On Mon, Oct 06, 2014 at 03:15:09PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
I think a few words about how these helpers help the transitions (and example maybe) would be great. Since without peeking ahead (which I didn't) and looking at the JIRA (which I've forgotten all about already right after writing it) this doesn't make a lot of sense ;-)
Quoted 38 lines
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/i915_drv.h | 12 ++++++++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 7 +++++++ 2 files changed, 19 insertions(+)drivers/gpu/drm/i915/i915_drv.h +12
-0@@ -1946,6 +1946,18 @@ struct drm_i915_gem_request {void i915_gem_request_free (struct kref *req_ref); +static inline uint32_t +i915_gem_request_get_seqno (struct drm_i915_gem_request *req) +{ +return req ? req->seqno :0 ; +} + +static inline struct intel_engine_cs * +i915_gem_request_get_ring (struct drm_i915_gem_request *req) +{ +return req ? req->ring :NULL ; +} +static inline void i915_gem_request_reference (struct drm_i915_gem_request *req) {drivers/gpu/drm/i915/intel_ringbuffer.h +0
-0@@ -435,6 +435,13 @@ static inline u32 intel_ring_get_seqno(struct intel_engine_cs *ring)return ring->outstanding_lazy_seqno; }
Quoted 7 lines
drivers/gpu/drm/i915/intel_ringbuffer.h static inline u32 intel_ring_get_seqno(struct intel_engine_cs *ring) · line 441
Quoted 4 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +4
-0+static inline struct drm_i915_gem_request * +intel_ring_get_request (struct intel_engine_cs *ring) +{ +BUG_ON (ring->preallocated_lazy_request ==NULL );Again, I'm not a fan of
BUG_ON. But if this is just a transitional thing and will disappear at the end of the series I'm ok with it. -Daniel
I just copied the existing 'intel_ring_get_seqno()' function and replaced 'seqno' with 'request'. The outgoing existing code has a BUG_ON therefore it seemed sensible to include a BUG_ON in the incoming new code.
Quoted 13 lines
Quoted 13 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +3
-0+return ring->preallocated_lazy_request; +} +static inline void i915_trace_irq_get (struct intel_engine_cs *ring,u32 seqno) {if (ring->trace_irq_seqno ==0 && ring->irq_get (ring))Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 03/21] drm/i915: Ensure OLS & PLR are always in sync
In reply to [email protected]Quoted 9 lines · On Mon, Oct 06, 2014 at 03:15:07PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
The new seqno alloction code pre-allocates a 'lazy' request structure and then tries to allocate the 'lazy' seqno. The seqno allocation can potential wrap around zero and when doing so, tries to idle the ring by waiting for all oustanding work to complete. With a scheduler in place, this can mean first submitting extra work to the ring. However, at this point in time, the lazy request is valid but the lazy seqno is not. Some existing code was getting confused by this state and Bad Things would happen.
I'm confused about this description. I think you need to go into more detail about what exactly you mean with "new seqno allocation code" (does this reference some future patches), "confused state" and "Bad Things".
I get a bit the impression that this is just a bad interaction between the scheduler and the seqno wrapping. Which might be due to allocating the seqno too early when the scheduler it active. So if that's the case then imo this shouldn't be in the s/seqno/request/ prep work series.
Quoted 30 lines
The safest solution is to still allocate the lazy request in advance (to avoid having to roll back in an out of memory sitation) but to save the pointer in a local variable rather than immediately updating the lazy pointer. Only after a valid seqno has been acquired is the lazy request pointer actually updated.
This guarantees that both lazy values are either invalid or both valid. There can no longer be an inconsistent state.
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/intel_lrc.c | 42 ++++++++++++++++++++----------- drivers/gpu/drm/i915/intel_ringbuffer.c | 29 +++++++++++++++------ 2 files changed, 48 insertions(+), 23 deletions(-)drivers/gpu/drm/i915/intel_lrc.c +2
-4@@ -796,27 +796,39 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf)static int logical_ring_alloc_seqno (struct intel_engine_cs *ring,struct intel_context *ctx) { -if (ring->outstanding_lazy_seqno) -return 0 ; +struct drm_i915_gem_request *request; +int ret; -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request;
drivers/gpu/drm/i915/intel_lrc.c void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf) · line 808
Quoted 10 lines
drivers/gpu/drm/i915/intel_lrc.c +6
-3+/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL );
Too many BUG_ON in this patch. This one here can easily be avoided with a if (WARN_ON) retrun -EIO; or so, the below one would just indicate a leak so a plain WARN_ON is good enough.
BUG_ON makes it a major pain to debug anything out there in the field, so the barrier to acceptance is really high.
Quoted 87 lines
drivers/gpu/drm/i915/intel_lrc.c +19
-8+return 0 ; + } +BUG_ON (ring->preallocated_lazy_request !=NULL ); -/* Hold a reference to the context this request belongs to -* (we will need it when the time comes to emit/retire the -* request). -*/ - request->ctx = ctx; -i915_gem_context_reference (request->ctx); + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +/* Hold a reference to the context this request belongs to +* (we will need it when the time comes to emit/retire the +* request). +*/ + request->ctx = ctx; +i915_gem_context_reference (request->ctx); + + ring->preallocated_lazy_request = request; +return 0 ; } static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf,drivers/gpu/drm/i915/intel_ringbuffer.c +21
-8@@ -2000,20 +2000,33 @@ int intel_ring_idle(struct intel_engine_cs *ring)static int intel_ring_alloc_seqno (struct intel_engine_cs *ring) { -if (ring->outstanding_lazy_seqno) +int ret; +struct drm_i915_gem_request *request; + +/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ + +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL );return 0 ; + } -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request; +BUG_ON (ring->preallocated_lazy_request !=NULL ); - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ring->preallocated_lazy_request = request; +return 0 ; } static int __intel_ring_prepare(struct intel_engine_cs *ring,Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 03/21] drm/i915: Ensure OLS & PLR are always in sync
Quoted 10 lines · Daniel Vetter
Quoted 9 lines · On Mon, Oct 06, 2014 at 03:15:07PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
The new seqno alloction code pre-allocates a 'lazy' request structure and then tries to allocate the 'lazy' seqno. The seqno allocation can potential wrap around zero and when doing so, tries to idle the ring by waiting for all oustanding work to complete. With a scheduler in place, this can mean first submitting extra work to the ring. However, at this point in time, the lazy request is valid but the lazy seqno is not. Some existing code was getting confused by this state and Bad Things would happen.
Quoted 8 lines
I'm confused about this description. I think you need to go into more detail about what exactly you mean with "new seqno allocation code" (does this reference some future patches), "confused state" and "Bad Things".
I get a bit the impression that this is just a bad interaction between the scheduler and the seqno wrapping. Which might be due to allocating the seqno too early when the scheduler it active. So if that's the case then imo this shouldn't be in the s/seqno/request/ prep work series.
This patch was originally part of the scheduler set. However, it made sense to pull this in earlier as part of this patch set as a safer step in the shift of focus from seqno to request. I didn't get around to updating the description for the new location. However, it might be simpler to not bother with this change and just jump straight to removing the seqno completely.
The Bad Things were either null pointer derefences when code used a null lazy request having only checked for the presence of a lazy seqno, or possibly attempting to process a zero seqno having only checked for the presence of a lazy request. The former was a kernel panic, the latter a silent failure to synchronise properly.
Quoted 44 lines
Quoted 30 lines
The safest solution is to still allocate the lazy request in advance (to avoid having to roll back in an out of memory sitation) but to save the pointer in a local variable rather than immediately updating the lazy pointer. Only after a valid seqno has been acquired is the lazy request pointer actually updated.
This guarantees that both lazy values are either invalid or both valid. There can no longer be an inconsistent state.
For: VIZ-4377
Signed-off-by [email protected]--- drivers/gpu/drm/i915/intel_lrc.c | 42 ++++++++++++++++++++----------- drivers/gpu/drm/i915/intel_ringbuffer.c | 29 +++++++++++++++------ 2 files changed, 48 insertions(+), 23 deletions(-)drivers/gpu/drm/i915/intel_lrc.c +2
-4@@ -796,27 +796,39 @@ void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf)static int logical_ring_alloc_seqno (struct intel_engine_cs *ring,struct intel_context *ctx) { -if (ring->outstanding_lazy_seqno) -return 0 ; +struct drm_i915_gem_request *request; +int ret; -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request;drivers/gpu/drm/i915/intel_lrc.c void intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf) · line 808
Quoted 10 lines
drivers/gpu/drm/i915/intel_lrc.c +6
-3+/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL );Too many
BUG_ONin this patch. This one here can easily be avoided with a if (WARN_ON) retrun-EIO; or so, the below one would just indicate a leak so a plainWARN_ONis good enough.
Quoted 2 lines
BUG_ONmakes it a major pain to debug anything out there in the field, so the barrier to acceptance is really high.
These BUG_ONs are purely temporary. They disappear when the seqno itself is removed and the function becomes '..._alloc_request()'.
Quoted 87 lines
Quoted 87 lines
drivers/gpu/drm/i915/intel_lrc.c +19
-8+return 0 ; + } +BUG_ON (ring->preallocated_lazy_request !=NULL ); -/* Hold a reference to the context this request belongs to -* (we will need it when the time comes to emit/retire the -* request). -*/ - request->ctx = ctx; -i915_gem_context_reference (request->ctx); + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +/* Hold a reference to the context this request belongs to +* (we will need it when the time comes to emit/retire the +* request). +*/ + request->ctx = ctx; +i915_gem_context_reference (request->ctx); + + ring->preallocated_lazy_request = request; +return 0 ; } static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf,drivers/gpu/drm/i915/intel_ringbuffer.c +21
-8@@ -2000,20 +2000,33 @@ int intel_ring_idle(struct intel_engine_cs *ring)static int intel_ring_alloc_seqno (struct intel_engine_cs *ring) { -if (ring->outstanding_lazy_seqno) +int ret; +struct drm_i915_gem_request *request; + +/* The aim is to replace seqno values with request structures. A step +* along the way is to switch to using the PLR in preference to the +* OLS. That requires the PLR to only be valid when the OLS is also +* valid. I.e., the two must be kept in step. */ + +if (ring->outstanding_lazy_seqno) { +BUG_ON (ring->preallocated_lazy_request ==NULL );return 0 ; + } -if (ring->preallocated_lazy_request ==NULL ) { -struct drm_i915_gem_request *request; +BUG_ON (ring->preallocated_lazy_request !=NULL ); - request =kmalloc (sizeof (*request),GFP_KERNEL ); -if (request ==NULL ) -return -ENOMEM ; + request =kmalloc (sizeof (*request),GFP_KERNEL ); +if (request ==NULL ) +return -ENOMEM ; - ring->preallocated_lazy_request = request; + ret =i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); +if (ret) { +kfree (request); +return ret; } -return i915_gem_get_seqno (ring->dev, &ring->outstanding_lazy_seqno); + ring->preallocated_lazy_request = request; +return 0 ; } static int __intel_ring_prepare(struct intel_engine_cs *ring,Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 02/21] drm/i915: Remove redundant parameter to i915_gem_object_wait_rendering__tail()
In reply to [email protected]Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:06PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Empty commit messages are a bit too thin. E.g. in this case this was an oversight from
commit c8725f3dc0911d4354315a65150aecd8b7d0d74a
Author: Chris Wilson <[email protected]> Date: Mon Mar 17 12:21:55 2014 +0000
drm/i915: Do not call retire_requests from wait_for_rendering
which means your patch here really should have Cc'ed patch author and revierer (i.e. Chris & Brad) of the offending patch, too.
Especially in GEM where some of the interactions are _really_ tricky you can't just remove code that looks funky. You really have to dig into the history and find solid evidence that this is really just an oversight and not something deeper.
Quoted 2 lines
For: VIZ-4377
Signed-off-by [email protected]
Youre sob line isn't quite up to spec, it should have a full name + mail address (like the From: line git send-email inserted). -Daniel
Quoted 43 lines
--- drivers/gpu/drm/i915/i915_gem.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)drivers/gpu/drm/i915/i915_gem.c +3
-4@@ -1281,8 +1281,7 @@ i915_wait_seqno(struct intel_engine_cs *ring, uint32_t seqno) }static int -i915_gem_object_wait_rendering__tail (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring) +i915_gem_object_wait_rendering__tail (struct drm_i915_gem_object *obj) {if (!obj->active)return 0 ;@@ -1319,7 +1318,7 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,if (ret)return ret; -return i915_gem_object_wait_rendering__tail (obj, ring); +return i915_gem_object_wait_rendering__tail (obj); }/* A nonblocking variant of the above wait. This is a highly dangerous routine @@ -1359,7 +1358,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,if (ret)return ret; -return i915_gem_object_wait_rendering__tail (obj, ring); +return i915_gem_object_wait_rendering__tail (obj); }/** Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 02/21] drm/i915: Remove redundant parameter to i915_gem_object_wait_rendering__tail()
Quoted 6 lines · Daniel Vetter
Quoted 1 line · On Mon, Oct 06, 2014 at 03:15:06PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Empty commit messages are a bit too thin. E.g. in this case this was an oversight from
Quoted 8 lines
commit
c8725f3dc0911d4354315a65150aecd8b7d0d74aAuthor: Chris Wilson <[email protected]> Date: Mon Mar 17 12:21:55 2014 +0000drm/i915: Do not call
retire_requestsfromwait_for_renderingwhich means your patch here really should have Cc'ed patch author and revierer (i.e. Chris & Brad) of the offending patch, too.
To make things clear: Cc'ing people isn't to assign blame for every single mistake they do (we're all racking up impressive amounts of them anyway), but to give them a learning opportunity and the chance to improve. It is pretty much impossible to follow intel-gfx and read everything that's on here even for me, so if you don't cc people for even the most mundane and small things they'll miss the mail. -Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 01/21] Bug: missing i915_seqno_passed() call?
In reply to [email protected]Quoted 11 lines · On Mon, Oct 06, 2014 at 03:15:05PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
--- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/i915_gem.c +0
-0
drivers/gpu/drm/i915/i915_gem.c i915_gem_object_sync(struct drm_i915_gem_object *obj, · line 2889
Quoted 5 lines
drivers/gpu/drm/i915/i915_gem.c +1
-1@@ -2886,7 +2886,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, seqno = obj->last_read_seqno;/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ -if (seqno <= from->semaphore.sync_seqno[idx]) +if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */
No, becuase hw semaphores don't wrap around so we actual need to check for <=, not the wrap-around logic seqno_passed has. If you feel like pls submit a patch to add a comment here.
btw you're threading looks a bit funny, instead of patches all being in-reply-to the cover letter they're all in-reply-to the previous patch. Presuming you don't have a funky .gitconfig recent git should do this in the preferred form by default.
-Daniel
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +0
-0return 0 ; ret =i915_gem_check_olr (obj->ring, seqno);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 01/21] Bug: missing i915_seqno_passed() call?
Quoted 7 lines · Daniel Vetter
Quoted 6 lines · On Mon, Oct 06, 2014 at 03:15:05PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
--- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 13 lines
drivers/gpu/drm/i915/i915_gem.c i915_gem_object_sync(struct drm_i915_gem_object *obj, · line 2889
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +1
-1@@ -2886,7 +2886,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj, seqno = obj->last_read_seqno;/* Optimization: Avoid semaphore sync when we are sure we already * waited for an object with higher seqno */ -if (seqno <= from->semaphore.sync_seqno[idx]) +if (seqno <= from->semaphore.sync_seqno[idx])/* <--- broken?! needs to use i915_seqno_passed()??? */ No, becuase hw semaphores don't wrap around so we actual need to check for <=, not the wrap-around logic
seqno_passedhas. If you feel like pls submit a patch to add a comment here.
I would prefer to remove the test completely and replace it with some request structure based operation. Unfortunately, I haven't worked out what yet.
Quoted 4 lines
btw you're threading looks a bit funny, instead of patches all being in-reply-to the cover letter they're all in-reply-to the previous patch. Presuming you don't have a funky .gitconfig recent git should do this in the preferred form by default.
Phooey. Git-sendmail kept spewing warning messages about chain-reply-to having changed and needs to be sent explicitly in the config file to keep it quiet. I guess I set it the wrong way.
Quoted 12 lines
-Daniel
Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +0
-0return 0 ; ret =i915_gem_check_olr (obj->ring, seqno);Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 3 lines · On Mon, Oct 06, 2014 at 03:15:04PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
You fail to end up with my earlier code. Nak.
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Quoted 6 lines · Chris Wilson
Quoted 3 lines · On Mon, Oct 06, 2014 at 03:15:04PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
You fail to end up with my earlier code. Nak.
Well I've tried to split up your patch into small independent changes and failed at that. And given how many changes there are in there I simply can't merge your patch as-is. I know that it does seem to fix some random hangs, but with massive behaviour changes like the read-read stuff in there this could very well just be a random fluke.
But if there's anything amiss in John's work for the plain s/seqno/request/ change, or anything else that we absolutely need to have then I very much want your opinion on this. But a flat outright nak without further content isn't useful to move forward.
Thanks, Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Quoted 6 lines · Daniel Vetter
Quoted 5 lines · Chris Wilson
Quoted 3 lines · On Mon, Oct 06, 2014 at 03:15:04PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
Quoted 7 lines
Quoted 1 line
You fail to end up with my earlier code. Nak.
Well I've tried to split up your patch into small independent changes and failed at that. And given how many changes there are in there I simply can't merge your patch as-is. I know that it does seem to fix some random hangs, but with massive behaviour changes like the read-read stuff in there this could very well just be a random fluke.
Nope. Try reading it again. The most invasive change is defining the order of engine/context/ppgtt/ring creation (and doing the setup/enable split). Everything else is about making a request a transaction and using that to fix bugs in our state tacking.
The fundamental change, that I am going to be a stickler for, is that the request is the ring access. It defines the context, owner and interface to both the ring and the engine/scheduler. In doing so it drops the inconsistent and unmaintainable approach of introducing separate but almost duplicate code paths inside the core of GEM.
Quoted 4 lines
But if there's anything amiss in John's work for the plain s/seqno/request/ change, or anything else that we absolutely need to have then I very much want your opinion on this. But a flat outright nak without further content isn't useful to move forward.
It doesn't reduce the technical debt of execlists, which was the motivation for the changes. It even brings a breath of sanity to contexts and ppgtt as well, a massive boon.
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Quoted 15 lines · Chris Wilson
Quoted 6 lines · Daniel Vetter
Quoted 5 lines · Chris Wilson
Quoted 3 lines · On Mon, Oct 06, 2014 at 03:15:04PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
Quoted 7 lines
Quoted 1 line
You fail to end up with my earlier code. Nak.
Well I've tried to split up your patch into small independent changes and failed at that. And given how many changes there are in there I simply can't merge your patch as-is. I know that it does seem to fix some random hangs, but with massive behaviour changes like the read-read stuff in there this could very well just be a random fluke.
Quoted 4 lines
Nope. Try reading it again. The most invasive change is defining the order of engine/context/ppgtt/ring creation (and doing the setup/enable split). Everything else is about making a request a transaction and using that to fix bugs in our state tacking.
I guess the setup changes only showed up in the revised and much bigger version. At least I don't remember seeing that in the first iteration I've tried to split up. Or maybe I've been blind. In any case I want to untangle that web too, just with a much more gradual approach.
Quoted 5 lines
The fundamental change, that I am going to be a stickler for, is that the request is the ring access. It defines the context, owner and interface to both the ring and the engine/scheduler. In doing so it drops the inconsistent and unmaintainable approach of introducing separate but almost duplicate code paths inside the core of GEM.
So for a clean slate design I agree with you. But given where we are I still think that the duplication with the execlist code in intel_lrc.c is a good approach. It will lead to a lot more ugliness, and there's a good chance that we can't get rid of the final bits in years. But given that execlist absolutely had to get in I really didn't see any other way to merge it, without the inevitable reworks totally wreaking the other parts of gem.
It's definitely not a great option, but there's also limits to how much flak I can absorb without crumbling completely.
Quoted 8 lines
Quoted 4 lines
But if there's anything amiss in John's work for the plain s/seqno/request/ change, or anything else that we absolutely need to have then I very much want your opinion on this. But a flat outright nak without further content isn't useful to move forward.
It doesn't reduce the technical debt of execlists, which was the motivation for the changes. It even brings a breath of sanity to contexts and ppgtt as well, a massive boon.
I agree on that. It should give us a start though to clean things up by shifting a lot tracking execlist does over to requests. Or at least I hope so. And there's still the issue that I really need to avoid another dept pile-up excercise with the scheduler, so for just that part I think we need this. Even when it doesn't address any of the issues with execlist. But I do think this will allow us to at least move into the right direction a bit, so please scream if I'm totally off the tracks.
Thanks, Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
[RFC 22/21] drm/i915: Cache request completion status
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 34 ++++++++++++++++--------------- drivers/gpu/drm/i915/i915_gem.c | 21 +++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +++ 5 files changed, 45 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +18 -16
@@ -1913,6 +1913,9 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,struct drm_i915_gem_request {struct kref ref; +/** Is this request known to be complete? */ +bool complete; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;
@@ -1943,6 +1946,8 @@ struct drm_i915_gem_request { };void i915_gem_request_free (struct kref *req_ref); +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency);static inline uint32_t i915_gem_request_get_seqno (struct drm_i915_gem_request *req)
@@ -1968,7 +1973,19 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } -/* ??? i915_gem_request_completed should be here ??? */ +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +if (req->complete) +return true; + +if (req->ring ==NULL ) +return false; + +i915_gem_complete_requests_ring (req->ring, lazy_coherency); + +return req->complete; +}struct drm_i915_file_private {struct drm_i915_private *dev_priv;
@@ -3019,19 +3036,4 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } -static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, -bool lazy_coherency) -{ -u32 seqno; - -BUG_ON (req ==NULL ); - -if (req->ring ==NULL ) -return false; - - seqno = req->ring->get_seqno (req->ring, lazy_coherency); - -return i915_seqno_passed (seqno, req->seqno); -} -#endif
drivers/gpu/drm/i915/i915_gem.c +21 -0
@@ -2641,6 +2641,27 @@ void i915_gem_reset(struct drm_device *dev)i915_gem_restore_fences (dev); } +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency) +{ +struct drm_i915_gem_request *req; +u32 seqno; + + seqno = ring->get_seqno (ring, lazy_coherency); +if (seqno == ring->last_read_seqno) +return ; + +list_for_each_entry (req, &ring->request_list, list) { +if (req->complete) +continue ; + +if (i915_seqno_passed (seqno, req->seqno)) + req->complete = true; + } + + ring->last_read_seqno = seqno; +} +/** * This function clears the request list as sequence numbers are passed. */
drivers/gpu/drm/i915/intel_lrc.c +1 -0
@@ -808,6 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -0
@@ -2023,6 +2023,7 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
@@ -2115,6 +2116,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)I915_WRITE (RING_SYNC_2 (ring->mmio_base),0 ); } + ring->last_read_seqno =0 ; ring->set_seqno (ring, seqno); ring->hangcheck.seqno = seqno; }
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
@@ -269,6 +269,9 @@ struct intel_engine_cs {bool gpu_caches_dirty;bool fbc_dirty; +/* For optimising request completion events */ +u32 last_read_seqno; +wait_queue_head_t irq_queue;struct intel_context *default_context;
Signature · 1 line
1.7.9.5
RFC 23/25 drm/i915: Cache request completion status
From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 32 +++++++++++++++---------------- drivers/gpu/drm/i915/i915_gem.c | 21 ++++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +++ 5 files changed, 42 insertions(+), 17 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +15 -17
@@ -1913,6 +1913,9 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,struct drm_i915_gem_request {struct kref ref; +/** Is this request known to be complete? */ +bool complete; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;
@@ -1943,6 +1946,8 @@ struct drm_i915_gem_request { };void i915_gem_request_free (struct kref *req_ref); +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency);static inline uint32_t i915_gem_request_get_seqno (struct drm_i915_gem_request *req)
@@ -1968,11 +1973,16 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } -/* -* NB: i915_gem_request_completed should be here but currently needs the -* definition of i915_seqno_passed() which is below. It will be moved in -* a later patch when the call to i915_seqno_passed() is obsoleted... -*/ +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +if (req->complete) +return true; + +i915_gem_complete_requests_ring (req->ring, lazy_coherency); + +return req->complete; +}struct drm_i915_file_private {struct drm_i915_private *dev_priv;
@@ -3023,16 +3033,4 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } -static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, -bool lazy_coherency) -{ -u32 seqno; - -BUG_ON (req ==NULL ); - - seqno = req->ring->get_seqno (req->ring, lazy_coherency); - -return i915_seqno_passed (seqno, req->seqno); -} -#endif
drivers/gpu/drm/i915/i915_gem.c +21 -0
@@ -2640,6 +2640,27 @@ void i915_gem_reset(struct drm_device *dev)i915_gem_restore_fences (dev); } +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency) +{ +struct drm_i915_gem_request *req; +u32 seqno; + + seqno = ring->get_seqno (ring, lazy_coherency); +if (seqno == ring->last_read_seqno) +return ; + +list_for_each_entry (req, &ring->request_list, list) { +if (req->complete) +continue ; + +if (i915_seqno_passed (seqno, req->seqno)) + req->complete = true; + } + + ring->last_read_seqno = seqno; +} +/** * This function clears the request list as sequence numbers are passed. */
drivers/gpu/drm/i915/intel_lrc.c +1 -0
@@ -808,6 +808,7 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring = ring; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -0
@@ -2023,6 +2023,7 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring = ring; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
@@ -2115,6 +2116,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)I915_WRITE (RING_SYNC_2 (ring->mmio_base),0 ); } + ring->last_read_seqno =0 ; ring->set_seqno (ring, seqno); ring->hangcheck.seqno = seqno; }
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
@@ -269,6 +269,9 @@ struct intel_engine_cs {bool gpu_caches_dirty;bool fbc_dirty; +/* For optimising request completion events */ +u32 last_read_seqno; +wait_queue_head_t irq_queue;struct intel_context *default_context;
Signature · 1 line
1.7.9.5
Re: [RFC 22/21] drm/i915: Cache request completion status
In reply to [email protected]Quoted 4 lines · On Tue, Oct 07, 2014 at 05:47:29PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
Why? If it's just for performance I think we should do this as part of the switch to struct fence, which already has this.
Quoted 39 lines
--- drivers/gpu/drm/i915/i915_drv.h | 34 ++++++++++++++++--------------- drivers/gpu/drm/i915/i915_gem.c | 21 +++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +++ 5 files changed, 45 insertions(+), 16 deletions(-)drivers/gpu/drm/i915/i915_drv.h +8
-1@@ -1913,6 +1913,9 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,struct drm_i915_gem_request {struct kref ref; +/** Is this request known to be complete? */ +bool complete; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;@@ -1943,6 +1946,8 @@ struct drm_i915_gem_request { };void i915_gem_request_free (struct kref *req_ref); +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency);static inline uint32_t i915_gem_request_get_seqno (struct drm_i915_gem_request *req)@@ -1968,7 +1973,19 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } -/* ??? i915_gem_request_completed should be here ??? */ +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{
drivers/gpu/drm/i915/i915_drv.h i915_gem_request_unreference(struct drm_i915_gem_request *req) · line 1988
Quoted 10 lines
drivers/gpu/drm/i915/i915_drv.h +10
-0+if (req->complete) +return true; + +if (req->ring ==NULL ) +return false; + +i915_gem_complete_requests_ring (req->ring, lazy_coherency); + +return req->complete; +}
Also, this is looking way too big now I think ;-) If you have a full non-inline function call in your inline it's always a net loss. -Daniel
Quoted 108 lines
drivers/gpu/drm/i915/i915_drv.h +0
-15struct drm_i915_file_private {struct drm_i915_private *dev_priv;@@ -3019,19 +3036,4 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } -static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, -bool lazy_coherency) -{ -u32 seqno; - -BUG_ON (req ==NULL ); - -if (req->ring ==NULL ) -return false; - - seqno = req->ring->get_seqno (req->ring, lazy_coherency); - -return i915_seqno_passed (seqno, req->seqno); -} -#endif drivers/gpu/drm/i915/i915_gem.c +21
-0@@ -2641,6 +2641,27 @@ void i915_gem_reset(struct drm_device *dev)i915_gem_restore_fences (dev); } +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency) +{ +struct drm_i915_gem_request *req; +u32 seqno; + + seqno = ring->get_seqno (ring, lazy_coherency); +if (seqno == ring->last_read_seqno) +return ; + +list_for_each_entry (req, &ring->request_list, list) { +if (req->complete) +continue ; + +if (i915_seqno_passed (seqno, req->seqno)) + req->complete = true; + } + + ring->last_read_seqno = seqno; +} +/** * This function clears the request list as sequence numbers are passed. */ drivers/gpu/drm/i915/intel_lrc.c +1
-0@@ -808,6 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {drivers/gpu/drm/i915/intel_ringbuffer.c +2
-0@@ -2023,6 +2023,7 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {@@ -2115,6 +2116,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)I915_WRITE (RING_SYNC_2 (ring->mmio_base),0 ); } + ring->last_read_seqno =0 ; ring->set_seqno (ring, seqno); ring->hangcheck.seqno = seqno; }drivers/gpu/drm/i915/intel_ringbuffer.h +3
-0@@ -269,6 +269,9 @@ struct intel_engine_cs {bool gpu_caches_dirty;bool fbc_dirty; +/* For optimising request completion events */ +u32 last_read_seqno; +wait_queue_head_t irq_queue;struct intel_context *default_context;Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 22/21] drm/i915: Cache request completion status
Quoted 7 lines · Daniel Vetter
Quoted 4 lines · On Tue, Oct 07, 2014 at 05:47:29PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]Why? If it's just for performance I think we should do this as part of the switch to struct fence, which already has this.
For performance and also as part of getting rid of all the i915_seqno_passed() calls.
Quoted 45 lines
Quoted 45 lines
--- drivers/gpu/drm/i915/i915_drv.h | 34 ++++++++++++++++--------------- drivers/gpu/drm/i915/i915_gem.c | 21 +++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +++ 5 files changed, 45 insertions(+), 16 deletions(-)drivers/gpu/drm/i915/i915_drv.h +13
-1@@ -1913,6 +1913,9 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,struct drm_i915_gem_request {struct kref ref; +/** Is this request known to be complete? */ +bool complete; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;@@ -1943,6 +1946,8 @@ struct drm_i915_gem_request { };void i915_gem_request_free (struct kref *req_ref); +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency);static inline uint32_t i915_gem_request_get_seqno (struct drm_i915_gem_request *req)@@ -1968,7 +1973,19 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } -/* ??? i915_gem_request_completed should be here ??? */ +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +if (req->complete) +return true; + +if (req->ring ==NULL ) +return false;
Quoted 8 lines
drivers/gpu/drm/i915/i915_drv.h i915_gem_request_unreference(struct drm_i915_gem_request *req) · line 1988
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +5
-0+ +i915_gem_complete_requests_ring (req->ring, lazy_coherency); + +return req->complete; +}Also, this is looking way too big now I think ;-) If you have a full non-inline function call in your inline it's always a net loss. -Daniel
That depends how you define gain/loss. In terms of performance, it can still be a gain because the function call is not always taken. Whereas the alternative is at least one function calls and possibly two. Either way, as noted already, the final intention is for this to become simply 'return req->complete' and not have any function calls at all.
Quoted 101 lines
Quoted 101 lines
drivers/gpu/drm/i915/i915_drv.h +0
-15struct drm_i915_file_private {struct drm_i915_private *dev_priv;@@ -3019,19 +3036,4 @@ wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms) } } -static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, -bool lazy_coherency) -{ -u32 seqno; - -BUG_ON (req ==NULL ); - -if (req->ring ==NULL ) -return false; - - seqno = req->ring->get_seqno (req->ring, lazy_coherency); - -return i915_seqno_passed (seqno, req->seqno); -} -#endif drivers/gpu/drm/i915/i915_gem.c +21
-0@@ -2641,6 +2641,27 @@ void i915_gem_reset(struct drm_device *dev)i915_gem_restore_fences (dev); } +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency) +{ +struct drm_i915_gem_request *req; +u32 seqno; + + seqno = ring->get_seqno (ring, lazy_coherency); +if (seqno == ring->last_read_seqno) +return ; + +list_for_each_entry (req, &ring->request_list, list) { +if (req->complete) +continue ; + +if (i915_seqno_passed (seqno, req->seqno)) + req->complete = true; + } + + ring->last_read_seqno = seqno; +} +/** * This function clears the request list as sequence numbers are passed. */ drivers/gpu/drm/i915/intel_lrc.c +1
-0@@ -808,6 +808,7 @@ static int logical_ring_alloc_seqno(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {drivers/gpu/drm/i915/intel_ringbuffer.c +2
-0@@ -2023,6 +2023,7 @@ intel_ring_alloc_seqno(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring =NULL ; + request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {@@ -2115,6 +2116,7 @@ void intel_ring_init_seqno(struct intel_engine_cs *ring, u32 seqno)I915_WRITE (RING_SYNC_2 (ring->mmio_base),0 ); } + ring->last_read_seqno =0 ; ring->set_seqno (ring, seqno); ring->hangcheck.seqno = seqno; }drivers/gpu/drm/i915/intel_ringbuffer.h +3
-0@@ -269,6 +269,9 @@ struct intel_engine_cs {bool gpu_caches_dirty;bool fbc_dirty; +/* For optimising request completion events */ +u32 last_read_seqno; +wait_queue_head_t irq_queue;
Quoted 8 lines
Quoted 8 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +0
-0struct intel_context *default_context;Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 22/21] drm/i915: Cache request completion status
Quoted 10 lines · John Harrison
Quoted 7 lines · Daniel Vetter
Quoted 4 lines · On Tue, Oct 07, 2014 at 05:47:29PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]Why? If it's just for performance I think we should do this as part of the switch to struct fence, which already has this.
For performance and also as part of getting rid of all the
i915_seqno_passed()calls.
The why was more a request for a proper commit message. If you claim perfomance, then the commit message also needs the relevant data.
Quoted 47 lines
Quoted 39 lines
Quoted 39 lines
--- drivers/gpu/drm/i915/i915_drv.h | 34 ++++++++++++++++--------------- drivers/gpu/drm/i915/i915_gem.c | 21 +++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +++ 5 files changed, 45 insertions(+), 16 deletions(-)drivers/gpu/drm/i915/i915_drv.h +13
-1@@ -1913,6 +1913,9 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,struct drm_i915_gem_request {struct kref ref; +/** Is this request known to be complete? */ +bool complete; +/** On Which ring this request was generated */ struct intel_engine_cs *ring;@@ -1943,6 +1946,8 @@ struct drm_i915_gem_request { };void i915_gem_request_free (struct kref *req_ref); +void i915_gem_complete_requests_ring (struct intel_engine_cs *ring, +bool lazy_coherency);static inline uint32_t i915_gem_request_get_seqno (struct drm_i915_gem_request *req)@@ -1968,7 +1973,19 @@ i915_gem_request_unreference(struct drm_i915_gem_request *req)kref_put (&req->ref, i915_gem_request_free); } -/* ??? i915_gem_request_completed should be here ??? */ +static inline bool i915_gem_request_completed (struct drm_i915_gem_request *req, +bool lazy_coherency) +{ +if (req->complete) +return true; + +if (req->ring ==NULL ) +return false;Quoted 8 lines
drivers/gpu/drm/i915/i915_drv.h i915_gem_request_unreference(struct drm_i915_gem_request *req) · line 1988
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +5
-0+ +i915_gem_complete_requests_ring (req->ring, lazy_coherency); + +return req->complete; +}Also, this is looking way too big now I think ;-) If you have a full non-inline function call in your inline it's always a net loss. -Daniel
Quoted 5 lines
That depends how you define gain/loss. In terms of performance, it can still be a gain because the function call is not always taken. Whereas the alternative is at least one function calls and possibly two. Either way, as noted already, the final intention is for this to become simply 'return
req->complete' and not have any function calls at all.
Thrashing instructions caches is usually what this does, so even when you microbenchmark a gain it might be a net loss. Rule-of-thumb is static inline is ok if it results in a net reduction of the binary, not for any other case.
Unless you supply solid performance data justifying it.
Also, static inlines in random headers are a pain for documentation. -Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
RFC 15/25 drm/i915: Connect requests to rings at creation not submission
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_gem.c | 1 - drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 1 + 3 files changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem.c +0 -1
@@ -2386,7 +2386,6 @@ int __i915_add_request(struct intel_engine_cs *ring,i915_gem_request_reference (request); - request->ring = ring; request->head = request_start; request->tail = request_ring_position;
drivers/gpu/drm/i915/intel_lrc.c +1 -0
@@ -807,6 +807,7 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,return -ENOMEM ;kref_init (&request->ref); + request->ring = ring; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
drivers/gpu/drm/i915/intel_ringbuffer.c +1 -0
@@ -2020,6 +2020,7 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)return -ENOMEM ;kref_init (&request->ref); + request->ring = ring; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
Signature · 1 line
1.7.9.5
RFC 24/25 drm/i915: Zero fill the request structure
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/intel_lrc.c | 3 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 3 +-- 2 files changed, 2 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +1 -2
@@ -802,13 +802,12 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
drivers/gpu/drm/i915/intel_ringbuffer.c +1 -2
@@ -2017,13 +2017,12 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {
Signature · 1 line
1.7.9.5
Re: [RFC 24/25] drm/i915: Zero fill the request structure
Quoted 4 lines · On Fri, Oct 10, 2014 at 12:41:12PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
I think this should be squashed (well, split first) into the relevant earlier patches. Generally I much prefer kzalloc, and we use that almost everywhere.
Or does this silently fix some issue and doesn't tell? -Daniel
Quoted 50 lines
--- drivers/gpu/drm/i915/intel_lrc.c | 3 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 3 +-- 2 files changed, 2 insertions(+), 4 deletions(-)drivers/gpu/drm/i915/intel_lrc.c +1
-2@@ -802,13 +802,12 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {drivers/gpu/drm/i915/intel_ringbuffer.c +1
-2@@ -2017,13 +2017,12 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [RFC 24/25] drm/i915: Zero fill the request structure
Quoted 11 lines · Daniel Vetter
Quoted 4 lines · On Fri, Oct 10, 2014 at 12:41:12PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]I think this should be squashed (well, split first) into the relevant earlier patches. Generally I much prefer kzalloc, and we use that almost everywhere.
Or does this silently fix some issue and doesn't tell? -Daniel
No, I just left it separate so as to change as little as possible in each individual patch. The original code did a kmalloc() of the request. That predates my changes. So if I did a km -> kz change in the middle of some other work I was assuming you would complain at me doing too much in a single patch. Whereas, the 'complete' field is the first time a zero fill becomes significant so it made sense (to me) to leave the switch to kz until here.
Quoted 46 lines
Quoted 46 lines
--- drivers/gpu/drm/i915/intel_lrc.c | 3 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 3 +-- 2 files changed, 2 insertions(+), 4 deletions(-)drivers/gpu/drm/i915/intel_lrc.c +1
-2@@ -802,13 +802,12 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {drivers/gpu/drm/i915/intel_ringbuffer.c +1
-2@@ -2017,13 +2017,12 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)if (ring->outstanding_lazy_request)return 0 ; - request =kmalloc (sizeof (*request),GFP_KERNEL ); + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;kref_init (&request->ref); request->ring = ring; - request->complete = false; ret =i915_gem_get_seqno (ring->dev, &request->seqno);if (ret) {Signature · 1 line
1.7.9.5
Quoted 4 lines
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [RFC 24/25] drm/i915: Zero fill the request structure
Quoted 12 lines · John Harrison
Quoted 11 lines · Daniel Vetter
Quoted 4 lines · On Fri, Oct 10, 2014 at 12:41:12PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]I think this should be squashed (well, split first) into the relevant earlier patches. Generally I much prefer kzalloc, and we use that almost everywhere.
Or does this silently fix some issue and doesn't tell? -Daniel
Quoted 7 lines
No, I just left it separate so as to change as little as possible in each individual patch. The original code did a
kmalloc()of the request. That predates my changes. So if I did a km -> kz change in the middle of some other work I was assuming you would complain at me doing too much in a single patch. Whereas, the 'complete' field is the first time a zero fill becomes significant so it made sense (to me) to leave the switch to kz until here.
Makes sense, but this kind of explanation really should be in the commit message.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
RFC 25/25 drm/i915: Defer seqno allocation until actual hardware submission time
In reply to [email protected]From: John Harrison <[email protected]>
For: VIZ-4377
--- drivers/gpu/drm/i915/i915_drv.h | 5 ++++- drivers/gpu/drm/i915/i915_gem.c | 28 +++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_lrc.c | 10 ++++------ drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++------ 4 files changed, 39 insertions(+), 14 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +4 -1
@@ -1979,6 +1979,9 @@ static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,if (req->complete)return true; +if (req->seqno ==0 ) +return false; +i915_gem_complete_requests_ring (req->ring, lazy_coherency);return req->complete;
@@ -2482,7 +2485,7 @@ i915_seqno_passed(uint32_t seq1, uint32_t seq2)return (int32_t )(seq1 - seq2) >=0 ; } -int __must_check i915_gem_get_seqno (struct drm_device *dev,u32 *seqno); +int __must_check i915_gem_prepare_next_seqno (struct drm_device *dev);int __must_check i915_gem_set_seqno (struct drm_device *dev,u32 seqno);int __must_check i915_gem_object_get_fence (struct drm_i915_gem_object *obj);int __must_check i915_gem_object_put_fence (struct drm_i915_gem_object *obj);
drivers/gpu/drm/i915/i915_gem.c +27 -1
@@ -2310,12 +2310,15 @@ int i915_gem_set_seqno(struct drm_device *dev, u32 seqno) }int -i915_gem_get_seqno (struct drm_device *dev,u32 *seqno) +i915_gem_prepare_next_seqno (struct drm_device *dev) {struct drm_i915_private *dev_priv = dev->dev_private;/* reserve 0 for non-seqno */ if (dev_priv->next_seqno ==0 ) { +/* Why is the full re-initialisation required? Is it only for +* hardware semaphores? If so, could skip it in the case where +* semaphores are disabled? */ int ret =i915_gem_init_seqno (dev,0 );if (ret)return ret;
@@ -2323,6 +2326,24 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno) dev_priv->next_seqno =1 ; } +return 0 ; +} + +static int +i915_gem_get_seqno (struct drm_device *dev,u32 *seqno) +{ +struct drm_i915_private *dev_priv = dev->dev_private; + +/* reserve 0 for non-seqno */ +if (dev_priv->next_seqno ==0 ) { +/* Should never get here! Must always call 'prepare_next' in +* advance. This code is called during request submission. +* Trying to wrap the seqno and the implicit idle() calls that +* the wrap code makes are a bad idea at this point! */ +DRM_ERROR ("Need to wrap seqno at inopportune moment!\n" ); +return -EBUSY ; + } + *seqno = dev_priv->last_seqno = dev_priv->next_seqno++;return 0 ; }
@@ -2366,6 +2387,11 @@ int __i915_add_request(struct intel_engine_cs *ring,return ret; } +/* Assign an identifier to track this request through the hardware: */ + ret =i915_gem_get_seqno (ring->dev, &request->seqno); +if (ret) +return ret; +/* Record the position of the start of the request so that * should we detect the updated seqno part-way through the * GPU processing the request, we never over-estimate the
drivers/gpu/drm/i915/intel_lrc.c +4 -6
@@ -802,6 +802,10 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,if (ring->outstanding_lazy_request)return 0 ; + ret =i915_gem_prepare_next_seqno (ring->dev); +if (ret) +return ret; + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;
@@ -809,12 +813,6 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring = ring; - ret =i915_gem_get_seqno (ring->dev, &request->seqno); -if (ret) { -kfree (request); -return ret; - } -/* Hold a reference to the context this request belongs to * (we will need it when the time comes to emit/retire the * request).
drivers/gpu/drm/i915/intel_ringbuffer.c +4 -6
@@ -2017,6 +2017,10 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)if (ring->outstanding_lazy_request)return 0 ; + ret =i915_gem_prepare_next_seqno (ring->dev); +if (ret) +return ret; + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;
@@ -2024,12 +2028,6 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring = ring; - ret =i915_gem_get_seqno (ring->dev, &request->seqno); -if (ret) { -kfree (request); -return ret; - } - ring->outstanding_lazy_request = request;return 0 ; }
Signature · 1 line
1.7.9.5
Re: [RFC 25/25] drm/i915: Defer seqno allocation until actual hardware submission time
Quoted 4 lines · On Fri, Oct 10, 2014 at 12:41:33PM +0100, [email protected] wrote:
From: John Harrison <[email protected]>
For: VIZ-4377
Signed-off-by [email protected]
Now I'm confused ... patch 3 made it sound like having the request and the seqno allocated at different points is a really fragile idea? Or is this now all save with everyone using struct request? Please elaborate.
I think the idea is solid, since with the scheduler we'll probably want to allocate the seqno even later (to avoid having to deal with out-of-order seqnos).
-Daniel
Quoted 151 lines
--- drivers/gpu/drm/i915/i915_drv.h | 5 ++++- drivers/gpu/drm/i915/i915_gem.c | 28 +++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_lrc.c | 10 ++++------ drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++------ 4 files changed, 39 insertions(+), 14 deletions(-)drivers/gpu/drm/i915/i915_drv.h +4
-1@@ -1979,6 +1979,9 @@ static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,if (req->complete)return true; +if (req->seqno ==0 ) +return false; +i915_gem_complete_requests_ring (req->ring, lazy_coherency);return req->complete;@@ -2482,7 +2485,7 @@ i915_seqno_passed(uint32_t seq1, uint32_t seq2)return (int32_t )(seq1 - seq2) >=0 ; } -int __must_check i915_gem_get_seqno (struct drm_device *dev,u32 *seqno); +int __must_check i915_gem_prepare_next_seqno (struct drm_device *dev);int __must_check i915_gem_set_seqno (struct drm_device *dev,u32 seqno);int __must_check i915_gem_object_get_fence (struct drm_i915_gem_object *obj);int __must_check i915_gem_object_put_fence (struct drm_i915_gem_object *obj);drivers/gpu/drm/i915/i915_gem.c +27
-1@@ -2310,12 +2310,15 @@ int i915_gem_set_seqno(struct drm_device *dev, u32 seqno) }int -i915_gem_get_seqno (struct drm_device *dev,u32 *seqno) +i915_gem_prepare_next_seqno (struct drm_device *dev) {struct drm_i915_private *dev_priv = dev->dev_private;/* reserve 0 for non-seqno */ if (dev_priv->next_seqno ==0 ) { +/* Why is the full re-initialisation required? Is it only for +* hardware semaphores? If so, could skip it in the case where +* semaphores are disabled? */ int ret =i915_gem_init_seqno (dev,0 );if (ret)return ret;@@ -2323,6 +2326,24 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno) dev_priv->next_seqno =1 ; } +return 0 ; +} + +static int +i915_gem_get_seqno (struct drm_device *dev,u32 *seqno) +{ +struct drm_i915_private *dev_priv = dev->dev_private; + +/* reserve 0 for non-seqno */ +if (dev_priv->next_seqno ==0 ) { +/* Should never get here! Must always call 'prepare_next' in +* advance. This code is called during request submission. +* Trying to wrap the seqno and the implicit idle() calls that +* the wrap code makes are a bad idea at this point! */ +DRM_ERROR ("Need to wrap seqno at inopportune moment!\n" ); +return -EBUSY ; + } + *seqno = dev_priv->last_seqno = dev_priv->next_seqno++;return 0 ; }@@ -2366,6 +2387,11 @@ int __i915_add_request(struct intel_engine_cs *ring,return ret; } +/* Assign an identifier to track this request through the hardware: */ + ret =i915_gem_get_seqno (ring->dev, &request->seqno); +if (ret) +return ret; +/* Record the position of the start of the request so that * should we detect the updated seqno part-way through the * GPU processing the request, we never over-estimate the drivers/gpu/drm/i915/intel_lrc.c +4
-6@@ -802,6 +802,10 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,if (ring->outstanding_lazy_request)return 0 ; + ret =i915_gem_prepare_next_seqno (ring->dev); +if (ret) +return ret; + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;@@ -809,12 +813,6 @@ static int logical_ring_alloc_request(struct intel_engine_cs *ring,kref_init (&request->ref); request->ring = ring; - ret =i915_gem_get_seqno (ring->dev, &request->seqno); -if (ret) { -kfree (request); -return ret; - } -/* Hold a reference to the context this request belongs to * (we will need it when the time comes to emit/retire the * request). drivers/gpu/drm/i915/intel_ringbuffer.c +4
-6@@ -2017,6 +2017,10 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)if (ring->outstanding_lazy_request)return 0 ; + ret =i915_gem_prepare_next_seqno (ring->dev); +if (ret) +return ret; + request =kzalloc (sizeof (*request),GFP_KERNEL );if (request ==NULL )return -ENOMEM ;@@ -2024,12 +2028,6 @@ intel_ring_alloc_request(struct intel_engine_cs *ring)kref_init (&request->ref); request->ring = ring; - ret =i915_gem_get_seqno (ring->dev, &request->seqno); -if (ret) { -kfree (request); -return ret; - } - ring->outstanding_lazy_request = request;return 0 ; }Signature · 6 lines
1.7.9.5 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
I have just posted an updated subset of the patch series. Note that one patch has been inserted in the middle and the first one has been dropped. The correct sequence is now:
01 drm/i915: Remove redundant parameter to
i915_gem_object_wait_rendering__tail()
02 drm/i915: Ensure OLS & PLR are always in sync 03 drm/i915: Add reference count to request structure 04 drm/i915: Add helper functions to aid seqno -> request transition 05 drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req 06 drm/i915: Ensure requests stick around during waits 07 drm/i915: Remove 'outstanding_lazy_seqno' 08 drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno 09 drm/i915: Convert 'last_flip_req' to be a request not a seqno 10 drm/i915: Convert i915_wait_seqno to i915_wait_request 11 drm/i915: Convert 'i915_add_request' to take a request not a seqno 12 drm/i915: Convert mmio_flip::seqno to struct request 13 drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' 14 drm/i915: Connect requests to rings at creation not submission 15 drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' 16 drm/i915: Convert __wait_seqno() to __wait_request() 17 drm/i915: Convert trace functions from seqno to request 18 drm/i915: Convert 'trace_irq' to use requests rather than seqnos 19 drm/i915: Convert semaphores to handle requests not seqnos 20 drm/i915: Convert 'ring_idle()' to use requests not seqnos 21 drm/i915: Remove 'obj->ring' 22 drm/i915: Cache request completion status 23 drm/i915: Zero fill the request structure 24 drm/i915: Defer seqno allocation until actual hardware
submission time
The whole set in its latest and greatest form has also been uploaded to the drm-private git as 'topic/seqno-request'.
Quoted 55 lines · On 06/10/2014 15:15, [email protected] wrote:
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
There is a general feeling that it is better to move away from using a simple integer 'seqno' value to track batch buffer completion. Instead, the request structure should be used. That provides for much more flexibility going forwards. Especially which things like a GPU scheduler (which can re-order batch buffers and hence seqnos after submission to the hardware), Android sync points and other such features which potentially make seqno usage more and more complex.
The current set of patches do most of the seqno to request structure conversion. There are still a couple of direct seqno comparisons in the semaphore code. The final conversion of a seqno test into a 'completed' flag inside the request structure is still do to as well. Along with whatever changes are required to maintain such a flag.
The patches are being posted now to make sure that the various people involved agree that it is heading in the right direction.
[Patches against drm-intel-nightly tree fetched 11/09/2014]
John Harrison (21):
Bug: missingi915_seqno_passed()call?
drm/i915: Remove redundant parameter toi915_gem_object_wait_rendering__tail()drm/i915: Ensure OLS & PLR are always in sync
drm/i915: Add reference count to request structure
drm/i915: Add helper functions to aid seqno -> request transition drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
drm/i915: Ensure requests stick around during waits
drm/i915: Remove 'outstanding_lazy_seqno'
drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno drm/i915: Convert 'last_flip_req' to be a request not a seqno drm/i915: Converti915_wait_seqnotoi915_wait_request
drm/i915: Convert 'i915_add_request' to take a request not a seqno drm/i915: Convertmmio_flip::seqnoto struct request
drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' drm/i915: Convert__wait_seqno()to__wait_request()
drm/i915: Convert trace functions from seqno to request
drm/i915: Convert 'trace_irq' to use requests rather than seqnos drm/i915: Convert semaphores to handle requests not seqnos
drm/i915: Convert 'ring_idle()' to use requests not seqnos
drm/i915: Remove 'obj->ring'drivers/gpu/drm/i915/i915_debugfs.c | 23 +-- drivers/gpu/drm/i915/i915_drv.h | 64 +++++-- drivers/gpu/drm/i915/i915_gem.c | 265 ++++++++++++++-------------- drivers/gpu/drm/i915/i915_gem_context.c | 12 +- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 +- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 19 +- drivers/gpu/drm/i915/i915_irq.c | 12 +-
Quoted 9 lines
drivers/gpu/drm/i915/i915_trace.h | 47 ++--- drivers/gpu/drm/i915/intel_display.c | 47 ++--- drivers/gpu/drm/i915/intel_drv.h | 4 +- drivers/gpu/drm/i915/intel_lrc.c | 49 ++--- drivers/gpu/drm/i915/intel_overlay.c | 17 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 88 +++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 24 ++- 16 files changed, 392 insertions(+), 293 deletions(-)
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 31 lines · John Harrison
I have just posted an updated subset of the patch series. Note that one patch has been inserted in the middle and the first one has been dropped. The correct sequence is now:
01 drm/i915: Remove redundant parameter to
i915_gem_object_wait_rendering__tail()02 drm/i915: Ensure OLS & PLR are always in sync 03 drm/i915: Add reference count to request structure 04 drm/i915: Add helper functions to aid seqno -> request transition 05 drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req 06 drm/i915: Ensure requests stick around during waits 07 drm/i915: Remove 'outstanding_lazy_seqno' 08 drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno 09 drm/i915: Convert 'last_flip_req' to be a request not a seqno 10 drm/i915: Convert i915_wait_seqno to i915_wait_request 11 drm/i915: Convert 'i915_add_request' to take a request not a seqno 12 drm/i915: Convert mmio_flip::seqno to struct request 13 drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' 14 drm/i915: Connect requests to rings at creation not submission 15 drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' 16 drm/i915: Convert __wait_seqno() to __wait_request() 17 drm/i915: Convert trace functions from seqno to request 18 drm/i915: Convert 'trace_irq' to use requests rather than seqnos 19 drm/i915: Convert semaphores to handle requests not seqnos 20 drm/i915: Convert 'ring_idle()' to use requests not seqnos 21 drm/i915: Remove 'obj->ring' 22 drm/i915: Cache request completion status 23 drm/i915: Zero fill the request structure 24 drm/i915: Defer seqno allocation until actual hardware
Quoted 5 lines
submission time
The whole set in its latest and greatest form has also been uploaded to the drm-private git as 'topic/seqno-request'.
Ok, read through the entire pile and looks good from a high level I think. Review summary is really just "please less BUG_ON and more commit message". I think even the few funky things can probably just be explained away with a good commit message.
That aside I think even for RFC the commit message really should be there, since that's what I need to be able to efficiently review patches. If the series is littered with FIXME comments all over that's ok (that's what an rfc is for really), but the justifications and tricky points should all be there. So I highly recommend to self-review patch series (using the review training slides) even for an rfc to make sure there's not too many bubbles in the pipeline.
Presuming there's no dragon hidden in any of the open issues I'll sign up someone else for the in-depth review.
Cheers, Daniel
Quoted 73 lines · On 06/10/2014 15:15, [email protected] wrote:
Quoted 55 lines
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
There is a general feeling that it is better to move away from using a simple integer 'seqno' value to track batch buffer completion. Instead, the request structure should be used. That provides for much more flexibility going forwards. Especially which things like a GPU scheduler (which can re-order batch buffers and hence seqnos after submission to the hardware), Android sync points and other such features which potentially make seqno usage more and more complex.
The current set of patches do most of the seqno to request structure conversion. There are still a couple of direct seqno comparisons in the semaphore code. The final conversion of a seqno test into a 'completed' flag inside the request structure is still do to as well. Along with whatever changes are required to maintain such a flag.
The patches are being posted now to make sure that the various people involved agree that it is heading in the right direction.
[Patches against drm-intel-nightly tree fetched 11/09/2014]
John Harrison (21):
Bug: missingi915_seqno_passed()call?
drm/i915: Remove redundant parameter toi915_gem_object_wait_rendering__tail()drm/i915: Ensure OLS & PLR are always in sync
drm/i915: Add reference count to request structure
drm/i915: Add helper functions to aid seqno -> request transition drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
drm/i915: Ensure requests stick around during waits
drm/i915: Remove 'outstanding_lazy_seqno'
drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno drm/i915: Convert 'last_flip_req' to be a request not a seqno drm/i915: Converti915_wait_seqnotoi915_wait_request
drm/i915: Convert 'i915_add_request' to take a request not a seqno drm/i915: Convertmmio_flip::seqnoto struct request
drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' drm/i915: Convert__wait_seqno()to__wait_request()
drm/i915: Convert trace functions from seqno to request
drm/i915: Convert 'trace_irq' to use requests rather than seqnos drm/i915: Convert semaphores to handle requests not seqnos
drm/i915: Convert 'ring_idle()' to use requests not seqnos
drm/i915: Remove 'obj->ring'drivers/gpu/drm/i915/i915_debugfs.c | 23 +-- drivers/gpu/drm/i915/i915_drv.h | 64 +++++-- drivers/gpu/drm/i915/i915_gem.c | 265 ++++++++++++++-------------- drivers/gpu/drm/i915/i915_gem_context.c | 12 +- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 +- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 19 +- drivers/gpu/drm/i915/i915_irq.c | 12 +-Quoted 9 lines
drivers/gpu/drm/i915/i915_trace.h | 47 ++--- drivers/gpu/drm/i915/intel_display.c | 47 ++--- drivers/gpu/drm/i915/intel_drv.h | 4 +- drivers/gpu/drm/i915/intel_lrc.c | 49 ++--- drivers/gpu/drm/i915/intel_overlay.c | 17 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 88 +++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 24 ++- 16 files changed, 392 insertions(+), 293 deletions(-)_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Quoted 35 lines · Daniel Vetter
Quoted 34 lines · John Harrison
I have just posted an updated subset of the patch series. Note that one patch has been inserted in the middle and the first one has been dropped. The correct sequence is now:
01 drm/i915: Remove redundant parameter to
i915_gem_object_wait_rendering__tail()
02 drm/i915: Ensure OLS & PLR are always in sync
03 drm/i915: Add reference count to request structure
04 drm/i915: Add helper functions to aid seqno -> request transition
05 drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
06 drm/i915: Ensure requests stick around during waits
07 drm/i915: Remove 'outstanding_lazy_seqno'
08 drm/i915: Make 'i915_gem_check_olr' actually check by request
not seqno
09 drm/i915: Convert 'last_flip_req' to be a request not a seqno
10 drm/i915: Converti915_wait_seqnotoi915_wait_request
11 drm/i915: Convert 'i915_add_request' to take a request not a seqno
12 drm/i915: Convertmmio_flip::seqnoto struct request
13 drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request'
14 drm/i915: Connect requests to rings at creation not submission
15 drm/i915: Convert most 'i915_seqno_passed' calls into
'i915_gem_request_completed'
16 drm/i915: Convert__wait_seqno()to__wait_request()
17 drm/i915: Convert trace functions from seqno to request
18 drm/i915: Convert 'trace_irq' to use requests rather than seqnos
19 drm/i915: Convert semaphores to handle requests not seqnos
20 drm/i915: Convert 'ring_idle()' to use requests not seqnos
21 drm/i915: Remove 'obj->ring'
22 drm/i915: Cache request completion status
23 drm/i915: Zero fill the request structure
24 drm/i915: Defer seqno allocation until actual hardware
submission time
Quoted 6 lines
Quoted 2 lines
The whole set in its latest and greatest form has also been uploaded to the drm-private git as 'topic/seqno-request'.
Ok, read through the entire pile and looks good from a high level I think. Review summary is really just "please less
BUG_ONand more commit message". I think even the few funky things can probably just be explained away with a good commit message.
A few of the BUG_ONs disappear again along the way and others are just converting existing BUG_ONs from seqnos to requests.
The minimal messaging was because the intention was to get something posted as soon as possible in order to be reviewed as soon as possible if only from a 'is this what you had in mind' point of view. I didn't realise you were going to be out of office for so long. There didn't seem much value in writing reams of description only to be told a day later that I've misunderstood your design spec and gone off in completely the wrong direction.
Quoted 84 lines
That aside I think even for RFC the commit message really should be there, since that's what I need to be able to efficiently review patches. If the series is littered with FIXME comments all over that's ok (that's what an rfc is for really), but the justifications and tricky points should all be there. So I highly recommend to self-review patch series (using the review training slides) even for an rfc to make sure there's not too many bubbles in the pipeline.
Presuming there's no dragon hidden in any of the open issues I'll sign up someone else for the in-depth review.
Cheers, Daniel
Quoted 70 lines · On 06/10/2014 15:15, [email protected] wrote:
Quoted 55 lines
From: John Harrison <[email protected]>
Work in progress for replacing seqno usage with requst structures.
There is a general feeling that it is better to move away from using a simple integer 'seqno' value to track batch buffer completion. Instead, the request structure should be used. That provides for much more flexibility going forwards. Especially which things like a GPU scheduler (which can re-order batch buffers and hence seqnos after submission to the hardware), Android sync points and other such features which potentially make seqno usage more and more complex.
The current set of patches do most of the seqno to request structure conversion. There are still a couple of direct seqno comparisons in the semaphore code. The final conversion of a seqno test into a 'completed' flag inside the request structure is still do to as well. Along with whatever changes are required to maintain such a flag.
The patches are being posted now to make sure that the various people involved agree that it is heading in the right direction.
[Patches against drm-intel-nightly tree fetched 11/09/2014]
John Harrison (21):
Bug: missingi915_seqno_passed()call?
drm/i915: Remove redundant parameter toi915_gem_object_wait_rendering__tail()drm/i915: Ensure OLS & PLR are always in sync
drm/i915: Add reference count to request structure
drm/i915: Add helper functions to aid seqno -> request transition drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req
drm/i915: Ensure requests stick around during waits
drm/i915: Remove 'outstanding_lazy_seqno'
drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno drm/i915: Convert 'last_flip_req' to be a request not a seqno drm/i915: Converti915_wait_seqnotoi915_wait_request
drm/i915: Convert 'i915_add_request' to take a request not a seqno drm/i915: Convertmmio_flip::seqnoto struct request
drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' drm/i915: Convert__wait_seqno()to__wait_request()
drm/i915: Convert trace functions from seqno to request
drm/i915: Convert 'trace_irq' to use requests rather than seqnos drm/i915: Convert semaphores to handle requests not seqnos
drm/i915: Convert 'ring_idle()' to use requests not seqnos
drm/i915: Remove 'obj->ring'drivers/gpu/drm/i915/i915_debugfs.c | 23 +-- drivers/gpu/drm/i915/i915_drv.h | 64 +++++-- drivers/gpu/drm/i915/i915_gem.c | 265 ++++++++++++++-------------- drivers/gpu/drm/i915/i915_gem_context.c | 12 +- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 +- drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 19 +- drivers/gpu/drm/i915/i915_irq.c | 12 +-Quoted 9 lines
drivers/gpu/drm/i915/i915_trace.h | 47 ++--- drivers/gpu/drm/i915/intel_display.c | 47 ++--- drivers/gpu/drm/i915/intel_drv.h | 4 +- drivers/gpu/drm/i915/intel_lrc.c | 49 ++--- drivers/gpu/drm/i915/intel_overlay.c | 17 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 88 +++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 24 ++- 16 files changed, 392 insertions(+), 293 deletions(-)_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 36 lines · John Harrison
Quoted 35 lines · Daniel Vetter
Quoted 34 lines · John Harrison
I have just posted an updated subset of the patch series. Note that one patch has been inserted in the middle and the first one has been dropped. The correct sequence is now:
01 drm/i915: Remove redundant parameter to
i915_gem_object_wait_rendering__tail()02 drm/i915: Ensure OLS & PLR are always in sync 03 drm/i915: Add reference count to request structure 04 drm/i915: Add helper functions to aid seqno -> request transition 05 drm/i915: Replace last_[rwf]_seqno with last_[rwf]_req 06 drm/i915: Ensure requests stick around during waits 07 drm/i915: Remove 'outstanding_lazy_seqno' 08 drm/i915: Make 'i915_gem_check_olr' actually check by request not seqno 09 drm/i915: Convert 'last_flip_req' to be a request not a seqno 10 drm/i915: Convert i915_wait_seqno to i915_wait_request 11 drm/i915: Convert 'i915_add_request' to take a request not a seqno 12 drm/i915: Convert mmio_flip::seqno to struct request 13 drm/i915: Convert 'flip_queued_seqno' into 'flip_queued_request' 14 drm/i915: Connect requests to rings at creation not submission 15 drm/i915: Convert most 'i915_seqno_passed' calls into 'i915_gem_request_completed' 16 drm/i915: Convert __wait_seqno() to __wait_request() 17 drm/i915: Convert trace functions from seqno to request 18 drm/i915: Convert 'trace_irq' to use requests rather than seqnos 19 drm/i915: Convert semaphores to handle requests not seqnos 20 drm/i915: Convert 'ring_idle()' to use requests not seqnos 21 drm/i915: Remove 'obj->ring' 22 drm/i915: Cache request completion status 23 drm/i915: Zero fill the request structure 24 drm/i915: Defer seqno allocation until actual hardwaresubmission time
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
The whole set in its latest and greatest form has also been uploaded to the drm-private git as 'topic/seqno-request'.
Ok, read through the entire pile and looks good from a high level I think. Review summary is really just "please less
BUG_ONand more commit message". I think even the few funky things can probably just be explained away with a good commit message.A few of the
BUG_ONsdisappear again along the way and others are just converting existingBUG_ONsfrom seqnos to requests.
Yeah, there's a bunch of preexisting ones that I've let slip through. But I've screamed at a BUG_ON that killed my machine once too often in the past few months, so I've started to be really strict about them. If they're just temporary then a WARN_ON should be about as informative, or just drop them since if a pointer is NULL you'll blow up anyway a few instructions later with an Oops.
Quoted 7 lines
The minimal messaging was because the intention was to get something posted as soon as possible in order to be reviewed as soon as possible if only from a 'is this what you had in mind' point of view. I didn't realise you were going to be out of office for so long. There didn't seem much value in writing reams of description only to be told a day later that I've misunderstood your design spec and gone off in completely the wrong direction.
Oh, I guess then I've looked too closely at it ;-) From the high-level point it's pretty much what I expected, so lgtm.
-Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch