diff options
author | Dave Airlie <airlied@redhat.com> | 2020-02-27 08:59:19 +1000 |
---|---|---|
committer | Dave Airlie <airlied@redhat.com> | 2020-02-27 09:00:25 +1000 |
commit | 4825b61a3d39eceef7db723808103aa60fc24520 (patch) | |
tree | 87e1a754aef38ef088a5ec8e1613790c5a17c078 /drivers/gpu/drm/i915/i915_request.c | |
parent | aaa9d265a21e7c4fcec12b1203cbfa516277e4ad (diff) | |
parent | 53e3ca6749186b5c147964bddc4eb47ba8b5f69e (diff) |
Merge tag 'drm-intel-next-2020-02-25' of git://anongit.freedesktop.org/drm/drm-intel into drm-next
- A backmerge of drm-next solving conflicts on i915/gt/intel_lrc.c
- Clean up shadow batch after I915_EXEC_SECURE
- Drop assertion that active->fence is unchanged
Here goes drm-intel-next-2020-02-25:
- A backmerge of drm-next solving conflicts on i915/gt/intel_lrc.c
- Clean up shadow batch after I915_EXEC_SECURE
- Drop assertion that active->fence is unchanged
drm-intel-next-2020-02-24-1:
- RC6 fixes - Chris
- Add extra slice common debug register - Lionel
- Align virtual engines uabi_class/instance with i915_drm.h - Tvrtko
- Avoid potential division by zero in computing CS timestamp - Chris
- Avoid using various globals - Michal Winiarski, Matt Auld
- Break up long lists of GEM object reclaim - Chris
- Check that the vma hasn't been closed before we insert it - Chris
- Consolidate SDVO HDMI force_dvi handling - Ville
- Conversion to new logging and warn macros and functions - Pankaj, Wambul, Chris
- DC3CO fixes - Jose
- Disable use of hwsp_cacheline for kernel_context - Chris
- Display IRQ pre/post uninstall refactor - Jani
- Display port sync refactor for robustness and fixes - Ville, Manasi
- Do not attempt to reprogram IA/ring frequencies for dgfx - Chris
- Drop alpha_support for good in favor of force_probe - Jani
- DSI ACPI related fixes and refactors - Vivek, Jani, Rajat
- Encoder refactor for flexibility to add more information, especiallly DSI related - Jani, Vandita
- Engine workarounds refactor for robustness around resue - Daniele
- FBC simplification and tracepoints
- Various fixes for build - Jani, Kees Cook, Chris, Zhang Xiaoxu
- Fix cmdparser - Chris
- Fix DRM_I915_GEM_MMAP_OFFFSET - Chris
- Fix i915_request flags - Chris
- Fix inconsistency between pfit enable and scaler freeing - Stanislav
- Fix inverted warn_on on display code - Chris
- Fix modeset locks in sanitize_watermarks - Ville
- Fix OA context id overlap with idle context id - Umesh
- Fix pipe and vblank enable for MST - Jani
- Fix VBT handling for timing parameters - Vandita
- Fixes o kernel doc - Chris, Ville
- Force full modeset whenever DSC is enabled at probe - Jani
- Various GEM locking simplification and fixes - Jani , Chris, Jose
- Including some changes in preparation for making GEM execbuf parallel - Chris
- Gen11 pcode error codes - Matt Roper
- Gen8+ interrupt handler refactor - Chris
- Many fixes and improvements around GuC code - Daniele, Michal Wajdeczko
- i915 parameters improvements sfor flexible input and better debugability - Chris, Jani
- Ice Lake and Elkhart Lake Fixes and workarounds - Matt Roper, Jose, Vivek, Matt Atwood
- Improvements on execlists, requests and other areas, fixing hangs and also
improving hang detection, recover and debugability - Chris
- Also introducing offline GT error capture - Chris
- Introduce encoder->compute_config_late() to help MST - Ville
- Make dbuf configuration const - Jani
- Few misc clean ups - Ville, Chris
- Never allow userptr into the new mapping types - Janusz
- Poison rings after use and GTT scratch pages - Chris
- Protect signaler walk with RCU - Chris
- PSR fixes - Jose
- Pull sseu context updates under gt - Chris
- Read rawclk_freq earlier - Chris
- Refactor around VBT handling to allow geting information through the encoder - Jani
- Refactor l3cc/mocs availability - Chris
- Refactor to use intel_connector over drm_connector - Ville
- Remove i915_energy_uJ from debugfs - Tvrtko
- Remove lite restore defines - Mika Kuoppala
- Remove prefault_disable modparam - Chris
- Many selftests fixes and improvements - Chris
- Set intel_dp_set_m_n() for MST slaves - Jose
- Simplify hot plug pin handling and other fixes around pin and polled modes - Ville
- Skip CPU synchronization on dma-buf attachments - chris
- Skip global serialization of clear_range for bxt vtd - Chris
- Skip rmw for marked register - Chris
- Some other GEM Fixes - Chris
- Some small changes for satisfying static code analysis - Colin, Chris
- Suppress warnings for unused debugging locals
- Tiger Lake enabling, including re-enable -f RPS, workarounds and other display fixes and changes - Chris, Matt Roper, Mika Kuoppala, Anshuman, Jose, Radhakrishna, Rafael.
- Track hw reported context runtime - Tvrtko
- Update bug filling URL - Jani
- Use async bind for PIN_USER into bsw/bxt ggtt - Chris
- Use the kernel_context to measuer the breadcrumb size - Chris
- Userptr fixes and robustness for big pages - Matt Auld
- Various Display refactors and clean-ups, specially around logs and use of drm_i915_private - Jani, Ville
- Various display refactors and fixes, especially around cdclk, modeset, and encoder - Chris, Jani
- Various eDP/DP fixes around DPCD - Lyude
- Various fixes and refactors for better Display watermark handling - Ville, Stanislav
- Various other display refactors - Ville
- Various refactor for better handling of display plane states - Ville
- Wean off drm_pci_alloc/drm_pci_free - Chris
- Correctly terminate connector iteration- Ville
- Downgrade gen7 (ivb, byt, hsw) back to aliasing-ppgtt - Chris
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200225185853.GA3282832@intel.com
Diffstat (limited to 'drivers/gpu/drm/i915/i915_request.c')
-rw-r--r-- | drivers/gpu/drm/i915/i915_request.c | 75 |
1 files changed, 42 insertions, 33 deletions
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 78a5f5d3c070..d53af93b919b 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -51,7 +51,6 @@ struct execute_cb { static struct i915_global_request { struct i915_global base; struct kmem_cache *slab_requests; - struct kmem_cache *slab_dependencies; struct kmem_cache *slab_execute_cbs; } global; @@ -203,6 +202,19 @@ static void free_capture_list(struct i915_request *request) } } +static void __i915_request_fill(struct i915_request *rq, u8 val) +{ + void *vaddr = rq->ring->vaddr; + u32 head; + + head = rq->infix; + if (rq->postfix < head) { + memset(vaddr + head, val, rq->ring->size - head); + head = 0; + } + memset(vaddr + head, val, rq->postfix - head); +} + static void remove_from_engine(struct i915_request *rq) { struct intel_engine_cs *engine, *locked; @@ -247,6 +259,9 @@ bool i915_request_retire(struct i915_request *rq) */ GEM_BUG_ON(!list_is_first(&rq->link, &i915_request_timeline(rq)->requests)); + if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) + /* Poison before we release our space in the ring */ + __i915_request_fill(rq, POISON_FREE); rq->ring->head = rq->postfix; /* @@ -595,6 +610,8 @@ static void __i915_request_ctor(void *arg) i915_sw_fence_init(&rq->submit, submit_notify); i915_sw_fence_init(&rq->semaphore, semaphore_notify); + dma_fence_init(&rq->fence, &i915_fence_ops, &rq->lock, 0, 0); + rq->file_priv = NULL; rq->capture_list = NULL; @@ -653,25 +670,30 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp) } } - ret = intel_timeline_get_seqno(tl, rq, &seqno); - if (ret) - goto err_free; - rq->i915 = ce->engine->i915; rq->context = ce; rq->engine = ce->engine; rq->ring = ce->ring; rq->execution_mask = ce->engine->mask; + kref_init(&rq->fence.refcount); + rq->fence.flags = 0; + rq->fence.error = 0; + INIT_LIST_HEAD(&rq->fence.cb_list); + + ret = intel_timeline_get_seqno(tl, rq, &seqno); + if (ret) + goto err_free; + + rq->fence.context = tl->fence_context; + rq->fence.seqno = seqno; + RCU_INIT_POINTER(rq->timeline, tl); RCU_INIT_POINTER(rq->hwsp_cacheline, tl->hwsp_cacheline); rq->hwsp_seqno = tl->hwsp_seqno; rq->rcustate = get_state_synchronize_rcu(); /* acts as smp_mb() */ - dma_fence_init(&rq->fence, &i915_fence_ops, &rq->lock, - tl->fence_context, seqno); - /* We bump the ref for the fence chain */ i915_sw_fence_reinit(&i915_request_get(rq)->submit); i915_sw_fence_reinit(&i915_request_get(rq)->semaphore); @@ -879,6 +901,12 @@ emit_semaphore_wait(struct i915_request *to, struct i915_request *from, gfp_t gfp) { + if (!intel_context_use_semaphores(to->context)) + goto await_fence; + + if (!rcu_access_pointer(from->hwsp_cacheline)) + goto await_fence; + /* Just emit the first semaphore we see as request space is limited. */ if (already_busywaiting(to) & from->engine->mask) goto await_fence; @@ -924,12 +952,8 @@ i915_request_await_request(struct i915_request *to, struct i915_request *from) ret = i915_sw_fence_await_sw_fence_gfp(&to->submit, &from->submit, I915_FENCE_GFP); - else if (intel_context_use_semaphores(to->context)) - ret = emit_semaphore_wait(to, from, I915_FENCE_GFP); else - ret = i915_sw_fence_await_dma_fence(&to->submit, - &from->fence, 0, - I915_FENCE_GFP); + ret = emit_semaphore_wait(to, from, I915_FENCE_GFP); if (ret < 0) return ret; @@ -1028,6 +1052,8 @@ __i915_request_await_execution(struct i915_request *to, { int err; + GEM_BUG_ON(intel_context_is_barrier(from->context)); + /* Submit both requests at the same time */ err = __await_execution(to, from, hook, I915_FENCE_GFP); if (err) @@ -1168,9 +1194,6 @@ i915_request_await_object(struct i915_request *to, void i915_request_skip(struct i915_request *rq, int error) { - void *vaddr = rq->ring->vaddr; - u32 head; - GEM_BUG_ON(!IS_ERR_VALUE((long)error)); dma_fence_set_error(&rq->fence, error); @@ -1182,12 +1205,7 @@ void i915_request_skip(struct i915_request *rq, int error) * context, clear out all the user operations leaving the * breadcrumb at the end (so we get the fence notifications). */ - head = rq->infix; - if (rq->postfix < head) { - memset(vaddr + head, 0, rq->ring->size - head); - head = 0; - } - memset(vaddr + head, 0, rq->postfix - head); + __i915_request_fill(rq, 0); rq->infix = rq->postfix; } @@ -1564,6 +1582,8 @@ long i915_request_wait(struct i915_request *rq, break; } + intel_engine_flush_submission(rq->engine); + if (signal_pending_state(state, current)) { timeout = -ERESTARTSYS; break; @@ -1574,7 +1594,6 @@ long i915_request_wait(struct i915_request *rq, break; } - intel_engine_flush_submission(rq->engine); timeout = io_schedule_timeout(timeout); } __set_current_state(TASK_RUNNING); @@ -1594,14 +1613,12 @@ out: static void i915_global_request_shrink(void) { - kmem_cache_shrink(global.slab_dependencies); kmem_cache_shrink(global.slab_execute_cbs); kmem_cache_shrink(global.slab_requests); } static void i915_global_request_exit(void) { - kmem_cache_destroy(global.slab_dependencies); kmem_cache_destroy(global.slab_execute_cbs); kmem_cache_destroy(global.slab_requests); } @@ -1631,17 +1648,9 @@ int __init i915_global_request_init(void) if (!global.slab_execute_cbs) goto err_requests; - global.slab_dependencies = KMEM_CACHE(i915_dependency, - SLAB_HWCACHE_ALIGN | - SLAB_RECLAIM_ACCOUNT); - if (!global.slab_dependencies) - goto err_execute_cbs; - i915_global_register(&global.base); return 0; -err_execute_cbs: - kmem_cache_destroy(global.slab_execute_cbs); err_requests: kmem_cache_destroy(global.slab_requests); return -ENOMEM; |