aboutsummaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/i915_request.c
diff options
context:
space:
mode:
authorDave Airlie <airlied@redhat.com>2020-02-27 08:59:19 +1000
committerDave Airlie <airlied@redhat.com>2020-02-27 09:00:25 +1000
commit4825b61a3d39eceef7db723808103aa60fc24520 (patch)
tree87e1a754aef38ef088a5ec8e1613790c5a17c078 /drivers/gpu/drm/i915/i915_request.c
parentaaa9d265a21e7c4fcec12b1203cbfa516277e4ad (diff)
parent53e3ca6749186b5c147964bddc4eb47ba8b5f69e (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.c75
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;