* [PATCH] drm/i915: Cancel all ready but queued requests when wedging
@ 2017-09-15 16:49 Chris Wilson
2017-09-15 16:57 ` Chris Wilson
` (5 more replies)
0 siblings, 6 replies; 10+ messages in thread
From: Chris Wilson @ 2017-09-15 16:49 UTC (permalink / raw)
To: intel-gfx
When wedging the hw, we want to mark all in-flight requests as -EIO.
This is made slightly more complex by execlists who store the ready but
not yet submitted-to-hw requests on a private queue (an rbtree
priolist). Call into execlists to cancel not only the ELSP tracking for
the submitted requests, but also the queue of unsubmitted requests.
Testcase: igt/gem_eio/in-flight-contexts
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 33 ++++++----------------
drivers/gpu/drm/i915/intel_lrc.c | 49 +++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/intel_ringbuffer.h | 8 ++++++
3 files changed, 66 insertions(+), 24 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a5d1a4a9a011..625351792089 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3072,41 +3072,26 @@ static void engine_set_wedged(struct intel_engine_cs *engine)
*/
engine->submit_request = nop_submit_request;
- /* Mark all executing requests as skipped */
spin_lock_irqsave(&engine->timeline->lock, flags);
- list_for_each_entry(request, &engine->timeline->requests, link)
+
+ /* Mark all executing requests as skipped */
+ list_for_each_entry(request, &engine->timeline->requests, link) {
+ GEM_BUG_ON(!request->global_seqno);
if (!i915_gem_request_completed(request))
dma_fence_set_error(&request->fence, -EIO);
- spin_unlock_irqrestore(&engine->timeline->lock, flags);
+ }
/*
* Clear the execlists queue up before freeing the requests, as those
* are the ones that keep the context and ringbuffer backing objects
* pinned in place.
*/
+ if (engine->cancel_all_requests)
+ engine->cancel_all_requests(engine);
- if (i915.enable_execlists) {
- struct execlist_port *port = engine->execlist_port;
- unsigned long flags;
- unsigned int n;
+ /* Remaing pending _unready_ requests will be nop'ed when submitted */
- spin_lock_irqsave(&engine->timeline->lock, flags);
-
- for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++)
- i915_gem_request_put(port_request(&port[n]));
- memset(engine->execlist_port, 0, sizeof(engine->execlist_port));
- engine->execlist_queue = RB_ROOT;
- engine->execlist_first = NULL;
-
- spin_unlock_irqrestore(&engine->timeline->lock, flags);
-
- /* The port is checked prior to scheduling a tasklet, but
- * just in case we have suspended the tasklet to do the
- * wedging make sure that when it wakes, it decides there
- * is no work to do by clearing the irq_posted bit.
- */
- clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
- }
+ spin_unlock_irqrestore(&engine->timeline->lock, flags);
/* Mark all pending requests as complete so that any concurrent
* (lockless) lookup doesn't try and wait upon the request as we
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 3151a13b5bf9..5f57d9c4a522 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -506,6 +506,54 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
execlists_submit_ports(engine);
}
+static void execlists_cancel_all_requests(struct intel_engine_cs *engine)
+{
+ struct execlist_port *port = engine->execlist_port;
+ struct rb_node *rb;
+ unsigned long n;
+
+ lockdep_assert_held(&engine->timeline->lock);
+
+ /* Cancel the requests on the HW and clear the ELSP tracker. */
+ for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++)
+ i915_gem_request_put(port_request(&port[n]));
+ memset(engine->execlist_port, 0, sizeof(engine->execlist_port));
+
+ /* Flush the queued _ready_ requests to the executing lists. */
+ rb = engine->execlist_first;
+ while (rb) {
+ struct i915_priolist *p = rb_entry(rb, typeof(*p), node);
+ struct drm_i915_gem_request *rq, *rn;
+
+ list_for_each_entry_safe(rq, rn, &p->requests, priotree.link) {
+ INIT_LIST_HEAD(&rq->priotree.link);
+ rq->priotree.priority = INT_MAX;
+
+ dma_fence_set_error(&rq->fence, -EIO);
+ __i915_gem_request_submit(rq);
+ }
+
+ rb = rb_next(rb);
+ rb_erase(&p->node, &engine->execlist_queue);
+ INIT_LIST_HEAD(&p->requests);
+ if (p->priority != I915_PRIORITY_NORMAL)
+ kmem_cache_free(engine->i915->priorities, p);
+ }
+
+ engine->execlist_queue = RB_ROOT;
+ engine->execlist_first = NULL;
+
+ GEM_BUG_ON(port_isset(&port[0]));
+
+ /*
+ * The port is checked prior to scheduling a tasklet, but
+ * just in case we have suspended the tasklet to do the
+ * wedging make sure that when it wakes, it decides there
+ * is no work to do by clearing the irq_posted bit.
+ */
+ clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
+}
+
static bool execlists_elsp_ready(const struct intel_engine_cs *engine)
{
const struct execlist_port *port = engine->execlist_port;
@@ -1687,6 +1735,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
static void execlists_set_default_submission(struct intel_engine_cs *engine)
{
engine->submit_request = execlists_submit_request;
+ engine->cancel_all_requests = execlists_cancel_all_requests;
engine->schedule = execlists_schedule;
engine->irq_tasklet.func = intel_lrc_irq_handler;
}
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index abf171c3cb9c..04fc50c993bf 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -306,6 +306,14 @@ struct intel_engine_cs {
void (*schedule)(struct drm_i915_gem_request *request,
int priority);
+ /*
+ * Cancel all requests on the hardware, or queued for execution.
+ *
+ * This is called under the engine->timeline->lock when marking
+ * the device as wedged.
+ */
+ void (*cancel_all_requests)(struct intel_engine_cs *engine);
+
/* Some chipsets are not quite as coherent as advertised and need
* an expensive kick to force a true read of the up-to-date seqno.
* However, the up-to-date seqno is not always required and the last
--
2.14.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
@ 2017-09-15 16:57 ` Chris Wilson
2017-09-15 16:59 ` Chris Wilson
` (4 subsequent siblings)
5 siblings, 0 replies; 10+ messages in thread
From: Chris Wilson @ 2017-09-15 16:57 UTC (permalink / raw)
To: intel-gfx
Quoting Chris Wilson (2017-09-15 17:49:16)
> When wedging the hw, we want to mark all in-flight requests as -EIO.
> This is made slightly more complex by execlists who store the ready but
> not yet submitted-to-hw requests on a private queue (an rbtree
> priolist). Call into execlists to cancel not only the ELSP tracking for
> the submitted requests, but also the queue of unsubmitted requests.
>
Reported-by: Michał Winiarski <michal.winiarski@intel.com>
> Testcase: igt/gem_eio/in-flight-contexts
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Michał Winiarski <michal.winiarski@intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
2017-09-15 16:57 ` Chris Wilson
@ 2017-09-15 16:59 ` Chris Wilson
2017-09-15 17:16 ` Michał Winiarski
2017-09-15 17:07 ` ✓ Fi.CI.BAT: success for " Patchwork
` (3 subsequent siblings)
5 siblings, 1 reply; 10+ messages in thread
From: Chris Wilson @ 2017-09-15 16:59 UTC (permalink / raw)
To: intel-gfx
Quoting Chris Wilson (2017-09-15 17:49:16)
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> index abf171c3cb9c..04fc50c993bf 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -306,6 +306,14 @@ struct intel_engine_cs {
> void (*schedule)(struct drm_i915_gem_request *request,
> int priority);
>
> + /*
> + * Cancel all requests on the hardware, or queued for execution.
> + *
> + * This is called under the engine->timeline->lock when marking
> + * the device as wedged.
> + */
> + void (*cancel_all_requests)(struct intel_engine_cs *engine);
cancel_all_requests is a bit too broad, could just shorten it to
cancel_requests with the doc explaining that we only cancel the requests
that have been submitted to the engine (not the not-yet-ready requests
still floating in the aether).
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* ✓ Fi.CI.BAT: success for drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
2017-09-15 16:57 ` Chris Wilson
2017-09-15 16:59 ` Chris Wilson
@ 2017-09-15 17:07 ` Patchwork
2017-09-15 17:31 ` [PATCH v2] " Chris Wilson
` (2 subsequent siblings)
5 siblings, 0 replies; 10+ messages in thread
From: Patchwork @ 2017-09-15 17:07 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: drm/i915: Cancel all ready but queued requests when wedging
URL : https://patchwork.freedesktop.org/series/30434/
State : success
== Summary ==
Series 30434v1 drm/i915: Cancel all ready but queued requests when wedging
https://patchwork.freedesktop.org/api/1.0/series/30434/revisions/1/mbox/
Test pm_rpm:
Subgroup basic-rte:
pass -> DMESG-WARN (fi-cfl-s) fdo#102294
fdo#102294 https://bugs.freedesktop.org/show_bug.cgi?id=102294
fi-bdw-5557u total:289 pass:268 dwarn:0 dfail:0 fail:0 skip:21 time:449s
fi-bdw-gvtdvm total:289 pass:265 dwarn:0 dfail:0 fail:0 skip:24 time:457s
fi-blb-e6850 total:289 pass:224 dwarn:1 dfail:0 fail:0 skip:64 time:383s
fi-bsw-n3050 total:289 pass:243 dwarn:0 dfail:0 fail:0 skip:46 time:524s
fi-bwr-2160 total:289 pass:184 dwarn:0 dfail:0 fail:0 skip:105 time:269s
fi-bxt-j4205 total:289 pass:260 dwarn:0 dfail:0 fail:0 skip:29 time:511s
fi-byt-j1900 total:289 pass:254 dwarn:1 dfail:0 fail:0 skip:34 time:501s
fi-byt-n2820 total:289 pass:250 dwarn:1 dfail:0 fail:0 skip:38 time:500s
fi-cfl-s total:289 pass:222 dwarn:35 dfail:0 fail:0 skip:32 time:552s
fi-elk-e7500 total:289 pass:230 dwarn:0 dfail:0 fail:0 skip:59 time:425s
fi-glk-2a total:289 pass:260 dwarn:0 dfail:0 fail:0 skip:29 time:594s
fi-hsw-4770 total:289 pass:263 dwarn:0 dfail:0 fail:0 skip:26 time:425s
fi-hsw-4770r total:289 pass:263 dwarn:0 dfail:0 fail:0 skip:26 time:410s
fi-ilk-650 total:289 pass:229 dwarn:0 dfail:0 fail:0 skip:60 time:438s
fi-ivb-3520m total:289 pass:261 dwarn:0 dfail:0 fail:0 skip:28 time:489s
fi-ivb-3770 total:289 pass:261 dwarn:0 dfail:0 fail:0 skip:28 time:464s
fi-kbl-7500u total:289 pass:264 dwarn:1 dfail:0 fail:0 skip:24 time:489s
fi-kbl-7560u total:289 pass:270 dwarn:0 dfail:0 fail:0 skip:19 time:573s
fi-kbl-r total:289 pass:262 dwarn:0 dfail:0 fail:0 skip:27 time:590s
fi-pnv-d510 total:289 pass:223 dwarn:1 dfail:0 fail:0 skip:65 time:552s
fi-skl-6260u total:289 pass:269 dwarn:0 dfail:0 fail:0 skip:20 time:456s
fi-skl-6700k total:289 pass:265 dwarn:0 dfail:0 fail:0 skip:24 time:521s
fi-skl-6770hq total:289 pass:269 dwarn:0 dfail:0 fail:0 skip:20 time:496s
fi-skl-gvtdvm total:289 pass:266 dwarn:0 dfail:0 fail:0 skip:23 time:460s
fi-skl-x1585l total:289 pass:268 dwarn:0 dfail:0 fail:0 skip:21 time:474s
fi-snb-2520m total:289 pass:251 dwarn:0 dfail:0 fail:0 skip:38 time:573s
fi-snb-2600 total:289 pass:248 dwarn:0 dfail:0 fail:2 skip:39 time:425s
9adc9e93d6243c82bcefd175c2d11770802de194 drm-tip: 2017y-09m-15d-11h-44m-46s UTC integration manifest
427f54436a91 drm/i915: Cancel all ready but queued requests when wedging
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5714/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 16:59 ` Chris Wilson
@ 2017-09-15 17:16 ` Michał Winiarski
2017-09-15 17:19 ` Chris Wilson
0 siblings, 1 reply; 10+ messages in thread
From: Michał Winiarski @ 2017-09-15 17:16 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
On Fri, Sep 15, 2017 at 05:59:30PM +0100, Chris Wilson wrote:
> Quoting Chris Wilson (2017-09-15 17:49:16)
> > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> > index abf171c3cb9c..04fc50c993bf 100644
> > --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> > @@ -306,6 +306,14 @@ struct intel_engine_cs {
> > void (*schedule)(struct drm_i915_gem_request *request,
> > int priority);
> >
> > + /*
> > + * Cancel all requests on the hardware, or queued for execution.
> > + *
> > + * This is called under the engine->timeline->lock when marking
> > + * the device as wedged.
> > + */
> > + void (*cancel_all_requests)(struct intel_engine_cs *engine);
>
> cancel_all_requests is a bit too broad, could just shorten it to
> cancel_requests with the doc explaining that we only cancel the requests
> that have been submitted to the engine (not the not-yet-ready requests
> still floating in the aether).
> -Chris
I agree.
Note, that we're still doing part of the work directly in submit_notify:
/* Mark all executing requests as skipped */
list_for_each_entry(request, &engine->timeline->requests, link) {
GEM_BUG_ON(!request->global_seqno);
if (!i915_gem_request_completed(request))
dma_fence_set_error(&request->fence, -EIO);
}
Perhaps it would be cleaner if we could extract that (cancel_requests_inflight?)
and us it as a cancel_requests for legacy ringbuf? (then just call into that in
cancel_requests for execlists)
Thoughts?
-Michał
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 17:16 ` Michał Winiarski
@ 2017-09-15 17:19 ` Chris Wilson
0 siblings, 0 replies; 10+ messages in thread
From: Chris Wilson @ 2017-09-15 17:19 UTC (permalink / raw)
To: Michał Winiarski; +Cc: intel-gfx
Quoting Michał Winiarski (2017-09-15 18:16:16)
> On Fri, Sep 15, 2017 at 05:59:30PM +0100, Chris Wilson wrote:
> > Quoting Chris Wilson (2017-09-15 17:49:16)
> > > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> > > index abf171c3cb9c..04fc50c993bf 100644
> > > --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> > > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> > > @@ -306,6 +306,14 @@ struct intel_engine_cs {
> > > void (*schedule)(struct drm_i915_gem_request *request,
> > > int priority);
> > >
> > > + /*
> > > + * Cancel all requests on the hardware, or queued for execution.
> > > + *
> > > + * This is called under the engine->timeline->lock when marking
> > > + * the device as wedged.
> > > + */
> > > + void (*cancel_all_requests)(struct intel_engine_cs *engine);
> >
> > cancel_all_requests is a bit too broad, could just shorten it to
> > cancel_requests with the doc explaining that we only cancel the requests
> > that have been submitted to the engine (not the not-yet-ready requests
> > still floating in the aether).
> > -Chris
>
> I agree.
>
> Note, that we're still doing part of the work directly in submit_notify:
>
> /* Mark all executing requests as skipped */
> list_for_each_entry(request, &engine->timeline->requests, link) {
> GEM_BUG_ON(!request->global_seqno);
> if (!i915_gem_request_completed(request))
> dma_fence_set_error(&request->fence, -EIO);
> }
>
> Perhaps it would be cleaner if we could extract that (cancel_requests_inflight?)
> and us it as a cancel_requests for legacy ringbuf? (then just call into that in
> cancel_requests for execlists)
> Thoughts?
But there is also the third phase of catching all the requests not on
the timeline (or expected to be on the timeline). I guess we just move
the entire engine_set_wedged to the backends and if we cared enough
about the duplication, pull the list_for_each_entry out to a helper.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v2] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
` (2 preceding siblings ...)
2017-09-15 17:07 ` ✓ Fi.CI.BAT: success for " Patchwork
@ 2017-09-15 17:31 ` Chris Wilson
2017-09-15 18:13 ` Michał Winiarski
2017-09-15 17:52 ` ✓ Fi.CI.BAT: success for drm/i915: Cancel all ready but queued requests when wedging (rev2) Patchwork
2017-09-15 18:53 ` ✗ Fi.CI.IGT: warning " Patchwork
5 siblings, 1 reply; 10+ messages in thread
From: Chris Wilson @ 2017-09-15 17:31 UTC (permalink / raw)
To: intel-gfx
When wedging the hw, we want to mark all in-flight requests as -EIO.
This is made slightly more complex by execlists who store the ready but
not yet submitted-to-hw requests on a private queue (an rbtree
priolist). Call into execlists to cancel not only the ELSP tracking for
the submitted requests, but also the queue of unsubmitted requests.
v2: Move the majority of engine_set_wedged to the backends (both legacy
ringbuffer and execlists handling their own lists).
Reported-by: Michał Winiarski <michal.winiarski@intel.com>
Testcase: igt/gem_eio/in-flight-contexts
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 38 +--------------------
drivers/gpu/drm/i915/intel_lrc.c | 60 +++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/intel_ringbuffer.c | 20 +++++++++++
drivers/gpu/drm/i915/intel_ringbuffer.h | 8 +++++
4 files changed, 89 insertions(+), 37 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a5d1a4a9a011..ce37a54b29d4 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3061,9 +3061,6 @@ static void nop_submit_request(struct drm_i915_gem_request *request)
static void engine_set_wedged(struct intel_engine_cs *engine)
{
- struct drm_i915_gem_request *request;
- unsigned long flags;
-
/* We need to be sure that no thread is running the old callback as
* we install the nop handler (otherwise we would submit a request
* to hardware that will never complete). In order to prevent this
@@ -3073,40 +3070,7 @@ static void engine_set_wedged(struct intel_engine_cs *engine)
engine->submit_request = nop_submit_request;
/* Mark all executing requests as skipped */
- spin_lock_irqsave(&engine->timeline->lock, flags);
- list_for_each_entry(request, &engine->timeline->requests, link)
- if (!i915_gem_request_completed(request))
- dma_fence_set_error(&request->fence, -EIO);
- spin_unlock_irqrestore(&engine->timeline->lock, flags);
-
- /*
- * Clear the execlists queue up before freeing the requests, as those
- * are the ones that keep the context and ringbuffer backing objects
- * pinned in place.
- */
-
- if (i915.enable_execlists) {
- struct execlist_port *port = engine->execlist_port;
- unsigned long flags;
- unsigned int n;
-
- spin_lock_irqsave(&engine->timeline->lock, flags);
-
- for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++)
- i915_gem_request_put(port_request(&port[n]));
- memset(engine->execlist_port, 0, sizeof(engine->execlist_port));
- engine->execlist_queue = RB_ROOT;
- engine->execlist_first = NULL;
-
- spin_unlock_irqrestore(&engine->timeline->lock, flags);
-
- /* The port is checked prior to scheduling a tasklet, but
- * just in case we have suspended the tasklet to do the
- * wedging make sure that when it wakes, it decides there
- * is no work to do by clearing the irq_posted bit.
- */
- clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
- }
+ engine->cancel_requests(engine);
/* Mark all pending requests as complete so that any concurrent
* (lockless) lookup doesn't try and wait upon the request as we
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 3151a13b5bf9..2f9ebd32025c 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -506,6 +506,65 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
execlists_submit_ports(engine);
}
+static void execlists_cancel_requests(struct intel_engine_cs *engine)
+{
+ struct execlist_port *port = engine->execlist_port;
+ struct drm_i915_gem_request *rq, *rn;
+ struct rb_node *rb;
+ unsigned long flags;
+ unsigned long n;
+
+ spin_lock_irqsave(&engine->timeline->lock, flags);
+
+ /* Cancel the requests on the HW and clear the ELSP tracker. */
+ for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++)
+ i915_gem_request_put(port_request(&port[n]));
+ memset(engine->execlist_port, 0, sizeof(engine->execlist_port));
+
+ /* Mark all executing requests as skipped. */
+ list_for_each_entry(rq, &engine->timeline->requests, link) {
+ GEM_BUG_ON(!rq->global_seqno);
+ if (!i915_gem_request_completed(rq))
+ dma_fence_set_error(&rq->fence, -EIO);
+ }
+
+ /* Flush the queued requests to the timeline list (for retiring). */
+ rb = engine->execlist_first;
+ while (rb) {
+ struct i915_priolist *p = rb_entry(rb, typeof(*p), node);
+
+ list_for_each_entry_safe(rq, rn, &p->requests, priotree.link) {
+ INIT_LIST_HEAD(&rq->priotree.link);
+ rq->priotree.priority = INT_MAX;
+
+ dma_fence_set_error(&rq->fence, -EIO);
+ __i915_gem_request_submit(rq);
+ }
+
+ rb = rb_next(rb);
+ rb_erase(&p->node, &engine->execlist_queue);
+ INIT_LIST_HEAD(&p->requests);
+ if (p->priority != I915_PRIORITY_NORMAL)
+ kmem_cache_free(engine->i915->priorities, p);
+ }
+
+ /* Remaining _unready_ requests will be nop'ed when submitted */
+
+ engine->execlist_queue = RB_ROOT;
+ engine->execlist_first = NULL;
+ GEM_BUG_ON(port_isset(&port[0]));
+
+ /*
+ * The port is checked prior to scheduling a tasklet, but
+ * just in case we have suspended the tasklet to do the
+ * wedging make sure that when it wakes, it decides there
+ * is no work to do by clearing the irq_posted bit.
+ */
+ clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
+
+ spin_unlock_irqrestore(&engine->timeline->lock, flags);
+}
+
static bool execlists_elsp_ready(const struct intel_engine_cs *engine)
{
const struct execlist_port *port = engine->execlist_port;
@@ -1687,6 +1746,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
static void execlists_set_default_submission(struct intel_engine_cs *engine)
{
engine->submit_request = execlists_submit_request;
+ engine->cancel_requests = execlists_cancel_requests;
engine->schedule = execlists_schedule;
engine->irq_tasklet.func = intel_lrc_irq_handler;
}
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 7e13bfa7e1cd..fe0e2674700f 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -779,6 +779,24 @@ static u32 *gen6_signal(struct drm_i915_gem_request *req, u32 *cs)
return cs;
}
+static void cancel_requests(struct intel_engine_cs *engine)
+{
+ struct drm_i915_gem_request *request;
+ unsigned long flags;
+
+ spin_lock_irqsave(&engine->timeline->lock, flags);
+
+ /* Mark all submitted requests as skipped. */
+ list_for_each_entry(request, &engine->timeline->requests, link) {
+ GEM_BUG_ON(!request->global_seqno);
+ if (!i915_gem_request_completed(request))
+ dma_fence_set_error(&request->fence, -EIO);
+ }
+ /* Remaining _unready_ requests will be nop'ed when submitted */
+
+ spin_unlock_irqrestore(&engine->timeline->lock, flags);
+}
+
static void i9xx_submit_request(struct drm_i915_gem_request *request)
{
struct drm_i915_private *dev_priv = request->i915;
@@ -1999,11 +2017,13 @@ static void intel_ring_init_irq(struct drm_i915_private *dev_priv,
static void i9xx_set_default_submission(struct intel_engine_cs *engine)
{
engine->submit_request = i9xx_submit_request;
+ engine->cancel_requests = cancel_requests;
}
static void gen6_bsd_set_default_submission(struct intel_engine_cs *engine)
{
engine->submit_request = gen6_bsd_submit_request;
+ engine->cancel_requests = cancel_requests;
}
static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv,
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index abf171c3cb9c..138116a3b537 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -306,6 +306,14 @@ struct intel_engine_cs {
void (*schedule)(struct drm_i915_gem_request *request,
int priority);
+ /*
+ * Cancel all requests on the hardware, or queued for execution.
+ * This should only cancel the ready requests that have been
+ * submitted to the engine (via the engine->submit_request callback).
+ * This is called when marking the device as wedged.
+ */
+ void (*cancel_requests)(struct intel_engine_cs *engine);
+
/* Some chipsets are not quite as coherent as advertised and need
* an expensive kick to force a true read of the up-to-date seqno.
* However, the up-to-date seqno is not always required and the last
--
2.14.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 10+ messages in thread
* ✓ Fi.CI.BAT: success for drm/i915: Cancel all ready but queued requests when wedging (rev2)
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
` (3 preceding siblings ...)
2017-09-15 17:31 ` [PATCH v2] " Chris Wilson
@ 2017-09-15 17:52 ` Patchwork
2017-09-15 18:53 ` ✗ Fi.CI.IGT: warning " Patchwork
5 siblings, 0 replies; 10+ messages in thread
From: Patchwork @ 2017-09-15 17:52 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: drm/i915: Cancel all ready but queued requests when wedging (rev2)
URL : https://patchwork.freedesktop.org/series/30434/
State : success
== Summary ==
Series 30434v2 drm/i915: Cancel all ready but queued requests when wedging
https://patchwork.freedesktop.org/api/1.0/series/30434/revisions/2/mbox/
Test kms_cursor_legacy:
Subgroup basic-busy-flip-before-cursor-legacy:
fail -> PASS (fi-snb-2600) fdo#100215
Test kms_flip:
Subgroup basic-flip-vs-modeset:
skip -> PASS (fi-skl-x1585l) fdo#101781
fdo#100215 https://bugs.freedesktop.org/show_bug.cgi?id=100215
fdo#101781 https://bugs.freedesktop.org/show_bug.cgi?id=101781
fi-bdw-5557u total:289 pass:268 dwarn:0 dfail:0 fail:0 skip:21 time:442s
fi-bdw-gvtdvm total:289 pass:265 dwarn:0 dfail:0 fail:0 skip:24 time:452s
fi-blb-e6850 total:289 pass:224 dwarn:1 dfail:0 fail:0 skip:64 time:378s
fi-bsw-n3050 total:289 pass:243 dwarn:0 dfail:0 fail:0 skip:46 time:537s
fi-bwr-2160 total:289 pass:184 dwarn:0 dfail:0 fail:0 skip:105 time:269s
fi-bxt-j4205 total:289 pass:260 dwarn:0 dfail:0 fail:0 skip:29 time:506s
fi-byt-j1900 total:289 pass:254 dwarn:1 dfail:0 fail:0 skip:34 time:498s
fi-byt-n2820 total:289 pass:250 dwarn:1 dfail:0 fail:0 skip:38 time:502s
fi-cfl-s total:289 pass:223 dwarn:34 dfail:0 fail:0 skip:32 time:553s
fi-elk-e7500 total:289 pass:230 dwarn:0 dfail:0 fail:0 skip:59 time:425s
fi-glk-2a total:289 pass:260 dwarn:0 dfail:0 fail:0 skip:29 time:594s
fi-hsw-4770 total:289 pass:263 dwarn:0 dfail:0 fail:0 skip:26 time:429s
fi-hsw-4770r total:289 pass:263 dwarn:0 dfail:0 fail:0 skip:26 time:406s
fi-ilk-650 total:289 pass:229 dwarn:0 dfail:0 fail:0 skip:60 time:439s
fi-ivb-3520m total:289 pass:261 dwarn:0 dfail:0 fail:0 skip:28 time:484s
fi-ivb-3770 total:289 pass:261 dwarn:0 dfail:0 fail:0 skip:28 time:469s
fi-kbl-7500u total:289 pass:264 dwarn:1 dfail:0 fail:0 skip:24 time:490s
fi-kbl-7560u total:289 pass:270 dwarn:0 dfail:0 fail:0 skip:19 time:576s
fi-kbl-r total:289 pass:262 dwarn:0 dfail:0 fail:0 skip:27 time:584s
fi-pnv-d510 total:289 pass:223 dwarn:1 dfail:0 fail:0 skip:65 time:555s
fi-skl-6260u total:289 pass:269 dwarn:0 dfail:0 fail:0 skip:20 time:457s
fi-skl-6700k total:289 pass:265 dwarn:0 dfail:0 fail:0 skip:24 time:517s
fi-skl-6770hq total:289 pass:269 dwarn:0 dfail:0 fail:0 skip:20 time:501s
fi-skl-gvtdvm total:289 pass:266 dwarn:0 dfail:0 fail:0 skip:23 time:456s
fi-skl-x1585l total:289 pass:269 dwarn:0 dfail:0 fail:0 skip:20 time:492s
fi-snb-2520m total:289 pass:251 dwarn:0 dfail:0 fail:0 skip:38 time:568s
fi-snb-2600 total:289 pass:249 dwarn:0 dfail:0 fail:1 skip:39 time:428s
9adc9e93d6243c82bcefd175c2d11770802de194 drm-tip: 2017y-09m-15d-11h-44m-46s UTC integration manifest
f7a3dce2b390 drm/i915: Cancel all ready but queued requests when wedging
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5715/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v2] drm/i915: Cancel all ready but queued requests when wedging
2017-09-15 17:31 ` [PATCH v2] " Chris Wilson
@ 2017-09-15 18:13 ` Michał Winiarski
0 siblings, 0 replies; 10+ messages in thread
From: Michał Winiarski @ 2017-09-15 18:13 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
On Fri, Sep 15, 2017 at 06:31:00PM +0100, Chris Wilson wrote:
> When wedging the hw, we want to mark all in-flight requests as -EIO.
> This is made slightly more complex by execlists who store the ready but
> not yet submitted-to-hw requests on a private queue (an rbtree
> priolist). Call into execlists to cancel not only the ELSP tracking for
> the submitted requests, but also the queue of unsubmitted requests.
>
> v2: Move the majority of engine_set_wedged to the backends (both legacy
> ringbuffer and execlists handling their own lists).
>
> Reported-by: Michał Winiarski <michal.winiarski@intel.com>
> Testcase: igt/gem_eio/in-flight-contexts
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Michał Winiarski <michal.winiarski@intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
-Michał
> ---
> drivers/gpu/drm/i915/i915_gem.c | 38 +--------------------
> drivers/gpu/drm/i915/intel_lrc.c | 60 +++++++++++++++++++++++++++++++++
> drivers/gpu/drm/i915/intel_ringbuffer.c | 20 +++++++++++
> drivers/gpu/drm/i915/intel_ringbuffer.h | 8 +++++
> 4 files changed, 89 insertions(+), 37 deletions(-)
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
* ✗ Fi.CI.IGT: warning for drm/i915: Cancel all ready but queued requests when wedging (rev2)
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
` (4 preceding siblings ...)
2017-09-15 17:52 ` ✓ Fi.CI.BAT: success for drm/i915: Cancel all ready but queued requests when wedging (rev2) Patchwork
@ 2017-09-15 18:53 ` Patchwork
5 siblings, 0 replies; 10+ messages in thread
From: Patchwork @ 2017-09-15 18:53 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: drm/i915: Cancel all ready but queued requests when wedging (rev2)
URL : https://patchwork.freedesktop.org/series/30434/
State : warning
== Summary ==
Test perf:
Subgroup polling:
pass -> FAIL (shard-hsw) fdo#102252
Test kms_plane:
Subgroup plane-panning-bottom-right-suspend-pipe-A-planes:
pass -> SKIP (shard-hsw)
Test kms_busy:
Subgroup extended-modeset-hang-newfb-render-B:
pass -> SKIP (shard-hsw)
Test kms_plane_multiple:
Subgroup atomic-pipe-C-tiling-x:
pass -> SKIP (shard-hsw)
fdo#102252 https://bugs.freedesktop.org/show_bug.cgi?id=102252
shard-hsw total:2313 pass:1241 dwarn:0 dfail:0 fail:14 skip:1058 time:9340s
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5715/shards.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2017-09-15 18:53 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-09-15 16:49 [PATCH] drm/i915: Cancel all ready but queued requests when wedging Chris Wilson
2017-09-15 16:57 ` Chris Wilson
2017-09-15 16:59 ` Chris Wilson
2017-09-15 17:16 ` Michał Winiarski
2017-09-15 17:19 ` Chris Wilson
2017-09-15 17:07 ` ✓ Fi.CI.BAT: success for " Patchwork
2017-09-15 17:31 ` [PATCH v2] " Chris Wilson
2017-09-15 18:13 ` Michał Winiarski
2017-09-15 17:52 ` ✓ Fi.CI.BAT: success for drm/i915: Cancel all ready but queued requests when wedging (rev2) Patchwork
2017-09-15 18:53 ` ✗ Fi.CI.IGT: warning " Patchwork
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.