Commit 23122a4d authored by Chris Wilson's avatar Chris Wilson

drm/i915/gt: Scrub execlists state on resume

Before we resume, we reset the HW so we restart from a known good state.
However, as a part of the reset process, we drain our pending CS event
queue -- and if we are resuming that does not correspond to internal
state. On setup, we are scrubbing the CS pointers, but alas only on
setup.

Apply the sanitization not just to setup, but to all resumes.
Reported-by: default avatarVenkata Ramana Nayana <venkata.ramana.nayana@intel.com>
Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
Cc: Venkata Ramana Nayana <venkata.ramana.nayana@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200416114117.3460-1-chris@chris-wilson.co.uk
parent 48b8b04c
...@@ -419,6 +419,7 @@ struct intel_engine_cs { ...@@ -419,6 +419,7 @@ struct intel_engine_cs {
void (*irq_enable)(struct intel_engine_cs *engine); void (*irq_enable)(struct intel_engine_cs *engine);
void (*irq_disable)(struct intel_engine_cs *engine); void (*irq_disable)(struct intel_engine_cs *engine);
void (*sanitize)(struct intel_engine_cs *engine);
int (*resume)(struct intel_engine_cs *engine); int (*resume)(struct intel_engine_cs *engine);
struct { struct {
......
...@@ -147,6 +147,10 @@ static void gt_sanitize(struct intel_gt *gt, bool force) ...@@ -147,6 +147,10 @@ static void gt_sanitize(struct intel_gt *gt, bool force)
if (intel_gt_is_wedged(gt)) if (intel_gt_is_wedged(gt))
intel_gt_unset_wedged(gt); intel_gt_unset_wedged(gt);
for_each_engine(engine, gt, id)
if (engine->sanitize)
engine->sanitize(engine);
intel_uc_sanitize(&gt->uc); intel_uc_sanitize(&gt->uc);
for_each_engine(engine, gt, id) for_each_engine(engine, gt, id)
......
...@@ -3615,6 +3615,43 @@ static int intel_init_workaround_bb(struct intel_engine_cs *engine) ...@@ -3615,6 +3615,43 @@ static int intel_init_workaround_bb(struct intel_engine_cs *engine)
return ret; return ret;
} }
static void reset_csb_pointers(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
const unsigned int reset_value = execlists->csb_size - 1;
ring_set_paused(engine, 0);
/*
* After a reset, the HW starts writing into CSB entry [0]. We
* therefore have to set our HEAD pointer back one entry so that
* the *first* entry we check is entry 0. To complicate this further,
* as we don't wait for the first interrupt after reset, we have to
* fake the HW write to point back to the last entry so that our
* inline comparison of our cached head position against the last HW
* write works even before the first interrupt.
*/
execlists->csb_head = reset_value;
WRITE_ONCE(*execlists->csb_write, reset_value);
wmb(); /* Make sure this is visible to HW (paranoia?) */
/*
* Sometimes Icelake forgets to reset its pointers on a GPU reset.
* Bludgeon them with a mmio update to be sure.
*/
ENGINE_WRITE(engine, RING_CONTEXT_STATUS_PTR,
reset_value << 8 | reset_value);
ENGINE_POSTING_READ(engine, RING_CONTEXT_STATUS_PTR);
invalidate_csb_entries(&execlists->csb_status[0],
&execlists->csb_status[reset_value]);
}
static void execlists_sanitize(struct intel_engine_cs *engine)
{
reset_csb_pointers(engine);
}
static void enable_error_interrupt(struct intel_engine_cs *engine) static void enable_error_interrupt(struct intel_engine_cs *engine)
{ {
u32 status; u32 status;
...@@ -3754,38 +3791,6 @@ static void execlists_reset_prepare(struct intel_engine_cs *engine) ...@@ -3754,38 +3791,6 @@ static void execlists_reset_prepare(struct intel_engine_cs *engine)
intel_engine_stop_cs(engine); intel_engine_stop_cs(engine);
} }
static void reset_csb_pointers(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
const unsigned int reset_value = execlists->csb_size - 1;
ring_set_paused(engine, 0);
/*
* After a reset, the HW starts writing into CSB entry [0]. We
* therefore have to set our HEAD pointer back one entry so that
* the *first* entry we check is entry 0. To complicate this further,
* as we don't wait for the first interrupt after reset, we have to
* fake the HW write to point back to the last entry so that our
* inline comparison of our cached head position against the last HW
* write works even before the first interrupt.
*/
execlists->csb_head = reset_value;
WRITE_ONCE(*execlists->csb_write, reset_value);
wmb(); /* Make sure this is visible to HW (paranoia?) */
/*
* Sometimes Icelake forgets to reset its pointers on a GPU reset.
* Bludgeon them with a mmio update to be sure.
*/
ENGINE_WRITE(engine, RING_CONTEXT_STATUS_PTR,
reset_value << 8 | reset_value);
ENGINE_POSTING_READ(engine, RING_CONTEXT_STATUS_PTR);
invalidate_csb_entries(&execlists->csb_status[0],
&execlists->csb_status[reset_value]);
}
static void __reset_stop_ring(u32 *regs, const struct intel_engine_cs *engine) static void __reset_stop_ring(u32 *regs, const struct intel_engine_cs *engine)
{ {
int x; int x;
...@@ -4545,6 +4550,7 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine) ...@@ -4545,6 +4550,7 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine)
{ {
/* Default vfuncs which can be overriden by each engine. */ /* Default vfuncs which can be overriden by each engine. */
engine->sanitize = execlists_sanitize;
engine->resume = execlists_resume; engine->resume = execlists_resume;
engine->cops = &execlists_context_ops; engine->cops = &execlists_context_ops;
...@@ -4659,8 +4665,6 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine) ...@@ -4659,8 +4665,6 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine)
else else
execlists->csb_size = GEN11_CSB_ENTRIES; execlists->csb_size = GEN11_CSB_ENTRIES;
reset_csb_pointers(engine);
/* Finally, take ownership and responsibility for cleanup! */ /* Finally, take ownership and responsibility for cleanup! */
engine->release = execlists_release; engine->release = execlists_release;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment