In order to avoid confusing the HW, we must never submit an empty ring
during lite-restore, that is we should always advance the RING_TAIL
before submitting to stay ahead of the RING_HEAD.
Normally this is prevented by keeping a couple of spare NOPs in the
request->wa_tail so that on resubmission we can advance the tail. This
relies on the request only being resubmitted once, which is the normal
condition as it is seen once for ELSP[1] and then later in ELSP[0]. On
preemption, the requests are unwound and the tail reset back to the
normal end point (as we know the request is incomplete and therefore its
RING_HEAD is even earlier).
However, if this w/a should fail we would try and resubmit the request
with the RING_TAIL already set to the location of this request's wa_tail
potentially causing a GPU hang. We can spot when we do try and
incorrectly resubmit without advancing the RING_TAIL and spare any
embarrassment by forcing the context restore.
In the case of preempt-to-busy, we leave the requests running on the HW
while we unwind. As the ring is still live, we cannot rewind our
rq->tail without forcing a reload so leave it set to rq->wa_tail and
only force a reload if we resubmit after a lite-restore. (Normally, the
forced reload will be a part of the preemption event.)
Fixes: 22b7a426bbe1 ("drm/i915/execlists: Preempt-to-busy")
Closes: https://gitlab.freedesktop.org/drm/intel/issues/673
Signed-off-by: Chris Wilson <chris(a)chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala(a)linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin(a)intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin(a)intel.com>
Cc: stable(a)vger.kernel.org
Link: https://patchwork.freedesktop.org/patch/msgid/20191209023215.3519970-1-chri…
(cherry picked from commit 82c69bf58650e644c61aa2bf5100b63a1070fd2f)
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 42 ++++++++++++++---------------
1 file changed, 20 insertions(+), 22 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 06a506c29463..6600b2e08fe3 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -471,12 +471,6 @@ lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine)
return desc;
}
-static void unwind_wa_tail(struct i915_request *rq)
-{
- rq->tail = intel_ring_wrap(rq->ring, rq->wa_tail - WA_TAIL_BYTES);
- assert_ring_tail_valid(rq->ring, rq->tail);
-}
-
static struct i915_request *
__unwind_incomplete_requests(struct intel_engine_cs *engine)
{
@@ -495,7 +489,6 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine)
continue; /* XXX */
__i915_request_unsubmit(rq);
- unwind_wa_tail(rq);
/*
* Push the request back into the queue for later resubmission.
@@ -649,13 +642,29 @@ execlists_schedule_out(struct i915_request *rq)
i915_request_put(rq);
}
-static u64 execlists_update_context(const struct i915_request *rq)
+static u64 execlists_update_context(struct i915_request *rq)
{
struct intel_context *ce = rq->hw_context;
- u64 desc;
+ u64 desc = ce->lrc_desc;
+ u32 tail;
- ce->lrc_reg_state[CTX_RING_TAIL + 1] =
- intel_ring_set_tail(rq->ring, rq->tail);
+ /*
+ * WaIdleLiteRestore:bdw,skl
+ *
+ * We should never submit the context with the same RING_TAIL twice
+ * just in case we submit an empty ring, which confuses the HW.
+ *
+ * We append a couple of NOOPs (gen8_emit_wa_tail) after the end of
+ * the normal request to be able to always advance the RING_TAIL on
+ * subsequent resubmissions (for lite restore). Should that fail us,
+ * and we try and submit the same tail again, force the context
+ * reload.
+ */
+ tail = intel_ring_set_tail(rq->ring, rq->tail);
+ if (unlikely(ce->lrc_reg_state[CTX_RING_TAIL + 1] == tail))
+ desc |= CTX_DESC_FORCE_RESTORE;
+ ce->lrc_reg_state[CTX_RING_TAIL + 1] = tail;
+ rq->tail = rq->wa_tail;
/*
* Make sure the context image is complete before we submit it to HW.
@@ -674,7 +683,6 @@ static u64 execlists_update_context(const struct i915_request *rq)
*/
mb();
- desc = ce->lrc_desc;
ce->lrc_desc &= ~CTX_DESC_FORCE_RESTORE;
return desc;
@@ -1149,16 +1157,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
if (!list_is_last(&last->sched.link,
&engine->active.requests))
return;
-
- /*
- * WaIdleLiteRestore:bdw,skl
- * Apply the wa NOOPs to prevent
- * ring:HEAD == rq:TAIL as we resubmit the
- * request. See gen8_emit_fini_breadcrumb() for
- * where we prepare the padding after the
- * end of the request.
- */
- last->tail = last->wa_tail;
}
}
--
2.25.0.rc0
link_startup_again flow in ufshcd_link_startup() is not necessary
since currently device can be moved to "active" power mode during
resume flow by commit as below,
"scsi: ufs: set device as default active power mode during
initialization only"
Fixes: 7caf489b99a4 (scsi: ufs: issue link starup 2 times if device isn't active)
Cc: Alim Akhtar <alim.akhtar(a)samsung.com>
Cc: Avri Altman <avri.altman(a)wdc.com>
Cc: Bart Van Assche <bvanassche(a)acm.org>
Cc: Bean Huo <beanhuo(a)micron.com>
Cc: Can Guo <cang(a)codeaurora.org>
Cc: Matthias Brugger <matthias.bgg(a)gmail.com>
Cc: Subhash Jadavani <subhashj(a)codeaurora.org>
Cc: stable(a)vger.kernel.org
Signed-off-by: Stanley Chu <stanley.chu(a)mediatek.com>
---
drivers/scsi/ufs/ufshcd.c | 15 ---------------
1 file changed, 15 deletions(-)
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 9abb7085a5d0..1900f811394a 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -4365,16 +4365,7 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
{
int ret;
int retries = DME_LINKSTARTUP_RETRIES;
- bool link_startup_again = false;
- /*
- * If UFS device isn't active then we will have to issue link startup
- * 2 times to make sure the device state move to active.
- */
- if (!ufshcd_is_ufs_dev_active(hba))
- link_startup_again = true;
-
-link_startup:
do {
ufshcd_vops_link_startup_notify(hba, PRE_CHANGE);
@@ -4408,12 +4399,6 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
goto out;
}
- if (link_startup_again) {
- link_startup_again = false;
- retries = DME_LINKSTARTUP_RETRIES;
- goto link_startup;
- }
-
/* Mark that link is up in PWM-G1, 1-lane, SLOW-AUTO mode */
ufshcd_init_pwr_info(hba);
ufshcd_print_pwr_info(hba);
--
2.18.0