drm/i915/execlists: Force write serialisation into context image vs execution

commit 0a823e8fd4fd67726697854578f3584ee3a49b1d upstream.

Ensure that the writes into the context image are completed prior to the
register mmio to trigger execution. Although previously we were assured
by the SDM that all writes are flushed before an uncached memory
transaction (our mmio write to submit the context to HW for execution),
we have empirical evidence to believe that this is not actually the
case.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108656
References: https://bugs.freedesktop.org/show_bug.cgi?id=108315
References: https://bugs.freedesktop.org/show_bug.cgi?id=106887
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181108081740.25615-1-chris@chris-wilson.co.uk
Cc: stable@vger.kernel.org
(cherry picked from commit 987abd5c62f92ee4970b45aa077f47949974e615)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index d638b641b76018a4c96fcfe78a5c013f36cf23ed..0775e71ea95b9121b596b6658d5b6c4153ca80ef 100644 (file)
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -325,7 +325,8 @@ static u64 execlists_update_context(struct drm_i915_gem_request *rq)
 
        reg_state[CTX_RING_TAIL+1] = intel_ring_set_tail(rq->ring, rq->tail);
 
-       /* True 32b PPGTT with dynamic page allocation: update PDP
+       /*
+        * True 32b PPGTT with dynamic page allocation: update PDP
         * registers and point the unallocated PDPs to scratch page.
         * PML4 is allocated during ppgtt init, so this is not needed
         * in 48-bit mode.
@@ -333,6 +334,17 @@ static u64 execlists_update_context(struct drm_i915_gem_request *rq)
        if (ppgtt && !i915_vm_is_48bit(&ppgtt->base))
                execlists_update_context_pdps(ppgtt, reg_state);
 
+       /*
+        * Make sure the context image is complete before we submit it to HW.
+        *
+        * Ostensibly, writes (including the WCB) should be flushed prior to
+        * an uncached write such as our mmio register access, the empirical
+        * evidence (esp. on Braswell) suggests that the WC write into memory
+        * may not be visible to the HW prior to the completion of the UC
+        * register write and that we may begin execution from the context
+        * before its image is complete leading to invalid PD chasing.
+        */
+       wmb();
        return ce->lrc_desc;
 }