Triggering a GPU reset for one engine affects another, notably
corrupting the context status buffer (CSB) effectively losing track of
inflight requests.

Adding a few printks:
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index ad41836fa5e5..a969456bc0fa 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1953,6 +1953,7 @@ int i915_reset_engine(struct intel_engine_cs *engine)
                goto out;
        }

+       pr_err("Resetting %s\n", engine->name);
        ret = intel_gpu_reset(engine->i915, intel_engine_flag(engine));
        if (ret) {
                /* If we fail here, we expect to fallback to a global reset */
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 716e5c9ea222..a72bc35d0870 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -355,6 +355,7 @@ static void execlists_submit_ports(struct intel_engine_cs *engine)
                                execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
                        port_set(&port[n], port_pack(rq, count));
                        desc = execlists_update_context(rq);
+                       pr_err("%s: in (rq=%x) ctx=%d\n", engine->name, rq->global_seqno, upper_32_bits(desc));
                        GEM_DEBUG_EXEC(port[n].context_id = upper_32_bits(desc));
                } else {
                        GEM_BUG_ON(!n);
@@ -594,9 +595,23 @@ static void intel_lrc_irq_handler(unsigned long data)
                        if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK))
                                continue;

+                       pr_err("%s: out CSB (%x head=%d, tail=%d), ctx=%d, rq=%d\n",
+                                       engine->name,
+                                       readl(csb_mmio),
+                                       head, tail,
+                                       readl(buf+2*head+1),
+                                       port->context_id);
+
                        /* Check the context/desc id for this event matches */
-                       GEM_DEBUG_BUG_ON(readl(buf + 2 * head + 1) !=
-                                        port->context_id);
+                       if (readl(buf + 2 * head + 1) != port->context_id) {
+                               pr_err("%s: BUG CSB (%x head=%d, tail=%d), ctx=%d, rq=%d\n",
+                                               engine->name,
+                                               readl(csb_mmio),
+                                               head, tail,
+                                               readl(buf+2*head+1),
+                                               port->context_id);
+                               BUG();
+                       }

                        rq = port_unpack(port, &count);
                        GEM_BUG_ON(count == 0);

Results in:

[ 6423.006602] Resetting rcs0
[ 6423.009080] rcs0: in (rq=fffffe70) ctx=1
[ 6423.009216] rcs0: in (rq=fffffe6f) ctx=3
[ 6423.009542] rcs0: out CSB (2 head=1, tail=2), ctx=3, rq=3
[ 6423.009619] Resetting bcs0
[ 6423.009980] rcs0: BUG CSB (0 head=1, tail=2), ctx=0, rq=3

Note that this bug may be affect all machines and not just Broxton,
Broxton is just the first machine on which I have confirmed this bug.

Fixes: 142bc7d9 ("drm/i915: Modify error handler for per engine hang recovery")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Acked-by: Michel Thierry <michel.thierry@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170721123238.16428-13-chris@chris-wilson.co.ukSigned-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>