| Message ID | 1456930109-21532-1-git-send-email-mika.kuoppala@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Mar 02, 2016 at 04:48:29PM +0200, Mika Kuoppala wrote: > With full-ppgtt, it takes the GPU an eon to traverse the entire 256PiB > address space, causing a loop to be detected. Under the current scheme, > if ACTHD walks off the end of a batch buffer and into an empty > address space, we "never" detect the hang. If we always increment the > score as the ACTHD is progressing then we will eventually timeout (after > ~46.5s (31 * 1.5s) without advancing onto a new batch). To counter act > this, increase the amount we reduce the score for good batches, so that > only a series of almost-bad batches trigger a full reset. DoS detection > suffers slightly but series of long running shader tests will benefit. > > Based on a patch from Chris Wilson. > > Testcase: igt/drv_hangman/hangcheck-unterminated > Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> > Cc: Chris Wilson <chris@chris-wilson.co.uk> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris
Patchwork <patchwork@emeril.freedesktop.org> writes: > == Series Details == > > Series: drm/i915/hangcheck: Prevent long walks across full-ppgtt > URL : https://patchwork.freedesktop.org/series/4023/ > State : warning > > == Summary == > > Series 4023v1 drm/i915/hangcheck: Prevent long walks across full-ppgtt > http://patchwork.freedesktop.org/api/1.0/series/4023/revisions/1/mbox/ > > Test drv_module_reload_basic: > pass -> DMESG-WARN (ilk-hp8440p) https://bugs.freedesktop.org/show_bug.cgi?id=94385 > Test kms_flip: > Subgroup basic-flip-vs-dpms: > pass -> DMESG-WARN (ilk-hp8440p) UNSTABLE > Subgroup basic-flip-vs-modeset: > pass -> INCOMPLETE (ilk-hp8440p) UNSTABLE > Test kms_force_connector_basic: > Subgroup force-load-detect: > skip -> PASS (ivb-t430s) > Test kms_pipe_crc_basic: > Subgroup nonblocking-crc-pipe-b-frame-sequence: > pass -> DMESG-WARN (snb-x220t) https://bugs.freedesktop.org/show_bug.cgi?id=94349 > dmesg-warn -> PASS (hsw-brixbox) > Subgroup suspend-read-crc-pipe-a: > incomplete -> PASS (hsw-gt2) > Subgroup suspend-read-crc-pipe-c: > dmesg-warn -> PASS (bsw-nuc-2) > Test pm_rpm: > Subgroup basic-rte: > pass -> DMESG-WARN (snb-dellxps) https://bugs.freedesktop.org/show_bug.cgi?id=94349 > > bdw-nuci7 total:169 pass:158 dwarn:0 dfail:0 fail:0 skip:11 > bdw-ultra total:169 pass:155 dwarn:0 dfail:0 fail:0 skip:14 > bsw-nuc-2 total:169 pass:138 dwarn:0 dfail:0 fail:1 skip:30 > byt-nuc total:169 pass:144 dwarn:0 dfail:0 fail:0 skip:25 > hsw-brixbox total:169 pass:154 dwarn:0 dfail:0 fail:0 skip:15 > hsw-gt2 total:169 pass:158 dwarn:1 dfail:0 fail:0 skip:10 > ilk-hp8440p total:156 pass:106 dwarn:2 dfail:0 fail:0 skip:47 > ivb-t430s total:169 pass:154 dwarn:0 dfail:0 fail:0 skip:15 > skl-i5k-2 total:169 pass:153 dwarn:0 dfail:0 fail:0 skip:16 > skl-i7k-2 total:169 pass:153 dwarn:0 dfail:0 fail:0 skip:16 > snb-dellxps total:169 pass:144 dwarn:2 dfail:0 fail:0 skip:23 > snb-x220t total:169 pass:144 dwarn:2 dfail:0 fail:1 skip:22 > > Results at /archive/results/CI_IGT_test/Patchwork_1517/ > > db506392f6706faffdc965c53c4cdea58cc16a02 drm-intel-nightly: 2016y-03m-02d-13h-47m-11s UTC integration manifest > 73a64b9e04a74b5bed5333823b5eebe930396689 drm/i915/hangcheck: Prevent long walks across full-ppgtt
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index a0f1bd711b53..15aacd0ee66f 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1367,8 +1367,6 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused) seq_printf(m, "\tACTHD = 0x%08llx [current 0x%08llx]\n", (long long)ring->hangcheck.acthd, (long long)acthd[i]); - seq_printf(m, "\tmax ACTHD = 0x%08llx\n", - (long long)ring->hangcheck.max_acthd); seq_printf(m, "\tscore = %d\n", ring->hangcheck.score); seq_printf(m, "\taction = %d\n", ring->hangcheck.action); diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 3b6bfbf35482..13b5f3aed01c 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -230,8 +230,6 @@ static const char *hangcheck_action_to_str(enum intel_ring_hangcheck_action a) return "wait"; case HANGCHECK_ACTIVE: return "active"; - case HANGCHECK_ACTIVE_LOOP: - return "active (loop)"; case HANGCHECK_KICK: return "kick"; case HANGCHECK_HUNG: diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index d1a46ef5ab3f..53e5104964b3 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3001,12 +3001,7 @@ head_stuck(struct intel_engine_cs *ring, u64 acthd) memset(ring->hangcheck.instdone, 0, sizeof(ring->hangcheck.instdone)); - if (acthd > ring->hangcheck.max_acthd) { - ring->hangcheck.max_acthd = acthd; - return HANGCHECK_ACTIVE; - } - - return HANGCHECK_ACTIVE_LOOP; + return HANGCHECK_ACTIVE; } if (!subunits_stuck(ring)) @@ -3083,6 +3078,7 @@ static void i915_hangcheck_elapsed(struct work_struct *work) #define BUSY 1 #define KICK 5 #define HUNG 20 +#define ACTIVE_DECAY 15 if (!i915.enable_hangcheck) return; @@ -3151,9 +3147,8 @@ static void i915_hangcheck_elapsed(struct work_struct *work) switch (ring->hangcheck.action) { case HANGCHECK_IDLE: case HANGCHECK_WAIT: - case HANGCHECK_ACTIVE: break; - case HANGCHECK_ACTIVE_LOOP: + case HANGCHECK_ACTIVE: ring->hangcheck.score += BUSY; break; case HANGCHECK_KICK: @@ -3172,10 +3167,12 @@ static void i915_hangcheck_elapsed(struct work_struct *work) * attempts across multiple batches. */ if (ring->hangcheck.score > 0) - ring->hangcheck.score--; + ring->hangcheck.score -= ACTIVE_DECAY; + if (ring->hangcheck.score < 0) + ring->hangcheck.score = 0; /* Clear head and subunit states on seqno movement */ - ring->hangcheck.acthd = ring->hangcheck.max_acthd = 0; + ring->hangcheck.acthd = 0; memset(ring->hangcheck.instdone, 0, sizeof(ring->hangcheck.instdone)); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index dd910d30a380..4b1439deb7fe 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -79,7 +79,6 @@ enum intel_ring_hangcheck_action { HANGCHECK_IDLE = 0, HANGCHECK_WAIT, HANGCHECK_ACTIVE, - HANGCHECK_ACTIVE_LOOP, HANGCHECK_KICK, HANGCHECK_HUNG, }; @@ -88,7 +87,6 @@ enum intel_ring_hangcheck_action { struct intel_ring_hangcheck { u64 acthd; - u64 max_acthd; u32 seqno; int score; enum intel_ring_hangcheck_action action;
With full-ppgtt, it takes the GPU an eon to traverse the entire 256PiB address space, causing a loop to be detected. Under the current scheme, if ACTHD walks off the end of a batch buffer and into an empty address space, we "never" detect the hang. If we always increment the score as the ACTHD is progressing then we will eventually timeout (after ~46.5s (31 * 1.5s) without advancing onto a new batch). To counter act this, increase the amount we reduce the score for good batches, so that only a series of almost-bad batches trigger a full reset. DoS detection suffers slightly but series of long running shader tests will benefit. Based on a patch from Chris Wilson. Testcase: igt/drv_hangman/hangcheck-unterminated Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> --- drivers/gpu/drm/i915/i915_debugfs.c | 2 -- drivers/gpu/drm/i915/i915_gpu_error.c | 2 -- drivers/gpu/drm/i915/i915_irq.c | 17 +++++++---------- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 -- 4 files changed, 7 insertions(+), 16 deletions(-)