It is possible for a residual tasklet to be pending execution as we resume (whether that's some prior test kicking off the tasklet, or if we are in a suspend/resume stress test). As such, we do not want that tasklet to execute in the middle of our sanitization, such that it sees the poisoned state. For example, <4>[ 449.386553] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI <4>[ 449.386555] CPU: 1 PID: 5115 Comm: i915_selftest Tainted: G U W 5.7.0-rc4-CI-CI_DRM_8472+ #1 <4>[ 449.386556] Hardware name: Intel Corporation Ice Lake Client Platform/IceLake U DDR4 SODIMM PD RVP TLC, BIOS ICLSFWR1.R00.3183.A00.1905020411 05/02/2019 <4>[ 449.386585] RIP: 0010:process_csb+0x6bf/0x830 [i915] <4>[ 449.386588] Code: 00 48 c7 c2 10 bc 4c a0 48 c7 c7 d4 75 34 a0 e8 87 0e e6 e0 bf 01 00 00 00 e8 9d e0 e5 e0 31 f6 bf 09 00 00 00 e8 e1 ba d6 e0 <0f> 0b 8b 87 10 05 00 00 85 c0 0f 85 5f f9 ff ff 48 c7 c1 70 a5 4f <4>[ 449.386591] RSP: 0018:ffffc90000170ea0 EFLAGS: 00010297 <4>[ 449.386594] RAX: 0000000080000101 RBX: 0000000000000000 RCX: 0000000000000000 <4>[ 449.386596] RDX: ffff88849d5bc040 RSI: 0000000000000000 RDI: 0000000000000009 <4>[ 449.386598] RBP: ffffc90000170f00 R08: 0000000000000000 R09: 0000000000000000 <4>[ 449.386600] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88843ccea018 <4>[ 449.386602] R13: ffff88843ccea658 R14: ffff88843ccea640 R15: ffff88843ccea000 <4>[ 449.386605] FS: 00007f826a813300(0000) GS:ffff88849fe80000(0000) knlGS:0000000000000000 <4>[ 449.386607] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 <4>[ 449.386609] CR2: 0000560366b94280 CR3: 000000048ba02002 CR4: 0000000000760ee0 <4>[ 449.386611] PKRU: 55555554 <4>[ 449.386613] Call Trace: <4>[ 449.386616] <IRQ> <4>[ 449.386646] ? execlists_submission_tasklet+0xcf/0x140 [i915] <4>[ 449.386674] execlists_submission_tasklet+0x2f/0x140 [i915] <4>[ 449.386679] tasklet_action_common.isra.16+0x6c/0x1c0 <4>[ 449.386684] __do_softirq+0xdf/0x49e <4>[ 449.386687] irq_exit+0xba/0xc0 <4>[ 449.386690] smp_apic_timer_interrupt+0xb7/0x280 <4>[ 449.386693] apic_timer_interrupt+0xf/0x20 <4>[ 449.386695] </IRQ> <4>[ 449.386698] RIP: 0010:_raw_spin_unlock_irqrestore+0x49/0x60 <4>[ 449.386701] Code: c7 02 75 1f 53 9d e8 26 ab 75 ff bf 01 00 00 00 e8 7c a3 69 ff 65 8b 05 7d 9b 5c 7e 85 c0 74 0c 5b 5d c3 e8 09 aa 75 ff 53 9d <eb> df e8 ca 39 5b ff 5b 5d c3 0f 1f 00 66 2e 0f 1f 84 00 00 00 00 <4>[ 449.386703] RSP: 0018:ffffc90000a6b950 EFLAGS: 00000202 ORIG_RAX: ffffffffffffff13 <4>[ 449.386706] RAX: 0000000080000001 RBX: 0000000000000202 RCX: 0000000000000000 <4>[ 449.386708] RDX: ffff88849d5bc040 RSI: ffff88849d5bc900 RDI: ffffffff82386f12 <4>[ 449.386710] RBP: ffff88847d400f00 R08: ffff88849d5bc900 R09: 0000000000000000 <4>[ 449.386712] R10: 0000000000000000 R11: 0000000000000000 R12: 00000000ffff0b0b <4>[ 449.386714] R13: 000000000000000c R14: ffff88847d40bf70 R15: ffff88847d40cef8 <4>[ 449.386742] reset_csb_pointers+0x59/0x140 [i915] <4>[ 449.386769] execlists_sanitize+0x3e/0x60 [i915] <4>[ 449.386797] gt_sanitize+0xd6/0x260 [i915] As part of the reset preparation, engine->reset.prepare() prevents the tasklet from running, so pull the sanitization inside the critical section for reset. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1812 Fixes: 23122a4d992b ("drm/i915/gt: Scrub execlists state on resume") Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> --- drivers/gpu/drm/i915/gt/intel_gt_pm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c index e59776485457..6bdb434a442d 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c @@ -150,10 +150,6 @@ static void gt_sanitize(struct intel_gt *gt, bool force) if (intel_gt_is_wedged(gt)) intel_gt_unset_wedged(gt); - for_each_engine(engine, gt, id) - if (engine->sanitize) - engine->sanitize(engine); - intel_uc_sanitize(>->uc); for_each_engine(engine, gt, id) @@ -162,6 +158,10 @@ static void gt_sanitize(struct intel_gt *gt, bool force) intel_uc_reset_prepare(>->uc); + for_each_engine(engine, gt, id) + if (engine->sanitize) + engine->sanitize(engine); + if (reset_engines(gt) || force) { for_each_engine(engine, gt, id) __intel_engine_reset(engine, false); -- 2.20.1 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
