Quoting Mika Kuoppala (2018-05-15 09:08:45)
> Chris Wilson <ch...@chris-wilson.co.uk> writes:
>
> > We call i915_gem_reset_prepare_engine() during reset and then upon
> > wedging if the reset fails. Unfortunately, kthread_park and similar do
> > not support being called recursively and so we must count the number of
> > times we prepare for reset and only actually prepare on the outermost
> > layer. (Similarly for finish on unwinding the onion.)
> >
> > [ 87.705581] WARNING: CPU: 2 PID: 1377 at kernel/kthread.c:505
> > kthread_park+0x55/0x60
> > [ 87.705583] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek
> > snd_hda_codec_generic i915 x86_pkg_temp_thermal intel_powerclamp coretemp
> > crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_intel
> > snd_hda_codec snd_hwdep snd_hda_core snd_pcm broadcom bcm_phy_lib tg3
> > mei_me prime_numbers mei lpc_ich
> > [ 87.705618] CPU: 2 PID: 1377 Comm: gem_eio Tainted: G U
> > 4.17.0-rc5-CI-CI_DRM_4177+ #1
> > [ 87.705620] Hardware name: Dell Inc. XPS 8300 /0Y2MRG, BIOS A06
> > 10/17/2011
> > [ 87.705622] RIP: 0010:kthread_park+0x55/0x60
> > [ 87.705624] RSP: 0018:ffffc9000051bac0 EFLAGS: 00010202
> > [ 87.705627] RAX: 0000000000000004 RBX: ffff88021ca13de8 RCX:
> > 0000000000000001
> > [ 87.705629] RDX: 0000000080000001 RSI: ffffffff821228a9 RDI:
> > ffff88020e8f0040
> > [ 87.705630] RBP: ffff880215937670 R08: 00000000bae32d65 R09:
> > 0000000000000000
> > [ 87.705632] R10: 0000000000000000 R11: 0000000000000000 R12:
> > ffff8802159376b0
> > [ 87.705634] R13: ffff880215937670 R14: ffff880215930000 R15:
> > ffffffffa01c8d60
> > [ 87.705636] FS: 00007f0c32061980(0000) GS:ffff88022fa80000(0000)
> > knlGS:0000000000000000
> > [ 87.705637] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [ 87.705639] CR2: 00007f0c32094000 CR3: 000000021a0d4004 CR4:
> > 00000000000606e0
> > [ 87.705641] Call Trace:
> > [ 87.705668] i915_gem_reset_prepare_engine+0x1d/0xa0 [i915]
> > [ 87.705694] i915_gem_set_wedged+0x7b/0x1e0 [i915]
> > [ 87.705699] ? __drm_printfn_info+0x20/0x20
> > [ 87.705722] i915_reset+0x14a/0x290 [i915]
> > [ 87.705743] i915_reset_device+0x1fb/0x290 [i915]
> > [ 87.705767] ? __intel_get_crtc_scanline+0x1c0/0x1c0 [i915]
> > [ 87.705772] ? work_on_cpu_safe+0x50/0x50
> > [ 87.705798] i915_handle_error+0x207/0x4a0 [i915]
> > [ 87.705810] ? __might_fault+0x39/0x90
> > [ 87.705835] i915_wedged_set+0x7f/0xc0 [i915]
> > [ 87.705841] simple_attr_write+0xb0/0xd0
> > [ 87.705847] full_proxy_write+0x51/0x80
> > [ 87.705852] __vfs_write+0x31/0x160
> > [ 87.705857] ? rcu_read_lock_sched_held+0x6f/0x80
> > [ 87.705860] ? rcu_sync_lockdep_assert+0x29/0x50
> > [ 87.705862] ? __sb_start_write+0x152/0x1f0
> > [ 87.705864] ? __sb_start_write+0x168/0x1f0
> > [ 87.705868] vfs_write+0xbd/0x1a0
> > [ 87.705872] ksys_write+0x50/0xc0
> > [ 87.705877] do_syscall_64+0x55/0x190
> > [ 87.705880] entry_SYSCALL_64_after_hwframe+0x49/0xbe
> > [ 87.705882] RIP: 0033:0x7f0c315df281
> > [ 87.705884] RSP: 002b:00007ffc9c990328 EFLAGS: 00000246 ORIG_RAX:
> > 0000000000000001
> > [ 87.705887] RAX: ffffffffffffffda RBX: 0000000000000000 RCX:
> > 00007f0c315df281
> > [ 87.705889] RDX: 0000000000000002 RSI: 000055a5e23ef276 RDI:
> > 0000000000000047
> > [ 87.705890] RBP: 00007ffc9c990350 R08: 0000000000000000 R09:
> > 0000000000000034
> > [ 87.705892] R10: 0000000000000000 R11: 0000000000000246 R12:
> > 000055a5e23ebc50
> > [ 87.705894] R13: 00007ffc9c990dc0 R14: 0000000000000000 R15:
> > 0000000000000000
> > [ 87.705902] Code: 00 31 ed 48 39 c7 74 0e e8 79 db 00 00 48 8d 7b 18 e8
> > a0 05 88 00 89 e8 5b 5d c3 0f 0b bd da ff ff ff 89 e8 5b 5d c3 0f 0b eb b7
> > <0f> 0b bd f0 ff ff ff eb e2 66 90 41 57 41 56 49 c7 c6 f4 ff ff
> >
> > References: 85f1abe0019f ("kthread, sched/wait: Fix kthread_parkme()
> > completion issue")
> > Signed-off-by: Chris Wilson <ch...@chris-wilson.co.uk>
> > Cc: Mika Kuoppala <mika.kuopp...@linux.intel.com>
> > ---
> > drivers/gpu/drm/i915/i915_gem.c | 52 ++++++++++++++++++-------
> > drivers/gpu/drm/i915/intel_engine_cs.c | 1 +
> > drivers/gpu/drm/i915/intel_ringbuffer.h | 3 ++
> > 3 files changed, 43 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c
> > b/drivers/gpu/drm/i915/i915_gem.c
> > index 0a2070112b66..b169b630bf78 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -2997,14 +2997,11 @@ i915_gem_find_active_request(struct intel_engine_cs
> > *engine)
> > return active;
> > }
> >
> > -/*
> > - * Ensure irq handler finishes, and not run again.
> > - * Also return the active request so that we only search for it once.
> > - */
> > -struct i915_request *
> > -i915_gem_reset_prepare_engine(struct intel_engine_cs *engine)
> > +static void __i915_gem_reset_prepare_engine(struct intel_engine_cs *engine)
> > {
> > - struct i915_request *request = NULL;
> > + mutex_lock(&engine->reset_lock);
> > + if (engine->reset_depth++)
> > + goto unlock;
> >
> > /*
> > * During the reset sequence, we must prevent the engine from
> > @@ -3057,6 +3054,38 @@ i915_gem_reset_prepare_engine(struct intel_engine_cs
> > *engine)
> > if (engine->i915->guc.preempt_wq)
> > flush_workqueue(engine->i915->guc.preempt_wq);
> >
> > +unlock:
> > + mutex_unlock(&engine->reset_lock);
> > +}
> > +
> > +static void __i915_gem_reset_finish_engine(struct intel_engine_cs *engine)
> > +{
> > + mutex_lock(&engine->reset_lock);
> > +
> > + GEM_BUG_ON(engine->reset_depth);
>
> !engine->reset_depth
Hah. I think I prefer the patchset to kill kthread_park entirely. Hmm,
should also then include the removal of parking from the signaler for a
dramatic saving of one if().
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
