[PATCH] drm/amdgpu: grab extra fence reference for drm_sched_job_add_dependency
Christian König
ckoenig.leichtzumerken at gmail.com
Thu Jan 5 10:03:14 UTC 2023
Am 05.01.23 um 02:44 schrieb Mikhail Gavrilov:
> On Tue, Jan 3, 2023 at 7:26 PM Alex Deucher <alexdeucher at gmail.com> wrote:
>> On Tue, Jan 3, 2023 at 3:34 AM Christian König
>> <ckoenig.leichtzumerken at gmail.com> wrote:
>>> I assume that this was already upstreamed while I was on sick leave?
>> Yes.
>>
>> Alex
>>
> What about commit 2fdb8a8f07c2f1353770a324fd19b8114e4329ac ?
That one should be fixed by:
commit 9f1ecfc5dcb47a7ca37be47b0eaca0f37f1ae93d
Author: Dmitry Osipenko <dmitry.osipenko at collabora.com>
Date: Wed Nov 23 03:13:03 2022 +0300
drm/scheduler: Fix lockup in drm_sched_entity_kill()
The drm_sched_entity_kill() is invoked twice by
drm_sched_entity_destroy()
while userspace process is exiting or being killed. First time it's
invoked
when sched entity is flushed and second time when entity is
released. This
causes a lockup within wait_for_completion(entity_idle) due to how
completion
API works.
Calling wait_for_completion() more times than complete() was
invoked is a
error condition that causes lockup because completion internally uses
counter for complete/wait calls. The complete_all() must be used
instead
in such cases.
This patch fixes lockup of Panfrost driver that is reproducible by
killing
any application in a middle of 3d drawing operation.
Fixes: 2fdb8a8f07c2 ("drm/scheduler: rework entity flush, kill and
fini")
Signed-off-by: Dmitry Osipenko <dmitry.osipenko at collabora.com>
Reviewed-by: Christian König <christian.koenig at amd.com>
Link:
https://patchwork.freedesktop.org/patch/msgid/20221123001303.533968-1-dmitry.osipenko@collabora.com
Regards,
Christian.
> I checked twice and I'm sure that this commit is the reason why I
> can't terminate some games (and others processes).
> Demonstration: https://youtu.be/O0AfjiMdFGw
> I also attached a full kernel log.
>
> INFO: task ZAT.exe:4745 blocked for more than 122 seconds.
> Tainted: G W L
> 6.1.0-rc1-13-2fdb8a8f07c2f1353770a324fd19b8114e4329ac+ #18
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:ZAT.exe state:D stack:12608 pid:4745 ppid:1 flags:0x20004006
> Call Trace:
> <TASK>
> __schedule+0x4c5/0x1740
> schedule+0x5d/0xe0
> schedule_timeout+0xf0/0x130
> __wait_for_common+0xa9/0x1f0
> ? usleep_range_state+0x90/0x90
> drm_sched_entity_kill.part.0+0x4d/0x210 [gpu_sched]
> drm_sched_entity_flush+0xa0/0x260 [gpu_sched]
> amdgpu_ctx_mgr_entity_flush+0x83/0xd0 [amdgpu]
> amdgpu_flush+0x25/0x40 [amdgpu]
> filp_close+0x31/0x70
> put_files_struct+0x78/0xf0
> do_exit+0x364/0xc30
> ? sched_clock_cpu+0xb/0xc0
> do_group_exit+0x33/0xa0
> get_signal+0xb41/0xb50
> arch_do_signal_or_restart+0x44/0x7a0
> exit_to_user_mode_prepare+0x17b/0x250
> syscall_exit_to_user_mode+0x16/0x50
> __do_fast_syscall_32+0x94/0xf0
> 2132]: Reached target exit.target - Exit the Session.
> 1]: user at 1000.service: Killing process 4402 (reaper) with signal SIGKILL.
> 1]: user at 1000.service: Killing process 4745 (ZAT.exe) with signal SIGKILL.
> 1]: Started plymouth-reboot.service - Show Plymouth Reboot Screen.
> : SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295
> subj=system_u:system_r:init_t:s0 msg='unit=plymouth-reboot
> comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=?
> terminal=? res=succe>
> 1]: plymouth-switch-root-initramfs.service - Tell Plymouth To Jump To
> initramfs was skipped because of an unmet condition check
> (ConditionPathExists=/run/initramfs/bin/sh).
> INFO: task ZAT.exe:4745 blocked for more than 122 seconds.
> Tainted: G W L
> 6.1.0-rc1-13-2fdb8a8f07c2f1353770a324fd19b8114e4329ac+ #18
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:ZAT.exe state:D stack:12608 pid:4745 ppid:1 flags:0x20004006
> Call Trace:
> <TASK>
> __schedule+0x4c5/0x1740
> schedule+0x5d/0xe0
> schedule_timeout+0xf0/0x130
> __wait_for_common+0xa9/0x1f0
> ? usleep_range_state+0x90/0x90
> drm_sched_entity_kill.part.0+0x4d/0x210 [gpu_sched]
> drm_sched_entity_flush+0xa0/0x260 [gpu_sched]
> amdgpu_ctx_mgr_entity_flush+0x83/0xd0 [amdgpu]
> amdgpu_flush+0x25/0x40 [amdgpu]
> filp_close+0x31/0x70
> put_files_struct+0x78/0xf0
> do_exit+0x364/0xc30
> ? sched_clock_cpu+0xb/0xc0
> do_group_exit+0x33/0xa0
> get_signal+0xb41/0xb50
> arch_do_signal_or_restart+0x44/0x7a0
> exit_to_user_mode_prepare+0x17b/0x250
> syscall_exit_to_user_mode+0x16/0x50
> __do_fast_syscall_32+0x94/0xf0
> ? __do_fast_syscall_32+0x94/0xf0
> ? lockdep_hardirqs_on+0x7d/0x100
> ? __do_fast_syscall_32+0x94/0xf0
> ? __do_fast_syscall_32+0x94/0xf0
> do_fast_syscall_32+0x2f/0x70
> entry_SYSCALL_compat_after_hwframe+0x62/0x6a
> RIP: 0023:0xf7f6b579
> RSP: 002b:00000000e8dffd40 EFLAGS: 00200282 ORIG_RAX: 00000000000000f0
> RAX: fffffffffffffe00 RBX: 00000000f0b54dcc RCX: 0000000000000189
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
> RBP: 00000000ffffffff R08: 00000000e8dffd40 R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000200282 R12: 0000000000000000
> R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
> </TASK>
>
> Showing all locks held in the system:
> 1 lock held by rcu_tasks_kthre/11:
> #0: ffffffffae368a20 (rcu_tasks.tasks_gp_mutex){+.+.}-{3:3}, at:
> rcu_tasks_one_gp+0x2b/0x3e0
> 1 lock held by rcu_tasks_rude_/12:
> #0: ffffffffae368760 (rcu_tasks_rude.tasks_gp_mutex){+.+.}-{3:3}, at:
> rcu_tasks_one_gp+0x2b/0x3e0
> 1 lock held by rcu_tasks_trace/13:
> #0: ffffffffae368460 (rcu_tasks_trace.tasks_gp_mutex){+.+.}-{3:3},
> at: rcu_tasks_one_gp+0x2b/0x3e0
> 1 lock held by khungtaskd/182:
> #0: ffffffffae369520 (rcu_read_lock){....}-{1:2}, at:
> debug_show_all_locks+0x15/0x16b
> 2 locks held by kworker/25:1/215:
> 1 lock held by systemd-journal/852:
> 1 lock held by ZAT.exe/4745:
> #0: ffff9b087c337cf8 (&mgr->lock#3){+.+.}-{3:3}, at:
> amdgpu_ctx_mgr_entity_flush+0x3a/0xd0 [amdgpu]
>
> =============================================
> 1]: user at 1000.service: Processes still around after final SIGKILL.
> Entering failed mode.
> 1]: user at 1000.service: Failed with result 'timeout'.
> 1]: Stopped user at 1000.service - User Manager for UID 1000.
>
>
More information about the amd-gfx
mailing list