Re: sched_setaffinity causing "rq->clock_update_flags < RQCF_ACT_SKIP" warning.

[Chris Wilson]

On Tue, Feb 28, 2017 at 01:43:02PM +0000, Chris Wilson wrote:
> On Mon, Feb 27, 2017 at 03:06:35PM +0000, Matt Fleming wrote:
> > On Fri, 24 Feb, at 05:19:03PM, Dave Jones wrote:
> > > Looks like fallout from cb42c9a3ebbbb23448c3f9a25417fae6309b1a92
> > > 
> > > WARNING: CPU: 1 PID: 561 at kernel/sched/sched.h:812 set_next_entity+0x11d/0x350
> > > rq->clock_update_flags < RQCF_ACT_SKIP
> > > CPU: 1 PID: 561 Comm: trinity-c36 Not tainted 4.10.0-think+ #6 
> > > Call Trace:
> > >  dump_stack+0x4f/0x73
> > >  __warn+0xcb/0xf0
> > >  warn_slowpath_fmt+0x5f/0x80
> > >  set_next_entity+0x11d/0x350
> > >  set_curr_task_fair+0x2b/0x60
> > >  do_set_cpus_allowed+0x138/0x180
> > >  __set_cpus_allowed_ptr+0x106/0x250
> > >  sched_setaffinity+0x24c/0x2b0
> > >  ? sched_setaffinity+0x5/0x2b0
> > >  SyS_sched_setaffinity+0x7a/0xc0
> > >  do_syscall_64+0x61/0x170
> > >  entry_SYSCALL64_slow_path+0x25/0x25
> > > RIP: 0033:0x7f461db010f9
> > > RSP: 002b:00007fff76d320a8 EFLAGS: 00000246
> > >  ORIG_RAX: 00000000000000cb
> > > RAX: ffffffffffffffda RBX: 00000000000000cb RCX: 00007f461db010f9
> > > RDX: 00007f461ba30000 RSI: 0000000000001000 RDI: 0000000000000234
> > > RBP: 00007f461e0d4000 R08: 00000000000000a6 R09: 00000000000000cf
> > > R10: fffffffffffffffb R11: 0000000000000246 R12: 0000000000000002
> > > R13: 00007f461e0d4048 R14: 00007f461e1d7ad8 R15: 00007f461e0d4000
> > 
> > Should be fixed here,
> > 
> >   https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?h=sched/urgent&id=a499c3ead88ccf147fc50689e85a530ad923ce36
> > 
> > It'll hopefully be sent to Linus soonish since it's in the
> > sched/urgent branch.
> 
> I pushed that 
> 	commit 7925851af123091a2590110e28ea268840ebd177
> 	Author: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
> 	Date:   Tue Feb 21 23:52:55 2017 -0800
> 
> 	    sched/fair: Update rq clock before changing a task's CPU affinity
> to our CI, and we still see the splat
> https://intel-gfx-ci.01.org/CI/CI_DRM_2251/fi-byt-n2820/igt@gem_exec_suspend@basic-s3.html
> 
> Anything else in that branch that might be the fix?

Ah,
sched/core: Fix update_rq_clock() splat on hotplug (and suspend/resume)
https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?h=sched/urgent&id=8cb68b343a66cf19834472012590490d34d31703
sounds likely.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre