Re: 2.6.14-rc4-rt7

[Lee Revell <rlrevell@joe-job.com>]

On Mon, 2005-10-24 at 16:39 -0400, Steven Rostedt wrote:
> > And this when starting Hydrogen for the first time (the next startup is
> > fine):
> > 
> > hydrogen:4610 userspace BUG: scheduling in user-atomic context!
> >  [<c034fd9a>] schedule+0xea/0x100 (8)
> >  [<c034ff74>] wait_for_completion+0xa4/0xe0 (28)
> >  [<c011c150>] default_wake_function+0x0/0x20 (12)
> >  [<c0177951>] coredump_wait+0xa1/0x100 (28)
> >  [<c01ec60c>] copy_from_user+0x4c/0xc0 (8)
> >  [<c0177aaa>] do_coredump+0xfa/0x270 (108)
> >  [<c015456d>] kmem_cache_free+0x4d/0xb0 (40)
> >  [<c012aab5>] __dequeue_signal+0xf5/0x1c0 (24)
> >  [<c012aba3>] dequeue_signal+0x23/0xe0 (32)
> >  [<c012ca58>] get_signal_to_deliver+0x298/0x310 (20)
> >                [<c0351ee0>] do_page_fault+0x0/0x590 (24)
> >  [<c0102f80>] do_signal+0x70/0x180 (8)
> >  [<c014f495>] free_pages_bulk+0x225/0x2a0 (28)
> >  [<c0129493>] try_to_del_timer_sync+0x43/0x50 (12)
> >  [<c0147735>] audit_filter_syscall+0x45/0xe0 (4)
> >  [<c014834b>] audit_syscall_exit+0x4b/0x400 (36)
> >  [<c035219d>] do_page_fault+0x2bd/0x590 (40)
> >  [<c0351ee0>] do_page_fault+0x0/0x590 (48)
> >  [<c01030b5>] do_notify_resume+0x25/0x34 (8)
> >  [<c0103294>] work_notifysig+0x13/0x1b (8)
> > 
> > No other BUG messages that I can see.
> > -- Fernando
> 
> This part is new, I'll take a look into this.

Hydrogen segfaulted in the RT thread, and it was decided a while back
that an RT thread should lose RT priority when it coredumps - previously
we were seeing long latencies if the highest priority thread caught a
sig 11.  So this looks like a false positive in the userspace atomicity
debugger.

Lee