Re: Sleeping thread (tid 100033, pid 16): panic in FreeBSD 10.0-CURRENT/amd64 r228662

2011/12/20 John Baldwin <jhb_at_freebsd.org>:
> On Tuesday, December 20, 2011 9:20:09 am Attilio Rao wrote:
>> 2011/12/20 John Baldwin <jhb_at_freebsd.org>:
>> > On Saturday, December 17, 2011 10:41:15 pm mdf_at_freebsd.org wrote:
>> >> On Sat, Dec 17, 2011 at 5:45 PM, Alexander Kabaev <kabaev_at_gmail.com> wrote:
>> >> > On Sun, 18 Dec 2011 01:09:00 +0100
>> >> > "O. Hartmann" <ohartman_at_zedat.fu-berlin.de> wrote:
>> >> >
>> >> >> Sleeping thread (tid 100033, pid 16) owns a non sleepable lock
>> >> >> panic: sleeping thread
>> >> >> cpuid = 0
>> >> >>
>> >> >> PID 16 is always USB on my box.
>> >> >
>> >> > You really need to give us a backtrace when you quote panics. It is
>> >> > impossible to make any sense of the above panic message without more
>> >> > context.
>> >>
>> >> In the case of this panic, the stack of the thread which panics is
>> >> useless; it's someone trying to propagate priority that discovered it.
>> >>  A backtrace on tid 100033 would be useful.
>> >>
>> >> With WITNESS enabled, it's possible to have this panic display the
>> >> stack of the incorrectly sleeping thread at the time it acquired the
>> >> lock, as well, but this code isn't in CURRENT or any release.  I have
>> >> a patch at $WORK I can dig up on Monday.
>> >
>> > Huh?  The stock kernel dumps a stack trace of the offending thread if you have
>> > DDB enabled:
>> >
>> >                /*
>> >                 * If the thread is asleep, then we are probably about
>> >                 * to deadlock.  To make debugging this easier, just
>> >                 * panic and tell the user which thread misbehaved so
>> >                 * they can hopefully get a stack trace from the truly
>> >                 * misbehaving thread.
>> >                 */
>> >                if (TD_IS_SLEEPING(td)) {
>> >                        printf(
>> >                "Sleeping thread (tid %d, pid %d) owns a non-sleepable lock\n",
>> >                            td->td_tid, td->td_proc->p_pid);
>> > #ifdef DDB
>> >                        db_trace_thread(td, -1);
>> > #endif
>> >                        panic("sleeping thread");
>> >                }
>> >
>> > It may be that we can make use of the STACK API here instead to output this
>> > trace even when DDB isn't enabled.  The patch below tries to do that
>> > (untested).  It does some odd thigns though since it is effectively running
>> > from a panic context already, so it uses a statically allocated 'struct stack'
>> > rather than using stack_create() and uses stack_print_ddb() since it is
>> > holding spin locks and can't possibly grab an sx lock:
>> >
>> > Index: subr_turnstile.c
>> > ===================================================================
>> > --- subr_turnstile.c    (revision 228534)
>> > +++ subr_turnstile.c    (working copy)
>> > _at__at_ -72,6 +72,7 _at__at_ __FBSDID("$FreeBSD$");
>> >  #include <sys/proc.h>
>> >  #include <sys/queue.h>
>> >  #include <sys/sched.h>
>> > +#include <sys/stack.h>
>> >  #include <sys/sysctl.h>
>> >  #include <sys/turnstile.h>
>> >
>> > _at__at_ -175,6 +176,7 _at__at_ static void turnstile_fini(void *mem, int size);
>> >  static void
>> >  propagate_priority(struct thread *td)
>> >  {
>> > +       static struct stack st;
>> >        struct turnstile *ts;
>> >        int pri;
>> >
>> > _at__at_ -217,8 +219,10 _at__at_ propagate_priority(struct thread *td)
>> >                        printf(
>> >                "Sleeping thread (tid %d, pid %d) owns a non-sleepable lock\n",
>> >                            td->td_tid, td->td_proc->p_pid);
>> > -#ifdef DDB
>> > -                       db_trace_thread(td, -1);
>> > +#ifdef STACK
>> > +                       stack_zero(&st);
>> > +                       stack_save_td(&st, td);
>> > +                       stack_print_ddb(&st);
>> >  #endif
>> >                        panic("sleeping thread");
>> >                }
>> >
>> > --
>>
>> I'm not sure it is a wise idea to trimm the DDB part, because it is
>> much more common than STACK enabled. Note that stack(9) is working if
>> you define DDB too, so I'd say to do that for both.
>> Also, I don't think you need the stack_zero() prior to set it.
>
> Err, STACK is enabled in GENERIC in release kernels but DDB is not, so I think
> STACK is the more common one.  As far as stack_zero(), I was just being paranoid.

And what is the point for not having
#ifdef STACK
as
#if defined(STACK) || defined(DDB)

?

>> As we are here, however, I have a question for Robert here: do you
>> think we should support the _ddb() variant of options even in the case
>> DDB is not enabled in the kernel?
>> Probabilly the way it is nowadays is easier to have stack(9) both
>> defined for DDB and STACK, but in general I wouldn't advertise that.
>
> The _ddb variants are always enabled by my reading.  They just use different
> entry points into the linker that don't use locking.

My question is different: why we define them anyway even when DDB is
not enabled?

Attilio


-- 
Peace can only be achieved by understanding - A. Einstein