'make -j16 universe' gives SIReset

[Marius Strobl]

On Thu, Jul 07, 2011 at 08:28:51AM +1000, Peter Jeremy wrote:
> On 2011-Jul-06 18:39:10 +0800, Marius Strobl <marius at alchemy.franken.de> wrote:
> >On Wed, Jul 06, 2011 at 02:26:34PM +1000, Peter Jeremy wrote:
> >> And DDB for one of the stuck processes shows
> >> db> trace 8881
> >> Tracing pid 8881 tid 195433 td 0xfffff8b0a2e72880
> >> mi_switch() at mi_switch+0x2a8
> >> sleepq_switch() at sleepq_switch+0x1cc
> >> sleepq_catch_signals() at sleepq_catch_signals+0x130
> >> sleepq_wait_sig() at sleepq_wait_sig+0x8
> >> _sleep() at _sleep+0x41c
> >> do_rw_rdlock() at do_rw_rdlock+0x7e4
> >> __umtx_op_rw_rdlock() at __umtx_op_rw_rdlock+0x1c
> >> _umtx_op() at _umtx_op+0x3c
> >> syscallenter() at syscallenter+0x270
> >> syscall() at syscall+0x74
> >> -- syscall (454, FreeBSD ELF64, _umtx_op) %o7=0x40479574 --
> >> userland() at 0x4047957c
> >> user trace: trap %o7=0x40479574
> >> pc 0x4047957c, sp 0x7fdffffc561
> >> pc 0x7fdffffd1c0, sp 0x40365a10
> >> pc 0x90000000000125a, sp 0xac00002d11220000
> >
> >What line does mi_switch+0x2a8 translate to?
> 
> 0xc0503628 <mi_switch+680>:     call  0xc0528ba0 <sched_switch>
>     448         sched_switch(td, newtd, flags);
> 
> The system is still running so I think the bigger issue is why
> none of the processes can grab the mutex.

Could you please give the below patch a try? This is just a shot in the
dark though.

> 
> >sparc64 package of gdb53, which still has the '-k' option:
> >http://people.freebsd.org/~marius/gdb-5.3_1%2c1.tbz
> 
> Unfortunately, it doesn't like me:
> 
> # gdb53 -k /usr/obj/usr/src/sys/GENERIC/kernel.debug /dev/mem
> GNU gdb 5.3 (FreeBSD)
> Copyright 2002 Free Software Foundation, Inc.
> GDB is free software, covered by the GNU General Public License, and you are
> welcome to change it and/or distribute copies of it under certain conditions.
> Type "show copying" to see the conditions.
> There is absolutely no warranty for GDB.  Type "show warranty" for details.
> This GDB was configured as "sparc64-unknown-freebsd9.0"...
> panic messages:
> ---
> dmesg: kvm_nlist: No such file or directory
> ---
> ---Can't read userspace from dump, or kernel process---
> 
> (kgdb) where
> ---Can't read userspace from dump, or kernel process---
> 
> (kgdb) disas mi_switch
> Segmentation fault (core dumped)
> 

FYI, as of r224687 I've basically unfucked thread debugging and kgdb(1)
on sparc64.

Marius

Index: lib/libthr/thread/thr_umtx.c
===================================================================
--- lib/libthr/thread/thr_umtx.c	(revision 223862)
+++ lib/libthr/thread/thr_umtx.c	(working copy)
@@ -152,13 +152,13 @@ __thr_umutex_timedlock(struct umutex *mtx, uint32_
 int
 __thr_umutex_unlock(struct umutex *mtx, uint32_t id)
 {
-#ifndef __ia64__
-	/* XXX this logic has a race-condition on ia64. */
+#if !defined(__ia64__) && !defined(__sparc64__)
+	/* XXX this logic has a race-condition on these architectures. */
 	if ((mtx->m_flags & (UMUTEX_PRIO_PROTECT | UMUTEX_PRIO_INHERIT)) == 0) {
 		atomic_cmpset_rel_32(&mtx->m_owner, id | UMUTEX_CONTESTED, UMUTEX_CONTESTED);
 		return _umtx_op_err(mtx, UMTX_OP_MUTEX_WAKE, 0, 0, 0);
 	}
-#endif /* __ia64__ */
+#endif /* !__ia64__ && !__sparc64__*/
 	return _umtx_op_err(mtx, UMTX_OP_MUTEX_UNLOCK, 0, 0, 0);
 }