svn commit: r202889 - head/sys/kern

Rob Farmer rfarmer at
Tue Jan 26 06:52:29 UTC 2010

On Sat, Jan 23, 2010 at 7:54 AM, Attilio Rao <attilio at> wrote:
> Author: attilio
> Date: Sat Jan 23 15:54:21 2010
> New Revision: 202889
> URL:
> Log:
>  - Fix a race in sched_switch() of sched_4bsd.
>    In the case of the thread being on a sleepqueue or a turnstile, the
>    sched_lock was acquired (without the aid of the td_lock interface) and
>    the td_lock was dropped. This was going to break locking rules on other
>    threads willing to access to the thread (via the td_lock interface) and
>    modify his flags (allowed as long as the container lock was different
>    by the one used in sched_switch).
>    In order to prevent this situation, while sched_lock is acquired there
>    the td_lock gets blocked. [0]
>  - Merge the ULE's internal function thread_block_switch() into the global
>    thread_lock_block() and make the former semantic as the default for
>    thread_lock_block(). This means that thread_lock_block() will not
>    disable interrupts when called (and consequently thread_unlock_block()
>    will not re-enabled them when called). This should be done manually
>    when necessary.
>    Note, however, that ULE's thread_unblock_switch() is not reaped
>    because it does reflect a difference in semantic due in ULE (the
>    td_lock may not be necessarilly still blocked_lock when calling this).
>    While asymmetric, it does describe a remarkable difference in semantic
>    that is good to keep in mind.
>  [0] Reported by:      Kohji Okuno
>                        <okuno dot kohji at jp dot panasonic dot com>
>  Tested by:            Giovanni Trematerra
>                        <giovanni dot trematerra at gmail dot com>
>  MFC:                  2 weeks
> Modified:
>  head/sys/kern/kern_mutex.c
>  head/sys/kern/sched_4bsd.c
>  head/sys/kern/sched_ule.c


This commit seems to be causing me a kernel panic on sparc64 - details
are in PR 143215. Could you take a look before MFCing this?

Rob Farmer

More information about the svn-src-all mailing list