svn commit: r278472 - in head/sys: netinet netinet6

John Baldwin jhb at freebsd.org
Mon Feb 9 20:34:40 UTC 2015


On Monday, February 09, 2015 07:28:12 PM Randall Stewart wrote:
> Author: rrs
> Date: Mon Feb  9 19:28:11 2015
> New Revision: 278472
> URL: https://svnweb.freebsd.org/changeset/base/278472
> 
> Log:
>   This fixes a bug in the way that the LLE timers for nd6
>   and arp were being used. They basically would pass in the
>   mutex to the callout_init. Because they used this method
>   to the callout system, it was possible to "stop" the callout.
>   When flushing the table and you stopped the running callout, the
>   callout_stop code would return 1 indicating that it was going
>   to stop the callout (that was about to run on the callout_wheel blocked
>   by the function calling the stop). Now when 1 was returned, it would
>   lower the reference count one extra time for the stopped timer, then
>   a few lines later delete the memory. Of course the callout_wheel was
>   stuck in the lock code and would then crash since it was accessing
>   freed memory. By using callout_init(c, 1) we always get a 0 back
>   and the reference counting bug does not rear its head. We do have
>   to make a few adjustments to the callouts themselves though to make
>   sure it does the proper thing if rescheduled as well as gets the lock.
> 
>   Commented upon by hiren and sbruno
>   See Phabricator D1777 for more details.
> 
>   Commented upon by hiren and sbruno
>   Reviewed by:	adrian, jhb and bz
>   Sponsored by:	Netflix Inc.

Eh, I looked at it, but I really, really don't like it.  I think 
callout_init_*() should be preferred to CALLOUT_MPSAFE whenever possible as it 
is less race-prone.  I think this should probably be fixed by adding Hans' 
callout_drain_async() instead, though this is fine as a temporary workaround.

-- 
John Baldwin


More information about the svn-src-head mailing list