svn commit: r250953 - head/sys/cddl/contrib/opensolaris/uts/common/dtrace

[asomers at gmail.com]

Great timing.  I think that's only a few hours more recent than my
latest src checkout.

On Wed, Jul 3, 2013 at 11:35 AM, Mark Johnston <markj at freebsd.org> wrote:
> On Wed, Jul 3, 2013 at 1:28 PM, Alan Somers <asomers at freebsd.org> wrote:
>> This creates another panic on module unload when WITNESS is enabled,
>> because the module exits while holding the fasttrap_cleanup_mtx.  This
>> patch fixes the problem.  I'm not sure if the mtx_destroy() is
>> necessary, but I would feel dirty to leave it out.  Does this patch
>> look good to you?
>
> Hi Alan,
>
> Sorry about that - I noticed this mistake recently and fixed it in r252493:
> http://svnweb.freebsd.org/base?view=revision&revision=252493
>
> DTrace module unloading is now working properly for me with WITNESS enabled.
>
> Thanks,
> -Mark
>
>>
>> Index: sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c
>> ===================================================================
>> --- sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c   (revision 252490)
>> +++ sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c   (working copy)
>> @@ -2434,6 +2434,7 @@
>>         wakeup(&fasttrap_cleanup_cv);
>>         mtx_sleep(&fasttrap_cleanup_drain, &fasttrap_cleanup_mtx, 0, "ftcld",
>>             0);
>> +       mtx_unlock(&fasttrap_cleanup_mtx);
>>         fasttrap_cleanup_proc = NULL;
>>
>>  #ifdef DEBUG
>> @@ -2473,6 +2474,7 @@
>>  #if !defined(sun)
>>         destroy_dev(fasttrap_cdev);
>>         mutex_destroy(&fasttrap_count_mtx);
>> +       mtx_destroy(&fasttrap_cleanup_mtx);
>>         CPU_FOREACH(i) {
>>                 mutex_destroy(&fasttrap_cpuc_pid_lock[i]);
>>         }