Re: LoR in slab allocator and swapper

On Sat, 3 Jul 2004, Jon Noack wrote:
> On 07/03/04 02:04, Nate Lawson wrote:
> > Just in case no one has reported this one:
> >
> > [...]
> > lock order reversal
> >  1st 0xc1a31108 vm object (vm object) _at_ /usr/src/sys/vm/swap_pager.c:1313
> >  2nd 0xc08c6300 swap_pager swhash (swap_pager swhash) _at_ /usr/src/sys/vm/swap_pager.c:1799
> >  3rd 0xc103fce4 vm object (vm object) _at_ /usr/src/sys/vm/uma_core.c:923
>
> Yeah, it's been reported a few dozen times.  It's a false positive and
> is harmless.  See #1 on the LOR status page:
> http://sources.zabbadoz.net/freebsd/lor.html
>
> We're working on getting a new FAQ entry for LORs
> (http://www.freebsd.org/cgi/query-pr.cgi?pr=68595) which will have a
> pointer to the LOR status page given above.  Once this is done, the idea
> is to update the lock order reversal message to point to this FAQ entry
> so we can better educate users on LORs and eliminate people reporting
> the same LOR tons of times.

Thanks, that's a good summary.  A better thing would be to add this to the
static WITNESS ordering so it knows that combination is ok.

-Nate