On Thu, May 17, 2018 at 10:07:34AM -0700, Conrad Meyer wrote: > On Wed, May 16, 2018 at 9:27 PM, Mark Johnston <ma...@freebsd.org> wrote: > > Author: markj > > Date: Thu May 17 04:27:08 2018 > > New Revision: 333703 > > URL: https://svnweb.freebsd.org/changeset/base/333703 > > > > Log: > > Fix a race in vm_page_pagequeue_lockptr(). > > > > The value of m->queue must be cached after comparing it with PQ_NONE, > > since it may be concurrently changing. > > > > Reported by: glebius > > What were the symptoms of this issue? The test plan in the linked > phabricator revision says: > > "Gleb reported seeing panics as a result of the use of a bogus index > into the pagequeue array, and also reported that this patch fixed the > panics." > > So an attempt to lock pagequeues[PQ_NONE=255].pq_mutex, which is > either something later in the vm_domain object, or bogus memory? One > of the mtx asserts trips?
I think it was "mtx_lock() of spin mutex"; I didn't get a lot of details. I failed to note in the commit message that this race was introduced in r332974. _______________________________________________ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
