On Thu, Jun 28, 2018 at 3:24 PM, Eric Covener <cove...@gmail.com> wrote: > On Thu, Jun 28, 2018 at 9:18 AM Yann Ylavic <ylavic....@gmail.com> wrote: >> >> On Wed, Jun 27, 2018 at 11:00 PM, <jaillet...@apache.org> wrote: >> > >> > + *) event: based on the commit log ("use a subpool of pchild since >> > skiplist >> > + allocations will happen across threads and are only protected from >> > + other skiplist operations.", per sf's review comments), it sounds >> > like this >> > + should be backported. >> > + trunk patch: http://svn.apache.org/r1588806 >> > + 2.4.x patch: >> > http://home.apache.org/~jailletc36/backport_1588806.patch >> > + +1: >> >> Hmm, timer_skiplist seems to be used only when g_timer_skiplist_mtx is >> locked, am I missing something? > > IIUC the risk is colliding with someone else using pchild w/ similar > locking around just their own use of it.
Oh I see, there shouldn't be much activity on pchild during runtime but indeed apr_skiplist could race with legal things like apr_pool_create(..., pchild). I wonder if apr_skiplist shouldn't handle a subpool by itself though, by the time this patch is simple enough so I'm +1.