Re: store_client assertion in 2.7STABLE3

From: Henrik Nordstrom <henrik_at_henriknordstrom.net>
Date: Sun, 20 Jul 2008 22:32:57 +0200

Just committed a patch for this. But I have not audited the code to tell
if this was the only place where this may happen..

Regards
Henrik

On sön, 2008-07-20 at 09:20 +0200, Henrik Nordstrom wrote:
> So clientCacheHit needs to check for ENTRY_ABORTED and fall back on miss
> as if it was a soft swapin failure.
>
>
> On lör, 2008-07-19 at 21:23 -0700, Mark Nottingham wrote:
> > Sorry it took so long; see below.
> >
> > (gdb) bt
> > #0 0x0000000800bcb89c in pthread_testcancel () from /lib/
> > libpthread.so.2
> > #1 0x0000000800bb95c3 in sigaction () from /lib/libpthread.so.2
> > #2 0x0000000800bbb0e2 in sigaction () from /lib/libpthread.so.2
> > #3 0x0000000800bb4db6 in pthread_kill () from /lib/libpthread.so.2
> > #4 0x0000000800bb4633 in raise () from /lib/libpthread.so.2
> > #5 0x0000000800fce63d in abort () from /lib/libc.so.6
> > #6 0x000000000043539e in xassert ()
> > #7 0x0000000000428652 in clientCacheHit ()
> > #8 0x0000000000422d19 in storeClientCopyHeadersCB ()
> > #9 0x000000000048c7e6 in storeClientCallback ()
> > #10 0x000000000048d0a3 in storeClientReadBody ()
> > #11 0x00000000004a4c56 in storeAufsReadDone ()
> > #12 0x00000000004a7272 in aioCheckCallbacks ()
> > #13 0x00000000004900f3 in storeDirCallback ()
> > #14 0x000000000043439a in comm_select ()
> > #15 0x0000000000467ac9 in main ()
> >
> >
> > Need anything more specific?
> >
> > Cheers,
> >
> >
> > On 08/07/2008, at 1:44 AM, Henrik Nordstrom wrote:
> >
> > > On tis, 2008-07-08 at 13:30 +1000, Mark Nottingham wrote:
> > >> Seen again.
> > >
> > >>> assertion failed: store_client.c:172: "!EBIT_TEST(e->flags,
> > >>> ENTRY_ABORTED)"
> > >>>
> > >> Perhaps this?
> > >>
> > >> http://www.squid-cache.org/Versions/v2/2.7/changesets/12220.patch
> > >
> > > Maybe. Quite likely related. Most likely there is some store client
> > > forgetting to check for abort. But yes, that patch males
> > > ENTRY_ABORTED a
> > > bit more likely condition than before.
> > >
> > > Should be very easy to fix if you could get a stack trace of the
> > > failure.
> > >
> > > Regards
> > > Henrik
> >
> > --
> > Mark Nottingham mnot_at_yahoo-inc.com
> >

Received on Sun Jul 20 2008 - 20:33:04 MDT

This archive was generated by hypermail 2.2.0 : Mon Jul 21 2008 - 12:00:07 MDT