Date: Sun, 7 May 2006 18:40:29 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Dmitry Morozovsky <marck@rinet.ru> Cc: stable@freebsd.org Subject: Re: quota deadlock on 6.1-RC1 Message-ID: <20060507154029.GA12655@deviant.kiev.zoral.com.ua> In-Reply-To: <20060507175330.H19670@woozle.rinet.ru> References: <44599732.1050905@rogers.com> <20060505080543.GD5466@garage.freebsd.pl> <35c231bf0605051049t2761281ar97b9634b8279b1fd@mail.gmail.com> <445B991F.3050600@rogers.com> <6.2.3.4.0.20060505145259.10d8dca0@64.7.153.2> <3aaaa3a0605061238ib77bb38x6f61eb6cc1c1c6c1@mail.gmail.com> <20060506204720.S46997@fledge.watson.org> <20060506233525.GA52731@xor.obsecurity.org> <20060507104355.K92932@woozle.rinet.ru> <20060507175330.H19670@woozle.rinet.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
--fdj2RfSjLxBAspz7 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sun, May 07, 2006 at 07:26:55PM +0400, Dmitry Morozovsky wrote: > On Sun, 7 May 2006, Dmitry Morozovsky wrote: >=20 > DM> KK> > Also, as an FYI, several quota-related snapshot fixes went into= CVS in the=20 > DM> KK> > last 24-48 hours. Once they've settled for a few weeks, and as= suming they=20 > DM> KK> > don't have complex dependencies, they will get merged to RELENG= _6.=20 > DM> KK> > Presumably if there is sufficient demand, and they adequately a= ddress the=20 > DM> KK> > reports of deadlocks with quotas+snapshots, they could be merge= d to=20 > DM> KK> > RELENG_6_1 as an errata update. > DM> KK>=20 > DM> KK> The fixes seem to indeed correct all the deadlocks I was seeing. = I > DM> KK> don't have time to produce a 6.x patch now though, but it should = be a > DM> KK> simple matter for someone else to do. > DM>=20 > DM> FWIW, for me it is much stabler too, at least in normal load patterns= . I'll=20 > DM> spend a bit of time to stress-test this on our Samba quota-enabled se= rver=20 > DM> during our long weekend (today-to-Tuesday) to catch possible rare rac= es. >=20 > Alas, it is locked again: >=20 > root@office:~# ps axlww | grep snap > 0 32 0 0 -4 0 0 8 snaplk DL ?? 0:04.55 [bufd= aemon] > 0 35 0 0 -4 0 0 8 snaplk DL ?? 0:01.11 [soft= depflush] > 0 45080 45061 0 -4 0 5980 2768 snaplk D ?? 0:00.09 /usr/= local/sbin/smbd -D > 0 46040 46039 0 -4 0 4492 3616 snaplk D+ p1 0:15.04 rsync= -av /ar/winroot/distrib/ /st/winroot/tmp/distrib-temp/ > 1002 46116 726 0 -4 0 1612 1144 snaplk D+ p3 0:00.00 ls -G= F -la /st/.snap/ >=20 > Unfortunately, this was with non-debug kernel. I'm trying to reproduce th= e lock=20 > with debug kernel right now, but still with no luck - sometimes machine i= s=20 > almost hangs but up to now revives 4 or 5 times. >=20 So, you are running CURRENT ? Or did you manually backported changes from CURRENT to 6-STABLE ? Note that none of the relevant commits where MFCed still. --fdj2RfSjLxBAspz7 Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (FreeBSD) iD8DBQFEXhTsC3+MBN1Mb4gRAs2ZAKCg37GAtn8fSt9lr3hhL1u7OxlXqACdHpiH vTRzbF30LwmF2aBcU4qFO1Y= =tvif -----END PGP SIGNATURE----- --fdj2RfSjLxBAspz7--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060507154029.GA12655>