Date: Tue, 25 Apr 2006 08:09:32 -0500 From: Chris Dillon <cdillon@wolves.k12.mo.us> To: Dmitry Morozovsky <marck@rinet.ru> Cc: stable@freebsd.org, Kris Kennaway <kris@obsecurity.org> Subject: Re: fsck_ufs locked in snaplk Message-ID: <20060425080932.1rv9hq0rcws4wc84@www.wolves.k12.mo.us> In-Reply-To: <20060425134418.J57625@woozle.rinet.ru> References: <20060423193208.N1187@woozle.rinet.ru> <20060423201732.GA74905@xor.obsecurity.org> <20060424091803.L20593@woozle.rinet.ru> <20060424215650.P36233@woozle.rinet.ru> <20060424181531.GA13774@xor.obsecurity.org> <20060425001751.S44618@woozle.rinet.ru> <20060424202859.GA18457@xor.obsecurity.org> <20060425004405.G44618@woozle.rinet.ru> <20060424205026.GA18844@xor.obsecurity.org> <20060425134418.J57625@woozle.rinet.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
Quoting Dmitry Morozovsky <marck@rinet.ru>: > On Mon, 24 Apr 2006, Kris Kennaway wrote: > > KK> Also you should add DEBUG_LOCKS and DEBUG_VFS_LOCKS on the off chance > KK> they catch the problem. > > I got one thought about the source of these hangs/crashes: this =20 > machine is the > only one with actively used quotas. I'll test the more thoroughly =20 > this evening. I had problems with snapshots and hangs in 5.x. For that, a daily =20 reboot would keep the problems at bay. I upgraded to 6.0 and the =20 problems completely disappeared. I kept 6.0-STABLE running for weeks. =20 Somewhere along the line, as 6.1 approached, similar problems =20 re-appeared, but not exactly the same as what I had in 5.x. Now =20 instead of a complete system hang, individual processes will hang =20 while attempting to access a certain filesystem. I'm running =20 6.1-PRERELEASE from April 2 and for some reason since this weekend it =20 has happened more often, but I'm not sure why since I haven't made any =20 system changes since April 2. I also am using quotas heavily with this system, and snapshoting every =20 filesystem once a day. The filesystem which processes will hang on =20 when attempting to access it happens to be the one with quotas enabled. I'm going to update to the latest 6.1 code this evening and enable =20 INVARIANTS, WITNESS, and the two DEBUG_LOCKS options to the kernel to =20 see if it catches anything.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060425080932.1rv9hq0rcws4wc84>