Date: Sun, 13 Oct 2002 13:55:33 -0700 From: Terry Lambert <tlambert2@mindspring.com> To: Ollivier Robert <roberto@keltia.freenix.fr> Cc: current@FreeBSD.org Subject: Re: current unable to exec processes under load Message-ID: <3DA9DDC5.1547947E@mindspring.com> References: <20021012213213.GC93425@xor.obsecurity.org> <20021013195623.GA33370@tara.freenix.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Ollivier Robert wrote: > According to Kris Kennaway: > > load: 0.00 cmd: tcsh 8403 [inode] 0.01u 0.00s 0% 1076k > > The dreaded "inode" problem. I've been seeing this from time to time where the > system will be blocked for all I/O on a given disk with all processes waiting > on "inode". > > Generally updating the system to a more recent "fixes" the problem till next > time. I don't know where it comes from :-( It's actually the name for "vp->v_vnlock" from FFS; specifically, line #1231 of /usr/src/sys/ufs/ffs/ffs_vfsops.c. It's actually the address of the vp->v_lock. Acquisition supposedly times out in 50ms. You may want to try compiling with: options DEBUG_LOCKS It's locked and unlocked in /usr/src/sys/kern/vfs_default.c in vop_stdlock(), vop_stdunlock() (which should also be instrumented for DEBUG_LOCKS, but isn't), vop_sharedlock(), and (sort of) in vop_nolock(). So try the option, if that doesn't work, instrument vop_stdlock() and try it again, and if that still gives no information, check with whoever wrote the code. -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3DA9DDC5.1547947E>