Date: Sat, 16 Dec 2006 06:20:13 GMT From: Bruce Evans <bde@zeta.org.au> To: freebsd-bugs@FreeBSD.org Subject: Re: kern/92785: Using exported filesystem on OS/2 NFS client causes filesystem freeze Message-ID: <200612160620.kBG6KDYH068259@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR kern/92785; it has been noted by GNATS. From: Bruce Evans <bde@zeta.org.au> To: Kostik Belousov <kostikbel@gmail.com> Cc: freebsd-gnats-submit@freebsd.org Subject: Re: kern/92785: Using exported filesystem on OS/2 NFS client causes filesystem freeze Date: Sat, 16 Dec 2006 17:16:58 +1100 (EST) On Fri, 15 Dec 2006, Kostik Belousov wrote: > Am I right that all you did was ls -l <root of nfs mount> ? Does OS/2 > supports the notion of ".." directory ? Could you do just "ls -l .." > from nfs client and then try "stat <root of exported fs>" on the server > (i think it shall hang) ? > > My hypothesis is that LOOKUP RPC for ".." causes directory vnode lock > leak in nfs_namei. After that, mountd hang is just consequence. Is this related to rev.1.103 of vfs_cache.c? That commit turns off caching of ".." in most (?) cases. This gives an especially large pessimization for nfs since for nfs there is no chance that ".." is cached elsewhere in the kernel, so physical i/o is always required to look up "..". I run with this commit backed out and haven't noticed any problems (but it seems to only be needed for the unusual operation of removing active current directories). Bruce.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200612160620.kBG6KDYH068259>