Date: Mon, 23 Jan 2012 07:38:12 +0400 From: Yuri Pankov <yuri.pankov@gmail.com> To: Kostik Belousov <kostikbel@gmail.com> Cc: freebsd-current@freebsd.org Subject: Re: panic: No NCF_TS Message-ID: <20120123033812.GC10149@sirius.xvoid.org> In-Reply-To: <20120123030513.GK31224@deviant.kiev.zoral.com.ua> References: <20120123013642.GB10149@sirius.xvoid.org> <20120123030513.GK31224@deviant.kiev.zoral.com.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Jan 23, 2012 at 05:05:13AM +0200, Kostik Belousov wrote: > On Mon, Jan 23, 2012 at 05:36:42AM +0400, Yuri Pankov wrote: > > Seems to be reproducible here running r230467 as the NFS client and > > r230135 as NFS server. NFSv4 not enabled. > > > > # mount > > [...] > > sirius:/data/distfiles on /usr/ports/distfiles (nfs) > > > > # /usr/bin/env /usr/bin/fetch -AFpr -S 4682084 -o /usr/ports/distfiles/sqlite-src-3071000.zip http://www.sqlite.org/sqlite-src-3071000.zip > > /usr/ports/distfiles/sqlite-src-3071000.zip 100% of 4572 kB 379 kBps 00m00s > > # rm /usr/ports/distfiles/sqlite-src-3071000.zip > > > > immediately followed by: > > > > panic: No NCF_TS > > cpuid = 2 > > KDB: enter: panic > > [ thread pid 1603 tid 100494 ] > > Stopped at kdb_enter+0x3e: movq $0,kdb_why > > db> bt > > Tracing pid 1603 tid 100494 td 0xfffffe0089585460 > > kdb_enter() at kdb_enter+0x3e > > panic() at panic+0x245 > > cache_lookup_times() at cache_lookup_times+0x6b5 > > nfs_lookup() at nfs_lookup+0x190 > > VOP_LOOKUP_APV() at VOP_LOOKUP_APV+0x8b > > lookup() at lookup+0x7e9 > > namei() at namei+0x74c > > kern_statat_vnhook() at kern_statat_vnhook+0x90 > > sys_lstat() at sys_lstat+0x30 > > amd64_syscall() at amd64_syscall+0x221 > > Xfast_syscall() at Xfast_syscall+0xfb > > --- syscall (190, FreeBSD ELF64, sys_lstat), rip = 0x80093ff3c, rsp = 0x7fffffffd8d8, rbp = 0x7fffffffd979 --- > > > > Yes, my bad. I wrote the r230441 with the assumption that filesystems > are consistent in their use of cache_enter_time(). And my net-booting > test box did not catched this, which is at least interesting. > > Please try the change below. Actually, it should be enough to only apply > the changes for fs/nfsclient (or nfsclient/ if you use old nfs). If this > does not help, then please try the whole patch. > > Even if not needed for correctness, I think vfs_cache.c change could > be made useful by adding KASSERT to it. Thanks, applying only the fs/nfsclient part of the patch helped. No problems after applying the full patch as well. Yuri
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120123033812.GC10149>