Broken NFS ?

From: Thierry Herbelot <thierry_at_herbelot.com>
Date: Thu, 3 Nov 2005 10:37:33 +0100
Hello,

the latest current kernel seems to have a broken NFS : it can't get a listing 
for a shared directory (itself hosted on a just recompiled 6-STABLE server) :

# ls /files2/net<TAB>nfs server tfh:/files2: not responding
-cvs

A pcap trace of the NFS messages is :
http://thierry.herbelot.free.fr/nfs-broken/nfs-broken.pcap

tfh# tcpdump -n -t -r nfs-broken.pcap
reading from file nfs-broken.pcap, link-type EN10MB (Ethernet)
IP 192.168.2.12.855699310 > 192.168.2.1.2049: 104 access fh 1056,330756/2 003f
IP 192.168.2.1.2049 > 192.168.2.12.855699310: reply ok 120 access c 003f
IP 192.168.2.12.855699311 > 192.168.2.1.2049: 104 access fh 1056,330756/2 003f
IP 192.168.2.1.2049 > 192.168.2.12.855699311: reply ok 120 access c 003f
IP 192.168.2.12.855699312 > 192.168.2.1.2049: 100 fsstat fh 1056,330756/2
IP 192.168.2.1.2049 > 192.168.2.12.855699312: reply ok 168 fsstat tbytes 161358067712 fbytes 9270075392 abytes 2815754240
IP 192.168.2.12.855699313 > 192.168.2.1.2049: 104 access fh 1056,330756/10633216 003f
IP 192.168.2.1.2049 > 192.168.2.12.855699313: reply ok 120 access c 003f
IP 192.168.2.12.855699314 > 192.168.2.1.2049: 104 access fh 1056,330756/2 003f
IP 192.168.2.1.2049 > 192.168.2.12.855699314: reply ok 120 access c 003f
IP 192.168.2.12.855699315 > 192.168.2.1.2049: 104 access fh 1056,330756/2 003f
IP 192.168.2.1.2049 > 192.168.2.12.855699315: reply ok 120 access c 003f
IP 192.168.2.12.855699316 > 192.168.2.1.2049: 100 fsstat fh 1056,330756/2
IP 192.168.2.1.2049 > 192.168.2.12.855699316: reply ok 168 fsstat tbytes 161358067712 fbytes 9270075392 abytes 2815754240
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624
IP 192.168.2.12.855699317 > 192.168.2.1.2049: 112 lookup fh 1056,330756/2 "net-cvs"
IP 192.168.2.1.2049 > 192.168.2.12.855699317: reply ok 236 lookup fh 1056,330756/18202624

the kernel needs 10 tries to read the directory listing ?

the kernel has been built via the make buildworld/make buildkernel procedure 
(the kernel is a straight GENERIC : FreeBSD XXX 7.0-CURRENT FreeBSD 
7.0-CURRENT #65: Thu Nov  3 09:22:29 CET 2005     
ZZZ_at_XXX:/files1/obj/files1/src/sys/GENERIC  i386)

the latest diffs I have in the non-working kernel include :
(kernel.good.ident are all the idents for the last working kernel, kernel.ident 
is the same for the first non-working kernel)

multi-cur# diff kernel.good.ident kernel.ident
1c1
< /boot/kernel.good/kernel:
---
> /boot/kernel/kernel:
336c336
< src/sys/kern/kern_mbuf.c,v 1.11 2005/09/27 20:28:43 rwatson Exp $
---
> src/sys/kern/kern_mbuf.c,v 1.12 2005/11/02 16:20:35 andre Exp $
410c410
< src/sys/kern/uipc_mbuf.c,v 1.155 2005/08/30 21:31:42 andre Exp $
---
> src/sys/kern/uipc_mbuf.c,v 1.156 2005/11/02 16:20:35 andre Exp $
413,414c413,414
< src/sys/kern/uipc_socket.c,v 1.251 2005/10/30 19:44:38 rwatson Exp $
< src/sys/kern/uipc_socket2.c,v 1.149 2005/10/30 19:44:39 rwatson Exp $
---
> src/sys/kern/uipc_socket.c,v 1.252 2005/11/02 13:46:31 andre Exp $
> src/sys/kern/uipc_socket2.c,v 1.150 2005/11/02 13:46:31 andre Exp $
463c463
< src/sys/net80211/ieee80211_freebsd.c,v 1.8 2005/08/08 18:46:35 sam Exp $
---
> src/sys/net80211/ieee80211_freebsd.c,v 1.9 2005/11/02 13:46:31 andre Exp $
467c467
< src/sys/net80211/ieee80211_output.c,v 1.34 2005/08/10 16:22:29 sam Exp $
---
> src/sys/net80211/ieee80211_output.c,v 1.35 2005/11/02 13:46:31 andre Exp $

I have re-synced the sources and recompiled the kernel to avoid a partial commit, 
but the bug is still there.

	TfH

PS : kernel modules loaded :
# kldstat
Id Refs Address    Size     Name
 1    6 0xc0400000 698da0   kernel
 2    1 0xc0a99000 70ec     snd_es137x.ko
 3    6 0xc0aa1000 23aa4    sound.ko
 4    1 0xc0ac5000 b59c     snd_mss.ko
 5    1 0xc0ad1000 2d14     joy.ko
 6    1 0xc0ad4000 147c0    geom_mirror.ko

PS2 : CC'd to andre_at_ as he was the latest commiter on the kernel
Received on Thu Nov 03 2005 - 08:37:49 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:47 UTC