kern/39777: NFS client triggers kernel panics

[tron%zhadum.org.uk@localhost]

>Number:         39777
>Category:       kern
>Synopsis:       NFS client triggers kernel panics
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Oct 21 18:20:01 +0000 2008
>Originator:     tron%zhadum.org.uk@localhost
>Release:        NetBSD 4.99.73
>Organization:
>Environment:
System: NetBSD lyssa.zhadum.org.uk 4.99.73 NetBSD 4.99.73 (COLWYN.64) #0: Tue 
Oct 21 11:09:59 BST 2008 
tron%lyssa.zhadum.org.uk@localhost:/src/sys/compile/COLWYN.64 amd64
Architecture: x86_64
Machine: amd64
>Description:
Using a kernel built from today's NetBSD-current sources the NFS client in
the kernel seems to be broken. I got a lot of spurious NFS error messages:

Oct 21 12:45:05 lyssa /netbsd: nfs server colwyn:/export/share: not responding
Oct 21 12:45:11 lyssa /netbsd: nfs server colwyn:/export/share: is alive again
Oct 21 12:45:11 lyssa /netbsd: nfs server colwyn:/src: not responding
Oct 21 12:45:17 lyssa /netbsd: nfs server colwyn:/src: is alive again
Oct 21 12:45:17 lyssa /netbsd: nfs server colwyn:/export/ftp: not responding
Oct 21 12:45:23 lyssa /netbsd: nfs server colwyn:/export/ftp: is alive again

A few seconds later the kernel crashed:

fatal double fault in supervisor mode
trap type 13 code 0 rip ffffffff8037d1f4 cs 8 rflags 10282 cr2  
ffff800056d74fa8 cpl 6 rsp ffff800056d74fb0
kernel: double fault trap, code=0
Stopped in pid 17674.1 (sh) at  netbsd:_bus_dmamap_load_mbuf+0x14:      movq    
%
rdi,0xffffffffffffffb0(%rbp)
db{1}> bt
_bus_dmamap_load_mbuf() at netbsd:_bus_dmamap_load_mbuf+0x14
bge_start() at netbsd:bge_start+0x26c
ifq_enqueue() at netbsd:ifq_enqueue+0xc5
ether_output() at netbsd:ether_output+0x55d
ip_output() at netbsd:ip_output+0xb52
tcp_output() at netbsd:tcp_output+0x14aa
tcp_usrreq() at netbsd:tcp_usrreq+0x1e0
tcp_usrreq_wrapper() at netbsd:tcp_usrreq_wrapper+0x51
sosend() at netbsd:sosend+0x4e4
nfs_send() at netbsd:nfs_send+0xa9
nfs_request() at netbsd:nfs_request+0x34a
nfs_readrpc() at netbsd:nfs_readrpc+0x364
nfs_doio() at netbsd:nfs_doio+0x73c
VOP_STRATEGY() at netbsd:VOP_STRATEGY+0x3d
genfs_getpages() at netbsd:genfs_getpages+0xf09
nfs_getpages() at netbsd:nfs_getpages+0xe9
VOP_GETPAGES() at netbsd:VOP_GETPAGES+0x59
ubc_fault() at netbsd:ubc_fault+0x148
uvm_fault_internal() at netbsd:uvm_fault_internal+0x5de
trap() at netbsd:trap+0x6c2
--- trap (number 0) ---
?() at 0
Bad frame pointer: 0xffff800000000000

I wasn't able to produce a crash dump so far. Martti Kuparinen has
similar problems on his NetBSD/amd64 system, too:

http://mail-index.netbsd.org/current-users/2008/10/21/msg005291.html

>How-To-Repeat:
Use a 4.99.73 kernel built from 2008-10-21 sources on a NetBSD/amd64 system,
mount a filesystem via NFS and access it.

>Fix:
None known.