Around 9:45 this morning, our mailserver (SunOS 5.11 snv_91 i86pc i386
i86pc) rebooted.
Looking at /var/crash/HOSTNAME, I saw the unix.0 and vmcore0 files.
Loading them up in MDB, I get the following:
> ::panicinfo
cpu 0
thread ffffff02d75b1ca0
message
BAD TRAP: type=d (#gp General protection) rp=ffffff00104fd610 addr=8
rdi ffffff033fdad800
rsi 0
rdx ffffff02d75b1ca0
rcx 0
r8 0
r9 c3180000000000fb
rax 0
rbx 0
rbp ffffff00104fd720
r10 ffffff036abb8a48
r10 ffffff036abb8a48
r11 0
r12 ffffff033fdad800
r13 ffffff033fdad800
r14 0
r15 8
fsbase 0
gsbase fffffffffbc287b0
ds 4b
es 4b
> ::msgbuf
[SNIP]
panic[cpu0]/thread=ffffff02d75b1ca0:
BAD TRAP: type=d (#gp General protection) rp=ffffff00104fd610 addr=8
imapd:
#gp General protection
addr=0x8
pid=9078, pc=0xfffffffff783ab10, sp=0xffffff00104fd700, eflags=0x10246
cr0: 8005003b<pg,wp,ne,et,ts,mp,pe> cr4:
6f8<xmme,fxsr,pge,mce,pae,pse,de>
cr2: fe5e0000
cr3: 1efa44000
cr8: c
rdi: ffffff033fdad800 rsi: 0 rdx:
ffffff02d75b1ca0
rcx: 0 r8: 0 r9:
c3180000000000fb
rax: 0 rbx: 0 rbp:
ffffff00104fd720
r10: ffffff036abb8a48 r11: 0 r12:
ffffff033fdad800
r13: ffffff033fdad800 r14: 0
r15: 8
fsb: 0 gsb: fffffffffbc287b0
ds: 4b
es: 4b fs: 0 gs:
1c3
trp: d err: 0 rip:
fffffffff783ab10
cs: 30 rfl: 10246 rsp:
ffffff00104fd700
ss: 0
ffffff00104fd4f0 unix:die+ea ()
ffffff00104fd600 unix:trap+3d0 ()
ffffff00104fd610 unix:_cmntrap+e9 ()
ffffff00104fd720 zfs:dbuf_evict_user+50 ()
ffffff00104fd770 zfs:dbuf_rele+a8 ()
ffffff00104fd7d0 zfs:dmu_buf_rele_array+4d ()
ffffff00104fd860 zfs:dmu_read+1a8 ()
ffffff00104fd910 zfs:zfs_fillpage+15e ()
ffffff00104fd9f0 zfs:zfs_getpage+187 ()
ffffff00104fdaa0 genunix:fop_getpage+9f ()
ffffff00104fdc60 genunix:segvn_fault+9ef ()
ffffff00104fdd70 genunix:as_fault+5ae ()
ffffff00104fddf0 unix:pagefault+95 ()
ffffff00104fdf00 unix:trap+bd3 ()
ffffff00104fdf10 unix:_cmntrap+e9 ()
syncing file systems...
23
done
dumping to /dev/zvol/dsk/rootpool/dump, offset 65536, content: kernel
Is there a way to tell if ZFS caused the kernel panic? I notice that
it says imapd: in the middle of the msgbuffer, does that mean imapd
caused the kernel panic? I''m just trying to figure out what to do
here and determine if a bug caused the panic so that I can submit the
proper information to get it fixed :^)
--
Michael Hale <mhale at
transcomus.com
>
Manager of Engineering Support Enterprise Engineering Group
Transcom Enhanced Services http://www.transcomus.com
Michael Hale wrote:> Around 9:45 this morning, our mailserver (SunOS 5.11 snv_91 i86pc i386 > i86pc) rebooted. > > Looking at /var/crash/HOSTNAME, I saw the unix.0 and vmcore0 files. > > Loading them up in MDB, I get the following: > > > ::panicinfoIn general, if you get a panic that cannot be directly attributed to hardware, please file a bug. -- richard
The stack trace makes it clear that it was ZFS that crashed. (The _cmntrap stack frame indicates that a trap happened; in this case it''s an access to bad memory by the kernel. The previous stack frame indicates that ZFS was active.) Now, it may not have been ZFS which caused the panic -- another part of the kernel could have corrupted ZFS''s memory -- but an engineer familiar with the code would have to look at the core file to know for sure. This message posted from opensolaris.org
Michael Hale wrote:> Around 9:45 this morning, our mailserver (SunOS 5.11 snv_91 i86pc i386 > i86pc) rebooted.[...]> dumping to /dev/zvol/dsk/rootpool/dump, offset 65536, content: kernel > > Is there a way to tell if ZFS caused the kernel panic? I notice that > it says imapd: in the middle of the msgbuffer, does that mean imapd > caused the kernel panic? I''m just trying to figure out what to do > here and determine if a bug caused the panic so that I can submit the > proper information to get it fixed :^)Is it really true that you run your companies mailserver on snv_91 and with root on ZFS? No offense, but in that case I think the proper thing to do is to switch to Solaris 10 5/08.
On Jul 15, 2008, at 4:31 PM, Richard Elling wrote:> Michael Hale wrote: >> Around 9:45 this morning, our mailserver (SunOS 5.11 snv_91 i86pc >> i386 i86pc) rebooted. >> >> Looking at /var/crash/HOSTNAME, I saw the unix.0 and vmcore0 files. >> >> Loading them up in MDB, I get the following: >> >> > ::panicinfo > > In general, if you get a panic that cannot be directly attributed > to hardware, please file a bug. > -- richard >What''s the proper way to file a bug report for opensolaris? Is there a web form and a way to upload the core file? -- Michael Hale <mhale at transcomus.com > Manager of Engineering Support Enterprise Engineering Group Transcom Enhanced Services http://www.transcomus.com