Den 19-Mar-00 05:20:57 skrev Gregory Maxwell f�lgende om "Reiserfs as root?
2.3.99-pre1":
> I decided to migrate a test system to 100% reiser fs.
Me too, although I've had to put / on a small ext2 partition due to lack
of raid boot support. The / partition is too small for reiserfs' minimum
requirement of 32 MB.
> /tmp, /var, and /usr went fine. They run great.
> /boot is being left as ext2 (It's smaller then the reiserfs log, and
> normally mounted read only).
> After copying root to a reiserfs volume and pointing the kernel to it, I
> see it get mounted fine, then the system hard locks after the "Freeing
> memory" message. This is booting with init=/bin/bash
I'm guessting this could be a corrupt partition which reiserfs chokes
on. I've had problems when /var/run is on reiserfs. It appears to be
corrupted somehow while the filesystem is mounted. Then if something tries
to list the directory after a reboot (which RedHat rc.sysinit does), the
process hangs in uninterruptible sleep (state "D" in top/ps). My /var/run
normally contains
[root@vissevasse raid]# ls -lLR /var/run
/var/run:
total 14
-rw-r--r-- 1 root root 4 Mar 17 18:47 atd.pid
-rw-r--r-- 1 root root 4 Mar 17 18:41 crond.pid
-rw-r--r-- 1 root root 4 Mar 17 18:41 inetd.pid
-rw-r--r-- 1 root root 4 Mar 17 18:41 klogd.pid
drwxr-xr-x 2 root root 1024 Mar 17 18:29 netreport
srwx------ 1 root root 0 Mar 17 18:41 pump.sock
-rw------- 1 root root 512 Mar 17 18:41 random-seed
-rw-r--r-- 1 root root 16 Mar 17 18:41 runlevel.dir
-rw-r--r-- 1 root root 4 Mar 17 18:41 sshd2_22.pid
-rw-r--r-- 1 root root 4 Mar 17 18:41 syslogd.pid
-rw-rw-r-- 1 root utmp 4992 Mar 20 02:31 utmp
/var/run/netreport:
total 0
where the only slightly unusual entry is the Unix domain socket. Dmesg
shows some ugly stuff:
ReiserFS version 3.6.1
[cut two Ethernet drives telling me they loaded fine]
uniqueness 252
kernel BUG at /usr/build/linux-2.3.51/include/linux/reiserfs_fs.h:419!
invalid operand: 0000
CPU: 0
EIP: 0010:[<c0160189>]
EFLAGS: 00010286
eax: 00000047 ebx: c05d5f24 ecx: c021a198 edx: c0aa5f84
esi: c027d308 edi: c027d6e8 ebp: c05d5f24 esp: c05d5e2c
ds: 0018 es: 0018 ss: 0018
Process slocate (pid: 759, stackpage=c05d5000)
Stack: c01f2850 c01f2920 000001a3 c05d5f24 c027d308 c027d6e8 c027d6fa 00000000
c0160424 c05d5f24 c027d308 c05d5e9c c01607b0 c027d308 c05d5f24 c059b1a0
c05d5f24 c05d5f00 00000000 c027d6e8 c05d5e84 c05d5e88 0000006b 00000000
Call Trace: [<c01f2850>] [<c01f2920>] [<c0160424>] [<c01607b0>] [<c0158bfd>]
[<c01effe0>] [<c0154784>]
[<c0134d0a>] [<c0134f84>] [<c0135078>] [<c0131eae>] [<c01096ac>] [<ffffdfb9>]
[<c010002b>]
Code: 0f 0b 83 c4 0c 89 f6 b8 0f 00 00 00 89 c6 89 f7 c1 ff 1f 89
UDP: short packet: 0/226
vs-13042: reiserfs_read_inode: [107114 107445 0x0 SD] not founduniqueness 107114
kernel BUG at /usr/build/linux-2.3.51/include/linux/reiserfs_fs.h:419!
invalid operand: 0000
CPU: 0
EIP: 0010:[<c0160189>]
EFLAGS: 00010286
eax: 00000047 ebx: c09eff24 ecx: c021a198 edx: c0aa5f84
esi: c027d304 edi: c027d6e8 ebp: c09eff24 esp: c09efe2c
ds: 0018 es: 0018 ss: 0018
Process make (pid: 9656, stackpage=c09ef000)
Stack: c01f2850 c01f2920 000001a3 c09eff24 c027d304 c027d6e8 c027d6fa 00000000
c0160424 c09eff24 c027d304 c09efe9c c01607b0 c027d304 c09eff24 c07b9350
c09eff24 c09eff00 00000000 c027d6e8 c09efe84 c09efe88 0000006b 00000000
Call Trace: [<c01f2850>] [<c01f2920>] [<c0160424>] [<c01607b0>] [<c0158bfd>]
[<c01effe0>] [<c0154784>]
[<c0134d0a>] [<c0134f84>] [<c0135078>] [<c0131dce>] [<c01096ac>] [<c010002b>]
Code: 0f 0b 83 c4 0c 89 f6 b8 0f 00 00 00 89 c6 89 f7 c1 ff 1f 89
The first part is the nightly file database update which runs into a
"death trap" directory. The last part is make dying in a copy of my kernel
build directory, before it even got as far as to write anything to the
screen. Hmm.
I think I saw a note somewhere about how to turn those angle bracketed
hex numbers into useful symbolic constants, so I'll have a go at that.
Regards,
/��������������������������������T���������������������������������������\
| Rask Ingemann Lambertsen | E-mail: mailto:[EMAIL PROTECTED] |
| Please do NOT Cc: to me or the | WWW: http://www.gbar.dtu.dk/~c948374/ |
| mailing list. I am on the list.| "ThrustMe" on XPilot, ARCnet and IRC |
|To err is human. To forgive is beyond the scope of the Operating System.|
