5.2-current does not allow login and panics with modifiedmemory
after free
Doug White
dwhite at gumbysoft.com
Fri Apr 2 19:11:14 PST 2004
On Fri, 2 Apr 2004, Artem Koutchine wrote:
> Hello!
>
> I am trying to run 5.2-CURRENT (cvsupped and rebuild apr 1) on the following
> server:
>
> Supermicro DPE-G2 motherboard
> DUAL XEON 2.66Ghz (HT enabled)
> 2GB RAM
ECC or non-ECC RAM?
> 4 SATA SEAGTE 120GB DRIVES
> 3WARE 8506-4LP SATA RAID5 CONTROLLER
>
> SATA drivers are joint into a single RAID5 array which is seens
> as twe0.
>
> 3dm is installed.
Hm, I haven't had 3dm working on my machines with -CURRENT, but maybe I'm
doing something wierd.
> The box ran half a day and a night just fine. After that it was shutdown
> correctly and after a couple of hours turned on.
>
> I have heared (leds do not work due to connector incompatibility with
> supermicro case) high hdd activity for some time, but all filesysrems
> were marked as clean. I tried to login but i could not because after i
> types 'root' at the login nothing happened at all. Just a blinking cursor.
> However, the box was not hanging - screensaver appeared after some
> time and i can type on tty and also swicth vttys and use scroll lock to
> scroll
> boot messages. 3DM did not respond via web. So, i waited 4 hours and
> rebooted.
>
> After reboot i can logon but after several requests to RAID status via
> web (3dm) i got the following on the console:
>
> twe0: TWEIO_GET_PARAM failed for 0x402/0x3/16
Sounds like one of the disks flaked out ... if you reboot and check the
3ware BIOS, does it complain one of the disks has failed?
>
> Then after a minute or two the following happened:
>
> Memory modified after free 0x788f400(508) val=20202020 @ 0xe788f400
Thats an ominous value. The usual overwrite value is 0xd0d0d0d0, but that
looks like it was shifted.
> panic: Most recently used by devbuf
> at line 128 in file /usr/src/sys/udm_dbg.c
> cpu=0;
> Debugger ("panic")
> Stopped at Debugger +0.46: xchgl %ebx, in_Debugger.0
I suspect this is a secondary issue with a drive problem. We'd need a
backtrace to get more info, however. ('tr' at hte ddb prompt.)
--
Doug White | FreeBSD: The Power to Serve
dwhite at gumbysoft.com | www.FreeBSD.org
More information about the freebsd-current
mailing list