panic: ffs_blkfree: freeing free frag

NAKAJI Hiroyuki nakaji at tutrp.tut.ac.jp
Sun May 4 19:58:16 PDT 2003


>>>>> In <Pine.NEB.3.96L.1030504122645.92658z-100000 at fledge.watson.org> 
>>>>>	Robert Watson <rwatson at freebsd.org> wrote:

> Could you boot to single user mode, manually fsck all of your file
> systems, and then try running your various load tests again?  Perhaps a
> free list problem persisted on disk after the fix left over from earlier
> operation with the bug...

The problem still exists unfortunately. Now I switched my kernel to
the one of Apr 12 2003.

I rebooted my pc98 box to single user mode, ran "fsck" (I have to do
this again and again because of panic) and exitted to multi-user
mode. At this time no panic happened.

Then, I ran "make update buildworld" in /usr/src (/dev/da2s1g) in
kterm on X through the night. No panic.

After I wake up, I checked my new arriving emails with Emacs and
Gnus. Retrieving emails did not cause panic. But while reading emails,
because email backend of Gnus is "nnml" like MH, accessing ~/Mail on
/dev/da2s1h occured and got panic. And next "make update" in /usr/src
was running in background.

I cannot beleive these two are the "heavy disk access" but it may be
so.

Backtrace from vmcore is almost same as I posted before in this
thread.
-- 
NAKAJI Hiroyuki


More information about the freebsd-current mailing list