DragonFlyBSD

I have similar dumping problems on two machines.

Sascha

:Hi all,
:My workstation panics upon downloading via bittorrent. Weird huh!
:Anyway, ive got vmcore dumps set to swap(which is twice the size of physical
:memory). Anyway, when I type panic in ddb, it sets of to create a core dump
:but stops on block 1022(which is like the first number that comes up anyway)
:and doesn't go any further. ive tried setting kern.sync_on_panic to 0 but
:that doesn't help at all.
:
:What to do?
:
:Petr

Can you do a 'trace' from the ddb> prompt ? That might give us enough
information to track it down. Plus the panic message itself and fault
address info.

Sometimes it's easier just to take a digital photo of your screen instead
of writing it all down :-)

When did this start happening? Any particular version or date when you
updated your sources and it began happening ?

-Matt
Matthew Dillon
<dillon@backplane.com>

Hi Matt,

What if the content of the trace is bigger than my screen and i cannot fit it
into the shot? Is there such thing as "scroll up" so i can see the whole
trace and not just last 15 lines of it?

I've got 2 hard drives. One with / /usr /var/ home etc on it and the other
mounting to /home/petr/data which is just one 80gb partition. it seems to
happen only when downloading to ~/data

must have begun about a month ago when i came back from 2 months vacation and
updated to latest -devel.

1) What could be causing the inability to dump the vmcore?
2) Would it be hard to implement FreeBSD's option DDB_UNATTENDED? I think it
would be a very nice thing to have for us who also run DF servers in a remote
location with DDB compiled in the kernel.

Thanks,

That's not a FreeBSD-specific option. Try it out if you want.

Do not know if it works after a panic, but usually pressing Scroll Lock
and then using Page Up/Down will work, press Scroll Lock again when you
are done (finally something that uses the key like it was meant to be used).

Matt, here is a picture of the trace:

http://www.punchyouremployer.com/images/p1000672_small.jpg

I didnt actually notice there was any panic message. The scrolling as suggest
by Eric worked though. Thanks Eric.

Petr

Here is another screenshot, this time with the panic message. I completely
ignored it previously...

http://www.punchyouremployer.com/images/p1000673_small.jpg

:Here is another screenshot, this time with the panic message. I completely
:ignored it previously...
:
:http://www.punchyouremployer.com/images/p1000673_small.jpg
:
:On Tue, 2 Oct 2007 01:02:56 pm Petr Janda wrote:
:> Matt, here is a picture of the trace:
:>
:> http://www.punchyouremployer.com/images/p1000672_small.jpg
:>
:> I didnt actually notice there was any panic message. The scrolling as
:> suggest by Eric worked though. Thanks Eric.
:>
:> Petr

I'm looking at it right now. Is this the latest HEAD?

-Matt
Matthew Dillon
<dillon@backplane.com>

::Here is another screenshot, this time with the panic message. I completely
::ignored it previously...
::
::http://www.punchyouremployer.com/images/p1000673_small.jpg
::
::On Tue, 2 Oct 2007 01:02:56 pm Petr Janda wrote:
::> Matt, here is a picture of the trace:
::>
::> http://www.punchyouremployer.com/images/p1000672_small.jpg
::>
::> I didnt actually notice there was any panic message. The scrolling as
::> suggest by Eric worked though. Thanks Eric.
::>
::> Petr

Ok, I could be missing something but that code path doesn't exist
in RELEASE or HEAD. Are you running an older kernel? The putpages
code does not call vm_page_wakeup() insofar as I can tell.

If you are running the latest release or development kernel I'll need
your kernel binary so I can track down those instruction offsets.

-Matt
Matthew Dillon
<dillon@backplane.com>

Hi Matt,
After upgrading to latest HEAD, the problem seems to be gone. Although. should
it happen again, I will let you know.

Petr

:Hi Matt,
:After upgrading to latest HEAD, the problem seems to be gone. Although. should
:it happen again, I will let you know.
:
:Petr

Ok. I'm pretty sure that particular problem was fixed. You must have
been running an older HEAD which was still undergoing transition.

-Matt
Matthew Dillon
<dillon@backplane.com>

I'll close this one for now as the problem seems fixed.