Dave Craft wrote:
>
> I saw this less than 10 minutes ago myself on the 2.4.0
> kernel. It still seems to produce a valid dump. I am
> not running with the kdb patch...only lkcd.
>
> Dave
> >
> >i just patched the latest lkcd patch into the SGI XFS tree (which
> >also contains kdb) ... it so far also works but after it has
> >written the dump it panics (now into kdb) on mm/memory.c line 634
> >
> > for (j = 0; j < nlast; ppage++, j++) {
> > page = *ppage;
> > if (!page)
> > continue;
> > UnlockPage(page); <-------
> > }
> > }
> > return 0;
> >}
> >
> >don't know what the reason for this is - kdb or 2.4.0 kernel - so
> >my question: did anyone else try this so far? - any solution in
> >sight?
The solution at this point is two-fold. First, I believe you can
remove the call to free_kiovec(), and that would eliminate the
problem (if you have a system ready to try it, give it a shot).
If that doesn't work, I can explore it in more detail. We don't
_really_ have to free the kiovec, as we're pretty much done with
I/O, and all we're going to do is reset the system anyway. I just
had that in there for completeness.
The _real_ solution is to not use kiobufs at all. I've got
an IDE dump driver sort of working, it's choking after about 700 pages
or so (not sure why yet), but it's coming along. That will be our
next patch. In the meantime, though, if the removal of free_kiovec()
fixes the problem, let me know and I'll release a patch in the meantime.
Thanks, guys.
--Matt
> >
> >it is absolutely reproducable if anyone wants to try it - but i
> >can also post an backtrace of the panic here if that might help
> >(but it does not look very interesting i think - it ends in
> >kiovec_unlock or something like) ... btw. i used a simple panic
> >module (calling panic() in init_module) via insmod to force the
> >panic ...
> >
> >any ideas? - thanks in advance
> >
> >t
> >
> >--
> >thomas.graichen@xxxxxxxxxxxxxx
> > innominate AG
> > the linux architects
> >tel: +49-30-308806-13 fax: -77 http://www.innominate.com
> >
>
> --
> --------- Opinions are mine and not IBM's ----------
> Mail : dave@xxxxxxxxxxxxxx Phone : 512-838-8248
> Wish I could fly like Superman.
|