Luiz Fernando N. Capitulino skrev:
Em Wed, 03 Jan 2007 21:39:16 +0200
Thomas Backlund <[EMAIL PROTECTED]> escreveu:
| Luiz Fernando N. Capitulino skrev:
| > Em Sat, 30 Dec 2006 17:25:35 +0200
| > Thomas Backlund <[EMAIL PROTECTED]> escreveu:
| >
| > | > I'll apply the fix, but I've submitted a mdv2007 kernel release already
| > | > and QA is finishing the tests, then mdv2007 will have to wait for the
| > | > next release (which will happen in a month or so) or the next UC
| > | > (which may happen in one week).
| > | >
| > | > Is that ok? Is someone hitting that bug?
| > | >
| > |
| > | I think it's quite OK to make it wait for next uc, so that you can get
| > | the current Security Fixed kernel out without any further delay...
| > |
| > | Then people can run the uc kernel if/when they have the problem, just to
| > | provide more testing ;-)
| >
| > I've tried to reproduce this following Linus' instructions[1] in three
| > different machines with latest mdv2007 and mdv2006 kernel w/o success.
| >
|
| Yeah...
| As noted it's harder to reproduce on older kernels, but if you read the
| whole thread on LKML, there were some reports of reproducing it as far
| back as 2.6.5 ...
Yes, but I wonder whether it was an unpatched kernel.
Sorry for the long delay in responding ...
I can't say about thr 2.6.5 kernel, but one of the 2.6.18 kernel that
hit this bug had th stuff that broke backported from 2.6.19...
So that's a dead end..
as for the other reports, it seems to be another bug (se below)...
| And if IRC it's easier to reproduce with either PREEMPT or SMP enabled...
I've tried on a SMP machine, but it has a fast SCSI disk.
| One thing I have noted on my 2007.0 x86_64 Server is that when I'm
| running my 2.6.19.1-4mdv (wich has the fix for this bug) I haven't been
| able so far to reproduce the "rpmdb: reference count went negative"
| error in my CHROOTS, and I set up 6 of them ... (2006 i586 & x86_64,
| 2007 i586 & x86_64, Cooker i586 & x86_64)
Can you try to reproduce it?
Today I actually got my first "rpmdb: referencecount went negative", so
it's still there ...
I really want to avoid applying a fix for a theoretical bug, or a bug
which won't affect us in real life (just because new changes can
potentially introduce new problems).
Actually the discussion on LKML has been quite active regarding this bug
and kernels <=2.6.19,
and the consensus is that there is most likely a bug in older kernels,
but it's hard to reproduce, so it
most likely wont be fixed in older kernels...
As for this patch that went into 2.6.19.2, if applied to older kernels
it only seems to hide the bug more,
not actually fix it...
So you can forget about this patch...
| Oh, and on a side note...
| Any movement on the *-latest kernels ??
| Should I post an updated patch ??
Would be nice to get an updated patch, but I'm going to speak with
Arnaud and GB. We need this stuff in cooker first.
I'll re-send a patch today or tomorrow as a last try, after that I wont
bother anymore...
We'we proven with kernel-tmb, kernel-mm, and now also kernel-linus that
this works great, so ...
--
Thomas
