2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Justin Piszcz]

Beginning at 2.6.17 to 2.6.17.6, there is a serious XFS bug that results 
in filesystem corruption, there was a 1 line bugfix patch that was 
released recently and I was wondering when 2.6.17.7 would be released with 
that patch?  It affected ALL my Linux machines (x86) running XFS and many 
people on the XFS mailing list who upgraded to 2.6.17.  I understand when 
there is a root exploit or DoS bug, the kernel is naturally patched by the 
-stable team and a new version is released immediately.  Does filesystem 
corruption not constitute an immediate new -stable release of the kernel?

Re: 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Justin Piszcz]

On Mon, 24 Jul 2006, Justin Piszcz wrote:

> Beginning at 2.6.17 to 2.6.17.6, there is a serious XFS bug that results in 
> filesystem corruption, there was a 1 line bugfix patch that was released 
> recently and I was wondering when 2.6.17.7 would be released with that patch? 
> It affected ALL my Linux machines (x86) running XFS and many people on the 
> XFS mailing list who upgraded to 2.6.17.  I understand when there is a root 
> exploit or DoS bug, the kernel is naturally patched by the -stable team and a 
> new version is released immediately.  Does filesystem corruption not 
> constitute an immediate new -stable release of the kernel?
>
>
>

This fix was available as of 2.6.17.2, but not currently in 2.6.17.6...

http://marc.theaimsgroup.com/?l=linux-kernel&m=115315508506996&w=2

I am running all of my machines with this patch and rebooted a couple of
them with KNOPPIX and checked the FS, it seems to be OK now.

Justin.

Re: 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Daniel Drake]

Justin Piszcz wrote:
> Beginning at 2.6.17 to 2.6.17.6, there is a serious XFS bug that results 
> in filesystem corruption, there was a 1 line bugfix patch that was 
> released recently and I was wondering when 2.6.17.7 would be released 
> with that patch?  It affected ALL my Linux machines (x86) running XFS 
> and many people on the XFS mailing list who upgraded to 2.6.17.  I 
> understand when there is a root exploit or DoS bug, the kernel is 
> naturally patched by the -stable team and a new version is released 
> immediately.  Does filesystem corruption not constitute an immediate new 
> -stable release of the kernel?

Greg has been too busy at OLS, expect it in the next few days.

Additionally, some problems have been reported with the forty-something 
patches that were posted for review. If you want to help, apply all of 
those patches on top of 2.6.17.6 and see if you can find any problems.

Daniel

Re: 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Nathan Scott]

On Mon, Jul 24, 2006 at 12:29:48PM -0400, Justin Piszcz wrote:
> Beginning at 2.6.17 to 2.6.17.6, there is a serious XFS bug that results 
> in filesystem corruption, there was a 1 line bugfix patch that was 
> released recently and I was wondering when 2.6.17.7 would be released with 
> that patch? ...

I think everyone's been real busy leading up to OLS, and many folks
are probably taking some time off in Canada too.  I know I would be
if I was over there. ;)

cheers.

-- 
Nathan

Re: 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Matthias Andree]

On Tue, 25 Jul 2006, Nathan Scott wrote:

> On Mon, Jul 24, 2006 at 12:29:48PM -0400, Justin Piszcz wrote:
> > Beginning at 2.6.17 to 2.6.17.6, there is a serious XFS bug that results 
> > in filesystem corruption, there was a 1 line bugfix patch that was 
> > released recently and I was wondering when 2.6.17.7 would be released with 
> > that patch? ...
> 
> I think everyone's been real busy leading up to OLS, and many folks
> are probably taking some time off in Canada too.  I know I would be
> if I was over there. ;)

I don't mean to be nasty, but the effect this has on readers, in spite
of the smiley, is that "Linux hasn't grown up yet if they find Ottawa so
much kewler than fixing a simple bug".  Meaning: the downside of public
discussion is that it's easy to be mistaken.

A serious suggestion is, providing that the arguments presented (people
busy with OLS preparations and long review queue for 2.6.17.N+1):

How about doing 2.6.17.7 as an interim release fixing just what is known
to be critical at the point of the release, and then review the rest for
2.6.17.8? That would nicely fit the release early, release often - users
like that particularly for bug fixes.

(Please remember to Cc: replies, particularly I'm not subscribed to stable)

-- 
Matthias "2.6.16.27 works for my Samba" Andree

Re: 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Matthias Andree]

On Tue, 25 Jul 2006, Matthias Andree wrote:

> A serious suggestion is, providing that the arguments presented (people
> busy with OLS preparations and long review queue for 2.6.17.N+1):
> 
> How about doing 2.6.17.7 as an interim release fixing just what is known
> to be critical at the point of the release, and then review the rest for
> 2.6.17.8? That would nicely fit the release early, release often - users
> like that particularly for bug fixes.

OK, 2.6.17.7 is out, but still - is this suggestion worthwhile
considering for future -stable release engineering or just crap?

-- 
Matthias Andree

Re: [stable] 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Greg KH]

On Tue, Jul 25, 2006 at 11:09:19PM +0200, Matthias Andree wrote:
> On Tue, 25 Jul 2006, Matthias Andree wrote:
> 
> > A serious suggestion is, providing that the arguments presented (people
> > busy with OLS preparations and long review queue for 2.6.17.N+1):
> > 
> > How about doing 2.6.17.7 as an interim release fixing just what is known
> > to be critical at the point of the release, and then review the rest for
> > 2.6.17.8? That would nicely fit the release early, release often - users
> > like that particularly for bug fixes.
> 
> OK, 2.6.17.7 is out, but still - is this suggestion worthwhile
> considering for future -stable release engineering or just crap?

.7 took a bit longer than expected, due to some security bugs that
needed to be added to the queue, combined with the fact that both Chris
and I were busy with OLS stuff.  Normally we both aren't travelling at
the same time, but right then, we were, so we couldn't respond as
quickly as it seems some people felt we should have.

Sorry about this, we'll try to do better next time.

greg k-h

Re: [stable] 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Dave Jones]

On Fri, Jul 28, 2006 at 04:26:54PM -0700, Greg Kroah-Hartman wrote:

 > > OK, 2.6.17.7 is out, but still - is this suggestion worthwhile
 > > considering for future -stable release engineering or just crap?
 > 
 > .7 took a bit longer than expected, due to some security bugs that
 > needed to be added to the queue, combined with the fact that both Chris
 > and I were busy with OLS stuff.  Normally we both aren't travelling at
 > the same time, but right then, we were, so we couldn't respond as
 > quickly as it seems some people felt we should have.
 > 
 > Sorry about this, we'll try to do better next time.

The flipside to this is that those patches had been posted for around a
week before you released .7, and *no-one* caught this problem until
after the release.

The burden of testing shouldn't solely be on the -stable team.
Perhaps a -pre release at the time of review would be a good idea.
Just a roll-up of the proposed patches, to save testers having
to save and apply 30 patches seperately ?

		Dave

-- 
http://www.codemonkey.org.uk

Re: [stable] 2.6.17.[1-6] XFS Filesystem Corruption, Where is 2.6.17.7?

[Greg KH]

On Fri, Jul 28, 2006 at 07:55:34PM -0400, Dave Jones wrote:
> On Fri, Jul 28, 2006 at 04:26:54PM -0700, Greg Kroah-Hartman wrote:
> 
>  > > OK, 2.6.17.7 is out, but still - is this suggestion worthwhile
>  > > considering for future -stable release engineering or just crap?
>  > 
>  > .7 took a bit longer than expected, due to some security bugs that
>  > needed to be added to the queue, combined with the fact that both Chris
>  > and I were busy with OLS stuff.  Normally we both aren't travelling at
>  > the same time, but right then, we were, so we couldn't respond as
>  > quickly as it seems some people felt we should have.
>  > 
>  > Sorry about this, we'll try to do better next time.
> 
> The flipside to this is that those patches had been posted for around a
> week before you released .7, and *no-one* caught this problem until
> after the release.
> 
> The burden of testing shouldn't solely be on the -stable team.
> Perhaps a -pre release at the time of review would be a good idea.
> Just a roll-up of the proposed patches, to save testers having
> to save and apply 30 patches seperately ?

I can do that, it's very simple to do using quilt.  Others have asked
for this too.  I'll try it out with the next release.

thanks,

greg k-h