Mailing List Archive: 2.6.19 going stable in 1 week

2.6.19 going stable in 1 week

Jan 7, 2007, 10:27 AM

Post #1 of 6 (266 views)

I will be asking the x86 and amd64 arch teams to mark the latest
revision of gentoo-sources-2.6.19 stable in 1 weeks time (Jan 14th).

This is a 1 week advance warning to maintainers of external kernel
modules which are still broken in the stable tree -- fix them, or file
new bugs asking for newer versions to go stable and block #156669

Actually, 1 week is the best-case scenario, since we have a couple of
minor regressions which need fixing. Additional pressure comes from the
mmap shared page corruption fix, a significant bug which has existed for
at least 3 years. Unfortunately the fix has not been backported to older
kernels and it's not clear how to do so.

Daniel
--
gentoo-dev@gentoo.org mailing list

Re: 2.6.19 going stable in 1 week [ In reply to ]

remi at gentoo

Jan 7, 2007, 12:35 PM

Post #2 of 6 (263 views)

Permalink

Daniel Drake wrote:
> Actually, 1 week is the best-case scenario, since we have a couple of
> minor regressions which need fixing. Additional pressure comes from the
> mmap shared page corruption fix, a significant bug which has existed for
> at least 3 years. Unfortunately the fix has not been backported to older
> kernels and it's not clear how to do so.

About that bug, has anyone filed anything for it in Gentoo or is it the
kind of bug that creeps up anywhere and can look like something else is
responsible for it?

Just wondering :)

Rémi

--
gentoo-dev@gentoo.org mailing list

Re: 2.6.19 going stable in 1 week [ In reply to ]

dsd at gentoo

Jan 7, 2007, 1:52 PM

Post #3 of 6 (261 views)

Permalink

Rémi Cardona wrote:
> About that bug, has anyone filed anything for it in Gentoo or is it the
> kind of bug that creeps up anywhere and can look like something else is
> responsible for it?

Now that the problem is understood, it can be reproduced trivially on
any kernel and can easily result in filesystem corruption.

Daniel
--
gentoo-dev@gentoo.org mailing list

Re: Re: 2.6.19 going stable in 1 week [ In reply to ]

chutz at gg3

Jan 22, 2007, 5:08 PM

Post #4 of 6 (264 views)

Permalink

Quoting Hendrik Boom <hendrik@topoi.pooq.com>:

> I would strongly advise against considering 2.6.19 stable
> until this bug has been fixed.

From the Changelog of 2.6.19.2

commit 54e25b0460e6b1100e7ef9c0ac801bdce83921c0
Author: Linus Torvalds <torvalds@macmini.osdl.org>
Date: Fri Dec 29 10:00:58 2006 -0800

[PATCH] VM: Fix nasty and subtle race in shared mmap'ed page writeback

Sounds like the one to me...

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

--
gentoo-dev@gentoo.org mailing list

Re: Re: 2.6.19 going stable in 1 week [ In reply to ]

hendrik at topoi

Jun 17, 2008, 1:15 AM

Post #5 of 6 (264 views)

Permalink

On Tue, 23 Jan 2007 09:08:47 +0900, Georgi Georgiev wrote:

> Quoting Hendrik Boom <hendrik@topoi.pooq.com>:
>
>> I would strongly advise against considering 2.6.19 stable
>> until this bug has been fixed.
>
> From the Changelog of 2.6.19.2
>
> commit 54e25b0460e6b1100e7ef9c0ac801bdce83921c0
> Author: Linus Torvalds <torvalds@macmini.osdl.org>
> Date: Fri Dec 29 10:00:58 2006 -0800
>
> [PATCH] VM: Fix nasty and subtle race in shared mmap'ed page writeback
>
> Sounds like the one to me...

Sure does!

-- hendrik

--
gentoo-dev@gentoo.org mailing list

Re: 2.6.19 going stable in 1 week [ In reply to ]

hendrik at topoi

Jun 17, 2008, 1:15 AM

Post #6 of 6 (262 views)

Permalink

On Sun, 07 Jan 2007 15:52:54 -0500, Daniel Drake wrote:

> Rémi Cardona wrote:
>> About that bug, has anyone filed anything for it in Gentoo or is it the
>> kind of bug that creeps up anywhere and can look like something else is
>> responsible for it?
>
>
> Now that the problem is understood, it can be reproduced trivially on
> any kernel and can easily result in filesystem corruption.
>
> Daniel
I've hit this bug in Debian, where they backported it from 2.6.19 to
2.6.18. It doesn't seem to hit everyone, but it can be a real disaster if
it hits you. Debian etch won't even install on one of my
machines because of severe file-system corruption during installation!

I would strongly advise against considering 2.6.19 stable
until this bug has been fixed.

-- hendrik

P.S. Actually Debian backported some useful code from 2.6.19 that
happened to activate the bug.

-- hendrik

--
gentoo-dev@gentoo.org mailing list