Re: kernel and d-i testing
reassign 401980 linux-2.6
retitle 401980 msync() fixes break NSLU2 (apt segfaults)
thanks
It seems the msync() LSB fixes applied in -6 break the NSLU2: during
d-i, apt-get segfaults. This does not happen on a system that is
already installed (only during d-i) and not on any other ARM machines.
Gordon will try to find out which of the msync() patches is the
culprint. Once we know that, we can decide what to do (i.e. whether
there's a fix for that patch or whether I'll just revert it on ARM
since LSB on ARM doesn't exist anyway afaik).
* Gordon Farquharson <gordonfarquharson@gmail.com> [2006-12-14 06:50]:
> The binary search is basically the approach I have taken, except I
> have been checking out revisions of the kernel from subversion, which
> amounts to the same thing, but is less efficient.
>
> And here is the result (drum roll). It appears that revision 7722 is
> the kernel where this problem starts. 7721 works, whereas 7722 and
> subsequent kernel revisions fail.
>
> Installer Kernel Subversion Result
> Date Version Revision
>
> 20061206 2.6.18-5 7720 Pass
> 20061206 7721 Pass
> 20061206 7722 Fail
> 20061206 7731 Fail
> 20061206 7815 Fail
> 20061206 2.6.18-6 7867 Fail
>
> Now for the bad news: there are six patch files applied to this
> reversion of the kernel, and they all seem to be related.
>
> 7722 changelog additions:
>
> * Fix msync() for LSB 3.1 compliance, backport fedora patches from
> 2.6.19
> - mm: tracking shared dirty pages
> - mm: balance dirty pages
> - mm: optimize the new mprotect() code a bit
> - mm: small cleanup of install_page()
> - mm: fixup do_wp_page()
> - mm: msync() cleanup (closes: 394392)
>
> + features/mm-tracking-shared-dirty-pages.patch
> + features/mm-balance-dirty-pages.patch
> + features/mm-optimize-mprotect.patch
> + features/mm-install_page-cleanup.patch
> + features/mm-do_wp_page-fixup.patch
> + features/mm-msync-cleanup.patch
>
> I don't have time this morning to check whether these patches can be
> applied independently. Could you have a look at them, and let me know.
> I may not be able to work on this again until Friday evening.
>
> Gordon
>
> --
> Gordon Farquharson
--
Martin Michlmayr
http://www.cyrius.com/
Reply to: