Re: [PATCH 0/5] reiserfs lock inversion fixes on xattr

Re: [PATCH 0/5] reiserfs lock inversion fixes on xattr

[Christian Kujau]

On Wed, 30 Dec 2009 at 21:42, Frederic Weisbecker wrote:
> I hope you can give it a try (and/or review) before I send the whole
> to Linus.

Without your patches, "dbench -x" was pretty much instantly causing the 
threads to lock up, forcing me to reboot the box. With LOCKDEP enabled, 
warnings were printed. After applying your patches (and a reboot) I've run 
"dbench -x" for quite a while now with different runtimes and number of
clients and it's not locking up any more. I've tested with:

CONFIG_REISERFS_FS=y
CONFIG_REISERFS_CHECK=y
CONFIG_REISERFS_PROC_INFO=y
CONFIG_REISERFS_FS_XATTR=y
CONFIG_REISERFS_FS_POSIX_ACL=y
# CONFIG_REISERFS_FS_SECURITY is not set

...and a few "Kernel hacking" options set. Full config and dmesg:

   http://nerdbynature.de/bits/2.6.33-rc1/reiserfs/t3/

   (although I'm not really testing -rc1 but latest mainline -git
    and your changes pulled)

So, feel free to add:

  Tested-by: Christian Kujau <lists@nerdbynature.de>

I can't help with the code review though :-\

Thanks!
Christian.
-- 
BOFH excuse #208:

Your mail is being routed through Germany ... and they're censoring us.

Re: [PATCH 0/5] reiserfs lock inversion fixes on xattr

[Ingo Molnar]

* Christian Kujau <lists@nerdbynature.de> wrote:

> On Wed, 30 Dec 2009 at 21:42, Frederic Weisbecker wrote:
> > I hope you can give it a try (and/or review) before I send the whole
> > to Linus.
> 
> Without your patches, "dbench -x" was pretty much instantly causing the 
> threads to lock up, forcing me to reboot the box. With LOCKDEP enabled, 
> warnings were printed. After applying your patches (and a reboot) I've run 
> "dbench -x" for quite a while now with different runtimes and number of
> clients and it's not locking up any more. I've tested with:
> 
> CONFIG_REISERFS_FS=y
> CONFIG_REISERFS_CHECK=y
> CONFIG_REISERFS_PROC_INFO=y
> CONFIG_REISERFS_FS_XATTR=y
> CONFIG_REISERFS_FS_POSIX_ACL=y
> # CONFIG_REISERFS_FS_SECURITY is not set
> 
> ...and a few "Kernel hacking" options set. Full config and dmesg:
> 
>    http://nerdbynature.de/bits/2.6.33-rc1/reiserfs/t3/
> 
>    (although I'm not really testing -rc1 but latest mainline -git
>     and your changes pulled)
> 
> So, feel free to add:
> 
>   Tested-by: Christian Kujau <lists@nerdbynature.de>

Frederic, i'd suggest you send this tree to Linus with Christian's tested-by 
tags added ASAP - we better have these fixes in -rc3.

Thanks,

	Ingo

Re: [PATCH 0/5] reiserfs lock inversion fixes on xattr

[Frederic Weisbecker]

On Wed, Dec 30, 2009 at 10:32:04PM -0800, Christian Kujau wrote:
> On Wed, 30 Dec 2009 at 21:42, Frederic Weisbecker wrote:
> > I hope you can give it a try (and/or review) before I send the whole
> > to Linus.
> 
> Without your patches, "dbench -x" was pretty much instantly causing the 
> threads to lock up, forcing me to reboot the box. With LOCKDEP enabled, 
> warnings were printed. After applying your patches (and a reboot) I've run 
> "dbench -x" for quite a while now with different runtimes and number of
> clients and it's not locking up any more. I've tested with:
> 
> CONFIG_REISERFS_FS=y
> CONFIG_REISERFS_CHECK=y
> CONFIG_REISERFS_PROC_INFO=y
> CONFIG_REISERFS_FS_XATTR=y
> CONFIG_REISERFS_FS_POSIX_ACL=y
> # CONFIG_REISERFS_FS_SECURITY is not set
> 
> ...and a few "Kernel hacking" options set. Full config and dmesg:
> 
>    http://nerdbynature.de/bits/2.6.33-rc1/reiserfs/t3/
> 
>    (although I'm not really testing -rc1 but latest mainline -git
>     and your changes pulled)
> 
> So, feel free to add:
> 
>   Tested-by: Christian Kujau <lists@nerdbynature.de>


Great, I've added this to the xattr related commits,
thanks a lot!


> 
> I can't help with the code review though :-\


Don't worry, both testing and review are very precious
and appreciated!

Re: [PATCH 0/5] reiserfs lock inversion fixes on xattr

[Frederic Weisbecker]

On Fri, Jan 01, 2010 at 10:29:44AM +0100, Ingo Molnar wrote:
> 
> * Christian Kujau <lists@nerdbynature.de> wrote:
> 
> > On Wed, 30 Dec 2009 at 21:42, Frederic Weisbecker wrote:
> > > I hope you can give it a try (and/or review) before I send the whole
> > > to Linus.
> > 
> > Without your patches, "dbench -x" was pretty much instantly causing the 
> > threads to lock up, forcing me to reboot the box. With LOCKDEP enabled, 
> > warnings were printed. After applying your patches (and a reboot) I've run 
> > "dbench -x" for quite a while now with different runtimes and number of
> > clients and it's not locking up any more. I've tested with:
> > 
> > CONFIG_REISERFS_FS=y
> > CONFIG_REISERFS_CHECK=y
> > CONFIG_REISERFS_PROC_INFO=y
> > CONFIG_REISERFS_FS_XATTR=y
> > CONFIG_REISERFS_FS_POSIX_ACL=y
> > # CONFIG_REISERFS_FS_SECURITY is not set
> > 
> > ...and a few "Kernel hacking" options set. Full config and dmesg:
> > 
> >    http://nerdbynature.de/bits/2.6.33-rc1/reiserfs/t3/
> > 
> >    (although I'm not really testing -rc1 but latest mainline -git
> >     and your changes pulled)
> > 
> > So, feel free to add:
> > 
> >   Tested-by: Christian Kujau <lists@nerdbynature.de>
> 
> Frederic, i'd suggest you send this tree to Linus with Christian's tested-by 
> tags added ASAP - we better have these fixes in -rc3.
> 
> Thanks,
> 
> 	Ingo


Yep, done :)