allocator options, kernel 2.6.7

allocator options, kernel 2.6.7

[Pierre Etchemaite]

	Hi all,

I've been using 2.6.5 kernel with data-logging, allocator options patch, and

	data=ordered,alloc=skip_busy:dirid_groups,packing_groups

mount options for a while. So far so good.


Now, two questions.

1. I came across this fix:
http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.6/2.6.6-mm5/broken-out/reiserfs-block-allocator-should-not-inherit-packing-locality.patch

Does that mean that I could have problem fixing some of my filesystems if
they happened to need reiserfsck ?

(From what I understand I can check if it's not too late already if

/sbin/debugreiserfs -d /dev/xxxx |grep '^|.*SD'|sed 's/^.....//'|awk '$1==1'

returns only one line, beginning with 1 2 ...)


2. kernel 2.6.7 seems to have data-logging patch applied (which is a very
good news!), but not advanced allocator options patch. Is there a patch for
2.6.7, or would you recommend me forgetting about those options for now ?
Allocator tweaks looks a bit more experimental than I thought.

Best regards,
Pierre.

Re: allocator options, kernel 2.6.7

[Chris Mason]

[-- Attachment #1: Type: text/plain, Size: 1425 bytes --]

On Wed, 2004-06-16 at 19:29, Pierre Etchemaite wrote:
> 	Hi all,
> 
> I've been using 2.6.5 kernel with data-logging, allocator options patch, and
> 
> 	data=ordered,alloc=skip_busy:dirid_groups,packing_groups
> 
> mount options for a while. So far so good.
> 
> 
> Now, two questions.
> 
> 1. I came across this fix:
> http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.6/2.6.6-mm5/broken-out/reiserfs-block-allocator-should-not-inherit-packing-locality.patch
> 
> Does that mean that I could have problem fixing some of my filesystems if
> they happened to need reiserfsck ?
> 
There's a reiserfsprogs patch as well, which was sent to Vitaly.  I've
attached it to this message.

> (From what I understand I can check if it's not too late already if
> 
> /sbin/debugreiserfs -d /dev/xxxx |grep '^|.*SD'|sed 's/^.....//'|awk '$1==1'
> 
> returns only one line, beginning with 1 2 ...)
> 
> 
> 2. kernel 2.6.7 seems to have data-logging patch applied (which is a very
> good news!), but not advanced allocator options patch. Is there a patch for
> 2.6.7, or would you recommend me forgetting about those options for now ?
> Allocator tweaks looks a bit more experimental than I thought.

It seemed best to keep the changes for 2.6.7 somewhat more limited.  The
allocator patches are going through a big QA round at SUSE as well right
now, so any remaining problems should get ironed out very shortly.

-chris


[-- Attachment #2: reiserfsck-packing-groups.diff --]
[-- Type: text/x-patch, Size: 659 bytes --]

--- reiserfsck-3.6.17/fsck/pass0.c	2004-05-20 20:15:19.312305008 -0400
+++ reiserfsck-3.6.17.fix/fsck/pass0.c	2004-05-20 20:16:41.366830824 -0400
@@ -565,7 +565,8 @@
     if (get_key_dirid (key) == 0 || get_key_objectid (key) == 0 || get_key_objectid (key) == 1 ||
 	get_key_dirid (key) == ~(__u32)0 || get_key_objectid (key) == ~(__u32)0 ||
 	get_key_dirid (key) == get_key_objectid (key) ||
-	(get_key_dirid (key) == 1 && get_key_objectid (key) != 2) ||
+        /* the alloc=packing_groups used to allow dirid = 1
+	(get_key_dirid (key) == 1 && get_key_objectid (key) != 2) || */
 	(get_key_dirid (key) != 1 && get_key_objectid (key) == 2) )
 	return 1;
 

Re: allocator options, kernel 2.6.7

[Pierre Etchemaite]

Le Wed, 16 Jun 2004 19:38:53 -0400, Chris Mason <mason@suse.com> a écrit :

> > Does that mean that I could have problem fixing some of my filesystems
> > if they happened to need reiserfsck ?
> > 
> There's a reiserfsprogs patch as well, which was sent to Vitaly.  I've
> attached it to this message.

Ok, so in the worst case filesystem can be checked, given that
used reiserfsck is recent enough. Getting last reiserfsck is always a good
advice anyway ;)

> It seemed best to keep the changes for 2.6.7 somewhat more limited.  The
> allocator patches are going through a big QA round at SUSE as well right
> now, so any remaining problems should get ironed out very shortly.

Once again doing a great work (and answering lightning fast ;) )

Many thanks,
Pierre.

Re: allocator options, kernel 2.6.7

[Mike Benoit]

If it helps at all, I've been running the new allocator on my
"Desktop" (used constantly for development, games, and as a NFS server
for two very busy MythTV boxes) since about May 11th with no problems
whatsoever.

Kernel: 2.6.7-rc2-mm2

Fstab:

/dev/hda5 / reiserfs notail,noatime,nodiratime,alloc=skip_busy:
dirid_groups:packing_groups 1 1
/dev/hda6 /home reiserfs notail,noatime,nodiratime,alloc=skip_busy:
dirid_groups:packing_groups 1 2


On Wed, 2004-06-16 at 19:38 -0400, Chris Mason wrote:
> On Wed, 2004-06-16 at 19:29, Pierre Etchemaite wrote:
> > 	Hi all,
> > 
> > I've been using 2.6.5 kernel with data-logging, allocator options patch, and
> > 
> > 	data=ordered,alloc=skip_busy:dirid_groups,packing_groups
> > 
> > mount options for a while. So far so good.
> > 
> > 
> > Now, two questions.
> > 
> > 1. I came across this fix:
> > http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.6/2.6.6-mm5/broken-out/reiserfs-block-allocator-should-not-inherit-packing-locality.patch
> > 
> > Does that mean that I could have problem fixing some of my filesystems if
> > they happened to need reiserfsck ?
> > 
> There's a reiserfsprogs patch as well, which was sent to Vitaly.  I've
> attached it to this message.
> 
> > (From what I understand I can check if it's not too late already if
> > 
> > /sbin/debugreiserfs -d /dev/xxxx |grep '^|.*SD'|sed 's/^.....//'|awk '$1==1'
> > 
> > returns only one line, beginning with 1 2 ...)
> > 
> > 
> > 2. kernel 2.6.7 seems to have data-logging patch applied (which is a very
> > good news!), but not advanced allocator options patch. Is there a patch for
> > 2.6.7, or would you recommend me forgetting about those options for now ?
> > Allocator tweaks looks a bit more experimental than I thought.
> 
> It seemed best to keep the changes for 2.6.7 somewhat more limited.  The
> allocator patches are going through a big QA round at SUSE as well right
> now, so any remaining problems should get ironed out very shortly.
> 
> -chris
> 
-- 
Mike Benoit <ipso@snappymail.ca>

Re: allocator options, kernel 2.6.7

[Dieter Nützel]

Chris Mason schrieb:
> On Wed, 2004-06-16 at 19:29, Pierre Etchemaite wrote:
> 
>>	Hi all,
>>
>>I've been using 2.6.5 kernel with data-logging, allocator options patch, and
>>
>>	data=ordered,alloc=skip_busy:dirid_groups,packing_groups
>>
>>mount options for a while. So far so good.
>>
>>
>>Now, two questions.
>>
>>1. I came across this fix:
>>http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.6/2.6.6-mm5/broken-out/reiserfs-block-allocator-should-not-inherit-packing-locality.patch
>>
>>Does that mean that I could have problem fixing some of my filesystems if
>>they happened to need reiserfsck ?
>>
> 
> There's a reiserfsprogs patch as well, which was sent to Vitaly.  I've
> attached it to this message.
> 
> 
>>(From what I understand I can check if it's not too late already if
>>
>>/sbin/debugreiserfs -d /dev/xxxx |grep '^|.*SD'|sed 's/^.....//'|awk '$1==1'
>>
>>returns only one line, beginning with 1 2 ...)
>>
>>
>>2. kernel 2.6.7 seems to have data-logging patch applied (which is a very
>>good news!), but not advanced allocator options patch. Is there a patch for
>>2.6.7, or would you recommend me forgetting about those options for now ?
>>Allocator tweaks looks a bit more experimental than I thought.
> 
> 
> It seemed best to keep the changes for 2.6.7 somewhat more limited.  The
> allocator patches are going through a big QA round at SUSE as well right
> now, so any remaining problems should get ironed out very shortly.
> 
> -chris
> 
> 
> 
> ------------------------------------------------------------------------
> 
> --- reiserfsck-3.6.17/fsck/pass0.c	2004-05-20 20:15:19.312305008 -0400
> +++ reiserfsck-3.6.17.fix/fsck/pass0.c	2004-05-20 20:16:41.366830824 -0400
> @@ -565,7 +565,8 @@
>      if (get_key_dirid (key) == 0 || get_key_objectid (key) == 0 || get_key_objectid (key) == 1 ||
>  	get_key_dirid (key) == ~(__u32)0 || get_key_objectid (key) == ~(__u32)0 ||
>  	get_key_dirid (key) == get_key_objectid (key) ||
> -	(get_key_dirid (key) == 1 && get_key_objectid (key) != 2) ||
> +        /* the alloc=packing_groups used to allow dirid = 1
> +	(get_key_dirid (key) == 1 && get_key_objectid (key) != 2) || */
>  	(get_key_dirid (key) != 1 && get_key_objectid (key) == 2) )
>  	return 1;

Hello Chris,

is there a reiserfsck-3.6.17 version (*.src.rpm) for SuSE 9.0/9.1 available?

Thanks,
	Dieter