linux-next: slab tree build failure

linux-next: slab tree build failure

[Stephen Rothwell]

Hi all,

Today's linux-next build (powerpc ppc64_defconfig) failed like this:

fs/xfs/linux-2.6/xfs_super.c: In function 'xfs_init_zones':
fs/xfs/linux-2.6/xfs_super.c:1714: error: 'xfs_vnode_zone' undeclared (first use in this function)

It is a mismerge (in the linux-next tree) between commits
7836dda725873a5ce05d216295a452f16ccaab72 ("[XFS] Combine the XFS and
Linux inodes") form the xfs tree and
200d770048651d9a03266dc1bf031bb6f9e1363a ("Filesystem: XFS slab
defragmentation") from the slab tree.

I guess the latter patch needs to be dropped since its target is going
away.

I fixed up the merge (see below) and can carry the fix for a few days.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index 23e0afb..6bd3e2a 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -1711,8 +1711,6 @@ STATIC int __init
 xfs_init_zones(void)
 {
 
-	kmem_cache_setup_defrag(xfs_vnode_zone, get_inodes, kick_inodes);
-
 	xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
 	if (!xfs_ioend_zone)
 		goto out;

Re: linux-next: slab tree build failure

[Dave Chinner]

On Mon, Nov 03, 2008 at 03:42:21PM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> 
> fs/xfs/linux-2.6/xfs_super.c: In function 'xfs_init_zones':
> fs/xfs/linux-2.6/xfs_super.c:1714: error: 'xfs_vnode_zone' undeclared (first use in this function)
> 
> It is a mismerge (in the linux-next tree) between commits
> 7836dda725873a5ce05d216295a452f16ccaab72 ("[XFS] Combine the XFS and
> Linux inodes") form the xfs tree and
> 200d770048651d9a03266dc1bf031bb6f9e1363a ("Filesystem: XFS slab
> defragmentation") from the slab tree.

I didn't even know that the slab defrag stuff was being carried in
a tree anywhere. The xfs_vnode_zone has gone away, so I guess that
hunk can be dropped. defrag should really now target the xfs_inode_zone,
but it's going to need custom get and kick methods because it's
far different to the generic inode code now....

> I guess the latter patch needs to be dropped since its target is going
> away.
> 
> I fixed up the merge (see below) and can carry the fix for a few days.

Yeah, that will work for the moment. It's probably best that we
don't try to add defrag back into the XFS inode code until we've
sorted all the niggling problems as a result of the "combine..."
patch.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: linux-next: slab tree build failure

[Pekka Enberg]

On Mon, 2008-11-03 at 16:14 +1100, Dave Chinner wrote:
> Yeah, that will work for the moment. It's probably best that we
> don't try to add defrag back into the XFS inode code until we've
> sorted all the niggling problems as a result of the "combine..."
> patch.

OK, I dropped the XFS defrag patch from my tree for now. Thanks guys!

		Pekka

linux-next: slab tree build failure

[Stephen Rothwell]

Hi all,

Today's linux-next build (powerpc ppc64_defconfig) failed like this:

mm/slqb.c:75: error: 'SLAB_GFP_BOOT_MASK' undeclared here (not in a function)

Caused by commit df548b0bbaf2d4d16ce52d2f5ccdc1cfded70e06 ("SLQB: Fix
early boot allocations") from the slab tree interacting with commit
dcce284a259373f9e5570f2e33f79eca84fcf565 ("mm: Extend gfp masking to the
page allocator") that entered Linus' tree today.

I have applied this patch for today:

From: Stephen Rothwell <sfr@canb.auug.org.au>
Date: Fri, 19 Jun 2009 14:13:21 +1000
Subject: [PATCH] slqb: fix for macro name change

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 mm/slqb.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/mm/slqb.c b/mm/slqb.c
index b61ad01..4c9051d 100644
--- a/mm/slqb.c
+++ b/mm/slqb.c
@@ -72,7 +72,7 @@ static int slqb_min_objects = 1;
  * The slab allocator is initialized with interrupts disabled. Therefore, make
  * sure early boot allocations don't accidentally enable interrupts.
  */
-static gfp_t slab_gfp_mask __read_mostly = SLAB_GFP_BOOT_MASK;
+static gfp_t slab_gfp_mask __read_mostly = GFP_BOOT_MASK;
 
 #ifdef CONFIG_NUMA
 static inline int slab_numa(struct kmem_cache *s)
-- 
1.6.3.1

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

Re: linux-next: slab tree build failure

[Benjamin Herrenschmidt]

On Fri, 2009-06-19 at 14:16 +1000, Stephen Rothwell wrote:
> Hi all,
> 
> Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> 
> mm/slqb.c:75: error: 'SLAB_GFP_BOOT_MASK' undeclared here (not in a function)
> 
> Caused by commit df548b0bbaf2d4d16ce52d2f5ccdc1cfded70e06 ("SLQB: Fix
> early boot allocations") from the slab tree interacting with commit
> dcce284a259373f9e5570f2e33f79eca84fcf565 ("mm: Extend gfp masking to the
> page allocator") that entered Linus' tree today.
> 
> I have applied this patch for today:

The right fix is to remove slab_gfp_mask from slqb.c and use
the global gfp_allowed_mask instead.

Cheers,
Ben.

> From: Stephen Rothwell <sfr@canb.auug.org.au>
> Date: Fri, 19 Jun 2009 14:13:21 +1000
> Subject: [PATCH] slqb: fix for macro name change
> 
> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
> ---
>  mm/slqb.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/mm/slqb.c b/mm/slqb.c
> index b61ad01..4c9051d 100644
> --- a/mm/slqb.c
> +++ b/mm/slqb.c
> @@ -72,7 +72,7 @@ static int slqb_min_objects = 1;
>   * The slab allocator is initialized with interrupts disabled. Therefore, make
>   * sure early boot allocations don't accidentally enable interrupts.
>   */
> -static gfp_t slab_gfp_mask __read_mostly = SLAB_GFP_BOOT_MASK;
> +static gfp_t slab_gfp_mask __read_mostly = GFP_BOOT_MASK;
>  
>  #ifdef CONFIG_NUMA
>  static inline int slab_numa(struct kmem_cache *s)
> -- 
> 1.6.3.1
> 

Re: linux-next: slab tree build failure

[Pekka Enberg]

On Fri, 2009-06-19 at 20:40 +1000, Benjamin Herrenschmidt wrote:
> On Fri, 2009-06-19 at 14:16 +1000, Stephen Rothwell wrote:
> > Hi all,
> > 
> > Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> > 
> > mm/slqb.c:75: error: 'SLAB_GFP_BOOT_MASK' undeclared here (not in a function)
> > 
> > Caused by commit df548b0bbaf2d4d16ce52d2f5ccdc1cfded70e06 ("SLQB: Fix
> > early boot allocations") from the slab tree interacting with commit
> > dcce284a259373f9e5570f2e33f79eca84fcf565 ("mm: Extend gfp masking to the
> > page allocator") that entered Linus' tree today.
> > 
> > I have applied this patch for today:
> 
> The right fix is to remove slab_gfp_mask from slqb.c and use
> the global gfp_allowed_mask instead.

This should be fixed now in slab.git. Something in SQLB is still
enabling interrupts during early boot but I don't have the time to look
at it in more detail right now.

			Pekka

linux-next: slab tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 653 bytes --]

Hi all,

Today's linux-next build (x86_64 allmodconfig) failed like this:

net/wireless/scan.c: In function 'cfg80211_inform_bss':
net/wireless/scan.c:499: error: implicit declaration of function 'kmemleak_ignore'

Caused by commit 74ce35c0d2261896ed4d49799311a894520f7a94 ("cfg80211:
clear cfg80211_inform_bss() from kmemleak reports") from the wireless
tree (which should have included linux/kmemleak.h into
net/wireless/scan.c).  Some interaction with the slab tree has exposed
this.

I have reverted the above commit for today.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: slab tree build failure

[Pekka Enberg]

Hi Stephen,

On Mon, Sep 7, 2009 at 8:37 AM, Stephen Rothwell<sfr@canb.auug.org.au> wrote:
> Hi all,
>
> Today's linux-next build (x86_64 allmodconfig) failed like this:
>
> net/wireless/scan.c: In function 'cfg80211_inform_bss':
> net/wireless/scan.c:499: error: implicit declaration of function 'kmemleak_ignore'
>
> Caused by commit 74ce35c0d2261896ed4d49799311a894520f7a94 ("cfg80211:
> clear cfg80211_inform_bss() from kmemleak reports") from the wireless
> tree (which should have included linux/kmemleak.h into
> net/wireless/scan.c).  Some interaction with the slab tree has exposed
> this.
>
> I have reverted the above commit for today.

AFAICT, you're interested in kmemleak.git, not slab.git. Catalin?

                                Pekka

Re: linux-next: slab tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 372 bytes --]

Hi Pekka,

On Mon, 7 Sep 2009 09:10:18 +0300 Pekka Enberg <penberg@cs.helsinki.fi> wrote:
>
> AFAICT, you're interested in kmemleak.git, not slab.git. Catalin?

I first got the error after merging the slab tree ... the kmemleak tree
is merged much later.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: slab tree build failure

[Pekka Enberg]

Hi Stephen,

On Mon, 7 Sep 2009 09:10:18 +0300 Pekka Enberg <penberg@cs.helsinki.fi> wrote:
> > AFAICT, you're interested in kmemleak.git, not slab.git. Catalin?

On Mon, 2009-09-07 at 17:31 +1000, Stephen Rothwell wrote:
> I first got the error after merging the slab tree ... the kmemleak tree
> is merged much later.

Aah, let me guess, CONFIG_SLQB? I guess I'll just rebase the damn tree
and make it non-default. We need that anyway if we want to merge it to
2.6.32. I'll take a stab at it later this evening (my time zone).

			Pekka

Re: linux-next: slab tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 526 bytes --]

Hi Pekka,

On Mon, 07 Sep 2009 10:43:46 +0300 Pekka Enberg <penberg@cs.helsinki.fi> wrote:
>
> Aah, let me guess, CONFIG_SLQB? I guess I'll just rebase the damn tree

That would be my guess.

> and make it non-default. We need that anyway if we want to merge it to
> 2.6.32. I'll take a stab at it later this evening (my time zone).

OK.  This breakage still needs to be fixed in the wireless tree, though.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: slab tree build failure

[Catalin Marinas]

On Mon, 2009-09-07 at 09:10 +0300, Pekka Enberg wrote:
> Hi Stephen,
> 
> On Mon, Sep 7, 2009 at 8:37 AM, Stephen Rothwell<sfr@canb.auug.org.au> wrote:
> > Hi all,
> >
> > Today's linux-next build (x86_64 allmodconfig) failed like this:
> >
> > net/wireless/scan.c: In function 'cfg80211_inform_bss':
> > net/wireless/scan.c:499: error: implicit declaration of function 'kmemleak_ignore'
> >
> > Caused by commit 74ce35c0d2261896ed4d49799311a894520f7a94 ("cfg80211:
> > clear cfg80211_inform_bss() from kmemleak reports") from the wireless
> > tree (which should have included linux/kmemleak.h into
> > net/wireless/scan.c).  Some interaction with the slab tree has exposed
> > this.
> >
> > I have reverted the above commit for today.
> 
> AFAICT, you're interested in kmemleak.git, not slab.git. Catalin?

I think that's a commit from Luis. It shouldn't have been merged in the
first instance anyway (we shouldn't randomly ignore kmemleak reports
unless we know for sure why kmemleak thinks it is a leak and we should
write a comment explaining why).

-- 
Catalin

Re: linux-next: slab tree build failure

[Heiko Carstens]

On Mon, Sep 07, 2009 at 10:43:46AM +0300, Pekka Enberg wrote:
> Aah, let me guess, CONFIG_SLQB? I guess I'll just rebase the damn tree
> and make it non-default. We need that anyway if we want to merge it to
> 2.6.32. I'll take a stab at it later this evening (my time zone).

Will SLQB be merged during the next merge window? Just realized that s390
doesn't boot with CONFIG_SLQB (crashes before the console is active).

I tend to ignore bugs which seem to be related to SLQB in linux-next,
but if it gets merged I might need to look deeper into this.

Re: linux-next: slab tree build failure

[Pekka Enberg]

Hi Heiko,

On Mon, Sep 07, 2009 at 10:43:46AM +0300, Pekka Enberg wrote:
>> Aah, let me guess, CONFIG_SLQB? I guess I'll just rebase the damn tree
>> and make it non-default. We need that anyway if we want to merge it to
>> 2.6.32. I'll take a stab at it later this evening (my time zone).

On Mon, Sep 7, 2009 at 12:04 PM, Heiko
Carstens<heiko.carstens@de.ibm.com> wrote:
> Will SLQB be merged during the next merge window? Just realized that s390
> doesn't boot with CONFIG_SLQB (crashes before the console is active).
>
> I tend to ignore bugs which seem to be related to SLQB in linux-next,
> but if it gets merged I might need to look deeper into this.

Yes, that's the plan. That said, there's a long-standing PPC boot-time
bug in SLQB that's still unfixed which could affect other archs as
well AFAICT. Help in tracking that down would be appreciated as Nick
seems to be busy atm.

I am not completely happy with sending something that's known to be
broken to Linus. However, the problem is not getting fixed in slab.git
and everyone seems to be in favor of merging the allocator and mark it
as "experimental" so I'll probably just send it Linus' way for 2.6.32
and let him decide whether to take it or not.

                        Pekka

Re: linux-next: slab tree build failure

[Luis R. Rodriguez]

On Mon, Sep 07, 2009 at 01:35:09AM -0700, Catalin Marinas wrote:
> On Mon, 2009-09-07 at 09:10 +0300, Pekka Enberg wrote:
> > Hi Stephen,
> >
> > On Mon, Sep 7, 2009 at 8:37 AM, Stephen Rothwell<sfr@canb.auug.org.au> wrote:
> > > Hi all,
> > >
> > > Today's linux-next build (x86_64 allmodconfig) failed like this:
> > >
> > > net/wireless/scan.c: In function 'cfg80211_inform_bss':
> > > net/wireless/scan.c:499: error: implicit declaration of function 'kmemleak_ignore'
> > >
> > > Caused by commit 74ce35c0d2261896ed4d49799311a894520f7a94 ("cfg80211:
> > > clear cfg80211_inform_bss() from kmemleak reports") from the wireless
> > > tree (which should have included linux/kmemleak.h into
> > > net/wireless/scan.c).  Some interaction with the slab tree has exposed
> > > this.
> > >
> > > I have reverted the above commit for today.
> >
> > AFAICT, you're interested in kmemleak.git, not slab.git. Catalin?
> 
> I think that's a commit from Luis. It shouldn't have been merged in the
> first instance anyway (we shouldn't randomly ignore kmemleak reports
> unless we know for sure why kmemleak thinks it is a leak and we should
> write a comment explaining why).

I asked John to revert it, feel free to do the same.

  Luis

linux-next: slab tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 752 bytes --]

Hi all,

Today's linux-next build (x86_64 allmodconfig) failed like this:

tests/slab_test.c: In function 'test_func':
tests/slab_test.c:176: error: implicit declaration of function 'set_cpus_allowed'

Caused by commit 1c999557258d251e2e5790de4b723d5996c06d5d ("Add slab
performance tests").

set_cpus_allowed it now only defined if CONFIG_CPUMASK_OFFSTACK is not
set.  This was changed by commit e0ad955680878998ff7dc51ce06ddad12260423a
("cpumask: don't define set_cpus_allowed() if CONFIG_CPUMASK_OFFSTACK=y")
which entered Linus' tree before 2.6.32-rc1.

I have used the version of the slab tree from next-20091204 for today.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

linux-next: slab tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 421 bytes --]

Hi all,

Today's linux-next build (x86_64 allmodconfig) failed like this:

scripts/Makefile.headersinst:15: include/tests/Kbuild: No such file or directory

Caused by commit 1c999557258d251e2e5790de4b723d5996c06d5d ("Add slab
performance tests").

I have used the slab tree from next-20091204 for today.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]