linux-next: xfs tree build failure

linux-next: xfs tree build failure

[Stephen Rothwell]

Hi David,

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

ERROR: "inode_wait" [fs/xfs/xfs.ko] undefined!

Caused by commit 676a7711ee281c3bf41db18e29e296584de82793 ("[XFS] Fix
race when looking up reclaimable inodes") which introduced a use of
wait_on_inode() which is an inline that calls inode_wait().

I applied the following patch (probably wrong :-)).
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

From: Stephen Rothwell <sfr@canb.auug.org.au>
Date: Thu, 30 Oct 2008 13:19:46 +1100
Subject: [PATCH] fs: xfs needs inode_wait to be exported

Since wait_on_inode() references it.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 fs/inode.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index f84ba33..098a244 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -1334,6 +1334,7 @@ int inode_wait(void *word)
 	schedule();
 	return 0;
 }
+EXPORT_SYMBOL(inode_wait);
 
 /*
  * If we try to find an inode in the inode hash while it is being
-- 
1.5.6.5

Re: linux-next: xfs tree build failure

[Dave Chinner]

On Thu, Oct 30, 2008 at 01:23:24PM +1100, Stephen Rothwell wrote:
> Hi David,
> 
> Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> 
> ERROR: "inode_wait" [fs/xfs/xfs.ko] undefined!
> 
> Caused by commit 676a7711ee281c3bf41db18e29e296584de82793 ("[XFS] Fix
> race when looking up reclaimable inodes") which introduced a use of
> wait_on_inode() which is an inline that calls inode_wait().

Putting a non-exported symbol in an inline function in a header file
that is full of exported functions.  That's pretty anti-social. ;)

> I applied the following patch (probably wrong :-)).

I think it's probably the only solution - we need to wait until
the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
is the only way to do it.

Christoph?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: linux-next: xfs tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1025 bytes --]

Hi Dave,

On Thu, 30 Oct 2008 20:08:32 +1100 Dave Chinner <david@fromorbit.com> wrote:
>
> On Thu, Oct 30, 2008 at 01:23:24PM +1100, Stephen Rothwell wrote:
> > 
> > Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> > 
> > ERROR: "inode_wait" [fs/xfs/xfs.ko] undefined!
> > 
> > Caused by commit 676a7711ee281c3bf41db18e29e296584de82793 ("[XFS] Fix
> > race when looking up reclaimable inodes") which introduced a use of
> > wait_on_inode() which is an inline that calls inode_wait().
> 
> Putting a non-exported symbol in an inline function in a header file
> that is full of exported functions.  That's pretty anti-social. ;)
> 
> > I applied the following patch (probably wrong :-)).
> 
> I think it's probably the only solution - we need to wait until
> the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
> is the only way to do it.

Any resolution?

-- 
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: xfs tree build failure

[Dave Chinner]

On Wed, Nov 05, 2008 at 01:20:00PM +1100, Stephen Rothwell wrote:
> Hi Dave,
> 
> On Thu, 30 Oct 2008 20:08:32 +1100 Dave Chinner <david@fromorbit.com> wrote:
> >
> > On Thu, Oct 30, 2008 at 01:23:24PM +1100, Stephen Rothwell wrote:
> > > 
> > > Today's linux-next build (powerpc ppc64_defconfig) failed like this:
> > > 
> > > ERROR: "inode_wait" [fs/xfs/xfs.ko] undefined!
> > > 
> > > Caused by commit 676a7711ee281c3bf41db18e29e296584de82793 ("[XFS] Fix
> > > race when looking up reclaimable inodes") which introduced a use of
> > > wait_on_inode() which is an inline that calls inode_wait().
> > 
> > Putting a non-exported symbol in an inline function in a header file
> > that is full of exported functions.  That's pretty anti-social. ;)
> > 
> > > I applied the following patch (probably wrong :-)).
> > 
> > I think it's probably the only solution - we need to wait until
> > the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
> > is the only way to do it.
> 
> Any resolution?

I was waiting on a comment from Christoph. I'll cc him directly this
time ;)

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: linux-next: xfs tree build failure

[Christoph Hellwig]

On Thu, Nov 06, 2008 at 07:47:54AM +1100, Dave Chinner wrote:
> > > I think it's probably the only solution - we need to wait until
> > > the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
> > > is the only way to do it.
> > 
> > Any resolution?
> 
> I was waiting on a comment from Christoph. I'll cc him directly this
> time ;)

What comment do you expect from me?  The fix looks correct, but at some
point you should probably run all these exports past Al to get another
review.  Especially as he's doing major work in this area currently.

Re: linux-next: xfs tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 883 bytes --]

Hi all,

On Thu, 6 Nov 2008 10:27:53 -0500 Christoph Hellwig <hch@infradead.org> wrote:
>
> On Thu, Nov 06, 2008 at 07:47:54AM +1100, Dave Chinner wrote:
> > > > I think it's probably the only solution - we need to wait until
> > > > the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
> > > > is the only way to do it.
> > > 
> > > Any resolution?
> > 
> > I was waiting on a comment from Christoph. I'll cc him directly this
> > time ;)
> 
> What comment do you expect from me?  The fix looks correct, but at some
> point you should probably run all these exports past Al to get another
> review.  Especially as he's doing major work in this area currently.

OK, I have dropped the xfs tree for today.  Let me know when it is sorted
out.
-- 
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: xfs tree build failure

[Lachlan McIlroy]

Stephen Rothwell wrote:
> Hi all,
> 
> On Thu, 6 Nov 2008 10:27:53 -0500 Christoph Hellwig <hch@infradead.org> wrote:
>> On Thu, Nov 06, 2008 at 07:47:54AM +1100, Dave Chinner wrote:
>>>>> I think it's probably the only solution - we need to wait until
>>>>> the I_LOCK bit is cleared from the inode and AFAICT wait_on_inode()
>>>>> is the only way to do it.
>>>> Any resolution?
>>> I was waiting on a comment from Christoph. I'll cc him directly this
>>> time ;)
>> What comment do you expect from me?  The fix looks correct, but at some
>> point you should probably run all these exports past Al to get another
>> review.  Especially as he's doing major work in this area currently.
> 
> OK, I have dropped the xfs tree for today.  Let me know when it is sorted
> out.

Okay, your fix is in and it is building again.

Re: linux-next: xfs tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 338 bytes --]

Hi Lachlan,

On Tue, 11 Nov 2008 13:05:14 +1100 Lachlan McIlroy <lachlan@sgi.com> wrote:
>
> Okay, your fix is in and it is building again.

Thanks, I saw that this morning, so the xfs tree should be in linux-next
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 --]