Re: linux-next: Tree for April 1 [BROKEN ubifs when CONFIG_DEBUG_SECTION_MISMATCH=y]s

[Jens Axboe <jaxboe@fusionio.com>]

On 2011-04-04 00:38, Sedat Dilek wrote:
> On Mon, Apr 4, 2011 at 12:36 AM, Jens Axboe <jaxboe@fusionio.com> wrote:
>> On 2011-04-04 00:33, Sedat Dilek wrote:
>>> On Mon, Apr 4, 2011 at 12:21 AM, Jens Axboe <jaxboe@fusionio.com> wrote:
>>>> On 2011-04-04 00:19, Sedat Dilek wrote:
>>>>> On Mon, Apr 4, 2011 at 12:16 AM, Jens Axboe <jaxboe@fusionio.com> wrote:
>>>>>> On 2011-04-02 13:02, Sedat Dilek wrote:
>>>>>>> On Sat, Apr 2, 2011 at 12:14 PM, Sedat Dilek <sedat.dilek@googlemail.com> wrote:
>>>>>>>> On Sat, Apr 2, 2011 at 2:20 AM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>>>>>>>>> cc'ing Jens ...
>>>>>>>>>
>>>>>>>>> On Fri, 1 Apr 2011 20:22:41 +0200 Sedat Dilek <sedat.dilek@googlemail.com> wrote:
>>>>>>>>>>
>>>>>>>>>> On Fri, Apr 1, 2011 at 7:02 PM, Artem Bityutskiy <dedekind1@gmail.com> wrote:
>>>>>>>>>>> On Fri, 2011-04-01 at 18:10 +0200, Sedat Dilek wrote:
>>>>>>>>>>>> On Fri, Apr 1, 2011 at 6:06 PM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>>>>>>>>>>>>> Cc'ing Artem,
>>>>>>>>>>>>>
>>>>>>>>>>>>> On Fri, 1 Apr 2011 17:55:52 +0200 Sedat Dilek <sedat.dilek@googlemail.com> wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> With CONFIG_DEBUG_SECTION_MISMATCH=y set, I see in my build.log:
>>>>>>>>>>>>>> ...
>>>>>>>>>>>>>> MODPOST 2742 modules
>>>>>>>>>>>>>> ...
>>>>>>>>>>>>>> ERROR: "empty_aops" [fs/ubifs/ubifs.ko] undefined!
>>>>>>>>>>>>>> make[5]: *** [__modpost] Error 1
>>>>>>>>>>>>>> make[4]: *** [modules] Error 2
>>>>>>>>>>>>>> make[3]: *** [sub-make] Error 2
>>>>>>>>>>>>>> make[2]: *** [all] Error 2
>>>>>>>>>>>>>> make[2]: Leaving directory
>>>>>>>>>>>>>> `/home/sd/src/linux-2.6/linux-2.6.39-rc1/debian/build/build_i386_none_686-iniza'
>>>>>>>>>>>>>>
>>>>>>> [...]
>>>>>>>> Just FYI:
>>>>>>>> I contacted Jens last night and he refreshed his for-linus GIT branch.
>>>>>>>> Adding missing include <linux/fs.h> did not fix the issue.
>>>>>>>> I am trying with the attached one.
>>>>>>>>
>>>>>>>> - Sedat -
>>>>>>>>
>>>>>>>
>>>>>>> I have split the single patch into two, first reflects ther build-error.
>>>>>>> The second considers {inode,file}_operations have also undefined
>>>>>>> functions by using "unified" empty_{iops,fops} as used in other fs/*
>>>>>>> files.
>>>>>>
>>>>>> What are these patches against? Not for-next nor my for-linus.
>>>>>>
>>>>>
>>>>> I tested with linux-next (next-20110401) as base and pulled-in your
>>>>> for-linus GIT branch.
>>>>
>>>> Then perhaps there was some merge error. There's no empty_aops defined
>>>> in my tree in nilfs_mapping_init(), for instance.
>>>>
>>>> Are you using an old for-linus?
>>>>
>>>
>>> I dropped the idea of exporting empty_aops via include/linux/fs.h (&
>>> changes in fs/inode.c) as it did not work as intended.
>>> As an alternative I used empty_{aops,iops,fops} only in
>>> fs/nilfs2/page.c and fs/ubifs/xattr.c where it is only needed (for
>>> example for aops: static const struct address_space_operations
>>> empty_aops {}; etc.)
>>
>> I'm asking one thing, you are replying with something else. The patches
>> you sent do NOT apply to for-linus.
>>
> 
> Oh, I am sorry. I checkout your tree and adapt.

I believe that you need is this, plain and simple.

diff --git a/fs/inode.c b/fs/inode.c
index b818730..33c963d 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -130,6 +130,7 @@ static DECLARE_RWSEM(iprune_sem);
  */
 const struct address_space_operations empty_aops = {
 };
+EXPORT_SYMBOL(empty_aops);
 
 /*
  * Statistics gathering..

-- 
Jens Axboe