[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Joseph Qi]

If we use le32_add_cpu to set ocfs2_dinode i_flags, it may lead to the
corresponding flag corrupted. So we should change it to bitwise and/or
operation.

Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
---
 fs/ocfs2/namei.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index 04ee1b5..f53471d 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -522,7 +522,7 @@ static int __ocfs2_mknod_locked(struct inode *dir,

 	fe->i_last_eb_blk = 0;
 	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
-	le32_add_cpu(&fe->i_flags, OCFS2_VALID_FL);
+	fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL);
 	fe->i_atime = fe->i_ctime = fe->i_mtime =
 		cpu_to_le64(CURRENT_TIME.tv_sec);
 	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
@@ -2044,7 +2044,7 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
 		goto leave;
 	}

-	le32_add_cpu(&fe->i_flags, OCFS2_ORPHANED_FL);
+	fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL);
 	OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;

 	/* Record which orphan dir our inode now resides
@@ -2434,7 +2434,7 @@ int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
 	}

 	di = (struct ocfs2_dinode *)di_bh->b_data;
-	le32_add_cpu(&di->i_flags, -OCFS2_ORPHANED_FL);
+	di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL);
 	di->i_orphaned_slot = 0;
 	set_nlink(inode, 1);
 	ocfs2_set_links_count(di, inode->i_nlink);
-- 
1.7.9.7

[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Jeff Liu]

On 05/30/2013 12:49 PM, Joseph Qi wrote:

> If we use le32_add_cpu to set ocfs2_dinode i_flags, it may lead to the
> corresponding flag corrupted. So we should change it to bitwise and/or
> operation.
> 
> Signed-off-by: Joseph Qi <joseph.qi@huawei.com>

Reviewed-by: Jie Liu <jeff.liu@oracle.com>

> ---
>  fs/ocfs2/namei.c |    6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
> index 04ee1b5..f53471d 100644
> --- a/fs/ocfs2/namei.c
> +++ b/fs/ocfs2/namei.c
> @@ -522,7 +522,7 @@ static int __ocfs2_mknod_locked(struct inode *dir,
> 
>  	fe->i_last_eb_blk = 0;
>  	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
> -	le32_add_cpu(&fe->i_flags, OCFS2_VALID_FL);
> +	fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL);
>  	fe->i_atime = fe->i_ctime = fe->i_mtime =
>  		cpu_to_le64(CURRENT_TIME.tv_sec);
>  	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
> @@ -2044,7 +2044,7 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
>  		goto leave;
>  	}
> 
> -	le32_add_cpu(&fe->i_flags, OCFS2_ORPHANED_FL);
> +	fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL);
>  	OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;
> 
>  	/* Record which orphan dir our inode now resides
> @@ -2434,7 +2434,7 @@ int ocfs2_mv_orphaned_inode_to_new(struct inode *dir,
>  	}
> 
>  	di = (struct ocfs2_dinode *)di_bh->b_data;
> -	le32_add_cpu(&di->i_flags, -OCFS2_ORPHANED_FL);
> +	di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL);
>  	di->i_orphaned_slot = 0;
>  	set_nlink(inode, 1);
>  	ocfs2_set_links_count(di, inode->i_nlink);

[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Andrew Morton]

On Thu, 30 May 2013 12:49:46 +0800 Joseph Qi <joseph.qi@huawei.com> wrote:

> If we use le32_add_cpu to set ocfs2_dinode i_flags, it may lead to the
> corresponding flag corrupted. So we should change it to bitwise and/or
> operation.
> 

The usual question: what is the end-user impact of the bug which was
just fixed?

For the usual reason: so I and others can decide which kernel versions
need the fix.

Thanks.

[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Jeff Liu]

Hi Andrew,

Sorry for the late response because I just returned from the LinuxCon/Japan.

I'm afraid that Joesph can not follow this thread up in time since today is the
weekend in China.  So please let me answer your question on behalf of Joesph
as a quick response.

On 06/01/2013 06:25 AM, Andrew Morton wrote:

> On Thu, 30 May 2013 12:49:46 +0800 Joseph Qi <joseph.qi@huawei.com> wrote:
> 
>> If we use le32_add_cpu to set ocfs2_dinode i_flags, it may lead to the
>> corresponding flag corrupted. So we should change it to bitwise and/or
>> operation.
>>
> 
> The usual question: what is the end-user impact of the bug which was
> just fixed?

This patch can be treated as a trivial fix.  Yes, i_flags is operated in
bitwise context, but those assignments are only happened at the time of
initializing the corresponding OCFS2 private on-disk inode info.  Hence
there is no impact from the end-user's perspective.

However, it's better to replace those three assignments with bitwise operations
since they essentially should be.  In this way, we can immediately know that the
business performed on i_flags are in bitwise rather than increasing or decreasing
a count value. 

> For the usual reason: so I and others can decide which kernel versions
> need the fix.

The current fix comments looks too horrible than it would be.
Can it make your life easier if I repost this patch with the revised
descriptions if Joseph agrees to me?


Thanks,
-Jeff

[Ocfs2-devel] [PATCH v2] ocfs2: should not use le32_add_cpu to set ocfs2_dinode i_flags

[Joseph Qi]

Hi Jeff & Andrew,
We found this bug when we do analysis for a read-only filesystem bug
that is directly caused by dinode flags mismatch.
Of course Jeff you could fix the comments and make it comprehensible.

On 2013/6/1 22:34, Jeff Liu wrote:
> Hi Andrew,
> 
> Sorry for the late response because I just returned from the LinuxCon/Japan.
> 
> I'm afraid that Joesph can not follow this thread up in time since today is the
> weekend in China.  So please let me answer your question on behalf of Joesph
> as a quick response.
> 
> On 06/01/2013 06:25 AM, Andrew Morton wrote:
> 
>> On Thu, 30 May 2013 12:49:46 +0800 Joseph Qi <joseph.qi@huawei.com> wrote:
>>
>>> If we use le32_add_cpu to set ocfs2_dinode i_flags, it may lead to the
>>> corresponding flag corrupted. So we should change it to bitwise and/or
>>> operation.
>>>
>>
>> The usual question: what is the end-user impact of the bug which was
>> just fixed?
> 
> This patch can be treated as a trivial fix.  Yes, i_flags is operated in
> bitwise context, but those assignments are only happened at the time of
> initializing the corresponding OCFS2 private on-disk inode info.  Hence
> there is no impact from the end-user's perspective.
> 
> However, it's better to replace those three assignments with bitwise operations
> since they essentially should be.  In this way, we can immediately know that the
> business performed on i_flags are in bitwise rather than increasing or decreasing
> a count value. 
> 
>> For the usual reason: so I and others can decide which kernel versions
>> need the fix.
> 
> The current fix comments looks too horrible than it would be.
> Can it make your life easier if I repost this patch with the revised
> descriptions if Joseph agrees to me?
> 
> 
> Thanks,
> -Jeff
> 
> .
>