[Ocfs2-devel] [PATCH] ocfs2: free meta_ac and data_ac when ocfs2_start_trans fails in ocfs2_xattr_set()

[Jeff Liu]

On 08/07/2013 10:03 AM, Younger Liu wrote:

> On 2013/8/7 6:44, Jeff Liu wrote:
>> On 08/01/2013 05:02 PM, Younger Liu wrote:
>>
>>> In ocfs2_xattr_set(), if ocfs2_start_trans failed, 
>>> meta_ac and data_ac should be free. 
>>> Otherwise, It would lead to  memory leak.
>>>
>>> Signed-off-by: Younger Liu <younger.liu at huawei.com>
>>> Cc: Joseph Qi <joseph.qi at huawei.com>
>>> ---
>>>  fs/ocfs2/xattr.c |    5 +++--
>>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
>>> index 317ef0a..1c2a892 100644
>>> --- a/fs/ocfs2/xattr.c
>>> +++ b/fs/ocfs2/xattr.c
>>> @@ -3505,7 +3505,7 @@ int ocfs2_xattr_set(struct inode *inode,
>>>  	int ret, credits, ref_meta = 0, ref_credits = 0;
>>>  	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
>>>  	struct inode *tl_inode = osb->osb_tl_inode;
>>> -	struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
>>> +	struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, NULL, };
>>>  	struct ocfs2_refcount_tree *ref_tree = NULL;
>>>  
>>>  	struct ocfs2_xattr_info xi = {
>>> @@ -3609,13 +3609,14 @@ int ocfs2_xattr_set(struct inode *inode,
>>>  	if (IS_ERR(ctxt.handle)) {
>>>  		ret = PTR_ERR(ctxt.handle);
>>>  		mlog_errno(ret);
>>> -		goto cleanup;
>>> +		goto out_free_ac;
>>>  	}
>>>  
>>>  	ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
>>>  
>>>  	ocfs2_commit_trans(osb, ctxt.handle);
>>>  
>>> +out_free_ac:
>>>  	if (ctxt.data_ac)
>>>  		ocfs2_free_alloc_context(ctxt.data_ac);
>>>  	if (ctxt.meta_ac)
>>
>> Yes, we should free the allocated contex in this case.
>> However, we don't performs the log truncation as well as the delayed allocation

s/delayed allocation/delalloc/. Sorry, I habitually typed above wrong.

>> business as per the old logic.
>>
>> Srinivas,
>>
>> How do you think about this fix?
>>
>> Thanks,
>> -Jeff
>>
> In ocfs2_init_xattr_set_ctxt(), the allocated contex is initialized, 
> (&ctxt.dealloc)->c_global_allocator and (&ctxt.dealloc)->
> c_first_suballocator are set to NULL.
> 
> After applying the patch, if ocfs2_start_trans fails, we would free 
> meta_ac/data_ac, but not perform truncate_log_flush/ocfs2_run_deallocs.
> So, it would take no side effect to the function.

So this fix looks good, thanks.

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