linux-kernel - Re: [PATCH] ext4: direct return when jinode allocate failed

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <EA9B5AA4-DA21-4B1D-9E89-8C9F16EB8986@dilger.ca>
Date:   Wed, 17 Oct 2018 17:28:55 -0600
From:   Andreas Dilger <adilger@...ger.ca>
To:     liu.song11@....com.cn
Cc:     "Theodore Y. Ts'o" <tytso@....edu>, fishland@...yun.com,
        wang.yi59@....com.cn,
        Ext4 Developers List <linux-ext4@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] ext4: direct return when jinode allocate failed


> On Oct 16, 2018, at 8:26 PM, liu.song11@....com.cn wrote:
> 
>> On Tue, Oct 16, 2018 at 10:55:26PM +0800, fishland wrote:
>>> The jinode does not need protected by *i_lock*, we can return
>>> directly if memory allocation fails.
>>> 
>> 
>> I don't see anything wrong with this patch, but at the same time, I'm
>> not sure I see the benefit, either.  Checking for the allocation
>> failure is cheap, and moving it out spinlock doesn't buy much; not to
>> mention that the allocation failure is going to be highly uncommon.
>> 
>> What inspired this commit?
>> 
>> - Ted
> 
> Hi, Ted
> I found this during code reading. All code logic is very smooth, when
> reading here, need to spin_unlock before "return -ENOMEM", which I
> feel this could more close to the allocate. I refer to other processing
> logic for this situation in "inode.c", found that they all return immediately
> after memory application failed. I agree with you that the benefit of this
> patch is almost none, however it can slightly improve the code in rare case.

Looking at the patch there are two effects that it has:
- optimize a very rare case where there is an allocation failure before locking
- return an unnecessary error if "ei->jinode" is allocated before locking

I don't think it is worthwhile to optimize this case, since allocation failures
will have a serious impact on the application, and I'd rather avoid the rare
case where we don't return an unnecessary error than make the error case faster.

Cheers, Andreas

> 
>>> Signed-off-by: Liu Song <liu.song11@....com.cn>
>>> Reviewed-by: Wang Yi <wang.yi59@....com.cn>
>>> ---
>>> fs/ext4/inode.c | 7 +++----
>>> 1 file changed, 3 insertions(+), 4 deletions(-)
>>> 
>>> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
>>> index d767e993591d..67ba6f062de5 100644
>>> --- a/fs/ext4/inode.c
>>> +++ b/fs/ext4/inode.c
>>> @@ -4384,12 +4384,11 @@ int ext4_inode_attach_jinode(struct inode *inode)
>>>         return 0;
>>> 
>>>  	jinode = jbd2_alloc_inode(GFP_KERNEL);
>>> +	if (!jinode)
>>> +	return -ENOMEM;
>>> +
>>>  	spin_lock(&inode->i_lock);
>>>  	if (!ei->jinode) {
>>> -		if (!jinode) {
>>> -			spin_unlock(&inode->i_lock);
>>> -			return -ENOMEM;
>>> -		}
>>>  		ei->jinode = jinode;
>>>  		jbd2_journal_init_jbd_inode(ei->jinode, inode);
>>>  		jinode = NULL;
>>> --
>>> 2.17.1


Cheers, Andreas






Download attachment "signature.asc" of type "application/pgp-signature" (874 bytes)