Re: PATCH? trace_remove_event_call() should fail if call is active

[Masami Hiramatsu]

(2013/07/04 6:02), Steven Rostedt wrote:
> On Wed, 2013-07-03 at 21:17 +0200, Oleg Nesterov wrote:
> 
>> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
>> index 4888cb3..c23d41e 100644
>> --- a/fs/debugfs/inode.c
>> +++ b/fs/debugfs/inode.c
>> @@ -475,6 +475,7 @@ static int __debugfs_remove(struct dentry *dentry, struct dentry *parent)
>>  				kfree(dentry->d_inode->i_private);
>>  				/* fall through */
>>  			default:
>> +				dentry->d_inode->i_private = NULL;
>>  				simple_unlink(parent->d_inode, dentry);
>>  				break;
>>  			}
> 
> Ah, I see what you are saying. If the file is being opened just as it is
> being deleted, it can up the dentry refcount and prevent it from
> actually being deleted at that point. :-p

Yeah, that's actually what I'd like to point out.,,

> 
> There's a slight race that we can get to the open call when the dentry
> was deleted. But can it? Seems that there would be other places that
> have issues as I would think it would be common to do something like:
> 
> 	data = kmalloc(size, GFP_KERNEL);
> 	dentry = debugfs_create_file("file", 0644, parent, data, &ops);
> 
> [...]
> 
> 	debugfs_remove(dentry);
> 	kfree(data);
> 
> Any thing like this would have issues if the open referenced the data.

I'm not so sure about vfs layer, but yes, I think that pattern may be
always unsafe. Perhaps, we need to wait the entry is surely removed
before freeing the data.

Thanks,

-- 
Masami HIRAMATSU
IT Management Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@xxxxxxxxxxx


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/