Re: [V9fs-developer] [PATCH] net/9p: fix potential refcnt problem of trans module

[jiangyiwen <jiangyiwen@huawei.com>]

On 2018/3/28 10:52, cgxu519@gmx.com wrote:
> 在 2018年3月28日，上午10:10，jiangyiwen <jiangyiwen@huawei.com> 写道：
>>
>> On 2018/3/27 20:49, Chengguang Xu wrote:
>>> When specifying trans_mod multiple times in a mount,
>>> it may cause inaccurate refcount of trans module. Also,
>>> in the error case of option parsing, we should put the
>>> trans module if we have already got.
>>>
>>> Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
>>> ---
>>> net/9p/client.c | 5 ++++-
>>> 1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/net/9p/client.c b/net/9p/client.c
>>> index b433aff..7ccfb4b 100644
>>> --- a/net/9p/client.c
>>> +++ b/net/9p/client.c
>>> @@ -190,7 +190,9 @@ static int parse_opts(char *opts, struct p9_client *clnt)
>>> 				p9_debug(P9_DEBUG_ERROR,
>>> 					 "problem allocating copy of trans arg\n");
>>> 				goto free_and_return;
>>> -			 }
>>> +			}
>>> +
>>> +			v9fs_put_trans(clnt->trans_mod);
>>
>> I think this should return error if using multiple times
>> in a mount.
> 
> Fail or retake are just different policies how to deal with this kind of situation,
> most filesystems in kernel choose retaking new value, so we’d better keep consistence
> with others unless there is a necessary reason.
> 
> Thanks,
> Chengguang.
> 

Yes, you're right, most filesystems retake indeed the new value.

Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>

> 
>>
>>> 			clnt->trans_mod = v9fs_get_trans_by_name(s);
>>> 			if (clnt->trans_mod == NULL) {
>>> 				pr_info("Could not find request transport: %s\n",
>>> @@ -226,6 +228,7 @@ static int parse_opts(char *opts, struct p9_client *clnt)
>>> 	}
>>>
>>> free_and_return:
>>> +	v9fs_put_trans(clnt->trans_mod);
>>
>> This looks good.
>>
>>> 	kfree(tmp_options);
>>> 	return ret;
>>> }
>>>
>>
>>
> 
> 
> .
>