Re: [PATCH] ceph: avoid call to strlen() in ceph_mds_auth_match()

[Xiubo Li <xiubli@redhat.com>]

On 6/19/24 20:37, Luis Henriques wrote:
> On Tue 18 Jun 2024 05:36:40 PM +03, Dmitry Antipov wrote;
>
>> Since 'snprintf()' returns the number of characters emitted,
>> an extra call to 'strlen()' in 'ceph_mds_auth_match()' may
>> be dropped. Compile tested only.
>>
>> Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
>> ---
>>   fs/ceph/mds_client.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
>> index c2157f6e0c69..7224283046a7 100644
>> --- a/fs/ceph/mds_client.c
>> +++ b/fs/ceph/mds_client.c
>> @@ -5665,9 +5665,9 @@ static int ceph_mds_auth_match(struct ceph_mds_client *mdsc,
>>   				if (!_tpath)
>>   					return -ENOMEM;
>>   				/* remove the leading '/' */
>> -				snprintf(_tpath, n, "%s/%s", spath + 1, tpath);
>> +				tlen = snprintf(_tpath, n, "%s/%s",
>> +						spath + 1, tpath);
>>   				free_tpath = true;
>> -				tlen = strlen(_tpath);
>>   			}
> Unless I'm missing something, this patch is incorrect.  snprintf() may not
> return the actual string length *if* the output is truncated.  For
> example:
>
> 	snprintf(str, 5, "%s", "0123456789");
>
> snprintf() will return 10, while strlen(str) will return 4.

Yeah, you are correct.

Thanks Luis.


> Cheers,