Re: [PATCH] fuse: Put leaked request on error path of fuse_retrieve()

[Kirill Tkhai]

On 06.11.2018 12:23, Miklos Szeredi wrote:
> On Tue, Nov 6, 2018 at 10:13 AM, Kirill Tkhai <ktkhai@xxxxxxxxxxxxx> wrote:
>> fuse_request_send_notify_reply() may fail, and this case
>> it remains leaked (fuse_retrieve_end(), which is called
>> on error path, does not do that). Also, fc->num_waiting,
>> will never be decremented, and fuse_wait_aborted() will
>> never finish. So, put the request patently.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai@xxxxxxxxxxxxx>
> 
> Posted same patch yesterday for a syzbot report.   How did you notice this?

I've found this by code review. I did this last week and I have 10 patches more
on different theme. I was waiting for when the merge window opens.
 
> 
>> ---
>>  fs/fuse/dev.c |    4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
>> index ae813e609932..6fe330cc9709 100644
>> --- a/fs/fuse/dev.c
>> +++ b/fs/fuse/dev.c
>> @@ -1768,8 +1768,10 @@ static int fuse_retrieve(struct fuse_conn *fc, struct inode *inode,
>>         req->in.args[1].size = total_len;
>>
>>         err = fuse_request_send_notify_reply(fc, req, outarg->notify_unique);
>> -       if (err)
>> +       if (err) {
>>                 fuse_retrieve_end(fc, req);
>> +               fuse_put_request(fc, req);
>> +       }
>>
>>         return err;
>>  }
>>