Re: [PATCH 2/4] io_uring: move CLOSE req->file checking into handler

[Jens Axboe <axboe@kernel.dk>]

On 3/4/20 6:07 AM, Pavel Begunkov wrote:
> On 04/03/2020 02:50, Jens Axboe wrote:
>> In preparation for not needing req->file in on the prep side at all.
>>
>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>> ---
>>  fs/io_uring.c | 8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index 0464efbeba25..9d5e49a39dba 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -3367,10 +3367,6 @@ static int io_close_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>>  		return -EBADF;
>>  
>>  	req->close.fd = READ_ONCE(sqe->fd);
>> -	if (req->file->f_op == &io_uring_fops ||
>> -	    req->close.fd == req->ctx->ring_fd)
>> -		return -EBADF;
>> -
>>  	return 0;
>>  }
>>  
>> @@ -3400,6 +3396,10 @@ static int io_close(struct io_kiocb *req, bool force_nonblock)
>>  {
>>  	int ret;
>>  
>> +	if (req->file->f_op == &io_uring_fops ||
>> +	    req->close.fd == req->ctx->ring_fd)
>> +		return -EBADF;
>> +
> 
> @ring_fd's and @ring_file's lifetimes are bound by call to io_submit_sqes(), and
> they're undefined outside. For the same reason both of them should be used with
> ctx->uring_lock hold.

I've fixed this by splitting the check.


-- 
Jens Axboe