Re: [RFC PATCH 07/13] fs/userfaultfd: support read_iter to use io_uring

[Nadav Amit]

> On Nov 30, 2020, at 10:20 AM, Jens Axboe <axboe@xxxxxxxxx> wrote:
> 
> On 11/28/20 5:45 PM, Nadav Amit wrote:
>> From: Nadav Amit <namit@xxxxxxxxxx>
>> 
>> iouring with userfaultfd cannot currently be used fixed buffers since
>> userfaultfd does not provide read_iter(). This is required to allow
>> asynchronous (queued) reads from userfaultfd.
>> 
>> To support async-reads of userfaultfd provide read_iter() instead of
>> read().
>> 
>> Cc: Jens Axboe <axboe@xxxxxxxxx>
>> Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
>> Cc: Peter Xu <peterx@xxxxxxxxxx>
>> Cc: Alexander Viro <viro@xxxxxxxxxxxxxxxxxx>
>> Cc: io-uring@xxxxxxxxxxxxxxx
>> Cc: linux-fsdevel@xxxxxxxxxxxxxxx
>> Cc: linux-kernel@xxxxxxxxxxxxxxx
>> Cc: linux-mm@xxxxxxxxx
>> Signed-off-by: Nadav Amit <namit@xxxxxxxxxx>
>> ---
>> fs/userfaultfd.c | 18 ++++++++++--------
>> 1 file changed, 10 insertions(+), 8 deletions(-)
>> 
>> diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
>> index b6a04e526025..6333b4632742 100644
>> --- a/fs/userfaultfd.c
>> +++ b/fs/userfaultfd.c
>> @@ -1195,9 +1195,9 @@ static ssize_t userfaultfd_ctx_read(struct userfaultfd_ctx *ctx, int no_wait,
>> 	return ret;
>> }
>> 
>> -static ssize_t userfaultfd_read(struct file *file, char __user *buf,
>> -				size_t count, loff_t *ppos)
>> +static ssize_t userfaultfd_read_iter(struct kiocb *iocb, struct iov_iter *to)
>> {
>> +	struct file *file = iocb->ki_filp;
>> 	struct userfaultfd_ctx *ctx = file->private_data;
>> 	ssize_t _ret, ret = 0;
>> 	struct uffd_msg msg;
>> @@ -1207,16 +1207,18 @@ static ssize_t userfaultfd_read(struct file *file, char __user *buf,
>> 		return -EINVAL;
>> 
>> 	for (;;) {
>> -		if (count < sizeof(msg))
>> +		if (iov_iter_count(to) < sizeof(msg))
>> 			return ret ? ret : -EINVAL;
>> 		_ret = userfaultfd_ctx_read(ctx, no_wait, &msg);
> 
> 'no_wait' should be changed to factor in iocb->ki_flags & IOCB_NOWAIT as well,
> not just f_flags & O_NONBLOCK.
> 
> I didn't check your write_iter, but if appropriate, that should do that
> too.

Thanks, I completely missed this point and will fix it in v1 (if I get
a positive feedback on the rest from the userfaultfd people).