On Wed, Jan 03, 2018 at 02:08:05AM +0000, Al Viro wrote:
> On Wed, Dec 27, 2017 at 04:19:58PM +0800, Sean Fu wrote:
> > generic_file_read_iter has done the count test.
> > So ext4_file_read_iter don't need to test the count repeatedly.
>
> Huh? You do realize that generic_file_read_iter() is not the
> only variant possible there, right?
>
Yes, I know that generic_file_read_iter() is not the only variant possible.
The other possible dax_iomap_rw() with zero count would return zero.
> static ssize_t ext4_dax_read_iter(struct kiocb *iocb, struct iov_iter *to)
> {
> struct inode *inode = file_inode(iocb->ki_filp);
> ssize_t ret;
>
> if (!inode_trylock_shared(inode)) {
> if (iocb->ki_flags & IOCB_NOWAIT)
> return -EAGAIN;
> inode_lock_shared(inode);
> }
>
> ... and now IOCB_NOWAIT read with zero count can fail with -EAGAIN.
>
Correct, IOCB_NOWAIT read with zero count can return -EAGAIN, But I
think that it is reasonable. while it got lock, zero would be returned
in this case.
