On Thu, Nov 23, 2023 at 9:47 AM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
>
> > +ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from,
> > + loff_t *pos);
> > +ssize_t do_iter_writev(struct file *file, struct iov_iter *iter, loff_t *ppos,
> > + rwf_t flags);
>
> So I first stumbled on the name because I kinda hate do_* functions,
> especially if they are not static, and then went down a little rathole:
>
>
> - first obviously the name, based on the other functions it probably
> should be in the __kernel_* namespace unless I'm missing something.
Not sure about the best name.
I just derived the name from do_iter_readv_writev(), which would be the
name of this helper if we split up do_iter_readv_writev() as you suggested.
> - second can you add a little comment when it is to be used? Our
> maze of read and write helpers is becoming a little too confusing
> even for someone who thought he knew the code (and wrote some if it).
I could try, but I can't say that I know how to document the maze.
I was trying to make small changes that make the maze a bit
more consistent - after my series, in most cases we have:
vfs_XXX()
rw_verify_area()
file_start_write()
do_XXX()
But obviously, there is still a wide variety of do_* helpers
with different conventions.
> - can we just split do_iter_readv_writev instead of adding a wrapper
> Yes, that'll duplicate a little boiler plate code, but it'll make
> things much easier to follow.
> (- same probably for do_loop_readv_writev, although not directly
> relevant to this series)
>
We can do that, it's easy, but I also got opposite comments from
Christian about opportunities to factor out more common code, so
I will do it if there is consensus.
Thanks,
Amir.
