Re: [PATCH v3 1/5] fs,io_uring: add infrastructure for uring-cmd

[Kanchan Joshi <joshiiitr@gmail.com>]

> > @@ -27,7 +27,6 @@ struct io_uring_sqe {
> >         union {
> >                 __u64   addr;   /* pointer to buffer or iovecs */
> >                 __u64   splice_off_in;
> > -               __u16   cmd_len;
> >         };
> >         __u32   len;            /* buffer size or number of iovecs */
> >         union {
> > @@ -64,16 +63,19 @@ struct io_uring_sqe {
> >                 __u32   file_index;
> >         };
> >         union {
> > -               __u64   addr3;
> > -               __u64   cmd;
> > +               struct {
> > +                       __u64   addr3;
> > +                       __u64   __pad2[1];
> > +               } small;
>
> Thinking if this can cause any issue for existing users of addr3, i.e.
> in the userspace side? Since it needs to access this field with
> small.addr3.
> Jens - is xattr infra already frozen?

If this breaks anything, we can avoid that by not touching addr3 at all.
And instead keep "__u64   __pad2[1]" and  "__u8    cmd[0]" in the union.
That means 72 bytes of space, but it is enough for the new nvme struct
(nvme_uring_cmd) which is also 72 bytes.
Does this sound fine?