Re: [PATCH net-next v5] net: ioctl: Use kernel memory on protocol ioctl callbacks

[Willem de Bruijn]

> On Sat, Jun 03, 2023 at 10:21:50AM +0200, Willem de Bruijn wrote:
> > On Fri, Jun 2, 2023 at 6:31 PM Breno Leitao <leitao@xxxxxxxxxx> wrote:
> > > Signed-off-by: Breno Leitao <leitao@xxxxxxxxxx>
> >
> > Please check the checkpatch output
> >
> > https://patchwork.hopto.org/static/nipa/753609/13265673/checkpatch/stdout
>
> I am checking my current checkpatch before sending the patch, but I am
> not seeing the problems above.
>
> My tree is at 44c026a73be8038 ("Linux 6.4-rc3"), and I am not able to
> reproduce the problems above.
>
>         $ scripts/checkpatch.pl v5/v5-0001-net-ioctl-Use-kernel-memory-on-protocol-ioctl-cal.patch
>         total: 0 errors, 0 warnings, 0 checks, 806 lines checked
>         v5/v5-0001-net-ioctl-Use-kernel-memory-on-protocol-ioctl-cal.patch has no obvious style problems and is ready for submission.
>
> Let me investigate what options I am missing when running checkpatch.

The reference is to the checkpatch as referenced by patchwork:

https://patchwork.kernel.org/project/netdevbpf/patch/20230602163044.1820619-1-leitao@xxxxxxxxxx/

The 80 character limit is a soft limit. But also note the CHECK
statements on whitespace.

>
> > > +/* A wrapper around sock ioctls, which copies the data from userspace
> > > + * (depending on the protocol/ioctl), and copies back the result to userspace.
> > > + * The main motivation for this function is to pass kernel memory to the
> > > + * protocol ioctl callbacks, instead of userspace memory.
> > > + */
> > > +int sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
> > > +{
> > > +       int rc = 1;
> > > +
> > > +       if (sk_is_ipmr(sk))
> > > +               rc = ipmr_sk_ioctl(sk, cmd, arg);
> > > +       else if (sk_is_icmpv6(sk))
> > > +               rc = ip6mr_sk_ioctl(sk, cmd, arg);
> > > +       else if (sk_is_phonet(sk))
> > > +               rc = phonet_sk_ioctl(sk, cmd, arg);
> >
> > Does this handle all phonet ioctl cases correctly?
> >
> > Notably pn_socket_ioctl has a SIOCPNGETOBJECT that reads and writes a u16.
>
> We are not touching  "struct proto_ops" in this patch at all.  And
> pn_socket_ioctl() is part of "struct proto_ops".
>
>         const struct proto_ops phonet_stream_ops = {
>                   ...
>                   .ioctl          = pn_socket_ioctl,
>         }
>
> That said, all the "struct proto_ops" ioctl calls backs continue to use
> "unsigned long arg" with userspace information, at least for now.

Ok. Perhaps good to call out in the commit message that this does not
convert all protocol ioctl callbacks.