Re: [PATCH 15/24] compat_ioctl: scsi: move ioctl handling into drivers

[Arnd Bergmann]

On Thu, Dec 12, 2019 at 1:28 AM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> On 12/12/19 00:05, Michael S. Tsirkin wrote:
> >> @@ -405,6 +405,9 @@ static int virtblk_getgeo(struct block_device *bd, struct hd_geometry *geo)
> >>
> >>  static const struct block_device_operations virtblk_fops = {
> >>      .ioctl  = virtblk_ioctl,
> >> +#ifdef CONFIG_COMPAT
> >> +    .compat_ioctl = blkdev_compat_ptr_ioctl,
> >> +#endif
> >>      .owner  = THIS_MODULE,
> >>      .getgeo = virtblk_getgeo,
> >>  };
> > Hmm - is virtio blk lumped in with scsi things intentionally?
>
> I think it's because the only ioctl for virtio-blk is SG_IO.  It makes
> sense to lump it in with scsi, but I wouldn't mind getting rid of
> CONFIG_VIRTIO_BLK_SCSI altogether.

It currently calls scsi_cmd_blk_ioctl(), which implements a bunch of ioctl
commands, including some that are unrelated to SG_IO:

                case SG_GET_VERSION_NUM:
                case SCSI_IOCTL_GET_IDLUN:
                case SCSI_IOCTL_GET_BUS_NUMBER:
                case SG_SET_TIMEOUT:
                case SG_GET_TIMEOUT:
                case SG_GET_RESERVED_SIZE:
                case SG_SET_RESERVED_SIZE:
                case SG_EMULATED_HOST:
                case SG_IO: {
                case CDROM_SEND_PACKET:
                case SCSI_IOCTL_SEND_COMMAND:
                case CDROMCLOSETRAY:
                case CDROMEJECT:

My patch changes all callers of this function, and the idea is
to preserve the existing behavior through my series, so I think
it makes sense to keep my patch as is.

I would assume that calling scsi_cmd_blk_ioctl() is harmless
here, but if you want to remove it or limit the set of supported
commands, that should be independent of my change.

       Arnd