RE: [PATCH v3 5/5] RFC: fw_cfg: add DMA write operation in sysfs

[Hatayama, Daisuke]

> -----Original Message-----
> From: linux-kernel-owner@xxxxxxxxxxxxxxx
> [mailto:linux-kernel-owner@xxxxxxxxxxxxxxx] On Behalf Of Michael S. Tsirkin
> Sent: Tuesday, October 31, 2017 1:19 AM
> To: Hatayama, Daisuke <d.hatayama@xxxxxxxxxxxxxx>
> Cc: 'marcandre.lureau@xxxxxxxxxx' <marcandre.lureau@xxxxxxxxxx>;
> 'somlo@xxxxxxx' <somlo@xxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx;
> 'qemu-devel@xxxxxxxxxx' <qemu-devel@xxxxxxxxxx>
> Subject: Re: [PATCH v3 5/5] RFC: fw_cfg: add DMA write operation in sysfs
> 
> On Mon, Oct 30, 2017 at 11:07:20AM +0000, Hatayama, Daisuke wrote:
> > > From: Marc-André Lureau <marcandre.lureau@xxxxxxxxxx>
> > >
> > > Since qemu 2.9, DMA write operations are allowed. However, usage of this
> > > interface from kernel or user-space is strongly discouraged by the
> > > maintainers. This patch is meant for experimentations for now.
> > >
> >
> > Could you (or maintainers?) tell me how experimental the DMA write
> > operations from kernel are?
> >
> > From some technical reason?
> > Or simply there has not been enough test yet so far?
> 
> The concern is security, talking from userspace to hypervisor
> might become a trivial DOS vector.
> 
> If there's need for a specific entry to be accessible from userspace,
> I'd rather white-list it.
> 

Thanks for the explanation.

If so, I guess the white-list is done in userspace qemu side,
and no additional change would be needed for this kernel-side patch set for the white-list feature.
Is this understanding correct?

> > > Signed-off-by: Marc-André Lureau <marcandre.lureau@xxxxxxxxxx>
> > > ---
> > >  drivers/firmware/qemu_fw_cfg.c | 21 ++++++++++++++++++++-
> > >  1 file changed, 20 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/firmware/qemu_fw_cfg.c
> b/drivers/firmware/qemu_fw_cfg.c
> > > index 54b569da3257..e2f2ad1c9c0c 100644
> > > --- a/drivers/firmware/qemu_fw_cfg.c
> > > +++ b/drivers/firmware/qemu_fw_cfg.c
> > > @@ -524,9 +524,28 @@ static ssize_t fw_cfg_sysfs_read_raw(struct file *filp,
> struct kobject *kobj,
> > >       return fw_cfg_read_blob(entry->f.select, buf, pos, count, true);
> > >  }
> > >
> > > +static ssize_t fw_cfg_sysfs_write_raw(struct file *filp, struct kobject
> *kobj,
> > > +                                   struct bin_attribute *bin_attr,
> > > +                                   char *buf, loff_t pos, size_t count)
> > > +{
> > > +     struct fw_cfg_sysfs_entry *entry = to_entry(kobj);
> > > +
> > > +     if (!fw_cfg_dma_enabled())
> > > +             return -ENOTSUPP;
> > > +
> > > +     if (pos > entry->f.size)
> > > +             return -EINVAL;
> > > +
> > > +     if (count > entry->f.size - pos)
> > > +             count = entry->f.size - pos;
> > > +
> > > +     return fw_cfg_write_blob(entry->f.select, buf, pos, count);
> > > +}
> > > +
> > >  static struct bin_attribute fw_cfg_sysfs_attr_raw = {
> > > -     .attr = { .name = "raw", .mode = S_IRUSR },
> > > +     .attr = { .name = "raw", .mode = S_IRUSR | S_IWUSR },
> > >       .read = fw_cfg_sysfs_read_raw,
> > > +     .write = fw_cfg_sysfs_write_raw,
> > >  };
> > >
> > >  /*
> > > --
> > > 2.14.1.146.gd35faa819
> >
> > Thanks.
> > HATAYAMA, Daisuke
> >
>