On Tue, 21 Nov 2017 08:27:51 +0100 Thomas Huth <th...@redhat.com> wrote:
> On 20.11.2017 08:14, David Gibson wrote: > > The spapr-vty device implements the PAPR defined virtual console, > > which is also implemented by IBM's proprietary PowerVM hypervisor. > > > > PowerVM's implementation has a bug where it inserts an extra \0 after > > every \r going to the guest. Because of that Linux's guest side > > driver has a workaround which strips \0 characters that appear > > immediately after a \r. > > Ouch. > > I wonder whether it would make more sense to change the guest side > driver to apply the workaround only if it's really running under > PowerVM...? E.g. what if the PowerVM bug ever gets fixed one day? > The bug has been around forever (already there in initial linux git commit 1da177e4c3f4), so I'm not sure PowerVM will ever fix it. It is legacy API now :) Also I'm not sure it is worth the pain to introduce extra complexity in the existing linux driver... I guess we'd rather encourage people to switch to virtio serial instead. Anyway, Reviewed-by: Greg Kurz <gr...@kaod.org> > > That means that when running under qemu, sending a binary stream from > > host to guest via spapr-vty which happens to include a \r\0 sequence > > will get corrupted by that workaround. > > > > To deal with that, this patch duplicates PowerVM's bug, inserting an > > extra \0 after each \r. Ugly, but the best option available. > > > > Signed-off-by: David Gibson <da...@gibson.dropbear.id.au> > > --- > > hw/char/spapr_vty.c | 18 ++++++++++++++++++ > > 1 file changed, 18 insertions(+) > > > > diff --git a/hw/char/spapr_vty.c b/hw/char/spapr_vty.c > > index 0fa416ca6b..a95e5e91a7 100644 > > --- a/hw/char/spapr_vty.c > > +++ b/hw/char/spapr_vty.c > > @@ -58,6 +58,24 @@ static int vty_getchars(VIOsPAPRDevice *sdev, uint8_t > > *buf, int max) > > > > while ((n < max) && (dev->out != dev->in)) { > > buf[n++] = dev->buf[dev->out++ % VTERM_BUFSIZE]; > > + > > + /* PowerVM's vty implementation has a bug where it inserts a > > + * \0 after every \r going to the guest. Existing guests have > > + * a workaround for this which removes every \0 immediately > > + * following a \r, so here we make ourselves bug-for-bug > > + * compatible, so that the guest won't drop a real \0-after-\r > > + * that happens to occur in a binary stream. */ > > + if (buf[n-1] == '\r') { > > + if (n < max) { > > + buf[n++] = '\0'; > > + } else { > > + /* No room for the extra \0, roll back and try again > > + * next time */ > > + dev->out--; > > + n--; > > + break; > > + } > > + } > > } > > > > qemu_chr_fe_accept_input(&dev->chardev); > > > > Code looks fine to me, so: > > Reviewed-by: Thomas Huth <th...@redhat.com> >