Re: [PATCH 4/4] serial: Allow unaligned i/o access

[Arwed Meyer]

Am 08.09.22 um 13:15 schrieb Michael S. Tsirkin:
> On Thu, Sep 08, 2022 at 02:11:28PM +0400, Marc-André Lureau wrote:
> > Hi
> >
> > On Wed, Sep 7, 2022 at 2:03 AM Arwed Meyer <arwed.meyer@gmx.de> wrote:
> >
> >      Unaligned i/o access on serial UART works on real PCs.
> >      This is used for example by FreeDOS CTMouse driver. Without this it
> >      can't reset and detect serial mice.
> >
> >      Resolves: https://gitlab.com/qemu-project/qemu/-/issues/77
> >      Signed-off-by: Arwed Meyer <arwed.meyer@gmx.de>
> >      ---
> >       hw/char/serial.c | 3 +++
> >       1 file changed, 3 insertions(+)
> >
> >      diff --git a/hw/char/serial.c b/hw/char/serial.c
> >      index 7061aacbce..41b5e61977 100644
> >      --- a/hw/char/serial.c
> >      +++ b/hw/char/serial.c
> >      @@ -961,6 +961,9 @@ void serial_set_frequency(SerialState *s, uint32_t
> >      frequency)
> >       const MemoryRegionOps serial_io_ops = {
> >           .read = serial_ioport_read,
> >           .write = serial_ioport_write,
> >      +    .valid = {
> >      +        .unaligned = 1,
> >      +    },
> >
> >
> > I don't get how this can help if both min_access_size & max_access_size are 1.
> >
> >
> >           .impl = {
> >               .min_access_size = 1,
> >               .max_access_size = 1,
> >      --
> >      2.34.1
>
>
> Because that's .impl. If access is invalid we don't get as far
> as breaking it up to chunks.
>
> > --
> > Marc-André Lureau
Exactly. Not really knowing the serial/chardev code much it took me a
while to figure out why calling FreeDOS CTMouse/protocol.com would never
execute the ioctl mouse reset code in msmouse.c.