On Wed, Mar 16, 2011 at 05:39:32PM +0100, Jes Sorensen wrote:
> On 03/16/11 16:52, Alon Levy wrote:
> > This moves the int pipe[2] and pthread_t main data from the
> > PCIQXLDevice struct to the SimpleSpiceDisplay. This will let us
> > reuse it in the next patch for both -spice with no -qxl usage and
> > for vga mode from qxl.
> >
> > Also move the pipe creation function (which is effectively completely
> > rewritten
> > by this patch anyways) from hw/qxl.c to ui/spice-display.c, since
> > spice-display will depend on it after the next patch and qemu can be build
> > with ui/spice-display.c in combination with no hw/qxl.c.
> > ---
> > hw/qxl.c | 22 +++++-----------------
> > hw/qxl.h | 4 ----
> > ui/spice-display.c | 21 +++++++++++++++++++++
> > ui/spice-display.h | 8 ++++++++
> > 4 files changed, 34 insertions(+), 21 deletions(-)
> >
> > diff --git a/hw/qxl.c b/hw/qxl.c
> > index fe4212b..201698f 100644
> > --- a/hw/qxl.c
> > +++ b/hw/qxl.c
> > @@ -1062,7 +1062,7 @@ static void pipe_read(void *opaque)
> > int len;
> >
> > do {
> > - len = read(d->pipe[0], &dummy, sizeof(dummy));
> > + len = read(d->ssd.pipe[0], &dummy, sizeof(dummy));
> > } while (len == sizeof(dummy));
> > qxl_set_irq(d);
> > }
> > @@ -1078,10 +1078,11 @@ static void qxl_send_events(PCIQXLDevice *d,
> > uint32_t events)
> > if ((old_pending & le_events) == le_events) {
> > return;
> > }
> > - if (pthread_self() == d->main) {
> > + if (pthread_self() == d->ssd.main) {
> > + /* running in io_thread thread */
> [snip]
> > diff --git a/ui/spice-display.h b/ui/spice-display.h
> > index aef0464..3e6cf7c 100644
> > --- a/ui/spice-display.h
> > +++ b/ui/spice-display.h
> > @@ -43,6 +43,11 @@ typedef struct SimpleSpiceDisplay {
> > QXLRect dirty;
> > int notify;
> > int running;
> > +
> > + /* thread signaling - used both in qxl (in vga mode
> > + * and in native mode) and without qxl */
> > + pthread_t main;
> > + int pipe[2]; /* to iothread */
> > } SimpleSpiceDisplay;
>
> This really should be using QemuThread rather than pthread_t directly. I
> am not quite sure what impact it will have on the use of pthread_self()
> above, there might be a need for an additional wrapper there.
>
Agreed, but it's not introduced by this patch, that code is just being moved
from qxl.h to spice-display.h, from PCIQXLDevice to SimpleSpiceDisplay (which
is embedded in PCIQXLDevice as well as used standalone when there is no qxl
device).
I'll make a separate patch for this.
> Cheers,
> Jes
