[Qemu-devel] [PATCH] No keyboard / mouse after reboot

[Qemu-devel] [PATCH] No keyboard / mouse after reboot

[Dinesh Subhraveti]

[-- Attachment #1: Type: text/plain, Size: 1935 bytes --]

At system initialization, some guests conclude that i8042 controller is missing when the PS2 buffer is full.  As a result, the guest comes up with no keyboard or mouse.  The behavior is seen when the user types or mouses over the VNC window during the BIOS phase of a reboot, causing the PS2 buffer to fill up.  It doesn't occur when the system boots the first time, since the device is initially disabled and doesn't accept events.

This patch tunes Qemu parameters to reasonable values which make the guest happy.  Linux assumes the size of PS2 buffer to be 16 bytes, while Qemu defines it as 256.  Reducing PS2_QUEUE_SIZE to 15 ensures that the guest never sees the buffer to be full.

However, PS2 mouse event handler drops mouse events unless there is at least 16 bytes of buffer space left in the queue (to account for mouse movements that consume multiple events).  With PS2_QUEUE_SIZE reduced to 15, there will never be 16 bytes left in the queue and no mouse event ever gets in.  So the size of reserved buffer space is reduced to 8.

With these changes, both keyboard and mouse remain responsive after reboot, even when typed or moused over during reboot.  Tested with SLES 11 GM 64-bit guest.

---
diff --git a/hw/ps2.c b/hw/ps2.c
index fb77005..51a3ab8 100644
--- a/hw/ps2.c
+++ b/hw/ps2.c
@@ -70,7 +70,7 @@
 #define MOUSE_STATUS_ENABLED    0x20
 #define MOUSE_STATUS_SCALE21    0x10

-#define PS2_QUEUE_SIZE 256
+#define PS2_QUEUE_SIZE 15

 typedef struct {
     uint8_t data[PS2_QUEUE_SIZE];
@@ -346,7 +346,7 @@ static void ps2_mouse_event(void *opaque,
     s->mouse_buttons = buttons_state;

     if (!(s->mouse_status & MOUSE_STATUS_REMOTE) &&
-        (s->common.queue.count < (PS2_QUEUE_SIZE - 16))) {
+        (s->common.queue.count < (PS2_QUEUE_SIZE - 8))) {
         for(;;) {
             /* if not remote, send event. Multiple events are sent if
                too big deltas */

[-- Attachment #2: Type: text/html, Size: 2920 bytes --]

Re: [Qemu-devel] [PATCH] No keyboard / mouse after reboot

[Anthony Liguori]

Needs a Signed-off-by.  However...

Dinesh Subhraveti wrote:
> At system initialization, some guests conclude that i8042 controller 
> is missing when the PS2 buffer is full.  As a result, the guest comes 
> up with no keyboard or mouse.  The behavior is seen when the user 
> types or mouses over the VNC window during the BIOS phase of a reboot, 
> causing the PS2 buffer to fill up.  It doesn't occur when the system 
> boots the first time, since the device is initially disabled and 
> doesn't accept events.
>  
> This patch tunes Qemu parameters to reasonable values which make the 
> guest happy.  Linux assumes the size of PS2 buffer to be 16 bytes, 
> while Qemu defines it as 256.  Reducing PS2_QUEUE_SIZE to 15 ensures 
> that the guest never sees the buffer to be full.

Why is Linux making assumptions about queue size being 16?  It sounds 
broken to me.  In fact, I looked in one of my PC hardware books and it 
claims that the typical KBD buffer size is around 20 bytes so if Linux 
is assuming 16, it's definitely broken.

But there's speculation in the previous thread that we're masking a BIOS 
bug.  Has that been ruled out?

Regards,

Anthony Liguori

[Qemu-devel] Re: [PATCH] No keyboard / mouse after reboot

[Dinesh Subhraveti]

"Anthony Liguori" <anthony@codemonkey.ws> wrote in message 
news:4A5BDB00.8030100@codemonkey.ws...
> Needs a Signed-off-by.  However...
>
> Dinesh Subhraveti wrote:
>> At system initialization, some guests conclude that i8042 controller is 
>> missing when the PS2 buffer is full.  As a result, the guest comes up 
>> with no keyboard or mouse.  The behavior is seen when the user types or 
>> mouses over the VNC window during the BIOS phase of a reboot, causing the 
>> PS2 buffer to fill up.  It doesn't occur when the system boots the first 
>> time, since the device is initially disabled and doesn't accept events.
>>  This patch tunes Qemu parameters to reasonable values which make the 
>> guest happy.  Linux assumes the size of PS2 buffer to be 16 bytes, while 
>> Qemu defines it as 256.  Reducing PS2_QUEUE_SIZE to 15 ensures that the 
>> guest never sees the buffer to be full.
>
> Why is Linux making assumptions about queue size being 16?  It sounds

Good question, for which I don't know the answer.  But regardless of the 
buffer size, it fills up quite quickly with mouse events, and the guest 
would be left without a keyboard.

> broken to me.  In fact, I looked in one of my PC hardware books and it 
> claims that the typical KBD buffer size is around 20 bytes so if Linux is 
> assuming 16, it's definitely broken.
>
> But there's speculation in the previous thread that we're masking a BIOS 
> bug.  Has that been ruled out?
>

Yeah, this doesn't seem to be a BIOS issue.  It is also reproducible 
with -kernel option.

Thanks,
Dinesh