diff mbox

serial: poll the serial console with G_IO_HUP

Message ID 1398186535-66539-1-git-send-email-roger.pau@citrix.com
State New
Headers show

Commit Message

Roger Pau Monné April 22, 2014, 5:08 p.m. UTC
On FreeBSD polling a master pty while the other end is not connected
with G_IO_OUT only results in an endless wait. This is different from
the Linux behaviour, that returns immediately. In order to demonstrate
this, I have the following example code:

http://xenbits.xen.org/people/royger/test_poll.c

When executed on Linux:

$ ./test_poll
In callback

On FreeBSD instead, the callback never gets called:

$ ./test_poll

So, in order to workaround this, poll the source with G_IO_HUP (which
makes the code behave the same way on both Linux and FreeBSD).

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Cc: Michael Tokarev <mjt@tls.msk.ru>
Cc: "Andreas Färber" <afaerber@suse.de>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: xen-devel@lists.xenproject.org
---
 hw/char/serial.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

Comments

Paolo Bonzini April 28, 2014, 9:19 a.m. UTC | #1
Il 22/04/2014 19:08, Roger Pau Monne ha scritto:
> On FreeBSD polling a master pty while the other end is not connected
> with G_IO_OUT only results in an endless wait. This is different from
> the Linux behaviour, that returns immediately. In order to demonstrate
> this, I have the following example code:
> 
> http://xenbits.xen.org/people/royger/test_poll.c
> 
> When executed on Linux:
> 
> $ ./test_poll
> In callback
> 
> On FreeBSD instead, the callback never gets called:
> 
> $ ./test_poll
> 
> So, in order to workaround this, poll the source with G_IO_HUP (which
> makes the code behave the same way on both Linux and FreeBSD).
> 
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> Cc: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
> Cc: Michael Tokarev <mjt@tls.msk.ru>
> Cc: "Andreas Färber" <afaerber@suse.de>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: xen-devel@lists.xenproject.org
> ---
>  hw/char/serial.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/hw/char/serial.c b/hw/char/serial.c
> index 6025592..ab9c40f 100644
> --- a/hw/char/serial.c
> +++ b/hw/char/serial.c
> @@ -243,7 +243,7 @@ static gboolean serial_xmit(GIOChannel *chan, GIOCondition cond, void *opaque)
>          serial_receive1(s, &s->tsr, 1);
>      } else if (qemu_chr_fe_write(s->chr, &s->tsr, 1) != 1) {
>          if (s->tsr_retry >= 0 && s->tsr_retry < MAX_XMIT_RETRY &&
> -            qemu_chr_fe_add_watch(s->chr, G_IO_OUT, serial_xmit, s) > 0) {
> +            qemu_chr_fe_add_watch(s->chr, G_IO_OUT|G_IO_HUP, serial_xmit, s) > 0) {
>              s->tsr_retry++;
>              return FALSE;
>          }
> 

The patch looks good.  However, the same problem is most likely there
in other places:

hw/char/cadence_uart.c:        int r = qemu_chr_fe_add_watch(s->chr, G_IO_OUT, cadence_uart_xmit, s);
hw/char/serial.c:            qemu_chr_fe_add_watch(s->chr, G_IO_OUT, serial_xmit, s) > 0) {
hw/char/virtio-console.c:                vcon->watch = qemu_chr_fe_add_watch(vcon->chr, G_IO_OUT,
hw/usb/redirect.c:            dev->watch = qemu_chr_fe_add_watch(dev->cs, G_IO_OUT,
monitor.c:            mon->watch = qemu_chr_fe_add_watch(mon->chr, G_IO_OUT,

Should we add G_IO_HUP everywhere we have G_IO_OUT?

Or perhaps even in qemu_chr_fe_add_watch (I don't really like this,
but I will shoot it out...)?

Paolo
Roger Pau Monné April 28, 2014, 9:32 a.m. UTC | #2
On 28/04/14 11:19, Paolo Bonzini wrote:
> Il 22/04/2014 19:08, Roger Pau Monne ha scritto:
>> On FreeBSD polling a master pty while the other end is not connected
>> with G_IO_OUT only results in an endless wait. This is different from
>> the Linux behaviour, that returns immediately. In order to demonstrate
>> this, I have the following example code:
>>
>> http://xenbits.xen.org/people/royger/test_poll.c
>>
>> When executed on Linux:
>>
>> $ ./test_poll
>> In callback
>>
>> On FreeBSD instead, the callback never gets called:
>>
>> $ ./test_poll
>>
>> So, in order to workaround this, poll the source with G_IO_HUP (which
>> makes the code behave the same way on both Linux and FreeBSD).
>>
>> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
>> Cc: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
>> Cc: Michael Tokarev <mjt@tls.msk.ru>
>> Cc: "Andreas Färber" <afaerber@suse.de>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: xen-devel@lists.xenproject.org
>> ---
>>  hw/char/serial.c |    2 +-
>>  1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/hw/char/serial.c b/hw/char/serial.c
>> index 6025592..ab9c40f 100644
>> --- a/hw/char/serial.c
>> +++ b/hw/char/serial.c
>> @@ -243,7 +243,7 @@ static gboolean serial_xmit(GIOChannel *chan, GIOCondition cond, void *opaque)
>>          serial_receive1(s, &s->tsr, 1);
>>      } else if (qemu_chr_fe_write(s->chr, &s->tsr, 1) != 1) {
>>          if (s->tsr_retry >= 0 && s->tsr_retry < MAX_XMIT_RETRY &&
>> -            qemu_chr_fe_add_watch(s->chr, G_IO_OUT, serial_xmit, s) > 0) {
>> +            qemu_chr_fe_add_watch(s->chr, G_IO_OUT|G_IO_HUP, serial_xmit, s) > 0) {
>>              s->tsr_retry++;
>>              return FALSE;
>>          }
>>
> 
> The patch looks good.  However, the same problem is most likely there
> in other places:
> 
> hw/char/cadence_uart.c:        int r = qemu_chr_fe_add_watch(s->chr, G_IO_OUT, cadence_uart_xmit, s);
> hw/char/serial.c:            qemu_chr_fe_add_watch(s->chr, G_IO_OUT, serial_xmit, s) > 0) {

This is the one I've fixed.

> hw/char/virtio-console.c:                vcon->watch = qemu_chr_fe_add_watch(vcon->chr, G_IO_OUT,
> hw/usb/redirect.c:            dev->watch = qemu_chr_fe_add_watch(dev->cs, G_IO_OUT,
> monitor.c:            mon->watch = qemu_chr_fe_add_watch(mon->chr, G_IO_OUT,
> 
> Should we add G_IO_HUP everywhere we have G_IO_OUT?

IMOH it seems better to add G_IO_HUP to those other callers rather than
appending it in qemu_chr_fe_add_watch unconditionally, but I don't have
a strong opinion. If no one else expresses an interest into having it
added to qemu_chr_fe_add_watch I will send a new version that also fixes
the other callers listed above.

Roger.
diff mbox

Patch

diff --git a/hw/char/serial.c b/hw/char/serial.c
index 6025592..ab9c40f 100644
--- a/hw/char/serial.c
+++ b/hw/char/serial.c
@@ -243,7 +243,7 @@  static gboolean serial_xmit(GIOChannel *chan, GIOCondition cond, void *opaque)
         serial_receive1(s, &s->tsr, 1);
     } else if (qemu_chr_fe_write(s->chr, &s->tsr, 1) != 1) {
         if (s->tsr_retry >= 0 && s->tsr_retry < MAX_XMIT_RETRY &&
-            qemu_chr_fe_add_watch(s->chr, G_IO_OUT, serial_xmit, s) > 0) {
+            qemu_chr_fe_add_watch(s->chr, G_IO_OUT|G_IO_HUP, serial_xmit, s) > 0) {
             s->tsr_retry++;
             return FALSE;
         }