Patchwork QEMUFileBuffered: indicate that we're ready when the underlying file is ready

login
register
mail settings
Submitter Avi Kivity
Date July 7, 2010, 4:44 p.m.
Message ID <1278521062-13795-1-git-send-email-avi@redhat.com>
Download mbox | patch
Permalink /patch/58135/
State New
Headers show

Comments

Avi Kivity - July 7, 2010, 4:44 p.m.
QEMUFileBuffered stops writing when the underlying QEMUFile is not ready,
and tells its producer so.  However, when the underlying QEMUFile becomes
ready, it neglects to pass that information along, resulting in stoppage
of all data until the next tick (a tenths of a second).

Usually this doesn't matter, because most QEMUFiles used with QEMUFileBuffered
are almost always ready, but in the case of exec: migration this is not true,
due to the small pipe buffers used to connect to the target process.  The
result is very slow migration.

Fix by detecting the readiness notification and propagating it.  The detection
is a little ugly since QEMUFile overloads put_buffer() to send it, but that's
the suject for a different patch.

Signed-off-by: Avi Kivity <avi@redhat.com>
---
 buffered_file.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
Avi Kivity - Aug. 3, 2010, 6:12 a.m.
On 07/07/2010 07:44 PM, Avi Kivity wrote:
> QEMUFileBuffered stops writing when the underlying QEMUFile is not ready,
> and tells its producer so.  However, when the underlying QEMUFile becomes
> ready, it neglects to pass that information along, resulting in stoppage
> of all data until the next tick (a tenths of a second).
>
> Usually this doesn't matter, because most QEMUFiles used with QEMUFileBuffered
> are almost always ready, but in the case of exec: migration this is not true,
> due to the small pipe buffers used to connect to the target process.  The
> result is very slow migration.
>
> Fix by detecting the readiness notification and propagating it.  The detection
> is a little ugly since QEMUFile overloads put_buffer() to send it, but that's
> the suject for a different patch.

Ping.
Luiz Capitulino - Aug. 3, 2010, 1 p.m.
On Tue, 03 Aug 2010 09:12:52 +0300
Avi Kivity <avi@redhat.com> wrote:

>   On 07/07/2010 07:44 PM, Avi Kivity wrote:
> > QEMUFileBuffered stops writing when the underlying QEMUFile is not ready,
> > and tells its producer so.  However, when the underlying QEMUFile becomes
> > ready, it neglects to pass that information along, resulting in stoppage
> > of all data until the next tick (a tenths of a second).
> >
> > Usually this doesn't matter, because most QEMUFiles used with QEMUFileBuffered
> > are almost always ready, but in the case of exec: migration this is not true,
> > due to the small pipe buffers used to connect to the target process.  The
> > result is very slow migration.
> >
> > Fix by detecting the readiness notification and propagating it.  The detection
> > is a little ugly since QEMUFile overloads put_buffer() to send it, but that's
> > the suject for a different patch.
> 
> Ping.

I've queued this fix and another one in a 'for-0.13' branch and was planning
to send a pull request as soon as Anthony called for 0.13 fixes, however
this didn't happen yet and I forgot.
Avi Kivity - Aug. 19, 2010, 1:12 p.m.
On 08/03/2010 04:00 PM, Luiz Capitulino wrote:
> On Tue, 03 Aug 2010 09:12:52 +0300
> Avi Kivity<avi@redhat.com>  wrote:
>
>>    On 07/07/2010 07:44 PM, Avi Kivity wrote:
>>> QEMUFileBuffered stops writing when the underlying QEMUFile is not ready,
>>> and tells its producer so.  However, when the underlying QEMUFile becomes
>>> ready, it neglects to pass that information along, resulting in stoppage
>>> of all data until the next tick (a tenths of a second).
>>>
>>> Usually this doesn't matter, because most QEMUFiles used with QEMUFileBuffered
>>> are almost always ready, but in the case of exec: migration this is not true,
>>> due to the small pipe buffers used to connect to the target process.  The
>>> result is very slow migration.
>>>
>>> Fix by detecting the readiness notification and propagating it.  The detection
>>> is a little ugly since QEMUFile overloads put_buffer() to send it, but that's
>>> the suject for a different patch.
>> Ping.
> I've queued this fix and another one in a 'for-0.13' branch and was planning
> to send a pull request as soon as Anthony called for 0.13 fixes, however
> this didn't happen yet and I forgot.

Ping again.

I don't want to take over Cam's longest-time-to-apply record, so someone 
apply this please.  0.13 and mainline.
Anthony Liguori - Aug. 19, 2010, 3:19 p.m.
On 07/07/2010 11:44 AM, Avi Kivity wrote:
> QEMUFileBuffered stops writing when the underlying QEMUFile is not ready,
> and tells its producer so.  However, when the underlying QEMUFile becomes
> ready, it neglects to pass that information along, resulting in stoppage
> of all data until the next tick (a tenths of a second).
>
> Usually this doesn't matter, because most QEMUFiles used with QEMUFileBuffered
> are almost always ready, but in the case of exec: migration this is not true,
> due to the small pipe buffers used to connect to the target process.  The
> result is very slow migration.
>
> Fix by detecting the readiness notification and propagating it.  The detection
> is a little ugly since QEMUFile overloads put_buffer() to send it, but that's
> the suject for a different patch.
>
> Signed-off-by: Avi Kivity<avi@redhat.com>\
>    

Applied.  Thanks.

Regards,

Anthony Liguori
> ---
>   buffered_file.c |    8 ++++++++
>   1 files changed, 8 insertions(+), 0 deletions(-)
>
> diff --git a/buffered_file.c b/buffered_file.c
> index 54dc6c2..a79264f 100644
> --- a/buffered_file.c
> +++ b/buffered_file.c
> @@ -156,6 +156,14 @@ static int buffered_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, in
>           offset = size;
>       }
>
> +    if (pos == 0&&  size == 0) {
> +        DPRINTF("file is ready\n");
> +        if (s->bytes_xfer<= s->xfer_limit) {
> +            DPRINTF("notifying client\n");
> +            s->put_ready(s->opaque);
> +        }
> +    }
> +
>       return offset;
>   }
>
>
Avi Kivity - Aug. 29, 2010, 9 a.m.
On 08/19/2010 06:19 PM, Anthony Liguori wrote:
> On 07/07/2010 11:44 AM, Avi Kivity wrote:
>> QEMUFileBuffered stops writing when the underlying QEMUFile is not 
>> ready,
>> and tells its producer so.  However, when the underlying QEMUFile 
>> becomes
>> ready, it neglects to pass that information along, resulting in stoppage
>> of all data until the next tick (a tenths of a second).
>>
>> Usually this doesn't matter, because most QEMUFiles used with 
>> QEMUFileBuffered
>> are almost always ready, but in the case of exec: migration this is 
>> not true,
>> due to the small pipe buffers used to connect to the target process.  
>> The
>> result is very slow migration.
>>
>> Fix by detecting the readiness notification and propagating it.  The 
>> detection
>> is a little ugly since QEMUFile overloads put_buffer() to send it, 
>> but that's
>> the suject for a different patch.
>>
>> Signed-off-by: Avi Kivity<avi@redhat.com>\
>
> Applied.  Thanks.

Should be applied to stable-0.13 as well.

Patch

diff --git a/buffered_file.c b/buffered_file.c
index 54dc6c2..a79264f 100644
--- a/buffered_file.c
+++ b/buffered_file.c
@@ -156,6 +156,14 @@  static int buffered_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, in
         offset = size;
     }
 
+    if (pos == 0 && size == 0) {
+        DPRINTF("file is ready\n");
+        if (s->bytes_xfer <= s->xfer_limit) {
+            DPRINTF("notifying client\n");
+            s->put_ready(s->opaque);
+        }
+    }
+
     return offset;
 }