ing its lifetime regardless of demand, and destroyed
upon freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spike.
Signed-off-by: Nicolas Saenz Julienne
---
The fix I propose here works for my specific use-case, but I'm pretty
sure it
Hi Stefan, thanks for the review. I took note of your comments.
On Thu, 2022-02-03 at 10:53 +, Stefan Hajnoczi wrote:
> > Some background on my workload: I'm using IDE emulation, the guest is an
> > old RTOS that doesn't support virtio, using 'aio=native' isn't possible
> > either (unaligned I
On Thu, 2022-02-03 at 14:19 +, Stefan Hajnoczi wrote:
> Yep, I didn't comment on this because I don't have a good suggestion.
>
> In terms of semantics I think we should have:
>
> 1. A global default value that all new AioContext take. The QEMU main
>loop's qemu_aio_context will use this
using the
'-object' keyword[1]. Only one instance of 'MainLoopClass' might be
created at any time.
'EventLoopBackendClass' learns a new callback, 'can_be_deleted()' so as
to mark 'MainLoop' as non-deletable.
Signed-off-by: Nicolas Saenz Julienne
tialization, remain available during its lifetime
regardless of demand, and destroyed upon freeing it. A properly
characterized workload will then be able to configure the pool to avoid
any latency spike.
Signed-off-by: Nicolas Saenz Julienne
---
include/block/aio.h | 11 +++
qap
attach
themselves to the user creatable class 'complete' function.
The new 'event-loop-backend' class will live in the util directory, and
will be packed into the qom static library.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
---
MAINTAINERS
iendly and pass all the tests.
[1]
https://patchwork.ozlabs.org/project/qemu-devel/patch/20220202175234.656711-1-nsaen...@redhat.com/
---
Nicolas Saenz Julienne (3):
util & iothread: Introduce event-loop abstract class
util/main-loop: Introduce the main loop into QOM
util/ev
Hi Stefan, thanks for the review.
On Thu, 2022-02-24 at 09:48 +, Stefan Hajnoczi wrote:
> On Mon, Feb 21, 2022 at 06:08:43PM +0100, Nicolas Saenz Julienne wrote:
> > diff --git a/qom/meson.build b/qom/meson.build
> > index 062a3789d8..c20e5dd1cb 100644
> > --- a/qom/meso
On Thu, 2022-02-24 at 10:01 +, Stefan Hajnoczi wrote:
> On Mon, Feb 21, 2022 at 06:08:44PM +0100, Nicolas Saenz Julienne wrote:
> > diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
> > index 8dbc6fcb89..fea5a3e9d4 100644
> > --- a/include/qemu/main-loop.h
&
On Thu, 2022-02-24 at 10:40 +, Stefan Hajnoczi wrote:
> On Mon, Feb 21, 2022 at 06:08:45PM +0100, Nicolas Saenz Julienne wrote:
> > The thread pool regulates itself: when idle, it kills threads until
> > empty, when in demand, it creates new threads until full. This behaviour
&
iendly and pass all the tests.
[1]
https://patchwork.ozlabs.org/project/qemu-devel/patch/20220202175234.656711-1-nsaen...@redhat.com/
---
Changes since v1:
- Address all Stefan's comments
- Introduce new fix
Nicolas Saenz Julienne (4):
util/thread-pool: Fix thread pool freeing locking
using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
Signed-off-by: Nicolas Saenz Julienne
ively blocking the workers from exiting.
So let's release the thread pool lock after signaling a worker thread
that it's time to exit to give it a chance to do so.
Fixes: f7311ccc63 ("threadpool: add thread_pool_new() and thread_pool_free()")
Signed-off-by: Nicolas Saenz Julienne
ordering in meson.build, among other things the 'hw'
subdir is processed earlier as it introduces files into the 'qom' source
set.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
---
Changes since v1:
- Rename to event-loop-base
- Move event-loop-base int
lization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne
---
Changes s
On Thu, 2022-03-10 at 09:20 +, Stefan Hajnoczi wrote:
> On Thu, Mar 03, 2022 at 03:58:19PM +0100, Nicolas Saenz Julienne wrote:
> > Upon freeing a thread pool we need to get rid of any remaining worker.
> > This is achieved by setting the thread pool's topping flag, wakin
On Thu, 2022-03-10 at 10:25 +, Stefan Hajnoczi wrote:
> On Thu, Mar 03, 2022 at 03:58:20PM +0100, Nicolas Saenz Julienne wrote:
> > @@ -2935,13 +2947,6 @@ qemu_syms = custom_target('qemu.syms', output:
> > 'qemu.syms',
> >
On Thu, 2022-03-10 at 10:45 +, Stefan Hajnoczi wrote:
> On Thu, Mar 03, 2022 at 04:13:07PM +0100, Nicolas Saenz Julienne wrote:
> > @@ -537,10 +546,19 @@
> > # 0 means that the engine will use its default
> > # (default:0, since 6.1)
>
On Mon, 2022-03-14 at 13:33 +, Stefan Hajnoczi wrote:
> On Fri, Mar 11, 2022 at 11:17:22AM +0100, Nicolas Saenz Julienne wrote:
> > On Thu, 2022-03-10 at 10:25 +, Stefan Hajnoczi wrote:
> > > On Thu, Mar 03, 2022 at 03:58:20PM +0100, Nicolas Saenz Julienne wrote:
&
On Mon, 2022-03-14 at 13:35 +, Stefan Hajnoczi wrote:
> On Fri, Mar 11, 2022 at 11:40:30AM +0100, Nicolas Saenz Julienne wrote:
> > On Thu, 2022-03-10 at 10:45 +, Stefan Hajnoczi wrote:
> > > On Thu, Mar 03, 2022 at 04:13:07PM +0100, Nicolas Saenz Julienne wrote:
> >
using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
[1] For example:
-object main-loop,id=mai
two depend on it.
- Process the 'hw' subdir earlier, as it introduces files into the
'qom' source set.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
---
Changes since v2:
- reword commit message to better explain compilation dependencies.
iendly and pass all the tests.
[1]
https://patchwork.ozlabs.org/project/qemu-devel/patch/20220202175234.656711-1-nsaen...@redhat.com/
---
Changes since v2:
- Get rid of wrong locking/waiting
- Fix qapi versioning
- Better commit messages
Changes since v1:
- Address all Stefan's commen
lization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne
---
Changes
Hi Markus,
On Wed, 2022-03-16 at 15:28 +0100, Markus Armbruster wrote:
> Nicolas Saenz Julienne writes:
>
> > 'event-loop-base' provides basic property handling for all 'AioContext'
> > based event loops. So let's define a new 'MainLoopClass'
two depend on it.
- Process the 'hw' subdir earlier, as it introduces files into the
'qom' source set.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
Reviewed-by: Stefan Hajnoczi
---
Changes since v3:
- Fix event-loop-base compilation so it de
using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
[1] For example:
-object main-loop,id=mai
pBaseProperties.
- Fix failures on first compilation due to race between
event-loop-base.o and qapi header generation.
Changes since v2:
- Get rid of wrong locking/waiting
- Fix qapi versioning
- Better commit messages
Changes since v1:
- Address all Stefan's comments
- Introduce new
lization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne
Review
On Fri, 2022-04-01 at 11:35 +0200, Nicolas Saenz Julienne wrote:
Subject says 0/4 where is should've been 0/3.
> As discussed on the previous RFC[1] the thread-pool's dynamic thread
> management doesn't play well with real-time and latency sensitive
> systems. This se
On Mon, 2022-04-04 at 10:29 +0100, Stefan Hajnoczi wrote:
> On Fri, Apr 01, 2022 at 11:35:20AM +0200, Nicolas Saenz Julienne wrote:
> > As discussed on the previous RFC[1] the thread-pool's dynamic thread
> > management doesn't play well with real-time and latency se
On Fri, 2022-04-22 at 13:13 +0200, Markus Armbruster wrote:
> Nicolas Saenz Julienne writes:
>
> > 'event-loop-base' provides basic property handling for all 'AioContext'
> > based event loops. So let's define a new 'MainLoopClass' that inh
On Fri, 2022-04-22 at 13:15 +0200, Markus Armbruster wrote:
> Nicolas Saenz Julienne writes:
>
> > The thread pool regulates itself: when idle, it kills threads until
> > empty, when in demand, it creates new threads until full. This behaviour
> > doesn't pla
On Fri, 2022-04-22 at 13:40 +0200, Nicolas Saenz Julienne wrote:
> > > +##
> > > +{ 'struct': 'EventLoopBaseProperties',
> > > + 'data': { '*aio-max-batch': 'int' } }
> > > +
> > > ##
> > >
two depend on it.
- Process the 'hw' subdir earlier, as it introduces files into the
'qom' source set.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
Reviewed-by: Stefan Hajnoczi
---
Changes since v4:
- Introduce relevant qom.json changes, l
using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
[1] For example:
-object main-loop,id=mai
nce v1:
- Address all Stefan's comments
- Introduce new fix
Nicolas Saenz Julienne (3):
Introduce event-loop-base abstract class
util/main-loop: Introduce the main loop into QOM
util/event-loop-base: Introduce options to set the thread pool size
event-loop-base.c|
lization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne
Review
On Mon, 2022-04-25 at 07:22 +0200, Markus Armbruster wrote:
> Nicolas Saenz Julienne writes:
>
> > As discussed on the previous RFC[1] the thread-pool's dynamic thread
> > management doesn't play well with real-time and latency sensitive
> > systems. This s
v3:
- Avoid duplication in qom.json by creating EventLoopBaseProperties.
- Fix failures on first compilation due to race between
event-loop-base.o and qapi header generation.
Changes since v2:
- Get rid of wrong locking/waiting
- Fix qapi versioning
- Better commit messages
Changes since v1:
using the '-object' keyword[1].
Only one instance of 'MainLoopClass' might be created at any time.
'EventLoopBaseClass' learns a new callback, 'can_be_deleted()' so as to
mark 'MainLoop' as non-deletable.
[1] For example:
-object main-loop,id=mai
lization or upon updating the property's value, remain
available during its lifetime regardless of demand, and destroyed upon
freeing it. A properly characterized workload will then be able to
configure the pool to avoid any latency spikes.
Signed-off-by: Nicolas Saenz Julienne
Reviewed-by:
two depend on it.
- Process the 'hw' subdir earlier, as it introduces files into the
'qom' source set.
No functional changes intended.
Signed-off-by: Nicolas Saenz Julienne
Reviewed-by: Stefan Hajnoczi
Acked-by: Markus Armbruster
---
Changes since v4:
- Introdu
43 matches
Mail list logo
