Skip to contentSkip to conversation

Future of icount discussion for next KVM call?

DISCUSSION Alex Bennée 5 participants 9 messages Feb 16, 2023 – Mar 3, 2023 1 message not in the archive KVM

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

(replying all because qemu-devel rejected my email again)

Quoted 40 lines · Alex Bennée

Hi Juan,

Do we have an agenda for next weeks KVM call yet? If there is space I'd like to take some time to discuss the future direction of icount.

Specifically I believe there might be some proposals for how we could support icount with MTTCG worth discussing. From my point of view icount provides too things:

- a sense of time vaguely related to execution rather than wall clock
- determinism

I would love to divorce the former from icount and punt it to plugins. The plugin would be free to instrument as heavily or lightly as it sees fit and provide its best guess as to guest time on demand. I wrote this idea up as a card in Linaro's JIRA if anyone is interested:

https://linaro.atlassian.net/browse/QEMU-481

Being able to punt cost modelling and sense of time into plugins would allow the core icount support to concentrate on determinism. Then any attempt to enable icount for MTTCG would then have to ensure it stays deterministic.

Richard and I have discussed the problem a few times and weren't sure it was solvable but I'm totally open to hearing ideas on how to do it. Fundamentally I think we would have to ensure any TB's doing IO would have to execute in an exclusive context. The TCG code already has mechanisms to ensure all IO is only done at the end of blocks so it doesn't seem a huge leap to ensure we execute those blocks exclusively. However there is still the problem of what to do about other pure computation threads getting ahead or behind of the IO blocks on subsequent runs.

Anyway does anyone else have ideas to bring to the discussion?

--
Alex Bennée
Virtualisation Tech Lead @ Linaro

Signature · 2 lines
Alex Bennée
Emulation and Virtualisation Tech Lead @ Linaro
#2 lore raw
Quoted 8 lines · Alex Bennée

(replying all because qemu-devel rejected my email again)

Quoted 4 lines · Alex Bennée

Hi Juan,

Do we have an agenda for next weeks KVM call yet? If there is space I'd like to take some time to discuss the future direction of icount.

For next week we have:
- more single binary qemu (philippe?)
- TDX migration from intel.
We asked them on the previous call to change their design to transfer stuff through migration channels and not create a new channel. But I haven't heard from intel. (wei?)
They agreed to send the slides and post the code before continue discussion.

And now I like the title of you topic

- Future Direction of icount

O:-)

So, I will recommend 20 minutes each if Wei shows up, or 30/30 for the rest.

What do the rest of the people think.

Quoted 29 lines
Quoted 29 lines

Specifically I believe there might be some proposals for how we could support icount with MTTCG worth discussing. From my point of view icount provides too things:

- a sense of time vaguely related to execution rather than wall clock
- determinism

I would love to divorce the former from icount and punt it to plugins. The plugin would be free to instrument as heavily or lightly as it sees fit and provide its best guess as to guest time on demand. I wrote this idea up as a card in Linaro's JIRA if anyone is interested:

https://linaro.atlassian.net/browse/QEMU-481

Being able to punt cost modelling and sense of time into plugins would allow the core icount support to concentrate on determinism. Then any attempt to enable icount for MTTCG would then have to ensure it stays deterministic.

Richard and I have discussed the problem a few times and weren't sure it was solvable but I'm totally open to hearing ideas on how to do it. Fundamentally I think we would have to ensure any TB's doing IO would have to execute in an exclusive context. The TCG code already has mechanisms to ensure all IO is only done at the end of blocks so it doesn't seem a huge leap to ensure we execute those blocks exclusively. However there is still the problem of what to do about other pure computation threads getting ahead or behind of the IO blocks on subsequent runs.

Quoted 1 line
Quoted 1 line

Anyway does anyone else have ideas to bring to the discussion?

Hat on to you O:-)

Open discussion with a Jira Epic and a good introduction.
I am sorry that I am not an expert (or even newbie) on that part of qemu to give apport anything.

Thanks, Juan.

#3 lore raw
Quoted 21 lines · Juan Quintela
Quoted 8 lines · Alex Bennée

(replying all because qemu-devel rejected my email again)

Quoted 4 lines · Alex Bennée

Hi Juan,

Do we have an agenda for next weeks KVM call yet? If there is space I'd like to take some time to discuss the future direction of icount.

For next week we have:
- more single binary qemu (philippe?)
- TDX migration from intel.
We asked them on the previous call to change their design to transfer stuff through migration channels and not create a new channel. But I haven't heard from intel. (wei?)
They agreed to send the slides and post the code before continue discussion.

And now I like the title of you topic

Quoted 8 lines

- Future Direction of icount

O:-)

So, I will recommend 20 minutes each if Wei shows up, or 30/30 for the rest.

What do the rest of the people think.

I think we either need fewer topics per call (ideally one), or strictly enforced time limits per topic. I don't fancy meetings where the topic that made me attend falls off the end.

The former may necessitate more frequent calls.

#4 lore raw
Quoted 7 lines · Markus Armbruster
Quoted 5 lines · Juan Quintela
Quoted 4 lines · Alex Bennée

(replying all because qemu-devel rejected my email again)

On Thu, 16 Feb 2023 at 10:19, Alex Bennée <[email protected]> wrote:

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines
Quoted 4 lines

Hi Juan,

Do we have an agenda for next weeks KVM call yet? If there is space I'd like to take some time to discuss the future direction of icount.

For next week we have:
- more single binary qemu (philippe?)

I'd rather skip "Single qemu-system binary" for next week agenda. (In 2 weeks we could discuss CPU topology and shared buses.)

Quoted 15 lines
Quoted 15 lines

- TDX migration from intel.
We asked them on the previous call to change their design to transfer stuff through migration channels and not create a new channel. But I haven't heard from intel. (wei?)
They agreed to send the slides and post the code before continue discussion. >>
And now I like the title of you topic

- Future Direction of icount

O:-)

So, I will recommend 20 minutes each if Wei shows up, or 30/30 for the rest.

Quoted 7 lines
Quoted 1 line

What do the rest of the people think.

I think we either need fewer topics per call (ideally one), or strictly enforced time limits per topic. I don't fancy meetings where the topic that made me attend falls off the end.

The former may necessitate more frequent calls.

IIRC it was said we can have more than every 2 weeks per community request, otherwise we should try to respect the current cadence to maintain the current inertia.

#5 lore raw
In reply to Markus Armbruster
Quoted 27 lines · Markus Armbruster
Quoted 25 lines · Juan Quintela
Quoted 8 lines · Alex Bennée

(replying all because qemu-devel rejected my email again)

Quoted 4 lines · Alex Bennée

Hi Juan,

Do we have an agenda for next weeks KVM call yet? If there is space I'd like to take some time to discuss the future direction of icount.

For next week we have:
- more single binary qemu (philippe?)
- TDX migration from intel.
We asked them on the previous call to change their design to transfer stuff through migration channels and not create a new channel. But I haven't heard from intel. (wei?)
They agreed to send the slides and post the code before continue discussion.

And now I like the title of you topic

- Future Direction of icount

O:-)

Quoted 8 lines
Quoted 4 lines

So, I will recommend 20 minutes each if Wei shows up, or 30/30 for the rest.

What do the rest of the people think.

I think we either need fewer topics per call (ideally one), or strictly enforced time limits per topic. I don't fancy meetings where the topic that made me attend falls off the end.

I'm happy to push this out to the next KVM meeting if you want. The main driver for posting this now was to ensure availability for interested participants and maybe get some pre-discussion on the list.

Quoted 1 line

The former may necessitate more frequent calls.

Maybe - the number of agenda items tends to ebb and flow and we seem to be making good use of the slot now to discuss a wide array of topics.

Signature · 2 lines
Alex Bennée
Virtualisation Tech Lead @ Linaro
#6 lore raw
In reply to Alex Bennée
Quoted 1 line · Alex Bennée

(replying all because qemu-devel rejected my email again)

Just to see what we are having now:

- single qemu binary moved to next slot (moved to next week?) Phillipe proposal
- TDX migration: we have the slides, but no code
So I guess we can move it to the following slot, when we have a chance to look at the code, Wei?
- Markus asked to have only one topic per call
(and it is reasonable)
- the future of icount cames with documentation to read before the call.

So I think we can do the icount call this week, and discuss here on list if we move to a weekly call on the same slot.

What do you think?

Later, Juan.

#7 lore raw
Quoted 7 lines · Juan Quintela

Just to see what we are having now:

- single qemu binary moved to next slot (moved to next week?) Phillipe proposal
- TDX migration: we have the slides, but no code
So I guess we can move it to the following slot, when we have a chance to look at the code, Wei?

It's ok to me to continue the discussion on either Feb 21st or March 7th, and I plan to finish some update and share the code before end of next week.

Thanks,
Wei

#8 lore raw
Quoted 19 lines

-----Original Message-----
From: Wang, Wei W
Sent: Thursday, February 16, 2023 10:36 PM
To: [email protected]; Alex Bennée <[email protected]>
Cc: Paolo Bonzini <[email protected]>; Pavel Dovgalyuk <[email protected]>; [email protected]; Richard Henderson <[email protected]>; Mark Burton <[email protected]>; Bill Mills <[email protected]>; Marco Liebel <[email protected]>; Alexandre Iooss <[email protected]>; Mahmoud Mandour <[email protected]>; Emilio Cota <[email protected]>; kvm-devel <[email protected]>; Philippe Mathieu- Daudé <[email protected]>
Subject: RE: Future of icount discussion for next KVM call?

Quoted 2 lines · Juan Quintela

Just to see what we are having now:

Quoted 8 lines
Quoted 5 lines

- single qemu binary moved to next slot (moved to next week?) Phillipe proposal
- TDX migration: we have the slides, but no code
So I guess we can move it to the following slot, when we have a chance to look at the code, Wei?

It's ok to me to continue the discussion on either Feb 21st or March 7th, and I plan to finish some update and share the code before end of next week.

KVM code can be read here: https://github.com/intel/tdx/ tdx-mig-wip QEMU code will be shared soon.

#9 lore raw
Quoted 9 lines · Wei W
Quoted 9 lines · Juan Quintela
Quoted 7 lines

Just to see what we are having now:

- single qemu binary moved to next slot (moved to next week?) Phillipe proposal
- TDX migration: we have the slides, but no code
So I guess we can move it to the following slot, when we have a chance to look at the code, Wei?

Quoted 6 lines
Quoted 2 lines

It's ok to me to continue the discussion on either Feb 21st or March 7th, and I plan to finish some update and share the code before end of

next week.

KVM code can be read here: https://github.com/intel/tdx/ tdx-mig-wip QEMU code will be shared soon.

QEMU code can be read here: https://github.com/intel/qemu-tdx/   tdx-mig-wip