Re: [PATCH v6 18/19] x86: Add support for generic vDSO

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Thomas Gleixner]

On Fri, 14 Jun 2019, Sasha Levin wrote:
> On Fri, Jun 14, 2019 at 01:15:23PM +0200, Thomas Gleixner wrote:
> > On Thu, 30 May 2019, Michael Kelley wrote:
> > > Vincenzo -- these changes for Hyper-V are a subset of a larger patch set
> > > I have that moves all of the Hyper-V clock/timer code into a separate
> > > clocksource driver in drivers/clocksource, with an include file in
> > > includes/clocksource.  That new include file should be able to work
> > > instead of your new mshyperv-tsc.h.  It also has the benefit of being
> > > ISA neutral, so it will work with my in-progress patch set to support
> > > Linux on Hyper-V on ARM64.  See https://lkml.org/lkml/2019/5/27/231
> > > for the new clocksource driver patch set.
> > 
> > Grrr. That's queued in hyperv-next for whatever reasons.
> 
> I queue up our future pull requests there to give them some soaking in
> -next.

What? You queue completely unreviewed stuff which touches two other
subsystems to let it soak in next?

> > Sasha, can you please provide me the branch to pull from so I can have a
> > common base for all the various changes floating around?
> 
> I'll send you a unified pull request for these changes.

Which has not materialized yet.

TBH, I'm pretty grumpy about those clocksource changes. Here is the
diffstat:

 MAINTAINERS                          |    2 
 arch/x86/entry/vdso/vclock_gettime.c |    1 
 arch/x86/entry/vdso/vma.c            |    2 
 arch/x86/hyperv/hv_init.c            |   91 ---------
 arch/x86/include/asm/hyperv-tlfs.h   |    6 
 arch/x86/include/asm/mshyperv.h      |   81 +-------
 arch/x86/kernel/cpu/mshyperv.c       |    2 
 arch/x86/kvm/x86.c                   |    1 
 drivers/clocksource/Makefile         |    1 
 drivers/clocksource/hyperv_timer.c   |  322 +++++++++++++++++++++++++++++++++++
 drivers/hv/Kconfig                   |    3 
 drivers/hv/hv.c                      |  156 ----------------
 drivers/hv/hv_util.c                 |    1 
 drivers/hv/hyperv_vmbus.h            |    3 
 drivers/hv/vmbus_drv.c               |   42 ++--
 include/clocksource/hyperv_timer.h   |  105 +++++++++++

While the world and some more people have been CC'ed on those patches,
neither the clocksource nor the x86 maintainer have been.

When I gave Vincenzo the advise to base his code on that hyper-v branch, I
expected that I find the related patches in my mail backlog. No, they have
not been there because I was not on CC.

Folks, please stop chosing Cc lists as you like. We have well established
rules for that. And please stop queueing random unreviewed patches in
next. Next is not a playground for not ready and unreviewed stuff. No, the
hyper-v inbreed Reviewed-by is not sufficient for anything x86 and
clocksource related.

After chasing and looking at those patches, which have horrible subject
lines and changelogs btw, I was not able to judge quickly whether that
stuff is self contained or not. So no, I fixed up the fallout and rebased
Vincenzos VDSO stuff on mainline w/o those hyperv changes simply because if
they are not self contained they will break bisection badly.

I'm going to push out the VDSO series later today. That will nicely break
in combination with the hyper-next branch. Stephen, please drop that and do
not try to handle the fallout. That stuff needs to go through the proper
channels or at least be acked/reviewed by the relevant maintainers. So the
hyper-v folks can rebase themself and post it proper.

Yours grumpy,

	tglx

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Sasha Levin]

On Sat, Jun 22, 2019 at 04:46:28PM +0200, Thomas Gleixner wrote:
>On Fri, 14 Jun 2019, Sasha Levin wrote:
>> On Fri, Jun 14, 2019 at 01:15:23PM +0200, Thomas Gleixner wrote:
>> > On Thu, 30 May 2019, Michael Kelley wrote:
>> > > Vincenzo -- these changes for Hyper-V are a subset of a larger patch set
>> > > I have that moves all of the Hyper-V clock/timer code into a separate
>> > > clocksource driver in drivers/clocksource, with an include file in
>> > > includes/clocksource.  That new include file should be able to work
>> > > instead of your new mshyperv-tsc.h.  It also has the benefit of being
>> > > ISA neutral, so it will work with my in-progress patch set to support
>> > > Linux on Hyper-V on ARM64.  See https://lkml.org/lkml/2019/5/27/231
>> > > for the new clocksource driver patch set.
>> >
>> > Grrr. That's queued in hyperv-next for whatever reasons.
>>
>> I queue up our future pull requests there to give them some soaking in
>> -next.
>
>What? You queue completely unreviewed stuff which touches two other
>subsystems to let it soak in next?

It was out on LKML for 2+ weeks before I've pulled it in. As it mostly
touches hyperv bits I felt comfortable to give it time in -next (but not
actually to try and merge it until it gets a few acks).

>> > Sasha, can you please provide me the branch to pull from so I can have a
>> > common base for all the various changes floating around?
>>
>> I'll send you a unified pull request for these changes.
>
>Which has not materialized yet.

Appologies about this. I ended up with way more travel than I would have
liked (writing this from an airport). I've reset our hyperv-next branch
to remove these 3 commits until we figure this out.

>TBH, I'm pretty grumpy about those clocksource changes. Here is the
>diffstat:
>
> MAINTAINERS                          |    2
> arch/x86/entry/vdso/vclock_gettime.c |    1
> arch/x86/entry/vdso/vma.c            |    2
> arch/x86/hyperv/hv_init.c            |   91 ---------
> arch/x86/include/asm/hyperv-tlfs.h   |    6
> arch/x86/include/asm/mshyperv.h      |   81 +-------
> arch/x86/kernel/cpu/mshyperv.c       |    2
> arch/x86/kvm/x86.c                   |    1
> drivers/clocksource/Makefile         |    1
> drivers/clocksource/hyperv_timer.c   |  322 +++++++++++++++++++++++++++++++++++
> drivers/hv/Kconfig                   |    3
> drivers/hv/hv.c                      |  156 ----------------
> drivers/hv/hv_util.c                 |    1
> drivers/hv/hyperv_vmbus.h            |    3
> drivers/hv/vmbus_drv.c               |   42 ++--
> include/clocksource/hyperv_timer.h   |  105 +++++++++++
>
>While the world and some more people have been CC'ed on those patches,
>neither the clocksource nor the x86 maintainer have been.
>
>When I gave Vincenzo the advise to base his code on that hyper-v branch, I
>expected that I find the related patches in my mail backlog. No, they have
>not been there because I was not on CC.
>
>Folks, please stop chosing Cc lists as you like. We have well established
>rules for that. And please stop queueing random unreviewed patches in
>next. Next is not a playground for not ready and unreviewed stuff. No, the
>hyper-v inbreed Reviewed-by is not sufficient for anything x86 and
>clocksource related.

I'm sorry for this, you were supposed to be Cc'ed on these patches and I
see that you were not.

>After chasing and looking at those patches, which have horrible subject
>lines and changelogs btw, I was not able to judge quickly whether that
>stuff is self contained or not. So no, I fixed up the fallout and rebased
>Vincenzos VDSO stuff on mainline w/o those hyperv changes simply because if
>they are not self contained they will break bisection badly.
>
>I'm going to push out the VDSO series later today. That will nicely break
>in combination with the hyper-next branch. Stephen, please drop that and do
>not try to handle the fallout. That stuff needs to go through the proper
>channels or at least be acked/reviewed by the relevant maintainers. So the
>hyper-v folks can rebase themself and post it proper.

Okay, thank you. We'll rebase and resend.

--
Thanks,
Sasha

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 359 bytes --]

Hi Sasha,

On Sun, 23 Jun 2019 15:09:29 -0400 Sasha Levin <sashal@kernel.org> wrote:
>
> Appologies about this. I ended up with way more travel than I would have
> liked (writing this from an airport). I've reset our hyperv-next branch
> to remove these 3 commits until we figure this out.

But not pushed out, yet?

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Sasha Levin]

On Mon, Jun 24, 2019 at 07:58:34AM +1000, Stephen Rothwell wrote:
>Hi Sasha,
>
>On Sun, 23 Jun 2019 15:09:29 -0400 Sasha Levin <sashal@kernel.org> wrote:
>>
>> Appologies about this. I ended up with way more travel than I would have
>> liked (writing this from an airport). I've reset our hyperv-next branch
>> to remove these 3 commits until we figure this out.
>
>But not pushed out, yet?

Pushed now. For some reason the airport wifi was blocking ssh :/

--
Thanks,
Sasha

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 204 bytes --]

Hi Sasha,

On Sun, 23 Jun 2019 20:24:30 -0400 Sasha Levin <sashal@kernel.org> wrote:
>
> Pushed now. For some reason the airport wifi was blocking ssh :/

Thanks.

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 18/19] x86: Add support for generic vDSO

[Thomas Gleixner]

Sasha,

On Sun, 23 Jun 2019, Sasha Levin wrote:
> On Sat, Jun 22, 2019 at 04:46:28PM +0200, Thomas Gleixner wrote:
> > Folks, please stop chosing Cc lists as you like. We have well established
> > rules for that. And please stop queueing random unreviewed patches in
> > next. Next is not a playground for not ready and unreviewed stuff. No, the
> > hyper-v inbreed Reviewed-by is not sufficient for anything x86 and
> > clocksource related.
> 
> I'm sorry for this, you were supposed to be Cc'ed on these patches and I
> see that you were not.

All good. I've vented steam and am back to normal pressure :)

> > After chasing and looking at those patches, which have horrible subject
> > lines and changelogs btw, I was not able to judge quickly whether that
> > stuff is self contained or not. So no, I fixed up the fallout and rebased
> > Vincenzos VDSO stuff on mainline w/o those hyperv changes simply because if
> > they are not self contained they will break bisection badly.
> > 
> > I'm going to push out the VDSO series later today. That will nicely break

Not yet, but soon :)

> > in combination with the hyper-next branch. Stephen, please drop that and do
> > not try to handle the fallout. That stuff needs to go through the proper
> > channels or at least be acked/reviewed by the relevant maintainers. So the
> > hyper-v folks can rebase themself and post it proper.
> 
> Okay, thank you. We'll rebase and resend.

I have no objections if you collect hyper-v stuff, quite the contrary, but
changes which touch other subsystems need to be coordinated upfront. That's
all I'm asking for.

Btw, that clocksource stuff looks good code wise, just the change logs need
some care and after the VDSO stuff hits next we need to sort out the
logistics. I hope these changes are completely self contained. If not we'll
find a solution.

Thanks,

	tglx

RE: [PATCH v6 18/19] x86: Add support for generic vDSO

[Michael Kelley]

From: Thomas Gleixner <tglx@linutronix.de> Sent: Sunday, June 23, 2019 3:13 PM
> 
> I have no objections if you collect hyper-v stuff, quite the contrary, but
> changes which touch other subsystems need to be coordinated upfront. That's
> all I'm asking for.
> 
> Btw, that clocksource stuff looks good code wise, just the change logs need
> some care and after the VDSO stuff hits next we need to sort out the
> logistics. I hope these changes are completely self contained. If not we'll
> find a solution.
>

In my view, the only thing that potentially needs a solution is where the
Hyper-V clock code used by VDSO ends up in the code tree.  I think the
right long term place is include/clocksource/hyperv_timer.h.   That location
is architecture neutral, and the same Hyper-V clock code will be shared by
the Hyper-V on ARM64 support that's in process.

Vincenzo's patch set creates a new file arch/x86/include/asm/mshyperv-tsc.h,
which I will want to move when creating the separate Hyper-V clocksource
driver.   If you're OK with that file existing for a release and then going away,
that's fine.  Alternatively, put the code in include/clocksource/hyperv_timer.h
now as part of the VDSO patch set so it's in the right place from the start.  My
subsequent patch set will add a few additional tweaks to remove x86-isms
and fully integrate with the separate Hyper-V clocksource driver.

Michael 

RE: [PATCH v6 18/19] x86: Add support for generic vDSO

[Thomas Gleixner]

On Mon, 24 Jun 2019, Michael Kelley wrote:
> From: Thomas Gleixner <tglx@linutronix.de> Sent: Sunday, June 23, 2019 3:13 PM
> > 
> > I have no objections if you collect hyper-v stuff, quite the contrary, but
> > changes which touch other subsystems need to be coordinated upfront. That's
> > all I'm asking for.
> > 
> > Btw, that clocksource stuff looks good code wise, just the change logs need
> > some care and after the VDSO stuff hits next we need to sort out the
> > logistics. I hope these changes are completely self contained. If not we'll
> > find a solution.
> >
> 
> In my view, the only thing that potentially needs a solution is where the
> Hyper-V clock code used by VDSO ends up in the code tree.  I think the
> right long term place is include/clocksource/hyperv_timer.h.   That location
> is architecture neutral, and the same Hyper-V clock code will be shared by
> the Hyper-V on ARM64 support that's in process.
> 
> Vincenzo's patch set creates a new file arch/x86/include/asm/mshyperv-tsc.h,
> which I will want to move when creating the separate Hyper-V clocksource
> driver.   If you're OK with that file existing for a release and then going away,
> that's fine.  Alternatively, put the code in include/clocksource/hyperv_timer.h
> now as part of the VDSO patch set so it's in the right place from the start.  My
> subsequent patch set will add a few additional tweaks to remove x86-isms
> and fully integrate with the separate Hyper-V clocksource driver.

I don't care whether this goes into 5.3 or later. If you can provide me
rebased self contained patches on top of

  git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git timers/vdso

I'm happy to pull them in on top.

Thanks,

	tglx

RE: [PATCH v6 18/19] x86: Add support for generic vDSO

[Michael Kelley]

From: Thomas Gleixner <tglx@linutronix.de> Sent: Sunday, June 23, 2019 5:25 PM
> 
> I don't care whether this goes into 5.3 or later. If you can provide me
> rebased self contained patches on top of
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git timers/vdso
> 
> I'm happy to pull them in on top.
> 

I've sent out "v4" of the patch set to create a Hyper-V clocksource, based
on the above tree.  It is contained to Hyper-V code, plus updating a #include
statement in two of the VDSO files and in one KVM file.  If the KVM file
update is problematic, the patch set can just wait until 5.3-rc1.

Michael