Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Josh Boyer]

On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
<paulmck@linux.vnet.ibm.com> wrote:
> From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
>

Why would one want to do this?  The commit log doesn't explain the
switch and why it's great.  Mostly because the commit log doesn't
exist.  (I'll stop channeling Andrew now)

It also might be a good idea to CC the linuxppc-dev list for this
patch.  Or at least Kumar for the mpc86xx changes.

josh

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

On Sat, Apr 04, 2009 at 09:21:58AM -0400, Josh Boyer wrote:
> On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
> <paulmck@linux.vnet.ibm.com> wrote:
> > From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> 
> Why would one want to do this?  The commit log doesn't explain the
> switch and why it's great.  Mostly because the commit log doesn't
> exist.  (I'll stop channeling Andrew now)

Indeed, I am sure that I will get the hang the relevant git commands
at some point, but clearly am not yet there.

The thing that is great about this patch is that it is a step on
the way to retiring Classic RCU entirely.

> It also might be a good idea to CC the linuxppc-dev list for this
> patch.  Or at least Kumar for the mpc86xx changes.

Good point -- I will do so on the next version.

							Thanx, Paul

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Ingo Molnar]

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

> On Sat, Apr 04, 2009 at 09:21:58AM -0400, Josh Boyer wrote:
> > On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
> > <paulmck@linux.vnet.ibm.com> wrote:
> > > From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > 
> > Why would one want to do this?  The commit log doesn't explain the
> > switch and why it's great.  Mostly because the commit log doesn't
> > exist.  (I'll stop channeling Andrew now)
> 
> Indeed, I am sure that I will get the hang the relevant git 
> commands at some point, but clearly am not yet there.
> 
> The thing that is great about this patch is that it is a step on 
> the way to retiring Classic RCU entirely.

I'd suggest you dont do these dozens of defconfig updates at all - 
arch maintainers can do this themselves, once the core kernel 
default has been flipped.

	Ingo

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Andi Kleen]

> I'd suggest you dont do these dozens of defconfig updates at all - 
> arch maintainers can do this themselves, once the core kernel 
> default has been flipped.

AFAIK they don't. There's no make updatemyconfigwithdefaultstatementsconfig

-Andi
-- 
ak@linux.intel.com -- Speaking for myself only.

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

On Sun, Apr 05, 2009 at 12:37:59PM +0200, Andi Kleen wrote:
> > I'd suggest you dont do these dozens of defconfig updates at all - 
> > arch maintainers can do this themselves, once the core kernel 
> > default has been flipped.
> 
> AFAIK they don't. There's no make updatemyconfigwithdefaultstatementsconfig

My guess is that I would in fact have to send a few patches.  But I
suppose that the real question is "how long should we wait to drop
Classic RCU once the last arch has changed its defconfig?"  I have
been assuming that we would wait one release cycle.

							Thanx, Paul

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Andi Kleen]

On Sun, Apr 05, 2009 at 09:49:04AM -0700, Paul E. McKenney wrote:
> On Sun, Apr 05, 2009 at 12:37:59PM +0200, Andi Kleen wrote:
> > > I'd suggest you dont do these dozens of defconfig updates at all - 
> > > arch maintainers can do this themselves, once the core kernel 
> > > default has been flipped.
> > 
> > AFAIK they don't. There's no make updatemyconfigwithdefaultstatementsconfig
> 
> My guess is that I would in fact have to send a few patches.  But I
> suppose that the real question is "how long should we wait to drop
> Classic RCU once the last arch has changed its defconfig?"  I have
> been assuming that we would wait one release cycle.

I don't think you have to wait, stale entries in defconfigs happen
all the time and Kconfig just asks then.

-Andi
-- 
ak@linux.intel.com -- Speaking for myself only.

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

On Sun, Apr 05, 2009 at 07:01:13PM +0200, Andi Kleen wrote:
> On Sun, Apr 05, 2009 at 09:49:04AM -0700, Paul E. McKenney wrote:
> > On Sun, Apr 05, 2009 at 12:37:59PM +0200, Andi Kleen wrote:
> > > > I'd suggest you dont do these dozens of defconfig updates at all - 
> > > > arch maintainers can do this themselves, once the core kernel 
> > > > default has been flipped.
> > > 
> > > AFAIK they don't. There's no make updatemyconfigwithdefaultstatementsconfig
> > 
> > My guess is that I would in fact have to send a few patches.  But I
> > suppose that the real question is "how long should we wait to drop
> > Classic RCU once the last arch has changed its defconfig?"  I have
> > been assuming that we would wait one release cycle.
> 
> I don't think you have to wait, stale entries in defconfigs happen
> all the time and Kconfig just asks then.

Good point!

							Thanx, Paul

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

[-- Attachment #1: Type: text/plain, Size: 1499 bytes --]

On Sun, Apr 05, 2009 at 12:17:56PM +0200, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> 
> > On Sat, Apr 04, 2009 at 09:21:58AM -0400, Josh Boyer wrote:
> > > On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
> > > <paulmck@linux.vnet.ibm.com> wrote:
> > > > From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > 
> > > Why would one want to do this?  The commit log doesn't explain the
> > > switch and why it's great.  Mostly because the commit log doesn't
> > > exist.  (I'll stop channeling Andrew now)
> > 
> > Indeed, I am sure that I will get the hang the relevant git 
> > commands at some point, but clearly am not yet there.
> > 
> > The thing that is great about this patch is that it is a step on 
> > the way to retiring Classic RCU entirely.
> 
> I'd suggest you dont do these dozens of defconfig updates at all - 
> arch maintainers can do this themselves, once the core kernel 
> default has been flipped.

I am of course personally quite happy with this approach in the short
term, as I would only need to worry about one patch.

My one concern would be that we cannot retire Classic RCU until all the
arch maintainers remove CONFIG_CLASSIC_RCU=y from their defconfig files.
However, given that the attached script suffices, I am not too worried
about that.

Of course, none of this matters unless people agree that Tree RCU has
proven reliable enough to be made default.  So, any objections to such
a change at this point?

							Thanx, Paul

[-- Attachment #2: 2rcutree.sh --]
[-- Type: application/x-sh, Size: 1448 bytes --]

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Ingo Molnar]

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

> On Sun, Apr 05, 2009 at 12:17:56PM +0200, Ingo Molnar wrote:
> > 
> > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> > 
> > > On Sat, Apr 04, 2009 at 09:21:58AM -0400, Josh Boyer wrote:
> > > > On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
> > > > <paulmck@linux.vnet.ibm.com> wrote:
> > > > > From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > > 
> > > > Why would one want to do this?  The commit log doesn't explain the
> > > > switch and why it's great.  Mostly because the commit log doesn't
> > > > exist.  (I'll stop channeling Andrew now)
> > > 
> > > Indeed, I am sure that I will get the hang the relevant git 
> > > commands at some point, but clearly am not yet there.
> > > 
> > > The thing that is great about this patch is that it is a step on 
> > > the way to retiring Classic RCU entirely.
> > 
> > I'd suggest you dont do these dozens of defconfig updates at all - 
> > arch maintainers can do this themselves, once the core kernel 
> > default has been flipped.
> 
> I am of course personally quite happy with this approach in the 
> short term, as I would only need to worry about one patch.
> 
> My one concern would be that we cannot retire Classic RCU until 
> all the arch maintainers remove CONFIG_CLASSIC_RCU=y from their 
> defconfig files. However, given that the attached script suffices, 
> I am not too worried about that.

ah, that's not a real concern. Most defconfigs are stale all the 
time, and arch maintainers typically refresh them at around rc7.
(if at all) Kconfig will warn about (and ignore) stale entries.

So no, there's no requirement to touch defconfigs - to the contrary: 
it has way too wide of a cross section with other trees to be really 
practical to carry in the RCU tree.

	Ingo

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

On Wed, Apr 08, 2009 at 02:46:23PM +0200, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> 
> > On Sun, Apr 05, 2009 at 12:17:56PM +0200, Ingo Molnar wrote:
> > > 
> > > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> > > 
> > > > On Sat, Apr 04, 2009 at 09:21:58AM -0400, Josh Boyer wrote:
> > > > > On Fri, Apr 3, 2009 at 8:27 PM, Paul E. McKenney
> > > > > <paulmck@linux.vnet.ibm.com> wrote:
> > > > > > From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > > > 
> > > > > Why would one want to do this?  The commit log doesn't explain the
> > > > > switch and why it's great.  Mostly because the commit log doesn't
> > > > > exist.  (I'll stop channeling Andrew now)
> > > > 
> > > > Indeed, I am sure that I will get the hang the relevant git 
> > > > commands at some point, but clearly am not yet there.
> > > > 
> > > > The thing that is great about this patch is that it is a step on 
> > > > the way to retiring Classic RCU entirely.
> > > 
> > > I'd suggest you dont do these dozens of defconfig updates at all - 
> > > arch maintainers can do this themselves, once the core kernel 
> > > default has been flipped.
> > 
> > I am of course personally quite happy with this approach in the 
> > short term, as I would only need to worry about one patch.
> > 
> > My one concern would be that we cannot retire Classic RCU until 
> > all the arch maintainers remove CONFIG_CLASSIC_RCU=y from their 
> > defconfig files. However, given that the attached script suffices, 
> > I am not too worried about that.
> 
> ah, that's not a real concern. Most defconfigs are stale all the 
> time, and arch maintainers typically refresh them at around rc7.
> (if at all) Kconfig will warn about (and ignore) stale entries.
> 
> So no, there's no requirement to touch defconfigs - to the contrary: 
> it has way too wide of a cross section with other trees to be really 
> practical to carry in the RCU tree.

Sounds good to me!!!  So, is http://lkml.org/lkml/2009/4/3/561
sufficient, or should I repost?

							Thanx, Paul

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Ingo Molnar]

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

> > So no, there's no requirement to touch defconfigs - to the 
> > contrary: it has way too wide of a cross section with other 
> > trees to be really practical to carry in the RCU tree.
> 
> Sounds good to me!!!  So, is http://lkml.org/lkml/2009/4/3/561 
> sufficient, or should I repost?

I have queued up your patch:

 31c9a24: RCU: make treercu be default

... a week ago :-)

	Ingo

Re: [PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

On Wed, Apr 08, 2009 at 08:10:05PM +0200, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> 
> > > So no, there's no requirement to touch defconfigs - to the 
> > > contrary: it has way too wide of a cross section with other 
> > > trees to be really practical to carry in the RCU tree.
> > 
> > Sounds good to me!!!  So, is http://lkml.org/lkml/2009/4/3/561 
> > sufficient, or should I repost?
> 
> I have queued up your patch:
> 
>  31c9a24: RCU: make treercu be default
> 
> ... a week ago :-)

Color me slow, stupid, and oblivious!  ;-)

Thank you for applying it!

							Thanx, Paul

[PATCH -tip] Make TREE_RCU be default (was CLASSIC_RCU).

[Paul E. McKenney]

Given that there have been no TREE_RCU bugs reported for some months,
it is time to make TREE_RCU be the default RCU implementation.  The rest
of this series updates defconfig files that: (1) have CONFIG_SMP=y and
(2) explicitly mention CONFIG_CLASSIC_RCU.

Given that I don't have access to most of the relevant hardware, much
of this is untested.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 init/Kconfig |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/init/Kconfig b/init/Kconfig
index 236a793..ef9d030 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -302,7 +302,7 @@ menu "RCU Subsystem"
 
 choice
 	prompt "RCU Implementation"
-	default CLASSIC_RCU
+	default TREE_RCU
 
 config CLASSIC_RCU
 	bool "Classic RCU"
-- 
1.5.2.5

[PATCH -tip] Convert CONFIG_SMP=y powerpc defconfigs to TREE_RCU.

[Paul E. McKenney]

From: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 arch/powerpc/configs/cell_defconfig    |    2 +-
 arch/powerpc/configs/celleb_defconfig  |    2 +-
 arch/powerpc/configs/chrp32_defconfig  |    2 +-
 arch/powerpc/configs/g5_defconfig      |    2 +-
 arch/powerpc/configs/iseries_defconfig |    2 +-
 arch/powerpc/configs/maple_defconfig   |    2 +-
 arch/powerpc/configs/mpc86xx_defconfig |    4 ++--
 arch/powerpc/configs/pasemi_defconfig  |    2 +-
 arch/powerpc/configs/ppc64_defconfig   |    2 +-
 arch/powerpc/configs/ps3_defconfig     |    4 ++--
 arch/powerpc/configs/pseries_defconfig |    2 +-
 11 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/arch/powerpc/configs/cell_defconfig b/arch/powerpc/configs/cell_defconfig
index c6d2baa..00bc3dc 100644
--- a/arch/powerpc/configs/cell_defconfig
+++ b/arch/powerpc/configs/cell_defconfig
@@ -155,7 +155,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 
 #
 # Platform support
diff --git a/arch/powerpc/configs/celleb_defconfig b/arch/powerpc/configs/celleb_defconfig
index d212377..e133858 100644
--- a/arch/powerpc/configs/celleb_defconfig
+++ b/arch/powerpc/configs/celleb_defconfig
@@ -148,7 +148,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 
 #
 # Platform support
diff --git a/arch/powerpc/configs/chrp32_defconfig b/arch/powerpc/configs/chrp32_defconfig
index 5094a65..33f69b0 100644
--- a/arch/powerpc/configs/chrp32_defconfig
+++ b/arch/powerpc/configs/chrp32_defconfig
@@ -157,7 +157,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 # CONFIG_FREEZER is not set
 
 #
diff --git a/arch/powerpc/configs/g5_defconfig b/arch/powerpc/configs/g5_defconfig
index fc90592..2c31eec 100644
--- a/arch/powerpc/configs/g5_defconfig
+++ b/arch/powerpc/configs/g5_defconfig
@@ -162,7 +162,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 # CONFIG_FREEZER is not set
 CONFIG_PPC_MSI_BITMAP=y
 
diff --git a/arch/powerpc/configs/iseries_defconfig b/arch/powerpc/configs/iseries_defconfig
index f925c55..affb125 100644
--- a/arch/powerpc/configs/iseries_defconfig
+++ b/arch/powerpc/configs/iseries_defconfig
@@ -160,7 +160,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 # CONFIG_FREEZER is not set
 
 #
diff --git a/arch/powerpc/configs/maple_defconfig b/arch/powerpc/configs/maple_defconfig
index 8b24400..17af126 100644
--- a/arch/powerpc/configs/maple_defconfig
+++ b/arch/powerpc/configs/maple_defconfig
@@ -158,7 +158,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 # CONFIG_FREEZER is not set
 CONFIG_PPC_MSI_BITMAP=y
 
diff --git a/arch/powerpc/configs/mpc86xx_defconfig b/arch/powerpc/configs/mpc86xx_defconfig
index adc756e..12aa6cd 100644
--- a/arch/powerpc/configs/mpc86xx_defconfig
+++ b/arch/powerpc/configs/mpc86xx_defconfig
@@ -157,8 +157,8 @@ CONFIG_IOSCHED_CFQ=y
 CONFIG_DEFAULT_CFQ=y
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="cfq"
-CONFIG_CLASSIC_RCU=y
-# CONFIG_TREE_RCU is not set
+CONFIG_TREE_RCU=y
+# CONFIG_CLASSIC_RCU is not set
 # CONFIG_PREEMPT_RCU is not set
 # CONFIG_TREE_RCU_TRACE is not set
 # CONFIG_PREEMPT_RCU_TRACE is not set
diff --git a/arch/powerpc/configs/pasemi_defconfig b/arch/powerpc/configs/pasemi_defconfig
index 4f8681c..e320a56 100644
--- a/arch/powerpc/configs/pasemi_defconfig
+++ b/arch/powerpc/configs/pasemi_defconfig
@@ -150,7 +150,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 
 #
 # Platform support
diff --git a/arch/powerpc/configs/ppc64_defconfig b/arch/powerpc/configs/ppc64_defconfig
index 2524018..e4dc721 100644
--- a/arch/powerpc/configs/ppc64_defconfig
+++ b/arch/powerpc/configs/ppc64_defconfig
@@ -176,7 +176,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 # CONFIG_FREEZER is not set
 CONFIG_PPC_MSI_BITMAP=y
 
diff --git a/arch/powerpc/configs/ps3_defconfig b/arch/powerpc/configs/ps3_defconfig
index ac14f52..606e0ba 100644
--- a/arch/powerpc/configs/ps3_defconfig
+++ b/arch/powerpc/configs/ps3_defconfig
@@ -79,8 +79,8 @@ CONFIG_POSIX_MQUEUE=y
 #
 # RCU Subsystem
 #
-CONFIG_CLASSIC_RCU=y
-# CONFIG_TREE_RCU is not set
+CONFIG_TREE_RCU=y
+# CONFIG_CLASSIC_RCU is not set
 # CONFIG_PREEMPT_RCU is not set
 # CONFIG_TREE_RCU_TRACE is not set
 # CONFIG_PREEMPT_RCU_TRACE is not set
diff --git a/arch/powerpc/configs/pseries_defconfig b/arch/powerpc/configs/pseries_defconfig
index 61b1008..d023f33 100644
--- a/arch/powerpc/configs/pseries_defconfig
+++ b/arch/powerpc/configs/pseries_defconfig
@@ -176,7 +176,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_CFQ is not set
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
-CONFIG_CLASSIC_RCU=y
+CONFIG_TREE_RCU=y
 CONFIG_FREEZER=y
 CONFIG_PPC_MSI_BITMAP=y
 
-- 
1.5.2.5