[patch 0/2] make kcmp own config entry

[patch 0/2] make kcmp own config entry

[Cyrill Gorcunov]

Hi guys, kcmp syscall (according to git log) is being wired up for a number
of architectures other than x86 (where it was developed initially). But the
main Makefile which controling its compilation is still x86 dependant.

Thus I think kcmp may deserve own config entry unrelated to arch type it
is being compiled for. In case if some particular arch doesn't support it
ENOSYS should be returned. This small set also wires up kcmp for ARM.

Please take a look, thanks!

	Cyrill

[patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

[-- Attachment #1: kcmp-config --]
[-- Type: text/plain, Size: 2322 bytes --]

Since kcmp syscall has been implemented (initially on
x86 architecture) a number of other archs wire it up
as well: xtensa, sparc, sh, s390, mips, microblaze,
m68k (not taking into account those who uses
<asm-generic/unistd.h> for syscall numbers
definitions).

But the Makefile, which turns kcmp.o generation on
still depends on former config-x86. Thus get rid
of this limitation and make kcmp.o depend on CONFIG_KCMP
option.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Valdis.Kletnieks@vt.edu
Cc: Michal Marek <mmarek@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 init/Kconfig    |    9 +++++++++
 kernel/Makefile |    4 +---
 2 files changed, 10 insertions(+), 3 deletions(-)

Index: linux-2.6.git/init/Kconfig
===================================================================
--- linux-2.6.git.orig/init/Kconfig
+++ linux-2.6.git/init/Kconfig
@@ -279,6 +279,15 @@ config FHANDLE
 	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
 	  syscalls.
 
+config KCMP
+	bool "kcmp syscall"
+	default n
+	help
+	  If you say Y here, a user level program will be able to use
+	  kcmp(2) syscall.
+
+	  If unsure, say N.
+
 config AUDIT
 	bool "Auditing support"
 	depends on NET
Index: linux-2.6.git/kernel/Makefile
===================================================================
--- linux-2.6.git.orig/kernel/Makefile
+++ linux-2.6.git/kernel/Makefile
@@ -25,9 +25,7 @@ endif
 obj-y += sched/
 obj-y += power/
 
-ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
-obj-$(CONFIG_X86) += kcmp.o
-endif
+obj-$(CONFIG_KCMP) += kcmp.o
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o

[patch 2/2] arm: Wire up kcmp syscall

[Cyrill Gorcunov]

[-- Attachment #1: kcmp-arm --]
[-- Type: text/plain, Size: 1293 bytes --]

Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Cc: Russell King <linux@arm.linux.org.uk>
---
 arch/arm/include/uapi/asm/unistd.h |    2 +-
 arch/arm/kernel/calls.S            |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Index: linux-2.6.git/arch/arm/include/uapi/asm/unistd.h
===================================================================
--- linux-2.6.git.orig/arch/arm/include/uapi/asm/unistd.h
+++ linux-2.6.git/arch/arm/include/uapi/asm/unistd.h
@@ -404,7 +404,7 @@
 #define __NR_setns			(__NR_SYSCALL_BASE+375)
 #define __NR_process_vm_readv		(__NR_SYSCALL_BASE+376)
 #define __NR_process_vm_writev		(__NR_SYSCALL_BASE+377)
-					/* 378 for kcmp */
+#define __NR_kcmp			(__NR_SYSCALL_BASE+378)
 #define __NR_finit_module		(__NR_SYSCALL_BASE+379)
 
 /*
Index: linux-2.6.git/arch/arm/kernel/calls.S
===================================================================
--- linux-2.6.git.orig/arch/arm/kernel/calls.S
+++ linux-2.6.git/arch/arm/kernel/calls.S
@@ -387,7 +387,7 @@
 /* 375 */	CALL(sys_setns)
 		CALL(sys_process_vm_readv)
 		CALL(sys_process_vm_writev)
-		CALL(sys_ni_syscall)	/* reserved for sys_kcmp */
+		CALL(sys_kcmp)
 		CALL(sys_finit_module)
 #ifndef syscalls_counted
 .equ syscalls_padding, ((NR_syscalls + 3) & ~3) - NR_syscalls

Re: [patch 2/2] arm: Wire up kcmp syscall

[Cyrill Gorcunov]

quilt eat original header, here is the former, sorry
for incovenience.
---
From: Alexander Kartashov <alekskartashov@parallels.com>
Subject: arm: Wire up kcmp syscall

Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Cc: Russell King <linux@arm.linux.org.uk>
---
 arch/arm/include/uapi/asm/unistd.h |    2 +-
 arch/arm/kernel/calls.S            |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Index: linux-2.6.git/arch/arm/include/uapi/asm/unistd.h
===================================================================
--- linux-2.6.git.orig/arch/arm/include/uapi/asm/unistd.h
+++ linux-2.6.git/arch/arm/include/uapi/asm/unistd.h
@@ -404,7 +404,7 @@
 #define __NR_setns			(__NR_SYSCALL_BASE+375)
 #define __NR_process_vm_readv		(__NR_SYSCALL_BASE+376)
 #define __NR_process_vm_writev		(__NR_SYSCALL_BASE+377)
-					/* 378 for kcmp */
+#define __NR_kcmp			(__NR_SYSCALL_BASE+378)
 #define __NR_finit_module		(__NR_SYSCALL_BASE+379)
 
 /*
Index: linux-2.6.git/arch/arm/kernel/calls.S
===================================================================
--- linux-2.6.git.orig/arch/arm/kernel/calls.S
+++ linux-2.6.git/arch/arm/kernel/calls.S
@@ -387,7 +387,7 @@
 /* 375 */	CALL(sys_setns)
 		CALL(sys_process_vm_readv)
 		CALL(sys_process_vm_writev)
-		CALL(sys_ni_syscall)	/* reserved for sys_kcmp */
+		CALL(sys_kcmp)
 		CALL(sys_finit_module)
 #ifndef syscalls_counted
 .equ syscalls_padding, ((NR_syscalls + 3) & ~3) - NR_syscalls

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[KOSAKI Motohiro]

(2/19/13 1:48 AM), Cyrill Gorcunov wrote:
> Since kcmp syscall has been implemented (initially on
> x86 architecture) a number of other archs wire it up
> as well: xtensa, sparc, sh, s390, mips, microblaze,
> m68k (not taking into account those who uses
> <asm-generic/unistd.h> for syscall numbers
> definitions).
> 
> But the Makefile, which turns kcmp.o generation on
> still depends on former config-x86. Thus get rid
> of this limitation and make kcmp.o depend on CONFIG_KCMP
> option.

Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Michal Marek]

On 19.2.2013 07:48, Cyrill Gorcunov wrote:
> --- linux-2.6.git.orig/init/Kconfig
> +++ linux-2.6.git/init/Kconfig
> @@ -279,6 +279,15 @@ config FHANDLE
>  	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
>  	  syscalls.
>  
> +config KCMP
> +	bool "kcmp syscall"
> +	default n

Previously, kcmp was enabled with CONFIG_CHECKPOINT_RESTORE. Maybe
change it to 'default CHECKPOINT_RESTORE', so that make oldconfig does
the right thing?

Michal

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Tue, Feb 19, 2013 at 10:22:21AM +0100, Michal Marek wrote:
> On 19.2.2013 07:48, Cyrill Gorcunov wrote:
> > --- linux-2.6.git.orig/init/Kconfig
> > +++ linux-2.6.git/init/Kconfig
> > @@ -279,6 +279,15 @@ config FHANDLE
> >  	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
> >  	  syscalls.
> >  
> > +config KCMP
> > +	bool "kcmp syscall"
> > +	default n
> 
> Previously, kcmp was enabled with CONFIG_CHECKPOINT_RESTORE. Maybe
> change it to 'default CHECKPOINT_RESTORE', so that make oldconfig does
> the right thing?

Yes, thanks, good point! I've been adviced that make it obj-y by default
may be a good option as well, not sure.
---
From: Cyrill Gorcunov <gorcunov@openvz.org>
Subject: kcmp: Make it to depend on CONFIG_KCMP

Since kcmp syscall has been implemented (initially on
x86 architecture) a number of other archs wire it up
as well: xtensa, sparc, sh, s390, mips, microblaze,
m68k (not taking into account those who uses
<asm-generic/unistd.h> for syscall numbers
definitions).

But the Makefile, which turns kcmp.o generation on
still depends on former config-x86. Thus get rid
of this limitation and make kcmp.o depend on CONFIG_KCMP
option.

v2:
 - As Michal pointed the old configs might already use of
   CHECKPOINT_RESTORE, so make "default" accordingly.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Valdis.Kletnieks@vt.edu
Cc: Michal Marek <mmarek@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 init/Kconfig    |    9 +++++++++
 kernel/Makefile |    4 +---
 2 files changed, 10 insertions(+), 3 deletions(-)

Index: linux-2.6.git/init/Kconfig
===================================================================
--- linux-2.6.git.orig/init/Kconfig
+++ linux-2.6.git/init/Kconfig
@@ -279,6 +279,15 @@ config FHANDLE
 	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
 	  syscalls.
 
+config KCMP
+	bool "kcmp syscall"
+	default CHECKPOINT_RESTORE
+	help
+	  If you say Y here, a user level program will be able to use
+	  kcmp(2) syscall.
+
+	  If unsure, say N.
+
 config AUDIT
 	bool "Auditing support"
 	depends on NET
Index: linux-2.6.git/kernel/Makefile
===================================================================
--- linux-2.6.git.orig/kernel/Makefile
+++ linux-2.6.git/kernel/Makefile
@@ -25,9 +25,7 @@ endif
 obj-y += sched/
 obj-y += power/
 
-ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
-obj-$(CONFIG_X86) += kcmp.o
-endif
+obj-$(CONFIG_KCMP) += kcmp.o
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[H. Peter Anvin]

On 02/19/2013 01:31 AM, Cyrill Gorcunov wrote:
> +
> +	  If unsure, say N.
> +

Wrong advice.  In this particular case, Y is the safe alternative.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Tue, Feb 19, 2013 at 09:53:47AM -0800, H. Peter Anvin wrote:
> On 02/19/2013 01:31 AM, Cyrill Gorcunov wrote:
> >+
> >+	  If unsure, say N.
> >+
> 
> Wrong advice.  In this particular case, Y is the safe alternative.
---
From: Cyrill Gorcunov <gorcunov@openvz.org>
Subject: kcmp: Make it to depend on CONFIG_KCMP

Since kcmp syscall has been implemented (initially on
x86 architecture) a number of other archs wire it up
as well: xtensa, sparc, sh, s390, mips, microblaze,
m68k (not taking into account those who uses
<asm-generic/unistd.h> for syscall numbers
definitions).

But the Makefile, which turns kcmp.o generation on
still depends on former config-x86. Thus get rid
of this limitation and make kcmp.o depend on CONFIG_KCMP
option.

v2:
 - As Michal pointed the old configs might already use of
   CHECKPOINT_RESTORE, so make "default" accordingly.
 - Advice Y if unsure

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Valdis.Kletnieks@vt.edu
Cc: Michal Marek <mmarek@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 init/Kconfig    |    9 +++++++++
 kernel/Makefile |    4 +---
 2 files changed, 10 insertions(+), 3 deletions(-)

Index: linux-2.6.git/init/Kconfig
===================================================================
--- linux-2.6.git.orig/init/Kconfig
+++ linux-2.6.git/init/Kconfig
@@ -279,6 +279,15 @@ config FHANDLE
 	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
 	  syscalls.
 
+config KCMP
+	bool "kcmp syscall"
+	default CHECKPOINT_RESTORE
+	help
+	  If you say Y here, a user level program will be able to use
+	  kcmp(2) syscall.
+
+	  If unsure, say Y.
+
 config AUDIT
 	bool "Auditing support"
 	depends on NET
Index: linux-2.6.git/kernel/Makefile
===================================================================
--- linux-2.6.git.orig/kernel/Makefile
+++ linux-2.6.git/kernel/Makefile
@@ -25,9 +25,7 @@ endif
 obj-y += sched/
 obj-y += power/
 
-ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
-obj-$(CONFIG_X86) += kcmp.o
-endif
+obj-$(CONFIG_KCMP) += kcmp.o
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Andrew Morton]

On Tue, 19 Feb 2013 22:28:38 +0400
Cyrill Gorcunov <gorcunov@openvz.org> wrote:

> Since kcmp syscall has been implemented (initially on
> x86 architecture) a number of other archs wire it up
> as well: xtensa, sparc, sh, s390, mips, microblaze,
> m68k (not taking into account those who uses
> <asm-generic/unistd.h> for syscall numbers
> definitions).
> 
> But the Makefile, which turns kcmp.o generation on
> still depends on former config-x86. Thus get rid
> of this limitation and make kcmp.o depend on CONFIG_KCMP
> option.
> 
> ...
>
> --- linux-2.6.git.orig/init/Kconfig
> +++ linux-2.6.git/init/Kconfig
> @@ -279,6 +279,15 @@ config FHANDLE
>  	  get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
>  	  syscalls.
>  
> +config KCMP
> +	bool "kcmp syscall"
> +	default CHECKPOINT_RESTORE
> +	help
> +	  If you say Y here, a user level program will be able to use
> +	  kcmp(2) syscall.
> +
> +	  If unsure, say Y.
> +
>  config AUDIT
>  	bool "Auditing support"
>  	depends on NET
> Index: linux-2.6.git/kernel/Makefile
> ===================================================================
> --- linux-2.6.git.orig/kernel/Makefile
> +++ linux-2.6.git/kernel/Makefile
> @@ -25,9 +25,7 @@ endif
>  obj-y += sched/
>  obj-y += power/
>  
> -ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
> -obj-$(CONFIG_X86) += kcmp.o
> -endif
> +obj-$(CONFIG_KCMP) += kcmp.o
>  obj-$(CONFIG_FREEZER) += freezer.o
>  obj-$(CONFIG_PROFILING) += profile.o
>  obj-$(CONFIG_STACKTRACE) += stacktrace.o

This permits people to select kcmp with CONFIG_CHECKPOINT_RESTORE=n. 
Is there any point in doing that?

What would be wrong with just doing

	obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o

?

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[H. Peter Anvin]

On 02/19/2013 01:42 PM, Andrew Morton wrote:
>> Index: linux-2.6.git/kernel/Makefile
>> ===================================================================
>> --- linux-2.6.git.orig/kernel/Makefile
>> +++ linux-2.6.git/kernel/Makefile
>> @@ -25,9 +25,7 @@ endif
>>   obj-y += sched/
>>   obj-y += power/
>>
>> -ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
>> -obj-$(CONFIG_X86) += kcmp.o
>> -endif
>> +obj-$(CONFIG_KCMP) += kcmp.o
>>   obj-$(CONFIG_FREEZER) += freezer.o
>>   obj-$(CONFIG_PROFILING) += profile.o
>>   obj-$(CONFIG_STACKTRACE) += stacktrace.o
>
> This permits people to select kcmp with CONFIG_CHECKPOINT_RESTORE=n.
> Is there any point in doing that?
>
> What would be wrong with just doing
>
> 	obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o
>

The real question is if there are any potential use cases of kcmp() 
outside checkpoint/restore.  It is actually a very general facility.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Tue, Feb 19, 2013 at 01:42:56PM -0800, Andrew Morton wrote:
> > Index: linux-2.6.git/kernel/Makefile
> > ===================================================================
> > --- linux-2.6.git.orig/kernel/Makefile
> > +++ linux-2.6.git/kernel/Makefile
> > @@ -25,9 +25,7 @@ endif
> >  obj-y += sched/
> >  obj-y += power/
> >  
> > -ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
> > -obj-$(CONFIG_X86) += kcmp.o
> > -endif
> > +obj-$(CONFIG_KCMP) += kcmp.o
> >  obj-$(CONFIG_FREEZER) += freezer.o
> >  obj-$(CONFIG_PROFILING) += profile.o
> >  obj-$(CONFIG_STACKTRACE) += stacktrace.o
> 
> This permits people to select kcmp with CONFIG_CHECKPOINT_RESTORE=n. 
> Is there any point in doing that?
> 
> What would be wrong with just doing
> 
> 	obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o

I think this syscall is usefull even without c/r stuff.
That's why I made it with separate config option.

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Andrew Morton]

On Wed, 20 Feb 2013 01:54:32 +0400
Cyrill Gorcunov <gorcunov@openvz.org> wrote:

> On Tue, Feb 19, 2013 at 01:42:56PM -0800, Andrew Morton wrote:
> > > Index: linux-2.6.git/kernel/Makefile
> > > ===================================================================
> > > --- linux-2.6.git.orig/kernel/Makefile
> > > +++ linux-2.6.git/kernel/Makefile
> > > @@ -25,9 +25,7 @@ endif
> > >  obj-y += sched/
> > >  obj-y += power/
> > >  
> > > -ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
> > > -obj-$(CONFIG_X86) += kcmp.o
> > > -endif
> > > +obj-$(CONFIG_KCMP) += kcmp.o
> > >  obj-$(CONFIG_FREEZER) += freezer.o
> > >  obj-$(CONFIG_PROFILING) += profile.o
> > >  obj-$(CONFIG_STACKTRACE) += stacktrace.o
> > 
> > This permits people to select kcmp with CONFIG_CHECKPOINT_RESTORE=n. 
> > Is there any point in doing that?
> > 
> > What would be wrong with just doing
> > 
> > 	obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o
> 
> I think this syscall is usefull even without c/r stuff.
> That's why I made it with separate config option.

hm, OK.

But the patch also permits CONFIG_CHECKPOINT_RESTORE=y, CONFIG_KCMP=n
which surely isn't something which CRIU wants to support?

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Tue, Feb 19, 2013 at 01:48:27PM -0800, H. Peter Anvin wrote:
> >
> >This permits people to select kcmp with CONFIG_CHECKPOINT_RESTORE=n.
> >Is there any point in doing that?
> >
> >What would be wrong with just doing
> >
> >	obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o
> >
> 
> The real question is if there are any potential use cases of kcmp()
> outside checkpoint/restore.  It is actually a very general facility.

As far as I remember Eric was pointing to such potential use at early
time when kcmp being only started developing (I'm trying to find this
email in archive, if only my memory doesn't betray me and it was about
something else ;) Nevertheless, one may write utility to output
statistics on shared "resource" used by a task (don't know if it's
that useful since we use kcmp for c/r sake, but still).

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Tue, Feb 19, 2013 at 02:00:35PM -0800, Andrew Morton wrote:
> > 
> > I think this syscall is usefull even without c/r stuff.
> > That's why I made it with separate config option.
> 
> hm, OK.
> 
> But the patch also permits CONFIG_CHECKPOINT_RESTORE=y, CONFIG_KCMP=n
> which surely isn't something which CRIU wants to support?

Hmm, yes from one pov this feature is useful for out-of-c/r user,
from another pov -- we will have to ask users to turn on additional
CONFIG entries (which i'm sure not set by default in wide range of distros).
Thus it seems less paiful way is either make it obj-(CHECKPOINT_RESTORE)
as you proposed, or obj-y by default. The last can't be undone, so I'll
prepare the patch for obj-(CHECKPOINT_RESTORE) I think.

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Wed, Feb 20, 2013 at 02:11:46AM +0400, Cyrill Gorcunov wrote:
> On Tue, Feb 19, 2013 at 02:00:35PM -0800, Andrew Morton wrote:
> > > 
> > > I think this syscall is usefull even without c/r stuff.
> > > That's why I made it with separate config option.
> > 
> > hm, OK.
> > 
> > But the patch also permits CONFIG_CHECKPOINT_RESTORE=y, CONFIG_KCMP=n
> > which surely isn't something which CRIU wants to support?
> 
> Hmm, yes from one pov this feature is useful for out-of-c/r user,
> from another pov -- we will have to ask users to turn on additional
> CONFIG entries (which i'm sure not set by default in wide range of distros).
> Thus it seems less paiful way is either make it obj-(CHECKPOINT_RESTORE)
> as you proposed, or obj-y by default. The last can't be undone, so I'll
> prepare the patch for obj-(CHECKPOINT_RESTORE) I think.

Here is update. Peter, does it look good for you too?
---
From: Cyrill Gorcunov <gorcunov@openvz.org>
Subject: kcmp: Make it to depend on CHECKPOINT_RESTORE

Since kcmp syscall has been implemented (initially on x86
architecture) a number of other archs wire it up as well:
xtensa, sparc, sh, s390, mips, microblaze, m68k (not taking
into account those who uses <asm-generic/unistd.h> for syscall
numbers definitions).

But the Makefile, which turns kcmp.o generation on still
depends on former config-x86. Thus get rid of this
limitation and make kcmp.o depend on CHECKPOINT_RESTORE
option.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Valdis.Kletnieks@vt.edu
Cc: Michal Marek <mmarek@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 kernel/Makefile |    4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

Index: linux-2.6.git/kernel/Makefile
===================================================================
--- linux-2.6.git.orig/kernel/Makefile
+++ linux-2.6.git/kernel/Makefile
@@ -25,9 +25,7 @@ endif
 obj-y += sched/
 obj-y += power/
 
-ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
-obj-$(CONFIG_X86) += kcmp.o
-endif
+obj-$(CHECKPOINT_RESTORE) += kcmp.o
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[Cyrill Gorcunov]

On Wed, Feb 20, 2013 at 02:15:48AM +0400, Cyrill Gorcunov wrote:
> 
> Here is update. Peter, does it look good for you too?
Sigh, -ENOTQUILTREFRESHED, sorry
---
From: Cyrill Gorcunov <gorcunov@openvz.org>
Subject: kcmp: Make it to depend on CHECKPOINT_RESTORE

Since kcmp syscall has been implemented (initially on
x86 architecture) a number of other archs wire it up
as well: xtensa, sparc, sh, s390, mips, microblaze,
m68k (not taking into account those who uses
<asm-generic/unistd.h> for syscall numbers
definitions).

But the Makefile, which turns kcmp.o generation on
still depends on former config-x86. Thus get rid
of this limitation and make kcmp.o depend on CHECKPOINT_RESTORE
option.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Vasiliy Kulikov <segoon@openwall.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Valdis.Kletnieks@vt.edu
Cc: Michal Marek <mmarek@suse.cz>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 kernel/Makefile |    4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

Index: linux-2.6.git/kernel/Makefile
===================================================================
--- linux-2.6.git.orig/kernel/Makefile
+++ linux-2.6.git/kernel/Makefile
@@ -25,9 +25,7 @@ endif
 obj-y += sched/
 obj-y += power/
 
-ifeq ($(CONFIG_CHECKPOINT_RESTORE),y)
-obj-$(CONFIG_X86) += kcmp.o
-endif
+obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o

Re: [patch 1/2] kcmp: Make it to depend on CONFIG_KCMP

[H. Peter Anvin]

On 02/19/2013 02:11 PM, Cyrill Gorcunov wrote:
> On Tue, Feb 19, 2013 at 02:00:35PM -0800, Andrew Morton wrote:
>>>
>>> I think this syscall is usefull even without c/r stuff.
>>> That's why I made it with separate config option.
>>
>> hm, OK.
>>
>> But the patch also permits CONFIG_CHECKPOINT_RESTORE=y, CONFIG_KCMP=n
>> which surely isn't something which CRIU wants to support?
>
> Hmm, yes from one pov this feature is useful for out-of-c/r user,
> from another pov -- we will have to ask users to turn on additional
> CONFIG entries (which i'm sure not set by default in wide range of distros).
> Thus it seems less paiful way is either make it obj-(CHECKPOINT_RESTORE)
> as you proposed, or obj-y by default. The last can't be undone, so I'll
> prepare the patch for obj-(CHECKPOINT_RESTORE) I think.
>

Well, that's what dependencies are for.

Either way we might just want to wait for such a use case to appear, I 
don't know.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Re: [patch 2/2] arm: Wire up kcmp syscall

[Andrew Morton]

On Tue, 19 Feb 2013 11:07:03 +0400
Cyrill Gorcunov <gorcunov@openvz.org> wrote:

> From: Alexander Kartashov <alekskartashov@parallels.com>
> Subject: arm: Wire up kcmp syscall
> 
> Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
> Cc: Russell King <linux@arm.linux.org.uk>

This should have had signed-off-by:you, as you were on the patch's
delivery path.

Re: [patch 2/2] arm: Wire up kcmp syscall

[Cyrill Gorcunov]

On Wed, Feb 20, 2013 at 03:17:23PM -0800, Andrew Morton wrote:
> On Tue, 19 Feb 2013 11:07:03 +0400
> Cyrill Gorcunov <gorcunov@openvz.org> wrote:
> 
> > From: Alexander Kartashov <alekskartashov@parallels.com>
> > Subject: arm: Wire up kcmp syscall
> > 
> > Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
> > Cc: Russell King <linux@arm.linux.org.uk>
> 
> This should have had signed-off-by:you, as you were on the patch's
> delivery path.

Ouch, sorry Andrew! Thanks!

Re: [patch 2/2] arm: Wire up kcmp syscall

[Arnd Bergmann]

On Tuesday 19 February 2013, Cyrill Gorcunov wrote:
> From: Alexander Kartashov <alekskartashov@parallels.com>
> Subject: arm: Wire up kcmp syscall
> 
> Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
> Cc: Russell King <linux@arm.linux.org.uk>

Acked-by: Arnd Bergmann <arnd@arndb.de>

Looks uncontroversial since the number is already reserved and
there is a cond_syscall() line, which means it does not
even depend on patch 1.

Russell, are you picking this one up?