[patch 012/265] timerfd: wire the new timerfd API to the x86 family

[patch 012/265] timerfd: wire the new timerfd API to the x86 family

[akpm]

From: Davide Libenzi <davidel@xmailserver.org>

Wires up the new timerfd API to the x86 family.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 arch/x86/ia32/ia32entry.S          |    4 +++-
 arch/x86/kernel/syscall_table_32.S |    4 +++-
 include/asm-x86/unistd_32.h        |    4 +++-
 include/asm-x86/unistd_64.h        |    9 +++++++--
 4 files changed, 16 insertions(+), 5 deletions(-)

diff -puN arch/x86/ia32/ia32entry.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/ia32/ia32entry.S
--- a/arch/x86/ia32/ia32entry.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
+++ a/arch/x86/ia32/ia32entry.S
@@ -722,7 +722,9 @@ ia32_sys_call_table:
 	.quad sys_epoll_pwait
 	.quad compat_sys_utimensat	/* 320 */
 	.quad compat_sys_signalfd
-	.quad compat_sys_timerfd
+	.quad sys_timerfd_create
 	.quad sys_eventfd
 	.quad sys32_fallocate
+	.quad compat_sys_timerfd_settime	/* 325 */
+	.quad compat_sys_timerfd_gettime
 ia32_syscall_end:
diff -puN arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/kernel/syscall_table_32.S
--- a/arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
+++ a/arch/x86/kernel/syscall_table_32.S
@@ -321,6 +321,8 @@ ENTRY(sys_call_table)
 	.long sys_epoll_pwait
 	.long sys_utimensat		/* 320 */
 	.long sys_signalfd
-	.long sys_timerfd
+	.long sys_timerfd_create
 	.long sys_eventfd
 	.long sys_fallocate
+	.long sys_timerfd_settime	/* 325 */
+	.long sys_timerfd_gettime
diff -puN include/asm-x86/unistd_32.h~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family include/asm-x86/unistd_32.h
--- a/include/asm-x86/unistd_32.h~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
+++ a/include/asm-x86/unistd_32.h
@@ -327,9 +327,11 @@
 #define __NR_epoll_pwait	319
 #define __NR_utimensat		320
 #define __NR_signalfd		321
-#define __NR_timerfd		322
+#define __NR_timerfd_create	322
 #define __NR_eventfd		323
 #define __NR_fallocate		324
+#define __NR_timerfd_settime	325
+#define __NR_timerfd_gettime	326
 
 #ifdef __KERNEL__
 
diff -puN include/asm-x86/unistd_64.h~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family include/asm-x86/unistd_64.h
--- a/include/asm-x86/unistd_64.h~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
+++ a/include/asm-x86/unistd_64.h
@@ -629,12 +629,17 @@ __SYSCALL(__NR_utimensat, sys_utimensat)
 __SYSCALL(__NR_epoll_pwait, sys_epoll_pwait)
 #define __NR_signalfd				282
 __SYSCALL(__NR_signalfd, sys_signalfd)
-#define __NR_timerfd				283
-__SYSCALL(__NR_timerfd, sys_timerfd)
+#define __NR_timerfd_create			283
+__SYSCALL(__NR_timerfd_create, sys_timerfd_create)
 #define __NR_eventfd				284
 __SYSCALL(__NR_eventfd, sys_eventfd)
 #define __NR_fallocate				285
 __SYSCALL(__NR_fallocate, sys_fallocate)
+#define __NR_timerfd_settime			286
+__SYSCALL(__NR_timerfd_settime, sys_timerfd_settime)
+#define __NR_timerfd_gettime			287
+__SYSCALL(__NR_timerfd_gettime, sys_timerfd_gettime)
+
 
 #ifndef __NO_STUBS
 #define __ARCH_WANT_OLD_READDIR
_

[PATCH] Fix timerfd breakage on avr32

[Haavard Skinnemoen]

sys_timerfd() has been removed, but avr32 still references it from its
syscall table.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
---
On Mon, 04 Feb 2008 22:27:28 -0800
akpm@linux-foundation.org wrote:

> From: Davide Libenzi <davidel@xmailserver.org>
> 
> Wires up the new timerfd API to the x86 family.

Just one thing...

> diff -puN arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/kernel/syscall_table_32.S
> --- a/arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
> +++ a/arch/x86/kernel/syscall_table_32.S
> @@ -321,6 +321,8 @@ ENTRY(sys_call_table)
>  	.long sys_epoll_pwait
>  	.long sys_utimensat		/* 320 */
>  	.long sys_signalfd
> -	.long sys_timerfd

The next time you go and remove a system call, could you _please_ post
a HUGE warning to linux-arch? Or just do a quick grep and fix it up.

Judging by the number of -fix patches in -mm,
timerfd-v3-new-timerfd-api.patch broke almost every single architecture
in the tree. And not only is this potential breakage never mentioned in
the patch description, the patch that actually removes the
sys_timerfd() implementation was never Cc'ed linux-arch.

While I agree that new syscalls don't necessarily have to come wired up
and ready to use on all architectures, you need to take a lot more care
when removing syscalls that have already been wired up.

The patch below fixes avr32. Please apply.

 arch/avr32/kernel/syscall_table.S |    2 +-
 include/asm-avr32/unistd.h        |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff -puN arch/avr32/kernel/syscall_table.S~fix-timerfd-breakage-on-avr32-was-re-fix-variable-use-in-avr32-pte_alloc_one arch/avr32/kernel/syscall_table.S
--- a/arch/avr32/kernel/syscall_table.S~fix-timerfd-breakage-on-avr32-was-re-fix-variable-use-in-avr32-pte_alloc_one
+++ a/arch/avr32/kernel/syscall_table.S
@@ -293,6 +293,6 @@ sys_call_table:
 	.long	sys_shmctl
 	.long	sys_utimensat
 	.long	sys_signalfd
-	.long	sys_timerfd		/* 280 */
+	.long	sys_ni_syscall		/* 280, was sys_timerfd */
 	.long	sys_eventfd
 	.long	sys_ni_syscall		/* r8 is saturated at nr_syscalls */
diff -puN include/asm-avr32/unistd.h~fix-timerfd-breakage-on-avr32-was-re-fix-variable-use-in-avr32-pte_alloc_one include/asm-avr32/unistd.h
--- a/include/asm-avr32/unistd.h~fix-timerfd-breakage-on-avr32-was-re-fix-variable-use-in-avr32-pte_alloc_one
+++ a/include/asm-avr32/unistd.h
@@ -297,7 +297,7 @@
 
 #define __NR_utimensat		278
 #define __NR_signalfd		279
-#define __NR_timerfd		280
+/* 280 was __NR_timerfd */
 #define __NR_eventfd		281
 
 #ifdef __KERNEL__

Re: [PATCH] Fix timerfd breakage on avr32

[Haavard Skinnemoen]

On Tue, 5 Feb 2008 10:54:54 +0100
Haavard Skinnemoen <hskinnemoen@atmel.com> wrote:

> The next time you go and remove a system call, could you _please_ post
> a HUGE warning to linux-arch? Or just do a quick grep and fix it up.

Btw, a quick grep through 2.6.24-mm1 shows

arch/frv/kernel/entry.S:	.long sys_timerfd
arch/m32r/kernel/syscall_table.S:	.long sys_timerfd
arch/sh/kernel/syscalls_32.S:	.long sys_timerfd
arch/sh/kernel/syscalls_64.S:	.long sys_timerfd		/* 350 */

so frv, m32r and sh are probably broken too.

Haavard

Re: [PATCH] Fix timerfd breakage on avr32

[Andrew Morton]

On Tue, 5 Feb 2008 11:03:47 +0100 Haavard Skinnemoen <hskinnemoen@atmel.com> wrote:

> On Tue, 5 Feb 2008 10:54:54 +0100
> Haavard Skinnemoen <hskinnemoen@atmel.com> wrote:
> 
> > The next time you go and remove a system call, could you _please_ post
> > a HUGE warning to linux-arch? Or just do a quick grep and fix it up.
> 
> Btw, a quick grep through 2.6.24-mm1 shows
> 
> arch/frv/kernel/entry.S:	.long sys_timerfd
> arch/m32r/kernel/syscall_table.S:	.long sys_timerfd
> arch/sh/kernel/syscalls_32.S:	.long sys_timerfd
> arch/sh/kernel/syscalls_64.S:	.long sys_timerfd		/* 350 */
> 
> so frv, m32r and sh are probably broken too.
> 

argh. After fixing so many architectures I asked for a grepped
fix :(


This?


 arch/frv/kernel/entry.S          |    2 +-
 arch/m32r/kernel/syscall_table.S |    2 +-
 arch/sh/kernel/syscalls_32.S     |    2 +-
 arch/sh/kernel/syscalls_64.S     |    2 +-
 include/asm-frv/unistd.h         |    2 +-
 include/asm-m32r/unistd.h        |    2 +-
 include/asm-sh/unistd_32.h       |    2 +-
 include/asm-sh/unistd_64.h       |    2 +-
 8 files changed, 8 insertions(+), 8 deletions(-)

diff -puN arch/frv/kernel/entry.S~timerfd-fix-remaining-architectures arch/frv/kernel/entry.S
--- a/arch/frv/kernel/entry.S~timerfd-fix-remaining-architectures
+++ a/arch/frv/kernel/entry.S
@@ -1494,7 +1494,7 @@ sys_call_table:
 	.long sys_epoll_pwait
 	.long sys_utimensat		/* 320 */
 	.long sys_signalfd
-	.long sys_timerfd
+	.long sys_ni_syscall
 	.long sys_eventfd
 	.long sys_fallocate
 
diff -puN arch/m32r/kernel/syscall_table.S~timerfd-fix-remaining-architectures arch/m32r/kernel/syscall_table.S
--- a/arch/m32r/kernel/syscall_table.S~timerfd-fix-remaining-architectures
+++ a/arch/m32r/kernel/syscall_table.S
@@ -321,6 +321,6 @@ ENTRY(sys_call_table)
 	.long sys_epoll_pwait
 	.long sys_utimensat		/* 320 */
 	.long sys_signalfd
-	.long sys_timerfd
+	.long sys_ni_syscall
 	.long sys_eventfd
 	.long sys_fallocate
diff -puN arch/sh/kernel/syscalls_32.S~timerfd-fix-remaining-architectures arch/sh/kernel/syscalls_32.S
--- a/arch/sh/kernel/syscalls_32.S~timerfd-fix-remaining-architectures
+++ a/arch/sh/kernel/syscalls_32.S
@@ -338,6 +338,6 @@ ENTRY(sys_call_table)
 	.long sys_epoll_pwait
 	.long sys_utimensat		/* 320 */
 	.long sys_signalfd
-	.long sys_timerfd
+	.long sys_ni_syscall
 	.long sys_eventfd
 	.long sys_fallocate
diff -puN arch/sh/kernel/syscalls_64.S~timerfd-fix-remaining-architectures arch/sh/kernel/syscalls_64.S
--- a/arch/sh/kernel/syscalls_64.S~timerfd-fix-remaining-architectures
+++ a/arch/sh/kernel/syscalls_64.S
@@ -376,6 +376,6 @@ sys_call_table:
 	.long sys_epoll_pwait
 	.long sys_utimensat
 	.long sys_signalfd
-	.long sys_timerfd		/* 350 */
+	.long sys_ni_syscall		/* 350 */
 	.long sys_eventfd
 	.long sys_fallocate
diff -puN include/asm-frv/unistd.h~timerfd-fix-remaining-architectures include/asm-frv/unistd.h
--- a/include/asm-frv/unistd.h~timerfd-fix-remaining-architectures
+++ a/include/asm-frv/unistd.h
@@ -328,7 +328,7 @@
 #define __NR_epoll_pwait	319
 #define __NR_utimensat		320
 #define __NR_signalfd		321
-#define __NR_timerfd		322
+/* #define __NR_timerfd		322 removed */
 #define __NR_eventfd		323
 #define __NR_fallocate		324
 
diff -puN include/asm-m32r/unistd.h~timerfd-fix-remaining-architectures include/asm-m32r/unistd.h
--- a/include/asm-m32r/unistd.h~timerfd-fix-remaining-architectures
+++ a/include/asm-m32r/unistd.h
@@ -327,7 +327,7 @@
 #define __NR_epoll_pwait	319
 #define __NR_utimensat		320
 #define __NR_signalfd		321
-#define __NR_timerfd		322
+/* #define __NR_timerfd		322 removed */
 #define __NR_eventfd		323
 #define __NR_fallocate		324
 
diff -puN include/asm-sh/unistd.h~timerfd-fix-remaining-architectures include/asm-sh/unistd.h
diff -puN include/asm-sh/unistd_32.h~timerfd-fix-remaining-architectures include/asm-sh/unistd_32.h
--- a/include/asm-sh/unistd_32.h~timerfd-fix-remaining-architectures
+++ a/include/asm-sh/unistd_32.h
@@ -330,7 +330,7 @@
 #define __NR_epoll_pwait	319
 #define __NR_utimensat		320
 #define __NR_signalfd		321
-#define __NR_timerfd		322
+/* #define __NR_timerfd		322 removed */
 #define __NR_eventfd		323
 #define __NR_fallocate		324
 
diff -puN include/asm-sh/unistd_64.h~timerfd-fix-remaining-architectures include/asm-sh/unistd_64.h
--- a/include/asm-sh/unistd_64.h~timerfd-fix-remaining-architectures
+++ a/include/asm-sh/unistd_64.h
@@ -370,7 +370,7 @@
 #define __NR_epoll_pwait	347
 #define __NR_utimensat		348
 #define __NR_signalfd		349
-#define __NR_timerfd		350
+/* #define __NR_timerfd		350 removed */
 #define __NR_eventfd		351
 #define __NR_fallocate		352
 
_

Re: [PATCH] Fix timerfd breakage on avr32

[Paul Mundt]

On Tue, Feb 05, 2008 at 02:13:47AM -0800, Andrew Morton wrote:
> On Tue, 5 Feb 2008 11:03:47 +0100 Haavard Skinnemoen <hskinnemoen@atmel.com> wrote:
> 
> > On Tue, 5 Feb 2008 10:54:54 +0100
> > Haavard Skinnemoen <hskinnemoen@atmel.com> wrote:
> > 
> > > The next time you go and remove a system call, could you _please_ post
> > > a HUGE warning to linux-arch? Or just do a quick grep and fix it up.
> > 
> > Btw, a quick grep through 2.6.24-mm1 shows
> > 
> > arch/frv/kernel/entry.S:	.long sys_timerfd
> > arch/m32r/kernel/syscall_table.S:	.long sys_timerfd
> > arch/sh/kernel/syscalls_32.S:	.long sys_timerfd
> > arch/sh/kernel/syscalls_64.S:	.long sys_timerfd		/* 350 */
> > 
> > so frv, m32r and sh are probably broken too.
> > 
> 
> argh. After fixing so many architectures I asked for a grepped
> fix :(
> 
> 
> This?
> 
At least if this patch goes in first then bisect won't needlessly break,
though the previous patch has already gone to Linus, so maybe it's too
late already. Once this patch and the previous one are in, then the new
interface can be wired up -- preferably in that order.

Re: [PATCH] Fix timerfd breakage on avr32

[Russell King]

On Tue, Feb 05, 2008 at 10:54:54AM +0100, Haavard Skinnemoen wrote:
> sys_timerfd() has been removed, but avr32 still references it from its
> syscall table.
> 
> Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
> ---
> On Mon, 04 Feb 2008 22:27:28 -0800
> akpm@linux-foundation.org wrote:
> 
> > From: Davide Libenzi <davidel@xmailserver.org>
> > 
> > Wires up the new timerfd API to the x86 family.
> 
> Just one thing...
> 
> > diff -puN arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/kernel/syscall_table_32.S
> > --- a/arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
> > +++ a/arch/x86/kernel/syscall_table_32.S
> > @@ -321,6 +321,8 @@ ENTRY(sys_call_table)
> >  	.long sys_epoll_pwait
> >  	.long sys_utimensat		/* 320 */
> >  	.long sys_signalfd
> > -	.long sys_timerfd
> 
> The next time you go and remove a system call, could you _please_ post
> a HUGE warning to linux-arch? Or just do a quick grep and fix it up.

Wasn't there a decision at a kernel summit that anything which adds
new syscalls should have a test program included so that architecture
maintainers can test the functionality on their architectures?

I seem to remember that it came up because the merged timerfd was a
pile of utter crap which didn't have a hope in hells chance of working.

So... where is the new timerfd test program?

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:

Re: [PATCH] Fix timerfd breakage on avr32

[Heiko Carstens]

> Wasn't there a decision at a kernel summit that anything which adds
> new syscalls should have a test program included so that architecture
> maintainers can test the functionality on their architectures?
> 
> I seem to remember that it came up because the merged timerfd was a
> pile of utter crap which didn't have a hope in hells chance of working.
> 
> So... where is the new timerfd test program?

As described in git commit 4d672e7ac79b5ec5cdc90e450823441e20464691 it's here:
http://www.xmailserver.org/timerfd-test2.c
I used it to test the s390 backend.

Re: [PATCH] Fix timerfd breakage on avr32

[Geert Uytterhoeven]

On Wed, 6 Feb 2008, Heiko Carstens wrote:
> > Wasn't there a decision at a kernel summit that anything which adds
> > new syscalls should have a test program included so that architecture
> > maintainers can test the functionality on their architectures?
> > 
> > I seem to remember that it came up because the merged timerfd was a
> > pile of utter crap which didn't have a hope in hells chance of working.
> > 
> > So... where is the new timerfd test program?
> 
> As described in git commit 4d672e7ac79b5ec5cdc90e450823441e20464691 it's here:
> http://www.xmailserver.org/timerfd-test2.c
> I used it to test the s390 backend.

Thanks!

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [PATCH] Fix timerfd breakage on avr32

[Randy Dunlap]

On Wed, 6 Feb 2008 11:04:25 +0000 Russell King wrote:

> On Tue, Feb 05, 2008 at 10:54:54AM +0100, Haavard Skinnemoen wrote:
> > sys_timerfd() has been removed, but avr32 still references it from its
> > syscall table.
> > 
> > Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
> > ---
> > On Mon, 04 Feb 2008 22:27:28 -0800
> > akpm@linux-foundation.org wrote:
> > 
> > > From: Davide Libenzi <davidel@xmailserver.org>
> > > 
> > > Wires up the new timerfd API to the x86 family.
> > 
> > Just one thing...
> > 
> > > diff -puN arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/kernel/syscall_table_32.S
> > > --- a/arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
> > > +++ a/arch/x86/kernel/syscall_table_32.S
> > > @@ -321,6 +321,8 @@ ENTRY(sys_call_table)
> > >  	.long sys_epoll_pwait
> > >  	.long sys_utimensat		/* 320 */
> > >  	.long sys_signalfd
> > > -	.long sys_timerfd
> > 
> > The next time you go and remove a system call, could you _please_ post
> > a HUGE warning to linux-arch? Or just do a quick grep and fix it up.
> 
> Wasn't there a decision at a kernel summit that anything which adds
> new syscalls should have a test program included so that architecture
> maintainers can test the functionality on their architectures?

Thanks for sharing that useful info.  (seriously)


> I seem to remember that it came up because the merged timerfd was a
> pile of utter crap which didn't have a hope in hells chance of working.
> 
> So... where is the new timerfd test program?

---
~Randy

Re: [PATCH] Fix timerfd breakage on avr32

[Davide Libenzi]

On Wed, 6 Feb 2008, Russell King wrote:

> On Tue, Feb 05, 2008 at 10:54:54AM +0100, Haavard Skinnemoen wrote:
> > sys_timerfd() has been removed, but avr32 still references it from its
> > syscall table.
> > 
> > Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
> > ---
> > On Mon, 04 Feb 2008 22:27:28 -0800
> > akpm@linux-foundation.org wrote:
> > 
> > > From: Davide Libenzi <davidel@xmailserver.org>
> > > 
> > > Wires up the new timerfd API to the x86 family.
> > 
> > Just one thing...
> > 
> > > diff -puN arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family arch/x86/kernel/syscall_table_32.S
> > > --- a/arch/x86/kernel/syscall_table_32.S~timerfd-v3-wire-the-new-timerfd-api-to-the-x86-family
> > > +++ a/arch/x86/kernel/syscall_table_32.S
> > > @@ -321,6 +321,8 @@ ENTRY(sys_call_table)
> > >  	.long sys_epoll_pwait
> > >  	.long sys_utimensat		/* 320 */
> > >  	.long sys_signalfd
> > > -	.long sys_timerfd
> > 
> > The next time you go and remove a system call, could you _please_ post
> > a HUGE warning to linux-arch? Or just do a quick grep and fix it up.

Haavard, that was my fault not Andrew's.  I thought that the patch setting 
it CONFIG_BROKEN carried a syscall unlink from tables.  I should have 
grepped anyway.



> Wasn't there a decision at a kernel summit that anything which adds
> new syscalls should have a test program included so that architecture
> maintainers can test the functionality on their architectures?

It was always there, even before KS:

http://www.xmailserver.org/timerfd-test2.c



> I seem to remember that it came up because the merged timerfd was a
> pile of utter crap which didn't have a hope in hells chance of working.

Your pile of utter crap, was a typo that made the syscall return a 
truncated 32 bit "expired ticks" value (instead of 64 bit), and a wish 
from the manpages maintainer to have a way to retrieve the remaining time.



- Davide