* [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
2010-09-01 1:49 [PATCH] sh_net: Do not implement socketcall, use specific syscalls Carmelo AMOROSO
@ 2010-09-01 1:57 ` Carmelo AMOROSO
2010-09-02 2:53 ` Paul Mundt
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Carmelo AMOROSO @ 2010-09-01 1:57 UTC (permalink / raw)
To: linux-sh
Socket syscalls entries are now available in the syscall table,
so it does not need to implement the multipler syscall.
It is mapped into the sys_ni_syscall.
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
---
arch/sh/include/asm/unistd_32.h | 1 -
arch/sh/include/asm/unistd_64.h | 1 -
2 files changed, 0 insertions(+), 2 deletions(-)
diff --git a/arch/sh/include/asm/unistd_32.h b/arch/sh/include/asm/unistd_32.h
index 2000709..a683d2f 100644
--- a/arch/sh/include/asm/unistd_32.h
+++ b/arch/sh/include/asm/unistd_32.h
@@ -387,7 +387,6 @@
#define __ARCH_WANT_SYS_TIME
#define __ARCH_WANT_SYS_UTIME
#define __ARCH_WANT_SYS_WAITPID
-#define __ARCH_WANT_SYS_SOCKETCALL
#define __ARCH_WANT_SYS_FADVISE64
#define __ARCH_WANT_SYS_GETPGRP
#define __ARCH_WANT_SYS_LLSEEK
diff --git a/arch/sh/include/asm/unistd_64.h b/arch/sh/include/asm/unistd_64.h
index 09aa93f..2a91bb2 100644
--- a/arch/sh/include/asm/unistd_64.h
+++ b/arch/sh/include/asm/unistd_64.h
@@ -408,7 +408,6 @@
#define __ARCH_WANT_SYS_TIME
#define __ARCH_WANT_SYS_UTIME
#define __ARCH_WANT_SYS_WAITPID
-#define __ARCH_WANT_SYS_SOCKETCALL
#define __ARCH_WANT_SYS_FADVISE64
#define __ARCH_WANT_SYS_GETPGRP
#define __ARCH_WANT_SYS_LLSEEK
--
1.5.5.6
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
2010-09-01 1:49 [PATCH] sh_net: Do not implement socketcall, use specific syscalls Carmelo AMOROSO
2010-09-01 1:57 ` Carmelo AMOROSO
@ 2010-09-02 2:53 ` Paul Mundt
2010-09-02 4:27 ` [PATCH] sh_net: Do not implement socketcall, use specific Carmelo AMOROSO
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2010-09-02 2:53 UTC (permalink / raw)
To: linux-sh
On Wed, Sep 01, 2010 at 03:57:09AM +0200, Carmelo AMOROSO wrote:
> Socket syscalls entries are now available in the syscall table,
> so it does not need to implement the multipler syscall.
> It is mapped into the sys_ni_syscall.
>
I don't see how we can make this sort of change safely, it's part of the
existing ABI. Going forward we can gradually migrate people over to using
the broken out syscalls directly, but we will need to maintain
compatability with old binaries almost indefinitely.
^ permalink raw reply [flat|nested] 6+ messages in thread* RE: [PATCH] sh_net: Do not implement socketcall, use specific
2010-09-01 1:49 [PATCH] sh_net: Do not implement socketcall, use specific syscalls Carmelo AMOROSO
2010-09-01 1:57 ` Carmelo AMOROSO
2010-09-02 2:53 ` Paul Mundt
@ 2010-09-02 4:27 ` Carmelo AMOROSO
2010-09-02 6:14 ` [PATCH] sh_net: Do not implement socketcall, use specific syscalls Paul Mundt
2010-09-02 6:27 ` [PATCH] sh_net: Do not implement socketcall, use specific Carmelo AMOROSO
4 siblings, 0 replies; 6+ messages in thread
From: Carmelo AMOROSO @ 2010-09-02 4:27 UTC (permalink / raw)
To: linux-sh
-----Original Message-----
From: Paul Mundt [mailto:lethal@linux-sh.org]
Sent: Thursday, September 02, 2010 4:54 AM
To: Carmelo AMOROSO
Cc: linux-sh@vger.kernel.org
Subject: Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
On Wed, Sep 01, 2010 at 03:57:09AM +0200, Carmelo AMOROSO wrote:
> Socket syscalls entries are now available in the syscall table,
> so it does not need to implement the multipler syscall.
> It is mapped into the sys_ni_syscall.
>
I don't see how we can make this sort of change safely, it's part of the
existing ABI. Going forward we can gradually migrate people over to using
the broken out syscalls directly, but we will need to maintain
compatability with old binaries almost indefinitely.
Hi Paul,
I understand your point regarding the backward compatibility, you're right in keeping
the socketcall implemented, but I think we should add the broken-out syscalls, so an
updated version of the C library could make use of them.
Carmelo
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
2010-09-01 1:49 [PATCH] sh_net: Do not implement socketcall, use specific syscalls Carmelo AMOROSO
` (2 preceding siblings ...)
2010-09-02 4:27 ` [PATCH] sh_net: Do not implement socketcall, use specific Carmelo AMOROSO
@ 2010-09-02 6:14 ` Paul Mundt
2010-09-02 6:27 ` [PATCH] sh_net: Do not implement socketcall, use specific Carmelo AMOROSO
4 siblings, 0 replies; 6+ messages in thread
From: Paul Mundt @ 2010-09-02 6:14 UTC (permalink / raw)
To: linux-sh
On Thu, Sep 02, 2010 at 06:27:41AM +0200, Carmelo AMOROSO wrote:
>
>
> -----Original Message-----
> From: Paul Mundt [mailto:lethal@linux-sh.org]
> Sent: Thursday, September 02, 2010 4:54 AM
> To: Carmelo AMOROSO
> Cc: linux-sh@vger.kernel.org
> Subject: Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
>
> On Wed, Sep 01, 2010 at 03:57:09AM +0200, Carmelo AMOROSO wrote:
> > Socket syscalls entries are now available in the syscall table,
> > so it does not need to implement the multipler syscall.
> > It is mapped into the sys_ni_syscall.
> >
> I don't see how we can make this sort of change safely, it's part of the
> existing ABI. Going forward we can gradually migrate people over to using
> the broken out syscalls directly, but we will need to maintain
> compatability with old binaries almost indefinitely.
>
> Hi Paul,
> I understand your point regarding the backward compatibility, you're right in keeping
> the socketcall implemented, but I think we should add the broken-out syscalls, so an
> updated version of the C library could make use of them.
>
Agreed, I had thought about doing this before but I didn't have the time
to look in to what was needed on the libc side. If other people are
taking care of this, I'm of course more than happy to add them in on the
kernel side.
^ permalink raw reply [flat|nested] 6+ messages in thread* RE: [PATCH] sh_net: Do not implement socketcall, use specific
2010-09-01 1:49 [PATCH] sh_net: Do not implement socketcall, use specific syscalls Carmelo AMOROSO
` (3 preceding siblings ...)
2010-09-02 6:14 ` [PATCH] sh_net: Do not implement socketcall, use specific syscalls Paul Mundt
@ 2010-09-02 6:27 ` Carmelo AMOROSO
4 siblings, 0 replies; 6+ messages in thread
From: Carmelo AMOROSO @ 2010-09-02 6:27 UTC (permalink / raw)
To: linux-sh
-----Original Message-----
From: Paul Mundt [mailto:lethal@linux-sh.org]
Sent: Thursday, September 02, 2010 8:14 AM
To: Carmelo AMOROSO
Cc: linux-sh@vger.kernel.org
Subject: Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
On Thu, Sep 02, 2010 at 06:27:41AM +0200, Carmelo AMOROSO wrote:
>
>
> -----Original Message-----
> From: Paul Mundt [mailto:lethal@linux-sh.org]
> Sent: Thursday, September 02, 2010 4:54 AM
> To: Carmelo AMOROSO
> Cc: linux-sh@vger.kernel.org
> Subject: Re: [PATCH] sh_net: Do not implement socketcall, use specific syscalls.
>
> On Wed, Sep 01, 2010 at 03:57:09AM +0200, Carmelo AMOROSO wrote:
> > Socket syscalls entries are now available in the syscall table,
> > so it does not need to implement the multipler syscall.
> > It is mapped into the sys_ni_syscall.
> >
> I don't see how we can make this sort of change safely, it's part of the
> existing ABI. Going forward we can gradually migrate people over to using
> the broken out syscalls directly, but we will need to maintain
> compatability with old binaries almost indefinitely.
>
> Hi Paul,
> I understand your point regarding the backward compatibility, you're right in keeping
> the socketcall implemented, but I think we should add the broken-out syscalls, so an
> updated version of the C library could make use of them.
>
Agreed, I had thought about doing this before but I didn't have the time
to look in to what was needed on the libc side. If other people are
taking care of this, I'm of course more than happy to add them in on the
kernel side.
Paul,
uClibc is ready to support broken-out socket syscalls, it's just a matter to be
rebuild it against updated kernel headers.
I'm going to add cancellation handling (currently is done globally in the socketcall only)
and commit it to uClibc mainstream.
I won't care of glibc for now.
Carmelo
^ permalink raw reply [flat|nested] 6+ messages in thread