Re: [PATCH] remove timerd() syscall number

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 17 Sep 2007 18:58:58 +0200
Michael Kerrisk <mtk-manpages@gmx.net> wrote:

> I hope I've done this right...  This patch, against 2.6.23-rc6,
> removes the timerfd() syscall (which in any case had a bug
> on its 2.6.22 release) from all architectures, so that we
> can have some breathing space to think about the API design.
> All of the existing timerfd() code is left intact.

I think a minimally-intrusive implementation would be

--- a/fs/timerfd.c~a
+++ a/fs/timerfd.c
@@ -159,6 +159,8 @@ asmlinkage long sys_timerfd(int ufd, int
 	struct inode *inode;
 	struct itimerspec ktmr;
 
+	return -ENOSYS;
+
 	if (copy_from_user(&ktmr, utmr, sizeof(ktmr)))
 		return -EFAULT;
 
_

or, better,

--- a/init/Kconfig~a
+++ a/init/Kconfig
@@ -491,6 +491,7 @@ config SIGNALFD
 
 config TIMERFD
 	bool "Enable timerfd() system call" if EMBEDDED
+	depends on BROKEN
 	select ANON_INODES
 	default y
 	help
_