Re: [PATCH 3/4] x86_64 ia32 syscall audit fast-path

[Linus Torvalds]

On Mon, 21 Jul 2008, Roland McGrath wrote:
> 
> Here is a GIT branch with freshly-rebased patches from today's upstream.
> There was some merging fixup to be done after the x86/step changes went in.

Ok, so I decided to try this, because my new toy (Nehalem) showed some 
absolutely horrid performance with the Fedora kernel that went away when I 
compiled my own sane kernel.

And I suspect it's due to the audit path and auditd being enabled (for no 
good reason that I can tell, except to slow everything down, since nobody 
sane cares) by default.

HOWEVER. When I pull from your tree, I can no longer compile with the only 
sane default (namely "CONFIG_AUDIT is stupid, nobody should use it"), 
because I get

	arch/x86/kernel/built-in.o: In function `auditsys':
	signal_64.c:(.text+0x1f40): undefined reference to `audit_syscall_entry'
	arch/x86/kernel/built-in.o: In function `sysret_audit':
	signal_64.c:(.text+0x1f85): undefined reference to `audit_syscall_exit'
	arch/x86/ia32/built-in.o: In function `sysenter_auditsys':
	(.text+0xdd): undefined reference to `audit_syscall_entry'
	arch/x86/ia32/built-in.o: In function `sysexit_audit':
	(.text+0x128): undefined reference to `audit_syscall_exit'
	arch/x86/ia32/built-in.o: In function `cstar_auditsys':
	(.text+0x303): undefined reference to `audit_syscall_entry'
	arch/x86/ia32/built-in.o: In function `sysretl_audit':
	(.text+0x350): undefined reference to `audit_syscall_exit'
	make: *** [.tmp_vmlinux1] Error 1

so while I would like to merge this, I really can't.

I'll test the performance anyway (with CONFIG_AUDIT enabled, of course).

		Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/