Re: [RFC PATCH v2 2/3] thread_local_abi: wire up x86 32/64 system call

----- On Dec 23, 2015, at 10:42 PM, Josh Triplett josh-iaAMLnmF4UmaiuxdJuQwMA@public.gmane.org wrote:

On December 22, 2015 10:02:12 AM PST, Mathieu Desnoyers
[off-list ref] wrote:

quoted

Wire up the thread local ABI on x86 32/64. Call the
thread_local_abi_handle_notify_resume() function on return to
userspace if TIF_NOTIFY_RESUME thread flag is set.

This provides an ABI improving the speed of a getcpu operation
on x86 by removing the need to perform a function call, "lsl"
instruction, or system call on the fast path.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>
CC: Russell King <linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org>
CC: Catalin Marinas <redacted>
CC: Will Deacon <redacted>
CC: Thomas Gleixner <redacted>
CC: Paul Turner <redacted>
CC: Andrew Hunter <redacted>
CC: Peter Zijlstra <redacted>
CC: Andy Lutomirski <redacted>
CC: Andi Kleen <andi-Vw/NltI1exuRpAAqCnN02g@public.gmane.org>
CC: Dave Watson <redacted>
CC: Chris Lameter <redacted>
CC: Ingo Molnar <redacted>
CC: Ben Maurer <redacted>
CC: Steven Rostedt <redacted>
CC: "Paul E. McKenney" <redacted>
CC: Josh Triplett <redacted>
CC: Linus Torvalds <torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
CC: Thomas Gleixner <redacted>
CC: linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
---
arch/x86/entry/common.c                | 2 ++
arch/x86/entry/syscalls/syscall_32.tbl | 1 +
arch/x86/entry/syscalls/syscall_64.tbl | 1 +
3 files changed, 4 insertions(+)

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index a89fdbc..222cacf 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c

@@ -249,6 +249,8 @@ static void exit_to_usermode_loop(struct pt_regs

*regs, u32 cached_flags)
		if (cached_flags & _TIF_NOTIFY_RESUME) {
			clear_thread_flag(TIF_NOTIFY_RESUME);
			tracehook_notify_resume(regs);
+			if (thread_local_abi_active(current))
+				thread_local_abi_handle_notify_resume(current);

Every caller seems likely to duplicate this pattern; why not make the call
itself a static inline containing this check and call (or no-op if compiled
out)?

Very good point, I'll do that.

Thanks!

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com