Re: [PATCH 00/12] x86: Rewrite 64-bit syscall code

[Andy Lutomirski]

On Mon, Dec 7, 2015 at 1:51 PM, Andy Lutomirski <luto@xxxxxxxxxx> wrote:
> This is kind of like the 32-bit and compat code, except that I
> preserved the fast path this time.  I was unable to measure any
> significant performance change on my laptop in the fast path.
>
> What do you all think?

For completeness, if I zap the fast path entirely (see attached), I
lose 20 cycles (148 cycles vs 128 cycles) on Skylake.  Switching
between movq and pushq for stack setup makes no difference whatsoever,
interestingly.  I haven't tried to figure out exactly where those 20
cycles go.

--Andy
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index 1ab5362f241d..a97981f0d9ce 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -163,71 +163,17 @@ GLOBAL(entry_SYSCALL_64_after_swapgs)
 	pushq	%r9				/* pt_regs->r9 */
 	pushq	%r10				/* pt_regs->r10 */
 	pushq	%r11				/* pt_regs->r11 */
-	sub	$(6*8), %rsp			/* pt_regs->bp, bx, r12-15 not saved */
+	pushq	%rbx				/* pt_regs->rbx */
+	pushq	%rbp				/* pt_regs->rbp */
+	pushq	%r12				/* pt_regs->r12 */
+	pushq	%r13				/* pt_regs->r13 */
+	pushq	%r14				/* pt_regs->r14 */
+	pushq	%r15				/* pt_regs->r15 */
 
-	/*
-	 * If we need to do entry work or if we guess we'll need to do
-	 * exit work, go straight to the slow path.
-	 */
-	testl	$_TIF_WORK_SYSCALL_ENTRY|_TIF_ALLWORK_MASK, ASM_THREAD_INFO(TI_flags, %rsp, SIZEOF_PTREGS)
-	jnz	entry_SYSCALL64_slow_path
-
-entry_SYSCALL_64_fastpath:
-	/*
-	 * Easy case: enable interrupts and issue the syscall.  If the syscall
-	 * needs pt_regs, we'll call a stub that disables interrupts again
-	 * and jumps to the slow path.
-	 */
-	TRACE_IRQS_ON
-	ENABLE_INTERRUPTS(CLBR_NONE)
-#if __SYSCALL_MASK == ~0
-	cmpq	$__NR_syscall_max, %rax
-#else
-	andl	$__SYSCALL_MASK, %eax
-	cmpl	$__NR_syscall_max, %eax
-#endif
-	ja	1f				/* return -ENOSYS (already in pt_regs->ax) */
-	movq	%r10, %rcx
-	call	*sys_call_table_fastpath_64(, %rax, 8)
-	movq	%rax, RAX(%rsp)
-1:
-
-	/*
-	 * If we get here, then we know that pt_regs is clean for SYSRET64.
-	 * If we see that no exit work is required (which we are required
-	 * to check with IRQs off), then we can go straight to SYSRET64.
-	 */
-	DISABLE_INTERRUPTS(CLBR_NONE)
-	TRACE_IRQS_OFF
-	testl	$_TIF_ALLWORK_MASK, ASM_THREAD_INFO(TI_flags, %rsp, SIZEOF_PTREGS)
-	jnz	1f
-
-	LOCKDEP_SYS_EXIT
-	TRACE_IRQS_ON		/* user mode is traced as IRQs on */
-	RESTORE_C_REGS
-	movq	RSP(%rsp), %rsp
-	USERGS_SYSRET64
-
-1:
-	/*
-	 * The fast path looked good when we started, but something changed
-	 * along the way and we need to switch to the slow path.  Calling
-	 * raise(3) will trigger this, for example.  IRQs are off.
-	 */
-	TRACE_IRQS_ON
-	ENABLE_INTERRUPTS(CLBR_NONE)
-	SAVE_EXTRA_REGS
-	movq	%rsp, %rdi
-	call	syscall_return_slowpath	/* returns with IRQs disabled */
-	jmp	return_from_SYSCALL_64
-
-entry_SYSCALL64_slow_path:
 	/* IRQs are off. */
-	SAVE_EXTRA_REGS
 	movq	%rsp, %rdi
 	call	do_syscall_64		/* returns with IRQs disabled */
 
-return_from_SYSCALL_64:
 	RESTORE_EXTRA_REGS
 	TRACE_IRQS_IRETQ		/* we're about to change IF */
 
@@ -305,14 +251,7 @@ opportunistic_sysret_failed:
 END(entry_SYSCALL_64)
 
 ENTRY(stub_ptregs_64)
-	/*
-	 * Syscalls marked as needing ptregs that go through the fast path
-	 * land here.  We transfer to the slow path.
-	 */
-	DISABLE_INTERRUPTS(CLBR_NONE)
-	TRACE_IRQS_OFF
-	addq	$8, %rsp
-	jmp	entry_SYSCALL64_slow_path
+	ud2
 END(stub_ptregs_64)
 
 /*