All of lore.kernel.org
 help / color / mirror / Atom feed
From: Brian Gerst <brgerst@gmail.com>
To: Ingo Molnar <mingo@elte.hu>
Cc: linux-kernel@vger.kernel.org, Brian Gerst <brgerst@gmail.com>
Subject: [PATCH 2/3] x86-64: Unify x86_*_percpu() functions.
Date: Tue, 23 Dec 2008 12:15:05 -0500	[thread overview]
Message-ID: <1230052506-5041-2-git-send-email-brgerst@gmail.com> (raw)
In-Reply-To: <1230052506-5041-1-git-send-email-brgerst@gmail.com>

Merge the 32-bit and 64-bit versions of these functions.  Unlike 32-bit,
the segment base is the current cpu's PDA instead of the offset from the
original per-cpu area.  This is because GCC hardcodes the stackprotector
canary at %gs:40.  Since the assembler is incapable of relocating against
multiple symbols, the code ends up looking like:

	movq $per_cpu__var, reg
	subq $per_cpu__pda, reg
	movq %gs:(reg), reg

This is still atomic since the offset is a constant (just calculated at
runtime) and not dependant on the cpu number.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
---
 arch/x86/include/asm/percpu.h |   92 +++++++++++++++++-----------------------
 1 files changed, 39 insertions(+), 53 deletions(-)

diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 6f866fd..f704243 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -1,54 +1,9 @@
 #ifndef _ASM_X86_PERCPU_H
 #define _ASM_X86_PERCPU_H
 
-#ifdef CONFIG_X86_64
-#include <linux/compiler.h>
-
-/* Same as asm-generic/percpu.h, except that we store the per cpu offset
-   in the PDA. Longer term the PDA and every per cpu variable
-   should be just put into a single section and referenced directly
-   from %gs */
-
-#ifdef CONFIG_SMP
-#include <asm/pda.h>
-
-#define __my_cpu_offset read_pda(data_offset)
-
-#endif
-#include <asm-generic/percpu.h>
-
-DECLARE_PER_CPU(struct x8664_pda, pda);
-
-/*
- * These are supposed to be implemented as a single instruction which
- * operates on the per-cpu data base segment.  x86-64 doesn't have
- * that yet, so this is a fairly inefficient workaround for the
- * meantime.  The single instruction is atomic with respect to
- * preemption and interrupts, so we need to explicitly disable
- * interrupts here to achieve the same effect.  However, because it
- * can be used from within interrupt-disable/enable, we can't actually
- * disable interrupts; disabling preemption is enough.
- */
-#define x86_read_percpu(var)						\
-	({								\
-		typeof(per_cpu_var(var)) __tmp;				\
-		preempt_disable();					\
-		__tmp = __get_cpu_var(var);				\
-		preempt_enable();					\
-		__tmp;							\
-	})
-
-#define x86_write_percpu(var, val)					\
-	do {								\
-		preempt_disable();					\
-		__get_cpu_var(var) = (val);				\
-		preempt_enable();					\
-	} while(0)
-
-#else /* CONFIG_X86_64 */
-
 #ifdef __ASSEMBLY__
 
+#ifdef CONFIG_X86_32
 /*
  * PER_CPU finds an address of a per-cpu variable.
  *
@@ -72,6 +27,8 @@ DECLARE_PER_CPU(struct x8664_pda, pda);
 #define PER_CPU_VAR(var)	per_cpu__##var
 #endif	/* SMP */
 
+#endif	/* X86_32 */
+
 #else /* ...!ASSEMBLY */
 
 /*
@@ -88,19 +45,37 @@ DECLARE_PER_CPU(struct x8664_pda, pda);
  */
 #ifdef CONFIG_SMP
 
+#ifdef CONFIG_X86_32
+
 #define __my_cpu_offset x86_read_percpu(this_cpu_off)
 
 /* fs segment starts at (positive) offset == __per_cpu_offset[cpu] */
 #define __percpu_seg "%%fs:"
+#define __percpu_seg_off(x) (x)
+
+#else
+
+#define __my_cpu_offset read_pda(data_offset)
+
+#define __percpu_seg "%%gs:"
+#define __percpu_seg_off(x) RELOC_HIDE((x), -(unsigned long)&per_cpu__pda)
+
+#endif
 
 #else  /* !SMP */
 
 #define __percpu_seg ""
+#define __percpu_seg_off(x) (x)
 
 #endif	/* SMP */
 
 #include <asm-generic/percpu.h>
 
+#ifdef CONFIG_X86_64
+#include <asm/pda.h>
+DECLARE_PER_CPU(struct x8664_pda, pda);
+#endif
+
 /* We can use this directly for local CPU (faster). */
 DECLARE_PER_CPU(unsigned long, this_cpu_off);
 
@@ -111,6 +86,7 @@ extern void __bad_percpu_size(void);
 #define percpu_to_op(op, var, val)			\
 do {							\
 	typedef typeof(var) T__;			\
+	typeof(var) *var__ = __percpu_seg_off(&var);	\
 	if (0) {					\
 		T__ tmp__;				\
 		tmp__ = (val);				\
@@ -118,17 +94,22 @@ do {							\
 	switch (sizeof(var)) {				\
 	case 1:						\
 		asm(op "b %1,"__percpu_seg"%0"		\
-		    : "+m" (var)			\
+		    : "+m" (*var__)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	case 2:						\
 		asm(op "w %1,"__percpu_seg"%0"		\
-		    : "+m" (var)			\
+		    : "+m" (*var__)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	case 4:						\
 		asm(op "l %1,"__percpu_seg"%0"		\
-		    : "+m" (var)			\
+		    : "+m" (*var__)			\
+		    : "ri" ((T__)val));			\
+		break;					\
+	case 8:						\
+		asm(op "q %1,"__percpu_seg"%0"		\
+		    : "+m" (*var__)			\
 		    : "ri" ((T__)val));			\
 		break;					\
 	default: __bad_percpu_size();			\
@@ -138,21 +119,27 @@ do {							\
 #define percpu_from_op(op, var)				\
 ({							\
 	typeof(var) ret__;				\
+	typeof(var) *var__ = __percpu_seg_off(&var);	\
 	switch (sizeof(var)) {				\
 	case 1:						\
 		asm(op "b "__percpu_seg"%1,%0"		\
 		    : "=r" (ret__)			\
-		    : "m" (var));			\
+		    : "m" (*var__));			\
 		break;					\
 	case 2:						\
 		asm(op "w "__percpu_seg"%1,%0"		\
 		    : "=r" (ret__)			\
-		    : "m" (var));			\
+		    : "m" (*var__));			\
 		break;					\
 	case 4:						\
 		asm(op "l "__percpu_seg"%1,%0"		\
 		    : "=r" (ret__)			\
-		    : "m" (var));			\
+		    : "m" (*var__));			\
+		break;					\
+	case 8:						\
+		asm(op "q "__percpu_seg"%1,%0"		\
+		    : "=r" (ret__)			\
+		    : "m" (*var__));			\
 		break;					\
 	default: __bad_percpu_size();			\
 	}						\
@@ -165,7 +152,6 @@ do {							\
 #define x86_sub_percpu(var, val) percpu_to_op("sub", per_cpu__##var, val)
 #define x86_or_percpu(var, val) percpu_to_op("or", per_cpu__##var, val)
 #endif /* !__ASSEMBLY__ */
-#endif /* !CONFIG_X86_64 */
 
 #ifdef CONFIG_SMP
 
-- 
1.6.1.rc1


  reply	other threads:[~2008-12-23 17:15 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-12-23 17:15 [PATCH 1/3] x86-64: Convert the PDA to percpu Brian Gerst
2008-12-23 17:15 ` Brian Gerst [this message]
2008-12-23 17:15   ` [PATCH 3/3] x86-64: Move cpu number from PDA to per-cpu and consolidate with 32-bit Brian Gerst
2008-12-27 11:03   ` [PATCH 2/3] x86-64: Unify x86_*_percpu() functions Jeremy Fitzhardinge
2008-12-27 11:09     ` Ingo Molnar
2008-12-27 11:20       ` Jeremy Fitzhardinge
2008-12-27 10:41 ` [PATCH 1/3] x86-64: Convert the PDA to percpu Ingo Molnar
2008-12-27 15:30   ` Brian Gerst
2008-12-27 15:53     ` Ingo Molnar
2008-12-27 17:16       ` Brian Gerst

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1230052506-5041-2-git-send-email-brgerst@gmail.com \
    --to=brgerst@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.