Skip to contentSkip to conversation

OMAP clock and PM architecture fixes for 2.6.32-rc6

PATCH New Paul Walmsley 4 participants 12 messages Nov 16, 2009 – Nov 18, 2009

Subsystems (5)
lore
Share & tools
lore

Paul Walmsley

author maintainer
#0 lore raw

cover OMAP clock and PM architecture fixes for 2.6.32-rc6

Hi,

this series contains fixes for some clock and PM architecture components for the 2.6.32-rc series.

- Paul

---

Paul Walmsley (1):
OMAP clock/hwmod: fix off-by-one errors

Rajendra Nayak (1):
OMAP3 clock: Fix the DPLL freqsel computations

Tero Kristo (1):
OMAP3: Clock: Fixed dpll3_m2x2 rate calculation

 arch/arm/mach-omap2/clock34xx.c  |    2 +-
 arch/arm/mach-omap2/clock34xx.h  |    4 ++--
 arch/arm/mach-omap2/cm.c         |    2 +-
 arch/arm/mach-omap2/omap_hwmod.c |   12 +++++-------
 arch/arm/mach-omap2/prcm.c       |    4 ++--
 5 files changed, 11 insertions(+), 13 deletions(-)
author maintainer
#1 lore raw

PATCH 1/3 OMAP3 clock: Fix the DPLL freqsel computations

From: Rajendra Nayak <[email protected]>

Fix the freqsel value computation. Use n instead of (n+1)

The formula in the TRM uses a zero-based N, hence the (n+1); however at this point in the clock34xx.c code, N is one-based.

Hayati Bayrakdar <[email protected]> and Nishanth Menon <[email protected]> helped track down this bug.

[paul at pwsan.com: modified commit message]
Cc Hayati Bayrakdar <[email protected]>
Cc Nishanth Menon <[email protected]>
---
 arch/arm/mach-omap2/clock34xx.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/arm/mach-omap2/clock34xx.c +1 -1
index c258f87..225c1a7 100644
@@ -473,7 +473,7 @@ static u16 _omap3_dpll_compute_freqsel(struct clk *clk, u8 n)
 	unsigned long fint;
 	u16 f = 0;
 
-	fint = clk->dpll_data->clk_ref->rate / (n + 1);
+	fint = clk->dpll_data->clk_ref->rate / n;
 
 	pr_debug("clock: fint is %lu\n", fint);
 
maintainer
#2 lore raw

PATCH 1/3 OMAP3 clock: Fix the DPLL freqsel computations

This patch has been applied to the linux-omap
by youw fwiendly patch wobot.

Branch in linux-omap: omap-fixes

Initial commit ID (Likely to change): 8c3b128f6b50f9b8a6b015afeead77fd6de366a6

PatchWorks
http://patchwork.kernel.org/patch/60367/

Git (Likely to change, and takes a while to get mirrored)
http://git.kernel.org/?p=linux/kernel/git/tmlind/linux-omap-2.6.git;a=commit;h=8c3b128f6b50f9b8a6b015afeead77fd6de366a6

author maintainer
#3 lore raw

PATCH 2/3 OMAP3: Clock: Fixed dpll3_m2x2 rate calculation

In reply to Paul Walmsley

From: Tero Kristo <[email protected]>

Current calculation does not take into account any changes to M2 divisor, and thus when we change VDD2 OPP, dpll3_m2x2 rate does not change. Fixed by re-routing dpll3_m2x2 parent to dpll3_m2.

---
 arch/arm/mach-omap2/clock34xx.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/arm/mach-omap2/clock34xx.h +2 -2
index a1b3de7..8fe1bcb 100644
@@ -489,9 +489,9 @@ static struct clk core_ck = {
 static struct clk dpll3_m2x2_ck = {
 	.name		= "dpll3_m2x2_ck",
 	.ops		= &clkops_null,
-	.parent		= &dpll3_x2_ck,
+	.parent		= &dpll3_m2_ck,
 	.clkdm_name	= "dpll3_clkdm",
-	.recalc		= &followparent_recalc,
+	.recalc		= &omap3_clkoutx2_recalc,
 };
 
 /* The PWRDN bit is apparently only available on 3430ES2 and above */
maintainer
#4 lore raw

PATCH 2/3 OMAP3: Clock: Fixed dpll3_m2x2 rate calculation

This patch has been applied to the linux-omap
by youw fwiendly patch wobot.

Branch in linux-omap: omap-fixes

Initial commit ID (Likely to change): 2dcbc9ebbe46c92e748e6180133d10cce69acb4d

PatchWorks
http://patchwork.kernel.org/patch/60368/

Git (Likely to change, and takes a while to get mirrored)
http://git.kernel.org/?p=linux/kernel/git/tmlind/linux-omap-2.6.git;a=commit;h=2dcbc9ebbe46c92e748e6180133d10cce69acb4d

author maintainer
#5 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

In reply to Paul Walmsley

Fix loop bailout off-by-one bugs reported by Juha Lepp?nen <[email protected]>.

---
 arch/arm/mach-omap2/cm.c         |    2 +-
 arch/arm/mach-omap2/omap_hwmod.c |   12 +++++-------
 arch/arm/mach-omap2/prcm.c       |    4 ++--
 3 files changed, 8 insertions(+), 10 deletions(-)
arch/arm/mach-omap2/cm.c +1 -1
index 8eb2dab..0766e52 100644
@@ -65,6 +65,6 @@ int omap2_cm_wait_module_ready(s16 prcm_mod, u8 idlest_id, u8 idlest_shift)
 	       (i++ < MAX_MODULE_READY_TIME))
 		udelay(1);
 
-	return (i < MAX_MODULE_READY_TIME) ? 0 : -EBUSY;
+	return (i <= MAX_MODULE_READY_TIME) ? 0 : -EBUSY;
 }
 
arch/arm/mach-omap2/omap_hwmod.c +5 -7
index 633b216..a4a9518 100644
@@ -759,14 +759,12 @@ static int _reset(struct omap_hwmod *oh)
 	_write_sysconfig(v, oh);
 
 	c = 0;
-	while (c < MAX_MODULE_RESET_WAIT &&
-	       !(omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
-		 SYSS_RESETDONE_MASK)) {
+	while (!(omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
+		 SYSS_RESETDONE_MASK) &&
+	       (c++ < MAX_MODULE_RESET_WAIT))
 		udelay(1);
-		c++;
-	}
 
-	if (c == MAX_MODULE_RESET_WAIT)
+	if (c > MAX_MODULE_RESET_WAIT)
 		WARN(1, "omap_hwmod: %s: failed to reset in %d usec\n",
 		     oh->name, MAX_MODULE_RESET_WAIT);
 	else
@@ -777,7 +775,7 @@ static int _reset(struct omap_hwmod *oh)
 	 * _wait_target_ready() or _reset()
 	 */
 
-	return (c == MAX_MODULE_RESET_WAIT) ? -ETIMEDOUT : 0;
+	return (c > MAX_MODULE_RESET_WAIT) ? -ETIMEDOUT : 0;
 }
 
 /**
arch/arm/mach-omap2/prcm.c +2 -2
index 029d376..d486709 100644
@@ -251,14 +251,14 @@ int omap2_cm_wait_idlest(void __iomem *reg, u32 mask, const char *name)
 	       (i++ < MAX_MODULE_ENABLE_WAIT))
 		udelay(1);
 
-	if (i < MAX_MODULE_ENABLE_WAIT)
+	if (i <= MAX_MODULE_ENABLE_WAIT)
 		pr_debug("cm: Module associated with clock %s ready after %d "
 			 "loops\n", name, i);
 	else
 		pr_err("cm: Module associated with clock %s didn't enable in "
 		       "%d tries\n", name, MAX_MODULE_ENABLE_WAIT);
 
-	return (i < MAX_MODULE_ENABLE_WAIT) ? 1 : 0;
+	return (i <= MAX_MODULE_ENABLE_WAIT) ? 1 : 0;
 };
 
 void __init omap2_set_globals_prcm(struct omap_globals *omap2_globals)
#6 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Quoted 2 lines · Paul Walmsley

Fix loop bailout off-by-one bugs reported by Juha Lepp?nen <[email protected]>.

I'm not sure the new code is any easier to read than the old code. And what with some checks post-loop being <= and others being >, it's a recipe for cut'n'paste errors happening.

Quoted 13 lines
arch/arm/mach-omap2/omap_hwmod.c +2 -3
index 633b216..a4a9518 100644
@@ -759,14 +759,12 @@ static int _reset(struct omap_hwmod *oh)
 	_write_sysconfig(v, oh);
 
 	c = 0;
-	while (c < MAX_MODULE_RESET_WAIT &&
-	       !(omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
-		 SYSS_RESETDONE_MASK)) {
+	while (!(omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
+		 SYSS_RESETDONE_MASK) &&

arch/arm/mach-omap2/omap_hwmod.c static int _reset(struct omap_hwmod *oh) · line 770

Quoted 10 lines
arch/arm/mach-omap2/omap_hwmod.c +2 -3
+	       (c++ < MAX_MODULE_RESET_WAIT))
 		udelay(1);
-		c++;
-	}
 
-	if (c == MAX_MODULE_RESET_WAIT)
+	if (c > MAX_MODULE_RESET_WAIT)
 		WARN(1, "omap_hwmod: %s: failed to reset in %d usec\n",
 		     oh->name, MAX_MODULE_RESET_WAIT);
 	else

How about:

	for (c = 0; c <= MAX_MODULE_RESET_WAIT; c++) {
		if (omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
		    SYSS_RESETDONE_MASK) {
			pr_debug("omap_hwmod: %s: reset in %d usec\n",
				oh->name, c);
			return 0;
		}
	}

	WARN(1, "omap_hwmod: %s: failed to reset in %d usec\n",
                     oh->name, c - 1);
	return -ETIMEDOUT;

?

Even better would be to turn this into a helper much like wait_event(), which would stop silly mistakes happening.

author maintainer
#7 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Hi Russell,

Quoted 7 lines · Russell King - ARM Linux
Quoted 2 lines · Paul Walmsley

Fix loop bailout off-by-one bugs reported by Juha Leppänen <[email protected]>.

I'm not sure the new code is any easier to read than the old code. And what with some checks post-loop being <= and others being >, it's a recipe for cut'n'paste errors happening.

...

Quoted 15 lines

How about:

	for (c = 0; c <= MAX_MODULE_RESET_WAIT; c++) {
		if (omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
		    SYSS_RESETDONE_MASK) {
			pr_debug("omap_hwmod: %s: reset in %d usec\n",
				oh->name, c);
			return 0;
		}
	}

	WARN(1, "omap_hwmod: %s: failed to reset in %d usec\n",
                     oh->name, c - 1);
	return -ETIMEDOUT;
Quoted 4 lines

?

Even better would be to turn this into a helper much like wait_event(), which would stop silly mistakes happening.

Agreed. How about this?

arch/arm/plat-omap/include/plat/common.h probably isn't the right place for it, since it's not OMAP-specific...

- Paul

From 972b3bbf2c210086b862aedb35587484bcefe998 Mon Sep 17 00:00:00 2001
From: Paul Walmsley <[email protected]>
Date: Mon, 16 Nov 2009 06:13:15 -0700
Subject: [PATCH] OMAP clock/hwmod: fix off-by-one errors
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Fix loop bailout off-by-one bugs reported by Juha Leppänen <[email protected]>.

This second version incorporates comments from Russell King <[email protected]>. A new macro, 'omap_test_timeout', has been created, with cleaner code, and existing code has been converted to use it.

Cc Juha Leppänen <[email protected]>
Cc Russell King <[email protected]>
---
 arch/arm/mach-omap2/cm.c                 |    7 ++++---
 arch/arm/mach-omap2/omap_hwmod.c         |   13 +++++--------
 arch/arm/mach-omap2/prcm.c               |    5 ++---
 arch/arm/plat-omap/include/plat/common.h |   20 ++++++++++++++++++++
 4 files changed, 31 insertions(+), 14 deletions(-)
arch/arm/mach-omap2/cm.c +4 -3
index 8eb2dab..58e4a1c 100644
@@ -21,6 +21,8 @@
 
 #include <asm/atomic.h>
 
+#include <plat/common.h>
+
 #include "cm.h"
 #include "cm-regbits-24xx.h"
 #include "cm-regbits-34xx.h"
@@ -61,9 +63,8 @@ int omap2_cm_wait_module_ready(s16 prcm_mod, u8 idlest_id, u8 idlest_shift)
 	mask = 1 << idlest_shift;
 
 	/* XXX should be OMAP2 CM */
-	while (((cm_read_mod_reg(prcm_mod, cm_idlest_reg) & mask) != ena) &&
-	       (i++ < MAX_MODULE_READY_TIME))
-		udelay(1);
+	omap_test_timeout(((cm_read_mod_reg(prcm_mod, cm_idlest_reg) & mask) == ena),
+			  MAX_MODULE_READY_TIME, i);
 
 	return (i < MAX_MODULE_READY_TIME) ? 0 : -EBUSY;
 }
arch/arm/mach-omap2/omap_hwmod.c +5 -8
index 633b216..7aaf5f1 100644
@@ -45,6 +45,7 @@
 #include <linux/mutex.h>
 #include <linux/bootmem.h>
 
+#include <plat/common.h>
 #include <plat/cpu.h>
 #include <plat/clockdomain.h>
 #include <plat/powerdomain.h>
@@ -736,7 +737,7 @@ static int _wait_target_ready(struct omap_hwmod *oh)
 static int _reset(struct omap_hwmod *oh)
 {
 	u32 r, v;
-	int c;
+	int c = 0;
 
 	if (!oh->sysconfig ||
 	    !(oh->sysconfig->sysc_flags & SYSC_HAS_SOFTRESET) ||
@@ -758,13 +759,9 @@ static int _reset(struct omap_hwmod *oh)
 		return r;
 	_write_sysconfig(v, oh);
 
-	c = 0;
-	while (c < MAX_MODULE_RESET_WAIT &&
-	       !(omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
-		 SYSS_RESETDONE_MASK)) {
-		udelay(1);
-		c++;
-	}
+	omap_test_timeout((omap_hwmod_readl(oh, oh->sysconfig->syss_offs) &
+			   SYSS_RESETDONE_MASK),
+			  MAX_MODULE_RESET_WAIT, c);
 
 	if (c == MAX_MODULE_RESET_WAIT)
 		WARN(1, "omap_hwmod: %s: failed to reset in %d usec\n",
arch/arm/mach-omap2/prcm.c +2 -3
index 029d376..7acd03d 100644
@@ -247,9 +247,8 @@ int omap2_cm_wait_idlest(void __iomem *reg, u32 mask, const char *name)
 		BUG();
 
 	/* Wait for lock */
-	while (((__raw_readl(reg) & mask) != ena) &&
-	       (i++ < MAX_MODULE_ENABLE_WAIT))
-		udelay(1);
+	omap_test_timeout(((__raw_readl(reg) & mask) == ena),
+			  MAX_MODULE_ENABLE_WAIT, i);
 
 	if (i < MAX_MODULE_ENABLE_WAIT)
 		pr_debug("cm: Module associated with clock %s ready after %d "
arch/arm/plat-omap/include/plat/common.h +20 -0
index 064f173..e977967 100644
@@ -71,4 +71,24 @@ void omap2_set_globals_sdrc(struct omap_globals *);
 void omap2_set_globals_control(struct omap_globals *);
 void omap2_set_globals_prcm(struct omap_globals *);
 
+/**
+ * omap_test_timeout - busy-loop, testing a condition
+ * @cond: condition to test until it evaluates to true
+ * @timeout: maximum number of microseconds in the timeout
+ * @index: loop index (integer)
+ *
+ * Loop waiting for @cond to become true or until at least @timeout
+ * microseconds have passed.  To use, define some integer @index in the
+ * calling code.  After running, if @index == @timeout, then the loop has
+ * timed out.
+ */
+#define omap_test_timeout(cond, timeout, index)			\
+({								\
+	for (index = 0; index < timeout; index++) {		\
+		if (cond)					\
+			break;					\
+		udelay(1);					\
+	}							\
+})
+
 #endif /* __ARCH_ARM_MACH_OMAP_COMMON_H */
Signature · 1 line
1.6.5.GIT
#8 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Paul

Quoted 1 line

-----Original Message-----

From: [email protected] [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of Paul

Quoted 5 lines

Walmsley
Sent: Tuesday, November 17, 2009 11:39 AM
To: Russell King - ARM Linux
Cc: linux-arm-kernel at lists.infradead.org; Juha Lepp?nen; linux-omap at vger.kernel.org
Subject: Re: [PATCH 3/3] OMAP clock/hwmod: fix off-by-one errors

<snip>

Quoted 11 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+/**
+ * omap_test_timeout - busy-loop, testing a condition
+ * @cond: condition to test until it evaluates to true
+ * @timeout: maximum number of microseconds in the timeout
+ * @index: loop index (integer)
+ *
+ * Loop waiting for @cond to become true or until at least @timeout
+ * microseconds have passed.  To use, define some integer @index in the
+ * calling code.  After running, if @index == @timeout, then the loop has
+ * timed out.

arch/arm/plat-omap/include/plat/common.h void omap2_set_globals_sdrc(struct omap_globals *); · line 93

Quoted 10 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+ */
+#define omap_test_timeout(cond, timeout, index)			\
+({								\
+	for (index = 0; index < timeout; index++) {		\
+		if (cond)					\
+			break;					\
+		udelay(1);					\
+	}							\
+})
+

There is a similar kind of function implemented in USB host, that returns -ETIMEDOUT Which makes more sense for failure case.

Maybe such a generic function for all omaps is needed.

Refer:
drivers/usb/host/ehci-hcd.c
static int handshake (struct ehci_hcd *ehci, void __iomem *ptr,
u32 mask, u32 done, int usec)

{
        u32     result;

        do {
                result = ehci_readl(ehci, ptr);
                if (result == ~(u32)0)          /* card removed */
                        return -ENODEV;
                result &= mask;
                if (result == done)
                        return 0;
                udelay (1);
                usec--;
        } while (usec > 0);
        return -ETIMEDOUT;
}
Quoted 3 lines

#endif /* __ARCH_ARM_MACH_OMAP_COMMON_H */
--
1.6.5.GIT

maintainer
#9 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Quoted 41 lines · Pandita, Vikram

Paul

Quoted 7 lines

-----Original Message-----
From: linux-omap-owner at vger.kernel.org [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of Paul Walmsley
Sent: Tuesday, November 17, 2009 11:39 AM
To: Russell King - ARM Linux
Cc: linux-arm-kernel at lists.infradead.org; Juha Lepp?nen; linux-omap at vger.kernel.org
Subject: Re: [PATCH 3/3] OMAP clock/hwmod: fix off-by-one errors

<snip>

Quoted 11 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+/**
+ * omap_test_timeout - busy-loop, testing a condition
+ * @cond: condition to test until it evaluates to true
+ * @timeout: maximum number of microseconds in the timeout
+ * @index: loop index (integer)
+ *
+ * Loop waiting for @cond to become true or until at least @timeout
+ * microseconds have passed.  To use, define some integer @index in the
+ * calling code.  After running, if @index == @timeout, then the loop has
+ * timed out.

arch/arm/plat-omap/include/plat/common.h void omap2_set_globals_sdrc(struct omap_globals *); · line 93

Quoted 10 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+ */
+#define omap_test_timeout(cond, timeout, index)			\
+({								\
+	for (index = 0; index < timeout; index++) {		\
+		if (cond)					\
+			break;					\
+		udelay(1);					\
+	}							\
+})
+

There is a similar kind of function implemented in USB host, that returns -ETIMEDOUT Which makes more sense for failure case.

Maybe such a generic function for all omaps is needed.

Refer:
drivers/usb/host/ehci-hcd.c
static int handshake (struct ehci_hcd *ehci, void __iomem *ptr,
u32 mask, u32 done, int usec)

Quoted 18 lines
{
        u32     result;

        do {
                result = ehci_readl(ehci, ptr);
                if (result == ~(u32)0)          /* card removed */
                        return -ENODEV;
                result &= mask;
                if (result == done)
                        return 0;
                udelay (1);
                usec--;
        } while (usec > 0);
        return -ETIMEDOUT;
}
Quoted 1 line

#endif /* __ARCH_ARM_MACH_OMAP_COMMON_H */

Well we need to stick to minimal changes if we plan to send this in as a fix. Anything more complex we should do the next merge window.

Tony

maintainer
#10 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Quoted 59 lines · Tony Lindgren
Quoted 58 lines · Pandita, Vikram

Paul

Quoted 7 lines

-----Original Message-----
From: linux-omap-owner at vger.kernel.org [mailto:linux-omap-owner at vger.kernel.org] On Behalf Of Paul Walmsley
Sent: Tuesday, November 17, 2009 11:39 AM
To: Russell King - ARM Linux
Cc: linux-arm-kernel at lists.infradead.org; Juha Lepp?nen; linux-omap at vger.kernel.org
Subject: Re: [PATCH 3/3] OMAP clock/hwmod: fix off-by-one errors

<snip>

Quoted 11 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+/**
+ * omap_test_timeout - busy-loop, testing a condition
+ * @cond: condition to test until it evaluates to true
+ * @timeout: maximum number of microseconds in the timeout
+ * @index: loop index (integer)
+ *
+ * Loop waiting for @cond to become true or until at least @timeout
+ * microseconds have passed.  To use, define some integer @index in the
+ * calling code.  After running, if @index == @timeout, then the loop has
+ * timed out.

arch/arm/plat-omap/include/plat/common.h void omap2_set_globals_sdrc(struct omap_globals *); · line 93

Quoted 10 lines
arch/arm/plat-omap/include/plat/common.h +10 -0
+ */
+#define omap_test_timeout(cond, timeout, index)			\
+({								\
+	for (index = 0; index < timeout; index++) {		\
+		if (cond)					\
+			break;					\
+		udelay(1);					\
+	}							\
+})
+

There is a similar kind of function implemented in USB host, that returns -ETIMEDOUT Which makes more sense for failure case.

Maybe such a generic function for all omaps is needed.

Refer:
drivers/usb/host/ehci-hcd.c
static int handshake (struct ehci_hcd *ehci, void __iomem *ptr,
u32 mask, u32 done, int usec)

{
        u32     result;

        do {
                result = ehci_readl(ehci, ptr);
                if (result == ~(u32)0)          /* card removed */
                        return -ENODEV;
                result &= mask;
                if (result == done)
                        return 0;
                udelay (1);
                usec--;
        } while (usec > 0);
        return -ETIMEDOUT;
}
Quoted 4 lines
Quoted 1 line
Quoted 1 line

#endif /* __ARCH_ARM_MACH_OMAP_COMMON_H */

Well we need to stick to minimal changes if we plan to send this in as a fix. Anything more complex we should do the next merge window.

To me it sounds like this can wait until the merge window, so not adding to omap-fixes for now.

Regards,

Tony

author maintainer
#11 lore raw

PATCH 3/3 OMAP clock/hwmod: fix off-by-one errors

Quoted 2 lines · Tony Lindgren

To me it sounds like this can wait until the merge window, so not adding to omap-fixes for now.

Indeed, there's no rush on this one, so let's set it up for 2.6.33. I'll keep that macro in plat/system.h unless you or Russell have a better suggestion. It might be worth looking at the rest of the Linux codebase to see if there are other potential users; maybe someone might be willing to pick up a "kernel-janitor" task for this?

Vikram, I agree with your comment. Maybe someone can send a patch to clean up ehci-hcd.c and any other users of the same idiom to use this macro for 2.6.33? Looks like also plat-omap/mailbox.c:__mbox_msg_send() and mach-omap2/powerdomain.c:pwrdm_wait_transition() would be good candidates for this, under arch/arm/*omap*.

- Paul