Skip to contentSkip to conversation

introduce macro spin_event_timeout()

PATCH Superseded Timur Tabi 4 participants 7 messages Apr 27, 2009 – May 1, 2009 PowerPC

v6 of 9 Superseded Latest is v9 →

lore
Share & tools
lore

Timur Tabi

author
#0 lore raw

PATCH v6 1/1 introduce macro spin_event_timeout()

The macro spin_event_timeout() takes a condition and timeout value (in microseconds) as parameters. It spins until either the condition is true or the timeout expires. It returns the result of the condition when the loop was terminated.

This primary purpose of this macro is to poll on a hardware register until a status bit changes. The timeout ensures that the loop still terminates if the bit doesn't change as expected. This macro makes it easier for driver developers to perform this kind of operation properly.

---

I'm making this a PowerPC-specific patch because I want to use tb_ticks_per_usec, which does not exist on all other platforms. I don't want to use jiffies because jiffies works only when interrupts are enabled, and the resolution may not be fine enough.

 arch/powerpc/include/asm/delay.h |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)
arch/powerpc/include/asm/delay.h +27 -0
index f9200a6..1939e0f 100644
@@ -2,6 +2,8 @@
 #define _ASM_POWERPC_DELAY_H
 #ifdef __KERNEL__
 
+#include <asm/time.h>
+
 /*
  * Copyright 1996, Paul Mackerras.
  *
@@ -30,5 +32,30 @@ extern void udelay(unsigned long usecs);
 #define mdelay(n)	udelay((n) * 1000)
 #endif
 
+/**
+ * spin_event_timeout - spin until a condition gets true or a timeout elapses
+ * @condition: a C expression to evalate
+ * @timeout: timeout, in microseconds
+ * @delay: the number of microseconds to delay between eache evaluation of
+ *         @condition
+ * @rc: the last value of the condition
+ *
+ * The process spins until the condition evaluates to true (non-zero) or the
+ * timeout elapses.  Upon exit, @rc contains the value of the condition. This
+ * allows you to test the condition without incurring any side effects.
+ *
+ * This primary purpose of this macro is to poll on a hardware register
+ * until a status bit changes.  The timeout ensures that the loop still
+ * terminates if the bit never changes.  The delay is for devices that need a
+ * delay in between successive reads.
+ */
+#define spin_event_timeout(condition, timeout, delay, rc)                   \
+{                                                                           \
+	unsigned long __loops = tb_ticks_per_usec * timeout;                \
+	unsigned long __start = get_tbl();                                  \
+	while (!(rc = (condition)) && (tb_ticks_since(__start) <= __loops)) \
+		udelay(delay);                                      	    \
+}
+
 #endif /* __KERNEL__ */
 #endif /* _ASM_POWERPC_DELAY_H */
Signature · 1 line
1.6.0.6
author
#1 lore raw
Quoted 2 lines · Timur Tabi

The macro spin_event_timeout() takes a condition and timeout value (in microseconds) as parameters. =A0It spins until either the condition i=

s true

Quoted 1 line

or the timeout expires. =A0It returns the result of the condition when th=

e loop

Quoted 3 lines

was terminated.

This primary purpose of this macro is to poll on a hardware register unti=

l a

Quoted 1 line

status bit changes. =A0The timeout ensures that the loop still terminates=

if the

Quoted 2 lines

bit doesn't change as expected. =A0This macro makes it easier for driver developers to perform this kind of operation properly.

Any comments on this patch? I'd like to see this macro added to 2.6.31.

--=20
Timur Tabi
Linux kernel developer at Freescale

#2 lore raw
In reply to Timur Tabi
Quoted 73 lines · Timur Tabi

The macro spin_event_timeout() takes a condition and timeout value (in microseconds) as parameters. It spins until either the condition is true or the timeout expires. It returns the result of the condition when the loop was terminated.

This primary purpose of this macro is to poll on a hardware register until a status bit changes. The timeout ensures that the loop still terminates if the bit doesn't change as expected. This macro makes it easier for driver developers to perform this kind of operation properly.

---

I'm making this a PowerPC-specific patch because I want to use tb_ticks_per_usec, which does not exist on all other platforms. I don't want to use jiffies because jiffies works only when interrupts are enabled, and the resolution may not be fine enough.

arch/powerpc/include/asm/delay.h | 27 +++++++++++++++++++++++++++

1 files changed, 27 insertions(+), 0 deletions(-)
arch/powerpc/include/asm/delay.h +4 -0
index f9200a6..1939e0f 100644
@@ -2,6 +2,8 @@
 #define _ASM_POWERPC_DELAY_H
 #ifdef __KERNEL__
 
+#include <asm/time.h>
+
 /*
 * Copyright 1996, Paul Mackerras.
 *
@@ -30,5 +32,30 @@ extern void udelay(unsigned long usecs);
 #define mdelay(n)	udelay((n) * 1000)
 #endif
 
+/**
+ * spin_event_timeout - spin until a condition gets true or a  

timeout elapses
+ * @condition: a C expression to evalate
+ * @timeout: timeout, in microseconds
+ * @delay: the number of microseconds to delay between eache evaluation of
+ * @condition
+ * @rc: the last value of the condition
+ *
+ * The process spins until the condition evaluates to true (non- zero) or the
+ * timeout elapses. Upon exit, @rc contains the value of the condition. This
+ * allows you to test the condition without incurring any side effects.
+ *
+ * This primary purpose of this macro is to poll on a hardware register
+ * until a status bit changes. The timeout ensures that the loop still
+ * terminates if the bit never changes. The delay is for devices that need a
+ * delay in between successive reads.
+ */
+#define spin_event_timeout(condition, timeout, delay,
rc) \
+
{ \

arch/powerpc/include/asm/delay.h +1 -0
+	unsigned long __loops = tb_ticks_per_usec *  

timeout; \

arch/powerpc/include/asm/delay.h +1 -0
+	unsigned long __start =  

get_tbl(); \

arch/powerpc/include/asm/delay.h +1 -0
+	while (!(rc = (condition)) && (tb_ticks_since(__start) <=  

__loops)) \

arch/powerpc/include/asm/delay.h extern void udelay(unsigned long usecs); · line 59

Quoted 3 lines
arch/powerpc/include/asm/delay.h +3 -0
+		udelay(delay);                                      	    \
+}
+

I wouldn't call it spin_event_timeout as its a bit too generic of a name for something that is arch specific.

- k

#3 lore raw
Quoted 1 line · Kumar Gala

I wouldn't call it spin_event_timeout as its a bit too generic of a name for something that is arch specific.

The concept is not arch-specific -- ideally, other architectures could add their own implementations (or it could be made non-architecture-specific in the future).

-Scott

author
#4 lore raw
In reply to Kumar Gala
Quoted 2 lines · Kumar Gala

I wouldn't call it spin_event_timeout as its a bit too generic of a name for something that is arch specific.

Well, I'm hoping it won't be arch-specific forever. I'd like to see other architectures implement something like tb_ticks_per_usec. I'm only tackling PowerPC for now.

Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale
#5 lore raw
In reply to Timur Tabi
Quoted 11 lines · On Mon, 27 Apr 2009 14:05:44 -0500 "Timur Tabi" <[email protected]> wrote:

+#define spin_event_timeout(condition, timeout, delay,
rc) \

(quoted diff) +2 -0
+{                                                                           \
+	unsigned long __loops = tb_ticks_per_usec *

timeout; \

(quoted diff) +1 -0
+	unsigned long __start =

get_tbl(); \

(quoted diff) +1 -0
+	while (!(rc = (condition)) && (tb_ticks_since(__start) <=

__loops)) \

(quoted diff) +1 -0
+		udelay(delay);
	    \ +}
Quoted 1 line

+

Would cpu_relax be a good thing to put here?

Something like:

while (!(rc = (condition)) && (tb_ticks_since(__start) <= __loops)) \
	if (delay) \
		udelay(delay); \
	else \
		cpu_relax(); \

Cheers,
Sean

author
#6 lore raw
Quoted 10 lines · Sean MacLennan

Would cpu_relax be a good thing to put here?

Something like:

while (!(rc = (condition)) && (tb_ticks_since(__start) <= __loops)) \
	if (delay) \
		udelay(delay); \
	else \
		cpu_relax(); \

I had that at one point, but then I looked at the code for udelay(), and it appears that if I do udelay(0), it does something similar to cpu_relax:

	start = get_tbl();
	while (get_tbl() - start < loops)
		HMT_low();
	HMT_medium();

cpu_relax does this:

do { HMT_low(); HMT_medium(); barrier(); } while (0)

Well, now that I look at it, cpu_relax() changes the thread priority to low and then back to medium, whereas udelay() never sets it to low if 'loops' is 0.

I'm okay with changing my code, but I wonder if udelay() should look like this:

	start = get_tbl();
	HMT_low();
	while (get_tbl() - start < loops)
		HMT_low();
	HMT_medium();
Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale