[patch 2/6] 12/34: cdrom/cdu31a: replace interruptible_sleep_on() with wait_event_interruptible()

[patch 2/6] 12/34: cdrom/cdu31a: replace interruptible_sleep_on() with wait_event_interruptible()

[domen]

Use wait_event_interruptible() instead of the deprecated
interruptible_sleep_on(). The patch is straight-forward as the macros should 
result in the same execution. Patch is compile-tested (still throws out warnings
regarding {save,restore}_flags()).

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Domen Puncer <domen@coderock.org>
---


 kj-domen/drivers/cdrom/cdu31a.c |   40 +++++++++++++++++-----------------------
 1 files changed, 17 insertions(+), 23 deletions(-)

diff -puN drivers/cdrom/cdu31a.c~int_sleep_on-drivers_cdrom_cdu31a drivers/cdrom/cdu31a.c
--- kj/drivers/cdrom/cdu31a.c~int_sleep_on-drivers_cdrom_cdu31a	2005-03-05 16:11:33.000000000 +0100
+++ kj-domen/drivers/cdrom/cdu31a.c	2005-03-05 16:11:33.000000000 +0100
@@ -166,6 +166,7 @@
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/interrupt.h>
+#include <linux/wait.h>
 
 #include <asm/system.h>
 #include <asm/io.h>
@@ -865,15 +866,13 @@ do_sony_cd_cmd(unsigned char cmd,
 	save_flags(flags);
 	cli();
 	if (current != has_cd_task) {	/* Allow recursive calls to this routine */
-		while (sony_inuse) {
-			interruptible_sleep_on(&sony_wait);
-			if (signal_pending(current)) {
-				result_buffer[0] = 0x20;
-				result_buffer[1] = SONY_SIGNAL_OP_ERR;
-				*result_size = 2;
-				restore_flags(flags);
-				return;
-			}
+		wait_event_interruptible(sony_wait, !sony_inuse);
+		if (signal_pending(current)) {
+			result_buffer[0] = 0x20;
+			result_buffer[1] = SONY_SIGNAL_OP_ERR;
+			*result_size = 2;
+			restore_flags(flags);
+			return;
 		}
 		sony_inuse = 1;
 		has_cd_task = current;
@@ -1372,16 +1371,13 @@ static void do_cdu31a_request(request_qu
 	 */
 	save_flags(flags);
 	cli();
-	while (sony_inuse) {
-		interruptible_sleep_on(&sony_wait);
-		if (signal_pending(current)) {
-			restore_flags(flags);
+	wait_event_interruptible(sony_wait, !sony_inuse);
+	if (signal_pending(current)) {
+		restore_flags(flags);
 #if DEBUG
-			printk("Leaving do_cdu31a_request at %d\n",
-			       __LINE__);
+		printk("Leaving do_cdu31a_request at %d\n", __LINE__);
 #endif
-			return;
-		}
+		return;
 	}
 	sony_inuse = 1;
 	has_cd_task = current;
@@ -2326,12 +2322,10 @@ static int read_audio(struct cdrom_read_
 	 */
 	save_flags(flags);
 	cli();
-	while (sony_inuse) {
-		interruptible_sleep_on(&sony_wait);
-		if (signal_pending(current)) {
-			restore_flags(flags);
-			return -EAGAIN;
-		}
+	wait_event_interruptible(sony_wait, sony_inuse);
+	if (signal_pending(current)) {
+		restore_flags(flags);
+		return -EAGAIN;
 	}
 	sony_inuse = 1;
 	has_cd_task = current;
_

Re: [patch 2/6] 12/34: cdrom/cdu31a: replace interruptible_sleep_on() with wait_event_interruptible()

[Ondrej Zary]

domen@coderock.org wrote:
> Use wait_event_interruptible() instead of the deprecated
> interruptible_sleep_on(). The patch is straight-forward as the macros should 
> result in the same execution. Patch is compile-tested (still throws out warnings
> regarding {save,restore}_flags()).
> 
> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
> Signed-off-by: Domen Puncer <domen@coderock.org>

I've posted a patch for the cdu31a driver some time ago that removes 
almost all usage of interruptible_sleep_on() and also 
{save,restore}_flags() - it uses semaphore instead.
The only remaining code is in sony_sleep() function when using 
IRQ-driven operation.

See http://lkml.org/lkml/2004/12/18/107
The patch is big because I've messed with the formatting...

-- 
Ondrej Zary

Re: [patch 2/6] 12/34: cdrom/cdu31a: replace interruptible_sleep_on() with wait_event_interruptible()

[Domen Puncer]

On 06/03/05 12:42 +0100, Ondrej Zary wrote:
> domen@coderock.org wrote:
> >Use wait_event_interruptible() instead of the deprecated
> >interruptible_sleep_on(). The patch is straight-forward as the macros 
> >should result in the same execution. Patch is compile-tested (still throws 
> >out warnings
> >regarding {save,restore}_flags()).
> >
> >Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
> >Signed-off-by: Domen Puncer <domen@coderock.org>
> 

Uh, this one escaped me, as schedule() (sleep_on) after cli() is clearly
wrong. Btw. what was the reason for this?

> I've posted a patch for the cdu31a driver some time ago that removes 
> almost all usage of interruptible_sleep_on() and also 
> {save,restore}_flags() - it uses semaphore instead.
> The only remaining code is in sony_sleep() function when using 
> IRQ-driven operation.
> 
> See http://lkml.org/lkml/2004/12/18/107
> The patch is big because I've messed with the formatting...

I looked at it, and rewrote some of it into smaller patches. If you don't
mind, can i send them to you for review and testing?


	Domen

Re: [patch 2/6] 12/34: cdrom/cdu31a: replace interruptible_sleep_on() with wait_event_interruptible()

[Ondrej Zary]

Domen Puncer wrote:
> On 06/03/05 12:42 +0100, Ondrej Zary wrote:
> 
>>domen@coderock.org wrote:
>>
>>>Use wait_event_interruptible() instead of the deprecated
>>>interruptible_sleep_on(). The patch is straight-forward as the macros 
>>>should result in the same execution. Patch is compile-tested (still throws 
>>>out warnings
>>>regarding {save,restore}_flags()).
>>>
>>>Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
>>>Signed-off-by: Domen Puncer <domen@coderock.org>
>>
> 
> Uh, this one escaped me, as schedule() (sleep_on) after cli() is clearly
> wrong. Btw. what was the reason for this?

I've never tried the IRQ-driven operation so I don't know if that code 
still works (it certainly used to for the original author some years 
ago). I think that it should wait until the IRQ from the drive arrives.

>>I've posted a patch for the cdu31a driver some time ago that removes 
>>almost all usage of interruptible_sleep_on() and also 
>>{save,restore}_flags() - it uses semaphore instead.
>>The only remaining code is in sony_sleep() function when using 
>>IRQ-driven operation.
>>
>>See http://lkml.org/lkml/2004/12/18/107
>>The patch is big because I've messed with the formatting...
> 
> 
> I looked at it, and rewrote some of it into smaller patches. If you don't
> mind, can i send them to you for review and testing?

Thanks. I'll test them. I guess that these smaller patches are more 
likely to be accepted into mainline.

-- 
Ondrej Zary