[PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Kumar Gala]

Since we can only open the watchdog once having a spinlock to protect
multiple access is pointless.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>

---
commit 8852c088c0624f5534b08a769b5887c42a211694
tree 04ad05af51efb90604ff86af48197b6b7b45f77a
parent 1161d00ace361999d5b98bbe3082da4c8d457ba3
author Kumar Gala <galak@kernel.crashing.org> Thu, 19 Jan 2006 01:03:09 -0600
committer Kumar Gala <galak@kernel.crashing.org> Thu, 19 Jan 2006 01:03:09 -0600

 drivers/char/watchdog/mpc83xx_wdt.c |    5 -----
 1 files changed, 0 insertions(+), 5 deletions(-)

diff --git a/drivers/char/watchdog/mpc83xx_wdt.c b/drivers/char/watchdog/mpc83xx_wdt.c
index 5d6f506..b310144 100644
--- a/drivers/char/watchdog/mpc83xx_wdt.c
+++ b/drivers/char/watchdog/mpc83xx_wdt.c
@@ -57,15 +57,12 @@ static int prescale = 1;
 static unsigned int timeout_sec;
 
 static unsigned long wdt_is_open;
-static spinlock_t wdt_spinlock;
 
 static void mpc83xx_wdt_keepalive(void)
 {
 	/* Ping the WDT */
-	spin_lock(&wdt_spinlock);
 	out_be16(&wd_base->swsrr, 0x556c);
 	out_be16(&wd_base->swsrr, 0xaa39);
-	spin_unlock(&wdt_spinlock);
 }
 
 static ssize_t mpc83xx_wdt_write(struct file *file, const char __user *buf,
@@ -184,8 +181,6 @@ static int __devinit mpc83xx_wdt_probe(s
 		"mode:%s timeout=%d (%d seconds)\n",
 		reset ? "reset":"interrupt", timeout, timeout_sec);
 
-	spin_lock_init(&wdt_spinlock);
-
 	return 0;
 
 err_unmap:

Re: [PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Alan Cox]

On Iau, 2006-01-19 at 00:58 -0600, Kumar Gala wrote:
> Since we can only open the watchdog once having a spinlock to protect
> multiple access is pointless.
> 
> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>

NAK

This is a common mistake.

open is called on the open() call and is indeed in this case 'single
open', but file handles can be inherited and many users may have access
to a single file handle.

eg

	f = open("/dev/watchdog", O_RDWR);
	fork();
	while(1) {
		write(f, "Boing", 5);
	}

Alan

Re: [PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Kumar Gala]

On Jan 19, 2006, at 3:49 AM, Alan Cox wrote:

> On Iau, 2006-01-19 at 00:58 -0600, Kumar Gala wrote:
>> Since we can only open the watchdog once having a spinlock to protect
>> multiple access is pointless.
>>
>> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
>
> NAK
>
> This is a common mistake.
>
> open is called on the open() call and is indeed in this case 'single
> open', but file handles can be inherited and many users may have  
> access
> to a single file handle.
>
> eg
>
> 	f = open("/dev/watchdog", O_RDWR);
> 	fork();
> 	while(1) {
> 		write(f, "Boing", 5);
> 	}
>
> Alan

Thanks, you learn something new every day.

- kumar

Re: [PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Marcelo Tosatti]

On Thu, Jan 19, 2006 at 09:49:16AM +0000, Alan Cox wrote:
> On Iau, 2006-01-19 at 00:58 -0600, Kumar Gala wrote:
> > Since we can only open the watchdog once having a spinlock to protect
> > multiple access is pointless.
> > 
> > Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
> 
> NAK
> 
> This is a common mistake.
> 
> open is called on the open() call and is indeed in this case 'single
> open', but file handles can be inherited and many users may have access
> to a single file handle.
> 
> eg
> 
> 	f = open("/dev/watchdog", O_RDWR);
> 	fork();
> 	while(1) {
> 		write(f, "Boing", 5);
> 	}

Oops.

At least 50% of the watchdog drivers rely solely on the "wdt_is_open"
atomic variable and are broken with respect to synchronization.

Re: [PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Alan Cox]

On Iau, 2006-01-19 at 14:48 -0200, Marcelo Tosatti wrote:
> On Thu, Jan 19, 2006 at 09:49:16AM +0000, Alan Cox wrote:
> > 
> > 	f = open("/dev/watchdog", O_RDWR);
> > 	fork();
> > 	while(1) {
> > 		write(f, "Boing", 5);
> > 	}
> 
> Oops.
> 
> At least 50% of the watchdog drivers rely solely on the "wdt_is_open"
> atomic variable and are broken with respect to synchronization.

What an excellent janitors project

Re: [PATCH] powerpc: remove useless spinlock from mpc83xx watchdog

[Wim Van Sebroeck]

All,

> On Iau, 2006-01-19 at 14:48 -0200, Marcelo Tosatti wrote:
> > On Thu, Jan 19, 2006 at 09:49:16AM +0000, Alan Cox wrote:
> > > 
> > > 	f = open("/dev/watchdog", O_RDWR);
> > > 	fork();
> > > 	while(1) {
> > > 		write(f, "Boing", 5);
> > > 	}
> > 
> > Oops.
> > 
> > At least 50% of the watchdog drivers rely solely on the "wdt_is_open"
> > atomic variable and are broken with respect to synchronization.
> 
> What an excellent janitors project

I'll have a look at it from a global point of view.

Greetings,
Wim.