[PATCH 2.6] Fix dev_printk to work with unclaimed devices

[PATCH 2.6] Fix dev_printk to work with unclaimed devices

[Deepak Saxena]

I need to do some fixup in platform_notify() and when trying to 
use the dev_* print functions for informational messages, they OOPs 
b/c the current code assumes that dev->driver exists. This is not the 
case since platform_notify() is called before a device has been attached
to any driver. 

--- linux-2.5-bk/include/linux/device.h	2004-02-10 14:51:49.000000000 -0700
+++ linux-2.6-ds/include/linux/device.h	2004-02-26 11:10:38.000000000 -0700
@@ -395,7 +395,13 @@
 
 /* debugging and troubleshooting/diagnostic helpers. */
 #define dev_printk(level, dev, format, arg...)	\
-	printk(level "%s %s: " format , (dev)->driver->name , (dev)->bus_id , ## arg)
+	do {					\
+		if ((dev)->driver) {		\
+			printk(level "%s %s: " format , (dev)->driver->name , (dev)->bus_id , ## arg);				\
+		} else {			\
+			printk(level "%s (Unclaimed %s bus device): " format , (dev)->bus_id, (dev)->bus->name , ## arg);					\
+		}				\
+	} while (0)
 
 #ifdef DEBUG
 #define dev_dbg(dev, format, arg...)		\

Re: [PATCH 2.6] Fix dev_printk to work with unclaimed devices

[Linus Torvalds]

On Thu, 26 Feb 2004, Deepak Saxena wrote:
> 
> I need to do some fixup in platform_notify() and when trying to 
> use the dev_* print functions for informational messages, they OOPs 
> b/c the current code assumes that dev->driver exists. This is not the 
> case since platform_notify() is called before a device has been attached
> to any driver. 

Make it a real function with varags, please.

On the other hand, it also is probably just _wrong_ to use "dev_printk()" 
if you aren't the driver for the device.

		Linus

Re: [PATCH 2.6] Fix dev_printk to work with unclaimed devices

[Greg KH]

On Thu, Feb 26, 2004 at 11:34:39AM -0700, Deepak Saxena wrote:
> 
> I need to do some fixup in platform_notify() and when trying to 
> use the dev_* print functions for informational messages, they OOPs 
> b/c the current code assumes that dev->driver exists. This is not the 
> case since platform_notify() is called before a device has been attached
> to any driver. 

Yeah, this "limitation" of the dev_* printks have been known for a
while, and it was determined that for situations like this, it's not
worth using those calls.

I have a patch somewhere in my tree that will give you a nice WARN()
output if this ever happens, so as to help when trying to port a new bus
to the driver model, but it's too ugly for mainline.  Ah, found it, it's
below...

thanks,

greg k-h


diff -Nru a/include/linux/device.h b/include/linux/device.h
--- a/include/linux/device.h	Thu Feb 26 10:48:37 2004
+++ b/include/linux/device.h	Thu Feb 26 10:48:37 2004
@@ -394,8 +394,20 @@
 extern void firmware_unregister(struct subsystem *);
 
 /* debugging and troubleshooting/diagnostic helpers. */
+#ifdef CONFIG_DEBUG_DEV_PRINTK
+#define dev_printk(level, dev, format, arg...)			\
+	do {							\
+		if (!(dev) || !(dev)->driver)			\
+			WARN_ON(1);				\
+		else						\
+			printk(level "%s %s: " format , 	\
+				(dev)->driver->name , 		\
+				(dev)->bus_id , ## arg);	\
+	} while (0)
+#else
 #define dev_printk(level, dev, format, arg...)	\
 	printk(level "%s %s: " format , (dev)->driver->name , (dev)->bus_id , ## arg)
+#endif
 
 #ifdef DEBUG
 #define dev_dbg(dev, format, arg...)		\

Re: [PATCH 2.6] Fix dev_printk to work with unclaimed devices

[Deepak Saxena]

On Feb 26 2004, at 10:53, Greg KH was caught saying:
> On Thu, Feb 26, 2004 at 11:34:39AM -0700, Deepak Saxena wrote:
> > 
> > I need to do some fixup in platform_notify() and when trying to 
> > use the dev_* print functions for informational messages, they OOPs 
> > b/c the current code assumes that dev->driver exists. This is not the 
> > case since platform_notify() is called before a device has been attached
> > to any driver. 
> 
> Yeah, this "limitation" of the dev_* printks have been known for a
> while, and it was determined that for situations like this, it's not
> worth using those calls.

I can just use printks as it is only two quick log messages at
init time.

Tnx,
~Deepak

-- 
Deepak Saxena - dsaxena at plexity dot net - http://www.plexity.net/