Skip to contentSkip to conversation

RTC: mark if rtc-cmos drivers were successfully registered.

PATCH In review Thadeu Lima de Souza Cascardo 3 participants 5 messages Jul 5, 2009 – Jul 20, 2009 Real time clock (RTC)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 RTC: mark if rtc-cmos drivers were successfully registered.

rtc-cmos has two drivers, one PNP and one platform. When PNP has not succeeded probing, platform is registered. However, it tries to unregister both drivers unconditionally, instead of only unregistering those that were successfully registered. Fix that with a boolean variable for each driver indicating whether registering was successful.

---
 drivers/rtc/rtc-cmos.c |   23 ++++++++++++++++++-----
 1 files changed, 18 insertions(+), 5 deletions(-)
drivers/rtc/rtc-cmos.c +18 -5
index 23e10b6..f7a4701 100644
@@ -1174,23 +1174,34 @@ static struct platform_driver cmos_platform_driver = {
 	}
 };
 
+#ifdef CONFIG_PNP
+static bool pnp_driver_registered;
+#endif
+static bool platform_driver_registered;
+
 static int __init cmos_init(void)
 {
 	int retval = 0;
 
 #ifdef	CONFIG_PNP
-	pnp_register_driver(&cmos_pnp_driver);
+	retval = pnp_register_driver(&cmos_pnp_driver);
+	if (retval == 0)
+		pnp_driver_registered = true;
 #endif
 
-	if (!cmos_rtc.dev)
+	if (!cmos_rtc.dev) {
 		retval = platform_driver_probe(&cmos_platform_driver,
 					       cmos_platform_probe);
+		if (retval == 0)
+			platform_driver_registered = true;
+	}
 
 	if (retval == 0)
 		return 0;
 
 #ifdef	CONFIG_PNP
-	pnp_unregister_driver(&cmos_pnp_driver);
+	if (pnp_driver_registered)
+		pnp_unregister_driver(&cmos_pnp_driver);
 #endif
 	return retval;
 }
@@ -1199,9 +1210,11 @@ module_init(cmos_init);
 static void __exit cmos_exit(void)
 {
 #ifdef	CONFIG_PNP
-	pnp_unregister_driver(&cmos_pnp_driver);
+	if (pnp_driver_registered)
+		pnp_unregister_driver(&cmos_pnp_driver);
 #endif
-	platform_driver_unregister(&cmos_platform_driver);
+	if (platform_driver_registered)
+		platform_driver_unregister(&cmos_platform_driver);
 }
 module_exit(cmos_exit);
 
Signature · 1 line
1.6.3
#1 lore raw
Quoted 5 lines · On Sun, 5 Jul 2009 01:07:59 -0300 Thadeu Lima de Souza Cascardo…

rtc-cmos has two drivers, one PNP and one platform. When PNP has not succeeded probing, platform is registered. However, it tries to unregister both drivers unconditionally, instead of only unregistering those that were successfully registered. Fix that with a boolean variable for each driver indicating whether registering was successful.

OK, thanks.

This came up a few weeks ago - the kernel was actually crashing deep down in the driver core, when a not-registered device was unregistered.

I believe Kay was planning on making the driver core more robust, so that crash shouldn't be happening any more. Kay, can you please confirm that thsi got fixed?

But we still shouldn't be unregistering a not-registered driver, and the patch looks to be a suitable way of preventing that.

Alternatively, we could declare that unregistering a not-registered driver is a permissible thing for a subsytem to do. It doesn't sound like a good idea though - we'd lose runtime checking opportunities.

 drivers/rtc/rtc-cmos.c |   23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)
drivers/rtc/rtc-cmos.c renamed from drivers/rtc/rtc-cmos.c~rtc-mark-if-rtc-cmos-drivers-were-successfully-registered +18 -5
diff -puN drivers/rtc/rtc-cmos.c~rtc-mark-if-rtc-cmos-drivers-were-successfully-registered drivers/rtc/rtc-cmos.c
@@ -1174,23 +1174,34 @@ static struct platform_driver cmos_platf
 	}
 };
 
+#ifdef CONFIG_PNP
+static bool pnp_driver_registered;
+#endif
+static bool platform_driver_registered;
+
 static int __init cmos_init(void)
 {
 	int retval = 0;
 
 #ifdef	CONFIG_PNP
-	pnp_register_driver(&cmos_pnp_driver);
+	retval = pnp_register_driver(&cmos_pnp_driver);
+	if (retval == 0)
+		pnp_driver_registered = true;
 #endif
 
-	if (!cmos_rtc.dev)
+	if (!cmos_rtc.dev) {
 		retval = platform_driver_probe(&cmos_platform_driver,
 					       cmos_platform_probe);
+		if (retval == 0)
+			platform_driver_registered = true;
+	}
 
 	if (retval == 0)
 		return 0;
 
 #ifdef	CONFIG_PNP
-	pnp_unregister_driver(&cmos_pnp_driver);
+	if (pnp_driver_registered)
+		pnp_unregister_driver(&cmos_pnp_driver);
 #endif
 	return retval;
 }
@@ -1199,9 +1210,11 @@ module_init(cmos_init);
 static void __exit cmos_exit(void)
 {
 #ifdef	CONFIG_PNP
-	pnp_unregister_driver(&cmos_pnp_driver);
+	if (pnp_driver_registered)
+		pnp_unregister_driver(&cmos_pnp_driver);
 #endif
-	platform_driver_unregister(&cmos_platform_driver);
+	if (platform_driver_registered)
+		platform_driver_unregister(&cmos_platform_driver);
 }
 module_exit(cmos_exit);
 

_

#2 lore raw
Quoted 9 lines · Andrew Morton
Quoted 5 lines · On Sun,  5 Jul 2009 01:07:59 -0300 Thadeu Lima de Souza Cascardo…

rtc-cmos has two drivers, one PNP and one platform. When PNP has not succeeded probing, platform is registered. However, it tries to unregister both drivers unconditionally, instead of only unregistering those that were successfully registered. Fix that with a boolean variable for each driver indicating whether registering was successful.

Quoted 6 lines

This came up a few weeks ago - the kernel was actually crashing deep down in the driver core, when a not-registered device was unregistered.

I believe Kay was planning on making the driver core more robust, so that crash shouldn't be happening any more.  Kay, can you please confirm that thsi got fixed?

This is supposed to fix it:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=5c8563d773c0e9f0ac2a552e84806decd98ce732

Thanks,
Kay

#3 lore raw
Quoted 17 lines · Kay Sievers
Quoted 9 lines · Andrew Morton
Quoted 5 lines · On Sun,  5 Jul 2009 01:07:59 -0300 Thadeu Lima de Souza Cascardo…

rtc-cmos has two drivers, one PNP and one platform. When PNP has not succeeded probing, platform is registered. However, it tries to unregister both drivers unconditionally, instead of only unregistering those that were successfully registered. Fix that with a boolean variable for each driver indicating whether registering was successful.

Quoted 6 lines

This came up a few weeks ago - the kernel was actually crashing deep down in the driver core, when a not-registered device was unregistered.

I believe Kay was planning on making the driver core more robust, so that crash shouldn't be happening any more.  Kay, can you please confirm that thsi got fixed?

Quoted 5 lines

This is supposed to fix it:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=5c8563d773c0e9f0ac2a552e84806decd98ce732

Thanks,
Kay

Yes, but instead of an oops, we get a warning when the driver could just behave properly. That's the intention of this patch: do not warn the user when there's nothing to warn about, as long as the driver behaves itself. Doing an rmmod on a unused module should not give the user what looks like an oops. I should start being more verbose with my commit messages. :-D

My best and thanks for you work,
Thadeu Cascardo.

#4 lore raw
Quoted 7 lines · Thadeu Lima de Souza Cascardo
Quoted 6 lines · Kay Sievers
Quoted 3 lines · Andrew Morton

I believe Kay was planning on making the driver core more robust, so that crash shouldn't be happening any more.  Kay, can you please confirm that thsi got fixed?

Quoted 7 lines
Quoted 2 lines

This is supposed to fix it:
  http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=5c8563d773c0e9f0ac2a552e84806decd98ce732

Yes, but instead of an oops, we get a warning when the driver could just behave properly. That's the intention of this patch: do not warn the user when there's nothing to warn about, as long as the driver behaves itself.

Sure, it was not supposed to fix the rtc issue properly, it's just to make it not oops. :)

Thanks,
Kay