Home » Mailing list archives » linux-kernel » 2010 » May » 4

Re: [PATCH 2/2] lockdep: allow user to disable CONFIG_TRACE_IRQFLAGS

Previous thread: [PATCH 1/2] kernel:fork: fix dependency of hardirqs_enabled and softirqs_enabled by tom.leiming on Tuesday, May 4, 2010 - 9:01 am. (3 messages)

Next thread: [PATCH][GIT PULL] tracing: Convert nop macros to static inlines by Steven Rostedt on Tuesday, May 4, 2010 - 9:17 am. (2 messages)
[view original message]
From: tom.leiming
Subject: [PATCH 2/2] lockdep: allow user to disable CONFIG_TRACE_IRQFLAGS
Date: Tuesday, May 4, 2010 - 9:02 am

From: Ming Lei <tom.leiming@gmail.com>

There is still the lost irq-on tracing issue on ARM,
(see http://marc.info/?l=linux-arm-kernel&m=126047420005553&w=2)
which may disable lockdep after kernel boot, seems no
good solution for the issue up to now.

The patch exports the config option of CONFIG_TRACE_IRQFLAGS
and allows user to disable it for support of checking circular
dead lock only if irq flags tracing can't work correctly always.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
---
 lib/Kconfig.debug      |    3 +++
 lib/locking-selftest.c |    2 ++
 2 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 935248b..448f18f 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -550,11 +550,14 @@ config DEBUG_LOCKDEP
 	  of more runtime overhead.
 
 config TRACE_IRQFLAGS
+	bool "Lock prove: IRQ flags tracing"
 	depends on DEBUG_KERNEL
 	bool
 	default y
 	depends on TRACE_IRQFLAGS_SUPPORT
 	depends on PROVE_LOCKING
+	help
+	  If you say Y here, IRQ flags will be traced to prove locking. 
 
 config DEBUG_SPINLOCK_SLEEP
 	bool "Spinlock debugging: sleep-inside-spinlock checking"
diff --git a/lib/locking-selftest.c b/lib/locking-selftest.c
index 619313e..d9e5bfe 100644
--- a/lib/locking-selftest.c
+++ b/lib/locking-selftest.c
@@ -1176,6 +1176,7 @@ void locking_selftest(void)
 
 	printk("  --------------------------------------------------------------------------\n");
 
+#ifdef CONFIG_TRACE_IRQFLAGS
 	/*
 	 * irq-context testcases:
 	 */
@@ -1188,6 +1189,7 @@ void locking_selftest(void)
 
 	DO_TESTCASE_6x2("irq read-recursion", irq_read_recursion);
 //	DO_TESTCASE_6x2B("irq read-recursion #2", irq_read_recursion2);
+#endif
 
 	if (unexpected_testcase_failures) {
 		printk("-----------------------------------------------------------------\n");
-- 
1.6.2.5

--


[ message continues ]
[view original message]
From: Ming Lei
Subject: Re: [PATCH 2/2] lockdep: allow user to disable CONFIG_TRACE_IRQFLAGS
Date: Tuesday, May 4, 2010 - 9:06 am

CC: ARM




-- 
Lei Ming
--


[ message continues ]
[view original message]
From: Peter Zijlstra
Subject: Re: [PATCH 2/2] lockdep: allow user to disable CONFIG_TRACE_IRQFLAGS
Date: Tuesday, May 4, 2010 - 9:26 am

Well, do like x86 does in entry.S call into lockdep right before
returning to userspace.


--


[ message continues ]
Previous thread: [PATCH 1/2] kernel:fork: fix dependency of hardirqs_enabled and softirqs_enabled by tom.leiming on Tuesday, May 4, 2010 - 9:01 am. (3 messages)

Next thread: [PATCH][GIT PULL] tracing: Convert nop macros to static inlines by Steven Rostedt on Tuesday, May 4, 2010 - 9:17 am. (2 messages)