2007-06-06 10:53:26

by maximilian attems

[permalink] [raw]
Subject: [patch] Kconfig.debug DEBUG_PREEMPT dependency

allow to set DEBUG_PREEMPT on PREEMPT_BKL, but !PREEMPT
purpose is to get bonus of debug_smp_processor_id().

Signed-off-by: maximilian attems <[email protected]>

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index da95e10..1d4ab66 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -145,7 +145,7 @@ config DEBUG_SLAB_LEAK

config DEBUG_PREEMPT
bool "Debug preemptible kernel"
- depends on DEBUG_KERNEL && PREEMPT && TRACE_IRQFLAGS_SUPPORT
+ depends on DEBUG_KERNEL && (PREEMPT || PREEMPT_BKL) && TRACE_IRQFLAGS_SUPPORT
default y
help
If you say Y here then the kernel will use a debug variant of the


2007-06-06 11:12:25

by maximilian attems

[permalink] [raw]
Subject: Re: [patch] Kconfig.debug DEBUG_PREEMPT dependency

On Wed, 06 Jun 2007, maximilian attems wrote:

> - depends on DEBUG_KERNEL && PREEMPT && TRACE_IRQFLAGS_SUPPORT
> + depends on DEBUG_KERNEL && (PREEMPT || PREEMPT_BKL) && TRACE_IRQFLAGS_SUPPORT

upps please discard needs more work:
arch/i386/kernel/built-in.o: In function `show_registers':
(.text+0x2e8e): undefined reference to `add_preempt_count'
arch/i386/kernel/built-in.o: In function `show_registers':
(.text+0x2e9c): undefined reference to `sub_preempt_count'
arch/i386/kernel/built-in.o: In function `show_registers':
(.text+0x2ef2): undefined reference to `add_preempt_count'
arch/i386/kernel/built-in.o: In function `show_registers':
(.text+0x2f02): undefined reference to `sub_preempt_count'
arch/i386/kernel/built-in.o: In function `is_valid_bugaddr':
(.text+0x366c): undefined reference to `add_preempt_count'
arch/i386/kernel/built-in.o: In function `is_valid_bugaddr':
(.text+0x367b): undefined reference to `sub_preempt_count'
arch/i386/kernel/built-in.o: In function `romchecksum':
...

--
maks

2007-06-06 11:32:42

by Ingo Molnar

[permalink] [raw]
Subject: Re: [patch] Kconfig.debug DEBUG_PREEMPT dependency


* maximilian attems <[email protected]> wrote:

> allow to set DEBUG_PREEMPT on PREEMPT_BKL, but !PREEMPT purpose is to
> get bonus of debug_smp_processor_id().

hm, have you tested this? In the case of !PREEMPT there will be _lots_
of places that use smp_processor_id() but where the preempt_count() wont
be elevated - causing false positive warnings.

Ingo