Re: [PATCH] LOCKDEP: use depends on LOCKDEP_SUPPORT instead of $ARCH list

[Randy Dunlap]

On 5/24/21 2:31 PM, Randy Dunlap wrote:
> On 5/24/21 2:04 PM, Waiman Long wrote:
>> On 5/24/21 3:47 PM, Randy Dunlap wrote:
>>> On 5/17/21 7:02 AM, Waiman Long wrote:
>>>> On 5/17/21 3:11 AM, Ingo Molnar wrote:
>>>>> * Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
>>>>>
>>>>>> Both arch/um/ and arch/xtensa/ cause a Kconfig warning for LOCKDEP.
>>>>>> These arch-es select LOCKDEP_SUPPORT but they are not listed as one
>>>>>> of the arch-es that LOCKDEP depends on.
>>>>>>
>>>>>> Since (16) arch-es define the Kconfig symbol LOCKDEP_SUPPORT if they
>>>>>> intend to have LOCKDEP support, replace the awkward list of
>>>>>> arch-es that LOCKDEP depends on with the LOCKDEP_SUPPORT symbol.
>>>>>>
>>>>>> Fixes this kconfig warning: (for both um and xtensa)
>>>>>>
>>>>>> WARNING: unmet direct dependencies detected for LOCKDEP
>>>>>>     Depends on [n]: DEBUG_KERNEL [=y] && LOCK_DEBUGGING_SUPPORT [=y] && (FRAME_POINTER [=n] || MIPS || PPC || S390 || MICROBLAZE || ARM || ARC || X86)
>>>>>>     Selected by [y]:
>>>>>>     - PROVE_LOCKING [=y] && DEBUG_KERNEL [=y] && LOCK_DEBUGGING_SUPPORT [=y]
>>>>>>     - LOCK_STAT [=y] && DEBUG_KERNEL [=y] && LOCK_DEBUGGING_SUPPORT [=y]
>>>>>>     - DEBUG_LOCK_ALLOC [=y] && DEBUG_KERNEL [=y] && LOCK_DEBUGGING_SUPPORT [=y]
>>>>>>
>>>>>> Signed-off-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
>>>>>> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
>>>>>> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
>>>>>> Cc: Will Deacon <will@xxxxxxxxxx>
>>>>>> Cc: Waiman Long <longman@xxxxxxxxxx>
>>>>>> Cc: Boqun Feng <boqun.feng@xxxxxxxxx>
>>>>>> Cc: Chris Zankel <chris@xxxxxxxxxx>
>>>>>> Cc: Max Filippov <jcmvbkbc@xxxxxxxxx>
>>>>>> Cc: linux-xtensa@xxxxxxxxxxxxxxxx
>>>>>> Cc: Johannes Berg <johannes@xxxxxxxxxxxxxxxx>
>>>>>> Cc: Jeff Dike <jdike@xxxxxxxxxxx>
>>>>>> Cc: Richard Weinberger <richard@xxxxxx>
>>>>>> Cc: Anton Ivanov <anton.ivanov@xxxxxxxxxxxxxxxxxx>
>>>>>> Cc: linux-um@xxxxxxxxxxxxxxxxxxx
>>>>>> ---
>>>>>>    lib/Kconfig.debug |    2 +-
>>>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>>
>>>>>> --- linux-next-20210514.orig/lib/Kconfig.debug
>>>>>> +++ linux-next-20210514/lib/Kconfig.debug
>>>>>> @@ -1383,7 +1383,7 @@ config LOCKDEP
>>>>>>        bool
>>>>>>        depends on DEBUG_KERNEL && LOCK_DEBUGGING_SUPPORT
>>>>>>        select STACKTRACE
>>>>>> -    depends on FRAME_POINTER || MIPS || PPC || S390 || MICROBLAZE || ARM || ARC || X86
>>>>>> +    depends on FRAME_POINTER || LOCKDEP_SUPPORT
>>>>> Ok - the FRAME_POINTER bit is weird. Are there any architectures that have
>>>>> FRAME_POINTER defined but no LOCKDEP_SUPPORT?
>>>> LOCK_DEBUGGING_SUPPORT depends on LOCKDEP_SUPPORT. So this patch is equivalent to just delete the second depends-on line.
>>> Yes, if we disregard the FRAME_POINTER part.
>>
>> My understanding is that the 2 depends-on statements have an implicit AND. So it is like
> 
> Right (on the implicit AND).
> 
>> DEBUG_KERNEL && LOCK_DEBUGGING_SUPPORT && (FRAME_POINTER || LOCKDEP_SUPPORT). LOCK_DEBUGGING_SUPPORT is true means the (FRAME_POINTER || LOCKDEP_SUPPORT) will always be true. FRAME_POINTER is true doesn't mean the other dependencies are true. That is why I said it is equivalent to just "DEBUG_KERNEL && LOCK_DEBUGGING_SUPPORT". IOW, FRAME_POINTER will play no part here.
>>
> 
> Ack. I should have done that myself.
> 
>>>
>>>> Beside LOCKDEP, LATENCYTOP also have exactly the same depends-on line.
>>> True, but I don't get any implication that the same patch applies there.
>>> Do you?
>> It is just an observation that I stumble on. It is not related to your patch.
> 
> Got it.
> 
>>>> So isn't FRAME_POINTER used mainly to support STACK_TRACE? However, LOCK_DEBUGGING_SUPPORT has already included STACK_TRACE_SUPPORT in its dependency. So why there is a FRAME_POINTER dependency?
>>> FRAME_POINTER is one way but it does not seem to be required
>>> for STACKTRACE_SUPPORT.
>>>
>>> Do you have any patch suggestions?
>>
>> Is it possible to just get rid of the 2nd depends-on statement?
>>
>> The 2nd depends-on line was introduced by commit 7d37cb2c912d ("lib: fix kconfig dependency on ARCH_WANT_FRAME_POINTER"):
> 
> and I should have looked at that history too. Thanks.
> 
> Yes, I agree, we can just delete that line...
> 
> I'll send a v2 and copy the author of commit 7d37cb2c912d as well.

Hm, as I review that commit, I have to wonder if the previous 'select'
was correct (if we disregard the Kconfig warning).  If so, then
FRAME_POINTER is still wanted/needed for some arch-es.

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2779c29d9981..417c3d3e521b 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1363,7 +1363,7 @@ config LOCKDEP
        bool
        depends on DEBUG_KERNEL && LOCK_DEBUGGING_SUPPORT
        select STACKTRACE
-       select FRAME_POINTER if !MIPS && !PPC && !ARM && !S390 && !MICROBLAZE && !ARC && !X86 <<<<<<<<<<<


>> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
>> index 2779c29d9981..417c3d3e521b 100644
>> --- a/lib/Kconfig.debug
>> +++ b/lib/Kconfig.debug
>> @@ -1363,7 +1363,7 @@ config LOCKDEP
>>         bool
>>         depends on DEBUG_KERNEL && LOCK_DEBUGGING_SUPPORT
>>         select STACKTRACE
>> -       select FRAME_POINTER if !MIPS && !PPC && !ARM && !S390 && !MICROBLAZE &&
>> +       depends on FRAME_POINTER || MIPS || PPC || S390 || MICROBLAZE || ARM ||
>>         select KALLSYMS
>>         select KALLSYMS_ALL
>>
>> Since STACKTRACE is selected by lockdep, maybe we can just remove the 2nd depends-on line to see if anyone complain.


-- 
~Randy