On Mar 6, 2013, at 1:50 PM, H. Peter Anvin wrote:
> On 03/06/2013 11:37 AM, Brassow Jonathan wrote:
>>
>> On Mar 6, 2013, at 12:51 PM, H. Peter Anvin wrote:
>>
>>> On 03/06/2013 10:24 AM, Brassow Jonathan wrote:
>>>> I thought the same thing. I couldn't find my mistake though... Would it help if I attached my config file?
>>>>
>>>> brassow
>>>
>>> Yes, and the exact steps you took which resulted in the failure.
>>
>> I'm running RHEL6.3. To generate the .config file, I used 'make localmodconfig'. When prompted on new features, I simply went with the default (hit enter). This resulted in the attached config file, which errors out with the following messages:
>> LD arch/x86/kernel/acpi/built-in.o
>> CC arch/x86/kernel/apic/apic.o
>> arch/x86/kernel/apic/apic.c: In function ‘lapic_next_deadline’:
>> arch/x86/kernel/apic/apic.c:475: error: ‘MSR_IA32_TSC_DEADLINE’ undeclared (first use in this function)
>> arch/x86/kernel/apic/apic.c:475: error: (Each undeclared identifier is reported only once
>> arch/x86/kernel/apic/apic.c:475: error: for each function it appears in.)
>> make[3]: *** [arch/x86/kernel/apic/apic.o] Error 1
>> make[2]: *** [arch/x86/kernel/apic] Error 2
>> make[1]: *** [arch/x86/kernel] Error 2
>> make: *** [arch/x86] Error 2
>>
>> I tracked that down to the header file not being properly included. I wondered also, why is the compiler complaining about the lack of macro definitions instead of complaining about the missing header file?
>>
>
> That is extremely odd.
>
> The thing is, your change is actually buggy -- it breaks "make
> headers_install".
>
> Could you do "make arch/x86/kernel/apic/apic.i" and see what version of
> msr-index.h actually *does* get included?
>
> The only reason for putting <uapi/*> in is when there are namespace
> collisions between uapi and kernel proper...
I think I see the problem....
I've been rsync'ing differences between my work machine and test machines - compiling on the test machines. I have not been using '--delete' when rsync'ing, so there have been two copies of msr-index.h - the new one in uapi and the old one that didn't get removed. ('make mrproper' obviously doesn't clean that sort of thing up.) The collision is what is causing the problem and explains why the compiler never complains about a missing header. I didn't spot this sooner because I was looking at the sources on my work machine.
Sorry for the noise,
brassow
--
To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|