Re: [PATCH 1/1] intel_txt: to fix build errors of CONFIG_ACPI_SLEEP

[Ingo Molnar]

* H. Peter Anvin <hpa@xxxxxxxxx> wrote:

> On 08/12/2009 11:46 PM, Ingo Molnar wrote:
> > * Wang, Shane <shane.wang@xxxxxxxxx> wrote:
> > 
> >> Signed-off-by: Shane Wang <shane.wang@xxxxxxxxx>
> >>
> >> diff -r 7358cf1b3090 arch/x86/kernel/tboot.c
> >> --- a/arch/x86/kernel/tboot.c   Wed Aug 12 03:04:23 2009 -0700
> >> +++ b/arch/x86/kernel/tboot.c   Wed Aug 12 18:06:21 2009 -0700
> >> @@ -169,6 +169,7 @@ void tboot_create_trampoline(void)
> >>
> >>  static void set_mac_regions(void)
> >>  {
> >> +#ifdef CONFIG_ACPI_SLEEP
> >>         tboot->num_mac_regions = 3;
> >>         /* S3 resume code */
> >>         tboot->mac_regions[0].start = PFN_PHYS(PFN_DOWN(acpi_wakeup_address));
> >> @@ -181,6 +182,7 @@ static void set_mac_regions(void)
> >>         tboot->mac_regions[2].start = PFN_PHYS(PFN_DOWN(virt_to_phys(&_text)));
> >>         tboot->mac_regions[2].size = PFN_PHYS(PFN_UP(virt_to_phys(&_end))) -
> >>                                      PFN_PHYS(PFN_DOWN(virt_to_phys(&_text)));
> >> +#endif
> > 
> > These #ifdefs are quite ugly. Why not add a 'depends on ACPI_SLEEP' 
> > rule to the INTEL_TXT Kconfig section?
> 
> I *strongly* disagree with that kind of false dependencies.  It 
> makes it seem like there is something magic going on, which 
> invites cargo cult programming in the future.  I also think those 
> particular #ifdefs are fairly innocuous... it's not like they're 
> hiding flow of control or major chunks of code.

no argument about that. But i'd rather have some extra dependency 
on a Kconfig than ugly #ifdefs.

and while it's certainly not the same kind of dependency as 
'depends on ACPI', it's not really a false dependency either, it's 
just a different technical dependency: 'the ACPI_SLEEP interfaces 
suck a bit and we could only offer !ACPI_SLEEP support by uglifying 
the .c code - so we'll opt for always having it'.

> However, I think the actual code to set the sections is ugly as 
> hell, which is probably why the #ifdef sticks in your eyes.

Yeah.

> Consider the attached instead patch, which abstracts some of the 
> (way more complex than it should be) open-coded stuff and 
> therefore makes it stick out less.

it's better, but why not put these:

> +#ifdef CONFIG_ACPI_SLEEP
>  	tboot->acpi_sinfo.kernel_s3_resume_vector = acpi_wakeup_address;
> +#endif

into a helper inline in a header file and let the #ifdefs be there, 
or something like that. That way the .c file stays readable.

I.e. we can cleanly support !ACPI_SLEEP case too, because it's 
encapsulated and abstracted away.

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/