Re: [PATCH] x86, mtrr: mark range_new in mtrr_calc_range_state() as __initdata

[Ingo Molnar <mingo@kernel.org>]

* Rasmus Villemoes <linux@rasmusvillemoes.dk> wrote:

> On Fri, Nov 27 2015, Ingo Molnar <mingo@kernel.org> wrote:
> 
> > * Rasmus Villemoes <linux@rasmusvillemoes.dk> wrote:
> >
> >> range_new doesn't seem to be used after init. It is only passed to
> >> memset, sum_ranges, memcmp and x86_get_mtrr_mem_range, the latter of
> >> which also only passes it on to various *range* library functions. So
> >> mark it __initdata to free up an extra page after init.
> >> 
> >> nr_range_new is unconditionally assigned to before it is read, so
> >> there's no point in having it static.
> >> 
> >> Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
> >> ---
> >>  arch/x86/kernel/cpu/mtrr/cleanup.c | 4 ++--
> >>  1 file changed, 2 insertions(+), 2 deletions(-)
> >> 
> >> diff --git a/arch/x86/kernel/cpu/mtrr/cleanup.c b/arch/x86/kernel/cpu/mtrr/cleanup.c
> >> index 70d7c93f4550..b1a9ad366f67 100644
> >> --- a/arch/x86/kernel/cpu/mtrr/cleanup.c
> >> +++ b/arch/x86/kernel/cpu/mtrr/cleanup.c
> >> @@ -593,9 +593,9 @@ mtrr_calc_range_state(u64 chunk_size, u64 gran_size,
> >>  		      unsigned long x_remove_base,
> >>  		      unsigned long x_remove_size, int i)
> >>  {
> >> -	static struct range range_new[RANGE_NUM];
> >> +	static struct range range_new[RANGE_NUM] __initdata;
> >>  	unsigned long range_sums_new;
> >> -	static int nr_range_new;
> >> +	int nr_range_new;
> >>  	int num_reg;
> >>  
> >>  	/* Convert ranges to var ranges state: */
> >
> > So this static variable actually surprised me - I never realized it was there - 
> > and it's not some simple 'once' flag, but something that is essential semantics.
> >
> > So marking it __initdata is correct, but please also move it out of function local 
> > variables scope, into file scope - and name it properly as well, like 
> > mtrr_new_range[] or so?
> 
> I can certainly do that, but isn't the usual preference to keep the scope as 
> small as possible? IOW, why do you want to make this a file-scoped variable?

The preference is to keep code readable and obvious, and this one wasn't: relevant 
state/data was hidden via a non-commented local static variable.

> Also, I don't really see how the 'static' has 'essential semantics'. AFAICT, the 
> contents are wiped on every invocation of mtrr_calc_range_state, so the only 
> reason it's static is to avoid blowing the stack.

So this was another property that wasn't obvious from the limited context I saw in 
the patch, i.e. the variable definition. Another solution would be to add a 
comment explaining that this is a local variable to keep kernel stack size down, 
and explain why it's safe to do that.

Thanks,

	Ingo