num_physpages vs. totalram_pages

num_physpages vs. totalram_pages

[Jan Beulich]

Rusty,

in a mail from beginning of last year you indicated that the confusion between
these two variables should be cleaned up. Since I wasn't able to spot any
follow-up regarding this matter, I'm wondering what the plan here is. I'm
asking because we have got customers reporting inconsistencies mainly caused
by the use of num_physpages where totalram_pages would really be more
appropriate (other than in your mail, where you mainly point at the networking
code, this is with the determination of the number of files the system will use
without admin involvement), due to large sparse areas in the physical address
map.

Unfortunately, the meaning of num_physpages seems to also vary between
architectures (some treat it as being the same as totalram_pages) as well as
subsystems (memory hotplug increments/decrements num_physpages along
with totalram_pages).

Thanks, Jan

Re: num_physpages vs. totalram_pages

[Rusty Russell]

On Fri, 5 Jun 2009 10:54:35 pm Jan Beulich wrote:
> Rusty,
>
> in a mail from beginning of last year you indicated that the confusion
> between these two variables should be cleaned up. Since I wasn't able to
> spot any follow-up regarding this matter, I'm wondering what the plan here
> is. I'm asking because we have got customers reporting inconsistencies
> mainly caused by the use of num_physpages where totalram_pages would really
> be more appropriate (other than in your mail, where you mainly point at the
> networking code, this is with the determination of the number of files the
> system will use without admin involvement), due to large sparse areas in
> the physical address map.
>
> Unfortunately, the meaning of num_physpages seems to also vary between
> architectures (some treat it as being the same as totalram_pages) as well
> as subsystems (memory hotplug increments/decrements num_physpages along
> with totalram_pages).

Unfortunately, I have not even thought about it between then and now.  I look 
forward to your proposal!

Rusty.