Re: [PATCH] smaps: Report correct page sizes with THP

[Andi Kleen <ak@linux.intel.com>]

On Mon, Feb 09, 2026 at 08:04:38PM +0000, Matthew Wilcox wrote:
> On Mon, Feb 09, 2026 at 11:32:23AM -0800, Andi Kleen wrote:
> > +	} else {
> > +		unsigned ps = vma_mmu_pagesize(vma);
> > +		/* Will need adjustments when more THP page sizes are added. */
> > +		SEQ_PUT_DEC(" kB\nMMUPageSize:    ", ps);
> > +		if (mss.shmem_thp + mss.file_thp + mss.anonymous_thp > 0 &&
> > +		    ps != HPAGE_PMD_SIZE)
> > +			SEQ_PUT_DEC(" kB\nMMUPageSize2:   ", HPAGE_PMD_SIZE);
> > +	}
> 
> I'm not a fan of adding support for just two page sizes when we already
> know that we need to support many.  Particularly not with such an
> uninformative name as "MMUPageSize2".

What is uninformative about it? 

I intentionally used the number to make it extensible for the future,
you can add MMUPageSize3 and 4 and beyond, although the current code doesn't
implement it.

> 
> Something like MMUOtherPageSizes: 64Kib,256KiB,2MiB would work for me.
> But maybe other people have better ideas.

I considered just adding the numbers to a single line (the existing one),
but the risk of breaking some existing parser seemed too high. No other entry
in smaps has multiple numbers, and it seems to be against standard /proc
conventions.

-Andi