On 21 Mar 2022, at 15:54, Matthew Wilcox wrote:
On Mon, Mar 21, 2022 at 03:07:46PM -0400, Zi Yan wrote:
Yes. Will change it to new_nr to be consistent.
uh, you're going to call ilog2?
fortunately, no. Inside split_page_memcg(), I probably need to add VM_BUG_ON(nr % new_nr != 0) to make sure new_nr is a divisor of nr, since there are a couple of nr / new_nr operations. Otherwise, new_nr works.
I think this would look less inconsistent if 'nr' were an unsigned long (how long until we need 16GB pages? Think PPC already supports those)
-- Best Regards, Yan, Zi