Re: [PATCH V3] powerpc/mm: Fix Multi hit ERAT cause by recent THP update

Balbir Singh [off-list ref] writes:

quoted

Now we can't depend for mm_cpumask, a parallel find_linux_pte_hugepte
can happen outside that. Now i had a variant for kick_all_cpus_sync that
ignored idle cpus. But then that needs more verification.
=20
http://article.gmane.org/gmane.linux.ports.ppc.embedded/81105

Can be racy as a CPU moves from non-idle to idle

In

quoted

+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0pmd_hugepage_update(vma->vm_mm, address, =

pmdp, ~0UL, 0);

quoted

+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0/*
+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0* This ensures that generic code th=

at rely on IRQ disabling

quoted

+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0* to prevent a parallel THP split w=

ork as expected.

quoted

+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0*/
+=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0kick_all_cpus_sync();

pmdp_invalidate()->pmd_hugepage_update() can still run in parallel with=

=C2=A0

find_linux_pte_or_hugepte() and race.. Am I missing something?

Yes. But then we make sure that the pte_t returned by
find_linux_pte_or_hugepte doesn't change to a regular pmd entry by using
that kick. Now callers of find_lnux_pte_or_hugepte will check for
_PAGE_PRESENT. So if it called before
pmd_hugepage_update(_PAGE_PRESENT), we wait for the caller to finish the
usage (via kick()). Or they bail out after finding _PAGE_PRESENT cleared.

-aneesh