Re: [PATCH 2/3] mm/huge_memory.c: update tlb entry if pmd is changed

[Aneesh Kumar K.V]

On 6/25/20 10:16 PM, Mike Kravetz wrote:
> On 6/25/20 5:01 AM, Aneesh Kumar K.V wrote:
> > Mike Kravetz <mike.kravetz@xxxxxxxxxx> writes:
> >
> > > On 6/24/20 2:26 AM, Bibo Mao wrote:
> > > > When set_pmd_at is called in function do_huge_pmd_anonymous_page,
> > > > new tlb entry can be added by software on MIPS platform.
> > > >
> > > > Here add update_mmu_cache_pmd when pmd entry is set, and
> > > > update_mmu_cache_pmd is defined as empty excepts arc/mips platform.
> > > > This patch has no negative effect on other platforms except arc/mips
> > > > system.
> > >
> > > I am confused by this comment.  It appears that update_mmu_cache_pmd
> > > is defined as non-empty on arc, mips, powerpc and sparc architectures.
> > > Am I missing something?
> > >
> > > If those architectures do provide update_mmu_cache_pmd, then the previous
> > > patch and this one now call update_mmu_cache_pmd with the actual faulting
> > > address instead of the huge page aligned address.  This was intentional
> > > for mips.  However, are there any potential issues on the other architectures?
> > > I am no expert in any of those architectures.  arc looks like it could be
> > > problematic as update_mmu_cache_pmd calls update_mmu_cache and then
> > > operates on (address & PAGE_MASK).  That could now be different.
> >
> > Also we added update_mmu_cache_pmd to update a THP entry. That could be
> > different from a hugetlb entry on some architectures. If we need to do
> > hugetlb equivalent for update_mmu_cache, we should add a different
> > function.
>
> I do not know the mips architecture well enough or if the motivation for
> this patch was based on THP or hugetlb pages.  However, it will change
> the address passed to update_mmu_cache_pmd from huge page aligned to the
> actual faulting address.  Will such a change in the passed address impact
> the powerpc update_mmu_cache_pmd routine?

Right now powerpc update_mmu_cache_pmd() is a dummy function. But I 
agree we should audit arch to make sure such a change can work with 
architectures. My comment was related to the fact that mmu cache update 
w.r.t THP and hugetlb can be different on some platforms. So we may
want to avoid using the same function for both.

-aneesh