Message ID | 20170413051655.26113-1-bsingharora@gmail.com (mailing list archive) |
---|---|
State | Changes Requested |
Headers | show |
Balbir Singh <bsingharora@gmail.com> writes: > Do the checks that __flush_tlb_pending() does and check if > a local flush will do when batch->active is false inside of > hpte_need_flush(). I've checked the changes with tlbie tracing, > I see local flushes as applicable now and I've also run > some basic ltp testcases on top of these changes on a powernv > machine. > > Signed-off-by: Balbir Singh <bsingharora@gmail.com> > --- > arch/powerpc/mm/tlb_hash64.c | 13 +++++++++---- > 1 file changed, 9 insertions(+), 4 deletions(-) > > diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c > index 4517aa4..1268e3a 100644 > --- a/arch/powerpc/mm/tlb_hash64.c > +++ b/arch/powerpc/mm/tlb_hash64.c > @@ -93,12 +93,17 @@ void hpte_need_flush(struct mm_struct *mm, unsigned long addr, > > /* > * Check if we have an active batch on this CPU. If not, just > - * flush now and return. For now, we don global invalidates > - * in that case, might be worth testing the mm cpu mask though > - * and decide to use local invalidates instead... > + * flush now and return. > */ > if (!batch->active) { > - flush_hash_page(vpn, rpte, psize, ssize, 0); > + const struct cpumask *tmp; > + int local = 0; > + > + tmp = cpumask_of(smp_processor_id()); > + if (cpumask_equal(mm_cpumask(mm), tmp)) > + local = 1; > + > + flush_hash_page(vpn, rpte, psize, ssize, local); Can you use mm_is_thread_local(). Also switch similar open coded paths when doing that ? > put_cpu_var(ppc64_tlb_batch); > return; > } > -- > 2.9.3 -aneesh
diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c index 4517aa4..1268e3a 100644 --- a/arch/powerpc/mm/tlb_hash64.c +++ b/arch/powerpc/mm/tlb_hash64.c @@ -93,12 +93,17 @@ void hpte_need_flush(struct mm_struct *mm, unsigned long addr, /* * Check if we have an active batch on this CPU. If not, just - * flush now and return. For now, we don global invalidates - * in that case, might be worth testing the mm cpu mask though - * and decide to use local invalidates instead... + * flush now and return. */ if (!batch->active) { - flush_hash_page(vpn, rpte, psize, ssize, 0); + const struct cpumask *tmp; + int local = 0; + + tmp = cpumask_of(smp_processor_id()); + if (cpumask_equal(mm_cpumask(mm), tmp)) + local = 1; + + flush_hash_page(vpn, rpte, psize, ssize, local); put_cpu_var(ppc64_tlb_batch); return; }
Do the checks that __flush_tlb_pending() does and check if a local flush will do when batch->active is false inside of hpte_need_flush(). I've checked the changes with tlbie tracing, I see local flushes as applicable now and I've also run some basic ltp testcases on top of these changes on a powernv machine. Signed-off-by: Balbir Singh <bsingharora@gmail.com> --- arch/powerpc/mm/tlb_hash64.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)