2022-11-16 08:23:53

by Alexander Gordeev

Subject: mm: mmu_gather: do not expose delayed_rmap flag

Flag delayed_rmap of 'struct mmu_gather' is rather
a private member, but it is still accessed directly.
Instead, let the TLB gather code access the flag.

Signed-off-by: Alexander Gordeev <[email protected]>
---
mm/memory.c | 3 +--
mm/mmu_gather.c | 3 +++
2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index 42f10cc1de58..38b58cd07b52 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1465,8 +1465,7 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
/* Do the actual TLB flush before dropping ptl */
if (force_flush) {
tlb_flush_mmu_tlbonly(tlb);
- if (tlb->delayed_rmap)
- tlb_flush_rmaps(tlb, vma);
+ tlb_flush_rmaps(tlb, vma);
}
pte_unmap_unlock(start_pte, ptl);

diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c
index 79de59136cd2..9f22309affee 100644
--- a/mm/mmu_gather.c
+++ b/mm/mmu_gather.c
@@ -60,6 +60,9 @@ void tlb_flush_rmaps(struct mmu_gather *tlb, struct vm_area_struct *vma)
{
struct mmu_gather_batch *batch;

+ if (!tlb->delayed_rmap)
+ return;
+
batch = tlb->active;
for (int i = 0; i < batch->nr; i++) {
struct encoded_page *enc = batch->encoded_pages[i];
--
2.31.1

2022-11-16 18:02:01

by Linus Torvalds

Subject: Re: mm: mmu_gather: do not expose delayed_rmap flag

On Tue, Nov 15, 2022 at 11:49 PM Alexander Gordeev
<[email protected]> wrote:
>
> Flag delayed_rmap of 'struct mmu_gather' is rather
> a private member, but it is still accessed directly.
> Instead, let the TLB gather code access the flag.

Now, I set it up so that if you don't use delayed_rmap, the
tlb_flush_rmaps() function ends up being an empty inline function, and
as such the compiler should already have done this for you - including
optimizing out the test that then doesn't even matter.

But this patch shouldn't *matter*, but it also isn't wrong, so..

Acked-by: Linus Torvalds <[email protected]>

Thanks,

Linus