2021-11-10 08:32:09

by Peter Xu

[permalink] [raw]
Subject: [PATCH RFC 2/2] mm: Rework swap handling of zap_pte_range

Clean the code up by merging the device private/exclusive swap entry handling
with the rest, then we merge the pte clear operation too.

struct* page is defined in multiple places in the function, move it upward.

free_swap_and_cache() is only useful for !non_swap_entry() case, put it into
the condition.

No functional change intended.

Signed-off-by: Peter Xu <[email protected]>
---
mm/memory.c | 25 ++++++++-----------------
1 file changed, 8 insertions(+), 17 deletions(-)

diff --git a/mm/memory.c b/mm/memory.c
index e454f3c6aeb9..e5d59a6b6479 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1326,6 +1326,8 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
arch_enter_lazy_mmu_mode();
do {
pte_t ptent = *pte;
+ struct page *page;
+
if (pte_none(ptent))
continue;

@@ -1333,8 +1335,6 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
break;

if (pte_present(ptent)) {
- struct page *page;
-
page = vm_normal_page(vma, addr, ptent);
if (unlikely(zap_skip_check_mapping(details, page)))
continue;
@@ -1368,32 +1368,23 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
entry = pte_to_swp_entry(ptent);
if (is_device_private_entry(entry) ||
is_device_exclusive_entry(entry)) {
- struct page *page = pfn_swap_entry_to_page(entry);
-
+ page = pfn_swap_entry_to_page(entry);
if (unlikely(zap_skip_check_mapping(details, page)))
continue;
- pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
rss[mm_counter(page)]--;
-
if (is_device_private_entry(entry))
page_remove_rmap(page, false);
-
put_page(page);
- continue;
- }
-
- if (!non_swap_entry(entry))
- rss[MM_SWAPENTS]--;
- else if (is_migration_entry(entry)) {
- struct page *page;
-
+ } else if (is_migration_entry(entry)) {
page = pfn_swap_entry_to_page(entry);
if (unlikely(zap_skip_check_mapping(details, page)))
continue;
rss[mm_counter(page)]--;
+ } else if (!non_swap_entry(entry)) {
+ rss[MM_SWAPENTS]--;
+ if (unlikely(!free_swap_and_cache(entry)))
+ print_bad_pte(vma, addr, ptent, NULL);
}
- if (unlikely(!free_swap_and_cache(entry)))
- print_bad_pte(vma, addr, ptent, NULL);
pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
} while (pte++, addr += PAGE_SIZE, addr != end);

--
2.32.0


2021-11-15 11:22:01

by Alistair Popple

[permalink] [raw]
Subject: Re: [PATCH RFC 2/2] mm: Rework swap handling of zap_pte_range

Hi Peter,

I was having trouble applying this cleanly to any of my local trees so was
wondering which sha1 should I be applying this on top of? Thanks.

- Alistair

On Wednesday, 10 November 2021 7:29:52 PM AEDT Peter Xu wrote:
> Clean the code up by merging the device private/exclusive swap entry handling
> with the rest, then we merge the pte clear operation too.
>
> struct* page is defined in multiple places in the function, move it upward.
>
> free_swap_and_cache() is only useful for !non_swap_entry() case, put it into
> the condition.
>
> No functional change intended.
>
> Signed-off-by: Peter Xu <[email protected]>
> ---
> mm/memory.c | 25 ++++++++-----------------
> 1 file changed, 8 insertions(+), 17 deletions(-)
>
> diff --git a/mm/memory.c b/mm/memory.c
> index e454f3c6aeb9..e5d59a6b6479 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -1326,6 +1326,8 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
> arch_enter_lazy_mmu_mode();
> do {
> pte_t ptent = *pte;
> + struct page *page;
> +
> if (pte_none(ptent))
> continue;
>
> @@ -1333,8 +1335,6 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
> break;
>
> if (pte_present(ptent)) {
> - struct page *page;
> -
> page = vm_normal_page(vma, addr, ptent);
> if (unlikely(zap_skip_check_mapping(details, page)))
> continue;
> @@ -1368,32 +1368,23 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
> entry = pte_to_swp_entry(ptent);
> if (is_device_private_entry(entry) ||
> is_device_exclusive_entry(entry)) {
> - struct page *page = pfn_swap_entry_to_page(entry);
> -
> + page = pfn_swap_entry_to_page(entry);
> if (unlikely(zap_skip_check_mapping(details, page)))
> continue;
> - pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
> rss[mm_counter(page)]--;
> -
> if (is_device_private_entry(entry))
> page_remove_rmap(page, false);
> -
> put_page(page);
> - continue;
> - }
> -
> - if (!non_swap_entry(entry))
> - rss[MM_SWAPENTS]--;
> - else if (is_migration_entry(entry)) {
> - struct page *page;
> -
> + } else if (is_migration_entry(entry)) {
> page = pfn_swap_entry_to_page(entry);
> if (unlikely(zap_skip_check_mapping(details, page)))
> continue;
> rss[mm_counter(page)]--;
> + } else if (!non_swap_entry(entry)) {
> + rss[MM_SWAPENTS]--;
> + if (unlikely(!free_swap_and_cache(entry)))
> + print_bad_pte(vma, addr, ptent, NULL);
> }
> - if (unlikely(!free_swap_and_cache(entry)))
> - print_bad_pte(vma, addr, ptent, NULL);
> pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
> } while (pte++, addr += PAGE_SIZE, addr != end);
>
>





2021-11-15 11:37:42

by Peter Xu

[permalink] [raw]
Subject: Re: [PATCH RFC 2/2] mm: Rework swap handling of zap_pte_range

On Mon, Nov 15, 2021 at 10:21:18PM +1100, Alistair Popple wrote:
> Hi Peter,

Hi, Alistair,

>
> I was having trouble applying this cleanly to any of my local trees so was
> wondering which sha1 should I be applying this on top of? Thanks.

Thanks for considering trying it out. I thought it was easy to apply onto any
of the recent branches as long as with -mm's rc1 applied, and I just did it to
Linus's 5.16-rc1 in my uffd-wp rebase:

https://github.com/xzpeter/linux/commits/uffd-wp-shmem-hugetlbfs

This commit is here:

https://github.com/xzpeter/linux/commit/c32043436282bb352e6fe10eb5fa693340fe5281

It could be that "git rebase" is normally smarter so I didn't notice it's not
applicable directly.

I'll repost a new version soon, please also consider to fetch directly from the
git tree too before I do so.

Thanks,

--
Peter Xu