* [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic
@ 2021-03-17 6:47 Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 2/3] mm: add page_cpupid_reset_last atomic variant Nicholas Piggin
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Nicholas Piggin @ 2021-03-17 6:47 UTC (permalink / raw)
To: linux-mm; +Cc: Nicholas Piggin, Peter Zijlstra
Add a __ prefix to indicate that it should be used in non-atomic
situations (where the page is not subject to concurrent flags
access, following the pattern of the other page flags modifications).
This prepares for an atomic version.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
Rebase only. Any comments on this idea?
Thanks,
Nick
include/linux/mm.h | 6 +++---
mm/page_alloc.c | 4 ++--
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 64a71bf20536..086ab710383f 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1396,7 +1396,7 @@ static inline int page_cpupid_last(struct page *page)
{
return page->_last_cpupid;
}
-static inline void page_cpupid_reset_last(struct page *page)
+static inline void __page_cpupid_reset_last(struct page *page)
{
page->_last_cpupid = -1 & LAST_CPUPID_MASK;
}
@@ -1408,7 +1408,7 @@ static inline int page_cpupid_last(struct page *page)
extern int page_cpupid_xchg_last(struct page *page, int cpupid);
-static inline void page_cpupid_reset_last(struct page *page)
+static inline void __page_cpupid_reset_last(struct page *page)
{
page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
}
@@ -1449,7 +1449,7 @@ static inline bool cpupid_pid_unset(int cpupid)
return true;
}
-static inline void page_cpupid_reset_last(struct page *page)
+static inline void __page_cpupid_reset_last(struct page *page)
{
}
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index cfc72873961d..172368357e73 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1266,7 +1266,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
if (bad)
return false;
- page_cpupid_reset_last(page);
+ __page_cpupid_reset_last(page);
page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
reset_page_owner(page, order);
@@ -1463,7 +1463,7 @@ static void __meminit __init_single_page(struct page *page, unsigned long pfn,
set_page_links(page, zone, nid, pfn);
init_page_count(page);
page_mapcount_reset(page);
- page_cpupid_reset_last(page);
+ __page_cpupid_reset_last(page);
page_kasan_tag_reset(page);
INIT_LIST_HEAD(&page->lru);
--
2.23.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v2 2/3] mm: add page_cpupid_reset_last atomic variant
2021-03-17 6:47 [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
@ 2021-03-17 6:47 ` Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 3/3] mm: skip page_cpupid_reset_last atomic update if the value is already reset Nicholas Piggin
2021-06-01 5:06 ` [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2 siblings, 0 replies; 6+ messages in thread
From: Nicholas Piggin @ 2021-03-17 6:47 UTC (permalink / raw)
To: linux-mm; +Cc: Nicholas Piggin, Peter Zijlstra
This can be used when there is concurrency on the struct page. If cpupid
is not in page flags, it can use a store.
Use this to replace the open-coded reset in wp_page_reuse.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
include/linux/mm.h | 15 +++++++++++++++
mm/memory.c | 2 +-
2 files changed, 16 insertions(+), 1 deletion(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 086ab710383f..2524553ea898 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1400,6 +1400,11 @@ static inline void __page_cpupid_reset_last(struct page *page)
{
page->_last_cpupid = -1 & LAST_CPUPID_MASK;
}
+
+static inline void page_cpupid_reset_last(struct page *page)
+{
+ WRITE_ONCE(page->_last_cpupid, -1 & LAST_CPUPID_MASK);
+}
#else
static inline int page_cpupid_last(struct page *page)
{
@@ -1412,6 +1417,12 @@ static inline void __page_cpupid_reset_last(struct page *page)
{
page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
}
+
+static inline void page_cpupid_reset_last(struct page *page)
+{
+ page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);
+}
+
#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
#else /* !CONFIG_NUMA_BALANCING */
static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
@@ -1453,6 +1464,10 @@ static inline void __page_cpupid_reset_last(struct page *page)
{
}
+static inline void page_cpupid_reset_last(struct page *page)
+{
+}
+
static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
{
return false;
diff --git a/mm/memory.c b/mm/memory.c
index 5efa07fb6cdc..bb37dc0f4dac 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2804,7 +2804,7 @@ static inline void wp_page_reuse(struct vm_fault *vmf)
* unrelated process.
*/
if (page)
- page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);
+ page_cpupid_reset_last(page);
flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
entry = pte_mkyoung(vmf->orig_pte);
--
2.23.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH v2 3/3] mm: skip page_cpupid_reset_last atomic update if the value is already reset
2021-03-17 6:47 [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 2/3] mm: add page_cpupid_reset_last atomic variant Nicholas Piggin
@ 2021-03-17 6:47 ` Nicholas Piggin
2021-06-01 5:06 ` [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2 siblings, 0 replies; 6+ messages in thread
From: Nicholas Piggin @ 2021-03-17 6:47 UTC (permalink / raw)
To: linux-mm; +Cc: Nicholas Piggin, Peter Zijlstra
Improves performance of a PROT_READ -> PROT_READ|PROT_WRITE -> store
microbenchmark (which goes via the wp_page_reuse fault path) by about
10% on a POWER9.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
include/linux/mm.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 2524553ea898..1e1471235055 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1420,6 +1420,9 @@ static inline void __page_cpupid_reset_last(struct page *page)
static inline void page_cpupid_reset_last(struct page *page)
{
+ if (page_cpupid_last(page) == (-1 & LAST_CPUPID_MASK))
+ return;
+
page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);
}
--
2.23.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic
2021-03-17 6:47 [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 2/3] mm: add page_cpupid_reset_last atomic variant Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 3/3] mm: skip page_cpupid_reset_last atomic update if the value is already reset Nicholas Piggin
@ 2021-06-01 5:06 ` Nicholas Piggin
2021-06-01 6:42 ` Peter Zijlstra
2 siblings, 1 reply; 6+ messages in thread
From: Nicholas Piggin @ 2021-06-01 5:06 UTC (permalink / raw)
To: linux-mm; +Cc: Peter Zijlstra
Ping? Any concerns about these patches?
Thanks,
Nick
Excerpts from Nicholas Piggin's message of March 17, 2021 4:47 pm:
> Add a __ prefix to indicate that it should be used in non-atomic
> situations (where the page is not subject to concurrent flags
> access, following the pattern of the other page flags modifications).
>
> This prepares for an atomic version.
>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
> Rebase only. Any comments on this idea?
>
> Thanks,
> Nick
>
> include/linux/mm.h | 6 +++---
> mm/page_alloc.c | 4 ++--
> 2 files changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 64a71bf20536..086ab710383f 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -1396,7 +1396,7 @@ static inline int page_cpupid_last(struct page *page)
> {
> return page->_last_cpupid;
> }
> -static inline void page_cpupid_reset_last(struct page *page)
> +static inline void __page_cpupid_reset_last(struct page *page)
> {
> page->_last_cpupid = -1 & LAST_CPUPID_MASK;
> }
> @@ -1408,7 +1408,7 @@ static inline int page_cpupid_last(struct page *page)
>
> extern int page_cpupid_xchg_last(struct page *page, int cpupid);
>
> -static inline void page_cpupid_reset_last(struct page *page)
> +static inline void __page_cpupid_reset_last(struct page *page)
> {
> page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
> }
> @@ -1449,7 +1449,7 @@ static inline bool cpupid_pid_unset(int cpupid)
> return true;
> }
>
> -static inline void page_cpupid_reset_last(struct page *page)
> +static inline void __page_cpupid_reset_last(struct page *page)
> {
> }
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index cfc72873961d..172368357e73 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1266,7 +1266,7 @@ static __always_inline bool free_pages_prepare(struct page *page,
> if (bad)
> return false;
>
> - page_cpupid_reset_last(page);
> + __page_cpupid_reset_last(page);
> page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
> reset_page_owner(page, order);
>
> @@ -1463,7 +1463,7 @@ static void __meminit __init_single_page(struct page *page, unsigned long pfn,
> set_page_links(page, zone, nid, pfn);
> init_page_count(page);
> page_mapcount_reset(page);
> - page_cpupid_reset_last(page);
> + __page_cpupid_reset_last(page);
> page_kasan_tag_reset(page);
>
> INIT_LIST_HEAD(&page->lru);
> --
> 2.23.0
>
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic
2021-06-01 5:06 ` [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
@ 2021-06-01 6:42 ` Peter Zijlstra
2021-06-08 2:18 ` Nicholas Piggin
0 siblings, 1 reply; 6+ messages in thread
From: Peter Zijlstra @ 2021-06-01 6:42 UTC (permalink / raw)
To: Nicholas Piggin; +Cc: linux-mm
On Tue, Jun 01, 2021 at 03:06:01PM +1000, Nicholas Piggin wrote:
> Ping? Any concerns about these patches?
For all three:
Acked-by: Peter Zijlstra (Intel) <peterz@infradead>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic
2021-06-01 6:42 ` Peter Zijlstra
@ 2021-06-08 2:18 ` Nicholas Piggin
0 siblings, 0 replies; 6+ messages in thread
From: Nicholas Piggin @ 2021-06-08 2:18 UTC (permalink / raw)
To: Peter Zijlstra; +Cc: linux-mm
Excerpts from Peter Zijlstra's message of June 1, 2021 4:42 pm:
> On Tue, Jun 01, 2021 at 03:06:01PM +1000, Nicholas Piggin wrote:
>> Ping? Any concerns about these patches?
>
> For all three:
>
> Acked-by: Peter Zijlstra (Intel) <peterz@infradead>
>
Great, thanks for taking a look.
Thanks,
Nick
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-06-08 2:18 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-17 6:47 [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 2/3] mm: add page_cpupid_reset_last atomic variant Nicholas Piggin
2021-03-17 6:47 ` [PATCH v2 3/3] mm: skip page_cpupid_reset_last atomic update if the value is already reset Nicholas Piggin
2021-06-01 5:06 ` [PATCH v2 1/3] mm: add __ prefix to page_cpupid_reset_last because it is non-atomic Nicholas Piggin
2021-06-01 6:42 ` Peter Zijlstra
2021-06-08 2:18 ` Nicholas Piggin
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).