From: Barry Song <21cnbao@gmail.com>
To: yosryahmed@google.com
Cc: 21cnbao@gmail.com, akpm@linux-foundation.org,
chengming.zhou@linux.dev, david@redhat.com, hanchuanhua@oppo.com,
hannes@cmpxchg.org, hughd@google.com, kernel-team@meta.com,
linux-kernel@vger.kernel.org, linux-mm@kvack.org,
nphamcs@gmail.com, shakeel.butt@linux.dev,
usamaarif642@gmail.com, willy@infradead.org,
ying.huang@intel.com, Barry Song <v-songbaohua@oppo.com>
Subject: Re: [PATCH v4 1/2] mm: store zero pages to be swapped out in a bitmap
Date: Thu, 5 Sep 2024 12:29:26 +1200 [thread overview]
Message-ID: <20240905002926.1055-1-21cnbao@gmail.com> (raw)
In-Reply-To: <CAJD7tkYK3ecG7cuJcxSF-cw-0K=JPnoE3L4Y=7xUp_R8apqueg@mail.gmail.com>
On Thu, Sep 5, 2024 at 11:57 AM Yosry Ahmed <yosryahmed@google.com> wrote:
>
> [..]
> > well. i feel i have a much cheaper way to implement this, which
> > can entirely iteration even in your original code:
> >
> > +/*
> > + * Return the number of entries which are zero-filled according to
> > + * swap_info_struct->zeromap. It isn't precise if the return value
> > + * is 1 for nr > 1. In this case, it means entries have inconsistent
> > + * zeromap.
> > + */
> > +static inline unsigned int swap_zeromap_entries_count(swp_entry_t
> > entry, int nr)
>
> FWIW I am not really a fan of the count() function not returning an
> actual count. I think an enum with three states is more appropriate
> here, and renaming the function to swap_zeromap_entries_check() or
> similar.
>
Make sense to me, what about the below?
From 24228a1e8426b8b05711a5efcfaae70abeb012c4 Mon Sep 17 00:00:00 2001
From: Barry Song <v-songbaohua@oppo.com>
Date: Thu, 5 Sep 2024 11:56:03 +1200
Subject: [PATCH] mm: fix handling zero for large folios with partial zeromap
There could be a corner case where the first entry is non-zeromap,
but a subsequent entry is zeromap. In this case, we should not
return false.
Additionally, the iteration of test_bit() is unnecessary and
can be replaced with bitmap operations, which are more efficient.
Since swap_read_folio() can't handle reading a large folio that's
partially zeromap and partially non-zeromap, we've moved the code
to mm/swap.h so that others, like those working on swap-in, can
access it.
Fixes: 0ca0c24e3211 ("mm: store zero pages to be swapped out in a bitmap")
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
---
mm/page_io.c | 27 ++++-----------------------
mm/swap.h | 29 +++++++++++++++++++++++++++++
2 files changed, 33 insertions(+), 23 deletions(-)
diff --git a/mm/page_io.c b/mm/page_io.c
index 4bc77d1c6bfa..46907c9dd20b 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -226,26 +226,6 @@ static void swap_zeromap_folio_clear(struct folio *folio)
}
}
-/*
- * Return the index of the first subpage which is not zero-filled
- * according to swap_info_struct->zeromap.
- * If all pages are zero-filled according to zeromap, it will return
- * folio_nr_pages(folio).
- */
-static unsigned int swap_zeromap_folio_test(struct folio *folio)
-{
- struct swap_info_struct *sis = swp_swap_info(folio->swap);
- swp_entry_t entry;
- unsigned int i;
-
- for (i = 0; i < folio_nr_pages(folio); i++) {
- entry = page_swap_entry(folio_page(folio, i));
- if (!test_bit(swp_offset(entry), sis->zeromap))
- return i;
- }
- return i;
-}
-
/*
* We may have stale swap cache pages in memory: notice
* them here and get rid of the unnecessary final write.
@@ -524,9 +504,10 @@ static void sio_read_complete(struct kiocb *iocb, long ret)
static bool swap_read_folio_zeromap(struct folio *folio)
{
- unsigned int idx = swap_zeromap_folio_test(folio);
+ unsigned int nr_pages = folio_nr_pages(folio);
+ zeromap_stat_t stat = swap_zeromap_entries_check(folio->swap, nr_pages);
- if (idx == 0)
+ if (stat == SWAP_ZEROMAP_NON)
return false;
/*
@@ -534,7 +515,7 @@ static bool swap_read_folio_zeromap(struct folio *folio)
* currently handled. Return true without marking the folio uptodate so
* that an IO error is emitted (e.g. do_swap_page() will sigbus).
*/
- if (WARN_ON_ONCE(idx < folio_nr_pages(folio)))
+ if (WARN_ON_ONCE(stat == SWAP_ZEROMAP_PARTIAL))
return true;
folio_zero_range(folio, 0, folio_size(folio));
diff --git a/mm/swap.h b/mm/swap.h
index f8711ff82f84..f8e3fa061c1d 100644
--- a/mm/swap.h
+++ b/mm/swap.h
@@ -4,6 +4,12 @@
struct mempolicy;
+typedef enum {
+ SWAP_ZEROMAP_NON,
+ SWAP_ZEROMAP_FULL,
+ SWAP_ZEROMAP_PARTIAL
+} zeromap_stat_t;
+
#ifdef CONFIG_SWAP
#include <linux/swapops.h> /* for swp_offset */
#include <linux/blk_types.h> /* for bio_end_io_t */
@@ -80,6 +86,24 @@ static inline unsigned int folio_swap_flags(struct folio *folio)
{
return swp_swap_info(folio->swap)->flags;
}
+
+/*
+ * Check if nr entries are all zeromap, non-zeromap or partially zeromap
+ */
+static inline zeromap_stat_t swap_zeromap_entries_check(swp_entry_t entry, int nr)
+{
+ struct swap_info_struct *sis = swp_swap_info(entry);
+ unsigned long start = swp_offset(entry);
+ unsigned long end = start + nr;
+
+ if (find_next_bit(sis->zeromap, end, start) == end)
+ return SWAP_ZEROMAP_NON;
+ if (find_next_zero_bit(sis->zeromap, end, start) == end)
+ return SWAP_ZEROMAP_FULL;
+
+ return SWAP_ZEROMAP_PARTIAL;
+}
+
#else /* CONFIG_SWAP */
struct swap_iocb;
static inline void swap_read_folio(struct folio *folio, struct swap_iocb **plug)
@@ -171,6 +195,11 @@ static inline unsigned int folio_swap_flags(struct folio *folio)
{
return 0;
}
+
+static inline zeromap_stat_t swap_zeromap_entries_check(swp_entry_t entry, int nr)
+{
+ return SWAP_ZEROMAP_NONE;
+}
#endif /* CONFIG_SWAP */
#endif /* _MM_SWAP_H */
--
2.34.1
next prev parent reply other threads:[~2024-09-05 0:29 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-06-12 12:43 [PATCH v4 0/2] " Usama Arif
2024-06-12 12:43 ` [PATCH v4 1/2] " Usama Arif
2024-06-12 20:13 ` Yosry Ahmed
2024-06-13 11:37 ` Usama Arif
2024-06-13 16:38 ` Yosry Ahmed
2024-06-13 19:21 ` Usama Arif
2024-06-13 19:26 ` Yosry Ahmed
2024-06-13 19:38 ` Usama Arif
2024-09-04 5:55 ` Barry Song
2024-09-04 7:12 ` Yosry Ahmed
2024-09-04 7:17 ` Barry Song
2024-09-04 7:22 ` Yosry Ahmed
2024-09-04 7:54 ` Barry Song
2024-09-04 17:40 ` Yosry Ahmed
2024-09-05 7:03 ` Barry Song
2024-09-05 7:55 ` Yosry Ahmed
2024-09-05 8:49 ` Barry Song
2024-09-05 10:10 ` Barry Song
2024-09-05 10:33 ` Barry Song
2024-09-05 10:53 ` Usama Arif
2024-09-05 11:00 ` Barry Song
2024-09-05 19:19 ` Usama Arif
2024-09-05 17:36 ` Yosry Ahmed
2024-09-05 19:28 ` Yosry Ahmed
2024-09-06 10:22 ` Barry Song
2024-09-05 10:37 ` Usama Arif
2024-09-05 10:42 ` Barry Song
2024-09-05 10:50 ` Usama Arif
2024-09-04 11:14 ` Usama Arif
2024-09-04 23:44 ` Barry Song
2024-09-04 23:47 ` Barry Song
2024-09-04 23:57 ` Yosry Ahmed
2024-09-05 0:29 ` Barry Song [this message]
2024-09-05 7:38 ` Yosry Ahmed
2024-06-12 12:43 ` [PATCH v4 2/2] mm: remove code to handle same filled pages Usama Arif
2024-06-12 15:09 ` Nhat Pham
2024-06-12 16:34 ` Usama Arif
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240905002926.1055-1-21cnbao@gmail.com \
--to=21cnbao@gmail.com \
--cc=akpm@linux-foundation.org \
--cc=chengming.zhou@linux.dev \
--cc=david@redhat.com \
--cc=hanchuanhua@oppo.com \
--cc=hannes@cmpxchg.org \
--cc=hughd@google.com \
--cc=kernel-team@meta.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=nphamcs@gmail.com \
--cc=shakeel.butt@linux.dev \
--cc=usamaarif642@gmail.com \
--cc=v-songbaohua@oppo.com \
--cc=willy@infradead.org \
--cc=ying.huang@intel.com \
--cc=yosryahmed@google.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox