From: "Huang, Ying" <ying.huang@intel.com>
To: Khalid Aziz <khalid.aziz@oracle.com>
Cc: akpm@linux-foundation.org, willy@infradead.org,
steven.sistare@oracle.com, david@redhat.com,
mgorman@techsingularity.net, baolin.wang@linux.alibaba.com,
linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Khalid Aziz <khalid@kernel.org>
Subject: Re: [PATCH v4] mm, compaction: Skip all non-migratable pages during scan
Date: Mon, 29 May 2023 11:01:40 +0800 [thread overview]
Message-ID: <87ttvvx2ln.fsf@yhuang6-desk2.ccr.corp.intel.com> (raw)
In-Reply-To: <20230525191507.160076-1-khalid.aziz@oracle.com> (Khalid Aziz's message of "Thu, 25 May 2023 13:15:07 -0600")
Khalid Aziz <khalid.aziz@oracle.com> writes:
> Pages pinned in memory through extra refcounts can not be migrated.
> Currently as isolate_migratepages_block() scans pages for
> compaction, it skips any pinned anonymous pages. All non-migratable
> pages should be skipped and not just the anonymous pinned pages.
> This patch adds a check for extra refcounts on a page to determine
> if the page can be migrated. This was seen as a real issue on a
> customer workload where a large number of pages were pinned by vfio
> on the host and any attempts to allocate hugepages resulted in
> significant amount of cpu time spent in either direct compaction or
> in kcompactd scanning vfio pinned pages over and over again that can
> not be migrated. These are the changes in relevant stats with this
> patch for a test run of this scenario:
>
> Before After
> compact_migrate_scanned 329,798,858 370,984,387
> compact_free_scanned 40,478,406 25,843,262
> compact_isolated 135,470,452 777,235
> pgmigrate_success 544,255 507,325
> pgmigrate_fail 134,616,282 47
> kcompactd CPU time 5:12.81 0:12.28
>
> Before the patch, large number of pages were isolated but most of
> them failed to migrate.
>
> Signed-off-by: Khalid Aziz <khalid.aziz@oracle.com>
> Suggested-by: Steve Sistare <steven.sistare@oracle.com>
> Cc: Khalid Aziz <khalid@kernel.org>
> ---
> v4:
> - Use existing folio_expected_refs() function (Suggested
> by Huang, Ying)
> - Use folio functions
> - Take into account contig allocations when checking for
> long temr pinning and skip pages in ZONE_MOVABLE and
> MIGRATE_CMA type pages (Suggested by David Hildenbrand)
> - Use folio version of total_mapcount() instead of
> page_mapcount() (Suggested by Baolin Wang)
>
> v3:
> - Account for extra ref added by get_page_unless_zero() earlier
> in isolate_migratepages_block() (Suggested by Huang, Ying)
> - Clean up computation of extra refs to be consistent
> (Suggested by Huang, Ying)
>
> v2:
> - Update comments in the code (Suggested by Andrew)
> - Use PagePrivate() instead of page_has_private() (Suggested
> by Matthew)
> - Pass mapping to page_has_extrarefs() (Suggested by Matthew)
> - Use page_ref_count() (Suggested by Matthew)
> - Rename is_pinned_page() to reflect its function more
> accurately (Suggested by Matthew)
>
> include/linux/migrate.h | 16 +++++++++++++++
> mm/compaction.c | 44 +++++++++++++++++++++++++++++++++++++----
> mm/migrate.c | 14 -------------
> 3 files changed, 56 insertions(+), 18 deletions(-)
>
> diff --git a/include/linux/migrate.h b/include/linux/migrate.h
> index 6241a1596a75..4f59e15eae99 100644
> --- a/include/linux/migrate.h
> +++ b/include/linux/migrate.h
> @@ -141,6 +141,22 @@ const struct movable_operations *page_movable_ops(struct page *page)
> ((unsigned long)page->mapping - PAGE_MAPPING_MOVABLE);
> }
>
> +static inline
> +int folio_expected_refs(struct address_space *mapping,
> + struct folio *folio)
I don't think that it's necessary to make this function inline. It
isn't called in hot path.
> +{
> + int refs = 1;
> +
> + if (!mapping)
> + return refs;
> +
> + refs += folio_nr_pages(folio);
> + if (folio_test_private(folio))
> + refs++;
> +
> + return refs;
> +}
> +
> #ifdef CONFIG_NUMA_BALANCING
> int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
> int node);
> diff --git a/mm/compaction.c b/mm/compaction.c
> index 5a9501e0ae01..b548e05f0349 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -764,6 +764,42 @@ static bool too_many_isolated(pg_data_t *pgdat)
> return too_many;
> }
>
> +/*
> + * Check if this base page should be skipped from isolation because
> + * it has extra refcounts that will prevent it from being migrated.
> + * This code is inspired by similar code in migrate_vma_check_page(),
> + * can_split_folio() and folio_migrate_mapping()
> + */
> +static inline bool page_has_extra_refs(struct page *page,
> + struct address_space *mapping)
> +{
> + unsigned long extra_refs;
s/extra_refs/expected_refs/
?
> + struct folio *folio;
> +
> + /*
> + * Skip this check for pages in ZONE_MOVABLE or MIGRATE_CMA
> + * pages that can not be long term pinned
> + */
> + if (is_zone_movable_page(page) || is_migrate_cma_page(page))
> + return false;
I suggest to move these 2 checks out to the place before calling the
function. Or change the name of the function.
> +
> + folio = page_folio(page);
> +
> + /*
> + * caller holds a ref already from get_page_unless_zero()
> + * which is accounted for in folio_expected_refs()
> + */
> + extra_refs = folio_expected_refs(mapping, folio);
> +
> + /*
> + * This is an admittedly racy check but good enough to determine
> + * if a page is pinned and can not be migrated
> + */
> + if ((folio_ref_count(folio) - extra_refs) > folio_mapcount(folio))
> + return true;
> + return false;
> +}
> +
> /**
> * isolate_migratepages_block() - isolate all migrate-able pages within
> * a single pageblock
> @@ -992,12 +1028,12 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
> goto isolate_fail;
>
> /*
> - * Migration will fail if an anonymous page is pinned in memory,
> - * so avoid taking lru_lock and isolating it unnecessarily in an
> - * admittedly racy check.
> + * Migration will fail if a page has extra refcounts
> + * from long term pinning preventing it from migrating,
> + * so avoid taking lru_lock and isolating it unnecessarily.
> */
> mapping = page_mapping(page);
> - if (!mapping && (page_count(page) - 1) > total_mapcount(page))
> + if (!cc->alloc_contig && page_has_extra_refs(page, mapping))
> goto isolate_fail_put;
>
> /*
> diff --git a/mm/migrate.c b/mm/migrate.c
> index db3f154446af..a2f3e5834996 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -385,20 +385,6 @@ void pmd_migration_entry_wait(struct mm_struct *mm, pmd_t *pmd)
> }
> #endif
>
> -static int folio_expected_refs(struct address_space *mapping,
> - struct folio *folio)
> -{
> - int refs = 1;
> - if (!mapping)
> - return refs;
> -
> - refs += folio_nr_pages(folio);
> - if (folio_test_private(folio))
> - refs++;
> -
> - return refs;
> -}
> -
> /*
> * Replace the page in the mapping.
> *
Best Regards,
Huang, Ying
prev parent reply other threads:[~2023-05-29 3:02 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-05-25 19:15 [PATCH v4] mm, compaction: Skip all non-migratable pages during scan Khalid Aziz
2023-05-25 19:58 ` Matthew Wilcox
2023-05-25 20:15 ` Steven Sistare
2023-05-25 20:45 ` Matthew Wilcox
2023-05-25 21:31 ` Matthew Wilcox
2023-05-26 15:44 ` Khalid Aziz
2023-05-26 16:44 ` Matthew Wilcox
2023-05-26 16:46 ` David Hildenbrand
2023-05-26 18:46 ` Matthew Wilcox
2023-05-26 18:50 ` David Hildenbrand
2023-05-27 2:11 ` John Hubbard
2023-05-27 3:18 ` Matthew Wilcox
2023-05-28 23:49 ` John Hubbard
2023-05-29 0:31 ` Matthew Wilcox
2023-05-29 9:25 ` David Hildenbrand
2023-05-30 15:42 ` Khalid Aziz
2023-06-09 22:11 ` Andrew Morton
2023-06-09 23:28 ` Khalid Aziz
2023-05-25 20:41 ` Khalid Aziz
2023-05-29 3:01 ` Huang, Ying [this message]
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=87ttvvx2ln.fsf@yhuang6-desk2.ccr.corp.intel.com \
--to=ying.huang@intel.com \
--cc=akpm@linux-foundation.org \
--cc=baolin.wang@linux.alibaba.com \
--cc=david@redhat.com \
--cc=khalid.aziz@oracle.com \
--cc=khalid@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@techsingularity.net \
--cc=steven.sistare@oracle.com \
--cc=willy@infradead.org \
/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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).