]> git.baikalelectronics.ru Git - kernel.git/commitdiff
mm, thp: use head page in __migration_entry_wait()
authorXu Yu <xuyu@linux.alibaba.com>
Wed, 16 Jun 2021 01:23:42 +0000 (18:23 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 16 Jun 2021 16:24:42 +0000 (09:24 -0700)
We notice that hung task happens in a corner but practical scenario when
CONFIG_PREEMPT_NONE is enabled, as follows.

Process 0                       Process 1                     Process 2..Inf
split_huge_page_to_list
    unmap_page
        split_huge_pmd_address
                                __migration_entry_wait(head)
                                                              __migration_entry_wait(tail)
    remap_page (roll back)
        remove_migration_ptes
            rmap_walk_anon
                cond_resched

Where __migration_entry_wait(tail) is occurred in kernel space, e.g.,
copy_to_user in fstat, which will immediately fault again without
rescheduling, and thus occupy the cpu fully.

When there are too many processes performing __migration_entry_wait on
tail page, remap_page will never be done after cond_resched.

This makes __migration_entry_wait operate on the compound head page,
thus waits for remap_page to complete, whether the THP is split
successfully or roll back.

Note that put_and_wait_on_page_locked helps to drop the page reference
acquired with get_page_unless_zero, as soon as the page is on the wait
queue, before actually waiting.  So splitting the THP is only prevented
for a brief interval.

Link: https://lkml.kernel.org/r/b9836c1dd522e903891760af9f0c86a2cce987eb.1623144009.git.xuyu@linux.alibaba.com
Fixes: ba98828088ad ("thp: add option to setup migration entries during PMD split")
Suggested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Gang Deng <gavin.dg@linux.alibaba.com>
Signed-off-by: Xu Yu <xuyu@linux.alibaba.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/migrate.c

index b234c3f3acb7941abf1a558d6e0c5a28d2b1f3d6..41ff2c9896c4f99e5da575ca83d1952ba155a293 100644 (file)
@@ -295,6 +295,7 @@ void __migration_entry_wait(struct mm_struct *mm, pte_t *ptep,
                goto out;
 
        page = migration_entry_to_page(entry);
+       page = compound_head(page);
 
        /*
         * Once page cache replacement of page migration started, page_count