From: Hugh Dickins Date: Fri, 30 Nov 2018 22:10:25 +0000 (-0800) Subject: mm/khugepaged: collapse_shmem() stop if punched or truncated X-Git-Tag: android-x86-7.1-r3~7^2~75^2~46 X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=10e458e6eb59e533e499269258b6069392bec627;p=android-x86%2Fkernel.git mm/khugepaged: collapse_shmem() stop if punched or truncated commit 701270fa193aadf00bdcf607738f64997275d4c7 upstream. Huge tmpfs testing showed that although collapse_shmem() recognizes a concurrently truncated or hole-punched page correctly, its handling of holes was liable to refill an emptied extent. Add check to stop that. Link: http://lkml.kernel.org/r/alpine.LSU.2.11.1811261522040.2275@eggly.anvils Fixes: f3f0e1d2150b2 ("khugepaged: add support of collapse for tmpfs/shmem pages") Signed-off-by: Hugh Dickins Reviewed-by: Matthew Wilcox Cc: Kirill A. Shutemov Cc: Jerome Glisse Cc: Konstantin Khlebnikov Cc: [4.8+] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Sasha Levin --- diff --git a/mm/khugepaged.c b/mm/khugepaged.c index 1df37ee996d5..62de24194f24 100644 --- a/mm/khugepaged.c +++ b/mm/khugepaged.c @@ -1352,6 +1352,16 @@ static void collapse_shmem(struct mm_struct *mm, int n = min(iter.index, end) - index; /* + * Stop if extent has been hole-punched, and is now completely + * empty (the more obvious i_size_read() check would take an + * irq-unsafe seqlock on 32-bit). + */ + if (n >= HPAGE_PMD_NR) { + result = SCAN_TRUNCATED; + goto tree_locked; + } + + /* * Handle holes in the radix tree: charge it from shmem and * insert relevant subpage of new_page into the radix-tree. */ @@ -1462,6 +1472,11 @@ out_unlock: if (result == SCAN_SUCCEED && index < end) { int n = end - index; + /* Stop if extent has been truncated, and is now empty */ + if (n >= HPAGE_PMD_NR) { + result = SCAN_TRUNCATED; + goto tree_locked; + } if (!shmem_charge(mapping->host, n)) { result = SCAN_FAIL; goto tree_locked;