LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [Resend][PATCH -v2 1/3 -next] Unlock the locked hugetlb page without MF_COUNT_INCREASED.
@ 2011-02-07  4:47 Jin Dongming
  0 siblings, 0 replies; only message in thread
From: Jin Dongming @ 2011-02-07  4:47 UTC (permalink / raw)
  To: Andi Kleen
  Cc: Naoya Horiguchi, Huang Ying, Wu Fengguang, Hidetoshi Seto, LKLM

When the unmapped hugetlb page is poisoned, the page is locked
for checking some conditions of page. When one of the following condition
in __memory_failure()
    if (!PageHWPoison(hpage)
        || (hwpoison_filter(p) && TestClearPageHWPoison(p))
        || (p != hpage && TestSetPageHWPoison(hpage) {
        ...
    }

is matched, it just returns from __memory_failure() without unlocking
the locked hugetlb page. This will block the process which wants to
map the blocked hugetlb page.

If some process is blocked, error message could be outputted as following:
    kernel: INFO: task hmmap:3307 blocked for more than 120 seconds.
    kernel: "echo 0 > /proc/sys/    kernel/hung_task_timeout_secs" disables this message.
    kernel: hmmap         D ffff880196946100     0  3307      1 0x00000080
    kernel:  ffff880196946100 0000000000000086 ffff880195ca6040 0000000000000000
    kernel:  0000000000000046 ffffffff81036a7d ffffffff81d08980 0000000000000000
    kernel:  ffffffff8189c74c ffffffff814f6886 000000000000002e 0000000000000246
    kernel: Call Trace:
    kernel:  [<ffffffff81036a7d>] ? release_console_sem+0x174/0x18e
    kernel:  [<ffffffff810914f4>] ? sync_page+0x0/0x48
    kernel:  [<ffffffff813bcbdb>] ? io_schedule+0x56/0x8c
    kernel:  [<ffffffff81091538>] ? sync_page+0x44/0x48
    kernel:  [<ffffffff813bd0ab>] ? __wait_on_bit_lock+0x3e/0x83
    kernel:  [<ffffffff810914e1>] ? __lock_page+0x5e/0x64
    kernel:  [<ffffffff8104ea89>] ? wake_bit_function+0x0/0x23
    kernel:  [<ffffffff810bd723>] ? hugetlb_fault+0x38e/0x6ef
    kernel:  [<ffffffff813c11ac>] ? do_page_fault+0x283/0x3dd
    kernel:  [<ffffffff81032546>] ? __wake_up+0x30/0x44
    kernel:  [<ffffffff813be79f>] ? page_fault+0x1f/0x30

This is because the hugetlb page has been locked in __memory_failure(),
the process will be blocked by lock_page() in hugetlb_fault()
all the time.

So unlocked the locked hugetlb page before returning from
__memory_failure().

Signed-off-by: Jin Dongming <jin.dongming@np.css.fujitsu.com>
Reviewed-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
---
 mm/memory-failure.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 0207c2f..d2c2a7b 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1043,6 +1043,7 @@ int __memory_failure(unsigned long pfn, int trapno, int flags)
 			    || (hwpoison_filter(p) && TestClearPageHWPoison(p))
 			    || (p != hpage && TestSetPageHWPoison(hpage))) {
 				atomic_long_sub(nr_pages, &mce_bad_pages);
+				unlock_page(hpage);
 				return 0;
 			}
 			set_page_hwpoison_huge_page(hpage);
-- 
1.7.2.2



^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2011-02-07  4:44 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-02-07  4:47 [Resend][PATCH -v2 1/3 -next] Unlock the locked hugetlb page without MF_COUNT_INCREASED Jin Dongming

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).