Merge tag 'v4.6-rc6' into patchwork

Linux 4.6-rc6

* tag 'v4.6-rc6': (762 commits)
  Linux 4.6-rc6
  EDAC: i7core, sb_edac: Don't return NOTIFY_BAD from mce_decoder callback
  Documentation/sysctl/vm.txt: update numa_zonelist_order description
  lib/stackdepot.c: allow the stack trace hash to be zero
  rapidio: fix potential NULL pointer dereference
  mm/memory-failure: fix race with compound page split/merge
  ocfs2/dlm: return zero if deref_done message is successfully handled
  Ananth has moved
  kcov: don't profile branches in kcov
  kcov: don't trace the code coverage code
  mm: wake kcompactd before kswapd's short sleep
  .mailmap: add Frank Rowand
  mm/hwpoison: fix wrong num_poisoned_pages accounting
  mm: call swap_slot_free_notify() with page lock held
  mm: vmscan: reclaim highmem zone if buffer_heads is over limit
  numa: fix /proc/<pid>/numa_maps for THP
  mm/huge_memory: replace VM_NO_THP VM_BUG_ON with actual VMA check
  mailmap: fix Krzysztof Kozlowski's misspelled name
  thp: keep huge zero page pinned until tlb flush
  mm: exclude HugeTLB pages from THP page_mapped() logic
  ...

1 files changed, 29 insertions, 0 deletions

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 2232ae3e3ad6..3bfdff06eea7 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -666,6 +666,35 @@ static void set_work_pool_and_clear_pending(struct work_struct *work,
 	 */
 	smp_wmb();
 	set_work_data(work, (unsigned long)pool_id << WORK_OFFQ_POOL_SHIFT, 0);
+	/*
+	 * The following mb guarantees that previous clear of a PENDING bit
+	 * will not be reordered with any speculative LOADS or STORES from
+	 * work->current_func, which is executed afterwards.  This possible
+	 * reordering can lead to a missed execution on attempt to qeueue
+	 * the same @work.  E.g. consider this case:
+	 *
+	 *   CPU#0                         CPU#1
+	 *   ----------------------------  --------------------------------
+	 *
+	 * 1  STORE event_indicated
+	 * 2  queue_work_on() {
+	 * 3    test_and_set_bit(PENDING)
+	 * 4 }                             set_..._and_clear_pending() {
+	 * 5                                 set_work_data() # clear bit
+	 * 6                                 smp_mb()
+	 * 7                               work->current_func() {
+	 * 8				      LOAD event_indicated
+	 *				   }
+	 *
+	 * Without an explicit full barrier speculative LOAD on line 8 can
+	 * be executed before CPU#0 does STORE on line 1.  If that happens,
+	 * CPU#0 observes the PENDING bit is still set and new execution of
+	 * a @work is not queued in a hope, that CPU#1 will eventually
+	 * finish the queued @work.  Meanwhile CPU#1 does not see
+	 * event_indicated is set, because speculative LOAD was executed
+	 * before actual STORE.
+	 */
+	smp_mb();
 }
 
 static void clear_work_data(struct work_struct *work)