Merge tag 'v4.6-rc6' into locking/core, to pick up fixes

Signed-off-by: Ingo Molnar <mingo@kernel.org>

15 files changed, 188 insertions, 48 deletions

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 2e08f8e9b771..db2574e7b8b0 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -1374,6 +1374,7 @@ static int check_ld_abs(struct verifier_env *env, struct bpf_insn *insn)
 	}
 
 	if (insn->dst_reg != BPF_REG_0 || insn->off != 0 ||
+	    BPF_SIZE(insn->code) == BPF_DW ||
 	    (mode == BPF_ABS && insn->src_reg != BPF_REG_0)) {
 		verbose("BPF_LD_ABS uses reserved fields\n");
 		return -EINVAL;
@@ -2029,7 +2030,6 @@ static int replace_map_fd_with_map_ptr(struct verifier_env *env)
 			if (IS_ERR(map)) {
 				verbose("fd %d is not pointing to valid bpf_map\n",
 					insn->imm);
-				fdput(f);
 				return PTR_ERR(map);
 			}
 
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 671dc05c0b0f..909a7d31ffd3 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -2825,9 +2825,10 @@ static ssize_t __cgroup_procs_write(struct kernfs_open_file *of, char *buf,
 				    size_t nbytes, loff_t off, bool threadgroup)
 {
 	struct task_struct *tsk;
+	struct cgroup_subsys *ss;
 	struct cgroup *cgrp;
 	pid_t pid;
-	int ret;
+	int ssid, ret;
 
 	if (kstrtoint(strstrip(buf), 0, &pid) || pid < 0)
 		return -EINVAL;
@@ -2875,8 +2876,10 @@ out_unlock_rcu:
 	rcu_read_unlock();
 out_unlock_threadgroup:
 	percpu_up_write(&cgroup_threadgroup_rwsem);
+	for_each_subsys(ss, ssid)
+		if (ss->post_attach)
+			ss->post_attach();
 	cgroup_kn_unlock(of->kn);
-	cpuset_post_attach_flush();
 	return ret ?: nbytes;
 }
 
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 6ea42e8da861..3e3f6e49eabb 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -36,6 +36,7 @@
  * @target:	The target state
  * @thread:	Pointer to the hotplug thread
  * @should_run:	Thread should execute
+ * @rollback:	Perform a rollback
  * @cb_stat:	The state for a single callback (install/uninstall)
  * @cb:		Single callback function (install/uninstall)
  * @result:	Result of the operation
@@ -47,6 +48,7 @@ struct cpuhp_cpu_state {
 #ifdef CONFIG_SMP
 	struct task_struct	*thread;
 	bool			should_run;
+	bool			rollback;
 	enum cpuhp_state	cb_state;
 	int			(*cb)(unsigned int cpu);
 	int			result;
@@ -301,6 +303,11 @@ static int cpu_notify(unsigned long val, unsigned int cpu)
 	return __cpu_notify(val, cpu, -1, NULL);
 }
 
+static void cpu_notify_nofail(unsigned long val, unsigned int cpu)
+{
+	BUG_ON(cpu_notify(val, cpu));
+}
+
 /* Notifier wrappers for transitioning to state machine */
 static int notify_prepare(unsigned int cpu)
 {
@@ -477,6 +484,16 @@ static void cpuhp_thread_fun(unsigned int cpu)
 		} else {
 			ret = cpuhp_invoke_callback(cpu, st->cb_state, st->cb);
 		}
+	} else if (st->rollback) {
+		BUG_ON(st->state < CPUHP_AP_ONLINE_IDLE);
+
+		undo_cpu_down(cpu, st, cpuhp_ap_states);
+		/*
+		 * This is a momentary workaround to keep the notifier users
+		 * happy. Will go away once we got rid of the notifiers.
+		 */
+		cpu_notify_nofail(CPU_DOWN_FAILED, cpu);
+		st->rollback = false;
 	} else {
 		/* Cannot happen .... */
 		BUG_ON(st->state < CPUHP_AP_ONLINE_IDLE);
@@ -636,11 +653,6 @@ static inline void check_for_tasks(int dead_cpu)
 	read_unlock(&tasklist_lock);
 }
 
-static void cpu_notify_nofail(unsigned long val, unsigned int cpu)
-{
-	BUG_ON(cpu_notify(val, cpu));
-}
-
 static int notify_down_prepare(unsigned int cpu)
 {
 	int err, nr_calls = 0;
@@ -721,9 +733,10 @@ static int takedown_cpu(unsigned int cpu)
 	 */
 	err = stop_machine(take_cpu_down, NULL, cpumask_of(cpu));
 	if (err) {
-		/* CPU didn't die: tell everyone.  Can't complain. */
-		cpu_notify_nofail(CPU_DOWN_FAILED, cpu);
+		/* CPU refused to die */
 		irq_unlock_sparse();
+		/* Unpark the hotplug thread so we can rollback there */
+		kthread_unpark(per_cpu_ptr(&cpuhp_state, cpu)->thread);
 		return err;
 	}
 	BUG_ON(cpu_online(cpu));
@@ -832,6 +845,11 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen,
 	 * to do the further cleanups.
 	 */
 	ret = cpuhp_down_callbacks(cpu, st, cpuhp_bp_states, target);
+	if (ret && st->state > CPUHP_TEARDOWN_CPU && st->state < prev_state) {
+		st->target = prev_state;
+		st->rollback = true;
+		cpuhp_kick_ap_work(cpu);
+	}
 
 	hasdied = prev_state != st->state && st->state == CPUHP_OFFLINE;
 out:
@@ -1249,6 +1267,7 @@ static struct cpuhp_step cpuhp_ap_states[] = {
 		.name			= "notify:online",
 		.startup		= notify_online,
 		.teardown		= notify_down_prepare,
+		.skip_onerr		= true,
 	},
 #endif
 	/*
diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 00ab5c2b7c5b..1902956baba1 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -58,7 +58,6 @@
 #include <asm/uaccess.h>
 #include <linux/atomic.h>
 #include <linux/mutex.h>
-#include <linux/workqueue.h>
 #include <linux/cgroup.h>
 #include <linux/wait.h>
 
@@ -1016,7 +1015,7 @@ static void cpuset_migrate_mm(struct mm_struct *mm, const nodemask_t *from,
 	}
 }
 
-void cpuset_post_attach_flush(void)
+static void cpuset_post_attach(void)
 {
 	flush_workqueue(cpuset_migrate_mm_wq);
 }
@@ -2087,6 +2086,7 @@ struct cgroup_subsys cpuset_cgrp_subsys = {
 	.can_attach	= cpuset_can_attach,
 	.cancel_attach	= cpuset_cancel_attach,
 	.attach		= cpuset_attach,
+	.post_attach	= cpuset_post_attach,
 	.bind		= cpuset_bind,
 	.legacy_cftypes	= files,
 	.early_init	= true,
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 52bedc5a5aaa..4e2ebf6f2f1f 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -412,7 +412,8 @@ int perf_cpu_time_max_percent_handler(struct ctl_table *table, int write,
 	if (ret || !write)
 		return ret;
 
-	if (sysctl_perf_cpu_time_max_percent == 100) {
+	if (sysctl_perf_cpu_time_max_percent == 100 ||
+	    sysctl_perf_cpu_time_max_percent == 0) {
 		printk(KERN_WARNING
 		       "perf: Dynamic interrupt throttling disabled, can hang your system!\n");
 		WRITE_ONCE(perf_sample_allowed_ns, 0);
@@ -1105,6 +1106,7 @@ static void put_ctx(struct perf_event_context *ctx)
  * function.
  *
  * Lock order:
+ *    cred_guard_mutex
  *	task_struct::perf_event_mutex
  *	  perf_event_context::mutex
  *	    perf_event::child_mutex;
@@ -3420,7 +3422,6 @@ static struct task_struct *
 find_lively_task_by_vpid(pid_t vpid)
 {
 	struct task_struct *task;
-	int err;
 
 	rcu_read_lock();
 	if (!vpid)
@@ -3434,16 +3435,7 @@ find_lively_task_by_vpid(pid_t vpid)
 	if (!task)
 		return ERR_PTR(-ESRCH);
 
-	/* Reuse ptrace permission checks for now. */
-	err = -EACCES;
-	if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS))
-		goto errout;
-
 	return task;
-errout:
-	put_task_struct(task);
-	return ERR_PTR(err);
-
 }
 
 /*
@@ -8413,6 +8405,24 @@ SYSCALL_DEFINE5(perf_event_open,
 
 	get_online_cpus();
 
+	if (task) {
+		err = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
+		if (err)
+			goto err_cpus;
+
+		/*
+		 * Reuse ptrace permission checks for now.
+		 *
+		 * We must hold cred_guard_mutex across this and any potential
+		 * perf_install_in_context() call for this new event to
+		 * serialize against exec() altering our credentials (and the
+		 * perf_event_exit_task() that could imply).
+		 */
+		err = -EACCES;
+		if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS))
+			goto err_cred;
+	}
+
 	if (flags & PERF_FLAG_PID_CGROUP)
 		cgroup_fd = pid;
 
@@ -8420,7 +8430,7 @@ SYSCALL_DEFINE5(perf_event_open,
 				 NULL, NULL, cgroup_fd);
 	if (IS_ERR(event)) {
 		err = PTR_ERR(event);
-		goto err_cpus;
+		goto err_cred;
 	}
 
 	if (is_sampling_event(event)) {
@@ -8479,11 +8489,6 @@ SYSCALL_DEFINE5(perf_event_open,
 		goto err_context;
 	}
 
-	if (task) {
-		put_task_struct(task);
-		task = NULL;
-	}
-
 	/*
 	 * Look up the group leader (we will attach this event to it):
 	 */
@@ -8581,6 +8586,11 @@ SYSCALL_DEFINE5(perf_event_open,
 
 	WARN_ON_ONCE(ctx->parent_ctx);
 
+	/*
+	 * This is the point on no return; we cannot fail hereafter. This is
+	 * where we start modifying current state.
+	 */
+
 	if (move_group) {
 		/*
 		 * See perf_event_ctx_lock() for comments on the details
@@ -8652,6 +8662,11 @@ SYSCALL_DEFINE5(perf_event_open,
 		mutex_unlock(&gctx->mutex);
 	mutex_unlock(&ctx->mutex);
 
+	if (task) {
+		mutex_unlock(&task->signal->cred_guard_mutex);
+		put_task_struct(task);
+	}
+
 	put_online_cpus();
 
 	mutex_lock(&current->perf_event_mutex);
@@ -8684,6 +8699,9 @@ err_alloc:
 	 */
 	if (!event_file)
 		free_event(event);
+err_cred:
+	if (task)
+		mutex_unlock(&task->signal->cred_guard_mutex);
 err_cpus:
 	put_online_cpus();
 err_task:
@@ -8968,6 +8986,9 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn)
 
 /*
  * When a child task exits, feed back event values to parent events.
+ *
+ * Can be called with cred_guard_mutex held when called from
+ * install_exec_creds().
  */
 void perf_event_exit_task(struct task_struct *child)
 {
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 220fc17b9718..7edc95edfaee 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -321,7 +321,7 @@ retry:
 	copy_to_page(new_page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE);
 
 	ret = __replace_page(vma, vaddr, old_page, new_page);
-	page_cache_release(new_page);
+	put_page(new_page);
 put_old:
 	put_page(old_page);
 
@@ -539,14 +539,14 @@ static int __copy_insn(struct address_space *mapping, struct file *filp,
 	 * see uprobe_register().
 	 */
 	if (mapping->a_ops->readpage)
-		page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp);
+		page = read_mapping_page(mapping, offset >> PAGE_SHIFT, filp);
 	else
-		page = shmem_read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT);
+		page = shmem_read_mapping_page(mapping, offset >> PAGE_SHIFT);
 	if (IS_ERR(page))
 		return PTR_ERR(page);
 
 	copy_from_page(page, offset, insn, nbytes);
-	page_cache_release(page);
+	put_page(page);
 
 	return 0;
 }
diff --git a/kernel/futex.c b/kernel/futex.c
index a5d2e74c89e0..c20f06f38ef3 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -1295,10 +1295,20 @@ static int wake_futex_pi(u32 __user *uaddr, u32 uval, struct futex_q *this,
 	if (unlikely(should_fail_futex(true)))
 		ret = -EFAULT;
 
-	if (cmpxchg_futex_value_locked(&curval, uaddr, uval, newval))
+	if (cmpxchg_futex_value_locked(&curval, uaddr, uval, newval)) {
 		ret = -EFAULT;
-	else if (curval != uval)
-		ret = -EINVAL;
+	} else if (curval != uval) {
+		/*
+		 * If a unconditional UNLOCK_PI operation (user space did not
+		 * try the TID->0 transition) raced with a waiter setting the
+		 * FUTEX_WAITERS flag between get_user() and locking the hash
+		 * bucket lock, retry the operation.
+		 */
+		if ((FUTEX_TID_MASK & curval) == uval)
+			ret = -EAGAIN;
+		else
+			ret = -EINVAL;
+	}
 	if (ret) {
 		raw_spin_unlock_irq(&pi_state->pi_mutex.wait_lock);
 		return ret;
@@ -1525,8 +1535,8 @@ void requeue_futex(struct futex_q *q, struct futex_hash_bucket *hb1,
 	if (likely(&hb1->chain != &hb2->chain)) {
 		plist_del(&q->list, &hb1->chain);
 		hb_waiters_dec(hb1);
-		plist_add(&q->list, &hb2->chain);
 		hb_waiters_inc(hb2);
+		plist_add(&q->list, &hb2->chain);
 		q->lock_ptr = &hb2->lock;
 	}
 	get_futex_key_refs(key2);
@@ -2623,6 +2633,15 @@ retry:
 		if (ret == -EFAULT)
 			goto pi_faulted;
 		/*
+		 * A unconditional UNLOCK_PI op raced against a waiter
+		 * setting the FUTEX_WAITERS bit. Try again.
+		 */
+		if (ret == -EAGAIN) {
+			spin_unlock(&hb->lock);
+			put_futex_key(&key);
+			goto retry;
+		}
+		/*
 		 * wake_futex_pi has detected invalid state. Tell user
 		 * space.
 		 */
diff --git a/kernel/irq/ipi.c b/kernel/irq/ipi.c
index c37f34b00a11..14777af8e097 100644
--- a/kernel/irq/ipi.c
+++ b/kernel/irq/ipi.c
@@ -94,6 +94,7 @@ unsigned int irq_reserve_ipi(struct irq_domain *domain,
 		data = irq_get_irq_data(virq + i);
 		cpumask_copy(data->common->affinity, dest);
 		data->common->ipi_offset = offset;
+		irq_set_status_flags(virq + i, IRQ_NO_BALANCING);
 	}
 	return virq;
 
diff --git a/kernel/kcov.c b/kernel/kcov.c
index 3efbee0834a8..a02f2dddd1d7 100644
--- a/kernel/kcov.c
+++ b/kernel/kcov.c
@@ -1,5 +1,6 @@
 #define pr_fmt(fmt) "kcov: " fmt
 
+#define DISABLE_BRANCH_PROFILING
 #include <linux/compiler.h>
 #include <linux/types.h>
 #include <linux/file.h>
@@ -43,7 +44,7 @@ struct kcov {
  * Entry point from instrumented code.
  * This is called once per basic-block/edge.
  */
-void __sanitizer_cov_trace_pc(void)
+void notrace __sanitizer_cov_trace_pc(void)
 {
 	struct task_struct *t;
 	enum kcov_mode mode;
diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c
index 8d34308ea449..1391d3ee3b86 100644
--- a/kernel/kexec_core.c
+++ b/kernel/kexec_core.c
@@ -1415,6 +1415,9 @@ static int __init crash_save_vmcoreinfo_init(void)
 	VMCOREINFO_OFFSET(page, lru);
 	VMCOREINFO_OFFSET(page, _mapcount);
 	VMCOREINFO_OFFSET(page, private);
+	VMCOREINFO_OFFSET(page, compound_dtor);
+	VMCOREINFO_OFFSET(page, compound_order);
+	VMCOREINFO_OFFSET(page, compound_head);
 	VMCOREINFO_OFFSET(pglist_data, node_zones);
 	VMCOREINFO_OFFSET(pglist_data, nr_zones);
 #ifdef CONFIG_FLAT_NODE_MEM_MAP
@@ -1447,8 +1450,8 @@ static int __init crash_save_vmcoreinfo_init(void)
 #ifdef CONFIG_X86
 	VMCOREINFO_NUMBER(KERNEL_IMAGE_SIZE);
 #endif
-#ifdef CONFIG_HUGETLBFS
-	VMCOREINFO_SYMBOL(free_huge_page);
+#ifdef CONFIG_HUGETLB_PAGE
+	VMCOREINFO_NUMBER(HUGETLB_PAGE_DTOR);
 #endif
 
 	arch_crash_save_vmcoreinfo();
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 7cc43ef856c1..874d53eaf389 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -2176,15 +2176,37 @@ cache_hit:
 	chain->irq_context = hlock->irq_context;
 	i = get_first_held_lock(curr, hlock);
 	chain->depth = curr->lockdep_depth + 1 - i;
+
+	BUILD_BUG_ON((1UL << 24) <= ARRAY_SIZE(chain_hlocks));
+	BUILD_BUG_ON((1UL << 6)  <= ARRAY_SIZE(curr->held_locks));
+	BUILD_BUG_ON((1UL << 8*sizeof(chain_hlocks[0])) <= ARRAY_SIZE(lock_classes));
+
 	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
 		chain->base = nr_chain_hlocks;
-		nr_chain_hlocks += chain->depth;
 		for (j = 0; j < chain->depth - 1; j++, i++) {
 			int lock_id = curr->held_locks[i].class_idx - 1;
 			chain_hlocks[chain->base + j] = lock_id;
 		}
 		chain_hlocks[chain->base + j] = class - lock_classes;
 	}
+
+	if (nr_chain_hlocks < MAX_LOCKDEP_CHAIN_HLOCKS)
+		nr_chain_hlocks += chain->depth;
+
+#ifdef CONFIG_DEBUG_LOCKDEP
+	/*
+	 * Important for check_no_collision().
+	 */
+	if (unlikely(nr_chain_hlocks > MAX_LOCKDEP_CHAIN_HLOCKS)) {
+		if (debug_locks_off_graph_unlock())
+			return 0;
+
+		print_lockdep_off("BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!");
+		dump_stack();
+		return 0;
+	}
+#endif
+
 	hlist_add_head_rcu(&chain->entry, hash_head);
 	debug_atomic_inc(chain_lookup_misses);
 	inc_chains();
@@ -2932,6 +2954,11 @@ static int mark_irqflags(struct task_struct *curr, struct held_lock *hlock)
 	return 1;
 }
 
+static inline unsigned int task_irq_context(struct task_struct *task)
+{
+	return 2 * !!task->hardirq_context + !!task->softirq_context;
+}
+
 static int separate_irq_context(struct task_struct *curr,
 		struct held_lock *hlock)
 {
@@ -2940,8 +2967,6 @@ static int separate_irq_context(struct task_struct *curr,
 	/*
 	 * Keep track of points where we cross into an interrupt context:
 	 */
-	hlock->irq_context = 2*(curr->hardirq_context ? 1 : 0) +
-				curr->softirq_context;
 	if (depth) {
 		struct held_lock *prev_hlock;
 
@@ -2973,6 +2998,11 @@ static inline int mark_irqflags(struct task_struct *curr,
 	return 1;
 }
 
+static inline unsigned int task_irq_context(struct task_struct *task)
+{
+	return 0;
+}
+
 static inline int separate_irq_context(struct task_struct *curr,
 		struct held_lock *hlock)
 {
@@ -3241,6 +3271,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 	hlock->acquire_ip = ip;
 	hlock->instance = lock;
 	hlock->nest_lock = nest_lock;
+	hlock->irq_context = task_irq_context(curr);
 	hlock->trylock = trylock;
 	hlock->read = read;
 	hlock->check = check;
diff --git a/kernel/locking/lockdep_proc.c b/kernel/locking/lockdep_proc.c
index dbb61a302548..a0f61effad25 100644
--- a/kernel/locking/lockdep_proc.c
+++ b/kernel/locking/lockdep_proc.c
@@ -141,6 +141,8 @@ static int lc_show(struct seq_file *m, void *v)
 	int i;
 
 	if (v == SEQ_START_TOKEN) {
+		if (nr_chain_hlocks > MAX_LOCKDEP_CHAIN_HLOCKS)
+			seq_printf(m, "(buggered) ");
 		seq_printf(m, "all lock chains:\n");
 		return 0;
 	}
diff --git a/kernel/locking/qspinlock_stat.h b/kernel/locking/qspinlock_stat.h
index eb2a2c9bc3fc..d734b7502001 100644
--- a/kernel/locking/qspinlock_stat.h
+++ b/kernel/locking/qspinlock_stat.h
@@ -136,10 +136,12 @@ static ssize_t qstat_read(struct file *file, char __user *user_buf,
 	}
 
 	if (counter == qstat_pv_hash_hops) {
-		u64 frac;
+		u64 frac = 0;
 
-		frac = 100ULL * do_div(stat, kicks);
-		frac = DIV_ROUND_CLOSEST_ULL(frac, kicks);
+		if (kicks) {
+			frac = 100ULL * do_div(stat, kicks);
+			frac = DIV_ROUND_CLOSEST_ULL(frac, kicks);
+		}
 
 		/*
 		 * Return a X.XX decimal number
diff --git a/kernel/resource.c b/kernel/resource.c
index 2e78ead30934..9b5f04404152 100644
--- a/kernel/resource.c
+++ b/kernel/resource.c
@@ -105,16 +105,25 @@ static int r_show(struct seq_file *m, void *v)
 {
 	struct resource *root = m->private;
 	struct resource *r = v, *p;
+	unsigned long long start, end;
 	int width = root->end < 0x10000 ? 4 : 8;
 	int depth;
 
 	for (depth = 0, p = r; depth < MAX_IORES_LEVEL; depth++, p = p->parent)
 		if (p->parent == root)
 			break;
+
+	if (file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) {
+		start = r->start;
+		end = r->end;
+	} else {
+		start = end = 0;
+	}
+
 	seq_printf(m, "%*s%0*llx-%0*llx : %s\n",
 			depth * 2, "",
-			width, (unsigned long long) r->start,
-			width, (unsigned long long) r->end,
+			width, start,
+			width, end,
 			r->name ? r->name : "<BAD>");
 	return 0;
 }
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)