Merge branch 'master' into for-linus

Conflicts:
	fs/pipe.c

Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

1 files changed, 159 insertions, 68 deletions

diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 08f40a2f3fe..5d6fb339de0 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -119,7 +119,22 @@ struct mempolicy default_policy = {
 
 static const struct mempolicy_operations {
 	int (*create)(struct mempolicy *pol, const nodemask_t *nodes);
-	void (*rebind)(struct mempolicy *pol, const nodemask_t *nodes);
+	/*
+	 * If read-side task has no lock to protect task->mempolicy, write-side
+	 * task will rebind the task->mempolicy by two step. The first step is
+	 * setting all the newly nodes, and the second step is cleaning all the
+	 * disallowed nodes. In this way, we can avoid finding no node to alloc
+	 * page.
+	 * If we have a lock to protect task->mempolicy in read-side, we do
+	 * rebind directly.
+	 *
+	 * step:
+	 * 	MPOL_REBIND_ONCE - do rebind work at once
+	 * 	MPOL_REBIND_STEP1 - set all the newly nodes
+	 * 	MPOL_REBIND_STEP2 - clean all the disallowed nodes
+	 */
+	void (*rebind)(struct mempolicy *pol, const nodemask_t *nodes,
+			enum mpol_rebind_step step);
 } mpol_ops[MPOL_MAX];
 
 /* Check that the nodemask contains at least one populated zone */
@@ -127,9 +142,6 @@ static int is_valid_nodemask(const nodemask_t *nodemask)
 {
 	int nd, k;
 
-	/* Check that there is something useful in this mask */
-	k = policy_zone;
-
 	for_each_node_mask(nd, *nodemask) {
 		struct zone *z;
 
@@ -145,7 +157,7 @@ static int is_valid_nodemask(const nodemask_t *nodemask)
 
 static inline int mpol_store_user_nodemask(const struct mempolicy *pol)
 {
-	return pol->flags & (MPOL_F_STATIC_NODES | MPOL_F_RELATIVE_NODES);
+	return pol->flags & MPOL_MODE_FLAGS;
 }
 
 static void mpol_relative_nodemask(nodemask_t *ret, const nodemask_t *orig,
@@ -277,12 +289,19 @@ void __mpol_put(struct mempolicy *p)
 	kmem_cache_free(policy_cache, p);
 }
 
-static void mpol_rebind_default(struct mempolicy *pol, const nodemask_t *nodes)
+static void mpol_rebind_default(struct mempolicy *pol, const nodemask_t *nodes,
+				enum mpol_rebind_step step)
 {
 }
 
-static void mpol_rebind_nodemask(struct mempolicy *pol,
-				 const nodemask_t *nodes)
+/*
+ * step:
+ * 	MPOL_REBIND_ONCE  - do rebind work at once
+ * 	MPOL_REBIND_STEP1 - set all the newly nodes
+ * 	MPOL_REBIND_STEP2 - clean all the disallowed nodes
+ */
+static void mpol_rebind_nodemask(struct mempolicy *pol, const nodemask_t *nodes,
+				 enum mpol_rebind_step step)
 {
 	nodemask_t tmp;
 
@@ -291,12 +310,31 @@ static void mpol_rebind_nodemask(struct mempolicy *pol,
 	else if (pol->flags & MPOL_F_RELATIVE_NODES)
 		mpol_relative_nodemask(&tmp, &pol->w.user_nodemask, nodes);
 	else {
-		nodes_remap(tmp, pol->v.nodes, pol->w.cpuset_mems_allowed,
-			    *nodes);
-		pol->w.cpuset_mems_allowed = *nodes;
+		/*
+		 * if step == 1, we use ->w.cpuset_mems_allowed to cache the
+		 * result
+		 */
+		if (step == MPOL_REBIND_ONCE || step == MPOL_REBIND_STEP1) {
+			nodes_remap(tmp, pol->v.nodes,
+					pol->w.cpuset_mems_allowed, *nodes);
+			pol->w.cpuset_mems_allowed = step ? tmp : *nodes;
+		} else if (step == MPOL_REBIND_STEP2) {
+			tmp = pol->w.cpuset_mems_allowed;
+			pol->w.cpuset_mems_allowed = *nodes;
+		} else
+			BUG();
 	}
 
-	pol->v.nodes = tmp;
+	if (nodes_empty(tmp))
+		tmp = *nodes;
+
+	if (step == MPOL_REBIND_STEP1)
+		nodes_or(pol->v.nodes, pol->v.nodes, tmp);
+	else if (step == MPOL_REBIND_ONCE || step == MPOL_REBIND_STEP2)
+		pol->v.nodes = tmp;
+	else
+		BUG();
+
 	if (!node_isset(current->il_next, tmp)) {
 		current->il_next = next_node(current->il_next, tmp);
 		if (current->il_next >= MAX_NUMNODES)
@@ -307,7 +345,8 @@ static void mpol_rebind_nodemask(struct mempolicy *pol,
 }
 
 static void mpol_rebind_preferred(struct mempolicy *pol,
-				  const nodemask_t *nodes)
+				  const nodemask_t *nodes,
+				  enum mpol_rebind_step step)
 {
 	nodemask_t tmp;
 
@@ -330,16 +369,45 @@ static void mpol_rebind_preferred(struct mempolicy *pol,
 	}
 }
 
-/* Migrate a policy to a different set of nodes */
-static void mpol_rebind_policy(struct mempolicy *pol,
-			       const nodemask_t *newmask)
+/*
+ * mpol_rebind_policy - Migrate a policy to a different set of nodes
+ *
+ * If read-side task has no lock to protect task->mempolicy, write-side
+ * task will rebind the task->mempolicy by two step. The first step is
+ * setting all the newly nodes, and the second step is cleaning all the
+ * disallowed nodes. In this way, we can avoid finding no node to alloc
+ * page.
+ * If we have a lock to protect task->mempolicy in read-side, we do
+ * rebind directly.
+ *
+ * step:
+ * 	MPOL_REBIND_ONCE  - do rebind work at once
+ * 	MPOL_REBIND_STEP1 - set all the newly nodes
+ * 	MPOL_REBIND_STEP2 - clean all the disallowed nodes
+ */
+static void mpol_rebind_policy(struct mempolicy *pol, const nodemask_t *newmask,
+				enum mpol_rebind_step step)
 {
 	if (!pol)
 		return;
-	if (!mpol_store_user_nodemask(pol) &&
+	if (!mpol_store_user_nodemask(pol) && step == 0 &&
 	    nodes_equal(pol->w.cpuset_mems_allowed, *newmask))
 		return;
-	mpol_ops[pol->mode].rebind(pol, newmask);
+
+	if (step == MPOL_REBIND_STEP1 && (pol->flags & MPOL_F_REBINDING))
+		return;
+
+	if (step == MPOL_REBIND_STEP2 && !(pol->flags & MPOL_F_REBINDING))
+		BUG();
+
+	if (step == MPOL_REBIND_STEP1)
+		pol->flags |= MPOL_F_REBINDING;
+	else if (step == MPOL_REBIND_STEP2)
+		pol->flags &= ~MPOL_F_REBINDING;
+	else if (step >= MPOL_REBIND_NSTEP)
+		BUG();
+
+	mpol_ops[pol->mode].rebind(pol, newmask, step);
 }
 
 /*
@@ -349,9 +417,10 @@ static void mpol_rebind_policy(struct mempolicy *pol,
  * Called with task's alloc_lock held.
  */
 
-void mpol_rebind_task(struct task_struct *tsk, const nodemask_t *new)
+void mpol_rebind_task(struct task_struct *tsk, const nodemask_t *new,
+			enum mpol_rebind_step step)
 {
-	mpol_rebind_policy(tsk->mempolicy, new);
+	mpol_rebind_policy(tsk->mempolicy, new, step);
 }
 
 /*
@@ -366,7 +435,7 @@ void mpol_rebind_mm(struct mm_struct *mm, nodemask_t *new)
 
 	down_write(&mm->mmap_sem);
 	for (vma = mm->mmap; vma; vma = vma->vm_next)
-		mpol_rebind_policy(vma->vm_policy, new);
+		mpol_rebind_policy(vma->vm_policy, new, MPOL_REBIND_ONCE);
 	up_write(&mm->mmap_sem);
 }
 
@@ -859,7 +928,7 @@ static int migrate_to_node(struct mm_struct *mm, int source, int dest,
 	nodes_clear(nmask);
 	node_set(source, nmask);
 
-	check_range(mm, mm->mmap->vm_start, TASK_SIZE, &nmask,
+	check_range(mm, mm->mmap->vm_start, mm->task_size, &nmask,
 			flags | MPOL_MF_DISCONTIG_OK, &pagelist);
 
 	if (!list_empty(&pagelist))
@@ -1444,15 +1513,13 @@ static struct zonelist *policy_zonelist(gfp_t gfp, struct mempolicy *policy)
 		/*
 		 * Normally, MPOL_BIND allocations are node-local within the
 		 * allowed nodemask.  However, if __GFP_THISNODE is set and the
-		 * current node is part of the mask, we use the zonelist for
+		 * current node isn't part of the mask, we use the zonelist for
 		 * the first node in the mask instead.
 		 */
 		if (unlikely(gfp & __GFP_THISNODE) &&
 				unlikely(!node_isset(nd, policy->v.nodes)))
 			nd = first_node(policy->v.nodes);
 		break;
-	case MPOL_INTERLEAVE: /* should not happen */
-		break;
 	default:
 		BUG();
 	}
@@ -1572,6 +1639,8 @@ static inline unsigned interleave_nid(struct mempolicy *pol,
  * to the struct mempolicy for conditional unref after allocation.
  * If the effective policy is 'BIND, returns a pointer to the mempolicy's
  * @nodemask for filtering the zonelist.
+ *
+ * Must be protected by get_mems_allowed()
  */
 struct zonelist *huge_zonelist(struct vm_area_struct *vma, unsigned long addr,
 				gfp_t gfp_flags, struct mempolicy **mpol,
@@ -1617,6 +1686,7 @@ bool init_nodemask_of_mempolicy(nodemask_t *mask)
 	if (!(mask && current->mempolicy))
 		return false;
 
+	task_lock(current);
 	mempolicy = current->mempolicy;
 	switch (mempolicy->mode) {
 	case MPOL_PREFERRED:
@@ -1636,6 +1706,7 @@ bool init_nodemask_of_mempolicy(nodemask_t *mask)
 	default:
 		BUG();
 	}
+	task_unlock(current);
 
 	return true;
 }
@@ -1683,13 +1754,17 @@ alloc_page_vma(gfp_t gfp, struct vm_area_struct *vma, unsigned long addr)
 {
 	struct mempolicy *pol = get_vma_policy(current, vma, addr);
 	struct zonelist *zl;
+	struct page *page;
 
+	get_mems_allowed();
 	if (unlikely(pol->mode == MPOL_INTERLEAVE)) {
 		unsigned nid;
 
 		nid = interleave_nid(pol, vma, addr, PAGE_SHIFT);
 		mpol_cond_put(pol);
-		return alloc_page_interleave(gfp, 0, nid);
+		page = alloc_page_interleave(gfp, 0, nid);
+		put_mems_allowed();
+		return page;
 	}
 	zl = policy_zonelist(gfp, pol);
 	if (unlikely(mpol_needs_cond_ref(pol))) {
@@ -1699,12 +1774,15 @@ alloc_page_vma(gfp_t gfp, struct vm_area_struct *vma, unsigned long addr)
 		struct page *page =  __alloc_pages_nodemask(gfp, 0,
 						zl, policy_nodemask(gfp, pol));
 		__mpol_put(pol);
+		put_mems_allowed();
 		return page;
 	}
 	/*
 	 * fast path:  default or task policy
 	 */
-	return __alloc_pages_nodemask(gfp, 0, zl, policy_nodemask(gfp, pol));
+	page = __alloc_pages_nodemask(gfp, 0, zl, policy_nodemask(gfp, pol));
+	put_mems_allowed();
+	return page;
 }
 
 /**
@@ -1729,18 +1807,23 @@ alloc_page_vma(gfp_t gfp, struct vm_area_struct *vma, unsigned long addr)
 struct page *alloc_pages_current(gfp_t gfp, unsigned order)
 {
 	struct mempolicy *pol = current->mempolicy;
+	struct page *page;
 
 	if (!pol || in_interrupt() || (gfp & __GFP_THISNODE))
 		pol = &default_policy;
 
+	get_mems_allowed();
 	/*
 	 * No reference counting needed for current->mempolicy
 	 * nor system default_policy
 	 */
 	if (pol->mode == MPOL_INTERLEAVE)
-		return alloc_page_interleave(gfp, order, interleave_nodes(pol));
-	return __alloc_pages_nodemask(gfp, order,
+		page = alloc_page_interleave(gfp, order, interleave_nodes(pol));
+	else
+		page = __alloc_pages_nodemask(gfp, order,
 			policy_zonelist(gfp, pol), policy_nodemask(gfp, pol));
+	put_mems_allowed();
+	return page;
 }
 EXPORT_SYMBOL(alloc_pages_current);
 
@@ -1750,6 +1833,9 @@ EXPORT_SYMBOL(alloc_pages_current);
  * with the mems_allowed returned by cpuset_mems_allowed().  This
  * keeps mempolicies cpuset relative after its cpuset moves.  See
  * further kernel/cpuset.c update_nodemask().
+ *
+ * current's mempolicy may be rebinded by the other task(the task that changes
+ * cpuset's mems), so we needn't do rebind work for current task.
  */
 
 /* Slow path of a mempolicy duplicate */
@@ -1759,13 +1845,24 @@ struct mempolicy *__mpol_dup(struct mempolicy *old)
 
 	if (!new)
 		return ERR_PTR(-ENOMEM);
+
+	/* task's mempolicy is protected by alloc_lock */
+	if (old == current->mempolicy) {
+		task_lock(current);
+		*new = *old;
+		task_unlock(current);
+	} else
+		*new = *old;
+
 	rcu_read_lock();
 	if (current_cpuset_is_being_rebound()) {
 		nodemask_t mems = cpuset_mems_allowed(current);
-		mpol_rebind_policy(old, &mems);
+		if (new->flags & MPOL_F_REBINDING)
+			mpol_rebind_policy(new, &mems, MPOL_REBIND_STEP2);
+		else
+			mpol_rebind_policy(new, &mems, MPOL_REBIND_ONCE);
 	}
 	rcu_read_unlock();
-	*new = *old;
 	atomic_set(&new->refcnt, 1);
 	return new;
 }
@@ -1792,16 +1889,6 @@ struct mempolicy *__mpol_cond_copy(struct mempolicy *tompol,
 	return tompol;
 }
 
-static int mpol_match_intent(const struct mempolicy *a,
-			     const struct mempolicy *b)
-{
-	if (a->flags != b->flags)
-		return 0;
-	if (!mpol_store_user_nodemask(a))
-		return 1;
-	return nodes_equal(a->w.user_nodemask, b->w.user_nodemask);
-}
-
 /* Slow path of a mempolicy comparison */
 int __mpol_equal(struct mempolicy *a, struct mempolicy *b)
 {
@@ -1809,8 +1896,12 @@ int __mpol_equal(struct mempolicy *a, struct mempolicy *b)
 		return 0;
 	if (a->mode != b->mode)
 		return 0;
-	if (a->mode != MPOL_DEFAULT && !mpol_match_intent(a, b))
+	if (a->flags != b->flags)
 		return 0;
+	if (mpol_store_user_nodemask(a))
+		if (!nodes_equal(a->w.user_nodemask, b->w.user_nodemask))
+			return 0;
+
 	switch (a->mode) {
 	case MPOL_BIND:
 		/* Fall through */
@@ -2006,27 +2097,24 @@ void mpol_shared_policy_init(struct shared_policy *sp, struct mempolicy *mpol)
 			return;
 		/* contextualize the tmpfs mount point mempolicy */
 		new = mpol_new(mpol->mode, mpol->flags, &mpol->w.user_nodemask);
-		if (IS_ERR(new)) {
-			mpol_put(mpol);	/* drop our ref on sb mpol */
-			NODEMASK_SCRATCH_FREE(scratch);
-			return;		/* no valid nodemask intersection */
-		}
+		if (IS_ERR(new))
+			goto free_scratch; /* no valid nodemask intersection */
 
 		task_lock(current);
 		ret = mpol_set_nodemask(new, &mpol->w.user_nodemask, scratch);
 		task_unlock(current);
 		mpol_put(mpol);	/* drop our ref on sb mpol */
-		if (ret) {
-			NODEMASK_SCRATCH_FREE(scratch);
-			mpol_put(new);
-			return;
-		}
+		if (ret)
+			goto put_free;
 
 		/* Create pseudo-vma that contains just the policy */
 		memset(&pvma, 0, sizeof(struct vm_area_struct));
 		pvma.vm_end = TASK_SIZE;	/* policy covers entire file */
 		mpol_set_shared_policy(sp, &pvma, new); /* adds ref */
+
+put_free:
 		mpol_put(new);			/* drop initial ref */
+free_scratch:
 		NODEMASK_SCRATCH_FREE(scratch);
 	}
 }
@@ -2132,9 +2220,15 @@ void numa_default_policy(void)
  * "local" is pseudo-policy:  MPOL_PREFERRED with MPOL_F_LOCAL flag
  * Used only for mpol_parse_str() and mpol_to_str()
  */
-#define MPOL_LOCAL (MPOL_INTERLEAVE + 1)
-static const char * const policy_types[] =
-	{ "default", "prefer", "bind", "interleave", "local" };
+#define MPOL_LOCAL MPOL_MAX
+static const char * const policy_modes[] =
+{
+	[MPOL_DEFAULT]    = "default",
+	[MPOL_PREFERRED]  = "prefer",
+	[MPOL_BIND]       = "bind",
+	[MPOL_INTERLEAVE] = "interleave",
+	[MPOL_LOCAL]      = "local"
+};
 
 
 #ifdef CONFIG_TMPFS
@@ -2159,12 +2253,11 @@ static const char * const policy_types[] =
 int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context)
 {
 	struct mempolicy *new = NULL;
-	unsigned short uninitialized_var(mode);
+	unsigned short mode;
 	unsigned short uninitialized_var(mode_flags);
 	nodemask_t nodes;
 	char *nodelist = strchr(str, ':');
 	char *flags = strchr(str, '=');
-	int i;
 	int err = 1;
 
 	if (nodelist) {
@@ -2180,13 +2273,12 @@ int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context)
 	if (flags)
 		*flags++ = '\0';	/* terminate mode string */
 
-	for (i = 0; i <= MPOL_LOCAL; i++) {
-		if (!strcmp(str, policy_types[i])) {
-			mode = i;
+	for (mode = 0; mode <= MPOL_LOCAL; mode++) {
+		if (!strcmp(str, policy_modes[mode])) {
 			break;
 		}
 	}
-	if (i > MPOL_LOCAL)
+	if (mode > MPOL_LOCAL)
 		goto out;
 
 	switch (mode) {
@@ -2250,7 +2342,10 @@ int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context)
 	if (IS_ERR(new))
 		goto out;
 
-	{
+	if (no_context) {
+		/* save for contextualization */
+		new->w.user_nodemask = nodes;
+	} else {
 		int ret;
 		NODEMASK_SCRATCH(scratch);
 		if (scratch) {
@@ -2266,10 +2361,6 @@ int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context)
 		}
 	}
 	err = 0;
-	if (no_context) {
-		/* save for contextualization */
-		new->w.user_nodemask = nodes;
-	}
 
 out:
 	/* Restore string for error message */
@@ -2338,11 +2429,11 @@ int mpol_to_str(char *buffer, int maxlen, struct mempolicy *pol, int no_context)
 		BUG();
 	}
 
-	l = strlen(policy_types[mode]);
+	l = strlen(policy_modes[mode]);
 	if (buffer + maxlen < p + l + 1)
 		return -ENOSPC;
 
-	strcpy(p, policy_types[mode]);
+	strcpy(p, policy_modes[mode]);
 	p += l;
 
 	if (flags & MPOL_MODE_FLAGS) {