forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-09-20 cf4ce59b3b70238352c7f1729f0f7223214828ad 
 kernel/fs/kernfs/dir.c
....@@ -1,11 +1,10 @@
1
+// SPDX-License-Identifier: GPL-2.0-only
12 /*
23  * fs/kernfs/dir.c - kernfs directory implementation
34  *
45  * Copyright (c) 2001-3 Patrick Mochel
56  * Copyright (c) 2007 SUSE Linux Products GmbH
67  * Copyright (c) 2007, 2013 Tejun Heo <tj@kernel.org>
7
- *
8
- * This file is released under the GPLv2.
98  */
109 
1110 #include <linux/sched.h>
....@@ -20,7 +19,15 @@
2019 
2120 DEFINE_MUTEX(kernfs_mutex);
2221 static DEFINE_SPINLOCK(kernfs_rename_lock);	/* kn->parent and ->name */
23
-static char kernfs_pr_cont_buf[PATH_MAX];	/* protected by rename_lock */
22
+/*
23
+ * Don't use rename_lock to piggy back on pr_cont_buf. We don't want to
24
+ * call pr_cont() while holding rename_lock. Because sometimes pr_cont()
25
+ * will perform wakeups when releasing console_sem. Holding rename_lock
26
+ * will introduce deadlock if the scheduler reads the kernfs_name in the
27
+ * wakeup path.
28
+ */
29
+static DEFINE_SPINLOCK(kernfs_pr_cont_lock);
30
+static char kernfs_pr_cont_buf[PATH_MAX];	/* protected by pr_cont_lock */
2431 static DEFINE_SPINLOCK(kernfs_idr_lock);	/* root->ino_idr */
2532 
2633 #define rb_to_kn(X) rb_entry((X), struct kernfs_node, rb)
....@@ -138,6 +145,9 @@
138145 	if (kn_from == kn_to)
139146 		return strlcpy(buf, "/", buflen);
140147 
148
+	if (!buf)
149
+		return -EINVAL;
150
+
141151 	common = kernfs_common_ancestor(kn_from, kn_to);
142152 	if (WARN_ON(!common))
143153 		return -EINVAL;
....@@ -145,8 +155,7 @@
145155 	depth_to = kernfs_depth(common, kn_to);
146156 	depth_from = kernfs_depth(common, kn_from);
147157 
148
-	if (buf)
149
-		buf[0] = '\0';
158
+	buf[0] = '\0';
150159 
151160 	for (i = 0; i < depth_from; i++)
152161 		len += strlcpy(buf + len, parent_str,
....@@ -229,12 +238,12 @@
229238 {
230239 	unsigned long flags;
231240 
232
-	spin_lock_irqsave(&kernfs_rename_lock, flags);
241
+	spin_lock_irqsave(&kernfs_pr_cont_lock, flags);
233242 
234
-	kernfs_name_locked(kn, kernfs_pr_cont_buf, sizeof(kernfs_pr_cont_buf));
243
+	kernfs_name(kn, kernfs_pr_cont_buf, sizeof(kernfs_pr_cont_buf));
235244 	pr_cont("%s", kernfs_pr_cont_buf);
236245 
237
-	spin_unlock_irqrestore(&kernfs_rename_lock, flags);
246
+	spin_unlock_irqrestore(&kernfs_pr_cont_lock, flags);
238247 }
239248 
240249 /**
....@@ -248,10 +257,10 @@
248257 	unsigned long flags;
249258 	int sz;
250259 
251
-	spin_lock_irqsave(&kernfs_rename_lock, flags);
260
+	spin_lock_irqsave(&kernfs_pr_cont_lock, flags);
252261 
253
-	sz = kernfs_path_from_node_locked(kn, NULL, kernfs_pr_cont_buf,
254
-					  sizeof(kernfs_pr_cont_buf));
262
+	sz = kernfs_path_from_node(kn, NULL, kernfs_pr_cont_buf,
263
+				   sizeof(kernfs_pr_cont_buf));
255264 	if (sz < 0) {
256265 		pr_cont("(error)");
257266 		goto out;
....@@ -265,7 +274,7 @@
265274 	pr_cont("%s", kernfs_pr_cont_buf);
266275 
267276 out:
268
-	spin_unlock_irqrestore(&kernfs_rename_lock, flags);
277
+	spin_unlock_irqrestore(&kernfs_pr_cont_lock, flags);
269278 }
270279 
271280 /**
....@@ -431,19 +440,18 @@
431440  */
432441 void kernfs_put_active(struct kernfs_node *kn)
433442 {
434
-	struct kernfs_root *root = kernfs_root(kn);
435443 	int v;
436444 
437445 	if (unlikely(!kn))
438446 		return;
439447 
440448 	if (kernfs_lockdep(kn))
441
-		rwsem_release(&kn->dep_map, 1, _RET_IP_);
449
+		rwsem_release(&kn->dep_map, _RET_IP_);
442450 	v = atomic_dec_return(&kn->active);
443451 	if (likely(v != KN_DEACTIVATED_BIAS))
444452 		return;
445453 
446
-	wake_up_all(&root->deactivate_waitq);
454
+	wake_up_all(&kernfs_root(kn)->deactivate_waitq);
447455 }
448456 
449457 /**
....@@ -476,7 +484,7 @@
476484 
477485 	if (kernfs_lockdep(kn)) {
478486 		lock_acquired(&kn->dep_map, _RET_IP_);
479
-		rwsem_release(&kn->dep_map, 1, _RET_IP_);
487
+		rwsem_release(&kn->dep_map, _RET_IP_);
480488 	}
481489 
482490 	kernfs_drain_open_files(kn);
....@@ -508,10 +516,6 @@
508516 	struct kernfs_node *parent;
509517 	struct kernfs_root *root;
510518 
511
-	/*
512
-	 * kernfs_node is freed with ->count 0, kernfs_find_and_get_node_by_ino
513
-	 * depends on this to filter reused stale node
514
-	 */
515519 	if (!kn || !atomic_dec_and_test(&kn->count))
516520 		return;
517521 	root = kernfs_root(kn);
....@@ -532,14 +536,11 @@
532536 	kfree_const(kn->name);
533537 
534538 	if (kn->iattr) {
535
-		if (kn->iattr->ia_secdata)
536
-			security_release_secctx(kn->iattr->ia_secdata,
537
-						kn->iattr->ia_secdata_len);
538539 		simple_xattrs_free(&kn->iattr->xattrs);
540
+		kmem_cache_free(kernfs_iattrs_cache, kn->iattr);
539541 	}
540
-	kfree(kn->iattr);
541542 	spin_lock(&kernfs_idr_lock);
542
-	idr_remove(&root->ino_idr, kn->id.ino);
543
+	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
543544 	spin_unlock(&kernfs_idr_lock);
544545 	kmem_cache_free(kernfs_node_cache, kn);
545546 
....@@ -618,12 +619,13 @@
618619 }
619620 
620621 static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
622
+					     struct kernfs_node *parent,
621623 					     const char *name, umode_t mode,
622624 					     kuid_t uid, kgid_t gid,
623625 					     unsigned flags)
624626 {
625627 	struct kernfs_node *kn;
626
-	u32 gen;
628
+	u32 id_highbits;
627629 	int ret;
628630 
629631 	name = kstrdup_const(name, GFP_KERNEL);
....@@ -637,22 +639,18 @@
637639 	idr_preload(GFP_KERNEL);
638640 	spin_lock(&kernfs_idr_lock);
639641 	ret = idr_alloc_cyclic(&root->ino_idr, kn, 1, 0, GFP_ATOMIC);
640
-	if (ret >= 0 && ret < root->last_ino)
641
-		root->next_generation++;
642
-	gen = root->next_generation;
643
-	root->last_ino = ret;
642
+	if (ret >= 0 && ret < root->last_id_lowbits)
643
+		root->id_highbits++;
644
+	id_highbits = root->id_highbits;
645
+	root->last_id_lowbits = ret;
644646 	spin_unlock(&kernfs_idr_lock);
645647 	idr_preload_end();
646648 	if (ret < 0)
647649 		goto err_out2;
648
-	kn->id.ino = ret;
649
-	kn->id.generation = gen;
650650 
651
-	/*
652
-	 * set ino first. This RELEASE is paired with atomic_inc_not_zero in
653
-	 * kernfs_find_and_get_node_by_ino
654
-	 */
655
-	atomic_set_release(&kn->count, 1);
651
+	kn->id = (u64)id_highbits << 32 | ret;
652
+
653
+	atomic_set(&kn->count, 1);
656654 	atomic_set(&kn->active, KN_DEACTIVATED_BIAS);
657655 	RB_CLEAR_NODE(&kn->rb);
658656 
....@@ -672,10 +670,18 @@
672670 			goto err_out3;
673671 	}
674672 
673
+	if (parent) {
674
+		ret = security_kernfs_init_security(parent, kn);
675
+		if (ret)
676
+			goto err_out3;
677
+	}
678
+
675679 	return kn;
676680 
677681  err_out3:
678
-	idr_remove(&root->ino_idr, kn->id.ino);
682
+	spin_lock(&kernfs_idr_lock);
683
+	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
684
+	spin_unlock(&kernfs_idr_lock);
679685  err_out2:
680686 	kmem_cache_free(kernfs_node_cache, kn);
681687  err_out1:
....@@ -690,7 +696,7 @@
690696 {
691697 	struct kernfs_node *kn;
692698 
693
-	kn = __kernfs_new_node(kernfs_root(parent),
699
+	kn = __kernfs_new_node(kernfs_root(parent), parent,
694700 			       name, mode, uid, gid, flags);
695701 	if (kn) {
696702 		kernfs_get(parent);
....@@ -700,50 +706,52 @@
700706 }
701707 
702708 /*
703
- * kernfs_find_and_get_node_by_ino - get kernfs_node from inode number
709
+ * kernfs_find_and_get_node_by_id - get kernfs_node from node id
704710  * @root: the kernfs root
705
- * @ino: inode number
711
+ * @id: the target node id
712
+ *
713
+ * @id's lower 32bits encode ino and upper gen.  If the gen portion is
714
+ * zero, all generations are matched.
706715  *
707716  * RETURNS:
708717  * NULL on failure. Return a kernfs node with reference counter incremented
709718  */
710
-struct kernfs_node *kernfs_find_and_get_node_by_ino(struct kernfs_root *root,
711
-						    unsigned int ino)
719
+struct kernfs_node *kernfs_find_and_get_node_by_id(struct kernfs_root *root,
720
+						   u64 id)
712721 {
713722 	struct kernfs_node *kn;
723
+	ino_t ino = kernfs_id_ino(id);
724
+	u32 gen = kernfs_id_gen(id);
714725 
715
-	rcu_read_lock();
716
-	kn = idr_find(&root->ino_idr, ino);
726
+	spin_lock(&kernfs_idr_lock);
727
+
728
+	kn = idr_find(&root->ino_idr, (u32)ino);
717729 	if (!kn)
718
-		goto out;
730
+		goto err_unlock;
719731 
720
-	/*
721
-	 * Since kernfs_node is freed in RCU, it's possible an old node for ino
722
-	 * is freed, but reused before RCU grace period. But a freed node (see
723
-	 * kernfs_put) or an incompletedly initialized node (see
724
-	 * __kernfs_new_node) should have 'count' 0. We can use this fact to
725
-	 * filter out such node.
726
-	 */
727
-	if (!atomic_inc_not_zero(&kn->count)) {
728
-		kn = NULL;
729
-		goto out;
732
+	if (sizeof(ino_t) >= sizeof(u64)) {
733
+		/* we looked up with the low 32bits, compare the whole */
734
+		if (kernfs_ino(kn) != ino)
735
+			goto err_unlock;
736
+	} else {
737
+		/* 0 matches all generations */
738
+		if (unlikely(gen && kernfs_gen(kn) != gen))
739
+			goto err_unlock;
730740 	}
731741 
732742 	/*
733
-	 * The node could be a new node or a reused node. If it's a new node,
734
-	 * we are ok. If it's reused because of RCU (because of
735
-	 * SLAB_TYPESAFE_BY_RCU), the __kernfs_new_node always sets its 'ino'
736
-	 * before 'count'. So if 'count' is uptodate, 'ino' should be uptodate,
737
-	 * hence we can use 'ino' to filter stale node.
743
+	 * ACTIVATED is protected with kernfs_mutex but it was clear when
744
+	 * @kn was added to idr and we just wanna see it set.  No need to
745
+	 * grab kernfs_mutex.
738746 	 */
739
-	if (kn->id.ino != ino)
740
-		goto out;
741
-	rcu_read_unlock();
747
+	if (unlikely(!(kn->flags & KERNFS_ACTIVATED) ||
748
+		     !atomic_inc_not_zero(&kn->count)))
749
+		goto err_unlock;
742750 
751
+	spin_unlock(&kernfs_idr_lock);
743752 	return kn;
744
-out:
745
-	rcu_read_unlock();
746
-	kernfs_put(kn);
753
+err_unlock:
754
+	spin_unlock(&kernfs_idr_lock);
747755 	return NULL;
748756 }
749757 
....@@ -793,9 +801,8 @@
793801 	/* Update timestamps on the parent */
794802 	ps_iattr = parent->iattr;
795803 	if (ps_iattr) {
796
-		struct iattr *ps_iattrs = &ps_iattr->ia_iattr;
797
-		ktime_get_real_ts64(&ps_iattrs->ia_ctime);
798
-		ps_iattrs->ia_mtime = ps_iattrs->ia_ctime;
804
+		ktime_get_real_ts64(&ps_iattr->ia_ctime);
805
+		ps_iattr->ia_mtime = ps_iattr->ia_ctime;
799806 	}
800807 
801808 	mutex_unlock(&kernfs_mutex);
....@@ -867,13 +874,12 @@
867874 
868875 	lockdep_assert_held(&kernfs_mutex);
869876 
870
-	/* grab kernfs_rename_lock to piggy back on kernfs_pr_cont_buf */
871
-	spin_lock_irq(&kernfs_rename_lock);
877
+	spin_lock_irq(&kernfs_pr_cont_lock);
872878 
873879 	len = strlcpy(kernfs_pr_cont_buf, path, sizeof(kernfs_pr_cont_buf));
874880 
875881 	if (len >= sizeof(kernfs_pr_cont_buf)) {
876
-		spin_unlock_irq(&kernfs_rename_lock);
882
+		spin_unlock_irq(&kernfs_pr_cont_lock);
877883 		return NULL;
878884 	}
879885 
....@@ -885,7 +891,7 @@
885891 		parent = kernfs_find_ns(parent, name, ns);
886892 	}
887893 
888
-	spin_unlock_irq(&kernfs_rename_lock);
894
+	spin_unlock_irq(&kernfs_pr_cont_lock);
889895 
890896 	return parent;
891897 }
....@@ -958,9 +964,19 @@
958964 
959965 	idr_init(&root->ino_idr);
960966 	INIT_LIST_HEAD(&root->supers);
961
-	root->next_generation = 1;
962967 
963
-	kn = __kernfs_new_node(root, "", S_IFDIR | S_IRUGO | S_IXUGO,
968
+	/*
969
+	 * On 64bit ino setups, id is ino.  On 32bit, low 32bits are ino.
970
+	 * High bits generation.  The starting value for both ino and
971
+	 * genenration is 1.  Initialize upper 32bit allocation
972
+	 * accordingly.
973
+	 */
974
+	if (sizeof(ino_t) >= sizeof(u64))
975
+		root->id_highbits = 0;
976
+	else
977
+		root->id_highbits = 1;
978
+
979
+	kn = __kernfs_new_node(root, NULL, "", S_IFDIR | S_IRUGO | S_IXUGO,
964980 			       GLOBAL_ROOT_UID, GLOBAL_ROOT_GID,
965981 			       KERNFS_DIR);
966982 	if (!kn) {
....@@ -1259,7 +1275,7 @@
12591275 
12601276 	pos = NULL;
12611277 	while ((pos = kernfs_next_descendant_post(pos, kn))) {
1262
-		if (!pos || (pos->flags & KERNFS_ACTIVATED))
1278
+		if (pos->flags & KERNFS_ACTIVATED)
12631279 			continue;
12641280 
12651281 		WARN_ON_ONCE(pos->parent && RB_EMPTY_NODE(&pos->rb));
....@@ -1327,9 +1343,8 @@
13271343 
13281344 			/* update timestamps on the parent */
13291345 			if (ps_iattr) {
1330
-				ktime_get_real_ts64(&ps_iattr->ia_iattr.ia_ctime);
1331
-				ps_iattr->ia_iattr.ia_mtime =
1332
-					ps_iattr->ia_iattr.ia_ctime;
1346
+				ktime_get_real_ts64(&ps_iattr->ia_ctime);
1347
+				ps_iattr->ia_mtime = ps_iattr->ia_ctime;
13331348 			}
13341349 
13351350 			kernfs_put(pos);
....@@ -1506,8 +1521,11 @@
15061521 	mutex_lock(&kernfs_mutex);
15071522 
15081523 	kn = kernfs_find_ns(parent, name, ns);
1509
-	if (kn)
1524
+	if (kn) {
1525
+		kernfs_get(kn);
15101526 		__kernfs_remove(kn);
1527
+		kernfs_put(kn);
1528
+	}
15111529 
15121530 	mutex_unlock(&kernfs_mutex);
15131531 
....@@ -1675,7 +1693,7 @@
16751693 		const char *name = pos->name;
16761694 		unsigned int type = dt_type(pos);
16771695 		int len = strlen(name);
1678
-		ino_t ino = pos->id.ino;
1696
+		ino_t ino = kernfs_ino(pos);
16791697 
16801698 		ctx->pos = pos->hash;
16811699 		file->private_data = pos;