kernel/mutex-debug.h


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53

/*
 * Mutexes: blocking mutual exclusion locks
 *
 * started by Ingo Molnar:
 *
 *  Copyright (C) 2004, 2005, 2006 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
 *
 * This file contains mutex debugging related internal declarations,
 * prototypes and inline functions, for the CONFIG_DEBUG_MUTEXES case.
 * More details are in kernel/mutex-debug.c.
 */

/*
 * This must be called with lock->wait_lock held.
 */
extern void
debug_mutex_set_owner(struct mutex *lock, struct thread_info *new_owner);

static inline void debug_mutex_clear_owner(struct mutex *lock)
{
	lock->owner = NULL;
}

extern void debug_mutex_lock_common(struct mutex *lock,
				    struct mutex_waiter *waiter);
extern void debug_mutex_wake_waiter(struct mutex *lock,
				    struct mutex_waiter *waiter);
extern void debug_mutex_free_waiter(struct mutex_waiter *waiter);
extern void debug_mutex_add_waiter(struct mutex *lock,
				   struct mutex_waiter *waiter,
				   struct thread_info *ti);
extern void mutex_remove_waiter(struct mutex *lock, struct mutex_waiter *waiter,
				struct thread_info *ti);
extern void debug_mutex_unlock(struct mutex *lock);
extern void debug_mutex_init(struct mutex *lock, const char *name,
			     struct lock_class_key *key);

#define spin_lock_mutex(lock, flags)			\
	do {						\
		struct mutex *l = container_of(lock, struct mutex, wait_lock); \
							\
		DEBUG_LOCKS_WARN_ON(in_interrupt());	\
		local_irq_save(flags);			\
		__raw_spin_lock(&(lock)->raw_lock);	\
		DEBUG_LOCKS_WARN_ON(l->magic != l);	\
	} while (0)

#define spin_unlock_mutex(lock, flags)			\
	do {						\
		__raw_spin_unlock(&(lock)->raw_lock);	\
		local_irq_restore(flags);		\
		preempt_check_resched();		\
	} while (0)
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/pagemap.h>

#include "ctree.h"
#include "disk-io.h"
#include "free-space-cache.h"
#include "inode-map.h"
#include "transaction.h"

static int caching_kthread(void *data)
{
	struct btrfs_root *root = data;
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	u64 last = (u64)-1;
	int slot;
	int ret;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	/* Since the commit root is read-only, we can safely skip locking. */
	path->skip_locking = 1;
	path->search_commit_root = 1;
	path->reada = 2;

	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
	key.offset = 0;
	key.type = BTRFS_INODE_ITEM_KEY;
again:
	/* need to make sure the commit_root doesn't disappear */
	mutex_lock(&root->fs_commit_mutex);

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto out;

	while (1) {
		if (btrfs_fs_closing(fs_info))
			goto out;

		leaf = path->nodes[0];
		slot = path->slots[0];
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto out;
			else if (ret > 0)
				break;

			if (need_resched() ||
			    btrfs_transaction_in_commit(fs_info)) {
				leaf = path->nodes[0];

				if (btrfs_header_nritems(leaf) == 0) {
					WARN_ON(1);
					break;
				}

				/*
				 * Save the key so we can advances forward
				 * in the next search.
				 */
				btrfs_item_key_to_cpu(leaf, &key, 0);
				btrfs_release_path(path);
				root->cache_progress = last;
				mutex_unlock(&root->fs_commit_mutex);
				schedule_timeout(1);
				goto again;
			} else
				continue;
		}

		btrfs_item_key_to_cpu(leaf, &key, slot);

		if (key.type != BTRFS_INODE_ITEM_KEY)
			goto next;

		if (key.objectid >= root->highest_objectid)
			break;

		if (last != (u64)-1 && last + 1 != key.objectid) {
			__btrfs_add_free_space(ctl, last + 1,
					       key.objectid - last - 1);
			wake_up(&root->cache_wait);
		}

		last = key.objectid;
next:
		path->slots[0]++;
	}

	if (last < root->highest_objectid - 1) {
		__btrfs_add_free_space(ctl, last + 1,
				       root->highest_objectid - last - 1);
	}

	spin_lock(&root->cache_lock);
	root->cached = BTRFS_CACHE_FINISHED;
	spin_unlock(&root->cache_lock);

	root->cache_progress = (u64)-1;
	btrfs_unpin_free_ino(root);
out:
	wake_up(&root->cache_wait);
	mutex_unlock(&root->fs_commit_mutex);

	btrfs_free_path(path);

	return ret;
}

static void start_caching(struct btrfs_root *root)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct task_struct *tsk;
	int ret;
	u64 objectid;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

	spin_lock(&root->cache_lock);
	if (root->cached != BTRFS_CACHE_NO) {
		spin_unlock(&root->cache_lock);
		return;
	}

	root->cached = BTRFS_CACHE_STARTED;
	spin_unlock(&root->cache_lock);

	ret = load_free_ino_cache(root->fs_info, root);
	if (ret == 1) {
		spin_lock(&root->cache_lock);
		root->cached = BTRFS_CACHE_FINISHED;
		spin_unlock(&root->cache_lock);
		return;
	}

	/*
	 * It can be quite time-consuming to fill the cache by searching
	 * through the extent tree, and this can keep ino allocation path
	 * waiting. Therefore at start we quickly find out the highest
	 * inode number and we know we can use inode numbers which fall in
	 * [highest_ino + 1, BTRFS_LAST_FREE_OBJECTID].
	 */
	ret = btrfs_find_free_objectid(root, &objectid);
	if (!ret && objectid <= BTRFS_LAST_FREE_OBJECTID) {
		__btrfs_add_free_space(ctl, objectid,
				       BTRFS_LAST_FREE_OBJECTID - objectid + 1);
	}

	tsk = kthread_run(caching_kthread, root, "btrfs-ino-cache-%llu\n",
			  root->root_key.objectid);
	BUG_ON(IS_ERR(tsk));
}

int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid)
{
	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return btrfs_find_free_objectid(root, objectid);

again:
	*objectid = btrfs_find_ino_for_alloc(root);

	if (*objectid != 0)
		return 0;

	start_caching(root);

	wait_event(root->cache_wait,
		   root->cached == BTRFS_CACHE_FINISHED ||
		   root->free_ino_ctl->free_space > 0);

	if (root->cached == BTRFS_CACHE_FINISHED &&
	    root->free_ino_ctl->free_space == 0)
		return -ENOSPC;
	else
		goto again;
}

void btrfs_return_ino(struct btrfs_root *root, u64 objectid)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct btrfs_free_space_ctl *pinned = root->free_ino_pinned;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

again:
	if (root->cached == BTRFS_CACHE_FINISHED) {
		__btrfs_add_free_space(ctl, objectid, 1);
	} else {
		/*
		 * If we are in the process of caching free ino chunks,
		 * to avoid adding the same inode number to the free_ino
		 * tree twice due to cross transaction, we'll leave it
		 * in the pinned tree until a transaction is committed
		 * or the caching work is done.
		 */

		mutex_lock(&root->fs_commit_mutex);
		spin_lock(&root->cache_lock);
		if (root->cached == BTRFS_CACHE_FINISHED) {
			spin_unlock(&root->cache_lock);
			mutex_unlock(&root->fs_commit_mutex);
			goto again;
		}
		spin_unlock(&root->cache_lock);

		start_caching(root);

		if (objectid <= root->cache_progress ||
		    objectid > root->highest_objectid)
			__btrfs_add_free_space(ctl, objectid, 1);
		else
			__btrfs_add_free_space(pinned, objectid, 1);

		mutex_unlock(&root->fs_commit_mutex);
	}
}

/*
 * When a transaction is committed, we'll move those inode numbers which
 * are smaller than root->cache_progress from pinned tree to free_ino tree,
 * and others will just be dropped, because the commit root we were
 * searching has changed.
 *
 * Must be called with root->fs_commit_mutex held
 */
void btrfs_unpin_free_ino(struct btrfs_root *root)
{
	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
	struct rb_root *rbroot = &root->free_ino_pinned->free_space_offset;
	struct btrfs_free_space *info;
	struct rb_node *n;
	u64 count;

	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
		return;

	while (1) {
		n = rb_first(rbroot);
		if (!n)
			break;

		info = rb_entry(n, struct btrfs_free_space, offset_index);
		BUG_ON(info->bitmap);

		if (info->offset > root->cache_progress)
			goto free;
		else if (info->offset + info->bytes > root->cache_progress)
			count = root->cache_progress - info->offset + 1;
		else
			count = info->bytes;

		__btrfs_add_free_space(ctl, info->offset, count);
free:
		rb_erase(&info->offset_index, rbroot);
		kfree(info);
	}
}

#define INIT_THRESHOLD	(((1024 * 32) / 2) / sizeof(struct btrfs_free_space))
#define INODES_PER_BITMAP (PAGE_CACHE_SIZE * 8)

/*
 * The goal is to keep the memory used by the free_ino tree won't
 * exceed the memory if we use bitmaps only.
 */
static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
{
	struct btrfs_free_space *info;
	struct rb_node *n;
	int max_ino;
	int max_bitmaps;

	n = rb_last(&ctl->free_space_offset);
	if (!n) {
		ctl->extents_thresh = INIT_THRESHOLD;
		return;
	}
	info = rb_entry(n, struct btrfs_free_space, offset_index);

	/*
	 * Find the maximum inode number in the filesystem. Note we
	 * ignore the fact that this can be a bitmap, because we are
	 * not doing precise calculation.
	 */
	max_ino = info->bytes - 1;

	max_bitmaps = ALIGN(max_ino, INODES_PER_BITMAP) / INODES_PER_BITMAP;
	if (max_bitmaps <= ctl->total_bitmaps) {
		ctl->extents_thresh = 0;
		return;
	}

	ctl->extents_thresh = (max_bitmaps - ctl->total_bitmaps) *
				PAGE_CACHE_SIZE / sizeof(*info);
}

/*