linux/fs/hfs/bnode.c
Hin-Tak Leung 7cb74be6fd hfs,hfsplus: cache pages correctly between bnode_create and bnode_free
Pages looked up by __hfs_bnode_create() (called by hfs_bnode_create() and
hfs_bnode_find() for finding or creating pages corresponding to an inode)
are immediately kmap()'ed and used (both read and write) and kunmap()'ed,
and should not be page_cache_release()'ed until hfs_bnode_free().

This patch fixes a problem I first saw in July 2012: merely running "du"
on a large hfsplus-mounted directory a few times on a reasonably loaded
system would get the hfsplus driver all confused and complaining about
B-tree inconsistencies, and generates a "BUG: Bad page state".  Most
recently, I can generate this problem on up-to-date Fedora 22 with shipped
kernel 4.0.5, by running "du /" (="/" + "/home" + "/mnt" + other smaller
mounts) and "du /mnt" simultaneously on two windows, where /mnt is a
lightly-used QEMU VM image of the full Mac OS X 10.9:

$ df -i / /home /mnt
Filesystem                  Inodes   IUsed      IFree IUse% Mounted on
/dev/mapper/fedora-root    3276800  551665    2725135   17% /
/dev/mapper/fedora-home   52879360  716221   52163139    2% /home
/dev/nbd0p2             4294967295 1387818 4293579477    1% /mnt

After applying the patch, I was able to run "du /" (60+ times) and "du
/mnt" (150+ times) continuously and simultaneously for 6+ hours.

There are many reports of the hfsplus driver getting confused under load
and generating "BUG: Bad page state" or other similar issues over the
years.  [1]

The unpatched code [2] has always been wrong since it entered the kernel
tree.  The only reason why it gets away with it is that the
kmap/memcpy/kunmap follow very quickly after the page_cache_release() so
the kernel has not had a chance to reuse the memory for something else,
most of the time.

The current RW driver appears to have followed the design and development
of the earlier read-only hfsplus driver [3], where-by version 0.1 (Dec
2001) had a B-tree node-centric approach to
read_cache_page()/page_cache_release() per bnode_get()/bnode_put(),
migrating towards version 0.2 (June 2002) of caching and releasing pages
per inode extents.  When the current RW code first entered the kernel [2]
in 2005, there was an REF_PAGES conditional (and "//" commented out code)
to switch between B-node centric paging to inode-centric paging.  There
was a mistake with the direction of one of the REF_PAGES conditionals in
__hfs_bnode_create().  In a subsequent "remove debug code" commit [4], the
read_cache_page()/page_cache_release() per bnode_get()/bnode_put() were
removed, but a page_cache_release() was mistakenly left in (propagating
the "REF_PAGES <-> !REF_PAGE" mistake), and the commented-out
page_cache_release() in bnode_release() (which should be spanned by
!REF_PAGES) was never enabled.

References:
[1]:
Michael Fox, Apr 2013
http://www.spinics.net/lists/linux-fsdevel/msg63807.html
("hfsplus volume suddenly inaccessable after 'hfs: recoff %d too large'")

Sasha Levin, Feb 2015
http://lkml.org/lkml/2015/2/20/85 ("use after free")

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/740814
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1027887
https://bugzilla.kernel.org/show_bug.cgi?id=42342
https://bugzilla.kernel.org/show_bug.cgi?id=63841
https://bugzilla.kernel.org/show_bug.cgi?id=78761

[2]:
http://git.kernel.org/cgit/linux/kernel/git/tglx/history.git/commit/\
fs/hfs/bnode.c?id=d1081202f1d0ee35ab0beb490da4b65d4bc763db
commit d1081202f1d0ee35ab0beb490da4b65d4bc763db
Author: Andrew Morton <akpm@osdl.org>
Date:   Wed Feb 25 16:17:36 2004 -0800

    [PATCH] HFS rewrite

http://git.kernel.org/cgit/linux/kernel/git/tglx/history.git/commit/\
fs/hfsplus/bnode.c?id=91556682e0bf004d98a529bf829d339abb98bbbd

commit 91556682e0bf004d98a529bf829d339abb98bbbd
Author: Andrew Morton <akpm@osdl.org>
Date:   Wed Feb 25 16:17:48 2004 -0800

    [PATCH] HFS+ support

[3]:
http://sourceforge.net/projects/linux-hfsplus/

http://sourceforge.net/projects/linux-hfsplus/files/Linux%202.4.x%20patch/hfsplus%200.1/
http://sourceforge.net/projects/linux-hfsplus/files/Linux%202.4.x%20patch/hfsplus%200.2/

http://linux-hfsplus.cvs.sourceforge.net/viewvc/linux-hfsplus/linux/\
fs/hfsplus/bnode.c?r1=1.4&r2=1.5

Date:   Thu Jun 6 09:45:14 2002 +0000
Use buffer cache instead of page cache in bnode.c. Cache inode extents.

[4]:
http://git.kernel.org/cgit/linux/kernel/git/\
stable/linux-stable.git/commit/?id=a5e3985fa014029eb6795664c704953720cc7f7d

commit a5e3985fa0
Author: Roman Zippel <zippel@linux-m68k.org>
Date:   Tue Sep 6 15:18:47 2005 -0700

[PATCH] hfs: remove debug code

Signed-off-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: Sergei Antonov <saproj@gmail.com>
Reviewed-by: Anton Altaparmakov <anton@tuxera.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Vyacheslav Dubeyko <slava@dubeyko.com>
Cc: Sougata Santra <sougata@tuxera.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-10 13:29:01 -07:00

485 lines
11 KiB
C

/*
* linux/fs/hfs/bnode.c
*
* Copyright (C) 2001
* Brad Boyer (flar@allandria.com)
* (C) 2003 Ardis Technologies <roman@ardistech.com>
*
* Handle basic btree node operations
*/
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/swap.h>
#include "btree.h"
void hfs_bnode_read(struct hfs_bnode *node, void *buf,
int off, int len)
{
struct page *page;
off += node->page_offset;
page = node->page[0];
memcpy(buf, kmap(page) + off, len);
kunmap(page);
}
u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off)
{
__be16 data;
// optimize later...
hfs_bnode_read(node, &data, off, 2);
return be16_to_cpu(data);
}
u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off)
{
u8 data;
// optimize later...
hfs_bnode_read(node, &data, off, 1);
return data;
}
void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off)
{
struct hfs_btree *tree;
int key_len;
tree = node->tree;
if (node->type == HFS_NODE_LEAF ||
tree->attributes & HFS_TREE_VARIDXKEYS)
key_len = hfs_bnode_read_u8(node, off) + 1;
else
key_len = tree->max_key_len + 1;
hfs_bnode_read(node, key, off, key_len);
}
void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
{
struct page *page;
off += node->page_offset;
page = node->page[0];
memcpy(kmap(page) + off, buf, len);
kunmap(page);
set_page_dirty(page);
}
void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data)
{
__be16 v = cpu_to_be16(data);
// optimize later...
hfs_bnode_write(node, &v, off, 2);
}
void hfs_bnode_write_u8(struct hfs_bnode *node, int off, u8 data)
{
// optimize later...
hfs_bnode_write(node, &data, off, 1);
}
void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
{
struct page *page;
off += node->page_offset;
page = node->page[0];
memset(kmap(page) + off, 0, len);
kunmap(page);
set_page_dirty(page);
}
void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
struct hfs_bnode *src_node, int src, int len)
{
struct hfs_btree *tree;
struct page *src_page, *dst_page;
hfs_dbg(BNODE_MOD, "copybytes: %u,%u,%u\n", dst, src, len);
if (!len)
return;
tree = src_node->tree;
src += src_node->page_offset;
dst += dst_node->page_offset;
src_page = src_node->page[0];
dst_page = dst_node->page[0];
memcpy(kmap(dst_page) + dst, kmap(src_page) + src, len);
kunmap(src_page);
kunmap(dst_page);
set_page_dirty(dst_page);
}
void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
{
struct page *page;
void *ptr;
hfs_dbg(BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
if (!len)
return;
src += node->page_offset;
dst += node->page_offset;
page = node->page[0];
ptr = kmap(page);
memmove(ptr + dst, ptr + src, len);
kunmap(page);
set_page_dirty(page);
}
void hfs_bnode_dump(struct hfs_bnode *node)
{
struct hfs_bnode_desc desc;
__be32 cnid;
int i, off, key_off;
hfs_dbg(BNODE_MOD, "bnode: %d\n", node->this);
hfs_bnode_read(node, &desc, 0, sizeof(desc));
hfs_dbg(BNODE_MOD, "%d, %d, %d, %d, %d\n",
be32_to_cpu(desc.next), be32_to_cpu(desc.prev),
desc.type, desc.height, be16_to_cpu(desc.num_recs));
off = node->tree->node_size - 2;
for (i = be16_to_cpu(desc.num_recs); i >= 0; off -= 2, i--) {
key_off = hfs_bnode_read_u16(node, off);
hfs_dbg_cont(BNODE_MOD, " %d", key_off);
if (i && node->type == HFS_NODE_INDEX) {
int tmp;
if (node->tree->attributes & HFS_TREE_VARIDXKEYS)
tmp = (hfs_bnode_read_u8(node, key_off) | 1) + 1;
else
tmp = node->tree->max_key_len + 1;
hfs_dbg_cont(BNODE_MOD, " (%d,%d",
tmp, hfs_bnode_read_u8(node, key_off));
hfs_bnode_read(node, &cnid, key_off + tmp, 4);
hfs_dbg_cont(BNODE_MOD, ",%d)", be32_to_cpu(cnid));
} else if (i && node->type == HFS_NODE_LEAF) {
int tmp;
tmp = hfs_bnode_read_u8(node, key_off);
hfs_dbg_cont(BNODE_MOD, " (%d)", tmp);
}
}
hfs_dbg_cont(BNODE_MOD, "\n");
}
void hfs_bnode_unlink(struct hfs_bnode *node)
{
struct hfs_btree *tree;
struct hfs_bnode *tmp;
__be32 cnid;
tree = node->tree;
if (node->prev) {
tmp = hfs_bnode_find(tree, node->prev);
if (IS_ERR(tmp))
return;
tmp->next = node->next;
cnid = cpu_to_be32(tmp->next);
hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, next), 4);
hfs_bnode_put(tmp);
} else if (node->type == HFS_NODE_LEAF)
tree->leaf_head = node->next;
if (node->next) {
tmp = hfs_bnode_find(tree, node->next);
if (IS_ERR(tmp))
return;
tmp->prev = node->prev;
cnid = cpu_to_be32(tmp->prev);
hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, prev), 4);
hfs_bnode_put(tmp);
} else if (node->type == HFS_NODE_LEAF)
tree->leaf_tail = node->prev;
// move down?
if (!node->prev && !node->next) {
printk(KERN_DEBUG "hfs_btree_del_level\n");
}
if (!node->parent) {
tree->root = 0;
tree->depth = 0;
}
set_bit(HFS_BNODE_DELETED, &node->flags);
}
static inline int hfs_bnode_hash(u32 num)
{
num = (num >> 16) + num;
num += num >> 8;
return num & (NODE_HASH_SIZE - 1);
}
struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid)
{
struct hfs_bnode *node;
if (cnid >= tree->node_count) {
pr_err("request for non-existent node %d in B*Tree\n", cnid);
return NULL;
}
for (node = tree->node_hash[hfs_bnode_hash(cnid)];
node; node = node->next_hash) {
if (node->this == cnid) {
return node;
}
}
return NULL;
}
static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid)
{
struct super_block *sb;
struct hfs_bnode *node, *node2;
struct address_space *mapping;
struct page *page;
int size, block, i, hash;
loff_t off;
if (cnid >= tree->node_count) {
pr_err("request for non-existent node %d in B*Tree\n", cnid);
return NULL;
}
sb = tree->inode->i_sb;
size = sizeof(struct hfs_bnode) + tree->pages_per_bnode *
sizeof(struct page *);
node = kzalloc(size, GFP_KERNEL);
if (!node)
return NULL;
node->tree = tree;
node->this = cnid;
set_bit(HFS_BNODE_NEW, &node->flags);
atomic_set(&node->refcnt, 1);
hfs_dbg(BNODE_REFS, "new_node(%d:%d): 1\n",
node->tree->cnid, node->this);
init_waitqueue_head(&node->lock_wq);
spin_lock(&tree->hash_lock);
node2 = hfs_bnode_findhash(tree, cnid);
if (!node2) {
hash = hfs_bnode_hash(cnid);
node->next_hash = tree->node_hash[hash];
tree->node_hash[hash] = node;
tree->node_hash_cnt++;
} else {
spin_unlock(&tree->hash_lock);
kfree(node);
wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags));
return node2;
}
spin_unlock(&tree->hash_lock);
mapping = tree->inode->i_mapping;
off = (loff_t)cnid * tree->node_size;
block = off >> PAGE_CACHE_SHIFT;
node->page_offset = off & ~PAGE_CACHE_MASK;
for (i = 0; i < tree->pages_per_bnode; i++) {
page = read_mapping_page(mapping, block++, NULL);
if (IS_ERR(page))
goto fail;
if (PageError(page)) {
page_cache_release(page);
goto fail;
}
node->page[i] = page;
}
return node;
fail:
set_bit(HFS_BNODE_ERROR, &node->flags);
return node;
}
void hfs_bnode_unhash(struct hfs_bnode *node)
{
struct hfs_bnode **p;
hfs_dbg(BNODE_REFS, "remove_node(%d:%d): %d\n",
node->tree->cnid, node->this, atomic_read(&node->refcnt));
for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)];
*p && *p != node; p = &(*p)->next_hash)
;
BUG_ON(!*p);
*p = node->next_hash;
node->tree->node_hash_cnt--;
}
/* Load a particular node out of a tree */
struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
{
struct hfs_bnode *node;
struct hfs_bnode_desc *desc;
int i, rec_off, off, next_off;
int entry_size, key_size;
spin_lock(&tree->hash_lock);
node = hfs_bnode_findhash(tree, num);
if (node) {
hfs_bnode_get(node);
spin_unlock(&tree->hash_lock);
wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags));
if (test_bit(HFS_BNODE_ERROR, &node->flags))
goto node_error;
return node;
}
spin_unlock(&tree->hash_lock);
node = __hfs_bnode_create(tree, num);
if (!node)
return ERR_PTR(-ENOMEM);
if (test_bit(HFS_BNODE_ERROR, &node->flags))
goto node_error;
if (!test_bit(HFS_BNODE_NEW, &node->flags))
return node;
desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
node->prev = be32_to_cpu(desc->prev);
node->next = be32_to_cpu(desc->next);
node->num_recs = be16_to_cpu(desc->num_recs);
node->type = desc->type;
node->height = desc->height;
kunmap(node->page[0]);
switch (node->type) {
case HFS_NODE_HEADER:
case HFS_NODE_MAP:
if (node->height != 0)
goto node_error;
break;
case HFS_NODE_LEAF:
if (node->height != 1)
goto node_error;
break;
case HFS_NODE_INDEX:
if (node->height <= 1 || node->height > tree->depth)
goto node_error;
break;
default:
goto node_error;
}
rec_off = tree->node_size - 2;
off = hfs_bnode_read_u16(node, rec_off);
if (off != sizeof(struct hfs_bnode_desc))
goto node_error;
for (i = 1; i <= node->num_recs; off = next_off, i++) {
rec_off -= 2;
next_off = hfs_bnode_read_u16(node, rec_off);
if (next_off <= off ||
next_off > tree->node_size ||
next_off & 1)
goto node_error;
entry_size = next_off - off;
if (node->type != HFS_NODE_INDEX &&
node->type != HFS_NODE_LEAF)
continue;
key_size = hfs_bnode_read_u8(node, off) + 1;
if (key_size >= entry_size /*|| key_size & 1*/)
goto node_error;
}
clear_bit(HFS_BNODE_NEW, &node->flags);
wake_up(&node->lock_wq);
return node;
node_error:
set_bit(HFS_BNODE_ERROR, &node->flags);
clear_bit(HFS_BNODE_NEW, &node->flags);
wake_up(&node->lock_wq);
hfs_bnode_put(node);
return ERR_PTR(-EIO);
}
void hfs_bnode_free(struct hfs_bnode *node)
{
int i;
for (i = 0; i < node->tree->pages_per_bnode; i++)
if (node->page[i])
page_cache_release(node->page[i]);
kfree(node);
}
struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
{
struct hfs_bnode *node;
struct page **pagep;
int i;
spin_lock(&tree->hash_lock);
node = hfs_bnode_findhash(tree, num);
spin_unlock(&tree->hash_lock);
if (node) {
pr_crit("new node %u already hashed?\n", num);
WARN_ON(1);
return node;
}
node = __hfs_bnode_create(tree, num);
if (!node)
return ERR_PTR(-ENOMEM);
if (test_bit(HFS_BNODE_ERROR, &node->flags)) {
hfs_bnode_put(node);
return ERR_PTR(-EIO);
}
pagep = node->page;
memset(kmap(*pagep) + node->page_offset, 0,
min((int)PAGE_CACHE_SIZE, (int)tree->node_size));
set_page_dirty(*pagep);
kunmap(*pagep);
for (i = 1; i < tree->pages_per_bnode; i++) {
memset(kmap(*++pagep), 0, PAGE_CACHE_SIZE);
set_page_dirty(*pagep);
kunmap(*pagep);
}
clear_bit(HFS_BNODE_NEW, &node->flags);
wake_up(&node->lock_wq);
return node;
}
void hfs_bnode_get(struct hfs_bnode *node)
{
if (node) {
atomic_inc(&node->refcnt);
hfs_dbg(BNODE_REFS, "get_node(%d:%d): %d\n",
node->tree->cnid, node->this,
atomic_read(&node->refcnt));
}
}
/* Dispose of resources used by a node */
void hfs_bnode_put(struct hfs_bnode *node)
{
if (node) {
struct hfs_btree *tree = node->tree;
int i;
hfs_dbg(BNODE_REFS, "put_node(%d:%d): %d\n",
node->tree->cnid, node->this,
atomic_read(&node->refcnt));
BUG_ON(!atomic_read(&node->refcnt));
if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock))
return;
for (i = 0; i < tree->pages_per_bnode; i++) {
if (!node->page[i])
continue;
mark_page_accessed(node->page[i]);
}
if (test_bit(HFS_BNODE_DELETED, &node->flags)) {
hfs_bnode_unhash(node);
spin_unlock(&tree->hash_lock);
hfs_bmap_free(node);
hfs_bnode_free(node);
return;
}
spin_unlock(&tree->hash_lock);
}
}