tmp_suning_uos_patched/fs/ext3/xattr.c

1323 lines
34 KiB
C
Raw Normal View History

/*
* linux/fs/ext3/xattr.c
*
* Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
*
* Fix by Harrison Xing <harrison@mountainviewdata.com>.
* Ext3 code with a lot of help from Eric Jarman <ejarman@acm.org>.
* Extended attributes for symlinks and special files added per
* suggestion of Luka Renko <luka.renko@hermes.si>.
* xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
* Red Hat Inc.
* ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
* and Andreas Gruenbacher <agruen@suse.de>.
*/
/*
* Extended attributes are stored directly in inodes (on file systems with
* inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
* field contains the block number if an inode uses an additional block. All
* attributes must fit in the inode and one additional block. Blocks that
* contain the identical set of attributes may be shared among several inodes.
* Identical blocks are detected by keeping a cache of blocks that have
* recently been accessed.
*
* The attributes in inodes and on blocks have a different header; the entries
* are stored in the same format:
*
* +------------------+
* | header |
* | entry 1 | |
* | entry 2 | | growing downwards
* | entry 3 | v
* | four null bytes |
* | . . . |
* | value 1 | ^
* | value 3 | | growing upwards
* | value 2 | |
* +------------------+
*
* The header is followed by multiple entry descriptors. In disk blocks, the
* entry descriptors are kept sorted. In inodes, they are unsorted. The
* attribute values are aligned to the end of the block in no specific order.
*
* Locking strategy
* ----------------
* EXT3_I(inode)->i_file_acl is protected by EXT3_I(inode)->xattr_sem.
* EA blocks are only changed if they are exclusive to an inode, so
* holding xattr_sem also means that nothing but the EA block's reference
* count can change. Multiple writers to the same block are synchronized
* by the buffer lock.
*/
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/ext3_jbd.h>
#include <linux/ext3_fs.h>
#include <linux/mbcache.h>
#include <linux/quotaops.h>
#include <linux/rwsem.h>
#include "xattr.h"
#include "acl.h"
#define BHDR(bh) ((struct ext3_xattr_header *)((bh)->b_data))
#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr))
#define BFIRST(bh) ENTRY(BHDR(bh)+1)
#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
#define IHDR(inode, raw_inode) \
((struct ext3_xattr_ibody_header *) \
((void *)raw_inode + \
EXT3_GOOD_OLD_INODE_SIZE + \
EXT3_I(inode)->i_extra_isize))
#define IFIRST(hdr) ((struct ext3_xattr_entry *)((hdr)+1))
#ifdef EXT3_XATTR_DEBUG
# define ea_idebug(inode, f...) do { \
printk(KERN_DEBUG "inode %s:%lu: ", \
inode->i_sb->s_id, inode->i_ino); \
printk(f); \
printk("\n"); \
} while (0)
# define ea_bdebug(bh, f...) do { \
char b[BDEVNAME_SIZE]; \
printk(KERN_DEBUG "block %s:%lu: ", \
bdevname(bh->b_bdev, b), \
(unsigned long) bh->b_blocknr); \
printk(f); \
printk("\n"); \
} while (0)
#else
# define ea_idebug(f...)
# define ea_bdebug(f...)
#endif
static void ext3_xattr_cache_insert(struct buffer_head *);
static struct buffer_head *ext3_xattr_cache_find(struct inode *,
struct ext3_xattr_header *,
struct mb_cache_entry **);
static void ext3_xattr_rehash(struct ext3_xattr_header *,
struct ext3_xattr_entry *);
static struct mb_cache *ext3_xattr_cache;
static struct xattr_handler *ext3_xattr_handler_map[] = {
[EXT3_XATTR_INDEX_USER] = &ext3_xattr_user_handler,
#ifdef CONFIG_EXT3_FS_POSIX_ACL
[EXT3_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext3_xattr_acl_access_handler,
[EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler,
#endif
[EXT3_XATTR_INDEX_TRUSTED] = &ext3_xattr_trusted_handler,
#ifdef CONFIG_EXT3_FS_SECURITY
[EXT3_XATTR_INDEX_SECURITY] = &ext3_xattr_security_handler,
#endif
};
struct xattr_handler *ext3_xattr_handlers[] = {
&ext3_xattr_user_handler,
&ext3_xattr_trusted_handler,
#ifdef CONFIG_EXT3_FS_POSIX_ACL
&ext3_xattr_acl_access_handler,
&ext3_xattr_acl_default_handler,
#endif
#ifdef CONFIG_EXT3_FS_SECURITY
&ext3_xattr_security_handler,
#endif
NULL
};
static inline struct xattr_handler *
ext3_xattr_handler(int name_index)
{
struct xattr_handler *handler = NULL;
if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map))
handler = ext3_xattr_handler_map[name_index];
return handler;
}
/*
* Inode operation listxattr()
*
* dentry->d_inode->i_mutex: don't care
*/
ssize_t
ext3_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
return ext3_xattr_list(dentry->d_inode, buffer, size);
}
static int
ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end)
{
while (!IS_LAST_ENTRY(entry)) {
struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry);
if ((void *)next >= end)
return -EIO;
entry = next;
}
return 0;
}
static inline int
ext3_xattr_check_block(struct buffer_head *bh)
{
int error;
if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
BHDR(bh)->h_blocks != cpu_to_le32(1))
return -EIO;
error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
return error;
}
static inline int
ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size)
{
size_t value_size = le32_to_cpu(entry->e_value_size);
if (entry->e_value_block != 0 || value_size > size ||
le16_to_cpu(entry->e_value_offs) + value_size > size)
return -EIO;
return 0;
}
static int
ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index,
const char *name, size_t size, int sorted)
{
struct ext3_xattr_entry *entry;
size_t name_len;
int cmp = 1;
if (name == NULL)
return -EINVAL;
name_len = strlen(name);
entry = *pentry;
for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
cmp = name_index - entry->e_name_index;
if (!cmp)
cmp = name_len - entry->e_name_len;
if (!cmp)
cmp = memcmp(name, entry->e_name, name_len);
if (cmp <= 0 && (sorted || cmp == 0))
break;
}
*pentry = entry;
if (!cmp && ext3_xattr_check_entry(entry, size))
return -EIO;
return cmp ? -ENODATA : 0;
}
static int
ext3_xattr_block_get(struct inode *inode, int name_index, const char *name,
void *buffer, size_t buffer_size)
{
struct buffer_head *bh = NULL;
struct ext3_xattr_entry *entry;
size_t size;
int error;
ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
name_index, name, buffer, (long)buffer_size);
error = -ENODATA;
if (!EXT3_I(inode)->i_file_acl)
goto cleanup;
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
if (!bh)
goto cleanup;
ea_bdebug(bh, "b_count=%d, refcount=%d",
atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
if (ext3_xattr_check_block(bh)) {
bad_block: ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: bad block "E3FSBLK, inode->i_ino,
EXT3_I(inode)->i_file_acl);
error = -EIO;
goto cleanup;
}
ext3_xattr_cache_insert(bh);
entry = BFIRST(bh);
error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
if (error == -EIO)
goto bad_block;
if (error)
goto cleanup;
size = le32_to_cpu(entry->e_value_size);
if (buffer) {
error = -ERANGE;
if (size > buffer_size)
goto cleanup;
memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
size);
}
error = size;
cleanup:
brelse(bh);
return error;
}
static int
ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
void *buffer, size_t buffer_size)
{
struct ext3_xattr_ibody_header *header;
struct ext3_xattr_entry *entry;
struct ext3_inode *raw_inode;
struct ext3_iloc iloc;
size_t size;
void *end;
int error;
if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
return -ENODATA;
error = ext3_get_inode_loc(inode, &iloc);
if (error)
return error;
raw_inode = ext3_raw_inode(&iloc);
header = IHDR(inode, raw_inode);
entry = IFIRST(header);
end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
error = ext3_xattr_check_names(entry, end);
if (error)
goto cleanup;
error = ext3_xattr_find_entry(&entry, name_index, name,
end - (void *)entry, 0);
if (error)
goto cleanup;
size = le32_to_cpu(entry->e_value_size);
if (buffer) {
error = -ERANGE;
if (size > buffer_size)
goto cleanup;
memcpy(buffer, (void *)IFIRST(header) +
le16_to_cpu(entry->e_value_offs), size);
}
error = size;
cleanup:
brelse(iloc.bh);
return error;
}
/*
* ext3_xattr_get()
*
* Copy an extended attribute into the buffer
* provided, or compute the buffer size required.
* Buffer is NULL to compute the size of the buffer required.
*
* Returns a negative error number on failure, or the number of bytes
* used / required on success.
*/
int
ext3_xattr_get(struct inode *inode, int name_index, const char *name,
void *buffer, size_t buffer_size)
{
int error;
down_read(&EXT3_I(inode)->xattr_sem);
error = ext3_xattr_ibody_get(inode, name_index, name, buffer,
buffer_size);
if (error == -ENODATA)
error = ext3_xattr_block_get(inode, name_index, name, buffer,
buffer_size);
up_read(&EXT3_I(inode)->xattr_sem);
return error;
}
static int
ext3_xattr_list_entries(struct inode *inode, struct ext3_xattr_entry *entry,
char *buffer, size_t buffer_size)
{
size_t rest = buffer_size;
for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
struct xattr_handler *handler =
ext3_xattr_handler(entry->e_name_index);
if (handler) {
size_t size = handler->list(inode, buffer, rest,
entry->e_name,
entry->e_name_len);
if (buffer) {
if (size > rest)
return -ERANGE;
buffer += size;
}
rest -= size;
}
}
return buffer_size - rest;
}
static int
ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
{
struct buffer_head *bh = NULL;
int error;
ea_idebug(inode, "buffer=%p, buffer_size=%ld",
buffer, (long)buffer_size);
error = 0;
if (!EXT3_I(inode)->i_file_acl)
goto cleanup;
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
error = -EIO;
if (!bh)
goto cleanup;
ea_bdebug(bh, "b_count=%d, refcount=%d",
atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
if (ext3_xattr_check_block(bh)) {
ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: bad block "E3FSBLK, inode->i_ino,
EXT3_I(inode)->i_file_acl);
error = -EIO;
goto cleanup;
}
ext3_xattr_cache_insert(bh);
error = ext3_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
cleanup:
brelse(bh);
return error;
}
static int
ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
{
struct ext3_xattr_ibody_header *header;
struct ext3_inode *raw_inode;
struct ext3_iloc iloc;
void *end;
int error;
if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
return 0;
error = ext3_get_inode_loc(inode, &iloc);
if (error)
return error;
raw_inode = ext3_raw_inode(&iloc);
header = IHDR(inode, raw_inode);
end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
error = ext3_xattr_check_names(IFIRST(header), end);
if (error)
goto cleanup;
error = ext3_xattr_list_entries(inode, IFIRST(header),
buffer, buffer_size);
cleanup:
brelse(iloc.bh);
return error;
}
/*
* ext3_xattr_list()
*
* Copy a list of attribute names into the buffer
* provided, or compute the buffer size required.
* Buffer is NULL to compute the size of the buffer required.
*
* Returns a negative error number on failure, or the number of bytes
* used / required on success.
*/
int
ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
{
int i_error, b_error;
down_read(&EXT3_I(inode)->xattr_sem);
i_error = ext3_xattr_ibody_list(inode, buffer, buffer_size);
if (i_error < 0) {
b_error = 0;
} else {
if (buffer) {
buffer += i_error;
buffer_size -= i_error;
}
b_error = ext3_xattr_block_list(inode, buffer, buffer_size);
if (b_error < 0)
i_error = 0;
}
up_read(&EXT3_I(inode)->xattr_sem);
return i_error + b_error;
}
/*
* If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is
* not set, set it.
*/
static void ext3_xattr_update_super_block(handle_t *handle,
struct super_block *sb)
{
if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR))
return;
if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) {
EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR);
sb->s_dirt = 1;
ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh);
}
}
/*
* Release the xattr block BH: If the reference count is > 1, decrement
* it; otherwise free the block.
*/
static void
ext3_xattr_release_block(handle_t *handle, struct inode *inode,
struct buffer_head *bh)
{
struct mb_cache_entry *ce = NULL;
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
int error = 0;
ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr);
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
error = ext3_journal_get_write_access(handle, bh);
if (error)
goto out;
lock_buffer(bh);
if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
ea_bdebug(bh, "refcount now=0; freeing");
if (ce)
mb_cache_entry_free(ce);
ext3_free_blocks(handle, inode, bh->b_blocknr, 1);
get_bh(bh);
ext3_forget(handle, 1, inode, bh, bh->b_blocknr);
} else {
le32_add_cpu(&BHDR(bh)->h_refcount, -1);
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
error = ext3_journal_dirty_metadata(handle, bh);
if (IS_SYNC(inode))
handle->h_sync = 1;
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
DQUOT_FREE_BLOCK(inode, 1);
ea_bdebug(bh, "refcount now=%d; releasing",
le32_to_cpu(BHDR(bh)->h_refcount));
if (ce)
mb_cache_entry_release(ce);
}
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
unlock_buffer(bh);
out:
ext3_std_error(inode->i_sb, error);
return;
}
struct ext3_xattr_info {
int name_index;
const char *name;
const void *value;
size_t value_len;
};
struct ext3_xattr_search {
struct ext3_xattr_entry *first;
void *base;
void *end;
struct ext3_xattr_entry *here;
int not_found;
};
static int
ext3_xattr_set_entry(struct ext3_xattr_info *i, struct ext3_xattr_search *s)
{
struct ext3_xattr_entry *last;
size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
/* Compute min_offs and last. */
last = s->first;
for (; !IS_LAST_ENTRY(last); last = EXT3_XATTR_NEXT(last)) {
if (!last->e_value_block && last->e_value_size) {
size_t offs = le16_to_cpu(last->e_value_offs);
if (offs < min_offs)
min_offs = offs;
}
}
free = min_offs - ((void *)last - s->base) - sizeof(__u32);
if (!s->not_found) {
if (!s->here->e_value_block && s->here->e_value_size) {
size_t size = le32_to_cpu(s->here->e_value_size);
free += EXT3_XATTR_SIZE(size);
}
free += EXT3_XATTR_LEN(name_len);
}
if (i->value) {
if (free < EXT3_XATTR_SIZE(i->value_len) ||
free < EXT3_XATTR_LEN(name_len) +
EXT3_XATTR_SIZE(i->value_len))
return -ENOSPC;
}
if (i->value && s->not_found) {
/* Insert the new name. */
size_t size = EXT3_XATTR_LEN(name_len);
size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
memmove((void *)s->here + size, s->here, rest);
memset(s->here, 0, size);
s->here->e_name_index = i->name_index;
s->here->e_name_len = name_len;
memcpy(s->here->e_name, i->name, name_len);
} else {
if (!s->here->e_value_block && s->here->e_value_size) {
void *first_val = s->base + min_offs;
size_t offs = le16_to_cpu(s->here->e_value_offs);
void *val = s->base + offs;
size_t size = EXT3_XATTR_SIZE(
le32_to_cpu(s->here->e_value_size));
if (i->value && size == EXT3_XATTR_SIZE(i->value_len)) {
/* The old and the new value have the same
size. Just replace. */
s->here->e_value_size =
cpu_to_le32(i->value_len);
memset(val + size - EXT3_XATTR_PAD, 0,
EXT3_XATTR_PAD); /* Clear pad bytes. */
memcpy(val, i->value, i->value_len);
return 0;
}
/* Remove the old value. */
memmove(first_val + size, first_val, val - first_val);
memset(first_val, 0, size);
s->here->e_value_size = 0;
s->here->e_value_offs = 0;
min_offs += size;
/* Adjust all value offsets. */
last = s->first;
while (!IS_LAST_ENTRY(last)) {
size_t o = le16_to_cpu(last->e_value_offs);
if (!last->e_value_block &&
last->e_value_size && o < offs)
last->e_value_offs =
cpu_to_le16(o + size);
last = EXT3_XATTR_NEXT(last);
}
}
if (!i->value) {
/* Remove the old name. */
size_t size = EXT3_XATTR_LEN(name_len);
last = ENTRY((void *)last - size);
memmove(s->here, (void *)s->here + size,
(void *)last - (void *)s->here + sizeof(__u32));
memset(last, 0, size);
}
}
if (i->value) {
/* Insert the new value. */
s->here->e_value_size = cpu_to_le32(i->value_len);
if (i->value_len) {
size_t size = EXT3_XATTR_SIZE(i->value_len);
void *val = s->base + min_offs - size;
s->here->e_value_offs = cpu_to_le16(min_offs - size);
memset(val + size - EXT3_XATTR_PAD, 0,
EXT3_XATTR_PAD); /* Clear the pad bytes. */
memcpy(val, i->value, i->value_len);
}
}
return 0;
}
struct ext3_xattr_block_find {
struct ext3_xattr_search s;
struct buffer_head *bh;
};
static int
ext3_xattr_block_find(struct inode *inode, struct ext3_xattr_info *i,
struct ext3_xattr_block_find *bs)
{
struct super_block *sb = inode->i_sb;
int error;
ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
i->name_index, i->name, i->value, (long)i->value_len);
if (EXT3_I(inode)->i_file_acl) {
/* The inode already has an extended attribute block. */
bs->bh = sb_bread(sb, EXT3_I(inode)->i_file_acl);
error = -EIO;
if (!bs->bh)
goto cleanup;
ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
atomic_read(&(bs->bh->b_count)),
le32_to_cpu(BHDR(bs->bh)->h_refcount));
if (ext3_xattr_check_block(bs->bh)) {
ext3_error(sb, __FUNCTION__,
"inode %lu: bad block "E3FSBLK, inode->i_ino,
EXT3_I(inode)->i_file_acl);
error = -EIO;
goto cleanup;
}
/* Find the named attribute. */
bs->s.base = BHDR(bs->bh);
bs->s.first = BFIRST(bs->bh);
bs->s.end = bs->bh->b_data + bs->bh->b_size;
bs->s.here = bs->s.first;
error = ext3_xattr_find_entry(&bs->s.here, i->name_index,
i->name, bs->bh->b_size, 1);
if (error && error != -ENODATA)
goto cleanup;
bs->s.not_found = error;
}
error = 0;
cleanup:
return error;
}
static int
ext3_xattr_block_set(handle_t *handle, struct inode *inode,
struct ext3_xattr_info *i,
struct ext3_xattr_block_find *bs)
{
struct super_block *sb = inode->i_sb;
struct buffer_head *new_bh = NULL;
struct ext3_xattr_search *s = &bs->s;
struct mb_cache_entry *ce = NULL;
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
int error = 0;
#define header(x) ((struct ext3_xattr_header *)(x))
if (i->value && i->value_len > sb->s_blocksize)
return -ENOSPC;
if (s->base) {
ce = mb_cache_entry_get(ext3_xattr_cache, bs->bh->b_bdev,
bs->bh->b_blocknr);
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
error = ext3_journal_get_write_access(handle, bs->bh);
if (error)
goto cleanup;
lock_buffer(bs->bh);
if (header(s->base)->h_refcount == cpu_to_le32(1)) {
if (ce) {
mb_cache_entry_free(ce);
ce = NULL;
}
ea_bdebug(bs->bh, "modifying in-place");
error = ext3_xattr_set_entry(i, s);
if (!error) {
if (!IS_LAST_ENTRY(s->first))
ext3_xattr_rehash(header(s->base),
s->here);
ext3_xattr_cache_insert(bs->bh);
}
unlock_buffer(bs->bh);
if (error == -EIO)
goto bad_block;
if (!error)
error = ext3_journal_dirty_metadata(handle,
bs->bh);
if (error)
goto cleanup;
goto inserted;
} else {
int offset = (char *)s->here - bs->bh->b_data;
[PATCH] ext[34]: EA block reference count racing fix There are race issues around ext[34] xattr block release code. ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen. We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it). This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all. We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well). Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-03-01 12:13:35 +08:00
unlock_buffer(bs->bh);
journal_release_buffer(handle, bs->bh);
if (ce) {
mb_cache_entry_release(ce);
ce = NULL;
}
ea_bdebug(bs->bh, "cloning");
s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
error = -ENOMEM;
if (s->base == NULL)
goto cleanup;
memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
s->first = ENTRY(header(s->base)+1);
header(s->base)->h_refcount = cpu_to_le32(1);
s->here = ENTRY(s->base + offset);
s->end = s->base + bs->bh->b_size;
}
} else {
/* Allocate a buffer where we construct the new block. */
s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
/* assert(header == s->base) */
error = -ENOMEM;
if (s->base == NULL)
goto cleanup;
header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
header(s->base)->h_blocks = cpu_to_le32(1);
header(s->base)->h_refcount = cpu_to_le32(1);
s->first = ENTRY(header(s->base)+1);
s->here = ENTRY(header(s->base)+1);
s->end = s->base + sb->s_blocksize;
}
error = ext3_xattr_set_entry(i, s);
if (error == -EIO)
goto bad_block;
if (error)
goto cleanup;
if (!IS_LAST_ENTRY(s->first))
ext3_xattr_rehash(header(s->base), s->here);
inserted:
if (!IS_LAST_ENTRY(s->first)) {
new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce);
if (new_bh) {
/* We found an identical block in the cache. */
if (new_bh == bs->bh)
ea_bdebug(new_bh, "keeping");
else {
/* The old block is released after updating
the inode. */
error = -EDQUOT;
if (DQUOT_ALLOC_BLOCK(inode, 1))
goto cleanup;
error = ext3_journal_get_write_access(handle,
new_bh);
if (error)
goto cleanup_dquot;
lock_buffer(new_bh);
le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
ea_bdebug(new_bh, "reusing; refcount now=%d",
le32_to_cpu(BHDR(new_bh)->h_refcount));
unlock_buffer(new_bh);
error = ext3_journal_dirty_metadata(handle,
new_bh);
if (error)
goto cleanup_dquot;
}
mb_cache_entry_release(ce);
ce = NULL;
} else if (bs->bh && s->base == bs->bh->b_data) {
/* We were modifying this block in-place. */
ea_bdebug(bs->bh, "keeping this block");
new_bh = bs->bh;
get_bh(new_bh);
} else {
/* We need to allocate a new block */
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
ext3_fsblk_t goal = le32_to_cpu(
EXT3_SB(sb)->s_es->s_first_data_block) +
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
(ext3_fsblk_t)EXT3_I(inode)->i_block_group *
EXT3_BLOCKS_PER_GROUP(sb);
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
ext3_fsblk_t block = ext3_new_block(handle, inode,
goal, &error);
if (error)
goto cleanup;
ea_idebug(inode, "creating block %d", block);
new_bh = sb_getblk(sb, block);
if (!new_bh) {
getblk_failed:
ext3_free_blocks(handle, inode, block, 1);
error = -EIO;
goto cleanup;
}
lock_buffer(new_bh);
error = ext3_journal_get_create_access(handle, new_bh);
if (error) {
unlock_buffer(new_bh);
goto getblk_failed;
}
memcpy(new_bh->b_data, s->base, new_bh->b_size);
set_buffer_uptodate(new_bh);
unlock_buffer(new_bh);
ext3_xattr_cache_insert(new_bh);
error = ext3_journal_dirty_metadata(handle, new_bh);
if (error)
goto cleanup;
}
}
/* Update the inode. */
EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
/* Drop the previous xattr block. */
if (bs->bh && bs->bh != new_bh)
ext3_xattr_release_block(handle, inode, bs->bh);
error = 0;
cleanup:
if (ce)
mb_cache_entry_release(ce);
brelse(new_bh);
if (!(bs->bh && s->base == bs->bh->b_data))
kfree(s->base);
return error;
cleanup_dquot:
DQUOT_FREE_BLOCK(inode, 1);
goto cleanup;
bad_block:
ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: bad block "E3FSBLK, inode->i_ino,
EXT3_I(inode)->i_file_acl);
goto cleanup;
#undef header
}
struct ext3_xattr_ibody_find {
struct ext3_xattr_search s;
struct ext3_iloc iloc;
};
static int
ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i,
struct ext3_xattr_ibody_find *is)
{
struct ext3_xattr_ibody_header *header;
struct ext3_inode *raw_inode;
int error;
if (EXT3_I(inode)->i_extra_isize == 0)
return 0;
raw_inode = ext3_raw_inode(&is->iloc);
header = IHDR(inode, raw_inode);
is->s.base = is->s.first = IFIRST(header);
is->s.here = is->s.first;
is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
if (EXT3_I(inode)->i_state & EXT3_STATE_XATTR) {
error = ext3_xattr_check_names(IFIRST(header), is->s.end);
if (error)
return error;
/* Find the named attribute. */
error = ext3_xattr_find_entry(&is->s.here, i->name_index,
i->name, is->s.end -
(void *)is->s.base, 0);
if (error && error != -ENODATA)
return error;
is->s.not_found = error;
}
return 0;
}
static int
ext3_xattr_ibody_set(handle_t *handle, struct inode *inode,
struct ext3_xattr_info *i,
struct ext3_xattr_ibody_find *is)
{
struct ext3_xattr_ibody_header *header;
struct ext3_xattr_search *s = &is->s;
int error;
if (EXT3_I(inode)->i_extra_isize == 0)
return -ENOSPC;
error = ext3_xattr_set_entry(i, s);
if (error)
return error;
header = IHDR(inode, ext3_raw_inode(&is->iloc));
if (!IS_LAST_ENTRY(s->first)) {
header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
EXT3_I(inode)->i_state |= EXT3_STATE_XATTR;
} else {
header->h_magic = cpu_to_le32(0);
EXT3_I(inode)->i_state &= ~EXT3_STATE_XATTR;
}
return 0;
}
/*
* ext3_xattr_set_handle()
*
* Create, replace or remove an extended attribute for this inode. Buffer
* is NULL to remove an existing extended attribute, and non-NULL to
* either replace an existing extended attribute, or create a new extended
* attribute. The flags XATTR_REPLACE and XATTR_CREATE
* specify that an extended attribute must exist and must not exist
* previous to the call, respectively.
*
* Returns 0, or a negative error number on failure.
*/
int
ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
const char *name, const void *value, size_t value_len,
int flags)
{
struct ext3_xattr_info i = {
.name_index = name_index,
.name = name,
.value = value,
.value_len = value_len,
};
struct ext3_xattr_ibody_find is = {
.s = { .not_found = -ENODATA, },
};
struct ext3_xattr_block_find bs = {
.s = { .not_found = -ENODATA, },
};
int error;
if (!name)
return -EINVAL;
if (strlen(name) > 255)
return -ERANGE;
down_write(&EXT3_I(inode)->xattr_sem);
error = ext3_get_inode_loc(inode, &is.iloc);
if (error)
goto cleanup;
if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) {
struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc);
memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size);
EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW;
}
error = ext3_xattr_ibody_find(inode, &i, &is);
if (error)
goto cleanup;
if (is.s.not_found)
error = ext3_xattr_block_find(inode, &i, &bs);
if (error)
goto cleanup;
if (is.s.not_found && bs.s.not_found) {
error = -ENODATA;
if (flags & XATTR_REPLACE)
goto cleanup;
error = 0;
if (!value)
goto cleanup;
} else {
error = -EEXIST;
if (flags & XATTR_CREATE)
goto cleanup;
}
error = ext3_journal_get_write_access(handle, is.iloc.bh);
if (error)
goto cleanup;
if (!value) {
if (!is.s.not_found)
error = ext3_xattr_ibody_set(handle, inode, &i, &is);
else if (!bs.s.not_found)
error = ext3_xattr_block_set(handle, inode, &i, &bs);
} else {
error = ext3_xattr_ibody_set(handle, inode, &i, &is);
if (!error && !bs.s.not_found) {
i.value = NULL;
error = ext3_xattr_block_set(handle, inode, &i, &bs);
} else if (error == -ENOSPC) {
error = ext3_xattr_block_set(handle, inode, &i, &bs);
if (error)
goto cleanup;
if (!is.s.not_found) {
i.value = NULL;
error = ext3_xattr_ibody_set(handle, inode, &i,
&is);
}
}
}
if (!error) {
ext3_xattr_update_super_block(handle, inode->i_sb);
inode->i_ctime = CURRENT_TIME_SEC;
error = ext3_mark_iloc_dirty(handle, inode, &is.iloc);
/*
* The bh is consumed by ext3_mark_iloc_dirty, even with
* error != 0.
*/
is.iloc.bh = NULL;
if (IS_SYNC(inode))
handle->h_sync = 1;
}
cleanup:
brelse(is.iloc.bh);
brelse(bs.bh);
up_write(&EXT3_I(inode)->xattr_sem);
return error;
}
/*
* ext3_xattr_set()
*
* Like ext3_xattr_set_handle, but start from an inode. This extended
* attribute modification is a filesystem transaction by itself.
*
* Returns 0, or a negative error number on failure.
*/
int
ext3_xattr_set(struct inode *inode, int name_index, const char *name,
const void *value, size_t value_len, int flags)
{
handle_t *handle;
int error, retries = 0;
retry:
handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
if (IS_ERR(handle)) {
error = PTR_ERR(handle);
} else {
int error2;
error = ext3_xattr_set_handle(handle, inode, name_index, name,
value, value_len, flags);
error2 = ext3_journal_stop(handle);
if (error == -ENOSPC &&
ext3_should_retry_alloc(inode->i_sb, &retries))
goto retry;
if (error == 0)
error = error2;
}
return error;
}
/*
* ext3_xattr_delete_inode()
*
* Free extended attribute resources associated with this inode. This
* is called immediately before an inode is freed. We have exclusive
* access to the inode.
*/
void
ext3_xattr_delete_inode(handle_t *handle, struct inode *inode)
{
struct buffer_head *bh = NULL;
if (!EXT3_I(inode)->i_file_acl)
goto cleanup;
bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
if (!bh) {
ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: block "E3FSBLK" read error", inode->i_ino,
EXT3_I(inode)->i_file_acl);
goto cleanup;
}
if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
BHDR(bh)->h_blocks != cpu_to_le32(1)) {
ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: bad block "E3FSBLK, inode->i_ino,
EXT3_I(inode)->i_file_acl);
goto cleanup;
}
ext3_xattr_release_block(handle, inode, bh);
EXT3_I(inode)->i_file_acl = 0;
cleanup:
brelse(bh);
}
/*
* ext3_xattr_put_super()
*
* This is called when a file system is unmounted.
*/
void
ext3_xattr_put_super(struct super_block *sb)
{
mb_cache_shrink(sb->s_bdev);
}
/*
* ext3_xattr_cache_insert()
*
* Create a new entry in the extended attribute cache, and insert
* it unless such an entry is already in the cache.
*
* Returns 0, or a negative error number on failure.
*/
static void
ext3_xattr_cache_insert(struct buffer_head *bh)
{
__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
struct mb_cache_entry *ce;
int error;
ce = mb_cache_entry_alloc(ext3_xattr_cache);
if (!ce) {
ea_bdebug(bh, "out of memory");
return;
}
error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
if (error) {
mb_cache_entry_free(ce);
if (error == -EBUSY) {
ea_bdebug(bh, "already in cache");
error = 0;
}
} else {
ea_bdebug(bh, "inserting [%x]", (int)hash);
mb_cache_entry_release(ce);
}
}
/*
* ext3_xattr_cmp()
*
* Compare two extended attribute blocks for equality.
*
* Returns 0 if the blocks are equal, 1 if they differ, and
* a negative error number on errors.
*/
static int
ext3_xattr_cmp(struct ext3_xattr_header *header1,
struct ext3_xattr_header *header2)
{
struct ext3_xattr_entry *entry1, *entry2;
entry1 = ENTRY(header1+1);
entry2 = ENTRY(header2+1);
while (!IS_LAST_ENTRY(entry1)) {
if (IS_LAST_ENTRY(entry2))
return 1;
if (entry1->e_hash != entry2->e_hash ||
entry1->e_name_index != entry2->e_name_index ||
entry1->e_name_len != entry2->e_name_len ||
entry1->e_value_size != entry2->e_value_size ||
memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
return 1;
if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
return -EIO;
if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
(char *)header2 + le16_to_cpu(entry2->e_value_offs),
le32_to_cpu(entry1->e_value_size)))
return 1;
entry1 = EXT3_XATTR_NEXT(entry1);
entry2 = EXT3_XATTR_NEXT(entry2);
}
if (!IS_LAST_ENTRY(entry2))
return 1;
return 0;
}
/*
* ext3_xattr_cache_find()
*
* Find an identical extended attribute block.
*
* Returns a pointer to the block found, or NULL if such a block was
* not found or an error occurred.
*/
static struct buffer_head *
ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header,
struct mb_cache_entry **pce)
{
__u32 hash = le32_to_cpu(header->h_hash);
struct mb_cache_entry *ce;
if (!header->h_hash)
return NULL; /* never share */
ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
again:
ce = mb_cache_entry_find_first(ext3_xattr_cache, 0,
inode->i_sb->s_bdev, hash);
while (ce) {
struct buffer_head *bh;
if (IS_ERR(ce)) {
if (PTR_ERR(ce) == -EAGAIN)
goto again;
break;
}
bh = sb_bread(inode->i_sb, ce->e_block);
if (!bh) {
ext3_error(inode->i_sb, __FUNCTION__,
"inode %lu: block %lu read error",
inode->i_ino, (unsigned long) ce->e_block);
} else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
EXT3_XATTR_REFCOUNT_MAX) {
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes Some of the in-kernel ext3 block variable type are treated as signed 4 bytes int type, thus limited ext3 filesystem to 8TB (4kblock size based). While trying to fix them, it seems quite confusing in the ext3 code where some blocks are filesystem-wide blocks, some are group relative offsets that need to be signed value (as -1 has special meaning). So it seem saner to define two types of physical blocks: one is filesystem wide blocks, another is group-relative blocks. The following patches clarify these two types of blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3 filesystem limit to 8TB. With this series of patches and the percpu counter data type changes in the mm tree, we are able to extend exts filesystem limit to 16TB. This work is also a pre-request for the recent >32 bit ext3 work, and makes the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine ext3_fsblk_t from unsigned long to sector_t and redefine the format string for ext3 filesystem block corresponding. Two RFC with a series patches have been posted to ext2-devel list and have been reviewed and discussed: http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2 http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2 Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and >8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests includes overnight fsx, tiobench, dbench and fsstress. This patch: Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for filesystem wide blocks. This patch classifies all block group relative blocks, and ext3_fsblk_t blocks occurs in the same function where used to be confusing before. Also include kernel bug fixes for filesystem wide in-kernel block variables. There are some fileystem wide blocks are treated as int/unsigned int type in the kernel currently, especially in ext3 block allocation and reservation code. This patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned long) type. Signed-off-by: Mingming Cao <cmm@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-25 20:48:06 +08:00
ea_idebug(inode, "block %lu refcount %d>=%d",
(unsigned long) ce->e_block,
le32_to_cpu(BHDR(bh)->h_refcount),
EXT3_XATTR_REFCOUNT_MAX);
} else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) {
*pce = ce;
return bh;
}
brelse(bh);
ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
}
return NULL;
}
#define NAME_HASH_SHIFT 5
#define VALUE_HASH_SHIFT 16
/*
* ext3_xattr_hash_entry()
*
* Compute the hash of an extended attribute.
*/
static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header,
struct ext3_xattr_entry *entry)
{
__u32 hash = 0;
char *name = entry->e_name;
int n;
for (n=0; n < entry->e_name_len; n++) {
hash = (hash << NAME_HASH_SHIFT) ^
(hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
*name++;
}
if (entry->e_value_block == 0 && entry->e_value_size != 0) {
__le32 *value = (__le32 *)((char *)header +
le16_to_cpu(entry->e_value_offs));
for (n = (le32_to_cpu(entry->e_value_size) +
EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) {
hash = (hash << VALUE_HASH_SHIFT) ^
(hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
le32_to_cpu(*value++);
}
}
entry->e_hash = cpu_to_le32(hash);
}
#undef NAME_HASH_SHIFT
#undef VALUE_HASH_SHIFT
#define BLOCK_HASH_SHIFT 16
/*
* ext3_xattr_rehash()
*
* Re-compute the extended attribute hash value after an entry has changed.
*/
static void ext3_xattr_rehash(struct ext3_xattr_header *header,
struct ext3_xattr_entry *entry)
{
struct ext3_xattr_entry *here;
__u32 hash = 0;
ext3_xattr_hash_entry(header, entry);
here = ENTRY(header+1);
while (!IS_LAST_ENTRY(here)) {
if (!here->e_hash) {
/* Block is not shared if an entry's hash value == 0 */
hash = 0;
break;
}
hash = (hash << BLOCK_HASH_SHIFT) ^
(hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
le32_to_cpu(here->e_hash);
here = EXT3_XATTR_NEXT(here);
}
header->h_hash = cpu_to_le32(hash);
}
#undef BLOCK_HASH_SHIFT
int __init
init_ext3_xattr(void)
{
ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL,
sizeof(struct mb_cache_entry) +
sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
if (!ext3_xattr_cache)
return -ENOMEM;
return 0;
}
void
exit_ext3_xattr(void)
{
if (ext3_xattr_cache)
mb_cache_destroy(ext3_xattr_cache);
ext3_xattr_cache = NULL;
}