1
linux/fs/logfs/dir.c
Joern Engel e326068806 Prevent schedule while atomic in __logfs_readdir
Apparently filldir can sleep, which forbids kmap_atomic.

Signed-off-by: Joern Engel <joern@logfs.org>
2010-03-27 11:19:15 +01:00

828 lines
21 KiB
C

/*
* fs/logfs/dir.c - directory-related code
*
* As should be obvious for Linux kernel code, license is GPLv2
*
* Copyright (c) 2005-2008 Joern Engel <joern@logfs.org>
*/
#include "logfs.h"
/*
* Atomic dir operations
*
* Directory operations are by default not atomic. Dentries and Inodes are
* created/removed/altered in seperate operations. Therefore we need to do
* a small amount of journaling.
*
* Create, link, mkdir, mknod and symlink all share the same function to do
* the work: __logfs_create. This function works in two atomic steps:
* 1. allocate inode (remember in journal)
* 2. allocate dentry (clear journal)
*
* As we can only get interrupted between the two, when the inode we just
* created is simply stored in the anchor. On next mount, if we were
* interrupted, we delete the inode. From a users point of view the
* operation never happened.
*
* Unlink and rmdir also share the same function: unlink. Again, this
* function works in two atomic steps
* 1. remove dentry (remember inode in journal)
* 2. unlink inode (clear journal)
*
* And again, on the next mount, if we were interrupted, we delete the inode.
* From a users point of view the operation succeeded.
*
* Rename is the real pain to deal with, harder than all the other methods
* combined. Depending on the circumstances we can run into three cases.
* A "target rename" where the target dentry already existed, a "local
* rename" where both parent directories are identical or a "cross-directory
* rename" in the remaining case.
*
* Local rename is atomic, as the old dentry is simply rewritten with a new
* name.
*
* Cross-directory rename works in two steps, similar to __logfs_create and
* logfs_unlink:
* 1. Write new dentry (remember old dentry in journal)
* 2. Remove old dentry (clear journal)
*
* Here we remember a dentry instead of an inode. On next mount, if we were
* interrupted, we delete the dentry. From a users point of view, the
* operation succeeded.
*
* Target rename works in three atomic steps:
* 1. Attach old inode to new dentry (remember old dentry and new inode)
* 2. Remove old dentry (still remember the new inode)
* 3. Remove victim inode
*
* Here we remember both an inode an a dentry. If we get interrupted
* between steps 1 and 2, we delete both the dentry and the inode. If
* we get interrupted between steps 2 and 3, we delete just the inode.
* In either case, the remaining objects are deleted on next mount. From
* a users point of view, the operation succeeded.
*/
static int write_dir(struct inode *dir, struct logfs_disk_dentry *dd,
loff_t pos)
{
return logfs_inode_write(dir, dd, sizeof(*dd), pos, WF_LOCK, NULL);
}
static int write_inode(struct inode *inode)
{
return __logfs_write_inode(inode, WF_LOCK);
}
static s64 dir_seek_data(struct inode *inode, s64 pos)
{
s64 new_pos = logfs_seek_data(inode, pos);
return max(pos, new_pos - 1);
}
static int beyond_eof(struct inode *inode, loff_t bix)
{
loff_t pos = bix << inode->i_sb->s_blocksize_bits;
return pos >= i_size_read(inode);
}
/*
* Prime value was chosen to be roughly 256 + 26. r5 hash uses 11,
* so short names (len <= 9) don't even occupy the complete 32bit name
* space. A prime >256 ensures short names quickly spread the 32bit
* name space. Add about 26 for the estimated amount of information
* of each character and pick a prime nearby, preferrably a bit-sparse
* one.
*/
static u32 hash_32(const char *s, int len, u32 seed)
{
u32 hash = seed;
int i;
for (i = 0; i < len; i++)
hash = hash * 293 + s[i];
return hash;
}
/*
* We have to satisfy several conflicting requirements here. Small
* directories should stay fairly compact and not require too many
* indirect blocks. The number of possible locations for a given hash
* should be small to make lookup() fast. And we should try hard not
* to overflow the 32bit name space or nfs and 32bit host systems will
* be unhappy.
*
* So we use the following scheme. First we reduce the hash to 0..15
* and try a direct block. If that is occupied we reduce the hash to
* 16..255 and try an indirect block. Same for 2x and 3x indirect
* blocks. Lastly we reduce the hash to 0x800_0000 .. 0xffff_ffff,
* but use buckets containing eight entries instead of a single one.
*
* Using 16 entries should allow for a reasonable amount of hash
* collisions, so the 32bit name space can be packed fairly tight
* before overflowing. Oh and currently we don't overflow but return
* and error.
*
* How likely are collisions? Doing the appropriate math is beyond me
* and the Bronstein textbook. But running a test program to brute
* force collisions for a couple of days showed that on average the
* first collision occurs after 598M entries, with 290M being the
* smallest result. Obviously 21 entries could already cause a
* collision if all entries are carefully chosen.
*/
static pgoff_t hash_index(u32 hash, int round)
{
u32 i0_blocks = I0_BLOCKS;
u32 i1_blocks = I1_BLOCKS;
u32 i2_blocks = I2_BLOCKS;
u32 i3_blocks = I3_BLOCKS;
switch (round) {
case 0:
return hash % i0_blocks;
case 1:
return i0_blocks + hash % (i1_blocks - i0_blocks);
case 2:
return i1_blocks + hash % (i2_blocks - i1_blocks);
case 3:
return i2_blocks + hash % (i3_blocks - i2_blocks);
case 4 ... 19:
return i3_blocks + 16 * (hash % (((1<<31) - i3_blocks) / 16))
+ round - 4;
}
BUG();
}
static struct page *logfs_get_dd_page(struct inode *dir, struct dentry *dentry)
{
struct qstr *name = &dentry->d_name;
struct page *page;
struct logfs_disk_dentry *dd;
u32 hash = hash_32(name->name, name->len, 0);
pgoff_t index;
int round;
if (name->len > LOGFS_MAX_NAMELEN)
return ERR_PTR(-ENAMETOOLONG);
for (round = 0; round < 20; round++) {
index = hash_index(hash, round);
if (beyond_eof(dir, index))
return NULL;
if (!logfs_exist_block(dir, index))
continue;
page = read_cache_page(dir->i_mapping, index,
(filler_t *)logfs_readpage, NULL);
if (IS_ERR(page))
return page;
dd = kmap_atomic(page, KM_USER0);
BUG_ON(dd->namelen == 0);
if (name->len != be16_to_cpu(dd->namelen) ||
memcmp(name->name, dd->name, name->len)) {
kunmap_atomic(dd, KM_USER0);
page_cache_release(page);
continue;
}
kunmap_atomic(dd, KM_USER0);
return page;
}
return NULL;
}
static int logfs_remove_inode(struct inode *inode)
{
int ret;
inode->i_nlink--;
ret = write_inode(inode);
LOGFS_BUG_ON(ret, inode->i_sb);
return ret;
}
static void abort_transaction(struct inode *inode, struct logfs_transaction *ta)
{
if (logfs_inode(inode)->li_block)
logfs_inode(inode)->li_block->ta = NULL;
kfree(ta);
}
static int logfs_unlink(struct inode *dir, struct dentry *dentry)
{
struct logfs_super *super = logfs_super(dir->i_sb);
struct inode *inode = dentry->d_inode;
struct logfs_transaction *ta;
struct page *page;
pgoff_t index;
int ret;
ta = kzalloc(sizeof(*ta), GFP_KERNEL);
if (!ta)
return -ENOMEM;
ta->state = UNLINK_1;
ta->ino = inode->i_ino;
inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
page = logfs_get_dd_page(dir, dentry);
if (!page) {
kfree(ta);
return -ENOENT;
}
if (IS_ERR(page)) {
kfree(ta);
return PTR_ERR(page);
}
index = page->index;
page_cache_release(page);
mutex_lock(&super->s_dirop_mutex);
logfs_add_transaction(dir, ta);
ret = logfs_delete(dir, index, NULL);
if (!ret)
ret = write_inode(dir);
if (ret) {
abort_transaction(dir, ta);
printk(KERN_ERR"LOGFS: unable to delete inode\n");
goto out;
}
ta->state = UNLINK_2;
logfs_add_transaction(inode, ta);
ret = logfs_remove_inode(inode);
out:
mutex_unlock(&super->s_dirop_mutex);
return ret;
}
static inline int logfs_empty_dir(struct inode *dir)
{
u64 data;
data = logfs_seek_data(dir, 0) << dir->i_sb->s_blocksize_bits;
return data >= i_size_read(dir);
}
static int logfs_rmdir(struct inode *dir, struct dentry *dentry)
{
struct inode *inode = dentry->d_inode;
if (!logfs_empty_dir(inode))
return -ENOTEMPTY;
return logfs_unlink(dir, dentry);
}
/* FIXME: readdir currently has it's own dir_walk code. I don't see a good
* way to combine the two copies */
#define IMPLICIT_NODES 2
static int __logfs_readdir(struct file *file, void *buf, filldir_t filldir)
{
struct inode *dir = file->f_dentry->d_inode;
loff_t pos = file->f_pos - IMPLICIT_NODES;
struct page *page;
struct logfs_disk_dentry *dd;
int full;
BUG_ON(pos < 0);
for (;; pos++) {
if (beyond_eof(dir, pos))
break;
if (!logfs_exist_block(dir, pos)) {
/* deleted dentry */
pos = dir_seek_data(dir, pos);
continue;
}
page = read_cache_page(dir->i_mapping, pos,
(filler_t *)logfs_readpage, NULL);
if (IS_ERR(page))
return PTR_ERR(page);
dd = kmap(page);
BUG_ON(dd->namelen == 0);
full = filldir(buf, (char *)dd->name, be16_to_cpu(dd->namelen),
pos, be64_to_cpu(dd->ino), dd->type);
kunmap(page);
page_cache_release(page);
if (full)
break;
}
file->f_pos = pos + IMPLICIT_NODES;
return 0;
}
static int logfs_readdir(struct file *file, void *buf, filldir_t filldir)
{
struct inode *inode = file->f_dentry->d_inode;
ino_t pino = parent_ino(file->f_dentry);
int err;
if (file->f_pos < 0)
return -EINVAL;
if (file->f_pos == 0) {
if (filldir(buf, ".", 1, 1, inode->i_ino, DT_DIR) < 0)
return 0;
file->f_pos++;
}
if (file->f_pos == 1) {
if (filldir(buf, "..", 2, 2, pino, DT_DIR) < 0)
return 0;
file->f_pos++;
}
err = __logfs_readdir(file, buf, filldir);
return err;
}
static void logfs_set_name(struct logfs_disk_dentry *dd, struct qstr *name)
{
dd->namelen = cpu_to_be16(name->len);
memcpy(dd->name, name->name, name->len);
}
static struct dentry *logfs_lookup(struct inode *dir, struct dentry *dentry,
struct nameidata *nd)
{
struct page *page;
struct logfs_disk_dentry *dd;
pgoff_t index;
u64 ino = 0;
struct inode *inode;
page = logfs_get_dd_page(dir, dentry);
if (IS_ERR(page))
return ERR_CAST(page);
if (!page) {
d_add(dentry, NULL);
return NULL;
}
index = page->index;
dd = kmap_atomic(page, KM_USER0);
ino = be64_to_cpu(dd->ino);
kunmap_atomic(dd, KM_USER0);
page_cache_release(page);
inode = logfs_iget(dir->i_sb, ino);
if (IS_ERR(inode)) {
printk(KERN_ERR"LogFS: Cannot read inode #%llx for dentry (%lx, %lx)n",
ino, dir->i_ino, index);
return ERR_CAST(inode);
}
return d_splice_alias(inode, dentry);
}
static void grow_dir(struct inode *dir, loff_t index)
{
index = (index + 1) << dir->i_sb->s_blocksize_bits;
if (i_size_read(dir) < index)
i_size_write(dir, index);
}
static int logfs_write_dir(struct inode *dir, struct dentry *dentry,
struct inode *inode)
{
struct page *page;
struct logfs_disk_dentry *dd;
u32 hash = hash_32(dentry->d_name.name, dentry->d_name.len, 0);
pgoff_t index;
int round, err;
for (round = 0; round < 20; round++) {
index = hash_index(hash, round);
if (logfs_exist_block(dir, index))
continue;
page = find_or_create_page(dir->i_mapping, index, GFP_KERNEL);
if (!page)
return -ENOMEM;
dd = kmap_atomic(page, KM_USER0);
memset(dd, 0, sizeof(*dd));
dd->ino = cpu_to_be64(inode->i_ino);
dd->type = logfs_type(inode);
logfs_set_name(dd, &dentry->d_name);
kunmap_atomic(dd, KM_USER0);
err = logfs_write_buf(dir, page, WF_LOCK);
unlock_page(page);
page_cache_release(page);
if (!err)
grow_dir(dir, index);
return err;
}
/* FIXME: Is there a better return value? In most cases neither
* the filesystem nor the directory are full. But we have had
* too many collisions for this particular hash and no fallback.
*/
return -ENOSPC;
}
static int __logfs_create(struct inode *dir, struct dentry *dentry,
struct inode *inode, const char *dest, long destlen)
{
struct logfs_super *super = logfs_super(dir->i_sb);
struct logfs_inode *li = logfs_inode(inode);
struct logfs_transaction *ta;
int ret;
ta = kzalloc(sizeof(*ta), GFP_KERNEL);
if (!ta)
return -ENOMEM;
ta->state = CREATE_1;
ta->ino = inode->i_ino;
mutex_lock(&super->s_dirop_mutex);
logfs_add_transaction(inode, ta);
if (dest) {
/* symlink */
ret = logfs_inode_write(inode, dest, destlen, 0, WF_LOCK, NULL);
if (!ret)
ret = write_inode(inode);
} else {
/* creat/mkdir/mknod */
ret = write_inode(inode);
}
if (ret) {
abort_transaction(inode, ta);
li->li_flags |= LOGFS_IF_STILLBORN;
/* FIXME: truncate symlink */
inode->i_nlink--;
iput(inode);
goto out;
}
ta->state = CREATE_2;
logfs_add_transaction(dir, ta);
ret = logfs_write_dir(dir, dentry, inode);
/* sync directory */
if (!ret)
ret = write_inode(dir);
if (ret) {
logfs_del_transaction(dir, ta);
ta->state = CREATE_2;
logfs_add_transaction(inode, ta);
logfs_remove_inode(inode);
iput(inode);
goto out;
}
d_instantiate(dentry, inode);
out:
mutex_unlock(&super->s_dirop_mutex);
return ret;
}
static int logfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
{
struct inode *inode;
/*
* FIXME: why do we have to fill in S_IFDIR, while the mode is
* correct for mknod, creat, etc.? Smells like the vfs *should*
* do it for us but for some reason fails to do so.
*/
inode = logfs_new_inode(dir, S_IFDIR | mode);
if (IS_ERR(inode))
return PTR_ERR(inode);
inode->i_op = &logfs_dir_iops;
inode->i_fop = &logfs_dir_fops;
return __logfs_create(dir, dentry, inode, NULL, 0);
}
static int logfs_create(struct inode *dir, struct dentry *dentry, int mode,
struct nameidata *nd)
{
struct inode *inode;
inode = logfs_new_inode(dir, mode);
if (IS_ERR(inode))
return PTR_ERR(inode);
inode->i_op = &logfs_reg_iops;
inode->i_fop = &logfs_reg_fops;
inode->i_mapping->a_ops = &logfs_reg_aops;
return __logfs_create(dir, dentry, inode, NULL, 0);
}
static int logfs_mknod(struct inode *dir, struct dentry *dentry, int mode,
dev_t rdev)
{
struct inode *inode;
if (dentry->d_name.len > LOGFS_MAX_NAMELEN)
return -ENAMETOOLONG;
inode = logfs_new_inode(dir, mode);
if (IS_ERR(inode))
return PTR_ERR(inode);
init_special_inode(inode, mode, rdev);
return __logfs_create(dir, dentry, inode, NULL, 0);
}
static int logfs_symlink(struct inode *dir, struct dentry *dentry,
const char *target)
{
struct inode *inode;
size_t destlen = strlen(target) + 1;
if (destlen > dir->i_sb->s_blocksize)
return -ENAMETOOLONG;
inode = logfs_new_inode(dir, S_IFLNK | 0777);
if (IS_ERR(inode))
return PTR_ERR(inode);
inode->i_op = &logfs_symlink_iops;
inode->i_mapping->a_ops = &logfs_reg_aops;
return __logfs_create(dir, dentry, inode, target, destlen);
}
static int logfs_permission(struct inode *inode, int mask)
{
return generic_permission(inode, mask, NULL);
}
static int logfs_link(struct dentry *old_dentry, struct inode *dir,
struct dentry *dentry)
{
struct inode *inode = old_dentry->d_inode;
if (inode->i_nlink >= LOGFS_LINK_MAX)
return -EMLINK;
inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
atomic_inc(&inode->i_count);
inode->i_nlink++;
mark_inode_dirty_sync(inode);
return __logfs_create(dir, dentry, inode, NULL, 0);
}
static int logfs_get_dd(struct inode *dir, struct dentry *dentry,
struct logfs_disk_dentry *dd, loff_t *pos)
{
struct page *page;
void *map;
page = logfs_get_dd_page(dir, dentry);
if (IS_ERR(page))
return PTR_ERR(page);
*pos = page->index;
map = kmap_atomic(page, KM_USER0);
memcpy(dd, map, sizeof(*dd));
kunmap_atomic(map, KM_USER0);
page_cache_release(page);
return 0;
}
static int logfs_delete_dd(struct inode *dir, loff_t pos)
{
/*
* Getting called with pos somewhere beyond eof is either a goofup
* within this file or means someone maliciously edited the
* (crc-protected) journal.
*/
BUG_ON(beyond_eof(dir, pos));
dir->i_ctime = dir->i_mtime = CURRENT_TIME;
log_dir(" Delete dentry (%lx, %llx)\n", dir->i_ino, pos);
return logfs_delete(dir, pos, NULL);
}
/*
* Cross-directory rename, target does not exist. Just a little nasty.
* Create a new dentry in the target dir, then remove the old dentry,
* all the while taking care to remember our operation in the journal.
*/
static int logfs_rename_cross(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
struct logfs_super *super = logfs_super(old_dir->i_sb);
struct logfs_disk_dentry dd;
struct logfs_transaction *ta;
loff_t pos;
int err;
/* 1. locate source dd */
err = logfs_get_dd(old_dir, old_dentry, &dd, &pos);
if (err)
return err;
ta = kzalloc(sizeof(*ta), GFP_KERNEL);
if (!ta)
return -ENOMEM;
ta->state = CROSS_RENAME_1;
ta->dir = old_dir->i_ino;
ta->pos = pos;
/* 2. write target dd */
mutex_lock(&super->s_dirop_mutex);
logfs_add_transaction(new_dir, ta);
err = logfs_write_dir(new_dir, new_dentry, old_dentry->d_inode);
if (!err)
err = write_inode(new_dir);
if (err) {
super->s_rename_dir = 0;
super->s_rename_pos = 0;
abort_transaction(new_dir, ta);
goto out;
}
/* 3. remove source dd */
ta->state = CROSS_RENAME_2;
logfs_add_transaction(old_dir, ta);
err = logfs_delete_dd(old_dir, pos);
if (!err)
err = write_inode(old_dir);
LOGFS_BUG_ON(err, old_dir->i_sb);
out:
mutex_unlock(&super->s_dirop_mutex);
return err;
}
static int logfs_replace_inode(struct inode *dir, struct dentry *dentry,
struct logfs_disk_dentry *dd, struct inode *inode)
{
loff_t pos;
int err;
err = logfs_get_dd(dir, dentry, dd, &pos);
if (err)
return err;
dd->ino = cpu_to_be64(inode->i_ino);
dd->type = logfs_type(inode);
err = write_dir(dir, dd, pos);
if (err)
return err;
log_dir("Replace dentry (%lx, %llx) %s -> %llx\n", dir->i_ino, pos,
dd->name, be64_to_cpu(dd->ino));
return write_inode(dir);
}
/* Target dentry exists - the worst case. We need to attach the source
* inode to the target dentry, then remove the orphaned target inode and
* source dentry.
*/
static int logfs_rename_target(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
struct logfs_super *super = logfs_super(old_dir->i_sb);
struct inode *old_inode = old_dentry->d_inode;
struct inode *new_inode = new_dentry->d_inode;
int isdir = S_ISDIR(old_inode->i_mode);
struct logfs_disk_dentry dd;
struct logfs_transaction *ta;
loff_t pos;
int err;
BUG_ON(isdir != S_ISDIR(new_inode->i_mode));
if (isdir) {
if (!logfs_empty_dir(new_inode))
return -ENOTEMPTY;
}
/* 1. locate source dd */
err = logfs_get_dd(old_dir, old_dentry, &dd, &pos);
if (err)
return err;
ta = kzalloc(sizeof(*ta), GFP_KERNEL);
if (!ta)
return -ENOMEM;
ta->state = TARGET_RENAME_1;
ta->dir = old_dir->i_ino;
ta->pos = pos;
ta->ino = new_inode->i_ino;
/* 2. attach source inode to target dd */
mutex_lock(&super->s_dirop_mutex);
logfs_add_transaction(new_dir, ta);
err = logfs_replace_inode(new_dir, new_dentry, &dd, old_inode);
if (err) {
super->s_rename_dir = 0;
super->s_rename_pos = 0;
super->s_victim_ino = 0;
abort_transaction(new_dir, ta);
goto out;
}
/* 3. remove source dd */
ta->state = TARGET_RENAME_2;
logfs_add_transaction(old_dir, ta);
err = logfs_delete_dd(old_dir, pos);
if (!err)
err = write_inode(old_dir);
LOGFS_BUG_ON(err, old_dir->i_sb);
/* 4. remove target inode */
ta->state = TARGET_RENAME_3;
logfs_add_transaction(new_inode, ta);
err = logfs_remove_inode(new_inode);
out:
mutex_unlock(&super->s_dirop_mutex);
return err;
}
static int logfs_rename(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
if (new_dentry->d_inode)
return logfs_rename_target(old_dir, old_dentry,
new_dir, new_dentry);
return logfs_rename_cross(old_dir, old_dentry, new_dir, new_dentry);
}
/* No locking done here, as this is called before .get_sb() returns. */
int logfs_replay_journal(struct super_block *sb)
{
struct logfs_super *super = logfs_super(sb);
struct inode *inode;
u64 ino, pos;
int err;
if (super->s_victim_ino) {
/* delete victim inode */
ino = super->s_victim_ino;
printk(KERN_INFO"LogFS: delete unmapped inode #%llx\n", ino);
inode = logfs_iget(sb, ino);
if (IS_ERR(inode))
goto fail;
LOGFS_BUG_ON(i_size_read(inode) > 0, sb);
super->s_victim_ino = 0;
err = logfs_remove_inode(inode);
iput(inode);
if (err) {
super->s_victim_ino = ino;
goto fail;
}
}
if (super->s_rename_dir) {
/* delete old dd from rename */
ino = super->s_rename_dir;
pos = super->s_rename_pos;
printk(KERN_INFO"LogFS: delete unbacked dentry (%llx, %llx)\n",
ino, pos);
inode = logfs_iget(sb, ino);
if (IS_ERR(inode))
goto fail;
super->s_rename_dir = 0;
super->s_rename_pos = 0;
err = logfs_delete_dd(inode, pos);
iput(inode);
if (err) {
super->s_rename_dir = ino;
super->s_rename_pos = pos;
goto fail;
}
}
return 0;
fail:
LOGFS_BUG(sb);
return -EIO;
}
const struct inode_operations logfs_symlink_iops = {
.readlink = generic_readlink,
.follow_link = page_follow_link_light,
};
const struct inode_operations logfs_dir_iops = {
.create = logfs_create,
.link = logfs_link,
.lookup = logfs_lookup,
.mkdir = logfs_mkdir,
.mknod = logfs_mknod,
.rename = logfs_rename,
.rmdir = logfs_rmdir,
.permission = logfs_permission,
.symlink = logfs_symlink,
.unlink = logfs_unlink,
};
const struct file_operations logfs_dir_fops = {
.fsync = logfs_fsync,
.ioctl = logfs_ioctl,
.readdir = logfs_readdir,
.read = generic_read_dir,
};