static int ext2_alloc_block (struct inode * inode, unsigned long goal, int *err)
{
#ifdef EXT2FS_DEBUG
static unsigned long alloc_hits = 0, alloc_attempts = 0;
#endif
unsigned long result;
#ifdef EXT2_PREALLOCATE
/* Writer: ->i_prealloc* */
if (inode->u.ext2_i.i_prealloc_count &&
(goal == inode->u.ext2_i.i_prealloc_block ||
goal + 1 == inode->u.ext2_i.i_prealloc_block))
{
result = inode->u.ext2_i.i_prealloc_block++;
inode->u.ext2_i.i_prealloc_count--;
/* Writer: end */
ext2_debug ("preallocation hit (%lu/%lu).\n",
++alloc_hits, ++alloc_attempts);
} else {
ext2_discard_prealloc (inode);
ext2_debug ("preallocation miss (%lu/%lu).\n",
alloc_hits, ++alloc_attempts);
if (S_ISREG(inode->i_mode))
result = ext2_new_block (inode, goal,
&inode->u.ext2_i.i_prealloc_count,
&inode->u.ext2_i.i_prealloc_block, err);
else
result = ext2_new_block (inode, goal, 0, 0, err);
}
#else
result = ext2_new_block (inode, goal, 0, 0, err);
#endif
return result;
}
/*
* Free an inode.
*
* the maintenance of the actual bitmaps is again up to the linux code
*/
int
ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
{
struct ext2_sb_info *fs;
struct inode *pip;
mode_t save_i_mode;
pip = VTOI(pvp);
fs = pip->i_e2fs;
if ((u_int)ino > fs->s_inodes_per_group * fs->s_groups_count)
panic("ext2_vfree: range: dev = (%d, %d), ino = %"PRId64", fs = %s",
major(pip->i_dev), minor(pip->i_dev), ino, fs->fs_fsmnt);
/* ext2_debug("ext2_vfree (%d, %d) called\n", pip->i_number, mode);
*/
ext2_discard_prealloc(pip);
/* we need to make sure that ext2_free_inode can adjust the
used_dir_counts in the group summary information - I'd
really like to know what the rationale behind this
'set i_mode to zero to denote an unused inode' is
*/
save_i_mode = pip->i_mode;
pip->i_mode = mode;
ext2_free_inode(pip);
pip->i_mode = save_i_mode;
return (0);
}
/* Writes everything from NP's inode to the disk image, and returns a pointer
to it, or NULL if nothing need be done. */
static struct ext2_inode *
write_node (struct node *np)
{
error_t err;
struct stat *st = &np->dn_stat;
struct ext2_inode *di;
ext2_debug ("(%llu)", np->cache_id);
if (diskfs_node_disknode (np)->info.i_prealloc_count)
ext2_discard_prealloc (np);
if (np->dn_stat_dirty)
{
struct ext2_inode_info *info = &diskfs_node_disknode (np)->info;
assert (!diskfs_readonly);
ext2_debug ("writing inode %d to disk", np->cache_id);
err = diskfs_catch_exception ();
if (err)
return NULL;
di = dino_ref (np->cache_id);
di->i_generation = st->st_gen;
/* We happen to know that the stat mode bits are the same
as the ext2fs mode bits. */
/* XXX? */
/* Only the low 16 bits of these fields are standard across all ext2
implementations. */
di->i_mode = st->st_mode & 0xFFFF & ~S_ITRANS;
di->i_uid = st->st_uid & 0xFFFF;
di->i_gid = st->st_gid & 0xFFFF;
if (sblock->s_creator_os == EXT2_OS_HURD)
/* If this is a hurd-compatible filesystem, write the high bits too. */
{
di->i_mode_high = (st->st_mode >> 16) & 0xffff & ~S_ITRANS;
di->i_uid_high = st->st_uid >> 16;
di->i_gid_high = st->st_gid >> 16;
di->i_author = st->st_author;
}
else
/* No hurd extensions should be turned on. */
{
static void ext2_clear_inode(struct inode *inode)
{
#ifdef CONFIG_EXT2_FS_POSIX_ACL
struct ext2_inode_info *ei = EXT2_I(inode);
if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
posix_acl_release(ei->i_acl);
ei->i_acl = EXT2_ACL_NOT_CACHED;
}
if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
posix_acl_release(ei->i_default_acl);
ei->i_default_acl = EXT2_ACL_NOT_CACHED;
}
#endif
if (!is_bad_inode(inode))
ext2_discard_prealloc(inode);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*
* ext2_inactive(struct vnode *a_vp)
*/
int
ext2_inactive(struct vop_inactive_args *ap)
{
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
int mode, error = 0;
ext2_discard_prealloc(ip);
if (prtactive && vp->v_sysref.refcnt > 1)
vprint("ext2_inactive: pushing active", vp);
/*
* Ignore inodes related to stale file handles.
*/
if (ip == NULL || ip->i_mode == 0)
goto out;
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
#ifdef QUOTA
if (!ext2_getinoquota(ip))
(void)ext2_chkiq(ip, -1, NOCRED, FORCE);
#endif
error = EXT2_TRUNCATE(vp, (off_t)0, 0, NOCRED);
ip->i_rdev = 0;
mode = ip->i_mode;
ip->i_mode = 0;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
EXT2_VFREE(vp, ip->i_number, mode);
}
if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
EXT2_UPDATE(vp, 0);
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
if (ip == NULL || ip->i_mode == 0)
vrecycle(vp);
return (error);
}
static int ext2_alloc_block (struct inode * inode, unsigned long goal, int *err)
{
#ifdef EXT2FS_DEBUG
static unsigned long alloc_hits, alloc_attempts;
#endif
unsigned long result;
#ifdef EXT2_PREALLOCATE
struct ext2_inode_info *ei = EXT2_I(inode);
write_lock(&ei->i_meta_lock);
if (ei->i_prealloc_count &&
(goal == ei->i_prealloc_block || goal + 1 == ei->i_prealloc_block))
{
result = ei->i_prealloc_block++;
ei->i_prealloc_count--;
write_unlock(&ei->i_meta_lock);
ext2_debug ("preallocation hit (%lu/%lu).\n",
++alloc_hits, ++alloc_attempts);
} else {
write_unlock(&ei->i_meta_lock);
ext2_discard_prealloc (inode);
ext2_debug ("preallocation miss (%lu/%lu).\n",
alloc_hits, ++alloc_attempts);
if (S_ISREG(inode->i_mode))
result = ext2_new_block (inode, goal,
&ei->i_prealloc_count,
&ei->i_prealloc_block, err);
else
result = ext2_new_block(inode, goal, NULL, NULL, err);
}
#else
result = ext2_new_block (inode, goal, 0, 0, err);
#endif
return result;
}
/*
* Called at each iput().
*
* The inode may be "bad" if ext2_read_inode() saw an error from
* ext2_get_inode(), so we need to check that to avoid freeing random disk
* blocks.
*/
void ext2_put_inode(struct inode *inode)
{
if (!is_bad_inode(inode))
ext2_discard_prealloc(inode);
}
/*
* Called at each iput()
*/
void ext2_put_inode (struct inode * inode)
{
ext2_discard_prealloc (inode);
}
/*
* Allocate a block in the file system.
*
* this takes the framework from ffs_alloc. To implement the
* actual allocation, it calls ext2_new_block, the ported version
* of the same Linux routine.
*
* we note that this is always called in connection with ext2_blkpref
*
* preallocation is done as Linux does it
*/
int
ext2_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size,
struct ucred *cred, daddr_t *bnp)
{
struct ext2_sb_info *fs;
daddr_t bno;
#if QUOTA
int error;
#endif
*bnp = 0;
fs = ip->i_e2fs;
#if DIAGNOSTIC
if ((u_int)size > fs->s_blocksize || blkoff(fs, size) != 0) {
kprintf("dev = %s, bsize = %lu, size = %d, fs = %s\n",
devtoname(ip->i_dev), fs->s_blocksize, size, fs->fs_fsmnt);
panic("ext2_alloc: bad size");
}
if (cred == NOCRED)
panic("ext2_alloc: missing credential");
#endif /* DIAGNOSTIC */
if (size == fs->s_blocksize && fs->s_es->s_free_blocks_count == 0)
goto nospace;
if (cred->cr_uid != 0 &&
fs->s_es->s_free_blocks_count < fs->s_es->s_r_blocks_count)
goto nospace;
#if QUOTA
if ((error = ext2_chkdq(ip, (long)btodb(size), cred, 0)) != 0)
return (error);
#endif
if (bpref >= fs->s_es->s_blocks_count)
bpref = 0;
/* call the Linux code */
#ifdef EXT2_PREALLOCATE
/* To have a preallocation hit, we must
* - have at least one block preallocated
* - and our preferred block must have that block number or one below
*/
if (ip->i_prealloc_count &&
(bpref == ip->i_prealloc_block ||
bpref + 1 == ip->i_prealloc_block))
{
bno = ip->i_prealloc_block++;
ip->i_prealloc_count--;
/* ext2_debug ("preallocation hit (%lu/%lu).\n",
++alloc_hits, ++alloc_attempts); */
/* Linux gets, clears, and releases the buffer at this
point - we don't have to that; we leave it to the caller
*/
} else {
ext2_discard_prealloc (ip);
/* ext2_debug ("preallocation miss (%lu/%lu).\n",
alloc_hits, ++alloc_attempts); */
if (S_ISREG(ip->i_mode))
bno = ext2_new_block
(ITOV(ip)->v_mount, bpref,
&ip->i_prealloc_count,
&ip->i_prealloc_block);
else
bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount,
bpref, 0, 0);
}
#else
bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount, bpref, 0, 0);
#endif
if (bno > 0) {
/* set next_alloc fields as done in block_getblk */
ip->i_next_alloc_block = lbn;
ip->i_next_alloc_goal = bno;
ip->i_blocks += btodb(size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
#if QUOTA
/*
* Restore user's disk quota because allocation failed.
*/
ext2_chkdq(ip, (long)-btodb(size), cred, FORCE);
#endif
nospace:
ext2_fserr(fs, cred->cr_uid, "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
return (ENOSPC);
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred)
{
struct vnode *ovp = vp;
daddr_t lastblock;
struct inode *oip;
daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct ext2_sb_info *fs;
struct buf *bp;
int offset, size, level;
long count, nblocks, blocksreleased = 0;
int i;
int aflags, error, allerror;
off_t osize;
/*
kprintf("ext2_truncate called %d to %d\n", VTOI(ovp)->i_number, length);
*/ /*
* negative file sizes will totally break the code below and
* are not meaningful anyways.
*/
if (length < 0)
return EFBIG;
oip = VTOI(ovp);
if (ovp->v_type == VLNK &&
oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
#if DIAGNOSTIC
if (length != 0)
panic("ext2_truncate: partial truncate of symlink");
#endif
bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
oip->i_size = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (EXT2_UPDATE(ovp, 1));
}
if (oip->i_size == length) {
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (EXT2_UPDATE(ovp, 0));
}
#if QUOTA
if ((error = ext2_getinoquota(oip)) != 0)
return (error);
#endif
fs = oip->i_e2fs;
osize = oip->i_size;
ext2_discard_prealloc(oip);
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
offset = blkoff(fs, length - 1);
lbn = lblkno(fs, length - 1);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
vnode_pager_setsize(ovp, length);
error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, aflags);
if (error) {
vnode_pager_setsize(ovp, osize);
return (error);
}
oip->i_size = length;
if (aflags & IO_SYNC)
bwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (EXT2_UPDATE(ovp, 1));
}
/*
* Shorten the size of the file. If the file is not being
* truncated to a block boundry, the contents of the
* partial block following the end of the file must be
* zero'ed in case it ever become accessable again because
* of subsequent file growth.
*/
/* I don't understand the comment above */
offset = blkoff(fs, length);
if (offset == 0) {
oip->i_size = length;
} else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = ext2_balloc(oip, lbn, offset, cred, &bp, aflags);
if (error)
return (error);
oip->i_size = length;
size = blksize(fs, oip, lbn);
bzero((char *)bp->b_data + offset, (u_int)(size - offset));
allocbuf(bp, size);
if (aflags & IO_SYNC)
bwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->s_blocksize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->s_blocksize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ext2_indirtrunc below.
*/
bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
for (level = TRIPLE; level >= SINGLE; level--)
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
lastiblock[level] = -1;
}
for (i = NDADDR - 1; i > lastblock; i--)
oip->i_db[i] = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
allerror = EXT2_UPDATE(ovp, 1);
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
oip->i_size = osize;
error = vtruncbuf(ovp, length, (int)fs->s_blocksize);
if (error && (allerror == 0))
allerror = error;
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = oip->i_ib[level];
if (bn != 0) {
error = ext2_indirtrunc(oip, indir_lbn[level],
fsbtodoff(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
ext2_blkfree(oip, bn, fs->s_frag_size);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = oip->i_db[i];
if (bn == 0)
continue;
oip->i_db[i] = 0;
bsize = blksize(fs, oip, i);
ext2_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = oip->i_db[lastblock];
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
oip->i_size = length;
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("itrunc: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ext2_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#if DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != oip->i_ib[level])
panic("itrunc1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != oip->i_db[i])
panic("itrunc2");
if (length == 0 && (!RB_EMPTY(&ovp->v_rbdirty_tree) ||
!RB_EMPTY(&ovp->v_rbclean_tree)))
panic("itrunc3");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
oip->i_size = length;
oip->i_blocks -= blocksreleased;
if (oip->i_blocks < 0) /* sanity */
oip->i_blocks = 0;
oip->i_flag |= IN_CHANGE;
vnode_pager_setsize(ovp, length);
#if QUOTA
ext2_chkdq(oip, -blocksreleased, NOCRED, 0);
#endif
return (allerror);
}
/*
* Called when an inode is released. Note that this is different
* from ext2_open_file: open gets called at every open, but release
* gets called only when /all/ the files are closed.
*/
static int ext2_release_file (struct inode * inode, struct file * filp)
{
if (filp->f_mode & FMODE_WRITE)
ext2_discard_prealloc (inode);
return 0;
}
/*
* Called when an inode is released. Note that this is different
* from ext2_open_file: open gets called at every open, but release
* gets called only when /all/ the files are closed.
*/
static int ext2_release_file (struct _inode * inode, struct file * filp)
{
if (tx_cache_get_file_ro(filp)->f_mode & FMODE_WRITE)
ext2_discard_prealloc (parent(inode));
return 0;
}