/*
* Write a directory entry after a call to namei, using the parameters
* that it left in nameidata. The argument dirp is the new directory
* entry contents. Dvp is a pointer to the directory to be written,
* which was left locked by namei. Remaining parameters (dp->i_offset,
* dp->i_count) indicate how the space for the new entry is to be obtained.
* Non-null bp indicates that a directory is being created (for the
* soft dependency code).
*/
int
ufs_direnter(struct vnode *dvp, struct vnode *tvp, struct direct *dirp,
struct componentname *cnp, struct buf *newdirbp)
{
struct ucred *cr;
struct proc *p;
int newentrysize;
struct inode *dp;
struct buf *bp;
u_int dsize;
struct direct *ep, *nep;
int error, ret, blkoff, loc, spacefree, flags;
char *dirbuf;
UFS_WAPBL_JLOCK_ASSERT(dvp->v_mount);
error = 0;
cr = cnp->cn_cred;
p = cnp->cn_proc;
dp = VTOI(dvp);
newentrysize = DIRSIZ(FSFMT(dvp), dirp);
if (dp->i_count == 0) {
/*
* If dp->i_count is 0, then namei could find no
* space in the directory. Here, dp->i_offset will
* be on a directory block boundary and we will write the
* new entry into a fresh block.
*/
if (dp->i_offset & (DIRBLKSIZ - 1))
panic("ufs_direnter: newblk");
flags = B_CLRBUF;
if (!DOINGSOFTDEP(dvp))
flags |= B_SYNC;
if ((error = UFS_BUF_ALLOC(dp, (off_t)dp->i_offset, DIRBLKSIZ,
cr, flags, &bp)) != 0) {
if (DOINGSOFTDEP(dvp) && newdirbp != NULL)
bdwrite(newdirbp);
return (error);
}
DIP_ASSIGN(dp, size, dp->i_offset + DIRBLKSIZ);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(dvp, DIP(dp, size));
dirp->d_reclen = DIRBLKSIZ;
blkoff = dp->i_offset &
(VFSTOUFS(dvp->v_mount)->um_mountp->mnt_stat.f_iosize - 1);
memcpy(bp->b_data + blkoff, dirp, newentrysize);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL) {
ufsdirhash_newblk(dp, dp->i_offset);
ufsdirhash_add(dp, dirp, dp->i_offset);
ufsdirhash_checkblock(dp, (char *)bp->b_data + blkoff,
dp->i_offset);
}
#endif
if (DOINGSOFTDEP(dvp)) {
/*
* Ensure that the entire newly allocated block is a
* valid directory so that future growth within the
* block does not have to ensure that the block is
* written before the inode.
*/
blkoff += DIRBLKSIZ;
while (blkoff < bp->b_bcount) {
((struct direct *)
(bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
blkoff += DIRBLKSIZ;
}
if (softdep_setup_directory_add(bp, dp, dp->i_offset,
dirp->d_ino, newdirbp, 1) == 0) {
bdwrite(bp);
return (UFS_UPDATE(dp, 0));
}
/* We have just allocated a directory block in an
* indirect block. Rather than tracking when it gets
* claimed by the inode, we simply do a VOP_FSYNC
* now to ensure that it is there (in case the user
* does a future fsync). Note that we have to unlock
* the inode for the entry that we just entered, as
* the VOP_FSYNC may need to lock other inodes which
* can lead to deadlock if we also hold a lock on
* the newly entered node.
*/
if ((error = VOP_BWRITE(bp)))
return (error);
if (tvp != NULL)
VOP_UNLOCK(tvp, 0);
error = VOP_FSYNC(dvp, p->p_ucred, MNT_WAIT);
if (tvp != NULL)
vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p);
return (error);
}
error = VOP_BWRITE(bp);
ret = UFS_UPDATE(dp, !DOINGSOFTDEP(dvp));
if (error == 0)
return (ret);
return (error);
}
/*
* If dp->i_count is non-zero, then namei found space for the new
* entry in the range dp->i_offset to dp->i_offset + dp->i_count
* in the directory. To use this space, we may have to compact
* the entries located there, by copying them together towards the
* beginning of the block, leaving the free space in one usable
* chunk at the end.
*/
/*
* Increase size of directory if entry eats into new space.
* This should never push the size past a new multiple of
* DIRBLKSIZE.
*
* N.B. - THIS IS AN ARTIFACT OF 4.2 AND SHOULD NEVER HAPPEN.
*/
if (dp->i_offset + dp->i_count > DIP(dp, size)) {
DIP_ASSIGN(dp, size, dp->i_offset + dp->i_count);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
UFS_WAPBL_UPDATE(dp, MNT_WAIT);
}
/*
* Get the block containing the space for the new directory entry.
*/
if ((error = UFS_BUFATOFF(dp, (off_t)dp->i_offset, &dirbuf, &bp))
!= 0) {
if (DOINGSOFTDEP(dvp) && newdirbp != NULL)
bdwrite(newdirbp);
return (error);
}
/*
* Find space for the new entry. In the simple case, the entry at
* offset base will have the space. If it does not, then namei
* arranged that compacting the region dp->i_offset to
* dp->i_offset + dp->i_count would yield the space.
*/
ep = (struct direct *)dirbuf;
dsize = ep->d_ino ? DIRSIZ(FSFMT(dvp), ep) : 0;
spacefree = ep->d_reclen - dsize;
for (loc = ep->d_reclen; loc < dp->i_count; ) {
nep = (struct direct *)(dirbuf + loc);
/* Trim the existing slot (NB: dsize may be zero). */
ep->d_reclen = dsize;
ep = (struct direct *)((char *)ep + dsize);
/* Read nep->d_reclen now as the memmove() may clobber it. */
loc += nep->d_reclen;
if (nep->d_ino == 0) {
/*
* A mid-block unused entry. Such entries are
* never created by the kernel, but fsck_ffs
* can create them (and it doesn't fix them).
*
* Add up the free space, and initialise the
* relocated entry since we don't memmove it.
*/
spacefree += nep->d_reclen;
ep->d_ino = 0;
dsize = 0;
continue;
}
dsize = DIRSIZ(FSFMT(dvp), nep);
spacefree += nep->d_reclen - dsize;
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_move(dp, nep,
dp->i_offset + ((char *)nep - dirbuf),
dp->i_offset + ((char *)ep - dirbuf));
#endif
if (DOINGSOFTDEP(dvp))
softdep_change_directoryentry_offset(dp, dirbuf,
(caddr_t)nep, (caddr_t)ep, dsize);
else
memmove(ep, nep, dsize);
}
/*
* Here, `ep' points to a directory entry containing `dsize' in-use
* bytes followed by `spacefree' unused bytes. If ep->d_ino == 0,
* then the entry is completely unused (dsize == 0). The value
* of ep->d_reclen is always indeterminate.
*
* Update the pointer fields in the previous entry (if any),
* copy in the new entry, and write out the block.
*/
if (ep->d_ino == 0) {
if (spacefree + dsize < newentrysize)
panic("ufs_direnter: compact1");
dirp->d_reclen = spacefree + dsize;
} else {
if (spacefree < newentrysize)
panic("ufs_direnter: compact2");
dirp->d_reclen = spacefree;
ep->d_reclen = dsize;
ep = (struct direct *)((char *)ep + dsize);
}
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL && (ep->d_ino == 0 ||
dirp->d_reclen == spacefree))
ufsdirhash_add(dp, dirp, dp->i_offset + ((char *)ep - dirbuf));
#endif
memcpy(ep, dirp, newentrysize);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_checkblock(dp, dirbuf -
(dp->i_offset & (DIRBLKSIZ - 1)),
dp->i_offset & ~(DIRBLKSIZ - 1));
#endif
if (DOINGSOFTDEP(dvp)) {
(void)softdep_setup_directory_add(bp, dp,
dp->i_offset + (caddr_t)ep - dirbuf,
dirp->d_ino, newdirbp, 0);
bdwrite(bp);
} else {
error = VOP_BWRITE(bp);
}
dp->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* If all went well, and the directory can be shortened, proceed
* with the truncation. Note that we have to unlock the inode for
* the entry that we just entered, as the truncation may need to
* lock other inodes which can lead to deadlock if we also hold a
* lock on the newly entered node.
*/
if (error == 0 && dp->i_endoff && dp->i_endoff < DIP(dp, size)) {
if (tvp != NULL)
VOP_UNLOCK(tvp, 0);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_dirtrunc(dp, dp->i_endoff);
#endif
error = UFS_TRUNCATE(dp, (off_t)dp->i_endoff, IO_SYNC, cr);
if (tvp != NULL)
vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p);
}
UFS_WAPBL_UPDATE(dp, MNT_WAIT);
return (error);
}
// ffs文件系统的写入操作
int
ffs_write(void *v)
{
struct vop_write_args *ap = v;
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct fs *fs;
struct buf *bp;
daddr_t lbn;
off_t osize;
int blkoffset, error, extended, flags, ioflag, size, xfersize;
ssize_t resid, overrun;
extended = 0;
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
ip = VTOI(vp);
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_WRITE)
panic("ffs_write: mode");
#endif
/*
* If writing 0 bytes, succeed and do not change
* update time or file offset (standards compliance)
*/
if (uio->uio_resid == 0)
return (0);
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = DIP(ip, size);
if ((DIP(ip, flags) & APPEND) && uio->uio_offset != DIP(ip, size))
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
if ((ioflag & IO_SYNC) == 0)
panic("ffs_write: nonsync dir write");
break;
default:
panic("ffs_write: type");
}
fs = ip->i_fs;
if (uio->uio_offset < 0 ||
(u_int64_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
return (EFBIG);
/* do the filesize rlimit check */
if ((error = vn_fsizechk(vp, uio, ioflag, &overrun)))
return (error);
resid = uio->uio_resid;
osize = DIP(ip, size);
flags = ioflag & IO_SYNC ? B_SYNC : 0;
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (fs->fs_bsize > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
if ((error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize,
ap->a_cred, flags, &bp)) != 0)
break;
if (uio->uio_offset + xfersize > DIP(ip, size)) {
DIP_ASSIGN(ip, size, uio->uio_offset + xfersize);
uvm_vnp_setsize(vp, DIP(ip, size));
extended = 1;
}
(void)uvm_vnp_uncache(vp);
size = blksize(fs, ip, lbn) - bp->b_resid;
if (size < xfersize)
xfersize = size;
error =
uiomovei(bp->b_data + blkoffset, xfersize, uio);
if (error != 0)
memset(bp->b_data + blkoffset, 0, xfersize);
#if 0
if (ioflag & IO_NOCACHE)
bp->b_flags |= B_NOCACHE;
#endif
if (ioflag & IO_SYNC)
(void)bwrite(bp);
else if (xfersize + blkoffset == fs->fs_bsize) {
if (doclusterwrite)
cluster_write(bp, &ip->i_ci, DIP(ip, size));
else
bawrite(bp);
} else
bdwrite(bp);
if (error || xfersize == 0)
break;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
DIP_ASSIGN(ip, mode, DIP(ip, mode) & ~(ISUID | ISGID));
if (resid > uio->uio_resid)
VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
if (error) {
if (ioflag & IO_UNIT) {
(void)UFS_TRUNCATE(ip, osize,
ioflag & IO_SYNC, ap->a_cred);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
error = UFS_UPDATE(ip, 1);
}
/* correct the result for writes clamped by vn_fsizechk() */
uio->uio_resid += overrun;
return (error);
}
/*
* Vnode op for writing.
*/
int
ffs_write(void *v)
{
struct vop_write_args *ap = v;
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct fs *fs;
struct buf *bp;
struct proc *p;
daddr64_t lbn;
off_t osize;
int blkoffset, error, extended, flags, ioflag, resid, size, xfersize;
extended = 0;
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
ip = VTOI(vp);
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_WRITE)
panic("ffs_write: mode");
#endif
/*
* If writing 0 bytes, succeed and do not change
* update time or file offset (standards compliance)
*/
if (uio->uio_resid == 0)
return (0);
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = DIP(ip, size);
if ((DIP(ip, flags) & APPEND) && uio->uio_offset != DIP(ip, size))
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
if ((ioflag & IO_SYNC) == 0)
panic("ffs_write: nonsync dir write");
break;
default:
panic("ffs_write: type");
}
fs = ip->i_fs;
if (uio->uio_offset < 0 ||
(u_int64_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
return (EFBIG);
/*
* Maybe this should be above the vnode op call, but so long as
* file servers have no limits, I don't think it matters.
*/
p = uio->uio_procp;
if (vp->v_type == VREG && p && !(ioflag & IO_NOLIMIT) &&
uio->uio_offset + uio->uio_resid >
p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
psignal(p, SIGXFSZ);
return (EFBIG);
}
resid = uio->uio_resid;
osize = DIP(ip, size);
flags = ioflag & IO_SYNC ? B_SYNC : 0;
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (fs->fs_bsize > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
if ((error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize,
ap->a_cred, flags, &bp)) != 0)
break;
if (uio->uio_offset + xfersize > DIP(ip, size)) {
DIP_ASSIGN(ip, size, uio->uio_offset + xfersize);
uvm_vnp_setsize(vp, DIP(ip, size));
extended = 1;
}
(void)uvm_vnp_uncache(vp);
size = blksize(fs, ip, lbn) - bp->b_resid;
if (size < xfersize)
xfersize = size;
error =
uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
if (error != 0)
bzero((char *)bp->b_data + blkoffset, xfersize);
if (ioflag & IO_SYNC)
(void)bwrite(bp);
else if (xfersize + blkoffset == fs->fs_bsize) {
if (doclusterwrite)
cluster_write(bp, &ip->i_ci, DIP(ip, size));
else
bawrite(bp);
} else
bdwrite(bp);
if (error || xfersize == 0)
break;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
DIP_ASSIGN(ip, mode, DIP(ip, mode) & ~(ISUID | ISGID));
if (resid > uio->uio_resid)
VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
if (error) {
if (ioflag & IO_UNIT) {
(void)UFS_TRUNCATE(ip, osize,
ioflag & IO_SYNC, ap->a_cred);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
error = UFS_UPDATE(ip, MNT_WAIT);
}
return (error);
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
{
struct vnode *ovp;
daddr64_t lastblock;
daddr64_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr64_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct fs *fs;
struct buf *bp;
int offset, size, level;
long count, nblocks, vflags, blocksreleased = 0;
int i, aflags, error, allerror, indirect = 0;
off_t osize;
extern int num_indirdep;
extern int max_indirdep;
if (length < 0)
return (EINVAL);
ovp = ITOV(oip);
if (ovp->v_type != VREG &&
ovp->v_type != VDIR &&
ovp->v_type != VLNK)
return (0);
if (DIP(oip, size) == length)
return (0);
if (ovp->v_type == VLNK &&
(DIP(oip, size) < ovp->v_mount->mnt_maxsymlinklen ||
(ovp->v_mount->mnt_maxsymlinklen == 0 &&
oip->i_din1->di_blocks == 0))) {
#ifdef DIAGNOSTIC
if (length != 0)
panic("ffs_truncate: partial truncate of symlink");
#endif
memset(SHORTLINK(oip), 0, (size_t) DIP(oip, size));
DIP_ASSIGN(oip, size, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, MNT_WAIT));
}
if ((error = getinoquota(oip)) != 0)
return (error);
uvm_vnp_setsize(ovp, length);
oip->i_ci.ci_lasta = oip->i_ci.ci_clen
= oip->i_ci.ci_cstart = oip->i_ci.ci_lastw = 0;
if (DOINGSOFTDEP(ovp)) {
if (length > 0 || softdep_slowdown(ovp)) {
/*
* If a file is only partially truncated, then
* we have to clean up the data structures
* describing the allocation past the truncation
* point. Finding and deallocating those structures
* is a lot of work. Since partial truncation occurs
* rarely, we solve the problem by syncing the file
* so that it will have no data structures left.
*/
if ((error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
} else {
(void)ufs_quota_free_blocks(oip, DIP(oip, blocks),
NOCRED);
softdep_setup_freeblocks(oip, length);
(void) vinvalbuf(ovp, 0, cred, curproc, 0, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, 0));
}
}
fs = oip->i_fs;
osize = DIP(oip, size);
/*
* 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) {
if (length > fs->fs_maxfilesize)
return (EFBIG);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
if (bp->b_lblkno >= NDADDR)
indirect = 1;
DIP_ASSIGN(oip, size, length);
uvm_vnp_setsize(ovp, length);
(void) uvm_vnp_uncache(ovp);
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
error = UFS_UPDATE(oip, MNT_WAIT);
if (DOINGSOFTDEP(ovp) && num_indirdep > max_indirdep)
if (indirect) {
/*
* If the number of pending indirect block
* dependencies is sufficiently close to the
* maximum number of simultaneously mappable
* buffers force a sync on the vnode to prevent
* buffer cache exhaustion.
*/
VOP_FSYNC(ovp, curproc->p_ucred, MNT_WAIT);
}
return (error);
}
uvm_vnp_setsize(ovp, length);
/*
* Shorten the size of the file. If the file is not being
* truncated to a block boundary, the contents of the
* partial block following the end of the file must be
* zero'ed in case it ever becomes accessible again because
* of subsequent file growth. Directories however are not
* zero'ed as they should grow back initialized to empty.
*/
offset = blkoff(fs, length);
if (offset == 0) {
DIP_ASSIGN(oip, size, length);
} else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
/*
* When we are doing soft updates and the UFS_BALLOC
* above fills in a direct block hole with a full sized
* block that will be truncated down to a fragment below,
* we must flush out the block dependency with an FSYNC
* so that we do not get a soft updates inconsistency
* when we create the fragment below.
*/
if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
(error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
DIP_ASSIGN(oip, size, length);
size = blksize(fs, oip, lbn);
(void) uvm_vnp_uncache(ovp);
if (ovp->v_type != VDIR)
bzero((char *)bp->b_data + offset,
(u_int)(size - offset));
bp->b_bcount = size;
if (aflags & B_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->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* 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 ffs_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
oldblks[i] = DIP(oip, db[i]);
if (i > lastblock)
DIP_ASSIGN(oip, db[i], 0);
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
if ((error = UFS_UPDATE(oip, MNT_WAIT)) != 0)
allerror = error;
/*
* 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.
*/
for (i = 0; i < NDADDR; i++) {
newblks[i] = DIP(oip, db[i]);
DIP_ASSIGN(oip, db[i], oldblks[i]);
}
for (i = 0; i < NIADDR; i++) {
newblks[NDADDR + i] = DIP(oip, ib[i]);
DIP_ASSIGN(oip, ib[i], oldblks[NDADDR + i]);
}
DIP_ASSIGN(oip, size, osize);
vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
/*
* 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 = DIP(oip, ib[level]);
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
ffs_blkfree(oip, bn, fs->fs_bsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = DIP(oip, db[i]);
if (bn == 0)
continue;
DIP_ASSIGN(oip, db[i], 0);
bsize = blksize(fs, oip, i);
ffs_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 = DIP(oip, 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);
DIP_ASSIGN(oip, size, length);
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("ffs_truncate: 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);
ffs_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != DIP(oip, ib[level]))
panic("ffs_truncate1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != DIP(oip, db[i]))
panic("ffs_truncate2");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
DIP_ASSIGN(oip, size, length);
DIP_ADD(oip, blocks, -blocksreleased);
oip->i_flag |= IN_CHANGE;
(void)ufs_quota_free_blocks(oip, blocksreleased, NOCRED);
return (allerror);
}
/*
* Mkdir system call
*/
int
ufs_mkdir(void *v)
{
struct vop_mkdir_args *ap = v;
struct vnode *dvp = ap->a_dvp;
struct vattr *vap = ap->a_vap;
struct componentname *cnp = ap->a_cnp;
struct inode *ip, *dp;
struct vnode *tvp;
struct buf *bp;
struct direct newdir;
struct dirtemplate dirtemplate, *dtp;
int error, dmode, blkoff;
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ufs_mkdir: no name");
#endif
dp = VTOI(dvp);
if ((nlink_t) DIP(dp, nlink) >= LINK_MAX) {
error = EMLINK;
goto out;
}
dmode = vap->va_mode & 0777;
dmode |= IFDIR;
/*
* Must simulate part of ufs_makeinode here to acquire the inode,
* but not have it entered in the parent directory. The entry is
* made later after writing "." and ".." entries.
*/
if ((error = UFS_INODE_ALLOC(dp, dmode, cnp->cn_cred, &tvp)) != 0)
goto out;
ip = VTOI(tvp);
DIP_ASSIGN(ip, uid, cnp->cn_cred->cr_uid);
DIP_ASSIGN(ip, gid, DIP(dp, gid));
if ((error = getinoquota(ip)) ||
(error = ufs_quota_alloc_inode(ip, cnp->cn_cred))) {
pool_put(&namei_pool, cnp->cn_pnbuf);
UFS_INODE_FREE(ip, ip->i_number, dmode);
vput(tvp);
vput(dvp);
return (error);
}
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
DIP_ASSIGN(ip, mode, dmode);
tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
ip->i_effnlink = 2;
DIP_ASSIGN(ip, nlink, 2);
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
/*
* Bump link count in parent directory to reflect work done below.
* Should be done before reference is create so cleanup is
* possible if we crash.
*/
dp->i_effnlink++;
DIP_ADD(dp, nlink, 1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(dvp))
softdep_change_linkcnt(dp, 0);
if ((error = UFS_UPDATE(dp, !DOINGSOFTDEP(dvp))) != 0)
goto bad;
/*
* Initialize directory with "." and ".." from static template.
*/
if (dvp->v_mount->mnt_maxsymlinklen > 0)
dtp = &mastertemplate;
else
dtp = (struct dirtemplate *)&omastertemplate;
dirtemplate = *dtp;
dirtemplate.dot_ino = ip->i_number;
dirtemplate.dotdot_ino = dp->i_number;
if ((error = UFS_BUF_ALLOC(ip, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
B_CLRBUF, &bp)) != 0)
goto bad;
DIP_ASSIGN(ip, size, DIRBLKSIZ);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(tvp, DIP(ip, size));
bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
if (DOINGSOFTDEP(tvp)) {
/*
* Ensure that the entire newly allocated block is a
* valid directory so that future growth within the
* block does not have to ensure that the block is
* written before the inode
*/
blkoff = DIRBLKSIZ;
while (blkoff < bp->b_bcount) {
((struct direct *)
(bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
blkoff += DIRBLKSIZ;
}
}
if ((error = UFS_UPDATE(ip, !DOINGSOFTDEP(tvp))) != 0) {
(void)VOP_BWRITE(bp);
goto bad;
}
/*
* Directory set up, now install its entry in the parent directory.
*
* If we are not doing soft dependencies, then we must write out the
* buffer containing the new directory body before entering the new
* name in the parent. If we are doing soft dependencies, then the
* buffer containing the new directory body will be passed to and
* released in the soft dependency code after the code has attached
* an appropriate ordering dependency to the buffer which ensures that
* the buffer is written before the new name is written in the parent.
*/
if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp)) != 0))
goto bad;
ufs_makedirentry(ip, cnp, &newdir);
error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
bad:
if (error == 0) {
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
*ap->a_vpp = tvp;
} else {
dp->i_effnlink--;
DIP_ADD(dp, nlink, -1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(dvp))
softdep_change_linkcnt(dp, 0);
/*
* No need to do an explicit VOP_TRUNCATE here, vrele will
* do this for us because we set the link count to 0.
*/
ip->i_effnlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
vput(tvp);
}
out:
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
return (error);
}
/*
* Extended attribute area writing.
*/
int
ffs_ea_write(struct vnode *vp, struct uio *uio, int ioflag, struct ucred *ucred)
{
struct inode *ip;
struct ufs2_dinode *dp;
struct fs *fs;
struct buf *bp;
off_t osize;
size_t resid;
int error, flags, xfersize;
ip = VTOI(vp);
fs = ip->i_fs;
dp = ip->i_din2;
error = 0;
/* Don't cross the limit for the extended attribute area length. */
if ((unsigned)uio->uio_offset + uio->uio_resid > NXADDR * fs->fs_bsize)
return (EFBIG);
/*
* Save the original amount of data to be written ('uio->uio_resid') as
* well as the extended attribute area size ('dp->di_extsize') in case
* we need to truncate in the end due to an unpredicted error.
*/
resid = uio->uio_resid;
osize = dp->di_extsize;
/* Prepare flags to be passed to UFS_BUF_ALLOC(). */
flags = IO_EXT | B_CLRBUF;
if (!DOINGASYNC(vp))
flags |= B_SYNC;
/*
* Loop over the amount of data requested by the caller, stopping only
* if an error occurs. By default, we always try to write a file system
* block worth of bytes per iteration ('xfersize'). Check this value
* against what is left to be copied ('uio->uio_resid'). If we are
* writing less than a full block, ask the buffer to be cleaned first.
*/
while (uio->uio_resid > 0) {
xfersize = fs->fs_bsize;
if (uio->uio_resid < xfersize) {
xfersize = uio->uio_resid;
flags |= B_CLRBUF;
}
/* Ask the ffs_balloc.c code for the block. */
error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize, ucred,
flags, &bp);
if (error)
break;
/* Check if we are growing the extended attributes area. */
if (uio->uio_offset + xfersize > dp->di_extsize)
dp->di_extsize = uio->uio_offset + xfersize;
error = uiomove(bp->b_data, xfersize, uio);
if (error) {
clrbuf(bp); /* Get rid of stray contents. */
error = EIO;
break;
}
/*
* We use the same criteria for calling bwrite(), bawrite(), or
* bdwrite() as the rest of the FFS code does.
*/
if (ioflag & IO_SYNC)
bwrite(bp);
else if (xfersize == fs->fs_bsize)
bawrite(bp);
else
bdwrite(bp);
if (error || xfersize == 0)
break;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* If we successfully wrote any data, and we are not the superuser we
* clear the setuid and setgid bits as a precaution against tampering.
*/
if ((DIP(ip, mode) & (ISUID | ISGID)) && resid > uio->uio_resid)
if (ucred && suser_ucred(ucred))
DIP(ip, mode) &= ~(ISUID | ISGID);
if (error) {
/* Release blocks if we failed. */
if (ioflag & IO_UNIT) {
ffs_truncate(ip, osize, flags, ucred);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
return (error);
}
/* If needed, sync the inode now. */
if (resid > uio->uio_resid && (ioflag & IO_SYNC))
return (ffs_update(ip, NULL, NULL, MNT_WAIT));
return (0);
}
