STATIC int
xfs_qm_scall_trunc_qfiles(
xfs_mount_t *mp,
uint flags)
{
int error;
xfs_inode_t *qip;
if (!capable(CAP_SYS_ADMIN))
return XFS_ERROR(EPERM);
error = 0;
if (!XFS_SB_VERSION_HASQUOTA(&mp->m_sb) || flags == 0) {
qdprintk("qtrunc flags=%x m_qflags=%x\n", flags, mp->m_qflags);
return XFS_ERROR(EINVAL);
}
if ((flags & XFS_DQ_USER) && mp->m_sb.sb_uquotino != NULLFSINO) {
error = xfs_iget(mp, NULL, mp->m_sb.sb_uquotino, 0, 0, &qip, 0);
if (! error) {
(void) xfs_truncate_file(mp, qip);
VN_RELE(XFS_ITOV(qip));
}
}
if ((flags & (XFS_DQ_GROUP|XFS_DQ_PROJ)) &&
mp->m_sb.sb_gquotino != NULLFSINO) {
error = xfs_iget(mp, NULL, mp->m_sb.sb_gquotino, 0, 0, &qip, 0);
if (! error) {
(void) xfs_truncate_file(mp, qip);
VN_RELE(XFS_ITOV(qip));
}
}
return (error);
}
void
xfs_refcache_resize(int xfs_refcache_new_size)
{
int i;
xfs_inode_t *ip;
int iplist_index = 0;
xfs_inode_t **iplist;
int error;
/*
* If the new size is smaller than the current size,
* purge entries to create smaller cache, and
* reposition index if necessary.
* Don't bother if no refcache yet.
*/
if (xfs_refcache && (xfs_refcache_new_size < xfs_refcache_size)) {
iplist = (xfs_inode_t **)kmem_zalloc(XFS_REFCACHE_SIZE_MAX *
sizeof(xfs_inode_t *), KM_SLEEP);
spin_lock(&xfs_refcache_lock);
for (i = xfs_refcache_new_size; i < xfs_refcache_size; i++) {
ip = xfs_refcache[i];
if (ip != NULL) {
xfs_refcache[i] = NULL;
ip->i_refcache = NULL;
xfs_refcache_count--;
ASSERT(xfs_refcache_count >= 0);
iplist[iplist_index] = ip;
iplist_index++;
}
}
xfs_refcache_size = xfs_refcache_new_size;
/*
* Move index to beginning of cache if it's now past the end
*/
if (xfs_refcache_index >= xfs_refcache_new_size)
xfs_refcache_index = 0;
spin_unlock(&xfs_refcache_lock);
/*
* Now drop the inodes we collected.
*/
for (i = 0; i < iplist_index; i++) {
VOP_RELEASE(XFS_ITOV(iplist[i]), error);
VN_RELE(XFS_ITOV(iplist[i]));
}
kmem_free(iplist, XFS_REFCACHE_SIZE_MAX *
sizeof(xfs_inode_t *));
} else {
spin_lock(&xfs_refcache_lock);
xfs_refcache_size = xfs_refcache_new_size;
spin_unlock(&xfs_refcache_lock);
}
}
/*
* This is called from the XFS sync code to ensure that the refcache
* is emptied out over time. We purge a small number of entries with
* each call.
*/
void
xfs_refcache_purge_some(xfs_mount_t *mp)
{
int error, i;
xfs_inode_t *ip;
int iplist_index;
xfs_inode_t **iplist;
if ((xfs_refcache == NULL) || (xfs_refcache_count == 0)) {
return;
}
iplist_index = 0;
iplist = (xfs_inode_t **)kmem_zalloc(xfs_refcache_purge_count *
sizeof(xfs_inode_t *), KM_SLEEP);
spin_lock(&xfs_refcache_lock);
/*
* Store any inodes we find in the next several entries
* into the iplist array to be released after dropping
* the spinlock. We always start looking from the currently
* oldest place in the cache. We move the refcache index
* forward as we go so that we are sure to eventually clear
* out the entire cache when the system goes idle.
*/
for (i = 0; i < xfs_refcache_purge_count; i++) {
ip = xfs_refcache[xfs_refcache_index];
if (ip != NULL) {
xfs_refcache[xfs_refcache_index] = NULL;
ip->i_refcache = NULL;
xfs_refcache_count--;
ASSERT(xfs_refcache_count >= 0);
iplist[iplist_index] = ip;
iplist_index++;
}
xfs_refcache_index++;
if (xfs_refcache_index == xfs_refcache_size) {
xfs_refcache_index = 0;
}
}
spin_unlock(&xfs_refcache_lock);
/*
* Now drop the inodes we collected.
*/
for (i = 0; i < iplist_index; i++) {
VOP_RELEASE(XFS_ITOV(iplist[i]), error);
VN_RELE(XFS_ITOV(iplist[i]));
}
kmem_free(iplist, xfs_refcache_purge_count *
sizeof(xfs_inode_t *));
}
/* xfs_igetinode now returns an unlocked inode. This is fine, since we
* have a refcount on the holding vnode.
*/
int
xfs_igetinode(struct vfs *vfsp, dev_t dev, ino_t inode,
struct xfs_inode **ipp)
{
struct xfs_inode *ip;
vnode_t *vp;
vattr_t vattr;
int error;
AFS_STATCNT(igetinode);
*ipp = NULL;
if (error = xfs_getinode(vfsp, dev, inode, &ip)) {
return error;
}
xfs_iunlock(ip, XFS_ILOCK_SHARED);
vp = XFS_ITOV(ip);
vattr.va_mask = AT_STAT;
AFS_VOP_GETATTR(vp, &vattr, 0, OSI_GET_CURRENT_CRED(), error);
if (error) {
SET_XFS_ERROR(4, vp->v_vfsp->vfs_dev, inode);
VN_RELE(vp);
return error;
}
if (vattr.va_nlink == 0 || vattr.va_type != VREG) {
SET_XFS_ERROR(5, vp->v_vfsp->vfs_dev, inode);
VN_RELE(vp);
return ENOENT;
}
*ipp = ip;
return 0;
}
void
xfs_refcache_iunlock(
xfs_inode_t *ip,
uint lock_flags)
{
xfs_inode_t *release_ip;
int error;
release_ip = ip->i_release;
ip->i_release = NULL;
xfs_iunlock(ip, lock_flags);
if (release_ip != NULL) {
VOP_RELEASE(XFS_ITOV(release_ip), error);
VN_RELE(XFS_ITOV(release_ip));
}
}
/*
* xfs_root extracts the root vnode from a vfs.
*
* vfsp -- the vfs struct for the desired file system
* vpp -- address of the caller's vnode pointer which should be
* set to the desired fs root vnode
*/
STATIC int
xfs_root(
bhv_desc_t *bdp,
vnode_t **vpp)
{
vnode_t *vp;
vp = XFS_ITOV((XFS_BHVTOM(bdp))->m_rootip);
VN_HOLD(vp);
*vpp = vp;
return 0;
}
vnode_t *
afs_XFSIGetVnode(ino_t ainode)
{
struct xfs_inode *ip;
int error;
vnode_t *vp;
if ((error =
xfs_igetinode(afs_cacheVfsp, (dev_t) cacheDev.dev, ainode, &ip))) {
osi_Panic("afs_XFSIGetVnode: xfs_igetinode failed, error=%d", error);
}
vp = XFS_ITOV(ip);
return vp;
}
ssize_t
xfs_splice_read(
xfs_inode_t *ip,
struct file *infilp,
loff_t *ppos,
struct pipe_inode_info *pipe,
size_t count,
int flags,
int ioflags)
{
bhv_vnode_t *vp = XFS_ITOV(ip);
xfs_mount_t *mp = ip->i_mount;
ssize_t ret;
XFS_STATS_INC(xs_read_calls);
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(ip, DM_EVENT_READ) && !(ioflags & IO_INVIS)) {
bhv_vrwlock_t locktype = VRWLOCK_READ;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_READ, vp, *ppos, count,
FILP_DELAY_FLAG(infilp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return -error;
}
}
xfs_rw_enter_trace(XFS_SPLICE_READ_ENTER, ip,
pipe, count, *ppos, ioflags);
ret = generic_file_splice_read(infilp, ppos, pipe, count, flags);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return ret;
}
/*
* Special iput for brand-new inodes that are still locked
*/
void
xfs_iput_new(xfs_inode_t *ip,
uint lock_flags)
{
xfs_vnode_t *vp = XFS_ITOV(ip);
vn_trace_entry(vp, "xfs_iput_new", (inst_t *)__return_address);
printf("xfs_iput_new: ip %p\n",ip);
if ((ip->i_d.di_mode == 0)) {
ASSERT(!(ip->i_flags & XFS_IRECLAIMABLE));
//vn_mark_bad(vp);
printf("xfs_iput_new: ip %p di_mode == 0\n",ip);
/* mabe call vgone here? RMC */
}
if (lock_flags)
xfs_iunlock(ip, lock_flags);
ASSERT_VOP_LOCKED(vp->v_vnode, "xfs_iput_new");
vput(vp->v_vnode);
}
/*
* Change the requested timestamp in the given inode.
* We don't lock across timestamp updates, and we don't log them but
* we do record the fact that there is dirty information in core.
*
* NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
* with XFS_ICHGTIME_ACC to be sure that access time
* update will take. Calling first with XFS_ICHGTIME_ACC
* and then XFS_ICHGTIME_MOD may fail to modify the access
* timestamp if the filesystem is mounted noacctm.
*/
void
xfs_ichgtime(
xfs_inode_t *ip,
int flags)
{
struct inode *inode = vn_to_inode(XFS_ITOV(ip));
timespec_t tv;
nanotime(&tv);
if (flags & XFS_ICHGTIME_MOD) {
inode->i_mtime = tv;
ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
}
if (flags & XFS_ICHGTIME_ACC) {
inode->i_atime = tv;
ip->i_d.di_atime.t_sec = (__int32_t)tv.tv_sec;
ip->i_d.di_atime.t_nsec = (__int32_t)tv.tv_nsec;
}
if (flags & XFS_ICHGTIME_CHG) {
inode->i_ctime = tv;
ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
}
/*
* We update the i_update_core field _after_ changing
* the timestamps in order to coordinate properly with
* xfs_iflush() so that we don't lose timestamp updates.
* This keeps us from having to hold the inode lock
* while doing this. We use the SYNCHRONIZE macro to
* ensure that the compiler does not reorder the update
* of i_update_core above the timestamp updates above.
*/
SYNCHRONIZE();
ip->i_update_core = 1;
if (!(inode->i_state & I_NEW))
mark_inode_dirty_sync(inode);
}
/*
* This is called from the XFS unmount code to purge all entries for the
* given mount from the cache. It uses the refcache busy counter to
* make sure that new entries are not added to the cache as we purge them.
*/
void
xfs_refcache_purge_mp(
xfs_mount_t *mp)
{
vnode_t *vp;
int error, i;
xfs_inode_t *ip;
if (xfs_refcache == NULL) {
return;
}
spin_lock(&xfs_refcache_lock);
/*
* Bumping the busy counter keeps new entries from being added
* to the cache. We use a counter since multiple unmounts could
* be in here simultaneously.
*/
xfs_refcache_busy++;
for (i = 0; i < xfs_refcache_size; i++) {
ip = xfs_refcache[i];
if ((ip != NULL) && (ip->i_mount == mp)) {
xfs_refcache[i] = NULL;
ip->i_refcache = NULL;
xfs_refcache_count--;
ASSERT(xfs_refcache_count >= 0);
spin_unlock(&xfs_refcache_lock);
vp = XFS_ITOV(ip);
VOP_RELEASE(vp, error);
VN_RELE(vp);
spin_lock(&xfs_refcache_lock);
}
}
xfs_refcache_busy--;
ASSERT(xfs_refcache_busy >= 0);
spin_unlock(&xfs_refcache_lock);
}
/*
* If the given inode is in the reference cache, purge its entry and
* release the reference on the vnode.
*/
void
xfs_refcache_purge_ip(
xfs_inode_t *ip)
{
vnode_t *vp;
int error;
/*
* If we're not pointing to our entry in the cache, then
* we must not be in the cache.
*/
if (ip->i_refcache == NULL) {
return;
}
spin_lock(&xfs_refcache_lock);
if (ip->i_refcache == NULL) {
spin_unlock(&xfs_refcache_lock);
return;
}
/*
* Clear both our pointer to the cache entry and its pointer
* back to us.
*/
ASSERT(*(ip->i_refcache) == ip);
*(ip->i_refcache) = NULL;
ip->i_refcache = NULL;
xfs_refcache_count--;
ASSERT(xfs_refcache_count >= 0);
spin_unlock(&xfs_refcache_lock);
vp = XFS_ITOV(ip);
/* ASSERT(vp->v_count > 1); */
VOP_RELEASE(vp, error);
VN_RELE(vp);
}
int
xfs_read_file(xfs_mount_t *mp, xfs_inode_t *ip, struct uio *uio, int ioflag)
{
xfs_fileoff_t lbn, nextlbn;
xfs_fsize_t bytesinfile;
long size, xfersize, blkoffset;
struct buf *bp;
struct vnode *vp;
int error, orig_resid;
int seqcount;
seqcount = ioflag >> IO_SEQSHIFT;
orig_resid = uio->uio_resid;
if (orig_resid <= 0)
return (0);
vp = XFS_ITOV(ip)->v_vnode;
/*
* Ok so we couldn't do it all in one vm trick...
* so cycle around trying smaller bites..
*/
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = ip->i_d.di_size - uio->uio_offset) <= 0)
break;
lbn = XFS_B_TO_FSBT(mp, uio->uio_offset);
nextlbn = lbn + 1;
/*
* size of buffer. The buffer representing the
* end of the file is rounded up to the size of
* the block type ( fragment or full block,
* depending ).
*/
size = mp->m_sb.sb_blocksize;
blkoffset = XFS_B_FSB_OFFSET(mp, uio->uio_offset);
/*
* The amount we want to transfer in this iteration is
* one FS block less the amount of the data before
* our startpoint (duh!)
*/
xfersize = mp->m_sb.sb_blocksize - blkoffset;
/*
* But if we actually want less than the block,
* or the file doesn't have a whole block more of data,
* then use the lesser number.
*/
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (XFS_FSB_TO_B(mp, nextlbn) >= ip->i_d.di_size ) {
/*
* Don't do readahead if this is the end of the file.
*/
error = bread(vp, lbn, size, NOCRED, &bp);
} else if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
/*
* Otherwise if we are allowed to cluster,
* grab as much as we can.
*
* XXX This may not be a win if we are not
* doing sequential access.
*/
error = cluster_read(vp, ip->i_d.di_size, lbn,
size, NOCRED, uio->uio_resid, seqcount, &bp);
} else if (seqcount > 1) {
/*
* If we are NOT allowed to cluster, then
* if we appear to be acting sequentially,
* fire off a request for a readahead
* as well as a read. Note that the 4th and 5th
* arguments point to arrays of the size specified in
* the 6th argument.
*/
int nextsize = mp->m_sb.sb_blocksize;
error = breadn(vp, lbn,
size, &nextlbn, &nextsize, 1, NOCRED, &bp);
} else {
/*
* Failing all of the above, just read what the
* user asked for. Interestingly, the same as
* the first option above.
*/
error = bread(vp, lbn, size, NOCRED, &bp);
}
if (error) {
brelse(bp);
bp = NULL;
break;
}
/*
* If IO_DIRECT then set B_DIRECT for the buffer. This
* will cause us to attempt to release the buffer later on
* and will cause the buffer cache to attempt to free the
* underlying pages.
*/
if (ioflag & IO_DIRECT)
bp->b_flags |= B_DIRECT;
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
/*
* otherwise use the general form
*/
error = uiomove((char *)bp->b_data + blkoffset,
(int)xfersize, uio);
if (error)
break;
if (ioflag & (IO_VMIO|IO_DIRECT) ) {
/*
* If there are no dependencies, and it's VMIO,
* then we don't need the buf, mark it available
* for freeing. The VM has the data.
*/
bp->b_flags |= B_RELBUF;
brelse(bp);
} else {
/*
* Otherwise let whoever
* made the request take care of
* freeing it. We just queue
* it onto another list.
*/
bqrelse(bp);
}
}
/*
* This can only happen in the case of an error
* because the loop above resets bp to NULL on each iteration
* and on normal completion has not set a new value into it.
* so it must have come from a 'break' statement
*/
if (bp != NULL) {
if (ioflag & (IO_VMIO|IO_DIRECT)) {
bp->b_flags |= B_RELBUF;
brelse(bp);
} else
bqrelse(bp);
}
return (error);
}
/*
* Insert the given inode into the reference cache.
*/
void
xfs_refcache_insert(
xfs_inode_t *ip)
{
vnode_t *vp;
xfs_inode_t *release_ip;
xfs_inode_t **refcache;
ASSERT(ismrlocked(&(ip->i_iolock), MR_UPDATE));
/*
* If an unmount is busy blowing entries out of the cache,
* then don't bother.
*/
if (xfs_refcache_busy) {
return;
}
/*
* If we tuned the refcache down to zero, don't do anything.
*/
if (!xfs_refcache_size) {
return;
}
/*
* The inode is already in the refcache, so don't bother
* with it.
*/
if (ip->i_refcache != NULL) {
return;
}
vp = XFS_ITOV(ip);
/* ASSERT(vp->v_count > 0); */
VN_HOLD(vp);
/*
* We allocate the reference cache on use so that we don't
* waste the memory on systems not being used as NFS servers.
*/
if (xfs_refcache == NULL) {
refcache = (xfs_inode_t **)kmem_zalloc(XFS_REFCACHE_SIZE_MAX *
sizeof(xfs_inode_t *),
KM_SLEEP);
} else {
refcache = NULL;
}
spin_lock(&xfs_refcache_lock);
/*
* If we allocated memory for the refcache above and it still
* needs it, then use the memory we allocated. Otherwise we'll
* free the memory below.
*/
if (refcache != NULL) {
if (xfs_refcache == NULL) {
xfs_refcache = refcache;
refcache = NULL;
}
}
/*
* If an unmount is busy clearing out the cache, don't add new
* entries to it.
*/
if (xfs_refcache_busy) {
spin_unlock(&xfs_refcache_lock);
VN_RELE(vp);
/*
* If we allocated memory for the refcache above but someone
* else beat us to using it, then free the memory now.
*/
if (refcache != NULL) {
kmem_free(refcache,
XFS_REFCACHE_SIZE_MAX * sizeof(xfs_inode_t *));
}
return;
}
release_ip = xfs_refcache[xfs_refcache_index];
if (release_ip != NULL) {
release_ip->i_refcache = NULL;
xfs_refcache_count--;
ASSERT(xfs_refcache_count >= 0);
}
xfs_refcache[xfs_refcache_index] = ip;
ASSERT(ip->i_refcache == NULL);
ip->i_refcache = &(xfs_refcache[xfs_refcache_index]);
xfs_refcache_count++;
ASSERT(xfs_refcache_count <= xfs_refcache_size);
xfs_refcache_index++;
if (xfs_refcache_index == xfs_refcache_size) {
xfs_refcache_index = 0;
}
spin_unlock(&xfs_refcache_lock);
/*
* Save the pointer to the inode to be released so that we can
* VN_RELE it once we've dropped our inode locks in xfs_rwunlock().
* The pointer may be NULL, but that's OK.
*/
ip->i_release = release_ip;
/*
* If we allocated memory for the refcache above but someone
* else beat us to using it, then free the memory now.
*/
if (refcache != NULL) {
kmem_free(refcache,
XFS_REFCACHE_SIZE_MAX * sizeof(xfs_inode_t *));
}
}
/*
* xfs sync routine for internal use
*
* This routine supports all of the flags defined for the generic VFS_SYNC
* interface as explained above under xfs_sync. In the interests of not
* changing interfaces within the 6.5 family, additional internallly-
* required functions are specified within a separate xflags parameter,
* only available by calling this routine.
*
*/
STATIC int
xfs_sync_inodes(
xfs_mount_t *mp,
int flags,
int xflags,
int *bypassed)
{
xfs_inode_t *ip = NULL;
xfs_inode_t *ip_next;
xfs_buf_t *bp;
vnode_t *vp = NULL;
vmap_t vmap;
int error;
int last_error;
uint64_t fflag;
uint lock_flags;
uint base_lock_flags;
boolean_t mount_locked;
boolean_t vnode_refed;
int preempt;
xfs_dinode_t *dip;
xfs_iptr_t *ipointer;
#ifdef DEBUG
boolean_t ipointer_in = B_FALSE;
#define IPOINTER_SET ipointer_in = B_TRUE
#define IPOINTER_CLR ipointer_in = B_FALSE
#else
#define IPOINTER_SET
#define IPOINTER_CLR
#endif
/* Insert a marker record into the inode list after inode ip. The list
* must be locked when this is called. After the call the list will no
* longer be locked.
*/
#define IPOINTER_INSERT(ip, mp) { \
ASSERT(ipointer_in == B_FALSE); \
ipointer->ip_mnext = ip->i_mnext; \
ipointer->ip_mprev = ip; \
ip->i_mnext = (xfs_inode_t *)ipointer; \
ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
preempt = 0; \
XFS_MOUNT_IUNLOCK(mp); \
mount_locked = B_FALSE; \
IPOINTER_SET; \
}
/* Remove the marker from the inode list. If the marker was the only item
* in the list then there are no remaining inodes and we should zero out
* the whole list. If we are the current head of the list then move the head
* past us.
*/
#define IPOINTER_REMOVE(ip, mp) { \
ASSERT(ipointer_in == B_TRUE); \
if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
ip = ipointer->ip_mnext; \
ip->i_mprev = ipointer->ip_mprev; \
ipointer->ip_mprev->i_mnext = ip; \
if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
mp->m_inodes = ip; \
} \
} else { \
ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
mp->m_inodes = NULL; \
ip = NULL; \
} \
IPOINTER_CLR; \
}
#define XFS_PREEMPT_MASK 0x7f
if (bypassed)
*bypassed = 0;
if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
return 0;
error = 0;
last_error = 0;
preempt = 0;
/* Allocate a reference marker */
ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
fflag = XFS_B_ASYNC; /* default is don't wait */
if (flags & SYNC_BDFLUSH)
fflag = XFS_B_DELWRI;
if (flags & SYNC_WAIT)
fflag = 0; /* synchronous overrides all */
base_lock_flags = XFS_ILOCK_SHARED;
if (flags & (SYNC_DELWRI | SYNC_CLOSE)) {
/*
* We need the I/O lock if we're going to call any of
* the flush/inval routines.
*/
base_lock_flags |= XFS_IOLOCK_SHARED;
}
XFS_MOUNT_ILOCK(mp);
ip = mp->m_inodes;
mount_locked = B_TRUE;
vnode_refed = B_FALSE;
IPOINTER_CLR;
do {
ASSERT(ipointer_in == B_FALSE);
ASSERT(vnode_refed == B_FALSE);
lock_flags = base_lock_flags;
/*
* There were no inodes in the list, just break out
* of the loop.
*/
if (ip == NULL) {
break;
}
/*
* We found another sync thread marker - skip it
*/
if (ip->i_mount == NULL) {
ip = ip->i_mnext;
continue;
}
vp = XFS_ITOV_NULL(ip);
/*
* If the vnode is gone then this is being torn down,
* call reclaim if it is flushed, else let regular flush
* code deal with it later in the loop.
*/
if (vp == NULL) {
/* Skip ones already in reclaim */
if (ip->i_flags & XFS_IRECLAIM) {
ip = ip->i_mnext;
continue;
}
if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
ip = ip->i_mnext;
} else if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
IPOINTER_INSERT(ip, mp);
xfs_finish_reclaim(ip, 1,
XFS_IFLUSH_DELWRI_ELSE_ASYNC);
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
IPOINTER_REMOVE(ip, mp);
} else {
xfs_iunlock(ip, XFS_ILOCK_EXCL);
ip = ip->i_mnext;
}
continue;
}
if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
XFS_MOUNT_IUNLOCK(mp);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return 0;
}
/*
* If this is just vfs_sync() or pflushd() calling
* then we can skip inodes for which it looks like
* there is nothing to do. Since we don't have the
* inode locked this is racey, but these are periodic
* calls so it doesn't matter. For the others we want
* to know for sure, so we at least try to lock them.
*/
if (flags & SYNC_BDFLUSH) {
if (((ip->i_itemp == NULL) ||
!(ip->i_itemp->ili_format.ilf_fields &
XFS_ILOG_ALL)) &&
(ip->i_update_core == 0)) {
ip = ip->i_mnext;
continue;
}
}
/*
* Try to lock without sleeping. We're out of order with
* the inode list lock here, so if we fail we need to drop
* the mount lock and try again. If we're called from
* bdflush() here, then don't bother.
*
* The inode lock here actually coordinates with the
* almost spurious inode lock in xfs_ireclaim() to prevent
* the vnode we handle here without a reference from
* being freed while we reference it. If we lock the inode
* while it's on the mount list here, then the spurious inode
* lock in xfs_ireclaim() after the inode is pulled from
* the mount list will sleep until we release it here.
* This keeps the vnode from being freed while we reference
* it. It is also cheaper and simpler than actually doing
* a vn_get() for every inode we touch here.
*/
if (xfs_ilock_nowait(ip, lock_flags) == 0) {
if ((flags & SYNC_BDFLUSH) || (vp == NULL)) {
ip = ip->i_mnext;
continue;
}
/*
* We need to unlock the inode list lock in order
* to lock the inode. Insert a marker record into
* the inode list to remember our position, dropping
* the lock is now done inside the IPOINTER_INSERT
* macro.
*
* We also use the inode list lock to protect us
* in taking a snapshot of the vnode version number
* for use in calling vn_get().
*/
VMAP(vp, vmap);
IPOINTER_INSERT(ip, mp);
vp = vn_get(vp, &vmap);
if (vp == NULL) {
/*
* The vnode was reclaimed once we let go
* of the inode list lock. Skip to the
* next list entry. Remove the marker.
*/
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
vnode_refed = B_FALSE;
IPOINTER_REMOVE(ip, mp);
continue;
}
xfs_ilock(ip, lock_flags);
ASSERT(vp == XFS_ITOV(ip));
ASSERT(ip->i_mount == mp);
vnode_refed = B_TRUE;
}
/* From here on in the loop we may have a marker record
* in the inode list.
*/
if ((flags & SYNC_CLOSE) && (vp != NULL)) {
/*
* This is the shutdown case. We just need to
* flush and invalidate all the pages associated
* with the inode. Drop the inode lock since
* we can't hold it across calls to the buffer
* cache.
*
* We don't set the VREMAPPING bit in the vnode
* here, because we don't hold the vnode lock
* exclusively. It doesn't really matter, though,
* because we only come here when we're shutting
* down anyway.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (XFS_FORCED_SHUTDOWN(mp)) {
VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
} else {
VOP_FLUSHINVAL_PAGES(vp, 0, -1, FI_REMAPF);
}
xfs_ilock(ip, XFS_ILOCK_SHARED);
} else if ((flags & SYNC_DELWRI) && (vp != NULL)) {
if (VN_DIRTY(vp)) {
/* We need to have dropped the lock here,
* so insert a marker if we have not already
* done so.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
/*
* Drop the inode lock since we can't hold it
* across calls to the buffer cache.
*/
xfs_iunlock(ip, XFS_ILOCK_SHARED);
VOP_FLUSH_PAGES(vp, (xfs_off_t)0, -1,
fflag, FI_NONE, error);
xfs_ilock(ip, XFS_ILOCK_SHARED);
}
}
if (flags & SYNC_BDFLUSH) {
if ((flags & SYNC_ATTR) &&
((ip->i_update_core) ||
((ip->i_itemp != NULL) &&
(ip->i_itemp->ili_format.ilf_fields != 0)))) {
/* Insert marker and drop lock if not already
* done.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
/*
* We don't want the periodic flushing of the
* inodes by vfs_sync() to interfere with
* I/O to the file, especially read I/O
* where it is only the access time stamp
* that is being flushed out. To prevent
* long periods where we have both inode
* locks held shared here while reading the
* inode's buffer in from disk, we drop the
* inode lock while reading in the inode
* buffer. We have to release the buffer
* and reacquire the inode lock so that they
* are acquired in the proper order (inode
* locks first). The buffer will go at the
* end of the lru chain, though, so we can
* expect it to still be there when we go
* for it again in xfs_iflush().
*/
if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
xfs_ifunlock(ip);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
error = xfs_itobp(mp, NULL, ip,
&dip, &bp, 0);
if (!error) {
xfs_buf_relse(bp);
} else {
/* Bailing out, remove the
* marker and free it.
*/
XFS_MOUNT_ILOCK(mp);
IPOINTER_REMOVE(ip, mp);
XFS_MOUNT_IUNLOCK(mp);
ASSERT(!(lock_flags &
XFS_IOLOCK_SHARED));
kmem_free(ipointer,
sizeof(xfs_iptr_t));
return (0);
}
/*
* Since we dropped the inode lock,
* the inode may have been reclaimed.
* Therefore, we reacquire the mount
* lock and check to see if we were the
* inode reclaimed. If this happened
* then the ipointer marker will no
* longer point back at us. In this
* case, move ip along to the inode
* after the marker, remove the marker
* and continue.
*/
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
if (ip != ipointer->ip_mprev) {
IPOINTER_REMOVE(ip, mp);
ASSERT(!vnode_refed);
ASSERT(!(lock_flags &
XFS_IOLOCK_SHARED));
continue;
}
ASSERT(ip->i_mount == mp);
if (xfs_ilock_nowait(ip,
XFS_ILOCK_SHARED) == 0) {
ASSERT(ip->i_mount == mp);
/*
* We failed to reacquire
* the inode lock without
* sleeping, so just skip
* the inode for now. We
* clear the ILOCK bit from
* the lock_flags so that we
* won't try to drop a lock
* we don't hold below.
*/
lock_flags &= ~XFS_ILOCK_SHARED;
IPOINTER_REMOVE(ip_next, mp);
} else if ((xfs_ipincount(ip) == 0) &&
xfs_iflock_nowait(ip)) {
ASSERT(ip->i_mount == mp);
/*
* Since this is vfs_sync()
* calling we only flush the
* inode out if we can lock
* it without sleeping and
* it is not pinned. Drop
* the mount lock here so
* that we don't hold it for
* too long. We already have
* a marker in the list here.
*/
XFS_MOUNT_IUNLOCK(mp);
mount_locked = B_FALSE;
error = xfs_iflush(ip,
XFS_IFLUSH_DELWRI);
} else {
ASSERT(ip->i_mount == mp);
IPOINTER_REMOVE(ip_next, mp);
}
}
}
} else {
if ((flags & SYNC_ATTR) &&
((ip->i_update_core) ||
((ip->i_itemp != NULL) &&
(ip->i_itemp->ili_format.ilf_fields != 0)))) {
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
if (flags & SYNC_WAIT) {
xfs_iflock(ip);
error = xfs_iflush(ip,
XFS_IFLUSH_SYNC);
} else {
/*
* If we can't acquire the flush
* lock, then the inode is already
* being flushed so don't bother
* waiting. If we can lock it then
* do a delwri flush so we can
* combine multiple inode flushes
* in each disk write.
*/
if (xfs_iflock_nowait(ip)) {
error = xfs_iflush(ip,
XFS_IFLUSH_DELWRI);
}
else if (bypassed)
(*bypassed)++;
}
}
}
if (lock_flags != 0) {
xfs_iunlock(ip, lock_flags);
}
if (vnode_refed) {
/*
* If we had to take a reference on the vnode
* above, then wait until after we've unlocked
* the inode to release the reference. This is
* because we can be already holding the inode
* lock when VN_RELE() calls xfs_inactive().
*
* Make sure to drop the mount lock before calling
* VN_RELE() so that we don't trip over ourselves if
* we have to go for the mount lock again in the
* inactive code.
*/
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
VN_RELE(vp);
vnode_refed = B_FALSE;
}
if (error) {
last_error = error;
}
/*
* bail out if the filesystem is corrupted.
*/
if (error == EFSCORRUPTED) {
if (!mount_locked) {
XFS_MOUNT_ILOCK(mp);
IPOINTER_REMOVE(ip, mp);
}
XFS_MOUNT_IUNLOCK(mp);
ASSERT(ipointer_in == B_FALSE);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return XFS_ERROR(error);
}
/* Let other threads have a chance at the mount lock
* if we have looped many times without dropping the
* lock.
*/
if ((++preempt & XFS_PREEMPT_MASK) == 0) {
if (mount_locked) {
IPOINTER_INSERT(ip, mp);
}
}
if (mount_locked == B_FALSE) {
XFS_MOUNT_ILOCK(mp);
mount_locked = B_TRUE;
IPOINTER_REMOVE(ip, mp);
continue;
}
ASSERT(ipointer_in == B_FALSE);
ip = ip->i_mnext;
} while (ip != mp->m_inodes);
XFS_MOUNT_IUNLOCK(mp);
ASSERT(ipointer_in == B_FALSE);
kmem_free(ipointer, sizeof(xfs_iptr_t));
return XFS_ERROR(last_error);
}
ssize_t /* bytes read, or (-) error */
xfs_read(
xfs_inode_t *ip,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int segs,
loff_t *offset,
int ioflags)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
bhv_vnode_t *vp = XFS_ITOV(ip);
xfs_mount_t *mp = ip->i_mount;
size_t size = 0;
ssize_t ret = 0;
xfs_fsize_t n;
unsigned long seg;
XFS_STATS_INC(xs_read_calls);
/* START copy & waste from filemap.c */
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
size += iv->iov_len;
if (unlikely((ssize_t)(size|iv->iov_len) < 0))
return XFS_ERROR(-EINVAL);
}
/* END copy & waste from filemap.c */
if (unlikely(ioflags & IO_ISDIRECT)) {
xfs_buftarg_t *target =
XFS_IS_REALTIME_INODE(ip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((*offset & target->bt_smask) ||
(size & target->bt_smask)) {
if (*offset == ip->i_size) {
return (0);
}
return -XFS_ERROR(EINVAL);
}
}
n = XFS_MAXIOFFSET(mp) - *offset;
if ((n <= 0) || (size == 0))
return 0;
if (n < size)
size = n;
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
if (unlikely(ioflags & IO_ISDIRECT))
mutex_lock(&inode->i_mutex);
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (DM_EVENT_ENABLED(ip, DM_EVENT_READ) && !(ioflags & IO_INVIS)) {
bhv_vrwlock_t locktype = VRWLOCK_READ;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
ret = -XFS_SEND_DATA(mp, DM_EVENT_READ, vp, *offset, size,
dmflags, &locktype);
if (ret) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
if (unlikely(ioflags & IO_ISDIRECT))
mutex_unlock(&inode->i_mutex);
return ret;
}
}
if (unlikely(ioflags & IO_ISDIRECT)) {
if (VN_CACHED(vp))
ret = xfs_flushinval_pages(ip, (*offset & PAGE_CACHE_MASK),
-1, FI_REMAPF_LOCKED);
mutex_unlock(&inode->i_mutex);
if (ret) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return ret;
}
}
xfs_rw_enter_trace(XFS_READ_ENTER, ip,
(void *)iovp, segs, *offset, ioflags);
iocb->ki_pos = *offset;
ret = generic_file_aio_read(iocb, iovp, segs, *offset);
if (ret == -EIOCBQUEUED && !(ioflags & IO_ISAIO))
ret = wait_on_sync_kiocb(iocb);
if (ret > 0)
XFS_STATS_ADD(xs_read_bytes, ret);
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return ret;
}
ssize_t
xfs_splice_write(
xfs_inode_t *ip,
struct pipe_inode_info *pipe,
struct file *outfilp,
loff_t *ppos,
size_t count,
int flags,
int ioflags)
{
bhv_vnode_t *vp = XFS_ITOV(ip);
xfs_mount_t *mp = ip->i_mount;
ssize_t ret;
struct inode *inode = outfilp->f_mapping->host;
xfs_fsize_t isize, new_size;
XFS_STATS_INC(xs_write_calls);
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
return -EIO;
xfs_ilock(ip, XFS_IOLOCK_EXCL);
if (DM_EVENT_ENABLED(ip, DM_EVENT_WRITE) && !(ioflags & IO_INVIS)) {
bhv_vrwlock_t locktype = VRWLOCK_WRITE;
int error;
error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, vp, *ppos, count,
FILP_DELAY_FLAG(outfilp), &locktype);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return -error;
}
}
new_size = *ppos + count;
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (new_size > ip->i_size)
ip->i_new_size = new_size;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_rw_enter_trace(XFS_SPLICE_WRITE_ENTER, ip,
pipe, count, *ppos, ioflags);
ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
if (ret > 0)
XFS_STATS_ADD(xs_write_bytes, ret);
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && *ppos > isize))
*ppos = isize;
if (*ppos > ip->i_size) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
if (*ppos > ip->i_size)
ip->i_size = *ppos;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
if (ip->i_new_size) {
xfs_ilock(ip, XFS_ILOCK_EXCL);
ip->i_new_size = 0;
if (ip->i_d.di_size > ip->i_size)
ip->i_d.di_size = ip->i_size;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
}
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
return ret;
}
/*
* xfs_unmount_flush implements a set of flush operation on special
* inodes, which are needed as a separate set of operations so that
* they can be called as part of relocation process.
*/
int
xfs_unmount_flush(
xfs_mount_t *mp, /* Mount structure we are getting
rid of. */
int relocation) /* Called from vfs relocation. */
{
xfs_inode_t *rip = mp->m_rootip;
xfs_inode_t *rbmip;
xfs_inode_t *rsumip = NULL;
vnode_t *rvp = XFS_ITOV(rip);
int error;
xfs_ilock(rip, XFS_ILOCK_EXCL);
xfs_iflock(rip);
/*
* Flush out the real time inodes.
*/
if ((rbmip = mp->m_rbmip) != NULL) {
xfs_ilock(rbmip, XFS_ILOCK_EXCL);
xfs_iflock(rbmip);
error = xfs_iflush(rbmip, XFS_IFLUSH_SYNC);
xfs_iunlock(rbmip, XFS_ILOCK_EXCL);
if (error == EFSCORRUPTED)
goto fscorrupt_out;
ASSERT(vn_count(XFS_ITOV(rbmip)) == 1);
rsumip = mp->m_rsumip;
xfs_ilock(rsumip, XFS_ILOCK_EXCL);
xfs_iflock(rsumip);
error = xfs_iflush(rsumip, XFS_IFLUSH_SYNC);
xfs_iunlock(rsumip, XFS_ILOCK_EXCL);
if (error == EFSCORRUPTED)
goto fscorrupt_out;
ASSERT(vn_count(XFS_ITOV(rsumip)) == 1);
}
/*
* Synchronously flush root inode to disk
*/
error = xfs_iflush(rip, XFS_IFLUSH_SYNC);
if (error == EFSCORRUPTED)
goto fscorrupt_out2;
if (vn_count(rvp) != 1 && !relocation) {
xfs_iunlock(rip, XFS_ILOCK_EXCL);
return XFS_ERROR(EBUSY);
}
/*
* Release dquot that rootinode, rbmino and rsumino might be holding,
* flush and purge the quota inodes.
*/
error = XFS_QM_UNMOUNT(mp);
if (error == EFSCORRUPTED)
goto fscorrupt_out2;
if (rbmip) {
VN_RELE(XFS_ITOV(rbmip));
VN_RELE(XFS_ITOV(rsumip));
}
xfs_iunlock(rip, XFS_ILOCK_EXCL);
return 0;
fscorrupt_out:
xfs_ifunlock(rip);
fscorrupt_out2:
xfs_iunlock(rip, XFS_ILOCK_EXCL);
return XFS_ERROR(EFSCORRUPTED);
}
/*
* Convert userspace handle data into vnode (and inode).
* We [ab]use the fact that all the fsop_handlereq ioctl calls
* have a data structure argument whose first component is always
* a xfs_fsop_handlereq_t, so we can cast to and from this type.
* This allows us to optimise the copy_from_user calls and gives
* a handy, shared routine.
*
* If no error, caller must always VN_RELE the returned vp.
*/
STATIC int
xfs_vget_fsop_handlereq(
xfs_mount_t *mp,
struct inode *parinode, /* parent inode pointer */
xfs_fsop_handlereq_t *hreq,
vnode_t **vp,
struct inode **inode)
{
void __user *hanp;
size_t hlen;
xfs_fid_t *xfid;
xfs_handle_t *handlep;
xfs_handle_t handle;
xfs_inode_t *ip;
struct inode *inodep;
vnode_t *vpp;
xfs_ino_t ino;
__u32 igen;
int error;
/*
* Only allow handle opens under a directory.
*/
if (!S_ISDIR(parinode->i_mode))
return XFS_ERROR(ENOTDIR);
hanp = hreq->ihandle;
hlen = hreq->ihandlen;
handlep = &handle;
if (hlen < sizeof(handlep->ha_fsid) || hlen > sizeof(*handlep))
return XFS_ERROR(EINVAL);
if (copy_from_user(handlep, hanp, hlen))
return XFS_ERROR(EFAULT);
if (hlen < sizeof(*handlep))
memset(((char *)handlep) + hlen, 0, sizeof(*handlep) - hlen);
if (hlen > sizeof(handlep->ha_fsid)) {
if (handlep->ha_fid.xfs_fid_len !=
(hlen - sizeof(handlep->ha_fsid)
- sizeof(handlep->ha_fid.xfs_fid_len))
|| handlep->ha_fid.xfs_fid_pad)
return XFS_ERROR(EINVAL);
}
/*
* Crack the handle, obtain the inode # & generation #
*/
xfid = (struct xfs_fid *)&handlep->ha_fid;
if (xfid->xfs_fid_len == sizeof(*xfid) - sizeof(xfid->xfs_fid_len)) {
ino = xfid->xfs_fid_ino;
igen = xfid->xfs_fid_gen;
} else {
return XFS_ERROR(EINVAL);
}
/*
* Get the XFS inode, building a vnode to go with it.
*/
error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_SHARED, &ip, 0);
if (error)
return error;
if (ip == NULL)
return XFS_ERROR(EIO);
if (ip->i_d.di_mode == 0 || ip->i_d.di_gen != igen) {
xfs_iput_new(ip, XFS_ILOCK_SHARED);
return XFS_ERROR(ENOENT);
}
vpp = XFS_ITOV(ip);
inodep = vn_to_inode(vpp);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
*vp = vpp;
*inode = inodep;
return 0;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
struct xfs_inode *xip,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int nsegs,
loff_t *offset,
int ioflags)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
bhv_vnode_t *vp = XFS_ITOV(xip);
unsigned long segs = nsegs;
xfs_mount_t *mp;
ssize_t ret = 0, error = 0;
xfs_fsize_t isize, new_size;
int iolock;
int eventsent = 0;
bhv_vrwlock_t locktype;
size_t ocount = 0, count;
loff_t pos;
int need_i_mutex;
XFS_STATS_INC(xs_write_calls);
error = generic_segment_checks(iovp, &segs, &ocount, VERIFY_READ);
if (error)
return error;
count = ocount;
pos = *offset;
if (count == 0)
return 0;
mp = xip->i_mount;
xfs_wait_for_freeze(mp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp))
return -EIO;
relock:
if (ioflags & IO_ISDIRECT) {
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
need_i_mutex = 0;
} else {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
start:
error = -generic_write_checks(file, &pos, &count,
S_ISBLK(inode->i_mode));
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
goto out_unlock_mutex;
}
if ((DM_EVENT_ENABLED(xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
int dmflags = FILP_DELAY_FLAG(file);
if (need_i_mutex)
dmflags |= DM_FLAGS_IMUX;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
pos, count,
dmflags, &locktype);
if (error) {
goto out_unlock_internal;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reacquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) && pos != xip->i_size)
goto start;
}
if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target =
XFS_IS_REALTIME_INODE(xip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((pos & target->bt_smask) || (count & target->bt_smask)) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return XFS_ERROR(-EINVAL);
}
if (!need_i_mutex && (VN_CACHED(vp) || pos > xip->i_size)) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
need_i_mutex = 1;
mutex_lock(&inode->i_mutex);
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
goto start;
}
}
new_size = pos + count;
if (new_size > xip->i_size)
xip->i_new_size = new_size;
if (likely(!(ioflags & IO_INVIS))) {
file_update_time(file);
xfs_ichgtime_fast(xip, inode,
XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
}
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (pos > xip->i_size) {
error = xfs_zero_eof(xip, pos, xip->i_size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL);
goto out_unlock_internal;
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (likely(!error))
error = -remove_suid(file->f_path.dentry);
if (unlikely(error)) {
goto out_unlock_internal;
}
}
retry:
/* We can write back this queue in page reclaim */
current->backing_dev_info = mapping->backing_dev_info;
if ((ioflags & IO_ISDIRECT)) {
if (VN_CACHED(vp)) {
WARN_ON(need_i_mutex == 0);
xfs_inval_cached_trace(xip, pos, -1,
(pos & PAGE_CACHE_MASK), -1);
error = xfs_flushinval_pages(xip,
(pos & PAGE_CACHE_MASK),
-1, FI_REMAPF_LOCKED);
if (error)
goto out_unlock_internal;
}
if (need_i_mutex) {
/* demote the lock now the cached pages are gone */
xfs_ilock_demote(xip, XFS_IOLOCK_EXCL);
mutex_unlock(&inode->i_mutex);
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
need_i_mutex = 0;
}
xfs_rw_enter_trace(XFS_DIOWR_ENTER, xip, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_direct_write(iocb, iovp,
&segs, pos, offset, count, ocount);
/*
* direct-io write to a hole: fall through to buffered I/O
* for completing the rest of the request.
*/
if (ret >= 0 && ret != count) {
XFS_STATS_ADD(xs_write_bytes, ret);
pos += ret;
count -= ret;
ioflags &= ~IO_ISDIRECT;
xfs_iunlock(xip, iolock);
goto relock;
}
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, xip, (void *)iovp, segs,
*offset, ioflags);
ret = generic_file_buffered_write(iocb, iovp, segs,
pos, offset, count, ret);
}
current->backing_dev_info = NULL;
if (ret == -EIOCBQUEUED && !(ioflags & IO_ISAIO))
ret = wait_on_sync_kiocb(iocb);
if (ret == -ENOSPC &&
DM_EVENT_ENABLED(xip, DM_EVENT_NOSPACE) && !(ioflags & IO_INVIS)) {
xfs_rwunlock(xip, locktype);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_rwlock(xip, locktype);
if (error)
goto out_unlock_internal;
pos = xip->i_size;
ret = 0;
goto retry;
}
isize = i_size_read(inode);
if (unlikely(ret < 0 && ret != -EFAULT && *offset > isize))
*offset = isize;
if (*offset > xip->i_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_size)
xip->i_size = *offset;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
error = -ret;
if (ret <= 0)
goto out_unlock_internal;
XFS_STATS_ADD(xs_write_bytes, ret);
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
int error2;
xfs_rwunlock(xip, locktype);
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
error2 = sync_page_range(inode, mapping, pos, ret);
if (!error)
error = error2;
if (need_i_mutex)
mutex_lock(&inode->i_mutex);
xfs_rwlock(xip, locktype);
error2 = xfs_write_sync_logforce(mp, xip);
if (!error)
error = error2;
}
out_unlock_internal:
if (xip->i_new_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
xip->i_new_size = 0;
/*
* If this was a direct or synchronous I/O that failed (such
* as ENOSPC) then part of the I/O may have been written to
* disk before the error occured. In this case the on-disk
* file size may have been adjusted beyond the in-memory file
* size and now needs to be truncated back.
*/
if (xip->i_d.di_size > xip->i_size)
xip->i_d.di_size = xip->i_size;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
xfs_rwunlock(xip, locktype);
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
return -error;
}
/*
* Convert userspace handle data into vnode (and inode).
* We [ab]use the fact that all the fsop_handlereq ioctl calls
* have a data structure argument whose first component is always
* a xfs_fsop_handlereq_t, so we can cast to and from this type.
* This allows us to optimise the copy_from_user calls and gives
* a handy, shared routine.
*
* If no error, caller must always VN_RELE the returned vp.
*/
STATIC int
xfs_vget_fsop_handlereq(
xfs_mount_t *mp,
struct inode *parinode, /* parent inode pointer */
int cap, /* capability level for op */
unsigned long arg, /* userspace data pointer */
unsigned long size, /* size of expected struct */
/* output arguments */
xfs_fsop_handlereq_t *hreq,
vnode_t **vp,
struct inode **inode)
{
void *hanp;
size_t hlen;
xfs_fid_t *xfid;
xfs_handle_t *handlep;
xfs_handle_t handle;
xfs_inode_t *ip;
struct inode *inodep;
vnode_t *vpp;
xfs_ino_t ino;
__u32 igen;
int error;
if (!capable(cap))
return XFS_ERROR(EPERM);
/*
* Only allow handle opens under a directory.
*/
if (!S_ISDIR(parinode->i_mode))
return XFS_ERROR(ENOTDIR);
/*
* Copy the handle down from the user and validate
* that it looks to be in the correct format.
*/
if (copy_from_user(hreq, (struct xfs_fsop_handlereq *)arg, size))
return XFS_ERROR(EFAULT);
hanp = hreq->ihandle;
hlen = hreq->ihandlen;
handlep = &handle;
if (hlen < sizeof(handlep->ha_fsid) || hlen > sizeof(*handlep))
return XFS_ERROR(EINVAL);
if (copy_from_user(handlep, hanp, hlen))
return XFS_ERROR(EFAULT);
if (hlen < sizeof(*handlep))
memset(((char *)handlep) + hlen, 0, sizeof(*handlep) - hlen);
if (hlen > sizeof(handlep->ha_fsid)) {
if (handlep->ha_fid.xfs_fid_len !=
(hlen - sizeof(handlep->ha_fsid)
- sizeof(handlep->ha_fid.xfs_fid_len))
|| handlep->ha_fid.xfs_fid_pad)
return XFS_ERROR(EINVAL);
}
/*
* Crack the handle, obtain the inode # & generation #
*/
xfid = (struct xfs_fid *)&handlep->ha_fid;
if (xfid->xfs_fid_len == sizeof(*xfid) - sizeof(xfid->xfs_fid_len)) {
ino = xfid->xfs_fid_ino;
igen = xfid->xfs_fid_gen;
} else {
return XFS_ERROR(EINVAL);
}
/*
* Get the XFS inode, building a vnode to go with it.
*/
error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_SHARED, &ip, 0);
if (error)
return error;
if (ip == NULL)
return XFS_ERROR(EIO);
if (ip->i_d.di_mode == 0 || ip->i_d.di_gen != igen) {
xfs_iput_new(ip, XFS_ILOCK_SHARED);
return XFS_ERROR(ENOENT);
}
vpp = XFS_ITOV(ip);
inodep = LINVFS_GET_IP(vpp);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
*vp = vpp;
*inode = inodep;
return 0;
}
/*
* Allocate a block and fill it with dquots.
* This is called when the bmapi finds a hole.
*/
STATIC int
xfs_qm_dqalloc(
xfs_trans_t **tpp,
xfs_mount_t *mp,
xfs_dquot_t *dqp,
xfs_inode_t *quotip,
xfs_fileoff_t offset_fsb,
xfs_buf_t **O_bpp)
{
xfs_fsblock_t firstblock;
xfs_bmap_free_t flist;
xfs_bmbt_irec_t map;
int nmaps, error, committed;
xfs_buf_t *bp;
xfs_trans_t *tp = *tpp;
ASSERT(tp != NULL);
xfs_dqtrace_entry(dqp, "DQALLOC");
/*
* Initialize the bmap freelist prior to calling bmapi code.
*/
XFS_BMAP_INIT(&flist, &firstblock);
xfs_ilock(quotip, XFS_ILOCK_EXCL);
/*
* Return if this type of quotas is turned off while we didn't
* have an inode lock
*/
if (XFS_IS_THIS_QUOTA_OFF(dqp)) {
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
return (ESRCH);
}
/*
* xfs_trans_commit normally decrements the vnode ref count
* when it unlocks the inode. Since we want to keep the quota
* inode around, we bump the vnode ref count now.
*/
VN_HOLD(XFS_ITOV(quotip));
xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL);
nmaps = 1;
if ((error = xfs_bmapi(tp, quotip,
offset_fsb, XFS_DQUOT_CLUSTER_SIZE_FSB,
XFS_BMAPI_METADATA | XFS_BMAPI_WRITE,
&firstblock,
XFS_QM_DQALLOC_SPACE_RES(mp),
&map, &nmaps, &flist))) {
goto error0;
}
ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB);
ASSERT(nmaps == 1);
ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
(map.br_startblock != HOLESTARTBLOCK));
/*
* Keep track of the blkno to save a lookup later
*/
dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
/* now we can just get the buffer (there's nothing to read yet) */
bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
dqp->q_blkno,
XFS_QI_DQCHUNKLEN(mp),
0);
if (!bp || (error = XFS_BUF_GETERROR(bp)))
goto error1;
/*
* Make a chunk of dquots out of this buffer and log
* the entire thing.
*/
xfs_qm_init_dquot_blk(tp, mp, be32_to_cpu(dqp->q_core.d_id),
dqp->dq_flags & XFS_DQ_ALLTYPES, bp);
/*
* xfs_bmap_finish() may commit the current transaction and
* start a second transaction if the freelist is not empty.
*
* Since we still want to modify this buffer, we need to
* ensure that the buffer is not released on commit of
* the first transaction and ensure the buffer is added to the
* second transaction.
*
* If there is only one transaction then don't stop the buffer
* from being released when it commits later on.
*/
xfs_trans_bhold(tp, bp);
if ((error = xfs_bmap_finish(tpp, &flist, firstblock, &committed))) {
goto error1;
}
if (committed) {
tp = *tpp;
xfs_trans_bjoin(tp, bp);
} else {
xfs_trans_bhold_release(tp, bp);
}
*O_bpp = bp;
return 0;
error1:
xfs_bmap_cancel(&flist);
error0:
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
return (error);
}
/*
* Look up an inode by number in the given file system.
* The inode is looked up in the hash table for the file system
* represented by the mount point parameter mp. Each bucket of
* the hash table is guarded by an individual semaphore.
*
* If the inode is found in the hash table, its corresponding vnode
* is obtained with a call to vn_get(). This call takes care of
* coordination with the reclamation of the inode and vnode. Note
* that the vmap structure is filled in while holding the hash lock.
* This gives us the state of the inode/vnode when we found it and
* is used for coordination in vn_get().
*
* If it is not in core, read it in from the file system's device and
* add the inode into the hash table.
*
* The inode is locked according to the value of the lock_flags parameter.
* This flag parameter indicates how and if the inode's IO lock and inode lock
* should be taken.
*
* mp -- the mount point structure for the current file system. It points
* to the inode hash table.
* tp -- a pointer to the current transaction if there is one. This is
* simply passed through to the xfs_iread() call.
* ino -- the number of the inode desired. This is the unique identifier
* within the file system for the inode being requested.
* lock_flags -- flags indicating how to lock the inode. See the comment
* for xfs_ilock() for a list of valid values.
* bno -- the block number starting the buffer containing the inode,
* if known (as by bulkstat), else 0.
*/
int
xfs_iget(
xfs_mount_t *mp,
xfs_trans_t *tp,
xfs_ino_t ino,
uint flags,
uint lock_flags,
xfs_inode_t **ipp,
xfs_daddr_t bno)
{
xfs_ihash_t *ih;
xfs_inode_t *ip;
xfs_inode_t *iq;
xfs_vnode_t *vp;
ulong version;
int error;
/* REFERENCED */
int newnode;
xfs_chash_t *ch;
xfs_chashlist_t *chl, *chlnew;
vmap_t vmap;
SPLDECL(s);
XFS_STATS_INC(xs_ig_attempts);
ih = XFS_IHASH(mp, ino);
again:
read_lock(&ih->ih_lock);
for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) {
if (ip->i_ino == ino) {
vp = XFS_ITOV(ip);
VMAP(vp, vmap);
/*
* Inode cache hit: if ip is not at the front of
* its hash chain, move it there now.
* Do this with the lock held for update, but
* do statistics after releasing the lock.
*/
if (ip->i_prevp != &ih->ih_next
&& rwlock_trypromote(&ih->ih_lock)) {
if ((iq = ip->i_next)) {
iq->i_prevp = ip->i_prevp;
}
*ip->i_prevp = iq;
iq = ih->ih_next;
iq->i_prevp = &ip->i_next;
ip->i_next = iq;
ip->i_prevp = &ih->ih_next;
ih->ih_next = ip;
write_unlock(&ih->ih_lock);
} else {
read_unlock(&ih->ih_lock);
}
XFS_STATS_INC(xs_ig_found);
/*
* Get a reference to the vnode/inode.
* vn_get() takes care of coordination with
* the file system inode release and reclaim
* functions. If it returns NULL, the inode
* has been reclaimed so just start the search
* over again. We probably won't find it,
* but we could be racing with another cpu
* looking for the same inode so we have to at
* least look.
*/
if (!(vp = vn_get(vp, &vmap))) {
XFS_STATS_INC(xs_ig_frecycle);
goto again;
}
if (lock_flags != 0) {
ip->i_flags &= ~XFS_IRECLAIM;
xfs_ilock(ip, lock_flags);
}
newnode = (ip->i_d.di_mode == 0);
if (newnode) {
xfs_iocore_inode_reinit(ip);
}
ip->i_flags &= ~XFS_ISTALE;
vn_trace_exit(vp, "xfs_iget.found",
(inst_t *)__return_address);
goto return_ip;
}
}
/*
* Inode cache miss: save the hash chain version stamp and unlock
* the chain, so we don't deadlock in vn_alloc.
*/
XFS_STATS_INC(xs_ig_missed);
version = ih->ih_version;
read_unlock(&ih->ih_lock);
/*
* Read the disk inode attributes into a new inode structure and get
* a new vnode for it. This should also initialize i_ino and i_mount.
*/
error = xfs_iread(mp, tp, ino, &ip, bno);
if (error) {
return error;
}
error = xfs_vn_allocate(mp, ip, &vp);
if (error) {
return error;
}
vn_trace_exit(vp, "xfs_iget.alloc", (inst_t *)__return_address);
xfs_inode_lock_init(ip, vp);
xfs_iocore_inode_init(ip);
if (lock_flags != 0) {
xfs_ilock(ip, lock_flags);
}
/*
* Put ip on its hash chain, unless someone else hashed a duplicate
* after we released the hash lock.
*/
write_lock(&ih->ih_lock);
if (ih->ih_version != version) {
for (iq = ih->ih_next; iq != NULL; iq = iq->i_next) {
if (iq->i_ino == ino) {
write_unlock(&ih->ih_lock);
xfs_idestroy(ip);
XFS_STATS_INC(xs_ig_dup);
goto again;
}
}
}
/*
* These values _must_ be set before releasing ihlock!
*/
ip->i_hash = ih;
if ((iq = ih->ih_next)) {
iq->i_prevp = &ip->i_next;
}
ip->i_next = iq;
ip->i_prevp = &ih->ih_next;
ih->ih_next = ip;
ip->i_udquot = ip->i_gdquot = NULL;
ih->ih_version++;
write_unlock(&ih->ih_lock);
/*
* put ip on its cluster's hash chain
*/
ASSERT(ip->i_chash == NULL && ip->i_cprev == NULL &&
ip->i_cnext == NULL);
chlnew = NULL;
ch = XFS_CHASH(mp, ip->i_blkno);
chlredo:
s = mutex_spinlock(&ch->ch_lock);
for (chl = ch->ch_list; chl != NULL; chl = chl->chl_next) {
if (chl->chl_blkno == ip->i_blkno) {
/* insert this inode into the doubly-linked list
* where chl points */
if ((iq = chl->chl_ip)) {
ip->i_cprev = iq->i_cprev;
iq->i_cprev->i_cnext = ip;
iq->i_cprev = ip;
ip->i_cnext = iq;
} else {
ip->i_cnext = ip;
ip->i_cprev = ip;
}
chl->chl_ip = ip;
ip->i_chash = chl;
break;
}
}
/* no hash list found for this block; add a new hash list */
if (chl == NULL) {
if (chlnew == NULL) {
mutex_spinunlock(&ch->ch_lock, s);
ASSERT(xfs_chashlist_zone != NULL);
chlnew = (xfs_chashlist_t *)
kmem_zone_alloc(xfs_chashlist_zone,
KM_SLEEP);
ASSERT(chlnew != NULL);
goto chlredo;
} else {
ip->i_cnext = ip;
ip->i_cprev = ip;
ip->i_chash = chlnew;
chlnew->chl_ip = ip;
chlnew->chl_blkno = ip->i_blkno;
chlnew->chl_next = ch->ch_list;
ch->ch_list = chlnew;
chlnew = NULL;
}
} else {
if (chlnew != NULL) {
kmem_zone_free(xfs_chashlist_zone, chlnew);
}
}
mutex_spinunlock(&ch->ch_lock, s);
/*
* Link ip to its mount and thread it on the mount's inode list.
*/
XFS_MOUNT_ILOCK(mp);
if ((iq = mp->m_inodes)) {
ASSERT(iq->i_mprev->i_mnext == iq);
ip->i_mprev = iq->i_mprev;
iq->i_mprev->i_mnext = ip;
iq->i_mprev = ip;
ip->i_mnext = iq;
} else {
ip->i_mnext = ip;
ip->i_mprev = ip;
}
mp->m_inodes = ip;
XFS_MOUNT_IUNLOCK(mp);
newnode = 1;
return_ip:
ASSERT(ip->i_df.if_ext_max ==
XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t));
ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) ==
((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0));
*ipp = ip;
/*
* If we have a real type for an on-disk inode, we can set ops(&unlock)
* now. If it's a new inode being created, xfs_ialloc will handle it.
*/
XVFS_INIT_VNODE(XFS_MTOVFS(mp), vp, XFS_ITOBHV(ip), 1);
return 0;
}
/*
* Allocate a block and fill it with dquots.
* This is called when the bmapi finds a hole.
*/
STATIC int
xfs_qm_dqalloc(
xfs_trans_t *tp,
xfs_mount_t *mp,
xfs_dquot_t *dqp,
xfs_inode_t *quotip,
xfs_fileoff_t offset_fsb,
xfs_buf_t **O_bpp)
{
xfs_fsblock_t firstblock;
xfs_bmap_free_t flist;
xfs_bmbt_irec_t map;
int nmaps, error, committed;
xfs_buf_t *bp;
ASSERT(tp != NULL);
xfs_dqtrace_entry(dqp, "DQALLOC");
/*
* Initialize the bmap freelist prior to calling bmapi code.
*/
XFS_BMAP_INIT(&flist, &firstblock);
xfs_ilock(quotip, XFS_ILOCK_EXCL);
/*
* Return if this type of quotas is turned off while we didn't
* have an inode lock
*/
if (XFS_IS_THIS_QUOTA_OFF(dqp)) {
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
return (ESRCH);
}
/*
* xfs_trans_commit normally decrements the vnode ref count
* when it unlocks the inode. Since we want to keep the quota
* inode around, we bump the vnode ref count now.
*/
VN_HOLD(XFS_ITOV(quotip));
xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL);
nmaps = 1;
if ((error = xfs_bmapi(tp, quotip,
offset_fsb, XFS_DQUOT_CLUSTER_SIZE_FSB,
XFS_BMAPI_METADATA | XFS_BMAPI_WRITE,
&firstblock,
XFS_QM_DQALLOC_SPACE_RES(mp),
&map, &nmaps, &flist))) {
goto error0;
}
ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB);
ASSERT(nmaps == 1);
ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
(map.br_startblock != HOLESTARTBLOCK));
/*
* Keep track of the blkno to save a lookup later
*/
dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
/* now we can just get the buffer (there's nothing to read yet) */
bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
dqp->q_blkno,
XFS_QI_DQCHUNKLEN(mp),
0);
if (!bp || (error = XFS_BUF_GETERROR(bp)))
goto error1;
/*
* Make a chunk of dquots out of this buffer and log
* the entire thing.
*/
xfs_qm_init_dquot_blk(tp, mp, INT_GET(dqp->q_core.d_id, ARCH_CONVERT),
dqp->dq_flags & (XFS_DQ_USER|XFS_DQ_GROUP),
bp);
if ((error = xfs_bmap_finish(&tp, &flist, firstblock, &committed))) {
goto error1;
}
*O_bpp = bp;
return 0;
error1:
xfs_bmap_cancel(&flist);
error0:
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
return (error);
}
/*
* xfs_rename
*/
int
xfs_rename(
bhv_desc_t *src_dir_bdp,
bhv_vname_t *src_vname,
bhv_vnode_t *target_dir_vp,
bhv_vname_t *target_vname,
cred_t *credp)
{
xfs_trans_t *tp;
xfs_inode_t *src_dp, *target_dp, *src_ip, *target_ip;
xfs_mount_t *mp;
int new_parent; /* moving to a new dir */
int src_is_directory; /* src_name is a directory */
int error;
xfs_bmap_free_t free_list;
xfs_fsblock_t first_block;
int cancel_flags;
int committed;
xfs_inode_t *inodes[4];
int target_ip_dropped = 0; /* dropped target_ip link? */
bhv_vnode_t *src_dir_vp;
int spaceres;
int target_link_zero = 0;
int num_inodes;
char *src_name = VNAME(src_vname);
char *target_name = VNAME(target_vname);
int src_namelen = VNAMELEN(src_vname);
int target_namelen = VNAMELEN(target_vname);
src_dir_vp = BHV_TO_VNODE(src_dir_bdp);
vn_trace_entry(src_dir_vp, "xfs_rename", (inst_t *)__return_address);
vn_trace_entry(target_dir_vp, "xfs_rename", (inst_t *)__return_address);
/*
* Find the XFS behavior descriptor for the target directory
* vnode since it was not handed to us.
*/
target_dp = xfs_vtoi(target_dir_vp);
if (target_dp == NULL) {
return XFS_ERROR(EXDEV);
}
src_dp = XFS_BHVTOI(src_dir_bdp);
mp = src_dp->i_mount;
if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_RENAME) ||
DM_EVENT_ENABLED(target_dir_vp->v_vfsp,
target_dp, DM_EVENT_RENAME)) {
error = XFS_SEND_NAMESP(mp, DM_EVENT_RENAME,
src_dir_vp, DM_RIGHT_NULL,
target_dir_vp, DM_RIGHT_NULL,
src_name, target_name,
0, 0, 0);
if (error) {
return error;
}
}
/* Return through std_return after this point. */
/*
* Lock all the participating inodes. Depending upon whether
* the target_name exists in the target directory, and
* whether the target directory is the same as the source
* directory, we can lock from 2 to 4 inodes.
* xfs_lock_for_rename() will return ENOENT if src_name
* does not exist in the source directory.
*/
tp = NULL;
error = xfs_lock_for_rename(src_dp, target_dp, src_vname,
target_vname, &src_ip, &target_ip, inodes,
&num_inodes);
if (error) {
/*
* We have nothing locked, no inode references, and
* no transaction, so just get out.
*/
goto std_return;
}
ASSERT(src_ip != NULL);
if ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
/*
* Check for link count overflow on target_dp
*/
if (target_ip == NULL && (src_dp != target_dp) &&
target_dp->i_d.di_nlink >= XFS_MAXLINK) {
error = XFS_ERROR(EMLINK);
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
goto rele_return;
}
}
/*
* If we are using project inheritance, we only allow renames
* into our tree when the project IDs are the same; else the
* tree quota mechanism would be circumvented.
*/
if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
(target_dp->i_d.di_projid != src_ip->i_d.di_projid))) {
error = XFS_ERROR(EXDEV);
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
goto rele_return;
}
new_parent = (src_dp != target_dp);
src_is_directory = ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR);
/*
* Drop the locks on our inodes so that we can start the transaction.
*/
xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED);
XFS_BMAP_INIT(&free_list, &first_block);
tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME);
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
spaceres = XFS_RENAME_SPACE_RES(mp, target_namelen);
error = xfs_trans_reserve(tp, spaceres, XFS_RENAME_LOG_RES(mp), 0,
XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT);
if (error == ENOSPC) {
spaceres = 0;
error = xfs_trans_reserve(tp, 0, XFS_RENAME_LOG_RES(mp), 0,
XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT);
}
if (error) {
xfs_trans_cancel(tp, 0);
goto rele_return;
}
/*
* Attach the dquots to the inodes
*/
if ((error = XFS_QM_DQVOPRENAME(mp, inodes))) {
xfs_trans_cancel(tp, cancel_flags);
goto rele_return;
}
/*
* Reacquire the inode locks we dropped above.
*/
xfs_lock_inodes(inodes, num_inodes, 0, XFS_ILOCK_EXCL);
/*
* Join all the inodes to the transaction. From this point on,
* we can rely on either trans_commit or trans_cancel to unlock
* them. Note that we need to add a vnode reference to the
* directories since trans_commit & trans_cancel will decrement
* them when they unlock the inodes. Also, we need to be careful
* not to add an inode to the transaction more than once.
*/
VN_HOLD(src_dir_vp);
xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL);
if (new_parent) {
VN_HOLD(target_dir_vp);
xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL);
}
if ((src_ip != src_dp) && (src_ip != target_dp)) {
xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL);
}
if ((target_ip != NULL) &&
(target_ip != src_ip) &&
(target_ip != src_dp) &&
(target_ip != target_dp)) {
xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL);
}
/*
* Set up the target.
*/
if (target_ip == NULL) {
/*
* If there's no space reservation, check the entry will
* fit before actually inserting it.
*/
if (spaceres == 0 &&
(error = xfs_dir_canenter(tp, target_dp, target_name,
target_namelen)))
goto error_return;
/*
* If target does not exist and the rename crosses
* directories, adjust the target directory link count
* to account for the ".." reference from the new entry.
*/
error = xfs_dir_createname(tp, target_dp, target_name,
target_namelen, src_ip->i_ino,
&first_block, &free_list, spaceres);
if (error == ENOSPC)
goto error_return;
if (error)
goto abort_return;
xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
if (new_parent && src_is_directory) {
error = xfs_bumplink(tp, target_dp);
if (error)
goto abort_return;
}
} else { /* target_ip != NULL */
/*
* If target exists and it's a directory, check that both
* target and source are directories and that target can be
* destroyed, or that neither is a directory.
*/
if ((target_ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
/*
* Make sure target dir is empty.
*/
if (!(xfs_dir_isempty(target_ip)) ||
(target_ip->i_d.di_nlink > 2)) {
error = XFS_ERROR(EEXIST);
goto error_return;
}
}
/*
* Link the source inode under the target name.
* If the source inode is a directory and we are moving
* it across directories, its ".." entry will be
* inconsistent until we replace that down below.
*
* In case there is already an entry with the same
* name at the destination directory, remove it first.
*/
error = xfs_dir_replace(tp, target_dp, target_name,
target_namelen, src_ip->i_ino,
&first_block, &free_list, spaceres);
if (error)
goto abort_return;
xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* Decrement the link count on the target since the target
* dir no longer points to it.
*/
error = xfs_droplink(tp, target_ip);
if (error)
goto abort_return;
target_ip_dropped = 1;
if (src_is_directory) {
/*
* Drop the link from the old "." entry.
*/
error = xfs_droplink(tp, target_ip);
if (error)
goto abort_return;
}
/* Do this test while we still hold the locks */
target_link_zero = (target_ip)->i_d.di_nlink==0;
} /* target_ip != NULL */
/*
* Remove the source.
*/
if (new_parent && src_is_directory) {
/*
* Rewrite the ".." entry to point to the new
* directory.
*/
error = xfs_dir_replace(tp, src_ip, "..", 2, target_dp->i_ino,
&first_block, &free_list, spaceres);
ASSERT(error != EEXIST);
if (error)
goto abort_return;
xfs_ichgtime(src_ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
} else {
/*
* We always want to hit the ctime on the source inode.
* We do it in the if clause above for the 'new_parent &&
* src_is_directory' case, and here we get all the other
* cases. This isn't strictly required by the standards
* since the source inode isn't really being changed,
* but old unix file systems did it and some incremental
* backup programs won't work without it.
*/
xfs_ichgtime(src_ip, XFS_ICHGTIME_CHG);
}
/*
* Adjust the link count on src_dp. This is necessary when
* renaming a directory, either within one parent when
* the target existed, or across two parent directories.
*/
if (src_is_directory && (new_parent || target_ip != NULL)) {
/*
* Decrement link count on src_directory since the
* entry that's moved no longer points to it.
*/
error = xfs_droplink(tp, src_dp);
if (error)
goto abort_return;
}
error = xfs_dir_removename(tp, src_dp, src_name, src_namelen,
src_ip->i_ino, &first_block, &free_list, spaceres);
if (error)
goto abort_return;
xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* Update the generation counts on all the directory inodes
* that we're modifying.
*/
src_dp->i_gen++;
xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE);
if (new_parent) {
target_dp->i_gen++;
xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
}
/*
* If there was a target inode, take an extra reference on
* it here so that it doesn't go to xfs_inactive() from
* within the commit.
*/
if (target_ip != NULL) {
IHOLD(target_ip);
}
/*
* If this is a synchronous mount, make sure that the
* rename transaction goes to disk before returning to
* the user.
*/
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
xfs_trans_set_sync(tp);
}
/*
* Take refs. for vop_link_removed calls below. No need to worry
* about directory refs. because the caller holds them.
*
* Do holds before the xfs_bmap_finish since it might rele them down
* to zero.
*/
if (target_ip_dropped)
IHOLD(target_ip);
IHOLD(src_ip);
error = xfs_bmap_finish(&tp, &free_list, &committed);
if (error) {
xfs_bmap_cancel(&free_list);
xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES |
XFS_TRANS_ABORT));
if (target_ip != NULL) {
IRELE(target_ip);
}
if (target_ip_dropped) {
IRELE(target_ip);
}
IRELE(src_ip);
goto std_return;
}
/*
* trans_commit will unlock src_ip, target_ip & decrement
* the vnode references.
*/
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
if (target_ip != NULL) {
xfs_refcache_purge_ip(target_ip);
IRELE(target_ip);
}
/*
* Let interposed file systems know about removed links.
*/
if (target_ip_dropped) {
bhv_vop_link_removed(XFS_ITOV(target_ip), target_dir_vp,
target_link_zero);
IRELE(target_ip);
}
IRELE(src_ip);
/* Fall through to std_return with error = 0 or errno from
* xfs_trans_commit */
std_return:
if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_POSTRENAME) ||
DM_EVENT_ENABLED(target_dir_vp->v_vfsp,
target_dp, DM_EVENT_POSTRENAME)) {
(void) XFS_SEND_NAMESP (mp, DM_EVENT_POSTRENAME,
src_dir_vp, DM_RIGHT_NULL,
target_dir_vp, DM_RIGHT_NULL,
src_name, target_name,
0, error, 0);
}
return error;
abort_return:
cancel_flags |= XFS_TRANS_ABORT;
/* FALLTHROUGH */
error_return:
xfs_bmap_cancel(&free_list);
xfs_trans_cancel(tp, cancel_flags);
goto std_return;
rele_return:
IRELE(src_ip);
if (target_ip != NULL) {
IRELE(target_ip);
}
goto std_return;
}
int
xfs_write_file(xfs_inode_t *xip, struct uio *uio, int ioflag)
{
struct buf *bp;
//struct thread *td;
daddr_t lbn;
off_t osize = 0;
off_t offset= 0;
int blkoffset, error, resid, xfersize;
int fsblocksize;
int seqcount;
xfs_iomap_t iomap;
int maps = 1;
xfs_vnode_t *xvp = XFS_ITOV(xip);
struct vnode *vp = xvp->v_vnode;
xfs_mount_t *mp = (&xip->i_iocore)->io_mount;
seqcount = ioflag >> IO_SEQSHIFT;
memset(&iomap,0,sizeof(xfs_iomap_t));
/*
* Maybe this should be above the vnode op call, but so long as
* file servers have no limits, I don't think it matters.
*/
#if 0
td = uio->uio_td;
if (vp->v_type == VREG && td != NULL) {
PROC_LOCK(td->td_proc);
if (uio->uio_offset + uio->uio_resid >
lim_cur(td->td_proc, RLIMIT_FSIZE)) {
psignal(td->td_proc, SIGXFSZ);
PROC_UNLOCK(td->td_proc);
return (EFBIG);
}
PROC_UNLOCK(td->td_proc);
}
#endif
resid = uio->uio_resid;
offset = uio->uio_offset;
osize = xip->i_d.di_size;
/* xfs bmap wants bytes for both offset and size */
XVOP_BMAP(xvp,
uio->uio_offset,
uio->uio_resid,
BMAPI_WRITE|BMAPI_DIRECT,
&iomap, &maps, error);
if(error) {
printf("XVOP_BMAP failed\n");
goto error;
}
for (error = 0; uio->uio_resid > 0;) {
lbn = XFS_B_TO_FSBT(mp, offset);
blkoffset = XFS_B_FSB_OFFSET(mp, offset);
xfersize = mp->m_sb.sb_blocksize - blkoffset;
fsblocksize = mp->m_sb.sb_blocksize;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
/*
* getblk sets buf by blkno * bo->bo_bsize
* bo_bsize is set from the mnt point fsize
* so we call getblk in the case using fsblocks
* not basic blocks
*/
bp = getblk(vp, lbn, fsblocksize, 0, 0, 0);
if(!bp) {
printf("getblk failed\n");
error = EINVAL;
break;
}
if (!(bp->b_flags & B_CACHE) && fsblocksize > xfersize)
vfs_bio_clrbuf(bp);
if (offset + xfersize > xip->i_d.di_size) {
xip->i_d.di_size = offset + xfersize;
vnode_pager_setsize(vp, offset + fsblocksize);
}
/* move the offset for the next itteration of the loop */
offset += xfersize;
error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
if ((ioflag & IO_VMIO) &&
(LIST_FIRST(&bp->b_dep) == NULL)) /* in ext2fs? */
bp->b_flags |= B_RELBUF;
/* force to full direct for now */
bp->b_flags |= B_DIRECT;
/* and sync ... the delay path is not pushing data out */
ioflag |= IO_SYNC;
if (ioflag & IO_SYNC) {
(void)bwrite(bp);
} else if (0 /* RMC xfersize + blkoffset == fs->s_frag_size */) {
if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
bp->b_flags |= B_CLUSTEROK;
cluster_write(vp, bp, osize, seqcount);
} else {
bawrite(bp);
}
} else {
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
if (error || xfersize == 0)
break;
}
/*
* 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 0
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
ip->i_mode &= ~(ISUID | ISGID);
#endif
if (error) {
if (ioflag & IO_UNIT) {
#if 0
(void)ext2_truncate(vp, osize,
ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
#endif
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
/* Update the vnode here? */
}
error:
return error;
}
vnode_t *
xfs_itov(xfs_inode_t *ip)
{
return XFS_ITOV(ip);
}
/*
* Syssgi interface for swapext
*/
int
xfs_swapext(
xfs_swapext_t __user *sxp)
{
xfs_swapext_t sx;
xfs_inode_t *ip=NULL, *tip=NULL, *ips[2];
xfs_trans_t *tp;
xfs_mount_t *mp;
xfs_bstat_t *sbp;
struct file *fp = NULL, *tfp = NULL;
vnode_t *vp, *tvp;
bhv_desc_t *bdp, *tbdp;
vn_bhv_head_t *bhp, *tbhp;
uint lock_flags=0;
int ilf_fields, tilf_fields;
int error = 0;
xfs_ifork_t tempif, *ifp, *tifp;
__uint64_t tmp;
int aforkblks = 0;
int taforkblks = 0;
int locked = 0;
if (copy_from_user(&sx, sxp, sizeof(sx)))
return XFS_ERROR(EFAULT);
/* Pull information for the target fd */
if (((fp = fget((int)sx.sx_fdtarget)) == NULL) ||
((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
bhp = VN_BHV_HEAD(vp);
bdp = vn_bhv_lookup(bhp, &xfs_vnodeops);
if (bdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
ip = XFS_BHVTOI(bdp);
}
if (((tfp = fget((int)sx.sx_fdtmp)) == NULL) ||
((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
tbhp = VN_BHV_HEAD(tvp);
tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops);
if (tbdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
tip = XFS_BHVTOI(tbdp);
}
if (ip->i_mount != tip->i_mount) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (ip->i_ino == tip->i_ino) {
error = XFS_ERROR(EINVAL);
goto error0;
}
mp = ip->i_mount;
sbp = &sx.sx_stat;
if (XFS_FORCED_SHUTDOWN(mp)) {
error = XFS_ERROR(EIO);
goto error0;
}
locked = 1;
/* Lock in i_ino order */
if (ip->i_ino < tip->i_ino) {
ips[0] = ip;
ips[1] = tip;
} else {
ips[0] = tip;
ips[1] = ip;
}
lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
xfs_lock_inodes(ips, 2, 0, lock_flags);
/* Check permissions */
error = xfs_iaccess(ip, S_IWUSR, NULL);
if (error)
goto error0;
error = xfs_iaccess(tip, S_IWUSR, NULL);
if (error)
goto error0;
/* Verify that both files have the same format */
if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify both files are either real-time or non-realtime */
if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
(tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Should never get a local format */
if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (VN_CACHED(tvp) != 0)
xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore),
(xfs_off_t)0, 0, 0);
/* Verify O_DIRECT for ftmp */
if (VN_CACHED(tvp) != 0) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify all data are being swapped */
if (sx.sx_offset != 0 ||
sx.sx_length != ip->i_d.di_size ||
sx.sx_length != tip->i_d.di_size) {
error = XFS_ERROR(EFAULT);
goto error0;
}
/*
* If the target has extended attributes, the tmp file
* must also in order to ensure the correct data fork
* format.
*/
if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/*
* Compare the current change & modify times with that
* passed in. If they differ, we abort this swap.
* This is the mechanism used to ensure the calling
* process that the file was not changed out from
* under it.
*/
if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
(sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
(sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
(sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
/* We need to fail if the file is memory mapped. Once we have tossed
* all existing pages, the page fault will have no option
* but to go to the filesystem for pages. By making the page fault call
* VOP_READ (or write in the case of autogrow) they block on the iolock
* until we have switched the extents.
*/
if (VN_MAPPED(vp)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_iunlock(tip, XFS_ILOCK_EXCL);
/*
* There is a race condition here since we gave up the
* ilock. However, the data fork will not change since
* we have the iolock (locked for truncation too) so we
* are safe. We don't really care if non-io related
* fields change.
*/
VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
if ((error = xfs_trans_reserve(tp, 0,
XFS_ICHANGE_LOG_RES(mp), 0,
0, 0))) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
xfs_iunlock(tip, XFS_IOLOCK_EXCL);
xfs_trans_cancel(tp, 0);
return error;
}
xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
/*
* Count the number of extended attribute blocks
*/
if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
(tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
&taforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
/*
* Swap the data forks of the inodes
*/
ifp = &ip->i_df;
tifp = &tip->i_df;
tempif = *ifp; /* struct copy */
*ifp = *tifp; /* struct copy */
*tifp = tempif; /* struct copy */
/*
* Fix the on-disk inode values
*/
tmp = (__uint64_t)ip->i_d.di_nblocks;
ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
tmp = (__uint64_t) ip->i_d.di_nextents;
ip->i_d.di_nextents = tip->i_d.di_nextents;
tip->i_d.di_nextents = tmp;
tmp = (__uint64_t) ip->i_d.di_format;
ip->i_d.di_format = tip->i_d.di_format;
tip->i_d.di_format = tmp;
ilf_fields = XFS_ILOG_CORE;
switch(ip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
ifp->if_u1.if_extents =
ifp->if_u2.if_inline_ext;
}
ilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
ilf_fields |= XFS_ILOG_DBROOT;
break;
}
tilf_fields = XFS_ILOG_CORE;
switch(tip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
tifp->if_u1.if_extents =
tifp->if_u2.if_inline_ext;
}
tilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
tilf_fields |= XFS_ILOG_DBROOT;
break;
}
/*
* Increment vnode ref counts since xfs_trans_commit &
* xfs_trans_cancel will both unlock the inodes and
* decrement the associated ref counts.
*/
VN_HOLD(vp);
VN_HOLD(tvp);
xfs_trans_ijoin(tp, ip, lock_flags);
xfs_trans_ijoin(tp, tip, lock_flags);
xfs_trans_log_inode(tp, ip, ilf_fields);
xfs_trans_log_inode(tp, tip, tilf_fields);
/*
* If this is a synchronous mount, make sure that the
* transaction goes to disk before returning to the user.
*/
if (mp->m_flags & XFS_MOUNT_WSYNC) {
xfs_trans_set_sync(tp);
}
error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
fput(fp);
fput(tfp);
return error;
error0:
if (locked) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
}
if (fp != NULL) fput(fp);
if (tfp != NULL) fput(tfp);
return error;
}
/*
* Return quota status information, such as uquota-off, enforcements, etc.
*/
STATIC int
xfs_qm_scall_getqstat(
xfs_mount_t *mp,
fs_quota_stat_t *out)
{
xfs_inode_t *uip, *gip;
boolean_t tempuqip, tempgqip;
uip = gip = NULL;
tempuqip = tempgqip = B_FALSE;
memset(out, 0, sizeof(fs_quota_stat_t));
out->qs_version = FS_QSTAT_VERSION;
if (! XFS_SB_VERSION_HASQUOTA(&mp->m_sb)) {
out->qs_uquota.qfs_ino = NULLFSINO;
out->qs_gquota.qfs_ino = NULLFSINO;
return (0);
}
out->qs_flags = (__uint16_t) xfs_qm_export_flags(mp->m_qflags &
(XFS_ALL_QUOTA_ACCT|
XFS_ALL_QUOTA_ENFD));
out->qs_pad = 0;
out->qs_uquota.qfs_ino = mp->m_sb.sb_uquotino;
out->qs_gquota.qfs_ino = mp->m_sb.sb_gquotino;
if (mp->m_quotainfo) {
uip = mp->m_quotainfo->qi_uquotaip;
gip = mp->m_quotainfo->qi_gquotaip;
}
if (!uip && mp->m_sb.sb_uquotino != NULLFSINO) {
if (xfs_iget(mp, NULL, mp->m_sb.sb_uquotino,
0, 0, &uip, 0) == 0)
tempuqip = B_TRUE;
}
if (!gip && mp->m_sb.sb_gquotino != NULLFSINO) {
if (xfs_iget(mp, NULL, mp->m_sb.sb_gquotino,
0, 0, &gip, 0) == 0)
tempgqip = B_TRUE;
}
if (uip) {
out->qs_uquota.qfs_nblks = uip->i_d.di_nblocks;
out->qs_uquota.qfs_nextents = uip->i_d.di_nextents;
if (tempuqip)
VN_RELE(XFS_ITOV(uip));
}
if (gip) {
out->qs_gquota.qfs_nblks = gip->i_d.di_nblocks;
out->qs_gquota.qfs_nextents = gip->i_d.di_nextents;
if (tempgqip)
VN_RELE(XFS_ITOV(gip));
}
if (mp->m_quotainfo) {
out->qs_incoredqs = XFS_QI_MPLNDQUOTS(mp);
out->qs_btimelimit = XFS_QI_BTIMELIMIT(mp);
out->qs_itimelimit = XFS_QI_ITIMELIMIT(mp);
out->qs_rtbtimelimit = XFS_QI_RTBTIMELIMIT(mp);
out->qs_bwarnlimit = XFS_QI_BWARNLIMIT(mp);
out->qs_iwarnlimit = XFS_QI_IWARNLIMIT(mp);
}
return (0);
}
STATIC int
xfs_unmount(
bhv_desc_t *bdp,
int flags,
cred_t *credp)
{
struct vfs *vfsp = bhvtovfs(bdp);
xfs_mount_t *mp = XFS_BHVTOM(bdp);
xfs_inode_t *rip;
vnode_t *rvp;
int unmount_event_wanted = 0;
int unmount_event_flags = 0;
int xfs_unmountfs_needed = 0;
int error;
rip = mp->m_rootip;
rvp = XFS_ITOV(rip);
if (vfsp->vfs_flag & VFS_DMI) {
error = XFS_SEND_PREUNMOUNT(mp, vfsp,
rvp, DM_RIGHT_NULL, rvp, DM_RIGHT_NULL,
NULL, NULL, 0, 0,
(mp->m_dmevmask & (1<<DM_EVENT_PREUNMOUNT))?
0:DM_FLAGS_UNWANTED);
if (error)
return XFS_ERROR(error);
unmount_event_wanted = 1;
unmount_event_flags = (mp->m_dmevmask & (1<<DM_EVENT_UNMOUNT))?
0 : DM_FLAGS_UNWANTED;
}
/*
* First blow any referenced inode from this file system
* out of the reference cache, and delete the timer.
*/
xfs_refcache_purge_mp(mp);
XFS_bflush(mp->m_ddev_targp);
error = xfs_unmount_flush(mp, 0);
if (error)
goto out;
ASSERT(vn_count(rvp) == 1);
/*
* Drop the reference count
*/
VN_RELE(rvp);
/*
* If we're forcing a shutdown, typically because of a media error,
* we want to make sure we invalidate dirty pages that belong to
* referenced vnodes as well.
*/
if (XFS_FORCED_SHUTDOWN(mp)) {
error = xfs_sync(&mp->m_bhv,
(SYNC_WAIT | SYNC_CLOSE), credp);
ASSERT(error != EFSCORRUPTED);
}
xfs_unmountfs_needed = 1;
out:
/* Send DMAPI event, if required.
* Then do xfs_unmountfs() if needed.
* Then return error (or zero).
*/
if (unmount_event_wanted) {
/* Note: mp structure must still exist for
* XFS_SEND_UNMOUNT() call.
*/
XFS_SEND_UNMOUNT(mp, vfsp, error == 0 ? rvp : NULL,
DM_RIGHT_NULL, 0, error, unmount_event_flags);
}
if (xfs_unmountfs_needed) {
/*
* Call common unmount function to flush to disk
* and free the super block buffer & mount structures.
*/
xfs_unmountfs(mp, credp);
}
return XFS_ERROR(error);
}
