/*
* This gets called when the inode's version needs to be changed from 1 to 2.
* Currently this happens when the nlink field overflows the old 16-bit value
* or when chproj is called to change the project for the first time.
* As a side effect the superblock version will also get rev'd
* to contain the NLINK bit.
*/
void
xfs_bump_ino_vers2(
xfs_trans_t *tp,
xfs_inode_t *ip)
{
xfs_mount_t *mp;
unsigned long s;
ASSERT(ismrlocked (&ip->i_lock, MR_UPDATE));
ASSERT(ip->i_d.di_version == XFS_DINODE_VERSION_1);
ip->i_d.di_version = XFS_DINODE_VERSION_2;
ip->i_d.di_onlink = 0;
memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
mp = tp->t_mountp;
if (!XFS_SB_VERSION_HASNLINK(&mp->m_sb)) {
s = XFS_SB_LOCK(mp);
if (!XFS_SB_VERSION_HASNLINK(&mp->m_sb)) {
XFS_SB_VERSION_ADDNLINK(&mp->m_sb);
XFS_SB_UNLOCK(mp, s);
xfs_mod_sb(tp, XFS_SB_VERSIONNUM);
} else {
XFS_SB_UNLOCK(mp, s);
}
}
/* Caller must log the inode */
}
/*
* This is called to mark the fields indicated in fieldmask as needing
* to be logged when the transaction is committed. The inode must
* already be associated with the given transaction.
*
* The values for fieldmask are defined in xfs_inode_item.h. We always
* log all of the core inode if any of it has changed, and we always log
* all of the inline data/extents/b-tree root if any of them has changed.
*/
void
xfs_trans_log_inode(
xfs_trans_t *tp,
xfs_inode_t *ip,
uint flags)
{
xfs_log_item_desc_t *lidp;
ASSERT(ip->i_transp == tp);
ASSERT(ip->i_itemp != NULL);
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE));
lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)(ip->i_itemp));
ASSERT(lidp != NULL);
tp->t_flags |= XFS_TRANS_DIRTY;
lidp->lid_flags |= XFS_LID_DIRTY;
/*
* Always OR in the bits from the ili_last_fields field.
* This is to coordinate with the xfs_iflush() and xfs_iflush_done()
* routines in the eventual clearing of the ilf_fields bits.
* See the big comment in xfs_iflush() for an explanation of
* this coorination mechanism.
*/
flags |= ip->i_itemp->ili_last_fields;
ip->i_itemp->ili_format.ilf_fields |= flags;
}
/*
* This is called to pin the inode associated with the inode log
* item in memory so it cannot be written out. Do this by calling
* xfs_ipin() to bump the pin count in the inode while holding the
* inode pin lock.
*/
STATIC void
xfs_inode_item_pin(
xfs_inode_log_item_t *iip)
{
ASSERT(ismrlocked(&(iip->ili_inode->i_lock), MR_UPDATE));
xfs_ipin(iip->ili_inode);
}
/*
* This is called to asynchronously write the inode associated with this
* inode log item out to disk. The inode will already have been locked by
* a successful call to xfs_inode_item_trylock().
*/
STATIC void
xfs_inode_item_push(
xfs_inode_log_item_t *iip)
{
xfs_inode_t *ip;
ip = iip->ili_inode;
ASSERT(ismrlocked(&(ip->i_lock), MR_ACCESS));
ASSERT(valusema(&(ip->i_flock)) <= 0);
/*
* Since we were able to lock the inode's flush lock and
* we found it on the AIL, the inode must be dirty. This
* is because the inode is removed from the AIL while still
* holding the flush lock in xfs_iflush_done(). Thus, if
* we found it in the AIL and were able to obtain the flush
* lock without sleeping, then there must not have been
* anyone in the process of flushing the inode.
*/
ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) ||
iip->ili_format.ilf_fields != 0);
/*
* Write out the inode. The completion routine ('iflush_done') will
* pull it from the AIL, mark it clean, unlock the flush lock.
*/
(void) xfs_iflush(ip, XFS_IFLUSH_ASYNC);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}
/*
* This gets called when the inode's version needs to be changed from 1 to 2.
* Currently this happens when the nlink field overflows the old 16-bit value
* or when chproj is called to change the project for the first time.
* As a side effect the superblock version will also get rev'd
* to contain the NLINK bit.
*/
void
xfs_bump_ino_vers2(
xfs_trans_t *tp,
xfs_inode_t *ip)
{
xfs_mount_t *mp;
ASSERT(ismrlocked (&ip->i_lock, MR_UPDATE));
ASSERT(ip->i_d.di_version == XFS_DINODE_VERSION_1);
ip->i_d.di_version = XFS_DINODE_VERSION_2;
ip->i_d.di_onlink = 0;
memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
mp = tp->t_mountp;
if (!xfs_sb_version_hasnlink(&mp->m_sb)) {
spin_lock(&mp->m_sb_lock);
if (!xfs_sb_version_hasnlink(&mp->m_sb)) {
xfs_sb_version_addnlink(&mp->m_sb);
spin_unlock(&mp->m_sb_lock);
xfs_mod_sb(tp, XFS_SB_VERSIONNUM);
} else {
spin_unlock(&mp->m_sb_lock);
}
}
/* Caller must log the inode */
}
/*
* This routine is called to handle zeroing any space in the last
* block of the file that is beyond the EOF. We do this since the
* size is being increased without writing anything to that block
* and we don't want anyone to read the garbage on the disk.
*/
STATIC int /* error (positive) */
xfs_zero_last_block(
xfs_vnode_t *vp,
xfs_iocore_t *io,
xfs_fsize_t isize,
xfs_fsize_t end_size)
{
xfs_fileoff_t last_fsb;
xfs_mount_t *mp;
int nimaps;
int zero_offset;
int zero_len;
int error = 0;
xfs_bmbt_irec_t imap;
xfs_off_t loff;
ASSERT(ismrlocked(io->io_lock, MR_UPDATE) != 0);
mp = io->io_mount;
zero_offset = XFS_B_FSB_OFFSET(mp, isize);
if (zero_offset == 0) {
/*
* There are no extra bytes in the last block on disk to
* zero, so return.
*/
return 0;
}
last_fsb = XFS_B_TO_FSBT(mp, isize);
nimaps = 1;
error = XFS_BMAPI(mp, NULL, io, last_fsb, 1, 0, NULL, 0, &imap,
&nimaps, NULL, NULL);
if (error) {
return error;
}
ASSERT(nimaps > 0);
/*
* If the block underlying isize is just a hole, then there
* is nothing to zero.
*/
if (imap.br_startblock == HOLESTARTBLOCK) {
return 0;
}
/*
* Zero the part of the last block beyond the EOF, and write it
* out sync. We need to drop the ilock while we do this so we
* don't deadlock when the buffer cache calls back to us.
*/
XFS_IUNLOCK(mp, io, XFS_ILOCK_EXCL| XFS_EXTSIZE_RD);
loff = XFS_FSB_TO_B(mp, last_fsb);
zero_len = mp->m_sb.sb_blocksize - zero_offset;
error = xfs_iozero(vp, loff + zero_offset, zero_len, end_size);
XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
ASSERT(error >= 0);
return error;
}
/*ARGSUSED*/
void
xfs_trans_ihold(
xfs_trans_t *tp,
xfs_inode_t *ip)
{
ASSERT(ip->i_transp == tp);
ASSERT(ip->i_itemp != NULL);
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE));
ip->i_itemp->ili_flags |= XFS_ILI_HOLD;
}
/*
* Add the locked inode to the transaction.
* The inode must be locked, and it cannot be associated with any
* transaction. The caller must specify the locks already held
* on the inode.
*/
void
xfs_trans_ijoin(
xfs_trans_t *tp,
xfs_inode_t *ip,
uint lock_flags)
{
xfs_inode_log_item_t *iip;
ASSERT(ip->i_transp == NULL);
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE));
ASSERT(lock_flags & XFS_ILOCK_EXCL);
if (ip->i_itemp == NULL)
xfs_inode_item_init(ip, ip->i_mount);
iip = ip->i_itemp;
ASSERT(iip->ili_flags == 0);
ASSERT(iip->ili_ilock_recur == 0);
ASSERT(iip->ili_iolock_recur == 0);
/*
* Get a log_item_desc to point at the new item.
*/
(void) xfs_trans_add_item(tp, (xfs_log_item_t*)(iip));
xfs_trans_inode_broot_debug(ip);
/*
* If the IO lock is already held, mark that in the inode log item.
*/
if (lock_flags & XFS_IOLOCK_EXCL) {
iip->ili_flags |= XFS_ILI_IOLOCKED_EXCL;
} else if (lock_flags & XFS_IOLOCK_SHARED) {
iip->ili_flags |= XFS_ILI_IOLOCKED_SHARED;
}
/*
* Initialize i_transp so we can find it with xfs_inode_incore()
* in xfs_trans_iget() above.
*/
ip->i_transp = tp;
}
int /* error (positive) */
xfs_zero_eof(
vnode_t *vp,
xfs_iocore_t *io,
xfs_off_t offset, /* starting I/O offset */
xfs_fsize_t isize, /* current inode size */
xfs_fsize_t end_size) /* terminal inode size */
{
struct inode *ip = LINVFS_GET_IP(vp);
xfs_fileoff_t start_zero_fsb;
xfs_fileoff_t end_zero_fsb;
xfs_fileoff_t prev_zero_fsb;
xfs_fileoff_t zero_count_fsb;
xfs_fileoff_t last_fsb;
xfs_extlen_t buf_len_fsb;
xfs_extlen_t prev_zero_count;
xfs_mount_t *mp;
int nimaps;
int error = 0;
xfs_bmbt_irec_t imap;
loff_t loff;
size_t lsize;
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
mp = io->io_mount;
/*
* First handle zeroing the block on which isize resides.
* We only zero a part of that block so it is handled specially.
*/
error = xfs_zero_last_block(ip, io, offset, isize, end_size);
if (error) {
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
return error;
}
/*
* Calculate the range between the new size and the old
* where blocks needing to be zeroed may exist. To get the
* block where the last byte in the file currently resides,
* we need to subtract one from the size and truncate back
* to a block boundary. We subtract 1 in case the size is
* exactly on a block boundary.
*/
last_fsb = isize ? XFS_B_TO_FSBT(mp, isize - 1) : (xfs_fileoff_t)-1;
start_zero_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)isize);
end_zero_fsb = XFS_B_TO_FSBT(mp, offset - 1);
ASSERT((xfs_sfiloff_t)last_fsb < (xfs_sfiloff_t)start_zero_fsb);
if (last_fsb == end_zero_fsb) {
/*
* The size was only incremented on its last block.
* We took care of that above, so just return.
*/
return 0;
}
ASSERT(start_zero_fsb <= end_zero_fsb);
prev_zero_fsb = NULLFILEOFF;
prev_zero_count = 0;
while (start_zero_fsb <= end_zero_fsb) {
nimaps = 1;
zero_count_fsb = end_zero_fsb - start_zero_fsb + 1;
error = XFS_BMAPI(mp, NULL, io, start_zero_fsb, zero_count_fsb,
0, NULL, 0, &imap, &nimaps, NULL);
if (error) {
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
return error;
}
ASSERT(nimaps > 0);
if (imap.br_state == XFS_EXT_UNWRITTEN ||
imap.br_startblock == HOLESTARTBLOCK) {
/*
* This loop handles initializing pages that were
* partially initialized by the code below this
* loop. It basically zeroes the part of the page
* that sits on a hole and sets the page as P_HOLE
* and calls remapf if it is a mapped file.
*/
prev_zero_fsb = NULLFILEOFF;
prev_zero_count = 0;
start_zero_fsb = imap.br_startoff +
imap.br_blockcount;
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
continue;
}
/*
* There are blocks in the range requested.
* Zero them a single write at a time. We actually
* don't zero the entire range returned if it is
* too big and simply loop around to get the rest.
* That is not the most efficient thing to do, but it
* is simple and this path should not be exercised often.
*/
buf_len_fsb = XFS_FILBLKS_MIN(imap.br_blockcount,
mp->m_writeio_blocks << 8);
/*
* Drop the inode lock while we're doing the I/O.
* We'll still have the iolock to protect us.
*/
XFS_IUNLOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
loff = XFS_FSB_TO_B(mp, start_zero_fsb);
lsize = XFS_FSB_TO_B(mp, buf_len_fsb);
error = xfs_iozero(ip, loff, lsize, end_size);
if (error) {
goto out_lock;
}
prev_zero_fsb = start_zero_fsb;
prev_zero_count = buf_len_fsb;
start_zero_fsb = imap.br_startoff + buf_len_fsb;
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
}
return 0;
out_lock:
XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
ASSERT(error >= 0);
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 *));
}
}
/*
* Get and lock the inode for the caller if it is not already
* locked within the given transaction. If it is already locked
* within the transaction, just increment its lock recursion count
* and return a pointer to it.
*
* For an inode to be locked in a transaction, the inode lock, as
* opposed to the io lock, must be taken exclusively. This ensures
* that the inode can be involved in only 1 transaction at a time.
* Lock recursion is handled on the io lock, but only for lock modes
* of equal or lesser strength. That is, you can recur on the io lock
* held EXCL with a SHARED request but not vice versa. Also, if
* the inode is already a part of the transaction then you cannot
* go from not holding the io lock to having it EXCL or SHARED.
*
* Use the inode cache routine xfs_inode_incore() to find the inode
* if it is already owned by this transaction.
*
* If we don't already own the inode, use xfs_iget() to get it.
* Since the inode log item structure is embedded in the incore
* inode structure and is initialized when the inode is brought
* into memory, there is nothing to do with it here.
*
* If the given transaction pointer is NULL, just call xfs_iget().
* This simplifies code which must handle both cases.
*/
int
xfs_trans_iget(
xfs_mount_t *mp,
xfs_trans_t *tp,
xfs_ino_t ino,
uint lock_flags,
xfs_inode_t **ipp)
{
int error;
xfs_inode_t *ip;
xfs_inode_log_item_t *iip;
/*
* If the transaction pointer is NULL, just call the normal
* xfs_iget().
*/
if (tp == NULL) {
return (xfs_iget(mp, NULL, ino, lock_flags, ipp, 0));
}
/*
* If we find the inode in core with this transaction
* pointer in its i_transp field, then we know we already
* have it locked. In this case we just increment the lock
* recursion count and return the inode to the caller.
* Assert that the inode is already locked in the mode requested
* by the caller. We cannot do lock promotions yet, so
* die if someone gets this wrong.
*/
if ((ip = xfs_inode_incore(tp->t_mountp, ino, tp)) != NULL) {
/*
* Make sure that the inode lock is held EXCL and
* that the io lock is never upgraded when the inode
* is already a part of the transaction.
*/
ASSERT(ip->i_itemp != NULL);
ASSERT(lock_flags & XFS_ILOCK_EXCL);
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE));
ASSERT((!(lock_flags & XFS_IOLOCK_EXCL)) ||
ismrlocked(&ip->i_iolock, MR_UPDATE));
ASSERT((!(lock_flags & XFS_IOLOCK_EXCL)) ||
(ip->i_itemp->ili_flags & XFS_ILI_IOLOCKED_EXCL));
ASSERT((!(lock_flags & XFS_IOLOCK_SHARED)) ||
ismrlocked(&ip->i_iolock, (MR_UPDATE | MR_ACCESS)));
ASSERT((!(lock_flags & XFS_IOLOCK_SHARED)) ||
(ip->i_itemp->ili_flags & XFS_ILI_IOLOCKED_ANY));
if (lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) {
ip->i_itemp->ili_iolock_recur++;
}
if (lock_flags & XFS_ILOCK_EXCL) {
ip->i_itemp->ili_ilock_recur++;
}
*ipp = ip;
return 0;
}
ASSERT(lock_flags & XFS_ILOCK_EXCL);
error = xfs_iget(tp->t_mountp, tp, ino, lock_flags, &ip, 0);
if (error) {
return error;
}
ASSERT(ip != NULL);
/*
* Get a log_item_desc to point at the new item.
*/
if (ip->i_itemp == NULL)
xfs_inode_item_init(ip, mp);
iip = ip->i_itemp;
(void) xfs_trans_add_item(tp, (xfs_log_item_t *)(iip));
xfs_trans_inode_broot_debug(ip);
/*
* If the IO lock has been acquired, mark that in
* the inode log item so we'll know to unlock it
* when the transaction commits.
*/
ASSERT(iip->ili_flags == 0);
if (lock_flags & XFS_IOLOCK_EXCL) {
iip->ili_flags |= XFS_ILI_IOLOCKED_EXCL;
} else if (lock_flags & XFS_IOLOCK_SHARED) {
iip->ili_flags |= XFS_ILI_IOLOCKED_SHARED;
}
/*
* Initialize i_transp so we can find it with xfs_inode_incore()
* above.
*/
ip->i_transp = tp;
*ipp = ip;
return 0;
}
/*
* This gets called by xfs_trans_push_ail(), when IOP_TRYLOCK
* failed to get the inode flush lock but did get the inode locked SHARED.
* Here we're trying to see if the inode buffer is incore, and if so whether it's
* marked delayed write. If that's the case, we'll initiate a bawrite on that
* buffer to expedite the process.
*
* We aren't holding the AIL_LOCK (or the flush lock) when this gets called,
* so it is inherently race-y.
*/
STATIC void
xfs_inode_item_pushbuf(
xfs_inode_log_item_t *iip)
{
xfs_inode_t *ip;
xfs_mount_t *mp;
xfs_buf_t *bp;
uint dopush;
ip = iip->ili_inode;
ASSERT(ismrlocked(&(ip->i_lock), MR_ACCESS));
/*
* The ili_pushbuf_flag keeps others from
* trying to duplicate our effort.
*/
ASSERT(iip->ili_pushbuf_flag != 0);
ASSERT(iip->ili_push_owner == get_thread_id());
/*
* If flushlock isn't locked anymore, chances are that the
* inode flush completed and the inode was taken off the AIL.
* So, just get out.
*/
if ((valusema(&(ip->i_flock)) > 0) ||
((iip->ili_item.li_flags & XFS_LI_IN_AIL) == 0)) {
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}
mp = ip->i_mount;
bp = xfs_incore(mp->m_ddev_targp, iip->ili_format.ilf_blkno,
iip->ili_format.ilf_len, XFS_INCORE_TRYLOCK);
if (bp != NULL) {
if (XFS_BUF_ISDELAYWRITE(bp)) {
/*
* We were racing with iflush because we don't hold
* the AIL_LOCK or the flush lock. However, at this point,
* we have the buffer, and we know that it's dirty.
* So, it's possible that iflush raced with us, and
* this item is already taken off the AIL.
* If not, we can flush it async.
*/
dopush = ((iip->ili_item.li_flags & XFS_LI_IN_AIL) &&
(valusema(&(ip->i_flock)) <= 0));
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
xfs_buftrace("INODE ITEM PUSH", bp);
if (XFS_BUF_ISPINNED(bp)) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE);
}
if (dopush) {
xfs_bawrite(mp, bp);
} else {
xfs_buf_relse(bp);
}
} else {
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
xfs_buf_relse(bp);
}
return;
}
/*
* We have to be careful about resetting pushbuf flag too early (above).
* Even though in theory we can do it as soon as we have the buflock,
* we don't want others to be doing work needlessly. They'll come to
* this function thinking that pushing the buffer is their
* responsibility only to find that the buffer is still locked by
* another doing the same thing
*/
iip->ili_pushbuf_flag = 0;
xfs_iunlock(ip, XFS_ILOCK_SHARED);
return;
}
/*
* Unlock the inode associated with the inode log item.
* Clear the fields of the inode and inode log item that
* are specific to the current transaction. If the
* hold flags is set, do not unlock the inode.
*/
STATIC void
xfs_inode_item_unlock(
xfs_inode_log_item_t *iip)
{
uint hold;
uint iolocked;
uint lock_flags;
xfs_inode_t *ip;
ASSERT(iip != NULL);
ASSERT(iip->ili_inode->i_itemp != NULL);
ASSERT(ismrlocked(&(iip->ili_inode->i_lock), MR_UPDATE));
ASSERT((!(iip->ili_inode->i_itemp->ili_flags &
XFS_ILI_IOLOCKED_EXCL)) ||
ismrlocked(&(iip->ili_inode->i_iolock), MR_UPDATE));
ASSERT((!(iip->ili_inode->i_itemp->ili_flags &
XFS_ILI_IOLOCKED_SHARED)) ||
ismrlocked(&(iip->ili_inode->i_iolock), MR_ACCESS));
/*
* Clear the transaction pointer in the inode.
*/
ip = iip->ili_inode;
ip->i_transp = NULL;
/*
* If the inode needed a separate buffer with which to log
* its extents, then free it now.
*/
if (iip->ili_extents_buf != NULL) {
ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS);
ASSERT(ip->i_d.di_nextents > 0);
ASSERT(iip->ili_format.ilf_fields & XFS_ILOG_DEXT);
ASSERT(ip->i_df.if_bytes > 0);
kmem_free(iip->ili_extents_buf, ip->i_df.if_bytes);
iip->ili_extents_buf = NULL;
}
if (iip->ili_aextents_buf != NULL) {
ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS);
ASSERT(ip->i_d.di_anextents > 0);
ASSERT(iip->ili_format.ilf_fields & XFS_ILOG_AEXT);
ASSERT(ip->i_afp->if_bytes > 0);
kmem_free(iip->ili_aextents_buf, ip->i_afp->if_bytes);
iip->ili_aextents_buf = NULL;
}
/*
* Figure out if we should unlock the inode or not.
*/
hold = iip->ili_flags & XFS_ILI_HOLD;
/*
* Before clearing out the flags, remember whether we
* are holding the inode's IO lock.
*/
iolocked = iip->ili_flags & XFS_ILI_IOLOCKED_ANY;
/*
* Clear out the fields of the inode log item particular
* to the current transaction.
*/
iip->ili_ilock_recur = 0;
iip->ili_iolock_recur = 0;
iip->ili_flags = 0;
/*
* Unlock the inode if XFS_ILI_HOLD was not set.
*/
if (!hold) {
lock_flags = XFS_ILOCK_EXCL;
if (iolocked & XFS_ILI_IOLOCKED_EXCL) {
lock_flags |= XFS_IOLOCK_EXCL;
} else if (iolocked & XFS_ILI_IOLOCKED_SHARED) {
lock_flags |= XFS_IOLOCK_SHARED;
}
xfs_iput(iip->ili_inode, lock_flags);
}
}
int
xfs_iomap_write_delay(
xfs_inode_t *ip,
xfs_off_t offset,
size_t count,
int ioflag,
xfs_bmbt_irec_t *ret_imap,
int *nmaps)
{
xfs_mount_t *mp = ip->i_mount;
xfs_iocore_t *io = &ip->i_iocore;
xfs_fileoff_t offset_fsb;
xfs_fileoff_t last_fsb;
xfs_off_t aligned_offset;
xfs_fileoff_t ioalign;
xfs_fsblock_t firstblock;
xfs_extlen_t extsz;
xfs_fsize_t isize;
int nimaps;
xfs_bmbt_irec_t imap[XFS_WRITE_IMAPS];
int prealloc, fsynced = 0;
int error;
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE) != 0);
/*
* Make sure that the dquots are there. This doesn't hold
* the ilock across a disk read.
*/
error = XFS_QM_DQATTACH(mp, ip, XFS_QMOPT_ILOCKED);
if (error)
return XFS_ERROR(error);
if (XFS_IS_REALTIME_INODE(ip)) {
if (!(extsz = ip->i_d.di_extsize))
extsz = mp->m_sb.sb_rextsize;
} else {
extsz = ip->i_d.di_extsize;
}
offset_fsb = XFS_B_TO_FSBT(mp, offset);
retry:
isize = ip->i_d.di_size;
if (io->io_new_size > isize)
isize = io->io_new_size;
error = xfs_iomap_eof_want_preallocate(mp, io, isize, offset, count,
ioflag, imap, XFS_WRITE_IMAPS, &prealloc);
if (error)
return error;
if (prealloc) {
aligned_offset = XFS_WRITEIO_ALIGN(mp, (offset + count - 1));
ioalign = XFS_B_TO_FSBT(mp, aligned_offset);
last_fsb = ioalign + mp->m_writeio_blocks;
} else {
last_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)(offset + count)));
}
if (prealloc || extsz) {
error = xfs_iomap_eof_align_last_fsb(mp, io, isize, extsz,
&last_fsb);
if (error)
return error;
}
nimaps = XFS_WRITE_IMAPS;
firstblock = NULLFSBLOCK;
error = XFS_BMAPI(mp, NULL, io, offset_fsb,
(xfs_filblks_t)(last_fsb - offset_fsb),
XFS_BMAPI_DELAY | XFS_BMAPI_WRITE |
XFS_BMAPI_ENTIRE, &firstblock, 1, imap,
&nimaps, NULL, NULL);
if (error && (error != ENOSPC))
return XFS_ERROR(error);
/*
* If bmapi returned us nothing, and if we didn't get back EDQUOT,
* then we must have run out of space - flush delalloc, and retry..
*/
if (nimaps == 0) {
xfs_iomap_enter_trace(XFS_IOMAP_WRITE_NOSPACE,
io, offset, count);
if (xfs_flush_space(ip, &fsynced, &ioflag))
return XFS_ERROR(ENOSPC);
error = 0;
goto retry;
}
if (unlikely(!imap[0].br_startblock && !(io->io_flags & XFS_IOCORE_RT)))
return xfs_cmn_err_fsblock_zero(ip, &imap[0]);
*ret_imap = imap[0];
*nmaps = 1;
return 0;
}
int /* error (positive) */
xfs_zero_eof(
bhv_vnode_t *vp,
xfs_iocore_t *io,
xfs_off_t offset, /* starting I/O offset */
xfs_fsize_t isize) /* current inode size */
{
struct inode *ip = vn_to_inode(vp);
xfs_fileoff_t start_zero_fsb;
xfs_fileoff_t end_zero_fsb;
xfs_fileoff_t zero_count_fsb;
xfs_fileoff_t last_fsb;
xfs_fileoff_t zero_off;
xfs_fsize_t zero_len;
xfs_mount_t *mp = io->io_mount;
int nimaps;
int error = 0;
xfs_bmbt_irec_t imap;
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
ASSERT(offset > isize);
/*
* First handle zeroing the block on which isize resides.
* We only zero a part of that block so it is handled specially.
*/
error = xfs_zero_last_block(ip, io, offset, isize);
if (error) {
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
return error;
}
/*
* Calculate the range between the new size and the old
* where blocks needing to be zeroed may exist. To get the
* block where the last byte in the file currently resides,
* we need to subtract one from the size and truncate back
* to a block boundary. We subtract 1 in case the size is
* exactly on a block boundary.
*/
last_fsb = isize ? XFS_B_TO_FSBT(mp, isize - 1) : (xfs_fileoff_t)-1;
start_zero_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)isize);
end_zero_fsb = XFS_B_TO_FSBT(mp, offset - 1);
ASSERT((xfs_sfiloff_t)last_fsb < (xfs_sfiloff_t)start_zero_fsb);
if (last_fsb == end_zero_fsb) {
/*
* The size was only incremented on its last block.
* We took care of that above, so just return.
*/
return 0;
}
ASSERT(start_zero_fsb <= end_zero_fsb);
while (start_zero_fsb <= end_zero_fsb) {
nimaps = 1;
zero_count_fsb = end_zero_fsb - start_zero_fsb + 1;
error = XFS_BMAPI(mp, NULL, io, start_zero_fsb, zero_count_fsb,
0, NULL, 0, &imap, &nimaps, NULL, NULL);
if (error) {
ASSERT(ismrlocked(io->io_lock, MR_UPDATE));
ASSERT(ismrlocked(io->io_iolock, MR_UPDATE));
return error;
}
ASSERT(nimaps > 0);
if (imap.br_state == XFS_EXT_UNWRITTEN ||
imap.br_startblock == HOLESTARTBLOCK) {
/*
* This loop handles initializing pages that were
* partially initialized by the code below this
* loop. It basically zeroes the part of the page
* that sits on a hole and sets the page as P_HOLE
* and calls remapf if it is a mapped file.
*/
start_zero_fsb = imap.br_startoff + imap.br_blockcount;
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
continue;
}
/*
* There are blocks we need to zero.
* Drop the inode lock while we're doing the I/O.
* We'll still have the iolock to protect us.
*/
XFS_IUNLOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
zero_off = XFS_FSB_TO_B(mp, start_zero_fsb);
zero_len = XFS_FSB_TO_B(mp, imap.br_blockcount);
if ((zero_off + zero_len) > offset)
zero_len = offset - zero_off;
error = xfs_iozero(ip, zero_off, zero_len);
if (error) {
goto out_lock;
}
start_zero_fsb = imap.br_startoff + imap.br_blockcount;
ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
}
return 0;
out_lock:
XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
ASSERT(error >= 0);
return error;
}
int
xfs_iomap_write_delay(
xfs_inode_t *ip,
xfs_off_t offset,
size_t count,
int ioflag,
xfs_bmbt_irec_t *ret_imap,
int *nmaps)
{
xfs_mount_t *mp = ip->i_mount;
xfs_iocore_t *io = &ip->i_iocore;
xfs_fileoff_t offset_fsb;
xfs_fileoff_t last_fsb;
xfs_fsize_t isize;
xfs_fsblock_t firstblock;
int nimaps;
int error;
xfs_bmbt_irec_t imap[XFS_WRITE_IMAPS];
int aeof;
int fsynced = 0;
ASSERT(ismrlocked(&ip->i_lock, MR_UPDATE) != 0);
/*
* Make sure that the dquots are there. This doesn't hold
* the ilock across a disk read.
*/
error = XFS_QM_DQATTACH(mp, ip, XFS_QMOPT_ILOCKED);
if (error)
return XFS_ERROR(error);
retry:
isize = ip->i_d.di_size;
if (io->io_new_size > isize) {
isize = io->io_new_size;
}
aeof = 0;
offset_fsb = XFS_B_TO_FSBT(mp, offset);
last_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)(offset + count)));
/*
* If the caller is doing a write at the end of the file,
* then extend the allocation (and the buffer used for the write)
* out to the file system's write iosize. We clean up any extra
* space left over when the file is closed in xfs_inactive().
*
* For sync writes, we are flushing delayed allocate space to
* try to make additional space available for allocation near
* the filesystem full boundary - preallocation hurts in that
* situation, of course.
*/
if (!(ioflag & BMAPI_SYNC) && ((offset + count) > ip->i_d.di_size)) {
xfs_off_t aligned_offset;
xfs_filblks_t count_fsb;
unsigned int iosize;
xfs_fileoff_t ioalign;
int n;
xfs_fileoff_t start_fsb;
/*
* If there are any real blocks past eof, then don't
* do any speculative allocation.
*/
start_fsb = XFS_B_TO_FSBT(mp,
((xfs_ufsize_t)(offset + count - 1)));
count_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
while (count_fsb > 0) {
nimaps = XFS_WRITE_IMAPS;
error = XFS_BMAPI(mp, NULL, io, start_fsb, count_fsb,
0, &firstblock, 0, imap, &nimaps, NULL);
if (error) {
return error;
}
for (n = 0; n < nimaps; n++) {
if ( !(io->io_flags & XFS_IOCORE_RT) &&
!imap[n].br_startblock) {
cmn_err(CE_PANIC,"Access to block "
"zero: fs <%s> inode: %lld "
"start_block : %llx start_off "
": %llx blkcnt : %llx "
"extent-state : %x \n",
(ip->i_mount)->m_fsname,
(long long)ip->i_ino,
imap[n].br_startblock,
imap[n].br_startoff,
imap[n].br_blockcount,
imap[n].br_state);
}
if ((imap[n].br_startblock != HOLESTARTBLOCK) &&
(imap[n].br_startblock != DELAYSTARTBLOCK)) {
goto write_map;
}
start_fsb += imap[n].br_blockcount;
count_fsb -= imap[n].br_blockcount;
}
}
iosize = mp->m_writeio_blocks;
aligned_offset = XFS_WRITEIO_ALIGN(mp, (offset + count - 1));
ioalign = XFS_B_TO_FSBT(mp, aligned_offset);
last_fsb = ioalign + iosize;
aeof = 1;
}
write_map:
nimaps = XFS_WRITE_IMAPS;
firstblock = NULLFSBLOCK;
/*
* If mounted with the "-o swalloc" option, roundup the allocation
* request to a stripe width boundary if the file size is >=
* stripe width and we are allocating past the allocation eof.
*/
if (!(io->io_flags & XFS_IOCORE_RT) && mp->m_swidth
&& (mp->m_flags & XFS_MOUNT_SWALLOC)
&& (isize >= XFS_FSB_TO_B(mp, mp->m_swidth)) && aeof) {
int eof;
xfs_fileoff_t new_last_fsb;
new_last_fsb = roundup_64(last_fsb, mp->m_swidth);
error = xfs_bmap_eof(ip, new_last_fsb, XFS_DATA_FORK, &eof);
if (error) {
return error;
}
if (eof) {
last_fsb = new_last_fsb;
}
/*
* Roundup the allocation request to a stripe unit (m_dalign) boundary
* if the file size is >= stripe unit size, and we are allocating past
* the allocation eof.
*/
} else if (!(io->io_flags & XFS_IOCORE_RT) && mp->m_dalign &&
(isize >= XFS_FSB_TO_B(mp, mp->m_dalign)) && aeof) {
int eof;
xfs_fileoff_t new_last_fsb;
new_last_fsb = roundup_64(last_fsb, mp->m_dalign);
error = xfs_bmap_eof(ip, new_last_fsb, XFS_DATA_FORK, &eof);
if (error) {
return error;
}
if (eof) {
last_fsb = new_last_fsb;
}
/*
* Round up the allocation request to a real-time extent boundary
* if the file is on the real-time subvolume.
*/
} else if (io->io_flags & XFS_IOCORE_RT && aeof) {
int eof;
xfs_fileoff_t new_last_fsb;
new_last_fsb = roundup_64(last_fsb, mp->m_sb.sb_rextsize);
error = XFS_BMAP_EOF(mp, io, new_last_fsb, XFS_DATA_FORK, &eof);
if (error) {
return error;
}
if (eof)
last_fsb = new_last_fsb;
}
error = xfs_bmapi(NULL, ip, offset_fsb,
(xfs_filblks_t)(last_fsb - offset_fsb),
XFS_BMAPI_DELAY | XFS_BMAPI_WRITE |
XFS_BMAPI_ENTIRE, &firstblock, 1, imap,
&nimaps, NULL);
/*
* This can be EDQUOT, if nimaps == 0
*/
if (error && (error != ENOSPC)) {
return XFS_ERROR(error);
}
/*
* If bmapi returned us nothing, and if we didn't get back EDQUOT,
* then we must have run out of space.
*/
if (nimaps == 0) {
xfs_iomap_enter_trace(XFS_IOMAP_WRITE_NOSPACE,
io, offset, count);
if (xfs_flush_space(ip, &fsynced, &ioflag))
return XFS_ERROR(ENOSPC);
error = 0;
goto retry;
}
*ret_imap = imap[0];
*nmaps = 1;
if ( !(io->io_flags & XFS_IOCORE_RT) && !ret_imap->br_startblock) {
cmn_err(CE_PANIC,"Access to block zero: fs <%s> inode: %lld "
"start_block : %llx start_off : %llx blkcnt : %llx "
"extent-state : %x \n",
(ip->i_mount)->m_fsname,
(long long)ip->i_ino,
ret_imap->br_startblock, ret_imap->br_startoff,
ret_imap->br_blockcount,ret_imap->br_state);
}
return 0;
}
