/*
* Read in the in-core dquot's on-disk metadata and return the buffer.
* Returns ENOENT to signal a hole.
*/
STATIC int
xfs_dquot_disk_read(
struct xfs_mount *mp,
struct xfs_dquot *dqp,
struct xfs_buf **bpp)
{
struct xfs_bmbt_irec map;
struct xfs_buf *bp;
struct xfs_inode *quotip = xfs_quota_inode(mp, dqp->dq_flags);
uint lock_mode;
int nmaps = 1;
int error;
lock_mode = xfs_ilock_data_map_shared(quotip);
if (!xfs_this_quota_on(mp, dqp->dq_flags)) {
/*
* Return if this type of quotas is turned off while we
* didn't have the quota inode lock.
*/
xfs_iunlock(quotip, lock_mode);
return -ESRCH;
}
/*
* Find the block map; no allocations yet
*/
error = xfs_bmapi_read(quotip, dqp->q_fileoffset,
XFS_DQUOT_CLUSTER_SIZE_FSB, &map, &nmaps, 0);
xfs_iunlock(quotip, lock_mode);
if (error)
return error;
ASSERT(nmaps == 1);
ASSERT(map.br_blockcount >= 1);
ASSERT(map.br_startblock != DELAYSTARTBLOCK);
if (map.br_startblock == HOLESTARTBLOCK)
return -ENOENT;
trace_xfs_dqtobp_read(dqp);
/*
* store the blkno etc so that we don't have to do the
* mapping all the time
*/
dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen, 0, &bp,
&xfs_dquot_buf_ops);
if (error) {
ASSERT(bp == NULL);
return error;
}
ASSERT(xfs_buf_islocked(bp));
xfs_buf_set_ref(bp, XFS_DQUOT_REF);
*bpp = bp;
return 0;
}
/*
* Read the disk inode attributes into the in-core inode structure.
*
* For version 5 superblocks, if we are initialising a new inode and we are not
* utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
* inode core with a random generation number. If we are keeping inodes around,
* we need to read the inode cluster to get the existing generation number off
* disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
* format) then log recovery is dependent on the di_flushiter field being
* initialised from the current on-disk value and hence we must also read the
* inode off disk.
*/
int
xfs_iread(
xfs_mount_t *mp,
xfs_trans_t *tp,
xfs_inode_t *ip,
uint iget_flags)
{
xfs_buf_t *bp;
xfs_dinode_t *dip;
xfs_failaddr_t fa;
int error;
/*
* Fill in the location information in the in-core inode.
*/
error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
if (error)
return error;
/* shortcut IO on inode allocation if possible */
if ((iget_flags & XFS_IGET_CREATE) &&
xfs_sb_version_hascrc(&mp->m_sb) &&
!(mp->m_flags & XFS_MOUNT_IKEEP)) {
/* initialise the on-disk inode core */
memset(&ip->i_d, 0, sizeof(ip->i_d));
VFS_I(ip)->i_generation = prandom_u32();
ip->i_d.di_version = 3;
return 0;
}
/*
* Get pointers to the on-disk inode and the buffer containing it.
*/
error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
if (error)
return error;
/* even unallocated inodes are verified */
fa = xfs_dinode_verify(mp, ip->i_ino, dip);
if (fa) {
xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
sizeof(*dip), fa);
error = -EFSCORRUPTED;
goto out_brelse;
}
/*
* If the on-disk inode is already linked to a directory
* entry, copy all of the inode into the in-core inode.
* xfs_iformat_fork() handles copying in the inode format
* specific information.
* Otherwise, just get the truly permanent information.
*/
if (dip->di_mode) {
xfs_inode_from_disk(ip, dip);
error = xfs_iformat_fork(ip, dip);
if (error) {
#ifdef DEBUG
xfs_alert(mp, "%s: xfs_iformat() returned error %d",
__func__, error);
#endif /* DEBUG */
goto out_brelse;
}
} else {
/*
* Partial initialisation of the in-core inode. Just the bits
* that xfs_ialloc won't overwrite or relies on being correct.
*/
ip->i_d.di_version = dip->di_version;
VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
/*
* Make sure to pull in the mode here as well in
* case the inode is released without being used.
* This ensures that xfs_inactive() will see that
* the inode is already free and not try to mess
* with the uninitialized part of it.
*/
VFS_I(ip)->i_mode = 0;
}
ASSERT(ip->i_d.di_version >= 2);
ip->i_delayed_blks = 0;
/*
* Mark the buffer containing the inode as something to keep
* around for a while. This helps to keep recently accessed
* meta-data in-core longer.
*/
xfs_buf_set_ref(bp, XFS_INO_REF);
/*
* Use xfs_trans_brelse() to release the buffer containing the on-disk
* inode, because it was acquired with xfs_trans_read_buf() in
* xfs_imap_to_bp() above. If tp is NULL, this is just a normal
* brelse(). If we're within a transaction, then xfs_trans_brelse()
* will only release the buffer if it is not dirty within the
* transaction. It will be OK to release the buffer in this case,
* because inodes on disk are never destroyed and we will be locking the
* new in-core inode before putting it in the cache where other
* processes can find it. Thus we don't have to worry about the inode
* being changed just because we released the buffer.
*/
out_brelse:
xfs_trans_brelse(tp, bp);
return error;
}
/*
* Read the disk inode attributes into the in-core inode structure.
*
* For version 5 superblocks, if we are initialising a new inode and we are not
* utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
* inode core with a random generation number. If we are keeping inodes around,
* we need to read the inode cluster to get the existing generation number off
* disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
* format) then log recovery is dependent on the di_flushiter field being
* initialised from the current on-disk value and hence we must also read the
* inode off disk.
*/
int
xfs_iread(
xfs_mount_t *mp,
xfs_trans_t *tp,
xfs_inode_t *ip,
uint iget_flags)
{
xfs_buf_t *bp;
xfs_dinode_t *dip;
int error;
/*
* Fill in the location information in the in-core inode.
*/
error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
if (error)
return error;
/* shortcut IO on inode allocation if possible */
if ((iget_flags & XFS_IGET_CREATE) &&
xfs_sb_version_hascrc(&mp->m_sb) &&
!(mp->m_flags & XFS_MOUNT_IKEEP)) {
/* initialise the on-disk inode core */
memset(&ip->i_d, 0, sizeof(ip->i_d));
ip->i_d.di_magic = XFS_DINODE_MAGIC;
ip->i_d.di_gen = prandom_u32();
if (xfs_sb_version_hascrc(&mp->m_sb)) {
ip->i_d.di_version = 3;
ip->i_d.di_ino = ip->i_ino;
uuid_copy(&ip->i_d.di_uuid, &mp->m_sb.sb_uuid);
} else
ip->i_d.di_version = 2;
return 0;
}
/*
* Get pointers to the on-disk inode and the buffer containing it.
*/
error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
if (error)
return error;
/* even unallocated inodes are verified */
if (!xfs_dinode_verify(mp, ip, dip)) {
xfs_alert(mp, "%s: validation failed for inode %lld failed",
__func__, ip->i_ino);
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip);
error = -EFSCORRUPTED;
goto out_brelse;
}
/*
* If the on-disk inode is already linked to a directory
* entry, copy all of the inode into the in-core inode.
* xfs_iformat_fork() handles copying in the inode format
* specific information.
* Otherwise, just get the truly permanent information.
*/
if (dip->di_mode) {
xfs_dinode_from_disk(&ip->i_d, dip);
error = xfs_iformat_fork(ip, dip);
if (error) {
#ifdef DEBUG
xfs_alert(mp, "%s: xfs_iformat() returned error %d",
__func__, error);
#endif /* DEBUG */
goto out_brelse;
}
} else {
/*
* Partial initialisation of the in-core inode. Just the bits
* that xfs_ialloc won't overwrite or relies on being correct.
*/
ip->i_d.di_magic = be16_to_cpu(dip->di_magic);
ip->i_d.di_version = dip->di_version;
ip->i_d.di_gen = be32_to_cpu(dip->di_gen);
ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
if (dip->di_version == 3) {
ip->i_d.di_ino = be64_to_cpu(dip->di_ino);
uuid_copy(&ip->i_d.di_uuid, &dip->di_uuid);
}
/*
* Make sure to pull in the mode here as well in
* case the inode is released without being used.
* This ensures that xfs_inactive() will see that
* the inode is already free and not try to mess
* with the uninitialized part of it.
*/
ip->i_d.di_mode = 0;
}
/*
* Automatically convert version 1 inode formats in memory to version 2
* inode format. If the inode is modified, it will get logged and
* rewritten as a version 2 inode. We can do this because we set the
* superblock feature bit for v2 inodes unconditionally during mount
* and it means the reast of the code can assume the inode version is 2
* or higher.
*/
if (ip->i_d.di_version == 1) {
ip->i_d.di_version = 2;
memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
ip->i_d.di_nlink = ip->i_d.di_onlink;
ip->i_d.di_onlink = 0;
xfs_set_projid(ip, 0);
}
ip->i_delayed_blks = 0;
/*
* Mark the buffer containing the inode as something to keep
* around for a while. This helps to keep recently accessed
* meta-data in-core longer.
*/
xfs_buf_set_ref(bp, XFS_INO_REF);
/*
* Use xfs_trans_brelse() to release the buffer containing the on-disk
* inode, because it was acquired with xfs_trans_read_buf() in
* xfs_imap_to_bp() above. If tp is NULL, this is just a normal
* brelse(). If we're within a transaction, then xfs_trans_brelse()
* will only release the buffer if it is not dirty within the
* transaction. It will be OK to release the buffer in this case,
* because inodes on disk are never destroyed and we will be locking the
* new in-core inode before putting it in the cache where other
* processes can find it. Thus we don't have to worry about the inode
* being changed just because we released the buffer.
*/
out_brelse:
xfs_trans_brelse(tp, bp);
return error;
}
/*
* Read in the ondisk dquot using dqtobp() then copy it to an incore version,
* and release the buffer immediately.
*
* If XFS_QMOPT_DQALLOC is set, allocate a dquot on disk if it needed.
*/
int
xfs_qm_dqread(
struct xfs_mount *mp,
xfs_dqid_t id,
uint type,
uint flags,
struct xfs_dquot **O_dqpp)
{
struct xfs_dquot *dqp;
struct xfs_disk_dquot *ddqp;
struct xfs_buf *bp;
struct xfs_trans *tp = NULL;
int error;
int cancelflags = 0;
dqp = kmem_zone_zalloc(xfs_qm_dqzone, KM_SLEEP);
dqp->dq_flags = type;
dqp->q_core.d_id = cpu_to_be32(id);
dqp->q_mount = mp;
INIT_LIST_HEAD(&dqp->q_lru);
mutex_init(&dqp->q_qlock);
init_waitqueue_head(&dqp->q_pinwait);
/*
* Because we want to use a counting completion, complete
* the flush completion once to allow a single access to
* the flush completion without blocking.
*/
init_completion(&dqp->q_flush);
complete(&dqp->q_flush);
/*
* Make sure group quotas have a different lock class than user
* quotas.
*/
switch (type) {
case XFS_DQ_USER:
/* uses the default lock class */
break;
case XFS_DQ_GROUP:
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_group_class);
break;
case XFS_DQ_PROJ:
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_project_class);
break;
default:
ASSERT(0);
break;
}
XFS_STATS_INC(xs_qm_dquot);
trace_xfs_dqread(dqp);
if (flags & XFS_QMOPT_DQALLOC) {
tp = xfs_trans_alloc(mp, XFS_TRANS_QM_DQALLOC);
error = xfs_trans_reserve(tp, &M_RES(mp)->tr_qm_dqalloc,
XFS_QM_DQALLOC_SPACE_RES(mp), 0);
if (error)
goto error1;
cancelflags = XFS_TRANS_RELEASE_LOG_RES;
}
/*
* get a pointer to the on-disk dquot and the buffer containing it
* dqp already knows its own type (GROUP/USER).
*/
error = xfs_qm_dqtobp(&tp, dqp, &ddqp, &bp, flags);
if (error) {
/*
* This can happen if quotas got turned off (ESRCH),
* or if the dquot didn't exist on disk and we ask to
* allocate (ENOENT).
*/
trace_xfs_dqread_fail(dqp);
cancelflags |= XFS_TRANS_ABORT;
goto error1;
}
/* copy everything from disk dquot to the incore dquot */
memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
xfs_qm_dquot_logitem_init(dqp);
/*
* Reservation counters are defined as reservation plus current usage
* to avoid having to add every time.
*/
dqp->q_res_bcount = be64_to_cpu(ddqp->d_bcount);
dqp->q_res_icount = be64_to_cpu(ddqp->d_icount);
dqp->q_res_rtbcount = be64_to_cpu(ddqp->d_rtbcount);
/* initialize the dquot speculative prealloc thresholds */
xfs_dquot_set_prealloc_limits(dqp);
/* Mark the buf so that this will stay incore a little longer */
xfs_buf_set_ref(bp, XFS_DQUOT_REF);
/*
* We got the buffer with a xfs_trans_read_buf() (in dqtobp())
* So we need to release with xfs_trans_brelse().
* The strategy here is identical to that of inodes; we lock
* the dquot in xfs_qm_dqget() before making it accessible to
* others. This is because dquots, like inodes, need a good level of
* concurrency, and we don't want to take locks on the entire buffers
* for dquot accesses.
* Note also that the dquot buffer may even be dirty at this point, if
* this particular dquot was repaired. We still aren't afraid to
* brelse it because we have the changes incore.
*/
ASSERT(xfs_buf_islocked(bp));
xfs_trans_brelse(tp, bp);
if (tp) {
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
if (error)
goto error0;
}
*O_dqpp = dqp;
return error;
error1:
if (tp)
xfs_trans_cancel(tp, cancelflags);
error0:
xfs_qm_dqdestroy(dqp);
*O_dqpp = NULL;
return error;
}
/*
* Ensure that the given in-core dquot has a buffer on disk backing it, and
* return the buffer locked and held. This is called when the bmapi finds a
* hole.
*/
STATIC int
xfs_dquot_disk_alloc(
struct xfs_trans **tpp,
struct xfs_dquot *dqp,
struct xfs_buf **bpp)
{
struct xfs_bmbt_irec map;
struct xfs_trans *tp = *tpp;
struct xfs_mount *mp = tp->t_mountp;
struct xfs_buf *bp;
struct xfs_inode *quotip = xfs_quota_inode(mp, dqp->dq_flags);
int nmaps = 1;
int error;
trace_xfs_dqalloc(dqp);
xfs_ilock(quotip, XFS_ILOCK_EXCL);
if (!xfs_this_quota_on(dqp->q_mount, dqp->dq_flags)) {
/*
* Return if this type of quotas is turned off while we didn't
* have an inode lock
*/
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
return -ESRCH;
}
/* Create the block mapping. */
xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL);
error = xfs_bmapi_write(tp, quotip, dqp->q_fileoffset,
XFS_DQUOT_CLUSTER_SIZE_FSB, XFS_BMAPI_METADATA,
XFS_QM_DQALLOC_SPACE_RES(mp), &map, &nmaps);
if (error)
return error;
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,
mp->m_quotainfo->qi_dqchunklen, 0);
if (!bp)
return -ENOMEM;
bp->b_ops = &xfs_dquot_buf_ops;
/*
* 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_buf_set_ref(bp, XFS_DQUOT_REF);
/*
* Hold the buffer and join it to the dfops so that we'll still own
* the buffer when we return to the caller. The buffer disposal on
* error must be paid attention to very carefully, as it has been
* broken since commit efa092f3d4c6 "[XFS] Fixes a bug in the quota
* code when allocating a new dquot record" in 2005, and the later
* conversion to xfs_defer_ops in commit 310a75a3c6c747 failed to keep
* the buffer locked across the _defer_finish call. We can now do
* this correctly with xfs_defer_bjoin.
*
* Above, we allocated a disk block for the dquot information and used
* get_buf to initialize the dquot. If the _defer_finish fails, the old
* transaction is gone but the new buffer is not joined or held to any
* transaction, so we must _buf_relse it.
*
* If everything succeeds, the caller of this function is returned a
* buffer that is locked and held to the transaction. The caller
* is responsible for unlocking any buffer passed back, either
* manually or by committing the transaction. On error, the buffer is
* released and not passed back.
*/
xfs_trans_bhold(tp, bp);
error = xfs_defer_finish(tpp);
if (error) {
xfs_trans_bhold_release(*tpp, bp);
xfs_trans_brelse(*tpp, bp);
return error;
}
*bpp = bp;
return 0;
}