static int
xfs_dir_shortform_validate_ondisk(xfs_mount_t *mp, xfs_dinode_t *dp)
{
xfs_ino_t ino;
int namelen_sum;
int count;
xfs_dir_shortform_t *sf;
xfs_dir_sf_entry_t *sfe;
int i;
if ((INT_GET(dp->di_core.di_mode, ARCH_CONVERT) & IFMT) != IFDIR) {
return 0;
}
if (INT_GET(dp->di_core.di_format, ARCH_CONVERT) != XFS_DINODE_FMT_LOCAL) {
return 0;
}
if (INT_GET(dp->di_core.di_size, ARCH_CONVERT) < sizeof(sf->hdr)) {
xfs_fs_cmn_err(CE_WARN, mp, "Invalid shortform size: dp 0x%p\n",
dp);
return 1;
}
sf = (xfs_dir_shortform_t *)(&dp->di_u.di_dirsf);
ino = XFS_GET_DIR_INO_ARCH(mp, sf->hdr.parent, ARCH_CONVERT);
if (xfs_dir_ino_validate(mp, ino))
return 1;
count = sf->hdr.count;
if ((count < 0) || ((count * 10) > XFS_LITINO(mp))) {
xfs_fs_cmn_err(CE_WARN, mp,
"Invalid shortform count: dp 0x%p\n", dp);
return(1);
}
if (count == 0) {
return 0;
}
namelen_sum = 0;
sfe = &sf->list[0];
for (i = sf->hdr.count - 1; i >= 0; i--) {
ino = XFS_GET_DIR_INO_ARCH(mp, sfe->inumber, ARCH_CONVERT);
xfs_dir_ino_validate(mp, ino);
if (sfe->namelen >= XFS_LITINO(mp)) {
xfs_fs_cmn_err(CE_WARN, mp,
"Invalid shortform namelen: dp 0x%p\n", dp);
return 1;
}
namelen_sum += sfe->namelen;
sfe = XFS_DIR_SF_NEXTENTRY(sfe);
}
if (namelen_sum >= XFS_LITINO(mp)) {
xfs_fs_cmn_err(CE_WARN, mp,
"Invalid shortform namelen: dp 0x%p\n", dp);
return 1;
}
return 0;
}
/* ARGSUSED */
STATIC int
xfs_qm_dqread(
xfs_trans_t *tp,
xfs_dqid_t id,
xfs_dquot_t *dqp, /* dquot to get filled in */
uint flags)
{
xfs_disk_dquot_t *ddqp;
xfs_buf_t *bp;
int error;
/*
* get a pointer to the on-disk dquot and the buffer containing it
* dqp already knows its own type (GROUP/USER).
*/
xfs_dqtrace_entry(dqp, "DQREAD");
if ((error = xfs_qm_dqtobp(tp, dqp, &ddqp, &bp, flags))) {
return (error);
}
/* copy everything from disk dquot to the incore dquot */
memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
ASSERT(INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id);
xfs_qm_dquot_logitem_init(dqp);
/*
* Reservation counters are defined as reservation plus current usage
* to avoid having to add everytime.
*/
dqp->q_res_bcount = INT_GET(ddqp->d_bcount, ARCH_CONVERT);
dqp->q_res_icount = INT_GET(ddqp->d_icount, ARCH_CONVERT);
dqp->q_res_rtbcount = INT_GET(ddqp->d_rtbcount, ARCH_CONVERT);
/* Mark the buf so that this will stay incore a little longer */
XFS_BUF_SET_VTYPE_REF(bp, B_FS_DQUOT, 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_ISBUSY(bp));
ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
xfs_trans_brelse(tp, bp);
return (error);
}
static int
bnobt_key_count(
void *obj,
int startoff)
{
xfs_alloc_block_t *block;
ASSERT(startoff == 0);
block = obj;
if (INT_GET(block->bb_level, ARCH_CONVERT) == 0)
return 0;
return INT_GET(block->bb_numrecs, ARCH_CONVERT);
}
/*ARGSUSED*/
static int
dir_leaf_namelist_count(
void *obj,
int startoff)
{
xfs_dir_leafblock_t *block;
ASSERT(startoff == 0);
block = obj;
if (INT_GET(block->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC)
return 0;
return INT_GET(block->hdr.count, ARCH_CONVERT);
}
/*ARGSUSED*/
static int
dir_node_btree_count(
void *obj,
int startoff)
{
xfs_da_intnode_t *block;
ASSERT(startoff == 0); /* this is a base structure */
block = obj;
if (INT_GET(block->hdr.info.magic, ARCH_CONVERT) != XFS_DA_NODE_MAGIC)
return 0;
return INT_GET(block->hdr.count, ARCH_CONVERT);
}
/*
* uuid_getnodeuniq - obtain the node unique fields of a UUID.
*
* This is not in any way a standard or condoned UUID function;
* it just something that's needed for user-level file handles.
*/
void
uuid_getnodeuniq(uuid_t *uuid, int fsid [2])
{
char *uu=(char*)uuid;
/* on IRIX, this function assumes big-endian fields within
* the uuid, so we use INT_GET to get the same result on
* little-endian systems
*/
fsid[0] = (INT_GET(*(u_int16_t*)(uu+8), ARCH_CONVERT) << 16) +
INT_GET(*(u_int16_t*)(uu+4), ARCH_CONVERT);
fsid[1] = INT_GET(*(u_int32_t*)(uu ), ARCH_CONVERT);
}
STATIC int
xfs_dqtest_cmp2(
xfs_dqtest_t *d,
xfs_dquot_t *dqp)
{
int err = 0;
if (INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) != d->d_icount) {
xfs_qm_dqtest_failed(d, dqp, "icount mismatch",
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT),
d->d_icount, 0);
err++;
}
if (INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT) != d->d_bcount) {
xfs_qm_dqtest_failed(d, dqp, "bcount mismatch",
INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT),
d->d_bcount, 0);
err++;
}
if (INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT)) {
if (INT_ISZERO(dqp->q_core.d_btimer, ARCH_CONVERT) &&
!INT_ISZERO(dqp->q_core.d_id, ARCH_CONVERT)) {
cmn_err(CE_DEBUG,
"%d [%s] [0x%p] BLK TIMER NOT STARTED",
d->d_id, DQFLAGTO_TYPESTR(d), d->q_mount);
err++;
}
}
if (INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT)) {
if (INT_ISZERO(dqp->q_core.d_itimer, ARCH_CONVERT) &&
!INT_ISZERO(dqp->q_core.d_id, ARCH_CONVERT)) {
cmn_err(CE_DEBUG,
"%d [%s] [0x%p] INO TIMER NOT STARTED",
d->d_id, DQFLAGTO_TYPESTR(d), d->q_mount);
err++;
}
}
#ifdef QUOTADEBUG
if (!err) {
cmn_err(CE_DEBUG, "%d [%s] [0x%p] qchecked",
d->d_id, XFS_QM_ISUDQ(d) ? "USR" : "GRP", d->q_mount);
}
#endif
return (err);
}
/*
* Look up a name in a leaf directory structure, replace the inode number.
* This is the external routine.
*/
STATIC int
xfs_dir_leaf_replace(xfs_da_args_t *args)
{
int index, retval;
xfs_dabuf_t *bp;
xfs_ino_t inum;
xfs_dir_leafblock_t *leaf;
xfs_dir_leaf_entry_t *entry;
xfs_dir_leaf_name_t *namest;
inum = args->inumber;
retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
XFS_DATA_FORK);
if (retval)
return(retval);
ASSERT(bp != NULL);
retval = xfs_dir_leaf_lookup_int(bp, args, &index);
if (retval == EEXIST) {
leaf = bp->data;
entry = &leaf->entries[index];
namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
/* XXX - replace assert? */
XFS_DIR_SF_PUT_DIRINO_ARCH(&inum, &namest->inumber, ARCH_CONVERT);
xfs_da_log_buf(args->trans, bp,
XFS_DA_LOGRANGE(leaf, namest, sizeof(namest->inumber)));
xfs_da_buf_done(bp);
retval = 0;
} else
xfs_da_brelse(args->trans, bp);
return(retval);
}
/* ARGSUSED */
void
xfs_dqtrace_entry__(
xfs_dquot_t *dqp,
char *func,
void *retaddr,
xfs_inode_t *ip)
{
xfs_dquot_t *udqp = NULL;
int ino;
ASSERT(dqp->q_trace);
if (ip) {
ino = ip->i_ino;
udqp = ip->i_udquot;
}
ktrace_enter(dqp->q_trace,
(void *)(__psint_t)DQUOT_KTRACE_ENTRY,
(void *)func,
(void *)(__psint_t)dqp->q_nrefs,
(void *)(__psint_t)dqp->dq_flags,
(void *)(__psint_t)dqp->q_res_bcount,
(void *)(__psint_t)INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_icount, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT),
(void *)(__psint_t)INT_GET(dqp->q_core.d_id, ARCH_CONVERT), /* 11 */
(void *)(__psint_t)current_pid(),
(void *)(__psint_t)ino,
(void *)(__psint_t)retaddr,
(void *)(__psint_t)udqp);
return;
}
/*
* Look up a filename in an int directory, replace the inode number.
* Use an internal routine to actually do the lookup.
*/
STATIC int
xfs_dir_node_replace(xfs_da_args_t *args)
{
xfs_da_state_t *state;
xfs_da_state_blk_t *blk;
xfs_dir_leafblock_t *leaf;
xfs_dir_leaf_entry_t *entry;
xfs_dir_leaf_name_t *namest;
xfs_ino_t inum;
int retval, error, i;
xfs_dabuf_t *bp;
state = xfs_da_state_alloc();
state->args = args;
state->mp = args->dp->i_mount;
state->blocksize = state->mp->m_sb.sb_blocksize;
inum = args->inumber;
/*
* Search to see if name exists,
* and get back a pointer to it.
*/
error = xfs_da_node_lookup_int(state, &retval);
if (error) {
retval = error;
}
if (retval == EEXIST) {
blk = &state->path.blk[state->path.active - 1];
ASSERT(blk->magic == XFS_DIR_LEAF_MAGIC);
bp = blk->bp;
leaf = bp->data;
entry = &leaf->entries[blk->index];
namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
/* XXX - replace assert ? */
XFS_DIR_SF_PUT_DIRINO_ARCH(&inum, &namest->inumber, ARCH_CONVERT);
xfs_da_log_buf(args->trans, bp,
XFS_DA_LOGRANGE(leaf, namest, sizeof(namest->inumber)));
xfs_da_buf_done(bp);
blk->bp = NULL;
retval = 0;
} else {
i = state->path.active - 1;
xfs_da_brelse(args->trans, state->path.blk[i].bp);
state->path.blk[i].bp = NULL;
}
for (i = 0; i < state->path.active - 1; i++) {
xfs_da_brelse(args->trans, state->path.blk[i].bp);
state->path.blk[i].bp = NULL;
}
xfs_da_state_free(state);
return(retval);
}
/*ARGSUSED*/
static int
dir_leaf_hdr_count(
void *obj,
int startoff)
{
xfs_dir_leafblock_t *block;
ASSERT(startoff == 0);
block = obj;
return INT_GET(block->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC;
}
STATIC void
xfs_qm_dqtest_failed(
xfs_dqtest_t *d,
xfs_dquot_t *dqp,
char *reason,
xfs_qcnt_t a,
xfs_qcnt_t b,
int error)
{
qmtest_nfails++;
if (error)
cmn_err(CE_DEBUG, "quotacheck failed id=%d, err=%d\nreason: %s",
INT_GET(d->d_id, ARCH_CONVERT), error, reason);
else
cmn_err(CE_DEBUG, "quotacheck failed id=%d (%s) [%d != %d]",
INT_GET(d->d_id, ARCH_CONVERT), reason, (int)a, (int)b);
xfs_qm_dqtest_print(d);
if (dqp)
xfs_qm_dqprint(dqp);
}
/*ARGSUSED*/
static int
dir_node_hdr_count(
void *obj,
int startoff)
{
xfs_da_intnode_t *block;
ASSERT(startoff == 0);
block = obj;
return INT_GET(block->hdr.info.magic, ARCH_CONVERT) == XFS_DA_NODE_MAGIC;
}
static int
dir_leaf_name_count(
void *obj,
int startoff)
{
xfs_dir_leafblock_t *block;
xfs_dir_leaf_entry_t *e;
int i;
int off;
ASSERT(bitoffs(startoff) == 0);
off = byteize(startoff);
block = obj;
if (INT_GET(block->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC)
return 0;
for (i = 0; i < INT_GET(block->hdr.count, ARCH_CONVERT); i++) {
e = &block->entries[i];
if (INT_GET(e->nameidx, ARCH_CONVERT) == off)
return e->namelen;
}
return 0;
}
void
xfs_dqlock2(
xfs_dquot_t *d1,
xfs_dquot_t *d2)
{
if (d1 && d2) {
ASSERT(d1 != d2);
if (INT_GET(d1->q_core.d_id, ARCH_CONVERT) > INT_GET(d2->q_core.d_id, ARCH_CONVERT)) {
xfs_dqlock(d2);
xfs_dqlock(d1);
} else {
xfs_dqlock(d1);
xfs_dqlock(d2);
}
} else {
if (d1) {
xfs_dqlock(d1);
} else if (d2) {
xfs_dqlock(d2);
}
}
}
/* global_hdr_checksum_check - check the global media file header checksum.
* utility function for use by drive-specific strategies.
* returns BOOL_TRUE if ok, BOOL_FALSE if bad
*/
bool_t
global_hdr_checksum_check( global_hdr_t *hdrp )
{
u_int32_t *beginp = ( u_int32_t * )&hdrp[ 0 ];
u_int32_t *endp = ( u_int32_t * )&hdrp[ 1 ];
u_int32_t *p;
u_int32_t accum;
accum = 0;
for ( p = beginp ; p < endp ; p++ ) {
accum += INT_GET(*p, ARCH_CONVERT);
}
return accum == 0 ? BOOL_TRUE : BOOL_FALSE;
}
/*ARGSUSED*/
static int
dir_leaf_namelist_offset(
void *obj,
int startoff,
int idx)
{
xfs_dir_leafblock_t *block;
xfs_dir_leaf_entry_t *e;
ASSERT(startoff == 0);
block = obj;
e = &block->entries[idx];
return bitize(INT_GET(e->nameidx, ARCH_CONVERT));
}
/*
* Remove a name from the leaf directory structure
* This is the external routine.
*/
STATIC int
xfs_dir_leaf_removename(xfs_da_args_t *args, int *count, int *totallen)
{
xfs_dir_leafblock_t *leaf;
int index, retval;
xfs_dabuf_t *bp;
retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
XFS_DATA_FORK);
if (retval)
return(retval);
ASSERT(bp != NULL);
leaf = bp->data;
ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
retval = xfs_dir_leaf_lookup_int(bp, args, &index);
if (retval == EEXIST) {
(void)xfs_dir_leaf_remove(args->trans, bp, index);
*count = INT_GET(leaf->hdr.count, ARCH_CONVERT);
*totallen = INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
retval = 0;
}
xfs_da_buf_done(bp);
return(retval);
}
/* global_hdr_checksum_set - fill in the global media file header checksum.
* utility function for use by drive-specific strategies.
*/
void
global_hdr_checksum_set( global_hdr_t *hdrp )
{
u_int32_t *beginp = ( u_int32_t * )&hdrp[ 0 ];
u_int32_t *endp = ( u_int32_t * )&hdrp[ 1 ];
u_int32_t *p;
u_int32_t accum;
hdrp->gh_checksum = 0;
accum = 0;
for ( p = beginp ; p < endp ; p++ ) {
accum += INT_GET(*p, ARCH_CONVERT);
}
INT_SET(hdrp->gh_checksum, ARCH_CONVERT, (int32_t)(~accum + 1));
}
/*ARGSUSED*/
int
dir_leaf_name_size(
void *obj,
int startoff,
int idx)
{
xfs_dir_leafblock_t *block;
xfs_dir_leaf_entry_t *e;
ASSERT(startoff == 0);
block = obj;
if (INT_GET(block->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC)
return 0;
e = &block->entries[idx];
return bitize((int)XFS_DIR_LEAF_ENTSIZE_BYENTRY(e));
}
static int
bnobt_rec_offset(
void *obj,
int startoff,
int idx)
{
xfs_alloc_block_t *block;
xfs_alloc_rec_t *rp;
ASSERT(startoff == 0);
block = obj;
ASSERT(INT_GET(block->bb_level, ARCH_CONVERT) == 0);
rp = XFS_BTREE_REC_ADDR(mp->m_sb.sb_blocksize, xfs_alloc, block, idx,
XFS_BTREE_BLOCK_MAXRECS(mp->m_sb.sb_blocksize, xfs_alloc, 1));
return bitize((int)((char *)rp - (char *)block));
}
/*
* Initialize the dquot log item for a newly allocated dquot.
* The dquot isn't locked at this point, but it isn't on any of the lists
* either, so we don't care.
*/
void
xfs_qm_dquot_logitem_init(
struct xfs_dquot *dqp)
{
xfs_dq_logitem_t *lp;
lp = &dqp->q_logitem;
lp->qli_item.li_type = XFS_LI_DQUOT;
lp->qli_item.li_ops = &xfs_dquot_item_ops;
lp->qli_item.li_mountp = dqp->q_mount;
lp->qli_dquot = dqp;
lp->qli_format.qlf_type = XFS_LI_DQUOT;
lp->qli_format.qlf_id = INT_GET(dqp->q_core.d_id, ARCH_CONVERT);
lp->qli_format.qlf_blkno = dqp->q_blkno;
lp->qli_format.qlf_len = 1;
/*
* This is just the offset of this dquot within its buffer
* (which is currently 1 FSB and probably won't change).
* Hence 32 bits for this offset should be just fine.
* Alternatively, we can store (bufoffset / sizeof(xfs_dqblk_t))
* here, and recompute it at recovery time.
*/
lp->qli_format.qlf_boffset = (__uint32_t)dqp->q_bufoffset;
}
/*
* Lookup a dquot in the incore dquot hashtable. We keep two separate
* hashtables for user and group dquots; and, these are global tables
* inside the XQM, not per-filesystem tables.
* The hash chain must be locked by caller, and it is left locked
* on return. Returning dquot is locked.
*/
STATIC int
xfs_qm_dqlookup(
xfs_mount_t *mp,
xfs_dqid_t id,
xfs_dqhash_t *qh,
xfs_dquot_t **O_dqpp)
{
xfs_dquot_t *dqp;
uint flist_locked;
xfs_dquot_t *d;
ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
flist_locked = B_FALSE;
/*
* Traverse the hashchain looking for a match
*/
for (dqp = qh->qh_next; dqp != NULL; dqp = dqp->HL_NEXT) {
/*
* We already have the hashlock. We don't need the
* dqlock to look at the id field of the dquot, since the
* id can't be modified without the hashlock anyway.
*/
if (INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id && dqp->q_mount == mp) {
xfs_dqtrace_entry(dqp, "DQFOUND BY LOOKUP");
/*
* All in core dquots must be on the dqlist of mp
*/
ASSERT(dqp->MPL_PREVP != NULL);
xfs_dqlock(dqp);
if (dqp->q_nrefs == 0) {
ASSERT (XFS_DQ_IS_ON_FREELIST(dqp));
if (! xfs_qm_freelist_lock_nowait(xfs_Gqm)) {
xfs_dqtrace_entry(dqp, "DQLOOKUP: WANT");
/*
* We may have raced with dqreclaim_one()
* (and lost). So, flag that we don't
* want the dquot to be reclaimed.
*/
dqp->dq_flags |= XFS_DQ_WANT;
xfs_dqunlock(dqp);
xfs_qm_freelist_lock(xfs_Gqm);
xfs_dqlock(dqp);
dqp->dq_flags &= ~(XFS_DQ_WANT);
}
flist_locked = B_TRUE;
}
/*
* id couldn't have changed; we had the hashlock all
* along
*/
ASSERT(INT_GET(dqp->q_core.d_id, ARCH_CONVERT) == id);
if (flist_locked) {
if (dqp->q_nrefs != 0) {
xfs_qm_freelist_unlock(xfs_Gqm);
flist_locked = B_FALSE;
} else {
/*
* take it off the freelist
*/
xfs_dqtrace_entry(dqp,
"DQLOOKUP: TAKEOFF FL");
XQM_FREELIST_REMOVE(dqp);
/* xfs_qm_freelist_print(&(xfs_Gqm->
qm_dqfreelist),
"after removal"); */
}
}
/*
* grab a reference
*/
XFS_DQHOLD(dqp);
if (flist_locked)
xfs_qm_freelist_unlock(xfs_Gqm);
/*
* move the dquot to the front of the hashchain
*/
ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
if (dqp->HL_PREVP != &qh->qh_next) {
xfs_dqtrace_entry(dqp,
"DQLOOKUP: HASH MOVETOFRONT");
if ((d = dqp->HL_NEXT))
d->HL_PREVP = dqp->HL_PREVP;
*(dqp->HL_PREVP) = d;
d = qh->qh_next;
d->HL_PREVP = &dqp->HL_NEXT;
dqp->HL_NEXT = d;
dqp->HL_PREVP = &qh->qh_next;
qh->qh_next = dqp;
}
xfs_dqtrace_entry(dqp, "LOOKUP END");
*O_dqpp = dqp;
ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
return (0);
}
}
*O_dqpp = NULL;
ASSERT(XFS_DQ_IS_HASH_LOCKED(qh));
return (1);
}
/*
* Convert a block format directory to shortform.
* Caller has already checked that it will fit, and built us a header.
*/
int /* error */
xfs_dir2_block_to_sf(
xfs_da_args_t *args, /* operation arguments */
xfs_dabuf_t *bp, /* block buffer */
int size, /* shortform directory size */
xfs_dir2_sf_hdr_t *sfhp) /* shortform directory hdr */
{
xfs_dir2_block_t *block; /* block structure */
xfs_dir2_block_tail_t *btp; /* block tail pointer */
xfs_dir2_data_entry_t *dep; /* data entry pointer */
xfs_inode_t *dp; /* incore directory inode */
xfs_dir2_data_unused_t *dup; /* unused data pointer */
char *endptr; /* end of data entries */
int error; /* error return value */
int logflags; /* inode logging flags */
xfs_mount_t *mp; /* filesystem mount point */
char *ptr; /* current data pointer */
xfs_dir2_sf_entry_t *sfep; /* shortform entry */
xfs_dir2_sf_t *sfp; /* shortform structure */
xfs_ino_t temp;
xfs_dir2_trace_args_sb("block_to_sf", args, size, bp);
dp = args->dp;
mp = dp->i_mount;
/*
* Make a copy of the block data, so we can shrink the inode
* and add local data.
*/
block = kmem_alloc(mp->m_dirblksize, KM_SLEEP);
memcpy(block, bp->data, mp->m_dirblksize);
logflags = XFS_ILOG_CORE;
if ((error = xfs_dir2_shrink_inode(args, mp->m_dirdatablk, bp))) {
ASSERT(error != ENOSPC);
goto out;
}
/*
* The buffer is now unconditionally gone, whether
* xfs_dir2_shrink_inode worked or not.
*
* Convert the inode to local format.
*/
dp->i_df.if_flags &= ~XFS_IFEXTENTS;
dp->i_df.if_flags |= XFS_IFINLINE;
dp->i_d.di_format = XFS_DINODE_FMT_LOCAL;
ASSERT(dp->i_df.if_bytes == 0);
xfs_idata_realloc(dp, size, XFS_DATA_FORK);
logflags |= XFS_ILOG_DDATA;
/*
* Copy the header into the newly allocate local space.
*/
sfp = (xfs_dir2_sf_t *)dp->i_df.if_u1.if_data;
memcpy(sfp, sfhp, XFS_DIR2_SF_HDR_SIZE(sfhp->i8count));
dp->i_d.di_size = size;
/*
* Set up to loop over the block's entries.
*/
btp = XFS_DIR2_BLOCK_TAIL_P(mp, block);
ptr = (char *)block->u;
endptr = (char *)XFS_DIR2_BLOCK_LEAF_P(btp);
sfep = XFS_DIR2_SF_FIRSTENTRY(sfp);
/*
* Loop over the active and unused entries.
* Stop when we reach the leaf/tail portion of the block.
*/
while (ptr < endptr) {
/*
* If it's unused, just skip over it.
*/
dup = (xfs_dir2_data_unused_t *)ptr;
if (INT_GET(dup->freetag, ARCH_CONVERT) == XFS_DIR2_DATA_FREE_TAG) {
ptr += INT_GET(dup->length, ARCH_CONVERT);
continue;
}
dep = (xfs_dir2_data_entry_t *)ptr;
/*
* Skip .
*/
if (dep->namelen == 1 && dep->name[0] == '.')
ASSERT(INT_GET(dep->inumber, ARCH_CONVERT) == dp->i_ino);
/*
* Skip .., but make sure the inode number is right.
*/
else if (dep->namelen == 2 &&
dep->name[0] == '.' && dep->name[1] == '.')
ASSERT(INT_GET(dep->inumber, ARCH_CONVERT) ==
XFS_DIR2_SF_GET_INUMBER(sfp, &sfp->hdr.parent));
/*
* Normal entry, copy it into shortform.
*/
else {
sfep->namelen = dep->namelen;
XFS_DIR2_SF_PUT_OFFSET(sfep,
(xfs_dir2_data_aoff_t)
((char *)dep - (char *)block));
memcpy(sfep->name, dep->name, dep->namelen);
temp=INT_GET(dep->inumber, ARCH_CONVERT);
XFS_DIR2_SF_PUT_INUMBER(sfp, &temp,
XFS_DIR2_SF_INUMBERP(sfep));
sfep = XFS_DIR2_SF_NEXTENTRY(sfp, sfep);
}
ptr += XFS_DIR2_DATA_ENTSIZE(dep->namelen);
}
ASSERT((char *)sfep - (char *)sfp == size);
xfs_dir2_sf_check(args);
out:
xfs_trans_log_inode(args->trans, dp, logflags);
kmem_free(block, mp->m_dirblksize);
return error;
}
/*
* Given a block directory (dp/block), calculate its size as a shortform (sf)
* directory and a header for the sf directory, if it will fit it the
* space currently present in the inode. If it won't fit, the output
* size is too big (but not accurate).
*/
int /* size for sf form */
xfs_dir2_block_sfsize(
xfs_inode_t *dp, /* incore inode pointer */
xfs_dir2_block_t *block, /* block directory data */
xfs_dir2_sf_hdr_t *sfhp) /* output: header for sf form */
{
xfs_dir2_dataptr_t addr; /* data entry address */
xfs_dir2_leaf_entry_t *blp; /* leaf area of the block */
xfs_dir2_block_tail_t *btp; /* tail area of the block */
int count; /* shortform entry count */
xfs_dir2_data_entry_t *dep; /* data entry in the block */
int i; /* block entry index */
int i8count; /* count of big-inode entries */
int isdot; /* entry is "." */
int isdotdot; /* entry is ".." */
xfs_mount_t *mp; /* mount structure pointer */
int namelen; /* total name bytes */
xfs_ino_t parent = 0; /* parent inode number */
int size=0; /* total computed size */
mp = dp->i_mount;
count = i8count = namelen = 0;
btp = XFS_DIR2_BLOCK_TAIL_P(mp, block);
blp = XFS_DIR2_BLOCK_LEAF_P(btp);
/*
* Iterate over the block's data entries by using the leaf pointers.
*/
for (i = 0; i < INT_GET(btp->count, ARCH_CONVERT); i++) {
if ((addr = INT_GET(blp[i].address, ARCH_CONVERT)) == XFS_DIR2_NULL_DATAPTR)
continue;
/*
* Calculate the pointer to the entry at hand.
*/
dep = (xfs_dir2_data_entry_t *)
((char *)block + XFS_DIR2_DATAPTR_TO_OFF(mp, addr));
/*
* Detect . and .., so we can special-case them.
* . is not included in sf directories.
* .. is included by just the parent inode number.
*/
isdot = dep->namelen == 1 && dep->name[0] == '.';
isdotdot =
dep->namelen == 2 &&
dep->name[0] == '.' && dep->name[1] == '.';
#if XFS_BIG_INUMS
if (!isdot)
i8count += INT_GET(dep->inumber, ARCH_CONVERT) > XFS_DIR2_MAX_SHORT_INUM;
#endif
if (!isdot && !isdotdot) {
count++;
namelen += dep->namelen;
} else if (isdotdot)
parent = INT_GET(dep->inumber, ARCH_CONVERT);
/*
* Calculate the new size, see if we should give up yet.
*/
/* Add by Jerry 2007/10/25
Fixed bug that XFS file system corrupted under Winthrax test.
This is the "special" structure alignment on the ARM.
Port a workaround solution from a patch on
http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-March/020287.html
and http://www.nas-central.org/index.php/Buffalo_ARM9_Kernel_Port#XFS_Arm_Issues
This patch was never accepted by XFS because the code is incorrect for all other platforms.
*/
#if 0
size = XFS_DIR2_SF_HDR_SIZE(i8count) + /* header */
count + /* namelen */
count * (uint)sizeof(xfs_dir2_sf_off_t) + /* offset */
namelen + /* name */
(i8count ? /* inumber */
(uint)sizeof(xfs_dir2_ino8_t) * count :
(uint)sizeof(xfs_dir2_ino4_t) * count);
#else
size = XFS_DIR2_SF_HDR_SIZE(i8count) /* header */
+ namelen
+ (count * (sizeof(xfs_dir2_sf_entry_t) - 1))
- (count * (((i8count == 0) ? 1 : 0) *
(sizeof(xfs_dir2_ino8_t) - sizeof(xfs_dir2_ino4_t))));
#endif
if (size > XFS_IFORK_DSIZE(dp))
return size; /* size value is a failure */
}
/*
* Create the output header, if it worked.
*/
sfhp->count = count;
sfhp->i8count = i8count;
XFS_DIR2_SF_PUT_INUMBER((xfs_dir2_sf_t *)sfhp, &parent, &sfhp->parent);
return size;
}
/*
* Adjust quota limits, and start/stop timers accordingly.
*/
STATIC int
xfs_qm_scall_setqlim(
xfs_mount_t *mp,
xfs_dqid_t id,
uint type,
fs_disk_quota_t *newlim)
{
xfs_disk_dquot_t *ddq;
xfs_dquot_t *dqp;
xfs_trans_t *tp;
int error;
xfs_qcnt_t hard, soft;
if (!capable(CAP_SYS_ADMIN))
return XFS_ERROR(EPERM);
if ((newlim->d_fieldmask & (FS_DQ_LIMIT_MASK|FS_DQ_TIMER_MASK)) == 0)
return (0);
tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SETQLIM);
if ((error = xfs_trans_reserve(tp, 0, sizeof(xfs_disk_dquot_t) + 128,
0, 0, XFS_DEFAULT_LOG_COUNT))) {
xfs_trans_cancel(tp, 0);
return (error);
}
/*
* We don't want to race with a quotaoff so take the quotaoff lock.
* (We don't hold an inode lock, so there's nothing else to stop
* a quotaoff from happening). (XXXThis doesn't currently happen
* because we take the vfslock before calling xfs_qm_sysent).
*/
mutex_lock(&(XFS_QI_QOFFLOCK(mp)), PINOD);
/*
* Get the dquot (locked), and join it to the transaction.
* Allocate the dquot if this doesn't exist.
*/
if ((error = xfs_qm_dqget(mp, NULL, id, type, XFS_QMOPT_DQALLOC, &dqp))) {
xfs_trans_cancel(tp, XFS_TRANS_ABORT);
mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
ASSERT(error != ENOENT);
return (error);
}
xfs_dqtrace_entry(dqp, "Q_SETQLIM: AFT DQGET");
xfs_trans_dqjoin(tp, dqp);
ddq = &dqp->q_core;
/*
* Make sure that hardlimits are >= soft limits before changing.
*/
hard = (newlim->d_fieldmask & FS_DQ_BHARD) ?
(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_blk_hardlimit) :
INT_GET(ddq->d_blk_hardlimit, ARCH_CONVERT);
soft = (newlim->d_fieldmask & FS_DQ_BSOFT) ?
(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_blk_softlimit) :
INT_GET(ddq->d_blk_softlimit, ARCH_CONVERT);
if (hard == 0 || hard >= soft) {
INT_SET(ddq->d_blk_hardlimit, ARCH_CONVERT, hard);
INT_SET(ddq->d_blk_softlimit, ARCH_CONVERT, soft);
if (id == 0) {
mp->m_quotainfo->qi_bhardlimit = hard;
mp->m_quotainfo->qi_bsoftlimit = soft;
}
} else {
qdprintk("blkhard %Ld < blksoft %Ld\n", hard, soft);
}
hard = (newlim->d_fieldmask & FS_DQ_RTBHARD) ?
(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_rtb_hardlimit) :
INT_GET(ddq->d_rtb_hardlimit, ARCH_CONVERT);
soft = (newlim->d_fieldmask & FS_DQ_RTBSOFT) ?
(xfs_qcnt_t) XFS_BB_TO_FSB(mp, newlim->d_rtb_softlimit) :
INT_GET(ddq->d_rtb_softlimit, ARCH_CONVERT);
if (hard == 0 || hard >= soft) {
INT_SET(ddq->d_rtb_hardlimit, ARCH_CONVERT, hard);
INT_SET(ddq->d_rtb_softlimit, ARCH_CONVERT, soft);
if (id == 0) {
mp->m_quotainfo->qi_rtbhardlimit = hard;
mp->m_quotainfo->qi_rtbsoftlimit = soft;
}
} else {
qdprintk("rtbhard %Ld < rtbsoft %Ld\n", hard, soft);
}
hard = (newlim->d_fieldmask & FS_DQ_IHARD) ?
(xfs_qcnt_t) newlim->d_ino_hardlimit :
INT_GET(ddq->d_ino_hardlimit, ARCH_CONVERT);
soft = (newlim->d_fieldmask & FS_DQ_ISOFT) ?
(xfs_qcnt_t) newlim->d_ino_softlimit :
INT_GET(ddq->d_ino_softlimit, ARCH_CONVERT);
if (hard == 0 || hard >= soft) {
INT_SET(ddq->d_ino_hardlimit, ARCH_CONVERT, hard);
INT_SET(ddq->d_ino_softlimit, ARCH_CONVERT, soft);
if (id == 0) {
mp->m_quotainfo->qi_ihardlimit = hard;
mp->m_quotainfo->qi_isoftlimit = soft;
}
} else {
qdprintk("ihard %Ld < isoft %Ld\n", hard, soft);
}
if (id == 0) {
/*
* Timelimits for the super user set the relative time
* the other users can be over quota for this file system.
* If it is zero a default is used. Ditto for the default
* soft and hard limit values (already done, above).
*/
if (newlim->d_fieldmask & FS_DQ_BTIMER) {
mp->m_quotainfo->qi_btimelimit = newlim->d_btimer;
INT_SET(ddq->d_btimer, ARCH_CONVERT, newlim->d_btimer);
}
if (newlim->d_fieldmask & FS_DQ_ITIMER) {
mp->m_quotainfo->qi_itimelimit = newlim->d_itimer;
INT_SET(ddq->d_itimer, ARCH_CONVERT, newlim->d_itimer);
}
if (newlim->d_fieldmask & FS_DQ_RTBTIMER) {
mp->m_quotainfo->qi_rtbtimelimit = newlim->d_rtbtimer;
INT_SET(ddq->d_rtbtimer, ARCH_CONVERT, newlim->d_rtbtimer);
}
} else /* if (XFS_IS_QUOTA_ENFORCED(mp)) */ {
/*
* If the user is now over quota, start the timelimit.
* The user will not be 'warned'.
* Note that we keep the timers ticking, whether enforcement
* is on or off. We don't really want to bother with iterating
* over all ondisk dquots and turning the timers on/off.
*/
xfs_qm_adjust_dqtimers(mp, ddq);
}
dqp->dq_flags |= XFS_DQ_DIRTY;
xfs_trans_log_dquot(tp, dqp);
xfs_dqtrace_entry(dqp, "Q_SETQLIM: COMMIT");
xfs_trans_commit(tp, 0, NULL);
xfs_qm_dqprint(dqp);
xfs_qm_dqrele(dqp);
mutex_unlock(&(XFS_QI_QOFFLOCK(mp)));
return (0);
}
STATIC int
xfs_dir_node_getdents(xfs_trans_t *trans, xfs_inode_t *dp, uio_t *uio,
int *eofp, dirent_t *dbp, xfs_dir_put_t put)
{
xfs_da_intnode_t *node;
xfs_da_node_entry_t *btree;
xfs_dir_leafblock_t *leaf;
xfs_dablk_t bno, nextbno;
xfs_dahash_t cookhash;
xfs_mount_t *mp;
int error, eob, i;
xfs_dabuf_t *bp;
xfs_daddr_t nextda;
/*
* Pick up our context.
*/
mp = dp->i_mount;
bp = NULL;
bno = XFS_DA_COOKIE_BNO(mp, uio->uio_offset);
cookhash = XFS_DA_COOKIE_HASH(mp, uio->uio_offset);
xfs_dir_trace_g_du("node: start", dp, uio);
/*
* Re-find our place, even if we're confused about what our place is.
*
* First we check the block number from the magic cookie, it is a
* cache of where we ended last time. If we find a leaf block, and
* the starting hashval in that block is less than our desired
* hashval, then we run with it.
*/
if (bno > 0) {
error = xfs_da_read_buf(trans, dp, bno, -1, &bp, XFS_DATA_FORK);
if ((error != 0) && (error != EFSCORRUPTED))
return(error);
if (bp)
leaf = bp->data;
if (bp && INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC) {
xfs_dir_trace_g_dub("node: block not a leaf",
dp, uio, bno);
xfs_da_brelse(trans, bp);
bp = NULL;
}
if (bp && INT_GET(leaf->entries[0].hashval, ARCH_CONVERT) > cookhash) {
xfs_dir_trace_g_dub("node: leaf hash too large",
dp, uio, bno);
xfs_da_brelse(trans, bp);
bp = NULL;
}
if (bp &&
cookhash > INT_GET(leaf->entries[INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT)) {
xfs_dir_trace_g_dub("node: leaf hash too small",
dp, uio, bno);
xfs_da_brelse(trans, bp);
bp = NULL;
}
}
/*
* If we did not find a leaf block from the blockno in the cookie,
* or we there was no blockno in the cookie (eg: first time thru),
* the we start at the top of the Btree and re-find our hashval.
*/
if (bp == NULL) {
xfs_dir_trace_g_du("node: start at root" , dp, uio);
bno = 0;
for (;;) {
error = xfs_da_read_buf(trans, dp, bno, -1, &bp,
XFS_DATA_FORK);
if (error)
return(error);
if (bp == NULL)
return(XFS_ERROR(EFSCORRUPTED));
node = bp->data;
if (INT_GET(node->hdr.info.magic, ARCH_CONVERT) != XFS_DA_NODE_MAGIC)
break;
btree = &node->btree[0];
xfs_dir_trace_g_dun("node: node detail", dp, uio, node);
for (i = 0; i < INT_GET(node->hdr.count, ARCH_CONVERT); btree++, i++) {
if (INT_GET(btree->hashval, ARCH_CONVERT) >= cookhash) {
bno = INT_GET(btree->before, ARCH_CONVERT);
break;
}
}
if (i == INT_GET(node->hdr.count, ARCH_CONVERT)) {
xfs_da_brelse(trans, bp);
xfs_dir_trace_g_du("node: hash beyond EOF",
dp, uio);
uio->uio_offset = XFS_DA_MAKE_COOKIE(mp, 0, 0,
XFS_DA_MAXHASH);
*eofp = 1;
return(0);
}
xfs_dir_trace_g_dub("node: going to block",
dp, uio, bno);
xfs_da_brelse(trans, bp);
}
}
ASSERT(cookhash != XFS_DA_MAXHASH);
/*
* We've dropped down to the (first) leaf block that contains the
* hashval we are interested in. Continue rolling upward thru the
* leaf blocks until we fill up our buffer.
*/
for (;;) {
leaf = bp->data;
if (INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC) {
xfs_dir_trace_g_dul("node: not a leaf", dp, uio, leaf);
xfs_da_brelse(trans, bp);
return XFS_ERROR(EFSCORRUPTED);
}
xfs_dir_trace_g_dul("node: leaf detail", dp, uio, leaf);
if (nextbno = INT_GET(leaf->hdr.info.forw, ARCH_CONVERT)) {
nextda = xfs_da_reada_buf(trans, dp, nextbno,
XFS_DATA_FORK);
} else
nextda = -1;
error = xfs_dir_leaf_getdents_int(bp, dp, bno, uio, &eob, dbp,
put, nextda);
xfs_da_brelse(trans, bp);
bno = nextbno;
if (eob) {
xfs_dir_trace_g_dub("node: E-O-B", dp, uio, bno);
*eofp = 0;
return(error);
}
if (bno == 0)
break;
error = xfs_da_read_buf(trans, dp, bno, nextda, &bp,
XFS_DATA_FORK);
if (error)
return(error);
if (bp == NULL)
return(XFS_ERROR(EFSCORRUPTED));
}
*eofp = 1;
xfs_dir_trace_g_du("node: E-O-F", dp, uio);
return(0);
}
/*
* Translate an internal style on-disk-dquot to the exportable format.
* The main differences are that the counters/limits are all in Basic
* Blocks (BBs) instead of the internal FSBs, and all on-disk data has
* to be converted to the native endianness.
*/
STATIC void
xfs_qm_export_dquot(
xfs_mount_t *mp,
xfs_disk_dquot_t *src,
struct fs_disk_quota *dst)
{
memset(dst, 0, sizeof(*dst));
dst->d_version = FS_DQUOT_VERSION; /* different from src->d_version */
dst->d_flags =
xfs_qm_export_qtype_flags(INT_GET(src->d_flags, ARCH_CONVERT));
dst->d_id = INT_GET(src->d_id, ARCH_CONVERT);
dst->d_blk_hardlimit = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_blk_hardlimit, ARCH_CONVERT));
dst->d_blk_softlimit = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_blk_softlimit, ARCH_CONVERT));
dst->d_ino_hardlimit = (__uint64_t)
INT_GET(src->d_ino_hardlimit, ARCH_CONVERT);
dst->d_ino_softlimit = (__uint64_t)
INT_GET(src->d_ino_softlimit, ARCH_CONVERT);
dst->d_bcount = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_bcount, ARCH_CONVERT));
dst->d_icount = (__uint64_t) INT_GET(src->d_icount, ARCH_CONVERT);
dst->d_btimer = (__uint32_t) INT_GET(src->d_btimer, ARCH_CONVERT);
dst->d_itimer = (__uint32_t) INT_GET(src->d_itimer, ARCH_CONVERT);
dst->d_iwarns = INT_GET(src->d_iwarns, ARCH_CONVERT);
dst->d_bwarns = INT_GET(src->d_bwarns, ARCH_CONVERT);
dst->d_rtb_hardlimit = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_rtb_hardlimit, ARCH_CONVERT));
dst->d_rtb_softlimit = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_rtb_softlimit, ARCH_CONVERT));
dst->d_rtbcount = (__uint64_t)
XFS_FSB_TO_BB(mp, INT_GET(src->d_rtbcount, ARCH_CONVERT));
dst->d_rtbtimer = (__uint32_t) INT_GET(src->d_rtbtimer, ARCH_CONVERT);
dst->d_rtbwarns = INT_GET(src->d_rtbwarns, ARCH_CONVERT);
/*
* Internally, we don't reset all the timers when quota enforcement
* gets turned off. No need to confuse the userlevel code,
* so return zeroes in that case.
*/
if (! XFS_IS_QUOTA_ENFORCED(mp)) {
dst->d_btimer = 0;
dst->d_itimer = 0;
dst->d_rtbtimer = 0;
}
#ifdef DEBUG
if (XFS_IS_QUOTA_ENFORCED(mp) && dst->d_id != 0) {
if (((int) dst->d_bcount >= (int) dst->d_blk_softlimit) &&
(dst->d_blk_softlimit > 0)) {
ASSERT(dst->d_btimer != 0);
}
if (((int) dst->d_icount >= (int) dst->d_ino_softlimit) &&
(dst->d_ino_softlimit > 0)) {
ASSERT(dst->d_itimer != 0);
}
}
#endif
}
/*
* Called by xfs_trans_commit() and similar in spirit to
* xfs_trans_apply_sb_deltas().
* Go thru all the dquots belonging to this transaction and modify the
* INCORE dquot to reflect the actual usages.
* Unreserve just the reservations done by this transaction
* dquot is still left locked at exit.
*/
void
xfs_trans_apply_dquot_deltas(
xfs_trans_t *tp)
{
int i, j;
xfs_dquot_t *dqp;
xfs_dqtrx_t *qtrx, *qa;
xfs_disk_dquot_t *d;
long totalbdelta;
long totalrtbdelta;
ASSERT(tp->t_dqinfo);
qa = tp->t_dqinfo->dqa_usrdquots;
for (j = 0; j < 2; j++) {
if (qa[0].qt_dquot == NULL) {
qa = tp->t_dqinfo->dqa_grpdquots;
continue;
}
/*
* Lock all of the dquots and join them to the transaction.
*/
xfs_trans_dqlockedjoin(tp, qa);
for (i = 0; i < XFS_QM_TRANS_MAXDQS; i++) {
qtrx = &qa[i];
/*
* The array of dquots is filled
* sequentially, not sparsely.
*/
if ((dqp = qtrx->qt_dquot) == NULL)
break;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
ASSERT(XFS_DQ_IS_ADDEDTO_TRX(tp, dqp));
/*
* adjust the actual number of blocks used
*/
d = &dqp->q_core;
/*
* The issue here is - sometimes we don't make a blkquota
* reservation intentionally to be fair to users
* (when the amount is small). On the other hand,
* delayed allocs do make reservations, but that's
* outside of a transaction, so we have no
* idea how much was really reserved.
* So, here we've accumulated delayed allocation blks and
* non-delay blks. The assumption is that the
* delayed ones are always reserved (outside of a
* transaction), and the others may or may not have
* quota reservations.
*/
totalbdelta = qtrx->qt_bcount_delta +
qtrx->qt_delbcnt_delta;
totalrtbdelta = qtrx->qt_rtbcount_delta +
qtrx->qt_delrtb_delta;
#ifdef QUOTADEBUG
if (totalbdelta < 0)
ASSERT(INT_GET(d->d_bcount, ARCH_CONVERT) >=
(xfs_qcnt_t) -totalbdelta);
if (totalrtbdelta < 0)
ASSERT(INT_GET(d->d_rtbcount, ARCH_CONVERT) >=
(xfs_qcnt_t) -totalrtbdelta);
if (qtrx->qt_icount_delta < 0)
ASSERT(INT_GET(d->d_icount, ARCH_CONVERT) >=
(xfs_qcnt_t) -qtrx->qt_icount_delta);
#endif
if (totalbdelta)
INT_MOD(d->d_bcount, ARCH_CONVERT, (xfs_qcnt_t)totalbdelta);
if (qtrx->qt_icount_delta)
INT_MOD(d->d_icount, ARCH_CONVERT, (xfs_qcnt_t)qtrx->qt_icount_delta);
if (totalrtbdelta)
INT_MOD(d->d_rtbcount, ARCH_CONVERT, (xfs_qcnt_t)totalrtbdelta);
/*
* Start/reset the timer(s) if needed.
*/
xfs_qm_adjust_dqtimers(tp->t_mountp, d);
dqp->dq_flags |= XFS_DQ_DIRTY;
/*
* add this to the list of items to get logged
*/
xfs_trans_log_dquot(tp, dqp);
/*
* Take off what's left of the original reservation.
* In case of delayed allocations, there's no
* reservation that a transaction structure knows of.
*/
if (qtrx->qt_blk_res != 0) {
if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
if (qtrx->qt_blk_res >
qtrx->qt_blk_res_used)
dqp->q_res_bcount -= (xfs_qcnt_t)
(qtrx->qt_blk_res -
qtrx->qt_blk_res_used);
else
dqp->q_res_bcount -= (xfs_qcnt_t)
(qtrx->qt_blk_res_used -
qtrx->qt_blk_res);
}
} else {
/*
* These blks were never reserved, either inside
* a transaction or outside one (in a delayed
* allocation). Also, this isn't always a
* negative number since we sometimes
* deliberately skip quota reservations.
*/
if (qtrx->qt_bcount_delta) {
dqp->q_res_bcount +=
(xfs_qcnt_t)qtrx->qt_bcount_delta;
}
}
/*
* Adjust the RT reservation.
*/
if (qtrx->qt_rtblk_res != 0) {
if (qtrx->qt_blk_res != qtrx->qt_blk_res_used) {
if (qtrx->qt_rtblk_res >
qtrx->qt_rtblk_res_used)
dqp->q_res_rtbcount -= (xfs_qcnt_t)
(qtrx->qt_rtblk_res -
qtrx->qt_rtblk_res_used);
else
dqp->q_res_rtbcount -= (xfs_qcnt_t)
(qtrx->qt_rtblk_res_used -
qtrx->qt_rtblk_res);
}
} else {
if (qtrx->qt_rtbcount_delta)
dqp->q_res_rtbcount +=
(xfs_qcnt_t)qtrx->qt_rtbcount_delta;
}
/*
* Adjust the inode reservation.
*/
if (qtrx->qt_ino_res != 0) {
ASSERT(qtrx->qt_ino_res >=
qtrx->qt_ino_res_used);
if (qtrx->qt_ino_res > qtrx->qt_ino_res_used)
dqp->q_res_icount -= (xfs_qcnt_t)
(qtrx->qt_ino_res -
qtrx->qt_ino_res_used);
} else {
if (qtrx->qt_icount_delta)
dqp->q_res_icount +=
(xfs_qcnt_t)qtrx->qt_icount_delta;
}
#ifdef QUOTADEBUG
if (qtrx->qt_rtblk_res != 0)
printk("RT res %d for 0x%p\n",
(int) qtrx->qt_rtblk_res,
dqp);
#endif
ASSERT(dqp->q_res_bcount >= INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
ASSERT(dqp->q_res_icount >= INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
ASSERT(dqp->q_res_rtbcount >= INT_GET(dqp->q_core.d_rtbcount, ARCH_CONVERT));
}
/*
* Do the group quotas next
*/
qa = tp->t_dqinfo->dqa_grpdquots;
}
}
/*
* This reserves disk blocks and inodes against a dquot.
* Flags indicate if the dquot is to be locked here and also
* if the blk reservation is for RT or regular blocks.
* Sending in XFS_QMOPT_FORCE_RES flag skips the quota check.
* Returns EDQUOT if quota is exceeded.
*/
STATIC int
xfs_trans_dqresv(
xfs_trans_t *tp,
xfs_dquot_t *dqp,
long nblks,
long ninos,
uint flags)
{
int error;
xfs_qcnt_t hardlimit;
xfs_qcnt_t softlimit;
time_t btimer;
xfs_qcnt_t *resbcountp;
if (! (flags & XFS_QMOPT_DQLOCK)) {
xfs_dqlock(dqp);
}
ASSERT(XFS_DQ_IS_LOCKED(dqp));
if (flags & XFS_TRANS_DQ_RES_BLKS) {
hardlimit = INT_GET(dqp->q_core.d_blk_hardlimit, ARCH_CONVERT);
softlimit = INT_GET(dqp->q_core.d_blk_softlimit, ARCH_CONVERT);
btimer = INT_GET(dqp->q_core.d_btimer, ARCH_CONVERT);
resbcountp = &dqp->q_res_bcount;
} else {
ASSERT(flags & XFS_TRANS_DQ_RES_RTBLKS);
hardlimit = INT_GET(dqp->q_core.d_rtb_hardlimit, ARCH_CONVERT);
softlimit = INT_GET(dqp->q_core.d_rtb_softlimit, ARCH_CONVERT);
btimer = INT_GET(dqp->q_core.d_rtbtimer, ARCH_CONVERT);
resbcountp = &dqp->q_res_rtbcount;
}
error = 0;
if ((flags & XFS_QMOPT_FORCE_RES) == 0 &&
!INT_ISZERO(dqp->q_core.d_id, ARCH_CONVERT) &&
XFS_IS_QUOTA_ENFORCED(dqp->q_mount)) {
#ifdef QUOTADEBUG
printk("BLK Res: nblks=%ld + resbcount=%Ld > hardlimit=%Ld?\n",
nblks, *resbcountp, hardlimit);
#endif
if (nblks > 0) {
/*
* dquot is locked already. See if we'd go over the
* hardlimit or exceed the timelimit if we allocate
* nblks.
*/
if (hardlimit > 0ULL &&
(hardlimit <= nblks + *resbcountp)) {
error = EDQUOT;
goto error_return;
}
if (softlimit > 0ULL &&
(softlimit <= nblks + *resbcountp)) {
/*
* If timer or warnings has expired,
* return EDQUOT
*/
if ((btimer != 0 && CURRENT_TIME > btimer) ||
(!INT_ISZERO(dqp->q_core.d_bwarns, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_bwarns, ARCH_CONVERT) >=
XFS_QI_BWARNLIMIT(dqp->q_mount))) {
error = EDQUOT;
goto error_return;
}
}
}
if (ninos > 0) {
if (INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT) > 0ULL &&
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_ino_hardlimit, ARCH_CONVERT)) {
error = EDQUOT;
goto error_return;
} else if (INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT) > 0ULL &&
INT_GET(dqp->q_core.d_icount, ARCH_CONVERT) >=
INT_GET(dqp->q_core.d_ino_softlimit, ARCH_CONVERT)) {
/*
* If timer or warnings has expired,
* return EDQUOT
*/
if ((!INT_ISZERO(dqp->q_core.d_itimer, ARCH_CONVERT) &&
CURRENT_TIME > INT_GET(dqp->q_core.d_itimer, ARCH_CONVERT)) ||
(!INT_ISZERO(dqp->q_core.d_iwarns, ARCH_CONVERT) &&
INT_GET(dqp->q_core.d_iwarns, ARCH_CONVERT) >=
XFS_QI_IWARNLIMIT(dqp->q_mount))) {
error = EDQUOT;
goto error_return;
}
}
}
}
/*
* Change the reservation, but not the actual usage.
* Note that q_res_bcount = q_core.d_bcount + resv
*/
(*resbcountp) += (xfs_qcnt_t)nblks;
if (ninos != 0)
dqp->q_res_icount += (xfs_qcnt_t)ninos;
/*
* note the reservation amt in the trans struct too,
* so that the transaction knows how much was reserved by
* it against this particular dquot.
* We don't do this when we are reserving for a delayed allocation,
* because we don't have the luxury of a transaction envelope then.
*/
if (tp) {
ASSERT(tp->t_dqinfo);
ASSERT(flags & XFS_QMOPT_RESBLK_MASK);
if (nblks != 0)
xfs_trans_mod_dquot(tp, dqp,
flags & XFS_QMOPT_RESBLK_MASK,
nblks);
if (ninos != 0)
xfs_trans_mod_dquot(tp, dqp,
XFS_TRANS_DQ_RES_INOS,
ninos);
}
ASSERT(dqp->q_res_bcount >= INT_GET(dqp->q_core.d_bcount, ARCH_CONVERT));
ASSERT(dqp->q_res_rtbcount >= INT_GET(dqp->q_core.d_rtbcount, ARCH_CONVERT));
ASSERT(dqp->q_res_icount >= INT_GET(dqp->q_core.d_icount, ARCH_CONVERT));
error_return:
if (! (flags & XFS_QMOPT_DQLOCK)) {
xfs_dqunlock(dqp);
}
return (error);
}
