*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*/
#define DEBUG_SUBSYSTEM S_LMV
static struct lu_tgt_desc *lmv_locate_tgt_qos(struct lmv_obd *lmv, __u32 *mdt)
{
- struct lu_tgt_desc *tgt;
+ struct lu_tgt_desc *tgt, *cur = NULL;
+ __u64 total_avail = 0;
__u64 total_weight = 0;
__u64 cur_weight = 0;
+ int total_usable = 0;
__u64 rand;
int rc;
GOTO(unlock, tgt = ERR_PTR(rc));
lmv_foreach_tgt(lmv, tgt) {
- tgt->ltd_qos.ltq_usable = 0;
- if (!tgt->ltd_exp || !tgt->ltd_active)
+ if (!tgt->ltd_exp || !tgt->ltd_active) {
+ tgt->ltd_qos.ltq_usable = 0;
continue;
+ }
tgt->ltd_qos.ltq_usable = 1;
lu_tgt_qos_weight_calc(tgt);
+ if (tgt->ltd_index == *mdt)
+ cur = tgt;
+ total_avail += tgt->ltd_qos.ltq_avail;
total_weight += tgt->ltd_qos.ltq_weight;
+ total_usable++;
+ }
+
+ /* if current MDT has above-average space, within range of the QOS
+ * threshold, stay on the same MDT to avoid creating needless remote
+ * MDT directories.
+ */
+ rand = total_avail * (256 - lmv->lmv_qos.lq_threshold_rr) /
+ (total_usable * 256);
+ if (cur && cur->ltd_qos.ltq_avail >= rand) {
+ tgt = cur;
+ GOTO(unlock, rc = 0);
}
rand = lu_prandom_u64_max(total_weight);
ENTRY;
- spin_lock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+ spin_lock(&lmv->lmv_lock);
for (i = 0; i < lmv->lmv_mdt_descs.ltd_tgts_size; i++) {
index = (i + lmv->lmv_qos_rr_index) %
lmv->lmv_mdt_descs.ltd_tgts_size;
*mdt = tgt->ltd_index;
lmv->lmv_qos_rr_index = (*mdt + 1) %
lmv->lmv_mdt_descs.ltd_tgts_size;
- spin_unlock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+ spin_unlock(&lmv->lmv_lock);
RETURN(tgt);
}
- spin_unlock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+ spin_unlock(&lmv->lmv_lock);
RETURN(ERR_PTR(-ENODEV));
}
return rc;
}
+static inline bool lmv_op_user_qos_mkdir(const struct md_op_data *op_data)
+{
+ const struct lmv_user_md *lum = op_data->op_data;
+
+ return (op_data->op_cli_flags & CLI_SET_MEA) && lum &&
+ le32_to_cpu(lum->lum_magic) == LMV_USER_MAGIC &&
+ le32_to_cpu(lum->lum_stripe_offset) == LMV_OFFSET_DEFAULT;
+}
+
+static inline bool lmv_op_default_qos_mkdir(const struct md_op_data *op_data)
+{
+ const struct lmv_stripe_md *lsm = op_data->op_default_mea1;
+
+ return lsm && lsm->lsm_md_master_mdt_index == LMV_OFFSET_DEFAULT;
+}
+
/* mkdir by QoS in two cases:
* 1. 'lfs mkdir -i -1'
* 2. parent default LMV master_mdt_index is -1
*/
static inline bool lmv_op_qos_mkdir(const struct md_op_data *op_data)
{
- const struct lmv_stripe_md *lsm = op_data->op_default_mea1;
- const struct lmv_user_md *lum = op_data->op_data;
-
if (op_data->op_code != LUSTRE_OPC_MKDIR)
return false;
if (lmv_dir_striped(op_data->op_mea1))
return false;
- if (op_data->op_cli_flags & CLI_SET_MEA && lum &&
- (le32_to_cpu(lum->lum_magic) == LMV_USER_MAGIC ||
- le32_to_cpu(lum->lum_magic) == LMV_USER_MAGIC_SPECIFIC) &&
- le32_to_cpu(lum->lum_stripe_offset) == LMV_OFFSET_DEFAULT)
+ if (lmv_op_user_qos_mkdir(op_data))
return true;
- if (lsm && lsm->lsm_md_master_mdt_index == LMV_OFFSET_DEFAULT)
+ if (lmv_op_default_qos_mkdir(op_data))
return true;
return false;
}
+/* if default LMV is set, and its index is LMV_OFFSET_DEFAULT, and
+ * 1. max_inherit_rr is set and is not LMV_INHERIT_RR_NONE
+ * 2. or parent is ROOT
+ * mkdir roundrobin.
+ * NB, this also needs to check server is balanced, which is checked by caller.
+ */
+static inline bool lmv_op_default_rr_mkdir(const struct md_op_data *op_data)
+{
+ const struct lmv_stripe_md *lsm = op_data->op_default_mea1;
+
+ if (!lmv_op_default_qos_mkdir(op_data))
+ return false;
+
+ return lsm->lsm_md_max_inherit_rr != LMV_INHERIT_RR_NONE ||
+ fid_is_root(&op_data->op_fid1);
+}
+
/* 'lfs mkdir -i <specific_MDT>' */
static inline bool lmv_op_user_specific_mkdir(const struct md_op_data *op_data)
{
op_data->op_default_mea1->lsm_md_master_mdt_index !=
LMV_OFFSET_DEFAULT;
}
+
int lmv_create(struct obd_export *exp, struct md_op_data *op_data,
const void *data, size_t datalen, umode_t mode, uid_t uid,
gid_t gid, cfs_cap_t cap_effective, __u64 rdev,
if (!tgt)
RETURN(-ENODEV);
} else if (lmv_op_qos_mkdir(op_data)) {
+ struct lmv_tgt_desc *tmp = tgt;
+
tgt = lmv_locate_tgt_qos(lmv, &op_data->op_mds);
- if (tgt == ERR_PTR(-EAGAIN))
- tgt = lmv_locate_tgt_rr(lmv, &op_data->op_mds);
+ if (tgt == ERR_PTR(-EAGAIN)) {
+ if (ltd_qos_is_balanced(&lmv->lmv_mdt_descs) &&
+ !lmv_op_default_rr_mkdir(op_data) &&
+ !lmv_op_user_qos_mkdir(op_data))
+ /* if it's not necessary, don't create remote
+ * directory.
+ */
+ tgt = tmp;
+ else
+ tgt = lmv_locate_tgt_rr(lmv, &op_data->op_mds);
+ }
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
+
/*
* only update statfs after QoS mkdir, this means the cached
* statfs may be stale, and current mkdir may not follow QoS
lsm->lsm_md_stripe_count = le32_to_cpu(lmu->lum_stripe_count);
lsm->lsm_md_master_mdt_index = le32_to_cpu(lmu->lum_stripe_offset);
lsm->lsm_md_hash_type = le32_to_cpu(lmu->lum_hash_type);
+ lsm->lsm_md_max_inherit = lmu->lum_max_inherit;
+ lsm->lsm_md_max_inherit_rr = lmu->lum_max_inherit_rr;
lsm->lsm_md_pool_name[LOV_MAXPOOLNAME] = 0;
return 0;
}
if (lmv_dir_striped(lsm)) {
- for (i = 0; i < lsm->lsm_md_stripe_count; i++) {
- if (lsm->lsm_md_oinfo[i].lmo_root)
- iput(lsm->lsm_md_oinfo[i].lmo_root);
- }
+ for (i = 0; i < lsm->lsm_md_stripe_count; i++)
+ iput(lsm->lsm_md_oinfo[i].lmo_root);
lsm_size = lmv_stripe_md_size(lsm->lsm_md_stripe_count);
} else {
lsm_size = lmv_stripe_md_size(0);
}
static int
-lmv_get_lustre_md(struct obd_export *exp, struct ptlrpc_request *req,
+lmv_get_lustre_md(struct obd_export *exp, struct req_capsule *pill,
struct obd_export *dt_exp, struct obd_export *md_exp,
struct lustre_md *md)
{
if (!tgt || !tgt->ltd_exp)
return -EINVAL;
- return md_get_lustre_md(tgt->ltd_exp, req, dt_exp, md_exp, md);
+ return md_get_lustre_md(tgt->ltd_exp, pill, dt_exp, md_exp, md);
}
static int lmv_free_lustre_md(struct obd_export *exp, struct lustre_md *md)
}
static int lmv_intent_getattr_async(struct obd_export *exp,
- struct md_enqueue_info *minfo)
+ struct md_op_item *item)
{
- struct md_op_data *op_data = &minfo->mi_data;
+ struct md_op_data *op_data = &item->mop_data;
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *ptgt;
if (ctgt != ptgt)
RETURN(-EREMOTE);
- rc = md_intent_getattr_async(ptgt->ltd_exp, minfo);
+ rc = md_intent_getattr_async(ptgt->ltd_exp, item);
RETURN(rc);
}
static int __init lmv_init(void)
{
- return class_register_type(&lmv_obd_ops, &lmv_md_ops, true, NULL,
+ return class_register_type(&lmv_obd_ops, &lmv_md_ops, true,
LUSTRE_LMV_NAME, NULL);
}