--- /dev/null
+/*
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.gnu.org/licenses/gpl-2.0.html
+ *
+ * GPL HEADER END
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ *
+ * lustre/lmv/lmv_qos.c
+ *
+ * LMV QoS.
+ * These are the only exported functions, they provide some generic
+ * infrastructure for object allocation QoS
+ *
+ */
+
+#define DEBUG_SUBSYSTEM S_LMV
+
+#include <asm/div64.h>
+#include <libcfs/libcfs.h>
+#include <uapi/linux/lustre/lustre_idl.h>
+#include <lustre_swab.h>
+#include <obd_class.h>
+
+#include "lmv_internal.h"
+
+static inline __u64 tgt_statfs_bavail(struct lu_tgt_desc *tgt)
+{
+ struct obd_statfs *statfs = &tgt->ltd_statfs;
+
+ return statfs->os_bavail * statfs->os_bsize;
+}
+
+static inline __u64 tgt_statfs_iavail(struct lu_tgt_desc *tgt)
+{
+ return tgt->ltd_statfs.os_ffree;
+}
+
+/**
+ * Calculate penalties per-tgt and per-server
+ *
+ * Re-calculate penalties when the configuration changes, active targets
+ * change and after statfs refresh (all these are reflected by lq_dirty flag).
+ * On every MDT and MDS: decay the penalty by half for every 8x the update
+ * interval that the device has been idle. That gives lots of time for the
+ * statfs information to be updated (which the penalty is only a proxy for),
+ * and avoids penalizing MDS/MDTs under light load.
+ * See lmv_qos_calc_weight() for how penalties are factored into the weight.
+ *
+ * \param[in] lmv LMV device
+ *
+ * \retval 0 on success
+ * \retval -EAGAIN the number of MDTs isn't enough or all MDT spaces are
+ * almost the same
+ */
+static int lmv_qos_calc_ppts(struct lmv_obd *lmv)
+{
+ struct lu_qos *qos = &lmv->lmv_qos;
+ struct lu_tgt_desc *tgt;
+ struct lu_svr_qos *svr;
+ __u64 ba_max, ba_min, ba;
+ __u64 ia_max, ia_min, ia;
+ __u32 num_active;
+ unsigned int i;
+ int prio_wide;
+ time64_t now, age;
+ __u32 maxage = lmv->desc.ld_qos_maxage;
+ int rc;
+
+ ENTRY;
+
+ if (!qos->lq_dirty)
+ GOTO(out, rc = 0);
+
+ num_active = lmv->desc.ld_active_tgt_count;
+ if (num_active < 2)
+ GOTO(out, rc = -EAGAIN);
+
+ /* find bavail on each server */
+ list_for_each_entry(svr, &qos->lq_svr_list, lsq_svr_list) {
+ svr->lsq_bavail = 0;
+ svr->lsq_iavail = 0;
+ }
+ qos->lq_active_svr_count = 0;
+
+ /*
+ * How badly user wants to select targets "widely" (not recently chosen
+ * and not on recent MDS's). As opposed to "freely" (free space avail.)
+ * 0-256
+ */
+ prio_wide = 256 - qos->lq_prio_free;
+
+ ba_min = (__u64)(-1);
+ ba_max = 0;
+ ia_min = (__u64)(-1);
+ ia_max = 0;
+ now = ktime_get_real_seconds();
+
+ /* Calculate server penalty per object */
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ tgt = lmv->tgts[i];
+ if (!tgt || !tgt->ltd_exp || !tgt->ltd_active)
+ continue;
+
+ /* bavail >> 16 to avoid overflow */
+ ba = tgt_statfs_bavail(tgt) >> 16;
+ if (!ba)
+ continue;
+
+ ba_min = min(ba, ba_min);
+ ba_max = max(ba, ba_max);
+
+ /* iavail >> 8 to avoid overflow */
+ ia = tgt_statfs_iavail(tgt) >> 8;
+ if (!ia)
+ continue;
+
+ ia_min = min(ia, ia_min);
+ ia_max = max(ia, ia_max);
+
+ /* Count the number of usable MDS's */
+ if (tgt->ltd_qos.ltq_svr->lsq_bavail == 0)
+ qos->lq_active_svr_count++;
+ tgt->ltd_qos.ltq_svr->lsq_bavail += ba;
+ tgt->ltd_qos.ltq_svr->lsq_iavail += ia;
+
+ /*
+ * per-MDT penalty is
+ * prio * bavail * iavail / (num_tgt - 1) / 2
+ */
+ tgt->ltd_qos.ltq_penalty_per_obj = prio_wide * ba * ia;
+ do_div(tgt->ltd_qos.ltq_penalty_per_obj, num_active - 1);
+ tgt->ltd_qos.ltq_penalty_per_obj >>= 1;
+
+ age = (now - tgt->ltd_qos.ltq_used) >> 3;
+ if (qos->lq_reset || age > 32 * maxage)
+ tgt->ltd_qos.ltq_penalty = 0;
+ else if (age > maxage)
+ /* Decay tgt penalty. */
+ tgt->ltd_qos.ltq_penalty >>= (age / maxage);
+ }
+
+ num_active = qos->lq_active_svr_count;
+ if (num_active < 2) {
+ /*
+ * If there's only 1 MDS, we can't penalize it, so instead
+ * we have to double the MDT penalty
+ */
+ num_active = 2;
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ tgt = lmv->tgts[i];
+ if (!tgt || !tgt->ltd_exp || !tgt->ltd_active)
+ continue;
+
+ tgt->ltd_qos.ltq_penalty_per_obj <<= 1;
+ }
+ }
+
+ /*
+ * Per-MDS penalty is
+ * prio * bavail * iavail / server_tgts / (num_svr - 1) / 2
+ */
+ list_for_each_entry(svr, &qos->lq_svr_list, lsq_svr_list) {
+ ba = svr->lsq_bavail;
+ ia = svr->lsq_iavail;
+ svr->lsq_penalty_per_obj = prio_wide * ba * ia;
+ do_div(ba, svr->lsq_tgt_count * (num_active - 1));
+ svr->lsq_penalty_per_obj >>= 1;
+
+ age = (now - svr->lsq_used) >> 3;
+ if (qos->lq_reset || age > 32 * maxage)
+ svr->lsq_penalty = 0;
+ else if (age > maxage)
+ /* Decay server penalty. */
+ svr->lsq_penalty >>= age / maxage;
+ }
+
+ qos->lq_dirty = 0;
+ qos->lq_reset = 0;
+
+ /*
+ * If each MDT has almost same free space, do rr allocation for better
+ * creation performance
+ */
+ qos->lq_same_space = 0;
+ if ((ba_max * (256 - qos->lq_threshold_rr)) >> 8 < ba_min &&
+ (ia_max * (256 - qos->lq_threshold_rr)) >> 8 < ia_min) {
+ qos->lq_same_space = 1;
+ /* Reset weights for the next time we enter qos mode */
+ qos->lq_reset = 1;
+ }
+ rc = 0;
+
+out:
+ if (!rc && qos->lq_same_space)
+ RETURN(-EAGAIN);
+
+ RETURN(rc);
+}
+
+static inline bool lmv_qos_is_usable(struct lmv_obd *lmv)
+{
+ if (!lmv->lmv_qos.lq_dirty && lmv->lmv_qos.lq_same_space)
+ return false;
+
+ if (lmv->desc.ld_active_tgt_count < 2)
+ return false;
+
+ return true;
+}
+
+/**
+ * Calculate weight for a given MDT.
+ *
+ * The final MDT weight is bavail >> 16 * iavail >> 8 minus the MDT and MDS
+ * penalties. See lmv_qos_calc_ppts() for how penalties are calculated.
+ *
+ * \param[in] tgt MDT target descriptor
+ */
+static void lmv_qos_calc_weight(struct lu_tgt_desc *tgt)
+{
+ struct lu_tgt_qos *ltq = &tgt->ltd_qos;
+ __u64 temp, temp2;
+
+ temp = (tgt_statfs_bavail(tgt) >> 16) * (tgt_statfs_iavail(tgt) >> 8);
+ temp2 = ltq->ltq_penalty + ltq->ltq_svr->lsq_penalty;
+ if (temp < temp2)
+ ltq->ltq_weight = 0;
+ else
+ ltq->ltq_weight = temp - temp2;
+}
+
+/**
+ * Re-calculate weights.
+ *
+ * The function is called when some target was used for a new object. In
+ * this case we should re-calculate all the weights to keep new allocations
+ * balanced well.
+ *
+ * \param[in] lmv LMV device
+ * \param[in] tgt target where a new object was placed
+ * \param[out] total_wt new total weight for the pool
+ *
+ * \retval 0
+ */
+static int lmv_qos_used(struct lmv_obd *lmv, struct lu_tgt_desc *tgt,
+ __u64 *total_wt)
+{
+ struct lu_tgt_qos *ltq;
+ struct lu_svr_qos *svr;
+ unsigned int i;
+
+ ENTRY;
+
+ ltq = &tgt->ltd_qos;
+ LASSERT(ltq);
+
+ /* Don't allocate on this device anymore, until the next alloc_qos */
+ ltq->ltq_usable = 0;
+
+ svr = ltq->ltq_svr;
+
+ /*
+ * Decay old penalty by half (we're adding max penalty, and don't
+ * want it to run away.)
+ */
+ ltq->ltq_penalty >>= 1;
+ svr->lsq_penalty >>= 1;
+
+ /* mark the MDS and MDT as recently used */
+ ltq->ltq_used = svr->lsq_used = ktime_get_real_seconds();
+
+ /* Set max penalties for this MDT and MDS */
+ ltq->ltq_penalty += ltq->ltq_penalty_per_obj *
+ lmv->desc.ld_active_tgt_count;
+ svr->lsq_penalty += svr->lsq_penalty_per_obj *
+ lmv->lmv_qos.lq_active_svr_count;
+
+ /* Decrease all MDS penalties */
+ list_for_each_entry(svr, &lmv->lmv_qos.lq_svr_list, lsq_svr_list) {
+ if (svr->lsq_penalty < svr->lsq_penalty_per_obj)
+ svr->lsq_penalty = 0;
+ else
+ svr->lsq_penalty -= svr->lsq_penalty_per_obj;
+ }
+
+ *total_wt = 0;
+ /* Decrease all MDT penalties */
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ ltq = &lmv->tgts[i]->ltd_qos;
+ if (!tgt || !tgt->ltd_exp || !tgt->ltd_active)
+ continue;
+
+ if (ltq->ltq_penalty < ltq->ltq_penalty_per_obj)
+ ltq->ltq_penalty = 0;
+ else
+ ltq->ltq_penalty -= ltq->ltq_penalty_per_obj;
+
+ lmv_qos_calc_weight(lmv->tgts[i]);
+
+ /* Recalc the total weight of usable osts */
+ if (ltq->ltq_usable)
+ *total_wt += ltq->ltq_weight;
+
+ CDEBUG(D_OTHER, "recalc tgt %d usable=%d avail=%llu"
+ " tgtppo=%llu tgtp=%llu svrppo=%llu"
+ " svrp=%llu wt=%llu\n",
+ i, ltq->ltq_usable,
+ tgt_statfs_bavail(tgt) >> 10,
+ ltq->ltq_penalty_per_obj >> 10,
+ ltq->ltq_penalty >> 10,
+ ltq->ltq_svr->lsq_penalty_per_obj >> 10,
+ ltq->ltq_svr->lsq_penalty >> 10,
+ ltq->ltq_weight >> 10);
+ }
+
+ RETURN(0);
+}
+
+struct lu_tgt_desc *lmv_locate_tgt_qos(struct lmv_obd *lmv, __u32 *mdt)
+{
+ struct lu_tgt_desc *tgt;
+ __u64 total_weight = 0;
+ __u64 cur_weight = 0;
+ __u64 rand;
+ int i;
+ int rc;
+
+ ENTRY;
+
+ if (!lmv_qos_is_usable(lmv))
+ RETURN(ERR_PTR(-EAGAIN));
+
+ down_write(&lmv->lmv_qos.lq_rw_sem);
+
+ if (!lmv_qos_is_usable(lmv))
+ GOTO(unlock, tgt = ERR_PTR(-EAGAIN));
+
+ rc = lmv_qos_calc_ppts(lmv);
+ if (rc)
+ GOTO(unlock, tgt = ERR_PTR(rc));
+
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ tgt = lmv->tgts[i];
+ if (!tgt)
+ continue;
+
+ tgt->ltd_qos.ltq_usable = 0;
+ if (!tgt->ltd_exp || !tgt->ltd_active)
+ continue;
+
+ tgt->ltd_qos.ltq_usable = 1;
+ lmv_qos_calc_weight(tgt);
+ total_weight += tgt->ltd_qos.ltq_weight;
+ }
+
+ if (total_weight) {
+#if BITS_PER_LONG == 32
+ rand = cfs_rand() % (unsigned int)total_weight;
+ /*
+ * If total_weight > 32-bit, first generate the high
+ * 32 bits of the random number, then add in the low
+ * 32 bits (truncated to the upper limit, if needed)
+ */
+ if (total_weight > 0xffffffffULL)
+ rand = (__u64)(cfs_rand() %
+ (unsigned int)(total_weight >> 32)) << 32;
+ else
+ rand = 0;
+
+ if (rand == (total_weight & 0xffffffff00000000ULL))
+ rand |= cfs_rand() % (unsigned int)total_weight;
+ else
+ rand |= cfs_rand();
+
+#else
+ rand = ((__u64)cfs_rand() << 32 | cfs_rand()) % total_weight;
+#endif
+ } else {
+ rand = 0;
+ }
+
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ tgt = lmv->tgts[i];
+
+ if (!tgt || !tgt->ltd_qos.ltq_usable)
+ continue;
+
+ cur_weight += tgt->ltd_qos.ltq_weight;
+ if (cur_weight < rand)
+ continue;
+
+ *mdt = tgt->ltd_index;
+ lmv_qos_used(lmv, tgt, &total_weight);
+ GOTO(unlock, rc = 0);
+ }
+
+ /* no proper target found */
+ GOTO(unlock, tgt = ERR_PTR(-EAGAIN));
+unlock:
+ up_write(&lmv->lmv_qos.lq_rw_sem);
+
+ return tgt;
+}
+
+struct lu_tgt_desc *lmv_locate_tgt_rr(struct lmv_obd *lmv, __u32 *mdt)
+{
+ struct lu_tgt_desc *tgt;
+ int i;
+
+ ENTRY;
+
+ spin_lock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ tgt = lmv->tgts[(i + lmv->lmv_qos_rr_index) %
+ lmv->desc.ld_tgt_count];
+ if (tgt && tgt->ltd_exp && tgt->ltd_active) {
+ *mdt = tgt->ltd_index;
+ lmv->lmv_qos_rr_index =
+ (i + lmv->lmv_qos_rr_index + 1) %
+ lmv->desc.ld_tgt_count;
+ spin_unlock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+
+ RETURN(tgt);
+ }
+ }
+ spin_unlock(&lmv->lmv_qos.lq_rr.lqr_alloc);
+
+ RETURN(ERR_PTR(-ENODEV));
+}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff