X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Flod%2Flod_qos.c;h=644ced9673adcdcd21ff04eadccefb3e54e1ed2e;hp=8a4d1bbb82df2cde5b35d36c628450c64f4ccf19;hb=68825cdf81688faa8d090a3a5f2321b66c2cec63;hpb=a61ff5914bf6b784b22eea0625bb8b61d7bea950 diff --git a/lustre/lod/lod_qos.c b/lustre/lod/lod_qos.c index 8a4d1bb..644ced9 100644 --- a/lustre/lod/lod_qos.c +++ b/lustre/lod/lod_qos.c @@ -6,13 +6,13 @@ * 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 COPYING file that accompanied this code. - + * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA @@ -23,7 +23,7 @@ * Copyright 2009 Sun Microsystems, Inc. All rights reserved * Use is subject to license terms. * - * Copyright (c) 2012, 2013, Intel Corporation. + * Copyright (c) 2012, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -31,14 +31,20 @@ * * lustre/lod/lod_qos.c * + * Implementation of different allocation algorithm used + * to distribute objects and data among OSTs. */ #define DEBUG_SUBSYSTEM S_LOV +#include +#include + #include +#include +#include #include -#include -#include + #include "lod_internal.h" /* @@ -48,407 +54,212 @@ #define D_QOS D_OTHER -#if 0 -#define QOS_DEBUG(fmt, ...) CDEBUG(D_OTHER, fmt, ## __VA_ARGS__) -#define QOS_CONSOLE(fmt, ...) LCONSOLE(D_OTHER, fmt, ## __VA_ARGS__) -#else -#define QOS_DEBUG(fmt, ...) -#define QOS_CONSOLE(fmt, ...) -#endif +#define QOS_DEBUG(fmt, ...) CDEBUG(D_QOS, fmt, ## __VA_ARGS__) +#define QOS_CONSOLE(fmt, ...) LCONSOLE(D_QOS, fmt, ## __VA_ARGS__) #define TGT_BAVAIL(i) (OST_TGT(lod,i)->ltd_statfs.os_bavail * \ OST_TGT(lod,i)->ltd_statfs.os_bsize) -int qos_add_tgt(struct lod_device *lod, struct lod_tgt_desc *ost_desc) +static inline int lod_statfs_check(struct lu_tgt_descs *ltd, + struct lu_tgt_desc *tgt) { - struct lov_qos_oss *oss = NULL, *temposs; - struct obd_export *exp = ost_desc->ltd_exp; - int rc = 0, found = 0; - cfs_list_t *list; - ENTRY; - - down_write(&lod->lod_qos.lq_rw_sem); - /* - * a bit hacky approach to learn NID of corresponding connection - * but there is no official API to access information like this - * with OSD API. - */ - cfs_list_for_each_entry(oss, &lod->lod_qos.lq_oss_list, lqo_oss_list) { - if (obd_uuid_equals(&oss->lqo_uuid, - &exp->exp_connection->c_remote_uuid)) { - found++; - break; - } - } - - if (!found) { - OBD_ALLOC_PTR(oss); - if (!oss) - GOTO(out, rc = -ENOMEM); - memcpy(&oss->lqo_uuid, &exp->exp_connection->c_remote_uuid, - sizeof(oss->lqo_uuid)); - } else { - /* Assume we have to move this one */ - cfs_list_del(&oss->lqo_oss_list); - } - - oss->lqo_ost_count++; - ost_desc->ltd_qos.ltq_oss = oss; - - CDEBUG(D_QOS, "add tgt %s to OSS %s (%d OSTs)\n", - obd_uuid2str(&ost_desc->ltd_uuid), obd_uuid2str(&oss->lqo_uuid), - oss->lqo_ost_count); - - /* Add sorted by # of OSTs. Find the first entry that we're - bigger than... */ - list = &lod->lod_qos.lq_oss_list; - cfs_list_for_each_entry(temposs, list, lqo_oss_list) { - if (oss->lqo_ost_count > temposs->lqo_ost_count) - break; - } - /* ...and add before it. If we're the first or smallest, temposs - points to the list head, and we add to the end. */ - cfs_list_add_tail(&oss->lqo_oss_list, &temposs->lqo_oss_list); + struct obd_statfs *sfs = &tgt->ltd_statfs; - lod->lod_qos.lq_dirty = 1; - lod->lod_qos.lq_rr.lqr_dirty = 1; + if (((sfs->os_state & OS_STATFS_ENOSPC) || + (!ltd->ltd_is_mdt && sfs->os_state & OS_STATFS_ENOINO && + sfs->os_fprecreated == 0))) + return -ENOSPC; -out: - up_write(&lod->lod_qos.lq_rw_sem); - RETURN(rc); -} - -int qos_del_tgt(struct lod_device *lod, struct lod_tgt_desc *ost_desc) -{ - struct lov_qos_oss *oss; - int rc = 0; - ENTRY; + /* If the OST is readonly then we can't allocate objects there */ + if (sfs->os_state & OS_STATFS_READONLY) + return -EROFS; - down_write(&lod->lod_qos.lq_rw_sem); - oss = ost_desc->ltd_qos.ltq_oss; - if (!oss) - GOTO(out, rc = -ENOENT); - - oss->lqo_ost_count--; - if (oss->lqo_ost_count == 0) { - CDEBUG(D_QOS, "removing OSS %s\n", - obd_uuid2str(&oss->lqo_uuid)); - cfs_list_del(&oss->lqo_oss_list); - ost_desc->ltd_qos.ltq_oss = NULL; - OBD_FREE_PTR(oss); - } + /* object precreation is skipped on the OST with max_create_count=0 */ + if (!ltd->ltd_is_mdt && sfs->os_state & OS_STATFS_NOPRECREATE) + return -ENOBUFS; - lod->lod_qos.lq_dirty = 1; - lod->lod_qos.lq_rr.lqr_dirty = 1; -out: - up_write(&lod->lod_qos.lq_rw_sem); - RETURN(rc); + return 0; } +/** + * Check whether the target is available for new objects. + * + * Request statfs data from the given target and verify it's active and not + * read-only. If so, then it can be used to place new objects. This + * function also maintains the number of active/inactive targets and sets + * dirty flags if those numbers change so others can run re-balance procedures. + * No external locking is required. + * + * \param[in] env execution environment for this thread + * \param[in] d LOD device + * \param[in] ltd target table + * \param[in] tgt target + * + * \retval 0 if the target is good + * \retval negative negated errno on error + */ static int lod_statfs_and_check(const struct lu_env *env, struct lod_device *d, - int index, struct obd_statfs *sfs) + struct lu_tgt_descs *ltd, + struct lu_tgt_desc *tgt, __u64 reserve) { - struct lod_tgt_desc *ost; - int rc; + struct obd_statfs_info info = { 0 }; + struct lov_desc *desc = <d->ltd_lov_desc; + int rc; + ENTRY; LASSERT(d); - ost = OST_TGT(d,index); - LASSERT(ost); + LASSERT(tgt); - rc = dt_statfs(env, ost->ltd_ost, sfs); + info.os_enable_pre = 1; + rc = dt_statfs_info(env, tgt->ltd_tgt, &tgt->ltd_statfs, &info); if (rc && rc != -ENOTCONN) CERROR("%s: statfs: rc = %d\n", lod2obd(d)->obd_name, rc); + if (!rc) { + rc = lod_statfs_check(ltd, tgt); + if (rc == -ENOSPC) + return rc; + } + + if (reserve && + (reserve + (info.os_reserved_mb_low << 20) > + tgt->ltd_statfs.os_bavail * tgt->ltd_statfs.os_bsize)) + return -ENOSPC; + /* check whether device has changed state (active, inactive) */ - if (rc != 0 && ost->ltd_active) { + if (rc != 0 && tgt->ltd_active) { /* turned inactive? */ - spin_lock(&d->lod_desc_lock); - if (ost->ltd_active) { - ost->ltd_active = 0; - LASSERT(d->lod_desc.ld_active_tgt_count > 0); - d->lod_desc.ld_active_tgt_count--; - d->lod_qos.lq_dirty = 1; - d->lod_qos.lq_rr.lqr_dirty = 1; + spin_lock(&d->lod_lock); + if (tgt->ltd_active) { + tgt->ltd_active = 0; + if (rc == -ENOTCONN) + tgt->ltd_connecting = 1; + + LASSERT(desc->ld_active_tgt_count > 0); + desc->ld_active_tgt_count--; + ltd->ltd_qos.lq_dirty = 1; + ltd->ltd_qos.lq_rr.lqr_dirty = 1; CDEBUG(D_CONFIG, "%s: turns inactive\n", - ost->ltd_exp->exp_obd->obd_name); + tgt->ltd_exp->exp_obd->obd_name); } - spin_unlock(&d->lod_desc_lock); - } else if (rc == 0 && ost->ltd_active == 0) { + spin_unlock(&d->lod_lock); + } else if (rc == 0 && tgt->ltd_active == 0) { /* turned active? */ - LASSERTF(d->lod_desc.ld_active_tgt_count < d->lod_ostnr, - "active tgt count %d, ost nr %d\n", - d->lod_desc.ld_active_tgt_count, d->lod_ostnr); - spin_lock(&d->lod_desc_lock); - if (ost->ltd_active == 0) { - ost->ltd_active = 1; - d->lod_desc.ld_active_tgt_count++; - d->lod_qos.lq_dirty = 1; - d->lod_qos.lq_rr.lqr_dirty = 1; + LASSERTF(desc->ld_active_tgt_count < desc->ld_tgt_count, + "active tgt count %d, tgt nr %d\n", + desc->ld_active_tgt_count, desc->ld_tgt_count); + spin_lock(&d->lod_lock); + if (tgt->ltd_active == 0) { + tgt->ltd_active = 1; + tgt->ltd_connecting = 0; + desc->ld_active_tgt_count++; + ltd->ltd_qos.lq_dirty = 1; + ltd->ltd_qos.lq_rr.lqr_dirty = 1; CDEBUG(D_CONFIG, "%s: turns active\n", - ost->ltd_exp->exp_obd->obd_name); + tgt->ltd_exp->exp_obd->obd_name); } - spin_unlock(&d->lod_desc_lock); + spin_unlock(&d->lod_lock); } - - RETURN(rc); -} - -static void lod_qos_statfs_update(const struct lu_env *env, - struct lod_device *lod) -{ - struct obd_device *obd = lod2obd(lod); - struct ost_pool *osts = &(lod->lod_pool_info); - int i, idx, rc = 0; - __u64 max_age, avail; - ENTRY; - - max_age = cfs_time_shift_64(-2 * lod->lod_desc.ld_qos_maxage); - - if (cfs_time_beforeq_64(max_age, obd->obd_osfs_age)) - /* statfs data are quite recent, don't need to refresh it */ - RETURN_EXIT; - - down_write(&lod->lod_qos.lq_rw_sem); - if (cfs_time_beforeq_64(max_age, obd->obd_osfs_age)) - GOTO(out, rc = 0); - - for (i = 0; i < osts->op_count; i++) { - idx = osts->op_array[i]; - avail = OST_TGT(lod,idx)->ltd_statfs.os_bavail; - rc = lod_statfs_and_check(env, lod, idx, - &OST_TGT(lod,idx)->ltd_statfs); - if (rc) - break; - if (OST_TGT(lod,idx)->ltd_statfs.os_bavail != avail) - /* recalculate weigths */ - lod->lod_qos.lq_dirty = 1; + if (rc == -ENOTCONN) { + /* In case that the ENOTCONN for inactive OST state is + * mistreated as MDT disconnection state by the client, + * this error should be changed to someone else. + */ + rc = -EREMOTEIO; } - obd->obd_osfs_age = cfs_time_current_64(); -out: - up_write(&lod->lod_qos.lq_rw_sem); - EXIT; + RETURN(rc); } -/* Recalculate per-object penalties for OSSs and OSTs, - depends on size of each ost in an oss */ -static int lod_qos_calc_ppo(struct lod_device *lod) +static int lod_is_tgt_usable(struct lu_tgt_descs *ltd, struct lu_tgt_desc *tgt) { - struct lov_qos_oss *oss; - __u64 ba_max, ba_min, temp; - __u32 num_active; - int rc, i, prio_wide; - time_t now, age; - ENTRY; - - if (!lod->lod_qos.lq_dirty) - GOTO(out, rc = 0); - - num_active = lod->lod_desc.ld_active_tgt_count - 1; - if (num_active < 1) - GOTO(out, rc = -EAGAIN); - - /* find bavail on each OSS */ - cfs_list_for_each_entry(oss, &lod->lod_qos.lq_oss_list, lqo_oss_list) - oss->lqo_bavail = 0; - lod->lod_qos.lq_active_oss_count = 0; + int rc; - /* - * How badly user wants to select OSTs "widely" (not recently chosen - * and not on recent OSS's). As opposed to "freely" (free space - * avail.) 0-256 - */ - prio_wide = 256 - lod->lod_qos.lq_prio_free; - - ba_min = (__u64)(-1); - ba_max = 0; - now = cfs_time_current_sec(); - /* Calculate OST penalty per object - * (lod ref taken in lod_qos_prep_create()) */ - cfs_foreach_bit(lod->lod_ost_bitmap, i) { - LASSERT(OST_TGT(lod,i)); - temp = TGT_BAVAIL(i); - if (!temp) - continue; - ba_min = min(temp, ba_min); - ba_max = max(temp, ba_max); - - /* Count the number of usable OSS's */ - if (OST_TGT(lod,i)->ltd_qos.ltq_oss->lqo_bavail == 0) - lod->lod_qos.lq_active_oss_count++; - OST_TGT(lod,i)->ltd_qos.ltq_oss->lqo_bavail += temp; - - /* per-OST penalty is prio * TGT_bavail / (num_ost - 1) / 2 */ - temp >>= 1; - lov_do_div64(temp, num_active); - OST_TGT(lod,i)->ltd_qos.ltq_penalty_per_obj = - (temp * prio_wide) >> 8; - - age = (now - OST_TGT(lod,i)->ltd_qos.ltq_used) >> 3; - if (lod->lod_qos.lq_reset || - age > 32 * lod->lod_desc.ld_qos_maxage) - OST_TGT(lod,i)->ltd_qos.ltq_penalty = 0; - else if (age > lod->lod_desc.ld_qos_maxage) - /* 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 OSS/OSTs under light load. */ - OST_TGT(lod,i)->ltd_qos.ltq_penalty >>= - (age / lod->lod_desc.ld_qos_maxage); - } - - num_active = lod->lod_qos.lq_active_oss_count - 1; - if (num_active < 1) { - /* If there's only 1 OSS, we can't penalize it, so instead - we have to double the OST penalty */ - num_active = 1; - cfs_foreach_bit(lod->lod_ost_bitmap, i) - OST_TGT(lod,i)->ltd_qos.ltq_penalty_per_obj <<= 1; - } - - /* Per-OSS penalty is prio * oss_avail / oss_osts / (num_oss - 1) / 2 */ - cfs_list_for_each_entry(oss, &lod->lod_qos.lq_oss_list, lqo_oss_list) { - temp = oss->lqo_bavail >> 1; - lov_do_div64(temp, oss->lqo_ost_count * num_active); - oss->lqo_penalty_per_obj = (temp * prio_wide) >> 8; - - age = (now - oss->lqo_used) >> 3; - if (lod->lod_qos.lq_reset || - age > 32 * lod->lod_desc.ld_qos_maxage) - oss->lqo_penalty = 0; - else if (age > lod->lod_desc.ld_qos_maxage) - /* 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 OSS/OSTs under light load. */ - oss->lqo_penalty >>= age / lod->lod_desc.ld_qos_maxage; - } - - lod->lod_qos.lq_dirty = 0; - lod->lod_qos.lq_reset = 0; - - /* If each ost has almost same free space, - * do rr allocation for better creation performance */ - lod->lod_qos.lq_same_space = 0; - if ((ba_max * (256 - lod->lod_qos.lq_threshold_rr)) >> 8 < ba_min) { - lod->lod_qos.lq_same_space = 1; - /* Reset weights for the next time we enter qos mode */ - lod->lod_qos.lq_reset = 1; - } - rc = 0; + rc = lod_statfs_check(ltd, tgt); + if (rc) + return rc; -out: -#ifndef FORCE_QOS - if (!rc && lod->lod_qos.lq_same_space) - RETURN(-EAGAIN); -#endif - RETURN(rc); -} + if (!tgt->ltd_active) + return -ENOTCONN; -static int lod_qos_calc_weight(struct lod_device *lod, int i) -{ - __u64 temp, temp2; - - /* Final ost weight = TGT_BAVAIL - ost_penalty - oss_penalty */ - temp = TGT_BAVAIL(i); - temp2 = OST_TGT(lod,i)->ltd_qos.ltq_penalty + - OST_TGT(lod,i)->ltd_qos.ltq_oss->lqo_penalty; - if (temp < temp2) - OST_TGT(lod,i)->ltd_qos.ltq_weight = 0; - else - OST_TGT(lod,i)->ltd_qos.ltq_weight = temp - temp2; return 0; } -/* We just used this index for a stripe; adjust everyone's weights */ -static int lod_qos_used(struct lod_device *lod, struct ost_pool *osts, - __u32 index, __u64 *total_wt) +/** + * Maintain per-target statfs data. + * + * The function refreshes statfs data for all the targets every N seconds. + * The actual N is controlled via procfs and set to LOV_DESC_QOS_MAXAGE_DEFAULT + * initially. + * + * \param[in] env execution environment for this thread + * \param[in] lod LOD device + * \param[in] ltd tgt table + */ +void lod_qos_statfs_update(const struct lu_env *env, struct lod_device *lod, + struct lu_tgt_descs *ltd) { - struct lod_tgt_desc *ost; - struct lov_qos_oss *oss; - int j; + struct obd_device *obd = lod2obd(lod); + struct lu_tgt_desc *tgt; + time64_t max_age; + u64 avail; ENTRY; - ost = OST_TGT(lod,index); - LASSERT(ost); - - /* Don't allocate on this devuce anymore, until the next alloc_qos */ - ost->ltd_qos.ltq_usable = 0; - - oss = ost->ltd_qos.ltq_oss; + max_age = ktime_get_seconds() - 2 * ltd->ltd_lov_desc.ld_qos_maxage; - /* Decay old penalty by half (we're adding max penalty, and don't - want it to run away.) */ - ost->ltd_qos.ltq_penalty >>= 1; - oss->lqo_penalty >>= 1; - - /* mark the OSS and OST as recently used */ - ost->ltd_qos.ltq_used = oss->lqo_used = cfs_time_current_sec(); - - /* Set max penalties for this OST and OSS */ - ost->ltd_qos.ltq_penalty += - ost->ltd_qos.ltq_penalty_per_obj * lod->lod_ostnr; - oss->lqo_penalty += oss->lqo_penalty_per_obj * - lod->lod_qos.lq_active_oss_count; + if (obd->obd_osfs_age > max_age) + /* statfs data are quite recent, don't need to refresh it */ + RETURN_EXIT; - /* Decrease all OSS penalties */ - cfs_list_for_each_entry(oss, &lod->lod_qos.lq_oss_list, lqo_oss_list) { - if (oss->lqo_penalty < oss->lqo_penalty_per_obj) - oss->lqo_penalty = 0; - else - oss->lqo_penalty -= oss->lqo_penalty_per_obj; - } + down_write(<d->ltd_qos.lq_rw_sem); - *total_wt = 0; - /* Decrease all OST penalties */ - for (j = 0; j < osts->op_count; j++) { - int i; + if (obd->obd_osfs_age > max_age) + goto out; - i = osts->op_array[j]; - if (!cfs_bitmap_check(lod->lod_ost_bitmap, i)) + ltd_foreach_tgt(ltd, tgt) { + avail = tgt->ltd_statfs.os_bavail; + if (lod_statfs_and_check(env, lod, ltd, tgt, 0)) continue; - ost = OST_TGT(lod,i); - LASSERT(ost); - - if (ost->ltd_qos.ltq_penalty < - ost->ltd_qos.ltq_penalty_per_obj) - ost->ltd_qos.ltq_penalty = 0; - else - ost->ltd_qos.ltq_penalty -= - ost->ltd_qos.ltq_penalty_per_obj; - - lod_qos_calc_weight(lod, i); - - /* Recalc the total weight of usable osts */ - if (ost->ltd_qos.ltq_usable) - *total_wt += ost->ltd_qos.ltq_weight; - - QOS_DEBUG("recalc tgt %d usable=%d avail="LPU64 - " ostppo="LPU64" ostp="LPU64" ossppo="LPU64 - " ossp="LPU64" wt="LPU64"\n", - i, ost->ltd_qos.ltq_usable, TGT_BAVAIL(i) >> 10, - ost->ltd_qos.ltq_penalty_per_obj >> 10, - ost->ltd_qos.ltq_penalty >> 10, - ost->ltd_qos.ltq_oss->lqo_penalty_per_obj >> 10, - ost->ltd_qos.ltq_oss->lqo_penalty >> 10, - ost->ltd_qos.ltq_weight >> 10); + if (tgt->ltd_statfs.os_bavail != avail) + /* recalculate weigths */ + ltd->ltd_qos.lq_dirty = 1; } + obd->obd_osfs_age = ktime_get_seconds(); - RETURN(0); +out: + up_write(<d->ltd_qos.lq_rw_sem); + EXIT; } #define LOV_QOS_EMPTY ((__u32)-1) -/* compute optimal round-robin order, based on OSTs per OSS */ -static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, - struct lov_qos_rr *lqr) + +/** + * Calculate optimal round-robin order with regard to OSSes. + * + * Place all the OSTs from pool \a src_pool in a special array to be used for + * round-robin (RR) stripe allocation. The placement algorithm interleaves + * OSTs from the different OSSs so that RR allocation can balance OSSs evenly. + * Resorts the targets when the number of active targets changes (because of + * a new target or activation/deactivation). + * + * \param[in] lod LOD device + * \param[in] ltd tgt table + * \param[in] src_pool tgt pool + * \param[in] lqr round-robin list + * + * \retval 0 on success + * \retval -ENOMEM fails to allocate the array + */ +static int lod_qos_calc_rr(struct lod_device *lod, struct lu_tgt_descs *ltd, + const struct lu_tgt_pool *src_pool, + struct lu_qos_rr *lqr) { - struct lov_qos_oss *oss; - struct lod_tgt_desc *ost; + struct lu_svr_qos *svr; + struct lu_tgt_desc *tgt; unsigned placed, real_count; - int i, rc; + unsigned int i; + int rc; ENTRY; if (!lqr->lqr_dirty) { @@ -457,7 +268,7 @@ static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, } /* Do actual allocation. */ - down_write(&lod->lod_qos.lq_rw_sem); + down_write(<d->ltd_qos.lq_rw_sem); /* * Check again. While we were sleeping on @lq_rw_sem something could @@ -465,7 +276,7 @@ static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, */ if (!lqr->lqr_dirty) { LASSERT(lqr->lqr_pool.op_size); - up_write(&lod->lod_qos.lq_rw_sem); + up_write(<d->ltd_qos.lq_rw_sem); RETURN(0); } @@ -476,33 +287,33 @@ static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, deleting from the pool. The lq_rw_sem insures that nobody else is reading. */ lqr->lqr_pool.op_count = real_count; - rc = lod_ost_pool_extend(&lqr->lqr_pool, real_count); + rc = tgt_pool_extend(&lqr->lqr_pool, real_count); if (rc) { - up_write(&lod->lod_qos.lq_rw_sem); + up_write(<d->ltd_qos.lq_rw_sem); RETURN(rc); } for (i = 0; i < lqr->lqr_pool.op_count; i++) lqr->lqr_pool.op_array[i] = LOV_QOS_EMPTY; - /* Place all the OSTs from 1 OSS at the same time. */ + /* Place all the tgts from 1 svr at the same time. */ placed = 0; - cfs_list_for_each_entry(oss, &lod->lod_qos.lq_oss_list, lqo_oss_list) { + list_for_each_entry(svr, <d->ltd_qos.lq_svr_list, lsq_svr_list) { int j = 0; for (i = 0; i < lqr->lqr_pool.op_count; i++) { int next; - if (!cfs_bitmap_check(lod->lod_ost_bitmap, - src_pool->op_array[i])) + if (!test_bit(src_pool->op_array[i], + ltd->ltd_tgt_bitmap)) continue; - ost = OST_TGT(lod,src_pool->op_array[i]); - LASSERT(ost && ost->ltd_ost); - if (ost->ltd_qos.ltq_oss != oss) + tgt = LTD_TGT(ltd, src_pool->op_array[i]); + LASSERT(tgt && tgt->ltd_tgt); + if (tgt->ltd_qos.ltq_svr != svr) continue; - /* Evenly space these OSTs across arrayspace */ - next = j * lqr->lqr_pool.op_count / oss->lqo_ost_count; + /* Evenly space these tgts across arrayspace */ + next = j * lqr->lqr_pool.op_count / svr->lsq_tgt_count; while (lqr->lqr_pool.op_array[next] != LOV_QOS_EMPTY) next = (next + 1) % lqr->lqr_pool.op_count; @@ -513,15 +324,15 @@ static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, } lqr->lqr_dirty = 0; - up_write(&lod->lod_qos.lq_rw_sem); + up_write(<d->ltd_qos.lq_rw_sem); if (placed != real_count) { /* This should never happen */ - LCONSOLE_ERROR_MSG(0x14e, "Failed to place all OSTs in the " + LCONSOLE_ERROR_MSG(0x14e, "Failed to place all tgts in the " "round-robin list (%d of %d).\n", placed, real_count); for (i = 0; i < lqr->lqr_pool.op_count; i++) { - LCONSOLE(D_WARNING, "rr #%d ost idx=%d\n", i, + LCONSOLE(D_WARNING, "rr #%d tgt idx=%d\n", i, lqr->lqr_pool.op_array[i]); } lqr->lqr_dirty = 1; @@ -537,19 +348,28 @@ static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool, } /** - * A helper function to: - * create in-core lu object on the specified OSP - * declare creation of the object - * IMPORTANT: at this stage object is anonymouos - it has no fid assigned - * this is a workaround till we have natural FIDs on OST - * - * at this point we want to declare (reserve) object for us as - * we can't block at execution (when create method is called). - * otherwise we'd block whole transaction batch + * Instantiate and declare creation of a new object. + * + * The function instantiates LU representation for a new object on the + * specified device. Also it declares an intention to create that + * object on the storage target. + * + * Note lu_object_anon() is used which is a trick with regard to LU/OSD + * infrastructure - in the existing precreation framework we can't assign FID + * at this moment, we do this later once a transaction is started. So the + * special method instantiates FID-less object in the cache and later it + * will get a FID and proper placement in LU cache. + * + * \param[in] env execution environment for this thread + * \param[in] d LOD device + * \param[in] ost_idx OST target index where the object is being created + * \param[in] th transaction handle + * + * \retval object ptr on success, ERR_PTR() otherwise */ static struct dt_object *lod_qos_declare_object_on(const struct lu_env *env, struct lod_device *d, - int ost_idx, + __u32 ost_idx, struct thandle *th) { struct lod_tgt_desc *ost; @@ -560,13 +380,12 @@ static struct dt_object *lod_qos_declare_object_on(const struct lu_env *env, ENTRY; LASSERT(d); - LASSERT(ost_idx >= 0); - LASSERT(ost_idx < d->lod_osts_size); + LASSERT(ost_idx < d->lod_ost_descs.ltd_tgts_size); ost = OST_TGT(d,ost_idx); LASSERT(ost); - LASSERT(ost->ltd_ost); + LASSERT(ost->ltd_tgt); - nd = &ost->ltd_ost->dd_lu_dev; + nd = &ost->ltd_tgt->dd_lu_dev; /* * allocate anonymous object with zero fid, real fid @@ -575,7 +394,7 @@ static struct dt_object *lod_qos_declare_object_on(const struct lu_env *env, */ o = lu_object_anon(env, nd, NULL); if (IS_ERR(o)) - GOTO(out, dt = ERR_PTR(PTR_ERR(o))); + GOTO(out, dt = ERR_CAST(o)); n = lu_object_locate(o->lo_header, nd->ld_type); if (unlikely(n == NULL)) { @@ -586,8 +405,8 @@ static struct dt_object *lod_qos_declare_object_on(const struct lu_env *env, dt = container_of(n, struct dt_object, do_lu); - rc = dt_declare_create(env, dt, NULL, NULL, NULL, th); - if (rc) { + rc = lod_sub_declare_create(env, dt, NULL, NULL, NULL, th); + if (rc < 0) { CDEBUG(D_OTHER, "can't declare creation on #%u: %d\n", ost_idx, rc); lu_object_put(env, o); @@ -598,137 +417,398 @@ out: RETURN(dt); } -static int min_stripe_count(int stripe_cnt, int flags) +/** + * Calculate a minimum acceptable stripe count. + * + * Return an acceptable stripe count depending on flag LOV_USES_DEFAULT_STRIPE: + * all stripes or 3/4 of stripes. + * + * \param[in] stripe_count number of stripes requested + * \param[in] flags 0 or LOV_USES_DEFAULT_STRIPE + * + * \retval acceptable stripecount + */ +static int min_stripe_count(__u32 stripe_count, int flags) { return (flags & LOV_USES_DEFAULT_STRIPE ? - stripe_cnt - (stripe_cnt / 4) : stripe_cnt); + stripe_count - (stripe_count / 4) : stripe_count); } #define LOV_CREATE_RESEED_MULT 30 #define LOV_CREATE_RESEED_MIN 2000 -static int inline lod_qos_dev_is_full(struct obd_statfs *msfs) -{ - __u64 used; - int bs = msfs->os_bsize; - - LASSERT(((bs - 1) & bs) == 0); - - /* the minimum of 0.1% used blocks and 1GB bytes. */ - used = min_t(__u64, (msfs->os_blocks - msfs->os_bfree) >> 10, - 1 << (31 - ffs(bs))); - return (msfs->os_bavail < used); -} - -int lod_ea_store_resize(struct lod_thread_info *info, int size); - -static inline int lod_qos_ost_in_use_clear(const struct lu_env *env, int stripes) +/** + * Initialize temporary tgt-in-use array. + * + * Allocate or extend the array used to mark targets already assigned to a new + * striping so they are not used more than once. + * + * \param[in] env execution environment for this thread + * \param[in] stripes number of items needed in the array + * + * \retval 0 on success + * \retval -ENOMEM on error + */ +static inline int lod_qos_tgt_in_use_clear(const struct lu_env *env, + __u32 stripes) { struct lod_thread_info *info = lod_env_info(env); if (info->lti_ea_store_size < sizeof(int) * stripes) lod_ea_store_resize(info, stripes * sizeof(int)); if (info->lti_ea_store_size < sizeof(int) * stripes) { - CERROR("can't allocate memory for ost-in-use array\n"); + CERROR("can't allocate memory for tgt-in-use array\n"); return -ENOMEM; } memset(info->lti_ea_store, -1, sizeof(int) * stripes); return 0; } -static inline void lod_qos_ost_in_use(const struct lu_env *env, int idx, int ost) +/** + * Remember a target in the array of used targets. + * + * Mark the given target as used for a new striping being created. The status + * of an tgt in a striping can be checked with lod_qos_is_tgt_used(). + * + * \param[in] env execution environment for this thread + * \param[in] idx index in the array + * \param[in] tgt_idx target index to mark as used + */ +static inline void lod_qos_tgt_in_use(const struct lu_env *env, + int idx, int tgt_idx) { struct lod_thread_info *info = lod_env_info(env); - int *osts = info->lti_ea_store; + int *tgts = info->lti_ea_store; LASSERT(info->lti_ea_store_size >= idx * sizeof(int)); - osts[idx] = ost; + tgts[idx] = tgt_idx; } -static int lod_qos_is_ost_used(const struct lu_env *env, int ost, int stripes) +/** + * Check is tgt used in a striping. + * + * Checks whether tgt with the given index is marked as used in the temporary + * array (see lod_qos_tgt_in_use()). + * + * \param[in] env execution environment for this thread + * \param[in] tgt_idx target index to check + * \param[in] stripes the number of items used in the array already + * + * \retval 0 not used + * \retval 1 used + */ +static int lod_qos_is_tgt_used(const struct lu_env *env, int tgt_idx, + __u32 stripes) { struct lod_thread_info *info = lod_env_info(env); - int *osts = info->lti_ea_store; - int j; + int *tgts = info->lti_ea_store; + __u32 j; for (j = 0; j < stripes; j++) { - if (osts[j] == ost) + if (tgts[j] == tgt_idx) return 1; } return 0; } -/* Allocate objects on OSTs with round-robin algorithm */ -static int lod_alloc_rr(const struct lu_env *env, struct lod_object *lo, - struct dt_object **stripe, int flags, - struct thandle *th) +static inline bool +lod_obj_is_ost_use_skip_cb(const struct lu_env *env, struct lod_object *lo, + int comp_idx, struct lod_obj_stripe_cb_data *data) +{ + struct lod_layout_component *comp = &lo->ldo_comp_entries[comp_idx]; + + return comp->llc_ost_indices == NULL; +} + +static inline int +lod_obj_is_ost_use_cb(const struct lu_env *env, struct lod_object *lo, + int comp_idx, struct lod_obj_stripe_cb_data *data) +{ + struct lod_layout_component *comp = &lo->ldo_comp_entries[comp_idx]; + int i; + + for (i = 0; i < comp->llc_stripe_count; i++) { + if (comp->llc_ost_indices[i] == data->locd_ost_index) { + data->locd_ost_index = -1; + return -EEXIST; + } + } + + return 0; +} + +/** + * Check is OST used in a composite layout + * + * \param[in] lo lod object + * \param[in] ost OST target index to check + * + * \retval false not used + * \retval true used + */ +static inline bool lod_comp_is_ost_used(const struct lu_env *env, + struct lod_object *lo, int ost) +{ + struct lod_obj_stripe_cb_data data = { { 0 } }; + + data.locd_ost_index = ost; + data.locd_comp_skip_cb = lod_obj_is_ost_use_skip_cb; + data.locd_comp_cb = lod_obj_is_ost_use_cb; + + (void)lod_obj_for_each_stripe(env, lo, NULL, &data); + + return data.locd_ost_index == -1; +} + +static inline void lod_avoid_update(struct lod_object *lo, + struct lod_avoid_guide *lag) +{ + if (!lod_is_flr(lo)) + return; + + lag->lag_ost_avail--; +} + +static inline bool lod_should_avoid_ost(struct lod_object *lo, + struct lod_avoid_guide *lag, + __u32 index) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lod_tgt_desc *ost = OST_TGT(lod, index); + struct lu_svr_qos *lsq = ost->ltd_qos.ltq_svr; + bool used = false; + int i; + + if (!test_bit(index, lod->lod_ost_bitmap)) { + QOS_DEBUG("OST%d: been used in conflicting mirror component\n", + index); + return true; + } + + /** + * we've tried our best, all available OSTs have been used in + * overlapped components in the other mirror + */ + if (lag->lag_ost_avail == 0) + return false; + + /* check OSS use */ + for (i = 0; i < lag->lag_oaa_count; i++) { + if (lag->lag_oss_avoid_array[i] == lsq->lsq_id) { + used = true; + break; + } + } + /** + * if the OSS which OST[index] resides has not been used, we'd like to + * use it + */ + if (!used) + return false; + + /* if the OSS has been used, check whether the OST has been used */ + if (!cfs_bitmap_check(lag->lag_ost_avoid_bitmap, index)) + used = false; + else + QOS_DEBUG("OST%d: been used in conflicting mirror component\n", + index); + return used; +} + +static int lod_check_and_reserve_ost(const struct lu_env *env, + struct lod_object *lo, + struct lod_layout_component *lod_comp, + __u32 ost_idx, __u32 speed, __u32 *s_idx, + struct dt_object **stripe, + __u32 *ost_indices, + struct thandle *th, + bool *overstriped, + __u64 reserve) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lod_avoid_guide *lag = &lod_env_info(env)->lti_avoid; + struct lu_tgt_desc *ost = OST_TGT(lod, ost_idx); + struct dt_object *o; + __u32 stripe_idx = *s_idx; + int rc; + + ENTRY; + + rc = lod_statfs_and_check(env, lod, &lod->lod_ost_descs, ost, reserve); + if (rc) + RETURN(rc); + + /* + * We expect number of precreated objects in f_ffree at + * the first iteration, skip OSPs with no objects ready + */ + if (ost->ltd_statfs.os_fprecreated == 0 && speed == 0) { + QOS_DEBUG("#%d: precreation is empty\n", ost_idx); + RETURN(rc); + } + + /* + * try to use another OSP if this one is degraded + */ + if (ost->ltd_statfs.os_state & OS_STATFS_DEGRADED && speed < 2) { + QOS_DEBUG("#%d: degraded\n", ost_idx); + RETURN(rc); + } + + /* + * try not allocate on OST which has been used by other + * component + */ + if (speed == 0 && lod_comp_is_ost_used(env, lo, ost_idx)) { + QOS_DEBUG("iter %d: OST%d used by other component\n", + speed, ost_idx); + RETURN(rc); + } + + /** + * try not allocate OSTs used by conflicting component of other mirrors + * for the first and second time. + */ + if (speed < 2 && lod_should_avoid_ost(lo, lag, ost_idx)) { + QOS_DEBUG("iter %d: OST%d used by conflicting mirror component\n", + speed, ost_idx); + RETURN(rc); + } + + /* do not put >1 objects on a single OST, except for overstriping */ + if (lod_qos_is_tgt_used(env, ost_idx, stripe_idx)) { + if (lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING) + *overstriped = true; + else + RETURN(rc); + } + + o = lod_qos_declare_object_on(env, lod, ost_idx, th); + if (IS_ERR(o)) { + CDEBUG(D_OTHER, "can't declare new object on #%u: %d\n", + ost_idx, (int) PTR_ERR(o)); + rc = PTR_ERR(o); + RETURN(rc); + } + + /* + * We've successfully declared (reserved) an object + */ + lod_avoid_update(lo, lag); + lod_qos_tgt_in_use(env, stripe_idx, ost_idx); + stripe[stripe_idx] = o; + ost_indices[stripe_idx] = ost_idx; + OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_LOV_CREATE_RACE, 2); + stripe_idx++; + *s_idx = stripe_idx; + + RETURN(rc); +} + +/** + * Allocate a striping using round-robin algorithm. + * + * Allocates a new striping using round-robin algorithm. The function refreshes + * all the internal structures (statfs cache, array of available OSTs sorted + * with regard to OSS, etc). The number of stripes required is taken from the + * object (must be prepared by the caller), but can change if the flag + * LOV_USES_DEFAULT_STRIPE is supplied. The caller should ensure nobody else + * is trying to create a striping on the object in parallel. All the internal + * structures (like pools, etc) are protected and no additional locking is + * required. The function succeeds even if a single stripe is allocated. To save + * time we give priority to targets which already have objects precreated. + * Full OSTs are skipped (see lod_qos_dev_is_full() for the details). + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[out] stripe striping created + * \param[out] ost_indices ost indices of striping created + * \param[in] flags allocation flags (0 or LOV_USES_DEFAULT_STRIPE) + * \param[in] th transaction handle + * \param[in] comp_idx index of ldo_comp_entries + * + * \retval 0 on success + * \retval -ENOSPC if not enough OSTs are found + * \retval negative negated errno for other failures + */ +static int lod_ost_alloc_rr(const struct lu_env *env, struct lod_object *lo, + struct dt_object **stripe, __u32 *ost_indices, + int flags, struct thandle *th, int comp_idx, + __u64 reserve) { + struct lod_layout_component *lod_comp; struct lod_device *m = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); - struct obd_statfs *sfs = &lod_env_info(env)->lti_osfs; struct pool_desc *pool = NULL; - struct ost_pool *osts; - struct lov_qos_rr *lqr; - struct dt_object *o; - unsigned array_idx; - int i, rc; - int ost_start_idx_temp; - int speed = 0; - int stripe_idx = 0; - int stripe_cnt = lo->ldo_stripenr; - int stripe_cnt_min = min_stripe_count(stripe_cnt, flags); - __u32 ost_idx; + struct lu_tgt_pool *osts; + struct lu_qos_rr *lqr; + unsigned int i, array_idx; + __u32 ost_start_idx_temp; + __u32 stripe_idx = 0; + __u32 stripe_count, stripe_count_min, ost_idx; + int rc, speed = 0, ost_connecting = 0; + int stripes_per_ost = 1; + bool overstriped = false; ENTRY; - if (lo->ldo_pool) - pool = lod_find_pool(m, lo->ldo_pool); + LASSERT(lo->ldo_comp_cnt > comp_idx && lo->ldo_comp_entries != NULL); + lod_comp = &lo->ldo_comp_entries[comp_idx]; + stripe_count = lod_comp->llc_stripe_count; + stripe_count_min = min_stripe_count(stripe_count, flags); + + if (lod_comp->llc_pool != NULL) + pool = lod_find_pool(m, lod_comp->llc_pool); if (pool != NULL) { down_read(&pool_tgt_rw_sem(pool)); osts = &(pool->pool_obds); lqr = &(pool->pool_rr); } else { - osts = &(m->lod_pool_info); - lqr = &(m->lod_qos.lq_rr); + osts = &m->lod_ost_descs.ltd_tgt_pool; + lqr = &(m->lod_ost_descs.ltd_qos.lq_rr); } - rc = lod_qos_calc_rr(m, osts, lqr); + rc = lod_qos_calc_rr(m, &m->lod_ost_descs, osts, lqr); if (rc) GOTO(out, rc); - rc = lod_qos_ost_in_use_clear(env, lo->ldo_stripenr); + rc = lod_qos_tgt_in_use_clear(env, stripe_count); if (rc) GOTO(out, rc); + down_read(&m->lod_ost_descs.ltd_qos.lq_rw_sem); + spin_lock(&lqr->lqr_alloc); if (--lqr->lqr_start_count <= 0) { - lqr->lqr_start_idx = cfs_rand() % osts->op_count; + lqr->lqr_start_idx = prandom_u32_max(osts->op_count); lqr->lqr_start_count = (LOV_CREATE_RESEED_MIN / max(osts->op_count, 1U) + LOV_CREATE_RESEED_MULT) * max(osts->op_count, 1U); - } else if (stripe_cnt_min >= osts->op_count || + } else if (stripe_count_min >= osts->op_count || lqr->lqr_start_idx > osts->op_count) { /* If we have allocated from all of the OSTs, slowly * precess the next start if the OST/stripe count isn't * already doing this for us. */ lqr->lqr_start_idx %= osts->op_count; - if (stripe_cnt > 1 && (osts->op_count % stripe_cnt) != 1) + if (stripe_count > 1 && (osts->op_count % stripe_count) != 1) ++lqr->lqr_offset_idx; } - down_read(&m->lod_qos.lq_rw_sem); ost_start_idx_temp = lqr->lqr_start_idx; repeat_find: - array_idx = (lqr->lqr_start_idx + lqr->lqr_offset_idx) % - osts->op_count; - - QOS_DEBUG("pool '%s' want %d startidx %d startcnt %d offset %d " - "active %d count %d arrayidx %d\n", - lo->ldo_pool ? lo->ldo_pool : "", - stripe_cnt, lqr->lqr_start_idx, lqr->lqr_start_count, - lqr->lqr_offset_idx, osts->op_count, osts->op_count, - array_idx); - - for (i = 0; i < osts->op_count && stripe_idx < lo->ldo_stripenr; - i++, array_idx = (array_idx + 1) % osts->op_count) { + + QOS_DEBUG("pool '%s' want %d start_idx %d start_count %d offset %d " + "active %d count %d\n", + lod_comp->llc_pool ? lod_comp->llc_pool : "", + stripe_count, lqr->lqr_start_idx, lqr->lqr_start_count, + lqr->lqr_offset_idx, osts->op_count, osts->op_count); + + if (lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING) + stripes_per_ost = + (lod_comp->llc_stripe_count - 1)/osts->op_count + 1; + + for (i = 0; i < osts->op_count * stripes_per_ost + && stripe_idx < stripe_count; i++) { + array_idx = (lqr->lqr_start_idx + lqr->lqr_offset_idx) % + osts->op_count; ++lqr->lqr_start_idx; ost_idx = lqr->lqr_pool.op_array[array_idx]; @@ -737,7 +817,7 @@ repeat_find: stripe_idx, array_idx, ost_idx); if ((ost_idx == LOV_QOS_EMPTY) || - !cfs_bitmap_check(m->lod_ost_bitmap, ost_idx)) + !test_bit(ost_idx, m->lod_ost_bitmap)) continue; /* Fail Check before osc_precreate() is called @@ -745,75 +825,44 @@ repeat_find: if (OBD_FAIL_CHECK(OBD_FAIL_MDS_OSC_PRECREATE) && ost_idx == 0) continue; - rc = lod_statfs_and_check(env, m, ost_idx, sfs); - if (rc) { - /* this OSP doesn't feel well */ - continue; - } + spin_unlock(&lqr->lqr_alloc); + rc = lod_check_and_reserve_ost(env, lo, lod_comp, ost_idx, + speed, &stripe_idx, stripe, + ost_indices, th, &overstriped, + reserve); + spin_lock(&lqr->lqr_alloc); - /* - * skip full devices - */ - if (lod_qos_dev_is_full(sfs)) { - QOS_DEBUG("#%d is full\n", ost_idx); - continue; - } + if (rc != 0 && OST_TGT(m, ost_idx)->ltd_connecting) + ost_connecting = 1; + } + if ((speed < 2) && (stripe_idx < stripe_count_min)) { + /* Try again, allowing slower OSCs */ + speed++; + lqr->lqr_start_idx = ost_start_idx_temp; - /* - * We expect number of precreated objects in f_ffree at - * the first iteration, skip OSPs with no objects ready - */ - if (sfs->os_fprecreated == 0 && speed == 0) { - QOS_DEBUG("#%d: precreation is empty\n", ost_idx); - continue; - } + ost_connecting = 0; + goto repeat_find; + } - /* - * try to use another OSP if this one is degraded - */ - if (sfs->os_state == OS_STATE_DEGRADED && speed < 2) { - QOS_DEBUG("#%d: degraded\n", ost_idx); - continue; - } + spin_unlock(&lqr->lqr_alloc); + up_read(&m->lod_ost_descs.ltd_qos.lq_rw_sem); - /* - * do not put >1 objects on a single OST - */ - if (speed && lod_qos_is_ost_used(env, ost_idx, stripe_idx)) - continue; - - o = lod_qos_declare_object_on(env, m, ost_idx, th); - if (IS_ERR(o)) { - CDEBUG(D_OTHER, "can't declare new object on #%u: %d\n", - ost_idx, (int) PTR_ERR(o)); - rc = PTR_ERR(o); - continue; - } - - /* - * We've successfuly declared (reserved) an object - */ - lod_qos_ost_in_use(env, stripe_idx, ost_idx); - stripe[stripe_idx] = o; - stripe_idx++; - - } - if ((speed < 2) && (stripe_idx < stripe_cnt_min)) { - /* Try again, allowing slower OSCs */ - speed++; - lqr->lqr_start_idx = ost_start_idx_temp; - goto repeat_find; - } - - up_read(&m->lod_qos.lq_rw_sem); + /* If there are enough OSTs, a component with overstriping requested + * will not actually end up overstriped. The comp should reflect this. + */ + if (!overstriped) + lod_comp->llc_pattern &= ~LOV_PATTERN_OVERSTRIPING; if (stripe_idx) { - lo->ldo_stripenr = stripe_idx; + lod_comp->llc_stripe_count = stripe_idx; /* at least one stripe is allocated */ rc = 0; } else { /* nobody provided us with a single object */ - rc = -ENOSPC; + if (ost_connecting) + rc = -EINPROGRESS; + else + rc = -ENOSPC; } out: @@ -826,33 +875,375 @@ out: RETURN(rc); } -/* alloc objects on osts with specific stripe offset */ -static int lod_alloc_specific(const struct lu_env *env, struct lod_object *lo, - struct dt_object **stripe, int flags, - struct thandle *th) +static int +lod_qos_mdt_in_use_init(const struct lu_env *env, + const struct lu_tgt_descs *ltd, + u32 stripe_idx, u32 stripe_count, + const struct lu_tgt_pool *pool, + struct dt_object **stripes) +{ + u32 mdt_idx; + struct lu_tgt_desc *mdt; + int i, j; + int rc; + + rc = lod_qos_tgt_in_use_clear(env, stripe_count); + if (rc) + return rc; + + /* if stripe_idx > 1, we are splitting directory, mark existing stripes + * in_use. Because for either split or creation, stripe 0 is local, + * don't mark it in use. + */ + for (i = 1; i < stripe_idx; i++) { + LASSERT(stripes[i]); + for (j = 0; j < pool->op_count; j++) { + mdt_idx = pool->op_array[j]; + + if (!test_bit(mdt_idx, ltd->ltd_tgt_bitmap)) + continue; + + mdt = LTD_TGT(ltd, mdt_idx); + if (&mdt->ltd_tgt->dd_lu_dev == + stripes[i]->do_lu.lo_dev) + lod_qos_tgt_in_use(env, i, mdt_idx); + } + } + + return 0; +} + +/** + * Allocate a striping using round-robin algorithm. + * + * Allocates a new striping using round-robin algorithm. The function refreshes + * all the internal structures (statfs cache, array of available remote MDTs + * sorted with regard to MDS, etc). The number of stripes required is taken from + * the object (must be prepared by the caller). The caller should ensure nobody + * else is trying to create a striping on the object in parallel. All the + * internal structures (like pools, etc) are protected and no additional locking + * is required. The function succeeds even if a single stripe is allocated. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[out] stripes striping created + * + * \retval positive stripe objects allocated, including the first stripe + * allocated outside + * \retval -ENOSPC if not enough MDTs are found + * \retval negative negated errno for other failures + */ +int lod_mdt_alloc_rr(const struct lu_env *env, struct lod_object *lo, + struct dt_object **stripes, u32 stripe_idx, + u32 stripe_count) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lu_tgt_descs *ltd = &lod->lod_mdt_descs; + struct lu_tgt_pool *pool; + struct lu_qos_rr *lqr; + struct lu_tgt_desc *mdt; + struct lu_object_conf conf = { .loc_flags = LOC_F_NEW }; + struct lu_fid fid = { 0 }; + struct dt_object *dto; + unsigned int pool_idx; + unsigned int i; + u32 saved_idx = stripe_idx; + u32 start_mdt; + u32 mdt_idx; + bool use_degraded = false; + int tgt_connecting = 0; + int rc; + + ENTRY; + + pool = <d->ltd_tgt_pool; + lqr = <d->ltd_qos.lq_rr; + rc = lod_qos_calc_rr(lod, ltd, pool, lqr); + if (rc) + RETURN(rc); + + rc = lod_qos_mdt_in_use_init(env, ltd, stripe_idx, stripe_count, pool, + stripes); + if (rc) + RETURN(rc); + + down_read(<d->ltd_qos.lq_rw_sem); + spin_lock(&lqr->lqr_alloc); + if (--lqr->lqr_start_count <= 0) { + lqr->lqr_start_idx = prandom_u32_max(pool->op_count); + lqr->lqr_start_count = + (LOV_CREATE_RESEED_MIN / max(pool->op_count, 1U) + + LOV_CREATE_RESEED_MULT) * max(pool->op_count, 1U); + } else if (stripe_count - 1 >= pool->op_count || + lqr->lqr_start_idx > pool->op_count) { + /* If we have allocated from all of the tgts, slowly + * precess the next start if the tgt/stripe count isn't + * already doing this for us. */ + lqr->lqr_start_idx %= pool->op_count; + if (stripe_count - 1 > 1 && + (pool->op_count % (stripe_count - 1)) != 1) + ++lqr->lqr_offset_idx; + } + start_mdt = lqr->lqr_start_idx; + +repeat_find: + QOS_DEBUG("want=%d start_idx=%d start_count=%d offset=%d active=%d count=%d\n", + stripe_count - 1, lqr->lqr_start_idx, lqr->lqr_start_count, + lqr->lqr_offset_idx, pool->op_count, pool->op_count); + + for (i = 0; i < pool->op_count && stripe_idx < stripe_count; i++) { + pool_idx = (lqr->lqr_start_idx + lqr->lqr_offset_idx) % + pool->op_count; + ++lqr->lqr_start_idx; + mdt_idx = lqr->lqr_pool.op_array[pool_idx]; + mdt = LTD_TGT(ltd, mdt_idx); + + QOS_DEBUG("#%d strt %d act %d strp %d ary %d idx %d\n", + i, lqr->lqr_start_idx, /* XXX: active*/ 0, + stripe_idx, pool_idx, mdt_idx); + + if (mdt_idx == LOV_QOS_EMPTY || + !test_bit(mdt_idx, ltd->ltd_tgt_bitmap)) + continue; + + /* do not put >1 objects on one MDT */ + if (lod_qos_is_tgt_used(env, mdt_idx, stripe_idx)) + continue; + + rc = lod_is_tgt_usable(ltd, mdt); + if (rc) { + if (mdt->ltd_connecting) + tgt_connecting = 1; + continue; + } + + /* try to use another OSP if this one is degraded */ + if (mdt->ltd_statfs.os_state & OS_STATFS_DEGRADED && + !use_degraded) { + QOS_DEBUG("#%d: degraded\n", mdt_idx); + continue; + } + spin_unlock(&lqr->lqr_alloc); + + rc = dt_fid_alloc(env, mdt->ltd_tgt, &fid, NULL, NULL); + if (rc < 0) { + QOS_DEBUG("#%d: alloc FID failed: %dl\n", mdt_idx, rc); + spin_lock(&lqr->lqr_alloc); + continue; + } + + dto = dt_locate_at(env, mdt->ltd_tgt, &fid, + lo->ldo_obj.do_lu.lo_dev->ld_site->ls_top_dev, + &conf); + + spin_lock(&lqr->lqr_alloc); + if (IS_ERR(dto)) { + QOS_DEBUG("can't alloc stripe on #%u: %d\n", + mdt->ltd_index, (int) PTR_ERR(dto)); + + if (mdt->ltd_connecting) + tgt_connecting = 1; + continue; + } + + lod_qos_tgt_in_use(env, stripe_idx, mdt_idx); + stripes[stripe_idx++] = dto; + } + + if (!use_degraded && stripe_idx < stripe_count) { + /* Try again, allowing slower MDTs */ + use_degraded = true; + lqr->lqr_start_idx = start_mdt; + + tgt_connecting = 0; + goto repeat_find; + } + spin_unlock(&lqr->lqr_alloc); + up_read(<d->ltd_qos.lq_rw_sem); + + if (stripe_idx > saved_idx) + /* at least one stripe is allocated */ + RETURN(stripe_idx); + + /* nobody provided us with a single object */ + if (tgt_connecting) + RETURN(-EINPROGRESS); + + RETURN(-ENOSPC); +} + +/** + * Allocate a specific striping layout on a user defined set of OSTs. + * + * Allocates new striping using the OST index range provided by the data from + * the lmm_obejcts contained in the lov_user_md passed to this method. Full + * OSTs are not considered. The exact order of OSTs requested by the user + * is respected as much as possible depending on OST status. The number of + * stripes needed and stripe offset are taken from the object. If that number + * can not be met, then the function returns a failure and then it's the + * caller's responsibility to release the stripes allocated. All the internal + * structures are protected, but no concurrent allocation is allowed on the + * same objects. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[out] stripe striping created + * \param[out] ost_indices ost indices of striping created + * \param[in] th transaction handle + * \param[in] comp_idx index of ldo_comp_entries + * + * \retval 0 on success + * \retval -ENODEV OST index does not exist on file system + * \retval -EINVAL requested OST index is invalid + * \retval negative negated errno on error + */ +static int lod_alloc_ost_list(const struct lu_env *env, struct lod_object *lo, + struct dt_object **stripe, __u32 *ost_indices, + struct thandle *th, int comp_idx, __u64 reserve) { + struct lod_layout_component *lod_comp; + struct lod_device *m = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct dt_object *o; + unsigned int array_idx = 0; + int stripe_count = 0; + int i; + int rc = -EINVAL; + ENTRY; + + /* for specific OSTs layout */ + LASSERT(lo->ldo_comp_cnt > comp_idx && lo->ldo_comp_entries != NULL); + lod_comp = &lo->ldo_comp_entries[comp_idx]; + LASSERT(lod_comp->llc_ostlist.op_array); + LASSERT(lod_comp->llc_ostlist.op_count); + + rc = lod_qos_tgt_in_use_clear(env, lod_comp->llc_stripe_count); + if (rc < 0) + RETURN(rc); + + if (lod_comp->llc_stripe_offset == LOV_OFFSET_DEFAULT) + lod_comp->llc_stripe_offset = + lod_comp->llc_ostlist.op_array[0]; + + for (i = 0; i < lod_comp->llc_stripe_count; i++) { + if (lod_comp->llc_ostlist.op_array[i] == + lod_comp->llc_stripe_offset) { + array_idx = i; + break; + } + } + if (i == lod_comp->llc_stripe_count) { + CDEBUG(D_OTHER, + "%s: start index %d not in the specified list of OSTs\n", + lod2obd(m)->obd_name, lod_comp->llc_stripe_offset); + RETURN(-EINVAL); + } + + for (i = 0; i < lod_comp->llc_stripe_count; + i++, array_idx = (array_idx + 1) % lod_comp->llc_stripe_count) { + __u32 ost_idx = lod_comp->llc_ostlist.op_array[array_idx]; + + if (!test_bit(ost_idx, m->lod_ost_bitmap)) { + rc = -ENODEV; + break; + } + + /* do not put >1 objects on a single OST, except for + * overstriping + */ + if (lod_qos_is_tgt_used(env, ost_idx, stripe_count) && + !(lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING)) { + rc = -EINVAL; + break; + } + + rc = lod_statfs_and_check(env, m, &m->lod_ost_descs, + LTD_TGT(&m->lod_ost_descs, ost_idx), + reserve); + if (rc < 0) /* this OSP doesn't feel well */ + break; + + o = lod_qos_declare_object_on(env, m, ost_idx, th); + if (IS_ERR(o)) { + rc = PTR_ERR(o); + CDEBUG(D_OTHER, + "%s: can't declare new object on #%u: %d\n", + lod2obd(m)->obd_name, ost_idx, rc); + break; + } + + /* + * We've successfully declared (reserved) an object + */ + lod_qos_tgt_in_use(env, stripe_count, ost_idx); + stripe[stripe_count] = o; + ost_indices[stripe_count] = ost_idx; + stripe_count++; + } + + RETURN(rc); +} + +/** + * Allocate a striping on a predefined set of OSTs. + * + * Allocates new layout starting from OST index in lo->ldo_stripe_offset. + * Full OSTs are not considered. The exact order of OSTs is not important and + * varies depending on OST status. The allocation procedure prefers the targets + * with precreated objects ready. The number of stripes needed and stripe + * offset are taken from the object. If that number cannot be met, then the + * function returns an error and then it's the caller's responsibility to + * release the stripes allocated. All the internal structures are protected, + * but no concurrent allocation is allowed on the same objects. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[out] stripe striping created + * \param[out] ost_indices ost indices of striping created + * \param[in] flags not used + * \param[in] th transaction handle + * \param[in] comp_idx index of ldo_comp_entries + * + * \retval 0 on success + * \retval -ENOSPC if no OST objects are available at all + * \retval -EFBIG if not enough OST objects are found + * \retval -EINVAL requested offset is invalid + * \retval negative errno on failure + */ +static int lod_ost_alloc_specific(const struct lu_env *env, + struct lod_object *lo, + struct dt_object **stripe, __u32 *ost_indices, + int flags, struct thandle *th, int comp_idx, + __u64 reserve) +{ + struct lod_layout_component *lod_comp; struct lod_device *m = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); - struct obd_statfs *sfs = &lod_env_info(env)->lti_osfs; - struct dt_object *o; - unsigned ost_idx, array_idx, ost_count; - int i, rc, stripe_num = 0; - int speed = 0; - struct pool_desc *pool = NULL; - struct ost_pool *osts; + struct dt_object *o; + struct lu_tgt_desc *tgt; + __u32 ost_idx; + unsigned int i, array_idx, ost_count; + int rc, stripe_num = 0; + int speed = 0; + struct pool_desc *pool = NULL; + struct lu_tgt_pool *osts; + int stripes_per_ost = 1; + bool overstriped = false; ENTRY; - rc = lod_qos_ost_in_use_clear(env, lo->ldo_stripenr); + LASSERT(lo->ldo_comp_cnt > comp_idx && lo->ldo_comp_entries != NULL); + lod_comp = &lo->ldo_comp_entries[comp_idx]; + + rc = lod_qos_tgt_in_use_clear(env, lod_comp->llc_stripe_count); if (rc) GOTO(out, rc); - if (lo->ldo_pool) - pool = lod_find_pool(m, lo->ldo_pool); + if (lod_comp->llc_pool != NULL) + pool = lod_find_pool(m, lod_comp->llc_pool); if (pool != NULL) { down_read(&pool_tgt_rw_sem(pool)); osts = &(pool->pool_obds); } else { - osts = &(m->lod_pool_info); + osts = &m->lod_ost_descs.ltd_tgt_pool; } ost_count = osts->op_count; @@ -861,23 +1252,27 @@ repeat_find: /* search loi_ost_idx in ost array */ array_idx = 0; for (i = 0; i < ost_count; i++) { - if (osts->op_array[i] == lo->ldo_def_stripe_offset) { + if (osts->op_array[i] == lod_comp->llc_stripe_offset) { array_idx = i; break; } } if (i == ost_count) { CERROR("Start index %d not found in pool '%s'\n", - lo->ldo_def_stripe_offset, - lo->ldo_pool ? lo->ldo_pool : ""); + lod_comp->llc_stripe_offset, + lod_comp->llc_pool ? lod_comp->llc_pool : ""); GOTO(out, rc = -EINVAL); } - for (i = 0; i < ost_count; + if (lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING) + stripes_per_ost = + (lod_comp->llc_stripe_count - 1)/ost_count + 1; + + for (i = 0; i < ost_count * stripes_per_ost; i++, array_idx = (array_idx + 1) % ost_count) { ost_idx = osts->op_array[array_idx]; - if (!cfs_bitmap_check(m->lod_ost_bitmap, ost_idx)) + if (!test_bit(ost_idx, m->lod_ost_bitmap)) continue; /* Fail Check before osc_precreate() is called @@ -886,29 +1281,44 @@ repeat_find: continue; /* - * do not put >1 objects on a single OST + * do not put >1 objects on a single OST, except for + * overstriping, where it is intended */ - if (lod_qos_is_ost_used(env, ost_idx, stripe_num)) + if (lod_qos_is_tgt_used(env, ost_idx, stripe_num)) { + if (lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING) + overstriped = true; + else + continue; + } + + /* + * try not allocate on the OST used by other component + */ + if (speed == 0 && i != 0 && + lod_comp_is_ost_used(env, lo, ost_idx)) continue; + tgt = LTD_TGT(&m->lod_ost_descs, ost_idx); + /* Drop slow OSCs if we can, but not for requested start idx. * * This means "if OSC is slow and it is not the requested * start OST, then it can be skipped, otherwise skip it only * if it is inactive/recovering/out-of-space." */ - rc = lod_statfs_and_check(env, m, ost_idx, sfs); + rc = lod_statfs_and_check(env, m, &m->lod_ost_descs, + tgt, reserve); if (rc) { /* this OSP doesn't feel well */ continue; } /* - * We expect number of precreated objects in f_ffree at - * the first iteration, skip OSPs with no objects ready - * don't apply this logic to OST specified with stripe_offset + * We expect number of precreated objects at the first + * iteration. Skip OSPs with no objects ready. Don't apply + * this logic to OST specified with stripe_offset. */ - if (i != 0 && sfs->os_fprecreated == 0 && speed == 0) + if (i && !tgt->ltd_statfs.os_fprecreated && !speed) continue; o = lod_qos_declare_object_on(env, m, ost_idx, th); @@ -919,14 +1329,15 @@ repeat_find: } /* - * We've successfuly declared (reserved) an object + * We've successfully declared (reserved) an object */ - lod_qos_ost_in_use(env, stripe_num, ost_idx); + lod_qos_tgt_in_use(env, stripe_num, ost_idx); stripe[stripe_num] = o; + ost_indices[stripe_num] = ost_idx; stripe_num++; /* We have enough stripes */ - if (stripe_num == lo->ldo_stripenr) + if (stripe_num == lod_comp->llc_stripe_count) GOTO(out, rc = 0); } if (speed < 2) { @@ -938,13 +1349,18 @@ repeat_find: /* If we were passed specific striping params, then a failure to * meet those requirements is an error, since we can't reallocate * that memory (it might be part of a larger array or something). - * - * We can only get here if lsm_stripe_count was originally > 1. */ CERROR("can't lstripe objid "DFID": have %d want %u\n", PFID(lu_object_fid(lod2lu_obj(lo))), stripe_num, - lo->ldo_stripenr); - rc = -EFBIG; + lod_comp->llc_stripe_count); + rc = stripe_num == 0 ? -ENOSPC : -EFBIG; + + /* If there are enough OSTs, a component with overstriping requessted + * will not actually end up overstriped. The comp should reflect this. + */ + if (rc == 0 && !overstriped) + lod_comp->llc_pattern &= ~LOV_PATTERN_OVERSTRIPING; + out: if (pool != NULL) { up_read(&pool_tgt_rw_sem(pool)); @@ -955,106 +1371,132 @@ out: RETURN(rc); } -static inline int lod_qos_is_usable(struct lod_device *lod) -{ -#ifdef FORCE_QOS - /* to be able to debug QoS code */ - return 1; -#endif - - /* Detect -EAGAIN early, before expensive lock is taken. */ - if (!lod->lod_qos.lq_dirty && lod->lod_qos.lq_same_space) - return 0; - - if (lod->lod_desc.ld_active_tgt_count < 2) - return 0; - - return 1; -} - -/* Alloc objects on OSTs with optimization based on: - - free space - - network resources (shared OSS's) +/** + * Allocate a striping using an algorithm with weights. + * + * The function allocates OST objects to create a striping. The algorithm + * used is based on weights (currently only using the free space), and it's + * trying to ensure the space is used evenly by OSTs and OSSs. The striping + * configuration (# of stripes, offset, pool) is taken from the object and + * is prepared by the caller. + * + * If LOV_USES_DEFAULT_STRIPE is not passed and prepared configuration can't + * be met due to too few OSTs, then allocation fails. If the flag is passed + * fewer than 3/4 of the requested number of stripes can be allocated, then + * allocation fails. + * + * No concurrent allocation is allowed on the object and this must be ensured + * by the caller. All the internal structures are protected by the function. + * + * The algorithm has two steps: find available OSTs and calculate their + * weights, then select the OSTs with their weights used as the probability. + * An OST with a higher weight is proportionately more likely to be selected + * than one with a lower weight. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[out] stripe striping created + * \param[out] ost_indices ost indices of striping created + * \param[in] flags 0 or LOV_USES_DEFAULT_STRIPE + * \param[in] th transaction handle + * \param[in] comp_idx index of ldo_comp_entries + * + * \retval 0 on success + * \retval -EAGAIN not enough OSTs are found for specified stripe count + * \retval -EINVAL requested OST index is invalid + * \retval negative errno on failure */ -static int lod_alloc_qos(const struct lu_env *env, struct lod_object *lo, - struct dt_object **stripe, int flags, - struct thandle *th) +static int lod_ost_alloc_qos(const struct lu_env *env, struct lod_object *lo, + struct dt_object **stripe, __u32 *ost_indices, + int flags, struct thandle *th, int comp_idx, + __u64 reserve) { - struct lod_device *m = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); - struct obd_statfs *sfs = &lod_env_info(env)->lti_osfs; + struct lod_layout_component *lod_comp; + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lod_avoid_guide *lag = &lod_env_info(env)->lti_avoid; struct lod_tgt_desc *ost; - struct dt_object *o; - __u64 total_weight = 0; - int nfound, good_osts, i, rc = 0; - int stripe_cnt = lo->ldo_stripenr; - int stripe_cnt_min; - struct pool_desc *pool = NULL; - struct ost_pool *osts; + struct dt_object *o; + __u64 total_weight = 0; + struct pool_desc *pool = NULL; + struct lu_tgt_pool *osts; + unsigned int i; + __u32 nfound, good_osts, stripe_count, stripe_count_min; + bool overstriped = false; + int stripes_per_ost = 1; + int rc = 0; ENTRY; - stripe_cnt_min = min_stripe_count(stripe_cnt, flags); - if (stripe_cnt_min < 1) + LASSERT(lo->ldo_comp_cnt > comp_idx && lo->ldo_comp_entries != NULL); + lod_comp = &lo->ldo_comp_entries[comp_idx]; + stripe_count = lod_comp->llc_stripe_count; + stripe_count_min = min_stripe_count(stripe_count, flags); + if (stripe_count_min < 1) RETURN(-EINVAL); - if (lo->ldo_pool) - pool = lod_find_pool(m, lo->ldo_pool); + if (lod_comp->llc_pool != NULL) + pool = lod_find_pool(lod, lod_comp->llc_pool); if (pool != NULL) { down_read(&pool_tgt_rw_sem(pool)); osts = &(pool->pool_obds); } else { - osts = &(m->lod_pool_info); + osts = &lod->lod_ost_descs.ltd_tgt_pool; } /* Detect -EAGAIN early, before expensive lock is taken. */ - if (!lod_qos_is_usable(m)) + if (!ltd_qos_is_usable(&lod->lod_ost_descs)) GOTO(out_nolock, rc = -EAGAIN); + if (lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING) + stripes_per_ost = + (lod_comp->llc_stripe_count - 1)/osts->op_count + 1; + /* Do actual allocation, use write lock here. */ - down_write(&m->lod_qos.lq_rw_sem); + down_write(&lod->lod_ost_descs.ltd_qos.lq_rw_sem); /* * Check again, while we were sleeping on @lq_rw_sem things could * change. */ - if (!lod_qos_is_usable(m)) + if (!ltd_qos_is_usable(&lod->lod_ost_descs)) GOTO(out, rc = -EAGAIN); - rc = lod_qos_calc_ppo(m); + rc = ltd_qos_penalties_calc(&lod->lod_ost_descs); if (rc) GOTO(out, rc); - rc = lod_qos_ost_in_use_clear(env, lo->ldo_stripenr); + rc = lod_qos_tgt_in_use_clear(env, lod_comp->llc_stripe_count); if (rc) GOTO(out, rc); good_osts = 0; /* Find all the OSTs that are valid stripe candidates */ for (i = 0; i < osts->op_count; i++) { - if (!cfs_bitmap_check(m->lod_ost_bitmap, osts->op_array[i])) + if (!test_bit(osts->op_array[i], lod->lod_ost_bitmap)) continue; - rc = lod_statfs_and_check(env, m, osts->op_array[i], sfs); + ost = OST_TGT(lod, osts->op_array[i]); + ost->ltd_qos.ltq_usable = 0; + + rc = lod_statfs_and_check(env, lod, &lod->lod_ost_descs, + ost, reserve); if (rc) { /* this OSP doesn't feel well */ continue; } - /* - * skip full devices - */ - if (lod_qos_dev_is_full(sfs)) + if (ost->ltd_statfs.os_state & OS_STATFS_DEGRADED) continue; /* Fail Check before osc_precreate() is called - so we can only 'fail' single OSC. */ + * so we can only 'fail' single OSC. + */ if (OBD_FAIL_CHECK(OBD_FAIL_MDS_OSC_PRECREATE) && osts->op_array[i] == 0) continue; - ost = OST_TGT(m,osts->op_array[i]); ost->ltd_qos.ltq_usable = 1; - lod_qos_calc_weight(m, osts->op_array[i]); + lu_tgt_qos_weight_calc(ost); total_weight += ost->ltd_qos.ltq_weight; good_osts++; @@ -1062,85 +1504,78 @@ static int lod_alloc_qos(const struct lu_env *env, struct lod_object *lo, QOS_DEBUG("found %d good osts\n", good_osts); - if (good_osts < stripe_cnt_min) + if (good_osts < stripe_count_min) GOTO(out, rc = -EAGAIN); - /* We have enough osts */ - if (good_osts < stripe_cnt) - stripe_cnt = good_osts; + /* If we do not have enough OSTs for the requested stripe count, do not + * put more stripes per OST than requested. + */ + if (stripe_count / stripes_per_ost > good_osts) + stripe_count = good_osts * stripes_per_ost; /* Find enough OSTs with weighted random allocation. */ nfound = 0; - while (nfound < stripe_cnt) { - __u64 rand, cur_weight; + while (nfound < stripe_count) { + u64 rand, cur_weight; cur_weight = 0; rc = -ENOSPC; - if (total_weight) { -#if BITS_PER_LONG == 32 - rand = cfs_rand() % (unsigned)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)(total_weight >> 32)) << 32; - else - rand = 0; - - if (rand == (total_weight & 0xffffffff00000000ULL)) - rand |= cfs_rand() % (unsigned)total_weight; - else - rand |= cfs_rand(); - -#else - rand = ((__u64)cfs_rand() << 32 | cfs_rand()) % - total_weight; -#endif - } else { - rand = 0; - } + rand = lu_prandom_u64_max(total_weight); - /* On average, this will hit larger-weighted osts more often. - 0-weight osts will always get used last (only when rand=0) */ + /* On average, this will hit larger-weighted OSTs more often. + * 0-weight OSTs will always get used last (only when rand=0) + */ for (i = 0; i < osts->op_count; i++) { - int idx = osts->op_array[i]; + __u32 idx = osts->op_array[i]; - if (!cfs_bitmap_check(m->lod_ost_bitmap, idx)) + if (lod_should_avoid_ost(lo, lag, idx)) continue; - ost = OST_TGT(m,idx); + ost = OST_TGT(lod, idx); if (!ost->ltd_qos.ltq_usable) continue; cur_weight += ost->ltd_qos.ltq_weight; - QOS_DEBUG("stripe_cnt=%d nfound=%d cur_weight="LPU64 - " rand="LPU64" total_weight="LPU64"\n", - stripe_cnt, nfound, cur_weight, rand, + QOS_DEBUG("stripe_count=%d nfound=%d cur_weight=%llu " + "rand=%llu total_weight=%llu\n", + stripe_count, nfound, cur_weight, rand, total_weight); if (cur_weight < rand) continue; QOS_DEBUG("stripe=%d to idx=%d\n", nfound, idx); - /* - * do not put >1 objects on a single OST + * do not put >1 objects on a single OST, except for + * overstriping */ - if (lod_qos_is_ost_used(env, idx, nfound)) + if ((lod_comp_is_ost_used(env, lo, idx)) && + !(lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING)) continue; - lod_qos_ost_in_use(env, nfound, idx); - o = lod_qos_declare_object_on(env, m, idx, th); + if (lod_qos_is_tgt_used(env, idx, nfound)) { + if (lod_comp->llc_pattern & + LOV_PATTERN_OVERSTRIPING) + overstriped = true; + else + continue; + } + + o = lod_qos_declare_object_on(env, lod, idx, th); if (IS_ERR(o)) { QOS_DEBUG("can't declare object on #%u: %d\n", idx, (int) PTR_ERR(o)); continue; } - stripe[nfound++] = o; - lod_qos_used(m, osts, idx, &total_weight); + + lod_avoid_update(lo, lag); + lod_qos_tgt_in_use(env, nfound, idx); + stripe[nfound] = o; + ost_indices[nfound] = idx; + ltd_qos_update(&lod->lod_ost_descs, ost, &total_weight); + nfound++; rc = 0; break; } @@ -1151,7 +1586,7 @@ static int lod_alloc_qos(const struct lu_env *env, struct lod_object *lo, } } - if (unlikely(nfound != stripe_cnt)) { + if (unlikely(nfound != stripe_count)) { /* * when the decision to use weighted algorithm was made * we had enough appropriate OSPs, but this state can @@ -1159,22 +1594,29 @@ static int lod_alloc_qos(const struct lu_env *env, struct lod_object *lo, * so it's possible OSP won't be able to provide us with * an object due to just changed state */ - LCONSOLE_INFO("wanted %d, found %d\n", stripe_cnt, nfound); + QOS_DEBUG("%s: wanted %d objects, found only %d\n", + lod2obd(lod)->obd_name, stripe_count, nfound); for (i = 0; i < nfound; i++) { LASSERT(stripe[i] != NULL); - lu_object_put(env, &stripe[i]->do_lu); + dt_object_put(env, stripe[i]); stripe[i] = NULL; } /* makes sense to rebalance next time */ - m->lod_qos.lq_dirty = 1; - m->lod_qos.lq_same_space = 0; + lod->lod_ost_descs.ltd_qos.lq_dirty = 1; + lod->lod_ost_descs.ltd_qos.lq_same_space = 0; rc = -EAGAIN; } + /* If there are enough OSTs, a component with overstriping requessted + * will not actually end up overstriped. The comp should reflect this. + */ + if (rc == 0 && !overstriped) + lod_comp->llc_pattern &= ~LOV_PATTERN_OVERSTRIPING; + out: - up_write(&m->lod_qos.lq_rw_sem); + up_write(&lod->lod_ost_descs.ltd_qos.lq_rw_sem); out_nolock: if (pool != NULL) { @@ -1186,251 +1628,947 @@ out_nolock: RETURN(rc); } -/* Find the max stripecount we should use */ -static __u16 lod_get_stripecnt(struct lod_device *lod, __u32 magic, - __u16 stripe_count) +/** + * Allocate a striping using an algorithm with weights. + * + * The function allocates remote MDT objects to create a striping, the first + * object was already allocated on current MDT to ensure master object and + * the first object are on the same MDT. The algorithm used is based on weights + * (both free space and inodes), and it's trying to ensure the space/inodes are + * used evenly by MDTs and MDSs. The striping configuration (# of stripes, + * offset, pool) is taken from the object and is prepared by the caller. + * + * If prepared configuration can't be met due to too few MDTs, then allocation + * fails. + * + * No concurrent allocation is allowed on the object and this must be ensured + * by the caller. All the internal structures are protected by the function. + * + * The algorithm has two steps: find available MDTs and calculate their + * weights, then select the MDTs with their weights used as the probability. + * An MDT with a higher weight is proportionately more likely to be selected + * than one with a lower weight. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[in] stripe_idx starting stripe index to allocate, if it's not + * 0, we are restriping directory + * \param[in] stripe_count total stripe count + * \param[out] stripes striping created + * + * \retval positive stripes allocated, and it should be equal to + * lo->ldo_dir_stripe_count + * \retval -EAGAIN not enough tgts are found for specified stripe count + * \retval -EINVAL requested MDT index is invalid + * \retval negative errno on failure + */ +int lod_mdt_alloc_qos(const struct lu_env *env, struct lod_object *lo, + struct dt_object **stripes, u32 stripe_idx, + u32 stripe_count) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lu_tgt_descs *ltd = &lod->lod_mdt_descs; + struct lu_object_conf conf = { .loc_flags = LOC_F_NEW }; + struct lu_fid fid = { 0 }; + const struct lu_tgt_pool *pool; + struct lu_tgt_desc *mdt; + struct dt_object *dto; + u64 total_weight = 0; + u32 saved_idx = stripe_idx; + u32 mdt_idx; + unsigned int good_mdts; + unsigned int i; + int rc = 0; + + ENTRY; + + LASSERT(stripe_idx <= stripe_count); + if (stripe_idx == stripe_count) + RETURN(stripe_count); + + /* use MDT pool in @ltd, once MDT pool is supported in the future, it + * can be passed in as argument like OST object allocation. + */ + pool = <d->ltd_tgt_pool; + + /* Detect -EAGAIN early, before expensive lock is taken. */ + if (!ltd_qos_is_usable(ltd)) + RETURN(-EAGAIN); + + rc = lod_qos_mdt_in_use_init(env, ltd, stripe_idx, stripe_count, pool, + stripes); + if (rc) + RETURN(rc); + + /* Do actual allocation, use write lock here. */ + down_write(<d->ltd_qos.lq_rw_sem); + + /* + * Check again, while we were sleeping on @lq_rw_sem things could + * change. + */ + if (!ltd_qos_is_usable(ltd)) + GOTO(unlock, rc = -EAGAIN); + + rc = ltd_qos_penalties_calc(ltd); + if (rc) + GOTO(unlock, rc); + + good_mdts = 0; + /* Find all the MDTs that are valid stripe candidates */ + for (i = 0; i < pool->op_count; i++) { + if (!test_bit(pool->op_array[i], ltd->ltd_tgt_bitmap)) + continue; + + mdt = LTD_TGT(ltd, pool->op_array[i]); + mdt->ltd_qos.ltq_usable = 0; + + rc = lod_is_tgt_usable(ltd, mdt); + if (rc) + continue; + + if (mdt->ltd_statfs.os_state & OS_STATFS_DEGRADED) + continue; + + mdt->ltd_qos.ltq_usable = 1; + lu_tgt_qos_weight_calc(mdt); + total_weight += mdt->ltd_qos.ltq_weight; + + good_mdts++; + } + + QOS_DEBUG("found %d good MDTs\n", good_mdts); + + if (good_mdts < stripe_count - stripe_idx) + GOTO(unlock, rc = -EAGAIN); + + /* Find enough MDTs with weighted random allocation. */ + while (stripe_idx < stripe_count) { + u64 rand, cur_weight; + + cur_weight = 0; + rc = -ENOSPC; + + rand = lu_prandom_u64_max(total_weight); + + /* On average, this will hit larger-weighted MDTs more often. + * 0-weight MDT will always get used last (only when rand=0) */ + for (i = 0; i < pool->op_count; i++) { + int rc2; + + mdt_idx = pool->op_array[i]; + mdt = LTD_TGT(ltd, mdt_idx); + + if (!mdt->ltd_qos.ltq_usable) + continue; + + cur_weight += mdt->ltd_qos.ltq_weight; + + QOS_DEBUG("stripe_count=%d stripe_index=%d cur_weight=%llu rand=%llu total_weight=%llu\n", + stripe_count, stripe_idx, cur_weight, rand, + total_weight); + + if (cur_weight < rand) + continue; + + QOS_DEBUG("stripe=%d to idx=%d\n", + stripe_idx, mdt_idx); + + if (lod_qos_is_tgt_used(env, mdt_idx, stripe_idx)) + continue; + + rc2 = dt_fid_alloc(env, mdt->ltd_tgt, &fid, NULL, NULL); + if (rc2 < 0) { + QOS_DEBUG("can't alloc FID on #%u: %d\n", + mdt_idx, rc2); + continue; + } + + conf.loc_flags = LOC_F_NEW; + dto = dt_locate_at(env, mdt->ltd_tgt, &fid, + lo->ldo_obj.do_lu.lo_dev->ld_site->ls_top_dev, + &conf); + if (IS_ERR(dto)) { + QOS_DEBUG("can't alloc stripe on #%u: %d\n", + mdt_idx, (int) PTR_ERR(dto)); + continue; + } + + lod_qos_tgt_in_use(env, stripe_idx, mdt_idx); + stripes[stripe_idx] = dto; + ltd_qos_update(ltd, mdt, &total_weight); + stripe_idx++; + rc = 0; + break; + } + + /* no MDT found on this iteration, give up */ + if (rc) + break; + } + + if (unlikely(stripe_idx != stripe_count)) { + /* + * when the decision to use weighted algorithm was made + * we had enough appropriate OSPs, but this state can + * change anytime (no space on MDT, broken connection, etc) + * so it's possible OSP won't be able to provide us with + * an object due to just changed state + */ + QOS_DEBUG("%s: wanted %d objects, found only %d\n", + lod2obd(lod)->obd_name, stripe_count, stripe_idx); + for (i = saved_idx; i < stripe_idx; i++) { + LASSERT(stripes[i] != NULL); + dt_object_put(env, stripes[i]); + stripes[i] = NULL; + } + + /* makes sense to rebalance next time */ + ltd->ltd_qos.lq_dirty = 1; + ltd->ltd_qos.lq_same_space = 0; + + rc = -EAGAIN; + } else { + rc = stripe_idx; + } + +unlock: + up_write(<d->ltd_qos.lq_rw_sem); + + RETURN(rc); +} + +/** + * Check stripe count the caller can use. + * + * For new layouts (no initialized components), check the total size of the + * layout against the maximum EA size from the backing file system. This + * stops us from creating a layout which will be too large once initialized. + * + * For existing layouts (with initialized components): + * Find the maximal possible stripe count not greater than \a stripe_count. + * If the provided stripe count is 0, then the filesystem's default is used. + * + * \param[in] lod LOD device + * \param[in] lo The lod_object + * \param[in] stripe_count count the caller would like to use + * + * \retval the maximum usable stripe count + */ +__u16 lod_get_stripe_count(struct lod_device *lod, struct lod_object *lo, + __u16 stripe_count, bool overstriping) { __u32 max_stripes = LOV_MAX_STRIPE_COUNT_OLD; + /* max stripe count is based on OSD ea size */ + unsigned int easize = lod->lod_osd_max_easize; + int i; + if (!stripe_count) - stripe_count = lod->lod_desc.ld_default_stripe_count; - if (stripe_count > lod->lod_desc.ld_active_tgt_count) - stripe_count = lod->lod_desc.ld_active_tgt_count; + stripe_count = + lod->lod_ost_descs.ltd_lov_desc.ld_default_stripe_count; if (!stripe_count) stripe_count = 1; + /* Overstriping allows more stripes than targets */ + if (stripe_count > + lod->lod_ost_descs.ltd_lov_desc.ld_active_tgt_count && + !overstriping) + stripe_count = + lod->lod_ost_descs.ltd_lov_desc.ld_active_tgt_count; + + if (lo->ldo_is_composite) { + struct lod_layout_component *lod_comp; + unsigned int header_sz = sizeof(struct lov_comp_md_v1); + unsigned int init_comp_sz = 0; + unsigned int total_comp_sz = 0; + unsigned int comp_sz; + + header_sz += sizeof(struct lov_comp_md_entry_v1) * + lo->ldo_comp_cnt; + + for (i = 0; i < lo->ldo_comp_cnt; i++) { + lod_comp = &lo->ldo_comp_entries[i]; + comp_sz = lov_mds_md_size(lod_comp->llc_stripe_count, + LOV_MAGIC_V3); + total_comp_sz += comp_sz; + if (lod_comp->llc_flags & LCME_FL_INIT) + init_comp_sz += comp_sz; + } + + if (init_comp_sz > 0) + total_comp_sz = init_comp_sz; - /* stripe count is based on whether OSD can handle larger EA sizes */ - if (lod->lod_osd_max_easize > 0) - max_stripes = lov_mds_md_stripecnt(lod->lod_osd_max_easize, - magic); + header_sz += total_comp_sz; + + if (easize > header_sz) + easize -= header_sz; + else + easize = 0; + } + + max_stripes = lov_mds_md_max_stripe_count(easize, LOV_MAGIC_V3); + max_stripes = (max_stripes == 0) ? 0 : max_stripes - 1; return (stripe_count < max_stripes) ? stripe_count : max_stripes; } -static int lod_use_defined_striping(const struct lu_env *env, - struct lod_object *mo, - const struct lu_buf *buf) +/** + * Create in-core respresentation for a fully-defined striping + * + * When the caller passes a fully-defined striping (i.e. everything including + * OST object FIDs are defined), then we still need to instantiate LU-cache + * with the objects representing the stripes defined. This function completes + * that task. + * + * \param[in] env execution environment for this thread + * \param[in] mo LOD object + * \param[in] buf buffer containing the striping + * + * \retval 0 on success + * \retval negative negated errno on error + */ +int lod_use_defined_striping(const struct lu_env *env, + struct lod_object *mo, + const struct lu_buf *buf) { - struct lod_device *d = lu2lod_dev(lod2lu_obj(mo)->lo_dev); + struct lod_layout_component *lod_comp; struct lov_mds_md_v1 *v1 = buf->lb_buf; struct lov_mds_md_v3 *v3 = buf->lb_buf; + struct lov_comp_md_v1 *comp_v1 = NULL; struct lov_ost_data_v1 *objs; - __u32 magic; - int rc; + __u32 magic; + __u16 comp_cnt; + __u16 mirror_cnt; + int rc = 0, i; ENTRY; - rc = lod_verify_striping(d, buf, 1); - if (rc) - RETURN(rc); + mutex_lock(&mo->ldo_layout_mutex); + lod_striping_free_nolock(env, mo); + + magic = le32_to_cpu(v1->lmm_magic) & ~LOV_MAGIC_DEFINED; + + if (magic != LOV_MAGIC_V1 && magic != LOV_MAGIC_V3 && + magic != LOV_MAGIC_COMP_V1 && magic != LOV_MAGIC_FOREIGN) + GOTO(unlock, rc = -EINVAL); + + if (magic == LOV_MAGIC_COMP_V1) { + comp_v1 = buf->lb_buf; + comp_cnt = le16_to_cpu(comp_v1->lcm_entry_count); + if (comp_cnt == 0) + GOTO(unlock, rc = -EINVAL); + mirror_cnt = le16_to_cpu(comp_v1->lcm_mirror_count) + 1; + mo->ldo_flr_state = le16_to_cpu(comp_v1->lcm_flags) & + LCM_FL_FLR_MASK; + mo->ldo_is_composite = 1; + } else if (magic == LOV_MAGIC_FOREIGN) { + struct lov_foreign_md *foreign; + size_t length; + + if (buf->lb_len < offsetof(typeof(*foreign), lfm_value)) { + CDEBUG(D_LAYOUT, + "buf len %zu < min lov_foreign_md size (%zu)\n", + buf->lb_len, + offsetof(typeof(*foreign), lfm_value)); + GOTO(out, rc = -EINVAL); + } + foreign = (struct lov_foreign_md *)buf->lb_buf; + length = foreign_size_le(foreign); + if (buf->lb_len < length) { + CDEBUG(D_LAYOUT, + "buf len %zu < this lov_foreign_md size (%zu)\n", + buf->lb_len, length); + GOTO(out, rc = -EINVAL); + } - magic = le32_to_cpu(v1->lmm_magic); - if (magic == LOV_MAGIC_V1_DEF) { - objs = &v1->lmm_objects[0]; - } else if (magic == LOV_MAGIC_V3_DEF) { - objs = &v3->lmm_objects[0]; - lod_object_set_pool(mo, v3->lmm_pool_name); + /* just cache foreign LOV EA raw */ + rc = lod_alloc_foreign_lov(mo, length); + if (rc) + GOTO(out, rc); + memcpy(mo->ldo_foreign_lov, buf->lb_buf, length); + GOTO(out, rc); } else { - GOTO(out, rc = -EINVAL); + mo->ldo_is_composite = 0; + comp_cnt = 1; + mirror_cnt = 0; } - - mo->ldo_pattern = le32_to_cpu(v1->lmm_pattern); - mo->ldo_stripe_size = le32_to_cpu(v1->lmm_stripe_size); - mo->ldo_stripenr = le16_to_cpu(v1->lmm_stripe_count); mo->ldo_layout_gen = le16_to_cpu(v1->lmm_layout_gen); - LASSERT(buf->lb_len >= lov_mds_md_size(mo->ldo_stripenr, magic)); - rc = lod_initialize_objects(env, mo, objs); + rc = lod_alloc_comp_entries(mo, mirror_cnt, comp_cnt); + if (rc) + GOTO(unlock, rc); + + for (i = 0; i < comp_cnt; i++) { + struct lu_extent *ext; + char *pool_name; + __u32 offs; + + lod_comp = &mo->ldo_comp_entries[i]; + + if (mo->ldo_is_composite) { + offs = le32_to_cpu(comp_v1->lcm_entries[i].lcme_offset); + v1 = (struct lov_mds_md_v1 *)((char *)comp_v1 + offs); + v3 = (struct lov_mds_md_v3 *)v1; + magic = le32_to_cpu(v1->lmm_magic); + + ext = &comp_v1->lcm_entries[i].lcme_extent; + lod_comp->llc_extent.e_start = + le64_to_cpu(ext->e_start); + lod_comp->llc_extent.e_end = le64_to_cpu(ext->e_end); + lod_comp->llc_flags = + le32_to_cpu(comp_v1->lcm_entries[i].lcme_flags); + if (lod_comp->llc_flags & LCME_FL_NOSYNC) + lod_comp->llc_timestamp = le64_to_cpu( + comp_v1->lcm_entries[i].lcme_timestamp); + lod_comp->llc_id = + le32_to_cpu(comp_v1->lcm_entries[i].lcme_id); + if (lod_comp->llc_id == LCME_ID_INVAL) + GOTO(out, rc = -EINVAL); + } + + pool_name = NULL; + if (magic == LOV_MAGIC_V1) { + objs = &v1->lmm_objects[0]; + } else if (magic == LOV_MAGIC_V3) { + objs = &v3->lmm_objects[0]; + if (v3->lmm_pool_name[0] != '\0') + pool_name = v3->lmm_pool_name; + } else { + CDEBUG(D_LAYOUT, "Invalid magic %x\n", magic); + GOTO(out, rc = -EINVAL); + } + lod_comp->llc_pattern = le32_to_cpu(v1->lmm_pattern); + lod_comp->llc_stripe_size = le32_to_cpu(v1->lmm_stripe_size); + lod_comp->llc_stripe_count = le16_to_cpu(v1->lmm_stripe_count); + lod_comp->llc_layout_gen = le16_to_cpu(v1->lmm_layout_gen); + /** + * The stripe_offset of an uninit-ed component is stored in + * the lmm_layout_gen + */ + if (mo->ldo_is_composite && !lod_comp_inited(lod_comp)) + lod_comp->llc_stripe_offset = lod_comp->llc_layout_gen; + lod_obj_set_pool(mo, i, pool_name); + + if ((!mo->ldo_is_composite || lod_comp_inited(lod_comp)) && + !(lod_comp->llc_pattern & LOV_PATTERN_F_RELEASED) && + !(lod_comp->llc_pattern & LOV_PATTERN_MDT)) { + rc = lod_initialize_objects(env, mo, objs, i); + if (rc) + GOTO(out, rc); + } + } + + rc = lod_fill_mirrors(mo); + GOTO(out, rc); out: + if (rc) + lod_striping_free_nolock(env, mo); +unlock: + mutex_unlock(&mo->ldo_layout_mutex); + RETURN(rc); } -static int lod_qos_parse_config(const struct lu_env *env, - struct lod_object *lo, - const struct lu_buf *buf) +/** + * Parse suggested striping configuration. + * + * The caller gets a suggested striping configuration from a number of sources + * including per-directory default and applications. Then it needs to verify + * the suggested striping is valid, apply missing bits and store the resulting + * configuration in the object to be used by the allocator later. Must not be + * called concurrently against the same object. It's OK to provide a + * fully-defined striping. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[in] buf buffer containing the striping + * + * \retval 0 on success + * \retval negative negated errno on error + */ +int lod_qos_parse_config(const struct lu_env *env, struct lod_object *lo, + const struct lu_buf *buf) { - struct lod_device *d = lu2lod_dev(lod2lu_obj(lo)->lo_dev); + struct lod_layout_component *lod_comp; + struct lod_device *d = lu2lod_dev(lod2lu_obj(lo)->lo_dev); + struct lov_desc *desc = &d->lod_ost_descs.ltd_lov_desc; struct lov_user_md_v1 *v1 = NULL; struct lov_user_md_v3 *v3 = NULL; - struct pool_desc *pool; - __u32 magic; - int rc; + struct lov_comp_md_v1 *comp_v1 = NULL; + struct lov_foreign_md *lfm = NULL; + char def_pool[LOV_MAXPOOLNAME + 1]; + __u32 magic; + __u16 comp_cnt; + __u16 mirror_cnt; + int i, rc; ENTRY; if (buf == NULL || buf->lb_buf == NULL || buf->lb_len == 0) RETURN(0); + memset(def_pool, 0, sizeof(def_pool)); + if (lo->ldo_comp_entries != NULL) + lod_layout_get_pool(lo->ldo_comp_entries, lo->ldo_comp_cnt, + def_pool, sizeof(def_pool)); + + /* free default striping info */ + if (lo->ldo_is_foreign) + lod_free_foreign_lov(lo); + else + lod_free_comp_entries(lo); + + rc = lod_verify_striping(env, d, lo, buf, false); + if (rc) + RETURN(-EINVAL); + + v3 = buf->lb_buf; v1 = buf->lb_buf; + comp_v1 = buf->lb_buf; + /* {lmm,lfm}_magic position/length work for all LOV formats */ magic = v1->lmm_magic; - if (magic == __swab32(LOV_USER_MAGIC_V1)) { - lustre_swab_lov_user_md_v1(v1); - magic = v1->lmm_magic; - } else if (magic == __swab32(LOV_USER_MAGIC_V3)) { - v3 = buf->lb_buf; - lustre_swab_lov_user_md_v3(v3); - magic = v3->lmm_magic; - } - - if (unlikely(magic != LOV_MAGIC_V1 && magic != LOV_MAGIC_V3)) { + if (unlikely(le32_to_cpu(magic) & LOV_MAGIC_DEFINED)) { /* try to use as fully defined striping */ rc = lod_use_defined_striping(env, lo, buf); RETURN(rc); } - if (unlikely(buf->lb_len < sizeof(*v1))) { - CERROR("wrong size: %u\n", (unsigned) buf->lb_len); + switch (magic) { + case __swab32(LOV_USER_MAGIC_V1): + lustre_swab_lov_user_md_v1(v1); + magic = v1->lmm_magic; + /* fall through */ + case LOV_USER_MAGIC_V1: + break; + case __swab32(LOV_USER_MAGIC_V3): + lustre_swab_lov_user_md_v3(v3); + magic = v3->lmm_magic; + /* fall through */ + case LOV_USER_MAGIC_V3: + break; + case __swab32(LOV_USER_MAGIC_SPECIFIC): + lustre_swab_lov_user_md_v3(v3); + lustre_swab_lov_user_md_objects(v3->lmm_objects, + v3->lmm_stripe_count); + magic = v3->lmm_magic; + /* fall through */ + case LOV_USER_MAGIC_SPECIFIC: + break; + case __swab32(LOV_USER_MAGIC_COMP_V1): + lustre_swab_lov_comp_md_v1(comp_v1); + magic = comp_v1->lcm_magic; + /* fall trhough */ + case LOV_USER_MAGIC_COMP_V1: + break; + case __swab32(LOV_USER_MAGIC_FOREIGN): + lfm = buf->lb_buf; + __swab32s(&lfm->lfm_magic); + __swab32s(&lfm->lfm_length); + __swab32s(&lfm->lfm_type); + __swab32s(&lfm->lfm_flags); + magic = lfm->lfm_magic; + /* fall through */ + case LOV_USER_MAGIC_FOREIGN: + if (!lfm) + lfm = buf->lb_buf; + rc = lod_alloc_foreign_lov(lo, foreign_size(lfm)); + if (rc) + RETURN(rc); + memcpy(lo->ldo_foreign_lov, buf->lb_buf, foreign_size(lfm)); + RETURN(0); + default: + CERROR("%s: unrecognized magic %X\n", + lod2obd(d)->obd_name, magic); RETURN(-EINVAL); } - v1->lmm_magic = magic; - if (v1->lmm_pattern == 0) - v1->lmm_pattern = LOV_PATTERN_RAID0; - if (lov_pattern(v1->lmm_pattern) != LOV_PATTERN_RAID0) { - CERROR("invalid pattern: %x\n", v1->lmm_pattern); - RETURN(-EINVAL); + lustre_print_user_md(D_OTHER, v1, "parse config"); + + if (magic == LOV_USER_MAGIC_COMP_V1) { + comp_cnt = comp_v1->lcm_entry_count; + if (comp_cnt == 0) + RETURN(-EINVAL); + mirror_cnt = comp_v1->lcm_mirror_count + 1; + if (mirror_cnt > 1) + lo->ldo_flr_state = LCM_FL_RDONLY; + lo->ldo_is_composite = 1; + } else { + comp_cnt = 1; + mirror_cnt = 0; + lo->ldo_is_composite = 0; } - lo->ldo_pattern = v1->lmm_pattern; - if (v1->lmm_stripe_size) - lo->ldo_stripe_size = v1->lmm_stripe_size; - if (lo->ldo_stripe_size & (LOV_MIN_STRIPE_SIZE - 1)) - lo->ldo_stripe_size = LOV_MIN_STRIPE_SIZE; + rc = lod_alloc_comp_entries(lo, mirror_cnt, comp_cnt); + if (rc) + RETURN(rc); - if (v1->lmm_stripe_count) - lo->ldo_stripenr = v1->lmm_stripe_count; + LASSERT(lo->ldo_comp_entries); - if ((v1->lmm_stripe_offset >= d->lod_desc.ld_tgt_count) && - (v1->lmm_stripe_offset != (typeof(v1->lmm_stripe_offset))(-1))) { - CERROR("invalid offset: %x\n", v1->lmm_stripe_offset); - RETURN(-EINVAL); - } - lo->ldo_def_stripe_offset = v1->lmm_stripe_offset; + for (i = 0; i < comp_cnt; i++) { + struct pool_desc *pool; + struct lu_extent *ext; + char *pool_name; - CDEBUG(D_OTHER, "lsm: %u size, %u stripes, %u offset\n", - v1->lmm_stripe_size, v1->lmm_stripe_count, - v1->lmm_stripe_offset); + lod_comp = &lo->ldo_comp_entries[i]; - if (v1->lmm_magic == LOV_MAGIC_V3) { - if (buf->lb_len < sizeof(*v3)) { - CERROR("wrong size: %u\n", (unsigned) buf->lb_len); - RETURN(-EINVAL); + if (lo->ldo_is_composite) { + v1 = (struct lov_user_md *)((char *)comp_v1 + + comp_v1->lcm_entries[i].lcme_offset); + ext = &comp_v1->lcm_entries[i].lcme_extent; + lod_comp->llc_extent = *ext; + lod_comp->llc_flags = + comp_v1->lcm_entries[i].lcme_flags & + LCME_CL_COMP_FLAGS; } - v3 = buf->lb_buf; - lod_object_set_pool(lo, v3->lmm_pool_name); + pool_name = NULL; + if (v1->lmm_magic == LOV_USER_MAGIC_V3 || + v1->lmm_magic == LOV_USER_MAGIC_SPECIFIC) { + v3 = (struct lov_user_md_v3 *)v1; + if (v3->lmm_pool_name[0] != '\0') + pool_name = v3->lmm_pool_name; + + if (v3->lmm_magic == LOV_USER_MAGIC_SPECIFIC) { + rc = lod_comp_copy_ost_lists(lod_comp, v3); + if (rc) + GOTO(free_comp, rc); + } + } + + if (pool_name == NULL && def_pool[0] != '\0') + pool_name = def_pool; + + if (v1->lmm_pattern == 0) + v1->lmm_pattern = LOV_PATTERN_RAID0; + if (lov_pattern(v1->lmm_pattern) != LOV_PATTERN_RAID0 && + lov_pattern(v1->lmm_pattern) != LOV_PATTERN_MDT && + lov_pattern(v1->lmm_pattern) != + (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING)) { + CDEBUG(D_LAYOUT, "%s: invalid pattern: %x\n", + lod2obd(d)->obd_name, v1->lmm_pattern); + GOTO(free_comp, rc = -EINVAL); + } + + lod_comp->llc_pattern = v1->lmm_pattern; + lod_comp->llc_stripe_size = v1->lmm_stripe_size; + lod_adjust_stripe_size(lod_comp, desc->ld_default_stripe_size); + + lod_comp->llc_stripe_count = desc->ld_default_stripe_count; + if (v1->lmm_stripe_count || + lov_pattern(v1->lmm_pattern) == LOV_PATTERN_MDT) + lod_comp->llc_stripe_count = v1->lmm_stripe_count; + + if (lov_pattern(lod_comp->llc_pattern) == LOV_PATTERN_MDT && + lod_comp->llc_stripe_count != 0) { + CDEBUG(D_LAYOUT, "%s: invalid stripe count: %u\n", + lod2obd(d)->obd_name, + lod_comp->llc_stripe_count); + GOTO(free_comp, rc = -EINVAL); + } + + lod_comp->llc_stripe_offset = v1->lmm_stripe_offset; + lod_obj_set_pool(lo, i, pool_name); + + if (pool_name == NULL) + continue; /* In the function below, .hs_keycmp resolves to * pool_hashkey_keycmp() */ /* coverity[overrun-buffer-val] */ - pool = lod_find_pool(d, v3->lmm_pool_name); - if (pool != NULL) { - if (lo->ldo_def_stripe_offset != - (typeof(v1->lmm_stripe_offset))(-1)) { - rc = lo->ldo_def_stripe_offset; - rc = lod_check_index_in_pool(rc, pool); - if (rc < 0) { - lod_pool_putref(pool); - CERROR("invalid offset\n"); - RETURN(-EINVAL); - } + pool = lod_find_pool(d, pool_name); + if (pool == NULL) + continue; + + if (lod_comp->llc_stripe_offset != LOV_OFFSET_DEFAULT) { + rc = lod_check_index_in_pool( + lod_comp->llc_stripe_offset, pool); + if (rc < 0) { + lod_pool_putref(pool); + CDEBUG(D_LAYOUT, "%s: invalid offset, %u\n", + lod2obd(d)->obd_name, + lod_comp->llc_stripe_offset); + GOTO(free_comp, rc = -EINVAL); } + } + + if (lod_comp->llc_stripe_count > pool_tgt_count(pool) && + !(lod_comp->llc_pattern & LOV_PATTERN_OVERSTRIPING)) + lod_comp->llc_stripe_count = pool_tgt_count(pool); + + lod_pool_putref(pool); + } + + RETURN(0); - if (lo->ldo_stripenr > pool_tgt_count(pool)) - lo->ldo_stripenr= pool_tgt_count(pool); +free_comp: + lod_free_comp_entries(lo); + RETURN(rc); +} + +/** + * prepare enough OST avoidance bitmap space + */ +int lod_prepare_avoidance(const struct lu_env *env, struct lod_object *lo) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lod_avoid_guide *lag = &lod_env_info(env)->lti_avoid; + struct cfs_bitmap *bitmap = NULL; + __u32 *new_oss = NULL; + + lag->lag_ost_avail = lod->lod_ost_count; + + /* reset OSS avoid guide array */ + lag->lag_oaa_count = 0; + if (lag->lag_oss_avoid_array && + lag->lag_oaa_size < lod->lod_ost_count) { + OBD_FREE_PTR_ARRAY(lag->lag_oss_avoid_array, lag->lag_oaa_size); + lag->lag_oss_avoid_array = NULL; + lag->lag_oaa_size = 0; + } - lod_pool_putref(pool); + /* init OST avoid guide bitmap */ + if (lag->lag_ost_avoid_bitmap) { + if (lod->lod_ost_count <= lag->lag_ost_avoid_bitmap->size) { + CFS_RESET_BITMAP(lag->lag_ost_avoid_bitmap); + } else { + CFS_FREE_BITMAP(lag->lag_ost_avoid_bitmap); + lag->lag_ost_avoid_bitmap = NULL; } - } else - lod_object_set_pool(lo, NULL); + } - /* fixup for released file */ - if (lo->ldo_pattern & LOV_PATTERN_F_RELEASED) { - lo->ldo_released_stripenr = lo->ldo_stripenr; - lo->ldo_stripenr = 0; + if (!lag->lag_ost_avoid_bitmap) { + bitmap = CFS_ALLOCATE_BITMAP(lod->lod_ost_count); + if (!bitmap) + return -ENOMEM; } - RETURN(0); + if (!lag->lag_oss_avoid_array) { + /** + * usually there are multiple OSTs in one OSS, but we don't + * know the exact OSS number, so we choose a safe option, + * using OST count to allocate the array to store the OSS + * id. + */ + OBD_ALLOC_PTR_ARRAY(new_oss, lod->lod_ost_count); + if (!new_oss) { + CFS_FREE_BITMAP(bitmap); + return -ENOMEM; + } + } + + if (new_oss) { + lag->lag_oss_avoid_array = new_oss; + lag->lag_oaa_size = lod->lod_ost_count; + } + if (bitmap) + lag->lag_ost_avoid_bitmap = bitmap; + + return 0; } -/* - * buf should be NULL or contain striping settings +/** + * Collect information of used OSTs and OSSs in the overlapped components + * of other mirrors + */ +void lod_collect_avoidance(struct lod_object *lo, struct lod_avoid_guide *lag, + int comp_idx) +{ + struct lod_device *lod = lu2lod_dev(lo->ldo_obj.do_lu.lo_dev); + struct lod_layout_component *lod_comp = &lo->ldo_comp_entries[comp_idx]; + struct cfs_bitmap *bitmap = lag->lag_ost_avoid_bitmap; + int i, j; + + /* iterate mirrors */ + for (i = 0; i < lo->ldo_mirror_count; i++) { + struct lod_layout_component *comp; + + /** + * skip mirror containing component[comp_idx], we only + * collect OSTs info of conflicting component in other mirrors, + * so that during read, if OSTs of a mirror's component are + * not available, we still have other mirror with different + * OSTs to read the data. + */ + comp = &lo->ldo_comp_entries[lo->ldo_mirrors[i].lme_start]; + if (comp->llc_id != LCME_ID_INVAL && + mirror_id_of(comp->llc_id) == + mirror_id_of(lod_comp->llc_id)) + continue; + + /* iterate components of a mirror */ + lod_foreach_mirror_comp(comp, lo, i) { + /** + * skip non-overlapped or un-instantiated components, + * NOTE: don't use lod_comp_inited(comp) to judge + * whether @comp has been inited, since during + * declare phase, comp->llc_stripe has been allocated + * while it's init flag not been set until the exec + * phase. + */ + if (!lu_extent_is_overlapped(&comp->llc_extent, + &lod_comp->llc_extent) || + !comp->llc_stripe) + continue; + + /** + * collect used OSTs index and OSS info from a + * component + */ + for (j = 0; j < comp->llc_stripe_count; j++) { + struct lod_tgt_desc *ost; + struct lu_svr_qos *lsq; + int k; + + ost = OST_TGT(lod, comp->llc_ost_indices[j]); + lsq = ost->ltd_qos.ltq_svr; + + if (cfs_bitmap_check(bitmap, ost->ltd_index)) + continue; + + QOS_DEBUG("OST%d used in conflicting mirror " + "component\n", ost->ltd_index); + cfs_bitmap_set(bitmap, ost->ltd_index); + lag->lag_ost_avail--; + + for (k = 0; k < lag->lag_oaa_count; k++) { + if (lag->lag_oss_avoid_array[k] == + lsq->lsq_id) + break; + } + if (k == lag->lag_oaa_count) { + lag->lag_oss_avoid_array[k] = + lsq->lsq_id; + lag->lag_oaa_count++; + } + } + } + } +} + +/** + * Create a striping for an obejct. + * + * The function creates a new striping for the object. The function tries QoS + * algorithm first unless free space is distributed evenly among OSTs, but + * by default RR algorithm is preferred due to internal concurrency (QoS is + * serialized). The caller must ensure no concurrent calls to the function + * are made against the same object. + * + * \param[in] env execution environment for this thread + * \param[in] lo LOD object + * \param[in] attr attributes OST objects will be declared with + * \param[in] th transaction handle + * \param[in] comp_idx index of ldo_comp_entries + * + * \retval 0 on success + * \retval negative negated errno on error */ int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo, - struct lu_attr *attr, const struct lu_buf *buf, - struct thandle *th) + struct lu_attr *attr, struct thandle *th, + int comp_idx, __u64 reserve) { + struct lod_layout_component *lod_comp; struct lod_device *d = lu2lod_dev(lod2lu_obj(lo)->lo_dev); - struct dt_object **stripe; int stripe_len; int flag = LOV_USES_ASSIGNED_STRIPE; - int i, rc; + int i, rc = 0; + struct lod_avoid_guide *lag = &lod_env_info(env)->lti_avoid; + struct dt_object **stripe = NULL; + __u32 *ost_indices = NULL; ENTRY; LASSERT(lo); + LASSERT(lo->ldo_comp_cnt > comp_idx && lo->ldo_comp_entries != NULL); + lod_comp = &lo->ldo_comp_entries[comp_idx]; + LASSERT(!(lod_comp->llc_flags & LCME_FL_EXTENSION)); - /* no OST available */ - /* XXX: should we be waiting a bit to prevent failures during - * cluster initialization? */ - if (d->lod_ostnr == 0) - GOTO(out, rc = -EIO); - - /* - * by this time, the object's ldo_stripenr and ldo_stripe_size - * contain default value for striping: taken from the parent - * or from filesystem defaults - * - * in case the caller is passing lovea with new striping config, - * we may need to parse lovea and apply new configuration - */ - rc = lod_qos_parse_config(env, lo, buf); - if (rc) - GOTO(out, rc); + /* A released component is being created */ + if (lod_comp->llc_pattern & LOV_PATTERN_F_RELEASED) + RETURN(0); - /* A released file is being created */ - if (lo->ldo_stripenr == 0) - GOTO(out, rc = 0); + /* A Data-on-MDT component is being created */ + if (lov_pattern(lod_comp->llc_pattern) == LOV_PATTERN_MDT) + RETURN(0); - if (likely(lo->ldo_stripe == NULL)) { + if (likely(lod_comp->llc_stripe == NULL)) { /* * no striping has been created so far */ - LASSERT(lo->ldo_stripenr > 0); + LASSERT(lod_comp->llc_stripe_count); /* * statfs and check OST targets now, since ld_active_tgt_count * could be changed if some OSTs are [de]activated manually. */ - lod_qos_statfs_update(env, d); - lo->ldo_stripenr = lod_get_stripecnt(d, LOV_MAGIC, - lo->ldo_stripenr); - - stripe_len = lo->ldo_stripenr; - OBD_ALLOC(stripe, sizeof(stripe[0]) * stripe_len); + lod_qos_statfs_update(env, d, &d->lod_ost_descs); + stripe_len = lod_get_stripe_count(d, lo, + lod_comp->llc_stripe_count, + lod_comp->llc_pattern & + LOV_PATTERN_OVERSTRIPING); + + if (stripe_len == 0) + GOTO(out, rc = -ERANGE); + lod_comp->llc_stripe_count = stripe_len; + OBD_ALLOC_PTR_ARRAY(stripe, stripe_len); if (stripe == NULL) GOTO(out, rc = -ENOMEM); + OBD_ALLOC_PTR_ARRAY(ost_indices, stripe_len); + if (!ost_indices) + GOTO(out, rc = -ENOMEM); +repeat: lod_getref(&d->lod_ost_descs); /* XXX: support for non-0 files w/o objects */ - CDEBUG(D_OTHER, "tgt_count %d stripenr %d\n", - d->lod_desc.ld_tgt_count, stripe_len); - if (lo->ldo_def_stripe_offset >= d->lod_desc.ld_tgt_count) { - rc = lod_alloc_qos(env, lo, stripe, flag, th); + CDEBUG(D_OTHER, "tgt_count %d stripe_count %d\n", + d->lod_ost_count, stripe_len); + + if (lod_comp->llc_ostlist.op_array && + lod_comp->llc_ostlist.op_count) { + rc = lod_alloc_ost_list(env, lo, stripe, ost_indices, + th, comp_idx, reserve); + } else if (lod_comp->llc_stripe_offset == LOV_OFFSET_DEFAULT) { + /** + * collect OSTs and OSSs used in other mirrors whose + * components cross the ldo_comp_entries[comp_idx] + */ + rc = lod_prepare_avoidance(env, lo); + if (rc) + GOTO(put_ldts, rc); + + QOS_DEBUG("collecting conflict osts for comp[%d]\n", + comp_idx); + lod_collect_avoidance(lo, lag, comp_idx); + + rc = lod_ost_alloc_qos(env, lo, stripe, ost_indices, + flag, th, comp_idx, reserve); if (rc == -EAGAIN) - rc = lod_alloc_rr(env, lo, stripe, flag, th); + rc = lod_ost_alloc_rr(env, lo, stripe, + ost_indices, flag, th, + comp_idx, reserve); } else { - rc = lod_alloc_specific(env, lo, stripe, flag, th); + rc = lod_ost_alloc_specific(env, lo, stripe, + ost_indices, flag, th, + comp_idx, reserve); } +put_ldts: lod_putref(d, &d->lod_ost_descs); - if (rc < 0) { for (i = 0; i < stripe_len; i++) if (stripe[i] != NULL) - lu_object_put(env, &stripe[i]->do_lu); - - OBD_FREE(stripe, sizeof(stripe[0]) * stripe_len); + dt_object_put(env, stripe[i]); + + /* In case there is no space on any OST, let's ignore + * the @reserve space to avoid an error at the init + * time, probably the actual IO will be less than the + * given @reserve space (aka extension_size). */ + if (reserve) { + reserve = 0; + goto repeat; + } + lod_comp->llc_stripe_count = 0; } else { - lo->ldo_stripe = stripe; - lo->ldo_stripes_allocated = stripe_len; + lod_comp->llc_stripe = stripe; + lod_comp->llc_ost_indices = ost_indices; + lod_comp->llc_stripes_allocated = stripe_len; } } else { /* @@ -1438,21 +2576,89 @@ int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo, * striping (not a hint), so it allocated all the object * now we need to create them */ - for (i = 0; i < lo->ldo_stripenr; i++) { + for (i = 0; i < lod_comp->llc_stripe_count; i++) { struct dt_object *o; - o = lo->ldo_stripe[i]; + o = lod_comp->llc_stripe[i]; LASSERT(o); - rc = dt_declare_create(env, o, attr, NULL, NULL, th); - if (rc) { + rc = lod_sub_declare_create(env, o, attr, NULL, + NULL, th); + if (rc < 0) { CERROR("can't declare create: %d\n", rc); break; } } + /** + * Clear LCME_FL_INIT for the component so that + * lod_striping_create() can create the striping objects + * in replay. + */ + lod_comp_unset_init(lod_comp); } out: + if (rc < 0) { + if (stripe) + OBD_FREE_PTR_ARRAY(stripe, stripe_len); + if (ost_indices) + OBD_FREE_PTR_ARRAY(ost_indices, stripe_len); + } RETURN(rc); } +int lod_prepare_create(const struct lu_env *env, struct lod_object *lo, + struct lu_attr *attr, const struct lu_buf *buf, + struct thandle *th) + +{ + struct lod_device *d = lu2lod_dev(lod2lu_obj(lo)->lo_dev); + uint64_t size = 0; + int i; + int rc; + ENTRY; + + LASSERT(lo); + + /* no OST available */ + /* XXX: should we be waiting a bit to prevent failures during + * cluster initialization? */ + if (!d->lod_ost_count) + RETURN(-EIO); + + /* + * by this time, the object's ldo_stripe_count and ldo_stripe_size + * contain default value for striping: taken from the parent + * or from filesystem defaults + * + * in case the caller is passing lovea with new striping config, + * we may need to parse lovea and apply new configuration + */ + rc = lod_qos_parse_config(env, lo, buf); + if (rc) + RETURN(rc); + + if (attr->la_valid & LA_SIZE) + size = attr->la_size; + + /** + * prepare OST object creation for the component covering file's + * size, the 1st component (including plain layout file) is always + * instantiated. + */ + for (i = 0; i < lo->ldo_comp_cnt; i++) { + struct lod_layout_component *lod_comp; + struct lu_extent *extent; + + lod_comp = &lo->ldo_comp_entries[i]; + extent = &lod_comp->llc_extent; + QOS_DEBUG("comp[%d] %lld "DEXT"\n", i, size, PEXT(extent)); + if (!lo->ldo_is_composite || size >= extent->e_start) { + rc = lod_qos_prep_create(env, lo, attr, th, i, 0); + if (rc) + break; + } + } + + RETURN(rc); +}