X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Flov%2Flov_obd.c;h=8aed98224d083544caa2f73ce95aa05317546151;hp=9f84ece07d75ea2367a9e1fbb48d0ed43b75dbf6;hb=ab19dad1d77ca25856815aadf2cf10d3cf20d4a3;hpb=6e3ec5812ebd1b5ecf7cae584f429b013ffe7431 diff --git a/lustre/lov/lov_obd.c b/lustre/lov/lov_obd.c index 9f84ece..8aed982 100644 --- a/lustre/lov/lov_obd.c +++ b/lustre/lov/lov_obd.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -26,8 +24,10 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -41,9 +41,6 @@ * Author: Nathan Rutman */ -#ifndef EXPORT_SYMTAB -# define EXPORT_SYMTAB -#endif #define DEBUG_SUBSYSTEM S_LOV #ifdef __KERNEL__ #include @@ -64,7 +61,10 @@ #include #include #include +#include /* for cl_client_lru */ #include +#include +#include #include "lov_internal.h" @@ -75,9 +75,9 @@ static void lov_getref(struct obd_device *obd) struct lov_obd *lov = &obd->u.lov; /* nobody gets through here until lov_putref is done */ - cfs_mutex_down(&lov->lov_lock); + mutex_lock(&lov->lov_lock); cfs_atomic_inc(&lov->lov_refcount); - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); return; } @@ -87,7 +87,7 @@ static void lov_putref(struct obd_device *obd) { struct lov_obd *lov = &obd->u.lov; - cfs_mutex_down(&lov->lov_lock); + mutex_lock(&lov->lov_lock); /* ok to dec to 0 more than once -- ltd_exp's will be null */ if (cfs_atomic_dec_and_test(&lov->lov_refcount) && lov->lov_death_row) { CFS_LIST_HEAD(kill); @@ -108,7 +108,7 @@ static void lov_putref(struct obd_device *obd) lov->lov_tgts[i] = NULL; lov->lov_death_row--; } - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); cfs_list_for_each_entry_safe(tgt, n, &kill, ltd_kill) { cfs_list_del(&tgt->ltd_kill); @@ -116,12 +116,12 @@ static void lov_putref(struct obd_device *obd) __lov_del_obd(obd, tgt); } } else { - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); } } static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid, - int activate); + enum obd_notify_event ev); static int lov_notify(struct obd_device *obd, struct obd_device *watched, enum obd_notify_event ev, void *data); @@ -131,11 +131,10 @@ int lov_connect_obd(struct obd_device *obd, __u32 index, int activate, struct obd_connect_data *data) { struct lov_obd *lov = &obd->u.lov; - struct obd_uuid tgt_uuid; + struct obd_uuid *tgt_uuid; struct obd_device *tgt_obd; static struct obd_uuid lov_osc_uuid = { "LOV_OSC_UUID" }; struct obd_import *imp; - #ifdef __KERNEL__ cfs_proc_dir_entry_t *lov_proc_dir; #endif @@ -145,11 +144,11 @@ int lov_connect_obd(struct obd_device *obd, __u32 index, int activate, if (!lov->lov_tgts[index]) RETURN(-EINVAL); - tgt_uuid = lov->lov_tgts[index]->ltd_uuid; + tgt_uuid = &lov->lov_tgts[index]->ltd_uuid; tgt_obd = lov->lov_tgts[index]->ltd_obd; if (!tgt_obd->obd_set_up) { - CERROR("Target %s not set up\n", obd_uuid2str(&tgt_uuid)); + CERROR("Target %s not set up\n", obd_uuid2str(tgt_uuid)); RETURN(-EINVAL); } @@ -174,14 +173,14 @@ int lov_connect_obd(struct obd_device *obd, __u32 index, int activate, rc = obd_register_observer(tgt_obd, obd); if (rc) { CERROR("Target %s register_observer error %d\n", - obd_uuid2str(&tgt_uuid), rc); + obd_uuid2str(tgt_uuid), rc); RETURN(rc); } if (imp->imp_invalid) { - CERROR("not connecting OSC %s; administratively " - "disabled\n", obd_uuid2str(&tgt_uuid)); + CDEBUG(D_CONFIG, "not connecting OSC %s; administratively " + "disabled\n", obd_uuid2str(tgt_uuid)); RETURN(0); } @@ -189,44 +188,40 @@ int lov_connect_obd(struct obd_device *obd, __u32 index, int activate, &lov_osc_uuid, data, NULL); if (rc || !lov->lov_tgts[index]->ltd_exp) { CERROR("Target %s connect error %d\n", - obd_uuid2str(&tgt_uuid), rc); + obd_uuid2str(tgt_uuid), rc); RETURN(-ENODEV); } lov->lov_tgts[index]->ltd_reap = 0; CDEBUG(D_CONFIG, "Connected tgt idx %d %s (%s) %sactive\n", index, - obd_uuid2str(&tgt_uuid), tgt_obd->obd_name, activate ? "":"in"); + obd_uuid2str(tgt_uuid), tgt_obd->obd_name, activate ? "":"in"); #ifdef __KERNEL__ lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds"); if (lov_proc_dir) { struct obd_device *osc_obd = lov->lov_tgts[index]->ltd_exp->exp_obd; cfs_proc_dir_entry_t *osc_symlink; - char name[MAX_STRING_SIZE]; LASSERT(osc_obd != NULL); LASSERT(osc_obd->obd_magic == OBD_DEVICE_MAGIC); LASSERT(osc_obd->obd_type->typ_name != NULL); - snprintf(name, MAX_STRING_SIZE, "../../../%s/%s", - osc_obd->obd_type->typ_name, - osc_obd->obd_name); - osc_symlink = lprocfs_add_symlink(osc_obd->obd_name, lov_proc_dir, - name); + + osc_symlink = lprocfs_add_symlink(osc_obd->obd_name, + lov_proc_dir, + "../../../%s/%s", + osc_obd->obd_type->typ_name, + osc_obd->obd_name); if (osc_symlink == NULL) { CERROR("could not register LOV target " - "/proc/fs/lustre/%s/%s/target_obds/%s.", - obd->obd_type->typ_name, obd->obd_name, - osc_obd->obd_name); + "/proc/fs/lustre/%s/%s/target_obds/%s.", + obd->obd_type->typ_name, obd->obd_name, + osc_obd->obd_name); lprocfs_remove(&lov_proc_dir); } } #endif - rc = qos_add_tgt(obd, index); - if (rc) - CERROR("qos_add_tgt failed %d\n", rc); - RETURN(0); } @@ -336,8 +331,6 @@ static int lov_disconnect_obd(struct obd_device *obd, struct lov_tgt_desc *tgt) rc = 0; } - qos_del_tgt(obd, tgt); - tgt->ltd_exp = NULL; RETURN(0); } @@ -385,21 +378,34 @@ out: * any >= 0 : is log target index */ static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid, - int activate) + enum obd_notify_event ev) { struct lov_obd *lov = &obd->u.lov; struct lov_tgt_desc *tgt; - int index; + int index, activate, active; ENTRY; - CDEBUG(D_INFO, "Searching in lov %p for uuid %s (activate=%d)\n", - lov, uuid->uuid, activate); - - obd_getref(obd); - for (index = 0; index < lov->desc.ld_tgt_count; index++) { - tgt = lov->lov_tgts[index]; - if (!tgt || !tgt->ltd_exp) - continue; + CDEBUG(D_INFO, "Searching in lov %p for uuid %s event(%d)\n", + lov, uuid->uuid, ev); + + obd_getref(obd); + for (index = 0; index < lov->desc.ld_tgt_count; index++) { + tgt = lov->lov_tgts[index]; + if (!tgt) + continue; + /* + * LU-642, initially inactive OSC could miss the obd_connect, + * we make up for it here. + */ + if (ev == OBD_NOTIFY_ACTIVATE && tgt->ltd_exp == NULL && + obd_uuid_equals(uuid, &tgt->ltd_uuid)) { + struct obd_uuid lov_osc_uuid = {"LOV_OSC_UUID"}; + + obd_connect(NULL, &tgt->ltd_exp, tgt->ltd_obd, + &lov_osc_uuid, &lov->lov_ocd, NULL); + } + if (!tgt->ltd_exp) + continue; CDEBUG(D_INFO, "lov idx %d is %s conn "LPX64"\n", index, obd_uuid2str(&tgt->ltd_uuid), @@ -411,26 +417,41 @@ static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid, if (index == lov->desc.ld_tgt_count) GOTO(out, index = -EINVAL); - if (lov->lov_tgts[index]->ltd_active == activate) { - CDEBUG(D_INFO, "OSC %s already %sactive!\n", uuid->uuid, - activate ? "" : "in"); - GOTO(out, index); - } + if (ev == OBD_NOTIFY_DEACTIVATE || ev == OBD_NOTIFY_ACTIVATE) { + activate = (ev == OBD_NOTIFY_ACTIVATE) ? 1 : 0; - CDEBUG(D_CONFIG, "Marking OSC %s %sactive\n", obd_uuid2str(uuid), - activate ? "" : "in"); + if (lov->lov_tgts[index]->ltd_activate == activate) { + CDEBUG(D_INFO, "OSC %s already %sactivate!\n", + uuid->uuid, activate ? "" : "de"); + } else { + lov->lov_tgts[index]->ltd_activate = activate; + CDEBUG(D_CONFIG, "%sactivate OSC %s\n", + activate ? "" : "de", obd_uuid2str(uuid)); + } - lov->lov_tgts[index]->ltd_active = activate; + } else if (ev == OBD_NOTIFY_INACTIVE || ev == OBD_NOTIFY_ACTIVE) { + active = (ev == OBD_NOTIFY_ACTIVE) ? 1 : 0; - if (activate) { - lov->desc.ld_active_tgt_count++; - lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 0; + if (lov->lov_tgts[index]->ltd_active == active) { + CDEBUG(D_INFO, "OSC %s already %sactive!\n", + uuid->uuid, active ? "" : "in"); + GOTO(out, index); + } else { + CDEBUG(D_CONFIG, "Marking OSC %s %sactive\n", + obd_uuid2str(uuid), active ? "" : "in"); + } + + lov->lov_tgts[index]->ltd_active = active; + if (active) { + lov->desc.ld_active_tgt_count++; + lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 0; + } else { + lov->desc.ld_active_tgt_count--; + lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 1; + } } else { - lov->desc.ld_active_tgt_count--; - lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 1; + CERROR("Unknown event(%d) for uuid %s", ev, uuid->uuid); } - /* remove any old qos penalty */ - lov->lov_tgts[index]->ltd_qos.ltq_penalty = 0; out: obd_putref(obd); @@ -441,14 +462,23 @@ static int lov_notify(struct obd_device *obd, struct obd_device *watched, enum obd_notify_event ev, void *data) { int rc = 0; + struct lov_obd *lov = &obd->u.lov; ENTRY; - if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE) { + down_read(&lov->lov_notify_lock); + if (!lov->lov_connects) { + up_read(&lov->lov_notify_lock); + RETURN(rc); + } + + if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE || + ev == OBD_NOTIFY_ACTIVATE || ev == OBD_NOTIFY_DEACTIVATE) { struct obd_uuid *uuid; LASSERT(watched); if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME)) { + up_read(&lov->lov_notify_lock); CERROR("unexpected notification of %s %s!\n", watched->obd_type->typ_name, watched->obd_name); @@ -459,10 +489,10 @@ static int lov_notify(struct obd_device *obd, struct obd_device *watched, /* Set OSC as active before notifying the observer, so the * observer can use the OSC normally. */ - rc = lov_set_osc_active(obd, uuid, ev == OBD_NOTIFY_ACTIVE); + rc = lov_set_osc_active(obd, uuid, ev); if (rc < 0) { - CERROR("%sactivation of %s failed: %d\n", - (ev == OBD_NOTIFY_ACTIVE) ? "" : "de", + up_read(&lov->lov_notify_lock); + CERROR("event(%d) of %s failed: %d\n", ev, obd_uuid2str(uuid), rc); RETURN(rc); } @@ -477,39 +507,41 @@ static int lov_notify(struct obd_device *obd, struct obd_device *watched, /* NULL watched means all osc's in the lov (only for syncs) */ /* sync event should be send lov idx as data */ struct lov_obd *lov = &obd->u.lov; - struct obd_device *tgt_obd; - int i; + int i, is_sync; + + data = &i; + is_sync = (ev == OBD_NOTIFY_SYNC) || + (ev == OBD_NOTIFY_SYNC_NONBLOCK); + obd_getref(obd); for (i = 0; i < lov->desc.ld_tgt_count; i++) { + if (!lov->lov_tgts[i]) + continue; + /* don't send sync event if target not * connected/activated */ - if (!lov->lov_tgts[i] || - !lov->lov_tgts[i]->ltd_active) - continue; - - if ((ev == OBD_NOTIFY_SYNC) || - (ev == OBD_NOTIFY_SYNC_NONBLOCK)) - data = &i; - - tgt_obd = class_exp2obd(lov->lov_tgts[i]->ltd_exp); + if (is_sync && !lov->lov_tgts[i]->ltd_active) + continue; - rc = obd_notify_observer(obd, tgt_obd, ev, data); + rc = obd_notify_observer(obd, lov->lov_tgts[i]->ltd_obd, + ev, data); if (rc) { CERROR("%s: notify %s of %s failed %d\n", obd->obd_name, obd->obd_observer->obd_name, - tgt_obd->obd_name, rc); - break; + lov->lov_tgts[i]->ltd_obd->obd_name, + rc); } } obd_putref(obd); } + up_read(&lov->lov_notify_lock); RETURN(rc); } -int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, - __u32 index, int gen, int active) +static int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, + __u32 index, int gen, int active) { struct lov_obd *lov = &obd->u.lov; struct lov_tgt_desc *tgt; @@ -531,13 +563,13 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, if (tgt_obd == NULL) RETURN(-EINVAL); - cfs_mutex_down(&lov->lov_lock); + mutex_lock(&lov->lov_lock); if ((index < lov->lov_tgt_size) && (lov->lov_tgts[index] != NULL)) { tgt = lov->lov_tgts[index]; CERROR("UUID %s already assigned at LOV target index %d\n", obd_uuid2str(&tgt->ltd_uuid), index); - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); RETURN(-EEXIST); } @@ -551,7 +583,7 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, newsize = newsize << 1; OBD_ALLOC(newtgts, sizeof(*newtgts) * newsize); if (newtgts == NULL) { - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); RETURN(-ENOMEM); } @@ -562,13 +594,11 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, oldsize = lov->lov_tgt_size; } - lov->lov_tgts = newtgts; - lov->lov_tgt_size = newsize; -#ifdef __KERNEL__ - smp_rmb(); -#endif - if (old) - OBD_FREE(old, sizeof(*old) * oldsize); + lov->lov_tgts = newtgts; + lov->lov_tgt_size = newsize; + smp_rmb(); + if (old) + OBD_FREE(old, sizeof(*old) * oldsize); CDEBUG(D_CONFIG, "tgts: %p size: %d\n", lov->lov_tgts, lov->lov_tgt_size); @@ -576,18 +606,17 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, OBD_ALLOC_PTR(tgt); if (!tgt) { - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); RETURN(-ENOMEM); } rc = lov_ost_pool_add(&lov->lov_packed, index, lov->lov_tgt_size); if (rc) { - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); OBD_FREE_PTR(tgt); RETURN(rc); } - memset(tgt, 0, sizeof(*tgt)); tgt->ltd_uuid = *uuidp; tgt->ltd_obd = tgt_obd; /* XXX - add a sanity check on the generation number. */ @@ -598,11 +627,13 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, if (index >= lov->desc.ld_tgt_count) lov->desc.ld_tgt_count = index + 1; - cfs_mutex_up(&lov->lov_lock); + mutex_unlock(&lov->lov_lock); CDEBUG(D_CONFIG, "idx=%d ltd_gen=%d ld_tgt_count=%d\n", index, tgt->ltd_gen, lov->desc.ld_tgt_count); + rc = obd_notify(obd, tgt_obd, OBD_NOTIFY_CREATE, &index); + if (lov->lov_connects == 0) { /* lov_connect hasn't been called yet. We'll do the lov_connect_obd on this target when that fn first runs, @@ -620,6 +651,15 @@ int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp, if (!tgt->ltd_exp) GOTO(out, rc = 0); + if (lov->lov_cache != NULL) { + rc = obd_set_info_async(NULL, tgt->ltd_exp, + sizeof(KEY_CACHE_SET), KEY_CACHE_SET, + sizeof(struct cl_client_cache), lov->lov_cache, + NULL); + if (rc < 0) + GOTO(out, rc); + } + rc = lov_notify(obd, tgt->ltd_exp->exp_obd, active ? OBD_NOTIFY_CONNECT : OBD_NOTIFY_INACTIVE, (void *)&index); @@ -649,6 +689,8 @@ int lov_del_target(struct obd_device *obd, __u32 index, RETURN(-EINVAL); } + /* to make sure there's no ongoing lov_notify() now */ + down_write(&lov->lov_notify_lock); obd_getref(obd); if (!lov->lov_tgts[index]) { @@ -673,6 +715,7 @@ int lov_del_target(struct obd_device *obd, __u32 index, /* we really delete it from obd_putref */ out: obd_putref(obd); + up_write(&lov->lov_notify_lock); RETURN(rc); } @@ -704,16 +747,18 @@ static void __lov_del_obd(struct obd_device *obd, struct lov_tgt_desc *tgt) void lov_fix_desc_stripe_size(__u64 *val) { - if (*val < PTLRPC_MAX_BRW_SIZE) { - LCONSOLE_WARN("Increasing default stripe size to min %u\n", - PTLRPC_MAX_BRW_SIZE); - *val = PTLRPC_MAX_BRW_SIZE; - } else if (*val & (LOV_MIN_STRIPE_SIZE - 1)) { - *val &= ~(LOV_MIN_STRIPE_SIZE - 1); - LCONSOLE_WARN("Changing default stripe size to "LPU64" (a " - "multiple of %u)\n", - *val, LOV_MIN_STRIPE_SIZE); - } + if (*val < LOV_MIN_STRIPE_SIZE) { + if (*val != 0) + LCONSOLE_INFO("Increasing default stripe size to " + "minimum %u\n", + LOV_DEFAULT_STRIPE_SIZE); + *val = LOV_DEFAULT_STRIPE_SIZE; + } else if (*val & (LOV_MIN_STRIPE_SIZE - 1)) { + *val &= ~(LOV_MIN_STRIPE_SIZE - 1); + LCONSOLE_WARN("Changing default stripe size to "LPU64" (a " + "multiple of %u)\n", + *val, LOV_MIN_STRIPE_SIZE); + } } void lov_fix_desc_stripe_count(__u32 *val) @@ -785,37 +830,24 @@ int lov_setup(struct obd_device *obd, struct lustre_cfg *lcfg) lov->desc = *desc; lov->lov_tgt_size = 0; - cfs_sema_init(&lov->lov_lock, 1); + mutex_init(&lov->lov_lock); cfs_atomic_set(&lov->lov_refcount, 0); - CFS_INIT_LIST_HEAD(&lov->lov_qos.lq_oss_list); - cfs_init_rwsem(&lov->lov_qos.lq_rw_sem); lov->lov_sp_me = LUSTRE_SP_CLI; - lov->lov_qos.lq_dirty = 1; - lov->lov_qos.lq_rr.lqr_dirty = 1; - lov->lov_qos.lq_reset = 1; - /* Default priority is toward free space balance */ - lov->lov_qos.lq_prio_free = 232; - /* Default threshold for rr (roughly 17%) */ - lov->lov_qos.lq_threshold_rr = 43; - /* Init statfs fields */ - OBD_ALLOC_PTR(lov->lov_qos.lq_statfs_data); - if (NULL == lov->lov_qos.lq_statfs_data) - RETURN(-ENOMEM); - cfs_waitq_init(&lov->lov_qos.lq_statfs_waitq); + + init_rwsem(&lov->lov_notify_lock); lov->lov_pools_hash_body = cfs_hash_create("POOLS", HASH_POOLS_CUR_BITS, - HASH_POOLS_CUR_BITS, - &pool_hash_operations, 0); + HASH_POOLS_MAX_BITS, + HASH_POOLS_BKT_BITS, 0, + CFS_HASH_MIN_THETA, + CFS_HASH_MAX_THETA, + &pool_hash_operations, + CFS_HASH_DEFAULT); CFS_INIT_LIST_HEAD(&lov->lov_pool_list); lov->lov_pool_count = 0; rc = lov_ost_pool_init(&lov->lov_packed, 0); if (rc) - RETURN(rc); - rc = lov_ost_pool_init(&lov->lov_qos.lq_rr.lqr_pool, 0); - if (rc) { - lov_ost_pool_free(&lov->lov_packed); - RETURN(rc); - } + GOTO(out, rc); lprocfs_lov_init_vars(&lvars); lprocfs_obd_setup(obd, lvars.obd_vars); @@ -834,6 +866,9 @@ int lov_setup(struct obd_device *obd, struct lustre_cfg *lcfg) NULL, NULL); RETURN(0); + +out: + return rc; } static int lov_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage) @@ -868,17 +903,21 @@ static int lov_cleanup(struct obd_device *obd) struct lov_obd *lov = &obd->u.lov; cfs_list_t *pos, *tmp; struct pool_desc *pool; + ENTRY; cfs_list_for_each_safe(pos, tmp, &lov->lov_pool_list) { pool = cfs_list_entry(pos, struct pool_desc, pool_list); /* free pool structs */ CDEBUG(D_INFO, "delete pool %p\n", pool); + /* In the function below, .hs_keycmp resolves to + * pool_hashkey_keycmp() */ + /* coverity[overrun-buffer-val] */ lov_pool_del(obd, pool->pool_name); } - cfs_hash_destroy(lov->lov_pools_hash_body); - lov_ost_pool_free(&(lov->lov_qos.lq_rr.lqr_pool)); + cfs_hash_putref(lov->lov_pools_hash_body); lov_ost_pool_free(&lov->lov_packed); + lprocfs_obd_cleanup(obd); if (lov->lov_tgts) { int i; obd_getref(obd); @@ -903,11 +942,6 @@ static int lov_cleanup(struct obd_device *obd) lov->lov_tgt_size); lov->lov_tgt_size = 0; } - - /* clear pools parent proc entry only after all pools is killed */ - lprocfs_obd_cleanup(obd); - - OBD_FREE_PTR(lov->lov_qos.lq_statfs_data); RETURN(0); } @@ -976,82 +1010,6 @@ out: RETURN(rc); } -#ifndef log2 -#define log2(n) cfs_ffz(~(n)) -#endif - -static int lov_clear_orphans(struct obd_export *export, struct obdo *src_oa, - struct lov_stripe_md **ea, - struct obd_trans_info *oti) -{ - struct lov_obd *lov; - struct obdo *tmp_oa; - struct obd_uuid *ost_uuid = NULL; - int rc = 0, i; - ENTRY; - - LASSERT(src_oa->o_valid & OBD_MD_FLFLAGS && - src_oa->o_flags == OBD_FL_DELORPHAN); - - lov = &export->exp_obd->u.lov; - - OBDO_ALLOC(tmp_oa); - if (tmp_oa == NULL) - RETURN(-ENOMEM); - - if (oti->oti_ost_uuid) { - ost_uuid = oti->oti_ost_uuid; - CDEBUG(D_HA, "clearing orphans only for %s\n", - ost_uuid->uuid); - } - - obd_getref(export->exp_obd); - for (i = 0; i < lov->desc.ld_tgt_count; i++) { - struct lov_stripe_md obj_md; - struct lov_stripe_md *obj_mdp = &obj_md; - struct lov_tgt_desc *tgt; - int err; - - tgt = lov->lov_tgts[i]; - if (!tgt) - continue; - - /* if called for a specific target, we don't - care if it is not active. */ - if (!lov->lov_tgts[i]->ltd_active && ost_uuid == NULL) { - CDEBUG(D_HA, "lov idx %d inactive\n", i); - continue; - } - - if (ost_uuid && !obd_uuid_equals(ost_uuid, &tgt->ltd_uuid)) - continue; - - CDEBUG(D_CONFIG,"Clear orphans for %d:%s\n", i, - obd_uuid2str(ost_uuid)); - - memcpy(tmp_oa, src_oa, sizeof(*tmp_oa)); - - LASSERT(lov->lov_tgts[i]->ltd_exp); - /* XXX: LOV STACKING: use real "obj_mdp" sub-data */ - err = obd_create(lov->lov_tgts[i]->ltd_exp, - tmp_oa, &obj_mdp, oti); - if (err) { - /* This export will be disabled until it is recovered, - and then orphan recovery will be completed. */ - CERROR("error in orphan recovery on OST idx %d/%d: " - "rc = %d\n", i, lov->desc.ld_tgt_count, err); - rc = err; - } - - if (ost_uuid) - break; - } - obd_putref(export->exp_obd); - - OBDO_FREE(tmp_oa); - RETURN(rc); -} - static int lov_recreate(struct obd_export *exp, struct obdo *src_oa, struct lov_stripe_md **ea, struct obd_trans_info *oti) { @@ -1076,31 +1034,30 @@ static int lov_recreate(struct obd_export *exp, struct obdo *src_oa, !lov->lov_tgts[ost_idx]) GOTO(out, rc = -EINVAL); - for (i = 0; i < lsm->lsm_stripe_count; i++) { - if (lsm->lsm_oinfo[i]->loi_ost_idx == ost_idx) { - if (lsm->lsm_oinfo[i]->loi_id != src_oa->o_id) - GOTO(out, rc = -EINVAL); - break; - } - } + for (i = 0; i < lsm->lsm_stripe_count; i++) { + if (lsm->lsm_oinfo[i]->loi_ost_idx == ost_idx) { + if (ostid_id(&lsm->lsm_oinfo[i]->loi_oi) != + ostid_id(&src_oa->o_oi)) + GOTO(out, rc = -EINVAL); + break; + } + } if (i == lsm->lsm_stripe_count) GOTO(out, rc = -EINVAL); - rc = obd_create(lov->lov_tgts[ost_idx]->ltd_exp, src_oa, &obj_mdp, oti); + rc = obd_create(NULL, lov->lov_tgts[ost_idx]->ltd_exp, + src_oa, &obj_mdp, oti); out: OBD_FREE(obj_mdp, sizeof(*obj_mdp)); RETURN(rc); } /* the LOV expects oa->o_id to be set to the LOV object id */ -static int lov_create(struct obd_export *exp, struct obdo *src_oa, - struct lov_stripe_md **ea, struct obd_trans_info *oti) +static int lov_create(const struct lu_env *env, struct obd_export *exp, + struct obdo *src_oa, struct lov_stripe_md **ea, + struct obd_trans_info *oti) { struct lov_obd *lov; - struct obd_info oinfo; - struct lov_request_set *set = NULL; - struct lov_request *req; - struct l_wait_info lwi = { 0 }; int rc = 0; ENTRY; @@ -1110,8 +1067,8 @@ static int lov_create(struct obd_export *exp, struct obdo *src_oa, if ((src_oa->o_valid & OBD_MD_FLFLAGS) && src_oa->o_flags == OBD_FL_DELORPHAN) { - rc = lov_clear_orphans(exp, src_oa, ea, oti); - RETURN(rc); + /* should be used with LOV anymore */ + LBUG(); } lov = &exp->exp_obd->u.lov; @@ -1123,36 +1080,8 @@ static int lov_create(struct obd_export *exp, struct obdo *src_oa, if ((src_oa->o_valid & OBD_MD_FLFLAGS) && (src_oa->o_flags & OBD_FL_RECREATE_OBJS)) { rc = lov_recreate(exp, src_oa, ea, oti); - GOTO(out, rc); - } - - /* issue statfs rpcs if the osfs data is older than qos_maxage - 1s, - * later in alloc_qos(), we will wait for those rpcs to complete if - * the osfs age is older than 2 * qos_maxage */ - qos_statfs_update(exp->exp_obd, - cfs_time_shift_64(-lov->desc.ld_qos_maxage) + CFS_HZ, - 0); - - rc = lov_prep_create_set(exp, &oinfo, ea, src_oa, oti, &set); - if (rc) - GOTO(out, rc); - - cfs_list_for_each_entry(req, &set->set_list, rq_link) { - /* XXX: LOV STACKING: use real "obj_mdp" sub-data */ - rc = obd_create_async(lov->lov_tgts[req->rq_idx]->ltd_exp, - &req->rq_oi, &req->rq_oi.oi_md, oti); } - /* osc_create have timeout equ obd_timeout/2 so waiting don't be - * longer then this */ - l_wait_event(set->set_waitq, lov_finished_set(set), &lwi); - - /* we not have ptlrpc set for assign set->interpret and should - * be call interpret function himself. calling from cb_create_update - * not permited because lov_fini_create_set can sleep for long time, - * but we must avoid sleeping in ptlrpcd interpret function. */ - rc = lov_fini_create_set(set, ea); -out: obd_putref(exp->exp_obd); RETURN(rc); } @@ -1165,9 +1094,10 @@ do { "%p->lsm_magic=%x\n", (lsmp), (lsmp)->lsm_magic); \ } while (0) -static int lov_destroy(struct obd_export *exp, struct obdo *oa, - struct lov_stripe_md *lsm, struct obd_trans_info *oti, - struct obd_export *md_exp, void *capa) +static int lov_destroy(const struct lu_env *env, struct obd_export *exp, + struct obdo *oa, struct lov_stripe_md *lsm, + struct obd_trans_info *oti, struct obd_export *md_exp, + void *capa) { struct lov_request_set *set; struct obd_info oinfo; @@ -1199,17 +1129,18 @@ static int lov_destroy(struct obd_export *exp, struct obdo *oa, if (oa->o_valid & OBD_MD_FLCOOKIE) oti->oti_logcookies = set->set_cookies + req->rq_stripe; - err = obd_destroy(lov->lov_tgts[req->rq_idx]->ltd_exp, - req->rq_oi.oi_oa, NULL, oti, NULL, capa); - err = lov_update_common_set(set, req, err); - if (err) { - CERROR("error: destroying objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - oa->o_id, req->rq_oi.oi_oa->o_id, - req->rq_idx, err); - if (!rc) - rc = err; - } + err = obd_destroy(env, lov->lov_tgts[req->rq_idx]->ltd_exp, + req->rq_oi.oi_oa, NULL, oti, NULL, capa); + err = lov_update_common_set(set, req, err); + if (err) { + CERROR("%s: destroying objid "DOSTID" subobj " + DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, POSTID(&oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), + req->rq_idx, err); + if (!rc) + rc = err; + } } if (rc == 0) { @@ -1222,7 +1153,8 @@ out: RETURN(rc ? rc : err); } -static int lov_getattr(struct obd_export *exp, struct obd_info *oinfo) +static int lov_getattr(const struct lu_env *env, struct obd_export *exp, + struct obd_info *oinfo) { struct lov_request_set *set; struct lov_request *req; @@ -1246,20 +1178,22 @@ static int lov_getattr(struct obd_export *exp, struct obd_info *oinfo) cfs_list_for_each (pos, &set->set_list) { req = cfs_list_entry(pos, struct lov_request, rq_link); - CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx " - "%u\n", oinfo->oi_oa->o_id, req->rq_stripe, - req->rq_oi.oi_oa->o_id, req->rq_idx); + CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx" + " %u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe, + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx); - rc = obd_getattr(lov->lov_tgts[req->rq_idx]->ltd_exp, + rc = obd_getattr(env, lov->lov_tgts[req->rq_idx]->ltd_exp, &req->rq_oi); - err = lov_update_common_set(set, req, rc); - if (err) { - CERROR("error: getattr objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - oinfo->oi_oa->o_id, req->rq_oi.oi_oa->o_id, - req->rq_idx, err); - break; - } + err = lov_update_common_set(set, req, rc); + if (err) { + CERROR("%s: getattr objid "DOSTID" subobj " + DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, + POSTID(&oinfo->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), + req->rq_idx, err); + break; + } } rc = lov_fini_getattr_set(set); @@ -1277,7 +1211,7 @@ static int lov_getattr_interpret(struct ptlrpc_request_set *rqset, /* don't do attribute merge if this aysnc op failed */ if (rc) - lovset->set_completes = 0; + cfs_atomic_set(&lovset->set_completes, 0); err = lov_fini_getattr_set(lovset); RETURN(rc ? rc : err); } @@ -1304,26 +1238,28 @@ static int lov_getattr_async(struct obd_export *exp, struct obd_info *oinfo, if (rc) RETURN(rc); - CDEBUG(D_INFO, "objid "LPX64": %ux%u byte stripes\n", - oinfo->oi_md->lsm_object_id, oinfo->oi_md->lsm_stripe_count, - oinfo->oi_md->lsm_stripe_size); - - cfs_list_for_each (pos, &lovset->set_list) { - req = cfs_list_entry(pos, struct lov_request, rq_link); - - CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx " - "%u\n", oinfo->oi_oa->o_id, req->rq_stripe, - req->rq_oi.oi_oa->o_id, req->rq_idx); - rc = obd_getattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp, - &req->rq_oi, rqset); - if (rc) { - CERROR("error: getattr objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - oinfo->oi_oa->o_id, req->rq_oi.oi_oa->o_id, - req->rq_idx, rc); - GOTO(out, rc); - } - } + CDEBUG(D_INFO, "objid "DOSTID": %ux%u byte stripes\n", + POSTID(&oinfo->oi_md->lsm_oi), oinfo->oi_md->lsm_stripe_count, + oinfo->oi_md->lsm_stripe_size); + + cfs_list_for_each(pos, &lovset->set_list) { + req = cfs_list_entry(pos, struct lov_request, rq_link); + + CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx" + "%u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe, + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx); + rc = obd_getattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, rqset); + if (rc) { + CERROR("%s: getattr objid "DOSTID" subobj" + DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, + POSTID(&oinfo->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), + req->rq_idx, rc); + GOTO(out, rc); + } + } if (!cfs_list_empty(&rqset->set_requests)) { LASSERT(rc == 0); @@ -1334,13 +1270,13 @@ static int lov_getattr_async(struct obd_export *exp, struct obd_info *oinfo, } out: if (rc) - lovset->set_completes = 0; + cfs_atomic_set(&lovset->set_completes, 0); err = lov_fini_getattr_set(lovset); RETURN(rc ? rc : err); } -static int lov_setattr(struct obd_export *exp, struct obd_info *oinfo, - struct obd_trans_info *oti) +static int lov_setattr(const struct lu_env *env, struct obd_export *exp, + struct obd_info *oinfo, struct obd_trans_info *oti) { struct lov_request_set *set; struct lov_obd *lov; @@ -1371,17 +1307,19 @@ static int lov_setattr(struct obd_export *exp, struct obd_info *oinfo, cfs_list_for_each (pos, &set->set_list) { req = cfs_list_entry(pos, struct lov_request, rq_link); - rc = obd_setattr(lov->lov_tgts[req->rq_idx]->ltd_exp, - &req->rq_oi, NULL); - err = lov_update_setattr_set(set, req, rc); - if (err) { - CERROR("error: setattr objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - set->set_oi->oi_oa->o_id, - req->rq_oi.oi_oa->o_id, req->rq_idx, err); - if (!rc) - rc = err; - } + rc = obd_setattr(env, lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, NULL); + err = lov_update_setattr_set(set, req, rc); + if (err) { + CERROR("%s: setattr objid "DOSTID" subobj " + DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, + POSTID(&set->set_oi->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx, + err); + if (!rc) + rc = err; + } } err = lov_fini_setattr_set(set); if (!rc) @@ -1397,7 +1335,7 @@ static int lov_setattr_interpret(struct ptlrpc_request_set *rqset, ENTRY; if (rc) - lovset->set_completes = 0; + cfs_atomic_set(&lovset->set_completes, 0); err = lov_fini_setattr_set(lovset); RETURN(rc ? rc : err); } @@ -1430,37 +1368,38 @@ static int lov_setattr_async(struct obd_export *exp, struct obd_info *oinfo, if (rc) RETURN(rc); - CDEBUG(D_INFO, "objid "LPX64": %ux%u byte stripes\n", - oinfo->oi_md->lsm_object_id, oinfo->oi_md->lsm_stripe_count, - oinfo->oi_md->lsm_stripe_size); - - cfs_list_for_each (pos, &set->set_list) { - req = cfs_list_entry(pos, struct lov_request, rq_link); - - if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) - oti->oti_logcookies = set->set_cookies + req->rq_stripe; - - CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx " - "%u\n", oinfo->oi_oa->o_id, req->rq_stripe, - req->rq_oi.oi_oa->o_id, req->rq_idx); - - rc = obd_setattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp, - &req->rq_oi, oti, rqset); - if (rc) { - CERROR("error: setattr objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - set->set_oi->oi_oa->o_id, - req->rq_oi.oi_oa->o_id, - req->rq_idx, rc); - break; - } - } + CDEBUG(D_INFO, "objid "DOSTID": %ux%u byte stripes\n", + POSTID(&oinfo->oi_md->lsm_oi), + oinfo->oi_md->lsm_stripe_count, + oinfo->oi_md->lsm_stripe_size); + + cfs_list_for_each(pos, &set->set_list) { + req = cfs_list_entry(pos, struct lov_request, rq_link); + + if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) + oti->oti_logcookies = set->set_cookies + req->rq_stripe; + + CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx" + "%u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe, + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx); + + rc = obd_setattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, oti, rqset); + if (rc) { + CERROR("error: setattr objid "DOSTID" subobj" + DOSTID" on OST idx %d: rc = %d\n", + POSTID(&set->set_oi->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), + req->rq_idx, rc); + break; + } + } /* If we are not waiting for responses on async requests, return. */ if (rc || !rqset || cfs_list_empty(&rqset->set_requests)) { int err; if (rc) - set->set_completes = 0; + cfs_atomic_set(&set->set_completes, 0); err = lov_fini_setattr_set(set); RETURN(rc ? rc : err); } @@ -1480,7 +1419,7 @@ static int lov_punch_interpret(struct ptlrpc_request_set *rqset, ENTRY; if (rc) - lovset->set_completes = 0; + cfs_atomic_set(&lovset->set_completes, 0); err = lov_fini_punch_set(lovset); RETURN(rc ? rc : err); } @@ -1488,8 +1427,8 @@ static int lov_punch_interpret(struct ptlrpc_request_set *rqset, /* FIXME: maybe we'll just make one node the authoritative attribute node, then * we can send this 'punch' to just the authoritative node and the nodes * that the punch will affect. */ -static int lov_punch(struct obd_export *exp, struct obd_info *oinfo, - struct obd_trans_info *oti, +static int lov_punch(const struct lu_env *env, struct obd_export *exp, + struct obd_info *oinfo, struct obd_trans_info *oti, struct ptlrpc_request_set *rqset) { struct lov_request_set *set; @@ -1513,15 +1452,16 @@ static int lov_punch(struct obd_export *exp, struct obd_info *oinfo, cfs_list_for_each (pos, &set->set_list) { req = cfs_list_entry(pos, struct lov_request, rq_link); - rc = obd_punch(lov->lov_tgts[req->rq_idx]->ltd_exp, - &req->rq_oi, NULL, rqset); - if (rc) { - CERROR("error: punch objid "LPX64" subobj "LPX64 - " on OST idx %d: rc = %d\n", - set->set_oi->oi_oa->o_id, - req->rq_oi.oi_oa->o_id, req->rq_idx, rc); - break; - } + rc = obd_punch(env, lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, NULL, rqset); + if (rc) { + CERROR("%s: punch objid "DOSTID" subobj "DOSTID + " on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, + POSTID(&set->set_oi->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx, rc); + break; + } } if (rc || cfs_list_empty(&rqset->set_requests)) { @@ -1537,49 +1477,73 @@ static int lov_punch(struct obd_export *exp, struct obd_info *oinfo, RETURN(0); } -static int lov_sync(struct obd_export *exp, struct obdo *oa, - struct lov_stripe_md *lsm, obd_off start, obd_off end, - void *capa) +static int lov_sync_interpret(struct ptlrpc_request_set *rqset, + void *data, int rc) { - struct lov_request_set *set; - struct obd_info oinfo; + struct lov_request_set *lovset = data; + int err; + ENTRY; + + if (rc) + cfs_atomic_set(&lovset->set_completes, 0); + err = lov_fini_sync_set(lovset); + RETURN(rc ?: err); +} + +static int lov_sync(const struct lu_env *env, struct obd_export *exp, + struct obd_info *oinfo, obd_off start, obd_off end, + struct ptlrpc_request_set *rqset) +{ + struct lov_request_set *set = NULL; struct lov_obd *lov; cfs_list_t *pos; struct lov_request *req; - int err = 0, rc = 0; + int rc = 0; ENTRY; - ASSERT_LSM_MAGIC(lsm); + ASSERT_LSM_MAGIC(oinfo->oi_md); + LASSERT(rqset != NULL); if (!exp->exp_obd) RETURN(-ENODEV); lov = &exp->exp_obd->u.lov; - rc = lov_prep_sync_set(exp, &oinfo, oa, lsm, start, end, &set); + rc = lov_prep_sync_set(exp, oinfo, start, end, &set); if (rc) RETURN(rc); + CDEBUG(D_INFO, "fsync objid "DOSTID" ["LPX64", "LPX64"]\n", + POSTID(&set->set_oi->oi_oa->o_oi), start, end); + cfs_list_for_each (pos, &set->set_list) { req = cfs_list_entry(pos, struct lov_request, rq_link); - rc = obd_sync(lov->lov_tgts[req->rq_idx]->ltd_exp, - req->rq_oi.oi_oa, NULL, - req->rq_oi.oi_policy.l_extent.start, - req->rq_oi.oi_policy.l_extent.end, capa); - err = lov_update_common_set(set, req, rc); - if (err) { - CERROR("error: fsync objid "LPX64" subobj "LPX64 - " on OST idx %d: rc = %d\n", - set->set_oi->oi_oa->o_id, - req->rq_oi.oi_oa->o_id, req->rq_idx, rc); - if (!rc) - rc = err; - } + rc = obd_sync(env, lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, req->rq_oi.oi_policy.l_extent.start, + req->rq_oi.oi_policy.l_extent.end, rqset); + if (rc) { + CERROR("%s: fsync objid "DOSTID" subobj "DOSTID + " on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, + POSTID(&set->set_oi->oi_oa->o_oi), + POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx, + rc); + break; + } } - err = lov_fini_sync_set(set); - if (!rc) - rc = err; - RETURN(rc); + + /* If we are not waiting for responses on async requests, return. */ + if (rc || cfs_list_empty(&rqset->set_requests)) { + int err = lov_fini_sync_set(set); + + RETURN(rc ?: err); + } + + LASSERT(rqset->set_interpret == NULL); + rqset->set_interpret = lov_sync_interpret; + rqset->set_arg = (void *)set; + + RETURN(0); } static int lov_brw_check(struct lov_obd *lov, struct obd_info *lov_oinfo, @@ -1727,8 +1691,6 @@ static int lov_change_cbdata(struct obd_export *exp, if (!exp || !exp->exp_obd) RETURN(-ENODEV); - LASSERT_MDS_GROUP(lsm->lsm_object_gr); - lov = &exp->exp_obd->u.lov; for (i = 0; i < lsm->lsm_stripe_count; i++) { struct lov_stripe_md submd; @@ -1739,13 +1701,48 @@ static int lov_change_cbdata(struct obd_export *exp, continue; } - submd.lsm_object_id = loi->loi_id; - submd.lsm_object_gr = lsm->lsm_object_gr; - submd.lsm_stripe_count = 0; - rc = obd_change_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp, - &submd, it, data); - } - RETURN(rc); + submd.lsm_oi = loi->loi_oi; + submd.lsm_stripe_count = 0; + rc = obd_change_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp, + &submd, it, data); + } + RETURN(rc); +} + +/* find any ldlm lock of the inode in lov + * return 0 not find + * 1 find one + * < 0 error */ +static int lov_find_cbdata(struct obd_export *exp, + struct lov_stripe_md *lsm, ldlm_iterator_t it, + void *data) +{ + struct lov_obd *lov; + int rc = 0, i; + ENTRY; + + ASSERT_LSM_MAGIC(lsm); + + if (!exp || !exp->exp_obd) + RETURN(-ENODEV); + + lov = &exp->exp_obd->u.lov; + for (i = 0; i < lsm->lsm_stripe_count; i++) { + struct lov_stripe_md submd; + struct lov_oinfo *loi = lsm->lsm_oinfo[i]; + + if (!lov->lov_tgts[loi->loi_ost_idx]) { + CDEBUG(D_HA, "lov idx %d NULL \n", loi->loi_ost_idx); + continue; + } + submd.lsm_oi = loi->loi_oi; + submd.lsm_stripe_count = 0; + rc = obd_find_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp, + &submd, it, data); + if (rc != 0) + RETURN(rc); + } + RETURN(rc); } static int lov_cancel(struct obd_export *exp, struct lov_stripe_md *lsm, @@ -1765,37 +1762,36 @@ static int lov_cancel(struct obd_export *exp, struct lov_stripe_md *lsm, if (!exp || !exp->exp_obd) RETURN(-ENODEV); - LASSERT_MDS_GROUP(lsm->lsm_object_gr); LASSERT(lockh); lov = &exp->exp_obd->u.lov; rc = lov_prep_cancel_set(exp, &oinfo, lsm, mode, lockh, &set); if (rc) RETURN(rc); - cfs_list_for_each (pos, &set->set_list) { - req = cfs_list_entry(pos, struct lov_request, rq_link); - lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe; - - rc = obd_cancel(lov->lov_tgts[req->rq_idx]->ltd_exp, - req->rq_oi.oi_md, mode, lov_lockhp); - rc = lov_update_common_set(set, req, rc); - if (rc) { - CERROR("error: cancel objid "LPX64" subobj " - LPX64" on OST idx %d: rc = %d\n", - lsm->lsm_object_id, - req->rq_oi.oi_md->lsm_object_id, - req->rq_idx, rc); - err = rc; - } - - } - lov_fini_cancel_set(set); - RETURN(err); + cfs_list_for_each(pos, &set->set_list) { + req = cfs_list_entry(pos, struct lov_request, rq_link); + lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe; + + rc = obd_cancel(lov->lov_tgts[req->rq_idx]->ltd_exp, + req->rq_oi.oi_md, mode, lov_lockhp); + rc = lov_update_common_set(set, req, rc); + if (rc) { + CERROR("%s: cancel objid "DOSTID" subobj " + DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, POSTID(&lsm->lsm_oi), + POSTID(&req->rq_oi.oi_md->lsm_oi), + req->rq_idx, rc); + err = rc; + } + + } + lov_fini_cancel_set(set); + RETURN(err); } static int lov_cancel_unused(struct obd_export *exp, struct lov_stripe_md *lsm, - int flags, void *opaque) + ldlm_cancel_flags_t flags, void *opaque) { struct lov_obd *lov; int rc = 0, i; @@ -1821,34 +1817,34 @@ static int lov_cancel_unused(struct obd_export *exp, ASSERT_LSM_MAGIC(lsm); - LASSERT_MDS_GROUP(lsm->lsm_object_gr); - for (i = 0; i < lsm->lsm_stripe_count; i++) { - struct lov_stripe_md submd; - struct lov_oinfo *loi = lsm->lsm_oinfo[i]; - int err; - - if (!lov->lov_tgts[loi->loi_ost_idx]) { - CDEBUG(D_HA, "lov idx %d NULL\n", loi->loi_ost_idx); - continue; - } - - if (!lov->lov_tgts[loi->loi_ost_idx]->ltd_active) - CDEBUG(D_HA, "lov idx %d inactive\n", loi->loi_ost_idx); - - submd.lsm_object_id = loi->loi_id; - submd.lsm_object_gr = lsm->lsm_object_gr; - submd.lsm_stripe_count = 0; - err = obd_cancel_unused(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp, - &submd, flags, opaque); - if (err && lov->lov_tgts[loi->loi_ost_idx]->ltd_active) { - CERROR("error: cancel unused objid "LPX64" subobj "LPX64 - " on OST idx %d: rc = %d\n", lsm->lsm_object_id, - loi->loi_id, loi->loi_ost_idx, err); - if (!rc) - rc = err; - } - } - RETURN(rc); + for (i = 0; i < lsm->lsm_stripe_count; i++) { + struct lov_stripe_md submd; + struct lov_oinfo *loi = lsm->lsm_oinfo[i]; + int idx = loi->loi_ost_idx; + int err; + + if (!lov->lov_tgts[idx]) { + CDEBUG(D_HA, "lov idx %d NULL\n", idx); + continue; + } + + if (!lov->lov_tgts[idx]->ltd_active) + CDEBUG(D_HA, "lov idx %d inactive\n", idx); + + submd.lsm_oi = loi->loi_oi; + submd.lsm_stripe_count = 0; + err = obd_cancel_unused(lov->lov_tgts[idx]->ltd_exp, + &submd, flags, opaque); + if (err && lov->lov_tgts[idx]->ltd_active) { + CERROR("%s: cancel unused objid "DOSTID + " subobj "DOSTID" on OST idx %d: rc = %d\n", + exp->exp_obd->obd_name, POSTID(&lsm->lsm_oi), + POSTID(&loi->loi_oi), idx, err); + if (!rc) + rc = err; + } + } + RETURN(rc); } int lov_statfs_interpret(struct ptlrpc_request_set *rqset, void *data, int rc) @@ -1858,15 +1854,16 @@ int lov_statfs_interpret(struct ptlrpc_request_set *rqset, void *data, int rc) ENTRY; if (rc) - lovset->set_completes = 0; + cfs_atomic_set(&lovset->set_completes, 0); err = lov_fini_statfs_set(lovset); RETURN(rc ? rc : err); } -static int lov_statfs_async(struct obd_device *obd, struct obd_info *oinfo, +static int lov_statfs_async(struct obd_export *exp, struct obd_info *oinfo, __u64 max_age, struct ptlrpc_request_set *rqset) { + struct obd_device *obd = class_exp2obd(exp); struct lov_request_set *set; struct lov_request *req; cfs_list_t *pos; @@ -1883,12 +1880,9 @@ static int lov_statfs_async(struct obd_device *obd, struct obd_info *oinfo, RETURN(rc); cfs_list_for_each (pos, &set->set_list) { - struct obd_device *osc_obd; - req = cfs_list_entry(pos, struct lov_request, rq_link); - - osc_obd = class_exp2obd(lov->lov_tgts[req->rq_idx]->ltd_exp); - rc = obd_statfs_async(osc_obd, &req->rq_oi, max_age, rqset); + rc = obd_statfs_async(lov->lov_tgts[req->rq_idx]->ltd_exp, + &req->rq_oi, max_age, rqset); if (rc) break; } @@ -1896,7 +1890,7 @@ static int lov_statfs_async(struct obd_device *obd, struct obd_info *oinfo, if (rc || cfs_list_empty(&rqset->set_requests)) { int err; if (rc) - set->set_completes = 0; + cfs_atomic_set(&set->set_completes, 0); err = lov_fini_statfs_set(set); RETURN(rc ? rc : err); } @@ -1907,8 +1901,8 @@ static int lov_statfs_async(struct obd_device *obd, struct obd_info *oinfo, RETURN(0); } -static int lov_statfs(struct obd_device *obd, struct obd_statfs *osfs, - __u64 max_age, __u32 flags) +static int lov_statfs(const struct lu_env *env, struct obd_export *exp, + struct obd_statfs *osfs, __u64 max_age, __u32 flags) { struct ptlrpc_request_set *set = NULL; struct obd_info oinfo = { { { 0 } } }; @@ -1924,7 +1918,7 @@ static int lov_statfs(struct obd_device *obd, struct obd_statfs *osfs, oinfo.oi_osfs = osfs; oinfo.oi_flags = flags; - rc = lov_statfs_async(obd, &oinfo, max_age, set); + rc = lov_statfs_async(exp, &oinfo, max_age, set); if (rc == 0) rc = ptlrpc_set_wait(set); ptlrpc_set_destroy(set); @@ -1947,10 +1941,9 @@ static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len, struct obd_device *osc_obd; struct obd_statfs stat_buf = {0}; __u32 index; + __u32 flags; memcpy(&index, data->ioc_inlbuf2, sizeof(__u32)); - LASSERT(data->ioc_plen1 == sizeof(struct obd_statfs)); - if ((index >= count)) RETURN(-ENODEV); @@ -1964,17 +1957,22 @@ static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len, if (!osc_obd) RETURN(-EINVAL); + /* copy UUID */ + if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(osc_obd), + min((int)data->ioc_plen2, + (int)sizeof(struct obd_uuid)))) + RETURN(-EFAULT); + + flags = uarg ? *(__u32*)uarg : 0; /* got statfs data */ - rc = obd_statfs(osc_obd, &stat_buf, - cfs_time_current_64() - CFS_HZ, 0); + rc = obd_statfs(NULL, lov->lov_tgts[index]->ltd_exp, &stat_buf, + cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), + flags); if (rc) RETURN(rc); - if (cfs_copy_to_user(data->ioc_pbuf1, &stat_buf, - data->ioc_plen1)) - RETURN(-EFAULT); - /* copy UUID */ - if (cfs_copy_to_user(data->ioc_pbuf2, obd2cli_tgt(osc_obd), - data->ioc_plen2)) + if (copy_to_user(data->ioc_pbuf1, &stat_buf, + min((int) data->ioc_plen1, + (int) sizeof(stat_buf)))) RETURN(-EFAULT); break; } @@ -2018,7 +2016,7 @@ static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len, *genp = lov->lov_tgts[i]->ltd_gen; } - if (cfs_copy_to_user((void *)uarg, buf, len)) + if (copy_to_user((void *)uarg, buf, len)) rc = -EFAULT; obd_ioctl_freedata(buf, len); break; @@ -2037,13 +2035,13 @@ static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len, struct lov_tgt_desc *tgt = NULL; struct obd_quotactl *oqctl; - if (qctl->qc_valid == QC_OSTIDX) { - if (qctl->qc_idx < 0 || count <= qctl->qc_idx) - RETURN(-EINVAL); + if (qctl->qc_valid == QC_OSTIDX) { + if (count <= qctl->qc_idx) + RETURN(-EINVAL); - tgt = lov->lov_tgts[qctl->qc_idx]; - if (!tgt || !tgt->ltd_exp) - RETURN(-EINVAL); + tgt = lov->lov_tgts[qctl->qc_idx]; + if (!tgt || !tgt->ltd_exp) + RETURN(-EINVAL); } else if (qctl->qc_valid == QC_UUID) { for (i = 0; i < count; i++) { tgt = lov->lov_tgts[i]; @@ -2087,11 +2085,16 @@ static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len, for (i = 0; i < count; i++) { int err; + struct obd_device *osc_obd; /* OST was disconnected */ if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_exp) continue; + /* ll_umount_begin() sets force flag but for lov, not + * osc. Let's pass it through */ + osc_obd = class_exp2obd(lov->lov_tgts[i]->ltd_exp); + osc_obd->obd_force = obddev->obd_force; err = obd_iocontrol(cmd, lov->lov_tgts[i]->ltd_exp, len, karg, uarg); if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) { @@ -2159,6 +2162,8 @@ obd_size fiemap_calc_fm_end_offset(struct ll_user_fiemap *fiemap, break; } } + if (stripe_no == -1) + return -EINVAL; /* If we have finished mapping on previous device, shift logical * offset to start of next device */ @@ -2262,7 +2267,7 @@ static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key, int count_local; unsigned int get_num_extents = 0; int ost_index = 0, actual_start_stripe, start_stripe; - obd_size fm_start, fm_end, fm_length, fm_end_offset = 0; + obd_size fm_start, fm_end, fm_length, fm_end_offset; obd_size curr_loc; int current_extent = 0, rc = 0, i; int ost_eof = 0; /* EOF for object */ @@ -2271,13 +2276,34 @@ static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key, int cur_stripe = 0, cur_stripe_wrap = 0, stripe_count; unsigned int buffer_size = FIEMAP_BUFFER_SIZE; - if (lsm == NULL) - GOTO(out, rc = 0); + if (!lsm_has_objects(lsm)) { + if (lsm && lsm_is_released(lsm) && (fm_key->fiemap.fm_start < + fm_key->oa.o_size)) { + /* released file, return a minimal FIEMAP if + * request fits in file-size. + */ + fiemap->fm_mapped_extents = 1; + fiemap->fm_extents[0].fe_logical = + fm_key->fiemap.fm_start; + if (fm_key->fiemap.fm_start + fm_key->fiemap.fm_length < + fm_key->oa.o_size) + fiemap->fm_extents[0].fe_length = + fm_key->fiemap.fm_length; + else + fiemap->fm_extents[0].fe_length = + fm_key->oa.o_size - + fm_key->fiemap.fm_start; + fiemap->fm_extents[0].fe_flags |= + (FIEMAP_EXTENT_UNKNOWN | + FIEMAP_EXTENT_LAST); + } + GOTO(out, rc = 0); + } if (fiemap_count_to_size(fm_key->fiemap.fm_extent_count) < buffer_size) buffer_size = fiemap_count_to_size(fm_key->fiemap.fm_extent_count); - OBD_ALLOC(fm_local, buffer_size); + OBD_ALLOC_LARGE(fm_local, buffer_size); if (fm_local == NULL) GOTO(out, rc = -ENOMEM); lcl_fm_ext = &fm_local->fm_extents[0]; @@ -2298,8 +2324,10 @@ static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key, last_stripe = fiemap_calc_last_stripe(lsm, fm_start, fm_end, actual_start_stripe, &stripe_count); - fm_end_offset = fiemap_calc_fm_end_offset(fiemap, lsm, fm_start, fm_end, - &start_stripe); + fm_end_offset = fiemap_calc_fm_end_offset(fiemap, lsm, fm_start, + fm_end, &start_stripe); + if (fm_end_offset == -EINVAL) + GOTO(out, rc = -EINVAL); if (fiemap->fm_extent_count == 0) { get_num_extents = 1; @@ -2364,7 +2392,7 @@ static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key, fm_local->fm_mapped_extents = 0; fm_local->fm_flags = fiemap->fm_flags; - fm_key->oa.o_id = lsm->lsm_oinfo[cur_stripe]->loi_id; + fm_key->oa.o_oi = lsm->lsm_oinfo[cur_stripe]->loi_oi; ost_index = lsm->lsm_oinfo[cur_stripe]->loi_ost_idx; if (ost_index < 0 || ost_index >=lov->desc.ld_tgt_count) @@ -2387,7 +2415,8 @@ static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key, fm_local->fm_flags &= ~FIEMAP_FLAG_DEVICE_ORDER; memcpy(&fm_key->fiemap, fm_local, sizeof(*fm_local)); *vallen=fiemap_count_to_size(fm_local->fm_extent_count); - rc = obd_get_info(lov->lov_tgts[ost_index]->ltd_exp, + rc = obd_get_info(NULL, + lov->lov_tgts[ost_index]->ltd_exp, keylen, key, vallen, fm_local, lsm); if (rc != 0) GOTO(out, rc); @@ -2467,12 +2496,13 @@ skip_last_device_calc: fiemap->fm_mapped_extents = current_extent; out: - OBD_FREE(fm_local, buffer_size); + if (fm_local) + OBD_FREE_LARGE(fm_local, buffer_size); return rc; } -static int lov_get_info(struct obd_export *exp, __u32 keylen, - void *key, __u32 *vallen, void *val, +static int lov_get_info(const struct lu_env *env, struct obd_export *exp, + __u32 keylen, void *key, __u32 *vallen, void *val, struct lov_stripe_md *lsm) { struct obd_device *obddev = class_exp2obd(exp); @@ -2508,12 +2538,12 @@ static int lov_get_info(struct obd_export *exp, __u32 keylen, loi = lsm->lsm_oinfo[i]; if (!lov->lov_tgts[loi->loi_ost_idx]) continue; - if (lov->lov_tgts[loi->loi_ost_idx]->ltd_exp == - data->lock->l_conn_export && - osc_res_name_eq(loi->loi_id, loi->loi_gr, res_id)) { - *stripe = i; - GOTO(out, rc = 0); - } + if (lov->lov_tgts[loi->loi_ost_idx]->ltd_exp == + data->lock->l_conn_export && + ostid_res_name_eq(&loi->loi_oi, res_id)) { + *stripe = i; + GOTO(out, rc = 0); + } } LDLM_ERROR(data->lock, "lock on inode without such object"); dump_lsm(D_ERROR, lsm); @@ -2529,7 +2559,8 @@ static int lov_get_info(struct obd_export *exp, __u32 keylen, if (!tgt || !tgt->ltd_active) GOTO(out, rc = -ESRCH); - rc = obd_get_info(tgt->ltd_exp, keylen, key, &size, info->data, NULL); + rc = obd_get_info(env, tgt->ltd_exp, keylen, key, + &size, info->data, NULL); GOTO(out, rc = 0); } else if (KEY_IS(KEY_LOVDESC)) { struct lov_desc *desc_ret = val; @@ -2539,6 +2570,22 @@ static int lov_get_info(struct obd_export *exp, __u32 keylen, } else if (KEY_IS(KEY_FIEMAP)) { rc = lov_fiemap(lov, keylen, key, vallen, val, lsm); GOTO(out, rc); + } else if (KEY_IS(KEY_CONNECT_FLAG)) { + struct lov_tgt_desc *tgt; + __u64 ost_idx = *((__u64*)val); + + LASSERT(*vallen == sizeof(__u64)); + LASSERT(ost_idx < lov->desc.ld_tgt_count); + tgt = lov->lov_tgts[ost_idx]; + + if (!tgt || !tgt->ltd_exp) + GOTO(out, rc = -ESRCH); + + *((__u64 *)val) = exp_connect_flags(tgt->ltd_exp); + GOTO(out, rc = 0); + } else if (KEY_IS(KEY_TGT_COUNT)) { + *((int *)val) = lov->desc.ld_tgt_count; + GOTO(out, rc = 0); } rc = -EINVAL; @@ -2548,9 +2595,9 @@ out: RETURN(rc); } -static int lov_set_info_async(struct obd_export *exp, obd_count keylen, - void *key, obd_count vallen, void *val, - struct ptlrpc_request_set *set) +static int lov_set_info_async(const struct lu_env *env, struct obd_export *exp, + obd_count keylen, void *key, obd_count vallen, + void *val, struct ptlrpc_request_set *set) { struct obd_device *obddev = class_exp2obd(exp); struct lov_obd *lov = &obddev->u.lov; @@ -2587,7 +2634,11 @@ static int lov_set_info_async(struct obd_export *exp, obd_count keylen, mds_con = 1; } else if (KEY_IS(KEY_CAPA_KEY)) { capa = 1; - } + } else if (KEY_IS(KEY_CACHE_SET)) { + LASSERT(lov->lov_cache == NULL); + lov->lov_cache = val; + do_inactive = 1; + } for (i = 0; i < count; i++, val = (char *)val + incr) { if (next_id) { @@ -2614,11 +2665,11 @@ static int lov_set_info_async(struct obd_export *exp, obd_count keylen, &tgt->ltd_uuid)) continue; - err = obd_set_info_async(tgt->ltd_exp, + err = obd_set_info_async(env, tgt->ltd_exp, keylen, key, sizeof(int), &mgi->group, set); } else if (next_id) { - err = obd_set_info_async(tgt->ltd_exp, + err = obd_set_info_async(env, tgt->ltd_exp, keylen, key, vallen, ((struct obd_id_info*)val)->data, set); } else if (capa) { @@ -2631,17 +2682,16 @@ static int lov_set_info_async(struct obd_export *exp, obd_count keylen, !obd_uuid_equals(info->uuid, &tgt->ltd_uuid)) continue; - err = obd_set_info_async(tgt->ltd_exp, keylen, key, - sizeof(*info->capa), + err = obd_set_info_async(env, tgt->ltd_exp, keylen, + key, sizeof(*info->capa), info->capa, set); - } else { /* Only want a specific OSC */ if (check_uuid && !obd_uuid_equals(val, &tgt->ltd_uuid)) continue; - err = obd_set_info_async(tgt->ltd_exp, + err = obd_set_info_async(env, tgt->ltd_exp, keylen, key, vallen, val, set); } @@ -2659,31 +2709,15 @@ static int lov_set_info_async(struct obd_export *exp, obd_count keylen, RETURN(rc); } -int lov_test_and_clear_async_rc(struct lov_stripe_md *lsm) -{ - int i, rc = 0; - ENTRY; - - for (i = 0; i < lsm->lsm_stripe_count; i++) { - struct lov_oinfo *loi = lsm->lsm_oinfo[i]; - if (loi->loi_ar.ar_rc && !rc) - rc = loi->loi_ar.ar_rc; - loi->loi_ar.ar_rc = 0; - } - RETURN(rc); -} -EXPORT_SYMBOL(lov_test_and_clear_async_rc); - - static int lov_extent_calc(struct obd_export *exp, struct lov_stripe_md *lsm, int cmd, __u64 *offset) { __u32 ssize = lsm->lsm_stripe_size; __u64 start; - start = *offset; - do_div(start, ssize); - start = start * ssize; + start = *offset; + lov_do_div64(start, ssize); + start = start * ssize; CDEBUG(D_DLMTRACE, "offset "LPU64", stripe %u, start "LPU64 ", end "LPU64"\n", *offset, ssize, start, @@ -2701,21 +2735,126 @@ static int lov_extent_calc(struct obd_export *exp, struct lov_stripe_md *lsm, void lov_stripe_lock(struct lov_stripe_md *md) { - LASSERT(md->lsm_lock_owner != cfs_curproc_pid()); - cfs_spin_lock(&md->lsm_lock); - LASSERT(md->lsm_lock_owner == 0); - md->lsm_lock_owner = cfs_curproc_pid(); + LASSERT(md->lsm_lock_owner != current_pid()); + spin_lock(&md->lsm_lock); + LASSERT(md->lsm_lock_owner == 0); + md->lsm_lock_owner = current_pid(); } EXPORT_SYMBOL(lov_stripe_lock); void lov_stripe_unlock(struct lov_stripe_md *md) { - LASSERT(md->lsm_lock_owner == cfs_curproc_pid()); - md->lsm_lock_owner = 0; - cfs_spin_unlock(&md->lsm_lock); + LASSERT(md->lsm_lock_owner == current_pid()); + md->lsm_lock_owner = 0; + spin_unlock(&md->lsm_lock); } EXPORT_SYMBOL(lov_stripe_unlock); +static int lov_quotactl(struct obd_device *obd, struct obd_export *exp, + struct obd_quotactl *oqctl) +{ + struct lov_obd *lov = &obd->u.lov; + struct lov_tgt_desc *tgt; + __u64 curspace = 0; + __u64 bhardlimit = 0; + int i, rc = 0; + ENTRY; + + if (oqctl->qc_cmd != LUSTRE_Q_QUOTAON && + oqctl->qc_cmd != LUSTRE_Q_QUOTAOFF && + oqctl->qc_cmd != Q_GETOQUOTA && + oqctl->qc_cmd != Q_INITQUOTA && + oqctl->qc_cmd != LUSTRE_Q_SETQUOTA && + oqctl->qc_cmd != Q_FINVALIDATE) { + CERROR("bad quota opc %x for lov obd", oqctl->qc_cmd); + RETURN(-EFAULT); + } + + /* for lov tgt */ + obd_getref(obd); + for (i = 0; i < lov->desc.ld_tgt_count; i++) { + int err; + + tgt = lov->lov_tgts[i]; + + if (!tgt) + continue; + + if (!tgt->ltd_active || tgt->ltd_reap) { + if (oqctl->qc_cmd == Q_GETOQUOTA && + lov->lov_tgts[i]->ltd_activate) { + rc = -EREMOTEIO; + CERROR("ost %d is inactive\n", i); + } else { + CDEBUG(D_HA, "ost %d is inactive\n", i); + } + continue; + } + + err = obd_quotactl(tgt->ltd_exp, oqctl); + if (err) { + if (tgt->ltd_active && !rc) + rc = err; + continue; + } + + if (oqctl->qc_cmd == Q_GETOQUOTA) { + curspace += oqctl->qc_dqblk.dqb_curspace; + bhardlimit += oqctl->qc_dqblk.dqb_bhardlimit; + } + } + obd_putref(obd); + + if (oqctl->qc_cmd == Q_GETOQUOTA) { + oqctl->qc_dqblk.dqb_curspace = curspace; + oqctl->qc_dqblk.dqb_bhardlimit = bhardlimit; + } + RETURN(rc); +} + +static int lov_quotacheck(struct obd_device *obd, struct obd_export *exp, + struct obd_quotactl *oqctl) +{ + struct lov_obd *lov = &obd->u.lov; + int i, rc = 0; + ENTRY; + + obd_getref(obd); + + for (i = 0; i < lov->desc.ld_tgt_count; i++) { + if (!lov->lov_tgts[i]) + continue; + + /* Skip quota check on the administratively disabled OSTs. */ + if (!lov->lov_tgts[i]->ltd_activate) { + CWARN("lov idx %d was administratively disabled, " + "skip quotacheck on it.\n", i); + continue; + } + + if (!lov->lov_tgts[i]->ltd_active) { + CERROR("lov idx %d inactive\n", i); + rc = -EIO; + goto out; + } + } + + for (i = 0; i < lov->desc.ld_tgt_count; i++) { + int err; + + if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_activate) + continue; + + err = obd_quotacheck(lov->lov_tgts[i]->ltd_exp, oqctl); + if (err && !rc) + rc = err; + } + +out: + obd_putref(obd); + + RETURN(rc); +} struct obd_ops lov_obd_ops = { .o_owner = THIS_MODULE, @@ -2742,6 +2881,7 @@ struct obd_ops lov_obd_ops = { .o_sync = lov_sync, .o_enqueue = lov_enqueue, .o_change_cbdata = lov_change_cbdata, + .o_find_cbdata = lov_find_cbdata, .o_cancel = lov_cancel, .o_cancel_unused = lov_cancel_unused, .o_iocontrol = lov_iocontrol, @@ -2757,51 +2897,43 @@ struct obd_ops lov_obd_ops = { .o_pool_del = lov_pool_del, .o_getref = lov_getref, .o_putref = lov_putref, + .o_quotactl = lov_quotactl, + .o_quotacheck = lov_quotacheck, }; -static quota_interface_t *quota_interface; -extern quota_interface_t lov_quota_interface; - -cfs_mem_cache_t *lov_oinfo_slab; +struct kmem_cache *lov_oinfo_slab; extern struct lu_kmem_descr lov_caches[]; int __init lov_init(void) { struct lprocfs_static_vars lvars = { 0 }; - int rc, rc2; + int rc; ENTRY; /* print an address of _any_ initialized kernel symbol from this * module, to allow debugging with gdb that doesn't support data * symbols from modules.*/ - CDEBUG(D_CONSOLE, "Lustre LOV module (%p).\n", &lov_caches); + CDEBUG(D_INFO, "Lustre LOV module (%p).\n", &lov_caches); rc = lu_kmem_init(lov_caches); if (rc) return rc; - lov_oinfo_slab = cfs_mem_cache_create("lov_oinfo", - sizeof(struct lov_oinfo), - 0, CFS_SLAB_HWCACHE_ALIGN); + lov_oinfo_slab = kmem_cache_create("lov_oinfo", + sizeof(struct lov_oinfo), 0, + SLAB_HWCACHE_ALIGN, NULL); if (lov_oinfo_slab == NULL) { lu_kmem_fini(lov_caches); return -ENOMEM; } lprocfs_lov_init_vars(&lvars); - cfs_request_module("lquota"); - quota_interface = PORTAL_SYMBOL_GET(lov_quota_interface); - init_obd_quota_ops(quota_interface, &lov_obd_ops); - rc = class_register_type(&lov_obd_ops, NULL, lvars.module_vars, LUSTRE_LOV_NAME, &lov_device_type); if (rc) { - if (quota_interface) - PORTAL_SYMBOL_PUT(lov_quota_interface); - rc2 = cfs_mem_cache_destroy(lov_oinfo_slab); - LASSERT(rc2 == 0); + kmem_cache_destroy(lov_oinfo_slab); lu_kmem_fini(lov_caches); } @@ -2811,17 +2943,9 @@ int __init lov_init(void) #ifdef __KERNEL__ static void /*__exit*/ lov_exit(void) { - int rc; - - lu_device_type_fini(&lov_device_type); - lu_kmem_fini(lov_caches); - - if (quota_interface) - PORTAL_SYMBOL_PUT(lov_quota_interface); - - class_unregister_type(LUSTRE_LOV_NAME); - rc = cfs_mem_cache_destroy(lov_oinfo_slab); - LASSERT(rc == 0); + class_unregister_type(LUSTRE_LOV_NAME); + kmem_cache_destroy(lov_oinfo_slab); + lu_kmem_fini(lov_caches); } MODULE_AUTHOR("Sun Microsystems, Inc. ");