X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Flov%2Flov_io.c;h=7ac4590ff9226c6d5aaff75eb71726f7c032d3a6;hp=b21fd941a77803200cce30d6fcc624cfc09075e3;hb=893e5a99e766ab97e95217b7b201b9d293bed6e0;hpb=0b72da3f303132ec09d80c65156c7a20fa03c721 diff --git a/lustre/lov/lov_io.c b/lustre/lov/lov_io.c index b21fd94..7ac4590f 100644 --- a/lustre/lov/lov_io.c +++ b/lustre/lov/lov_io.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. @@ -17,17 +15,15 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -36,6 +32,7 @@ * Implementation of cl_io for LOV layer. * * Author: Nikita Danilov + * Author: Jinshan Xiong */ #define DEBUG_SUBSYSTEM S_LOV @@ -46,194 +43,145 @@ * @{ */ -static void lov_sub_enter(struct lov_io_sub *sub) +static inline struct lov_io_sub *lov_sub_alloc(struct lov_io *lio, int index) { - ENTRY; - if (sub->sub_reenter++ == 0) { - sub->sub_cookie = cl_env_reenter(); - cl_env_implant(sub->sub_env, &sub->sub_refcheck2); - } - EXIT; + struct lov_io_sub *sub; + + if (lio->lis_nr_subios == 0) { + LASSERT(lio->lis_single_subio_index == -1); + sub = &lio->lis_single_subio; + lio->lis_single_subio_index = index; + memset(sub, 0, sizeof(*sub)); + } else { + OBD_ALLOC_PTR(sub); + } + + if (sub != NULL) { + INIT_LIST_HEAD(&sub->sub_list); + INIT_LIST_HEAD(&sub->sub_linkage); + sub->sub_subio_index = index; + } + + return sub; } -static void lov_sub_exit(struct lov_io_sub *sub) +static inline void lov_sub_free(struct lov_io *lio, struct lov_io_sub *sub) { - ENTRY; - if (--sub->sub_reenter == 0) { - cl_env_unplant(sub->sub_env, &sub->sub_refcheck2); - cl_env_reexit(sub->sub_cookie); - } - EXIT; + if (sub->sub_subio_index == lio->lis_single_subio_index) { + LASSERT(sub == &lio->lis_single_subio); + lio->lis_single_subio_index = -1; + } else { + OBD_FREE_PTR(sub); + } } static void lov_io_sub_fini(const struct lu_env *env, struct lov_io *lio, - struct lov_io_sub *sub) -{ - ENTRY; - if (sub->sub_io != NULL) { - if (sub->sub_io_initialized) { - lov_sub_enter(sub); - cl_io_fini(sub->sub_env, sub->sub_io); - lov_sub_exit(sub); - sub->sub_io_initialized = 0; - lio->lis_active_subios--; - } - if (sub->sub_stripe == lio->lis_single_subio_index) - lio->lis_single_subio_index = -1; - else if (!sub->sub_borrowed) - OBD_FREE_PTR(sub->sub_io); - sub->sub_io = NULL; - } - if (sub->sub_env != NULL && !IS_ERR(sub->sub_env)) { - if (!sub->sub_borrowed) - cl_env_put(sub->sub_env, &sub->sub_refcheck); - sub->sub_env = NULL; - } - EXIT; -} - -static void lov_io_sub_inherit(struct cl_io *io, struct lov_io *lio, - int stripe, loff_t start, loff_t end) + struct lov_io_sub *sub) { - struct lov_stripe_md *lsm = lov_r0(lio->lis_object)->lo_lsm; - struct cl_io *parent = lio->lis_cl.cis_io; + ENTRY; - switch(io->ci_type) { - case CIT_TRUNC: { - loff_t new_size = parent->u.ci_truncate.tr_size; + cl_io_fini(sub->sub_env, &sub->sub_io); - new_size = lov_size_to_stripe(lsm, new_size, stripe); - io->u.ci_truncate.tr_capa = parent->u.ci_truncate.tr_capa; - io->u.ci_truncate.tr_size = new_size; - break; - } - case CIT_FAULT: { - struct cl_object *obj = parent->ci_obj; - loff_t off = cl_offset(obj, parent->u.ci_fault.ft_index); - - io->u.ci_fault = parent->u.ci_fault; - off = lov_size_to_stripe(lsm, off, stripe); - io->u.ci_fault.ft_index = cl_index(obj, off); - break; - } - case CIT_READ: - case CIT_WRITE: { - if (cl_io_is_append(parent)) { - io->u.ci_wr.wr_append = 1; - } else { - io->u.ci_rw.crw_pos = start; - io->u.ci_rw.crw_count = end - start; - } - break; - } - default: - break; - } + if (sub->sub_env != NULL && !IS_ERR(sub->sub_env)) { + cl_env_put(sub->sub_env, &sub->sub_refcheck); + sub->sub_env = NULL; + } + EXIT; } -static int lov_io_sub_init(const struct lu_env *env, struct lov_io *lio, - struct lov_io_sub *sub) +static inline bool +is_index_within_mirror(struct lov_object *lov, int index, int mirror_index) { - struct lov_object *lov = lio->lis_object; - struct lov_device *ld = lu2lov_dev(lov2cl(lov)->co_lu.lo_dev); - struct cl_io *sub_io; - struct cl_object *sub_obj; - struct cl_io *io = lio->lis_cl.cis_io; - - int stripe = sub->sub_stripe; - int result; - - LASSERT(sub->sub_io == NULL); - LASSERT(sub->sub_env == NULL); - LASSERT(sub->sub_stripe < lio->lis_stripe_count); - ENTRY; - - result = 0; - sub->sub_io_initialized = 0; - sub->sub_borrowed = 0; - - if (lio->lis_mem_frozen) { - LASSERT(cfs_mutex_is_locked(&ld->ld_mutex)); - sub->sub_io = &ld->ld_emrg[stripe]->emrg_subio; - sub->sub_env = ld->ld_emrg[stripe]->emrg_env; - sub->sub_borrowed = 1; - } else { - void *cookie; - - /* obtain new environment */ - cookie = cl_env_reenter(); - sub->sub_env = cl_env_get(&sub->sub_refcheck); - cl_env_reexit(cookie); - if (IS_ERR(sub->sub_env)) - result = PTR_ERR(sub->sub_env); - - if (result == 0) { - /* - * First sub-io. Use ->lis_single_subio to - * avoid dynamic allocation. - */ - if (lio->lis_active_subios == 0) { - sub->sub_io = &lio->lis_single_subio; - lio->lis_single_subio_index = stripe; - } else { - OBD_ALLOC_PTR(sub->sub_io); - if (sub->sub_io == NULL) - result = -ENOMEM; - } - } - } + struct lov_layout_composite *comp = &lov->u.composite; + struct lov_mirror_entry *lre = &comp->lo_mirrors[mirror_index]; - if (result == 0) { - sub_obj = lovsub2cl(lov_r0(lov)->lo_sub[stripe]); - sub_io = sub->sub_io; - - sub_io->ci_obj = sub_obj; - sub_io->ci_result = 0; - - sub_io->ci_parent = io; - sub_io->ci_lockreq = io->ci_lockreq; - sub_io->ci_type = io->ci_type; - sub_io->ci_no_srvlock = io->ci_no_srvlock; - - lov_sub_enter(sub); - result = cl_io_sub_init(sub->sub_env, sub_io, - io->ci_type, sub_obj); - lov_sub_exit(sub); - if (result >= 0) { - lio->lis_active_subios++; - sub->sub_io_initialized = 1; - result = 0; - } - } - if (result != 0) - lov_io_sub_fini(env, lio, sub); - RETURN(result); + return (index >= lre->lre_start && index <= lre->lre_end); } -struct lov_io_sub *lov_sub_get(const struct lu_env *env, - struct lov_io *lio, int stripe) +static int lov_io_sub_init(const struct lu_env *env, struct lov_io *lio, + struct lov_io_sub *sub) { - int rc; - struct lov_io_sub *sub = &lio->lis_subs[stripe]; - - LASSERT(stripe < lio->lis_stripe_count); - ENTRY; - - if (!sub->sub_io_initialized) { - sub->sub_stripe = stripe; - rc = lov_io_sub_init(env, lio, sub); - } else - rc = 0; - if (rc == 0) - lov_sub_enter(sub); - else - sub = ERR_PTR(rc); - RETURN(sub); + struct lov_object *lov = lio->lis_object; + struct cl_io *sub_io; + struct cl_object *sub_obj; + struct cl_io *io = lio->lis_cl.cis_io; + int index = lov_comp_entry(sub->sub_subio_index); + int stripe = lov_comp_stripe(sub->sub_subio_index); + int result = 0; + LASSERT(sub->sub_env == NULL); + ENTRY; + + if (unlikely(!lov_r0(lov, index)->lo_sub || + !lov_r0(lov, index)->lo_sub[stripe])) + RETURN(-EIO); + + LASSERTF(is_index_within_mirror(lov, index, lio->lis_mirror_index), + DFID "iot = %d, index = %d, mirror = %d\n", + PFID(lu_object_fid(lov2lu(lov))), io->ci_type, index, + lio->lis_mirror_index); + + /* obtain new environment */ + sub->sub_env = cl_env_get(&sub->sub_refcheck); + if (IS_ERR(sub->sub_env)) + result = PTR_ERR(sub->sub_env); + + sub_obj = lovsub2cl(lov_r0(lov, index)->lo_sub[stripe]); + sub_io = &sub->sub_io; + + sub_io->ci_obj = sub_obj; + sub_io->ci_result = 0; + + sub_io->ci_parent = io; + sub_io->ci_lockreq = io->ci_lockreq; + sub_io->ci_type = io->ci_type; + sub_io->ci_no_srvlock = io->ci_no_srvlock; + sub_io->ci_noatime = io->ci_noatime; + sub_io->ci_pio = io->ci_pio; + sub_io->ci_lock_no_expand = io->ci_lock_no_expand; + sub_io->ci_ndelay = io->ci_ndelay; + sub_io->ci_layout_version = io->ci_layout_version; + + result = cl_io_sub_init(sub->sub_env, sub_io, io->ci_type, sub_obj); + + if (result < 0) + lov_io_sub_fini(env, lio, sub); + + RETURN(result); } -void lov_sub_put(struct lov_io_sub *sub) +struct lov_io_sub *lov_sub_get(const struct lu_env *env, + struct lov_io *lio, int index) { - lov_sub_exit(sub); + struct lov_io_sub *sub; + int rc = 0; + + ENTRY; + + list_for_each_entry(sub, &lio->lis_subios, sub_list) { + if (sub->sub_subio_index == index) { + rc = 1; + break; + } + } + + if (rc == 0) { + sub = lov_sub_alloc(lio, index); + if (sub == NULL) + GOTO(out, rc = -ENOMEM); + + rc = lov_io_sub_init(env, lio, sub); + if (rc < 0) { + lov_sub_free(lio, sub); + GOTO(out, rc); + } + + list_add_tail(&sub->sub_list, &lio->lis_subios); + lio->lis_nr_subios++; + } +out: + if (rc < 0) + sub = ERR_PTR(rc); + RETURN(sub); } /***************************************************************************** @@ -242,92 +190,317 @@ void lov_sub_put(struct lov_io_sub *sub) * */ -static int lov_page_stripe(const struct cl_page *page) +int lov_page_index(const struct cl_page *page) { - struct lovsub_object *subobj; + const struct cl_page_slice *slice; + ENTRY; - ENTRY; - subobj = lu2lovsub( - lu_object_locate(page->cp_child->cp_obj->co_lu.lo_header, - &lovsub_device_type)); - LASSERT(subobj != NULL); - RETURN(subobj->lso_index); -} - -struct lov_io_sub *lov_page_subio(const struct lu_env *env, struct lov_io *lio, - const struct cl_page_slice *slice) -{ - struct lov_stripe_md *lsm = lov_r0(lio->lis_object)->lo_lsm; - struct cl_page *page = slice->cpl_page; - int stripe; + slice = cl_page_at(page, &lov_device_type); + LASSERT(slice != NULL); + LASSERT(slice->cpl_obj != NULL); - LASSERT(lio->lis_cl.cis_io != NULL); - LASSERT(cl2lov(slice->cpl_obj) == lio->lis_object); - LASSERT(lsm != NULL); - LASSERT(lio->lis_nr_subios > 0); - ENTRY; - - stripe = lov_page_stripe(page); - RETURN(lov_sub_get(env, lio, stripe)); + RETURN(cl2lov_page(slice)->lps_index); } - static int lov_io_subio_init(const struct lu_env *env, struct lov_io *lio, struct cl_io *io) { - struct lov_object *lov = lio->lis_object; - struct lov_stripe_md *lsm = lov_r0(lov)->lo_lsm; - int result; + ENTRY; - LASSERT(lio->lis_object != NULL); - ENTRY; + LASSERT(lio->lis_object != NULL); - /* - * Need to be optimized, we can't afford to allocate a piece of memory - * when writing a page. -jay - */ - OBD_ALLOC(lio->lis_subs, - lsm->lsm_stripe_count * sizeof lio->lis_subs[0]); - if (lio->lis_subs != NULL) { - lio->lis_nr_subios = lio->lis_stripe_count; - lio->lis_single_subio_index = -1; - lio->lis_active_subios = 0; - result = 0; - } else - result = -ENOMEM; - RETURN(result); + INIT_LIST_HEAD(&lio->lis_subios); + lio->lis_single_subio_index = -1; + lio->lis_nr_subios = 0; + + RETURN(0); } -static void lov_io_slice_init(struct lov_io *lio, - struct lov_object *obj, struct cl_io *io) +/** + * Decide if it will need write intent RPC + */ +static int lov_io_mirror_write_intent(struct lov_io *lio, + struct lov_object *obj, struct cl_io *io) { - struct lov_stripe_md *lsm = lov_r0(obj)->lo_lsm; + struct lov_layout_composite *comp = &obj->u.composite; + struct lu_extent *ext = &io->ci_write_intent; + struct lov_mirror_entry *lre; + struct lov_mirror_entry *primary; + struct lov_layout_entry *lle; + size_t count = 0; + ENTRY; + + *ext = (typeof(*ext)) { lio->lis_pos, lio->lis_endpos }; + io->ci_need_write_intent = 0; + + if (!(io->ci_type == CIT_WRITE || cl_io_is_trunc(io) || + cl_io_is_mkwrite(io))) + RETURN(0); + + /* FLR: check if it needs to send a write intent RPC to server. + * Writing to sync_pending file needs write intent RPC to change + * the file state back to write_pending, so that the layout version + * can be increased when the state changes to sync_pending at a later + * time. Otherwise there exists a chance that an evicted client may + * dirty the file data while resync client is working on it. + * Designated I/O is allowed for resync workload. + */ + if (lov_flr_state(obj) == LCM_FL_RDONLY || + (lov_flr_state(obj) == LCM_FL_SYNC_PENDING && + io->ci_designated_mirror == 0)) { + io->ci_need_write_intent = 1; + RETURN(0); + } + + LASSERT((lov_flr_state(obj) == LCM_FL_WRITE_PENDING)); + LASSERT(comp->lo_preferred_mirror >= 0); + + /* need to iterate all components to see if there are + * multiple components covering the writing component */ + primary = &comp->lo_mirrors[comp->lo_preferred_mirror]; + LASSERT(!primary->lre_stale); + lov_foreach_mirror_layout_entry(obj, lle, primary) { + LASSERT(lle->lle_valid); + if (!lu_extent_is_overlapped(ext, lle->lle_extent)) + continue; + + ext->e_start = MIN(ext->e_start, lle->lle_extent->e_start); + ext->e_end = MAX(ext->e_end, lle->lle_extent->e_end); + ++count; + } + if (count == 0) { + CERROR(DFID ": cannot find any valid components covering " + "file extent "DEXT", mirror: %d\n", + PFID(lu_object_fid(lov2lu(obj))), PEXT(ext), + primary->lre_mirror_id); + RETURN(-EIO); + } + + count = 0; + lov_foreach_mirror_entry(obj, lre) { + if (lre == primary) + continue; + + lov_foreach_mirror_layout_entry(obj, lle, lre) { + if (!lle->lle_valid) + continue; + + if (lu_extent_is_overlapped(ext, lle->lle_extent)) { + ++count; + break; + } + } + } + + CDEBUG(D_VFSTRACE, DFID "there are %zd components to be staled to " + "modify file extent "DEXT", iot: %d\n", + PFID(lu_object_fid(lov2lu(obj))), count, PEXT(ext), io->ci_type); + + io->ci_need_write_intent = count > 0; + + RETURN(0); +} - LASSERT(lsm != NULL); - ENTRY; +static int lov_io_mirror_init(struct lov_io *lio, struct lov_object *obj, + struct cl_io *io) +{ + struct lov_layout_composite *comp = &obj->u.composite; + int index; + int i; + int result; + ENTRY; + + if (!lov_is_flr(obj)) { + LASSERT(comp->lo_preferred_mirror == 0); + lio->lis_mirror_index = comp->lo_preferred_mirror; + io->ci_ndelay = 0; + RETURN(0); + } + + /* transfer the layout version for verification */ + if (io->ci_layout_version == 0) + io->ci_layout_version = obj->lo_lsm->lsm_layout_gen; + + /* find the corresponding mirror for designated mirror IO */ + if (io->ci_designated_mirror > 0) { + struct lov_mirror_entry *entry; + + LASSERT(!io->ci_ndelay); + + CDEBUG(D_LAYOUT, "designated I/O mirror state: %d\n", + lov_flr_state(obj)); + + if ((cl_io_is_trunc(io) || io->ci_type == CIT_WRITE) && + (io->ci_layout_version != obj->lo_lsm->lsm_layout_gen)) { + /* For resync I/O, the ci_layout_version was the layout + * version when resync starts. If it doesn't match the + * current object layout version, it means the layout + * has been changed */ + RETURN(-ESTALE); + } + + io->ci_layout_version |= LU_LAYOUT_RESYNC; + + index = 0; + lio->lis_mirror_index = -1; + lov_foreach_mirror_entry(obj, entry) { + if (entry->lre_mirror_id == + io->ci_designated_mirror) { + lio->lis_mirror_index = index; + break; + } + + index++; + } + + RETURN(lio->lis_mirror_index < 0 ? -EINVAL : 0); + } + + result = lov_io_mirror_write_intent(lio, obj, io); + if (result) + RETURN(result); + + if (io->ci_need_write_intent) { + CDEBUG(D_VFSTRACE, DFID " need write intent for [%llu, %llu)\n", + PFID(lu_object_fid(lov2lu(obj))), + lio->lis_pos, lio->lis_endpos); + + if (cl_io_is_trunc(io)) { + /** + * for truncate, we uses [size, EOF) to judge whether + * a write intent needs to be send, but we need to + * restore the write extent to [0, size). + */ + io->ci_write_intent.e_start = 0; + io->ci_write_intent.e_end = + io->u.ci_setattr.sa_attr.lvb_size; + } + /* stop cl_io_init() loop */ + RETURN(1); + } + + if (io->ci_ndelay_tried == 0 || /* first time to try */ + /* reset the mirror index if layout has changed */ + lio->lis_mirror_layout_gen != obj->lo_lsm->lsm_layout_gen) { + lio->lis_mirror_layout_gen = obj->lo_lsm->lsm_layout_gen; + index = lio->lis_mirror_index = comp->lo_preferred_mirror; + } else { + index = lio->lis_mirror_index; + LASSERT(index >= 0); + + /* move mirror index to the next one */ + index = (index + 1) % comp->lo_mirror_count; + } + + for (i = 0; i < comp->lo_mirror_count; i++) { + struct lu_extent ext = { .e_start = lio->lis_pos, + .e_end = lio->lis_pos + 1 }; + struct lov_mirror_entry *lre; + struct lov_layout_entry *lle; + bool found = false; + + lre = &comp->lo_mirrors[(index + i) % comp->lo_mirror_count]; + if (!lre->lre_valid) + continue; + + lov_foreach_mirror_layout_entry(obj, lle, lre) { + if (!lle->lle_valid) + continue; + + if (lu_extent_is_overlapped(&ext, lle->lle_extent)) { + found = true; + break; + } + } + + if (found) { + index = (index + i) % comp->lo_mirror_count; + break; + } + } + if (i == comp->lo_mirror_count) { + CERROR(DFID": failed to find a component covering " + "I/O region at %llu\n", + PFID(lu_object_fid(lov2lu(obj))), lio->lis_pos); + + dump_lsm(D_ERROR, obj->lo_lsm); + + RETURN(-EIO); + } + + CDEBUG(D_VFSTRACE, DFID ": flr state: %d, move mirror from %d to %d, " + "have retried: %d, mirror count: %d\n", + PFID(lu_object_fid(lov2lu(obj))), lov_flr_state(obj), + lio->lis_mirror_index, index, io->ci_ndelay_tried, + comp->lo_mirror_count); + + lio->lis_mirror_index = index; + + /* FLR: if all mirrors have been tried once, most likely the network + * of this client has been partitioned. We should relinquish CPU for + * a while before trying again. */ + ++io->ci_ndelay_tried; + if (io->ci_ndelay && io->ci_ndelay_tried >= comp->lo_mirror_count) { + set_current_state(TASK_INTERRUPTIBLE); + schedule_timeout(msecs_to_jiffies(MSEC_PER_SEC)); /* 10ms */ + if (signal_pending(current)) + RETURN(-EINTR); + + /* reset retry counter */ + io->ci_ndelay_tried = 1; + } + + CDEBUG(D_VFSTRACE, "use %sdelayed RPC state for this IO\n", + io->ci_ndelay ? "non-" : ""); + + RETURN(0); +} - io->ci_result = 0; - lio->lis_object = obj; - lio->lis_stripe_count = lsm->lsm_stripe_count; - - switch (io->ci_type) { - case CIT_READ: - case CIT_WRITE: - lio->lis_pos = io->u.ci_rw.crw_pos; - lio->lis_endpos = io->u.ci_rw.crw_pos + io->u.ci_rw.crw_count; - lio->lis_io_endpos = lio->lis_endpos; - if (cl_io_is_append(io)) { - LASSERT(io->ci_type == CIT_WRITE); +static int lov_io_slice_init(struct lov_io *lio, + struct lov_object *obj, struct cl_io *io) +{ + int index; + int result = 0; + ENTRY; + + io->ci_result = 0; + lio->lis_object = obj; + + LASSERT(obj->lo_lsm != NULL); + + switch (io->ci_type) { + case CIT_READ: + case CIT_WRITE: + lio->lis_pos = io->u.ci_rw.rw_range.cir_pos; + lio->lis_endpos = lio->lis_pos + io->u.ci_rw.rw_range.cir_count; + lio->lis_io_endpos = lio->lis_endpos; + if (cl_io_is_append(io)) { + LASSERT(io->ci_type == CIT_WRITE); + + /* If there is LOV EA hole, then we may cannot locate + * the current file-tail exactly. */ + if (unlikely(obj->lo_lsm->lsm_entries[0]->lsme_pattern & + LOV_PATTERN_F_HOLE)) + GOTO(out, result = -EIO); + + lio->lis_pos = 0; + lio->lis_endpos = OBD_OBJECT_EOF; + } + break; + + case CIT_SETATTR: + if (cl_io_is_trunc(io)) + lio->lis_pos = io->u.ci_setattr.sa_attr.lvb_size; + else lio->lis_pos = 0; - lio->lis_endpos = OBD_OBJECT_EOF; - } - break; - - case CIT_TRUNC: - lio->lis_pos = io->u.ci_truncate.tr_size; lio->lis_endpos = OBD_OBJECT_EOF; break; + case CIT_DATA_VERSION: + lio->lis_pos = 0; + lio->lis_endpos = OBD_OBJECT_EOF; + break; + case CIT_FAULT: { pgoff_t index = io->u.ci_fault.ft_index; lio->lis_pos = cl_offset(io->ci_obj, index); @@ -335,6 +508,28 @@ static void lov_io_slice_init(struct lov_io *lio, break; } + case CIT_FSYNC: { + lio->lis_pos = io->u.ci_fsync.fi_start; + lio->lis_endpos = io->u.ci_fsync.fi_end; + break; + } + + case CIT_LADVISE: { + lio->lis_pos = io->u.ci_ladvise.li_start; + lio->lis_endpos = io->u.ci_ladvise.li_end; + break; + } + + case CIT_GLIMPSE: + lio->lis_pos = 0; + lio->lis_endpos = OBD_OBJECT_EOF; + + if (lov_flr_state(obj) == LCM_FL_RDONLY && + !OBD_FAIL_CHECK(OBD_FAIL_FLR_GLIMPSE_IMMUTABLE)) + /* SoM is accurate, no need glimpse */ + GOTO(out, result = 1); + break; + case CIT_MISC: lio->lis_pos = 0; lio->lis_endpos = OBD_OBJECT_EOF; @@ -344,26 +539,166 @@ static void lov_io_slice_init(struct lov_io *lio, LBUG(); } - EXIT; + result = lov_io_mirror_init(lio, obj, io); + if (result) + GOTO(out, result); + + /* check if it needs to instantiate layout */ + if (!(io->ci_type == CIT_WRITE || cl_io_is_mkwrite(io) || + (cl_io_is_trunc(io) && io->u.ci_setattr.sa_attr.lvb_size > 0))) + GOTO(out, result = 0); + + /* for truncate, it only needs to instantiate the components + * before the truncated size. */ + if (cl_io_is_trunc(io)) { + io->ci_write_intent.e_start = 0; + io->ci_write_intent.e_end = io->u.ci_setattr.sa_attr.lvb_size; + } else { + io->ci_write_intent.e_start = lio->lis_pos; + io->ci_write_intent.e_end = lio->lis_endpos; + } + + index = 0; + lov_foreach_io_layout(index, lio, &io->ci_write_intent) { + if (!lsm_entry_inited(obj->lo_lsm, index)) { + io->ci_need_write_intent = 1; + break; + } + } + + if (io->ci_need_write_intent && io->ci_designated_mirror > 0) { + /* REINT_SYNC RPC has already tried to instantiate all of the + * components involved, obviously it didn't succeed. Skip this + * mirror for now. The server won't be able to figure out + * which mirror it should instantiate components */ + CERROR(DFID": trying to instantiate components for designated " + "I/O, file state: %d\n", + PFID(lu_object_fid(lov2lu(obj))), lov_flr_state(obj)); + + io->ci_need_write_intent = 0; + GOTO(out, result = -EIO); + } + + if (io->ci_need_write_intent) + GOTO(out, result = 1); + + EXIT; + +out: + return result; } static void lov_io_fini(const struct lu_env *env, const struct cl_io_slice *ios) { - struct lov_io *lio = cl2lov_io(env, ios); - int i; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_object *lov = cl2lov(ios->cis_obj); - ENTRY; - if (lio->lis_subs != NULL) { - for (i = 0; i < lio->lis_nr_subios; i++) - lov_io_sub_fini(env, lio, &lio->lis_subs[i]); - OBD_FREE(lio->lis_subs, - lio->lis_nr_subios * sizeof lio->lis_subs[0]); - lio->lis_nr_subios = 0; - } - EXIT; + ENTRY; + + LASSERT(list_empty(&lio->lis_active)); + + while (!list_empty(&lio->lis_subios)) { + struct lov_io_sub *sub = list_entry(lio->lis_subios.next, + struct lov_io_sub, + sub_list); + + list_del_init(&sub->sub_list); + lio->lis_nr_subios--; + + lov_io_sub_fini(env, lio, sub); + lov_sub_free(lio, sub); + } + LASSERT(lio->lis_nr_subios == 0); + + LASSERT(atomic_read(&lov->lo_active_ios) > 0); + if (atomic_dec_and_test(&lov->lo_active_ios)) + wake_up_all(&lov->lo_waitq); + EXIT; +} + +static void lov_io_sub_inherit(struct lov_io_sub *sub, struct lov_io *lio, + loff_t start, loff_t end) +{ + struct cl_io *io = &sub->sub_io; + struct lov_stripe_md *lsm = lio->lis_object->lo_lsm; + struct cl_io *parent = lio->lis_cl.cis_io; + int index = lov_comp_entry(sub->sub_subio_index); + int stripe = lov_comp_stripe(sub->sub_subio_index); + + io->ci_pio = parent->ci_pio; + switch (io->ci_type) { + case CIT_SETATTR: { + io->u.ci_setattr.sa_attr = parent->u.ci_setattr.sa_attr; + io->u.ci_setattr.sa_attr_flags = + parent->u.ci_setattr.sa_attr_flags; + io->u.ci_setattr.sa_valid = parent->u.ci_setattr.sa_valid; + io->u.ci_setattr.sa_stripe_index = stripe; + io->u.ci_setattr.sa_parent_fid = + parent->u.ci_setattr.sa_parent_fid; + if (cl_io_is_trunc(io)) { + loff_t new_size = parent->u.ci_setattr.sa_attr.lvb_size; + + new_size = lov_size_to_stripe(lsm, index, new_size, + stripe); + io->u.ci_setattr.sa_attr.lvb_size = new_size; + } + lov_lsm2layout(lsm, lsm->lsm_entries[index], + &io->u.ci_setattr.sa_layout); + break; + } + case CIT_DATA_VERSION: { + io->u.ci_data_version.dv_data_version = 0; + io->u.ci_data_version.dv_flags = + parent->u.ci_data_version.dv_flags; + break; + } + case CIT_FAULT: { + struct cl_object *obj = parent->ci_obj; + loff_t off = cl_offset(obj, parent->u.ci_fault.ft_index); + + io->u.ci_fault = parent->u.ci_fault; + off = lov_size_to_stripe(lsm, index, off, stripe); + io->u.ci_fault.ft_index = cl_index(obj, off); + break; + } + case CIT_FSYNC: { + io->u.ci_fsync.fi_start = start; + io->u.ci_fsync.fi_end = end; + io->u.ci_fsync.fi_fid = parent->u.ci_fsync.fi_fid; + io->u.ci_fsync.fi_mode = parent->u.ci_fsync.fi_mode; + break; + } + case CIT_READ: + case CIT_WRITE: { + io->u.ci_rw.rw_ptask = parent->u.ci_rw.rw_ptask; + io->u.ci_rw.rw_iter = parent->u.ci_rw.rw_iter; + io->u.ci_rw.rw_iocb = parent->u.ci_rw.rw_iocb; + io->u.ci_rw.rw_file = parent->u.ci_rw.rw_file; + io->u.ci_rw.rw_sync = parent->u.ci_rw.rw_sync; + if (cl_io_is_append(parent)) { + io->u.ci_rw.rw_append = 1; + } else { + io->u.ci_rw.rw_range.cir_pos = start; + io->u.ci_rw.rw_range.cir_count = end - start; + } + break; + } + case CIT_LADVISE: { + io->u.ci_ladvise.li_start = start; + io->u.ci_ladvise.li_end = end; + io->u.ci_ladvise.li_fid = parent->u.ci_ladvise.li_fid; + io->u.ci_ladvise.li_advice = parent->u.ci_ladvise.li_advice; + io->u.ci_ladvise.li_flags = parent->u.ci_ladvise.li_flags; + break; + } + case CIT_GLIMPSE: + case CIT_MISC: + default: + break; + } } -static obd_off lov_offset_mod(obd_off val, int delta) +static loff_t lov_offset_mod(loff_t val, int delta) { if (val != OBD_OBJECT_EOF) val += delta; @@ -371,96 +706,195 @@ static obd_off lov_offset_mod(obd_off val, int delta) } static int lov_io_iter_init(const struct lu_env *env, - const struct cl_io_slice *ios) + const struct cl_io_slice *ios) { - struct lov_io *lio = cl2lov_io(env, ios); - struct lov_stripe_md *lsm = lov_r0(lio->lis_object)->lo_lsm; - struct lov_io_sub *sub; - obd_off endpos; - obd_off start; - obd_off end; - int stripe; - int rc = 0; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_stripe_md *lsm = lio->lis_object->lo_lsm; + struct lov_io_sub *sub; + struct lu_extent ext; + int index; + int rc = 0; ENTRY; - endpos = lov_offset_mod(lio->lis_endpos, -1); - for (stripe = 0; stripe < lio->lis_stripe_count; stripe++) { - if (!lov_stripe_intersects(lsm, stripe, lio->lis_pos, - endpos, &start, &end)) - continue; - - end = lov_offset_mod(end, +1); - sub = lov_sub_get(env, lio, stripe); - if (!IS_ERR(sub)) { - lov_io_sub_inherit(sub->sub_io, lio, stripe, - start, end); - rc = cl_io_iter_init(sub->sub_env, sub->sub_io); - lov_sub_put(sub); - CDEBUG(D_VFSTRACE, "shrink: %i [%llu, %llu)\n", - stripe, start, end); - } else - rc = PTR_ERR(sub); - - if (!rc) - cfs_list_add_tail(&sub->sub_linkage, &lio->lis_active); - else - break; - } - RETURN(rc); + + ext.e_start = lio->lis_pos; + ext.e_end = lio->lis_endpos; + + lov_foreach_io_layout(index, lio, &ext) { + struct lov_layout_entry *le = lov_entry(lio->lis_object, index); + struct lov_layout_raid0 *r0 = &le->lle_raid0; + u64 start; + u64 end; + int stripe; + + CDEBUG(D_VFSTRACE, "component[%d] flags %#x\n", + index, lsm->lsm_entries[index]->lsme_flags); + if (!lsm_entry_inited(lsm, index)) { + /* Read from uninitialized components should return + * zero filled pages. */ + continue; + } + + if (!le->lle_valid && !ios->cis_io->ci_designated_mirror) { + CERROR("I/O to invalid component: %d, mirror: %d\n", + index, lio->lis_mirror_index); + RETURN(-EIO); + } + + for (stripe = 0; stripe < r0->lo_nr; stripe++) { + if (!lov_stripe_intersects(lsm, index, stripe, + &ext, &start, &end)) + continue; + + if (unlikely(r0->lo_sub[stripe] == NULL)) { + if (ios->cis_io->ci_type == CIT_READ || + ios->cis_io->ci_type == CIT_WRITE || + ios->cis_io->ci_type == CIT_FAULT) + RETURN(-EIO); + + continue; + } + + end = lov_offset_mod(end, 1); + sub = lov_sub_get(env, lio, + lov_comp_index(index, stripe)); + if (IS_ERR(sub)) { + rc = PTR_ERR(sub); + break; + } + + lov_io_sub_inherit(sub, lio, start, end); + rc = cl_io_iter_init(sub->sub_env, &sub->sub_io); + if (rc != 0) + cl_io_iter_fini(sub->sub_env, &sub->sub_io); + if (rc != 0) + break; + + CDEBUG(D_VFSTRACE, + "shrink stripe: {%d, %d} range: [%llu, %llu)\n", + index, stripe, start, end); + + list_add_tail(&sub->sub_linkage, &lio->lis_active); + } + if (rc != 0) + break; + } + RETURN(rc); } static int lov_io_rw_iter_init(const struct lu_env *env, - const struct cl_io_slice *ios) + const struct cl_io_slice *ios) { - struct lov_io *lio = cl2lov_io(env, ios); - struct cl_io *io = ios->cis_io; - struct lov_stripe_md *lsm = lov_r0(cl2lov(ios->cis_obj))->lo_lsm; - loff_t start = io->u.ci_rw.crw_pos; - loff_t next; - unsigned long ssize = lsm->lsm_stripe_size; - - LASSERT(io->ci_type == CIT_READ || io->ci_type == CIT_WRITE); - ENTRY; + struct cl_io *io = ios->cis_io; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_stripe_md_entry *lse; + struct cl_io_range *range = &io->u.ci_rw.rw_range; + loff_t start = range->cir_pos; + loff_t next; + int index; + + LASSERT(io->ci_type == CIT_READ || io->ci_type == CIT_WRITE); + ENTRY; + + if (cl_io_is_append(io)) + RETURN(lov_io_iter_init(env, ios)); + + index = lov_io_layout_at(lio, range->cir_pos); + if (index < 0) { /* non-existing layout component */ + if (io->ci_type == CIT_READ) { + /* TODO: it needs to detect the next component and + * then set the next pos */ + io->ci_continue = 0; + /* execute it in main thread */ + io->ci_pio = 0; + + RETURN(lov_io_iter_init(env, ios)); + } + + RETURN(-ENODATA); + } + + if (!lov_entry(lio->lis_object, index)->lle_valid && + !io->ci_designated_mirror) + RETURN(io->ci_type == CIT_READ ? -EAGAIN : -EIO); + + lse = lov_lse(lio->lis_object, index); + + next = MAX_LFS_FILESIZE; + if (lse->lsme_stripe_count > 1) { + unsigned long ssize = lse->lsme_stripe_size; + + lov_do_div64(start, ssize); + next = (start + 1) * ssize; + if (next <= start * ssize) + next = MAX_LFS_FILESIZE; + } + + LASSERTF(range->cir_pos >= lse->lsme_extent.e_start, + "pos %lld, [%lld, %lld)\n", range->cir_pos, + lse->lsme_extent.e_start, lse->lsme_extent.e_end); + next = min_t(__u64, next, lse->lsme_extent.e_end); + next = min_t(loff_t, next, lio->lis_io_endpos); + + io->ci_continue = next < lio->lis_io_endpos; + range->cir_count = next - range->cir_pos; + lio->lis_pos = range->cir_pos; + lio->lis_endpos = range->cir_pos + range->cir_count; + CDEBUG(D_VFSTRACE, + "stripe: {%d, %llu} range: [%llu, %llu) end: %llu, count: %zd\n", + index, start, lio->lis_pos, lio->lis_endpos, + lio->lis_io_endpos, range->cir_count); + + if (!io->ci_continue) { + /* the last piece of IO, execute it in main thread */ + io->ci_pio = 0; + } + + if (io->ci_pio) + RETURN(0); + + /* + * XXX The following call should be optimized: we know, that + * [lio->lis_pos, lio->lis_endpos) intersects with exactly one stripe. + */ + RETURN(lov_io_iter_init(env, ios)); +} - /* fast path for common case. */ - if (lio->lis_nr_subios != 1 && !cl_io_is_append(io)) { - - do_div(start, ssize); - next = (start + 1) * ssize; - if (next <= start * ssize) - next = ~0ull; - - io->ci_continue = next < lio->lis_io_endpos; - io->u.ci_rw.crw_count = min_t(loff_t, lio->lis_io_endpos, - next) - io->u.ci_rw.crw_pos; - lio->lis_pos = io->u.ci_rw.crw_pos; - lio->lis_endpos = io->u.ci_rw.crw_pos + io->u.ci_rw.crw_count; - CDEBUG(D_VFSTRACE, "stripe: %llu chunk: [%llu, %llu) %llu\n", - (__u64)start, lio->lis_pos, lio->lis_endpos, - (__u64)lio->lis_io_endpos); - } - /* - * XXX The following call should be optimized: we know, that - * [lio->lis_pos, lio->lis_endpos) intersects with exactly one stripe. - */ - RETURN(lov_io_iter_init(env, ios)); +static int lov_io_setattr_iter_init(const struct lu_env *env, + const struct cl_io_slice *ios) +{ + struct lov_io *lio = cl2lov_io(env, ios); + struct cl_io *io = ios->cis_io; + int index; + ENTRY; + + if (cl_io_is_trunc(io) && lio->lis_pos > 0) { + index = lov_io_layout_at(lio, lio->lis_pos - 1); + /* no entry found for such offset */ + if (index < 0) + RETURN(io->ci_result = -ENODATA); + } + + RETURN(lov_io_iter_init(env, ios)); } static int lov_io_call(const struct lu_env *env, struct lov_io *lio, - int (*iofunc)(const struct lu_env *, struct cl_io *)) + int (*iofunc)(const struct lu_env *, struct cl_io *)) { - struct lov_io_sub *sub; - int rc = 0; - - ENTRY; - cfs_list_for_each_entry(sub, &lio->lis_active, sub_linkage) { - lov_sub_enter(sub); - rc = iofunc(sub->sub_env, sub->sub_io); - lov_sub_exit(sub); - if (rc) - break; - } - RETURN(rc); + struct cl_io *parent = lio->lis_cl.cis_io; + struct lov_io_sub *sub; + int rc = 0; + + ENTRY; + list_for_each_entry(sub, &lio->lis_active, sub_linkage) { + rc = iofunc(sub->sub_env, &sub->sub_io); + if (rc) + break; + + if (parent->ci_result == 0) + parent->ci_result = sub->sub_io.ci_result; + } + RETURN(rc); } static int lov_io_lock(const struct lu_env *env, const struct cl_io_slice *ios) @@ -510,6 +944,31 @@ static void lov_io_end(const struct lu_env *env, const struct cl_io_slice *ios) LASSERT(rc == 0); } +static void +lov_io_data_version_end(const struct lu_env *env, const struct cl_io_slice *ios) +{ + struct lov_io *lio = cl2lov_io(env, ios); + struct cl_io *parent = lio->lis_cl.cis_io; + struct cl_data_version_io *pdv = &parent->u.ci_data_version; + struct lov_io_sub *sub; + + ENTRY; + list_for_each_entry(sub, &lio->lis_active, sub_linkage) { + struct cl_data_version_io *sdv = &sub->sub_io.u.ci_data_version; + + lov_io_end_wrapper(sub->sub_env, &sub->sub_io); + + pdv->dv_data_version += sdv->dv_data_version; + if (pdv->dv_layout_version > sdv->dv_layout_version) + pdv->dv_layout_version = sdv->dv_layout_version; + + if (parent->ci_result == 0) + parent->ci_result = sub->sub_io.ci_result; + } + + EXIT; +} + static void lov_io_iter_fini(const struct lu_env *env, const struct cl_io_slice *ios) { @@ -519,8 +978,8 @@ static void lov_io_iter_fini(const struct lu_env *env, ENTRY; rc = lov_io_call(env, lio, lov_io_iter_fini_wrapper); LASSERT(rc == 0); - while (!cfs_list_empty(&lio->lis_active)) - cfs_list_del_init(lio->lis_active.next); + while (!list_empty(&lio->lis_active)) + list_del_init(lio->lis_active.next); EXIT; } @@ -535,12 +994,85 @@ static void lov_io_unlock(const struct lu_env *env, EXIT; } - -static struct cl_page_list *lov_io_submit_qin(struct lov_device *ld, - struct cl_page_list *qin, - int idx, int alloc) +static int lov_io_read_ahead(const struct lu_env *env, + const struct cl_io_slice *ios, + pgoff_t start, struct cl_read_ahead *ra) { - return alloc ? &qin[idx] : &ld->ld_emrg[idx]->emrg_page_list; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_object *loo = lio->lis_object; + struct cl_object *obj = lov2cl(loo); + struct lov_layout_raid0 *r0; + struct lov_io_sub *sub; + loff_t offset; + loff_t suboff; + pgoff_t ra_end; + unsigned int pps; /* pages per stripe */ + int stripe; + int index; + int rc; + ENTRY; + + offset = cl_offset(obj, start); + index = lov_io_layout_at(lio, offset); + if (index < 0 || !lsm_entry_inited(loo->lo_lsm, index)) + RETURN(-ENODATA); + + /* avoid readahead to expand to stale components */ + if (!lov_entry(loo, index)->lle_valid) + RETURN(-EIO); + + stripe = lov_stripe_number(loo->lo_lsm, index, offset); + + r0 = lov_r0(loo, index); + if (unlikely(r0->lo_sub[stripe] == NULL)) + RETURN(-EIO); + + sub = lov_sub_get(env, lio, lov_comp_index(index, stripe)); + if (IS_ERR(sub)) + RETURN(PTR_ERR(sub)); + + lov_stripe_offset(loo->lo_lsm, index, offset, stripe, &suboff); + rc = cl_io_read_ahead(sub->sub_env, &sub->sub_io, + cl_index(lovsub2cl(r0->lo_sub[stripe]), suboff), + ra); + + CDEBUG(D_READA, DFID " cra_end = %lu, stripes = %d, rc = %d\n", + PFID(lu_object_fid(lov2lu(loo))), ra->cra_end, r0->lo_nr, rc); + if (rc != 0) + RETURN(rc); + + /** + * Adjust the stripe index by layout of comp. ra->cra_end is the + * maximum page index covered by an underlying DLM lock. + * This function converts cra_end from stripe level to file level, and + * make sure it's not beyond stripe and component boundary. + */ + + /* cra_end is stripe level, convert it into file level */ + ra_end = ra->cra_end; + if (ra_end != CL_PAGE_EOF) + ra->cra_end = lov_stripe_pgoff(loo->lo_lsm, index, + ra_end, stripe); + + /* boundary of current component */ + ra_end = cl_index(obj, (loff_t)lov_io_extent(lio, index)->e_end); + if (ra_end != CL_PAGE_EOF && ra->cra_end >= ra_end) + ra->cra_end = ra_end - 1; + + if (r0->lo_nr == 1) /* single stripe file */ + RETURN(0); + + pps = lov_lse(loo, index)->lsme_stripe_size >> PAGE_SHIFT; + + CDEBUG(D_READA, DFID " max_index = %lu, pps = %u, index = %u, " + "stripe_size = %u, stripe no = %u, start index = %lu\n", + PFID(lu_object_fid(lov2lu(loo))), ra->cra_end, pps, index, + lov_lse(loo, index)->lsme_stripe_size, stripe, start); + + /* never exceed the end of the stripe */ + ra->cra_end = min_t(pgoff_t, + ra->cra_end, start + pps - start % pps - 1); + RETURN(0); } /** @@ -559,175 +1091,177 @@ static struct cl_page_list *lov_io_submit_qin(struct lov_device *ld, * lov_device::ld_mutex mutex. */ static int lov_io_submit(const struct lu_env *env, - const struct cl_io_slice *ios, - enum cl_req_type crt, struct cl_2queue *queue, - enum cl_req_priority priority) -{ - struct lov_io *lio = cl2lov_io(env, ios); - struct lov_object *obj = lio->lis_object; - struct lov_device *ld = lu2lov_dev(lov2cl(obj)->co_lu.lo_dev); - struct cl_page_list *qin = &queue->c2_qin; - struct cl_2queue *cl2q = &lov_env_info(env)->lti_cl2q; - struct cl_page_list *stripes_qin = NULL; - struct cl_page *page; - struct cl_page *tmp; - int stripe; - -#define QIN(stripe) lov_io_submit_qin(ld, stripes_qin, stripe, alloc) - - int rc = 0; - int alloc = -#if defined(__KERNEL__) && defined(__linux__) - !(current->flags & PF_MEMALLOC); -#else - 1; -#endif - ENTRY; - if (lio->lis_active_subios == 1) { - int idx = lio->lis_single_subio_index; - struct lov_io_sub *sub; - - LASSERT(idx < lio->lis_nr_subios); - sub = lov_sub_get(env, lio, idx); - LASSERT(!IS_ERR(sub)); - LASSERT(sub->sub_io == &lio->lis_single_subio); - rc = cl_io_submit_rw(sub->sub_env, sub->sub_io, - crt, queue, priority); - lov_sub_put(sub); - RETURN(rc); - } - - LASSERT(lio->lis_subs != NULL); - if (alloc) { - OBD_ALLOC(stripes_qin, - sizeof(*stripes_qin) * lio->lis_nr_subios); - if (stripes_qin == NULL) - RETURN(-ENOMEM); - - for (stripe = 0; stripe < lio->lis_nr_subios; stripe++) - cl_page_list_init(&stripes_qin[stripe]); - } else { - /* - * If we get here, it means pageout & swap doesn't help. - * In order to not make things worse, even don't try to - * allocate the memory with __GFP_NOWARN. -jay - */ - cfs_mutex_lock(&ld->ld_mutex); - lio->lis_mem_frozen = 1; - } - - cl_2queue_init(cl2q); - cl_page_list_for_each_safe(page, tmp, qin) { - stripe = lov_page_stripe(page); - cl_page_list_move(QIN(stripe), qin, page); - } - - for (stripe = 0; stripe < lio->lis_nr_subios; stripe++) { - struct lov_io_sub *sub; - struct cl_page_list *sub_qin = QIN(stripe); - - if (cfs_list_empty(&sub_qin->pl_pages)) - continue; - - cl_page_list_splice(sub_qin, &cl2q->c2_qin); - sub = lov_sub_get(env, lio, stripe); - if (!IS_ERR(sub)) { - rc = cl_io_submit_rw(sub->sub_env, sub->sub_io, - crt, cl2q, priority); - lov_sub_put(sub); - } else - rc = PTR_ERR(sub); - cl_page_list_splice(&cl2q->c2_qin, &queue->c2_qin); - cl_page_list_splice(&cl2q->c2_qout, &queue->c2_qout); - if (rc != 0) - break; - } - - for (stripe = 0; stripe < lio->lis_nr_subios; stripe++) { - struct cl_page_list *sub_qin = QIN(stripe); - - if (cfs_list_empty(&sub_qin->pl_pages)) - continue; + const struct cl_io_slice *ios, + enum cl_req_type crt, struct cl_2queue *queue) +{ + struct cl_page_list *qin = &queue->c2_qin; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_io_sub *sub; + struct cl_page_list *plist = &lov_env_info(env)->lti_plist; + struct cl_page *page; + int index; + int rc = 0; + ENTRY; + + cl_page_list_init(plist); + while (qin->pl_nr > 0) { + struct cl_2queue *cl2q = &lov_env_info(env)->lti_cl2q; + + page = cl_page_list_first(qin); + if (lov_page_is_empty(page)) { + cl_page_list_move(&queue->c2_qout, qin, page); + + /* it could only be mirror read to get here therefore + * the pages will be transient. We don't care about + * the return code of cl_page_prep() at all. */ + (void) cl_page_prep(env, ios->cis_io, page, crt); + cl_page_completion(env, page, crt, 0); + continue; + } + + cl_2queue_init(cl2q); + cl_page_list_move(&cl2q->c2_qin, qin, page); + + index = lov_page_index(page); + while (qin->pl_nr > 0) { + page = cl_page_list_first(qin); + if (index != lov_page_index(page)) + break; + + cl_page_list_move(&cl2q->c2_qin, qin, page); + } + + sub = lov_sub_get(env, lio, index); + if (!IS_ERR(sub)) { + rc = cl_io_submit_rw(sub->sub_env, &sub->sub_io, + crt, cl2q); + } else { + rc = PTR_ERR(sub); + } + + cl_page_list_splice(&cl2q->c2_qin, plist); + cl_page_list_splice(&cl2q->c2_qout, &queue->c2_qout); + cl_2queue_fini(env, cl2q); + + if (rc != 0) + break; + } + + cl_page_list_splice(plist, qin); + cl_page_list_fini(env, plist); + + RETURN(rc); +} - cl_page_list_splice(sub_qin, qin); - } +static int lov_io_commit_async(const struct lu_env *env, + const struct cl_io_slice *ios, + struct cl_page_list *queue, int from, int to, + cl_commit_cbt cb) +{ + struct cl_page_list *plist = &lov_env_info(env)->lti_plist; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_io_sub *sub; + struct cl_page *page; + int rc = 0; + ENTRY; + + if (lio->lis_nr_subios == 1) { + int idx = lio->lis_single_subio_index; + + LASSERT(!lov_page_is_empty(cl_page_list_first(queue))); + + sub = lov_sub_get(env, lio, idx); + LASSERT(!IS_ERR(sub)); + LASSERT(sub == &lio->lis_single_subio); + rc = cl_io_commit_async(sub->sub_env, &sub->sub_io, queue, + from, to, cb); + RETURN(rc); + } + + cl_page_list_init(plist); + while (queue->pl_nr > 0) { + int stripe_to = to; + int index; + + LASSERT(plist->pl_nr == 0); + page = cl_page_list_first(queue); + LASSERT(!lov_page_is_empty(page)); + + cl_page_list_move(plist, queue, page); + + index = lov_page_index(page); + while (queue->pl_nr > 0) { + page = cl_page_list_first(queue); + if (index != lov_page_index(page)) + break; + + cl_page_list_move(plist, queue, page); + } + + if (queue->pl_nr > 0) /* still has more pages */ + stripe_to = PAGE_SIZE; + + sub = lov_sub_get(env, lio, index); + if (!IS_ERR(sub)) { + rc = cl_io_commit_async(sub->sub_env, &sub->sub_io, + plist, from, stripe_to, cb); + } else { + rc = PTR_ERR(sub); + break; + } + + if (plist->pl_nr > 0) /* short write */ + break; + + from = 0; + } + + /* for error case, add the page back into the qin list */ + LASSERT(ergo(rc == 0, plist->pl_nr == 0)); + while (plist->pl_nr > 0) { + /* error occurred, add the uncommitted pages back into queue */ + page = cl_page_list_last(plist); + cl_page_list_move_head(queue, plist, page); + } + + RETURN(rc); +} - if (alloc) { - OBD_FREE(stripes_qin, - sizeof(*stripes_qin) * lio->lis_nr_subios); - } else { - int i; +static int lov_io_fault_start(const struct lu_env *env, + const struct cl_io_slice *ios) +{ + struct cl_fault_io *fio; + struct lov_io *lio; + struct lov_io_sub *sub; - for (i = 0; i < lio->lis_nr_subios; i++) { - struct cl_io *cio = lio->lis_subs[i].sub_io; + ENTRY; - if (cio && cio == &ld->ld_emrg[i]->emrg_subio) - lov_io_sub_fini(env, lio, &lio->lis_subs[i]); - } - lio->lis_mem_frozen = 0; - cfs_mutex_unlock(&ld->ld_mutex); - } + fio = &ios->cis_io->u.ci_fault; + lio = cl2lov_io(env, ios); + sub = lov_sub_get(env, lio, lov_page_index(fio->ft_page)); + sub->sub_io.u.ci_fault.ft_nob = fio->ft_nob; - RETURN(rc); -#undef QIN + RETURN(lov_io_start(env, ios)); } -static int lov_io_prepare_write(const struct lu_env *env, - const struct cl_io_slice *ios, - const struct cl_page_slice *slice, - unsigned from, unsigned to) +static void lov_io_fsync_end(const struct lu_env *env, + const struct cl_io_slice *ios) { - struct lov_io *lio = cl2lov_io(env, ios); - struct cl_page *sub_page = lov_sub_page(slice); - struct lov_io_sub *sub; - int result; + struct lov_io *lio = cl2lov_io(env, ios); + struct lov_io_sub *sub; + unsigned int *written = &ios->cis_io->u.ci_fsync.fi_nr_written; + ENTRY; - ENTRY; - sub = lov_page_subio(env, lio, slice); - if (!IS_ERR(sub)) { - result = cl_io_prepare_write(sub->sub_env, sub->sub_io, - sub_page, from, to); - lov_sub_put(sub); - } else - result = PTR_ERR(sub); - RETURN(result); -} - -static int lov_io_commit_write(const struct lu_env *env, - const struct cl_io_slice *ios, - const struct cl_page_slice *slice, - unsigned from, unsigned to) -{ - struct lov_io *lio = cl2lov_io(env, ios); - struct cl_page *sub_page = lov_sub_page(slice); - struct lov_io_sub *sub; - int result; + *written = 0; + list_for_each_entry(sub, &lio->lis_active, sub_linkage) { + struct cl_io *subio = &sub->sub_io; - ENTRY; - sub = lov_page_subio(env, lio, slice); - if (!IS_ERR(sub)) { - result = cl_io_commit_write(sub->sub_env, sub->sub_io, - sub_page, from, to); - lov_sub_put(sub); - } else - result = PTR_ERR(sub); - RETURN(result); -} + lov_io_end_wrapper(sub->sub_env, subio); -static int lov_io_fault_start(const struct lu_env *env, - const struct cl_io_slice *ios) -{ - struct cl_fault_io *fio; - struct lov_io *lio; - struct lov_io_sub *sub; - - ENTRY; - fio = &ios->cis_io->u.ci_fault; - lio = cl2lov_io(env, ios); - sub = lov_sub_get(env, lio, lov_page_stripe(fio->ft_page)); - sub->sub_io->u.ci_fault.ft_nob = fio->ft_nob; - lov_sub_put(sub); - RETURN(lov_io_start(env, ios)); + if (subio->ci_result == 0) + *written += subio->u.ci_fsync.fi_nr_written; + } + RETURN_EXIT; } static const struct cl_io_operations lov_io_ops = { @@ -750,15 +1284,24 @@ static const struct cl_io_operations lov_io_ops = { .cio_start = lov_io_start, .cio_end = lov_io_end }, - [CIT_TRUNC] = { - .cio_fini = lov_io_fini, - .cio_iter_init = lov_io_iter_init, - .cio_iter_fini = lov_io_iter_fini, - .cio_lock = lov_io_lock, - .cio_unlock = lov_io_unlock, - .cio_start = lov_io_start, - .cio_end = lov_io_end - }, + [CIT_SETATTR] = { + .cio_fini = lov_io_fini, + .cio_iter_init = lov_io_setattr_iter_init, + .cio_iter_fini = lov_io_iter_fini, + .cio_lock = lov_io_lock, + .cio_unlock = lov_io_unlock, + .cio_start = lov_io_start, + .cio_end = lov_io_end + }, + [CIT_DATA_VERSION] = { + .cio_fini = lov_io_fini, + .cio_iter_init = lov_io_iter_init, + .cio_iter_fini = lov_io_iter_fini, + .cio_lock = lov_io_lock, + .cio_unlock = lov_io_unlock, + .cio_start = lov_io_start, + .cio_end = lov_io_data_version_end, + }, [CIT_FAULT] = { .cio_fini = lov_io_fini, .cio_iter_init = lov_io_iter_init, @@ -768,20 +1311,34 @@ static const struct cl_io_operations lov_io_ops = { .cio_start = lov_io_fault_start, .cio_end = lov_io_end }, - [CIT_MISC] = { - .cio_fini = lov_io_fini - } - }, - .req_op = { - [CRT_READ] = { - .cio_submit = lov_io_submit - }, - [CRT_WRITE] = { - .cio_submit = lov_io_submit - } - }, - .cio_prepare_write = lov_io_prepare_write, - .cio_commit_write = lov_io_commit_write + [CIT_FSYNC] = { + .cio_fini = lov_io_fini, + .cio_iter_init = lov_io_iter_init, + .cio_iter_fini = lov_io_iter_fini, + .cio_lock = lov_io_lock, + .cio_unlock = lov_io_unlock, + .cio_start = lov_io_start, + .cio_end = lov_io_fsync_end + }, + [CIT_LADVISE] = { + .cio_fini = lov_io_fini, + .cio_iter_init = lov_io_iter_init, + .cio_iter_fini = lov_io_iter_fini, + .cio_lock = lov_io_lock, + .cio_unlock = lov_io_unlock, + .cio_start = lov_io_start, + .cio_end = lov_io_end + }, + [CIT_GLIMPSE] = { + .cio_fini = lov_io_fini, + }, + [CIT_MISC] = { + .cio_fini = lov_io_fini + } + }, + .cio_read_ahead = lov_io_read_ahead, + .cio_submit = lov_io_submit, + .cio_commit_async = lov_io_commit_async, }; /***************************************************************************** @@ -793,8 +1350,19 @@ static const struct cl_io_operations lov_io_ops = { static void lov_empty_io_fini(const struct lu_env *env, const struct cl_io_slice *ios) { - ENTRY; - EXIT; + struct lov_object *lov = cl2lov(ios->cis_obj); + ENTRY; + + if (atomic_dec_and_test(&lov->lo_active_ios)) + wake_up_all(&lov->lo_waitq); + EXIT; +} + +static int lov_empty_io_submit(const struct lu_env *env, + const struct cl_io_slice *ios, + enum cl_req_type crt, struct cl_2queue *queue) +{ + return -EBADF; } static void lov_empty_impossible(const struct lu_env *env, @@ -811,8 +1379,8 @@ static void lov_empty_impossible(const struct lu_env *env, static const struct cl_io_operations lov_empty_io_ops = { .op = { [CIT_READ] = { -#if 0 .cio_fini = lov_empty_io_fini, +#if 0 .cio_iter_init = LOV_EMPTY_IMPOSSIBLE, .cio_lock = LOV_EMPTY_IMPOSSIBLE, .cio_start = LOV_EMPTY_IMPOSSIBLE, @@ -826,7 +1394,7 @@ static const struct cl_io_operations lov_empty_io_ops = { .cio_start = LOV_EMPTY_IMPOSSIBLE, .cio_end = LOV_EMPTY_IMPOSSIBLE }, - [CIT_TRUNC] = { + [CIT_SETATTR] = { .cio_fini = lov_empty_io_fini, .cio_iter_init = LOV_EMPTY_IMPOSSIBLE, .cio_lock = LOV_EMPTY_IMPOSSIBLE, @@ -840,67 +1408,178 @@ static const struct cl_io_operations lov_empty_io_ops = { .cio_start = LOV_EMPTY_IMPOSSIBLE, .cio_end = LOV_EMPTY_IMPOSSIBLE }, - [CIT_MISC] = { - .cio_fini = lov_empty_io_fini - } - }, - .req_op = { - [CRT_READ] = { - .cio_submit = LOV_EMPTY_IMPOSSIBLE - }, - [CRT_WRITE] = { - .cio_submit = LOV_EMPTY_IMPOSSIBLE - } - }, - .cio_commit_write = LOV_EMPTY_IMPOSSIBLE + [CIT_FSYNC] = { + .cio_fini = lov_empty_io_fini + }, + [CIT_LADVISE] = { + .cio_fini = lov_empty_io_fini + }, + [CIT_GLIMPSE] = { + .cio_fini = lov_empty_io_fini + }, + [CIT_MISC] = { + .cio_fini = lov_empty_io_fini + } + }, + .cio_submit = lov_empty_io_submit, + .cio_commit_async = LOV_EMPTY_IMPOSSIBLE }; -int lov_io_init_raid0(const struct lu_env *env, struct cl_object *obj, - struct cl_io *io) +int lov_io_init_composite(const struct lu_env *env, struct cl_object *obj, + struct cl_io *io) { - struct lov_io *lio = lov_env_io(env); - struct lov_object *lov = cl2lov(obj); - - ENTRY; - CFS_INIT_LIST_HEAD(&lio->lis_active); - lov_io_slice_init(lio, lov, io); - if (io->ci_result == 0) { - LASSERT(lov_r0(lov)->lo_lsm != NULL); - io->ci_result = lov_io_subio_init(env, lio, io); - if (io->ci_result == 0) - cl_io_slice_add(io, &lio->lis_cl, obj, &lov_io_ops); - } - RETURN(io->ci_result); + struct lov_io *lio = lov_env_io(env); + struct lov_object *lov = cl2lov(obj); + int result; + ENTRY; + + INIT_LIST_HEAD(&lio->lis_active); + result = lov_io_slice_init(lio, lov, io); + if (result) + GOTO(out, result); + + result = lov_io_subio_init(env, lio, io); + if (!result) { + cl_io_slice_add(io, &lio->lis_cl, obj, &lov_io_ops); + atomic_inc(&lov->lo_active_ios); + } + EXIT; +out: + io->ci_result = result < 0 ? result : 0; + return result; } int lov_io_init_empty(const struct lu_env *env, struct cl_object *obj, struct cl_io *io) { - struct lov_io *lio = lov_env_io(env); - int result; + struct lov_object *lov = cl2lov(obj); + struct lov_io *lio = lov_env_io(env); + int result; + ENTRY; + + lio->lis_object = lov; + switch (io->ci_type) { + default: + LBUG(); + case CIT_MISC: + case CIT_GLIMPSE: + case CIT_READ: + result = 0; + break; + case CIT_FSYNC: + case CIT_LADVISE: + case CIT_SETATTR: + case CIT_DATA_VERSION: + result = +1; + break; + case CIT_WRITE: + result = -EBADF; + break; + case CIT_FAULT: + result = -EFAULT; + CERROR("Page fault on a file without stripes: "DFID"\n", + PFID(lu_object_fid(&obj->co_lu))); + break; + } + if (result == 0) { + cl_io_slice_add(io, &lio->lis_cl, obj, &lov_empty_io_ops); + atomic_inc(&lov->lo_active_ios); + } + + io->ci_result = result < 0 ? result : 0; + RETURN(result); +} - ENTRY; - switch (io->ci_type) { - default: - LBUG(); - case CIT_MISC: - case CIT_READ: - result = 0; - break; - case CIT_WRITE: - case CIT_TRUNC: - result = -EBADF; - break; - case CIT_FAULT: - result = -EFAULT; - CERROR("Page fault on a file without stripes: "DFID"\n", - PFID(lu_object_fid(&obj->co_lu))); - break; - } - if (result == 0) - cl_io_slice_add(io, &lio->lis_cl, obj, &lov_empty_io_ops); - io->ci_result = result; - RETURN(result != 0); +int lov_io_init_released(const struct lu_env *env, struct cl_object *obj, + struct cl_io *io) +{ + struct lov_object *lov = cl2lov(obj); + struct lov_io *lio = lov_env_io(env); + int result; + ENTRY; + + LASSERT(lov->lo_lsm != NULL); + lio->lis_object = lov; + + switch (io->ci_type) { + default: + LASSERTF(0, "invalid type %d\n", io->ci_type); + result = -EOPNOTSUPP; + break; + case CIT_GLIMPSE: + case CIT_MISC: + case CIT_FSYNC: + case CIT_LADVISE: + case CIT_DATA_VERSION: + result = 1; + break; + case CIT_SETATTR: + /* the truncate to 0 is managed by MDT: + * - in open, for open O_TRUNC + * - in setattr, for truncate + */ + /* the truncate is for size > 0 so triggers a restore */ + if (cl_io_is_trunc(io)) { + io->ci_restore_needed = 1; + result = -ENODATA; + } else + result = 1; + break; + case CIT_READ: + case CIT_WRITE: + case CIT_FAULT: + io->ci_restore_needed = 1; + result = -ENODATA; + break; + } + + if (result == 0) { + cl_io_slice_add(io, &lio->lis_cl, obj, &lov_empty_io_ops); + atomic_inc(&lov->lo_active_ios); + } + + io->ci_result = result < 0 ? result : 0; + RETURN(result); +} + +/** + * Return the index in composite:lo_entries by the file offset + */ +int lov_io_layout_at(struct lov_io *lio, __u64 offset) +{ + struct lov_object *lov = lio->lis_object; + struct lov_layout_composite *comp = &lov->u.composite; + int start_index = 0; + int end_index = comp->lo_entry_count - 1; + int i; + + LASSERT(lov->lo_type == LLT_COMP); + + /* This is actual file offset so nothing can cover eof. */ + if (offset == LUSTRE_EOF) + return -1; + + if (lov_is_flr(lov)) { + struct lov_mirror_entry *lre; + + LASSERT(lio->lis_mirror_index >= 0); + + lre = &comp->lo_mirrors[lio->lis_mirror_index]; + start_index = lre->lre_start; + end_index = lre->lre_end; + } + + for (i = start_index; i <= end_index; i++) { + struct lov_layout_entry *lle = lov_entry(lov, i); + + if ((offset >= lle->lle_extent->e_start && + offset < lle->lle_extent->e_end) || + (offset == OBD_OBJECT_EOF && + lle->lle_extent->e_end == OBD_OBJECT_EOF)) + return i; + } + + return -1; } /** @} lov */