/* -*- 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. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 only, * as published by the Free Software Foundation. * * This program is distributed in the hope that it will be useful, but * WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * General Public License version 2 for more details (a copy is included * in the LICENSE file that accompanied this code). * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see * http://www.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. * * GPL HEADER END */ /* * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. */ /* * This file is part of Lustre, http://www.lustre.org/ * Lustre is a trademark of Sun Microsystems, Inc. * * lustre/mds/mds_lov.c * * Lustre Metadata Server (mds) handling of striped file data * * Author: Peter Braam */ #define DEBUG_SUBSYSTEM S_MDS #include #include #include #include #include #include #include #include #include "mds_internal.h" static void mds_lov_dump_objids(const char *label, struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; unsigned int i=0, j; if ((libcfs_debug & D_INFO) == 0) return; CDEBUG(D_INFO, "dump from %s\n", label); if (mds->mds_lov_page_dirty == NULL) { CERROR("NULL bitmap!\n"); GOTO(skip_bitmap, i); } for(i = 0; i < mds->mds_lov_page_dirty->size / BITS_PER_LONG + 1; i++) CDEBUG(D_INFO, "%u - %lx\n", i, mds->mds_lov_page_dirty->data[i]); skip_bitmap: if (mds->mds_lov_page_array == NULL) { CERROR("not init page array!\n"); GOTO(skip_array, i); } for(i = 0;i < MDS_LOV_OBJID_PAGES_COUNT; i++) { obd_id *data = mds->mds_lov_page_array[i]; if (data == NULL) continue; for(j=0; j < OBJID_PER_PAGE(); j++) { if (data[j] == 0) continue; CDEBUG(D_INFO,"objid page %u idx %u - "LPU64" \n", i, j, data[j]); } } skip_array: EXIT; } int mds_lov_init_objids(struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; int size = MDS_LOV_OBJID_PAGES_COUNT*sizeof(void *); struct file *file; int rc; ENTRY; CLASSERT(((MDS_LOV_ALLOC_SIZE % sizeof(obd_id)) == 0)); mds->mds_lov_page_dirty = CFS_ALLOCATE_BITMAP(MDS_LOV_OBJID_PAGES_COUNT); if (mds->mds_lov_page_dirty == NULL) RETURN(-ENOMEM); OBD_ALLOC(mds->mds_lov_page_array, size); if (mds->mds_lov_page_array == NULL) GOTO(err_free_bitmap, rc = -ENOMEM); /* open and test the lov objd file */ file = filp_open(LOV_OBJID, O_RDWR | O_CREAT, 0644); if (IS_ERR(file)) { rc = PTR_ERR(file); CERROR("cannot open/create %s file: rc = %d\n", LOV_OBJID, rc); GOTO(err_free, rc = PTR_ERR(file)); } if (!S_ISREG(file->f_dentry->d_inode->i_mode)) { CERROR("%s is not a regular file!: mode = %o\n", LOV_OBJID, file->f_dentry->d_inode->i_mode); GOTO(err_open, rc = -ENOENT); } mds->mds_lov_objid_filp = file; RETURN (0); err_open: if (filp_close((struct file *)file, 0)) CERROR("can't close %s after error\n", LOV_OBJID); err_free: OBD_FREE(mds->mds_lov_page_array, size); err_free_bitmap: CFS_FREE_BITMAP(mds->mds_lov_page_dirty); RETURN(rc); } void mds_lov_destroy_objids(struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; int i, rc; ENTRY; if (mds->mds_lov_page_array != NULL) { for(i=0;imds_lov_page_array[i]; if (data != NULL) OBD_FREE(data, MDS_LOV_ALLOC_SIZE); } OBD_FREE(mds->mds_lov_page_array, MDS_LOV_OBJID_PAGES_COUNT*sizeof(void *)); } if (mds->mds_lov_objid_filp) { rc = filp_close((struct file *)mds->mds_lov_objid_filp, NULL); mds->mds_lov_objid_filp = NULL; if (rc) CERROR("%s file won't close, rc=%d\n", LOV_OBJID, rc); } CFS_FREE_BITMAP(mds->mds_lov_page_dirty); EXIT; } /** * currently exist two ways for know about ost count and max ost index. * first - after ost is connected to mds and sync process finished * second - get from lmm in recovery process, in case when mds not have configs, * and ost isn't registered in mgs. * * \param mds pointer to mds structure * \param index maxium ost index * * \retval -ENOMEM is not hame memory for new page * \retval 0 is update passed */ static int mds_lov_update_max_ost(struct mds_obd *mds, obd_id index) { __u32 page = index / OBJID_PER_PAGE(); __u32 off = index % OBJID_PER_PAGE(); obd_id *data = mds->mds_lov_page_array[page]; if (data == NULL) { OBD_ALLOC(data, MDS_LOV_ALLOC_SIZE); if (data == NULL) RETURN(-ENOMEM); mds->mds_lov_page_array[page] = data; } if (index > mds->mds_lov_objid_max_index) { mds->mds_lov_objid_lastpage = page; mds->mds_lov_objid_lastidx = off; mds->mds_lov_objid_max_index = index; } /* workaround - New target not in objids file; increase mdsize */ /* ld_tgt_count is used as the max index everywhere, despite its name. */ if (data[off] == 0) { __u32 stripes; data[off] = 1; mds->mds_lov_objid_count++; stripes = min_t(__u32, LOV_MAX_STRIPE_COUNT, mds->mds_lov_objid_count); mds->mds_max_mdsize = lov_mds_md_size(stripes, LOV_MAGIC_V3); mds->mds_max_cookiesize = stripes * sizeof(struct llog_cookie); CDEBUG(D_CONFIG, "updated max_mdsize/max_cookiesize for %d" " stripes: %d/%d\n", stripes, mds->mds_max_mdsize, mds->mds_max_cookiesize); } EXIT; return 0; } static int mds_lov_objinit(struct mds_obd *mds, __u32 index) { __u32 page = index / OBJID_PER_PAGE(); __u32 off = index % OBJID_PER_PAGE(); obd_id *data = mds->mds_lov_page_array[page]; return (data[off] > 0); } int mds_lov_prepare_objids(struct obd_device *obd, struct lov_mds_md *lmm) { struct lov_ost_data_v1 *data; __u32 count; int rc = 0; __u32 j; /* if we create file without objects - lmm is NULL */ if (lmm == NULL) return 0; switch (le32_to_cpu(lmm->lmm_magic)) { case LOV_MAGIC_V1: count = le32_to_cpu(((struct lov_mds_md_v1*)lmm)->lmm_stripe_count); data = &(((struct lov_mds_md_v1*)lmm)->lmm_objects[0]); break; case LOV_MAGIC_V3: count = le32_to_cpu(((struct lov_mds_md_v3*)lmm)->lmm_stripe_count); data = &(((struct lov_mds_md_v3*)lmm)->lmm_objects[0]); break; default: CERROR("Unknow lmm type %X!\n", le32_to_cpu(lmm->lmm_magic)); RETURN(-EINVAL); } cfs_mutex_down(&obd->obd_dev_sem); for (j = 0; j < count; j++) { __u32 i = le32_to_cpu(data[j].l_ost_idx); if (mds_lov_update_max_ost(&obd->u.mds, i)) { rc = -ENOMEM; break; } } cfs_mutex_up(&obd->obd_dev_sem); RETURN(rc); } EXPORT_SYMBOL(mds_lov_prepare_objids); /* * write llog orphan record about lost ost object, * Special lsm is allocated with single stripe, caller should deallocated it * after use */ static int mds_log_lost_precreated(struct obd_device *obd, struct lov_stripe_md **lsmp, int *stripes, obd_id id, obd_count count, int idx) { struct lov_stripe_md *lsm = *lsmp; int rc; ENTRY; if (*lsmp == NULL) { rc = obd_alloc_memmd(obd->u.mds.mds_lov_exp, &lsm); if (rc < 0) RETURN(rc); /* need only one stripe, save old value */ *stripes = lsm->lsm_stripe_count; lsm->lsm_stripe_count = 1; *lsmp = lsm; } lsm->lsm_oinfo[0]->loi_id = id; lsm->lsm_oinfo[0]->loi_seq = mdt_to_obd_objseq(obd->u.mds.mds_id); lsm->lsm_oinfo[0]->loi_ost_idx = idx; rc = mds_log_op_orphan(obd, lsm, count); RETURN(rc); } void mds_lov_update_objids(struct obd_device *obd, struct lov_mds_md *lmm) { struct mds_obd *mds = &obd->u.mds; int j; struct lov_ost_data_v1 *obj; struct lov_stripe_md *lsm = NULL; int stripes = 0; int count; ENTRY; /* if we create file without objects - lmm is NULL */ if (lmm == NULL) return; switch (le32_to_cpu(lmm->lmm_magic)) { case LOV_MAGIC_V1: count = le32_to_cpu(((struct lov_mds_md_v1*)lmm)->lmm_stripe_count); obj = ((struct lov_mds_md_v1*)lmm)->lmm_objects; break; case LOV_MAGIC_V3: count = le32_to_cpu(((struct lov_mds_md_v3*)lmm)->lmm_stripe_count); obj = ((struct lov_mds_md_v3*)lmm)->lmm_objects; break; default: CERROR("Unknow lmm type %X !\n", le32_to_cpu(lmm->lmm_magic)); return; } for (j = 0; j < count; j++) { __u32 i = le32_to_cpu(obj[j].l_ost_idx); obd_id id = le64_to_cpu(obj[j].l_object_id); __u32 page = i / OBJID_PER_PAGE(); __u32 idx = i % OBJID_PER_PAGE(); obd_id *data; data = mds->mds_lov_page_array[page]; CDEBUG(D_INODE,"update last object for ost %u" " - new "LPU64" old "LPU64"\n", i, id, data[idx]); if (id > data[idx]) { int lost = id - data[idx] - 1; /* we might have lost precreated objects due to VBR */ if (lost > 0 && obd->obd_recovering) { CDEBUG(D_HA, "Gap in objids is %u\n", lost); if (!obd->obd_version_recov) CERROR("Unexpected gap in objids\n"); /* lsm is allocated if NULL */ mds_log_lost_precreated(obd, &lsm, &stripes, data[idx]+1, lost, i); } data[idx] = id; cfs_bitmap_set(mds->mds_lov_page_dirty, page); } } if (lsm) { /* restore stripes number */ lsm->lsm_stripe_count = stripes; obd_free_memmd(mds->mds_lov_exp, &lsm); } EXIT; return; } EXPORT_SYMBOL(mds_lov_update_objids); static int mds_lov_update_from_read(struct mds_obd *mds, obd_id *data, __u32 count) { __u32 i; __u32 stripes; for (i = 0; i < count; i++) { if (data[i] == 0) continue; mds->mds_lov_objid_count++; } stripes = min_t(__u32, LOV_MAX_STRIPE_COUNT, mds->mds_lov_objid_count); mds->mds_max_mdsize = lov_mds_md_size(stripes, LOV_MAGIC_V3); mds->mds_max_cookiesize = stripes * sizeof(struct llog_cookie); CDEBUG(D_CONFIG, "updated max_mdsize/max_cookiesize for %d stripes: " "%d/%d\n", stripes, mds->mds_max_mdsize, mds->mds_max_cookiesize); EXIT; return 0; } static int mds_lov_read_objids(struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; loff_t off = 0; int i, rc = 0, count = 0, page = 0; unsigned long size; ENTRY; /* Read everything in the file, even if our current lov desc has fewer targets. Old targets not in the lov descriptor during mds setup may still have valid objids. */ size = i_size_read(mds->mds_lov_objid_filp->f_dentry->d_inode); if (size == 0) RETURN(0); page = (size + MDS_LOV_ALLOC_SIZE - 1) / MDS_LOV_ALLOC_SIZE; CDEBUG(D_INFO, "file size %lu pages %d\n", size, page); for (i = 0; i < page; i++) { obd_id *data; loff_t off_old = off; LASSERT(mds->mds_lov_page_array[i] == NULL); OBD_ALLOC(mds->mds_lov_page_array[i], MDS_LOV_ALLOC_SIZE); if (mds->mds_lov_page_array[i] == NULL) GOTO(out, rc = -ENOMEM); data = mds->mds_lov_page_array[i]; rc = fsfilt_read_record(obd, mds->mds_lov_objid_filp, data, MDS_LOV_ALLOC_SIZE, &off); if (rc < 0) { CERROR("Error reading objids %d\n", rc); GOTO(out, rc); } if (off == off_old) /* hole is read */ off += MDS_LOV_ALLOC_SIZE; count = (off - off_old) / sizeof(obd_id); if (mds_lov_update_from_read(mds, data, count)) { CERROR("Can't update mds data\n"); GOTO(out, rc = -EIO); } } mds->mds_lov_objid_lastpage = page - 1; mds->mds_lov_objid_lastidx = count - 1; CDEBUG(D_INFO, "Read %u - %u %u objid\n", mds->mds_lov_objid_count, mds->mds_lov_objid_lastpage, mds->mds_lov_objid_lastidx); out: mds_lov_dump_objids("read",obd); RETURN(rc); } int mds_lov_write_objids(struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; int i = 0, rc = 0; ENTRY; if (cfs_bitmap_check_empty(mds->mds_lov_page_dirty)) RETURN(0); mds_lov_dump_objids("write", obd); cfs_foreach_bit(mds->mds_lov_page_dirty, i) { obd_id *data = mds->mds_lov_page_array[i]; unsigned int size = MDS_LOV_ALLOC_SIZE; loff_t off = i * size; LASSERT(data != NULL); if (!cfs_bitmap_test_and_clear(mds->mds_lov_page_dirty, i)) continue; /* check for particaly filled last page */ if (i == mds->mds_lov_objid_lastpage) size = (mds->mds_lov_objid_lastidx+1) * sizeof(obd_id); CDEBUG(D_INFO, "write %lld - %u\n", off, size); rc = fsfilt_write_record(obd, mds->mds_lov_objid_filp, data, size, &off, 0); if (rc < 0) { cfs_bitmap_set(mds->mds_lov_page_dirty, i); break; } } if (rc >= 0) rc = 0; RETURN(rc); } EXPORT_SYMBOL(mds_lov_write_objids); static int mds_lov_get_objid(struct obd_device * obd, obd_id idx) { struct mds_obd *mds = &obd->u.mds; struct obd_export *lov_exp = mds->mds_lov_exp; unsigned int page; unsigned int off; obd_id *data; __u32 size; int rc = 0; ENTRY; page = idx / OBJID_PER_PAGE(); off = idx % OBJID_PER_PAGE(); data = mds->mds_lov_page_array[page]; if (data[off] < 2) { /* We never read this lastid; ask the osc */ struct obd_id_info lastid; size = sizeof(lastid); lastid.idx = idx; lastid.data = &data[off]; rc = obd_get_info(lov_exp, sizeof(KEY_LAST_ID), KEY_LAST_ID, &size, &lastid, NULL); if (rc) GOTO(out, rc); /* workaround for clean filter */ if (data[off] == 0) data[off] = 1; cfs_bitmap_set(mds->mds_lov_page_dirty, page); } CDEBUG(D_INFO, "idx "LPU64" - %p - %d/%d - "LPU64"\n", idx, data, page, off, data[off]); out: RETURN(rc); } int mds_lov_clear_orphans(struct mds_obd *mds, struct obd_uuid *ost_uuid) { int rc; struct obdo oa = { 0 }; struct obd_trans_info oti = {0}; struct lov_stripe_md *empty_ea = NULL; ENTRY; LASSERT(mds->mds_lov_page_array != NULL); /* This create will in fact either create or destroy: If the OST is * missing objects below this ID, they will be created. If it finds * objects above this ID, they will be removed. */ memset(&oa, 0, sizeof(oa)); oa.o_flags = OBD_FL_DELORPHAN; oa.o_seq = mdt_to_obd_objseq(mds->mds_id); oa.o_valid = OBD_MD_FLFLAGS | OBD_MD_FLGROUP; if (ost_uuid != NULL) oti.oti_ost_uuid = ost_uuid; rc = obd_create(mds->mds_lov_exp, &oa, &empty_ea, &oti); RETURN(rc); } /* for one target */ static int mds_lov_set_one_nextid(struct obd_device *obd, __u32 idx, obd_id *id) { struct mds_obd *mds = &obd->u.mds; int rc; struct obd_id_info info; ENTRY; LASSERT(!obd->obd_recovering); info.idx = idx; info.data = id; rc = obd_set_info_async(mds->mds_lov_exp, sizeof(KEY_NEXT_ID), KEY_NEXT_ID, sizeof(info), &info, NULL); if (rc) CERROR ("%s: mds_lov_set_nextid failed (%d)\n", obd->obd_name, rc); RETURN(rc); } /* Update the lov desc for a new size lov. */ static int mds_lov_update_desc(struct obd_device *obd, int idx, struct obd_uuid *uuid) { struct mds_obd *mds = &obd->u.mds; struct lov_desc *ld; __u32 valsize = sizeof(mds->mds_lov_desc); int rc = 0; ENTRY; OBD_ALLOC(ld, sizeof(*ld)); if (!ld) RETURN(-ENOMEM); rc = obd_get_info(mds->mds_lov_exp, sizeof(KEY_LOVDESC), KEY_LOVDESC, &valsize, ld, NULL); if (rc) GOTO(out, rc); /* Don't change the mds_lov_desc until the objids size matches the count (paranoia) */ mds->mds_lov_desc = *ld; CDEBUG(D_CONFIG, "updated lov_desc, tgt_count: %d - idx %d / uuid %s\n", mds->mds_lov_desc.ld_tgt_count, idx, uuid->uuid); cfs_mutex_down(&obd->obd_dev_sem); rc = mds_lov_update_max_ost(mds, idx); cfs_mutex_up(&obd->obd_dev_sem); if (rc != 0) GOTO(out, rc ); /* If we added a target we have to reconnect the llogs */ /* We only _need_ to do this at first add (idx), or the first time after recovery. However, it should now be safe to call anytime. */ rc = obd_llog_init(obd, &obd->obd_olg, obd, &idx); if (rc) GOTO(out, rc); out: OBD_FREE(ld, sizeof(*ld)); RETURN(rc); } /* Inform MDS about new/updated target */ static int mds_lov_update_mds(struct obd_device *obd, struct obd_device *watched, __u32 idx) { struct mds_obd *mds = &obd->u.mds; int rc = 0; int page; int off; obd_id *data; ENTRY; LASSERT(mds_lov_objinit(mds, idx)); CDEBUG(D_CONFIG, "idx=%d, recov=%d/%d, cnt=%d\n", idx, obd->obd_recovering, obd->obd_async_recov, mds->mds_lov_desc.ld_tgt_count); /* idx is set as data from lov_notify. */ if (obd->obd_recovering) GOTO(out, rc); if (idx >= mds->mds_lov_desc.ld_tgt_count) { CERROR("index %d > count %d!\n", idx, mds->mds_lov_desc.ld_tgt_count); GOTO(out, rc = -EINVAL); } rc = mds_lov_get_objid(obd, idx); if (rc) GOTO(out, rc); page = idx / OBJID_PER_PAGE(); off = idx % OBJID_PER_PAGE(); data = mds->mds_lov_page_array[page]; /* We have read this lastid from disk; tell the osc. Don't call this during recovery. */ rc = mds_lov_set_one_nextid(obd, idx, &data[off]); if (rc) { CERROR("Failed to set next id, idx=%d rc=%d\n", idx,rc); /* Don't abort the rest of the sync */ rc = 0; } else { CDEBUG(D_CONFIG, "last object "LPU64" from OST %d rc=%d\n", data[off], idx, rc); } out: RETURN(rc); } /* update the LOV-OSC knowledge of the last used object id's */ int mds_lov_connect(struct obd_device *obd, char * lov_name) { struct mds_obd *mds = &obd->u.mds; struct obd_connect_data *data; int rc; ENTRY; if (IS_ERR(mds->mds_lov_obd)) RETURN(PTR_ERR(mds->mds_lov_obd)); if (mds->mds_lov_obd) RETURN(0); mds->mds_lov_obd = class_name2obd(lov_name); if (!mds->mds_lov_obd) { CERROR("MDS cannot locate LOV %s\n", lov_name); mds->mds_lov_obd = ERR_PTR(-ENOTCONN); RETURN(-ENOTCONN); } cfs_mutex_down(&obd->obd_dev_sem); rc = mds_lov_read_objids(obd); cfs_mutex_up(&obd->obd_dev_sem); if (rc) { CERROR("cannot read %s: rc = %d\n", "lov_objids", rc); GOTO(err_exit, rc); } rc = obd_register_observer(mds->mds_lov_obd, obd); if (rc) { CERROR("MDS cannot register as observer of LOV %s (%d)\n", lov_name, rc); GOTO(err_exit, rc); } /* ask lov to generate OBD_NOTIFY_CREATE events for already registered * targets */ obd_notify(mds->mds_lov_obd, NULL, OBD_NOTIFY_CREATE, NULL); mds->mds_lov_obd->u.lov.lov_sp_me = LUSTRE_SP_MDT; OBD_ALLOC(data, sizeof(*data)); if (data == NULL) GOTO(err_exit, rc = -ENOMEM); data->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_INDEX | OBD_CONNECT_REQPORTAL | OBD_CONNECT_QUOTA64 | OBD_CONNECT_OSS_CAPA | OBD_CONNECT_FID | OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CKSUM | OBD_CONNECT_CHANGE_QS | OBD_CONNECT_AT | OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | OBD_CONNECT_SOM | OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH; #ifdef HAVE_LRU_RESIZE_SUPPORT data->ocd_connect_flags |= OBD_CONNECT_LRU_RESIZE; #endif data->ocd_version = LUSTRE_VERSION_CODE; data->ocd_group = mdt_to_obd_objseq(mds->mds_id); /* send max bytes per rpc */ data->ocd_brw_size = PTLRPC_MAX_BRW_PAGES << CFS_PAGE_SHIFT; /* send the list of supported checksum types */ data->ocd_cksum_types = cksum_types_supported(); /* NB: lov_connect() needs to fill in .ocd_index for each OST */ rc = obd_connect(NULL, &mds->mds_lov_exp, mds->mds_lov_obd, &obd->obd_uuid, data, NULL); OBD_FREE(data, sizeof(*data)); if (rc) { CERROR("MDS cannot connect to LOV %s (%d)\n", lov_name, rc); mds->mds_lov_obd = ERR_PTR(rc); RETURN(rc); } /* I want to see a callback happen when the OBD moves to a * "For General Use" state, and that's when we'll call * set_nextid(). The class driver can help us here, because * it can use the obd_recovering flag to determine when the * the OBD is full available. */ /* MDD device will care about that if (!obd->obd_recovering) rc = mds_postrecov(obd); */ RETURN(rc); err_exit: mds->mds_lov_exp = NULL; mds->mds_lov_obd = ERR_PTR(rc); RETURN(rc); } int mds_lov_disconnect(struct obd_device *obd) { struct mds_obd *mds = &obd->u.mds; int rc = 0; ENTRY; if (!IS_ERR(mds->mds_lov_obd) && mds->mds_lov_exp != NULL) { obd_register_observer(mds->mds_lov_obd, NULL); /* The actual disconnect of the mds_lov will be called from * class_disconnect_exports from mds_lov_clean. So we have to * ensure that class_cleanup doesn't fail due to the extra ref * we're holding now. The mechanism to do that already exists - * the obd_force flag. We'll drop the final ref to the * mds_lov_exp in mds_cleanup. */ mds->mds_lov_obd->obd_force = 1; } RETURN(rc); } struct mds_lov_sync_info { struct obd_device *mlsi_obd; /* the lov device to sync */ struct obd_device *mlsi_watched; /* target osc */ __u32 mlsi_index; /* index of target */ }; static int mds_propagate_capa_keys(struct mds_obd *mds, struct obd_uuid *uuid) { struct mds_capa_info info = { .uuid = uuid }; struct lustre_capa_key *key; int i, rc = 0; ENTRY; if (!mds->mds_capa_keys) RETURN(0); OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_SYNC_CAPA_SL, 5); for (i = 0; i < 2; i++) { key = &mds->mds_capa_keys[i]; DEBUG_CAPA_KEY(D_SEC, key, "propagate"); info.capa = key; rc = obd_set_info_async(mds->mds_lov_exp, sizeof(KEY_CAPA_KEY), KEY_CAPA_KEY, sizeof(info), &info, NULL); if (rc) { DEBUG_CAPA_KEY(D_ERROR, key, "propagate failed (rc = %d) for", rc); RETURN(rc); } } RETURN(0); } /* We only sync one osc at a time, so that we don't have to hold any kind of lock on the whole mds_lov_desc, which may change (grow) as a result of mds_lov_add_ost. This also avoids any kind of mismatch between the lov_desc and the mds_lov_desc, which are not in lock-step during lov_add_obd */ static int __mds_lov_synchronize(void *data) { struct mds_lov_sync_info *mlsi = data; struct obd_device *obd = mlsi->mlsi_obd; struct obd_device *watched = mlsi->mlsi_watched; struct mds_obd *mds = &obd->u.mds; struct obd_uuid *uuid; __u32 idx = mlsi->mlsi_index; struct mds_group_info mgi; struct llog_ctxt *ctxt; int rc = 0; ENTRY; OBD_FREE_PTR(mlsi); LASSERT(obd); LASSERT(watched); uuid = &watched->u.cli.cl_target_uuid; LASSERT(uuid); cfs_down_read(&mds->mds_notify_lock); if (obd->obd_stopping || obd->obd_fail) GOTO(out, rc = -ENODEV); OBD_RACE(OBD_FAIL_MDS_LOV_SYNC_RACE); rc = mds_lov_update_mds(obd, watched, idx); if (rc != 0) { CERROR("%s failed at update_mds: %d\n", obd_uuid2str(uuid), rc); GOTO(out, rc); } mgi.group = mdt_to_obd_objseq(mds->mds_id); mgi.uuid = uuid; rc = obd_set_info_async(mds->mds_lov_exp, sizeof(KEY_MDS_CONN), KEY_MDS_CONN, sizeof(mgi), &mgi, NULL); if (rc != 0) GOTO(out, rc); /* propagate capability keys */ rc = mds_propagate_capa_keys(mds, uuid); if (rc) GOTO(out, rc); ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT); if (!ctxt) GOTO(out, rc = -ENODEV); OBD_FAIL_TIMEOUT(OBD_FAIL_MDS_LLOG_SYNC_TIMEOUT, 60); rc = llog_connect(ctxt, NULL, NULL, uuid); llog_ctxt_put(ctxt); if (rc != 0) { CERROR("%s failed at llog_origin_connect: %d\n", obd_uuid2str(uuid), rc); GOTO(out, rc); } LCONSOLE_INFO("MDS %s: %s now active, resetting orphans\n", obd->obd_name, obd_uuid2str(uuid)); rc = mds_lov_clear_orphans(mds, uuid); if (rc != 0) { CERROR("%s failed at mds_lov_clear_orphans: %d\n", obd_uuid2str(uuid), rc); GOTO(out, rc); } #ifdef HAVE_QUOTA_SUPPORT if (obd->obd_upcall.onu_owner) { /* * This is a hack for mds_notify->mdd_notify. When the mds obd * in mdd is removed, This hack should be removed. */ LASSERT(obd->obd_upcall.onu_upcall != NULL); rc = obd->obd_upcall.onu_upcall(obd, NULL, OBD_NOTIFY_QUOTA, obd->obd_upcall.onu_owner,NULL); } #endif EXIT; out: cfs_up_read(&mds->mds_notify_lock); if (rc) { /* Deactivate it for safety */ CERROR("%s sync failed %d, deactivating\n", obd_uuid2str(uuid), rc); if (!obd->obd_stopping && mds->mds_lov_obd && !mds->mds_lov_obd->obd_stopping && !watched->obd_stopping) obd_notify(mds->mds_lov_obd, watched, OBD_NOTIFY_INACTIVE, NULL); } class_decref(obd, "mds_lov_synchronize", obd); return rc; } int mds_lov_synchronize(void *data) { struct mds_lov_sync_info *mlsi = data; char name[20]; snprintf(name, sizeof(name), "ll_sync_%02u", mlsi->mlsi_index); cfs_daemonize_ctxt(name); RETURN(__mds_lov_synchronize(data)); } int mds_lov_start_synchronize(struct obd_device *obd, struct obd_device *watched, void *data, enum obd_notify_event ev) { struct mds_lov_sync_info *mlsi; int rc; struct obd_uuid *uuid; ENTRY; LASSERT(watched); uuid = &watched->u.cli.cl_target_uuid; OBD_ALLOC(mlsi, sizeof(*mlsi)); if (mlsi == NULL) RETURN(-ENOMEM); LASSERT(data); mlsi->mlsi_obd = obd; mlsi->mlsi_watched = watched; mlsi->mlsi_index = *(__u32 *)data; /* Although class_export_get(obd->obd_self_export) would lock the MDS in place, since it's only a self-export it doesn't lock the LOV in place. The LOV can be disconnected during MDS precleanup, leaving nothing for __mds_lov_synchronize. Simply taking an export ref on the LOV doesn't help, because it's still disconnected. Taking an obd reference insures that we don't disconnect the LOV. This of course means a cleanup won't finish for as long as the sync is blocking. */ class_incref(obd, "mds_lov_synchronize", obd); if (ev != OBD_NOTIFY_SYNC) { /* Synchronize in the background */ rc = cfs_create_thread(mds_lov_synchronize, mlsi, CFS_DAEMON_FLAGS); if (rc < 0) { CERROR("%s: error starting mds_lov_synchronize: %d\n", obd->obd_name, rc); class_decref(obd, "mds_lov_synchronize", obd); } else { CDEBUG(D_HA, "%s: mds_lov_synchronize idx=%d " "thread=%d\n", obd->obd_name, mlsi->mlsi_index, rc); rc = 0; } } else { rc = __mds_lov_synchronize((void *)mlsi); } RETURN(rc); } int mds_notify(struct obd_device *obd, struct obd_device *watched, enum obd_notify_event ev, void *data) { struct mds_obd *mds = &obd->u.mds; int rc = 0; ENTRY; CDEBUG(D_CONFIG, "notify %s ev=%d\n", watched->obd_name, ev); if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME) != 0) { CERROR("unexpected notification of %s %s!\n", watched->obd_type->typ_name, watched->obd_name); RETURN(-EINVAL); } /*XXX this notifies the MDD until lov handling use old mds code * must non block! */ if (obd->obd_upcall.onu_owner) { LASSERT(obd->obd_upcall.onu_upcall != NULL); rc = obd->obd_upcall.onu_upcall(obd, NULL, ev, obd->obd_upcall.onu_owner, &mds->mds_obt.obt_mount_count); } switch (ev) { /* We only handle these: */ case OBD_NOTIFY_CREATE: CWARN("MDS %s: add target %s\n",obd->obd_name, obd_uuid2str(&watched->u.cli.cl_target_uuid)); /* We still have to fix the lov descriptor for ost's */ LASSERT(data); rc = mds_lov_update_desc(obd, *(__u32 *)data, &watched->u.cli.cl_target_uuid); RETURN(rc); case OBD_NOTIFY_ACTIVE: /* lov want one or more _active_ targets for work */ /* activate event should be pass lov idx as argument */ case OBD_NOTIFY_SYNC: case OBD_NOTIFY_SYNC_NONBLOCK: /* sync event should be pass lov idx as argument */ break; default: RETURN(0); } if (obd->obd_recovering) { CWARN("MDS %s: in recovery, not resetting orphans on %s\n", obd->obd_name, obd_uuid2str(&watched->u.cli.cl_target_uuid)); /* We still have to fix the lov descriptor for ost's added after the mdt in the config log. They didn't make it into mds_lov_connect. */ rc = mds_lov_update_desc(obd, *(__u32 *)data, &watched->u.cli.cl_target_uuid); } else { rc = mds_lov_start_synchronize(obd, watched, data, ev); } RETURN(rc); }