X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Finclude%2Flustre_log.h;h=54f3f047419a1badaa169a75a3c20f4920901192;hb=39352577e47c7928d7e477dd5fcee04ddc5cbeaa;hp=31bd54306878ad92d591e8faf8c6a15dca87dc8b;hpb=c9842fdc5244e38593f0b12468e87f23853dba9f;p=fs%2Flustre-release.git diff --git a/lustre/include/lustre_log.h b/lustre/include/lustre_log.h index 31bd543..54f3f04 100644 --- a/lustre/include/lustre_log.h +++ b/lustre/include/lustre_log.h @@ -1,25 +1,41 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * Copyright (C) 2001 Cluster File Systems, Inc. + * GPL HEADER START * - * This file is part of Lustre, http://www.lustre.org. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * Lustre is free software; you can redistribute it and/or - * modify it under the terms of version 2 of the GNU General Public - * License as published by the Free Software Foundation. + * 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. * - * Lustre 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 for more details. + * 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 - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 * - * Generic infrastructure for managing a collection of logs. + * 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) 2007, 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/include/lustre_log.h + * + * Generic infrastructure for managing a collection of logs. * These logs are used for: * * - orphan recovery: OST adds record on create @@ -35,6 +51,11 @@ #ifndef _LUSTRE_LOG_H #define _LUSTRE_LOG_H +/** \defgroup log log + * + * @{ + */ + #if defined(__linux__) #include #elif defined(__APPLE__) @@ -45,7 +66,7 @@ #error Unsupported operating system. #endif -#include +#include #include #include @@ -54,20 +75,20 @@ #define LLOG_EEMPTY 4711 struct plain_handle_data { - struct list_head phd_entry; + cfs_list_t phd_entry; struct llog_handle *phd_cat_handle; struct llog_cookie phd_cookie; /* cookie of this log in its cat */ int phd_last_idx; }; struct cat_handle_data { - struct list_head chd_head; + cfs_list_t chd_head; struct llog_handle *chd_current_log; /* currently open log */ }; /* In-memory descriptor for a log object or log catalog */ struct llog_handle { - struct rw_semaphore lgh_lock; + cfs_rw_semaphore_t lgh_lock; struct llog_logid lgh_id; /* id of this log */ struct llog_log_hdr *lgh_hdr; struct file *lgh_file; @@ -90,22 +111,61 @@ int llog_init_handle(struct llog_handle *handle, int flags, extern void llog_free_handle(struct llog_handle *handle); int llog_process(struct llog_handle *loghandle, llog_cb_t cb, void *data, void *catdata); +int llog_process_flags(struct llog_handle *loghandle, llog_cb_t cb, + void *data, void *catdata, int flags); int llog_reverse_process(struct llog_handle *loghandle, llog_cb_t cb, void *data, void *catdata); extern int llog_cancel_rec(struct llog_handle *loghandle, int index); extern int llog_close(struct llog_handle *cathandle); extern int llog_get_size(struct llog_handle *loghandle); -/* llog_cat.c - catalog api */ +/* llog_process flags */ +#define LLOG_FLAG_NODEAMON 0x0001 + +/* llog_cat.c - catalog api */ struct llog_process_data { - void *lpd_data; - llog_cb_t lpd_cb; + /** + * Any useful data needed while processing catalog. This is + * passed later to process callback. + */ + void *lpd_data; + /** + * Catalog process callback function, called for each record + * in catalog. + */ + llog_cb_t lpd_cb; + /** + * Start processing the catalog from startcat/startidx + */ + int lpd_startcat; + int lpd_startidx; + int lpd_flags; /** llog_process flags */ }; struct llog_process_cat_data { - int first_idx; - int last_idx; - /* to process catalog across zero record */ + /** + * Temporary stored first_idx while scanning log. + */ + int lpcd_first_idx; + /** + * Temporary stored last_idx while scanning log. + */ + int lpcd_last_idx; +}; + +struct llog_process_cat_args { + /** + * Llog context used in recovery thread on OST (recov_thread.c) + */ + struct llog_ctxt *lpca_ctxt; + /** + * Llog callback used in recovery thread on OST (recov_thread.c) + */ + void *lpca_cb; + /** + * Data pointer for llog callback. + */ + void *lpca_arg; }; int llog_cat_put(struct llog_handle *cathandle); @@ -113,13 +173,21 @@ int llog_cat_add_rec(struct llog_handle *cathandle, struct llog_rec_hdr *rec, struct llog_cookie *reccookie, void *buf); int llog_cat_cancel_records(struct llog_handle *cathandle, int count, struct llog_cookie *cookies); -int llog_cat_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data); +int llog_cat_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data, + int startcat, int startidx); +int llog_cat_process_flags(struct llog_handle *cat_llh, llog_cb_t cb, void *data, + int flags, int startcat, int startidx); +int llog_cat_process_thread(void *data); int llog_cat_reverse_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data); int llog_cat_set_first_idx(struct llog_handle *cathandle, int index); /* llog_obd.c */ +int llog_setup_named(struct obd_device *obd, struct obd_llog_group *olg, + int index, struct obd_device *disk_obd, int count, + struct llog_logid *logid, const char *logname, + struct llog_operations *op); int llog_setup(struct obd_device *obd, struct obd_llog_group *olg, int index, - struct obd_device *disk_obd, int count, struct llog_logid *logid, + struct obd_device *disk_obd, int count, struct llog_logid *logid, struct llog_operations *op); int __llog_ctxt_put(struct llog_ctxt *ctxt); int llog_cleanup(struct llog_ctxt *); @@ -130,19 +198,16 @@ int llog_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec, int llog_cancel(struct llog_ctxt *, struct lov_stripe_md *lsm, int count, struct llog_cookie *cookies, int flags); -int llog_obd_origin_setup(struct obd_device *obd, struct obd_llog_group *olg, +int llog_obd_origin_setup(struct obd_device *obd, struct obd_llog_group *olg, int index, struct obd_device *disk_obd, int count, - struct llog_logid *logid); + struct llog_logid *logid, const char *name); int llog_obd_origin_cleanup(struct llog_ctxt *ctxt); int llog_obd_origin_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec, struct lov_stripe_md *lsm, struct llog_cookie *logcookies, int numcookies); -int llog_cat_initialize(struct obd_device *obd, struct obd_llog_group *olg, - int count, struct obd_uuid *uuid); -int obd_llog_init(struct obd_device *obd, int group, - struct obd_device *disk_obd, int count, - struct llog_catid *logid, struct obd_uuid *uuid); +int obd_llog_init(struct obd_device *obd, struct obd_llog_group *olg, + struct obd_device *disk_obd, int *idx); int obd_llog_finish(struct obd_device *obd, int count); @@ -154,7 +219,7 @@ int llog_catalog_list(struct obd_device *obd, int count, /* llog_net.c */ int llog_initiator_connect(struct llog_ctxt *ctxt); int llog_receptor_accept(struct llog_ctxt *ctxt, struct obd_import *imp); -int llog_origin_connect(struct llog_ctxt *ctxt, int count, +int llog_origin_connect(struct llog_ctxt *ctxt, struct llog_logid *logid, struct llog_gen *gen, struct obd_uuid *uuid); int llog_handle_connect(struct ptlrpc_request *req); @@ -164,9 +229,9 @@ int llog_obd_repl_cancel(struct llog_ctxt *ctxt, struct lov_stripe_md *lsm, int count, struct llog_cookie *cookies, int flags); int llog_obd_repl_sync(struct llog_ctxt *ctxt, struct obd_export *exp); -int llog_repl_connect(struct llog_ctxt *ctxt, int count, - struct llog_logid *logid, struct llog_gen *gen, - struct obd_uuid *uuid); +int llog_obd_repl_connect(struct llog_ctxt *ctxt, + struct llog_logid *logid, struct llog_gen *gen, + struct obd_uuid *uuid); struct llog_operations { int (*lop_write_rec)(struct llog_handle *loghandle, @@ -183,9 +248,9 @@ struct llog_operations { int (*lop_close)(struct llog_handle *handle); int (*lop_read_header)(struct llog_handle *handle); - int (*lop_setup)(struct obd_device *obd, struct obd_llog_group *olg, + int (*lop_setup)(struct obd_device *obd, struct obd_llog_group *olg, int ctxt_idx, struct obd_device *disk_obd, int count, - struct llog_logid *logid); + struct llog_logid *logid, const char *name); int (*lop_sync)(struct llog_ctxt *ctxt, struct obd_export *exp); int (*lop_cleanup)(struct llog_ctxt *ctxt); int (*lop_add)(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec, @@ -193,7 +258,7 @@ struct llog_operations { struct llog_cookie *logcookies, int numcookies); int (*lop_cancel)(struct llog_ctxt *ctxt, struct lov_stripe_md *lsm, int count, struct llog_cookie *cookies, int flags); - int (*lop_connect)(struct llog_ctxt *ctxt, int count, + int (*lop_connect)(struct llog_ctxt *ctxt, struct llog_logid *logid, struct llog_gen *gen, struct obd_uuid *uuid); /* XXX add 2 more: commit callbacks and llog recovery functions */ @@ -201,44 +266,132 @@ struct llog_operations { /* llog_lvfs.c */ extern struct llog_operations llog_lvfs_ops; -int llog_get_cat_list(struct obd_device *obd, struct obd_device *disk_obd, - char *name, int count, struct llog_catid *idarray); +int llog_get_cat_list(struct obd_device *disk_obd, + char *name, int idx, int count, + struct llog_catid *idarray); + +int llog_put_cat_list(struct obd_device *disk_obd, + char *name, int idx, int count, struct llog_catid *idarray); + +#define LLOG_CTXT_FLAG_UNINITIALIZED 0x00000001 struct llog_ctxt { int loc_idx; /* my index the obd array of ctxt's */ struct llog_gen loc_gen; struct obd_device *loc_obd; /* points back to the containing obd*/ - struct obd_llog_group *loc_olg; /* group containing that ctxt */ + struct obd_llog_group *loc_olg; /* group containing that ctxt */ struct obd_export *loc_exp; /* parent "disk" export (e.g. MDS) */ struct obd_import *loc_imp; /* to use in RPC's: can be backward pointing import */ struct llog_operations *loc_logops; struct llog_handle *loc_handle; - struct llog_canceld_ctxt *loc_llcd; - struct semaphore loc_sem; /* protects loc_llcd and loc_imp */ - atomic_t loc_refcount; struct llog_commit_master *loc_lcm; + struct llog_canceld_ctxt *loc_llcd; + cfs_semaphore_t loc_sem; /* protects loc_llcd and loc_imp */ + cfs_atomic_t loc_refcount; void *llog_proc_cb; + long loc_flags; /* flags, see above defines */ }; -#ifndef __KERNEL__ +#define LCM_NAME_SIZE 64 + +struct llog_commit_master { + /** + * Thread control flags (start, stop, etc.) + */ + long lcm_flags; + /** + * Number of llcds onthis lcm. + */ + cfs_atomic_t lcm_count; + /** + * The refcount for lcm + */ + cfs_atomic_t lcm_refcount; + /** + * Thread control structure. Used for control commit thread. + */ + struct ptlrpcd_ctl lcm_pc; + /** + * Lock protecting list of llcds. + */ + cfs_spinlock_t lcm_lock; + /** + * Llcds in flight for debugging purposes. + */ + cfs_list_t lcm_llcds; + /** + * Commit thread name buffer. Only used for thread start. + */ + char lcm_name[LCM_NAME_SIZE]; +}; -#define cap_raise(c, flag) do {} while(0) +static inline struct llog_commit_master +*lcm_get(struct llog_commit_master *lcm) +{ + cfs_atomic_inc(&lcm->lcm_refcount); + return lcm; +} + +static inline void +lcm_put(struct llog_commit_master *lcm) +{ + LASSERT_ATOMIC_POS(&lcm->lcm_refcount); + if (cfs_atomic_dec_and_test(&lcm->lcm_refcount)) + OBD_FREE_PTR(lcm); +} -#define CAP_SYS_RESOURCE 24 +struct llog_canceld_ctxt { + /** + * Llog context this llcd is attached to. Used for accessing + * ->loc_import and others in process of canceling cookies + * gathered in this llcd. + */ + struct llog_ctxt *llcd_ctxt; + /** + * Cancel thread control stucture pointer. Used for accessing + * it to see if should stop processing and other needs. + */ + struct llog_commit_master *llcd_lcm; + /** + * Maximal llcd size. Used in calculations on how much of room + * left in llcd to cookie comming cookies. + */ + int llcd_size; + /** + * Link to lcm llcds list. + */ + cfs_list_t llcd_list; + /** + * Current llcd size while gathering cookies. This should not be + * more than ->llcd_size. Used for determining if we need to + * send this llcd (if full) and allocate new one. This is also + * used for copying new cookie at the end of buffer. + */ + int llcd_cookiebytes; + /** + * Pointer to the start of cookies buffer. + */ + struct llog_cookie llcd_cookies[0]; +}; -#endif /* !__KERNEL__ */ +/* ptlrpc/recov_thread.c */ +extern struct llog_commit_master *llog_recov_thread_init(char *name); +extern void llog_recov_thread_fini(struct llog_commit_master *lcm, + int force); +extern int llog_recov_thread_start(struct llog_commit_master *lcm); +extern void llog_recov_thread_stop(struct llog_commit_master *lcm, + int force); static inline void llog_gen_init(struct llog_ctxt *ctxt) { struct obd_device *obd = ctxt->loc_exp->exp_obd; - if (!strcmp(obd->obd_type->typ_name, LUSTRE_MDS_NAME)) - ctxt->loc_gen.mnt_cnt = obd->u.mds.mds_mount_count; - else if (!strstr(obd->obd_type->typ_name, LUSTRE_OST_NAME)) - ctxt->loc_gen.mnt_cnt = obd->u.filter.fo_mount_count; - else - ctxt->loc_gen.mnt_cnt = 0; + LASSERTF(obd->u.obt.obt_magic == OBT_MAGIC, + "%s: wrong obt magic %#x\n", + obd->obd_name, obd->u.obt.obt_magic); + ctxt->loc_gen.mnt_cnt = obd->u.obt.obt_mount_count; + ctxt->loc_gen.conn_cnt++; } static inline int llog_gen_lt(struct llog_gen a, struct llog_gen b) @@ -250,7 +403,6 @@ static inline int llog_gen_lt(struct llog_gen a, struct llog_gen b) return(a.conn_cnt < b.conn_cnt ? 1 : 0); } -#define LLOG_GEN_INC(gen) ((gen).conn_cnt ++) #define LLOG_PROC_BREAK 0x0001 #define LLOG_DEL_RECORD 0x0002 @@ -278,15 +430,14 @@ static inline int llog_handle2ops(struct llog_handle *loghandle, static inline int llog_data_len(int len) { - return size_round(len); + return cfs_size_round(len); } static inline struct llog_ctxt *llog_ctxt_get(struct llog_ctxt *ctxt) { - LASSERT(atomic_read(&ctxt->loc_refcount) > 0); - atomic_inc(&ctxt->loc_refcount); + cfs_atomic_inc(&ctxt->loc_refcount); CDEBUG(D_INFO, "GETting ctxt %p : new refcount %d\n", ctxt, - atomic_read(&ctxt->loc_refcount)); + cfs_atomic_read(&ctxt->loc_refcount)); return ctxt; } @@ -294,32 +445,32 @@ static inline void llog_ctxt_put(struct llog_ctxt *ctxt) { if (ctxt == NULL) return; + LASSERT_ATOMIC_GT_LT(&ctxt->loc_refcount, 0, 0x5a5a5a); CDEBUG(D_INFO, "PUTting ctxt %p : new refcount %d\n", ctxt, - atomic_read(&ctxt->loc_refcount) - 1); - LASSERT(atomic_read(&ctxt->loc_refcount) > 0); - LASSERT(atomic_read(&ctxt->loc_refcount) < 0x5a5a5a); + cfs_atomic_read(&ctxt->loc_refcount) - 1); __llog_ctxt_put(ctxt); } static inline void llog_group_init(struct obd_llog_group *olg, int group) { cfs_waitq_init(&olg->olg_waitq); - spin_lock_init(&olg->olg_lock); - olg->olg_group = group; + cfs_spin_lock_init(&olg->olg_lock); + cfs_sema_init(&olg->olg_cat_processing, 1); + olg->olg_seq = group; } static inline void llog_group_set_export(struct obd_llog_group *olg, struct obd_export *exp) { LASSERT(exp != NULL); - - spin_lock(&olg->olg_lock); + + cfs_spin_lock(&olg->olg_lock); if (olg->olg_exp != NULL && olg->olg_exp != exp) CWARN("%s: export for group %d is changed: 0x%p -> 0x%p\n", - exp->exp_obd->obd_name, olg->olg_group, + exp->exp_obd->obd_name, olg->olg_seq, olg->olg_exp, exp); olg->olg_exp = exp; - spin_unlock(&olg->olg_lock); + cfs_spin_unlock(&olg->olg_lock); } static inline int llog_group_set_ctxt(struct obd_llog_group *olg, @@ -327,13 +478,13 @@ static inline int llog_group_set_ctxt(struct obd_llog_group *olg, { LASSERT(index >= 0 && index < LLOG_MAX_CTXTS); - spin_lock(&olg->olg_lock); + cfs_spin_lock(&olg->olg_lock); if (olg->olg_ctxts[index] != NULL) { - spin_unlock(&olg->olg_lock); + cfs_spin_unlock(&olg->olg_lock); return -EEXIST; } olg->olg_ctxts[index] = ctxt; - spin_unlock(&olg->olg_lock); + cfs_spin_unlock(&olg->olg_lock); return 0; } @@ -344,13 +495,13 @@ static inline struct llog_ctxt *llog_group_get_ctxt(struct obd_llog_group *olg, LASSERT(index >= 0 && index < LLOG_MAX_CTXTS); - spin_lock(&olg->olg_lock); + cfs_spin_lock(&olg->olg_lock); if (olg->olg_ctxts[index] == NULL) { ctxt = NULL; } else { ctxt = llog_ctxt_get(olg->olg_ctxts[index]); } - spin_unlock(&olg->olg_lock); + cfs_spin_unlock(&olg->olg_lock); return ctxt; } @@ -376,8 +527,7 @@ static inline int llog_write_rec(struct llog_handle *handle, int numcookies, void *buf, int idx) { struct llog_operations *lop; - __u32 cap; - int rc, buflen; + int raised, rc, buflen; ENTRY; rc = llog_handle2ops(handle, &lop); @@ -387,17 +537,20 @@ static inline int llog_write_rec(struct llog_handle *handle, if (lop->lop_write_rec == NULL) RETURN(-EOPNOTSUPP); + /* FIXME: Why doesn't caller just set the right lrh_len itself? */ if (buf) buflen = rec->lrh_len + sizeof(struct llog_rec_hdr) + sizeof(struct llog_rec_tail); else buflen = rec->lrh_len; - LASSERT(size_round(buflen) == buflen); + LASSERT(cfs_size_round(buflen) == buflen); - cap = current->cap_effective; - cap_raise(current->cap_effective, CAP_SYS_RESOURCE); + raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); + if (!raised) + cfs_cap_raise(CFS_CAP_SYS_RESOURCE); rc = lop->lop_write_rec(handle, rec, logcookies, numcookies, buf, idx); - current->cap_effective = cap; + if (!raised) + cfs_cap_lower(CFS_CAP_SYS_RESOURCE); RETURN(rc); } @@ -493,8 +646,7 @@ static inline int llog_create(struct llog_ctxt *ctxt, struct llog_handle **res, struct llog_logid *logid, char *name) { struct llog_operations *lop; - __u32 cap; - int rc; + int raised, rc; ENTRY; rc = llog_obd2ops(ctxt, &lop); @@ -503,14 +655,16 @@ static inline int llog_create(struct llog_ctxt *ctxt, struct llog_handle **res, if (lop->lop_create == NULL) RETURN(-EOPNOTSUPP); - cap = current->cap_effective; - cap_raise(current->cap_effective, CAP_SYS_RESOURCE); + raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); + if (!raised) + cfs_cap_raise(CFS_CAP_SYS_RESOURCE); rc = lop->lop_create(ctxt, res, logid, name); - current->cap_effective = cap; + if (!raised) + cfs_cap_lower(CFS_CAP_SYS_RESOURCE); RETURN(rc); } -static inline int llog_connect(struct llog_ctxt *ctxt, int count, +static inline int llog_connect(struct llog_ctxt *ctxt, struct llog_logid *logid, struct llog_gen *gen, struct obd_uuid *uuid) { @@ -524,8 +678,15 @@ static inline int llog_connect(struct llog_ctxt *ctxt, int count, if (lop->lop_connect == NULL) RETURN(-EOPNOTSUPP); - rc = lop->lop_connect(ctxt, count, logid, gen, uuid); + rc = lop->lop_connect(ctxt, logid, gen, uuid); RETURN(rc); } +int lustre_process_log(struct super_block *sb, char *logname, + struct config_llog_instance *cfg); +int lustre_end_log(struct super_block *sb, char *logname, + struct config_llog_instance *cfg); + +/** @} log */ + #endif