Whamcloud - gitweb
LU-3321 clio: revert LU-2622 for removing global env list
[fs/lustre-release.git] / lustre / liblustre / llite_lib.c
index 88c92d1..a4444af 100644 (file)
@@ -1,24 +1,41 @@
-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
  *
- * Lustre Light common routines
+ * 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
  *
- *  Copyright (c) 2002-2004 Cluster File Systems, Inc.
+ * 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.
  *
- *   This file is part of Lustre, http://www.lustre.org.
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
  *
- *   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.
+ * Copyright (c) 2012, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
  *
- *   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.
+ * lustre/liblustre/llite_lib.c
  *
- *   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.
+ * Lustre Light common routines
  */
 
 #include <stdlib.h>
 #include <assert.h>
 #include <signal.h>
 #include <sys/types.h>
+#include <sys/stat.h>
 #include <sys/queue.h>
 
-#include <sysio.h>
-#include <fs.h>
-#include <mount.h>
-#include <inode.h>
-#include <file.h>
-
-#ifdef REDSTORM
-#define CSTART_INIT
-#endif
-
-/* both sys/queue.h (libsysio require it) and portals/lists.h have definition
- * of 'LIST_HEAD'. undef it to suppress warnings
- */
-#undef LIST_HEAD
-#include <portals/ptlctl.h>
+#include <liblustre.h>
+#include <lnet/lnetctl.h>     /* needed for parse_dump */
+#include <lustre_log.h>
 
 #include "lutil.h"
 #include "llite_lib.h"
 
+int slp_global_init(void);
+
 static int lllib_init(void)
 {
-        liblustre_set_nal_nid();
-
-        if (liblustre_init_current("dummy") ||
-            init_obdclass() ||
+        if (liblustre_init_current("liblustre") ||
             init_lib_portals() ||
+            init_obdclass() ||
             ptlrpc_init() ||
+            mgc_init() ||
+            lmv_init() ||
             mdc_init() ||
             lov_init() ||
-            osc_init())
+            osc_init() ||
+            slp_global_init())
                 return -1;
 
-        return _sysio_fssw_register("llite", &llu_fssw_ops);
+        return _sysio_fssw_register("lustre", &llu_fssw_ops);
 }
-#ifndef CRAY_PORTALS
-#define LIBLUSTRE_NAL_NAME "tcp"
-#elif defined REDSTORM
-#define LIBLUSTRE_NAL_NAME "cray_qk_ernal"
-#else
-#define LIBLUSTRE_NAL_NAME "cray_pb_ernal"
-#endif
 
-int liblustre_process_log(struct config_llog_instance *cfg, int allow_recov)
+int liblustre_process_log(struct config_llog_instance *cfg,
+                          char *mgsnid, char *profile,
+                          int allow_recov)
 {
-        struct lustre_cfg lcfg;
-        char  *peer = "MDS_PEER_UUID";
+        struct lustre_cfg_bufs bufs;
+        struct lustre_cfg *lcfg;
+        char  *peer = "MGS_UUID";
         struct obd_device *obd;
-        struct lustre_handle mdc_conn = {0, };
         struct obd_export *exp;
-        char  *name = "mdc_dev";
+        char  *name = "mgc_dev";
         class_uuid_t uuid;
-        struct obd_uuid mdc_uuid;
+        struct obd_uuid mgc_uuid;
         struct llog_ctxt *ctxt;
-        ptl_nid_t nid = 0;
-        int nal, err, rc = 0;
+        lnet_nid_t nid = 0;
+        char *mdsnid;
+        int err, rc = 0;
+        struct obd_connect_data *ocd = NULL;
         ENTRY;
 
-        generate_random_uuid(uuid);
-        class_uuid_unparse(uuid, &mdc_uuid);
+        ll_generate_random_uuid(uuid);
+        class_uuid_unparse(uuid, &mgc_uuid);
 
-        if (ptl_parse_nid(&nid, g_zconf_mdsnid)) {
-                CERROR("Can't parse NID %s\n", g_zconf_mdsnid);
+        nid = libcfs_str2nid(mgsnid);
+        if (nid == LNET_NID_ANY) {
+                CERROR("Can't parse NID %s\n", mgsnid);
                 RETURN(-EINVAL);
         }
 
-        nal = ptl_name2nal(LIBLUSTRE_NAL_NAME);
-        if (nal <= 0) {
-                CERROR("Can't parse NAL %s\n", LIBLUSTRE_NAL_NAME);
-                RETURN(-EINVAL);
+        lustre_cfg_bufs_reset(&bufs, NULL);
+        lustre_cfg_bufs_set_string(&bufs, 1, peer);
+        lcfg = lustre_cfg_new(LCFG_ADD_UUID, &bufs);
+        lcfg->lcfg_nid = nid;
+        rc = class_process_config(lcfg);
+        lustre_cfg_free(lcfg);
+        if (rc < 0)
+                GOTO(out, rc);
+
+        lustre_cfg_bufs_reset(&bufs, name);
+        lustre_cfg_bufs_set_string(&bufs, 1, LUSTRE_MGC_NAME);
+        lustre_cfg_bufs_set_string(&bufs, 2, mgc_uuid.uuid);
+        lcfg = lustre_cfg_new(LCFG_ATTACH, &bufs);
+        rc = class_process_config(lcfg);
+        lustre_cfg_free(lcfg);
+        if (rc < 0)
+                GOTO(out_del_uuid, rc);
+
+        lustre_cfg_bufs_reset(&bufs, name);
+        lustre_cfg_bufs_set_string(&bufs, 1, LUSTRE_MGS_OBDNAME);
+        lustre_cfg_bufs_set_string(&bufs, 2, peer);
+        lcfg = lustre_cfg_new(LCFG_SETUP, &bufs);
+        rc = class_process_config(lcfg);
+        lustre_cfg_free(lcfg);
+        if (rc < 0)
+                GOTO(out_detach, rc);
+
+        while ((mdsnid = strsep(&mgsnid, ","))) {
+                nid = libcfs_str2nid(mdsnid);
+                lustre_cfg_bufs_reset(&bufs, NULL);
+                lustre_cfg_bufs_set_string(&bufs, 1, libcfs_nid2str(nid));
+                lcfg = lustre_cfg_new(LCFG_ADD_UUID, &bufs);
+                lcfg->lcfg_nid = nid;
+                rc = class_process_config(lcfg);
+                lustre_cfg_free(lcfg);
+                if (rc) {
+                        CERROR("Add uuid for %s failed %d\n",
+                               libcfs_nid2str(nid), rc);
+                        continue;
+                }
+
+                lustre_cfg_bufs_reset(&bufs, name);
+                lustre_cfg_bufs_set_string(&bufs, 1, libcfs_nid2str(nid));
+                lcfg = lustre_cfg_new(LCFG_ADD_CONN, &bufs);
+                lcfg->lcfg_nid = nid;
+                rc = class_process_config(lcfg);
+                lustre_cfg_free(lcfg);
+                if (rc) {
+                        CERROR("Add conn for %s failed %d\n",
+                               libcfs_nid2str(nid), rc);
+                        continue;
+                }
         }
-        LCFG_INIT(lcfg, LCFG_ADD_UUID, NULL);
-        lcfg.lcfg_nid = nid;
-        lcfg.lcfg_inllen1 = strlen(peer) + 1;
-        lcfg.lcfg_inlbuf1 = peer;
-        lcfg.lcfg_nal = nal;
-        err = class_process_config(&lcfg);
-        if (err < 0)
-                GOTO(out, err);
-
-        LCFG_INIT(lcfg, LCFG_ATTACH, name);
-        lcfg.lcfg_inlbuf1 = "mdc";
-        lcfg.lcfg_inllen1 = strlen(lcfg.lcfg_inlbuf1) + 1;
-        lcfg.lcfg_inlbuf2 = mdc_uuid.uuid;
-        lcfg.lcfg_inllen2 = strlen(lcfg.lcfg_inlbuf2) + 1;
-        err = class_process_config(&lcfg);
-        if (err < 0)
-                GOTO(out_del_uuid, err);
-
-        LCFG_INIT(lcfg, LCFG_SETUP, name);
-        lcfg.lcfg_inlbuf1 = g_zconf_mdsname;
-        lcfg.lcfg_inllen1 = strlen(lcfg.lcfg_inlbuf1) + 1;
-        lcfg.lcfg_inlbuf2 = peer;
-        lcfg.lcfg_inllen2 = strlen(lcfg.lcfg_inlbuf2) + 1;
-        err = class_process_config(&lcfg);
-        if (err < 0)
-                GOTO(out_detach, err);
-        
+
         obd = class_name2obd(name);
         if (obd == NULL)
-                GOTO(out_cleanup, err = -EINVAL);
+                GOTO(out_cleanup, rc = -EINVAL);
 
-        /* Disable initial recovery on this import */
-        err = obd_set_info(obd->obd_self_export,
-                           strlen("initial_recov"), "initial_recov",
-                           sizeof(allow_recov), &allow_recov);
+        OBD_ALLOC(ocd, sizeof(*ocd));
+        if (ocd == NULL)
+                GOTO(out_cleanup, rc = -ENOMEM);
 
-        err = obd_connect(&mdc_conn, obd, &mdc_uuid, 0);
-        if (err) {
-                CERROR("cannot connect to %s: rc = %d\n",
-                        g_zconf_mdsname, err);
-                GOTO(out_cleanup, err);
+       ocd->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_AT |
+                                OBD_CONNECT_FULL20;
+        ocd->ocd_version = LUSTRE_VERSION_CODE;
+
+        rc = obd_connect(NULL, &exp, obd, &mgc_uuid, ocd, NULL);
+        if (rc) {
+                CERROR("cannot connect to %s at %s: rc = %d\n",
+                       LUSTRE_MGS_OBDNAME, mgsnid, rc);
+                GOTO(out_cleanup, rc);
         }
-        
-        exp = class_conn2export(&mdc_conn);
-        
-        ctxt = exp->exp_obd->obd_llog_ctxt[LLOG_CONFIG_REPL_CTXT];
-        rc = class_config_process_llog(ctxt, g_zconf_profile, cfg);
+
+        ctxt = llog_get_context(exp->exp_obd, LLOG_CONFIG_REPL_CTXT);
+        cfg->cfg_flags |= CFG_F_COMPAT146;
+       rc = class_config_parse_llog(NULL, ctxt, profile, cfg);
+        llog_ctxt_put(ctxt);
         if (rc) {
                 CERROR("class_config_parse_llog failed: rc = %d\n", rc);
         }
 
-        err = obd_disconnect(exp, 0);
+        /* We don't so much care about errors in cleaning up the config llog
+         * connection, as we have already read the config by this point. */
+        err = obd_disconnect(exp);
+        if (err)
+                CERROR("obd_disconnect failed: rc = %d\n", err);
 
 out_cleanup:
-        LCFG_INIT(lcfg, LCFG_CLEANUP, name);
-        err = class_process_config(&lcfg);
-        if (err < 0)
-                GOTO(out, err);
+        if (ocd)
+                OBD_FREE(ocd, sizeof(*ocd));
+
+        lustre_cfg_bufs_reset(&bufs, name);
+        lcfg = lustre_cfg_new(LCFG_CLEANUP, &bufs);
+        err = class_process_config(lcfg);
+        lustre_cfg_free(lcfg);
+        if (err)
+                CERROR("md_cleanup failed: rc = %d\n", err);
 
 out_detach:
-        LCFG_INIT(lcfg, LCFG_DETACH, name);
-        err = class_process_config(&lcfg);
-        if (err < 0)
-                GOTO(out, err);
+        lustre_cfg_bufs_reset(&bufs, name);
+        lcfg = lustre_cfg_new(LCFG_DETACH, &bufs);
+        err = class_process_config(lcfg);
+        lustre_cfg_free(lcfg);
+        if (err)
+                CERROR("md_detach failed: rc = %d\n", err);
 
 out_del_uuid:
-        LCFG_INIT(lcfg, LCFG_DEL_UUID, name);
-        lcfg.lcfg_inllen1 = strlen(peer) + 1;
-        lcfg.lcfg_inlbuf1 = peer;
-        err = class_process_config(&lcfg);
-
+        lustre_cfg_bufs_reset(&bufs, name);
+        lustre_cfg_bufs_set_string(&bufs, 1, peer);
+        lcfg = lustre_cfg_new(LCFG_DEL_UUID, &bufs);
+        err = class_process_config(lcfg);
+        if (err)
+                CERROR("del MDC UUID failed: rc = %d\n", err);
+        lustre_cfg_free(lcfg);
 out:
-        if (rc == 0)
-                rc = err;
-        
+
         RETURN(rc);
 }
 
-/* parse host:/mdsname/profile string */
-int ll_parse_mount_target(const char *target, char **mdsnid,
-                          char **mdsname, char **profile)
+/* parse host:/fsname string */
+int ll_parse_mount_target(const char *target, char **mgsnid,
+                          char **fsname)
 {
         static char buf[256];
         char *s;
@@ -188,17 +230,15 @@ int ll_parse_mount_target(const char *target, char **mdsnid,
         strncpy(buf, target, 255);
 
         if ((s = strchr(buf, ':'))) {
-                *mdsnid = buf;
+                *mgsnid = buf;
                 *s = '\0';
-                                                                                                                        
+
                 while (*++s == '/')
                         ;
-                *mdsname = s;
-                if ((s = strchr(*mdsname, '/'))) {
-                        *s = '\0';
-                        *profile = s + 1;
-                        return 0;
-                }
+                sprintf(s + strlen(s), "-client");
+                *fsname = s;
+
+                return 0;
         }
 
         return -1;
@@ -214,7 +254,7 @@ int ll_parse_mount_target(const char *target, char **mdsnid,
  * or in the apps themselves.  The NAMESPACE_STRING specifying
  * the initial set of fs ops (creates, mounts, etc.) is passed
  * as an environment variable.
- * 
+ *
  *      _sysio_init();
  *      _sysio_incore_init();
  *      _sysio_native_init();
@@ -232,10 +272,9 @@ int ll_parse_mount_target(const char *target, char **mdsnid,
 int _sysio_lustre_init(void)
 {
         int err;
-
-#if 0
-        portal_debug = -1;
-        portal_subsystem_debug = -1;
+        char *envstr;
+#ifndef INIT_SYSIO
+        extern void __liblustre_cleanup_(void);
 #endif
 
         liblustre_init_random();
@@ -243,127 +282,129 @@ int _sysio_lustre_init(void)
         err = lllib_init();
         if (err) {
                 perror("init llite driver");
-        }       
+                return err;
+        }
+
+        envstr = getenv("LIBLUSTRE_TIMEOUT");
+        if (envstr != NULL) {
+                obd_timeout = (unsigned int)strtol(envstr, NULL, 0);
+                printf("LibLustre: obd timeout=%u seconds\n",
+                        obd_timeout);
+        }
+
+        /* debug peer on timeout? */
+        envstr = getenv("LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT");
+        if (envstr != NULL) {
+                obd_debug_peer_on_timeout = 
+                        (unsigned int)strtol(envstr, NULL, 0);
+                printf("LibLustre: debug peer on timeout=%d\n",
+                        obd_debug_peer_on_timeout ? 0 : 1);
+        }
+
+#ifndef INIT_SYSIO
+        (void)atexit(__liblustre_cleanup_);
+#endif
         return err;
 }
 
-/* env variables */
-#define ENV_LUSTRE_MNTPNT               "LIBLUSTRE_MOUNT_POINT"
-#define ENV_LUSTRE_MNTTGT               "LIBLUSTRE_MOUNT_TARGET"
-#define ENV_LUSTRE_TIMEOUT              "LIBLUSTRE_TIMEOUT"
-#define ENV_LUSTRE_DUMPFILE             "LIBLUSTRE_DUMPFILE"
-#define ENV_LUSTRE_DEBUG_MASK           "LIBLUSTRE_DEBUG_MASK"
-#define ENV_LUSTRE_DEBUG_SUBSYS         "LIBLUSTRE_DEBUG_SUBSYS"
-
 extern int _sysio_native_init();
-extern unsigned int obd_timeout;
-
-static char *lustre_path = NULL;
 
-/* global variables */
-char   *g_zconf_mdsname = NULL; /* mdsname, for zeroconf */
-char   *g_zconf_mdsnid = NULL;  /* mdsnid, for zeroconf */
-char   *g_zconf_profile = NULL; /* profile, for zeroconf */
+static int mnt_retry = 0;
 
+char *lustre_path = NULL;
 
 void __liblustre_setup_(void)
 {
         char *target = NULL;
-        char *timeout = NULL;
-        char *debug_mask = NULL;
-        char *debug_subsys = NULL;
-        char *root_driver = "native";
-        char *lustre_driver = "llite";
-        char *root_path = "/";
+        char *lustre_driver = "lustre";
         unsigned mntflgs = 0;
-       int err;
+        int err, count;
 
-       lustre_path = getenv(ENV_LUSTRE_MNTPNT);
-       if (!lustre_path) {
+        lustre_path = getenv("LIBLUSTRE_MOUNT_POINT");
+        if (!lustre_path) {
                 lustre_path = "/mnt/lustre";
-       }
+        }
+
+        target = getenv("LIBLUSTRE_MOUNT_RETRY");
+        if (target) {
+                mnt_retry = atoi(target);
+                if (mnt_retry < 0)
+                        mnt_retry = 0;
+        }
 
         /* mount target */
-        target = getenv(ENV_LUSTRE_MNTTGT);
+        target = getenv("LIBLUSTRE_MOUNT_TARGET");
         if (!target) {
                 printf("LibLustre: no mount target specified\n");
                 exit(1);
         }
-        if (ll_parse_mount_target(target,
-                                  &g_zconf_mdsnid,
-                                  &g_zconf_mdsname,
-                                  &g_zconf_profile)) {
-                CERROR("mal-formed target %s \n", target);
-                exit(1);
-        }
-        if (!g_zconf_mdsnid || !g_zconf_mdsname || !g_zconf_profile) {
-                printf("Liblustre: invalid target %s\n", target);
-                exit(1);
-        }
-        printf("LibLustre: mount point %s, target %s\n",
-                lustre_path, target);
 
-        timeout = getenv(ENV_LUSTRE_TIMEOUT);
-        if (timeout) {
-                obd_timeout = (unsigned int) strtol(timeout, NULL, 0);
-                printf("LibLustre: set obd timeout as %u seconds\n",
-                        obd_timeout);
-        }
+        CDEBUG(D_CONFIG, "LibLustre: mount point %s, target %s\n",
+               lustre_path, target);
 
-        /* debug masks */
-        debug_mask = getenv(ENV_LUSTRE_DEBUG_MASK);
-        if (debug_mask)
-                portal_debug = (unsigned int) strtol(debug_mask, NULL, 0);
-
-        debug_subsys = getenv(ENV_LUSTRE_DEBUG_SUBSYS);
-        if (debug_subsys)
-                portal_subsystem_debug =
-                                (unsigned int) strtol(debug_subsys, NULL, 0);
-
-#ifndef CSTART_INIT
+#ifdef INIT_SYSIO
         /* initialize libsysio & mount rootfs */
-       if (_sysio_init()) {
-               perror("init sysio");
-               exit(1);
-       }
+        if (_sysio_init()) {
+                perror("init sysio");
+                exit(1);
+        }
         _sysio_native_init();
 
-       err = _sysio_mount_root(root_path, root_driver, mntflgs, NULL);
-       if (err) {
-               perror(root_driver);
-               exit(1);
-       }
+        err = _sysio_mount_root("/", "native", mntflgs, NULL);
+        if (err) {
+                fprintf(stderr, "sysio mount failed: %s\n", strerror(errno));
+                exit(1);
+        }
 
         if (_sysio_lustre_init())
-               exit(1);
-#endif
-
-        err = mount("/", lustre_path, lustre_driver, mntflgs, NULL);
-       if (err) {
-               errno = -err;
-               perror(lustre_driver);
-               exit(1);
-       }
+                exit(1);
+#endif /* INIT_SYSIO */
+
+        count = mnt_retry;
+        do {
+                err = mount(target, lustre_path, lustre_driver, mntflgs, NULL);
+                if (err && mnt_retry && (-- count)) {
+                        fprintf(stderr, "Lustre mount failed: %s. "
+                                 "Will retry %d more times\n",
+                                strerror(errno), mnt_retry - count );
+                        sleep(2);
+                }
+        } while (err && count > 0);
+        if (err) {
+                fprintf(stderr, "Lustre mount failed: %s\n", strerror(errno));
+                exit(1);
+        }
 }
 
 void __liblustre_cleanup_(void)
 {
+#ifndef INIT_SYSIO
+        /* guard against being called multiple times */
+        static int cleaned = 0;
+
+        if (cleaned)
+                return;
+        cleaned++;
+#endif
+
         /* user app might chdir to a lustre directory, and leave busy pnode
          * during finaly libsysio cleanup. here we chdir back to "/".
          * but it can't fix the situation that liblustre is mounted
          * at "/".
          */
-        chdir("/");
+        if (!chdir("/")) {}
 #if 0
         umount(lustre_path);
 #endif
         /* we can't call umount here, because libsysio will not cleanup
          * opening files for us. _sysio_shutdown() will cleanup fds at
          * first but which will also close the sockets we need for umount
-         * liblutre. this delima lead to another hack in
+         * liblutre. this dilema lead to another hack in
          * libsysio/src/file_hack.c FIXME
          */
+#ifdef INIT_SYSIO
         _sysio_shutdown();
         cleanup_lib_portals();
-        PtlFini();
+        LNetFini();
+#endif
 }