1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/mds/handler.c
38 * Author: Peter Braam <braam@clusterfs.com>
39 * Author: Andreas Dilger <adilger@clusterfs.com>
40 * Author: Phil Schwan <phil@clusterfs.com>
41 * Author: Mike Shaver <shaver@clusterfs.com>
44 #define DEBUG_SUBSYSTEM S_MDS
46 #include <lustre_mds.h>
47 #include <linux/module.h>
48 #include <linux/init.h>
49 #include <linux/random.h>
51 #include <linux/jbd.h>
52 #include <linux/smp_lock.h>
53 #include <linux/buffer_head.h>
54 #include <linux/workqueue.h>
55 #include <linux/mount.h>
57 #include <lustre_acl.h>
58 #include <obd_class.h>
59 #include <lustre_dlm.h>
61 #include <lustre_fsfilt.h>
62 #include <lprocfs_status.h>
63 #include <lustre_quota.h>
64 #include <lustre_disk.h>
65 #include <lustre_param.h>
67 #include "mds_internal.h"
69 __u32 mds_max_ost_index=0xFFFF;
70 CFS_MODULE_PARM(mds_max_ost_index, "i", int, 0444,
73 /* Look up an entry by inode number. */
74 /* this function ONLY returns valid dget'd dentries with an initialized inode
76 struct dentry *mds_fid2dentry(struct mds_obd *mds, struct ll_fid *fid,
77 struct vfsmount **mnt)
80 unsigned long ino = fid->id;
81 __u32 generation = fid->generation;
83 struct dentry *result;
86 RETURN(ERR_PTR(-ESTALE));
88 snprintf(fid_name, sizeof(fid_name), "0x%lx", ino);
90 CDEBUG(D_DENTRY, "--> mds_fid2dentry: ino/gen %lu/%u, sb %p\n",
91 ino, generation, mds->mds_obt.obt_sb);
93 /* under ext3 this is neither supposed to return bad inodes
95 result = ll_lookup_one_len(fid_name, mds->mds_fid_de, strlen(fid_name));
99 inode = result->d_inode;
101 RETURN(ERR_PTR(-ENOENT));
103 if (inode->i_generation == 0 || inode->i_nlink == 0) {
104 LCONSOLE_WARN("Found inode with zero generation or link -- this"
105 " may indicate disk corruption (inode: %lu/%u, "
106 "link %lu, count %d)\n", inode->i_ino,
107 inode->i_generation,(unsigned long)inode->i_nlink,
108 atomic_read(&inode->i_count));
110 RETURN(ERR_PTR(-ENOENT));
113 if (generation && inode->i_generation != generation) {
114 /* we didn't find the right inode.. */
115 CDEBUG(D_INODE, "found wrong generation: inode %lu, link: %lu, "
116 "count: %d, generation %u/%u\n", inode->i_ino,
117 (unsigned long)inode->i_nlink,
118 atomic_read(&inode->i_count), inode->i_generation,
121 RETURN(ERR_PTR(-ENOENT));
125 *mnt = mds->mds_vfsmnt;
132 static int mds_lov_presetup (struct mds_obd *mds, struct lustre_cfg *lcfg)
137 if (lcfg->lcfg_bufcount >= 4 && LUSTRE_CFG_BUFLEN(lcfg, 3) > 0) {
140 ll_generate_random_uuid(uuid);
141 class_uuid_unparse(uuid, &mds->mds_lov_uuid);
143 OBD_ALLOC(mds->mds_profile, LUSTRE_CFG_BUFLEN(lcfg, 3));
144 if (mds->mds_profile == NULL)
147 strncpy(mds->mds_profile, lustre_cfg_string(lcfg, 3),
148 LUSTRE_CFG_BUFLEN(lcfg, 3));
153 static int mds_lov_clean(struct obd_device *obd)
155 struct mds_obd *mds = &obd->u.mds;
156 struct obd_device *osc = mds->mds_osc_obd;
159 if (mds->mds_profile) {
160 class_del_profile(mds->mds_profile);
161 OBD_FREE(mds->mds_profile, strlen(mds->mds_profile) + 1);
162 mds->mds_profile = NULL;
165 /* There better be a lov */
169 RETURN(PTR_ERR(osc));
171 obd_register_observer(osc, NULL);
173 /* Give lov our same shutdown flags */
174 osc->obd_force = obd->obd_force;
175 osc->obd_fail = obd->obd_fail;
177 /* Cleanup the lov */
178 obd_disconnect(mds->mds_osc_exp);
179 class_manual_cleanup(osc);
184 static int mds_postsetup(struct obd_device *obd)
186 struct mds_obd *mds = &obd->u.mds;
187 struct llog_ctxt *ctxt;
191 rc = llog_setup(obd, &obd->obd_olg, LLOG_CONFIG_ORIG_CTXT, obd, 0, NULL,
196 rc = llog_setup(obd, &obd->obd_olg, LLOG_LOVEA_ORIG_CTXT, obd, 0, NULL,
201 if (mds->mds_profile) {
202 struct lustre_profile *lprof;
203 /* The profile defines which osc and mdc to connect to, for a
204 client. We reuse that here to figure out the name of the
205 lov to use (and ignore lprof->lp_md).
206 The profile was set in the config log with
207 LCFG_MOUNTOPT profilenm oscnm mdcnm */
208 lprof = class_get_profile(mds->mds_profile);
210 CERROR("No profile found: %s\n", mds->mds_profile);
211 GOTO(err_cleanup, rc = -ENOENT);
213 rc = mds_lov_connect(obd, lprof->lp_dt);
215 GOTO(err_cleanup, rc);
222 ctxt = llog_get_context(obd, LLOG_LOVEA_ORIG_CTXT);
226 ctxt = llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
232 int mds_postrecov(struct obd_device *obd)
240 LASSERT(!obd->obd_recovering);
241 /* clean PENDING dir */
243 if (strncmp(obd->obd_name, MDD_OBD_NAME, strlen(MDD_OBD_NAME)))
244 rc = mds_cleanup_pending(obd);
248 /* FIXME Does target_finish_recovery really need this to block? */
249 /* Notify the LOV, which will in turn call mds_notify for each tgt */
250 /* This means that we have to hack obd_notify to think we're obd_set_up
251 during mds_lov_connect. */
252 obd_notify(obd->u.mds.mds_osc_obd, NULL,
253 obd->obd_async_recov ? OBD_NOTIFY_SYNC_NONBLOCK :
254 OBD_NOTIFY_SYNC, NULL);
257 lquota_recovery(mds_quota_interface_ref, obd);
262 /* We need to be able to stop an mds_lov_synchronize */
263 static int mds_lov_early_clean(struct obd_device *obd)
265 struct mds_obd *mds = &obd->u.mds;
266 struct obd_device *osc = mds->mds_osc_obd;
268 if (!osc || (!obd->obd_force && !obd->obd_fail))
271 CDEBUG(D_HA, "abort inflight\n");
272 return (obd_precleanup(osc, OBD_CLEANUP_EARLY));
275 static int mds_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
278 struct mds_obd *mds = &obd->u.mds;
282 case OBD_CLEANUP_EARLY:
284 case OBD_CLEANUP_EXPORTS:
285 mds_lov_early_clean(obd);
286 down_write(&mds->mds_notify_lock);
287 mds_lov_disconnect(obd);
289 llog_cleanup(llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT));
290 llog_cleanup(llog_get_context(obd, LLOG_LOVEA_ORIG_CTXT));
291 rc = obd_llog_finish(obd, 0);
292 mds->mds_osc_exp = NULL;
293 up_write(&mds->mds_notify_lock);
299 static struct dentry *mds_lvfs_fid2dentry(__u64 id, __u32 gen, __u64 gr,
302 struct obd_device *obd = data;
305 fid.generation = gen;
306 return mds_fid2dentry(&obd->u.mds, &fid, NULL);
310 struct lvfs_callback_ops mds_lvfs_ops = {
311 l_fid2dentry: mds_lvfs_fid2dentry,
314 quota_interface_t *mds_quota_interface_ref;
315 extern quota_interface_t mds_quota_interface;
317 static void mds_init_ctxt(struct obd_device *obd, struct vfsmount *mnt)
319 struct mds_obd *mds = &obd->u.mds;
321 mds->mds_vfsmnt = mnt;
322 /* why not mnt->mnt_sb instead of mnt->mnt_root->d_inode->i_sb? */
323 obd->u.obt.obt_sb = mnt->mnt_root->d_inode->i_sb;
325 fsfilt_setup(obd, obd->u.obt.obt_sb);
327 OBD_SET_CTXT_MAGIC(&obd->obd_lvfs_ctxt);
328 obd->obd_lvfs_ctxt.pwdmnt = mnt;
329 obd->obd_lvfs_ctxt.pwd = mnt->mnt_root;
330 obd->obd_lvfs_ctxt.fs = get_ds();
331 obd->obd_lvfs_ctxt.cb_ops = mds_lvfs_ops;
335 /*mds still need lov setup here*/
336 static int mds_cmd_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
338 struct mds_obd *mds = &obd->u.mds;
339 struct lvfs_run_ctxt saved;
341 struct vfsmount *mnt;
342 struct lustre_sb_info *lsi;
343 struct lustre_mount_info *lmi;
344 struct dentry *dentry;
348 CDEBUG(D_INFO, "obd %s setup \n", obd->obd_name);
349 if (strncmp(obd->obd_name, MDD_OBD_NAME, strlen(MDD_OBD_NAME)))
352 if (lcfg->lcfg_bufcount < 5) {
353 CERROR("invalid arg for setup %s\n", MDD_OBD_NAME);
356 dev = lustre_cfg_string(lcfg, 4);
357 lmi = server_get_mount(dev);
358 LASSERT(lmi != NULL);
360 lsi = s2lsi(lmi->lmi_sb);
362 /* FIXME: MDD LOV initialize objects.
363 * we need only lmi here but not get mount
364 * OSD did mount already, so put mount back
366 atomic_dec(&lsi->lsi_mounts);
368 init_rwsem(&mds->mds_notify_lock);
370 obd->obd_fsops = fsfilt_get_ops(MT_STR(lsi->lsi_ldd));
371 mds_init_ctxt(obd, mnt);
373 push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
374 dentry = simple_mkdir(current->fs->pwd, mnt, "OBJECTS", 0777, 1);
375 if (IS_ERR(dentry)) {
376 rc = PTR_ERR(dentry);
377 CERROR("cannot create OBJECTS directory: rc = %d\n", rc);
380 mds->mds_objects_dir = dentry;
382 dentry = lookup_one_len("__iopen__", current->fs->pwd,
383 strlen("__iopen__"));
384 if (IS_ERR(dentry)) {
385 rc = PTR_ERR(dentry);
386 CERROR("cannot lookup __iopen__ directory: rc = %d\n", rc);
387 GOTO(err_objects, rc);
390 mds->mds_fid_de = dentry;
391 if (!dentry->d_inode || is_bad_inode(dentry->d_inode)) {
393 CERROR("__iopen__ directory has no inode? rc = %d\n", rc);
396 rc = mds_lov_init_objids(obd);
398 CERROR("cannot init lov objid rc = %d\n", rc);
402 rc = mds_lov_presetup(mds, lcfg);
404 GOTO(err_objects, rc);
406 /* Don't wait for mds_postrecov trying to clear orphans */
407 obd->obd_async_recov = 1;
408 rc = mds_postsetup(obd);
409 /* Bug 11557 - allow async abort_recov start
410 FIXME can remove most of this obd_async_recov plumbing
411 obd->obd_async_recov = 0;
415 GOTO(err_objects, rc);
417 mds->mds_max_mdsize = sizeof(struct lov_mds_md_v3);
418 mds->mds_max_cookiesize = sizeof(struct llog_cookie);
421 pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
424 dput(mds->mds_fid_de);
426 dput(mds->mds_objects_dir);
428 fsfilt_put_ops(obd->obd_fsops);
432 static int mds_cmd_cleanup(struct obd_device *obd)
434 struct mds_obd *mds = &obd->u.mds;
435 struct lvfs_run_ctxt saved;
439 mds->mds_osc_exp = NULL;
442 LCONSOLE_WARN("%s: shutting down for failover; client state "
443 "will be preserved.\n", obd->obd_name);
445 push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
447 mds_lov_destroy_objids(obd);
449 if (mds->mds_objects_dir != NULL) {
450 l_dput(mds->mds_objects_dir);
451 mds->mds_objects_dir = NULL;
454 shrink_dcache_parent(mds->mds_fid_de);
455 dput(mds->mds_fid_de);
456 LL_DQUOT_OFF(obd->u.obt.obt_sb);
457 fsfilt_put_ops(obd->obd_fsops);
459 pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
464 static int mds_cmd_health_check(struct obd_device *obd)
469 static struct obd_ops mds_cmd_obd_ops = {
470 .o_owner = THIS_MODULE,
471 .o_setup = mds_cmd_setup,
472 .o_cleanup = mds_cmd_cleanup,
473 .o_precleanup = mds_precleanup,
474 .o_create = mds_obd_create,
475 .o_destroy = mds_obd_destroy,
476 .o_llog_init = mds_llog_init,
477 .o_llog_finish = mds_llog_finish,
478 .o_notify = mds_notify,
479 .o_postrecov = mds_postrecov,
480 // .o_health_check = mds_cmd_health_check,
483 static int __init mds_cmd_init(void)
485 struct lprocfs_static_vars lvars;
487 lprocfs_mds_init_vars(&lvars);
488 class_register_type(&mds_cmd_obd_ops, NULL, lvars.module_vars,
489 LUSTRE_MDS_NAME, NULL);
494 static void /*__exit*/ mds_cmd_exit(void)
496 class_unregister_type(LUSTRE_MDS_NAME);
499 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
500 MODULE_DESCRIPTION("Lustre Metadata Server (MDS)");
501 MODULE_LICENSE("GPL");
503 module_init(mds_cmd_init);
504 module_exit(mds_cmd_exit);