lustre/obdfilter/filter.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  *  linux/fs/obdfilter/filter.c
   5  *
   6  *  Copyright (c) 2001-2003 Cluster File Systems, Inc.
   7  *   Author: Peter Braam <braam@clusterfs.com>
   8  *   Author: Andreas Dilger <adilger@clusterfs.com>
   9  *
  10  *   This file is part of Lustre, http://www.lustre.org.
  11  *
  12  *   Lustre is free software; you can redistribute it and/or
  13  *   modify it under the terms of version 2 of the GNU General Public
  14  *   License as published by the Free Software Foundation.
  15  *
  16  *   Lustre is distributed in the hope that it will be useful,
  17  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  *   GNU General Public License for more details.
  20  *
  21  *   You should have received a copy of the GNU General Public License
  22  *   along with Lustre; if not, write to the Free Software
  23  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  24  */
  25
  26 /*
  27  * Invariant: Get O/R i_sem for lookup, if needed, before any journal ops
  28  *            (which need to get journal_lock, may block if journal full).
  29  *
  30  * Invariant: Call filter_start_transno() before any journal ops to avoid the
  31  *            same deadlock problem.  We can (and want) to get rid of the
  32  *            transno sem in favour of the dir/inode i_sem to avoid single
  33  *            threaded operation on the OST.
  34  */
  35
  36 #define DEBUG_SUBSYSTEM S_FILTER
  37
  38 #include <linux/config.h>
  39 #include <linux/module.h>
  40 #include <linux/fs.h>
  41 #include <linux/dcache.h>
  42 #include <linux/init.h>
  43 #include <linux/version.h>
  44 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
  45 # include <linux/mount.h>
  46 # include <linux/buffer_head.h>
  47 #endif
  48
  49 #include <linux/obd_class.h>
  50 #include <linux/lustre_dlm.h>
  51 #include <linux/lustre_fsfilt.h>
  52 #include <linux/lprocfs_status.h>
  53 #include <linux/lustre_log.h>
  54 #include <linux/lustre_commit_confd.h>
  55
  56 #include "filter_internal.h"
  57
  58 #define S_SHIFT 12
  59 static char *obd_type_by_mode[S_IFMT >> S_SHIFT] = {
  60         [0]                     NULL,
  61         [S_IFREG >> S_SHIFT]    "R",
  62         [S_IFDIR >> S_SHIFT]    "D",
  63         [S_IFCHR >> S_SHIFT]    "C",
  64         [S_IFBLK >> S_SHIFT]    "B",
  65         [S_IFIFO >> S_SHIFT]    "F",
  66         [S_IFSOCK >> S_SHIFT]   "S",
  67         [S_IFLNK >> S_SHIFT]    "L"
  68 };
  69
  70 static inline const char *obd_mode_to_type(int mode)
  71 {
  72         return obd_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
  73 }
  74
  75 static void filter_ffd_addref(void *ffdp)
  76 {
  77         struct filter_file_data *ffd = ffdp;
  78
  79         atomic_inc(&ffd->ffd_refcount);
  80         CDEBUG(D_INFO, "GETting ffd %p : new refcount %d\n", ffd,
  81                atomic_read(&ffd->ffd_refcount));
  82 }
  83
  84 static struct filter_file_data *filter_ffd_new(void)
  85 {
  86         struct filter_file_data *ffd;
  87
  88         OBD_ALLOC(ffd, sizeof *ffd);
  89         if (ffd == NULL) {
  90                 CERROR("out of memory\n");
  91                 return NULL;
  92         }
  93
  94         atomic_set(&ffd->ffd_refcount, 2);
  95
  96         INIT_LIST_HEAD(&ffd->ffd_handle.h_link);
  97         class_handle_hash(&ffd->ffd_handle, filter_ffd_addref);
  98
  99         return ffd;
 100 }
 101
 102 static struct filter_file_data *filter_handle2ffd(struct lustre_handle *handle)
 103 {
 104         struct filter_file_data *ffd = NULL;
 105         ENTRY;
 106         LASSERT(handle != NULL);
 107         ffd = class_handle2object(handle->cookie);
 108         if (ffd != NULL)
 109                 LASSERT(ffd->ffd_file->private_data == ffd);
 110         RETURN(ffd);
 111 }
 112
 113 static void filter_ffd_put(struct filter_file_data *ffd)
 114 {
 115         CDEBUG(D_INFO, "PUTting ffd %p : new refcount %d\n", ffd,
 116                atomic_read(&ffd->ffd_refcount) - 1);
 117         LASSERT(atomic_read(&ffd->ffd_refcount) > 0 &&
 118                 atomic_read(&ffd->ffd_refcount) < 0x5a5a);
 119         if (atomic_dec_and_test(&ffd->ffd_refcount)) {
 120                 LASSERT(list_empty(&ffd->ffd_handle.h_link));
 121                 OBD_FREE(ffd, sizeof *ffd);
 122         }
 123 }
 124
 125 static void filter_ffd_destroy(struct filter_file_data *ffd)
 126 {
 127         class_handle_unhash(&ffd->ffd_handle);
 128         filter_ffd_put(ffd);
 129 }
 130
 131 static void filter_commit_cb(struct obd_device *obd, __u64 transno,
 132                              void *cb_data, int error)
 133 {
 134         obd_transno_commit_cb(obd, transno, error);
 135 }
 136
 137 static int filter_client_log_cancel(struct lustre_handle *conn,
 138                                     struct lov_stripe_md *lsm, int count,
 139                                     struct llog_cookie *cookies, int flags)
 140 {
 141         struct obd_device *obd = class_conn2obd(conn);
 142         struct llog_commit_data *llcd;
 143         struct filter_obd *filter = &obd->u.filter;
 144         int rc = 0;
 145         ENTRY;
 146
 147         if (count == 0 || cookies == NULL) {
 148                 down(&filter->fo_sem);
 149                 if (filter->fo_llcd == NULL || !(flags & OBD_LLOG_FL_SENDNOW))
 150                         GOTO(out, rc);
 151
 152                 llcd = filter->fo_llcd;
 153                 GOTO(send_now, rc);
 154         }
 155
 156         down(&filter->fo_sem);
 157         llcd = filter->fo_llcd;
 158         if (llcd == NULL) {
 159                 llcd = llcd_grab();
 160                 if (llcd == NULL) {
 161                         CERROR("couldn't get an llcd - dropped "LPX64":%x+%u\n",
 162                                cookies->lgc_lgl.lgl_oid,
 163                                cookies->lgc_lgl.lgl_ogen, cookies->lgc_index);
 164                         GOTO(out, rc = -ENOMEM);
 165                 }
 166                 llcd->llcd_import = filter->fo_mdc_imp;
 167                 filter->fo_llcd = llcd;
 168         }
 169
 170         memcpy(llcd->llcd_cookies + llcd->llcd_cookiebytes, cookies,
 171                sizeof(*cookies));
 172         llcd->llcd_cookiebytes += sizeof(*cookies);
 173
 174         GOTO(send_now, rc);
 175 send_now:
 176         if ((PAGE_SIZE - llcd->llcd_cookiebytes < sizeof(*cookies) ||
 177              flags & OBD_LLOG_FL_SENDNOW)) {
 178                 filter->fo_llcd = NULL;
 179                 llcd_send(llcd);
 180         }
 181 out:
 182         up(&filter->fo_sem);
 183
 184         return rc;
 185 }
 186
 187 /* When this (destroy) operation is committed, return the cancel cookie */
 188 static void filter_cancel_cookies_cb(struct obd_device *obd, __u64 transno,
 189                                      void *cb_data, int error)
 190 {
 191         filter_client_log_cancel(&obd->u.filter.fo_mdc_conn, NULL, 1,
 192                                  cb_data, OBD_LLOG_FL_SENDNOW);
 193         OBD_FREE(cb_data, sizeof(struct llog_cookie));
 194 }
 195
 196 /* Assumes caller has already pushed us into the kernel context. */
 197 int filter_finish_transno(struct obd_export *exp, struct obd_trans_info *oti,
 198                           int rc)
 199 {
 200         struct filter_obd *filter = &exp->exp_obd->u.filter;
 201         struct filter_export_data *fed = &exp->exp_filter_data;
 202         struct filter_client_data *fcd = fed->fed_fcd;
 203         __u64 last_rcvd;
 204         loff_t off;
 205         ssize_t written;
 206
 207         /* Propagate error code. */
 208         if (rc)
 209                 RETURN(rc);
 210
 211         if (!exp->exp_obd->obd_replayable)
 212                 RETURN(rc);
 213
 214         /* we don't allocate new transnos for replayed requests */
 215         if (oti != NULL && oti->oti_transno == 0) {
 216                 spin_lock(&filter->fo_translock);
 217                 last_rcvd = le64_to_cpu(filter->fo_fsd->fsd_last_transno) + 1;
 218                 filter->fo_fsd->fsd_last_transno = cpu_to_le64(last_rcvd);
 219                 spin_unlock(&filter->fo_translock);
 220                 oti->oti_transno = last_rcvd;
 221                 fcd->fcd_last_rcvd = cpu_to_le64(last_rcvd);
 222                 fcd->fcd_mount_count = filter->fo_fsd->fsd_mount_count;
 223
 224                 /* could get xid from oti, if it's ever needed */
 225                 fcd->fcd_last_xid = 0;
 226
 227                 off = fed->fed_lr_off;
 228                 fsfilt_set_last_rcvd(exp->exp_obd, last_rcvd, oti->oti_handle,
 229                                      filter_commit_cb, NULL);
 230                 written = fsfilt_write_record(exp->exp_obd,
 231                                               filter->fo_rcvd_filp, (char *)fcd,
 232                                               sizeof(*fcd), &off);
 233                 CDEBUG(D_HA, "wrote trans #"LPD64" for client %s at #%d: "
 234                        "written = "LPSZ"\n", last_rcvd, fcd->fcd_uuid,
 235                        fed->fed_lr_idx, written);
 236
 237                 if (written == sizeof(*fcd))
 238                         RETURN(0);
 239                 CERROR("error writing to %s: rc = %d\n", LAST_RCVD,
 240                        (int)written);
 241                 if (written >= 0)
 242                         RETURN(-ENOSPC);
 243                 RETURN(written);
 244         }
 245
 246         RETURN(0);
 247 }
 248
 249 void f_dput(struct dentry *dentry)
 250 {
 251         /* Can't go inside filter_ddelete because it can block */
 252         CDEBUG(D_INODE, "putting %s: %p, count = %d\n",
 253                dentry->d_name.name, dentry, atomic_read(&dentry->d_count) - 1);
 254         LASSERT(atomic_read(&dentry->d_count) > 0);
 255
 256         dput(dentry);
 257 }
 258
 259 /* Not racy w.r.t. others, because we are the only user of this dentry */
 260 static void filter_drelease(struct dentry *dentry)
 261 {
 262         if (dentry->d_fsdata)
 263                 OBD_FREE(dentry->d_fsdata, sizeof(struct filter_dentry_data));
 264 }
 265
 266 struct dentry_operations filter_dops = {
 267         d_release: filter_drelease,
 268 };
 269
 270 /* Add client data to the FILTER.  We use a bitmap to locate a free space
 271  * in the last_rcvd file if cl_idx is -1 (i.e. a new client).
 272  * Otherwise, we have just read the data from the last_rcvd file and
 273  * we know its offset. */
 274 static int filter_client_add(struct obd_device *obd, struct filter_obd *filter,
 275                              struct filter_export_data *fed, int cl_idx)
 276 {
 277         unsigned long *bitmap = filter->fo_last_rcvd_slots;
 278         int new_client = (cl_idx == -1);
 279         ENTRY;
 280
 281         LASSERT(bitmap != NULL);
 282
 283         /* XXX if fcd_uuid were a real obd_uuid, I could use obd_uuid_equals */
 284         if (!strcmp(fed->fed_fcd->fcd_uuid, "OBD_CLASS_UUID"))
 285                 RETURN(0);
 286
 287         /* the bitmap operations can handle cl_idx > sizeof(long) * 8, so
 288          * there's no need for extra complication here
 289          */
 290         if (new_client) {
 291                 cl_idx = find_first_zero_bit(bitmap, FILTER_LR_MAX_CLIENTS);
 292         repeat:
 293                 if (cl_idx >= FILTER_LR_MAX_CLIENTS) {
 294                         CERROR("no client slots - fix FILTER_LR_MAX_CLIENTS\n");
 295                         RETURN(-ENOMEM);
 296                 }
 297                 if (test_and_set_bit(cl_idx, bitmap)) {
 298                         CERROR("FILTER client %d: found bit is set in bitmap\n",
 299                                cl_idx);
 300                         cl_idx = find_next_zero_bit(bitmap,
 301                                                     FILTER_LR_MAX_CLIENTS,
 302                                                     cl_idx);
 303                         goto repeat;
 304                 }
 305         } else {
 306                 if (test_and_set_bit(cl_idx, bitmap)) {
 307                         CERROR("FILTER client %d: bit already set in bitmap!\n",
 308                                cl_idx);
 309                         LBUG();
 310                 }
 311         }
 312
 313         fed->fed_lr_idx = cl_idx;
 314         fed->fed_lr_off = le32_to_cpu(filter->fo_fsd->fsd_client_start) +
 315                 cl_idx * le16_to_cpu(filter->fo_fsd->fsd_client_size);
 316
 317         CDEBUG(D_INFO, "client at index %d (%llu) with UUID '%s' added\n",
 318                fed->fed_lr_idx, fed->fed_lr_off, fed->fed_fcd->fcd_uuid);
 319
 320         if (new_client) {
 321                 struct obd_run_ctxt saved;
 322                 loff_t off = fed->fed_lr_off;
 323                 int written;
 324                 void *handle;
 325
 326                 CDEBUG(D_INFO, "writing client fcd at idx %u (%llu) (len %u)\n",
 327                        fed->fed_lr_idx,off,(unsigned int)sizeof(*fed->fed_fcd));
 328
 329                 push_ctxt(&saved, &filter->fo_ctxt, NULL);
 330                 /* Transaction needed to fix bug 1403 */
 331                 handle = fsfilt_start(obd,
 332                                       filter->fo_rcvd_filp->f_dentry->d_inode,
 333                                       FSFILT_OP_SETATTR, NULL);
 334                 if (IS_ERR(handle)) {
 335                         written = PTR_ERR(handle);
 336                         CERROR("unable to start transaction: rc %d\n",
 337                                (int)written);
 338                 } else {
 339                         written = fsfilt_write_record(obd, filter->fo_rcvd_filp,
 340                                                 (char *)fed->fed_fcd,
 341                                                 sizeof(*fed->fed_fcd), &off);
 342                         fsfilt_commit(obd,
 343                                       filter->fo_rcvd_filp->f_dentry->d_inode,
 344                                       handle, 0);
 345                 }
 346                 pop_ctxt(&saved, &filter->fo_ctxt, NULL);
 347
 348                 if (written != sizeof(*fed->fed_fcd)) {
 349                         CERROR("error writing %s client idx %u: rc %d\n",
 350                                LAST_RCVD, fed->fed_lr_idx, written);
 351                         if (written < 0)
 352                                 RETURN(written);
 353                         RETURN(-ENOSPC);
 354                 }
 355         }
 356         RETURN(0);
 357 }
 358
 359 static int filter_client_free(struct obd_export *exp, int flags)
 360 {
 361         struct filter_export_data *fed = &exp->exp_filter_data;
 362         struct filter_obd *filter = &exp->exp_obd->u.filter;
 363         struct obd_device *obd = exp->exp_obd;
 364         struct filter_client_data zero_fcd;
 365         struct obd_run_ctxt saved;
 366         int written;
 367         loff_t off;
 368         ENTRY;
 369
 370         if (fed->fed_fcd == NULL)
 371                 RETURN(0);
 372
 373         if (flags & OBD_OPT_FAILOVER)
 374                 GOTO(free, 0);
 375
 376         /* XXX if fcd_uuid were a real obd_uuid, I could use obd_uuid_equals */
 377         if (strcmp(fed->fed_fcd->fcd_uuid, "OBD_CLASS_UUID") == 0)
 378                 GOTO(free, 0);
 379
 380         LASSERT(filter->fo_last_rcvd_slots != NULL);
 381
 382         off = fed->fed_lr_off;
 383
 384         CDEBUG(D_INFO, "freeing client at idx %u (%lld) with UUID '%s'\n",
 385                fed->fed_lr_idx, fed->fed_lr_off, fed->fed_fcd->fcd_uuid);
 386
 387         if (!test_and_clear_bit(fed->fed_lr_idx, filter->fo_last_rcvd_slots)) {
 388                 CERROR("FILTER client %u: bit already clear in bitmap!!\n",
 389                        fed->fed_lr_idx);
 390                 LBUG();
 391         }
 392
 393         memset(&zero_fcd, 0, sizeof zero_fcd);
 394         push_ctxt(&saved, &filter->fo_ctxt, NULL);
 395         written = fsfilt_write_record(obd, filter->fo_rcvd_filp,
 396                                       (char *)&zero_fcd, sizeof(zero_fcd),
 397                                       &off);
 398
 399         /* XXX: this write gets lost sometimes, unless this sync is here. */
 400         if (written > 0)
 401                 file_fsync(filter->fo_rcvd_filp,
 402                            filter->fo_rcvd_filp->f_dentry, 1);
 403         pop_ctxt(&saved, &filter->fo_ctxt, NULL);
 404
 405         if (written != sizeof(zero_fcd)) {
 406                 CERROR("error zeroing out client %s idx %u (%llu) in %s: %d\n",
 407                        fed->fed_fcd->fcd_uuid, fed->fed_lr_idx, fed->fed_lr_off,
 408                        LAST_RCVD, written);
 409         } else {
 410                 CDEBUG(D_INFO,
 411                        "zeroed disconnecting client %s at idx %u (%llu)\n",
 412                        fed->fed_fcd->fcd_uuid, fed->fed_lr_idx,fed->fed_lr_off);
 413         }
 414
 415 free:
 416         OBD_FREE(fed->fed_fcd, sizeof(*fed->fed_fcd));
 417
 418         RETURN(0);
 419 }
 420
 421 static int filter_free_server_data(struct filter_obd *filter)
 422 {
 423         OBD_FREE(filter->fo_fsd, sizeof(*filter->fo_fsd));
 424         filter->fo_fsd = NULL;
 425         OBD_FREE(filter->fo_last_rcvd_slots,
 426                  FILTER_LR_MAX_CLIENT_WORDS * sizeof(unsigned long));
 427         filter->fo_last_rcvd_slots = NULL;
 428         return 0;
 429 }
 430
 431 /* assumes caller is already in kernel ctxt */
 432 int filter_update_server_data(struct obd_device *obd,
 433                               struct file *filp, struct filter_server_data *fsd)
 434 {
 435         loff_t off = 0;
 436         int rc;
 437         ENTRY;
 438
 439         CDEBUG(D_INODE, "server uuid      : %s\n", fsd->fsd_uuid);
 440         CDEBUG(D_INODE, "server last_objid: "LPU64"\n",
 441                le64_to_cpu(fsd->fsd_last_objid));
 442         CDEBUG(D_INODE, "server last_rcvd : "LPU64"\n",
 443                le64_to_cpu(fsd->fsd_last_transno));
 444         CDEBUG(D_INODE, "server last_mount: "LPU64"\n",
 445                le64_to_cpu(fsd->fsd_mount_count));
 446
 447         rc = fsfilt_write_record(obd, filp, (char *)fsd, sizeof(*fsd), &off);
 448         if (rc == sizeof(*fsd))
 449                 RETURN(0);
 450
 451         CDEBUG(D_INODE, "error writing filter_server_data: rc = %d\n", rc);
 452         if (rc >= 0)
 453                 RETURN(-ENOSPC);
 454         RETURN(rc);
 455 }
 456
 457 /* assumes caller has already in kernel ctxt */
 458 static int filter_init_server_data(struct obd_device *obd, struct file * filp,
 459                                    __u64 init_lastobjid)
 460 {
 461         struct filter_obd *filter = &obd->u.filter;
 462         struct filter_server_data *fsd;
 463         struct filter_client_data *fcd = NULL;
 464         struct inode *inode = filp->f_dentry->d_inode;
 465         unsigned long last_rcvd_size = inode->i_size;
 466         __u64 mount_count = 0;
 467         int cl_idx;
 468         loff_t off = 0;
 469         int rc;
 470
 471         /* ensure padding in the struct is the correct size */
 472         LASSERT (offsetof(struct filter_server_data, fsd_padding) +
 473                  sizeof(fsd->fsd_padding) == FILTER_LR_SERVER_SIZE);
 474         LASSERT (offsetof(struct filter_client_data, fcd_padding) +
 475                  sizeof(fcd->fcd_padding) == FILTER_LR_CLIENT_SIZE);
 476
 477         OBD_ALLOC(fsd, sizeof(*fsd));
 478         if (!fsd)
 479                 RETURN(-ENOMEM);
 480         filter->fo_fsd = fsd;
 481
 482         OBD_ALLOC(filter->fo_last_rcvd_slots,
 483                   FILTER_LR_MAX_CLIENT_WORDS * sizeof(unsigned long));
 484         if (filter->fo_last_rcvd_slots == NULL) {
 485                 OBD_FREE(fsd, sizeof(*fsd));
 486                 RETURN(-ENOMEM);
 487         }
 488
 489         if (last_rcvd_size == 0) {
 490                 CWARN("%s: initializing new %s\n", obd->obd_name, LAST_RCVD);
 491
 492                 memcpy(fsd->fsd_uuid, obd->obd_uuid.uuid,sizeof(fsd->fsd_uuid));
 493                 fsd->fsd_last_objid = cpu_to_le64(init_lastobjid);
 494                 fsd->fsd_last_transno = 0;
 495                 mount_count = fsd->fsd_mount_count = 0;
 496                 fsd->fsd_server_size = cpu_to_le32(FILTER_LR_SERVER_SIZE);
 497                 fsd->fsd_client_start = cpu_to_le32(FILTER_LR_CLIENT_START);
 498                 fsd->fsd_client_size = cpu_to_le16(FILTER_LR_CLIENT_SIZE);
 499                 fsd->fsd_subdir_count = cpu_to_le16(FILTER_SUBDIR_COUNT);
 500                 filter->fo_subdir_count = FILTER_SUBDIR_COUNT;
 501         } else {
 502                 int retval = fsfilt_read_record(obd, filp, (char *)fsd,
 503                                                 sizeof(*fsd), &off);
 504                 if (retval != sizeof(*fsd)) {
 505                         CDEBUG(D_INODE,"OBD filter: error reading %s: rc %d\n",
 506                                LAST_RCVD, retval);
 507                         GOTO(err_fsd, rc = -EIO);
 508                 }
 509                 mount_count = le64_to_cpu(fsd->fsd_mount_count);
 510                 filter->fo_subdir_count = le16_to_cpu(fsd->fsd_subdir_count);
 511                 fsd->fsd_last_objid =
 512                         cpu_to_le64(le64_to_cpu(fsd->fsd_last_objid) +
 513                                     FILTER_SKIP_OBJID);
 514         }
 515
 516         if (fsd->fsd_feature_incompat) {
 517                 CERROR("unsupported feature %x\n",
 518                        le32_to_cpu(fsd->fsd_feature_incompat));
 519                 GOTO(err_fsd, rc = -EINVAL);
 520         }
 521         if (fsd->fsd_feature_rocompat) {
 522                 CERROR("read-only feature %x\n",
 523                        le32_to_cpu(fsd->fsd_feature_rocompat));
 524                 /* Do something like remount filesystem read-only */
 525                 GOTO(err_fsd, rc = -EINVAL);
 526         }
 527
 528         CDEBUG(D_INODE, "%s: server last_objid: "LPU64"\n",
 529                obd->obd_name, le64_to_cpu(fsd->fsd_last_objid));
 530         CDEBUG(D_INODE, "%s: server last_rcvd : "LPU64"\n",
 531                obd->obd_name, le64_to_cpu(fsd->fsd_last_transno));
 532         CDEBUG(D_INODE, "%s: server last_mount: "LPU64"\n",
 533                obd->obd_name, mount_count);
 534         CDEBUG(D_INODE, "%s: server data size: %u\n",
 535                obd->obd_name, le32_to_cpu(fsd->fsd_server_size));
 536         CDEBUG(D_INODE, "%s: per-client data start: %u\n",
 537                obd->obd_name, le32_to_cpu(fsd->fsd_client_start));
 538         CDEBUG(D_INODE, "%s: per-client data size: %u\n",
 539                obd->obd_name, le32_to_cpu(fsd->fsd_client_size));
 540         CDEBUG(D_INODE, "%s: server subdir_count: %u\n",
 541                obd->obd_name, le16_to_cpu(fsd->fsd_subdir_count));
 542
 543         if (!obd->obd_replayable) {
 544                 CWARN("%s: recovery support OFF\n", obd->obd_name);
 545                 GOTO(out, rc = 0);
 546         }
 547
 548         for (cl_idx = 0; off < last_rcvd_size; cl_idx++) {
 549                 __u64 last_rcvd;
 550                 int mount_age;
 551
 552                 if (!fcd) {
 553                         OBD_ALLOC(fcd, sizeof(*fcd));
 554                         if (!fcd)
 555                                 GOTO(err_fsd, rc = -ENOMEM);
 556                 }
 557
 558                 /* Don't assume off is incremented properly, in case
 559                  * sizeof(fsd) isn't the same as fsd->fsd_client_size.
 560                  */
 561                 off = le32_to_cpu(fsd->fsd_client_start) +
 562                         cl_idx * le16_to_cpu(fsd->fsd_client_size);
 563                 rc = fsfilt_read_record(obd, filp, (char *)fcd, sizeof(*fcd),
 564                                         &off);
 565                 if (rc != sizeof(*fcd)) {
 566                         CERROR("error reading FILTER %s offset %d: rc = %d\n",
 567                                LAST_RCVD, cl_idx, rc);
 568                         if (rc > 0) /* XXX fatal error or just abort reading? */
 569                                 rc = -EIO;
 570                         break;
 571                 }
 572
 573                 if (fcd->fcd_uuid[0] == '\0') {
 574                         CDEBUG(D_INFO, "skipping zeroed client at offset %d\n",
 575                                cl_idx);
 576                         continue;
 577                 }
 578
 579                 last_rcvd = le64_to_cpu(fcd->fcd_last_rcvd);
 580
 581                 /* These exports are cleaned up by filter_disconnect(), so they
 582                  * need to be set up like real exports as filter_connect() does.
 583                  */
 584                 mount_age = mount_count - le64_to_cpu(fcd->fcd_mount_count);
 585                 if (mount_age < FILTER_MOUNT_RECOV) {
 586                         struct obd_export *exp = class_new_export(obd);
 587                         struct filter_export_data *fed;
 588                         CERROR("RCVRNG CLIENT uuid: %s idx: %d lr: "LPU64
 589                                " srv lr: "LPU64" mnt: "LPU64" last mount: "
 590                                LPU64"\n", fcd->fcd_uuid, cl_idx,
 591                                last_rcvd, le64_to_cpu(fsd->fsd_last_transno),
 592                                le64_to_cpu(fcd->fcd_mount_count), mount_count);
 593                         if (exp == NULL) {
 594                                 /* XXX this rc is ignored  */
 595                                 rc = -ENOMEM;
 596                                 break;
 597                         }
 598                         memcpy(&exp->exp_client_uuid.uuid, fcd->fcd_uuid,
 599                                sizeof exp->exp_client_uuid.uuid);
 600                         fed = &exp->exp_filter_data;
 601                         fed->fed_fcd = fcd;
 602                         filter_client_add(obd, filter, fed, cl_idx);
 603                         /* create helper if export init gets more complex */
 604                         INIT_LIST_HEAD(&fed->fed_open_head);
 605                         spin_lock_init(&fed->fed_lock);
 606
 607                         fcd = NULL;
 608                         obd->obd_recoverable_clients++;
 609                         class_export_put(exp);
 610                 } else {
 611                         CDEBUG(D_INFO,
 612                                "discarded client %d UUID '%s' count "LPU64"\n",
 613                                cl_idx, fcd->fcd_uuid,
 614                                le64_to_cpu(fcd->fcd_mount_count));
 615                 }
 616
 617                 CDEBUG(D_OTHER, "client at idx %d has last_rcvd = "LPU64"\n",
 618                        cl_idx, last_rcvd);
 619
 620                 if (last_rcvd > le64_to_cpu(filter->fo_fsd->fsd_last_transno))
 621                         filter->fo_fsd->fsd_last_transno=cpu_to_le64(last_rcvd);
 622
 623                 obd->obd_last_committed =
 624                         le64_to_cpu(filter->fo_fsd->fsd_last_transno);
 625
 626                 if (obd->obd_recoverable_clients) {
 627                         CERROR("RECOVERY: %d recoverable clients, last_rcvd "
 628                                LPU64"\n", obd->obd_recoverable_clients,
 629                                le64_to_cpu(filter->fo_fsd->fsd_last_transno));
 630                         obd->obd_next_recovery_transno =
 631                                 obd->obd_last_committed + 1;
 632                         obd->obd_recovering = 1;
 633                 }
 634
 635         }
 636
 637         if (fcd)
 638                 OBD_FREE(fcd, sizeof(*fcd));
 639
 640 out:
 641         fsd->fsd_mount_count = cpu_to_le64(mount_count + 1);
 642
 643         /* save it, so mount count and last_transno is current */
 644         rc = filter_update_server_data(obd, filp, filter->fo_fsd);
 645
 646         RETURN(rc);
 647
 648 err_fsd:
 649         filter_free_server_data(filter);
 650         RETURN(rc);
 651 }
 652
 653 /* setup the object store with correct subdirectories */
 654 static int filter_prep(struct obd_device *obd)
 655 {
 656         struct obd_run_ctxt saved;
 657         struct filter_obd *filter = &obd->u.filter;
 658         struct dentry *dentry, *O_dentry;
 659         struct file *file;
 660         struct inode *inode;
 661         int i;
 662         int rc = 0;
 663         int mode = 0;
 664
 665         push_ctxt(&saved, &filter->fo_ctxt, NULL);
 666         dentry = simple_mkdir(current->fs->pwd, "O", 0700);
 667         CDEBUG(D_INODE, "got/created O: %p\n", dentry);
 668         if (IS_ERR(dentry)) {
 669                 rc = PTR_ERR(dentry);
 670                 CERROR("cannot open/create O: rc = %d\n", rc);
 671                 GOTO(out, rc);
 672         }
 673         filter->fo_dentry_O = dentry;
 674
 675         /*
 676          * Create directories and/or get dentries for each object type.
 677          * This saves us from having to do multiple lookups for each one.
 678          */
 679         O_dentry = filter->fo_dentry_O;
 680         for (mode = 0; mode < (S_IFMT >> S_SHIFT); mode++) {
 681                 char *name = obd_type_by_mode[mode];
 682
 683                 if (!name) {
 684                         filter->fo_dentry_O_mode[mode] = NULL;
 685                         continue;
 686                 }
 687                 dentry = simple_mkdir(O_dentry, name, 0700);
 688                 CDEBUG(D_INODE, "got/created O/%s: %p\n", name, dentry);
 689                 if (IS_ERR(dentry)) {
 690                         rc = PTR_ERR(dentry);
 691                         CERROR("cannot create O/%s: rc = %d\n", name, rc);
 692                         GOTO(err_O_mode, rc);
 693                 }
 694                 filter->fo_dentry_O_mode[mode] = dentry;
 695         }
 696
 697         file = filp_open(LAST_RCVD, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 698         if (!file || IS_ERR(file)) {
 699                 rc = PTR_ERR(file);
 700                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 701                        LAST_RCVD, rc);
 702                 GOTO(err_O_mode, rc);
 703         }
 704
 705         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 706                 CERROR("%s is not a regular file!: mode = %o\n", LAST_RCVD,
 707                        file->f_dentry->d_inode->i_mode);
 708                 GOTO(err_filp, rc = -ENOENT);
 709         }
 710
 711         rc = fsfilt_journal_data(obd, file);
 712         if (rc) {
 713                 CERROR("cannot journal data on %s: rc = %d\n", LAST_RCVD, rc);
 714                 GOTO(err_filp, rc);
 715         }
 716         /* steal operations */
 717         inode = file->f_dentry->d_inode;
 718         filter->fo_fop = file->f_op;
 719         filter->fo_iop = inode->i_op;
 720         filter->fo_aops = inode->i_mapping->a_ops;
 721 #ifdef I_SKIP_PDFLUSH
 722         /*
 723          * we need this to protect from deadlock
 724          * pdflush vs. lustre_fwrite()
 725          */
 726         inode->i_flags |= I_SKIP_PDFLUSH;
 727 #endif
 728
 729         rc = filter_init_server_data(obd, file, FILTER_INIT_OBJID);
 730         if (rc) {
 731                 CERROR("cannot read %s: rc = %d\n", LAST_RCVD, rc);
 732                 GOTO(err_client, rc);
 733         }
 734         filter->fo_rcvd_filp = file;
 735
 736         if (filter->fo_subdir_count) {
 737                 O_dentry = filter->fo_dentry_O_mode[S_IFREG >> S_SHIFT];
 738                 OBD_ALLOC(filter->fo_dentry_O_sub,
 739                           filter->fo_subdir_count * sizeof(dentry));
 740                 if (!filter->fo_dentry_O_sub)
 741                         GOTO(err_client, rc = -ENOMEM);
 742
 743                 for (i = 0; i < filter->fo_subdir_count; i++) {
 744                         char dir[20];
 745                         snprintf(dir, sizeof(dir), "d%u", i);
 746
 747                         dentry = simple_mkdir(O_dentry, dir, 0700);
 748                         CDEBUG(D_INODE, "got/created O/R/%s: %p\n", dir,dentry);
 749                         if (IS_ERR(dentry)) {
 750                                 rc = PTR_ERR(dentry);
 751                                 CERROR("can't create O/R/%s: rc = %d\n",dir,rc);
 752                                 GOTO(err_O_sub, rc);
 753                         }
 754                         filter->fo_dentry_O_sub[i] = dentry;
 755                 }
 756         }
 757         rc = 0;
 758  out:
 759         pop_ctxt(&saved, &filter->fo_ctxt, NULL);
 760
 761         return(rc);
 762
 763 err_O_sub:
 764         while (i-- > 0) {
 765                 struct dentry *dentry = filter->fo_dentry_O_sub[i];
 766                 if (dentry) {
 767                         f_dput(dentry);
 768                         filter->fo_dentry_O_sub[i] = NULL;
 769                 }
 770         }
 771         OBD_FREE(filter->fo_dentry_O_sub,
 772                  filter->fo_subdir_count * sizeof(dentry));
 773 err_client:
 774         class_disconnect_exports(obd, 0);
 775 err_filp:
 776         if (filp_close(file, 0))
 777                 CERROR("can't close %s after error\n", LAST_RCVD);
 778         filter->fo_rcvd_filp = NULL;
 779 err_O_mode:
 780         while (mode-- > 0) {
 781                 struct dentry *dentry = filter->fo_dentry_O_mode[mode];
 782                 if (dentry) {
 783                         f_dput(dentry);
 784                         filter->fo_dentry_O_mode[mode] = NULL;
 785                 }
 786         }
 787         f_dput(filter->fo_dentry_O);
 788         filter->fo_dentry_O = NULL;
 789         goto out;
 790 }
 791
 792 /* cleanup the filter: write last used object id to status file */
 793 static void filter_post(struct obd_device *obd)
 794 {
 795         struct obd_run_ctxt saved;
 796         struct filter_obd *filter = &obd->u.filter;
 797         long rc;
 798         int mode;
 799
 800         /* XXX: filter_update_lastobjid used to call fsync_dev.  It might be
 801          * best to start a transaction with h_sync, because we removed this
 802          * from lastobjid */
 803
 804         push_ctxt(&saved, &filter->fo_ctxt, NULL);
 805         rc = filter_update_server_data(obd, filter->fo_rcvd_filp,
 806                                        filter->fo_fsd);
 807         if (rc)
 808                 CERROR("error writing lastobjid: rc = %ld\n", rc);
 809
 810
 811         if (filter->fo_rcvd_filp) {
 812                 rc = file_fsync(filter->fo_rcvd_filp,
 813                                 filter->fo_rcvd_filp->f_dentry, 1);
 814                 filp_close(filter->fo_rcvd_filp, 0);
 815                 filter->fo_rcvd_filp = NULL;
 816                 if (rc)
 817                         CERROR("error closing %s: rc = %ld\n", LAST_RCVD, rc);
 818         }
 819
 820         if (filter->fo_subdir_count) {
 821                 int i;
 822                 for (i = 0; i < filter->fo_subdir_count; i++) {
 823                         struct dentry *dentry = filter->fo_dentry_O_sub[i];
 824                         f_dput(dentry);
 825                         filter->fo_dentry_O_sub[i] = NULL;
 826                 }
 827                 OBD_FREE(filter->fo_dentry_O_sub,
 828                          filter->fo_subdir_count *
 829                          sizeof(*filter->fo_dentry_O_sub));
 830         }
 831         for (mode = 0; mode < (S_IFMT >> S_SHIFT); mode++) {
 832                 struct dentry *dentry = filter->fo_dentry_O_mode[mode];
 833                 if (dentry) {
 834                         f_dput(dentry);
 835                         filter->fo_dentry_O_mode[mode] = NULL;
 836                 }
 837         }
 838         f_dput(filter->fo_dentry_O);
 839         filter_free_server_data(filter);
 840         pop_ctxt(&saved, &filter->fo_ctxt, NULL);
 841 }
 842
 843 __u64 filter_next_id(struct filter_obd *filter)
 844 {
 845         obd_id id;
 846         LASSERT(filter->fo_fsd != NULL);
 847
 848         spin_lock(&filter->fo_objidlock);
 849         id = le64_to_cpu(filter->fo_fsd->fsd_last_objid);
 850         filter->fo_fsd->fsd_last_objid = cpu_to_le64(id + 1);
 851         spin_unlock(&filter->fo_objidlock);
 852
 853         return id;
 854 }
 855
 856 /* direct cut-n-paste of mds_blocking_ast() */
 857 static int filter_blocking_ast(struct ldlm_lock *lock,
 858                                struct ldlm_lock_desc *desc,
 859                                void *data, int flag)
 860 {
 861         int do_ast;
 862         ENTRY;
 863
 864         if (flag == LDLM_CB_CANCELING) {
 865                 /* Don't need to do anything here. */
 866                 RETURN(0);
 867         }
 868
 869         /* XXX layering violation!  -phil */
 870         l_lock(&lock->l_resource->lr_namespace->ns_lock);
 871         /* Get this: if filter_blocking_ast is racing with ldlm_intent_policy,
 872          * such that mds_blocking_ast is called just before l_i_p takes the
 873          * ns_lock, then by the time we get the lock, we might not be the
 874          * correct blocking function anymore.  So check, and return early, if
 875          * so. */
 876         if (lock->l_blocking_ast != filter_blocking_ast) {
 877                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 878                 RETURN(0);
 879         }
 880
 881         lock->l_flags |= LDLM_FL_CBPENDING;
 882         do_ast = (!lock->l_readers && !lock->l_writers);
 883         l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 884
 885         if (do_ast) {
 886                 struct lustre_handle lockh;
 887                 int rc;
 888
 889                 LDLM_DEBUG(lock, "already unused, calling ldlm_cli_cancel");
 890                 ldlm_lock2handle(lock, &lockh);
 891                 rc = ldlm_cli_cancel(&lockh);
 892                 if (rc < 0)
 893                         CERROR("ldlm_cli_cancel: %d\n", rc);
 894         } else {
 895                 LDLM_DEBUG(lock, "Lock still has references, will be "
 896                            "cancelled later");
 897         }
 898         RETURN(0);
 899 }
 900
 901 static int filter_lock_dentry(struct obd_device *obd, struct dentry *de,
 902                               ldlm_mode_t lock_mode,struct lustre_handle *lockh)
 903 {
 904         struct ldlm_res_id res_id = { .name = {0} };
 905         int flags = 0, rc;
 906         ENTRY;
 907
 908         res_id.name[0] = de->d_inode->i_ino;
 909         res_id.name[1] = de->d_inode->i_generation;
 910         rc = ldlm_cli_enqueue(NULL, NULL, obd->obd_namespace, NULL,
 911                               res_id, LDLM_PLAIN, NULL, 0, lock_mode,
 912                               &flags, ldlm_completion_ast,
 913                               filter_blocking_ast, NULL, lockh);
 914
 915         RETURN(rc == ELDLM_OK ? 0 : -ENOLCK);  /* XXX translate ldlm code */
 916 }
 917
 918 /* We never dget the object parent, so DON'T dput it either */
 919 static void filter_parent_unlock(struct dentry *dparent,
 920                                  struct lustre_handle *lockh,
 921                                  ldlm_mode_t lock_mode)
 922 {
 923         ldlm_lock_decref(lockh, lock_mode);
 924 }
 925
 926 /* We never dget the object parent, so DON'T dput it either */
 927 struct dentry *filter_parent(struct obd_device *obd, obd_mode mode,
 928                              obd_id objid)
 929 {
 930         struct filter_obd *filter = &obd->u.filter;
 931
 932         LASSERT(S_ISREG(mode));   /* only regular files for now */
 933         if (!S_ISREG(mode) || filter->fo_subdir_count == 0)
 934                 return filter->fo_dentry_O_mode[(mode & S_IFMT) >> S_SHIFT];
 935
 936         return filter->fo_dentry_O_sub[objid & (filter->fo_subdir_count - 1)];
 937 }
 938
 939 /* We never dget the object parent, so DON'T dput it either */
 940 struct dentry *filter_parent_lock(struct obd_device *obd, obd_mode mode,
 941                                   obd_id objid, ldlm_mode_t lock_mode,
 942                                   struct lustre_handle *lockh)
 943 {
 944         unsigned long now = jiffies;
 945         struct dentry *de = filter_parent(obd, mode, objid);
 946         int rc;
 947
 948         if (IS_ERR(de))
 949                 return de;
 950
 951         rc = filter_lock_dentry(obd, de, lock_mode, lockh);
 952         if (time_after(jiffies, now + 15 * HZ))
 953                 CERROR("slow parent lock %lus\n", (jiffies - now) / HZ);
 954         return rc ? ERR_PTR(rc) : de;
 955 }
 956
 957 /* How to get files, dentries, inodes from object id's.
 958  *
 959  * If dir_dentry is passed, the caller has already locked the parent
 960  * appropriately for this operation (normally a write lock).  If
 961  * dir_dentry is NULL, we do a read lock while we do the lookup to
 962  * avoid races with create/destroy and such changing the directory
 963  * internal to the filesystem code. */
 964 struct dentry *filter_fid2dentry(struct obd_device *obd,
 965                                  struct dentry *dir_dentry,
 966                                  obd_mode mode, obd_id id)
 967 {
 968         struct lustre_handle lockh;
 969         struct dentry *dparent = dir_dentry;
 970         struct dentry *dchild;
 971         char name[32];
 972         int len;
 973         ENTRY;
 974
 975         if (id == 0) {
 976                 CERROR("fatal: invalid object id 0\n");
 977                 LBUG();
 978                 RETURN(ERR_PTR(-ESTALE));
 979         }
 980
 981         len = sprintf(name, LPU64, id);
 982         if (dir_dentry == NULL) {
 983                 dparent = filter_parent_lock(obd, mode, id, LCK_PR, &lockh);
 984                 if (IS_ERR(dparent))
 985                         RETURN(dparent);
 986         }
 987         CDEBUG(D_INODE, "looking up object O/%*s/%s\n",
 988                dparent->d_name.len, dparent->d_name.name, name);
 989         dchild = ll_lookup_one_len(name, dparent, len);
 990         if (dir_dentry == NULL)
 991                 filter_parent_unlock(dparent, &lockh, LCK_PR);
 992         if (IS_ERR(dchild)) {
 993                 CERROR("child lookup error %ld\n", PTR_ERR(dchild));
 994                 RETURN(dchild);
 995         }
 996
 997         CDEBUG(D_INODE, "got child objid %s: %p, count = %d\n",
 998                name, dchild, atomic_read(&dchild->d_count));
 999
1000         LASSERT(atomic_read(&dchild->d_count) > 0);
1001
1002         RETURN(dchild);
1003 }
1004
1005 static struct file *filter_obj_open(struct obd_export *export,
1006                                     struct obd_trans_info *oti,
1007                                     __u64 id, __u32 type, int parent_mode,
1008                                     struct lustre_handle *parent_lockh)
1009 {
1010         struct obd_device *obd = export->exp_obd;
1011         struct filter_obd *filter = &obd->u.filter;
1012         struct dentry *dchild = NULL, *dparent = NULL;
1013         struct filter_export_data *fed = &export->exp_filter_data;
1014         struct filter_dentry_data *fdd = NULL;
1015         struct filter_file_data *ffd = NULL;
1016         struct obd_run_ctxt saved;
1017         char name[24];
1018         struct file *file;
1019         int len, cleanup_phase = 0;
1020         ENTRY;
1021
1022         push_ctxt(&saved, &filter->fo_ctxt, NULL);
1023
1024         if (!id) {
1025                 CERROR("fatal: invalid obdo "LPU64"\n", id);
1026                 GOTO(cleanup, file = ERR_PTR(-ESTALE));
1027         }
1028
1029         if (!(type & S_IFMT)) {
1030                 CERROR("OBD %s, object "LPU64" has bad type: %o\n",
1031                        __FUNCTION__, id, type);
1032                 GOTO(cleanup, file = ERR_PTR(-EINVAL));
1033         }
1034
1035         ffd = filter_ffd_new();
1036         if (ffd == NULL) {
1037                 CERROR("obdfilter: out of memory\n");
1038                 GOTO(cleanup, file = ERR_PTR(-ENOMEM));
1039         }
1040
1041         cleanup_phase = 1;
1042
1043         /* We preallocate this to avoid blocking while holding fo_fddlock */
1044         OBD_ALLOC(fdd, sizeof *fdd);
1045         if (fdd == NULL) {
1046                 CERROR("obdfilter: out of memory\n");
1047                 GOTO(cleanup, file = ERR_PTR(-ENOMEM));
1048         }
1049
1050         cleanup_phase = 2;
1051
1052         dparent = filter_parent_lock(obd, type, id, parent_mode, parent_lockh);
1053         if (IS_ERR(dparent))
1054                 GOTO(cleanup, file = (void *)dparent);
1055
1056         cleanup_phase = 3;
1057
1058         len = snprintf(name, sizeof(name), LPU64, id);
1059         dchild = ll_lookup_one_len(name, dparent, len);
1060         if (IS_ERR(dchild))
1061                 GOTO(cleanup, file = (void *)dchild);
1062
1063         cleanup_phase = 4;
1064
1065         if (dchild->d_inode == NULL) {
1066                 CERROR("opening non-existent object %s - O_CREAT?\n", name);
1067                 /* dput(dchild); call filter_create_internal here */
1068                 file = ERR_PTR(-ENOENT);
1069                 GOTO(cleanup, file);
1070         }
1071
1072         /* dentry_open does a dput(dchild) and mntput(mnt) on error */
1073         mntget(filter->fo_vfsmnt);
1074         file = dentry_open(dchild, filter->fo_vfsmnt, O_RDWR | O_LARGEFILE);
1075         if (IS_ERR(file)) {
1076                 dchild = NULL; /* prevent a double dput in step 4 */
1077                 CERROR("error opening %s: rc %ld\n", name, PTR_ERR(file));
1078                 GOTO(cleanup, file);
1079         }
1080
1081         spin_lock(&filter->fo_fddlock);
1082         if (dchild->d_fsdata) {
1083                 spin_unlock(&filter->fo_fddlock);
1084                 OBD_FREE(fdd, sizeof *fdd);
1085                 fdd = dchild->d_fsdata;
1086                 LASSERT(fdd->fdd_magic == FILTER_DENTRY_MAGIC);
1087                 /* should only happen during client recovery */
1088                 if (fdd->fdd_flags & FILTER_FLAG_DESTROY)
1089                         CDEBUG(D_INODE,"opening destroyed object "LPU64"\n",id);
1090                 atomic_inc(&fdd->fdd_open_count);
1091         } else {
1092                 atomic_set(&fdd->fdd_open_count, 1);
1093                 fdd->fdd_magic = FILTER_DENTRY_MAGIC;
1094                 fdd->fdd_flags = 0;
1095                 fdd->fdd_objid = id;
1096                 /* If this is racy, then we can use {cmp}xchg and atomic_add */
1097                 dchild->d_fsdata = fdd;
1098                 spin_unlock(&filter->fo_fddlock);
1099         }
1100
1101         ffd->ffd_file = file;
1102         LASSERT(file->private_data == NULL);
1103         file->private_data = ffd;
1104
1105         if (!dchild->d_op)
1106                 dchild->d_op = &filter_dops;
1107         else
1108                 LASSERT(dchild->d_op == &filter_dops);
1109
1110         spin_lock(&fed->fed_lock);
1111         list_add(&ffd->ffd_export_list, &fed->fed_open_head);
1112         spin_unlock(&fed->fed_lock);
1113
1114         CDEBUG(D_INODE, "opened objid "LPU64": rc = %p\n", id, file);
1115 cleanup:
1116         switch (cleanup_phase) {
1117         case 4:
1118                 if (IS_ERR(file))
1119                         f_dput(dchild);
1120         case 3:
1121                 if (IS_ERR(file))
1122                         filter_parent_unlock(dparent, parent_lockh,parent_mode);
1123         case 2:
1124                 if (IS_ERR(file))
1125                         OBD_FREE(fdd, sizeof *fdd);
1126         case 1:
1127                 if (IS_ERR(file))
1128                         filter_ffd_destroy(ffd);
1129                 filter_ffd_put(ffd);
1130         case 0:
1131                 pop_ctxt(&saved, &filter->fo_ctxt, NULL);
1132         }
1133         RETURN(file);
1134 }
1135
1136 /* Caller must hold LCK_PW on parent and push us into kernel context.
1137  * Caller is also required to ensure that dchild->d_inode exists. */
1138 static int filter_destroy_internal(struct obd_device *obd, obd_id objid,
1139                                    struct dentry *dparent,
1140                                    struct dentry *dchild)
1141 {
1142         struct inode *inode = dchild->d_inode;
1143         int rc;
1144         ENTRY;
1145
1146         if (inode->i_nlink != 1 || atomic_read(&inode->i_count) != 1) {
1147                 CERROR("destroying objid %*s nlink = %d, count = %d\n",
1148                        dchild->d_name.len, dchild->d_name.name,
1149                        inode->i_nlink, atomic_read(&inode->i_count));
1150         }
1151
1152
1153 #if 0
1154         /* Tell the clients that the object is gone now and that they should
1155          * throw away any cached pages.  We don't need to wait until they're
1156          * done, so just decref the lock right away and let ldlm_completion_ast
1157          * clean up when it's all over. */
1158         ldlm_cli_enqueue(..., LCK_PW, AST_INTENT_DESTROY, &lockh);
1159         ldlm_lock_decref(&lockh, LCK_PW);
1160 #endif
1161
1162         if (0) {
1163                 struct lustre_handle lockh;
1164                 int flags = 0, rc;
1165                 struct ldlm_res_id res_id = { .name = { objid } };
1166
1167                 /* This part is a wee bit iffy: we really only want to bust the
1168                  * locks on our stripe, so that we don't end up bouncing
1169                  * [0->EOF] locks around on each of the OSTs as the rest of the
1170                  * destroys get processed.  Because we're only talking to
1171                  * the local LDLM, though, we should only end up locking the
1172                  * whole of our stripe.  When bug 1425 (take all locks on OST
1173                  * for stripe 0) is fixed, this code should be revisited. */
1174                 struct ldlm_extent extent = { 0, OBD_OBJECT_EOF };
1175
1176                 rc = ldlm_cli_enqueue(NULL, NULL, obd->obd_namespace, NULL,
1177                                       res_id, LDLM_EXTENT, &extent,
1178                                       sizeof(extent), LCK_PW, &flags,
1179                                       ldlm_completion_ast, filter_blocking_ast,
1180                                       NULL, &lockh);
1181                 /* We only care about the side-effects, just drop the lock. */
1182                 ldlm_lock_decref(&lockh, LCK_PW);
1183         }
1184
1185         rc = vfs_unlink(dparent->d_inode, dchild);
1186
1187         if (rc)
1188                 CERROR("error unlinking objid %*s: rc %d\n",
1189                        dchild->d_name.len, dchild->d_name.name, rc);
1190
1191         RETURN(rc);
1192 }
1193
1194 /* If closing because we are failing this device, then
1195    don't do the unlink on close.
1196 */
1197 static int filter_close_internal(struct obd_export *exp,
1198                                  struct filter_file_data *ffd,
1199                                  struct obd_trans_info *oti, int flags)
1200 {
1201         struct obd_device *obd = exp->exp_obd;
1202         struct filter_obd *filter = &obd->u.filter;
1203         struct file *filp = ffd->ffd_file;
1204         struct dentry *dchild = dget(filp->f_dentry);
1205         struct filter_dentry_data *fdd = dchild->d_fsdata;
1206         struct lustre_handle parent_lockh;
1207         int rc, rc2, cleanup_phase = 0;
1208         struct dentry *dparent = NULL;
1209         struct obd_run_ctxt saved;
1210         ENTRY;
1211
1212         LASSERT(filp->private_data == ffd);
1213         LASSERT(fdd != NULL);
1214         LASSERT(fdd->fdd_magic == FILTER_DENTRY_MAGIC);
1215
1216         rc = filp_close(filp, 0);
1217
1218         if (atomic_dec_and_test(&fdd->fdd_open_count) &&
1219             (fdd->fdd_flags & FILTER_FLAG_DESTROY) &&
1220             !(flags & OBD_OPT_FAILOVER)) {
1221                 void *handle;
1222
1223                 push_ctxt(&saved, &filter->fo_ctxt, NULL);
1224                 cleanup_phase = 1;
1225
1226                 LASSERT(fdd->fdd_objid > 0);
1227                 dparent = filter_parent_lock(obd, S_IFREG, fdd->fdd_objid,
1228                                              LCK_PW, &parent_lockh);
1229                 if (IS_ERR(dparent))
1230                         GOTO(cleanup, rc = PTR_ERR(dparent));
1231                 cleanup_phase = 2;
1232
1233                 handle = fsfilt_start(obd, dparent->d_inode,
1234                                       FSFILT_OP_UNLINK_LOG, oti);
1235                 if (IS_ERR(handle))
1236                         GOTO(cleanup, rc = PTR_ERR(handle));
1237
1238                 if (oti != NULL) {
1239                         if (oti->oti_handle == NULL)
1240                                 oti->oti_handle = handle;
1241                         else
1242                                 LASSERT(oti->oti_handle == handle);
1243                 }
1244
1245 #ifdef ENABLE_ORPHANS
1246                 /* Remove orphan unlink record from log */
1247                 llog_cancel_records(filter->fo_catalog, 1, &fdd->fdd_cookie);
1248 #endif
1249                 /* XXX unlink from PENDING directory now too */
1250                 rc2 = filter_destroy_internal(obd, fdd->fdd_objid, dparent,
1251                                               dchild);
1252                 if (rc2 && !rc)
1253                         rc = rc2;
1254                 rc = filter_finish_transno(exp, oti, rc);
1255                 rc2 = fsfilt_commit(obd, dparent->d_inode, handle, 0);
1256                 if (rc2) {
1257                         CERROR("error on commit, err = %d\n", rc2);
1258                         if (!rc)
1259                                 rc = rc2;
1260                 }
1261         }
1262
1263 cleanup:
1264         switch(cleanup_phase) {
1265         case 2:
1266                 if (rc || oti == NULL) {
1267                         filter_parent_unlock(dparent, &parent_lockh, LCK_PW);
1268                 } else {
1269                         memcpy(&oti->oti_ack_locks[0].lock, &parent_lockh,
1270                                sizeof(parent_lockh));
1271                         oti->oti_ack_locks[0].mode = LCK_PW;
1272                 }
1273         case 1:
1274                 pop_ctxt(&saved, &filter->fo_ctxt, NULL);
1275         case 0:
1276                 f_dput(dchild);
1277                 filter_ffd_destroy(ffd);
1278                 break;
1279         default:
1280                 CERROR("invalid cleanup_phase %d\n", cleanup_phase);
1281                 LBUG();
1282         }
1283
1284         RETURN(rc);
1285 }
1286
1287 /* mount the file system (secretly) */
1288 int filter_common_setup(struct obd_device *obd, obd_count len, void *buf,
1289                         char *option)
1290 {
1291         struct obd_ioctl_data* data = buf;
1292         struct filter_obd *filter = &obd->u.filter;
1293         struct vfsmount *mnt;
1294         int rc = 0;
1295         ENTRY;
1296
1297         if (!data->ioc_inlbuf1 || !data->ioc_inlbuf2)
1298                 RETURN(-EINVAL);
1299
1300         obd->obd_fsops = fsfilt_get_ops(data->ioc_inlbuf2);
1301         if (IS_ERR(obd->obd_fsops))
1302                 RETURN(PTR_ERR(obd->obd_fsops));
1303
1304         mnt = do_kern_mount(data->ioc_inlbuf2, MS_NOATIME | MS_NODIRATIME,
1305                             data->ioc_inlbuf1, option);
1306         rc = PTR_ERR(mnt);
1307         if (IS_ERR(mnt))
1308                 GOTO(err_ops, rc);
1309
1310         if (data->ioc_inllen3 > 0 && data->ioc_inlbuf3) {
1311                 if (*data->ioc_inlbuf3 == 'f') {
1312                         obd->obd_replayable = 1;
1313                         obd_sync_filter = 1;
1314                         CERROR("%s: configured for recovery and sync write\n",
1315                                obd->obd_name);
1316                 } else {
1317                         if (*data->ioc_inlbuf3 != 'n') {
1318                                 CERROR("unrecognised flag '%c'\n",
1319                                        *data->ioc_inlbuf3);
1320                         }
1321                 }
1322         }
1323
1324         if (data->ioc_inllen4 > 0 && data->ioc_inlbuf4) {
1325                 if (*data->ioc_inlbuf4 == '/') {
1326                         CERROR("filter namespace mount: %s\n",
1327                                data->ioc_inlbuf4);
1328                         filter->fo_nspath = strdup(data->ioc_inlbuf4);
1329                 } else {
1330                         CERROR("namespace mount must be absolute path: '%s'\n",
1331                                data->ioc_inlbuf4);
1332                 }
1333         }
1334
1335         filter->fo_vfsmnt = mnt;
1336         filter->fo_sb = mnt->mnt_sb;
1337         filter->fo_fstype = mnt->mnt_sb->s_type->name;
1338         CDEBUG(D_SUPER, "%s: mnt = %p\n", filter->fo_fstype, mnt);
1339
1340         OBD_SET_CTXT_MAGIC(&filter->fo_ctxt);
1341         filter->fo_ctxt.pwdmnt = mnt;
1342         filter->fo_ctxt.pwd = mnt->mnt_root;
1343         filter->fo_ctxt.fs = get_ds();
1344
1345         rc = filter_prep(obd);
1346         if (rc)
1347                 GOTO(err_mntput, rc);
1348
1349         spin_lock_init(&filter->fo_translock);
1350         spin_lock_init(&filter->fo_fddlock);
1351         spin_lock_init(&filter->fo_objidlock);
1352         INIT_LIST_HEAD(&filter->fo_export_list);
1353
1354         ptlrpc_init_client(MDS_REQUEST_PORTAL, MDC_REPLY_PORTAL,
1355                            "filter_mdc", &filter->fo_mdc_client);
1356         sema_init(&filter->fo_sem, 1);
1357
1358         obd->obd_namespace = ldlm_namespace_new("filter-tgt",
1359                                                 LDLM_NAMESPACE_SERVER);
1360         if (obd->obd_namespace == NULL)
1361                 GOTO(err_post, rc = -ENOMEM);
1362
1363         ptlrpc_init_client(LDLM_CB_REQUEST_PORTAL, LDLM_CB_REPLY_PORTAL,
1364                            "filter_ldlm_cb_client", &obd->obd_ldlm_client);
1365
1366         /* Create a non-replaying connection for recovery logging, so that
1367          * we don't create a client entry for this local connection, and do
1368          * not log or assign transaction numbers for logging operations. */
1369 #ifdef ENABLE_ORPHANS
1370         filter->fo_catalog = filter_get_catalog(obd);
1371         if (IS_ERR(filter->fo_catalog))
1372                 GOTO(err_post, rc = PTR_ERR(filter->fo_catalog));
1373 #endif
1374
1375         RETURN(0);
1376
1377 err_post:
1378         filter_post(obd);
1379 err_mntput:
1380         unlock_kernel();
1381         mntput(mnt);
1382         filter->fo_sb = 0;
1383         lock_kernel();
1384 err_ops:
1385         fsfilt_put_ops(obd->obd_fsops);
1386         return rc;
1387 }
1388
1389 static int filter_setup(struct obd_device *obd, obd_count len, void *buf)
1390 {
1391         struct obd_ioctl_data* data = buf;
1392         char *option = NULL;
1393
1394 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
1395         /* bug 1577: implement async-delete for 2.5 */
1396         if (!strcmp(data->ioc_inlbuf2, "ext3"))
1397                 option = "asyncdel";
1398 #endif
1399
1400         return filter_common_setup(obd, len, buf, option);
1401 }
1402
1403 static int filter_cleanup(struct obd_device *obd, int flags)
1404 {
1405         struct filter_obd *filter = &obd->u.filter;
1406         ENTRY;
1407
1408         if (flags & OBD_OPT_FAILOVER)
1409                 CERROR("%s: shutting down for failover; client state will"
1410                        " be preserved.\n", obd->obd_name);
1411
1412         if (!list_empty(&obd->obd_exports)) {
1413                 CERROR("%s: still has clients!\n", obd->obd_name);
1414                 class_disconnect_exports(obd, flags);
1415                 if (!list_empty(&obd->obd_exports)) {
1416                         CERROR("still has exports after forced cleanup?\n");
1417                         RETURN(-EBUSY);
1418                 }
1419         }
1420
1421 #ifdef ENABLE_ORPHANS
1422         filter_put_catalog(filter->fo_catalog);
1423 #endif
1424
1425         ldlm_namespace_free(obd->obd_namespace);
1426
1427         if (filter->fo_sb == NULL)
1428                 RETURN(0);
1429
1430         filter_post(obd);
1431
1432         shrink_dcache_parent(filter->fo_sb->s_root);
1433         filter->fo_sb = 0;
1434
1435         if (atomic_read(&filter->fo_vfsmnt->mnt_count) > 1)
1436                 CERROR("%s: mount point busy, mnt_count: %d\n", obd->obd_name,
1437                        atomic_read(&filter->fo_vfsmnt->mnt_count));
1438
1439         unlock_kernel();
1440         mntput(filter->fo_vfsmnt);
1441         //destroy_buffers(filter->fo_sb->s_dev);
1442         filter->fo_sb = NULL;
1443         fsfilt_put_ops(obd->obd_fsops);
1444         lock_kernel();
1445
1446         RETURN(0);
1447 }
1448
1449 static int filter_attach(struct obd_device *obd, obd_count len, void *data)
1450 {
1451         struct lprocfs_static_vars lvars;
1452         int rc;
1453
1454         lprocfs_init_vars(filter, &lvars);
1455         rc = lprocfs_obd_attach(obd, lvars.obd_vars);
1456         if (rc != 0)
1457                 return rc;
1458
1459         rc = lprocfs_alloc_obd_stats(obd, LPROC_FILTER_LAST);
1460         if (rc != 0)
1461                 return rc;
1462
1463         /* Init obdfilter private stats here */
1464         lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_READ_BYTES,
1465                              LPROCFS_CNTR_AVGMINMAX, "read_bytes", "bytes");
1466         lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_WRITE_BYTES,
1467                              LPROCFS_CNTR_AVGMINMAX, "write_bytes", "bytes");
1468         return rc;
1469 }
1470
1471 static int filter_detach(struct obd_device *dev)
1472 {
1473         lprocfs_free_obd_stats(dev);
1474         return lprocfs_obd_detach(dev);
1475 }
1476
1477 /* nearly identical to mds_connect */
1478 static int filter_connect(struct lustre_handle *conn, struct obd_device *obd,
1479                           struct obd_uuid *cluuid)
1480 {
1481         struct obd_export *exp;
1482         struct filter_export_data *fed;
1483         struct filter_client_data *fcd;
1484         struct filter_obd *filter = &obd->u.filter;
1485         int rc;
1486         ENTRY;
1487
1488         if (conn == NULL || obd == NULL || cluuid == NULL)
1489                 RETURN(-EINVAL);
1490
1491         rc = class_connect(conn, obd, cluuid);
1492         if (rc)
1493                 RETURN(rc);
1494         exp = class_conn2export(conn);
1495         LASSERT(exp != NULL);
1496
1497         fed = &exp->exp_filter_data;
1498         class_export_put(exp);
1499
1500         INIT_LIST_HEAD(&fed->fed_open_head);
1501         spin_lock_init(&fed->fed_lock);
1502
1503         if (!obd->obd_replayable)
1504                 RETURN(0);
1505
1506         OBD_ALLOC(fcd, sizeof(*fcd));
1507         if (!fcd) {
1508                 CERROR("filter: out of memory for client data\n");
1509                 GOTO(out_export, rc = -ENOMEM);
1510         }
1511
1512         memcpy(fcd->fcd_uuid, cluuid, sizeof(fcd->fcd_uuid));
1513         fed->fed_fcd = fcd;
1514         fcd->fcd_mount_count = cpu_to_le64(filter->fo_fsd->fsd_mount_count);
1515
1516         rc = filter_client_add(obd, filter, fed, -1);
1517         if (rc)
1518                 GOTO(out_fcd, rc);
1519
1520         RETURN(rc);
1521
1522 out_fcd:
1523         OBD_FREE(fcd, sizeof(*fcd));
1524 out_export:
1525         class_disconnect(conn, 0);
1526
1527         RETURN(rc);
1528 }
1529
1530 static void filter_destroy_export(struct obd_export *exp)
1531 {
1532         struct filter_export_data *fed = &exp->exp_filter_data;
1533
1534         ENTRY;
1535         spin_lock(&fed->fed_lock);
1536         while (!list_empty(&fed->fed_open_head)) {
1537                 struct filter_file_data *ffd;
1538
1539                 ffd = list_entry(fed->fed_open_head.next, typeof(*ffd),
1540                                  ffd_export_list);
1541                 list_del(&ffd->ffd_export_list);
1542                 spin_unlock(&fed->fed_lock);
1543
1544                 CDEBUG(D_INFO, "force close file %*s (hdl %p:"LPX64") on "
1545                        "disconnect\n", ffd->ffd_file->f_dentry->d_name.len,
1546                        ffd->ffd_file->f_dentry->d_name.name,
1547                        ffd, ffd->ffd_handle.h_cookie);
1548
1549                 filter_close_internal(exp, ffd, NULL, exp->exp_flags);
1550                 spin_lock(&fed->fed_lock);
1551         }
1552         spin_unlock(&fed->fed_lock);
1553
1554         if (exp->exp_obd->obd_replayable)
1555                 filter_client_free(exp, exp->exp_flags);
1556         EXIT;
1557 }
1558
1559 /* also incredibly similar to mds_disconnect */
1560 static int filter_disconnect(struct lustre_handle *conn, int flags)
1561 {
1562         struct obd_export *exp = class_conn2export(conn);
1563         unsigned long irqflags;
1564         int rc;
1565         ENTRY;
1566
1567         LASSERT(exp);
1568         ldlm_cancel_locks_for_export(exp);
1569
1570         spin_lock_irqsave(&exp->exp_lock, irqflags);
1571         exp->exp_flags = flags;
1572         spin_unlock_irqrestore(&exp->exp_lock, irqflags);
1573
1574         rc = class_disconnect(conn, flags);
1575
1576         fsfilt_sync(exp->exp_obd, exp->exp_obd->u.filter.fo_sb);
1577         class_export_put(exp);
1578         /* XXX cleanup preallocated inodes */
1579         RETURN(rc);
1580 }
1581
1582 struct dentry *__filter_oa2dentry(struct obd_device *obd,
1583                                   struct obdo *oa, const char *what)
1584 {
1585         struct dentry *dchild = NULL;
1586
1587         if (oa->o_valid & OBD_MD_FLHANDLE) {
1588                 struct lustre_handle *ost_handle = obdo_handle(oa);
1589                 struct filter_file_data *ffd = filter_handle2ffd(ost_handle);
1590
1591                 if (ffd != NULL) {
1592                         struct filter_dentry_data *fdd;
1593                         dchild = dget(ffd->ffd_file->f_dentry);
1594                         fdd = dchild->d_fsdata;
1595                         LASSERT(fdd->fdd_magic == FILTER_DENTRY_MAGIC);
1596                         filter_ffd_put(ffd);
1597
1598                         CDEBUG(D_INODE,"%s got child objid %*s: %p, count %d\n",
1599                                what, dchild->d_name.len, dchild->d_name.name,
1600                                dchild, atomic_read(&dchild->d_count));
1601                 }
1602         }
1603
1604         if (!dchild)
1605                 dchild = filter_fid2dentry(obd, NULL, oa->o_mode, oa->o_id);
1606
1607         if (IS_ERR(dchild)) {
1608                 CERROR("%s error looking up object: "LPU64"\n", what, oa->o_id);
1609                 RETURN(dchild);
1610         }
1611
1612         if (!dchild->d_inode) {
1613                 CERROR("%s on non-existent object: "LPU64"\n", what, oa->o_id);
1614                 f_dput(dchild);
1615                 RETURN(ERR_PTR(-ENOENT));
1616         }
1617
1618         return dchild;
1619 }
1620
1621 static int filter_getattr(struct lustre_handle *conn, struct obdo *oa,
1622                           struct lov_stripe_md *md)
1623 {
1624         struct dentry *dentry = NULL;
1625         struct obd_device *obd;
1626         int rc = 0;
1627         ENTRY;
1628
1629         obd = class_conn2obd(conn);
1630         if (obd == NULL) {
1631                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",conn->cookie);
1632                 RETURN(-EINVAL);
1633         }
1634
1635         dentry = filter_oa2dentry(obd, oa);
1636         if (IS_ERR(dentry))
1637                 RETURN(PTR_ERR(dentry));
1638
1639         /* Limit the valid bits in the return data to what we actually use */
1640         oa->o_valid = OBD_MD_FLID;
1641         obdo_from_inode(oa, dentry->d_inode, FILTER_VALID_FLAGS);
1642
1643         f_dput(dentry);
1644         RETURN(rc);
1645 }
1646
1647 /* this is called from filter_truncate() until we have filter_punch() */
1648 static int filter_setattr(struct lustre_handle *conn, struct obdo *oa,
1649                           struct lov_stripe_md *md, struct obd_trans_info *oti)
1650 {
1651         struct obd_run_ctxt saved;
1652         struct obd_export *exp;
1653         struct filter_obd *filter;
1654         struct dentry *dentry;
1655         struct iattr iattr;
1656         void *handle;
1657         int rc, rc2;
1658         ENTRY;
1659
1660         LASSERT(oti != NULL);
1661         exp = class_conn2export(conn);
1662         if (!exp) {
1663                 CERROR("invalid client cookie "LPX64"\n", conn->cookie);
1664                 RETURN(-EINVAL);
1665         }
1666
1667         dentry = filter_oa2dentry(exp->exp_obd, oa);
1668         if (IS_ERR(dentry))
1669                 GOTO(out_exp, rc = PTR_ERR(dentry));
1670
1671         filter = &exp->exp_obd->u.filter;
1672
1673         iattr_from_obdo(&iattr, oa, oa->o_valid);
1674
1675         push_ctxt(&saved, &filter->fo_ctxt, NULL);
1676         lock_kernel();
1677
1678         /* XXX this could be a rwsem instead, if filter_preprw played along */
1679         if (iattr.ia_valid & ATTR_SIZE)
1680                 down(&dentry->d_inode->i_sem);
1681
1682         handle = fsfilt_start(exp->exp_obd, dentry->d_inode, FSFILT_OP_SETATTR,
1683                               oti);
1684         if (IS_ERR(handle))
1685                 GOTO(out_unlock, rc = PTR_ERR(handle));
1686
1687         rc = fsfilt_setattr(exp->exp_obd, dentry, handle, &iattr, 1);
1688         rc = filter_finish_transno(exp, oti, rc);
1689         rc2 = fsfilt_commit(exp->exp_obd, dentry->d_inode, handle, 0);
1690         if (rc2) {
1691                 CERROR("error on commit, err = %d\n", rc2);
1692                 if (!rc)
1693                         rc = rc2;
1694         }
1695
1696         if (iattr.ia_valid & ATTR_SIZE)
1697                 up(&dentry->d_inode->i_sem);
1698
1699         oa->o_valid = OBD_MD_FLID;
1700         obdo_from_inode(oa, dentry->d_inode, FILTER_VALID_FLAGS);
1701
1702 out_unlock:
1703         unlock_kernel();
1704         pop_ctxt(&saved, &filter->fo_ctxt, NULL);
1705
1706         f_dput(dentry);
1707  out_exp:
1708         class_export_put(exp);
1709         RETURN(rc);
1710 }
1711
1712 static int filter_open(struct lustre_handle *conn, struct obdo *oa,
1713                        struct lov_stripe_md *ea, struct obd_trans_info *oti,
1714                        struct obd_client_handle *och)
1715 {
1716         struct obd_export *exp;
1717         struct lustre_handle *handle;
1718         struct filter_file_data *ffd;
1719         struct file *filp;
1720         struct lustre_handle parent_lockh;
1721         int rc = 0;
1722         ENTRY;
1723
1724         exp = class_conn2export(conn);
1725         if (exp == NULL) {
1726                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",conn->cookie);
1727                 RETURN(-EINVAL);
1728         }
1729
1730         filp = filter_obj_open(exp, oti, oa->o_id, oa->o_mode,
1731                                LCK_PR, &parent_lockh);
1732         if (IS_ERR(filp))
1733                 GOTO(out, rc = PTR_ERR(filp));
1734
1735         oa->o_valid = OBD_MD_FLID;
1736         obdo_from_inode(oa, filp->f_dentry->d_inode, FILTER_VALID_FLAGS);
1737
1738         ffd = filp->private_data;
1739         handle = obdo_handle(oa);
1740         handle->cookie = ffd->ffd_handle.h_cookie;
1741         oa->o_valid |= OBD_MD_FLHANDLE;
1742
1743 out:
1744         class_export_put(exp);
1745         if (!rc) {
1746                 memcpy(&oti->oti_ack_locks[0].lock, &parent_lockh,
1747                        sizeof(parent_lockh));
1748                 oti->oti_ack_locks[0].mode = LCK_PR;
1749         }
1750         RETURN(rc);
1751 }
1752
1753 static int filter_close(struct lustre_handle *conn, struct obdo *oa,
1754                         struct lov_stripe_md *ea, struct obd_trans_info *oti)
1755 {
1756         struct obd_export *exp;
1757         struct filter_file_data *ffd;
1758         struct filter_export_data *fed;
1759         int rc;
1760         ENTRY;
1761
1762         exp = class_conn2export(conn);
1763         if (exp == NULL) {
1764                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",conn->cookie);
1765                 RETURN(-EINVAL);
1766         }
1767
1768         if (!(oa->o_valid & OBD_MD_FLHANDLE)) {
1769                 CERROR("no handle for close of objid "LPU64"\n", oa->o_id);
1770                 GOTO(out, rc = -EINVAL);
1771         }
1772
1773         ffd = filter_handle2ffd(obdo_handle(oa));
1774         if (ffd == NULL) {
1775                 CERROR("bad handle ("LPX64") for close\n",
1776                        obdo_handle(oa)->cookie);
1777                 GOTO(out, rc = -ESTALE);
1778         }
1779
1780         fed = &exp->exp_filter_data;
1781         spin_lock(&fed->fed_lock);
1782         list_del(&ffd->ffd_export_list);
1783         spin_unlock(&fed->fed_lock);
1784
1785         oa->o_valid = OBD_MD_FLID;
1786         obdo_from_inode(oa,ffd->ffd_file->f_dentry->d_inode,FILTER_VALID_FLAGS);
1787
1788         rc = filter_close_internal(exp, ffd, oti, 0);
1789         filter_ffd_put(ffd);
1790         GOTO(out, rc);
1791  out:
1792         class_export_put(exp);
1793         return rc;
1794 }
1795
1796 static int filter_create(struct lustre_handle *conn, struct obdo *oa,
1797                          struct lov_stripe_md **ea, struct obd_trans_info *oti)
1798 {
1799         struct obd_export *exp;
1800         struct obd_device *obd;
1801         struct filter_obd *filter;
1802         struct obd_run_ctxt saved;
1803         struct lustre_handle parent_lockh;
1804         struct dentry *dparent;
1805         struct ll_fid mds_fid = { .id = 0 };
1806         struct dentry *dchild = NULL;
1807         void *handle;
1808         int err, rc, cleanup_phase;
1809         ENTRY;
1810
1811         exp = class_conn2export(conn);
1812         if (exp == NULL) {
1813                 CDEBUG(D_IOCTL,"invalid client cookie "LPX64"\n", conn->cookie);
1814                 RETURN(-EINVAL);
1815         }
1816
1817         obd = exp->exp_obd;
1818         filter = &obd->u.filter;
1819         push_ctxt(&saved, &filter->fo_ctxt, NULL);
1820  retry:
1821         oa->o_id = filter_next_id(filter);
1822
1823         cleanup_phase = 0;
1824         dparent = filter_parent_lock(obd, S_IFREG, oa->o_id, LCK_PW,
1825                                      &parent_lockh);
1826         if (IS_ERR(dparent))
1827                 GOTO(cleanup, rc = PTR_ERR(dparent));
1828         cleanup_phase = 1;
1829
1830         dchild = filter_fid2dentry(obd, dparent, S_IFREG, oa->o_id);
1831         if (IS_ERR(dchild))
1832                 GOTO(cleanup, rc = PTR_ERR(dchild));
1833         if (dchild->d_inode) {
1834                 /* This would only happen if lastobjid was bad on disk */
1835                 CERROR("Serious error: objid %*s already exists; is this "
1836                        "filesystem corrupt?  I will try to work around it.\n",
1837                        dchild->d_name.len, dchild->d_name.name);
1838                 f_dput(dchild);
1839                 filter_parent_unlock(dparent, &parent_lockh, LCK_PW);
1840                 goto retry;
1841         }
1842
1843         cleanup_phase = 2;
1844         handle = fsfilt_start(obd, dparent->d_inode, FSFILT_OP_CREATE_LOG, oti);
1845         if (IS_ERR(handle))
1846                 GOTO(cleanup, rc = PTR_ERR(handle));
1847
1848         rc = vfs_create(dparent->d_inode, dchild, oa->o_mode);
1849         if (rc) {
1850                 CERROR("create failed rc = %d\n", rc);
1851         } else if (oa->o_valid & (OBD_MD_FLCTIME|OBD_MD_FLMTIME|OBD_MD_FLSIZE)){
1852                 struct iattr attr;
1853
1854                 iattr_from_obdo(&attr, oa, oa->o_valid);
1855                 rc = fsfilt_setattr(obd, dchild, handle, &attr, 1);
1856                 if (rc)
1857                         CERROR("create setattr failed rc = %d\n", rc);
1858         }
1859         rc = filter_finish_transno(exp, oti, rc);
1860         err = filter_update_server_data(obd, filter->fo_rcvd_filp,
1861                                         filter->fo_fsd);
1862         if (err)
1863                 CERROR("unable to write lastobjid but file created\n");
1864
1865         /* Set flags for fields we have set in the inode struct */
1866         if (!rc && mds_fid.id && (oa->o_valid & OBD_MD_FLCOOKIE)) {
1867                 err = filter_log_op_create(obd->u.filter.fo_catalog, &mds_fid,
1868                                            dchild->d_inode->i_ino,
1869                                            dchild->d_inode->i_generation,
1870                                            oti->oti_logcookies);
1871                 if (err) {
1872                         CERROR("error logging create record: rc %d\n", err);
1873                         oa->o_valid = OBD_MD_FLID;
1874                 } else {
1875                         oa->o_valid = OBD_MD_FLID | OBD_MD_FLCOOKIE;
1876                 }
1877         } else
1878                 oa->o_valid = OBD_MD_FLID;
1879
1880         err = fsfilt_commit(obd, dparent->d_inode, handle, 0);
1881         if (err) {
1882                 CERROR("error on commit, err = %d\n", err);
1883                 if (!rc)
1884                         rc = err;
1885         }
1886
1887         if (rc)
1888                 GOTO(cleanup, rc);
1889
1890         /* Set flags for fields we have set in the inode struct */
1891         obdo_from_inode(oa, dchild->d_inode, FILTER_VALID_FLAGS);
1892
1893         EXIT;
1894 cleanup:
1895         switch(cleanup_phase) {
1896         case 2:
1897                 f_dput(dchild);
1898         case 1: /* locked parent dentry */
1899                 if (rc || oti == NULL) {
1900                         filter_parent_unlock(dparent, &parent_lockh, LCK_PW);
1901                 } else {
1902                         memcpy(&oti->oti_ack_locks[0].lock, &parent_lockh,
1903                                sizeof(parent_lockh));
1904                         oti->oti_ack_locks[0].mode = LCK_PW;
1905                 }
1906         case 0:
1907                 pop_ctxt(&saved, &filter->fo_ctxt, NULL);
1908                 class_export_put(exp);
1909                 break;
1910         default:
1911                 CERROR("invalid cleanup_phase %d\n", cleanup_phase);
1912                 LBUG();
1913         }
1914
1915         RETURN(rc);
1916 }
1917
1918 static int filter_destroy(struct lustre_handle *conn, struct obdo *oa,
1919                           struct lov_stripe_md *ea, struct obd_trans_info *oti)
1920 {
1921         struct obd_export *exp;
1922         struct obd_device *obd;
1923         struct filter_obd *filter;
1924         struct dentry *dchild = NULL, *dparent = NULL;
1925         struct filter_dentry_data *fdd;
1926         struct obd_run_ctxt saved;
1927         void *handle = NULL;
1928         struct lustre_handle parent_lockh;
1929         struct llog_cookie *fcc = NULL;
1930         int rc, rc2, cleanup_phase = 0;
1931         ENTRY;
1932
1933         exp = class_conn2export(conn);
1934         if (exp == NULL) {
1935                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",conn->cookie);
1936                 RETURN(-EINVAL);
1937         }
1938
1939         obd = exp->exp_obd;
1940         filter = &obd->u.filter;
1941
1942         push_ctxt(&saved, &filter->fo_ctxt, NULL);
1943         dparent = filter_parent_lock(obd, oa->o_mode, oa->o_id,
1944                                      LCK_PW, &parent_lockh);
1945         if (IS_ERR(dparent))
1946                 GOTO(cleanup, rc = PTR_ERR(dparent));
1947         cleanup_phase = 1;
1948
1949         dchild = filter_fid2dentry(obd, dparent, S_IFREG, oa->o_id);
1950         if (IS_ERR(dchild))
1951                 GOTO(cleanup, rc = -ENOENT);
1952         cleanup_phase = 2;
1953
1954         if (dchild->d_inode == NULL) {
1955                 CERROR("destroying non-existent object "LPU64"\n", oa->o_id);
1956                 GOTO(cleanup, rc = -ENOENT);
1957         }
1958         handle = fsfilt_start(obd, dparent->d_inode, FSFILT_OP_UNLINK_LOG, oti);
1959         if (IS_ERR(handle))
1960                 GOTO(cleanup, rc = PTR_ERR(handle));
1961         cleanup_phase = 3;
1962
1963         fdd = dchild->d_fsdata;
1964
1965         /* Our MDC connection is established by the MDS to us */
1966         if ((oa->o_valid & OBD_MD_FLCOOKIE) && filter->fo_mdc_imp != NULL) {
1967                 OBD_ALLOC(fcc, sizeof(*fcc));
1968                 if (fcc != NULL)
1969                         memcpy(fcc, obdo_logcookie(oa), sizeof(*fcc));
1970         }
1971
1972         if (fdd != NULL && atomic_read(&fdd->fdd_open_count)) {
1973                 LASSERT(fdd->fdd_magic == FILTER_DENTRY_MAGIC);
1974                 if (!(fdd->fdd_flags & FILTER_FLAG_DESTROY)) {
1975                         fdd->fdd_flags |= FILTER_FLAG_DESTROY;
1976
1977 #ifdef ENABLE_ORPHANS
1978                         filter_log_op_orphan(filter->fo_catalog, oa->o_id,
1979                                              oa->o_generation,&fdd->fdd_cookie);
1980 #endif
1981                         CDEBUG(D_INODE,
1982                                "defer destroy of %dx open objid "LPU64"\n",
1983                                atomic_read(&fdd->fdd_open_count), oa->o_id);
1984                 } else {
1985                         CDEBUG(D_INODE,
1986                                "repeat destroy of %dx open objid "LPU64"\n",
1987                                atomic_read(&fdd->fdd_open_count), oa->o_id);
1988                 }
1989                 GOTO(cleanup, rc = 0);
1990         }
1991
1992         rc = filter_destroy_internal(obd, oa->o_id, dparent, dchild);
1993
1994 cleanup:
1995         switch(cleanup_phase) {
1996         case 3:
1997                 if (fcc != NULL)
1998                         fsfilt_set_last_rcvd(obd, 0, oti->oti_handle,
1999                                              filter_cancel_cookies_cb, fcc);
2000                 rc = filter_finish_transno(exp, oti, rc);
2001                 rc2 = fsfilt_commit(obd, dparent->d_inode, handle, 0);
2002                 if (rc2) {
2003                         CERROR("error on commit, err = %d\n", rc2);
2004                         if (!rc)
2005                                 rc = rc2;
2006                 }
2007         case 2:
2008                 f_dput(dchild);
2009         case 1:
2010                 if (rc || oti == NULL) {
2011                         filter_parent_unlock(dparent, &parent_lockh, LCK_PW);
2012                 } else {
2013                         memcpy(&oti->oti_ack_locks[0].lock, &parent_lockh,
2014                                sizeof(parent_lockh));
2015                         oti->oti_ack_locks[0].mode = LCK_PW;
2016                 }
2017         case 0:
2018                 pop_ctxt(&saved, &filter->fo_ctxt, NULL);
2019                 class_export_put(exp);
2020                 break;
2021         default:
2022                 CERROR("invalid cleanup_phase %d\n", cleanup_phase);
2023                 LBUG();
2024         }
2025
2026         RETURN(rc);
2027 }
2028
2029 /* NB start and end are used for punch, but not truncate */
2030 static int filter_truncate(struct lustre_handle *conn, struct obdo *oa,
2031                            struct lov_stripe_md *lsm,
2032                            obd_off start, obd_off end,
2033                            struct obd_trans_info *oti)
2034 {
2035         int error;
2036         ENTRY;
2037
2038         if (end != OBD_OBJECT_EOF)
2039                 CERROR("PUNCH not supported, only truncate: end = "LPX64"\n",
2040                        end);
2041
2042         CDEBUG(D_INODE, "calling truncate for object "LPU64", valid = %x, "
2043                "o_size = "LPD64"\n", oa->o_id, oa->o_valid, start);
2044         oa->o_size = start;
2045         error = filter_setattr(conn, oa, NULL, oti);
2046         RETURN(error);
2047 }
2048
2049 static int filter_syncfs(struct obd_export *exp)
2050 {
2051         ENTRY;
2052
2053         RETURN(fsfilt_sync(exp->exp_obd, exp->exp_obd->u.filter.fo_sb));
2054 }
2055
2056 static int filter_statfs(struct obd_device *obd, struct obd_statfs *osfs,
2057                          unsigned long max_age)
2058 {
2059         ENTRY;
2060         RETURN(fsfilt_statfs(obd, obd->u.filter.fo_sb, osfs));
2061 }
2062
2063 static int filter_get_info(struct lustre_handle *conn, __u32 keylen,
2064                            void *key, __u32 *vallen, void *val)
2065 {
2066         struct obd_device *obd;
2067         ENTRY;
2068
2069         obd = class_conn2obd(conn);
2070         if (obd == NULL) {
2071                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",
2072                        conn->cookie);
2073                 RETURN(-EINVAL);
2074         }
2075
2076         if (keylen == strlen("blocksize") &&
2077             memcmp(key, "blocksize", keylen) == 0) {
2078                 __u32 *blocksize = val;
2079                 *vallen = sizeof(*blocksize);
2080                 *blocksize = obd->u.filter.fo_sb->s_blocksize;
2081                 RETURN(0);
2082         }
2083
2084         if (keylen == strlen("blocksize_bits") &&
2085             memcmp(key, "blocksize_bits", keylen) == 0) {
2086                 __u32 *blocksize_bits = val;
2087                 *vallen = sizeof(*blocksize_bits);
2088                 *blocksize_bits = obd->u.filter.fo_sb->s_blocksize_bits;
2089                 RETURN(0);
2090         }
2091
2092         CDEBUG(D_IOCTL, "invalid key\n");
2093         RETURN(-EINVAL);
2094 }
2095
2096 static int filter_set_info(struct lustre_handle *conn, __u32 keylen,
2097                            void *key, __u32 vallen, void *val)
2098 {
2099         struct obd_device *obd;
2100         struct obd_export *exp;
2101         struct obd_import *imp;
2102         ENTRY;
2103
2104         obd = class_conn2obd(conn);
2105         if (obd == NULL) {
2106                 CDEBUG(D_IOCTL, "invalid client cookie "LPX64"\n",
2107                        conn->cookie);
2108                 RETURN(-EINVAL);
2109         }
2110
2111         if (keylen < strlen("mds_conn") ||
2112             memcmp(key, "mds_conn", keylen) != 0)
2113                 RETURN(-EINVAL);
2114
2115         CERROR("Received MDS connection ("LPX64")\n", conn->cookie);
2116         memcpy(&obd->u.filter.fo_mdc_conn, conn, sizeof(*conn));
2117
2118         imp = obd->u.filter.fo_mdc_imp = class_new_import();
2119
2120         exp = class_conn2export(conn);
2121         imp->imp_connection = ptlrpc_connection_addref(exp->exp_connection);
2122         class_export_put(exp);
2123
2124         imp->imp_client = &obd->u.filter.fo_mdc_client;
2125         imp->imp_remote_handle = *conn;
2126         imp->imp_obd = obd;
2127         imp->imp_dlm_fake = 1; /* XXX rename imp_dlm_fake to something else */
2128         imp->imp_level = LUSTRE_CONN_FULL;
2129         class_import_put(imp);
2130
2131         RETURN(0);
2132 }
2133
2134 int filter_iocontrol(unsigned int cmd, struct lustre_handle *conn,
2135                      int len, void *karg, void *uarg)
2136 {
2137         struct obd_device *obd = class_conn2obd(conn);
2138
2139         switch (cmd) {
2140         case OBD_IOC_ABORT_RECOVERY:
2141                 CERROR("aborting recovery for device %s\n", obd->obd_name);
2142                 target_abort_recovery(obd);
2143                 RETURN(0);
2144
2145         default:
2146                 RETURN(-EINVAL);
2147         }
2148         RETURN(0);
2149 }
2150
2151 static struct obd_ops filter_obd_ops = {
2152         o_owner:          THIS_MODULE,
2153         o_attach:         filter_attach,
2154         o_detach:         filter_detach,
2155         o_get_info:       filter_get_info,
2156         o_set_info:       filter_set_info,
2157         o_setup:          filter_setup,
2158         o_cleanup:        filter_cleanup,
2159         o_connect:        filter_connect,
2160         o_disconnect:     filter_disconnect,
2161         o_statfs:         filter_statfs,
2162         o_syncfs:         filter_syncfs,
2163         o_getattr:        filter_getattr,
2164         o_create:         filter_create,
2165         o_setattr:        filter_setattr,
2166         o_destroy:        filter_destroy,
2167         o_open:           filter_open,
2168         o_close:          filter_close,
2169         o_brw:            filter_brw,
2170         o_punch:          filter_truncate,
2171         o_preprw:         filter_preprw,
2172         o_commitrw:       filter_commitrw,
2173         o_log_cancel:     filter_log_cancel,
2174         o_destroy_export: filter_destroy_export,
2175         o_iocontrol:      filter_iocontrol,
2176 };
2177
2178 static struct obd_ops filter_sanobd_ops = {
2179         o_owner:          THIS_MODULE,
2180         o_attach:         filter_attach,
2181         o_detach:         filter_detach,
2182         o_get_info:       filter_get_info,
2183         o_set_info:       filter_set_info,
2184         o_setup:          filter_san_setup,
2185         o_cleanup:        filter_cleanup,
2186         o_connect:        filter_connect,
2187         o_disconnect:     filter_disconnect,
2188         o_statfs:         filter_statfs,
2189         o_getattr:        filter_getattr,
2190         o_create:         filter_create,
2191         o_setattr:        filter_setattr,
2192         o_destroy:        filter_destroy,
2193         o_open:           filter_open,
2194         o_close:          filter_close,
2195         o_brw:            filter_brw,
2196         o_punch:          filter_truncate,
2197         o_preprw:         filter_preprw,
2198         o_commitrw:       filter_commitrw,
2199         o_log_cancel:     filter_log_cancel,
2200         o_san_preprw:     filter_san_preprw,
2201         o_destroy_export: filter_destroy_export,
2202         o_iocontrol:      filter_iocontrol,
2203 };
2204
2205 static int __init obdfilter_init(void)
2206 {
2207         struct lprocfs_static_vars lvars;
2208         int rc;
2209
2210         printk(KERN_INFO "Lustre Filtering OBD driver; info@clusterfs.com\n");
2211
2212         lprocfs_init_vars(filter, &lvars);
2213
2214         rc = class_register_type(&filter_obd_ops, lvars.module_vars,
2215                                  OBD_FILTER_DEVICENAME);
2216         if (rc)
2217                 return rc;
2218
2219         rc = class_register_type(&filter_sanobd_ops, lvars.module_vars,
2220                                  OBD_FILTER_SAN_DEVICENAME);
2221         if (rc)
2222                 class_unregister_type(OBD_FILTER_DEVICENAME);
2223         return rc;
2224 }
2225
2226 static void __exit obdfilter_exit(void)
2227 {
2228         class_unregister_type(OBD_FILTER_SAN_DEVICENAME);
2229         class_unregister_type(OBD_FILTER_DEVICENAME);
2230 }
2231
2232 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
2233 MODULE_DESCRIPTION("Lustre Filtering OBD driver");
2234 MODULE_LICENSE("GPL");
2235
2236 module_init(obdfilter_init);
2237 module_exit(obdfilter_exit);