lustre/mds/handler.c

   1 /*
   2  *  linux/mds/handler.c
   3  *
   4  *  Lustre Metadata Server (mds) request handler
   5  *
   6  *  Copyright (C) 2001, 2002 Cluster File Systems, Inc.
   7  *
   8  *  This code is issued under the GNU General Public License.
   9  *  See the file COPYING in this distribution
  10  *
  11  *  by Peter Braam <braam@clusterfs.com>
  12  *
  13  *  This server is single threaded at present (but can easily be multi threaded).
  14  *
  15  */
  16
  17
  18 #define EXPORT_SYMTAB
  19
  20 #include <linux/version.h>
  21 #include <linux/module.h>
  22 #include <linux/fs.h>
  23 #include <linux/stat.h>
  24 #include <linux/locks.h>
  25 #include <linux/ext2_fs.h>
  26 #include <linux/quotaops.h>
  27 #include <asm/unistd.h>
  28 #include <asm/uaccess.h>
  29 #include <linux/obd_support.h>
  30 #include <linux/obd.h>
  31 #include <linux/lustre_lib.h>
  32 #include <linux/lustre_idl.h>
  33 #include <linux/lustre_mds.h>
  34 #include <linux/lustre_net.h>
  35 #include <linux/obd_class.h>
  36
  37 // XXX for testing
  38 static struct mds_obd *MDS;
  39
  40 // XXX make this networked!
  41 static int mds_queue_req(struct ptlrpc_request *req)
  42 {
  43         struct ptlrpc_request *srv_req;
  44
  45         if (!MDS) {
  46                 EXIT;
  47                 return -1;
  48         }
  49
  50         srv_req = kmalloc(sizeof(*srv_req), GFP_KERNEL);
  51         if (!srv_req) {
  52                 EXIT;
  53                 return -ENOMEM;
  54         }
  55
  56         printk("---> MDS at %d %p, incoming req %p, srv_req %p\n",
  57                __LINE__, MDS, req, srv_req);
  58
  59         memset(srv_req, 0, sizeof(*req));
  60
  61         /* move the request buffer */
  62         srv_req->rq_reqbuf = req->rq_reqbuf;
  63         srv_req->rq_reqlen    = req->rq_reqlen;
  64         srv_req->rq_obd = MDS;
  65
  66         /* remember where it came from */
  67         srv_req->rq_reply_handle = req;
  68
  69         list_add(&srv_req->rq_list, &MDS->mds_reqs);
  70         wake_up(&MDS->mds_waitq);
  71         return 0;
  72 }
  73
  74 int mds_sendpage(struct ptlrpc_request *req, struct file *file,
  75                     __u64 offset, struct niobuf *dst)
  76 {
  77         int rc;
  78         mm_segment_t oldfs = get_fs();
  79
  80         if (req->rq_peer.peer_nid == 0) {
  81                 /* dst->addr is a user address, but in a different task! */
  82                 set_fs(KERNEL_DS);
  83                 rc = generic_file_read(file, (char *)(long)dst->addr,
  84                                        PAGE_SIZE, &offset);
  85                 set_fs(oldfs);
  86
  87                 if (rc != PAGE_SIZE)
  88                         return -EIO;
  89         } else {
  90                 char *buf;
  91
  92                 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
  93                 if (!buf) {
  94                         return -ENOMEM;
  95                 }
  96
  97                 set_fs(KERNEL_DS);
  98                 rc = generic_file_read(file, buf, PAGE_SIZE, &offset);
  99                 set_fs(oldfs);
 100
 101                 if (rc != PAGE_SIZE)
 102                         return -EIO;
 103
 104                 req->rq_bulkbuf = buf;
 105                 req->rq_bulklen = PAGE_SIZE;
 106                 rc = ptl_send_buf(req, &req->rq_peer, MDS_BULK_PORTAL, 0);
 107                 init_waitqueue_head(&req->rq_wait_for_bulk);
 108                 sleep_on(&req->rq_wait_for_bulk);
 109                 kfree(buf);
 110                 req->rq_bulklen = 0; /* FIXME: eek. */
 111         }
 112
 113         return 0;
 114 }
 115
 116 int mds_reply(struct ptlrpc_request *req)
 117 {
 118         struct ptlrpc_request *clnt_req = req->rq_reply_handle;
 119
 120         ENTRY;
 121
 122         if (req->rq_obd->mds_service != NULL) {
 123                 /* This is a request that came from the network via portals. */
 124
 125                 /* FIXME: we need to increment the count of handled events */
 126                 ptl_send_buf(req, &req->rq_peer, MDS_REPLY_PORTAL, 0);
 127         } else {
 128                 /* This is a local request that came from another thread. */
 129
 130                 /* move the reply to the client */
 131                 clnt_req->rq_replen = req->rq_replen;
 132                 clnt_req->rq_repbuf = req->rq_repbuf;
 133                 req->rq_repbuf = NULL;
 134                 req->rq_replen = 0;
 135
 136                 /* free the request buffer */
 137                 kfree(req->rq_reqbuf);
 138                 req->rq_reqbuf = NULL;
 139
 140                 /* wake up the client */
 141                 wake_up_interruptible(&clnt_req->rq_wait_for_rep);
 142         }
 143
 144         EXIT;
 145         return 0;
 146 }
 147
 148 int mds_error(struct ptlrpc_request *req)
 149 {
 150         struct ptlrep_hdr *hdr;
 151
 152         ENTRY;
 153
 154         hdr = kmalloc(sizeof(*hdr), GFP_KERNEL);
 155         if (!hdr) {
 156                 EXIT;
 157                 return -ENOMEM;
 158         }
 159
 160         memset(hdr, 0, sizeof(*hdr));
 161
 162         hdr->seqno = req->rq_reqhdr->seqno;
 163         hdr->status = req->rq_status;
 164         hdr->type = MDS_TYPE_ERR;
 165
 166         req->rq_repbuf = (char *)hdr;
 167         req->rq_replen = sizeof(*hdr);
 168
 169         EXIT;
 170         return mds_reply(req);
 171 }
 172
 173 struct dentry *mds_fid2dentry(struct mds_obd *mds, struct ll_fid *fid, struct vfsmount **mnt)
 174 {
 175         /* stolen from NFS */
 176         struct super_block *sb = mds->mds_sb;
 177         unsigned long ino = fid->id;
 178         //__u32 generation = fid->generation;
 179         __u32 generation = 0;
 180         struct inode *inode;
 181         struct list_head *lp;
 182         struct dentry *result;
 183
 184         if (mnt) {
 185                 *mnt = mntget(mds->mds_vfsmnt);
 186         }
 187
 188         if (ino == 0)
 189                 return ERR_PTR(-ESTALE);
 190
 191         inode = iget(sb, ino);
 192         if (inode == NULL)
 193                 return ERR_PTR(-ENOMEM);
 194
 195         printk("--> mds_fid2dentry: sb %p\n", inode->i_sb);
 196
 197         if (is_bad_inode(inode)
 198             || (generation && inode->i_generation != generation)
 199                 ) {
 200                 /* we didn't find the right inode.. */
 201                 printk(__FUNCTION__
 202                        "bad inode %lu, link: %d ct: %d or version  %u/%u\n",
 203                         inode->i_ino,
 204                         inode->i_nlink, atomic_read(&inode->i_count),
 205                         inode->i_generation,
 206                         generation);
 207                 iput(inode);
 208                 return ERR_PTR(-ESTALE);
 209         }
 210
 211         /* now to find a dentry.
 212          * If possible, get a well-connected one
 213          */
 214         spin_lock(&dcache_lock);
 215         for (lp = inode->i_dentry.next; lp != &inode->i_dentry ; lp=lp->next) {
 216                 result = list_entry(lp,struct dentry, d_alias);
 217                 if (! (result->d_flags & DCACHE_NFSD_DISCONNECTED)) {
 218                         dget_locked(result);
 219                         result->d_vfs_flags |= DCACHE_REFERENCED;
 220                         spin_unlock(&dcache_lock);
 221                         iput(inode);
 222                         return result;
 223                 }
 224         }
 225         spin_unlock(&dcache_lock);
 226         result = d_alloc_root(inode);
 227         if (result == NULL) {
 228                 iput(inode);
 229                 return ERR_PTR(-ENOMEM);
 230         }
 231         result->d_flags |= DCACHE_NFSD_DISCONNECTED;
 232         return result;
 233 }
 234
 235 static inline void mds_get_objid(struct inode *inode, __u64 *id)
 236 {
 237         memcpy(id, &inode->u.ext2_i.i_data, sizeof(*id));
 238 }
 239
 240 int mds_getattr(struct ptlrpc_request *req)
 241 {
 242         struct dentry *de = mds_fid2dentry(req->rq_obd, &req->rq_req.mds->fid1,
 243                                            NULL);
 244         struct inode *inode;
 245         struct mds_rep *rep;
 246         int rc;
 247
 248         rc = mds_pack_rep(NULL, 0, NULL, 0, &req->rq_rephdr, &req->rq_rep.mds,
 249                           &req->rq_replen, &req->rq_repbuf);
 250         if (rc) {
 251                 EXIT;
 252                 printk("mds: out of memory\n");
 253                 req->rq_status = -ENOMEM;
 254                 return 0;
 255         }
 256
 257         req->rq_rephdr->seqno = req->rq_reqhdr->seqno;
 258         rep = req->rq_rep.mds;
 259
 260         if (!de) {
 261                 EXIT;
 262                 req->rq_rephdr->status = -ENOENT;
 263                 return 0;
 264         }
 265
 266         inode = de->d_inode;
 267         rep->ino = inode->i_ino;
 268         rep->atime = inode->i_atime;
 269         rep->ctime = inode->i_ctime;
 270         rep->mtime = inode->i_mtime;
 271         rep->uid = inode->i_uid;
 272         rep->gid = inode->i_gid;
 273         rep->size = inode->i_size;
 274         rep->mode = inode->i_mode;
 275         rep->nlink = inode->i_nlink;
 276         rep->valid = ~0;
 277         mds_get_objid(inode, &rep->objid);
 278         dput(de);
 279         return 0;
 280 }
 281
 282 int mds_readpage(struct ptlrpc_request *req)
 283 {
 284         struct vfsmount *mnt;
 285         struct dentry *de = mds_fid2dentry(req->rq_obd, &req->rq_req.mds->fid1,
 286                                            &mnt);
 287         struct file *file;
 288         struct niobuf *niobuf;
 289         struct mds_rep *rep;
 290         int rc;
 291
 292         printk("mds_readpage: ino %ld\n", de->d_inode->i_ino);
 293         rc = mds_pack_rep(NULL, 0, NULL, 0, &req->rq_rephdr, &req->rq_rep.mds,
 294                           &req->rq_replen, &req->rq_repbuf);
 295         if (rc) {
 296                 EXIT;
 297                 printk("mds: out of memory\n");
 298                 req->rq_status = -ENOMEM;
 299                 return 0;
 300         }
 301
 302         req->rq_rephdr->seqno = req->rq_reqhdr->seqno;
 303         rep = req->rq_rep.mds;
 304
 305         if (IS_ERR(de)) {
 306                 EXIT;
 307                 req->rq_rephdr->status = PTR_ERR(de);
 308                 return 0;
 309         }
 310
 311         file = dentry_open(de, mnt, O_RDONLY | O_LARGEFILE);
 312         /* note: in case of an error, dentry_open puts dentry */
 313         if (IS_ERR(file)) {
 314                 EXIT;
 315                 req->rq_rephdr->status = PTR_ERR(file);
 316                 return 0;
 317         }
 318
 319         niobuf = mds_req_tgt(req->rq_req.mds);
 320
 321         /* to make this asynchronous make sure that the handling function
 322            doesn't send a reply when this function completes. Instead a
 323            callback function would send the reply */
 324         rc = mds_sendpage(req, file, req->rq_req.mds->size, niobuf);
 325
 326         filp_close(file, 0);
 327         req->rq_rephdr->status = rc;
 328         EXIT;
 329         return 0;
 330 }
 331
 332 int mds_reint(struct ptlrpc_request *req)
 333 {
 334         int rc;
 335         char *buf = mds_req_tgt(req->rq_req.mds);
 336         int len = req->rq_req.mds->tgtlen;
 337         struct mds_update_record rec;
 338
 339         rc = mds_update_unpack(buf, len, &rec);
 340         if (rc) {
 341                 printk(__FUNCTION__ ": invalid record\n");
 342                 req->rq_status = -EINVAL;
 343                 return 0;
 344         }
 345         /* rc will be used to interrupt a for loop over multiple records */
 346         rc = mds_reint_rec(&rec, req);
 347         return 0;
 348 }
 349
 350 //int mds_handle(struct mds_conn *conn, int len, char *buf)
 351 int mds_handle(struct ptlrpc_request *req)
 352 {
 353         int rc;
 354         struct ptlreq_hdr *hdr;
 355
 356         ENTRY;
 357
 358         hdr = (struct ptlreq_hdr *)req->rq_reqbuf;
 359
 360         if (NTOH__u32(hdr->type) != MDS_TYPE_REQ) {
 361                 printk("lustre_mds: wrong packet type sent %d\n",
 362                        NTOH__u32(hdr->type));
 363                 rc = -EINVAL;
 364                 goto out;
 365         }
 366
 367         rc = mds_unpack_req(req->rq_reqbuf, req->rq_reqlen,
 368                             &req->rq_reqhdr, &req->rq_req.mds);
 369         if (rc) {
 370                 printk("lustre_mds: Invalid request\n");
 371                 EXIT;
 372                 goto out;
 373         }
 374
 375         switch (req->rq_reqhdr->opc) {
 376
 377         case MDS_GETATTR:
 378                 CDEBUG(D_INODE, "getattr\n");
 379                 rc = mds_getattr(req);
 380                 break;
 381
 382         case MDS_READPAGE:
 383                 CDEBUG(D_INODE, "readpage\n");
 384                 rc = mds_readpage(req);
 385                 break;
 386
 387         case MDS_REINT:
 388                 CDEBUG(D_INODE, "reint\n");
 389                 rc = mds_reint(req);
 390                 break;
 391
 392         default:
 393                 return mds_error(req);
 394         }
 395
 396 out:
 397         if (rc) {
 398                 printk(__FUNCTION__ ": no header\n");
 399                 return 0;
 400         }
 401
 402         if( req->rq_status) {
 403                 mds_error(req);
 404         } else {
 405                 CDEBUG(D_INODE, "sending reply\n");
 406                 mds_reply(req);
 407         }
 408
 409         return 0;
 410 }
 411
 412
 413 static void mds_timer_run(unsigned long __data)
 414 {
 415         struct task_struct * p = (struct task_struct *) __data;
 416
 417         wake_up_process(p);
 418 }
 419
 420 int mds_main(void *arg)
 421 {
 422         struct mds_obd *mds = (struct mds_obd *) arg;
 423         struct timer_list timer;
 424
 425         lock_kernel();
 426         daemonize();
 427         spin_lock_irq(&current->sigmask_lock);
 428         sigfillset(&current->blocked);
 429         recalc_sigpending(current);
 430         spin_unlock_irq(&current->sigmask_lock);
 431
 432         sprintf(current->comm, "lustre_mds");
 433
 434         /* Set up an interval timer which can be used to trigger a
 435            wakeup after the interval expires */
 436         init_timer(&timer);
 437         timer.data = (unsigned long) current;
 438         timer.function = mds_timer_run;
 439         mds->mds_timer = &timer;
 440
 441         /* Record that the  thread is running */
 442         mds->mds_thread = current;
 443         wake_up(&mds->mds_done_waitq);
 444
 445         printk(KERN_INFO "lustre_mds starting.  Commit interval %d seconds\n",
 446                         mds->mds_interval / HZ);
 447
 448         /* XXX maintain a list of all managed devices: insert here */
 449
 450         /* And now, wait forever for commit wakeup events. */
 451         while (1) {
 452                 int rc;
 453
 454                 if (mds->mds_flags & MDS_UNMOUNT)
 455                         break;
 456
 457                 wake_up(&mds->mds_done_waitq);
 458                 interruptible_sleep_on(&mds->mds_waitq);
 459
 460                 CDEBUG(D_INODE, "lustre_mds wakes\n");
 461                 CDEBUG(D_INODE, "pick up req here and continue\n");
 462
 463                 if (mds->mds_service != NULL) {
 464                         ptl_event_t ev;
 465
 466                         while (1) {
 467                                 struct ptlrpc_request request;
 468
 469                                 rc = PtlEQGet(mds->mds_service->srv_eq_h, &ev);
 470                                 if (rc != PTL_OK && rc != PTL_EQ_DROPPED)
 471                                         break;
 472                                 /* FIXME: If we move to an event-driven model,
 473                                  * we should put the request on the stack of
 474                                  * mds_handle instead. */
 475                                 memset(&request, 0, sizeof(request));
 476                                 request.rq_reqbuf = ev.mem_desc.start +
 477                                         ev.offset;
 478                                 request.rq_reqlen = ev.mem_desc.length;
 479                                 request.rq_obd = MDS;
 480                                 request.rq_xid = ev.match_bits;
 481
 482                                 request.rq_peer.peer_nid = ev.initiator.nid;
 483                                 /* FIXME: this NI should be the incoming NI.
 484                                  * We don't know how to find that from here. */
 485                                 request.rq_peer.peer_ni =
 486                                         mds->mds_service->srv_self.peer_ni;
 487                                 rc = mds_handle(&request);
 488                         }
 489                 } else {
 490                         struct ptlrpc_request *request;
 491
 492                         if (list_empty(&mds->mds_reqs)) {
 493                                 CDEBUG(D_INODE, "woke because of timer\n");
 494                         } else {
 495                                 request = list_entry(mds->mds_reqs.next,
 496                                                      struct ptlrpc_request,
 497                                                      rq_list);
 498                                 list_del(&request->rq_list);
 499                                 rc = mds_handle(request);
 500                         }
 501                 }
 502         }
 503
 504         del_timer_sync(mds->mds_timer);
 505
 506         /* XXX maintain a list of all managed devices: cleanup here */
 507
 508         mds->mds_thread = NULL;
 509         wake_up(&mds->mds_done_waitq);
 510         printk("lustre_mds: exiting\n");
 511         return 0;
 512 }
 513
 514 static void mds_stop_srv_thread(struct mds_obd *mds)
 515 {
 516         mds->mds_flags |= MDS_UNMOUNT;
 517
 518         while (mds->mds_thread) {
 519                 wake_up(&mds->mds_waitq);
 520                 sleep_on(&mds->mds_done_waitq);
 521         }
 522 }
 523
 524 static void mds_start_srv_thread(struct mds_obd *mds)
 525 {
 526         init_waitqueue_head(&mds->mds_waitq);
 527         init_waitqueue_head(&mds->mds_done_waitq);
 528         kernel_thread(mds_main, (void *)mds,
 529                       CLONE_VM | CLONE_FS | CLONE_FILES);
 530         while (!mds->mds_thread)
 531                 sleep_on(&mds->mds_done_waitq);
 532 }
 533
 534 /* mount the file system (secretly) */
 535 static int mds_setup(struct obd_device *obddev, obd_count len,
 536                         void *buf)
 537
 538 {
 539         struct obd_ioctl_data* data = buf;
 540         struct mds_obd *mds = &obddev->u.mds;
 541         struct vfsmount *mnt;
 542         struct lustre_peer peer;
 543         int err;
 544         ENTRY;
 545
 546
 547         mnt = do_kern_mount(data->ioc_inlbuf2, 0, data->ioc_inlbuf1, NULL);
 548         err = PTR_ERR(mnt);
 549         if (IS_ERR(mnt)) {
 550                 EXIT;
 551                 return err;
 552         }
 553
 554         mds->mds_sb = mnt->mnt_root->d_inode->i_sb;
 555         if (!obddev->u.mds.mds_sb) {
 556                 EXIT;
 557                 return -ENODEV;
 558         }
 559
 560         mds->mds_vfsmnt = mnt;
 561         obddev->u.mds.mds_fstype = strdup(data->ioc_inlbuf2);
 562
 563         mds->mds_ctxt.pwdmnt = mnt;
 564         mds->mds_ctxt.pwd = mnt->mnt_root;
 565         mds->mds_ctxt.fs = KERNEL_DS;
 566         mds->mds_remote_nid = 0;
 567
 568         INIT_LIST_HEAD(&mds->mds_reqs);
 569         mds->mds_thread = NULL;
 570         mds->mds_flags = 0;
 571         mds->mds_interval = 3 * HZ;
 572         MDS = mds;
 573
 574         spin_lock_init(&obddev->u.mds.mds_lock);
 575
 576         err = kportal_uuid_to_peer("self", &peer);
 577         if (err == 0) {
 578                 mds->mds_service = kmalloc(sizeof(*mds->mds_service),
 579                                                   GFP_KERNEL);
 580                 if (mds->mds_service == NULL)
 581                         return -ENOMEM;
 582                 mds->mds_service->srv_buf_size = 64 * 1024;
 583                 mds->mds_service->srv_portal = MDS_REQUEST_PORTAL;
 584                 memcpy(&mds->mds_service->srv_self, &peer, sizeof(peer));
 585                 mds->mds_service->srv_wait_queue = &mds->mds_waitq;
 586
 587                 rpc_register_service(mds->mds_service, "self");
 588         }
 589
 590         mds_start_srv_thread(mds);
 591
 592         MOD_INC_USE_COUNT;
 593         EXIT;
 594         return 0;
 595 }
 596
 597 static int mds_cleanup(struct obd_device * obddev)
 598 {
 599         struct super_block *sb;
 600         struct mds_obd *mds = &obddev->u.mds;
 601
 602         ENTRY;
 603
 604         if ( !(obddev->obd_flags & OBD_SET_UP) ) {
 605                 EXIT;
 606                 return 0;
 607         }
 608
 609         if ( !list_empty(&obddev->obd_gen_clients) ) {
 610                 printk(KERN_WARNING __FUNCTION__ ": still has clients!\n");
 611                 EXIT;
 612                 return -EBUSY;
 613         }
 614
 615         MDS = NULL;
 616         mds_stop_srv_thread(mds);
 617         sb = mds->mds_sb;
 618         if (!mds->mds_sb){
 619                 EXIT;
 620                 return 0;
 621         }
 622
 623         if (!list_empty(&mds->mds_reqs)) {
 624                 // XXX reply with errors and clean up
 625                 CDEBUG(D_INODE, "Request list not empty!\n");
 626         }
 627
 628         unlock_kernel();
 629         mntput(mds->mds_vfsmnt);
 630         mds->mds_sb = 0;
 631         kfree(mds->mds_fstype);
 632         lock_kernel();
 633
 634         MOD_DEC_USE_COUNT;
 635         EXIT;
 636         return 0;
 637 }
 638
 639 /* use obd ops to offer management infrastructure */
 640 static struct obd_ops mds_obd_ops = {
 641         o_setup:       mds_setup,
 642         o_cleanup:     mds_cleanup,
 643 };
 644
 645 static int __init mds_init(void)
 646 {
 647         obd_register_type(&mds_obd_ops, LUSTRE_MDS_NAME);
 648         return 0;
 649 }
 650
 651 static void __exit mds_exit(void)
 652 {
 653         obd_unregister_type(LUSTRE_MDS_NAME);
 654 }
 655
 656 MODULE_AUTHOR("Peter J. Braam <braam@clusterfs.com>");
 657 MODULE_DESCRIPTION("Lustre Metadata Server (MDS) v0.01");
 658 MODULE_LICENSE("GPL");
 659
 660
 661 // for testing (maybe this stays)
 662 EXPORT_SYMBOL(mds_queue_req);
 663
 664 module_init(mds_init);
 665 module_exit(mds_exit);