1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (c) 2002, 2003 Cluster File Systems, Inc.
5 * Author: Peter Braam <braam@clusterfs.com>
6 * Author: Phil Schwan <phil@clusterfs.com>
7 * Author: Andreas Dilger <adilger@clusterfs.com>
9 * This file is part of Lustre, http://www.lustre.org.
11 * Lustre is free software; you can redistribute it and/or
12 * modify it under the terms of version 2 of the GNU General Public
13 * License as published by the Free Software Foundation.
15 * Lustre is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with Lustre; if not, write to the Free Software
22 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
25 #define DEBUG_SUBSYSTEM S_LLITE
26 #include <linux/lustre_dlm.h>
27 #include <linux/lustre_lite.h>
28 #include <linux/pagemap.h>
29 #include <linux/file.h>
30 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
31 #include <linux/lustre_compat25.h>
33 #include "llite_internal.h"
34 #include <linux/obd_lov.h>
36 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
39 struct ll_file_data *fd = file->private_data;
40 struct ptlrpc_request *req = NULL;
41 struct obd_client_handle *och = &fd->fd_mds_och;
42 struct obdo *obdo = NULL;
46 /* clear group lock, if present */
47 if (fd->fd_flags & LL_FILE_GROUP_LOCKED) {
48 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
49 fd->fd_flags &= ~(LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK);
50 rc = ll_extent_unlock(fd, inode, lsm, LCK_GROUP,
58 obdo->o_id = inode->i_ino;
59 obdo->o_valid = OBD_MD_FLID;
60 obdo_from_inode(obdo, inode, (OBD_MD_FLTYPE | OBD_MD_FLMODE |
61 OBD_MD_FLSIZE | OBD_MD_FLBLOCKS |
62 OBD_MD_FLATIME | OBD_MD_FLMTIME |
64 if (0 /* ll_is_inode_dirty(inode) */) {
65 obdo->o_flags = MDS_BFLAG_UNCOMMITTED_WRITES;
66 obdo->o_valid |= OBD_MD_FLFLAGS;
68 obdo->o_mds = id_group(&ll_i2info(inode)->lli_id);
69 rc = md_close(md_exp, obdo, och, &req);
73 /* We are the last writer, so the MDS has instructed us to get
74 * the file size and any write cookies, then close again. */
75 //ll_queue_done_writing(inode);
78 CERROR("inode %lu mdc close failed: rc = %d\n",
82 rc = ll_objects_destroy(req, file->f_dentry->d_inode, 1);
84 CERROR("inode %lu ll_objects destroy: rc = %d\n",
88 mdc_clear_open_replay_data(md_exp, och);
89 ptlrpc_req_finished(req);
90 och->och_fh.cookie = DEAD_HANDLE_MAGIC;
91 file->private_data = NULL;
92 OBD_SLAB_FREE(fd, ll_file_data_slab, sizeof(*fd));
96 /* While this returns an error code, fput() the caller does not, so we need
97 * to make every effort to clean up all of our state here. Also, applications
98 * rarely check close errors and even if an error is returned they will not
99 * re-try the close call.
101 int ll_file_release(struct inode *inode, struct file *file)
103 struct ll_file_data *fd;
104 struct ll_sb_info *sbi = ll_i2sbi(inode);
108 CDEBUG(D_VFSTRACE, "VFS Op:inode="DLID4"(%p)\n",
109 OLID4(&ll_i2info(inode)->lli_id), inode);
111 /* don't do anything for / */
112 if (inode->i_sb->s_root == file->f_dentry)
115 lprocfs_counter_incr(sbi->ll_stats, LPROC_LL_RELEASE);
116 fd = (struct ll_file_data *)file->private_data;
119 rc = ll_md_close(sbi->ll_md_exp, inode, file);
123 static int ll_intent_file_open(struct file *file, void *lmm,
124 int lmmsize, struct lookup_intent *itp)
126 struct ll_sb_info *sbi = ll_i2sbi(file->f_dentry->d_inode);
127 struct dentry *parent = file->f_dentry->d_parent;
128 const char *name = file->f_dentry->d_name.name;
129 const int len = file->f_dentry->d_name.len;
130 struct lustre_handle lockh;
131 struct mdc_op_data *op_data;
137 OBD_ALLOC(op_data, sizeof(*op_data));
140 ll_prepare_mdc_data(op_data, parent->d_inode, NULL, name, len, O_RDWR);
142 rc = md_enqueue(sbi->ll_md_exp, LDLM_IBITS, itp, LCK_PR, op_data,
143 &lockh, lmm, lmmsize, ldlm_completion_ast,
144 ll_mdc_blocking_ast, NULL);
145 OBD_FREE(op_data, sizeof(*op_data));
147 if (itp->d.lustre.it_lock_mode)
148 memcpy(&itp->d.lustre.it_lock_handle,
149 &lockh, sizeof(lockh));
151 CERROR("lock enqueue: err: %d\n", rc);
157 int ll_local_open(struct file *file, struct lookup_intent *it)
159 struct ptlrpc_request *req = it->d.lustre.it_data;
160 struct ll_inode_info *lli = ll_i2info(file->f_dentry->d_inode);
161 struct obd_export *md_exp = ll_i2mdexp(file->f_dentry->d_inode);
162 struct ll_file_data *fd;
163 struct mds_body *body;
166 body = lustre_msg_buf (req->rq_repmsg, 1, sizeof (*body));
167 LASSERT (body != NULL); /* reply already checked out */
168 LASSERT_REPSWABBED (req, 1); /* and swabbed down */
170 LASSERTF(file->private_data == NULL, "file %*s/%*s ino %lu/%u (%o)\n",
171 file->f_dentry->d_name.len, file->f_dentry->d_name.name,
172 file->f_dentry->d_parent->d_name.len,
173 file->f_dentry->d_parent->d_name.name,
174 file->f_dentry->d_inode->i_ino,
175 file->f_dentry->d_inode->i_generation,
176 file->f_dentry->d_inode->i_mode);
179 OBD_SLAB_ALLOC(fd, ll_file_data_slab, SLAB_KERNEL, sizeof *fd);
181 /* We can't handle this well without reorganizing ll_file_open and
182 * ll_md_close(), so don't even try right now. */
185 memcpy(&fd->fd_mds_och.och_fh, &body->handle, sizeof(body->handle));
186 fd->fd_mds_och.och_magic = OBD_CLIENT_HANDLE_MAGIC;
187 file->private_data = fd;
188 ll_readahead_init(file->f_dentry->d_inode, &fd->fd_ras);
190 lli->lli_io_epoch = body->io_epoch;
192 mdc_set_open_replay_data(md_exp, &fd->fd_mds_och, it->d.lustre.it_data);
197 /* Open a file, and (for the very first open) create objects on the OSTs at
198 * this time. If opened with O_LOV_DELAY_CREATE, then we don't do the object
199 * creation or open until ll_lov_setstripe() ioctl is called. We grab
200 * lli_open_sem to ensure no other process will create objects, send the
201 * stripe MD to the MDS, or try to destroy the objects if that fails.
203 * If we already have the stripe MD locally then we don't request it in
204 * mdc_open(), by passing a lmm_size = 0.
206 * It is up to the application to ensure no other processes open this file
207 * in the O_LOV_DELAY_CREATE case, or the default striping pattern will be
208 * used. We might be able to avoid races of that sort by getting lli_open_sem
209 * before returning in the O_LOV_DELAY_CREATE case and dropping it here
210 * or in ll_file_release(), but I'm not sure that is desirable/necessary.
212 int ll_file_open(struct inode *inode, struct file *file)
214 struct ll_inode_info *lli = ll_i2info(inode);
215 struct lookup_intent *it, oit = { .it_op = IT_OPEN,
216 .it_flags = file->f_flags };
217 struct lov_stripe_md *lsm;
218 struct ptlrpc_request *req;
222 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), flags %o\n", inode->i_ino,
223 inode->i_generation, inode, file->f_flags);
225 /* don't do anything for / */
226 if (inode->i_sb->s_root == file->f_dentry)
231 if (!it || !it->d.lustre.it_disposition) {
233 rc = ll_intent_file_open(file, NULL, 0, it);
238 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_OPEN);
239 /* mdc_intent_lock() didn't get a request ref if there was an open
240 * error, so don't do cleanup on the request here (bug 3430) */
241 rc = it_open_error(DISP_OPEN_OPEN, it);
245 rc = ll_local_open(file, it);
247 LASSERTF(rc == 0, "rc = %d\n", rc);
249 if (!S_ISREG(inode->i_mode))
254 if (file->f_flags & O_LOV_DELAY_CREATE ||
255 !(file->f_mode & FMODE_WRITE)) {
256 CDEBUG(D_INODE, "object creation was delayed\n");
260 file->f_flags &= ~O_LOV_DELAY_CREATE;
263 req = it->d.lustre.it_data;
264 ptlrpc_req_finished(req);
266 ll_open_complete(inode);
270 /* Fills the obdo with the attributes for the inode defined by lsm */
271 int ll_lsm_getattr(struct obd_export *exp, struct lov_stripe_md *lsm,
274 struct ptlrpc_request_set *set;
278 LASSERT(lsm != NULL);
280 memset(oa, 0, sizeof *oa);
281 oa->o_id = lsm->lsm_object_id;
282 oa->o_gr = lsm->lsm_object_gr;
283 oa->o_mode = S_IFREG;
284 oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLSIZE |
285 OBD_MD_FLBLOCKS | OBD_MD_FLBLKSZ | OBD_MD_FLMTIME |
286 OBD_MD_FLCTIME | OBD_MD_FLGROUP;
288 set = ptlrpc_prep_set();
292 rc = obd_getattr_async(exp, oa, lsm, set);
294 rc = ptlrpc_set_wait(set);
295 ptlrpc_set_destroy(set);
300 oa->o_valid &= (OBD_MD_FLBLOCKS | OBD_MD_FLBLKSZ | OBD_MD_FLMTIME |
301 OBD_MD_FLCTIME | OBD_MD_FLSIZE);
305 static inline void ll_remove_suid(struct inode *inode)
309 /* set S_IGID if S_IXGRP is set, and always set S_ISUID */
310 mode = (inode->i_mode & S_IXGRP)*(S_ISGID/S_IXGRP) | S_ISUID;
312 /* was any of the uid bits set? */
313 mode &= inode->i_mode;
314 if (mode && !capable(CAP_FSETID)) {
315 inode->i_mode &= ~mode;
316 // XXX careful here - we cannot change the size
320 static int ll_lock_to_stripe_offset(struct inode *inode, struct ldlm_lock *lock)
322 struct ll_inode_info *lli = ll_i2info(inode);
323 struct lov_stripe_md *lsm = lli->lli_smd;
324 struct obd_export *exp = ll_i2dtexp(inode);
327 struct ldlm_lock *lock;
328 struct lov_stripe_md *lsm;
329 } key = { .name = "lock_to_stripe", .lock = lock, .lsm = lsm };
330 __u32 stripe, vallen = sizeof(stripe);
334 if (lsm->lsm_stripe_count == 1)
335 GOTO(check, stripe = 0);
337 /* get our offset in the lov */
338 rc = obd_get_info(exp, sizeof(key), &key, &vallen, &stripe);
340 CERROR("obd_get_info: rc = %d\n", rc);
343 LASSERT(stripe < lsm->lsm_stripe_count);
346 if (lsm->lsm_oinfo[stripe].loi_id != lock->l_resource->lr_name.name[0]||
347 lsm->lsm_oinfo[stripe].loi_gr != lock->l_resource->lr_name.name[2]){
348 LDLM_ERROR(lock, "resource doesn't match object "LPU64"/"LPU64
349 " inode=%lu/%u (%p)\n",
350 lsm->lsm_oinfo[stripe].loi_id,
351 lsm->lsm_oinfo[stripe].loi_gr,
352 inode->i_ino, inode->i_generation, inode);
353 return -ELDLM_NO_LOCK_DATA;
359 /* Flush the page cache for an extent as its canceled. When we're on an LOV,
360 * we get a lock cancellation for each stripe, so we have to map the obd's
361 * region back onto the stripes in the file that it held.
363 * No one can dirty the extent until we've finished our work and they can
364 * enqueue another lock. The DLM protects us from ll_file_read/write here,
365 * but other kernel actors could have pages locked.
367 * Called with the DLM lock held. */
368 void ll_pgcache_remove_extent(struct inode *inode, struct lov_stripe_md *lsm,
369 struct ldlm_lock *lock, __u32 stripe)
371 ldlm_policy_data_t tmpex;
372 unsigned long start, end, count, skip, i, j;
374 int rc, rc2, discard = lock->l_flags & LDLM_FL_DISCARD_DATA;
375 struct lustre_handle lockh;
378 memcpy(&tmpex, &lock->l_policy_data, sizeof(tmpex));
379 CDEBUG(D_INODE|D_PAGE, "inode %lu(%p) ["LPU64"->"LPU64"] size: %llu\n",
380 inode->i_ino, inode, tmpex.l_extent.start, tmpex.l_extent.end,
383 /* our locks are page granular thanks to osc_enqueue, we invalidate the
385 LASSERT((tmpex.l_extent.start & ~PAGE_CACHE_MASK) == 0);
386 LASSERT(((tmpex.l_extent.end + 1) & ~PAGE_CACHE_MASK) == 0);
390 start = tmpex.l_extent.start >> PAGE_CACHE_SHIFT;
391 end = tmpex.l_extent.end >> PAGE_CACHE_SHIFT;
392 if (lsm->lsm_stripe_count > 1) {
393 count = lsm->lsm_stripe_size >> PAGE_CACHE_SHIFT;
394 skip = (lsm->lsm_stripe_count - 1) * count;
395 start += start/count * skip + stripe * count;
397 end += end/count * skip + stripe * count;
399 if (end < tmpex.l_extent.end >> PAGE_CACHE_SHIFT)
402 i = inode->i_size ? (inode->i_size - 1) >> PAGE_CACHE_SHIFT : 0;
406 CDEBUG(D_INODE|D_PAGE, "walking page indices start: %lu j: %lu "
407 "count: %lu skip: %lu end: %lu%s\n", start, start % count,
408 count, skip, end, discard ? " (DISCARDING)" : "");
410 /* walk through the vmas on the inode and tear down mmaped pages that
411 * intersect with the lock. this stops immediately if there are no
412 * mmap()ed regions of the file. This is not efficient at all and
413 * should be short lived. We'll associate mmap()ed pages with the lock
414 * and will be able to find them directly */
416 for (i = start; i <= end; i += (j + skip)) {
417 j = min(count - (i % count), end - i + 1);
418 LASSERT(inode->i_mapping);
419 if (ll_teardown_mmaps(inode->i_mapping, i << PAGE_CACHE_SHIFT,
420 ((i+j) << PAGE_CACHE_SHIFT) - 1) )
424 /* this is the simplistic implementation of page eviction at
425 * cancelation. It is careful to get races with other page
426 * lockers handled correctly. fixes from bug 20 will make it
427 * more efficient by associating locks with pages and with
428 * batching writeback under the lock explicitly. */
429 for (i = start, j = start % count; i <= end;
430 j++, i++, tmpex.l_extent.start += PAGE_CACHE_SIZE) {
432 CDEBUG(D_PAGE, "skip index %lu to %lu\n", i, i + skip);
438 LASSERTF(tmpex.l_extent.start< lock->l_policy_data.l_extent.end,
439 LPU64" >= "LPU64" start %lu i %lu end %lu\n",
440 tmpex.l_extent.start, lock->l_policy_data.l_extent.end,
443 if (!mapping_has_pages(inode->i_mapping)) {
444 CDEBUG(D_INODE|D_PAGE, "nothing left\n");
450 page = find_get_page(inode->i_mapping, i);
453 LL_CDEBUG_PAGE(D_PAGE, page, "lock page idx %lu ext "LPU64"\n",
454 i, tmpex.l_extent.start);
457 /* page->mapping to check with racing against teardown */
458 if (!discard && clear_page_dirty_for_io(page)) {
459 rc = ll_call_writepage(inode, page);
461 CERROR("writepage of page %p failed: %d\n",
463 /* either waiting for io to complete or reacquiring
464 * the lock that the failed writepage released */
468 tmpex.l_extent.end = tmpex.l_extent.start + PAGE_CACHE_SIZE - 1;
469 /* check to see if another DLM lock covers this page */
470 rc2 = ldlm_lock_match(lock->l_resource->lr_namespace,
471 LDLM_FL_BLOCK_GRANTED|LDLM_FL_CBPENDING |
473 &lock->l_resource->lr_name, LDLM_EXTENT,
474 &tmpex, LCK_PR | LCK_PW, &lockh);
475 if (rc2 == 0 && page->mapping != NULL) {
476 // checking again to account for writeback's lock_page()
477 LL_CDEBUG_PAGE(D_PAGE, page, "truncating\n");
478 ll_ra_accounting(page, inode->i_mapping);
479 ll_truncate_complete_page(page);
482 page_cache_release(page);
484 LASSERTF(tmpex.l_extent.start <=
485 (lock->l_policy_data.l_extent.end == ~0ULL ? ~0ULL :
486 lock->l_policy_data.l_extent.end + 1),
487 "loop too long "LPU64" > "LPU64" start %lu i %lu end %lu\n",
488 tmpex.l_extent.start, lock->l_policy_data.l_extent.end,
493 static int ll_extent_lock_callback(struct ldlm_lock *lock,
494 struct ldlm_lock_desc *new, void *data,
497 struct lustre_handle lockh = { 0 };
501 if ((unsigned long)data > 0 && (unsigned long)data < 0x1000) {
502 LDLM_ERROR(lock, "cancelling lock with bad data %p", data);
507 case LDLM_CB_BLOCKING:
508 ldlm_lock2handle(lock, &lockh);
509 rc = ldlm_cli_cancel(&lockh);
511 CERROR("ldlm_cli_cancel failed: %d\n", rc);
513 case LDLM_CB_CANCELING: {
515 struct ll_inode_info *lli;
516 struct lov_stripe_md *lsm;
520 /* This lock wasn't granted, don't try to evict pages */
521 if (lock->l_req_mode != lock->l_granted_mode)
524 inode = ll_inode_from_lock(lock);
527 lli = ll_i2info(inode);
530 if (lli->lli_smd == NULL)
534 stripe = ll_lock_to_stripe_offset(inode, lock);
537 ll_pgcache_remove_extent(inode, lsm, lock, stripe);
539 /* grabbing the i_sem will wait for write() to complete. ns
540 * lock hold times should be very short as ast processing
541 * requires them and has a short timeout. so, i_sem before ns
545 l_lock(&lock->l_resource->lr_namespace->ns_lock);
546 kms = ldlm_extent_shift_kms(lock,
547 lsm->lsm_oinfo[stripe].loi_kms);
549 if (lsm->lsm_oinfo[stripe].loi_kms != kms)
550 LDLM_DEBUG(lock, "updating kms from "LPU64" to "LPU64,
551 lsm->lsm_oinfo[stripe].loi_kms, kms);
552 lsm->lsm_oinfo[stripe].loi_kms = kms;
553 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
555 //ll_try_done_writing(inode);
568 int ll_async_completion_ast(struct ldlm_lock *lock, int flags, void *data)
570 /* XXX ALLOCATE - 160 bytes */
571 struct inode *inode = ll_inode_from_lock(lock);
572 struct ll_inode_info *lli = ll_i2info(inode);
573 struct lustre_handle lockh = { 0 };
578 if (flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED |
579 LDLM_FL_BLOCK_CONV)) {
580 LBUG(); /* not expecting any blocked async locks yet */
581 LDLM_DEBUG(lock, "client-side async enqueue returned a blocked "
583 ldlm_lock_dump(D_OTHER, lock, 0);
584 ldlm_reprocess_all(lock->l_resource);
588 LDLM_DEBUG(lock, "client-side async enqueue: granted/glimpsed");
590 stripe = ll_lock_to_stripe_offset(inode, lock);
594 if (lock->l_lvb_len) {
595 struct lov_stripe_md *lsm = lli->lli_smd;
597 lvb = lock->l_lvb_data;
598 lsm->lsm_oinfo[stripe].loi_rss = lvb->lvb_size;
600 l_lock(&lock->l_resource->lr_namespace->ns_lock);
602 kms = MAX(lsm->lsm_oinfo[stripe].loi_kms, lvb->lvb_size);
603 kms = ldlm_extent_shift_kms(NULL, kms);
604 if (lsm->lsm_oinfo[stripe].loi_kms != kms)
605 LDLM_DEBUG(lock, "updating kms from "LPU64" to "LPU64,
606 lsm->lsm_oinfo[stripe].loi_kms, kms);
607 lsm->lsm_oinfo[stripe].loi_kms = kms;
609 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
614 wake_up(&lock->l_waitq);
616 ldlm_lock2handle(lock, &lockh);
617 ldlm_lock_decref(&lockh, LCK_PR);
622 static int ll_glimpse_callback(struct ldlm_lock *lock, void *reqp)
624 struct ptlrpc_request *req = reqp;
625 struct inode *inode = ll_inode_from_lock(lock);
626 struct ll_inode_info *lli;
628 struct lov_stripe_md *lsm;
629 int rc, size = sizeof(*lvb), stripe;
633 GOTO(out, rc = -ELDLM_NO_LOCK_DATA);
634 lli = ll_i2info(inode);
636 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
640 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
642 /* First, find out which stripe index this lock corresponds to. */
643 stripe = ll_lock_to_stripe_offset(inode, lock);
645 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
647 rc = lustre_pack_reply(req, 1, &size, NULL);
649 CERROR("lustre_pack_reply: %d\n", rc);
653 lvb = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*lvb));
654 lvb->lvb_size = lli->lli_smd->lsm_oinfo[stripe].loi_kms;
656 LDLM_DEBUG(lock, "i_size: %llu -> stripe number %u -> kms "LPU64,
657 inode->i_size, stripe, lvb->lvb_size);
663 /* These errors are normal races, so we don't want to fill the console
664 * with messages by calling ptlrpc_error() */
665 if (rc == -ELDLM_NO_LOCK_DATA)
666 lustre_pack_reply(req, 0, NULL, NULL);
672 __u64 lov_merge_size(struct lov_stripe_md *lsm, int kms);
673 __u64 lov_merge_blocks(struct lov_stripe_md *lsm);
674 __u64 lov_merge_mtime(struct lov_stripe_md *lsm, __u64 current_time);
676 /* NB: lov_merge_size will prefer locally cached writes if they extend the
677 * file (because it prefers KMS over RSS when larger) */
678 int ll_glimpse_size(struct inode *inode)
680 struct ll_inode_info *lli = ll_i2info(inode);
681 struct ll_sb_info *sbi = ll_i2sbi(inode);
682 ldlm_policy_data_t policy = { .l_extent = { 0, OBD_OBJECT_EOF } };
683 struct lustre_handle lockh = { 0 };
684 int rc, flags = LDLM_FL_HAS_INTENT;
687 CDEBUG(D_DLMTRACE, "Glimpsing inode %lu\n", inode->i_ino);
689 rc = obd_enqueue(sbi->ll_dt_exp, lli->lli_smd, LDLM_EXTENT, &policy,
690 LCK_PR, &flags, ll_extent_lock_callback,
691 ldlm_completion_ast, ll_glimpse_callback, inode,
692 sizeof(struct ost_lvb), lustre_swab_ost_lvb, &lockh);
697 CERROR("obd_enqueue returned rc %d, returning -EIO\n", rc);
698 RETURN(rc > 0 ? -EIO : rc);
701 inode->i_size = lov_merge_size(lli->lli_smd, 0);
702 inode->i_blocks = lov_merge_blocks(lli->lli_smd);
703 //inode->i_mtime = lov_merge_mtime(lli->lli_smd, inode->i_mtime);
705 CDEBUG(D_DLMTRACE, "glimpse: size: "LPU64", blocks: "LPU64"\n",
706 (__u64)inode->i_size, (__u64)inode->i_blocks);
707 obd_cancel(sbi->ll_dt_exp, lli->lli_smd, LCK_PR, &lockh);
711 void ll_stime_record(struct ll_sb_info *sbi, struct timeval *start,
712 struct obd_service_time *stime)
715 do_gettimeofday(&stop);
717 spin_lock(&sbi->ll_lock);
718 lprocfs_stime_record(stime, &stop, start);
719 spin_unlock(&sbi->ll_lock);
722 int ll_extent_lock(struct ll_file_data *fd, struct inode *inode,
723 struct lov_stripe_md *lsm, int mode,
724 ldlm_policy_data_t *policy, struct lustre_handle *lockh,
725 int ast_flags, struct obd_service_time *stime)
727 struct ll_sb_info *sbi = ll_i2sbi(inode);
728 struct timeval start;
732 LASSERT(lockh->cookie == 0);
734 /* XXX phil: can we do this? won't it screw the file size up? */
735 if ((fd && (fd->fd_flags & LL_FILE_IGNORE_LOCK)) ||
736 (sbi->ll_flags & LL_SBI_NOLCK))
739 CDEBUG(D_DLMTRACE, "Locking inode %lu, start "LPU64" end "LPU64"\n",
740 inode->i_ino, policy->l_extent.start, policy->l_extent.end);
742 do_gettimeofday(&start);
743 rc = obd_enqueue(sbi->ll_dt_exp, lsm, LDLM_EXTENT, policy, mode,
744 &ast_flags, ll_extent_lock_callback,
745 ldlm_completion_ast, ll_glimpse_callback, inode,
746 sizeof(struct ost_lvb), lustre_swab_ost_lvb, lockh);
750 ll_stime_record(sbi, &start, stime);
752 if (policy->l_extent.start == 0 &&
753 policy->l_extent.end == OBD_OBJECT_EOF) {
754 /* vmtruncate()->ll_truncate() first sets the i_size and then
755 * the kms under both a DLM lock and the i_sem. If we don't
756 * get the i_sem here we can match the DLM lock and reset
757 * i_size from the kms before the truncating path has updated
758 * the kms. generic_file_write can then trust the stale i_size
759 * when doing appending writes and effectively cancel the
760 * result of the truncate. Getting the i_sem after the enqueue
761 * maintains the DLM -> i_sem acquiry order. */
763 inode->i_size = lov_merge_size(lsm, 1);
766 //inode->i_mtime = lov_merge_mtime(lsm, inode->i_mtime);
771 int ll_extent_unlock(struct ll_file_data *fd, struct inode *inode,
772 struct lov_stripe_md *lsm, int mode,
773 struct lustre_handle *lockh)
775 struct ll_sb_info *sbi = ll_i2sbi(inode);
779 /* XXX phil: can we do this? won't it screw the file size up? */
780 if ((fd && (fd->fd_flags & LL_FILE_IGNORE_LOCK)) ||
781 (sbi->ll_flags & LL_SBI_NOLCK))
784 rc = obd_cancel(sbi->ll_dt_exp, lsm, mode, lockh);
789 static ssize_t ll_file_read(struct file *file, char *buf, size_t count,
792 struct inode *inode = file->f_dentry->d_inode;
793 struct ll_inode_info *lli = ll_i2info(inode);
794 struct lov_stripe_md *lsm = lli->lli_smd;
795 struct ll_lock_tree tree;
796 struct ll_lock_tree_node *node;
801 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),size="LPSZ",offset=%Ld\n",
802 inode->i_ino, inode->i_generation, inode, count, *ppos);
804 /* "If nbyte is 0, read() will return 0 and have no other results."
805 * -- Single Unix Spec */
809 lprocfs_counter_add(ll_i2sbi(inode)->ll_stats, LPROC_LL_READ_BYTES,
815 node = ll_node_from_inode(inode, *ppos, *ppos + count - 1,
818 tree.lt_fd = file->private_data;
820 rc = ll_tree_lock(&tree, node, inode, buf, count,
821 file->f_flags & O_NONBLOCK ? LDLM_FL_BLOCK_NOWAIT :0);
825 kms = lov_merge_size(lsm, 1);
826 if (*ppos + count - 1 > kms) {
827 /* A glimpse is necessary to determine whether we return a short
828 * read or some zeroes at the end of the buffer */
829 retval = ll_glimpse_size(inode);
836 CDEBUG(D_INFO, "Read ino %lu, "LPSZ" bytes, offset %lld, i_size %llu\n",
837 inode->i_ino, count, *ppos, inode->i_size);
839 /* turn off the kernel's read-ahead */
840 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
843 file->f_ra.ra_pages = 0;
845 retval = generic_file_read(file, buf, count, ppos);
848 ll_tree_unlock(&tree, inode);
853 * Write to a file (through the page cache).
855 static ssize_t ll_file_write(struct file *file, const char *buf,
856 size_t count, loff_t *ppos)
858 struct inode *inode = file->f_dentry->d_inode;
859 loff_t maxbytes = ll_file_maxbytes(inode);
860 struct ll_lock_tree tree;
861 struct ll_lock_tree_node *node;
866 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),size="LPSZ",offset=%Ld\n",
867 inode->i_ino, inode->i_generation, inode, count, *ppos);
869 SIGNAL_MASK_ASSERT(); /* XXX BUG 1511 */
871 /* POSIX, but surprised the VFS doesn't check this already */
875 /* If file was opened for LL_IOC_LOV_SETSTRIPE but the ioctl wasn't
876 * called on the file, don't fail the below assertion (bug 2388). */
877 if (file->f_flags & O_LOV_DELAY_CREATE &&
878 ll_i2info(inode)->lli_smd == NULL)
881 LASSERT(ll_i2info(inode)->lli_smd != NULL);
883 if (file->f_flags & O_APPEND)
884 node = ll_node_from_inode(inode, 0, OBD_OBJECT_EOF, LCK_PW);
886 node = ll_node_from_inode(inode, *ppos, *ppos + count - 1,
890 RETURN(PTR_ERR(node));
892 tree.lt_fd = file->private_data;
894 rc = ll_tree_lock(&tree, node, inode, buf, count,
895 file->f_flags & O_NONBLOCK ? LDLM_FL_BLOCK_NOWAIT :0);
899 /* this is ok, g_f_w will overwrite this under i_sem if it races
900 * with a local truncate, it just makes our maxbyte checking easier */
901 if (file->f_flags & O_APPEND)
902 *ppos = inode->i_size;
904 if (*ppos >= maxbytes) {
905 if (count || *ppos > maxbytes) {
906 send_sig(SIGXFSZ, current, 0);
907 GOTO(out, retval = -EFBIG);
910 if (*ppos + count > maxbytes)
911 count = maxbytes - *ppos;
913 CDEBUG(D_INFO, "Writing inode %lu, "LPSZ" bytes, offset %Lu\n",
914 inode->i_ino, count, *ppos);
916 /* generic_file_write handles O_APPEND after getting i_sem */
917 retval = generic_file_write(file, buf, count, ppos);
920 ll_tree_unlock(&tree, inode);
921 /* serialize with mmap/munmap/mremap */
922 lprocfs_counter_add(ll_i2sbi(inode)->ll_stats, LPROC_LL_WRITE_BYTES,
923 retval > 0 ? retval : 0);
927 static int ll_lov_recreate_obj(struct inode *inode, struct file *file,
930 struct ll_inode_info *lli = ll_i2info(inode);
931 struct obd_export *exp = ll_i2dtexp(inode);
932 struct ll_recreate_obj ucreatp;
933 struct obd_trans_info oti = { 0 };
934 struct obdo *oa = NULL;
937 struct lov_stripe_md *lsm, *lsm2;
940 if (!capable (CAP_SYS_ADMIN))
943 rc = copy_from_user(&ucreatp, (struct ll_recreate_obj *)arg,
944 sizeof(struct ll_recreate_obj));
952 down(&lli->lli_open_sem);
955 GOTO(out, rc = -ENOENT);
956 lsm_size = sizeof(*lsm) + (sizeof(struct lov_oinfo) *
957 (lsm->lsm_stripe_count));
959 OBD_ALLOC(lsm2, lsm_size);
961 GOTO(out, rc = -ENOMEM);
963 oa->o_id = ucreatp.lrc_id;
964 oa->o_nlink = ucreatp.lrc_ost_idx;
965 oa->o_gr = ucreatp.lrc_group;
966 oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP | OBD_MD_FLFLAGS;
967 oa->o_flags |= OBD_FL_RECREATE_OBJS;
968 obdo_from_inode(oa, inode, OBD_MD_FLTYPE | OBD_MD_FLATIME |
969 OBD_MD_FLMTIME | OBD_MD_FLCTIME);
971 oti.oti_objid = NULL;
972 memcpy(lsm2, lsm, lsm_size);
973 rc = obd_create(exp, oa, &lsm2, &oti);
975 OBD_FREE(lsm2, lsm_size);
978 up(&lli->lli_open_sem);
983 static int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
984 int flags, struct lov_user_md *lum,
987 struct ll_inode_info *lli = ll_i2info(inode);
989 struct obd_export *exp = ll_i2dtexp(inode);
990 struct lov_stripe_md *lsm;
991 struct lookup_intent oit = {.it_op = IT_OPEN, .it_flags = flags};
992 struct ptlrpc_request *req = NULL;
997 down(&lli->lli_open_sem);
1000 up(&lli->lli_open_sem);
1001 CDEBUG(D_IOCTL, "stripe already exists for ino %lu\n",
1006 f = get_empty_filp();
1010 f->f_dentry = file->f_dentry;
1011 f->f_vfsmnt = file->f_vfsmnt;
1013 rc = ll_intent_file_open(f, lum, lum_size, &oit);
1016 if (it_disposition(&oit, DISP_LOOKUP_NEG))
1018 req = oit.d.lustre.it_data;
1019 rc = oit.d.lustre.it_status;
1024 rc = mdc_req2lustre_md(ll_i2mdexp(inode), req, 1, exp, &md);
1027 ll_update_inode(f->f_dentry->d_inode, &md);
1029 rc = ll_local_open(f, &oit);
1032 ll_intent_release(&oit);
1034 rc = ll_file_release(f->f_dentry->d_inode, f);
1039 up(&lli->lli_open_sem);
1041 ptlrpc_req_finished(req);
1045 static int ll_lov_setea(struct inode *inode, struct file *file,
1048 int flags = MDS_OPEN_HAS_OBJS | FMODE_WRITE;
1049 struct lov_user_md *lump;
1050 int lum_size = sizeof(struct lov_user_md) +
1051 sizeof(struct lov_user_ost_data);
1055 if (!capable (CAP_SYS_ADMIN))
1058 OBD_ALLOC(lump, lum_size);
1062 rc = copy_from_user(lump, (struct lov_user_md *)arg, lum_size);
1064 OBD_FREE(lump, lum_size);
1068 rc = ll_lov_setstripe_ea_info(inode, file, flags, lump, lum_size);
1070 OBD_FREE(lump, lum_size);
1074 static int ll_lov_setstripe(struct inode *inode, struct file *file,
1077 struct lov_user_md lum, *lump = (struct lov_user_md *)arg;
1079 int flags = FMODE_WRITE;
1082 /* Bug 1152: copy properly when this is no longer true */
1083 LASSERT(sizeof(lum) == sizeof(*lump));
1084 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lump->lmm_objects[0]));
1085 rc = copy_from_user(&lum, lump, sizeof(lum));
1089 rc = ll_lov_setstripe_ea_info(inode, file, flags, &lum, sizeof(lum));
1091 put_user(0, &lump->lmm_stripe_count);
1092 rc = obd_iocontrol(LL_IOC_LOV_GETSTRIPE, ll_i2dtexp(inode),
1093 0, ll_i2info(inode)->lli_smd, lump);
1098 static int ll_lov_getstripe(struct inode *inode, unsigned long arg)
1100 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1105 return obd_iocontrol(LL_IOC_LOV_GETSTRIPE, ll_i2dtexp(inode), 0, lsm,
1109 static int ll_get_grouplock(struct inode *inode, struct file *file,
1112 struct ll_file_data *fd = file->private_data;
1113 ldlm_policy_data_t policy = { .l_extent = { .start = 0,
1114 .end = OBD_OBJECT_EOF}};
1115 struct lustre_handle lockh = { 0 };
1116 struct ll_inode_info *lli = ll_i2info(inode);
1117 struct lov_stripe_md *lsm = lli->lli_smd;
1121 if (fd->fd_flags & LL_FILE_GROUP_LOCKED) {
1125 policy.l_extent.gid = arg;
1126 if (file->f_flags & O_NONBLOCK)
1127 flags = LDLM_FL_BLOCK_NOWAIT;
1129 rc = ll_extent_lock(fd, inode, lsm, LCK_GROUP, &policy, &lockh, flags,
1130 &ll_i2sbi(inode)->ll_grouplock_stime);
1134 fd->fd_flags |= LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK;
1136 memcpy(&fd->fd_cwlockh, &lockh, sizeof(lockh));
1141 static int ll_put_grouplock(struct inode *inode, struct file *file,
1144 struct ll_file_data *fd = file->private_data;
1145 struct ll_inode_info *lli = ll_i2info(inode);
1146 struct lov_stripe_md *lsm = lli->lli_smd;
1150 if (!(fd->fd_flags & LL_FILE_GROUP_LOCKED)) {
1151 /* Ugh, it's already unlocked. */
1155 if (fd->fd_gid != arg) /* Ugh? Unlocking with different gid? */
1158 fd->fd_flags &= ~(LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK);
1160 rc = ll_extent_unlock(fd, inode, lsm, LCK_GROUP, &fd->fd_cwlockh);
1165 memset(&fd->fd_cwlockh, 0, sizeof(fd->fd_cwlockh));
1170 int ll_file_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
1173 struct ll_file_data *fd = file->private_data;
1174 struct ll_sb_info *sbi = ll_i2sbi(inode);
1178 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),cmd=%x\n", inode->i_ino,
1179 inode->i_generation, inode, cmd);
1181 if (_IOC_TYPE(cmd) == 'T') /* tty ioctls */
1184 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_IOCTL);
1186 case LL_IOC_GETFLAGS:
1187 /* Get the current value of the file flags */
1188 return put_user(fd->fd_flags, (int *)arg);
1189 case LL_IOC_SETFLAGS:
1190 case LL_IOC_CLRFLAGS:
1191 /* Set or clear specific file flags */
1192 /* XXX This probably needs checks to ensure the flags are
1193 * not abused, and to handle any flag side effects.
1195 if (get_user(flags, (int *) arg))
1198 if (cmd == LL_IOC_SETFLAGS)
1199 fd->fd_flags |= flags;
1201 fd->fd_flags &= ~flags;
1203 case LL_IOC_LOV_SETSTRIPE:
1204 RETURN(ll_lov_setstripe(inode, file, arg));
1205 case LL_IOC_LOV_SETEA:
1206 RETURN(ll_lov_setea(inode, file, arg));
1207 case IOC_MDC_SHOWFID: {
1208 struct lustre_id *idp = (struct lustre_id *)arg;
1209 struct lustre_id id;
1213 filename = getname((const char *)arg);
1214 if (IS_ERR(filename))
1215 RETURN(PTR_ERR(filename));
1217 ll_inode2id(&id, inode);
1219 rc = ll_get_fid(sbi->ll_md_exp, &id, filename, &id);
1221 GOTO(out_filename, rc);
1223 rc = copy_to_user(idp, &id, sizeof(*idp));
1225 GOTO(out_filename, rc = -EFAULT);
1232 case LL_IOC_LOV_GETSTRIPE:
1233 RETURN(ll_lov_getstripe(inode, arg));
1234 case LL_IOC_RECREATE_OBJ:
1235 RETURN(ll_lov_recreate_obj(inode, file, arg));
1236 case EXT3_IOC_GETFLAGS:
1237 case EXT3_IOC_SETFLAGS:
1238 RETURN( ll_iocontrol(inode, file, cmd, arg) );
1239 case LL_IOC_GROUP_LOCK:
1240 RETURN(ll_get_grouplock(inode, file, arg));
1241 case LL_IOC_GROUP_UNLOCK:
1242 RETURN(ll_put_grouplock(inode, file, arg));
1243 case EXT3_IOC_GETVERSION_OLD:
1244 case EXT3_IOC_GETVERSION:
1245 return put_user(inode->i_generation, (int *) arg);
1246 /* We need to special case any other ioctls we want to handle,
1247 * to send them to the MDS/OST as appropriate and to properly
1248 * network encode the arg field.
1249 case EXT2_IOC_GETVERSION_OLD:
1250 case EXT2_IOC_GETVERSION_NEW:
1251 case EXT2_IOC_SETVERSION_OLD:
1252 case EXT2_IOC_SETVERSION_NEW:
1253 case EXT3_IOC_SETVERSION_OLD:
1254 case EXT3_IOC_SETVERSION:
1257 RETURN( obd_iocontrol(cmd, ll_i2dtexp(inode), 0, NULL,
1262 loff_t ll_file_seek(struct file *file, loff_t offset, int origin)
1264 struct inode *inode = file->f_dentry->d_inode;
1265 struct ll_file_data *fd = file->private_data;
1266 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1267 struct lustre_handle lockh = {0};
1270 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),to=%llu\n", inode->i_ino,
1271 inode->i_generation, inode,
1272 offset + ((origin==2) ? inode->i_size : file->f_pos));
1274 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_LLSEEK);
1275 if (origin == 2) { /* SEEK_END */
1276 ldlm_policy_data_t policy = { .l_extent = {0, OBD_OBJECT_EOF }};
1277 int nonblock = 0, rc;
1279 if (file->f_flags & O_NONBLOCK)
1280 nonblock = LDLM_FL_BLOCK_NOWAIT;
1282 rc = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
1283 nonblock, &ll_i2sbi(inode)->ll_seek_stime);
1287 offset += inode->i_size;
1288 } else if (origin == 1) { /* SEEK_CUR */
1289 offset += file->f_pos;
1293 if (offset >= 0 && offset <= ll_file_maxbytes(inode)) {
1294 if (offset != file->f_pos) {
1295 file->f_pos = offset;
1296 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
1298 file->f_version = ++event;
1305 ll_extent_unlock(fd, inode, lsm, LCK_PR, &lockh);
1309 int ll_fsync(struct file *file, struct dentry *dentry, int data)
1311 struct inode *inode = dentry->d_inode;
1312 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1313 struct lustre_id id;
1314 struct ptlrpc_request *req;
1317 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n", inode->i_ino,
1318 inode->i_generation, inode);
1320 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_FSYNC);
1322 /* fsync's caller has already called _fdata{sync,write}, we want
1323 * that IO to finish before calling the osc and mdc sync methods */
1324 rc = filemap_fdatawait(inode->i_mapping);
1326 ll_inode2id(&id, inode);
1327 err = md_sync(ll_i2sbi(inode)->ll_md_exp, &id, &req);
1331 ptlrpc_req_finished(req);
1334 struct obdo *oa = obdo_alloc();
1337 RETURN(rc ? rc : -ENOMEM);
1339 oa->o_id = lsm->lsm_object_id;
1340 oa->o_gr = lsm->lsm_object_gr;
1341 oa->o_valid = OBD_MD_FLID;
1342 obdo_from_inode(oa, inode, (OBD_MD_FLTYPE | OBD_MD_FLATIME |
1343 OBD_MD_FLMTIME | OBD_MD_FLCTIME |
1346 err = obd_sync(ll_i2sbi(inode)->ll_dt_exp, oa, lsm,
1356 int ll_file_flock(struct file *file, int cmd, struct file_lock *file_lock)
1358 struct inode *inode = file->f_dentry->d_inode;
1359 struct ll_inode_info *li = ll_i2info(inode);
1360 struct ll_sb_info *sbi = ll_i2sbi(inode);
1361 struct obd_device *obddev;
1362 struct ldlm_res_id res_id =
1363 { .name = {id_fid(&li->lli_id), id_group(&li->lli_id), LDLM_FLOCK} };
1364 struct lustre_handle lockh = {0};
1365 ldlm_policy_data_t flock;
1366 ldlm_mode_t mode = 0;
1371 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu file_lock=%p\n",
1372 inode->i_ino, file_lock);
1374 flock.l_flock.pid = file_lock->fl_pid;
1375 flock.l_flock.start = file_lock->fl_start;
1376 flock.l_flock.end = file_lock->fl_end;
1378 switch (file_lock->fl_type) {
1383 /* An unlock request may or may not have any relation to
1384 * existing locks so we may not be able to pass a lock handle
1385 * via a normal ldlm_lock_cancel() request. The request may even
1386 * unlock a byte range in the middle of an existing lock. In
1387 * order to process an unlock request we need all of the same
1388 * information that is given with a normal read or write record
1389 * lock request. To avoid creating another ldlm unlock (cancel)
1390 * message we'll treat a LCK_NL flock request as an unlock. */
1397 CERROR("unknown fcntl lock type: %d\n", file_lock->fl_type);
1412 flags = LDLM_FL_BLOCK_NOWAIT;
1418 flags = LDLM_FL_TEST_LOCK;
1419 /* Save the old mode so that if the mode in the lock changes we
1420 * can decrement the appropriate reader or writer refcount. */
1421 file_lock->fl_type = mode;
1424 CERROR("unknown fcntl lock command: %d\n", cmd);
1428 CDEBUG(D_DLMTRACE, "inode=%lu, pid="LPU64", flags=%#x, mode=%u, "
1429 "start="LPU64", end="LPU64"\n", inode->i_ino, flock.l_flock.pid,
1430 flags, mode, flock.l_flock.start, flock.l_flock.end);
1432 obddev = md_get_real_obd(sbi->ll_md_exp, &li->lli_id);
1433 rc = ldlm_cli_enqueue(obddev->obd_self_export, NULL,
1434 obddev->obd_namespace,
1435 res_id, LDLM_FLOCK, &flock, mode, &flags,
1436 NULL, ldlm_flock_completion_ast, NULL, file_lock,
1437 NULL, 0, NULL, &lockh);
1441 int ll_inode_revalidate_it(struct dentry *dentry, struct lookup_intent *it)
1443 struct lookup_intent oit = { .it_op = IT_GETATTR };
1444 struct inode *inode = dentry->d_inode;
1445 struct ptlrpc_request *req = NULL;
1446 struct ll_inode_info *lli;
1447 struct lov_stripe_md *lsm;
1448 struct ll_sb_info *sbi;
1449 struct lustre_id id;
1455 CERROR("REPORT THIS LINE TO PETER\n");
1459 sbi = ll_i2sbi(inode);
1461 ll_inode2id(&id, inode);
1462 lli = ll_i2info(inode);
1463 LASSERT(id_fid(&id) != 0);
1465 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), name=%s, intent=%s\n",
1466 inode->i_ino, inode->i_generation, inode, dentry->d_name.name,
1469 #if (LINUX_VERSION_CODE <= KERNEL_VERSION(2,5,0))
1470 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_REVALIDATE);
1473 rc = md_intent_lock(sbi->ll_md_exp, &id, NULL, 0, NULL, 0, &id,
1474 &oit, 0, &req, ll_mdc_blocking_ast);
1478 rc = revalidate_it_finish(req, 1, &oit, dentry);
1480 ll_intent_release(&oit);
1484 ll_lookup_finish_locks(&oit, dentry);
1487 if (lsm == NULL) /* object not yet allocated, don't validate size */
1491 * ll_glimpse_size() will prefer locally cached writes if they extend
1494 rc = ll_glimpse_size(inode);
1498 ptlrpc_req_finished(req);
1502 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
1503 int ll_getattr(struct vfsmount *mnt, struct dentry *de,
1504 struct lookup_intent *it, struct kstat *stat)
1507 struct inode *inode = de->d_inode;
1509 res = ll_inode_revalidate_it(de, it);
1510 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_GETATTR);
1515 stat->ino = inode->i_ino;
1516 stat->mode = inode->i_mode;
1517 stat->nlink = inode->i_nlink;
1518 stat->uid = inode->i_uid;
1519 stat->gid = inode->i_gid;
1520 stat->atime = inode->i_atime;
1521 stat->mtime = inode->i_mtime;
1522 stat->ctime = inode->i_ctime;
1523 stat->size = inode->i_size;
1524 stat->blksize = inode->i_blksize;
1525 stat->blocks = inode->i_blocks;
1526 stat->rdev = kdev_t_to_nr(inode->i_rdev);
1527 stat->dev = id_group(&ll_i2info(inode)->lli_id);
1532 struct file_operations ll_file_operations = {
1533 .read = ll_file_read,
1534 .write = ll_file_write,
1535 .ioctl = ll_file_ioctl,
1536 .open = ll_file_open,
1537 .release = ll_file_release,
1538 .mmap = ll_file_mmap,
1539 .llseek = ll_file_seek,
1540 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
1541 .sendfile = generic_file_sendfile,
1544 .lock = ll_file_flock
1547 struct inode_operations ll_file_inode_operations = {
1548 .setattr_raw = ll_setattr_raw,
1549 .setattr = ll_setattr,
1550 .truncate = ll_truncate,
1551 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
1552 .getattr_it = ll_getattr,
1554 .revalidate_it = ll_inode_revalidate_it,