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 #include <linux/lustre_acl.h>
31 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
32 #include <linux/lustre_compat25.h>
34 #include "llite_internal.h"
35 #include <linux/obd_lov.h>
37 __u64 lov_merge_size(struct lov_stripe_md *lsm, int kms);
38 __u64 lov_merge_blocks(struct lov_stripe_md *lsm);
39 __u64 lov_merge_mtime(struct lov_stripe_md *lsm, __u64 current_time);
41 int ll_validate_size(struct inode *inode, __u64 *size, __u64 *blocks)
43 ldlm_policy_data_t extent = { .l_extent = { 0, OBD_OBJECT_EOF } };
44 struct obd_export *exp = ll_i2sbi(inode)->ll_dt_exp;
45 struct ll_inode_info *lli = ll_i2info(inode);
46 struct lustre_handle match_lockh = {0};
50 if (lli->lli_smd == NULL)
53 LASSERT(size != NULL && blocks != NULL);
55 flags = LDLM_FL_TEST_LOCK | LDLM_FL_CBPENDING | LDLM_FL_BLOCK_GRANTED;
56 rc = obd_match(exp, lli->lli_smd, LDLM_EXTENT, &extent,
57 LCK_PR | LCK_PW, &flags, inode, &match_lockh);
59 /* we have no all needed locks,
60 * so we don't know actual size */
64 /* we know actual size! */
65 down(&lli->lli_size_sem);
66 *size = lov_merge_size(lli->lli_smd, 0);
67 *blocks = lov_merge_blocks(lli->lli_smd);
68 up(&lli->lli_size_sem);
74 int ll_md_och_close(struct obd_export *md_exp, struct inode *inode,
75 struct obd_client_handle *och, int dirty)
77 struct ptlrpc_request *req = NULL;
78 struct obdo *obdo = NULL;
79 struct obd_device *obd;
83 obd = class_exp2obd(md_exp);
85 CERROR("Invalid MDC connection handle "LPX64"\n",
86 md_exp->exp_handle.h_cookie);
92 * here we check if this is forced umount. If so this is called on
93 * canceling "open lock" and we do not call md_close() in this case , as
94 * it will not successful, as import is already deactivated.
96 if (obd->obd_no_recov)
99 /* closing opened file */
104 obdo->o_id = inode->i_ino;
105 obdo->o_generation = inode->i_generation;
106 obdo->o_valid = OBD_MD_FLID;
107 obdo_from_inode(obdo, inode, (OBD_MD_FLTYPE | OBD_MD_FLMODE |
108 OBD_MD_FLATIME | OBD_MD_FLMTIME |
110 if (0 /* ll_is_inode_dirty(inode) */) {
111 obdo->o_flags = MDS_BFLAG_UNCOMMITTED_WRITES;
112 obdo->o_valid |= OBD_MD_FLFLAGS;
114 obdo->o_fid = id_fid(&ll_i2info(inode)->lli_id);
115 obdo->o_mds = id_group(&ll_i2info(inode)->lli_id);
118 obdo->o_valid |= OBD_MD_FLEPOCH;
119 obdo->o_easize = ll_i2info(inode)->lli_io_epoch;
122 /* we modified data through this handle */
123 obdo->o_flags |= MDS_BFLAG_DIRTY_EPOCH;
124 obdo->o_valid |= OBD_MD_FLFLAGS;
125 if (ll_validate_size(inode, &obdo->o_size, &obdo->o_blocks))
126 obdo->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
129 rc = md_close(md_exp, obdo, och, &req);
134 * we are the last writer, so the MDS has instructed us to get
135 * the file size and any write cookies, then close again.
138 //ll_queue_done_writing(inode);
141 CERROR("inode %lu mdc close failed: rc = %d\n",
142 (unsigned long)inode->i_ino, rc);
145 ptlrpc_req_finished(req);
148 mdc_clear_open_replay_data(md_exp, och);
149 och->och_fh.cookie = DEAD_HANDLE_MAGIC;
150 OBD_FREE(och, sizeof *och);
154 int ll_md_real_close(struct obd_export *md_exp,
155 struct inode *inode, int flags)
157 struct ll_inode_info *lli = ll_i2info(inode);
158 int freeing = inode->i_state & I_FREEING;
159 struct obd_client_handle **och_p;
160 struct obd_client_handle *och;
162 int rc = 0, dirty = 0;
165 if (flags & FMODE_WRITE) {
166 och_p = &lli->lli_mds_write_och;
167 och_usecount = &lli->lli_open_fd_write_count;
168 } else if (flags & FMODE_EXEC) {
169 och_p = &lli->lli_mds_exec_och;
170 och_usecount = &lli->lli_open_fd_exec_count;
172 och_p = &lli->lli_mds_read_och;
173 och_usecount = &lli->lli_open_fd_read_count;
176 down(&lli->lli_och_sem);
177 if (*och_usecount) { /* There are still users of this handle, so
179 up(&lli->lli_och_sem);
182 if (ll_is_inode_dirty(inode)) {
183 /* the inode still has dirty pages, let's close later */
184 CDEBUG(D_INODE, "inode %lu/%u still has dirty pages\n",
185 inode->i_ino, inode->i_generation);
186 LASSERT(freeing == 0);
187 ll_queue_done_writing(inode);
188 up(&lli->lli_och_sem);
192 if (LLI_DIRTY_HANDLE(inode) && (flags & FMODE_WRITE)) {
193 clear_bit(LLI_F_DIRTY_HANDLE, &lli->lli_flags);
195 } else if (0 && !(flags & FMODE_SYNC) && !freeing) {
196 /* in order to speed up creation rate we pass
197 * closing to dedicated thread so we don't need
198 * to wait for close reply here -bzzz */
199 ll_queue_done_writing(inode);
200 up(&lli->lli_och_sem);
208 up(&lli->lli_och_sem);
211 * there might be a race and somebody have freed this och
212 * already. Another way to have this twice called is if file closing
213 * will fail due to netwok problems and on umount lock will be canceled
214 * and this will be called from block_ast callack.
216 if (och && och->och_fh.cookie != DEAD_HANDLE_MAGIC)
217 rc = ll_md_och_close(md_exp, inode, och, dirty);
222 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
225 struct ll_file_data *fd = file->private_data;
226 struct ll_inode_info *lli = ll_i2info(inode);
230 /* clear group lock, if present */
231 if (fd->fd_flags & LL_FILE_GROUP_LOCKED) {
232 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
233 fd->fd_flags &= ~(LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK);
234 rc = ll_extent_unlock(fd, inode, lsm, LCK_GROUP,
238 /* Let's see if we have good enough OPEN lock on the file and if
239 we can skip talking to MDS */
240 if (file->f_dentry->d_inode) {
242 struct obd_device *obddev;
243 struct lustre_handle lockh;
244 int flags = LDLM_FL_BLOCK_GRANTED;
245 struct ldlm_res_id file_res_id = {.name = {id_fid(&lli->lli_id),
246 id_group(&lli->lli_id)}};
247 ldlm_policy_data_t policy = {.l_inodebits={MDS_INODELOCK_OPEN}};
249 down(&lli->lli_och_sem);
250 if (fd->fd_omode & FMODE_WRITE) {
252 LASSERT(lli->lli_open_fd_write_count);
253 lli->lli_open_fd_write_count--;
254 } else if (fd->fd_omode & FMODE_EXEC) {
256 LASSERT(lli->lli_open_fd_exec_count);
257 lli->lli_open_fd_exec_count--;
260 LASSERT(lli->lli_open_fd_read_count);
261 lli->lli_open_fd_read_count--;
263 up(&lli->lli_och_sem);
265 obddev = md_get_real_obd(md_exp, &lli->lli_id);
266 if (!ldlm_lock_match(obddev->obd_namespace, flags, &file_res_id,
267 LDLM_IBITS, &policy, lockmode, &lockh))
269 rc = ll_md_real_close(md_exp, file->f_dentry->d_inode,
272 ldlm_lock_decref(&lockh, lockmode);
276 file->private_data = NULL;
277 OBD_SLAB_FREE(fd, ll_file_data_slab, sizeof(*fd));
281 /* While this returns an error code, fput() the caller does not, so we need
282 * to make every effort to clean up all of our state here. Also, applications
283 * rarely check close errors and even if an error is returned they will not
284 * re-try the close call.
286 int ll_file_release(struct inode *inode, struct file *file)
288 struct ll_file_data *fd;
289 struct ll_sb_info *sbi = ll_i2sbi(inode);
293 CDEBUG(D_VFSTRACE, "VFS Op:inode="DLID4"(%p)\n",
294 OLID4(&ll_i2info(inode)->lli_id), inode);
296 /* don't do anything for / */
297 if (inode->i_sb->s_root == file->f_dentry)
300 lprocfs_counter_incr(sbi->ll_stats, LPROC_LL_RELEASE);
301 fd = (struct ll_file_data *)file->private_data;
304 rc = ll_md_close(sbi->ll_md_exp, inode, file);
308 static int ll_intent_file_open(struct file *file, void *lmm,
309 int lmmsize, struct lookup_intent *itp)
311 struct ll_sb_info *sbi = ll_i2sbi(file->f_dentry->d_inode);
312 const char *name = (char *)file->f_dentry->d_name.name;
313 struct dentry *parent = file->f_dentry->d_parent;
314 const int len = file->f_dentry->d_name.len;
315 struct lustre_handle lockh;
316 struct mdc_op_data *op_data;
322 OBD_ALLOC(op_data, sizeof(*op_data));
326 ll_prepare_mdc_data(op_data, parent->d_inode, NULL,
329 rc = md_enqueue(sbi->ll_md_exp, LDLM_IBITS, itp, LCK_PR, op_data,
330 &lockh, lmm, lmmsize, ldlm_completion_ast,
331 ll_mdc_blocking_ast, NULL);
332 OBD_FREE(op_data, sizeof(*op_data));
334 if (LUSTRE_IT(itp)->it_lock_mode)
335 memcpy(&LUSTRE_IT(itp)->it_lock_handle,
336 &lockh, sizeof(lockh));
339 CERROR("lock enqueue: err: %d\n", rc);
344 void ll_och_fill(struct inode *inode, struct lookup_intent *it,
345 struct obd_client_handle *och)
347 struct ptlrpc_request *req = LUSTRE_IT(it)->it_data;
348 struct ll_inode_info *lli = ll_i2info(inode);
349 struct mds_body *body;
352 body = lustre_msg_buf (req->rq_repmsg, 1, sizeof (*body));
353 LASSERT (body != NULL); /* reply already checked out */
354 LASSERT_REPSWABBED (req, 1); /* and swabbed down */
356 memcpy(&och->och_fh, &body->handle, sizeof(body->handle));
357 och->och_magic = OBD_CLIENT_HANDLE_MAGIC;
358 lli->lli_io_epoch = body->io_epoch;
359 mdc_set_open_replay_data(ll_i2mdexp(inode), och,
360 LUSTRE_IT(it)->it_data);
363 int ll_local_open(struct file *file, struct lookup_intent *it,
364 struct obd_client_handle *och)
366 struct ll_file_data *fd;
370 ll_och_fill(file->f_dentry->d_inode, it, och);
372 LASSERTF(file->private_data == NULL, "file %.*s/%.*s ino %lu/%u (%o)\n",
373 file->f_dentry->d_name.len, file->f_dentry->d_name.name,
374 file->f_dentry->d_parent->d_name.len,
375 file->f_dentry->d_parent->d_name.name,
376 file->f_dentry->d_inode->i_ino,
377 file->f_dentry->d_inode->i_generation,
378 file->f_dentry->d_inode->i_mode);
380 OBD_SLAB_ALLOC(fd, ll_file_data_slab, SLAB_KERNEL, sizeof *fd);
382 /* We can't handle this well without reorganizing ll_file_open and
383 * ll_md_close(), so don't even try right now. */
386 file->private_data = fd;
387 ll_readahead_init(file->f_dentry->d_inode, &fd->fd_ras);
388 fd->fd_omode = it->it_flags;
392 /* Open a file, and (for the very first open) create objects on the OSTs at
393 * this time. If opened with O_LOV_DELAY_CREATE, then we don't do the object
394 * creation or open until ll_lov_setstripe() ioctl is called. We grab
395 * lli_open_sem to ensure no other process will create objects, send the
396 * stripe MD to the MDS, or try to destroy the objects if that fails.
398 * If we already have the stripe MD locally then we don't request it in
399 * mdc_open(), by passing a lmm_size = 0.
401 * It is up to the application to ensure no other processes open this file
402 * in the O_LOV_DELAY_CREATE case, or the default striping pattern will be
403 * used. We might be able to avoid races of that sort by getting lli_open_sem
404 * before returning in the O_LOV_DELAY_CREATE case and dropping it here
405 * or in ll_file_release(), but I'm not sure that is desirable/necessary.
407 int ll_file_open(struct inode *inode, struct file *file)
409 struct ll_inode_info *lli = ll_i2info(inode);
410 struct lookup_intent *it, oit = { .it_op = IT_OPEN,
411 .it_flags = file->f_flags };
412 struct lov_stripe_md *lsm;
413 struct ptlrpc_request *req;
415 struct obd_client_handle **och_p = NULL;
416 __u64 *och_usecount = NULL;
419 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), flags %o\n",
420 inode->i_ino, inode->i_generation, inode, file->f_flags);
422 /* don't do anything for / */
423 if (inode->i_sb->s_root == file->f_dentry)
426 if ((file->f_flags+1) & O_ACCMODE)
428 if (file->f_flags & O_TRUNC)
434 * sometimes LUSTRE_IT(it) may not be allocated like opening file by
435 * dentry_open() from GNS stuff.
437 if (!it || !LUSTRE_IT(it)) {
439 rc = ll_intent_alloc(it);
444 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_OPEN);
447 * mdc_intent_lock() didn't get a request ref if there was an open
448 * error, so don't do cleanup on the * request here (bug 3430)
450 if (LUSTRE_IT(it)->it_disposition) {
451 rc = it_open_error(DISP_OPEN_OPEN, it);
456 /* Let's see if we have file open on MDS already. */
457 if (it->it_flags & FMODE_WRITE) {
458 och_p = &lli->lli_mds_write_och;
459 och_usecount = &lli->lli_open_fd_write_count;
460 } else if (it->it_flags & FMODE_EXEC) {
461 och_p = &lli->lli_mds_exec_och;
462 och_usecount = &lli->lli_open_fd_exec_count;
464 och_p = &lli->lli_mds_read_och;
465 och_usecount = &lli->lli_open_fd_read_count;
468 down(&lli->lli_och_sem);
469 if (*och_p) { /* Open handle is present */
470 if (it_disposition(it, DISP_LOOKUP_POS) && /* Positive lookup */
471 it_disposition(it, DISP_OPEN_OPEN)) { /* & OPEN happened */
472 struct obd_client_handle *och;
473 /* Well, there's extra open request that we do not need,
474 let's close it somehow*/
475 OBD_ALLOC(och, sizeof (struct obd_client_handle));
477 up(&lli->lli_och_sem);
481 ll_och_fill(inode, it, och);
482 /* ll_md_och_close() will free och */
483 ll_md_och_close(ll_i2mdexp(inode), inode, och, 0);
487 rc = ll_local_open(file, it, NULL);
491 LASSERT(*och_usecount == 0);
492 OBD_ALLOC(*och_p, sizeof (struct obd_client_handle));
494 GOTO(out, rc = -ENOMEM);
497 if (!it || !LUSTRE_IT(it) || !LUSTRE_IT(it)->it_disposition) {
499 * we are going to replace intent here, and that may
500 * possibly change access mode (FMODE_EXEC can only be
501 * set in intent), but I hope it never happens (I was
502 * not able to trigger it yet at least) -- green
505 /* FIXME: FMODE_EXEC is not covered by O_ACCMODE! */
506 LASSERT(!(it->it_flags & FMODE_EXEC));
507 LASSERTF((it->it_flags & O_ACCMODE) ==
508 (oit.it_flags & O_ACCMODE), "Changing intent "
509 "flags %x to incompatible %x\n", it->it_flags,
512 rc = ll_intent_file_open(file, NULL, 0, it);
515 rc = it_open_error(DISP_OPEN_OPEN, it);
517 GOTO(out_och_free, rc);
519 mdc_set_lock_data(NULL, &LUSTRE_IT(it)->it_lock_handle,
520 file->f_dentry->d_inode);
522 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_OPEN);
523 rc = ll_local_open(file, it, *och_p);
524 LASSERTF(rc == 0, "rc = %d\n", rc);
526 up(&lli->lli_och_sem);
529 * must do this outside lli_och_sem lock to prevent deadlock where
530 * different kind of OPEN lock for this same inode gets cancelled by
534 if (!S_ISREG(inode->i_mode))
539 if (file->f_flags & O_LOV_DELAY_CREATE ||
540 !(file->f_mode & FMODE_WRITE)) {
541 CDEBUG(D_INODE, "object creation was delayed\n");
545 file->f_flags &= ~O_LOV_DELAY_CREATE;
548 req = LUSTRE_IT(it)->it_data;
549 ll_intent_drop_lock(it);
550 ll_intent_release(it);
551 ptlrpc_req_finished(req);
553 ll_open_complete(inode);
557 OBD_FREE(*och_p, sizeof (struct obd_client_handle));
558 *och_p = NULL; /* OBD_FREE writes some magic there */
561 up(&lli->lli_och_sem);
567 /* Fills the obdo with the attributes for the inode defined by lsm */
568 int ll_lsm_getattr(struct obd_export *exp, struct lov_stripe_md *lsm,
571 struct ptlrpc_request_set *set;
575 LASSERT(lsm != NULL);
577 memset(oa, 0, sizeof *oa);
578 oa->o_id = lsm->lsm_object_id;
579 oa->o_gr = lsm->lsm_object_gr;
580 oa->o_mode = S_IFREG;
581 oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLSIZE |
582 OBD_MD_FLBLOCKS | OBD_MD_FLBLKSZ | OBD_MD_FLMTIME |
583 OBD_MD_FLCTIME | OBD_MD_FLGROUP;
585 set = ptlrpc_prep_set();
589 rc = obd_getattr_async(exp, oa, lsm, set);
591 rc = ptlrpc_set_wait(set);
592 ptlrpc_set_destroy(set);
597 oa->o_valid &= (OBD_MD_FLBLOCKS | OBD_MD_FLBLKSZ | OBD_MD_FLMTIME |
598 OBD_MD_FLCTIME | OBD_MD_FLSIZE);
602 static inline void ll_remove_suid(struct inode *inode)
606 /* set S_IGID if S_IXGRP is set, and always set S_ISUID */
607 mode = (inode->i_mode & S_IXGRP)*(S_ISGID/S_IXGRP) | S_ISUID;
609 /* was any of the uid bits set? */
610 mode &= inode->i_mode;
611 if (mode && !capable(CAP_FSETID)) {
612 inode->i_mode &= ~mode;
613 // XXX careful here - we cannot change the size
617 static int ll_lock_to_stripe_offset(struct inode *inode, struct ldlm_lock *lock)
619 struct ll_inode_info *lli = ll_i2info(inode);
620 struct lov_stripe_md *lsm = lli->lli_smd;
621 struct obd_export *exp = ll_i2dtexp(inode);
624 struct ldlm_lock *lock;
625 struct lov_stripe_md *lsm;
626 } key = { .name = "lock_to_stripe", .lock = lock, .lsm = lsm };
627 __u32 stripe, vallen = sizeof(stripe);
631 if (lsm->lsm_stripe_count == 1)
632 GOTO(check, stripe = 0);
634 /* get our offset in the lov */
635 rc = obd_get_info(exp, sizeof(key), &key, &vallen, &stripe);
637 CERROR("obd_get_info: rc = %d\n", rc);
640 LASSERT(stripe < lsm->lsm_stripe_count);
643 if (lsm->lsm_oinfo[stripe].loi_id != lock->l_resource->lr_name.name[0]||
644 lsm->lsm_oinfo[stripe].loi_gr != lock->l_resource->lr_name.name[2]){
645 LDLM_ERROR(lock, "resource doesn't match object "LPU64"/"LPU64
646 " inode=%lu/%u (%p)\n",
647 lsm->lsm_oinfo[stripe].loi_id,
648 lsm->lsm_oinfo[stripe].loi_gr,
649 inode->i_ino, inode->i_generation, inode);
650 return -ELDLM_NO_LOCK_DATA;
656 /* Flush the page cache for an extent as its canceled. When we're on an LOV,
657 * we get a lock cancellation for each stripe, so we have to map the obd's
658 * region back onto the stripes in the file that it held.
660 * No one can dirty the extent until we've finished our work and they can
661 * enqueue another lock. The DLM protects us from ll_file_read/write here,
662 * but other kernel actors could have pages locked.
664 * Called with the DLM lock held. */
665 void ll_pgcache_remove_extent(struct inode *inode, struct lov_stripe_md *lsm,
666 struct ldlm_lock *lock, __u32 stripe)
668 ldlm_policy_data_t tmpex;
669 unsigned long start, end, count, skip, i, j;
671 int rc, rc2, discard = lock->l_flags & LDLM_FL_DISCARD_DATA;
672 struct lustre_handle lockh;
675 memcpy(&tmpex, &lock->l_policy_data, sizeof(tmpex));
676 CDEBUG(D_INODE|D_PAGE, "inode %lu(%p) ["LPU64"->"LPU64"] size: %llu\n",
677 inode->i_ino, inode, tmpex.l_extent.start, tmpex.l_extent.end,
680 /* our locks are page granular thanks to osc_enqueue, we invalidate the
682 LASSERT((tmpex.l_extent.start & ~PAGE_CACHE_MASK) == 0);
683 LASSERT(((tmpex.l_extent.end + 1) & ~PAGE_CACHE_MASK) == 0);
687 start = tmpex.l_extent.start >> PAGE_CACHE_SHIFT;
688 end = tmpex.l_extent.end >> PAGE_CACHE_SHIFT;
689 if (lsm->lsm_stripe_count > 1) {
690 count = lsm->lsm_stripe_size >> PAGE_CACHE_SHIFT;
691 skip = (lsm->lsm_stripe_count - 1) * count;
692 start += start/count * skip + stripe * count;
694 end += end/count * skip + stripe * count;
696 if (end < tmpex.l_extent.end >> PAGE_CACHE_SHIFT)
699 i = inode->i_size ? (inode->i_size - 1) >> PAGE_CACHE_SHIFT : 0;
703 CDEBUG(D_INODE|D_PAGE, "walking page indices start: %lu j: %lu "
704 "count: %lu skip: %lu end: %lu%s\n", start, start % count,
705 count, skip, end, discard ? " (DISCARDING)" : "");
707 /* walk through the vmas on the inode and tear down mmaped pages that
708 * intersect with the lock. this stops immediately if there are no
709 * mmap()ed regions of the file. This is not efficient at all and
710 * should be short lived. We'll associate mmap()ed pages with the lock
711 * and will be able to find them directly */
713 for (i = start; i <= end; i += (j + skip)) {
714 j = min(count - (i % count), end - i + 1);
716 LASSERT(inode->i_mapping);
717 if (ll_teardown_mmaps(inode->i_mapping,
718 (__u64)i << PAGE_CACHE_SHIFT,
719 ((__u64)(i+j) << PAGE_CACHE_SHIFT) - 1) )
723 /* this is the simplistic implementation of page eviction at
724 * cancelation. It is careful to get races with other page
725 * lockers handled correctly. fixes from bug 20 will make it
726 * more efficient by associating locks with pages and with
727 * batching writeback under the lock explicitly. */
728 for (i = start, j = start % count; i <= end;
729 j++, i++, tmpex.l_extent.start += PAGE_CACHE_SIZE) {
731 CDEBUG(D_PAGE, "skip index %lu to %lu\n", i, i + skip);
737 LASSERTF(tmpex.l_extent.start< lock->l_policy_data.l_extent.end,
738 LPU64" >= "LPU64" start %lu i %lu end %lu\n",
739 tmpex.l_extent.start, lock->l_policy_data.l_extent.end,
742 if (!mapping_has_pages(inode->i_mapping)) {
743 CDEBUG(D_INODE|D_PAGE, "nothing left\n");
749 page = find_get_page(inode->i_mapping, i);
752 LL_CDEBUG_PAGE(D_PAGE, page, "lock page idx %lu ext "LPU64"\n",
753 i, tmpex.l_extent.start);
756 /* page->mapping to check with racing against teardown */
757 if (!discard && clear_page_dirty_for_io(page)) {
758 rc = ll_call_writepage(inode, page);
760 CERROR("writepage of page %p failed: %d\n",
762 /* either waiting for io to complete or reacquiring
763 * the lock that the failed writepage released */
767 tmpex.l_extent.end = tmpex.l_extent.start + PAGE_CACHE_SIZE - 1;
768 /* check to see if another DLM lock covers this page */
769 rc2 = ldlm_lock_match(lock->l_resource->lr_namespace,
770 LDLM_FL_BLOCK_GRANTED|LDLM_FL_CBPENDING |
772 &lock->l_resource->lr_name, LDLM_EXTENT,
773 &tmpex, LCK_PR | LCK_PW, &lockh);
774 if (rc2 == 0 && page->mapping != NULL) {
775 // checking again to account for writeback's lock_page()
776 LL_CDEBUG_PAGE(D_PAGE, page, "truncating\n");
777 ll_ra_accounting(page, inode->i_mapping);
778 ll_truncate_complete_page(page);
781 page_cache_release(page);
783 LASSERTF(tmpex.l_extent.start <=
784 (lock->l_policy_data.l_extent.end == ~0ULL ? ~0ULL :
785 lock->l_policy_data.l_extent.end + 1),
786 "loop too long "LPU64" > "LPU64" start %lu i %lu end %lu\n",
787 tmpex.l_extent.start, lock->l_policy_data.l_extent.end,
792 static int ll_extent_lock_callback(struct ldlm_lock *lock,
793 struct ldlm_lock_desc *new, void *data,
796 struct lustre_handle lockh = { 0 };
800 if ((unsigned long)data > 0 && (unsigned long)data < 0x1000) {
801 LDLM_ERROR(lock, "cancelling lock with bad data %p", data);
806 case LDLM_CB_BLOCKING:
807 ldlm_lock2handle(lock, &lockh);
808 rc = ldlm_cli_cancel(&lockh);
810 CERROR("ldlm_cli_cancel failed: %d\n", rc);
812 case LDLM_CB_CANCELING: {
814 struct ll_inode_info *lli;
815 struct lov_stripe_md *lsm;
819 /* This lock wasn't granted, don't try to evict pages */
820 if (lock->l_req_mode != lock->l_granted_mode)
823 inode = ll_inode_from_lock(lock);
826 lli = ll_i2info(inode);
829 if (lli->lli_smd == NULL)
833 stripe = ll_lock_to_stripe_offset(inode, lock);
836 ll_pgcache_remove_extent(inode, lsm, lock, stripe);
838 down(&lli->lli_size_sem);
839 lock_res_and_lock(lock);
840 kms = ldlm_extent_shift_kms(lock,
841 lsm->lsm_oinfo[stripe].loi_kms);
843 if (lsm->lsm_oinfo[stripe].loi_kms != kms)
844 LDLM_DEBUG(lock, "updating kms from "LPU64" to "LPU64,
845 lsm->lsm_oinfo[stripe].loi_kms, kms);
846 lsm->lsm_oinfo[stripe].loi_kms = kms;
847 unlock_res_and_lock(lock);
848 up(&lli->lli_size_sem);
849 //ll_try_done_writing(inode);
862 int ll_async_completion_ast(struct ldlm_lock *lock, int flags, void *data)
864 /* XXX ALLOCATE - 160 bytes */
865 struct inode *inode = ll_inode_from_lock(lock);
866 struct ll_inode_info *lli = ll_i2info(inode);
867 struct lustre_handle lockh = { 0 };
872 if (flags & (LDLM_FL_BLOCK_WAIT | LDLM_FL_BLOCK_GRANTED |
873 LDLM_FL_BLOCK_CONV)) {
874 LBUG(); /* not expecting any blocked async locks yet */
875 LDLM_DEBUG(lock, "client-side async enqueue returned a blocked "
877 ldlm_lock_dump(D_OTHER, lock, 0);
878 ldlm_reprocess_all(lock->l_resource);
882 LDLM_DEBUG(lock, "client-side async enqueue: granted/glimpsed");
884 stripe = ll_lock_to_stripe_offset(inode, lock);
888 if (lock->l_lvb_len) {
889 struct lov_stripe_md *lsm = lli->lli_smd;
891 lvb = lock->l_lvb_data;
892 lsm->lsm_oinfo[stripe].loi_rss = lvb->lvb_size;
895 lock_res_and_lock(lock);
896 kms = MAX(lsm->lsm_oinfo[stripe].loi_kms, lvb->lvb_size);
897 kms = ldlm_extent_shift_kms(NULL, kms);
898 if (lsm->lsm_oinfo[stripe].loi_kms != kms)
899 LDLM_DEBUG(lock, "updating kms from "LPU64" to "LPU64,
900 lsm->lsm_oinfo[stripe].loi_kms, kms);
901 lsm->lsm_oinfo[stripe].loi_kms = kms;
902 unlock_res_and_lock(lock);
908 wake_up(&lock->l_waitq);
910 ldlm_lock2handle(lock, &lockh);
911 ldlm_lock_decref(&lockh, LCK_PR);
916 static int ll_glimpse_callback(struct ldlm_lock *lock, void *reqp)
918 struct ptlrpc_request *req = reqp;
919 struct inode *inode = ll_inode_from_lock(lock);
920 struct ll_inode_info *lli;
922 struct lov_stripe_md *lsm;
923 int rc, size = sizeof(*lvb), stripe;
927 GOTO(out, rc = -ELDLM_NO_LOCK_DATA);
928 lli = ll_i2info(inode);
930 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
934 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
936 /* First, find out which stripe index this lock corresponds to. */
937 stripe = ll_lock_to_stripe_offset(inode, lock);
939 GOTO(iput, rc = -ELDLM_NO_LOCK_DATA);
941 rc = lustre_pack_reply(req, 1, &size, NULL);
943 CERROR("lustre_pack_reply: %d\n", rc);
947 lvb = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*lvb));
948 lvb->lvb_size = lli->lli_smd->lsm_oinfo[stripe].loi_kms;
949 lvb->lvb_mtime = LTIME_S(inode->i_mtime);
950 lvb->lvb_atime = LTIME_S(inode->i_atime);
951 lvb->lvb_ctime = LTIME_S(inode->i_ctime);
953 LDLM_DEBUG(lock, "i_size: %llu -> stripe number %u -> kms "LPU64,
954 inode->i_size, stripe, lvb->lvb_size);
960 /* These errors are normal races, so we don't want to fill the console
961 * with messages by calling ptlrpc_error() */
962 if (rc == -ELDLM_NO_LOCK_DATA)
963 lustre_pack_reply(req, 0, NULL, NULL);
969 /* NB: lov_merge_size will prefer locally cached writes if they extend the
970 * file (because it prefers KMS over RSS when larger) */
971 int ll_glimpse_size(struct inode *inode)
973 struct ll_inode_info *lli = ll_i2info(inode);
974 struct ll_sb_info *sbi = ll_i2sbi(inode);
975 ldlm_policy_data_t policy = { .l_extent = { 0, OBD_OBJECT_EOF } };
976 struct lustre_handle lockh = { 0 };
977 int rc, flags = LDLM_FL_HAS_INTENT;
980 CDEBUG(D_DLMTRACE, "Glimpsing inode %lu\n", inode->i_ino);
982 rc = obd_enqueue(sbi->ll_dt_exp, lli->lli_smd, LDLM_EXTENT, &policy,
983 LCK_PR, &flags, ll_extent_lock_callback,
984 ldlm_completion_ast, ll_glimpse_callback, inode,
985 sizeof(struct ost_lvb), lustre_swab_ost_lvb, &lockh);
990 CERROR("obd_enqueue returned rc %d, returning -EIO\n", rc);
991 RETURN(rc > 0 ? -EIO : rc);
994 down(&lli->lli_size_sem);
995 inode->i_size = lov_merge_size(lli->lli_smd, 0);
996 inode->i_blocks = lov_merge_blocks(lli->lli_smd);
997 up(&lli->lli_size_sem);
999 LTIME_S(inode->i_mtime) = lov_merge_mtime(lli->lli_smd,
1000 LTIME_S(inode->i_mtime));
1002 CDEBUG(D_DLMTRACE, "glimpse: size: "LPU64", blocks: "LPU64"\n",
1003 (__u64)inode->i_size, (__u64)inode->i_blocks);
1005 obd_cancel(sbi->ll_dt_exp, lli->lli_smd, LCK_PR, &lockh);
1009 void ll_stime_record(struct ll_sb_info *sbi, struct timeval *start,
1010 struct obd_service_time *stime)
1012 struct timeval stop;
1013 do_gettimeofday(&stop);
1015 spin_lock(&sbi->ll_lock);
1016 lprocfs_stime_record(stime, &stop, start);
1017 spin_unlock(&sbi->ll_lock);
1020 int ll_extent_lock(struct ll_file_data *fd, struct inode *inode,
1021 struct lov_stripe_md *lsm, int mode,
1022 ldlm_policy_data_t *policy, struct lustre_handle *lockh,
1023 int ast_flags, struct obd_service_time *stime)
1025 struct ll_inode_info *lli = ll_i2info(inode);
1026 struct ll_sb_info *sbi = ll_i2sbi(inode);
1027 struct timeval start;
1031 LASSERT(lockh->cookie == 0);
1033 /* XXX phil: can we do this? won't it screw the file size up? */
1034 if ((fd && (fd->fd_flags & LL_FILE_IGNORE_LOCK)) ||
1035 (sbi->ll_flags & LL_SBI_NOLCK))
1038 CDEBUG(D_DLMTRACE, "Locking inode %lu, start "LPU64" end "LPU64"\n",
1039 inode->i_ino, policy->l_extent.start, policy->l_extent.end);
1041 do_gettimeofday(&start);
1042 rc = obd_enqueue(sbi->ll_dt_exp, lsm, LDLM_EXTENT, policy, mode,
1043 &ast_flags, ll_extent_lock_callback,
1044 ldlm_completion_ast, ll_glimpse_callback, inode,
1045 sizeof(struct ost_lvb), lustre_swab_ost_lvb, lockh);
1049 ll_stime_record(sbi, &start, stime);
1051 if (policy->l_extent.start == 0 &&
1052 policy->l_extent.end == OBD_OBJECT_EOF) {
1053 /* vmtruncate()->ll_truncate() first sets the i_size and then
1054 * the kms under both a DLM lock and the i_sem. If we don't
1055 * get the i_sem here we can match the DLM lock and reset
1056 * i_size from the kms before the truncating path has updated
1057 * the kms. generic_file_write can then trust the stale i_size
1058 * when doing appending writes and effectively cancel the
1059 * result of the truncate. Getting the i_sem after the enqueue
1060 * maintains the DLM -> i_sem acquiry order. */
1061 down(&lli->lli_size_sem);
1062 inode->i_size = lov_merge_size(lsm, 1);
1063 up(&lli->lli_size_sem);
1067 LTIME_S(inode->i_mtime) =
1068 lov_merge_mtime(lsm, LTIME_S(inode->i_mtime));
1074 int ll_extent_unlock(struct ll_file_data *fd, struct inode *inode,
1075 struct lov_stripe_md *lsm, int mode,
1076 struct lustre_handle *lockh)
1078 struct ll_sb_info *sbi = ll_i2sbi(inode);
1082 /* XXX phil: can we do this? won't it screw the file size up? */
1083 if ((fd && (fd->fd_flags & LL_FILE_IGNORE_LOCK)) ||
1084 (sbi->ll_flags & LL_SBI_NOLCK))
1087 rc = obd_cancel(sbi->ll_dt_exp, lsm, mode, lockh);
1092 static ssize_t ll_file_read(struct file *file, char *buf, size_t count,
1095 struct inode *inode = file->f_dentry->d_inode;
1096 struct ll_inode_info *lli = ll_i2info(inode);
1097 struct lov_stripe_md *lsm = lli->lli_smd;
1098 struct ll_lock_tree tree;
1099 struct ll_lock_tree_node *node;
1104 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),size="LPSZ",offset=%Ld\n",
1105 inode->i_ino, inode->i_generation, inode, count, *ppos);
1107 /* "If nbyte is 0, read() will return 0 and have no other results."
1108 * -- Single Unix Spec */
1112 lprocfs_counter_add(ll_i2sbi(inode)->ll_stats, LPROC_LL_READ_BYTES,
1118 node = ll_node_from_inode(inode, *ppos, *ppos + count - 1,
1121 tree.lt_fd = file->private_data;
1123 rc = ll_tree_lock(&tree, node, inode, buf, count,
1124 file->f_flags & O_NONBLOCK ? LDLM_FL_BLOCK_NOWAIT :0);
1128 down(&lli->lli_size_sem);
1129 kms = lov_merge_size(lsm, 1);
1130 if (*ppos + count - 1 > kms) {
1131 /* A glimpse is necessary to determine whether we return a short
1132 * read or some zeroes at the end of the buffer */
1133 up(&lli->lli_size_sem);
1134 retval = ll_glimpse_size(inode);
1138 inode->i_size = kms;
1139 up(&lli->lli_size_sem);
1142 CDEBUG(D_INFO, "Read ino %lu, "LPSZ" bytes, offset %lld, i_size %llu\n",
1143 inode->i_ino, count, *ppos, inode->i_size);
1145 /* turn off the kernel's read-ahead */
1146 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
1149 file->f_ra.ra_pages = 0;
1151 retval = generic_file_read(file, buf, count, ppos);
1154 ll_tree_unlock(&tree, inode);
1159 * Write to a file (through the page cache).
1161 static ssize_t ll_file_write(struct file *file, const char *buf,
1162 size_t count, loff_t *ppos)
1164 struct inode *inode = file->f_dentry->d_inode;
1165 loff_t maxbytes = ll_file_maxbytes(inode);
1166 struct ll_lock_tree tree;
1167 struct ll_lock_tree_node *node;
1172 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),size="LPSZ",offset=%Ld\n",
1173 inode->i_ino, inode->i_generation, inode, count, *ppos);
1175 SIGNAL_MASK_ASSERT(); /* XXX BUG 1511 */
1177 /* POSIX, but surprised the VFS doesn't check this already */
1181 /* If file was opened for LL_IOC_LOV_SETSTRIPE but the ioctl wasn't
1182 * called on the file, don't fail the below assertion (bug 2388). */
1183 if (file->f_flags & O_LOV_DELAY_CREATE &&
1184 ll_i2info(inode)->lli_smd == NULL)
1187 LASSERT(ll_i2info(inode)->lli_smd != NULL);
1189 if (file->f_flags & O_APPEND)
1190 node = ll_node_from_inode(inode, 0, OBD_OBJECT_EOF, LCK_PW);
1192 node = ll_node_from_inode(inode, *ppos, *ppos + count - 1,
1196 RETURN(PTR_ERR(node));
1198 tree.lt_fd = file->private_data;
1200 rc = ll_tree_lock(&tree, node, inode, buf, count,
1201 file->f_flags & O_NONBLOCK ? LDLM_FL_BLOCK_NOWAIT :0);
1205 /* this is ok, g_f_w will overwrite this under i_sem if it races
1206 * with a local truncate, it just makes our maxbyte checking easier */
1207 if (file->f_flags & O_APPEND)
1208 *ppos = inode->i_size;
1210 if (*ppos >= maxbytes) {
1211 if (count || *ppos > maxbytes) {
1212 send_sig(SIGXFSZ, current, 0);
1213 GOTO(out, retval = -EFBIG);
1216 if (*ppos + count > maxbytes)
1217 count = maxbytes - *ppos;
1219 CDEBUG(D_INFO, "Writing inode %lu, "LPSZ" bytes, offset %Lu\n",
1220 inode->i_ino, count, *ppos);
1222 /* mark open handle dirty */
1223 set_bit(LLI_F_DIRTY_HANDLE, &(ll_i2info(inode)->lli_flags));
1225 /* generic_file_write handles O_APPEND after getting i_sem */
1226 retval = generic_file_write(file, buf, count, ppos);
1229 ll_tree_unlock(&tree, inode);
1230 /* serialize with mmap/munmap/mremap */
1231 lprocfs_counter_add(ll_i2sbi(inode)->ll_stats, LPROC_LL_WRITE_BYTES,
1232 retval > 0 ? retval : 0);
1236 static int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
1237 int flags, struct lov_user_md *lum,
1240 struct ll_inode_info *lli = ll_i2info(inode);
1242 struct obd_export *exp = ll_i2dtexp(inode);
1243 struct lov_stripe_md *lsm;
1244 struct lookup_intent oit = {.it_op = IT_OPEN, .it_flags = flags};
1245 struct ptlrpc_request *req = NULL;
1247 struct lustre_md md;
1248 struct obd_client_handle *och;
1252 if ((file->f_flags+1) & O_ACCMODE)
1254 if (file->f_flags & O_TRUNC)
1257 down(&lli->lli_open_sem);
1260 up(&lli->lli_open_sem);
1261 CDEBUG(D_IOCTL, "stripe already exists for ino %lu\n",
1266 f = get_empty_filp();
1270 f->f_dentry = file->f_dentry;
1271 f->f_vfsmnt = file->f_vfsmnt;
1274 rc = ll_intent_alloc(&oit);
1278 rc = ll_intent_file_open(f, lum, lum_size, &oit);
1281 if (it_disposition(&oit, DISP_LOOKUP_NEG))
1284 req = LUSTRE_IT(&oit)->it_data;
1285 rc = LUSTRE_IT(&oit)->it_status;
1290 rc = mdc_req2lustre_md(ll_i2mdexp(inode), req, 1, exp, &md);
1293 ll_update_inode(f->f_dentry->d_inode, &md);
1295 OBD_ALLOC(och, sizeof(struct obd_client_handle));
1296 rc = ll_local_open(f, &oit, och);
1297 if (rc) { /* Actually ll_local_open cannot fail! */
1300 if (LUSTRE_IT(&oit)->it_lock_mode) {
1301 ldlm_lock_decref_and_cancel((struct lustre_handle *)
1302 &LUSTRE_IT(&oit)->it_lock_handle,
1303 LUSTRE_IT(&oit)->it_lock_mode);
1304 LUSTRE_IT(&oit)->it_lock_mode = 0;
1307 ll_intent_release(&oit);
1309 /* ll_file_release will decrease the count, but won't free anything
1310 because we have at least one more reference coming from actual open
1312 down(&lli->lli_och_sem);
1313 lli->lli_open_fd_write_count++;
1314 up(&lli->lli_och_sem);
1315 rc = ll_file_release(f->f_dentry->d_inode, f);
1317 /* Now also destroy our supplemental och */
1318 ll_md_och_close(ll_i2mdexp(inode), f->f_dentry->d_inode, och, 0);
1321 ll_intent_release(&oit);
1324 up(&lli->lli_open_sem);
1326 ptlrpc_req_finished(req);
1330 static int ll_lov_setea(struct inode *inode, struct file *file,
1333 int flags = MDS_OPEN_HAS_OBJS | FMODE_WRITE;
1334 struct lov_user_md *lump;
1335 int lum_size = sizeof(struct lov_user_md) +
1336 sizeof(struct lov_user_ost_data);
1340 if (!capable (CAP_SYS_ADMIN))
1343 OBD_ALLOC(lump, lum_size);
1347 rc = copy_from_user(lump, (struct lov_user_md *)arg, lum_size);
1349 OBD_FREE(lump, lum_size);
1353 rc = ll_lov_setstripe_ea_info(inode, file, flags, lump, lum_size);
1355 OBD_FREE(lump, lum_size);
1359 static int ll_lov_setstripe(struct inode *inode, struct file *file,
1362 struct lov_user_md lum, *lump = (struct lov_user_md *)arg;
1364 int flags = FMODE_WRITE;
1367 /* Bug 1152: copy properly when this is no longer true */
1368 LASSERT(sizeof(lum) == sizeof(*lump));
1369 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lump->lmm_objects[0]));
1370 rc = copy_from_user(&lum, lump, sizeof(lum));
1374 rc = ll_lov_setstripe_ea_info(inode, file, flags, &lum, sizeof(lum));
1376 put_user(0, &lump->lmm_stripe_count);
1377 rc = obd_iocontrol(LL_IOC_LOV_GETSTRIPE, ll_i2dtexp(inode),
1378 0, ll_i2info(inode)->lli_smd, lump);
1383 static int ll_lov_getstripe(struct inode *inode, unsigned long arg)
1385 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1390 return obd_iocontrol(LL_IOC_LOV_GETSTRIPE, ll_i2dtexp(inode), 0, lsm,
1394 static int ll_get_grouplock(struct inode *inode, struct file *file,
1397 struct ll_file_data *fd = file->private_data;
1398 ldlm_policy_data_t policy = { .l_extent = { .start = 0,
1399 .end = OBD_OBJECT_EOF}};
1400 struct lustre_handle lockh = { 0 };
1401 struct ll_inode_info *lli = ll_i2info(inode);
1402 struct lov_stripe_md *lsm = lli->lli_smd;
1406 if (fd->fd_flags & LL_FILE_GROUP_LOCKED) {
1410 policy.l_extent.gid = arg;
1411 if (file->f_flags & O_NONBLOCK)
1412 flags = LDLM_FL_BLOCK_NOWAIT;
1414 rc = ll_extent_lock(fd, inode, lsm, LCK_GROUP, &policy, &lockh, flags,
1415 &ll_i2sbi(inode)->ll_grouplock_stime);
1419 fd->fd_flags |= LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK;
1421 memcpy(&fd->fd_cwlockh, &lockh, sizeof(lockh));
1426 static int ll_put_grouplock(struct inode *inode, struct file *file,
1429 struct ll_file_data *fd = file->private_data;
1430 struct ll_inode_info *lli = ll_i2info(inode);
1431 struct lov_stripe_md *lsm = lli->lli_smd;
1435 if (!(fd->fd_flags & LL_FILE_GROUP_LOCKED)) {
1436 /* Ugh, it's already unlocked. */
1440 if (fd->fd_gid != arg) /* Ugh? Unlocking with different gid? */
1443 fd->fd_flags &= ~(LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK);
1445 rc = ll_extent_unlock(fd, inode, lsm, LCK_GROUP, &fd->fd_cwlockh);
1450 memset(&fd->fd_cwlockh, 0, sizeof(fd->fd_cwlockh));
1455 int ll_file_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
1458 struct ll_file_data *fd = file->private_data;
1459 struct ll_sb_info *sbi = ll_i2sbi(inode);
1463 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),cmd=%x\n", inode->i_ino,
1464 inode->i_generation, inode, cmd);
1466 if (_IOC_TYPE(cmd) == 'T') /* tty ioctls */
1469 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_IOCTL);
1471 case LL_IOC_GETFLAGS:
1472 /* Get the current value of the file flags */
1473 return put_user(fd->fd_flags, (int *)arg);
1474 case LL_IOC_SETFLAGS:
1475 case LL_IOC_CLRFLAGS:
1476 /* Set or clear specific file flags */
1477 /* XXX This probably needs checks to ensure the flags are
1478 * not abused, and to handle any flag side effects.
1480 if (get_user(flags, (int *) arg))
1483 if (cmd == LL_IOC_SETFLAGS)
1484 fd->fd_flags |= flags;
1486 fd->fd_flags &= ~flags;
1488 case LL_IOC_LOV_SETSTRIPE:
1489 RETURN(ll_lov_setstripe(inode, file, arg));
1490 case LL_IOC_LOV_SETEA:
1491 RETURN(ll_lov_setea(inode, file, arg));
1492 case IOC_MDC_SHOWFID: {
1493 struct lustre_id *idp = (struct lustre_id *)arg;
1494 struct lustre_id id;
1498 filename = getname((const char *)arg);
1499 if (IS_ERR(filename))
1500 RETURN(PTR_ERR(filename));
1502 ll_inode2id(&id, inode);
1504 rc = ll_get_fid(sbi->ll_md_exp, &id, filename, &id);
1506 GOTO(out_filename, rc);
1508 rc = copy_to_user(idp, &id, sizeof(*idp));
1510 GOTO(out_filename, rc = -EFAULT);
1517 case LL_IOC_LOV_GETSTRIPE:
1518 RETURN(ll_lov_getstripe(inode, arg));
1519 case EXT3_IOC_GETFLAGS:
1520 case EXT3_IOC_SETFLAGS:
1521 RETURN( ll_iocontrol(inode, file, cmd, arg) );
1522 case LL_IOC_GROUP_LOCK:
1523 RETURN(ll_get_grouplock(inode, file, arg));
1524 case LL_IOC_GROUP_UNLOCK:
1525 RETURN(ll_put_grouplock(inode, file, arg));
1526 case EXT3_IOC_GETVERSION_OLD:
1527 case EXT3_IOC_GETVERSION:
1528 return put_user(inode->i_generation, (int *) arg);
1529 /* We need to special case any other ioctls we want to handle,
1530 * to send them to the MDS/OST as appropriate and to properly
1531 * network encode the arg field.
1532 case EXT2_IOC_GETVERSION_OLD:
1533 case EXT2_IOC_GETVERSION_NEW:
1534 case EXT2_IOC_SETVERSION_OLD:
1535 case EXT2_IOC_SETVERSION_NEW:
1536 case EXT3_IOC_SETVERSION_OLD:
1537 case EXT3_IOC_SETVERSION:
1539 case LL_IOC_FLUSH_CRED:
1540 RETURN(ll_flush_cred(inode));
1542 RETURN( obd_iocontrol(cmd, ll_i2dtexp(inode), 0, NULL,
1547 loff_t ll_file_seek(struct file *file, loff_t offset, int origin)
1549 struct inode *inode = file->f_dentry->d_inode;
1550 struct ll_file_data *fd = file->private_data;
1551 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1552 struct lustre_handle lockh = {0};
1555 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),to=%llu\n", inode->i_ino,
1556 inode->i_generation, inode,
1557 offset + ((origin==2) ? inode->i_size : file->f_pos));
1559 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_LLSEEK);
1560 if (origin == 2) { /* SEEK_END */
1561 ldlm_policy_data_t policy = { .l_extent = {0, OBD_OBJECT_EOF }};
1562 struct ll_inode_info *lli = ll_i2info(inode);
1563 int nonblock = 0, rc;
1565 if (file->f_flags & O_NONBLOCK)
1566 nonblock = LDLM_FL_BLOCK_NOWAIT;
1568 rc = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
1569 nonblock, &ll_i2sbi(inode)->ll_seek_stime);
1573 down(&lli->lli_size_sem);
1574 offset += inode->i_size;
1575 up(&lli->lli_size_sem);
1576 } else if (origin == 1) { /* SEEK_CUR */
1577 offset += file->f_pos;
1581 if (offset >= 0 && offset <= ll_file_maxbytes(inode)) {
1582 if (offset != file->f_pos) {
1583 file->f_pos = offset;
1584 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
1586 file->f_version = ++event;
1593 ll_extent_unlock(fd, inode, lsm, LCK_PR, &lockh);
1597 int ll_fsync(struct file *file, struct dentry *dentry, int data)
1599 struct inode *inode = dentry->d_inode;
1600 struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
1601 struct lustre_id id;
1602 struct ptlrpc_request *req;
1605 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n", inode->i_ino,
1606 inode->i_generation, inode);
1608 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_FSYNC);
1610 /* fsync's caller has already called _fdata{sync,write}, we want
1611 * that IO to finish before calling the osc and mdc sync methods */
1612 rc = filemap_fdatawait(inode->i_mapping);
1614 ll_inode2id(&id, inode);
1615 err = md_sync(ll_i2sbi(inode)->ll_md_exp, &id, &req);
1619 ptlrpc_req_finished(req);
1622 struct obdo *oa = obdo_alloc();
1625 RETURN(rc ? rc : -ENOMEM);
1627 oa->o_id = lsm->lsm_object_id;
1628 oa->o_gr = lsm->lsm_object_gr;
1629 oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
1631 obdo_from_inode(oa, inode, (OBD_MD_FLTYPE | OBD_MD_FLATIME |
1632 OBD_MD_FLMTIME | OBD_MD_FLCTIME |
1635 err = obd_sync(ll_i2sbi(inode)->ll_dt_exp, oa, lsm,
1645 int ll_file_flock(struct file *file, int cmd, struct file_lock *file_lock)
1647 struct inode *inode = file->f_dentry->d_inode;
1648 struct ll_inode_info *li = ll_i2info(inode);
1649 struct ll_sb_info *sbi = ll_i2sbi(inode);
1650 struct obd_device *obddev;
1651 struct ldlm_res_id res_id =
1652 { .name = {id_fid(&li->lli_id), id_group(&li->lli_id), LDLM_FLOCK} };
1653 struct lustre_handle lockh = {0};
1654 ldlm_policy_data_t flock;
1655 ldlm_mode_t mode = 0;
1660 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu file_lock=%p\n",
1661 inode->i_ino, file_lock);
1663 flock.l_flock.pid = file_lock->fl_pid;
1664 flock.l_flock.start = file_lock->fl_start;
1665 flock.l_flock.end = file_lock->fl_end;
1667 switch (file_lock->fl_type) {
1672 /* An unlock request may or may not have any relation to
1673 * existing locks so we may not be able to pass a lock handle
1674 * via a normal ldlm_lock_cancel() request. The request may even
1675 * unlock a byte range in the middle of an existing lock. In
1676 * order to process an unlock request we need all of the same
1677 * information that is given with a normal read or write record
1678 * lock request. To avoid creating another ldlm unlock (cancel)
1679 * message we'll treat a LCK_NL flock request as an unlock. */
1686 CERROR("unknown fcntl lock type: %d\n", file_lock->fl_type);
1701 flags = LDLM_FL_BLOCK_NOWAIT;
1707 flags = LDLM_FL_TEST_LOCK;
1708 /* Save the old mode so that if the mode in the lock changes we
1709 * can decrement the appropriate reader or writer refcount. */
1710 file_lock->fl_type = mode;
1713 CERROR("unknown fcntl lock command: %d\n", cmd);
1717 CDEBUG(D_DLMTRACE, "inode=%lu, pid="LPU64", flags=%#x, mode=%u, "
1718 "start="LPU64", end="LPU64"\n", inode->i_ino, flock.l_flock.pid,
1719 flags, mode, flock.l_flock.start, flock.l_flock.end);
1721 obddev = md_get_real_obd(sbi->ll_md_exp, &li->lli_id);
1722 rc = ldlm_cli_enqueue(obddev->obd_self_export, NULL,
1723 obddev->obd_namespace,
1724 res_id, LDLM_FLOCK, &flock, mode, &flags,
1725 NULL, ldlm_flock_completion_ast, NULL, file_lock,
1726 NULL, 0, NULL, &lockh);
1730 int ll_inode_revalidate_it(struct dentry *dentry)
1732 struct lookup_intent oit = { .it_op = IT_GETATTR };
1733 struct inode *inode = dentry->d_inode;
1734 struct ptlrpc_request *req = NULL;
1735 struct ll_inode_info *lli;
1736 struct lov_stripe_md *lsm;
1737 struct ll_sb_info *sbi;
1738 struct lustre_id id;
1743 CERROR("REPORT THIS LINE TO PETER\n");
1747 sbi = ll_i2sbi(inode);
1749 ll_inode2id(&id, inode);
1750 lli = ll_i2info(inode);
1751 LASSERT(id_fid(&id) != 0);
1753 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p), name=%s(%p)\n",
1754 inode->i_ino, inode->i_generation, inode, dentry->d_name.name,
1757 #if (LINUX_VERSION_CODE <= KERNEL_VERSION(2,5,0))
1758 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_REVALIDATE);
1761 rc = ll_intent_alloc(&oit);
1765 rc = md_intent_lock(sbi->ll_md_exp, &id, NULL, 0, NULL, 0, &id,
1766 &oit, 0, &req, ll_mdc_blocking_ast);
1770 rc = revalidate_it_finish(req, 1, &oit, dentry);
1775 ll_lookup_finish_locks(&oit, dentry);
1777 if (!LLI_HAVE_FLSIZE(inode)) {
1778 /* if object not yet allocated, don't validate size */
1781 /* ll_glimpse_size() will prefer locally cached
1782 * writes if they extend the file */
1783 rc = ll_glimpse_size(inode);
1788 ll_intent_release(&oit);
1790 ptlrpc_req_finished(req);
1794 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
1795 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat)
1798 struct inode *inode = de->d_inode;
1799 struct ll_inode_info *lli = ll_i2info(inode);
1801 res = ll_inode_revalidate_it(de);
1802 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_GETATTR);
1807 stat->ino = inode->i_ino;
1808 stat->mode = inode->i_mode;
1809 stat->nlink = inode->i_nlink;
1810 stat->uid = inode->i_uid;
1811 stat->gid = inode->i_gid;
1812 stat->atime = inode->i_atime;
1813 stat->mtime = inode->i_mtime;
1814 stat->ctime = inode->i_ctime;
1815 stat->blksize = inode->i_blksize;
1817 down(&lli->lli_size_sem);
1818 stat->size = inode->i_size;
1819 stat->blocks = inode->i_blocks;
1820 up(&lli->lli_size_sem);
1822 stat->rdev = kdev_t_to_nr(inode->i_rdev);
1823 stat->dev = id_group(&ll_i2info(inode)->lli_id);
1829 int ll_setxattr_internal(struct inode *inode, const char *name,
1830 const void *value, size_t size, int flags,
1833 struct ll_sb_info *sbi = ll_i2sbi(inode);
1834 struct ptlrpc_request *request = NULL;
1835 struct mdc_op_data op_data;
1840 CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu\n", inode->i_ino);
1841 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_SETXATTR);
1843 if (sbi->ll_remote && !strcmp(name, XATTR_NAME_ACL_ACCESS))
1844 RETURN(-EOPNOTSUPP);
1846 memset(&attr, 0x0, sizeof(attr));
1847 attr.ia_valid |= valid;
1848 attr.ia_attr_flags = flags;
1850 ll_prepare_mdc_data(&op_data, inode, NULL, NULL, 0, 0);
1852 rc = md_setattr(sbi->ll_md_exp, &op_data, &attr,
1853 (void*) name, strnlen(name, XATTR_NAME_MAX) + 1,
1854 (void*) value, size, &request);
1859 ptlrpc_req_finished(request);
1863 int ll_setxattr(struct dentry *dentry, const char *name, const void *value,
1864 size_t size, int flags)
1867 struct posix_acl *acl;
1868 struct ll_inode_info *lli;
1871 rc = ll_setxattr_internal(dentry->d_inode, name, value, size,
1874 /* update inode's acl info */
1875 if (rc == 0 && strcmp(name, XATTR_NAME_ACL_ACCESS) == 0) {
1877 acl = posix_acl_from_xattr(value, size);
1879 CERROR("convert from xattr to acl error: %ld",
1883 error = posix_acl_valid(acl);
1885 CERROR("acl valid error: %d", error);
1886 posix_acl_release(acl);
1894 lli = ll_i2info(dentry->d_inode);
1895 spin_lock(&lli->lli_lock);
1896 if (lli->lli_posix_acl != NULL)
1897 posix_acl_release(lli->lli_posix_acl);
1898 lli->lli_posix_acl = acl;
1899 spin_unlock(&lli->lli_lock);
1906 int ll_removexattr(struct dentry *dentry, const char *name)
1908 return ll_setxattr_internal(dentry->d_inode, name, NULL, 0, 0,
1913 int ll_getxattr_internal(struct inode *inode, const char *name,
1914 void *value, size_t size, __u64 valid)
1916 struct ptlrpc_request *request = NULL;
1917 struct ll_sb_info *sbi = ll_i2sbi(inode);
1918 struct lustre_id id;
1919 struct mds_body *body;
1924 lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_GETXATTR);
1926 if (sbi->ll_remote && !strcmp(name, XATTR_NAME_ACL_ACCESS))
1927 RETURN(-EOPNOTSUPP);
1929 ll_inode2id(&id, inode);
1930 rc = md_getattr(sbi->ll_md_exp, &id, valid, name, NULL, 0,
1933 if (rc != -ENODATA && rc != -EOPNOTSUPP)
1934 CERROR("rc = %d\n", rc);
1938 body = lustre_msg_buf(request->rq_repmsg, 0, sizeof(*body));
1939 LASSERT(body != NULL);
1940 LASSERT_REPSWABBED(request, 0);
1942 ea_size = body->eadatasize;
1944 GOTO(out, rc = ea_size);
1946 LASSERT(ea_size <= request->rq_repmsg->buflens[1]);
1947 ea_data = lustre_msg_buf(request->rq_repmsg, 1, ea_size);
1948 LASSERT(ea_data != NULL);
1949 LASSERT_REPSWABBED(request, 1);
1952 memcpy(value, ea_data, ea_size);
1956 ptlrpc_req_finished(request);
1960 int ll_getxattr(struct dentry *dentry, const char *name, void *value,
1963 return ll_getxattr_internal(dentry->d_inode, name,
1964 value, size, OBD_MD_FLXATTR);
1967 int ll_listxattr(struct dentry *dentry, char *list, size_t size)
1969 return ll_getxattr_internal(dentry->d_inode, NULL, list, size,
1970 OBD_MD_FLXATTRLIST);
1974 * XXX We could choose not to check DLM lock. Leave the decision
1975 * to remote acl handling.
1978 lustre_check_acl(struct inode *inode, int mask)
1980 struct lookup_intent it = { .it_op = IT_GETATTR };
1981 struct dentry de = { .d_inode = inode };
1982 struct ll_sb_info *sbi;
1983 struct lustre_id id;
1984 struct ptlrpc_request *req = NULL;
1985 struct ll_inode_info *lli = ll_i2info(inode);
1986 struct posix_acl *acl;
1990 sbi = ll_i2sbi(inode);
1991 ll_inode2id(&id, inode);
1993 if (ll_intent_alloc(&it))
1996 rc = md_intent_lock(sbi->ll_md_exp, &id, NULL, 0, NULL, 0, &id,
1997 &it, 0, &req, ll_mdc_blocking_ast);
1999 ll_intent_free(&it);
2003 rc = revalidate_it_finish(req, 1, &it, &de);
2005 ll_intent_release(&it);
2009 if (sbi->ll_remote) {
2010 rc = ll_remote_acl_permission(inode, mask);
2012 spin_lock(&lli->lli_lock);
2013 acl = posix_acl_dup(ll_i2info(inode)->lli_posix_acl);
2014 spin_unlock(&lli->lli_lock);
2019 rc = posix_acl_permission(inode, acl, mask);
2020 posix_acl_release(acl);
2024 ll_lookup_finish_locks(&it, &de);
2025 ll_intent_free(&it);
2029 ptlrpc_req_finished(req);
2034 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd)
2036 return generic_permission(inode, mask, lustre_check_acl);
2039 struct file_operations ll_file_operations = {
2040 .read = ll_file_read,
2041 .write = ll_file_write,
2042 .ioctl = ll_file_ioctl,
2043 .open = ll_file_open,
2044 .release = ll_file_release,
2045 .mmap = ll_file_mmap,
2046 .llseek = ll_file_seek,
2047 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
2048 .sendfile = generic_file_sendfile,
2051 .lock = ll_file_flock
2054 struct inode_operations ll_file_inode_operations = {
2055 .setattr = ll_setattr,
2056 .truncate = ll_truncate,
2057 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
2058 .getattr = ll_getattr,
2060 .revalidate_it = ll_inode_revalidate_it,
2062 .setxattr = ll_setxattr,
2063 .getxattr = ll_getxattr,
2064 .listxattr = ll_listxattr,
2065 .removexattr = ll_removexattr,
2066 .permission = ll_inode_permission,