block: avoid unconditionally freeing previously allocated request_queue
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / nilfs2 / ioctl.c
blobf90a33d9a5b097c06038f614cbbcecb1abfab191
1 /*
2 * ioctl.c - NILFS ioctl operations.
4 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 * Written by Koji Sato <koji@osrg.net>.
23 #include <linux/fs.h>
24 #include <linux/wait.h>
25 #include <linux/smp_lock.h> /* lock_kernel(), unlock_kernel() */
26 #include <linux/slab.h>
27 #include <linux/capability.h> /* capable() */
28 #include <linux/uaccess.h> /* copy_from_user(), copy_to_user() */
29 #include <linux/vmalloc.h>
30 #include <linux/mount.h> /* mnt_want_write(), mnt_drop_write() */
31 #include <linux/nilfs2_fs.h>
32 #include "nilfs.h"
33 #include "segment.h"
34 #include "bmap.h"
35 #include "cpfile.h"
36 #include "sufile.h"
37 #include "dat.h"
40 static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs,
41 struct nilfs_argv *argv, int dir,
42 ssize_t (*dofunc)(struct the_nilfs *,
43 __u64 *, int,
44 void *, size_t, size_t))
46 void *buf;
47 void __user *base = (void __user *)(unsigned long)argv->v_base;
48 size_t maxmembs, total, n;
49 ssize_t nr;
50 int ret, i;
51 __u64 pos, ppos;
53 if (argv->v_nmembs == 0)
54 return 0;
56 if (argv->v_size > PAGE_SIZE)
57 return -EINVAL;
59 buf = (void *)__get_free_pages(GFP_NOFS, 0);
60 if (unlikely(!buf))
61 return -ENOMEM;
62 maxmembs = PAGE_SIZE / argv->v_size;
64 ret = 0;
65 total = 0;
66 pos = argv->v_index;
67 for (i = 0; i < argv->v_nmembs; i += n) {
68 n = (argv->v_nmembs - i < maxmembs) ?
69 argv->v_nmembs - i : maxmembs;
70 if ((dir & _IOC_WRITE) &&
71 copy_from_user(buf, base + argv->v_size * i,
72 argv->v_size * n)) {
73 ret = -EFAULT;
74 break;
76 ppos = pos;
77 nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size,
78 n);
79 if (nr < 0) {
80 ret = nr;
81 break;
83 if ((dir & _IOC_READ) &&
84 copy_to_user(base + argv->v_size * i, buf,
85 argv->v_size * nr)) {
86 ret = -EFAULT;
87 break;
89 total += nr;
90 if ((size_t)nr < n)
91 break;
92 if (pos == ppos)
93 pos += n;
95 argv->v_nmembs = total;
97 free_pages((unsigned long)buf, 0);
98 return ret;
101 static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
102 unsigned int cmd, void __user *argp)
104 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
105 struct inode *cpfile = nilfs->ns_cpfile;
106 struct nilfs_transaction_info ti;
107 struct nilfs_cpmode cpmode;
108 int ret;
110 if (!capable(CAP_SYS_ADMIN))
111 return -EPERM;
113 ret = mnt_want_write(filp->f_path.mnt);
114 if (ret)
115 return ret;
117 ret = -EFAULT;
118 if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
119 goto out;
121 mutex_lock(&nilfs->ns_mount_mutex);
123 nilfs_transaction_begin(inode->i_sb, &ti, 0);
124 ret = nilfs_cpfile_change_cpmode(
125 cpfile, cpmode.cm_cno, cpmode.cm_mode);
126 if (unlikely(ret < 0))
127 nilfs_transaction_abort(inode->i_sb);
128 else
129 nilfs_transaction_commit(inode->i_sb); /* never fails */
131 mutex_unlock(&nilfs->ns_mount_mutex);
132 out:
133 mnt_drop_write(filp->f_path.mnt);
134 return ret;
137 static int
138 nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp,
139 unsigned int cmd, void __user *argp)
141 struct inode *cpfile = NILFS_SB(inode->i_sb)->s_nilfs->ns_cpfile;
142 struct nilfs_transaction_info ti;
143 __u64 cno;
144 int ret;
146 if (!capable(CAP_SYS_ADMIN))
147 return -EPERM;
149 ret = mnt_want_write(filp->f_path.mnt);
150 if (ret)
151 return ret;
153 ret = -EFAULT;
154 if (copy_from_user(&cno, argp, sizeof(cno)))
155 goto out;
157 nilfs_transaction_begin(inode->i_sb, &ti, 0);
158 ret = nilfs_cpfile_delete_checkpoint(cpfile, cno);
159 if (unlikely(ret < 0))
160 nilfs_transaction_abort(inode->i_sb);
161 else
162 nilfs_transaction_commit(inode->i_sb); /* never fails */
163 out:
164 mnt_drop_write(filp->f_path.mnt);
165 return ret;
168 static ssize_t
169 nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
170 void *buf, size_t size, size_t nmembs)
172 int ret;
174 down_read(&nilfs->ns_segctor_sem);
175 ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf,
176 size, nmembs);
177 up_read(&nilfs->ns_segctor_sem);
178 return ret;
181 static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp,
182 unsigned int cmd, void __user *argp)
184 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
185 struct nilfs_cpstat cpstat;
186 int ret;
188 down_read(&nilfs->ns_segctor_sem);
189 ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat);
190 up_read(&nilfs->ns_segctor_sem);
191 if (ret < 0)
192 return ret;
194 if (copy_to_user(argp, &cpstat, sizeof(cpstat)))
195 ret = -EFAULT;
196 return ret;
199 static ssize_t
200 nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
201 void *buf, size_t size, size_t nmembs)
203 int ret;
205 down_read(&nilfs->ns_segctor_sem);
206 ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size,
207 nmembs);
208 up_read(&nilfs->ns_segctor_sem);
209 return ret;
212 static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp,
213 unsigned int cmd, void __user *argp)
215 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
216 struct nilfs_sustat sustat;
217 int ret;
219 down_read(&nilfs->ns_segctor_sem);
220 ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat);
221 up_read(&nilfs->ns_segctor_sem);
222 if (ret < 0)
223 return ret;
225 if (copy_to_user(argp, &sustat, sizeof(sustat)))
226 ret = -EFAULT;
227 return ret;
230 static ssize_t
231 nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
232 void *buf, size_t size, size_t nmembs)
234 int ret;
236 down_read(&nilfs->ns_segctor_sem);
237 ret = nilfs_dat_get_vinfo(nilfs_dat_inode(nilfs), buf, size, nmembs);
238 up_read(&nilfs->ns_segctor_sem);
239 return ret;
242 static ssize_t
243 nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags,
244 void *buf, size_t size, size_t nmembs)
246 struct inode *dat = nilfs_dat_inode(nilfs);
247 struct nilfs_bmap *bmap = NILFS_I(dat)->i_bmap;
248 struct nilfs_bdesc *bdescs = buf;
249 int ret, i;
251 down_read(&nilfs->ns_segctor_sem);
252 for (i = 0; i < nmembs; i++) {
253 ret = nilfs_bmap_lookup_at_level(bmap,
254 bdescs[i].bd_offset,
255 bdescs[i].bd_level + 1,
256 &bdescs[i].bd_blocknr);
257 if (ret < 0) {
258 if (ret != -ENOENT) {
259 up_read(&nilfs->ns_segctor_sem);
260 return ret;
262 bdescs[i].bd_blocknr = 0;
265 up_read(&nilfs->ns_segctor_sem);
266 return nmembs;
269 static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp,
270 unsigned int cmd, void __user *argp)
272 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
273 struct nilfs_argv argv;
274 int ret;
276 if (copy_from_user(&argv, argp, sizeof(argv)))
277 return -EFAULT;
279 if (argv.v_size != sizeof(struct nilfs_bdesc))
280 return -EINVAL;
282 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd),
283 nilfs_ioctl_do_get_bdescs);
284 if (ret < 0)
285 return ret;
287 if (copy_to_user(argp, &argv, sizeof(argv)))
288 ret = -EFAULT;
289 return ret;
292 static int nilfs_ioctl_move_inode_block(struct inode *inode,
293 struct nilfs_vdesc *vdesc,
294 struct list_head *buffers)
296 struct buffer_head *bh;
297 int ret;
299 if (vdesc->vd_flags == 0)
300 ret = nilfs_gccache_submit_read_data(
301 inode, vdesc->vd_offset, vdesc->vd_blocknr,
302 vdesc->vd_vblocknr, &bh);
303 else
304 ret = nilfs_gccache_submit_read_node(
305 inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh);
307 if (unlikely(ret < 0)) {
308 if (ret == -ENOENT)
309 printk(KERN_CRIT
310 "%s: invalid virtual block address (%s): "
311 "ino=%llu, cno=%llu, offset=%llu, "
312 "blocknr=%llu, vblocknr=%llu\n",
313 __func__, vdesc->vd_flags ? "node" : "data",
314 (unsigned long long)vdesc->vd_ino,
315 (unsigned long long)vdesc->vd_cno,
316 (unsigned long long)vdesc->vd_offset,
317 (unsigned long long)vdesc->vd_blocknr,
318 (unsigned long long)vdesc->vd_vblocknr);
319 return ret;
321 if (unlikely(!list_empty(&bh->b_assoc_buffers))) {
322 printk(KERN_CRIT "%s: conflicting %s buffer: ino=%llu, "
323 "cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu\n",
324 __func__, vdesc->vd_flags ? "node" : "data",
325 (unsigned long long)vdesc->vd_ino,
326 (unsigned long long)vdesc->vd_cno,
327 (unsigned long long)vdesc->vd_offset,
328 (unsigned long long)vdesc->vd_blocknr,
329 (unsigned long long)vdesc->vd_vblocknr);
330 brelse(bh);
331 return -EEXIST;
333 list_add_tail(&bh->b_assoc_buffers, buffers);
334 return 0;
337 static int nilfs_ioctl_move_blocks(struct the_nilfs *nilfs,
338 struct nilfs_argv *argv, void *buf)
340 size_t nmembs = argv->v_nmembs;
341 struct inode *inode;
342 struct nilfs_vdesc *vdesc;
343 struct buffer_head *bh, *n;
344 LIST_HEAD(buffers);
345 ino_t ino;
346 __u64 cno;
347 int i, ret;
349 for (i = 0, vdesc = buf; i < nmembs; ) {
350 ino = vdesc->vd_ino;
351 cno = vdesc->vd_cno;
352 inode = nilfs_gc_iget(nilfs, ino, cno);
353 if (unlikely(inode == NULL)) {
354 ret = -ENOMEM;
355 goto failed;
357 do {
358 ret = nilfs_ioctl_move_inode_block(inode, vdesc,
359 &buffers);
360 if (unlikely(ret < 0))
361 goto failed;
362 vdesc++;
363 } while (++i < nmembs &&
364 vdesc->vd_ino == ino && vdesc->vd_cno == cno);
367 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
368 ret = nilfs_gccache_wait_and_mark_dirty(bh);
369 if (unlikely(ret < 0)) {
370 WARN_ON(ret == -EEXIST);
371 goto failed;
373 list_del_init(&bh->b_assoc_buffers);
374 brelse(bh);
376 return nmembs;
378 failed:
379 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
380 list_del_init(&bh->b_assoc_buffers);
381 brelse(bh);
383 return ret;
386 static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs,
387 struct nilfs_argv *argv, void *buf)
389 size_t nmembs = argv->v_nmembs;
390 struct inode *cpfile = nilfs->ns_cpfile;
391 struct nilfs_period *periods = buf;
392 int ret, i;
394 for (i = 0; i < nmembs; i++) {
395 ret = nilfs_cpfile_delete_checkpoints(
396 cpfile, periods[i].p_start, periods[i].p_end);
397 if (ret < 0)
398 return ret;
400 return nmembs;
403 static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs,
404 struct nilfs_argv *argv, void *buf)
406 size_t nmembs = argv->v_nmembs;
407 int ret;
409 ret = nilfs_dat_freev(nilfs_dat_inode(nilfs), buf, nmembs);
411 return (ret < 0) ? ret : nmembs;
414 static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs,
415 struct nilfs_argv *argv, void *buf)
417 size_t nmembs = argv->v_nmembs;
418 struct inode *dat = nilfs_dat_inode(nilfs);
419 struct nilfs_bmap *bmap = NILFS_I(dat)->i_bmap;
420 struct nilfs_bdesc *bdescs = buf;
421 int ret, i;
423 for (i = 0; i < nmembs; i++) {
424 /* XXX: use macro or inline func to check liveness */
425 ret = nilfs_bmap_lookup_at_level(bmap,
426 bdescs[i].bd_offset,
427 bdescs[i].bd_level + 1,
428 &bdescs[i].bd_blocknr);
429 if (ret < 0) {
430 if (ret != -ENOENT)
431 return ret;
432 bdescs[i].bd_blocknr = 0;
434 if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr)
435 /* skip dead block */
436 continue;
437 if (bdescs[i].bd_level == 0) {
438 ret = nilfs_mdt_mark_block_dirty(dat,
439 bdescs[i].bd_offset);
440 if (ret < 0) {
441 WARN_ON(ret == -ENOENT);
442 return ret;
444 } else {
445 ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset,
446 bdescs[i].bd_level);
447 if (ret < 0) {
448 WARN_ON(ret == -ENOENT);
449 return ret;
453 return nmembs;
456 int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
457 struct nilfs_argv *argv, void **kbufs)
459 const char *msg;
460 int ret;
462 ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
463 if (ret < 0) {
465 * can safely abort because checkpoints can be removed
466 * independently.
468 msg = "cannot delete checkpoints";
469 goto failed;
471 ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]);
472 if (ret < 0) {
474 * can safely abort because DAT file is updated atomically
475 * using a copy-on-write technique.
477 msg = "cannot delete virtual blocks from DAT file";
478 goto failed;
480 ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]);
481 if (ret < 0) {
483 * can safely abort because the operation is nondestructive.
485 msg = "cannot mark copying blocks dirty";
486 goto failed;
488 return 0;
490 failed:
491 printk(KERN_ERR "NILFS: GC failed during preparation: %s: err=%d\n",
492 msg, ret);
493 return ret;
496 static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
497 unsigned int cmd, void __user *argp)
499 struct nilfs_argv argv[5];
500 static const size_t argsz[5] = {
501 sizeof(struct nilfs_vdesc),
502 sizeof(struct nilfs_period),
503 sizeof(__u64),
504 sizeof(struct nilfs_bdesc),
505 sizeof(__u64),
507 void __user *base;
508 void *kbufs[5];
509 struct the_nilfs *nilfs;
510 size_t len, nsegs;
511 int n, ret;
513 if (!capable(CAP_SYS_ADMIN))
514 return -EPERM;
516 ret = mnt_want_write(filp->f_path.mnt);
517 if (ret)
518 return ret;
520 ret = -EFAULT;
521 if (copy_from_user(argv, argp, sizeof(argv)))
522 goto out;
524 ret = -EINVAL;
525 nsegs = argv[4].v_nmembs;
526 if (argv[4].v_size != argsz[4])
527 goto out;
530 * argv[4] points to segment numbers this ioctl cleans. We
531 * use kmalloc() for its buffer because memory used for the
532 * segment numbers is enough small.
534 kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base,
535 nsegs * sizeof(__u64));
536 if (IS_ERR(kbufs[4])) {
537 ret = PTR_ERR(kbufs[4]);
538 goto out;
540 nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
542 for (n = 0; n < 4; n++) {
543 ret = -EINVAL;
544 if (argv[n].v_size != argsz[n])
545 goto out_free;
547 if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment)
548 goto out_free;
550 len = argv[n].v_size * argv[n].v_nmembs;
551 base = (void __user *)(unsigned long)argv[n].v_base;
552 if (len == 0) {
553 kbufs[n] = NULL;
554 continue;
557 kbufs[n] = vmalloc(len);
558 if (!kbufs[n]) {
559 ret = -ENOMEM;
560 goto out_free;
562 if (copy_from_user(kbufs[n], base, len)) {
563 ret = -EFAULT;
564 vfree(kbufs[n]);
565 goto out_free;
570 * nilfs_ioctl_move_blocks() will call nilfs_gc_iget(),
571 * which will operates an inode list without blocking.
572 * To protect the list from concurrent operations,
573 * nilfs_ioctl_move_blocks should be atomic operation.
575 if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
576 ret = -EBUSY;
577 goto out_free;
580 ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]);
581 if (ret < 0)
582 printk(KERN_ERR "NILFS: GC failed during preparation: "
583 "cannot read source blocks: err=%d\n", ret);
584 else
585 ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
587 if (ret < 0)
588 nilfs_remove_all_gcinode(nilfs);
589 clear_nilfs_gc_running(nilfs);
591 out_free:
592 while (--n >= 0)
593 vfree(kbufs[n]);
594 kfree(kbufs[4]);
595 out:
596 mnt_drop_write(filp->f_path.mnt);
597 return ret;
600 static int nilfs_ioctl_sync(struct inode *inode, struct file *filp,
601 unsigned int cmd, void __user *argp)
603 __u64 cno;
604 int ret;
605 struct the_nilfs *nilfs;
607 ret = nilfs_construct_segment(inode->i_sb);
608 if (ret < 0)
609 return ret;
611 if (argp != NULL) {
612 nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
613 down_read(&nilfs->ns_segctor_sem);
614 cno = nilfs->ns_cno - 1;
615 up_read(&nilfs->ns_segctor_sem);
616 if (copy_to_user(argp, &cno, sizeof(cno)))
617 return -EFAULT;
619 return 0;
622 static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp,
623 unsigned int cmd, void __user *argp,
624 size_t membsz,
625 ssize_t (*dofunc)(struct the_nilfs *,
626 __u64 *, int,
627 void *, size_t, size_t))
630 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs;
631 struct nilfs_argv argv;
632 int ret;
634 if (copy_from_user(&argv, argp, sizeof(argv)))
635 return -EFAULT;
637 if (argv.v_size < membsz)
638 return -EINVAL;
640 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc);
641 if (ret < 0)
642 return ret;
644 if (copy_to_user(argp, &argv, sizeof(argv)))
645 ret = -EFAULT;
646 return ret;
649 long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
651 struct inode *inode = filp->f_dentry->d_inode;
652 void __user *argp = (void __user *)arg;
654 switch (cmd) {
655 case NILFS_IOCTL_CHANGE_CPMODE:
656 return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp);
657 case NILFS_IOCTL_DELETE_CHECKPOINT:
658 return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp);
659 case NILFS_IOCTL_GET_CPINFO:
660 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
661 sizeof(struct nilfs_cpinfo),
662 nilfs_ioctl_do_get_cpinfo);
663 case NILFS_IOCTL_GET_CPSTAT:
664 return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp);
665 case NILFS_IOCTL_GET_SUINFO:
666 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
667 sizeof(struct nilfs_suinfo),
668 nilfs_ioctl_do_get_suinfo);
669 case NILFS_IOCTL_GET_SUSTAT:
670 return nilfs_ioctl_get_sustat(inode, filp, cmd, argp);
671 case NILFS_IOCTL_GET_VINFO:
672 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
673 sizeof(struct nilfs_vinfo),
674 nilfs_ioctl_do_get_vinfo);
675 case NILFS_IOCTL_GET_BDESCS:
676 return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp);
677 case NILFS_IOCTL_CLEAN_SEGMENTS:
678 return nilfs_ioctl_clean_segments(inode, filp, cmd, argp);
679 case NILFS_IOCTL_SYNC:
680 return nilfs_ioctl_sync(inode, filp, cmd, argp);
681 default:
682 return -ENOTTY;