Commit a9bae189542e71f91e61a4428adf6e5a7dfe8063

Authored by Vyacheslav Dubeyko
Committed by Linus Torvalds
1 parent 6edacf05c8

nilfs2: fix fix very long mount time issue

There exists a situation when GC can work in background alone without
any other filesystem activity during significant time.

The nilfs_clean_segments() method calls nilfs_segctor_construct() that
updates superblocks in the case of NILFS_SC_SUPER_ROOT and
THE_NILFS_DISCONTINUED flags are set.  But when GC is working alone the
nilfs_clean_segments() is called with unset THE_NILFS_DISCONTINUED flag.
As a result, the update of superblocks doesn't occurred all this time
and in the case of SPOR superblocks keep very old values of last super
root placement.

SYMPTOMS:

Trying to mount a NILFS2 volume after SPOR in such environment ends with
very long mounting time (it can achieve about several hours in some
cases).

REPRODUCING PATH:

1. It needs to use external USB HDD, disable automount and doesn't
   make any additional filesystem activity on the NILFS2 volume.

2. Generate temporary file with size about 100 - 500 GB (for example,
   dd if=/dev/zero of=<file_name> bs=1073741824 count=200).  The size of
   file defines duration of GC working.

3. Then it needs to delete file.

4. Start GC manually by means of command "nilfs-clean -p 0".  When you
   start GC by means of such way then, at the end, superblocks is updated
   by once.  So, for simulation of SPOR, it needs to wait sometime (15 -
   40 minutes) and simply switch off USB HDD manually.

5. Switch on USB HDD again and try to mount NILFS2 volume.  As a
   result, NILFS2 volume will mount during very long time.

REPRODUCIBILITY: 100%

FIX:

This patch adds checking that superblocks need to update and set
THE_NILFS_DISCONTINUED flag before nilfs_clean_segments() call.

Reported-by: Sergey Alexandrov <splavgm@gmail.com>
Signed-off-by: Vyacheslav Dubeyko <slava@dubeyko.com>
Tested-by: Vyacheslav Dubeyko <slava@dubeyko.com>
Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 1 changed file with 4 additions and 1 deletions Inline Diff

1 /* 1 /*
2 * ioctl.c - NILFS ioctl operations. 2 * ioctl.c - NILFS ioctl operations.
3 * 3 *
4 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation. 4 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation.
5 * 5 *
6 * This program is free software; you can redistribute it and/or modify 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 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 8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version. 9 * (at your option) any later version.
10 * 10 *
11 * This program is distributed in the hope that it will be useful, 11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details. 14 * GNU General Public License for more details.
15 * 15 *
16 * You should have received a copy of the GNU General Public License 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 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 18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 * 19 *
20 * Written by Koji Sato <koji@osrg.net>. 20 * Written by Koji Sato <koji@osrg.net>.
21 */ 21 */
22 22
23 #include <linux/fs.h> 23 #include <linux/fs.h>
24 #include <linux/wait.h> 24 #include <linux/wait.h>
25 #include <linux/slab.h> 25 #include <linux/slab.h>
26 #include <linux/capability.h> /* capable() */ 26 #include <linux/capability.h> /* capable() */
27 #include <linux/uaccess.h> /* copy_from_user(), copy_to_user() */ 27 #include <linux/uaccess.h> /* copy_from_user(), copy_to_user() */
28 #include <linux/vmalloc.h> 28 #include <linux/vmalloc.h>
29 #include <linux/compat.h> /* compat_ptr() */ 29 #include <linux/compat.h> /* compat_ptr() */
30 #include <linux/mount.h> /* mnt_want_write_file(), mnt_drop_write_file() */ 30 #include <linux/mount.h> /* mnt_want_write_file(), mnt_drop_write_file() */
31 #include <linux/buffer_head.h> 31 #include <linux/buffer_head.h>
32 #include <linux/nilfs2_fs.h> 32 #include <linux/nilfs2_fs.h>
33 #include "nilfs.h" 33 #include "nilfs.h"
34 #include "segment.h" 34 #include "segment.h"
35 #include "bmap.h" 35 #include "bmap.h"
36 #include "cpfile.h" 36 #include "cpfile.h"
37 #include "sufile.h" 37 #include "sufile.h"
38 #include "dat.h" 38 #include "dat.h"
39 39
40 40
41 static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs, 41 static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs,
42 struct nilfs_argv *argv, int dir, 42 struct nilfs_argv *argv, int dir,
43 ssize_t (*dofunc)(struct the_nilfs *, 43 ssize_t (*dofunc)(struct the_nilfs *,
44 __u64 *, int, 44 __u64 *, int,
45 void *, size_t, size_t)) 45 void *, size_t, size_t))
46 { 46 {
47 void *buf; 47 void *buf;
48 void __user *base = (void __user *)(unsigned long)argv->v_base; 48 void __user *base = (void __user *)(unsigned long)argv->v_base;
49 size_t maxmembs, total, n; 49 size_t maxmembs, total, n;
50 ssize_t nr; 50 ssize_t nr;
51 int ret, i; 51 int ret, i;
52 __u64 pos, ppos; 52 __u64 pos, ppos;
53 53
54 if (argv->v_nmembs == 0) 54 if (argv->v_nmembs == 0)
55 return 0; 55 return 0;
56 56
57 if (argv->v_size > PAGE_SIZE) 57 if (argv->v_size > PAGE_SIZE)
58 return -EINVAL; 58 return -EINVAL;
59 59
60 buf = (void *)__get_free_pages(GFP_NOFS, 0); 60 buf = (void *)__get_free_pages(GFP_NOFS, 0);
61 if (unlikely(!buf)) 61 if (unlikely(!buf))
62 return -ENOMEM; 62 return -ENOMEM;
63 maxmembs = PAGE_SIZE / argv->v_size; 63 maxmembs = PAGE_SIZE / argv->v_size;
64 64
65 ret = 0; 65 ret = 0;
66 total = 0; 66 total = 0;
67 pos = argv->v_index; 67 pos = argv->v_index;
68 for (i = 0; i < argv->v_nmembs; i += n) { 68 for (i = 0; i < argv->v_nmembs; i += n) {
69 n = (argv->v_nmembs - i < maxmembs) ? 69 n = (argv->v_nmembs - i < maxmembs) ?
70 argv->v_nmembs - i : maxmembs; 70 argv->v_nmembs - i : maxmembs;
71 if ((dir & _IOC_WRITE) && 71 if ((dir & _IOC_WRITE) &&
72 copy_from_user(buf, base + argv->v_size * i, 72 copy_from_user(buf, base + argv->v_size * i,
73 argv->v_size * n)) { 73 argv->v_size * n)) {
74 ret = -EFAULT; 74 ret = -EFAULT;
75 break; 75 break;
76 } 76 }
77 ppos = pos; 77 ppos = pos;
78 nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size, 78 nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size,
79 n); 79 n);
80 if (nr < 0) { 80 if (nr < 0) {
81 ret = nr; 81 ret = nr;
82 break; 82 break;
83 } 83 }
84 if ((dir & _IOC_READ) && 84 if ((dir & _IOC_READ) &&
85 copy_to_user(base + argv->v_size * i, buf, 85 copy_to_user(base + argv->v_size * i, buf,
86 argv->v_size * nr)) { 86 argv->v_size * nr)) {
87 ret = -EFAULT; 87 ret = -EFAULT;
88 break; 88 break;
89 } 89 }
90 total += nr; 90 total += nr;
91 if ((size_t)nr < n) 91 if ((size_t)nr < n)
92 break; 92 break;
93 if (pos == ppos) 93 if (pos == ppos)
94 pos += n; 94 pos += n;
95 } 95 }
96 argv->v_nmembs = total; 96 argv->v_nmembs = total;
97 97
98 free_pages((unsigned long)buf, 0); 98 free_pages((unsigned long)buf, 0);
99 return ret; 99 return ret;
100 } 100 }
101 101
102 static int nilfs_ioctl_getflags(struct inode *inode, void __user *argp) 102 static int nilfs_ioctl_getflags(struct inode *inode, void __user *argp)
103 { 103 {
104 unsigned int flags = NILFS_I(inode)->i_flags & FS_FL_USER_VISIBLE; 104 unsigned int flags = NILFS_I(inode)->i_flags & FS_FL_USER_VISIBLE;
105 105
106 return put_user(flags, (int __user *)argp); 106 return put_user(flags, (int __user *)argp);
107 } 107 }
108 108
109 static int nilfs_ioctl_setflags(struct inode *inode, struct file *filp, 109 static int nilfs_ioctl_setflags(struct inode *inode, struct file *filp,
110 void __user *argp) 110 void __user *argp)
111 { 111 {
112 struct nilfs_transaction_info ti; 112 struct nilfs_transaction_info ti;
113 unsigned int flags, oldflags; 113 unsigned int flags, oldflags;
114 int ret; 114 int ret;
115 115
116 if (!inode_owner_or_capable(inode)) 116 if (!inode_owner_or_capable(inode))
117 return -EACCES; 117 return -EACCES;
118 118
119 if (get_user(flags, (int __user *)argp)) 119 if (get_user(flags, (int __user *)argp))
120 return -EFAULT; 120 return -EFAULT;
121 121
122 ret = mnt_want_write_file(filp); 122 ret = mnt_want_write_file(filp);
123 if (ret) 123 if (ret)
124 return ret; 124 return ret;
125 125
126 flags = nilfs_mask_flags(inode->i_mode, flags); 126 flags = nilfs_mask_flags(inode->i_mode, flags);
127 127
128 mutex_lock(&inode->i_mutex); 128 mutex_lock(&inode->i_mutex);
129 129
130 oldflags = NILFS_I(inode)->i_flags; 130 oldflags = NILFS_I(inode)->i_flags;
131 131
132 /* 132 /*
133 * The IMMUTABLE and APPEND_ONLY flags can only be changed by the 133 * The IMMUTABLE and APPEND_ONLY flags can only be changed by the
134 * relevant capability. 134 * relevant capability.
135 */ 135 */
136 ret = -EPERM; 136 ret = -EPERM;
137 if (((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) && 137 if (((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) &&
138 !capable(CAP_LINUX_IMMUTABLE)) 138 !capable(CAP_LINUX_IMMUTABLE))
139 goto out; 139 goto out;
140 140
141 ret = nilfs_transaction_begin(inode->i_sb, &ti, 0); 141 ret = nilfs_transaction_begin(inode->i_sb, &ti, 0);
142 if (ret) 142 if (ret)
143 goto out; 143 goto out;
144 144
145 NILFS_I(inode)->i_flags = (oldflags & ~FS_FL_USER_MODIFIABLE) | 145 NILFS_I(inode)->i_flags = (oldflags & ~FS_FL_USER_MODIFIABLE) |
146 (flags & FS_FL_USER_MODIFIABLE); 146 (flags & FS_FL_USER_MODIFIABLE);
147 147
148 nilfs_set_inode_flags(inode); 148 nilfs_set_inode_flags(inode);
149 inode->i_ctime = CURRENT_TIME; 149 inode->i_ctime = CURRENT_TIME;
150 if (IS_SYNC(inode)) 150 if (IS_SYNC(inode))
151 nilfs_set_transaction_flag(NILFS_TI_SYNC); 151 nilfs_set_transaction_flag(NILFS_TI_SYNC);
152 152
153 nilfs_mark_inode_dirty(inode); 153 nilfs_mark_inode_dirty(inode);
154 ret = nilfs_transaction_commit(inode->i_sb); 154 ret = nilfs_transaction_commit(inode->i_sb);
155 out: 155 out:
156 mutex_unlock(&inode->i_mutex); 156 mutex_unlock(&inode->i_mutex);
157 mnt_drop_write_file(filp); 157 mnt_drop_write_file(filp);
158 return ret; 158 return ret;
159 } 159 }
160 160
161 static int nilfs_ioctl_getversion(struct inode *inode, void __user *argp) 161 static int nilfs_ioctl_getversion(struct inode *inode, void __user *argp)
162 { 162 {
163 return put_user(inode->i_generation, (int __user *)argp); 163 return put_user(inode->i_generation, (int __user *)argp);
164 } 164 }
165 165
166 static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp, 166 static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
167 unsigned int cmd, void __user *argp) 167 unsigned int cmd, void __user *argp)
168 { 168 {
169 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 169 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
170 struct nilfs_transaction_info ti; 170 struct nilfs_transaction_info ti;
171 struct nilfs_cpmode cpmode; 171 struct nilfs_cpmode cpmode;
172 int ret; 172 int ret;
173 173
174 if (!capable(CAP_SYS_ADMIN)) 174 if (!capable(CAP_SYS_ADMIN))
175 return -EPERM; 175 return -EPERM;
176 176
177 ret = mnt_want_write_file(filp); 177 ret = mnt_want_write_file(filp);
178 if (ret) 178 if (ret)
179 return ret; 179 return ret;
180 180
181 ret = -EFAULT; 181 ret = -EFAULT;
182 if (copy_from_user(&cpmode, argp, sizeof(cpmode))) 182 if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
183 goto out; 183 goto out;
184 184
185 mutex_lock(&nilfs->ns_snapshot_mount_mutex); 185 mutex_lock(&nilfs->ns_snapshot_mount_mutex);
186 186
187 nilfs_transaction_begin(inode->i_sb, &ti, 0); 187 nilfs_transaction_begin(inode->i_sb, &ti, 0);
188 ret = nilfs_cpfile_change_cpmode( 188 ret = nilfs_cpfile_change_cpmode(
189 nilfs->ns_cpfile, cpmode.cm_cno, cpmode.cm_mode); 189 nilfs->ns_cpfile, cpmode.cm_cno, cpmode.cm_mode);
190 if (unlikely(ret < 0)) 190 if (unlikely(ret < 0))
191 nilfs_transaction_abort(inode->i_sb); 191 nilfs_transaction_abort(inode->i_sb);
192 else 192 else
193 nilfs_transaction_commit(inode->i_sb); /* never fails */ 193 nilfs_transaction_commit(inode->i_sb); /* never fails */
194 194
195 mutex_unlock(&nilfs->ns_snapshot_mount_mutex); 195 mutex_unlock(&nilfs->ns_snapshot_mount_mutex);
196 out: 196 out:
197 mnt_drop_write_file(filp); 197 mnt_drop_write_file(filp);
198 return ret; 198 return ret;
199 } 199 }
200 200
201 static int 201 static int
202 nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp, 202 nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp,
203 unsigned int cmd, void __user *argp) 203 unsigned int cmd, void __user *argp)
204 { 204 {
205 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 205 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
206 struct nilfs_transaction_info ti; 206 struct nilfs_transaction_info ti;
207 __u64 cno; 207 __u64 cno;
208 int ret; 208 int ret;
209 209
210 if (!capable(CAP_SYS_ADMIN)) 210 if (!capable(CAP_SYS_ADMIN))
211 return -EPERM; 211 return -EPERM;
212 212
213 ret = mnt_want_write_file(filp); 213 ret = mnt_want_write_file(filp);
214 if (ret) 214 if (ret)
215 return ret; 215 return ret;
216 216
217 ret = -EFAULT; 217 ret = -EFAULT;
218 if (copy_from_user(&cno, argp, sizeof(cno))) 218 if (copy_from_user(&cno, argp, sizeof(cno)))
219 goto out; 219 goto out;
220 220
221 nilfs_transaction_begin(inode->i_sb, &ti, 0); 221 nilfs_transaction_begin(inode->i_sb, &ti, 0);
222 ret = nilfs_cpfile_delete_checkpoint(nilfs->ns_cpfile, cno); 222 ret = nilfs_cpfile_delete_checkpoint(nilfs->ns_cpfile, cno);
223 if (unlikely(ret < 0)) 223 if (unlikely(ret < 0))
224 nilfs_transaction_abort(inode->i_sb); 224 nilfs_transaction_abort(inode->i_sb);
225 else 225 else
226 nilfs_transaction_commit(inode->i_sb); /* never fails */ 226 nilfs_transaction_commit(inode->i_sb); /* never fails */
227 out: 227 out:
228 mnt_drop_write_file(filp); 228 mnt_drop_write_file(filp);
229 return ret; 229 return ret;
230 } 230 }
231 231
232 static ssize_t 232 static ssize_t
233 nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 233 nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
234 void *buf, size_t size, size_t nmembs) 234 void *buf, size_t size, size_t nmembs)
235 { 235 {
236 int ret; 236 int ret;
237 237
238 down_read(&nilfs->ns_segctor_sem); 238 down_read(&nilfs->ns_segctor_sem);
239 ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf, 239 ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf,
240 size, nmembs); 240 size, nmembs);
241 up_read(&nilfs->ns_segctor_sem); 241 up_read(&nilfs->ns_segctor_sem);
242 return ret; 242 return ret;
243 } 243 }
244 244
245 static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp, 245 static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp,
246 unsigned int cmd, void __user *argp) 246 unsigned int cmd, void __user *argp)
247 { 247 {
248 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 248 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
249 struct nilfs_cpstat cpstat; 249 struct nilfs_cpstat cpstat;
250 int ret; 250 int ret;
251 251
252 down_read(&nilfs->ns_segctor_sem); 252 down_read(&nilfs->ns_segctor_sem);
253 ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat); 253 ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat);
254 up_read(&nilfs->ns_segctor_sem); 254 up_read(&nilfs->ns_segctor_sem);
255 if (ret < 0) 255 if (ret < 0)
256 return ret; 256 return ret;
257 257
258 if (copy_to_user(argp, &cpstat, sizeof(cpstat))) 258 if (copy_to_user(argp, &cpstat, sizeof(cpstat)))
259 ret = -EFAULT; 259 ret = -EFAULT;
260 return ret; 260 return ret;
261 } 261 }
262 262
263 static ssize_t 263 static ssize_t
264 nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 264 nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
265 void *buf, size_t size, size_t nmembs) 265 void *buf, size_t size, size_t nmembs)
266 { 266 {
267 int ret; 267 int ret;
268 268
269 down_read(&nilfs->ns_segctor_sem); 269 down_read(&nilfs->ns_segctor_sem);
270 ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size, 270 ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size,
271 nmembs); 271 nmembs);
272 up_read(&nilfs->ns_segctor_sem); 272 up_read(&nilfs->ns_segctor_sem);
273 return ret; 273 return ret;
274 } 274 }
275 275
276 static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp, 276 static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp,
277 unsigned int cmd, void __user *argp) 277 unsigned int cmd, void __user *argp)
278 { 278 {
279 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 279 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
280 struct nilfs_sustat sustat; 280 struct nilfs_sustat sustat;
281 int ret; 281 int ret;
282 282
283 down_read(&nilfs->ns_segctor_sem); 283 down_read(&nilfs->ns_segctor_sem);
284 ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat); 284 ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat);
285 up_read(&nilfs->ns_segctor_sem); 285 up_read(&nilfs->ns_segctor_sem);
286 if (ret < 0) 286 if (ret < 0)
287 return ret; 287 return ret;
288 288
289 if (copy_to_user(argp, &sustat, sizeof(sustat))) 289 if (copy_to_user(argp, &sustat, sizeof(sustat)))
290 ret = -EFAULT; 290 ret = -EFAULT;
291 return ret; 291 return ret;
292 } 292 }
293 293
294 static ssize_t 294 static ssize_t
295 nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 295 nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
296 void *buf, size_t size, size_t nmembs) 296 void *buf, size_t size, size_t nmembs)
297 { 297 {
298 int ret; 298 int ret;
299 299
300 down_read(&nilfs->ns_segctor_sem); 300 down_read(&nilfs->ns_segctor_sem);
301 ret = nilfs_dat_get_vinfo(nilfs->ns_dat, buf, size, nmembs); 301 ret = nilfs_dat_get_vinfo(nilfs->ns_dat, buf, size, nmembs);
302 up_read(&nilfs->ns_segctor_sem); 302 up_read(&nilfs->ns_segctor_sem);
303 return ret; 303 return ret;
304 } 304 }
305 305
306 static ssize_t 306 static ssize_t
307 nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags, 307 nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags,
308 void *buf, size_t size, size_t nmembs) 308 void *buf, size_t size, size_t nmembs)
309 { 309 {
310 struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap; 310 struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
311 struct nilfs_bdesc *bdescs = buf; 311 struct nilfs_bdesc *bdescs = buf;
312 int ret, i; 312 int ret, i;
313 313
314 down_read(&nilfs->ns_segctor_sem); 314 down_read(&nilfs->ns_segctor_sem);
315 for (i = 0; i < nmembs; i++) { 315 for (i = 0; i < nmembs; i++) {
316 ret = nilfs_bmap_lookup_at_level(bmap, 316 ret = nilfs_bmap_lookup_at_level(bmap,
317 bdescs[i].bd_offset, 317 bdescs[i].bd_offset,
318 bdescs[i].bd_level + 1, 318 bdescs[i].bd_level + 1,
319 &bdescs[i].bd_blocknr); 319 &bdescs[i].bd_blocknr);
320 if (ret < 0) { 320 if (ret < 0) {
321 if (ret != -ENOENT) { 321 if (ret != -ENOENT) {
322 up_read(&nilfs->ns_segctor_sem); 322 up_read(&nilfs->ns_segctor_sem);
323 return ret; 323 return ret;
324 } 324 }
325 bdescs[i].bd_blocknr = 0; 325 bdescs[i].bd_blocknr = 0;
326 } 326 }
327 } 327 }
328 up_read(&nilfs->ns_segctor_sem); 328 up_read(&nilfs->ns_segctor_sem);
329 return nmembs; 329 return nmembs;
330 } 330 }
331 331
332 static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp, 332 static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp,
333 unsigned int cmd, void __user *argp) 333 unsigned int cmd, void __user *argp)
334 { 334 {
335 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 335 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
336 struct nilfs_argv argv; 336 struct nilfs_argv argv;
337 int ret; 337 int ret;
338 338
339 if (copy_from_user(&argv, argp, sizeof(argv))) 339 if (copy_from_user(&argv, argp, sizeof(argv)))
340 return -EFAULT; 340 return -EFAULT;
341 341
342 if (argv.v_size != sizeof(struct nilfs_bdesc)) 342 if (argv.v_size != sizeof(struct nilfs_bdesc))
343 return -EINVAL; 343 return -EINVAL;
344 344
345 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), 345 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd),
346 nilfs_ioctl_do_get_bdescs); 346 nilfs_ioctl_do_get_bdescs);
347 if (ret < 0) 347 if (ret < 0)
348 return ret; 348 return ret;
349 349
350 if (copy_to_user(argp, &argv, sizeof(argv))) 350 if (copy_to_user(argp, &argv, sizeof(argv)))
351 ret = -EFAULT; 351 ret = -EFAULT;
352 return ret; 352 return ret;
353 } 353 }
354 354
355 static int nilfs_ioctl_move_inode_block(struct inode *inode, 355 static int nilfs_ioctl_move_inode_block(struct inode *inode,
356 struct nilfs_vdesc *vdesc, 356 struct nilfs_vdesc *vdesc,
357 struct list_head *buffers) 357 struct list_head *buffers)
358 { 358 {
359 struct buffer_head *bh; 359 struct buffer_head *bh;
360 int ret; 360 int ret;
361 361
362 if (vdesc->vd_flags == 0) 362 if (vdesc->vd_flags == 0)
363 ret = nilfs_gccache_submit_read_data( 363 ret = nilfs_gccache_submit_read_data(
364 inode, vdesc->vd_offset, vdesc->vd_blocknr, 364 inode, vdesc->vd_offset, vdesc->vd_blocknr,
365 vdesc->vd_vblocknr, &bh); 365 vdesc->vd_vblocknr, &bh);
366 else 366 else
367 ret = nilfs_gccache_submit_read_node( 367 ret = nilfs_gccache_submit_read_node(
368 inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh); 368 inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh);
369 369
370 if (unlikely(ret < 0)) { 370 if (unlikely(ret < 0)) {
371 if (ret == -ENOENT) 371 if (ret == -ENOENT)
372 printk(KERN_CRIT 372 printk(KERN_CRIT
373 "%s: invalid virtual block address (%s): " 373 "%s: invalid virtual block address (%s): "
374 "ino=%llu, cno=%llu, offset=%llu, " 374 "ino=%llu, cno=%llu, offset=%llu, "
375 "blocknr=%llu, vblocknr=%llu\n", 375 "blocknr=%llu, vblocknr=%llu\n",
376 __func__, vdesc->vd_flags ? "node" : "data", 376 __func__, vdesc->vd_flags ? "node" : "data",
377 (unsigned long long)vdesc->vd_ino, 377 (unsigned long long)vdesc->vd_ino,
378 (unsigned long long)vdesc->vd_cno, 378 (unsigned long long)vdesc->vd_cno,
379 (unsigned long long)vdesc->vd_offset, 379 (unsigned long long)vdesc->vd_offset,
380 (unsigned long long)vdesc->vd_blocknr, 380 (unsigned long long)vdesc->vd_blocknr,
381 (unsigned long long)vdesc->vd_vblocknr); 381 (unsigned long long)vdesc->vd_vblocknr);
382 return ret; 382 return ret;
383 } 383 }
384 if (unlikely(!list_empty(&bh->b_assoc_buffers))) { 384 if (unlikely(!list_empty(&bh->b_assoc_buffers))) {
385 printk(KERN_CRIT "%s: conflicting %s buffer: ino=%llu, " 385 printk(KERN_CRIT "%s: conflicting %s buffer: ino=%llu, "
386 "cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu\n", 386 "cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu\n",
387 __func__, vdesc->vd_flags ? "node" : "data", 387 __func__, vdesc->vd_flags ? "node" : "data",
388 (unsigned long long)vdesc->vd_ino, 388 (unsigned long long)vdesc->vd_ino,
389 (unsigned long long)vdesc->vd_cno, 389 (unsigned long long)vdesc->vd_cno,
390 (unsigned long long)vdesc->vd_offset, 390 (unsigned long long)vdesc->vd_offset,
391 (unsigned long long)vdesc->vd_blocknr, 391 (unsigned long long)vdesc->vd_blocknr,
392 (unsigned long long)vdesc->vd_vblocknr); 392 (unsigned long long)vdesc->vd_vblocknr);
393 brelse(bh); 393 brelse(bh);
394 return -EEXIST; 394 return -EEXIST;
395 } 395 }
396 list_add_tail(&bh->b_assoc_buffers, buffers); 396 list_add_tail(&bh->b_assoc_buffers, buffers);
397 return 0; 397 return 0;
398 } 398 }
399 399
400 static int nilfs_ioctl_move_blocks(struct super_block *sb, 400 static int nilfs_ioctl_move_blocks(struct super_block *sb,
401 struct nilfs_argv *argv, void *buf) 401 struct nilfs_argv *argv, void *buf)
402 { 402 {
403 size_t nmembs = argv->v_nmembs; 403 size_t nmembs = argv->v_nmembs;
404 struct the_nilfs *nilfs = sb->s_fs_info; 404 struct the_nilfs *nilfs = sb->s_fs_info;
405 struct inode *inode; 405 struct inode *inode;
406 struct nilfs_vdesc *vdesc; 406 struct nilfs_vdesc *vdesc;
407 struct buffer_head *bh, *n; 407 struct buffer_head *bh, *n;
408 LIST_HEAD(buffers); 408 LIST_HEAD(buffers);
409 ino_t ino; 409 ino_t ino;
410 __u64 cno; 410 __u64 cno;
411 int i, ret; 411 int i, ret;
412 412
413 for (i = 0, vdesc = buf; i < nmembs; ) { 413 for (i = 0, vdesc = buf; i < nmembs; ) {
414 ino = vdesc->vd_ino; 414 ino = vdesc->vd_ino;
415 cno = vdesc->vd_cno; 415 cno = vdesc->vd_cno;
416 inode = nilfs_iget_for_gc(sb, ino, cno); 416 inode = nilfs_iget_for_gc(sb, ino, cno);
417 if (IS_ERR(inode)) { 417 if (IS_ERR(inode)) {
418 ret = PTR_ERR(inode); 418 ret = PTR_ERR(inode);
419 goto failed; 419 goto failed;
420 } 420 }
421 if (list_empty(&NILFS_I(inode)->i_dirty)) { 421 if (list_empty(&NILFS_I(inode)->i_dirty)) {
422 /* 422 /*
423 * Add the inode to GC inode list. Garbage Collection 423 * Add the inode to GC inode list. Garbage Collection
424 * is serialized and no two processes manipulate the 424 * is serialized and no two processes manipulate the
425 * list simultaneously. 425 * list simultaneously.
426 */ 426 */
427 igrab(inode); 427 igrab(inode);
428 list_add(&NILFS_I(inode)->i_dirty, 428 list_add(&NILFS_I(inode)->i_dirty,
429 &nilfs->ns_gc_inodes); 429 &nilfs->ns_gc_inodes);
430 } 430 }
431 431
432 do { 432 do {
433 ret = nilfs_ioctl_move_inode_block(inode, vdesc, 433 ret = nilfs_ioctl_move_inode_block(inode, vdesc,
434 &buffers); 434 &buffers);
435 if (unlikely(ret < 0)) { 435 if (unlikely(ret < 0)) {
436 iput(inode); 436 iput(inode);
437 goto failed; 437 goto failed;
438 } 438 }
439 vdesc++; 439 vdesc++;
440 } while (++i < nmembs && 440 } while (++i < nmembs &&
441 vdesc->vd_ino == ino && vdesc->vd_cno == cno); 441 vdesc->vd_ino == ino && vdesc->vd_cno == cno);
442 442
443 iput(inode); /* The inode still remains in GC inode list */ 443 iput(inode); /* The inode still remains in GC inode list */
444 } 444 }
445 445
446 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) { 446 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
447 ret = nilfs_gccache_wait_and_mark_dirty(bh); 447 ret = nilfs_gccache_wait_and_mark_dirty(bh);
448 if (unlikely(ret < 0)) { 448 if (unlikely(ret < 0)) {
449 WARN_ON(ret == -EEXIST); 449 WARN_ON(ret == -EEXIST);
450 goto failed; 450 goto failed;
451 } 451 }
452 list_del_init(&bh->b_assoc_buffers); 452 list_del_init(&bh->b_assoc_buffers);
453 brelse(bh); 453 brelse(bh);
454 } 454 }
455 return nmembs; 455 return nmembs;
456 456
457 failed: 457 failed:
458 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) { 458 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
459 list_del_init(&bh->b_assoc_buffers); 459 list_del_init(&bh->b_assoc_buffers);
460 brelse(bh); 460 brelse(bh);
461 } 461 }
462 return ret; 462 return ret;
463 } 463 }
464 464
465 static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs, 465 static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs,
466 struct nilfs_argv *argv, void *buf) 466 struct nilfs_argv *argv, void *buf)
467 { 467 {
468 size_t nmembs = argv->v_nmembs; 468 size_t nmembs = argv->v_nmembs;
469 struct inode *cpfile = nilfs->ns_cpfile; 469 struct inode *cpfile = nilfs->ns_cpfile;
470 struct nilfs_period *periods = buf; 470 struct nilfs_period *periods = buf;
471 int ret, i; 471 int ret, i;
472 472
473 for (i = 0; i < nmembs; i++) { 473 for (i = 0; i < nmembs; i++) {
474 ret = nilfs_cpfile_delete_checkpoints( 474 ret = nilfs_cpfile_delete_checkpoints(
475 cpfile, periods[i].p_start, periods[i].p_end); 475 cpfile, periods[i].p_start, periods[i].p_end);
476 if (ret < 0) 476 if (ret < 0)
477 return ret; 477 return ret;
478 } 478 }
479 return nmembs; 479 return nmembs;
480 } 480 }
481 481
482 static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs, 482 static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs,
483 struct nilfs_argv *argv, void *buf) 483 struct nilfs_argv *argv, void *buf)
484 { 484 {
485 size_t nmembs = argv->v_nmembs; 485 size_t nmembs = argv->v_nmembs;
486 int ret; 486 int ret;
487 487
488 ret = nilfs_dat_freev(nilfs->ns_dat, buf, nmembs); 488 ret = nilfs_dat_freev(nilfs->ns_dat, buf, nmembs);
489 489
490 return (ret < 0) ? ret : nmembs; 490 return (ret < 0) ? ret : nmembs;
491 } 491 }
492 492
493 static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs, 493 static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs,
494 struct nilfs_argv *argv, void *buf) 494 struct nilfs_argv *argv, void *buf)
495 { 495 {
496 size_t nmembs = argv->v_nmembs; 496 size_t nmembs = argv->v_nmembs;
497 struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap; 497 struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
498 struct nilfs_bdesc *bdescs = buf; 498 struct nilfs_bdesc *bdescs = buf;
499 int ret, i; 499 int ret, i;
500 500
501 for (i = 0; i < nmembs; i++) { 501 for (i = 0; i < nmembs; i++) {
502 /* XXX: use macro or inline func to check liveness */ 502 /* XXX: use macro or inline func to check liveness */
503 ret = nilfs_bmap_lookup_at_level(bmap, 503 ret = nilfs_bmap_lookup_at_level(bmap,
504 bdescs[i].bd_offset, 504 bdescs[i].bd_offset,
505 bdescs[i].bd_level + 1, 505 bdescs[i].bd_level + 1,
506 &bdescs[i].bd_blocknr); 506 &bdescs[i].bd_blocknr);
507 if (ret < 0) { 507 if (ret < 0) {
508 if (ret != -ENOENT) 508 if (ret != -ENOENT)
509 return ret; 509 return ret;
510 bdescs[i].bd_blocknr = 0; 510 bdescs[i].bd_blocknr = 0;
511 } 511 }
512 if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr) 512 if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr)
513 /* skip dead block */ 513 /* skip dead block */
514 continue; 514 continue;
515 if (bdescs[i].bd_level == 0) { 515 if (bdescs[i].bd_level == 0) {
516 ret = nilfs_mdt_mark_block_dirty(nilfs->ns_dat, 516 ret = nilfs_mdt_mark_block_dirty(nilfs->ns_dat,
517 bdescs[i].bd_offset); 517 bdescs[i].bd_offset);
518 if (ret < 0) { 518 if (ret < 0) {
519 WARN_ON(ret == -ENOENT); 519 WARN_ON(ret == -ENOENT);
520 return ret; 520 return ret;
521 } 521 }
522 } else { 522 } else {
523 ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset, 523 ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset,
524 bdescs[i].bd_level); 524 bdescs[i].bd_level);
525 if (ret < 0) { 525 if (ret < 0) {
526 WARN_ON(ret == -ENOENT); 526 WARN_ON(ret == -ENOENT);
527 return ret; 527 return ret;
528 } 528 }
529 } 529 }
530 } 530 }
531 return nmembs; 531 return nmembs;
532 } 532 }
533 533
534 int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs, 534 int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
535 struct nilfs_argv *argv, void **kbufs) 535 struct nilfs_argv *argv, void **kbufs)
536 { 536 {
537 const char *msg; 537 const char *msg;
538 int ret; 538 int ret;
539 539
540 ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]); 540 ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
541 if (ret < 0) { 541 if (ret < 0) {
542 /* 542 /*
543 * can safely abort because checkpoints can be removed 543 * can safely abort because checkpoints can be removed
544 * independently. 544 * independently.
545 */ 545 */
546 msg = "cannot delete checkpoints"; 546 msg = "cannot delete checkpoints";
547 goto failed; 547 goto failed;
548 } 548 }
549 ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]); 549 ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]);
550 if (ret < 0) { 550 if (ret < 0) {
551 /* 551 /*
552 * can safely abort because DAT file is updated atomically 552 * can safely abort because DAT file is updated atomically
553 * using a copy-on-write technique. 553 * using a copy-on-write technique.
554 */ 554 */
555 msg = "cannot delete virtual blocks from DAT file"; 555 msg = "cannot delete virtual blocks from DAT file";
556 goto failed; 556 goto failed;
557 } 557 }
558 ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]); 558 ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]);
559 if (ret < 0) { 559 if (ret < 0) {
560 /* 560 /*
561 * can safely abort because the operation is nondestructive. 561 * can safely abort because the operation is nondestructive.
562 */ 562 */
563 msg = "cannot mark copying blocks dirty"; 563 msg = "cannot mark copying blocks dirty";
564 goto failed; 564 goto failed;
565 } 565 }
566 return 0; 566 return 0;
567 567
568 failed: 568 failed:
569 printk(KERN_ERR "NILFS: GC failed during preparation: %s: err=%d\n", 569 printk(KERN_ERR "NILFS: GC failed during preparation: %s: err=%d\n",
570 msg, ret); 570 msg, ret);
571 return ret; 571 return ret;
572 } 572 }
573 573
574 static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp, 574 static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
575 unsigned int cmd, void __user *argp) 575 unsigned int cmd, void __user *argp)
576 { 576 {
577 struct nilfs_argv argv[5]; 577 struct nilfs_argv argv[5];
578 static const size_t argsz[5] = { 578 static const size_t argsz[5] = {
579 sizeof(struct nilfs_vdesc), 579 sizeof(struct nilfs_vdesc),
580 sizeof(struct nilfs_period), 580 sizeof(struct nilfs_period),
581 sizeof(__u64), 581 sizeof(__u64),
582 sizeof(struct nilfs_bdesc), 582 sizeof(struct nilfs_bdesc),
583 sizeof(__u64), 583 sizeof(__u64),
584 }; 584 };
585 void __user *base; 585 void __user *base;
586 void *kbufs[5]; 586 void *kbufs[5];
587 struct the_nilfs *nilfs; 587 struct the_nilfs *nilfs;
588 size_t len, nsegs; 588 size_t len, nsegs;
589 int n, ret; 589 int n, ret;
590 590
591 if (!capable(CAP_SYS_ADMIN)) 591 if (!capable(CAP_SYS_ADMIN))
592 return -EPERM; 592 return -EPERM;
593 593
594 ret = mnt_want_write_file(filp); 594 ret = mnt_want_write_file(filp);
595 if (ret) 595 if (ret)
596 return ret; 596 return ret;
597 597
598 ret = -EFAULT; 598 ret = -EFAULT;
599 if (copy_from_user(argv, argp, sizeof(argv))) 599 if (copy_from_user(argv, argp, sizeof(argv)))
600 goto out; 600 goto out;
601 601
602 ret = -EINVAL; 602 ret = -EINVAL;
603 nsegs = argv[4].v_nmembs; 603 nsegs = argv[4].v_nmembs;
604 if (argv[4].v_size != argsz[4]) 604 if (argv[4].v_size != argsz[4])
605 goto out; 605 goto out;
606 if (nsegs > UINT_MAX / sizeof(__u64)) 606 if (nsegs > UINT_MAX / sizeof(__u64))
607 goto out; 607 goto out;
608 608
609 /* 609 /*
610 * argv[4] points to segment numbers this ioctl cleans. We 610 * argv[4] points to segment numbers this ioctl cleans. We
611 * use kmalloc() for its buffer because memory used for the 611 * use kmalloc() for its buffer because memory used for the
612 * segment numbers is enough small. 612 * segment numbers is enough small.
613 */ 613 */
614 kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base, 614 kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base,
615 nsegs * sizeof(__u64)); 615 nsegs * sizeof(__u64));
616 if (IS_ERR(kbufs[4])) { 616 if (IS_ERR(kbufs[4])) {
617 ret = PTR_ERR(kbufs[4]); 617 ret = PTR_ERR(kbufs[4]);
618 goto out; 618 goto out;
619 } 619 }
620 nilfs = inode->i_sb->s_fs_info; 620 nilfs = inode->i_sb->s_fs_info;
621 621
622 for (n = 0; n < 4; n++) { 622 for (n = 0; n < 4; n++) {
623 ret = -EINVAL; 623 ret = -EINVAL;
624 if (argv[n].v_size != argsz[n]) 624 if (argv[n].v_size != argsz[n])
625 goto out_free; 625 goto out_free;
626 626
627 if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment) 627 if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment)
628 goto out_free; 628 goto out_free;
629 629
630 if (argv[n].v_nmembs >= UINT_MAX / argv[n].v_size) 630 if (argv[n].v_nmembs >= UINT_MAX / argv[n].v_size)
631 goto out_free; 631 goto out_free;
632 632
633 len = argv[n].v_size * argv[n].v_nmembs; 633 len = argv[n].v_size * argv[n].v_nmembs;
634 base = (void __user *)(unsigned long)argv[n].v_base; 634 base = (void __user *)(unsigned long)argv[n].v_base;
635 if (len == 0) { 635 if (len == 0) {
636 kbufs[n] = NULL; 636 kbufs[n] = NULL;
637 continue; 637 continue;
638 } 638 }
639 639
640 kbufs[n] = vmalloc(len); 640 kbufs[n] = vmalloc(len);
641 if (!kbufs[n]) { 641 if (!kbufs[n]) {
642 ret = -ENOMEM; 642 ret = -ENOMEM;
643 goto out_free; 643 goto out_free;
644 } 644 }
645 if (copy_from_user(kbufs[n], base, len)) { 645 if (copy_from_user(kbufs[n], base, len)) {
646 ret = -EFAULT; 646 ret = -EFAULT;
647 vfree(kbufs[n]); 647 vfree(kbufs[n]);
648 goto out_free; 648 goto out_free;
649 } 649 }
650 } 650 }
651 651
652 /* 652 /*
653 * nilfs_ioctl_move_blocks() will call nilfs_iget_for_gc(), 653 * nilfs_ioctl_move_blocks() will call nilfs_iget_for_gc(),
654 * which will operates an inode list without blocking. 654 * which will operates an inode list without blocking.
655 * To protect the list from concurrent operations, 655 * To protect the list from concurrent operations,
656 * nilfs_ioctl_move_blocks should be atomic operation. 656 * nilfs_ioctl_move_blocks should be atomic operation.
657 */ 657 */
658 if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) { 658 if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
659 ret = -EBUSY; 659 ret = -EBUSY;
660 goto out_free; 660 goto out_free;
661 } 661 }
662 662
663 ret = nilfs_ioctl_move_blocks(inode->i_sb, &argv[0], kbufs[0]); 663 ret = nilfs_ioctl_move_blocks(inode->i_sb, &argv[0], kbufs[0]);
664 if (ret < 0) 664 if (ret < 0)
665 printk(KERN_ERR "NILFS: GC failed during preparation: " 665 printk(KERN_ERR "NILFS: GC failed during preparation: "
666 "cannot read source blocks: err=%d\n", ret); 666 "cannot read source blocks: err=%d\n", ret);
667 else 667 else {
668 if (nilfs_sb_need_update(nilfs))
669 set_nilfs_discontinued(nilfs);
668 ret = nilfs_clean_segments(inode->i_sb, argv, kbufs); 670 ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
671 }
669 672
670 nilfs_remove_all_gcinodes(nilfs); 673 nilfs_remove_all_gcinodes(nilfs);
671 clear_nilfs_gc_running(nilfs); 674 clear_nilfs_gc_running(nilfs);
672 675
673 out_free: 676 out_free:
674 while (--n >= 0) 677 while (--n >= 0)
675 vfree(kbufs[n]); 678 vfree(kbufs[n]);
676 kfree(kbufs[4]); 679 kfree(kbufs[4]);
677 out: 680 out:
678 mnt_drop_write_file(filp); 681 mnt_drop_write_file(filp);
679 return ret; 682 return ret;
680 } 683 }
681 684
682 static int nilfs_ioctl_sync(struct inode *inode, struct file *filp, 685 static int nilfs_ioctl_sync(struct inode *inode, struct file *filp,
683 unsigned int cmd, void __user *argp) 686 unsigned int cmd, void __user *argp)
684 { 687 {
685 __u64 cno; 688 __u64 cno;
686 int ret; 689 int ret;
687 struct the_nilfs *nilfs; 690 struct the_nilfs *nilfs;
688 691
689 ret = nilfs_construct_segment(inode->i_sb); 692 ret = nilfs_construct_segment(inode->i_sb);
690 if (ret < 0) 693 if (ret < 0)
691 return ret; 694 return ret;
692 695
693 nilfs = inode->i_sb->s_fs_info; 696 nilfs = inode->i_sb->s_fs_info;
694 if (nilfs_test_opt(nilfs, BARRIER)) { 697 if (nilfs_test_opt(nilfs, BARRIER)) {
695 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); 698 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
696 if (ret == -EIO) 699 if (ret == -EIO)
697 return ret; 700 return ret;
698 } 701 }
699 702
700 if (argp != NULL) { 703 if (argp != NULL) {
701 down_read(&nilfs->ns_segctor_sem); 704 down_read(&nilfs->ns_segctor_sem);
702 cno = nilfs->ns_cno - 1; 705 cno = nilfs->ns_cno - 1;
703 up_read(&nilfs->ns_segctor_sem); 706 up_read(&nilfs->ns_segctor_sem);
704 if (copy_to_user(argp, &cno, sizeof(cno))) 707 if (copy_to_user(argp, &cno, sizeof(cno)))
705 return -EFAULT; 708 return -EFAULT;
706 } 709 }
707 return 0; 710 return 0;
708 } 711 }
709 712
710 static int nilfs_ioctl_resize(struct inode *inode, struct file *filp, 713 static int nilfs_ioctl_resize(struct inode *inode, struct file *filp,
711 void __user *argp) 714 void __user *argp)
712 { 715 {
713 __u64 newsize; 716 __u64 newsize;
714 int ret = -EPERM; 717 int ret = -EPERM;
715 718
716 if (!capable(CAP_SYS_ADMIN)) 719 if (!capable(CAP_SYS_ADMIN))
717 goto out; 720 goto out;
718 721
719 ret = mnt_want_write_file(filp); 722 ret = mnt_want_write_file(filp);
720 if (ret) 723 if (ret)
721 goto out; 724 goto out;
722 725
723 ret = -EFAULT; 726 ret = -EFAULT;
724 if (copy_from_user(&newsize, argp, sizeof(newsize))) 727 if (copy_from_user(&newsize, argp, sizeof(newsize)))
725 goto out_drop_write; 728 goto out_drop_write;
726 729
727 ret = nilfs_resize_fs(inode->i_sb, newsize); 730 ret = nilfs_resize_fs(inode->i_sb, newsize);
728 731
729 out_drop_write: 732 out_drop_write:
730 mnt_drop_write_file(filp); 733 mnt_drop_write_file(filp);
731 out: 734 out:
732 return ret; 735 return ret;
733 } 736 }
734 737
735 static int nilfs_ioctl_set_alloc_range(struct inode *inode, void __user *argp) 738 static int nilfs_ioctl_set_alloc_range(struct inode *inode, void __user *argp)
736 { 739 {
737 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 740 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
738 __u64 range[2]; 741 __u64 range[2];
739 __u64 minseg, maxseg; 742 __u64 minseg, maxseg;
740 unsigned long segbytes; 743 unsigned long segbytes;
741 int ret = -EPERM; 744 int ret = -EPERM;
742 745
743 if (!capable(CAP_SYS_ADMIN)) 746 if (!capable(CAP_SYS_ADMIN))
744 goto out; 747 goto out;
745 748
746 ret = -EFAULT; 749 ret = -EFAULT;
747 if (copy_from_user(range, argp, sizeof(__u64[2]))) 750 if (copy_from_user(range, argp, sizeof(__u64[2])))
748 goto out; 751 goto out;
749 752
750 ret = -ERANGE; 753 ret = -ERANGE;
751 if (range[1] > i_size_read(inode->i_sb->s_bdev->bd_inode)) 754 if (range[1] > i_size_read(inode->i_sb->s_bdev->bd_inode))
752 goto out; 755 goto out;
753 756
754 segbytes = nilfs->ns_blocks_per_segment * nilfs->ns_blocksize; 757 segbytes = nilfs->ns_blocks_per_segment * nilfs->ns_blocksize;
755 758
756 minseg = range[0] + segbytes - 1; 759 minseg = range[0] + segbytes - 1;
757 do_div(minseg, segbytes); 760 do_div(minseg, segbytes);
758 maxseg = NILFS_SB2_OFFSET_BYTES(range[1]); 761 maxseg = NILFS_SB2_OFFSET_BYTES(range[1]);
759 do_div(maxseg, segbytes); 762 do_div(maxseg, segbytes);
760 maxseg--; 763 maxseg--;
761 764
762 ret = nilfs_sufile_set_alloc_range(nilfs->ns_sufile, minseg, maxseg); 765 ret = nilfs_sufile_set_alloc_range(nilfs->ns_sufile, minseg, maxseg);
763 out: 766 out:
764 return ret; 767 return ret;
765 } 768 }
766 769
767 static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp, 770 static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp,
768 unsigned int cmd, void __user *argp, 771 unsigned int cmd, void __user *argp,
769 size_t membsz, 772 size_t membsz,
770 ssize_t (*dofunc)(struct the_nilfs *, 773 ssize_t (*dofunc)(struct the_nilfs *,
771 __u64 *, int, 774 __u64 *, int,
772 void *, size_t, size_t)) 775 void *, size_t, size_t))
773 776
774 { 777 {
775 struct the_nilfs *nilfs = inode->i_sb->s_fs_info; 778 struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
776 struct nilfs_argv argv; 779 struct nilfs_argv argv;
777 int ret; 780 int ret;
778 781
779 if (copy_from_user(&argv, argp, sizeof(argv))) 782 if (copy_from_user(&argv, argp, sizeof(argv)))
780 return -EFAULT; 783 return -EFAULT;
781 784
782 if (argv.v_size < membsz) 785 if (argv.v_size < membsz)
783 return -EINVAL; 786 return -EINVAL;
784 787
785 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc); 788 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc);
786 if (ret < 0) 789 if (ret < 0)
787 return ret; 790 return ret;
788 791
789 if (copy_to_user(argp, &argv, sizeof(argv))) 792 if (copy_to_user(argp, &argv, sizeof(argv)))
790 ret = -EFAULT; 793 ret = -EFAULT;
791 return ret; 794 return ret;
792 } 795 }
793 796
794 long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 797 long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
795 { 798 {
796 struct inode *inode = filp->f_dentry->d_inode; 799 struct inode *inode = filp->f_dentry->d_inode;
797 void __user *argp = (void __user *)arg; 800 void __user *argp = (void __user *)arg;
798 801
799 switch (cmd) { 802 switch (cmd) {
800 case FS_IOC_GETFLAGS: 803 case FS_IOC_GETFLAGS:
801 return nilfs_ioctl_getflags(inode, argp); 804 return nilfs_ioctl_getflags(inode, argp);
802 case FS_IOC_SETFLAGS: 805 case FS_IOC_SETFLAGS:
803 return nilfs_ioctl_setflags(inode, filp, argp); 806 return nilfs_ioctl_setflags(inode, filp, argp);
804 case FS_IOC_GETVERSION: 807 case FS_IOC_GETVERSION:
805 return nilfs_ioctl_getversion(inode, argp); 808 return nilfs_ioctl_getversion(inode, argp);
806 case NILFS_IOCTL_CHANGE_CPMODE: 809 case NILFS_IOCTL_CHANGE_CPMODE:
807 return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp); 810 return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp);
808 case NILFS_IOCTL_DELETE_CHECKPOINT: 811 case NILFS_IOCTL_DELETE_CHECKPOINT:
809 return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp); 812 return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp);
810 case NILFS_IOCTL_GET_CPINFO: 813 case NILFS_IOCTL_GET_CPINFO:
811 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 814 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
812 sizeof(struct nilfs_cpinfo), 815 sizeof(struct nilfs_cpinfo),
813 nilfs_ioctl_do_get_cpinfo); 816 nilfs_ioctl_do_get_cpinfo);
814 case NILFS_IOCTL_GET_CPSTAT: 817 case NILFS_IOCTL_GET_CPSTAT:
815 return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp); 818 return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp);
816 case NILFS_IOCTL_GET_SUINFO: 819 case NILFS_IOCTL_GET_SUINFO:
817 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 820 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
818 sizeof(struct nilfs_suinfo), 821 sizeof(struct nilfs_suinfo),
819 nilfs_ioctl_do_get_suinfo); 822 nilfs_ioctl_do_get_suinfo);
820 case NILFS_IOCTL_GET_SUSTAT: 823 case NILFS_IOCTL_GET_SUSTAT:
821 return nilfs_ioctl_get_sustat(inode, filp, cmd, argp); 824 return nilfs_ioctl_get_sustat(inode, filp, cmd, argp);
822 case NILFS_IOCTL_GET_VINFO: 825 case NILFS_IOCTL_GET_VINFO:
823 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 826 return nilfs_ioctl_get_info(inode, filp, cmd, argp,
824 sizeof(struct nilfs_vinfo), 827 sizeof(struct nilfs_vinfo),
825 nilfs_ioctl_do_get_vinfo); 828 nilfs_ioctl_do_get_vinfo);
826 case NILFS_IOCTL_GET_BDESCS: 829 case NILFS_IOCTL_GET_BDESCS:
827 return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp); 830 return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp);
828 case NILFS_IOCTL_CLEAN_SEGMENTS: 831 case NILFS_IOCTL_CLEAN_SEGMENTS:
829 return nilfs_ioctl_clean_segments(inode, filp, cmd, argp); 832 return nilfs_ioctl_clean_segments(inode, filp, cmd, argp);
830 case NILFS_IOCTL_SYNC: 833 case NILFS_IOCTL_SYNC:
831 return nilfs_ioctl_sync(inode, filp, cmd, argp); 834 return nilfs_ioctl_sync(inode, filp, cmd, argp);
832 case NILFS_IOCTL_RESIZE: 835 case NILFS_IOCTL_RESIZE:
833 return nilfs_ioctl_resize(inode, filp, argp); 836 return nilfs_ioctl_resize(inode, filp, argp);
834 case NILFS_IOCTL_SET_ALLOC_RANGE: 837 case NILFS_IOCTL_SET_ALLOC_RANGE:
835 return nilfs_ioctl_set_alloc_range(inode, argp); 838 return nilfs_ioctl_set_alloc_range(inode, argp);
836 default: 839 default:
837 return -ENOTTY; 840 return -ENOTTY;
838 } 841 }
839 } 842 }
840 843
841 #ifdef CONFIG_COMPAT 844 #ifdef CONFIG_COMPAT
842 long nilfs_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 845 long nilfs_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
843 { 846 {
844 switch (cmd) { 847 switch (cmd) {
845 case FS_IOC32_GETFLAGS: 848 case FS_IOC32_GETFLAGS:
846 cmd = FS_IOC_GETFLAGS; 849 cmd = FS_IOC_GETFLAGS;
847 break; 850 break;
848 case FS_IOC32_SETFLAGS: 851 case FS_IOC32_SETFLAGS:
849 cmd = FS_IOC_SETFLAGS; 852 cmd = FS_IOC_SETFLAGS;
850 break; 853 break;
851 case FS_IOC32_GETVERSION: 854 case FS_IOC32_GETVERSION:
852 cmd = FS_IOC_GETVERSION; 855 cmd = FS_IOC_GETVERSION;
853 break; 856 break;
854 case NILFS_IOCTL_CHANGE_CPMODE: 857 case NILFS_IOCTL_CHANGE_CPMODE:
855 case NILFS_IOCTL_DELETE_CHECKPOINT: 858 case NILFS_IOCTL_DELETE_CHECKPOINT:
856 case NILFS_IOCTL_GET_CPINFO: 859 case NILFS_IOCTL_GET_CPINFO:
857 case NILFS_IOCTL_GET_CPSTAT: 860 case NILFS_IOCTL_GET_CPSTAT:
858 case NILFS_IOCTL_GET_SUINFO: 861 case NILFS_IOCTL_GET_SUINFO:
859 case NILFS_IOCTL_GET_SUSTAT: 862 case NILFS_IOCTL_GET_SUSTAT:
860 case NILFS_IOCTL_GET_VINFO: 863 case NILFS_IOCTL_GET_VINFO:
861 case NILFS_IOCTL_GET_BDESCS: 864 case NILFS_IOCTL_GET_BDESCS:
862 case NILFS_IOCTL_CLEAN_SEGMENTS: 865 case NILFS_IOCTL_CLEAN_SEGMENTS:
863 case NILFS_IOCTL_SYNC: 866 case NILFS_IOCTL_SYNC:
864 case NILFS_IOCTL_RESIZE: 867 case NILFS_IOCTL_RESIZE:
865 case NILFS_IOCTL_SET_ALLOC_RANGE: 868 case NILFS_IOCTL_SET_ALLOC_RANGE:
866 break; 869 break;
867 default: 870 default:
868 return -ENOIOCTLCMD; 871 return -ENOIOCTLCMD;
869 } 872 }
870 return nilfs_ioctl(filp, cmd, (unsigned long)compat_ptr(arg)); 873 return nilfs_ioctl(filp, cmd, (unsigned long)compat_ptr(arg));
871 } 874 }
872 #endif 875 #endif
873 876