blob: 0a512aa81bf7505ab8803cc90191249f91cb4a27 [file] [log] [blame]
Johann Lombardic5e06d12011-05-24 18:31:25 -04001#include <linux/fs.h>
2#include <linux/random.h>
3#include <linux/buffer_head.h>
4#include <linux/utsname.h>
5#include <linux/kthread.h>
6
7#include "ext4.h"
8
Darrick J. Wong5c359a42012-04-29 18:47:10 -04009/* Checksumming functions */
Dmitry Monakhov171a7f22013-04-09 23:56:48 -040010static __le32 ext4_mmp_csum(struct super_block *sb, struct mmp_struct *mmp)
Darrick J. Wong5c359a42012-04-29 18:47:10 -040011{
12 struct ext4_sb_info *sbi = EXT4_SB(sb);
13 int offset = offsetof(struct mmp_struct, mmp_checksum);
14 __u32 csum;
15
16 csum = ext4_chksum(sbi, sbi->s_csum_seed, (char *)mmp, offset);
17
18 return cpu_to_le32(csum);
19}
20
Stephen Hemmingerc1978552014-05-12 10:50:23 -040021static int ext4_mmp_csum_verify(struct super_block *sb, struct mmp_struct *mmp)
Darrick J. Wong5c359a42012-04-29 18:47:10 -040022{
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -040023 if (!ext4_has_metadata_csum(sb))
Darrick J. Wong5c359a42012-04-29 18:47:10 -040024 return 1;
25
26 return mmp->mmp_checksum == ext4_mmp_csum(sb, mmp);
27}
28
Stephen Hemmingerc1978552014-05-12 10:50:23 -040029static void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp)
Darrick J. Wong5c359a42012-04-29 18:47:10 -040030{
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -040031 if (!ext4_has_metadata_csum(sb))
Darrick J. Wong5c359a42012-04-29 18:47:10 -040032 return;
33
34 mmp->mmp_checksum = ext4_mmp_csum(sb, mmp);
35}
36
Johann Lombardic5e06d12011-05-24 18:31:25 -040037/*
38 * Write the MMP block using WRITE_SYNC to try to get the block on-disk
39 * faster.
40 */
Darrick J. Wong5c359a42012-04-29 18:47:10 -040041static int write_mmp_block(struct super_block *sb, struct buffer_head *bh)
Johann Lombardic5e06d12011-05-24 18:31:25 -040042{
Darrick J. Wong5c359a42012-04-29 18:47:10 -040043 struct mmp_struct *mmp = (struct mmp_struct *)(bh->b_data);
44
Jan Kara8e8ad8a2012-06-12 16:20:38 +020045 /*
46 * We protect against freezing so that we don't create dirty buffers
47 * on frozen filesystem.
48 */
49 sb_start_write(sb);
Darrick J. Wong5c359a42012-04-29 18:47:10 -040050 ext4_mmp_csum_set(sb, mmp);
Johann Lombardic5e06d12011-05-24 18:31:25 -040051 mark_buffer_dirty(bh);
52 lock_buffer(bh);
53 bh->b_end_io = end_buffer_write_sync;
54 get_bh(bh);
Theodore Ts'o9f203502013-04-20 15:46:17 -040055 submit_bh(WRITE_SYNC | REQ_META | REQ_PRIO, bh);
Johann Lombardic5e06d12011-05-24 18:31:25 -040056 wait_on_buffer(bh);
Jan Kara8e8ad8a2012-06-12 16:20:38 +020057 sb_end_write(sb);
Johann Lombardic5e06d12011-05-24 18:31:25 -040058 if (unlikely(!buffer_uptodate(bh)))
59 return 1;
60
61 return 0;
62}
63
64/*
65 * Read the MMP block. It _must_ be read from disk and hence we clear the
66 * uptodate flag on the buffer.
67 */
68static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
69 ext4_fsblk_t mmp_block)
70{
71 struct mmp_struct *mmp;
Dan Carpenter98104462015-08-15 11:30:31 -040072 int ret;
Johann Lombardic5e06d12011-05-24 18:31:25 -040073
74 if (*bh)
75 clear_buffer_uptodate(*bh);
76
77 /* This would be sb_bread(sb, mmp_block), except we need to be sure
78 * that the MD RAID device cache has been bypassed, and that the read
79 * is not blocked in the elevator. */
Dan Carpenter98104462015-08-15 11:30:31 -040080 if (!*bh) {
Johann Lombardic5e06d12011-05-24 18:31:25 -040081 *bh = sb_getblk(sb, mmp_block);
Dan Carpenter98104462015-08-15 11:30:31 -040082 if (!*bh) {
83 ret = -ENOMEM;
84 goto warn_exit;
Johann Lombardic5e06d12011-05-24 18:31:25 -040085 }
86 }
Dan Carpenter98104462015-08-15 11:30:31 -040087
88 get_bh(*bh);
89 lock_buffer(*bh);
90 (*bh)->b_end_io = end_buffer_read_sync;
91 submit_bh(READ_SYNC | REQ_META | REQ_PRIO, *bh);
92 wait_on_buffer(*bh);
93 if (!buffer_uptodate(*bh)) {
94 brelse(*bh);
95 *bh = NULL;
96 ret = -EIO;
97 goto warn_exit;
Johann Lombardic5e06d12011-05-24 18:31:25 -040098 }
99
100 mmp = (struct mmp_struct *)((*bh)->b_data);
Darrick J. Wong6a797d22015-10-17 16:16:04 -0400101 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC)
102 ret = -EFSCORRUPTED;
103 else if (!ext4_mmp_csum_verify(sb, mmp))
104 ret = -EFSBADCRC;
105 else
Dan Carpenter98104462015-08-15 11:30:31 -0400106 return 0;
Johann Lombardic5e06d12011-05-24 18:31:25 -0400107
Dan Carpenter98104462015-08-15 11:30:31 -0400108warn_exit:
109 ext4_warning(sb, "Error %d while reading MMP block %llu",
110 ret, mmp_block);
111 return ret;
Johann Lombardic5e06d12011-05-24 18:31:25 -0400112}
113
114/*
115 * Dump as much information as possible to help the admin.
116 */
117void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
118 const char *function, unsigned int line, const char *msg)
119{
Dan Carpenterda0b5e42015-08-15 11:38:13 -0400120 __ext4_warning(sb, function, line, "%s", msg);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400121 __ext4_warning(sb, function, line,
122 "MMP failure info: last update time: %llu, last update "
123 "node: %s, last update device: %s\n",
124 (long long unsigned int) le64_to_cpu(mmp->mmp_time),
125 mmp->mmp_nodename, mmp->mmp_bdevname);
126}
127
128/*
129 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
130 */
131static int kmmpd(void *data)
132{
133 struct super_block *sb = ((struct mmpd_data *) data)->sb;
134 struct buffer_head *bh = ((struct mmpd_data *) data)->bh;
135 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
136 struct mmp_struct *mmp;
137 ext4_fsblk_t mmp_block;
138 u32 seq = 0;
139 unsigned long failed_writes = 0;
140 int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
141 unsigned mmp_check_interval;
142 unsigned long last_update_time;
143 unsigned long diff;
144 int retval;
145
146 mmp_block = le64_to_cpu(es->s_mmp_block);
147 mmp = (struct mmp_struct *)(bh->b_data);
148 mmp->mmp_time = cpu_to_le64(get_seconds());
149 /*
150 * Start with the higher mmp_check_interval and reduce it if
151 * the MMP block is being updated on time.
152 */
153 mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
154 EXT4_MMP_MIN_CHECK_INTERVAL);
155 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
156 bdevname(bh->b_bdev, mmp->mmp_bdevname);
157
Nikitas Angelinas215fc6a2011-10-18 10:49:51 -0400158 memcpy(mmp->mmp_nodename, init_utsname()->nodename,
Johann Lombardic5e06d12011-05-24 18:31:25 -0400159 sizeof(mmp->mmp_nodename));
160
161 while (!kthread_should_stop()) {
162 if (++seq > EXT4_MMP_SEQ_MAX)
163 seq = 1;
164
165 mmp->mmp_seq = cpu_to_le32(seq);
166 mmp->mmp_time = cpu_to_le64(get_seconds());
167 last_update_time = jiffies;
168
Darrick J. Wong5c359a42012-04-29 18:47:10 -0400169 retval = write_mmp_block(sb, bh);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400170 /*
171 * Don't spew too many error messages. Print one every
172 * (s_mmp_update_interval * 60) seconds.
173 */
Nikitas Angelinasbdfc2302011-10-18 10:51:51 -0400174 if (retval) {
175 if ((failed_writes % 60) == 0)
176 ext4_error(sb, "Error writing to MMP block");
Johann Lombardic5e06d12011-05-24 18:31:25 -0400177 failed_writes++;
178 }
179
180 if (!(le32_to_cpu(es->s_feature_incompat) &
181 EXT4_FEATURE_INCOMPAT_MMP)) {
182 ext4_warning(sb, "kmmpd being stopped since MMP feature"
183 " has been disabled.");
184 EXT4_SB(sb)->s_mmp_tsk = NULL;
185 goto failed;
186 }
187
188 if (sb->s_flags & MS_RDONLY) {
189 ext4_warning(sb, "kmmpd being stopped since filesystem "
190 "has been remounted as readonly.");
191 EXT4_SB(sb)->s_mmp_tsk = NULL;
192 goto failed;
193 }
194
195 diff = jiffies - last_update_time;
196 if (diff < mmp_update_interval * HZ)
197 schedule_timeout_interruptible(mmp_update_interval *
198 HZ - diff);
199
200 /*
201 * We need to make sure that more than mmp_check_interval
202 * seconds have not passed since writing. If that has happened
203 * we need to check if the MMP block is as we left it.
204 */
205 diff = jiffies - last_update_time;
206 if (diff > mmp_check_interval * HZ) {
207 struct buffer_head *bh_check = NULL;
208 struct mmp_struct *mmp_check;
209
210 retval = read_mmp_block(sb, &bh_check, mmp_block);
211 if (retval) {
212 ext4_error(sb, "error reading MMP data: %d",
213 retval);
214
215 EXT4_SB(sb)->s_mmp_tsk = NULL;
216 goto failed;
217 }
218
219 mmp_check = (struct mmp_struct *)(bh_check->b_data);
220 if (mmp->mmp_seq != mmp_check->mmp_seq ||
221 memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
222 sizeof(mmp->mmp_nodename))) {
223 dump_mmp_msg(sb, mmp_check,
224 "Error while updating MMP info. "
225 "The filesystem seems to have been"
226 " multiply mounted.");
227 ext4_error(sb, "abort");
228 goto failed;
229 }
230 put_bh(bh_check);
231 }
232
233 /*
234 * Adjust the mmp_check_interval depending on how much time
235 * it took for the MMP block to be written.
236 */
237 mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
238 EXT4_MMP_MAX_CHECK_INTERVAL),
239 EXT4_MMP_MIN_CHECK_INTERVAL);
240 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
241 }
242
243 /*
244 * Unmount seems to be clean.
245 */
246 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
247 mmp->mmp_time = cpu_to_le64(get_seconds());
248
Darrick J. Wong5c359a42012-04-29 18:47:10 -0400249 retval = write_mmp_block(sb, bh);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400250
251failed:
252 kfree(data);
253 brelse(bh);
254 return retval;
255}
256
257/*
258 * Get a random new sequence number but make sure it is not greater than
259 * EXT4_MMP_SEQ_MAX.
260 */
261static unsigned int mmp_new_seq(void)
262{
263 u32 new_seq;
264
265 do {
Theodore Ts'odd1f7232013-11-08 00:14:53 -0500266 new_seq = prandom_u32();
Johann Lombardic5e06d12011-05-24 18:31:25 -0400267 } while (new_seq > EXT4_MMP_SEQ_MAX);
268
269 return new_seq;
270}
271
272/*
273 * Protect the filesystem from being mounted more than once.
274 */
275int ext4_multi_mount_protect(struct super_block *sb,
276 ext4_fsblk_t mmp_block)
277{
278 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
279 struct buffer_head *bh = NULL;
280 struct mmp_struct *mmp = NULL;
281 struct mmpd_data *mmpd_data;
282 u32 seq;
283 unsigned int mmp_check_interval = le16_to_cpu(es->s_mmp_update_interval);
284 unsigned int wait_time = 0;
285 int retval;
286
287 if (mmp_block < le32_to_cpu(es->s_first_data_block) ||
288 mmp_block >= ext4_blocks_count(es)) {
289 ext4_warning(sb, "Invalid MMP block in superblock");
290 goto failed;
291 }
292
293 retval = read_mmp_block(sb, &bh, mmp_block);
294 if (retval)
295 goto failed;
296
297 mmp = (struct mmp_struct *)(bh->b_data);
298
299 if (mmp_check_interval < EXT4_MMP_MIN_CHECK_INTERVAL)
300 mmp_check_interval = EXT4_MMP_MIN_CHECK_INTERVAL;
301
302 /*
303 * If check_interval in MMP block is larger, use that instead of
304 * update_interval from the superblock.
305 */
Santosh Nayak85d21652012-02-27 01:09:03 -0500306 if (le16_to_cpu(mmp->mmp_check_interval) > mmp_check_interval)
307 mmp_check_interval = le16_to_cpu(mmp->mmp_check_interval);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400308
309 seq = le32_to_cpu(mmp->mmp_seq);
310 if (seq == EXT4_MMP_SEQ_CLEAN)
311 goto skip;
312
313 if (seq == EXT4_MMP_SEQ_FSCK) {
314 dump_mmp_msg(sb, mmp, "fsck is running on the filesystem");
315 goto failed;
316 }
317
318 wait_time = min(mmp_check_interval * 2 + 1,
319 mmp_check_interval + 60);
320
321 /* Print MMP interval if more than 20 secs. */
322 if (wait_time > EXT4_MMP_MIN_CHECK_INTERVAL * 4)
323 ext4_warning(sb, "MMP interval %u higher than expected, please"
324 " wait.\n", wait_time * 2);
325
326 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
327 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
328 goto failed;
329 }
330
331 retval = read_mmp_block(sb, &bh, mmp_block);
332 if (retval)
333 goto failed;
334 mmp = (struct mmp_struct *)(bh->b_data);
335 if (seq != le32_to_cpu(mmp->mmp_seq)) {
336 dump_mmp_msg(sb, mmp,
337 "Device is already active on another node.");
338 goto failed;
339 }
340
341skip:
342 /*
343 * write a new random sequence number.
344 */
Darrick J. Wongf6f96fd2011-10-18 10:53:51 -0400345 seq = mmp_new_seq();
346 mmp->mmp_seq = cpu_to_le32(seq);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400347
Darrick J. Wong5c359a42012-04-29 18:47:10 -0400348 retval = write_mmp_block(sb, bh);
Johann Lombardic5e06d12011-05-24 18:31:25 -0400349 if (retval)
350 goto failed;
351
352 /*
353 * wait for MMP interval and check mmp_seq.
354 */
355 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
356 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
357 goto failed;
358 }
359
360 retval = read_mmp_block(sb, &bh, mmp_block);
361 if (retval)
362 goto failed;
363 mmp = (struct mmp_struct *)(bh->b_data);
364 if (seq != le32_to_cpu(mmp->mmp_seq)) {
365 dump_mmp_msg(sb, mmp,
366 "Device is already active on another node.");
367 goto failed;
368 }
369
370 mmpd_data = kmalloc(sizeof(struct mmpd_data), GFP_KERNEL);
371 if (!mmpd_data) {
372 ext4_warning(sb, "not enough memory for mmpd_data");
373 goto failed;
374 }
375 mmpd_data->sb = sb;
376 mmpd_data->bh = bh;
377
378 /*
379 * Start a kernel thread to update the MMP block periodically.
380 */
381 EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, mmpd_data, "kmmpd-%s",
382 bdevname(bh->b_bdev,
383 mmp->mmp_bdevname));
384 if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
385 EXT4_SB(sb)->s_mmp_tsk = NULL;
386 kfree(mmpd_data);
387 ext4_warning(sb, "Unable to create kmmpd thread for %s.",
388 sb->s_id);
389 goto failed;
390 }
391
392 return 0;
393
394failed:
395 brelse(bh);
396 return 1;
397}
398
399