Loading...
1#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
9/* Checksumming functions */
10static __le32 ext4_mmp_csum(struct super_block *sb, struct mmp_struct *mmp)
11{
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
21static int ext4_mmp_csum_verify(struct super_block *sb, struct mmp_struct *mmp)
22{
23 if (!ext4_has_metadata_csum(sb))
24 return 1;
25
26 return mmp->mmp_checksum == ext4_mmp_csum(sb, mmp);
27}
28
29static void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp)
30{
31 if (!ext4_has_metadata_csum(sb))
32 return;
33
34 mmp->mmp_checksum = ext4_mmp_csum(sb, mmp);
35}
36
37/*
38 * Write the MMP block using WRITE_SYNC to try to get the block on-disk
39 * faster.
40 */
41static int write_mmp_block(struct super_block *sb, struct buffer_head *bh)
42{
43 struct mmp_struct *mmp = (struct mmp_struct *)(bh->b_data);
44
45 /*
46 * We protect against freezing so that we don't create dirty buffers
47 * on frozen filesystem.
48 */
49 sb_start_write(sb);
50 ext4_mmp_csum_set(sb, mmp);
51 mark_buffer_dirty(bh);
52 lock_buffer(bh);
53 bh->b_end_io = end_buffer_write_sync;
54 get_bh(bh);
55 submit_bh(WRITE_SYNC | REQ_META | REQ_PRIO, bh);
56 wait_on_buffer(bh);
57 sb_end_write(sb);
58 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;
72 int ret;
73
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. */
80 if (!*bh) {
81 *bh = sb_getblk(sb, mmp_block);
82 if (!*bh) {
83 ret = -ENOMEM;
84 goto warn_exit;
85 }
86 }
87
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 ret = -EIO;
95 goto warn_exit;
96 }
97 mmp = (struct mmp_struct *)((*bh)->b_data);
98 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC) {
99 ret = -EFSCORRUPTED;
100 goto warn_exit;
101 }
102 if (!ext4_mmp_csum_verify(sb, mmp)) {
103 ret = -EFSBADCRC;
104 goto warn_exit;
105 }
106 return 0;
107warn_exit:
108 brelse(*bh);
109 *bh = NULL;
110 ext4_warning(sb, "Error %d while reading MMP block %llu",
111 ret, mmp_block);
112 return ret;
113}
114
115/*
116 * Dump as much information as possible to help the admin.
117 */
118void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
119 const char *function, unsigned int line, const char *msg)
120{
121 __ext4_warning(sb, function, line, "%s", msg);
122 __ext4_warning(sb, function, line,
123 "MMP failure info: last update time: %llu, last update "
124 "node: %s, last update device: %s\n",
125 (long long unsigned int) le64_to_cpu(mmp->mmp_time),
126 mmp->mmp_nodename, mmp->mmp_bdevname);
127}
128
129/*
130 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
131 */
132static int kmmpd(void *data)
133{
134 struct super_block *sb = ((struct mmpd_data *) data)->sb;
135 struct buffer_head *bh = ((struct mmpd_data *) data)->bh;
136 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
137 struct mmp_struct *mmp;
138 ext4_fsblk_t mmp_block;
139 u32 seq = 0;
140 unsigned long failed_writes = 0;
141 int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
142 unsigned mmp_check_interval;
143 unsigned long last_update_time;
144 unsigned long diff;
145 int retval;
146
147 mmp_block = le64_to_cpu(es->s_mmp_block);
148 mmp = (struct mmp_struct *)(bh->b_data);
149 mmp->mmp_time = cpu_to_le64(get_seconds());
150 /*
151 * Start with the higher mmp_check_interval and reduce it if
152 * the MMP block is being updated on time.
153 */
154 mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
155 EXT4_MMP_MIN_CHECK_INTERVAL);
156 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
157 bdevname(bh->b_bdev, mmp->mmp_bdevname);
158
159 memcpy(mmp->mmp_nodename, init_utsname()->nodename,
160 sizeof(mmp->mmp_nodename));
161
162 while (!kthread_should_stop()) {
163 if (++seq > EXT4_MMP_SEQ_MAX)
164 seq = 1;
165
166 mmp->mmp_seq = cpu_to_le32(seq);
167 mmp->mmp_time = cpu_to_le64(get_seconds());
168 last_update_time = jiffies;
169
170 retval = write_mmp_block(sb, bh);
171 /*
172 * Don't spew too many error messages. Print one every
173 * (s_mmp_update_interval * 60) seconds.
174 */
175 if (retval) {
176 if ((failed_writes % 60) == 0)
177 ext4_error(sb, "Error writing to MMP block");
178 failed_writes++;
179 }
180
181 if (!(le32_to_cpu(es->s_feature_incompat) &
182 EXT4_FEATURE_INCOMPAT_MMP)) {
183 ext4_warning(sb, "kmmpd being stopped since MMP feature"
184 " has been disabled.");
185 goto exit_thread;
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 goto exit_thread;
192 }
193
194 diff = jiffies - last_update_time;
195 if (diff < mmp_update_interval * HZ)
196 schedule_timeout_interruptible(mmp_update_interval *
197 HZ - diff);
198
199 /*
200 * We need to make sure that more than mmp_check_interval
201 * seconds have not passed since writing. If that has happened
202 * we need to check if the MMP block is as we left it.
203 */
204 diff = jiffies - last_update_time;
205 if (diff > mmp_check_interval * HZ) {
206 struct buffer_head *bh_check = NULL;
207 struct mmp_struct *mmp_check;
208
209 retval = read_mmp_block(sb, &bh_check, mmp_block);
210 if (retval) {
211 ext4_error(sb, "error reading MMP data: %d",
212 retval);
213 goto exit_thread;
214 }
215
216 mmp_check = (struct mmp_struct *)(bh_check->b_data);
217 if (mmp->mmp_seq != mmp_check->mmp_seq ||
218 memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
219 sizeof(mmp->mmp_nodename))) {
220 dump_mmp_msg(sb, mmp_check,
221 "Error while updating MMP info. "
222 "The filesystem seems to have been"
223 " multiply mounted.");
224 ext4_error(sb, "abort");
225 put_bh(bh_check);
226 retval = -EBUSY;
227 goto exit_thread;
228 }
229 put_bh(bh_check);
230 }
231
232 /*
233 * Adjust the mmp_check_interval depending on how much time
234 * it took for the MMP block to be written.
235 */
236 mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
237 EXT4_MMP_MAX_CHECK_INTERVAL),
238 EXT4_MMP_MIN_CHECK_INTERVAL);
239 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
240 }
241
242 /*
243 * Unmount seems to be clean.
244 */
245 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
246 mmp->mmp_time = cpu_to_le64(get_seconds());
247
248 retval = write_mmp_block(sb, bh);
249
250exit_thread:
251 EXT4_SB(sb)->s_mmp_tsk = NULL;
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 {
266 new_seq = prandom_u32();
267 } 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 */
306 if (le16_to_cpu(mmp->mmp_check_interval) > mmp_check_interval)
307 mmp_check_interval = le16_to_cpu(mmp->mmp_check_interval);
308
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 */
345 seq = mmp_new_seq();
346 mmp->mmp_seq = cpu_to_le32(seq);
347
348 retval = write_mmp_block(sb, bh);
349 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
1#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
9/* Checksumming functions */
10static __le32 ext4_mmp_csum(struct super_block *sb, struct mmp_struct *mmp)
11{
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
21int ext4_mmp_csum_verify(struct super_block *sb, struct mmp_struct *mmp)
22{
23 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
24 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
25 return 1;
26
27 return mmp->mmp_checksum == ext4_mmp_csum(sb, mmp);
28}
29
30void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp)
31{
32 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
33 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
34 return;
35
36 mmp->mmp_checksum = ext4_mmp_csum(sb, mmp);
37}
38
39/*
40 * Write the MMP block using WRITE_SYNC to try to get the block on-disk
41 * faster.
42 */
43static int write_mmp_block(struct super_block *sb, struct buffer_head *bh)
44{
45 struct mmp_struct *mmp = (struct mmp_struct *)(bh->b_data);
46
47 /*
48 * We protect against freezing so that we don't create dirty buffers
49 * on frozen filesystem.
50 */
51 sb_start_write(sb);
52 ext4_mmp_csum_set(sb, mmp);
53 mark_buffer_dirty(bh);
54 lock_buffer(bh);
55 bh->b_end_io = end_buffer_write_sync;
56 get_bh(bh);
57 submit_bh(WRITE_SYNC | REQ_META | REQ_PRIO, bh);
58 wait_on_buffer(bh);
59 sb_end_write(sb);
60 if (unlikely(!buffer_uptodate(bh)))
61 return 1;
62
63 return 0;
64}
65
66/*
67 * Read the MMP block. It _must_ be read from disk and hence we clear the
68 * uptodate flag on the buffer.
69 */
70static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
71 ext4_fsblk_t mmp_block)
72{
73 struct mmp_struct *mmp;
74
75 if (*bh)
76 clear_buffer_uptodate(*bh);
77
78 /* This would be sb_bread(sb, mmp_block), except we need to be sure
79 * that the MD RAID device cache has been bypassed, and that the read
80 * is not blocked in the elevator. */
81 if (!*bh)
82 *bh = sb_getblk(sb, mmp_block);
83 if (!*bh)
84 return -ENOMEM;
85 if (*bh) {
86 get_bh(*bh);
87 lock_buffer(*bh);
88 (*bh)->b_end_io = end_buffer_read_sync;
89 submit_bh(READ_SYNC | REQ_META | REQ_PRIO, *bh);
90 wait_on_buffer(*bh);
91 if (!buffer_uptodate(*bh)) {
92 brelse(*bh);
93 *bh = NULL;
94 }
95 }
96 if (unlikely(!*bh)) {
97 ext4_warning(sb, "Error while reading MMP block %llu",
98 mmp_block);
99 return -EIO;
100 }
101
102 mmp = (struct mmp_struct *)((*bh)->b_data);
103 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC ||
104 !ext4_mmp_csum_verify(sb, mmp))
105 return -EINVAL;
106
107 return 0;
108}
109
110/*
111 * Dump as much information as possible to help the admin.
112 */
113void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
114 const char *function, unsigned int line, const char *msg)
115{
116 __ext4_warning(sb, function, line, msg);
117 __ext4_warning(sb, function, line,
118 "MMP failure info: last update time: %llu, last update "
119 "node: %s, last update device: %s\n",
120 (long long unsigned int) le64_to_cpu(mmp->mmp_time),
121 mmp->mmp_nodename, mmp->mmp_bdevname);
122}
123
124/*
125 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
126 */
127static int kmmpd(void *data)
128{
129 struct super_block *sb = ((struct mmpd_data *) data)->sb;
130 struct buffer_head *bh = ((struct mmpd_data *) data)->bh;
131 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
132 struct mmp_struct *mmp;
133 ext4_fsblk_t mmp_block;
134 u32 seq = 0;
135 unsigned long failed_writes = 0;
136 int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
137 unsigned mmp_check_interval;
138 unsigned long last_update_time;
139 unsigned long diff;
140 int retval;
141
142 mmp_block = le64_to_cpu(es->s_mmp_block);
143 mmp = (struct mmp_struct *)(bh->b_data);
144 mmp->mmp_time = cpu_to_le64(get_seconds());
145 /*
146 * Start with the higher mmp_check_interval and reduce it if
147 * the MMP block is being updated on time.
148 */
149 mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
150 EXT4_MMP_MIN_CHECK_INTERVAL);
151 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
152 bdevname(bh->b_bdev, mmp->mmp_bdevname);
153
154 memcpy(mmp->mmp_nodename, init_utsname()->nodename,
155 sizeof(mmp->mmp_nodename));
156
157 while (!kthread_should_stop()) {
158 if (++seq > EXT4_MMP_SEQ_MAX)
159 seq = 1;
160
161 mmp->mmp_seq = cpu_to_le32(seq);
162 mmp->mmp_time = cpu_to_le64(get_seconds());
163 last_update_time = jiffies;
164
165 retval = write_mmp_block(sb, bh);
166 /*
167 * Don't spew too many error messages. Print one every
168 * (s_mmp_update_interval * 60) seconds.
169 */
170 if (retval) {
171 if ((failed_writes % 60) == 0)
172 ext4_error(sb, "Error writing to MMP block");
173 failed_writes++;
174 }
175
176 if (!(le32_to_cpu(es->s_feature_incompat) &
177 EXT4_FEATURE_INCOMPAT_MMP)) {
178 ext4_warning(sb, "kmmpd being stopped since MMP feature"
179 " has been disabled.");
180 EXT4_SB(sb)->s_mmp_tsk = NULL;
181 goto failed;
182 }
183
184 if (sb->s_flags & MS_RDONLY) {
185 ext4_warning(sb, "kmmpd being stopped since filesystem "
186 "has been remounted as readonly.");
187 EXT4_SB(sb)->s_mmp_tsk = NULL;
188 goto failed;
189 }
190
191 diff = jiffies - last_update_time;
192 if (diff < mmp_update_interval * HZ)
193 schedule_timeout_interruptible(mmp_update_interval *
194 HZ - diff);
195
196 /*
197 * We need to make sure that more than mmp_check_interval
198 * seconds have not passed since writing. If that has happened
199 * we need to check if the MMP block is as we left it.
200 */
201 diff = jiffies - last_update_time;
202 if (diff > mmp_check_interval * HZ) {
203 struct buffer_head *bh_check = NULL;
204 struct mmp_struct *mmp_check;
205
206 retval = read_mmp_block(sb, &bh_check, mmp_block);
207 if (retval) {
208 ext4_error(sb, "error reading MMP data: %d",
209 retval);
210
211 EXT4_SB(sb)->s_mmp_tsk = NULL;
212 goto failed;
213 }
214
215 mmp_check = (struct mmp_struct *)(bh_check->b_data);
216 if (mmp->mmp_seq != mmp_check->mmp_seq ||
217 memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
218 sizeof(mmp->mmp_nodename))) {
219 dump_mmp_msg(sb, mmp_check,
220 "Error while updating MMP info. "
221 "The filesystem seems to have been"
222 " multiply mounted.");
223 ext4_error(sb, "abort");
224 goto failed;
225 }
226 put_bh(bh_check);
227 }
228
229 /*
230 * Adjust the mmp_check_interval depending on how much time
231 * it took for the MMP block to be written.
232 */
233 mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
234 EXT4_MMP_MAX_CHECK_INTERVAL),
235 EXT4_MMP_MIN_CHECK_INTERVAL);
236 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
237 }
238
239 /*
240 * Unmount seems to be clean.
241 */
242 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
243 mmp->mmp_time = cpu_to_le64(get_seconds());
244
245 retval = write_mmp_block(sb, bh);
246
247failed:
248 kfree(data);
249 brelse(bh);
250 return retval;
251}
252
253/*
254 * Get a random new sequence number but make sure it is not greater than
255 * EXT4_MMP_SEQ_MAX.
256 */
257static unsigned int mmp_new_seq(void)
258{
259 u32 new_seq;
260
261 do {
262 new_seq = prandom_u32();
263 } while (new_seq > EXT4_MMP_SEQ_MAX);
264
265 return new_seq;
266}
267
268/*
269 * Protect the filesystem from being mounted more than once.
270 */
271int ext4_multi_mount_protect(struct super_block *sb,
272 ext4_fsblk_t mmp_block)
273{
274 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
275 struct buffer_head *bh = NULL;
276 struct mmp_struct *mmp = NULL;
277 struct mmpd_data *mmpd_data;
278 u32 seq;
279 unsigned int mmp_check_interval = le16_to_cpu(es->s_mmp_update_interval);
280 unsigned int wait_time = 0;
281 int retval;
282
283 if (mmp_block < le32_to_cpu(es->s_first_data_block) ||
284 mmp_block >= ext4_blocks_count(es)) {
285 ext4_warning(sb, "Invalid MMP block in superblock");
286 goto failed;
287 }
288
289 retval = read_mmp_block(sb, &bh, mmp_block);
290 if (retval)
291 goto failed;
292
293 mmp = (struct mmp_struct *)(bh->b_data);
294
295 if (mmp_check_interval < EXT4_MMP_MIN_CHECK_INTERVAL)
296 mmp_check_interval = EXT4_MMP_MIN_CHECK_INTERVAL;
297
298 /*
299 * If check_interval in MMP block is larger, use that instead of
300 * update_interval from the superblock.
301 */
302 if (le16_to_cpu(mmp->mmp_check_interval) > mmp_check_interval)
303 mmp_check_interval = le16_to_cpu(mmp->mmp_check_interval);
304
305 seq = le32_to_cpu(mmp->mmp_seq);
306 if (seq == EXT4_MMP_SEQ_CLEAN)
307 goto skip;
308
309 if (seq == EXT4_MMP_SEQ_FSCK) {
310 dump_mmp_msg(sb, mmp, "fsck is running on the filesystem");
311 goto failed;
312 }
313
314 wait_time = min(mmp_check_interval * 2 + 1,
315 mmp_check_interval + 60);
316
317 /* Print MMP interval if more than 20 secs. */
318 if (wait_time > EXT4_MMP_MIN_CHECK_INTERVAL * 4)
319 ext4_warning(sb, "MMP interval %u higher than expected, please"
320 " wait.\n", wait_time * 2);
321
322 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
323 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
324 goto failed;
325 }
326
327 retval = read_mmp_block(sb, &bh, mmp_block);
328 if (retval)
329 goto failed;
330 mmp = (struct mmp_struct *)(bh->b_data);
331 if (seq != le32_to_cpu(mmp->mmp_seq)) {
332 dump_mmp_msg(sb, mmp,
333 "Device is already active on another node.");
334 goto failed;
335 }
336
337skip:
338 /*
339 * write a new random sequence number.
340 */
341 seq = mmp_new_seq();
342 mmp->mmp_seq = cpu_to_le32(seq);
343
344 retval = write_mmp_block(sb, bh);
345 if (retval)
346 goto failed;
347
348 /*
349 * wait for MMP interval and check mmp_seq.
350 */
351 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
352 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
353 goto failed;
354 }
355
356 retval = read_mmp_block(sb, &bh, mmp_block);
357 if (retval)
358 goto failed;
359 mmp = (struct mmp_struct *)(bh->b_data);
360 if (seq != le32_to_cpu(mmp->mmp_seq)) {
361 dump_mmp_msg(sb, mmp,
362 "Device is already active on another node.");
363 goto failed;
364 }
365
366 mmpd_data = kmalloc(sizeof(struct mmpd_data), GFP_KERNEL);
367 if (!mmpd_data) {
368 ext4_warning(sb, "not enough memory for mmpd_data");
369 goto failed;
370 }
371 mmpd_data->sb = sb;
372 mmpd_data->bh = bh;
373
374 /*
375 * Start a kernel thread to update the MMP block periodically.
376 */
377 EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, mmpd_data, "kmmpd-%s",
378 bdevname(bh->b_bdev,
379 mmp->mmp_bdevname));
380 if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
381 EXT4_SB(sb)->s_mmp_tsk = NULL;
382 kfree(mmpd_data);
383 ext4_warning(sb, "Unable to create kmmpd thread for %s.",
384 sb->s_id);
385 goto failed;
386 }
387
388 return 0;
389
390failed:
391 brelse(bh);
392 return 1;
393}
394
395