1 // SPDX-License-Identifier: GPL-2.0
3 #include <linux/random.h>
4 #include <linux/buffer_head.h>
5 #include <linux/utsname.h>
6 #include <linux/kthread.h>
10 /* Checksumming functions */
11 static __le32 ext4_mmp_csum(struct super_block *sb, struct mmp_struct *mmp)
13 struct ext4_sb_info *sbi = EXT4_SB(sb);
14 int offset = offsetof(struct mmp_struct, mmp_checksum);
17 csum = ext4_chksum(sbi, sbi->s_csum_seed, (char *)mmp, offset);
19 return cpu_to_le32(csum);
22 static int ext4_mmp_csum_verify(struct super_block *sb, struct mmp_struct *mmp)
24 if (!ext4_has_metadata_csum(sb))
27 return mmp->mmp_checksum == ext4_mmp_csum(sb, mmp);
30 static void ext4_mmp_csum_set(struct super_block *sb, struct mmp_struct *mmp)
32 if (!ext4_has_metadata_csum(sb))
35 mmp->mmp_checksum = ext4_mmp_csum(sb, mmp);
39 * Write the MMP block using REQ_SYNC to try to get the block on-disk
42 static int write_mmp_block(struct super_block *sb, struct buffer_head *bh)
44 struct mmp_struct *mmp = (struct mmp_struct *)(bh->b_data);
47 * We protect against freezing so that we don't create dirty buffers
48 * on frozen filesystem.
51 ext4_mmp_csum_set(sb, mmp);
53 bh->b_end_io = end_buffer_write_sync;
55 submit_bh(REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, bh);
58 if (unlikely(!buffer_uptodate(bh)))
65 * Read the MMP block. It _must_ be read from disk and hence we clear the
66 * uptodate flag on the buffer.
68 static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
69 ext4_fsblk_t mmp_block)
71 struct mmp_struct *mmp;
75 clear_buffer_uptodate(*bh);
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. */
81 *bh = sb_getblk(sb, mmp_block);
89 ret = ext4_read_bh(*bh, REQ_META | REQ_PRIO, NULL);
93 mmp = (struct mmp_struct *)((*bh)->b_data);
94 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC) {
98 if (!ext4_mmp_csum_verify(sb, mmp)) {
106 ext4_warning(sb, "Error %d while reading MMP block %llu",
112 * Dump as much information as possible to help the admin.
114 void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
115 const char *function, unsigned int line, const char *msg)
117 __ext4_warning(sb, function, line, "%s", msg);
118 __ext4_warning(sb, function, line,
119 "MMP failure info: last update time: %llu, last update node: %.*s, last update device: %.*s",
120 (unsigned long long)le64_to_cpu(mmp->mmp_time),
121 (int)sizeof(mmp->mmp_nodename), mmp->mmp_nodename,
122 (int)sizeof(mmp->mmp_bdevname), mmp->mmp_bdevname);
126 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
128 static int kmmpd(void *data)
130 struct super_block *sb = data;
131 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
132 struct buffer_head *bh = EXT4_SB(sb)->s_mmp_bh;
133 struct mmp_struct *mmp;
134 ext4_fsblk_t mmp_block;
136 unsigned long failed_writes = 0;
137 int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
138 unsigned mmp_check_interval;
139 unsigned long last_update_time;
143 mmp_block = le64_to_cpu(es->s_mmp_block);
144 mmp = (struct mmp_struct *)(bh->b_data);
145 mmp->mmp_time = cpu_to_le64(ktime_get_real_seconds());
147 * Start with the higher mmp_check_interval and reduce it if
148 * the MMP block is being updated on time.
150 mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
151 EXT4_MMP_MIN_CHECK_INTERVAL);
152 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
154 memcpy(mmp->mmp_nodename, init_utsname()->nodename,
155 sizeof(mmp->mmp_nodename));
157 while (!kthread_should_stop() && !sb_rdonly(sb)) {
158 if (!ext4_has_feature_mmp(sb)) {
159 ext4_warning(sb, "kmmpd being stopped since MMP feature"
160 " has been disabled.");
163 if (++seq > EXT4_MMP_SEQ_MAX)
166 mmp->mmp_seq = cpu_to_le32(seq);
167 mmp->mmp_time = cpu_to_le64(ktime_get_real_seconds());
168 last_update_time = jiffies;
170 retval = write_mmp_block(sb, bh);
172 * Don't spew too many error messages. Print one every
173 * (s_mmp_update_interval * 60) seconds.
176 if ((failed_writes % 60) == 0) {
177 ext4_error_err(sb, -retval,
178 "Error writing to MMP block");
183 diff = jiffies - last_update_time;
184 if (diff < mmp_update_interval * HZ)
185 schedule_timeout_interruptible(mmp_update_interval *
189 * We need to make sure that more than mmp_check_interval
190 * seconds have not passed since writing. If that has happened
191 * we need to check if the MMP block is as we left it.
193 diff = jiffies - last_update_time;
194 if (diff > mmp_check_interval * HZ) {
195 struct buffer_head *bh_check = NULL;
196 struct mmp_struct *mmp_check;
198 retval = read_mmp_block(sb, &bh_check, mmp_block);
200 ext4_error_err(sb, -retval,
201 "error reading MMP data: %d",
206 mmp_check = (struct mmp_struct *)(bh_check->b_data);
207 if (mmp->mmp_seq != mmp_check->mmp_seq ||
208 memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
209 sizeof(mmp->mmp_nodename))) {
210 dump_mmp_msg(sb, mmp_check,
211 "Error while updating MMP info. "
212 "The filesystem seems to have been"
213 " multiply mounted.");
214 ext4_error_err(sb, EBUSY, "abort");
223 * Adjust the mmp_check_interval depending on how much time
224 * it took for the MMP block to be written.
226 mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
227 EXT4_MMP_MAX_CHECK_INTERVAL),
228 EXT4_MMP_MIN_CHECK_INTERVAL);
229 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
233 * Unmount seems to be clean.
235 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
236 mmp->mmp_time = cpu_to_le64(ktime_get_real_seconds());
238 retval = write_mmp_block(sb, bh);
241 while (!kthread_should_stop()) {
242 set_current_state(TASK_INTERRUPTIBLE);
243 if (!kthread_should_stop())
246 set_current_state(TASK_RUNNING);
250 void ext4_stop_mmpd(struct ext4_sb_info *sbi)
252 if (sbi->s_mmp_tsk) {
253 kthread_stop(sbi->s_mmp_tsk);
254 brelse(sbi->s_mmp_bh);
255 sbi->s_mmp_tsk = NULL;
260 * Get a random new sequence number but make sure it is not greater than
263 static unsigned int mmp_new_seq(void)
268 new_seq = prandom_u32();
269 } while (new_seq > EXT4_MMP_SEQ_MAX);
275 * Protect the filesystem from being mounted more than once.
277 int ext4_multi_mount_protect(struct super_block *sb,
278 ext4_fsblk_t mmp_block)
280 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
281 struct buffer_head *bh = NULL;
282 struct mmp_struct *mmp = NULL;
284 unsigned int mmp_check_interval = le16_to_cpu(es->s_mmp_update_interval);
285 unsigned int wait_time = 0;
288 if (mmp_block < le32_to_cpu(es->s_first_data_block) ||
289 mmp_block >= ext4_blocks_count(es)) {
290 ext4_warning(sb, "Invalid MMP block in superblock");
294 retval = read_mmp_block(sb, &bh, mmp_block);
298 mmp = (struct mmp_struct *)(bh->b_data);
300 if (mmp_check_interval < EXT4_MMP_MIN_CHECK_INTERVAL)
301 mmp_check_interval = EXT4_MMP_MIN_CHECK_INTERVAL;
304 * If check_interval in MMP block is larger, use that instead of
305 * update_interval from the superblock.
307 if (le16_to_cpu(mmp->mmp_check_interval) > mmp_check_interval)
308 mmp_check_interval = le16_to_cpu(mmp->mmp_check_interval);
310 seq = le32_to_cpu(mmp->mmp_seq);
311 if (seq == EXT4_MMP_SEQ_CLEAN)
314 if (seq == EXT4_MMP_SEQ_FSCK) {
315 dump_mmp_msg(sb, mmp, "fsck is running on the filesystem");
319 wait_time = min(mmp_check_interval * 2 + 1,
320 mmp_check_interval + 60);
322 /* Print MMP interval if more than 20 secs. */
323 if (wait_time > EXT4_MMP_MIN_CHECK_INTERVAL * 4)
324 ext4_warning(sb, "MMP interval %u higher than expected, please"
325 " wait.\n", wait_time * 2);
327 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
328 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
332 retval = read_mmp_block(sb, &bh, mmp_block);
335 mmp = (struct mmp_struct *)(bh->b_data);
336 if (seq != le32_to_cpu(mmp->mmp_seq)) {
337 dump_mmp_msg(sb, mmp,
338 "Device is already active on another node.");
344 * write a new random sequence number.
347 mmp->mmp_seq = cpu_to_le32(seq);
349 retval = write_mmp_block(sb, bh);
354 * wait for MMP interval and check mmp_seq.
356 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
357 ext4_warning(sb, "MMP startup interrupted, failing mount");
361 retval = read_mmp_block(sb, &bh, mmp_block);
364 mmp = (struct mmp_struct *)(bh->b_data);
365 if (seq != le32_to_cpu(mmp->mmp_seq)) {
366 dump_mmp_msg(sb, mmp,
367 "Device is already active on another node.");
371 EXT4_SB(sb)->s_mmp_bh = bh;
373 BUILD_BUG_ON(sizeof(mmp->mmp_bdevname) < BDEVNAME_SIZE);
374 snprintf(mmp->mmp_bdevname, sizeof(mmp->mmp_bdevname),
378 * Start a kernel thread to update the MMP block periodically.
380 EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, sb, "kmmpd-%.*s",
381 (int)sizeof(mmp->mmp_bdevname),
383 if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
384 EXT4_SB(sb)->s_mmp_tsk = NULL;
385 ext4_warning(sb, "Unable to create kmmpd thread for %s.",