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);
90 (*bh)->b_end_io = end_buffer_read_sync;
91 submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, *bh);
93 if (!buffer_uptodate(*bh)) {
97 mmp = (struct mmp_struct *)((*bh)->b_data);
98 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC) {
102 if (!ext4_mmp_csum_verify(sb, mmp)) {
110 ext4_warning(sb, "Error %d while reading MMP block %llu",
116 * Dump as much information as possible to help the admin.
118 void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
119 const char *function, unsigned int line, const char *msg)
121 __ext4_warning(sb, function, line, "%s", msg);
122 __ext4_warning(sb, function, line,
123 "MMP failure info: last update time: %llu, last update node: %.*s, last update device: %.*s",
124 (unsigned long long)le64_to_cpu(mmp->mmp_time),
125 (int)sizeof(mmp->mmp_nodename), mmp->mmp_nodename,
126 (int)sizeof(mmp->mmp_bdevname), mmp->mmp_bdevname);
130 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
132 static int kmmpd(void *data)
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;
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;
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(ktime_get_real_seconds());
151 * Start with the higher mmp_check_interval and reduce it if
152 * the MMP block is being updated on time.
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 BUILD_BUG_ON(sizeof(mmp->mmp_bdevname) < BDEVNAME_SIZE);
158 bdevname(bh->b_bdev, mmp->mmp_bdevname);
160 memcpy(mmp->mmp_nodename, init_utsname()->nodename,
161 sizeof(mmp->mmp_nodename));
163 while (!kthread_should_stop()) {
164 if (++seq > EXT4_MMP_SEQ_MAX)
167 mmp->mmp_seq = cpu_to_le32(seq);
168 mmp->mmp_time = cpu_to_le64(ktime_get_real_seconds());
169 last_update_time = jiffies;
171 retval = write_mmp_block(sb, bh);
173 * Don't spew too many error messages. Print one every
174 * (s_mmp_update_interval * 60) seconds.
177 if ((failed_writes % 60) == 0) {
178 ext4_set_errno(sb, -retval);
179 ext4_error(sb, "Error writing to MMP block");
184 if (!(le32_to_cpu(es->s_feature_incompat) &
185 EXT4_FEATURE_INCOMPAT_MMP)) {
186 ext4_warning(sb, "kmmpd being stopped since MMP feature"
187 " has been disabled.");
194 diff = jiffies - last_update_time;
195 if (diff < mmp_update_interval * HZ)
196 schedule_timeout_interruptible(mmp_update_interval *
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.
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;
209 retval = read_mmp_block(sb, &bh_check, mmp_block);
211 ext4_set_errno(sb, -retval);
212 ext4_error(sb, "error reading MMP data: %d",
217 mmp_check = (struct mmp_struct *)(bh_check->b_data);
218 if (mmp->mmp_seq != mmp_check->mmp_seq ||
219 memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
220 sizeof(mmp->mmp_nodename))) {
221 dump_mmp_msg(sb, mmp_check,
222 "Error while updating MMP info. "
223 "The filesystem seems to have been"
224 " multiply mounted.");
225 ext4_set_errno(sb, EBUSY);
226 ext4_error(sb, "abort");
235 * Adjust the mmp_check_interval depending on how much time
236 * it took for the MMP block to be written.
238 mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
239 EXT4_MMP_MAX_CHECK_INTERVAL),
240 EXT4_MMP_MIN_CHECK_INTERVAL);
241 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
245 * Unmount seems to be clean.
247 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
248 mmp->mmp_time = cpu_to_le64(ktime_get_real_seconds());
250 retval = write_mmp_block(sb, bh);
253 EXT4_SB(sb)->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;
283 struct mmpd_data *mmpd_data;
285 unsigned int mmp_check_interval = le16_to_cpu(es->s_mmp_update_interval);
286 unsigned int wait_time = 0;
289 if (mmp_block < le32_to_cpu(es->s_first_data_block) ||
290 mmp_block >= ext4_blocks_count(es)) {
291 ext4_warning(sb, "Invalid MMP block in superblock");
295 retval = read_mmp_block(sb, &bh, mmp_block);
299 mmp = (struct mmp_struct *)(bh->b_data);
301 if (mmp_check_interval < EXT4_MMP_MIN_CHECK_INTERVAL)
302 mmp_check_interval = EXT4_MMP_MIN_CHECK_INTERVAL;
305 * If check_interval in MMP block is larger, use that instead of
306 * update_interval from the superblock.
308 if (le16_to_cpu(mmp->mmp_check_interval) > mmp_check_interval)
309 mmp_check_interval = le16_to_cpu(mmp->mmp_check_interval);
311 seq = le32_to_cpu(mmp->mmp_seq);
312 if (seq == EXT4_MMP_SEQ_CLEAN)
315 if (seq == EXT4_MMP_SEQ_FSCK) {
316 dump_mmp_msg(sb, mmp, "fsck is running on the filesystem");
320 wait_time = min(mmp_check_interval * 2 + 1,
321 mmp_check_interval + 60);
323 /* Print MMP interval if more than 20 secs. */
324 if (wait_time > EXT4_MMP_MIN_CHECK_INTERVAL * 4)
325 ext4_warning(sb, "MMP interval %u higher than expected, please"
326 " wait.\n", wait_time * 2);
328 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
329 ext4_warning(sb, "MMP startup interrupted, failing mount\n");
333 retval = read_mmp_block(sb, &bh, mmp_block);
336 mmp = (struct mmp_struct *)(bh->b_data);
337 if (seq != le32_to_cpu(mmp->mmp_seq)) {
338 dump_mmp_msg(sb, mmp,
339 "Device is already active on another node.");
345 * write a new random sequence number.
348 mmp->mmp_seq = cpu_to_le32(seq);
350 retval = write_mmp_block(sb, bh);
355 * wait for MMP interval and check mmp_seq.
357 if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
358 ext4_warning(sb, "MMP startup interrupted, failing mount");
362 retval = read_mmp_block(sb, &bh, mmp_block);
365 mmp = (struct mmp_struct *)(bh->b_data);
366 if (seq != le32_to_cpu(mmp->mmp_seq)) {
367 dump_mmp_msg(sb, mmp,
368 "Device is already active on another node.");
372 mmpd_data = kmalloc(sizeof(*mmpd_data), GFP_KERNEL);
374 ext4_warning(sb, "not enough memory for mmpd_data");
381 * Start a kernel thread to update the MMP block periodically.
383 EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, mmpd_data, "kmmpd-%.*s",
384 (int)sizeof(mmp->mmp_bdevname),
387 if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
388 EXT4_SB(sb)->s_mmp_tsk = NULL;
390 ext4_warning(sb, "Unable to create kmmpd thread for %s.",