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);
52 mark_buffer_dirty(bh);
54 bh->b_end_io = end_buffer_write_sync;
56 submit_bh(REQ_OP_WRITE, REQ_SYNC | REQ_META | REQ_PRIO, bh);
59 if (unlikely(!buffer_uptodate(bh)))
66 * Read the MMP block. It _must_ be read from disk and hence we clear the
67 * uptodate flag on the buffer.
69 static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
70 ext4_fsblk_t mmp_block)
72 struct mmp_struct *mmp;
76 clear_buffer_uptodate(*bh);
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. */
82 *bh = sb_getblk(sb, mmp_block);
91 (*bh)->b_end_io = end_buffer_read_sync;
92 submit_bh(REQ_OP_READ, REQ_META | REQ_PRIO, *bh);
94 if (!buffer_uptodate(*bh)) {
98 mmp = (struct mmp_struct *)((*bh)->b_data);
99 if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC) {
103 if (!ext4_mmp_csum_verify(sb, mmp)) {
111 ext4_warning(sb, "Error %d while reading MMP block %llu",
117 * Dump as much information as possible to help the admin.
119 void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
120 const char *function, unsigned int line, const char *msg)
122 __ext4_warning(sb, function, line, "%s", msg);
123 __ext4_warning(sb, function, line,
124 "MMP failure info: last update time: %llu, last update "
125 "node: %s, last update device: %s",
126 (long long unsigned int) le64_to_cpu(mmp->mmp_time),
127 mmp->mmp_nodename, mmp->mmp_bdevname);
131 * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
133 static int kmmpd(void *data)
135 struct super_block *sb = ((struct mmpd_data *) data)->sb;
136 struct buffer_head *bh = ((struct mmpd_data *) data)->bh;
137 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
138 struct mmp_struct *mmp;
139 ext4_fsblk_t mmp_block;
141 unsigned long failed_writes = 0;
142 int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
143 unsigned mmp_check_interval;
144 unsigned long last_update_time;
148 mmp_block = le64_to_cpu(es->s_mmp_block);
149 mmp = (struct mmp_struct *)(bh->b_data);
150 mmp->mmp_time = cpu_to_le64(get_seconds());
152 * Start with the higher mmp_check_interval and reduce it if
153 * the MMP block is being updated on time.
155 mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
156 EXT4_MMP_MIN_CHECK_INTERVAL);
157 mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
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(get_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_error(sb, "Error writing to MMP block");
182 if (!(le32_to_cpu(es->s_feature_incompat) &
183 EXT4_FEATURE_INCOMPAT_MMP)) {
184 ext4_warning(sb, "kmmpd being stopped since MMP feature"
185 " has been disabled.");
190 ext4_warning(sb, "kmmpd being stopped since filesystem "
191 "has been remounted as readonly.");
195 diff = jiffies - last_update_time;
196 if (diff < mmp_update_interval * HZ)
197 schedule_timeout_interruptible(mmp_update_interval *
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.
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;
210 retval = read_mmp_block(sb, &bh_check, mmp_block);
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_error(sb, "abort");
234 * Adjust the mmp_check_interval depending on how much time
235 * it took for the MMP block to be written.
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);
244 * Unmount seems to be clean.
246 mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
247 mmp->mmp_time = cpu_to_le64(get_seconds());
249 retval = write_mmp_block(sb, bh);
252 EXT4_SB(sb)->s_mmp_tsk = NULL;
259 * Get a random new sequence number but make sure it is not greater than
262 static unsigned int mmp_new_seq(void)
267 new_seq = prandom_u32();
268 } while (new_seq > EXT4_MMP_SEQ_MAX);
274 * Protect the filesystem from being mounted more than once.
276 int ext4_multi_mount_protect(struct super_block *sb,
277 ext4_fsblk_t mmp_block)
279 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
280 struct buffer_head *bh = NULL;
281 struct mmp_struct *mmp = NULL;
282 struct mmpd_data *mmpd_data;
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 mmpd_data = kmalloc(sizeof(*mmpd_data), GFP_KERNEL);
373 ext4_warning(sb, "not enough memory for mmpd_data");
380 * Start a kernel thread to update the MMP block periodically.
382 EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, mmpd_data, "kmmpd-%s",
385 if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
386 EXT4_SB(sb)->s_mmp_tsk = NULL;
388 ext4_warning(sb, "Unable to create kmmpd thread for %s.",