/*
* calc_chksums calculates the checksums for the blocks described in the
* descriptor block.
*/
static int calc_chksums(journal_t *journal, struct buffer_head *bh,
unsigned long long *next_log_block, __u32 *crc32_sum)
{
int i, num_blks, err;
unsigned long long io_block;
struct buffer_head *obh;
num_blks = count_tags(journal, bh);
/* Calculate checksum of the descriptor block. */
*crc32_sum = crc32_be(*crc32_sum, (void *)bh->b_data, bh->b_size);
for (i = 0; i < num_blks; i++) {
io_block = (*next_log_block)++;
wrap(journal, *next_log_block);
err = jread(&obh, journal, io_block);
if (err) {
printk(KERN_ERR "JBD: IO error %d recovering block "
"%llu in log\n", err, io_block);
return 1;
} else {
*crc32_sum = crc32_be(*crc32_sum, (void *)obh->b_data,
obh->b_size);
}
brelse(obh);
}
return 0;
}
test_exe_head(int file, Exe_head *exe)
{
if (jread(file, exe, sizeof(*exe)) != sizeof(*exe) )
return(0);
if (exe->M != 'M' || exe->Z != 'Z' || exe->one != 1)
return(0);
return(1);
}
/* fread() wrapper */
size_t jfread(void *ptr, size_t size, size_t nmemb, struct jfs *stream)
{
int rv;
rv = jread(stream, ptr, size * nmemb);
if (rv <= 0)
return 0;
return rv / size;
}
static int do_one_pass(journal_t *journal,
struct recovery_info *info, enum passtype pass)
{
unsigned int first_commit_ID, next_commit_ID;
unsigned long next_log_block;
int err, success = 0;
journal_superblock_t * sb;
journal_header_t * tmp;
struct buffer_head * bh;
unsigned int sequence;
int blocktype;
/* Precompute the maximum metadata descriptors in a descriptor block */
int MAX_BLOCKS_PER_DESC;
MAX_BLOCKS_PER_DESC = ((journal->j_blocksize-sizeof(journal_header_t))
/ sizeof(journal_block_tag_t));
/*
* First thing is to establish what we expect to find in the log
* (in terms of transaction IDs), and where (in terms of log
* block offsets): query the superblock.
*/
sb = journal->j_superblock;
next_commit_ID = be32_to_cpu(sb->s_sequence);
next_log_block = be32_to_cpu(sb->s_start);
first_commit_ID = next_commit_ID;
if (pass == PASS_SCAN)
info->start_transaction = first_commit_ID;
jbd_debug(1, "Starting recovery pass %d\n", pass);
/*
* Now we walk through the log, transaction by transaction,
* making sure that each transaction has a commit block in the
* expected place. Each complete transaction gets replayed back
* into the main filesystem.
*/
while (1) {
int flags;
char * tagp;
journal_block_tag_t * tag;
struct buffer_head * obh;
struct buffer_head * nbh;
cond_resched(); /* We're under lock_kernel() */
/* If we already know where to stop the log traversal,
* check right now that we haven't gone past the end of
* the log. */
if (pass != PASS_SCAN)
if (tid_geq(next_commit_ID, info->end_transaction))
break;
jbd_debug(2, "Scanning for sequence ID %u at %lu/%lu\n",
next_commit_ID, next_log_block, journal->j_last);
/* Skip over each chunk of the transaction looking
* either the next descriptor block or the final commit
* record. */
jbd_debug(3, "JBD: checking block %ld\n", next_log_block);
err = jread(&bh, journal, next_log_block);
if (err)
goto failed;
next_log_block++;
wrap(journal, next_log_block);
/* What kind of buffer is it?
*
* If it is a descriptor block, check that it has the
* expected sequence number. Otherwise, we're all done
* here. */
tmp = (journal_header_t *)bh->b_data;
if (tmp->h_magic != cpu_to_be32(JFS_MAGIC_NUMBER)) {
brelse(bh);
break;
}
blocktype = be32_to_cpu(tmp->h_blocktype);
sequence = be32_to_cpu(tmp->h_sequence);
jbd_debug(3, "Found magic %d, sequence %d\n",
blocktype, sequence);
if (sequence != next_commit_ID) {
brelse(bh);
break;
}
/* OK, we have a valid descriptor block which matches
* all of the sequence number checks. What are we going
* to do with it? That depends on the pass... */
switch(blocktype) {
case JFS_DESCRIPTOR_BLOCK:
/* If it is a valid descriptor block, replay it
* in pass REPLAY; otherwise, just skip over the
* blocks it describes. */
if (pass != PASS_REPLAY) {
next_log_block +=
count_tags(bh, journal->j_blocksize);
wrap(journal, next_log_block);
brelse(bh);
continue;
}
/* A descriptor block: we can now write all of
* the data blocks. Yay, useful work is finally
* getting done here! */
tagp = &bh->b_data[sizeof(journal_header_t)];
while ((tagp - bh->b_data +sizeof(journal_block_tag_t))
<= journal->j_blocksize) {
unsigned long io_block;
tag = (journal_block_tag_t *) tagp;
flags = be32_to_cpu(tag->t_flags);
io_block = next_log_block++;
wrap(journal, next_log_block);
err = jread(&obh, journal, io_block);
if (err) {
/* Recover what we can, but
* report failure at the end. */
success = err;
printk (KERN_ERR
"JBD: IO error %d recovering "
"block %ld in log\n",
err, io_block);
} else {
unsigned long blocknr;
J_ASSERT(obh != NULL);
blocknr = be32_to_cpu(tag->t_blocknr);
/* If the block has been
* revoked, then we're all done
* here. */
if (journal_test_revoke
(journal, blocknr,
next_commit_ID)) {
brelse(obh);
++info->nr_revoke_hits;
goto skip_write;
}
/* Find a buffer for the new
* data being restored */
nbh = __getblk(journal->j_fs_dev,
blocknr,
journal->j_blocksize);
if (nbh == NULL) {
printk(KERN_ERR
"JBD: Out of memory "
"during recovery.\n");
err = -ENOMEM;
brelse(bh);
brelse(obh);
goto failed;
}
lock_buffer(nbh);
memcpy(nbh->b_data, obh->b_data,
journal->j_blocksize);
if (flags & JFS_FLAG_ESCAPE) {
*((__be32 *)bh->b_data) =
cpu_to_be32(JFS_MAGIC_NUMBER);
}
BUFFER_TRACE(nbh, "marking dirty");
set_buffer_uptodate(nbh);
mark_buffer_dirty(nbh);
BUFFER_TRACE(nbh, "marking uptodate");
++info->nr_replays;
/* ll_rw_block(WRITE, 1, &nbh); */
unlock_buffer(nbh);
brelse(obh);
brelse(nbh);
}
skip_write:
tagp += sizeof(journal_block_tag_t);
if (!(flags & JFS_FLAG_SAME_UUID))
tagp += 16;
if (flags & JFS_FLAG_LAST_TAG)
break;
}
brelse(bh);
continue;
case JFS_COMMIT_BLOCK:
/* Found an expected commit block: not much to
* do other than move on to the next sequence
* number. */
brelse(bh);
next_commit_ID++;
continue;
case JFS_REVOKE_BLOCK:
/* If we aren't in the REVOKE pass, then we can
* just skip over this block. */
if (pass != PASS_REVOKE) {
brelse(bh);
continue;
}
err = scan_revoke_records(journal, bh,
next_commit_ID, info);
brelse(bh);
if (err)
goto failed;
continue;
default:
jbd_debug(3, "Unrecognised magic %d, end of scan.\n",
blocktype);
goto done;
}
}
done:
/*
* We broke out of the log scan loop: either we came to the
* known end of the log or we found an unexpected block in the
* log. If the latter happened, then we know that the "current"
* transaction marks the end of the valid log.
*/
if (pass == PASS_SCAN)
info->end_transaction = next_commit_ID;
else {
/* It's really bad news if different passes end up at
* different places (but possible due to IO errors). */
if (info->end_transaction != next_commit_ID) {
printk (KERN_ERR "JBD: recovery pass %d ended at "
"transaction %u, expected %u\n",
pass, next_commit_ID, info->end_transaction);
if (!success)
success = -EIO;
}
}
return success;
failed:
return err;
}
static int do_one_pass(journal_t *journal,
struct recovery_info *info, enum passtype pass)
{
unsigned int first_commit_ID, next_commit_ID;
unsigned long long next_log_block;
int err, success = 0;
journal_superblock_t * sb;
journal_header_t * tmp;
struct buffer_head * bh;
unsigned int sequence;
int blocktype;
int tag_bytes = journal_tag_bytes(journal);
__u32 crc32_sum = ~0; /* Transactional Checksums */
/* Precompute the maximum metadata descriptors in a descriptor block */
int MAX_BLOCKS_PER_DESC;
MAX_BLOCKS_PER_DESC = ((journal->j_blocksize-sizeof(journal_header_t))
/ tag_bytes);
/*
* First thing is to establish what we expect to find in the log
* (in terms of transaction IDs), and where (in terms of log
* block offsets): query the superblock.
*/
sb = journal->j_superblock;
next_commit_ID = be32_to_cpu(sb->s_sequence);
next_log_block = be32_to_cpu(sb->s_start);
first_commit_ID = next_commit_ID;
if (pass == PASS_SCAN)
info->start_transaction = first_commit_ID;
jbd_debug(1, "Starting recovery pass %d\n", pass);
/*
* Now we walk through the log, transaction by transaction,
* making sure that each transaction has a commit block in the
* expected place. Each complete transaction gets replayed back
* into the main filesystem.
*/
while (1) {
int flags;
char * tagp;
journal_block_tag_t * tag;
struct buffer_head * obh;
struct buffer_head * nbh;
cond_resched();
/* If we already know where to stop the log traversal,
* check right now that we haven't gone past the end of
* the log. */
if (pass != PASS_SCAN)
if (tid_geq(next_commit_ID, info->end_transaction))
break;
jbd_debug(2, "Scanning for sequence ID %u at %lu/%lu\n",
next_commit_ID, next_log_block, journal->j_last);
/* Skip over each chunk of the transaction looking
* either the next descriptor block or the final commit
* record. */
jbd_debug(3, "JBD: checking block %ld\n", next_log_block);
err = jread(&bh, journal, next_log_block);
if (err)
goto failed;
next_log_block++;
wrap(journal, next_log_block);
/* What kind of buffer is it?
*
* If it is a descriptor block, check that it has the
* expected sequence number. Otherwise, we're all done
* here. */
tmp = (journal_header_t *)bh->b_data;
if (tmp->h_magic != cpu_to_be32(JFS_MAGIC_NUMBER)) {
brelse(bh);
break;
}
blocktype = be32_to_cpu(tmp->h_blocktype);
sequence = be32_to_cpu(tmp->h_sequence);
jbd_debug(3, "Found magic %d, sequence %d\n",
blocktype, sequence);
if (sequence != next_commit_ID) {
brelse(bh);
break;
}
/* OK, we have a valid descriptor block which matches
* all of the sequence number checks. What are we going
* to do with it? That depends on the pass... */
switch(blocktype) {
case JFS_DESCRIPTOR_BLOCK:
/* If it is a valid descriptor block, replay it
* in pass REPLAY; if journal_checksums enabled, then
* calculate checksums in PASS_SCAN, otherwise,
* just skip over the blocks it describes. */
if (pass != PASS_REPLAY) {
if (pass == PASS_SCAN &&
JFS_HAS_COMPAT_FEATURE(journal,
JFS_FEATURE_COMPAT_CHECKSUM) &&
!info->end_transaction) {
if (calc_chksums(journal, bh,
&next_log_block,
&crc32_sum)) {
brelse(bh);
break;
}
brelse(bh);
continue;
}
next_log_block += count_tags(journal, bh);
wrap(journal, next_log_block);
brelse(bh);
continue;
}
/* A descriptor block: we can now write all of
* the data blocks. Yay, useful work is finally
* getting done here! */
tagp = &bh->b_data[sizeof(journal_header_t)];
while ((tagp - bh->b_data + tag_bytes)
<= journal->j_blocksize) {
unsigned long long io_block;
tag = (journal_block_tag_t *) tagp;
flags = be32_to_cpu(tag->t_flags);
io_block = next_log_block++;
wrap(journal, next_log_block);
err = jread(&obh, journal, io_block);
if (err) {
/* Recover what we can, but
* report failure at the end. */
success = err;
printk (KERN_ERR
"JBD: IO error %d recovering "
"block %llu in log\n",
err, io_block);
} else {
unsigned long long blocknr;
J_ASSERT(obh != NULL);
blocknr = read_tag_block(tag_bytes,
tag);
/* If the block has been
* revoked, then we're all done
* here. */
if (journal_test_revoke
(journal, blocknr,
next_commit_ID)) {
brelse(obh);
++info->nr_revoke_hits;
goto skip_write;
}
/* Find a buffer for the new
* data being restored */
nbh = __getblk(journal->j_fs_dev,
blocknr,
journal->j_blocksize);
if (nbh == NULL) {
printk(KERN_ERR
"JBD: Out of memory "
"during recovery.\n");
err = -ENOMEM;
brelse(bh);
brelse(obh);
goto failed;
}
lock_buffer(nbh);
memcpy(nbh->b_data, obh->b_data,
journal->j_blocksize);
if (flags & JFS_FLAG_ESCAPE) {
journal_header_t *header;
header = (journal_header_t *) &nbh->b_data[0];
header->h_magic = cpu_to_be32(JFS_MAGIC_NUMBER);
}
BUFFER_TRACE(nbh, "marking dirty");
set_buffer_uptodate(nbh);
mark_buffer_dirty(nbh);
BUFFER_TRACE(nbh, "marking uptodate");
++info->nr_replays;
/* ll_rw_block(WRITE, 1, &nbh); */
unlock_buffer(nbh);
brelse(obh);
brelse(nbh);
}
skip_write:
tagp += tag_bytes;
if (!(flags & JFS_FLAG_SAME_UUID))
tagp += 16;
if (flags & JFS_FLAG_LAST_TAG)
break;
}
brelse(bh);
continue;
case JFS_COMMIT_BLOCK:
jbd_debug(3, "Commit block for #%u found\n",
next_commit_ID);
/* How to differentiate between interrupted commit
* and journal corruption ?
*
* {nth transaction}
* Checksum Verification Failed
* |
* ____________________
* | |
* async_commit sync_commit
* | |
* | GO TO NEXT "Journal Corruption"
* | TRANSACTION
* |
* {(n+1)th transanction}
* |
* _______|______________
* | |
* Commit block found Commit block not found
* | |
* "Journal Corruption" |
* _____________|_________
* | |
* nth trans corrupt OR nth trans
* and (n+1)th interrupted interrupted
* before commit block
* could reach the disk.
* (Cannot find the difference in above
* mentioned conditions. Hence assume
* "Interrupted Commit".)
*/
/* Found an expected commit block: if checksums
* are present verify them in PASS_SCAN; else not
* much to do other than move on to the next sequence
* number. */
if (pass == PASS_SCAN &&
JFS_HAS_COMPAT_FEATURE(journal,
JFS_FEATURE_COMPAT_CHECKSUM)) {
int chksum_err, chksum_seen;
struct commit_header *cbh =
(struct commit_header *)bh->b_data;
unsigned found_chksum =
be32_to_cpu(cbh->h_chksum[0]);
chksum_err = chksum_seen = 0;
jbd_debug(3, "Checksums %x %x\n",
crc32_sum, found_chksum);
if (info->end_transaction) {
journal->j_failed_commit =
info->end_transaction;
brelse(bh);
break;
}
if (crc32_sum == found_chksum &&
cbh->h_chksum_type == JBD2_CRC32_CHKSUM &&
cbh->h_chksum_size ==
JBD2_CRC32_CHKSUM_SIZE)
chksum_seen = 1;
else if (!(cbh->h_chksum_type == 0 &&
cbh->h_chksum_size == 0 &&
found_chksum == 0 &&
!chksum_seen))
/*
* If fs is mounted using an old kernel and then
* kernel with journal_chksum is used then we
* get a situation where the journal flag has
* checksum flag set but checksums are not
* present i.e chksum = 0, in the individual
* commit blocks.
* Hence to avoid checksum failures, in this
* situation, this extra check is added.
*/
chksum_err = 1;
if (chksum_err) {
info->end_transaction = next_commit_ID;
jbd_debug(1, "Checksum_err %x %x\n",
crc32_sum, found_chksum);
if (!JFS_HAS_INCOMPAT_FEATURE(journal,
JFS_FEATURE_INCOMPAT_ASYNC_COMMIT)){
journal->j_failed_commit =
next_commit_ID;
brelse(bh);
break;
}
}
crc32_sum = ~0;
}
brelse(bh);
next_commit_ID++;
continue;
case JFS_REVOKE_BLOCK:
/* If we aren't in the REVOKE pass, then we can
* just skip over this block. */
if (pass != PASS_REVOKE) {
brelse(bh);
continue;
}
err = scan_revoke_records(journal, bh,
next_commit_ID, info);
brelse(bh);
if (err)
goto failed;
continue;
default:
jbd_debug(3, "Unrecognised magic %d, end of scan.\n",
blocktype);
brelse(bh);
goto done;
}
}
done:
/*
* We broke out of the log scan loop: either we came to the
* known end of the log or we found an unexpected block in the
* log. If the latter happened, then we know that the "current"
* transaction marks the end of the valid log.
*/
if (pass == PASS_SCAN) {
if (!info->end_transaction)
info->end_transaction = next_commit_ID;
} else {
/* It's really bad news if different passes end up at
* different places (but possible due to IO errors). */
if (info->end_transaction != next_commit_ID) {
printk (KERN_ERR "JBD: recovery pass %d ended at "
"transaction %u, expected %u\n",
pass, next_commit_ID, info->end_transaction);
if (!success)
success = -EIO;
}
}
return success;
failed:
return err;
}
void *load_exe(char *filename, Exe_head *eh)
{
long retval;
char *alligned_buf;
char *alloc_buf = NULL;
long (*rfunc)();
unsigned long code_offset;
unsigned long init_size;
unsigned long bss_size;
unsigned long total_size;
void *v;
unsigned long fixup_offset;
UWORD fixup[2];
UWORD *segpt;
UWORD code_seg;
int f;
unsigned i;
int ok = 0;
if ((f = jopen(filename, JREADONLY)) == 0)
{
cant_find(filename);
return(NULL);
}
if (!verify_exe_head(f, eh))
goto OUT;
code_offset = eh->head_size;
code_offset *= 16; /* make it a paragraph */
init_size = eh->blocks;
init_size *= 512;
init_size += eh->mod512;
if (eh->mod512 != 0)
init_size -= 512;
init_size -= code_offset;
bss_size = eh->min_data;
bss_size *= 16;
total_size = init_size + bss_size;
if ((alloc_buf = begmem((unsigned)total_size+16)) == NULL)
goto OUT;
code_seg = ptr_seg(alloc_buf) + 1;
alligned_buf = make_ptr(0, code_seg);
zero_structure(alligned_buf, (unsigned)total_size);
jseek(f, code_offset, JSEEK_START);
if (jread(f, alligned_buf, init_size) < init_size)
{
truncated(filename);
goto OUT;
}
v = alligned_buf;
eh->entry_point = v;
if (eh->reloc_count > 0)
{
fixup_offset = eh->reloc_list;
jseek(f, fixup_offset, JSEEK_START);
for (i=0; i<eh->reloc_count; i++)
{
if (jread(f, fixup, sizeof(fixup)) != sizeof(fixup))
{
truncated(filename);
goto OUT;
}
segpt = make_ptr(fixup[0], code_seg + fixup[1]);
segpt[0] += code_seg;
}
}
ok = 1;
OUT:
if (!ok)
{
gentle_freemem(alloc_buf);
alloc_buf = NULL;
}
jclose(f);
return(alloc_buf);
}
int
main(int ac, char **av)
{
const char *input_prefix = NULL;
char *output_transid_file = NULL;
char *mirror_transid_file = NULL;
const char *mirror_directory = ".";
char *record_prefix = NULL;
char *record_transid_file = NULL;
struct jsession jsdebug;
struct jsession jsoutput;
struct jsession jsmirror;
char *ptr;
int64_t mirror_transid;
int64_t output_transid;
int64_t record_transid;
int64_t transid;
int input_fd;
struct stat st;
struct jfile *jf;
struct jdata *jd;
int ch;
while ((ch = getopt(ac, av, "2c:dfm:o:s:uvw:D:O:W:F")) != -1) {
switch(ch) {
case '2':
jmodes |= JMODEF_INPUT_FULL;
break;
case 'c':
trans_count = strtoll(optarg, &ptr, 0);
switch(*ptr) {
case 't':
trans_count *= 1024;
/* fall through */
case 'g':
trans_count *= 1024;
/* fall through */
case 'm':
trans_count *= 1024;
/* fall through */
case 'k':
trans_count *= 1024;
break;
case 0:
break;
default:
fprintf(stderr, "Bad suffix for value specified with -c, use 'k', 'm', 'g', 't', or nothing\n");
usage(av[0]);
}
break;
case 'd':
jmodes |= JMODEF_DEBUG;
break;
case 'f':
jmodes |= JMODEF_LOOP_FOREVER;
break;
case 'v':
++verbose_opt;
break;
case 'm':
jmodes |= JMODEF_MIRROR;
if (strcmp(optarg, "none") != 0)
mirror_transid_file = optarg;
break;
case 'O':
jmodes |= JMODEF_OUTPUT_FULL;
/* fall through */
case 'o':
jmodes |= JMODEF_OUTPUT;
if (strcmp(optarg, "none") != 0)
output_transid_file = optarg;
break;
case 's':
prefix_file_size = strtoll(optarg, &ptr, 0);
switch(*ptr) {
case 't':
prefix_file_size *= 1024;
/* fall through */
case 'g':
prefix_file_size *= 1024;
/* fall through */
case 'm':
prefix_file_size *= 1024;
/* fall through */
case 'k':
prefix_file_size *= 1024;
break;
case 0:
break;
default:
fprintf(stderr, "Bad suffix for value specified with -s, use 'k', 'm', 'g', 't', or nothing\n");
usage(av[0]);
}
break;
case 'u':
jdirection = JD_BACKWARDS;
break;
case 'W':
jmodes |= JMODEF_RECORD_TMP;
/* fall through */
case 'w':
jmodes |= JMODEF_RECORD;
record_prefix = optarg;
asprintf(&record_transid_file, "%s.transid", record_prefix);
break;
case 'D':
mirror_directory = optarg;
break;
case 'F':
++fsync_opt;
break;
default:
fprintf(stderr, "unknown option: -%c\n", optopt);
usage(av[0]);
}
}
/*
* Sanity checks
*/
if ((jmodes & JMODEF_COMMAND_MASK) == 0)
usage(av[0]);
if (optind > ac + 1) {
fprintf(stderr, "Only one input file or prefix may be specified,\n"
"or zero if stdin is to be the input.\n");
usage(av[0]);
}
if (strcmp(mirror_directory, ".") != 0) {
struct stat sb;
if (stat(mirror_directory, &sb) != 0) {
perror ("Could not stat mirror directory");
usage(av[0]);
}
if (!S_ISDIR(sb.st_mode))
{
fprintf (stderr, "Mirror directory '%s' is not a directory\n", mirror_directory);
usage(av[0]);
}
}
if (jdirection == JD_BACKWARDS && (jmodes & (JMODEF_RECORD|JMODEF_OUTPUT))) {
fprintf(stderr, "Undo mode is only good in mirroring mode and "
"cannot be mixed with other modes.\n");
exit(1);
}
/*
* STEP1 - OPEN INPUT
*
* The input will either be a pipe, a regular file, or a journaling
* file prefix.
*/
jf = NULL;
if (optind == ac) {
input_prefix = "<stdin>";
input_fd = 0;
if (fstat(0, &st) < 0 || !S_ISREG(st.st_mode)) {
jmodes |= JMODEF_INPUT_PIPE;
if (jdirection == JD_BACKWARDS) {
fprintf(stderr, "Cannot scan journals on pipes backwards\n");
usage(av[0]);
}
}
jf = jopen_fd(input_fd);
} else if (stat(av[optind], &st) == 0 && S_ISREG(st.st_mode)) {
input_prefix = av[optind];
if ((input_fd = open(av[optind], O_RDONLY)) != 0) {
jf = jopen_fd(input_fd);
} else {
jf = NULL;
}
} else {
input_prefix = av[optind];
jf = jopen_prefix(input_prefix, 0);
jmodes |= JMODEF_INPUT_PREFIX;
}
if (jf == NULL) {
fprintf(stderr, "Unable to open input %s: %s\n",
input_prefix, strerror(errno));
exit(1);
}
/*
* STEP 1 - SYNCHRONIZING THE INPUT STREAM
*
* Figure out the starting point for our various output modes. Figure
* out the earliest transaction id and try to seek to that point,
* otherwise we might have to scan through terrabytes of data.
*
* Invalid transid's will be set to 0, but it should also be noted
* that 0 is also a valid transid.
*/
get_transid_from_file(output_transid_file, &output_transid,
JMODEF_OUTPUT_TRANSID_GOOD);
get_transid_from_file(mirror_transid_file, &mirror_transid,
JMODEF_MIRROR_TRANSID_GOOD);
get_transid_from_file(record_transid_file, &record_transid,
JMODEF_RECORD_TRANSID_GOOD);
transid = LLONG_MAX;
if ((jmodes & JMODEF_OUTPUT_TRANSID_GOOD) && output_transid < transid)
transid = output_transid;
if ((jmodes & JMODEF_MIRROR_TRANSID_GOOD) && mirror_transid < transid)
transid = mirror_transid;
if ((jmodes & JMODEF_RECORD_TRANSID_GOOD) && record_transid < transid)
transid = record_transid;
if ((jmodes & JMODEF_TRANSID_GOOD_MASK) == 0)
transid = 0;
if (verbose_opt) {
if (jmodes & JMODEF_OUTPUT) {
fprintf(stderr, "Starting transid for OUTPUT: %016jx\n",
(uintmax_t)output_transid);
}
if (jmodes & JMODEF_MIRROR) {
fprintf(stderr, "Starting transid for MIRROR: %016jx\n",
(uintmax_t)mirror_transid);
}
if (jmodes & JMODEF_RECORD) {
fprintf(stderr, "Starting transid for RECORD: %016jx\n",
(uintmax_t)record_transid);
}
}
if (strcmp(mirror_directory, ".") != 0) {
if (chdir (mirror_directory) != 0) {
perror ("Could not enter mirror directory");
exit (1);
}
}
/*
* Now it gets more difficult. If we are recording then the input
* could be representative of continuing data and not have any
* prior, older data that the output or mirror modes might need. Those
* modes must work off the recording data even as we write to it.
* In that case we fork and have the sub-processes work off the
* record output.
*
* Then we take the input and start recording.
*/
if (jmodes & JMODEF_RECORD) {
if (jrecord_init(record_prefix) < 0) {
fprintf(stderr, "Unable to initialize file set for: %s\n",
record_prefix);
exit(1);
}
if (jmodes & JMODEF_MIRROR) {
fork_subprocess(jf, jscan_do_mirror, record_prefix,
mirror_transid_file,
mirror_directory, mirror_transid);
/* XXX ack stream for temporary record file removal */
}
if (jmodes & JMODEF_OUTPUT) {
fork_subprocess(jf, jscan_do_output, record_prefix,
record_transid_file,
NULL, output_transid);
/* XXX ack stream for temporary record file removal */
}
jscan_do_record(jf, record_transid_file, record_prefix, record_transid);
exit(0);
}
/*
* If the input is a prefix set we can just pass it to the appropriate
* jscan_do_*() function. If we are doing both output and mirroring
* we fork the mirror and do the output in the foreground since that
* is going to stdout.
*/
if (jmodes & JMODEF_INPUT_PREFIX) {
if ((jmodes & JMODEF_OUTPUT) && (jmodes & JMODEF_MIRROR)) {
fork_subprocess(jf, jscan_do_mirror, input_prefix,
mirror_transid_file,
mirror_directory, mirror_transid);
jscan_do_output(jf, output_transid_file, NULL, output_transid);
} else if (jmodes & JMODEF_OUTPUT) {
jscan_do_output(jf, output_transid_file, NULL, output_transid);
} else if (jmodes & JMODEF_MIRROR) {
jscan_do_mirror(jf, mirror_transid_file, mirror_directory,
mirror_transid);
} else if (jmodes & JMODEF_DEBUG) {
jscan_do_debug(jf, NULL, NULL, 0);
}
exit(0);
}
/*
* The input is not a prefix set and we are not recording, which means
* we have to transfer the data on the input pipe to the output and
* mirroring code on the fly. This also means that we must keep track
* of meta-data records in-memory. However, if the input is a regular
* file we *CAN* try to optimize where we start reading.
*
* NOTE: If the mirroring code encounters a transaction record that is
* not marked begin, and it does not have the begin record, it will
* attempt to locate the begin record if the input is not a pipe, then
* seek back.
*/
if ((jmodes & JMODEF_TRANSID_GOOD_MASK) && !(jmodes & JMODEF_INPUT_PIPE))
jd = jseek(jf, transid, jdirection);
else
jd = jread(jf, NULL, jdirection);
jmodes |= JMODEF_MEMORY_TRACKING;
jsession_init(&jsdebug, jf, jdirection,
NULL, 0);
jsession_init(&jsoutput, jf, jdirection,
output_transid_file, output_transid);
jsession_init(&jsmirror, jf, jdirection,
mirror_transid_file, mirror_transid);
jsmirror.ss_mirror_directory = mirror_directory;
while (jd != NULL) {
if ((jmodes & JMODEF_DEBUG) && jsession_check(&jsdebug, jd))
dump_debug(&jsdebug, jd);
if ((jmodes & JMODEF_OUTPUT) && jsession_check(&jsoutput, jd))
dump_output(&jsoutput, jd);
if ((jmodes & JMODEF_MIRROR) && jsession_check(&jsmirror, jd))
dump_mirror(&jsmirror, jd);
if (donecheck(jdirection, jd, transid)) {
jfree(jf, jd);
break;
}
jd = jread(jf, jd, jdirection);
}
jclose(jf);
jsession_term(&jsdebug);
jsession_term(&jsoutput);
jsession_term(&jsmirror);
return(0);
}
static int do_one_pass(journal_t *journal,
struct recovery_info *info, enum passtype pass)
{
unsigned int first_commit_ID, next_commit_ID;
unsigned long next_log_block;
int err, success = 0;
journal_superblock_t * sb;
journal_header_t * tmp;
struct buffer_head * bh;
unsigned int sequence;
int blocktype;
int tag_bytes = journal_tag_bytes(journal);
__u32 crc32_sum = ~0;
sb = journal->j_superblock;
next_commit_ID = be32_to_cpu(sb->s_sequence);
next_log_block = be32_to_cpu(sb->s_start);
first_commit_ID = next_commit_ID;
if (pass == PASS_SCAN)
info->start_transaction = first_commit_ID;
jbd_debug(1, "Starting recovery pass %d\n", pass);
while (1) {
int flags;
char * tagp;
journal_block_tag_t * tag;
struct buffer_head * obh;
struct buffer_head * nbh;
cond_resched();
if (pass != PASS_SCAN)
if (tid_geq(next_commit_ID, info->end_transaction))
break;
jbd_debug(2, "Scanning for sequence ID %u at %lu/%lu\n",
next_commit_ID, next_log_block, journal->j_last);
jbd_debug(3, "JBD2: checking block %ld\n", next_log_block);
err = jread(&bh, journal, next_log_block);
if (err)
goto failed;
next_log_block++;
wrap(journal, next_log_block);
tmp = (journal_header_t *)bh->b_data;
if (tmp->h_magic != cpu_to_be32(JBD2_MAGIC_NUMBER)) {
brelse(bh);
break;
}
blocktype = be32_to_cpu(tmp->h_blocktype);
sequence = be32_to_cpu(tmp->h_sequence);
jbd_debug(3, "Found magic %d, sequence %d\n",
blocktype, sequence);
if (sequence != next_commit_ID) {
brelse(bh);
break;
}
switch(blocktype) {
case JBD2_DESCRIPTOR_BLOCK:
if (pass != PASS_REPLAY) {
if (pass == PASS_SCAN &&
JBD2_HAS_COMPAT_FEATURE(journal,
JBD2_FEATURE_COMPAT_CHECKSUM) &&
!info->end_transaction) {
if (calc_chksums(journal, bh,
&next_log_block,
&crc32_sum)) {
put_bh(bh);
break;
}
put_bh(bh);
continue;
}
next_log_block += count_tags(journal, bh);
wrap(journal, next_log_block);
put_bh(bh);
continue;
}
tagp = &bh->b_data[sizeof(journal_header_t)];
while ((tagp - bh->b_data + tag_bytes)
<= journal->j_blocksize) {
unsigned long io_block;
tag = (journal_block_tag_t *) tagp;
flags = be32_to_cpu(tag->t_flags);
io_block = next_log_block++;
wrap(journal, next_log_block);
err = jread(&obh, journal, io_block);
if (err) {
success = err;
printk(KERN_ERR
"JBD2: IO error %d recovering "
"block %ld in log\n",
err, io_block);
} else {
unsigned long long blocknr;
J_ASSERT(obh != NULL);
blocknr = read_tag_block(tag_bytes,
tag);
if (jbd2_journal_test_revoke
(journal, blocknr,
next_commit_ID)) {
brelse(obh);
++info->nr_revoke_hits;
goto skip_write;
}
nbh = __getblk(journal->j_fs_dev,
blocknr,
journal->j_blocksize);
if (nbh == NULL) {
printk(KERN_ERR
"JBD2: Out of memory "
"during recovery.\n");
err = -ENOMEM;
brelse(bh);
brelse(obh);
goto failed;
}
lock_buffer(nbh);
memcpy(nbh->b_data, obh->b_data,
journal->j_blocksize);
if (flags & JBD2_FLAG_ESCAPE) {
*((__be32 *)nbh->b_data) =
cpu_to_be32(JBD2_MAGIC_NUMBER);
}
BUFFER_TRACE(nbh, "marking dirty");
set_buffer_uptodate(nbh);
mark_buffer_dirty(nbh);
BUFFER_TRACE(nbh, "marking uptodate");
++info->nr_replays;
unlock_buffer(nbh);
brelse(obh);
brelse(nbh);
}
skip_write:
tagp += tag_bytes;
if (!(flags & JBD2_FLAG_SAME_UUID))
tagp += 16;
if (flags & JBD2_FLAG_LAST_TAG)
break;
}
brelse(bh);
continue;
case JBD2_COMMIT_BLOCK:
if (pass == PASS_SCAN &&
JBD2_HAS_COMPAT_FEATURE(journal,
JBD2_FEATURE_COMPAT_CHECKSUM)) {
int chksum_err, chksum_seen;
struct commit_header *cbh =
(struct commit_header *)bh->b_data;
unsigned found_chksum =
be32_to_cpu(cbh->h_chksum[0]);
chksum_err = chksum_seen = 0;
if (info->end_transaction) {
journal->j_failed_commit =
info->end_transaction;
brelse(bh);
break;
}
if (crc32_sum == found_chksum &&
cbh->h_chksum_type == JBD2_CRC32_CHKSUM &&
cbh->h_chksum_size ==
JBD2_CRC32_CHKSUM_SIZE)
chksum_seen = 1;
else if (!(cbh->h_chksum_type == 0 &&
cbh->h_chksum_size == 0 &&
found_chksum == 0 &&
!chksum_seen))
chksum_err = 1;
if (chksum_err) {
info->end_transaction = next_commit_ID;
if (!JBD2_HAS_INCOMPAT_FEATURE(journal,
JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)){
journal->j_failed_commit =
next_commit_ID;
brelse(bh);
break;
}
}
crc32_sum = ~0;
}
brelse(bh);
next_commit_ID++;
continue;
case JBD2_REVOKE_BLOCK:
if (pass != PASS_REVOKE) {
brelse(bh);
continue;
}
err = scan_revoke_records(journal, bh,
next_commit_ID, info);
brelse(bh);
if (err)
goto failed;
continue;
default:
jbd_debug(3, "Unrecognised magic %d, end of scan.\n",
blocktype);
brelse(bh);
goto done;
}
}
done:
if (pass == PASS_SCAN) {
if (!info->end_transaction)
info->end_transaction = next_commit_ID;
} else {
if (info->end_transaction != next_commit_ID) {
printk(KERN_ERR "JBD2: recovery pass %d ended at "
"transaction %u, expected %u\n",
pass, next_commit_ID, info->end_transaction);
if (!success)
success = -EIO;
}
}
return success;
failed:
return err;
}