void lustre_swab_llog_hdr (struct llog_log_hdr *h)
{
print_llog_hdr(h);
lustre_swab_llog_rec(&h->llh_hdr);
print_llog_hdr(h);
}
void lustre_swab_llog_hdr (struct llog_log_hdr *h)
{
ENTRY;
print_llog_hdr(h);
lustre_swab_llog_rec(&h->llh_hdr);
print_llog_hdr(h);
EXIT;
}
/**
* Implementation of the llog_operations::lop_prev_block
*
* This function finds the llog block to return which contains
* record with required index but in reverse order - from end of llog
* to the beginning.
* It is main part of reverse llog processing.
*
* \param[in] env execution environment
* \param[in] loghandle llog handle of the current llog
* \param[in] prev_idx target index to find
* \param[in] buf pointer to data buffer to fill
* \param[in] len required len to read, it is LLOG_CHUNK_SIZE usually.
*
* \retval 0 on successful buffer read
* \retval negative value on error
*/
static int llog_osd_prev_block(const struct lu_env *env,
struct llog_handle *loghandle,
int prev_idx, void *buf, int len)
{
struct llog_thread_info *lgi = llog_info(env);
struct dt_object *o;
struct dt_device *dt;
loff_t cur_offset;
int rc;
ENTRY;
if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
RETURN(-EINVAL);
CDEBUG(D_OTHER, "looking for log index %u\n", prev_idx);
LASSERT(loghandle);
LASSERT(loghandle->lgh_ctxt);
o = loghandle->lgh_obj;
LASSERT(o);
LASSERT(dt_object_exists(o));
dt = lu2dt_dev(o->do_lu.lo_dev);
LASSERT(dt);
cur_offset = LLOG_CHUNK_SIZE;
llog_skip_over(&cur_offset, 0, prev_idx);
rc = dt_attr_get(env, o, &lgi->lgi_attr, BYPASS_CAPA);
if (rc)
GOTO(out, rc);
while (cur_offset < lgi->lgi_attr.la_size) {
struct llog_rec_hdr *rec, *last_rec;
struct llog_rec_tail *tail;
lgi->lgi_buf.lb_len = len;
lgi->lgi_buf.lb_buf = buf;
rc = dt_read(env, o, &lgi->lgi_buf, &cur_offset);
if (rc < 0) {
CERROR("%s: can't read llog block from log "DFID
" offset "LPU64": rc = %d\n",
o->do_lu.lo_dev->ld_obd->obd_name,
PFID(lu_object_fid(&o->do_lu)), cur_offset, rc);
GOTO(out, rc);
}
if (rc == 0) /* end of file, nothing to do */
GOTO(out, rc);
if (rc < sizeof(*tail)) {
CERROR("%s: invalid llog block at log id "DOSTID"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, cur_offset);
GOTO(out, rc = -EINVAL);
}
rec = buf;
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
tail = (struct llog_rec_tail *)((char *)buf + rc -
sizeof(struct llog_rec_tail));
/* get the last record in block */
last_rec = (struct llog_rec_hdr *)((char *)buf + rc -
le32_to_cpu(tail->lrt_len));
if (LLOG_REC_HDR_NEEDS_SWABBING(last_rec))
lustre_swab_llog_rec(last_rec);
LASSERT(last_rec->lrh_index == tail->lrt_index);
/* this shouldn't happen */
if (tail->lrt_index == 0) {
CERROR("%s: invalid llog tail at log id "DOSTID"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, cur_offset);
GOTO(out, rc = -EINVAL);
}
if (tail->lrt_index < prev_idx)
continue;
/* sanity check that the start of the new buffer is no farther
* than the record that we wanted. This shouldn't happen. */
if (rec->lrh_index > prev_idx) {
CERROR("%s: missed desired record? %u > %u\n",
o->do_lu.lo_dev->ld_obd->obd_name,
rec->lrh_index, prev_idx);
GOTO(out, rc = -ENOENT);
}
GOTO(out, rc = 0);
}
GOTO(out, rc = -EIO);
out:
return rc;
}
/* sets:
* - cur_offset to the furthest point read in the log file
* - cur_idx to the log index preceeding cur_offset
* returns -EIO/-EINVAL on error
*/
static int llog_osd_next_block(const struct lu_env *env,
struct llog_handle *loghandle, int *cur_idx,
int next_idx, __u64 *cur_offset, void *buf,
int len)
{
struct llog_thread_info *lgi = llog_info(env);
struct dt_object *o;
struct dt_device *dt;
int rc;
ENTRY;
LASSERT(env);
LASSERT(lgi);
if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
RETURN(-EINVAL);
CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
next_idx, *cur_idx, *cur_offset);
LASSERT(loghandle);
LASSERT(loghandle->lgh_ctxt);
o = loghandle->lgh_obj;
LASSERT(o);
LASSERT(dt_object_exists(o));
dt = lu2dt_dev(o->do_lu.lo_dev);
LASSERT(dt);
rc = dt_attr_get(env, o, &lgi->lgi_attr, BYPASS_CAPA);
if (rc)
GOTO(out, rc);
while (*cur_offset < lgi->lgi_attr.la_size) {
struct llog_rec_hdr *rec, *last_rec;
struct llog_rec_tail *tail;
llog_skip_over(cur_offset, *cur_idx, next_idx);
/* read up to next LLOG_CHUNK_SIZE block */
lgi->lgi_buf.lb_len = LLOG_CHUNK_SIZE -
(*cur_offset & (LLOG_CHUNK_SIZE - 1));
lgi->lgi_buf.lb_buf = buf;
/* Note: read lock is not needed around la_size get above at
* the time of dt_attr_get(). There are only two cases that
* matter. Either la_size == cur_offset, in which case the
* entire read is skipped, or la_size > cur_offset and the loop
* is entered and this thread is blocked at dt_read_lock()
* until the write is completed. When the write completes, then
* the dt_read() will be done with the full length, and will
* get the full data.
*/
dt_read_lock(env, o, 0);
rc = dt_read(env, o, &lgi->lgi_buf, cur_offset);
dt_read_unlock(env, o);
if (rc < 0) {
CERROR("%s: can't read llog block from log "DFID
" offset "LPU64": rc = %d\n",
o->do_lu.lo_dev->ld_obd->obd_name,
PFID(lu_object_fid(&o->do_lu)), *cur_offset,
rc);
GOTO(out, rc);
}
if (rc < len) {
/* signal the end of the valid buffer to
* llog_process */
memset(buf + rc, 0, len - rc);
}
if (rc == 0) /* end of file, nothing to do */
GOTO(out, rc);
if (rc < sizeof(*tail)) {
CERROR("%s: invalid llog block at log id "LPU64"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
loghandle->lgh_id.lgl_oid,
loghandle->lgh_id.lgl_ogen, *cur_offset);
GOTO(out, rc = -EINVAL);
}
rec = buf;
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
tail = (struct llog_rec_tail *)((char *)buf + rc -
sizeof(struct llog_rec_tail));
/* get the last record in block */
last_rec = (struct llog_rec_hdr *)((char *)buf + rc -
le32_to_cpu(tail->lrt_len));
if (LLOG_REC_HDR_NEEDS_SWABBING(last_rec))
lustre_swab_llog_rec(last_rec);
LASSERT(last_rec->lrh_index == tail->lrt_index);
*cur_idx = tail->lrt_index;
/* this shouldn't happen */
if (tail->lrt_index == 0) {
CERROR("%s: invalid llog tail at log id "LPU64"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
loghandle->lgh_id.lgl_oid,
loghandle->lgh_id.lgl_ogen, *cur_offset);
GOTO(out, rc = -EINVAL);
}
if (tail->lrt_index < next_idx)
continue;
/* sanity check that the start of the new buffer is no farther
* than the record that we wanted. This shouldn't happen. */
if (rec->lrh_index > next_idx) {
CERROR("%s: missed desired record? %u > %u\n",
o->do_lu.lo_dev->ld_obd->obd_name,
rec->lrh_index, next_idx);
GOTO(out, rc = -ENOENT);
}
GOTO(out, rc = 0);
}
GOTO(out, rc = -EIO);
out:
return rc;
}
int llog_reverse_process(const struct lu_env *env,
struct llog_handle *loghandle, llog_cb_t cb,
void *data, void *catdata)
{
struct llog_log_hdr *llh = loghandle->lgh_hdr;
struct llog_process_cat_data *cd = catdata;
void *buf;
int rc = 0, first_index = 1, index, idx;
OBD_ALLOC(buf, LLOG_CHUNK_SIZE);
if (!buf)
return -ENOMEM;
if (cd != NULL)
first_index = cd->lpcd_first_idx + 1;
if (cd != NULL && cd->lpcd_last_idx)
index = cd->lpcd_last_idx;
else
index = LLOG_BITMAP_BYTES * 8 - 1;
while (rc == 0) {
struct llog_rec_hdr *rec;
struct llog_rec_tail *tail;
/* skip records not set in bitmap */
while (index >= first_index &&
!ext2_test_bit(index, llh->llh_bitmap))
--index;
LASSERT(index >= first_index - 1);
if (index == first_index - 1)
break;
/* get the buf with our target record; avoid old garbage */
memset(buf, 0, LLOG_CHUNK_SIZE);
rc = llog_prev_block(env, loghandle, index, buf,
LLOG_CHUNK_SIZE);
if (rc)
GOTO(out, rc);
rec = buf;
idx = rec->lrh_index;
CDEBUG(D_RPCTRACE, "index %u : idx %u\n", index, idx);
while (idx < index) {
rec = (void *)rec + rec->lrh_len;
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
idx ++;
}
LASSERT(idx == index);
tail = (void *)rec + rec->lrh_len - sizeof(*tail);
/* process records in buffer, starting where we found one */
while ((void *)tail > buf) {
if (tail->lrt_index == 0)
GOTO(out, rc = 0); /* no more records */
/* if set, process the callback on this record */
if (ext2_test_bit(index, llh->llh_bitmap)) {
rec = (void *)tail - tail->lrt_len +
sizeof(*tail);
rc = cb(env, loghandle, rec, data);
if (rc == LLOG_PROC_BREAK) {
GOTO(out, rc);
} else if (rc == LLOG_DEL_RECORD) {
llog_cancel_rec(env, loghandle,
tail->lrt_index);
rc = 0;
}
if (rc)
GOTO(out, rc);
}
/* previous record, still in buffer? */
--index;
if (index < first_index)
GOTO(out, rc = 0);
tail = (void *)tail - tail->lrt_len;
}
}
out:
if (buf)
OBD_FREE(buf, LLOG_CHUNK_SIZE);
return rc;
}
static int llog_process_thread(void *arg)
{
struct llog_process_info *lpi = arg;
struct llog_handle *loghandle = lpi->lpi_loghandle;
struct llog_log_hdr *llh = loghandle->lgh_hdr;
struct llog_process_cat_data *cd = lpi->lpi_catdata;
char *buf;
__u64 cur_offset = LLOG_CHUNK_SIZE;
__u64 last_offset;
int rc = 0, index = 1, last_index;
int saved_index = 0;
int last_called_index = 0;
LASSERT(llh);
OBD_ALLOC(buf, LLOG_CHUNK_SIZE);
if (!buf) {
lpi->lpi_rc = -ENOMEM;
return 0;
}
if (cd != NULL) {
last_called_index = cd->lpcd_first_idx;
index = cd->lpcd_first_idx + 1;
}
if (cd != NULL && cd->lpcd_last_idx)
last_index = cd->lpcd_last_idx;
else
last_index = LLOG_BITMAP_BYTES * 8 - 1;
while (rc == 0) {
struct llog_rec_hdr *rec;
/* skip records not set in bitmap */
while (index <= last_index &&
!ext2_test_bit(index, llh->llh_bitmap))
++index;
LASSERT(index <= last_index + 1);
if (index == last_index + 1)
break;
repeat:
CDEBUG(D_OTHER, "index: %d last_index %d\n",
index, last_index);
/* get the buf with our target record; avoid old garbage */
memset(buf, 0, LLOG_CHUNK_SIZE);
last_offset = cur_offset;
rc = llog_next_block(lpi->lpi_env, loghandle, &saved_index,
index, &cur_offset, buf, LLOG_CHUNK_SIZE);
if (rc)
GOTO(out, rc);
/* NB: when rec->lrh_len is accessed it is already swabbed
* since it is used at the "end" of the loop and the rec
* swabbing is done at the beginning of the loop. */
for (rec = (struct llog_rec_hdr *)buf;
(char *)rec < buf + LLOG_CHUNK_SIZE;
rec = (struct llog_rec_hdr *)((char *)rec + rec->lrh_len)){
CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n",
rec, rec->lrh_type);
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
rec->lrh_type, rec->lrh_index);
if (rec->lrh_index == 0) {
/* probably another rec just got added? */
if (index <= loghandle->lgh_last_idx)
GOTO(repeat, rc = 0);
GOTO(out, rc = 0); /* no more records */
}
if (rec->lrh_len == 0 ||
rec->lrh_len > LLOG_CHUNK_SIZE) {
CWARN("invalid length %d in llog record for "
"index %d/%d\n", rec->lrh_len,
rec->lrh_index, index);
GOTO(out, rc = -EINVAL);
}
if (rec->lrh_index < index) {
CDEBUG(D_OTHER, "skipping lrh_index %d\n",
rec->lrh_index);
continue;
}
CDEBUG(D_OTHER,
"lrh_index: %d lrh_len: %d (%d remains)\n",
rec->lrh_index, rec->lrh_len,
(int)(buf + LLOG_CHUNK_SIZE - (char *)rec));
loghandle->lgh_cur_idx = rec->lrh_index;
loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
last_offset;
/* if set, process the callback on this record */
if (ext2_test_bit(index, llh->llh_bitmap)) {
rc = lpi->lpi_cb(lpi->lpi_env, loghandle, rec,
lpi->lpi_cbdata);
last_called_index = index;
if (rc == LLOG_PROC_BREAK) {
GOTO(out, rc);
} else if (rc == LLOG_DEL_RECORD) {
llog_cancel_rec(lpi->lpi_env,
loghandle,
rec->lrh_index);
rc = 0;
}
if (rc)
GOTO(out, rc);
} else {
CDEBUG(D_OTHER, "Skipped index %d\n", index);
}
/* next record, still in buffer? */
++index;
if (index > last_index)
GOTO(out, rc = 0);
}
}
out:
if (cd != NULL)
cd->lpcd_last_idx = last_called_index;
OBD_FREE(buf, LLOG_CHUNK_SIZE);
lpi->lpi_rc = rc;
return 0;
}
static int llog_process_thread(void *arg)
{
struct llog_process_info *lpi = arg;
struct llog_handle *loghandle = lpi->lpi_loghandle;
struct llog_log_hdr *llh = loghandle->lgh_hdr;
struct llog_process_cat_data *cd = lpi->lpi_catdata;
char *buf;
size_t chunk_size;
__u64 cur_offset, tmp_offset;
int rc = 0, index = 1, last_index;
int saved_index = 0;
int last_called_index = 0;
ENTRY;
if (llh == NULL)
RETURN(-EINVAL);
cur_offset = chunk_size = llh->llh_hdr.lrh_len;
/* expect chunk_size to be power of two */
LASSERT(is_power_of_2(chunk_size));
OBD_ALLOC_LARGE(buf, chunk_size);
if (buf == NULL) {
lpi->lpi_rc = -ENOMEM;
RETURN(0);
}
if (cd != NULL) {
last_called_index = cd->lpcd_first_idx;
index = cd->lpcd_first_idx + 1;
}
if (cd != NULL && cd->lpcd_last_idx)
last_index = cd->lpcd_last_idx;
else
last_index = LLOG_HDR_BITMAP_SIZE(llh) - 1;
while (rc == 0) {
struct llog_rec_hdr *rec;
off_t chunk_offset;
unsigned int buf_offset = 0;
bool partial_chunk;
/* skip records not set in bitmap */
while (index <= last_index &&
!ext2_test_bit(index, LLOG_HDR_BITMAP(llh)))
++index;
/* There are no indices prior the last_index */
if (index > last_index)
break;
CDEBUG(D_OTHER, "index: %d last_index %d\n", index,
last_index);
repeat:
/* get the buf with our target record; avoid old garbage */
memset(buf, 0, chunk_size);
rc = llog_next_block(lpi->lpi_env, loghandle, &saved_index,
index, &cur_offset, buf, chunk_size);
if (rc != 0)
GOTO(out, rc);
/* NB: after llog_next_block() call the cur_offset is the
* offset of the next block after read one.
* The absolute offset of the current chunk is calculated
* from cur_offset value and stored in chunk_offset variable.
*/
tmp_offset = cur_offset;
if (do_div(tmp_offset, chunk_size) != 0) {
partial_chunk = true;
chunk_offset = cur_offset & ~(chunk_size - 1);
} else {
partial_chunk = false;
chunk_offset = cur_offset - chunk_size;
}
/* NB: when rec->lrh_len is accessed it is already swabbed
* since it is used at the "end" of the loop and the rec
* swabbing is done at the beginning of the loop. */
for (rec = (struct llog_rec_hdr *)(buf + buf_offset);
(char *)rec < buf + chunk_size;
rec = llog_rec_hdr_next(rec)) {
CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n",
rec, rec->lrh_type);
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
rec->lrh_type, rec->lrh_index);
/* for partial chunk the end of it is zeroed, check
* for index 0 to distinguish it. */
if (partial_chunk && rec->lrh_index == 0) {
/* concurrent llog_add() might add new records
* while llog_processing, check this is not
* the case and re-read the current chunk
* otherwise. */
if (index > loghandle->lgh_last_idx)
GOTO(out, rc = 0);
CDEBUG(D_OTHER, "Re-read last llog buffer for "
"new records, index %u, last %u\n",
index, loghandle->lgh_last_idx);
/* save offset inside buffer for the re-read */
buf_offset = (char *)rec - (char *)buf;
cur_offset = chunk_offset;
goto repeat;
}
if (rec->lrh_len == 0 || rec->lrh_len > chunk_size) {
CWARN("invalid length %d in llog record for "
"index %d/%d\n", rec->lrh_len,
rec->lrh_index, index);
GOTO(out, rc = -EINVAL);
}
if (rec->lrh_index < index) {
CDEBUG(D_OTHER, "skipping lrh_index %d\n",
rec->lrh_index);
continue;
}
if (rec->lrh_index != index) {
CERROR("%s: Invalid record: index %u but "
"expected %u\n",
loghandle->lgh_ctxt->loc_obd->obd_name,
rec->lrh_index, index);
GOTO(out, rc = -ERANGE);
}
CDEBUG(D_OTHER,
"lrh_index: %d lrh_len: %d (%d remains)\n",
rec->lrh_index, rec->lrh_len,
(int)(buf + chunk_size - (char *)rec));
loghandle->lgh_cur_idx = rec->lrh_index;
loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
chunk_offset;
/* if set, process the callback on this record */
if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) {
rc = lpi->lpi_cb(lpi->lpi_env, loghandle, rec,
lpi->lpi_cbdata);
last_called_index = index;
if (rc == LLOG_PROC_BREAK) {
GOTO(out, rc);
} else if (rc == LLOG_DEL_RECORD) {
rc = llog_cancel_rec(lpi->lpi_env,
loghandle,
rec->lrh_index);
}
if (rc)
GOTO(out, rc);
}
/* exit if the last index is reached */
if (index >= last_index)
GOTO(out, rc = 0);
++index;
}
}
out:
if (cd != NULL)
cd->lpcd_last_idx = last_called_index;
if (unlikely(rc == -EIO && loghandle->lgh_obj != NULL)) {
if (dt_object_remote(loghandle->lgh_obj)) {
/* If it is remote object, then -EIO might means
* disconnection or eviction, let's return -EAGAIN,
* so for update recovery log processing, it will
* retry until the umount or abort recovery, see
* lod_sub_recovery_thread() */
CERROR("%s retry remote llog process\n",
loghandle->lgh_ctxt->loc_obd->obd_name);
rc = -EAGAIN;
} else {
/* something bad happened to the processing of a local
* llog file, probably I/O error or the log got
* corrupted to be able to finally release the log we
* discard any remaining bits in the header */
CERROR("Local llog found corrupted\n");
while (index <= last_index) {
if (ext2_test_bit(index,
LLOG_HDR_BITMAP(llh)) != 0)
llog_cancel_rec(lpi->lpi_env, loghandle,
index);
index++;
}
rc = 0;
}
}
OBD_FREE_LARGE(buf, chunk_size);
lpi->lpi_rc = rc;
return 0;
}
/**
* Implementation of the llog_operations::lop_next_block
*
* This function finds the the next llog block to return which contains
* record with required index. It is main part of llog processing.
*
* \param[in] env execution environment
* \param[in] loghandle llog handle of the current llog
* \param[in,out] cur_idx index preceeding cur_offset
* \param[in] next_idx target index to find
* \param[in,out] cur_offset furtherst point read in the file
* \param[in] buf pointer to data buffer to fill
* \param[in] len required len to read, it is
* LLOG_CHUNK_SIZE usually.
*
* \retval 0 on successful buffer read
* \retval negative value on error
*/
static int llog_osd_next_block(const struct lu_env *env,
struct llog_handle *loghandle, int *cur_idx,
int next_idx, __u64 *cur_offset, void *buf,
int len)
{
struct llog_thread_info *lgi = llog_info(env);
struct dt_object *o;
struct dt_device *dt;
int rc;
ENTRY;
LASSERT(env);
LASSERT(lgi);
if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
RETURN(-EINVAL);
CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
next_idx, *cur_idx, *cur_offset);
LASSERT(loghandle);
LASSERT(loghandle->lgh_ctxt);
o = loghandle->lgh_obj;
LASSERT(o);
LASSERT(dt_object_exists(o));
dt = lu2dt_dev(o->do_lu.lo_dev);
LASSERT(dt);
rc = dt_attr_get(env, o, &lgi->lgi_attr, BYPASS_CAPA);
if (rc)
GOTO(out, rc);
while (*cur_offset < lgi->lgi_attr.la_size) {
struct llog_rec_hdr *rec, *last_rec;
struct llog_rec_tail *tail;
llog_skip_over(cur_offset, *cur_idx, next_idx);
/* read up to next LLOG_CHUNK_SIZE block */
lgi->lgi_buf.lb_len = LLOG_CHUNK_SIZE -
(*cur_offset & (LLOG_CHUNK_SIZE - 1));
lgi->lgi_buf.lb_buf = buf;
rc = dt_read(env, o, &lgi->lgi_buf, cur_offset);
if (rc < 0) {
CERROR("%s: can't read llog block from log "DFID
" offset "LPU64": rc = %d\n",
o->do_lu.lo_dev->ld_obd->obd_name,
PFID(lu_object_fid(&o->do_lu)), *cur_offset,
rc);
GOTO(out, rc);
}
if (rc < len) {
/* signal the end of the valid buffer to
* llog_process */
memset(buf + rc, 0, len - rc);
}
if (rc == 0) /* end of file, nothing to do */
GOTO(out, rc);
if (rc < sizeof(*tail)) {
CERROR("%s: invalid llog block at log id "DOSTID"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, *cur_offset);
GOTO(out, rc = -EINVAL);
}
rec = buf;
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
tail = (struct llog_rec_tail *)((char *)buf + rc -
sizeof(struct llog_rec_tail));
/* get the last record in block */
last_rec = (struct llog_rec_hdr *)((char *)buf + rc -
tail->lrt_len);
if (LLOG_REC_HDR_NEEDS_SWABBING(last_rec))
lustre_swab_llog_rec(last_rec);
LASSERT(last_rec->lrh_index == tail->lrt_index);
*cur_idx = tail->lrt_index;
/* this shouldn't happen */
if (tail->lrt_index == 0) {
CERROR("%s: invalid llog tail at log id "DOSTID"/%u "
"offset "LPU64"\n",
o->do_lu.lo_dev->ld_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, *cur_offset);
GOTO(out, rc = -EINVAL);
}
if (tail->lrt_index < next_idx)
continue;
/* sanity check that the start of the new buffer is no farther
* than the record that we wanted. This shouldn't happen. */
if (rec->lrh_index > next_idx) {
CERROR("%s: missed desired record? %u > %u\n",
o->do_lu.lo_dev->ld_obd->obd_name,
rec->lrh_index, next_idx);
GOTO(out, rc = -ENOENT);
}
/* Trim unsupported extensions for compat w/ older clients */
if (!(loghandle->lgh_hdr->llh_flags & LLOG_F_EXT_JOBID))
changelog_block_trim_ext(rec, last_rec,
CLF_VERSION | CLF_RENAME);
GOTO(out, rc = 0);
}
GOTO(out, rc = -EIO);
out:
return rc;
}
static int llog_process_thread(void *arg)
{
struct llog_process_info *lpi = arg;
struct llog_handle *loghandle = lpi->lpi_loghandle;
struct llog_log_hdr *llh = loghandle->lgh_hdr;
struct llog_process_cat_data *cd = lpi->lpi_catdata;
char *buf;
u64 cur_offset, tmp_offset;
int chunk_size;
int rc = 0, index = 1, last_index;
int saved_index = 0;
int last_called_index = 0;
if (!llh)
return -EINVAL;
cur_offset = llh->llh_hdr.lrh_len;
chunk_size = llh->llh_hdr.lrh_len;
/* expect chunk_size to be power of two */
LASSERT(is_power_of_2(chunk_size));
buf = libcfs_kvzalloc(chunk_size, GFP_NOFS);
if (!buf) {
lpi->lpi_rc = -ENOMEM;
return 0;
}
if (cd) {
last_called_index = cd->lpcd_first_idx;
index = cd->lpcd_first_idx + 1;
}
if (cd && cd->lpcd_last_idx)
last_index = cd->lpcd_last_idx;
else
last_index = LLOG_HDR_BITMAP_SIZE(llh) - 1;
while (rc == 0) {
unsigned int buf_offset = 0;
struct llog_rec_hdr *rec;
bool partial_chunk;
off_t chunk_offset;
/* skip records not set in bitmap */
while (index <= last_index &&
!ext2_test_bit(index, LLOG_HDR_BITMAP(llh)))
++index;
if (index > last_index)
break;
CDEBUG(D_OTHER, "index: %d last_index %d\n",
index, last_index);
repeat:
/* get the buf with our target record; avoid old garbage */
memset(buf, 0, chunk_size);
rc = llog_next_block(lpi->lpi_env, loghandle, &saved_index,
index, &cur_offset, buf, chunk_size);
if (rc)
goto out;
/*
* NB: after llog_next_block() call the cur_offset is the
* offset of the next block after read one.
* The absolute offset of the current chunk is calculated
* from cur_offset value and stored in chunk_offset variable.
*/
tmp_offset = cur_offset;
if (do_div(tmp_offset, chunk_size)) {
partial_chunk = true;
chunk_offset = cur_offset & ~(chunk_size - 1);
} else {
partial_chunk = false;
chunk_offset = cur_offset - chunk_size;
}
/* NB: when rec->lrh_len is accessed it is already swabbed
* since it is used at the "end" of the loop and the rec
* swabbing is done at the beginning of the loop.
*/
for (rec = (struct llog_rec_hdr *)(buf + buf_offset);
(char *)rec < buf + chunk_size;
rec = llog_rec_hdr_next(rec)) {
CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n",
rec, rec->lrh_type);
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
rec->lrh_type, rec->lrh_index);
/*
* for partial chunk the end of it is zeroed, check
* for index 0 to distinguish it.
*/
if (partial_chunk && !rec->lrh_index) {
/* concurrent llog_add() might add new records
* while llog_processing, check this is not
* the case and re-read the current chunk
* otherwise.
*/
if (index > loghandle->lgh_last_idx) {
rc = 0;
goto out;
}
CDEBUG(D_OTHER, "Re-read last llog buffer for new records, index %u, last %u\n",
index, loghandle->lgh_last_idx);
/* save offset inside buffer for the re-read */
buf_offset = (char *)rec - (char *)buf;
cur_offset = chunk_offset;
goto repeat;
}
if (!rec->lrh_len || rec->lrh_len > chunk_size) {
CWARN("invalid length %d in llog record for index %d/%d\n",
rec->lrh_len,
rec->lrh_index, index);
rc = -EINVAL;
goto out;
}
if (rec->lrh_index < index) {
CDEBUG(D_OTHER, "skipping lrh_index %d\n",
rec->lrh_index);
continue;
}
if (rec->lrh_index != index) {
CERROR("%s: Invalid record: index %u but expected %u\n",
loghandle->lgh_ctxt->loc_obd->obd_name,
rec->lrh_index, index);
rc = -ERANGE;
goto out;
}
CDEBUG(D_OTHER,
"lrh_index: %d lrh_len: %d (%d remains)\n",
rec->lrh_index, rec->lrh_len,
(int)(buf + chunk_size - (char *)rec));
loghandle->lgh_cur_idx = rec->lrh_index;
loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
chunk_offset;
/* if set, process the callback on this record */
if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) {
rc = lpi->lpi_cb(lpi->lpi_env, loghandle, rec,
lpi->lpi_cbdata);
last_called_index = index;
if (rc)
goto out;
}
/* exit if the last index is reached */
if (index >= last_index) {
rc = 0;
goto out;
}
index++;
}
}
out:
if (cd)
cd->lpcd_last_idx = last_called_index;
kfree(buf);
lpi->lpi_rc = rc;
return 0;
}
/* sets:
* - cur_offset to the furthest point read in the log file
* - cur_idx to the log index preceeding cur_offset
* returns -EIO/-EINVAL on error
*/
static int llog_lvfs_next_block(struct llog_handle *loghandle, int *cur_idx,
int next_idx, __u64 *cur_offset, void *buf,
int len)
{
int rc;
ENTRY;
if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
RETURN(-EINVAL);
CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
next_idx, *cur_idx, *cur_offset);
while (*cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
struct llog_rec_hdr *rec;
struct llog_rec_tail *tail;
loff_t ppos;
llog_skip_over(cur_offset, *cur_idx, next_idx);
ppos = *cur_offset;
rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
loghandle->lgh_file, buf, len,
&ppos);
if (rc) {
CERROR("Cant read llog block at log id "LPU64
"/%u offset "LPU64"\n",
loghandle->lgh_id.lgl_oid,
loghandle->lgh_id.lgl_ogen,
*cur_offset);
RETURN(rc);
}
/* put number of bytes read into rc to make code simpler */
rc = ppos - *cur_offset;
*cur_offset = ppos;
if (rc < len) {
/* signal the end of the valid buffer to llog_process */
memset(buf + rc, 0, len - rc);
}
if (rc == 0) /* end of file, nothing to do */
RETURN(0);
if (rc < sizeof(*tail)) {
CERROR("Invalid llog block at log id "LPU64"/%u offset "
LPU64"\n", loghandle->lgh_id.lgl_oid,
loghandle->lgh_id.lgl_ogen, *cur_offset);
RETURN(-EINVAL);
}
rec = buf;
tail = (struct llog_rec_tail *)((char *)buf + rc -
sizeof(struct llog_rec_tail));
if (LLOG_REC_HDR_NEEDS_SWABBING(rec)) {
lustre_swab_llog_rec(rec, tail);
}
*cur_idx = tail->lrt_index;
/* this shouldn't happen */
if (tail->lrt_index == 0) {
CERROR("Invalid llog tail at log id "LPU64"/%u offset "
LPU64"\n", loghandle->lgh_id.lgl_oid,
loghandle->lgh_id.lgl_ogen, *cur_offset);
RETURN(-EINVAL);
}
if (tail->lrt_index < next_idx)
continue;
/* sanity check that the start of the new buffer is no farther
* than the record that we wanted. This shouldn't happen. */
if (rec->lrh_index > next_idx) {
CERROR("missed desired record? %u > %u\n",
rec->lrh_index, next_idx);
RETURN(-ENOENT);
}
RETURN(0);
}
RETURN(-EIO);
}
static int llog_lvfs_prev_block(const struct lu_env *env,
struct llog_handle *loghandle,
int prev_idx, void *buf, int len)
{
__u64 cur_offset;
int rc;
if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
return -EINVAL;
CDEBUG(D_OTHER, "looking for log index %u\n", prev_idx);
cur_offset = LLOG_CHUNK_SIZE;
llog_skip_over(&cur_offset, 0, prev_idx);
while (cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
struct llog_rec_hdr *rec, *last_rec;
struct llog_rec_tail *tail;
loff_t ppos = cur_offset;
rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
loghandle->lgh_file, buf, len,
&cur_offset);
if (rc < 0) {
CERROR("Cant read llog block at log id "DOSTID
"/%u offset %llu\n",
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen,
cur_offset);
return rc;
}
/* put number of bytes read into rc to make code simpler */
rc = cur_offset - ppos;
if (rc == 0) /* end of file, nothing to do */
return 0;
if (rc < sizeof(*tail)) {
CERROR("Invalid llog block at log id "DOSTID"/%u offset%llu\n",
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, cur_offset);
return -EINVAL;
}
rec = buf;
if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
lustre_swab_llog_rec(rec);
tail = (struct llog_rec_tail *)(buf + rc -
sizeof(struct llog_rec_tail));
/* get the last record in block */
last_rec = (struct llog_rec_hdr *)(buf + rc -
le32_to_cpu(tail->lrt_len));
if (LLOG_REC_HDR_NEEDS_SWABBING(last_rec))
lustre_swab_llog_rec(last_rec);
LASSERT(last_rec->lrh_index == tail->lrt_index);
/* this shouldn't happen */
if (tail->lrt_index == 0) {
CERROR("Invalid llog tail at log id "DOSTID"/%u offset%llu\n",
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, cur_offset);
return -EINVAL;
}
if (tail->lrt_index < prev_idx)
continue;
/* sanity check that the start of the new buffer is no farther
* than the record that we wanted. This shouldn't happen. */
if (rec->lrh_index > prev_idx) {
CERROR("missed desired record? %u > %u\n",
rec->lrh_index, prev_idx);
return -ENOENT;
}
return 0;
}
return -EIO;
}
