* GPL HEADER END
*/
/*
- * Copyright (c) 2013, 2014, Intel Corporation.
+ * Copyright (c) 2013, 2015, Intel Corporation.
*/
/*
* lustre/lfsck/lfsck_engine.c
iops->put(env, di);
}
-static int lfsck_update_lma(const struct lu_env *env,
- struct lfsck_instance *lfsck, struct dt_object *obj)
-{
- struct lfsck_thread_info *info = lfsck_env_info(env);
- struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
- struct dt_device *dev = lfsck_obj2dev(obj);
- struct lustre_mdt_attrs *lma = &info->lti_lma;
- struct lu_buf *buf;
- struct thandle *th;
- int fl;
- int rc;
- ENTRY;
-
- if (bk->lb_param & LPF_DRYRUN)
- RETURN(0);
-
- buf = lfsck_buf_get(env, info->lti_lma_old, LMA_OLD_SIZE);
- rc = dt_xattr_get(env, obj, buf, XATTR_NAME_LMA);
- if (rc < 0) {
- if (rc != -ENODATA)
- RETURN(rc);
-
- fl = LU_XATTR_CREATE;
- lustre_lma_init(lma, lfsck_dto2fid(obj), LMAC_FID_ON_OST, 0);
- } else {
- if (rc != LMA_OLD_SIZE && rc != sizeof(struct lustre_mdt_attrs))
- RETURN(-EINVAL);
-
- fl = LU_XATTR_REPLACE;
- lustre_lma_swab(lma);
- lustre_lma_init(lma, lfsck_dto2fid(obj),
- lma->lma_compat | LMAC_FID_ON_OST,
- lma->lma_incompat);
- }
- lustre_lma_swab(lma);
-
- th = dt_trans_create(env, dev);
- if (IS_ERR(th))
- RETURN(PTR_ERR(th));
-
- buf = lfsck_buf_get(env, lma, sizeof(*lma));
- rc = dt_declare_xattr_set(env, obj, buf, XATTR_NAME_LMA, fl, th);
- if (rc != 0)
- GOTO(stop, rc);
-
- rc = dt_trans_start_local(env, dev, th);
- if (rc != 0)
- GOTO(stop, rc);
-
- rc = dt_xattr_set(env, obj, buf, XATTR_NAME_LMA, fl, th);
-
- GOTO(stop, rc);
-
-stop:
- dt_trans_stop(env, dev, th);
- return rc;
-}
-
static int lfsck_parent_fid(const struct lu_env *env, struct dt_object *obj,
struct lu_fid *fid)
{
if (CFS_FAIL_TIMEOUT(OBD_FAIL_LFSCK_DELAY2, cfs_fail_val) &&
unlikely(!thread_is_running(thread))) {
CDEBUG(D_LFSCK, "%s: scan dir exit for engine stop, "
- "parent "DFID", cookie "LPX64"\n",
+ "parent "DFID", cookie %#llx\n",
lfsck_lfsck2name(lfsck),
PFID(lfsck_dto2fid(dir)), lfsck->li_cookie_dir);
if (rc != 0) {
CDEBUG(D_LFSCK, "%s: scan dir failed at rec(), "
- "parent "DFID", cookie "LPX64": rc = %d\n",
+ "parent "DFID", cookie %#llx: rc = %d\n",
lfsck_lfsck2name(lfsck),
PFID(lfsck_dto2fid(dir)),
lfsck->li_cookie_dir, rc);
goto checkpoint;
}
- if (ent->lde_attrs & LUDA_IGNORE &&
- strcmp(ent->lde_name, dotdot) != 0)
+ if (ent->lde_attrs & LUDA_IGNORE)
goto checkpoint;
/* skip dot entry. */
lfsck_control_speed(lfsck);
if (unlikely(!thread_is_running(thread))) {
CDEBUG(D_LFSCK, "%s: scan dir exit for engine stop, "
- "parent "DFID", cookie "LPX64"\n",
+ "parent "DFID", cookie %#llx\n",
lfsck_lfsck2name(lfsck),
PFID(lfsck_dto2fid(dir)),
lfsck->li_cookie_dir);
}
rc = iops->next(env, di);
+ if (rc < 0)
+ CDEBUG(D_LFSCK, "%s dir engine fail to locate next "
+ "for the directory "DFID": rc = %d\n",
+ lfsck_lfsck2name(lfsck),
+ PFID(&lfsck->li_pos_current.lp_dir_parent), rc);
} while (rc == 0);
if (rc > 0 && !lfsck->li_oit_over)
do {
struct dt_object *target;
- bool update_lma = false;
if (lfsck->li_di_dir != NULL) {
rc = lfsck_master_dir_engine(env, lfsck);
if (CFS_FAIL_TIMEOUT(OBD_FAIL_LFSCK_DELAY1, cfs_fail_val) &&
unlikely(!thread_is_running(thread))) {
CDEBUG(D_LFSCK, "%s: OIT scan exit for engine stop, "
- "cookie "LPU64"\n",
+ "cookie %llu\n",
lfsck_lfsck2name(lfsck), iops->store(env, di));
RETURN(0);
LASSERT(!lfsck->li_master);
- /* It is an old format device, update the LMA. */
if (idx != idx1) {
struct ost_id *oi = &info->lti_oi;
+ if (unlikely(idx1 != 0)) {
+ CDEBUG(D_LFSCK, "%s: invalid IDIF "DFID
+ ", not match device index %u\n",
+ lfsck_lfsck2name(lfsck),
+ PFID(fid), idx);
+
+ goto checkpoint;
+ }
+
+ /* rebuild the IDIF with index to
+ * avoid double instances for the
+ * same object. */
fid_to_ostid(fid, oi);
ostid_to_fid(fid, oi, idx);
- update_lma = true;
}
} else if (!fid_is_norm(fid) && !fid_is_igif(fid) &&
!fid_is_last_id(fid) &&
target = lfsck_object_find_bottom(env, lfsck, fid);
if (IS_ERR(target)) {
CDEBUG(D_LFSCK, "%s: OIT scan failed at find target "
- DFID", cookie "LPU64": rc = %d\n",
+ DFID", cookie %llu: rc = %d\n",
lfsck_lfsck2name(lfsck), PFID(fid),
iops->store(env, di), rc);
lfsck_fail(env, lfsck, true);
goto checkpoint;
}
- if (dt_object_exists(target)) {
- if (update_lma) {
- rc = lfsck_update_lma(env, lfsck, target);
- if (rc != 0)
- CDEBUG(D_LFSCK, "%s: fail to update "
- "LMA for "DFID": rc = %d\n",
- lfsck_lfsck2name(lfsck),
- PFID(lfsck_dto2fid(target)), rc);
- }
- if (rc == 0)
- rc = lfsck_exec_oit(env, lfsck, target);
- }
+ if (dt_object_exists(target))
+ rc = lfsck_exec_oit(env, lfsck, target);
+
lfsck_object_put(env, target);
if (rc != 0 && bk->lb_param & LPF_FAILOUT)
RETURN(rc);
lfsck->li_oit_over = 1;
else if (likely(rc == 0))
lfsck->li_current_oit_processed = 0;
+ else
+ CDEBUG(D_LFSCK, "%s oit engine fail to locate next at "
+ "%llu: rc = %d\n", lfsck_lfsck2name(lfsck),
+ iops->store(env, di), rc);
if (unlikely(!thread_is_running(thread))) {
CDEBUG(D_LFSCK, "%s: OIT scan exit for engine stop, "
- "cookie "LPU64"\n", lfsck_lfsck2name(lfsck),
+ "cookie %llu\n", lfsck_lfsck2name(lfsck),
iops->store(env, di));
RETURN(0);
}
int rc;
ENTRY;
+ /* There will be some objects verification during the LFSCK start,
+ * such as the subsequent lfsck_verify_lpf(). Trigger low layer OI
+ * OI scrub before that to handle the potential inconsistence. */
+ oit_di = oit_iops->init(env, oit_obj, lfsck->li_args_oit);
+ if (IS_ERR(oit_di)) {
+ rc = PTR_ERR(oit_di);
+ CDEBUG(D_LFSCK, "%s: master engine fail to init iteration: "
+ "rc = %d\n", lfsck_lfsck2name(lfsck), rc);
+
+ GOTO(fini_args, rc);
+ }
+
if (lfsck->li_master &&
(!list_empty(&lfsck->li_list_scan) ||
!list_empty(&lfsck->li_list_double_scan))) {
lfsck_lfsck2name(lfsck), rc);
}
- oit_di = oit_iops->init(env, oit_obj, lfsck->li_args_oit);
- if (IS_ERR(oit_di)) {
- rc = PTR_ERR(oit_di);
- CDEBUG(D_LFSCK, "%s: master engine fail to init iteration: "
- "rc = %d\n", lfsck_lfsck2name(lfsck), rc);
-
- GOTO(fini_args, rc);
- }
-
spin_lock(&lfsck->li_lock);
lfsck->li_di_oit = oit_di;
spin_unlock(&lfsck->li_lock);
GOTO(fini_oit, rc);
CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = %#x, dir_flags = %#x, "
- "oit_cookie = "LPU64", dir_cookie = "LPX64", parent = "DFID
+ "oit_cookie = %llu, dir_cookie = %#llx, parent = "DFID
", pid = %d\n", lfsck->li_args_oit, lfsck->li_args_dir,
lfsck->li_pos_checkpoint.lp_oit_cookie,
lfsck->li_pos_checkpoint.lp_dir_cookie,
lfsck_pos_fill(env, lfsck, &lfsck->li_pos_checkpoint, false);
CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = %#x, dir_flags = %#x, "
- "oit_cookie = "LPU64", dir_cookie = "LPX64", parent = "DFID
+ "oit_cookie = %llu, dir_cookie = %#llx, parent = "DFID
", pid = %d, rc = %d\n", lfsck->li_args_oit, lfsck->li_args_dir,
lfsck->li_pos_checkpoint.lp_oit_cookie,
lfsck->li_pos_checkpoint.lp_dir_cookie,
memset(lr, 0, sizeof(*lr));
lr->lr_event = LE_QUERY;
lr->lr_active = com->lc_type;
+
+ memset(laia, 0, sizeof(*laia));
laia->laia_com = com;
laia->laia_lr = lr;
- laia->laia_shared = 0;
if (!list_empty(&lad->lad_mdt_phase1_list)) {
ltds = &lfsck->li_mdt_descs;
lr->lr_index = lfsck_dev_idx(lfsck);
lr->lr_active = com->lc_type;
+
+ memset(laia, 0, sizeof(*laia));
laia->laia_com = com;
laia->laia_lr = lr;
- laia->laia_shared = 0;
switch (lr->lr_event) {
case LE_START:
LASSERT(ltd != NULL);
laia->laia_ltd = ltd;
- ltd->ltd_layout_done = 0;
- ltd->ltd_synced_failures = 0;
rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
lfsck_async_interpret_common,
laia, LFSCK_NOTIFY);
spin_lock(<ds->ltd_lock);
if (com->lc_type == LFSCK_TYPE_LAYOUT) {
cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
- ltd = LTD_TGT(ltds, idx);
+ ltd = lfsck_ltd2tgt(ltds, idx);
LASSERT(ltd != NULL);
if (!list_empty(<d->ltd_layout_list))
}
} else {
cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
- ltd = LTD_TGT(ltds, idx);
+ ltd = lfsck_ltd2tgt(ltds, idx);
LASSERT(ltd != NULL);
if (!list_empty(<d->ltd_namespace_list))
}
spin_lock(&lad->lad_lock);
+ lad->lad_task = current;
thread_set_flags(athread, SVC_RUNNING);
spin_unlock(&lad->lad_lock);
wake_up_all(&mthread->t_ctl_waitq);
if (unlikely(lad->lad_exit ||
!thread_is_running(mthread)))
- GOTO(cleanup1, rc = lad->lad_post_result);
+ GOTO(cleanup, rc = lad->lad_post_result);
lar = list_entry(lad->lad_req_list.next,
struct lfsck_assistant_req,
lao->la_req_fini(env, lar);
if (rc < 0 && bk->lb_param & LPF_FAILOUT)
- GOTO(cleanup1, rc);
+ GOTO(cleanup, rc);
}
l_wait_event(athread->t_ctl_waitq,
&lwi);
if (unlikely(lad->lad_exit))
- GOTO(cleanup1, rc = lad->lad_post_result);
+ GOTO(cleanup, rc = lad->lad_post_result);
if (!list_empty(&lad->lad_req_list))
continue;
lfsck_lfsck2name(lfsck), lad->lad_name);
if (unlikely(lad->lad_exit))
- GOTO(cleanup1, rc = lad->lad_post_result);
+ GOTO(cleanup, rc = lad->lad_post_result);
lad->lad_to_post = 0;
LASSERT(lad->lad_post_result > 0);
lfsck_lfsck2name(lfsck), rc2);
if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_DOUBLESCAN))
- GOTO(cleanup2, rc = 0);
+ GOTO(cleanup, rc = 0);
while (lad->lad_in_double_scan) {
rc = lfsck_assistant_query_others(env, com);
goto p2_next;
if (rc < 0)
- GOTO(cleanup2, rc);
+ GOTO(cleanup, rc);
/* Pull LFSCK status on related targets once
* per 30 seconds if we are not notified. */
if (unlikely(lad->lad_exit ||
!thread_is_running(mthread)))
- GOTO(cleanup2, rc = 0);
+ GOTO(cleanup, rc = 0);
if (rc == -ETIMEDOUT)
continue;
if (rc < 0)
- GOTO(cleanup2, rc);
+ GOTO(cleanup, rc);
p2_next:
rc = lao->la_handler_p2(env, com);
if (rc != 0)
- GOTO(cleanup2, rc);
+ GOTO(cleanup, rc);
if (unlikely(lad->lad_exit ||
!thread_is_running(mthread)))
- GOTO(cleanup2, rc = 0);
+ GOTO(cleanup, rc = 0);
}
}
}
-cleanup1:
+cleanup:
/* Cleanup the unfinished requests. */
spin_lock(&lad->lad_lock);
if (rc < 0)
if (lad->lad_exit && lad->lad_post_result <= 0)
lao->la_fill_pos(env, com, &lfsck->li_pos_checkpoint);
+ thread_set_flags(athread, SVC_STOPPING);
while (!list_empty(&lad->lad_req_list)) {
lar = list_entry(lad->lad_req_list.next,
struct lfsck_assistant_req,
LASSERTF(lad->lad_prefetched == 0, "unmatched prefeteched objs %d\n",
lad->lad_prefetched);
-cleanup2:
memset(lr, 0, sizeof(*lr));
if (rc > 0) {
lr->lr_event = LE_PHASE2_DONE;
lad->lad_assistant_status = (rc1 != 0 ? rc1 : rc);
thread_set_flags(athread, SVC_STOPPED);
wake_up_all(&mthread->t_ctl_waitq);
+ lad->lad_task = NULL;
spin_unlock(&lad->lad_lock);
CDEBUG(D_LFSCK, "%s: %s LFSCK assistant thread exit: rc = %d\n",