int osd_oi_lookup(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, struct osd_inode_id *id,
enum oi_check_flags flags)
{
if (unlikely(fid_is_last_id(fid)))
return osd_obj_spec_lookup(info, osd, fid, id);
if (fid_is_on_ost(info, osd, fid, flags) || fid_is_llog(fid))
return osd_obj_map_lookup(info, osd, fid, id);
if (unlikely(fid_seq(fid) == FID_SEQ_LOCAL_FILE)) {
int rc;
if (fid_is_fs_root(fid)) {
osd_id_gen(id, osd_sb(osd)->s_root->d_inode->i_ino,
osd_sb(osd)->s_root->d_inode->i_generation);
return 0;
}
if (unlikely(fid_is_acct(fid)))
return osd_acct_obj_lookup(info, osd, fid, id);
/* For other special FIDs, try OI first, then do spec lookup */
rc = __osd_oi_lookup(info, osd, fid, id);
if (rc == -ENOENT)
return osd_obj_spec_lookup(info, osd, fid, id);
return rc;
}
if (!osd->od_igif_inoi && fid_is_igif(fid)) {
osd_id_gen(id, lu_igif_ino(fid), lu_igif_gen(fid));
return 0;
}
return __osd_oi_lookup(info, osd, fid, id);
}
int fid_is_on_ost(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, enum oi_check_flags flags)
{
struct lu_seq_range *range = &info->oti_seq_range;
int rc;
ENTRY;
if (flags & OI_KNOWN_ON_OST)
RETURN(1);
if (unlikely(fid_is_local_file(fid) || fid_is_igif(fid) ||
fid_is_llog(fid)) || fid_is_name_llog(fid) ||
fid_is_quota(fid))
RETURN(0);
if (fid_is_idif(fid) || fid_is_last_id(fid))
RETURN(1);
if (!(flags & OI_CHECK_FLD))
RETURN(0);
rc = osd_fld_lookup(info->oti_env, osd, fid_seq(fid), range);
if (rc != 0) {
if (rc != -ENOENT)
CERROR("%s: lookup FLD "DFID": rc = %d\n",
osd_name(osd), PFID(fid), rc);
RETURN(0);
}
if (fld_range_is_ost(range))
RETURN(1);
RETURN(0);
}
int osd_oi_lookup(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, struct osd_inode_id *id,
enum oi_check_flags flags)
{
if (unlikely(fid_is_last_id(fid)))
return osd_obj_spec_lookup(info, osd, fid, id);
if (fid_is_on_ost(info, osd, fid, flags) || fid_is_llog(fid))
return osd_obj_map_lookup(info, osd, fid, id);
if (fid_is_fs_root(fid)) {
osd_id_gen(id, osd_sb(osd)->s_root->d_inode->i_ino,
osd_sb(osd)->s_root->d_inode->i_generation);
return 0;
}
if (unlikely(fid_is_acct(fid)))
return osd_acct_obj_lookup(info, osd, fid, id);
if (!osd->od_igif_inoi && fid_is_igif(fid)) {
osd_id_gen(id, lu_igif_ino(fid), lu_igif_gen(fid));
return 0;
}
return __osd_oi_lookup(info, osd, fid, id);
}
/**
* Lookup method for "fid" object. Only filenames with correct SEQ:OID format
* are valid. We also check if object with passed fid exists or not.
*/
static int obf_lookup(const struct lu_env *env, struct md_object *p,
const struct lu_name *lname, struct lu_fid *f,
struct md_op_spec *spec)
{
char *name = (char *)lname->ln_name;
struct mdd_device *mdd = mdo2mdd(p);
struct mdd_object *child;
int rc = 0;
while (*name == '[')
name++;
sscanf(name, SFID, RFID(f));
if (!fid_is_sane(f)) {
CWARN("%s: Trying to lookup invalid FID [%s] in %s/%s, FID "
"format should be "DFID"\n", mdd2obd_dev(mdd)->obd_name,
lname->ln_name, dot_lustre_name, mdd_obf_dir_name,
(__u64)FID_SEQ_NORMAL, 1, 0);
GOTO(out, rc = -EINVAL);
}
if (!fid_is_norm(f) && !fid_is_igif(f) && !fid_is_root(f) &&
!fid_seq_is_dot(f->f_seq)) {
CWARN("%s: Trying to lookup invalid FID "DFID" in %s/%s, "
"sequence should be >= "LPX64" or within ["LPX64","
""LPX64"].\n", mdd2obd_dev(mdd)->obd_name, PFID(f),
dot_lustre_name, mdd_obf_dir_name, (__u64)FID_SEQ_NORMAL,
(__u64)FID_SEQ_IGIF, (__u64)FID_SEQ_IGIF_MAX);
GOTO(out, rc = -EINVAL);
}
/* Check if object with this fid exists */
child = mdd_object_find(env, mdd, f);
if (child == NULL)
GOTO(out, rc = 0);
if (IS_ERR(child))
GOTO(out, rc = PTR_ERR(child));
if (mdd_object_exists(child) == 0)
rc = -ENOENT;
mdd_object_put(env, child);
out:
return rc;
}
static int lfsck_master_oit_engine(const struct lu_env *env,
struct lfsck_instance *lfsck)
{
struct lfsck_thread_info *info = lfsck_env_info(env);
const struct dt_it_ops *iops =
&lfsck->li_obj_oit->do_index_ops->dio_it;
struct dt_it *di = lfsck->li_di_oit;
struct lu_fid *fid = &info->lti_fid;
struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
struct ptlrpc_thread *thread = &lfsck->li_thread;
__u32 idx =
lfsck_dev_idx(lfsck->li_bottom);
int rc;
ENTRY;
do {
struct dt_object *target;
bool update_lma = false;
if (lfsck->li_di_dir != NULL) {
rc = lfsck_master_dir_engine(env, lfsck);
if (rc <= 0)
RETURN(rc);
}
if (unlikely(lfsck->li_oit_over))
RETURN(1);
if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY1) &&
cfs_fail_val > 0) {
struct l_wait_info lwi;
lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
NULL, NULL);
l_wait_event(thread->t_ctl_waitq,
!thread_is_running(thread),
&lwi);
}
if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
RETURN(0);
lfsck->li_current_oit_processed = 1;
lfsck->li_new_scanned++;
rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
if (rc != 0) {
lfsck_fail(env, lfsck, true);
if (rc < 0 && bk->lb_param & LPF_FAILOUT)
RETURN(rc);
else
goto checkpoint;
}
if (fid_is_idif(fid)) {
__u32 idx1 = fid_idif_ost_idx(fid);
LASSERT(!lfsck->li_master);
/* It is an old format device, update the LMA. */
if (idx != idx1) {
struct ost_id *oi = &info->lti_oi;
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) && !fid_is_root(fid) &&
!fid_seq_is_dot(fid_seq(fid))) {
/* If the FID/object is only used locally and invisible
* to external nodes, then LFSCK will not handle it. */
goto checkpoint;
}
target = lfsck_object_find(env, lfsck, fid);
if (target == NULL) {
goto checkpoint;
} else if (IS_ERR(target)) {
lfsck_fail(env, lfsck, true);
if (bk->lb_param & LPF_FAILOUT)
RETURN(PTR_ERR(target));
else
goto checkpoint;
}
/* XXX: Currently, skip remote object, the consistency for
* remote object will be processed in LFSCK phase III. */
if (dt_object_exists(target) && !dt_object_remote(target)) {
if (update_lma)
rc = lfsck_update_lma(env, lfsck, target);
if (rc == 0)
rc = lfsck_exec_oit(env, lfsck, target);
}
lfsck_object_put(env, target);
if (rc != 0 && bk->lb_param & LPF_FAILOUT)
RETURN(rc);
checkpoint:
rc = lfsck_checkpoint(env, lfsck);
if (rc != 0 && bk->lb_param & LPF_FAILOUT)
RETURN(rc);
/* Rate control. */
lfsck_control_speed(lfsck);
if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL1)) {
spin_lock(&lfsck->li_lock);
thread_set_flags(thread, SVC_STOPPING);
spin_unlock(&lfsck->li_lock);
RETURN(-EINVAL);
}
rc = iops->next(env, di);
if (unlikely(rc > 0))
lfsck->li_oit_over = 1;
else if (likely(rc == 0))
lfsck->li_current_oit_processed = 0;
if (unlikely(!thread_is_running(thread)))
RETURN(0);
} while (rc == 0 || lfsck->li_di_dir != NULL);
RETURN(rc);
}
int mdd_compat_fixes(const struct lu_env *env, struct mdd_device *mdd)
{
struct mdd_thread_info *info = mdd_env_info(env);
struct mdd_object *root;
struct dt_object *o;
struct lustre_mdt_attrs *lma;
struct lu_buf buf;
int rc;
ENTRY;
/* IGIF FIDS are valid for old 1.8 and 2.[123] ROOT and are kept.
* Normal FIDs used by Xyratex 1.8->2.1 upgrade tool are also kept. */
if (fid_is_igif(&mdd->mdd_root_fid) || fid_is_norm(&mdd->mdd_root_fid))
RETURN(0);
/*
* FID is supposed to be FID_SEQ_ROOT for:
* - new ldiskfs fs
* - new ZFS fs
* - old ZFS fs, by now processed with osd_convert_root_to_new_seq()
*/
if (fid_seq(&mdd->mdd_root_fid) != FID_SEQ_ROOT) {
CERROR("%s: wrong FID "DFID" is used for /ROOT\n",
mdd2obd_dev(mdd)->obd_name,
PFID(&mdd->mdd_root_fid));
RETURN(-EINVAL);
}
root = mdd_object_find(env, mdd, &mdd->mdd_root_fid);
if (IS_ERR(root))
RETURN(PTR_ERR(root));
o = mdd_object_child(root);
CDEBUG(D_OTHER, "/ROOT = "DFID"\n", PFID(&mdd->mdd_root_fid));
if (dt_try_as_dir(env, o) == 0) {
CERROR("%s: not a directory\n", mdd2obd_dev(mdd)->obd_name);
GOTO(out, rc = -ENOTDIR);
}
lma = (struct lustre_mdt_attrs *)&info->mti_xattr_buf;
CLASSERT(sizeof(info->mti_xattr_buf) >= LMA_OLD_SIZE);
buf.lb_len = LMA_OLD_SIZE;
buf.lb_buf = lma;
rc = mdo_xattr_get(env, root, &buf, XATTR_NAME_LMA, BYPASS_CAPA);
if (rc < 0 && rc != -ENODATA) {
CERROR("%s: can't fetch LMA: rc = %d\n",
mdd2obd_dev(mdd)->obd_name, rc);
GOTO(out, rc);
}
lustre_lma_swab(lma);
if (lu_fid_eq(&lma->lma_self_fid, &mdd->mdd_root_fid)) {
/* /ROOT has been converted already
* or was correct from the beginning */
CDEBUG(D_OTHER, "%s: converted already\n",
mdd2obd_dev(mdd)->obd_name);
GOTO(out, rc = 0);
}
/* this is supposed to happen only on pre-production ZFS backend */
if (strcmp(mdd->mdd_bottom->dd_lu_dev.ld_type->ldt_name,
LUSTRE_OSD_ZFS_NAME) != 0) {
CERROR("%s: "DFID" is used on ldiskfs?!\n",
mdd2obd_dev(mdd)->obd_name, PFID(&mdd->mdd_root_fid));
GOTO(out, rc = -ENOTSUPP);
}
LCONSOLE_INFO("%s: FID of /ROOT has been changed. "
"Please remount the clients.\n",
mdd2obd_dev(mdd)->obd_name);
/* Fill FLDB first */
rc = mdd_fill_fldb(env, mdd);
if (rc)
GOTO(out, rc);
/* remove ./.. from /ROOT */
rc = mdd_convert_remove_dots(env, mdd, root);
if (rc)
GOTO(out, rc);
/* go over the directory, fix all the objects */
rc = mdd_fix_children(env, mdd, o);
if (rc)
GOTO(out, rc);
/* Update LMA on /ROOT. Done for simplicity in MDD, not in osd-zfs.
* Correct LMA will imply the whole directory has been coverted
* successfully, otherwise it will be retried on next mount. */
rc = mdd_convert_lma(env, mdd, root);
out:
mdd_object_put(env, root);
RETURN(rc);
}
