* GPL HEADER END
*/
/*
- * Copyright (c) 2013, 2014, Intel Corporation.
+ * Copyright (c) 2013, 2017, Intel Corporation.
*/
/*
* lustre/lfsck/lfsck_lib.c
#define DEBUG_SUBSYSTEM S_LFSCK
-#include <libcfs/list.h>
+#include <linux/kthread.h>
+#include <linux/sched.h>
+#include <linux/list.h>
#include <lu_object.h>
#include <dt_object.h>
#include <md_object.h>
#include <lustre_lib.h>
#include <lustre_net.h>
#include <lustre_lfsck.h>
-#include <lustre/lustre_lfsck_user.h>
#include "lfsck_internal.h"
static struct list_head lfsck_mdt_orphan_list;
static DEFINE_SPINLOCK(lfsck_instance_lock);
-static const char *lfsck_status_names[] = {
- [LS_INIT] = "init",
- [LS_SCANNING_PHASE1] = "scanning-phase1",
- [LS_SCANNING_PHASE2] = "scanning-phase2",
- [LS_COMPLETED] = "completed",
- [LS_FAILED] = "failed",
- [LS_STOPPED] = "stopped",
- [LS_PAUSED] = "paused",
- [LS_CRASHED] = "crashed",
- [LS_PARTIAL] = "partial",
- [LS_CO_FAILED] = "co-failed",
- [LS_CO_STOPPED] = "co-stopped",
- [LS_CO_PAUSED] = "co-paused"
-};
-
const char *lfsck_flags_names[] = {
"scanned-once",
"inconsistent",
"orphan",
"create_ostobj",
"create_mdtobj",
+ NULL,
+ "delay_create_ostobj",
NULL
};
LVLT_BY_NAMEENTRY = 1,
};
-const char *lfsck_status2names(enum lfsck_status status)
+static inline void
+lfsck_reset_ltd_status(struct lfsck_tgt_desc *ltd, enum lfsck_type type)
{
- if (unlikely(status < 0 || status >= LS_MAX))
- return "unknown";
-
- return lfsck_status_names[status];
+ if (type == LFSCK_TYPE_LAYOUT) {
+ ltd->ltd_layout_status = LS_MAX;
+ ltd->ltd_layout_repaired = 0;
+ } else {
+ ltd->ltd_namespace_status = LS_MAX;
+ ltd->ltd_namespace_repaired = 0;
+ }
}
static int lfsck_tgt_descs_init(struct lfsck_tgt_descs *ltds)
}
cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
- ltd = LTD_TGT(ltds, idx);
+ ltd = lfsck_ltd2tgt(ltds, idx);
if (likely(ltd != NULL)) {
LASSERT(list_empty(<d->ltd_layout_list));
LASSERT(list_empty(<d->ltd_layout_phase_list));
ltds->ltd_tgtnr--;
cfs_bitmap_clear(ltds->ltd_tgts_bitmap, idx);
- LTD_TGT(ltds, idx) = NULL;
+ lfsck_assign_tgt(ltds, NULL, idx);
lfsck_tgt_put(ltd);
}
}
if (index >= ltds->ltd_tgts_bitmap->size) {
__u32 newsize = max((__u32)ltds->ltd_tgts_bitmap->size,
(__u32)BITS_PER_LONG);
- cfs_bitmap_t *old_bitmap = ltds->ltd_tgts_bitmap;
- cfs_bitmap_t *new_bitmap;
+ struct cfs_bitmap *old_bitmap = ltds->ltd_tgts_bitmap;
+ struct cfs_bitmap *new_bitmap;
while (newsize < index + 1)
newsize <<= 1;
GOTO(unlock, rc = -ENOMEM);
}
- LTD_TGT(ltds, index) = ltd;
+ lfsck_assign_tgt(ltds, ltd, index);
cfs_bitmap_set(ltds->ltd_tgts_bitmap, index);
ltds->ltd_tgtnr++;
static int __lfsck_ibits_lock(const struct lu_env *env,
struct lfsck_instance *lfsck,
- struct dt_object *obj,
- struct ldlm_res_id *resid,
- struct lustre_handle *lh,
- __u64 bits, ldlm_mode_t mode)
+ struct dt_object *obj, struct ldlm_res_id *resid,
+ struct lustre_handle *lh, __u64 bits,
+ enum ldlm_mode mode)
{
struct lfsck_thread_info *info = lfsck_env_info(env);
- ldlm_policy_data_t *policy = &info->lti_policy;
+ union ldlm_policy_data *policy = &info->lti_policy;
__u64 flags = LDLM_FL_ATOMIC_CB;
int rc;
einfo->ei_res_id = resid;
rc = dt_object_lock(env, obj, lh, einfo, policy);
+ /* for regular checks LFSCK doesn't use LDLM locking,
+ * so the state isn't coherent. here we just took LDLM
+ * lock for coherency and it's time to invalidate
+ * previous state */
+ if (rc == ELDLM_OK)
+ dt_invalidate(env, obj);
} else {
- rc = ldlm_cli_enqueue_local(lfsck->li_namespace, resid,
+ rc = ldlm_cli_enqueue_local(env, lfsck->li_namespace, resid,
LDLM_IBITS, policy, mode,
&flags, ldlm_blocking_ast,
ldlm_completion_ast, NULL, NULL,
*/
int lfsck_ibits_lock(const struct lu_env *env, struct lfsck_instance *lfsck,
struct dt_object *obj, struct lustre_handle *lh,
- __u64 bits, ldlm_mode_t mode)
+ __u64 bits, enum ldlm_mode mode)
{
struct ldlm_res_id *resid = &lfsck_env_info(env)->lti_resid;
}
/**
+ * Request the remote LOOKUP lock for the given object.
+ *
+ * If \a pobj is remote, the LOOKUP lock of \a obj is on the MDT where
+ * \a pobj is, acquire LOOKUP lock there.
+ *
+ * \param[in] env pointer to the thread context
+ * \param[in] lfsck pointer to the lfsck instance
+ * \param[in] pobj pointer to parent dt_object
+ * \param[in] obj pointer to the dt_object to be locked
+ * \param[out] lh pointer to the lock handle
+ * \param[in] mode the mode for the ldlm lock to be acquired
+ *
+ * \retval 0 for success
+ * \retval negative error number on failure
+ */
+int lfsck_remote_lookup_lock(const struct lu_env *env,
+ struct lfsck_instance *lfsck,
+ struct dt_object *pobj, struct dt_object *obj,
+ struct lustre_handle *lh, enum ldlm_mode mode)
+{
+ struct ldlm_res_id *resid = &lfsck_env_info(env)->lti_resid;
+
+ LASSERT(!lustre_handle_is_used(lh));
+
+ fid_build_reg_res_name(lfsck_dto2fid(obj), resid);
+ return __lfsck_ibits_lock(env, lfsck, pobj, resid, lh,
+ MDS_INODELOCK_LOOKUP, mode);
+}
+
+/**
* Release the the specified ibits lock.
*
* If the lock has been acquired before, release it
* \param[in] lh pointer to the lock handle
* \param[in] mode the mode for the ldlm lock to be released
*/
-void lfsck_ibits_unlock(struct lustre_handle *lh, ldlm_mode_t mode)
+void lfsck_ibits_unlock(struct lustre_handle *lh, enum ldlm_mode mode)
{
if (lustre_handle_is_used(lh)) {
ldlm_lock_decref(lh, mode);
* \retval negative error number on failure
*/
int lfsck_lock(const struct lu_env *env, struct lfsck_instance *lfsck,
- struct dt_object *obj, const char *name,
- struct lfsck_lock_handle *llh, __u64 bits, ldlm_mode_t mode)
+ struct dt_object *obj, const char *name,
+ struct lfsck_lock_handle *llh, __u64 bits, enum ldlm_mode mode)
{
struct ldlm_res_id *resid = &lfsck_env_info(env)->lti_resid;
int rc;
LASSERT(S_ISDIR(lfsck_object_type(obj)));
LASSERT(name != NULL);
+ LASSERT(name[0] != 0);
LASSERT(!lustre_handle_is_used(&llh->llh_pdo_lh));
LASSERT(!lustre_handle_is_used(&llh->llh_reg_lh));
return rc;
llh->llh_reg_mode = mode;
- resid->name[LUSTRE_RES_ID_HSH_OFF] = full_name_hash(name, strlen(name));
+ resid->name[LUSTRE_RES_ID_HSH_OFF] = ll_full_name_hash(NULL, name,
+ strlen(name));
+ LASSERT(resid->name[LUSTRE_RES_ID_HSH_OFF] != 0);
rc = __lfsck_ibits_lock(env, lfsck, obj, resid, &llh->llh_reg_lh,
bits, llh->llh_reg_mode);
if (rc != 0)
struct lu_seq_range *range = &lfsck_env_info(env)->lti_range;
int rc;
+ if (unlikely(fid_seq(fid) == FID_SEQ_LOCAL_FILE)) {
+ /* "ROOT" is always on the MDT0. */
+ if (lu_fid_eq(fid, &lfsck->li_global_root_fid))
+ return 0;
+
+ return lfsck_dev_idx(lfsck);
+ }
+
fld_range_set_mdt(range);
rc = fld_server_lookup(env, ss->ss_server_fld, fid_seq(fid), range);
if (rc == 0)
if (rc != 0)
GOTO(stop, rc);
- rc = dt_delete(env, parent, (const struct dt_key *)name, th,
- BYPASS_CAPA);
+ rc = dt_delete(env, parent, (const struct dt_key *)name, th);
if (rc != 0)
GOTO(stop, rc);
int rc;
ENTRY;
- rc = linkea_data_new(&ldata,
- &lfsck_env_info(env)->lti_linkea_buf2);
- if (rc != 0)
- RETURN(rc);
-
cname = lfsck_name_get_const(env, name, strlen(name));
- rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
+ rc = linkea_links_new(&ldata, &lfsck_env_info(env)->lti_linkea_buf2,
+ cname, lfsck_dto2fid(parent));
if (rc != 0)
RETURN(rc);
if (rc != 0)
GOTO(stop, rc);
+ if (!dt_try_as_dir(env, child))
+ GOTO(stop, rc = -ENOTDIR);
+
/* 2a. increase child nlink */
rc = dt_declare_ref_add(env, child, th);
if (rc != 0)
GOTO(stop, rc);
- /* 3a. insert linkEA for child */
+ /* 3a. insert dot into child dir */
+ rec->rec_type = S_IFDIR;
+ rec->rec_fid = cfid;
+ rc = dt_declare_insert(env, child, (const struct dt_rec *)rec,
+ (const struct dt_key *)dot, th);
+ if (rc != 0)
+ GOTO(stop, rc);
+
+ /* 4a. insert dotdot into child dir */
+ rec->rec_fid = &LU_LPF_FID;
+ rc = dt_declare_insert(env, child, (const struct dt_rec *)rec,
+ (const struct dt_key *)dotdot, th);
+ if (rc != 0)
+ GOTO(stop, rc);
+
+ /* 5a. insert linkEA for child */
lfsck_buf_init(&linkea_buf, ldata.ld_buf->lb_buf,
ldata.ld_leh->leh_len);
rc = dt_declare_xattr_set(env, child, &linkea_buf,
if (rc != 0)
GOTO(stop, rc);
- /* 4a. insert name into parent dir */
+ /* 6a. insert name into parent dir */
rec->rec_type = S_IFDIR;
rec->rec_fid = cfid;
rc = dt_declare_insert(env, parent, (const struct dt_rec *)rec,
if (rc != 0)
GOTO(stop, rc);
- /* 5a. increase parent nlink */
+ /* 7a. increase parent nlink */
rc = dt_declare_ref_add(env, parent, th);
if (rc != 0)
GOTO(stop, rc);
- /* 6a. update bookmark */
+ /* 8a. update bookmark */
rc = dt_declare_record_write(env, bk_obj,
lfsck_buf_get(env, bk, len), 0, th);
if (rc != 0)
GOTO(stop, rc);
dt_write_lock(env, child, 0);
- /* 1b.1. create child */
+ /* 1b. create child */
rc = dt_create(env, child, la, NULL, dof, th);
if (rc != 0)
GOTO(unlock, rc);
- if (unlikely(!dt_try_as_dir(env, child)))
- GOTO(unlock, rc = -ENOTDIR);
+ /* 2b. increase child nlink */
+ rc = dt_ref_add(env, child, th);
+ if (rc != 0)
+ GOTO(unlock, rc);
- /* 1b.2. insert dot into child dir */
+ /* 3b. insert dot into child dir */
rec->rec_fid = cfid;
rc = dt_insert(env, child, (const struct dt_rec *)rec,
- (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
+ (const struct dt_key *)dot, th);
if (rc != 0)
GOTO(unlock, rc);
- /* 1b.3. insert dotdot into child dir */
+ /* 4b. insert dotdot into child dir */
rec->rec_fid = &LU_LPF_FID;
rc = dt_insert(env, child, (const struct dt_rec *)rec,
- (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
- if (rc != 0)
- GOTO(unlock, rc);
-
- /* 2b. increase child nlink */
- rc = dt_ref_add(env, child, th);
+ (const struct dt_key *)dotdot, th);
if (rc != 0)
GOTO(unlock, rc);
- /* 3b. insert linkEA for child. */
+ /* 5b. insert linkEA for child. */
rc = dt_xattr_set(env, child, &linkea_buf,
- XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
+ XATTR_NAME_LINK, 0, th);
dt_write_unlock(env, child);
if (rc != 0)
GOTO(stop, rc);
- /* 4b. insert name into parent dir */
+ /* 6b. insert name into parent dir */
rec->rec_fid = cfid;
rc = dt_insert(env, parent, (const struct dt_rec *)rec,
- (const struct dt_key *)name, th, BYPASS_CAPA, 1);
+ (const struct dt_key *)name, th);
if (rc != 0)
GOTO(stop, rc);
dt_write_lock(env, parent, 0);
- /* 5b. increase parent nlink */
+ /* 7b. increase parent nlink */
rc = dt_ref_add(env, parent, th);
dt_write_unlock(env, parent);
if (rc != 0)
bk->lb_lpf_fid = *cfid;
lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
- /* 6b. update bookmark */
+ /* 8b. update bookmark */
rc = dt_record_write(env, bk_obj,
lfsck_buf_get(env, bk, len), &pos, th);
int rc;
ENTRY;
- rc = linkea_data_new(&ldata,
- &lfsck_env_info(env)->lti_linkea_buf2);
- if (rc != 0)
- RETURN(rc);
-
cname = lfsck_name_get_const(env, name, strlen(name));
- rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
+ rc = linkea_links_new(&ldata, &lfsck_env_info(env)->lti_linkea_buf2,
+ cname, lfsck_dto2fid(parent));
if (rc != 0)
RETURN(rc);
if (rc != 0)
GOTO(stop, rc);
+ if (!dt_try_as_dir(env, child))
+ GOTO(stop, rc = -ENOTDIR);
+
/* 2a. increase child nlink */
rc = dt_declare_ref_add(env, child, th);
if (rc != 0)
GOTO(stop, rc);
- /* 3a. insert linkEA for child */
+ /* 3a. insert dot into child dir */
+ rec->rec_type = S_IFDIR;
+ rec->rec_fid = cfid;
+ rc = dt_declare_insert(env, child, (const struct dt_rec *)rec,
+ (const struct dt_key *)dot, th);
+ if (rc != 0)
+ GOTO(stop, rc);
+
+ /* 4a. insert dotdot into child dir */
+ rec->rec_fid = &LU_LPF_FID;
+ rc = dt_declare_insert(env, child, (const struct dt_rec *)rec,
+ (const struct dt_key *)dotdot, th);
+ if (rc != 0)
+ GOTO(stop, rc);
+
+ /* 5a. insert linkEA for child */
lfsck_buf_init(&linkea_buf, ldata.ld_buf->lb_buf,
ldata.ld_leh->leh_len);
rc = dt_declare_xattr_set(env, child, &linkea_buf,
if (rc != 0)
GOTO(stop, rc);
- /* 4a. update bookmark */
+ /* 6a. update bookmark */
rc = dt_declare_record_write(env, bk_obj,
lfsck_buf_get(env, bk, len), 0, th);
if (rc != 0)
GOTO(stop, rc);
dt_write_lock(env, child, 0);
- /* 1b.1. create child */
+ /* 1b. create child */
rc = dt_create(env, child, la, NULL, dof, th);
if (rc != 0)
GOTO(unlock, rc);
- if (unlikely(!dt_try_as_dir(env, child)))
- GOTO(unlock, rc = -ENOTDIR);
+ /* 2b. increase child nlink */
+ rc = dt_ref_add(env, child, th);
+ if (rc != 0)
+ GOTO(unlock, rc);
- /* 1b.2. insert dot into child dir */
+ /* 3b. insert dot into child dir */
rec->rec_type = S_IFDIR;
rec->rec_fid = cfid;
rc = dt_insert(env, child, (const struct dt_rec *)rec,
- (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
+ (const struct dt_key *)dot, th);
if (rc != 0)
GOTO(unlock, rc);
- /* 1b.3. insert dotdot into child dir */
+ /* 4b. insert dotdot into child dir */
rec->rec_fid = &LU_LPF_FID;
rc = dt_insert(env, child, (const struct dt_rec *)rec,
- (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
- if (rc != 0)
- GOTO(unlock, rc);
-
- /* 2b. increase child nlink */
- rc = dt_ref_add(env, child, th);
+ (const struct dt_key *)dotdot, th);
if (rc != 0)
GOTO(unlock, rc);
- /* 3b. insert linkEA for child */
+ /* 5b. insert linkEA for child */
rc = dt_xattr_set(env, child, &linkea_buf,
- XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
+ XATTR_NAME_LINK, 0, th);
if (rc != 0)
GOTO(unlock, rc);
bk->lb_lpf_fid = *cfid;
lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
- /* 4b. update bookmark */
+ /* 6b. update bookmark */
rc = dt_record_write(env, bk_obj,
lfsck_buf_get(env, bk, len), &pos, th);
/* 5b. insert name into parent dir */
rc = dt_insert(env, parent, (const struct dt_rec *)rec,
- (const struct dt_key *)name, th, BYPASS_CAPA, 1);
+ (const struct dt_key *)name, th);
if (rc != 0)
GOTO(stop, rc);
LASSERT(parent != NULL);
LASSERT(lfsck->li_lpf_obj == NULL);
+ snprintf(name, 8, "MDT%04x", node);
rc = lfsck_lock(env, lfsck, parent, name, llh,
MDS_INODELOCK_UPDATE, LCK_PW);
if (rc != 0)
RETURN(rc);
- snprintf(name, 8, "MDT%04x", node);
if (fid_is_zero(&bk->lb_lpf_fid)) {
/* There is corner case that: in former LFSCK scanning we have
* created the .lustre/lost+found/MDTxxxx but failed to update
* the lfsck_bookmark::lb_lpf_fid successfully. So need lookup
* it from MDT0 firstly. */
rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
- (const struct dt_key *)name, BYPASS_CAPA);
+ (const struct dt_key *)name);
if (rc != 0 && rc != -ENOENT)
GOTO(unlock, rc);
*cfid = bk->lb_lpf_fid;
}
- child = lfsck_object_find_bottom(env, lfsck, cfid);
+ child = lfsck_object_find_bottom_new(env, lfsck, cfid);
if (IS_ERR(child))
GOTO(unlock, rc = PTR_ERR(child));
}
memset(la, 0, sizeof(*la));
- la->la_atime = la->la_mtime = la->la_ctime = cfs_time_current_sec();
+ la->la_atime = la->la_mtime = la->la_ctime = ktime_get_real_seconds();
la->la_mode = S_IFDIR | S_IRWXU;
la->la_valid = LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE |
LA_UID | LA_GID;
int rc;
ENTRY;
- it = iops->init(env, parent, LUDA_64BITHASH, BYPASS_CAPA);
+ it = iops->init(env, parent, LUDA_64BITHASH);
if (IS_ERR(it))
RETURN(PTR_ERR(it));
fid_zero(fid);
rc = dt_lookup(env, child, (struct dt_rec *)fid,
- (const struct dt_key *)dotdot, BYPASS_CAPA);
+ (const struct dt_key *)dotdot);
if (rc != 0)
GOTO(linkea, rc);
}
rc = dt_lookup(env, parent2, (struct dt_rec *)fid,
- (const struct dt_key *)name2, BYPASS_CAPA);
+ (const struct dt_key *)name2);
dt_read_unlock(env, child);
lfsck_ibits_unlock(&lh, LCK_PR);
if (rc != 0 && rc != -ENOENT)
lfsck_lfsck2name(lfsck), rc);
}
- if (!fid_is_zero(&bk->lb_lpf_fid)) {
- if (unlikely(!fid_is_norm(&bk->lb_lpf_fid))) {
- struct lu_fid tfid = bk->lb_lpf_fid;
-
- /* Invalid FID record in the bookmark file, reset it. */
- fid_zero(&bk->lb_lpf_fid);
- rc = lfsck_bookmark_store(env, lfsck);
-
- CDEBUG(D_LFSCK, "%s: reset invalid LPF fid "DFID
- " in the bookmark file: rc = %d\n",
- lfsck_lfsck2name(lfsck), PFID(&tfid), rc);
-
- if (rc != 0)
- GOTO(put, rc);
- } else {
- child1 = lfsck_object_find_bottom(env, lfsck,
- &bk->lb_lpf_fid);
- if (IS_ERR(child1)) {
- child1 = NULL;
- goto find_child2;
- }
-
- if (unlikely(!dt_object_exists(child1) ||
- dt_object_remote(child1)) ||
- !S_ISDIR(lfsck_object_type(child1))) {
- /* Invalid FID record in the bookmark file,
- * reset it. */
- fid_zero(&bk->lb_lpf_fid);
- rc = lfsck_bookmark_store(env, lfsck);
-
- CDEBUG(D_LFSCK, "%s: reset invalid LPF fid "DFID
- " in the bookmark file: rc = %d\n",
- lfsck_lfsck2name(lfsck),
- PFID(lfsck_dto2fid(child1)), rc);
-
- if (rc != 0)
- GOTO(put, rc);
-
- lfsck_object_put(env, child1);
- child1 = NULL;
- } else if (unlikely(!dt_try_as_dir(env, child1))) {
- GOTO(put, rc = -ENOTDIR);
- }
- }
- }
-
-find_child2:
+ /* child2 */
snprintf(name, 8, "MDT%04x", node);
rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
- (const struct dt_key *)name, BYPASS_CAPA);
+ (const struct dt_key *)name);
if (rc == -ENOENT) {
- if (!fid_is_zero(&bk->lb_lpf_fid))
- goto check_child1;
-
- GOTO(put, rc = 0);
+ rc = 0;
+ goto find_child1;
}
if (rc != 0)
if (rc != 0)
GOTO(put, rc);
- goto check_child1;
+ goto find_child1;
}
child2 = lfsck_object_find_bottom(env, lfsck, cfid);
if (rc != 0)
GOTO(put, rc);
- goto check_child1;
+ goto find_child1;
}
- if (unlikely(!dt_try_as_dir(env, child2)))
- GOTO(put, rc = -ENOTDIR);
+ if (unlikely(!dt_try_as_dir(env, child2))) {
+ lfsck_object_put(env, child2);
+ child2 = NULL;
+ rc = -ENOTDIR;
+ }
- if (child1 == NULL) {
- rc = lfsck_verify_lpf_pairs(env, lfsck, child2, name,
- pfid, LVLT_BY_NAMEENTRY);
- } else if (!lu_fid_eq(cfid, &bk->lb_lpf_fid)) {
- rc = lfsck_verify_lpf_pairs(env, lfsck, child1, name,
- pfid, LVLT_BY_BOOKMARK);
- if (!lu_fid_eq(pfid, &LU_LPF_FID))
- rc = lfsck_verify_lpf_pairs(env, lfsck, child2,
- name, pfid,
- LVLT_BY_NAMEENTRY);
- } else {
+find_child1:
+ if (fid_is_zero(&bk->lb_lpf_fid))
+ goto check_child2;
+
+ if (likely(lu_fid_eq(cfid, &bk->lb_lpf_fid))) {
if (lfsck->li_lpf_obj == NULL) {
lu_object_get(&child2->do_lu);
lfsck->li_lpf_obj = child2;
cname = lfsck_name_get_const(env, name, strlen(name));
rc = lfsck_verify_linkea(env, child2, cname, &LU_LPF_FID);
+
+ GOTO(put, rc);
}
- GOTO(put, rc);
+ if (unlikely(!fid_is_norm(&bk->lb_lpf_fid))) {
+ struct lu_fid tfid = bk->lb_lpf_fid;
+
+ /* Invalid FID record in the bookmark file, reset it. */
+ fid_zero(&bk->lb_lpf_fid);
+ rc = lfsck_bookmark_store(env, lfsck);
+
+ CDEBUG(D_LFSCK, "%s: reset invalid LPF fid "DFID
+ " in the bookmark file: rc = %d\n",
+ lfsck_lfsck2name(lfsck), PFID(&tfid), rc);
-check_child1:
- if (child1 != NULL)
- rc = lfsck_verify_lpf_pairs(env, lfsck, child1, name,
- pfid, LVLT_BY_BOOKMARK);
+ if (rc != 0)
+ GOTO(put, rc);
+
+ goto check_child2;
+ }
+
+ child1 = lfsck_object_find_bottom(env, lfsck, &bk->lb_lpf_fid);
+ if (IS_ERR(child1)) {
+ child1 = NULL;
+ goto check_child2;
+ }
+
+ if (unlikely(!dt_object_exists(child1) ||
+ dt_object_remote(child1)) ||
+ !S_ISDIR(lfsck_object_type(child1))) {
+ /* Invalid FID record in the bookmark file, reset it. */
+ fid_zero(&bk->lb_lpf_fid);
+ rc = lfsck_bookmark_store(env, lfsck);
+
+ CDEBUG(D_LFSCK, "%s: reset invalid LPF fid "DFID
+ " in the bookmark file: rc = %d\n",
+ lfsck_lfsck2name(lfsck),
+ PFID(lfsck_dto2fid(child1)), rc);
+
+ if (rc != 0)
+ GOTO(put, rc);
+
+ lfsck_object_put(env, child1);
+ child1 = NULL;
+ goto check_child2;
+ }
+
+ if (unlikely(!dt_try_as_dir(env, child1))) {
+ lfsck_object_put(env, child1);
+ child1 = NULL;
+ rc = -ENOTDIR;
+ goto check_child2;
+ }
+
+ rc = lfsck_verify_lpf_pairs(env, lfsck, child1, name, pfid,
+ LVLT_BY_BOOKMARK);
+ if (lu_fid_eq(pfid, &LU_LPF_FID))
+ GOTO(put, rc);
+
+check_child2:
+ if (child2 != NULL)
+ rc = lfsck_verify_lpf_pairs(env, lfsck, child2, name,
+ pfid, LVLT_BY_NAMEENTRY);
GOTO(put, rc);
return 0;
}
-int lfsck_bits_dump(struct seq_file *m, int bits, const char *names[],
- const char *prefix)
+void lfsck_bits_dump(struct seq_file *m, int bits, const char *names[],
+ const char *prefix)
{
int flag;
int i;
bool newline = (bits != 0 ? false : true);
- int rc;
- rc = seq_printf(m, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
- if (rc < 0)
- return rc;
+ seq_printf(m, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
if (flag & bits) {
if (bits == 0)
newline = true;
- rc = seq_printf(m, "%s%c", names[i],
- newline ? '\n' : ',');
- if (rc < 0)
- return rc;
+ seq_printf(m, "%s%c", names[i],
+ newline ? '\n' : ',');
}
}
}
if (!newline)
- rc = seq_printf(m, "\n");
-
- return rc;
+ seq_putc(m, '\n');
}
-int lfsck_time_dump(struct seq_file *m, __u64 time, const char *name)
+void lfsck_time_dump(struct seq_file *m, time64_t time, const char *name)
{
- int rc;
-
if (time == 0) {
- rc = seq_printf(m, "%s_time: N/A\n", name);
- if (rc == 0)
- rc = seq_printf(m, "time_since_%s: N/A\n", name);
-
- return rc;
+ seq_printf(m, "%s_time: N/A\n", name);
+ seq_printf(m, "time_since_%s: N/A\n", name);
+ } else {
+ seq_printf(m, "%s_time: %lld\n", name, time);
+ seq_printf(m, "time_since_%s: %lld seconds\n",
+ name, ktime_get_real_seconds() - time);
}
-
- rc = seq_printf(m, "%s_time: "LPU64"\n", name, time);
- if (rc == 0)
- rc = seq_printf(m, "time_since_%s: "LPU64" seconds\n",
- name, cfs_time_current_sec() - time);
-
- return rc;
}
-int lfsck_pos_dump(struct seq_file *m, struct lfsck_position *pos,
- const char *prefix)
+void lfsck_pos_dump(struct seq_file *m, struct lfsck_position *pos,
+ const char *prefix)
{
if (fid_is_zero(&pos->lp_dir_parent)) {
- if (pos->lp_oit_cookie == 0)
- return seq_printf(m, "%s: N/A, N/A, N/A\n", prefix);
-
- return seq_printf(m, "%s: "LPU64", N/A, N/A\n",
- prefix, pos->lp_oit_cookie);
+ if (pos->lp_oit_cookie == 0) {
+ seq_printf(m, "%s: N/A, N/A, N/A\n", prefix);
+ return;
+ }
+ seq_printf(m, "%s: %llu, N/A, N/A\n",
+ prefix, pos->lp_oit_cookie);
+ } else {
+ seq_printf(m, "%s: %llu, "DFID", %#llx\n",
+ prefix, pos->lp_oit_cookie,
+ PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
}
-
- return seq_printf(m, "%s: "LPU64", "DFID", "LPX64"\n",
- prefix, pos->lp_oit_cookie,
- PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
}
void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
if (!lfsck->li_current_oit_processed && !init)
pos->lp_oit_cookie--;
- LASSERT(pos->lp_oit_cookie > 0);
+ if (unlikely(pos->lp_oit_cookie == 0))
+ pos->lp_oit_cookie = 1;
if (lfsck->li_di_dir != NULL) {
struct dt_object *dto = lfsck->li_obj_dir;
bool dirty = false;
if (limit != LFSCK_SPEED_NO_LIMIT) {
- if (limit > msecs_to_jiffies(MSEC_PER_SEC)) {
- lfsck->li_sleep_rate = limit /
- msecs_to_jiffies(MSEC_PER_SEC);
+ if (limit > cfs_time_seconds(1)) {
+ lfsck->li_sleep_rate = limit / cfs_time_seconds(1);
lfsck->li_sleep_jif = 1;
} else {
lfsck->li_sleep_rate = 1;
- lfsck->li_sleep_jif = msecs_to_jiffies(MSEC_PER_SEC) /
- limit;
+ lfsck->li_sleep_jif = cfs_time_seconds(1) / limit;
}
} else {
lfsck->li_sleep_jif = 0;
return lad;
}
+struct lfsck_assistant_object *
+lfsck_assistant_object_init(const struct lu_env *env, const struct lu_fid *fid,
+ const struct lu_attr *attr, __u64 cookie,
+ bool is_dir)
+{
+ struct lfsck_assistant_object *lso;
+
+ OBD_ALLOC_PTR(lso);
+ if (lso == NULL)
+ return ERR_PTR(-ENOMEM);
+
+ lso->lso_fid = *fid;
+ if (attr != NULL)
+ lso->lso_attr = *attr;
+
+ atomic_set(&lso->lso_ref, 1);
+ lso->lso_oit_cookie = cookie;
+ if (is_dir)
+ lso->lso_is_dir = 1;
+
+ return lso;
+}
+
+struct dt_object *
+lfsck_assistant_object_load(const struct lu_env *env,
+ struct lfsck_instance *lfsck,
+ struct lfsck_assistant_object *lso)
+{
+ struct dt_object *obj;
+
+ obj = lfsck_object_find_bottom(env, lfsck, &lso->lso_fid);
+ if (IS_ERR(obj))
+ return obj;
+
+ if (unlikely(!dt_object_exists(obj) || lfsck_is_dead_obj(obj))) {
+ lso->lso_dead = 1;
+ lfsck_object_put(env, obj);
+
+ return ERR_PTR(-ENOENT);
+ }
+
+ if (lso->lso_is_dir && unlikely(!dt_try_as_dir(env, obj))) {
+ lfsck_object_put(env, obj);
+
+ return ERR_PTR(-ENOTDIR);
+ }
+
+ return obj;
+}
+
/**
* Generic LFSCK asynchronous communication interpretor function.
* The LFSCK RPC reply for both the event notification and status
switch (lr->lr_event) {
case LE_START:
+ if (unlikely(rc == -EINPROGRESS)) {
+ ltd->ltd_retry_start = 1;
+ break;
+ }
+
if (rc != 0) {
CDEBUG(D_LFSCK, "%s: fail to notify %s %x for %s "
"start: rc = %d\n",
}
if (rc != 0) {
+ if (lr->lr_flags & LEF_QUERY_ALL) {
+ lfsck_reset_ltd_status(ltd, com->lc_type);
+ break;
+ }
+
spin_lock(<ds->ltd_lock);
list_del_init(phase_list);
list_del_init(list);
CDEBUG(D_LFSCK, "%s: invalid query reply for %s: "
"rc = %d\n", lfsck_lfsck2name(com->lc_lfsck),
lad->lad_name, rc);
+
+ if (lr->lr_flags & LEF_QUERY_ALL) {
+ lfsck_reset_ltd_status(ltd, com->lc_type);
+ break;
+ }
+
spin_lock(<ds->ltd_lock);
list_del_init(phase_list);
list_del_init(list);
break;
}
+ if (lr->lr_flags & LEF_QUERY_ALL) {
+ if (com->lc_type == LFSCK_TYPE_LAYOUT) {
+ ltd->ltd_layout_status = reply->lr_status;
+ ltd->ltd_layout_repaired = reply->lr_repaired;
+ } else {
+ ltd->ltd_namespace_status = reply->lr_status;
+ ltd->ltd_namespace_repaired =
+ reply->lr_repaired;
+ }
+ break;
+ }
+
switch (reply->lr_status) {
case LS_SCANNING_PHASE1:
break;
if (ltds == &lfsck->li_ost_descs)
lr->lr_flags = LEF_TO_OST;
+ memset(laia, 0, sizeof(*laia));
laia->laia_com = com;
laia->laia_ltds = ltds;
atomic_inc(<d->ltd_ref);
laia->laia_ltd = ltd;
laia->laia_lr = lr;
- laia->laia_shared = 0;
rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
lfsck_async_interpret_common,
ltd->ltd_index, lad->lad_name, rc);
lfsck_tgt_put(ltd);
} else {
- rc = ptlrpc_set_wait(set);
+ rc = ptlrpc_set_wait(env, set);
}
ptlrpc_set_destroy(set);
if (laia->laia_com != NULL)
lfsck_component_get(laia->laia_com);
req->rq_interpret_reply = interpreter;
+ req->rq_allow_intr = 1;
+ req->rq_no_delay = 1;
ptlrpc_set_add_req(set, req);
return 0;
}
+int lfsck_query_all(const struct lu_env *env, struct lfsck_component *com)
+{
+ struct lfsck_thread_info *info = lfsck_env_info(env);
+ struct lfsck_request *lr = &info->lti_lr;
+ struct lfsck_async_interpret_args *laia = &info->lti_laia;
+ struct lfsck_instance *lfsck = com->lc_lfsck;
+ struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
+ struct lfsck_tgt_desc *ltd;
+ struct ptlrpc_request_set *set;
+ int idx;
+ int rc;
+ ENTRY;
+
+ memset(lr, 0, sizeof(*lr));
+ lr->lr_event = LE_QUERY;
+ lr->lr_active = com->lc_type;
+ lr->lr_flags = LEF_QUERY_ALL;
+
+ memset(laia, 0, sizeof(*laia));
+ laia->laia_com = com;
+ laia->laia_lr = lr;
+
+ set = ptlrpc_prep_set();
+ if (set == NULL)
+ RETURN(-ENOMEM);
+
+again:
+ laia->laia_ltds = ltds;
+ down_read(<ds->ltd_rw_sem);
+ cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
+ ltd = lfsck_tgt_get(ltds, idx);
+ LASSERT(ltd != NULL);
+
+ laia->laia_ltd = ltd;
+ up_read(<ds->ltd_rw_sem);
+ rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
+ lfsck_async_interpret_common,
+ laia, LFSCK_QUERY);
+ if (rc != 0) {
+ struct lfsck_assistant_data *lad = com->lc_data;
+
+ CDEBUG(D_LFSCK, "%s: Fail to query %s %x for stat %s: "
+ "rc = %d\n", lfsck_lfsck2name(lfsck),
+ (lr->lr_flags & LEF_TO_OST) ? "OST" : "MDT",
+ ltd->ltd_index, lad->lad_name, rc);
+ lfsck_reset_ltd_status(ltd, com->lc_type);
+ lfsck_tgt_put(ltd);
+ }
+ down_read(<ds->ltd_rw_sem);
+ }
+ up_read(<ds->ltd_rw_sem);
+
+ if (com->lc_type == LFSCK_TYPE_LAYOUT && !(lr->lr_flags & LEF_TO_OST)) {
+ ltds = &lfsck->li_ost_descs;
+ lr->lr_flags |= LEF_TO_OST;
+ goto again;
+ }
+
+ rc = ptlrpc_set_wait(env, set);
+ ptlrpc_set_destroy(set);
+
+ RETURN(rc);
+}
+
int lfsck_start_assistant(const struct lu_env *env, struct lfsck_component *com,
struct lfsck_start_param *lsp)
{
lad->lad_assistant_status = 0;
lad->lad_post_result = 0;
- lad->lad_to_post = 0;
- lad->lad_to_double_scan = 0;
- lad->lad_in_double_scan = 0;
- lad->lad_exit = 0;
+ lad->lad_flags = 0;
+ lad->lad_advance_lock = false;
thread_set_flags(athread, 0);
lta = lfsck_thread_args_init(lfsck, com, lsp);
l_wait_event(mthread->t_ctl_waitq,
thread_is_running(athread) ||
- thread_is_stopped(athread),
+ thread_is_stopped(athread) ||
+ !thread_is_starting(mthread),
&lwi);
- if (unlikely(!thread_is_running(athread)))
+ if (unlikely(!thread_is_starting(mthread)))
+ /* stopped by race */
+ rc = -ESRCH;
+ else if (unlikely(!thread_is_running(athread)))
rc = lad->lad_assistant_status;
else
rc = 0;
lad->lad_post_result = *result;
if (*result <= 0)
- lad->lad_exit = 1;
- lad->lad_to_post = 1;
+ set_bit(LAD_EXIT, &lad->lad_flags);
+ set_bit(LAD_TO_POST, &lad->lad_flags);
+
+ CDEBUG(D_LFSCK, "%s: waiting for assistant to do %s post, rc = %d\n",
+ lfsck_lfsck2name(com->lc_lfsck), lad->lad_name, *result);
wake_up_all(&athread->t_ctl_waitq);
l_wait_event(mthread->t_ctl_waitq,
if (lad->lad_assistant_status < 0)
*result = lad->lad_assistant_status;
+
+ CDEBUG(D_LFSCK, "%s: the assistant has done %s post, rc = %d\n",
+ lfsck_lfsck2name(com->lc_lfsck), lad->lad_name, *result);
}
int lfsck_double_scan_generic(const struct lu_env *env,
struct l_wait_info lwi = { 0 };
if (status != LS_SCANNING_PHASE2)
- lad->lad_exit = 1;
+ set_bit(LAD_EXIT, &lad->lad_flags);
else
- lad->lad_to_double_scan = 1;
+ set_bit(LAD_TO_DOUBLE_SCAN, &lad->lad_flags);
+
+ CDEBUG(D_LFSCK, "%s: waiting for assistant to do %s double_scan, "
+ "status %d\n",
+ lfsck_lfsck2name(com->lc_lfsck), lad->lad_name, status);
wake_up_all(&athread->t_ctl_waitq);
l_wait_event(mthread->t_ctl_waitq,
- lad->lad_in_double_scan ||
+ test_bit(LAD_IN_DOUBLE_SCAN, &lad->lad_flags) ||
thread_is_stopped(athread),
&lwi);
+ CDEBUG(D_LFSCK, "%s: the assistant has done %s double_scan, "
+ "status %d\n", lfsck_lfsck2name(com->lc_lfsck), lad->lad_name,
+ lad->lad_assistant_status);
+
if (lad->lad_assistant_status < 0)
return lad->lad_assistant_status;
struct ptlrpc_thread *athread = &lad->lad_thread;
struct l_wait_info lwi = { 0 };
- lad->lad_exit = 1;
+ set_bit(LAD_EXIT, &lad->lad_flags);
wake_up_all(&athread->t_ctl_waitq);
l_wait_event(mthread->t_ctl_waitq,
thread_is_init(athread) ||
&lwi);
}
-/* external interfaces */
+int lfsck_load_one_trace_file(const struct lu_env *env,
+ struct lfsck_component *com,
+ struct dt_object *parent,
+ struct dt_object **child,
+ const struct dt_index_features *ft,
+ const char *name, bool reset)
+{
+ struct lfsck_instance *lfsck = com->lc_lfsck;
+ struct dt_object *obj;
+ int rc;
+ ENTRY;
+
+ if (*child != NULL) {
+ struct dt_it *it;
+ const struct dt_it_ops *iops;
+ struct lu_fid *fid = &lfsck_env_info(env)->lti_fid3;
+
+ if (!reset)
+ RETURN(0);
-int lfsck_get_speed(struct seq_file *m, struct dt_device *key)
+ obj = *child;
+ rc = obj->do_ops->do_index_try(env, obj, ft);
+ if (rc)
+ /* unlink by force */
+ goto unlink;
+
+ iops = &obj->do_index_ops->dio_it;
+ it = iops->init(env, obj, 0);
+ if (IS_ERR(it))
+ /* unlink by force */
+ goto unlink;
+
+ fid_zero(fid);
+ rc = iops->get(env, it, (const struct dt_key *)fid);
+ if (rc >= 0) {
+ rc = iops->next(env, it);
+ iops->put(env, it);
+ }
+ iops->fini(env, it);
+ if (rc > 0)
+ /* "rc > 0" means the index file is empty. */
+ RETURN(0);
+
+unlink:
+ /* The old index is not empty, remove it firstly. */
+ rc = local_object_unlink(env, lfsck->li_bottom, parent, name);
+ CDEBUG_LIMIT(rc ? D_ERROR : D_LFSCK,
+ "%s: unlink lfsck sub trace file %s: rc = %d\n",
+ lfsck_lfsck2name(com->lc_lfsck), name, rc);
+ if (rc)
+ RETURN(rc);
+
+ if (*child) {
+ lfsck_object_put(env, *child);
+ *child = NULL;
+ }
+ } else if (reset) {
+ goto unlink;
+ }
+
+ obj = local_index_find_or_create(env, lfsck->li_los, parent, name,
+ S_IFREG | S_IRUGO | S_IWUSR, ft);
+ if (IS_ERR(obj))
+ RETURN(PTR_ERR(obj));
+
+ rc = obj->do_ops->do_index_try(env, obj, ft);
+ if (rc) {
+ lfsck_object_put(env, obj);
+ CDEBUG(D_LFSCK, "%s: LFSCK fail to load "
+ "sub trace file %s: rc = %d\n",
+ lfsck_lfsck2name(com->lc_lfsck), name, rc);
+ } else {
+ *child = obj;
+ }
+
+ RETURN(rc);
+}
+
+int lfsck_load_sub_trace_files(const struct lu_env *env,
+ struct lfsck_component *com,
+ const struct dt_index_features *ft,
+ const char *prefix, bool reset)
+{
+ char *name = lfsck_env_info(env)->lti_key;
+ struct lfsck_sub_trace_obj *lsto;
+ int rc;
+ int i;
+
+ for (i = 0, rc = 0, lsto = &com->lc_sub_trace_objs[0];
+ i < LFSCK_STF_COUNT && rc == 0; i++, lsto++) {
+ snprintf(name, NAME_MAX, "%s_%02d", prefix, i);
+ rc = lfsck_load_one_trace_file(env, com,
+ com->lc_lfsck->li_lfsck_dir,
+ &lsto->lsto_obj, ft, name, reset);
+ }
+
+ return rc;
+}
+
+/* external interfaces */
+int lfsck_get_speed(char *buf, struct dt_device *key)
{
struct lu_env env;
struct lfsck_instance *lfsck;
RETURN(rc);
lfsck = lfsck_instance_find(key, true, false);
- if (likely(lfsck != NULL)) {
- seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_speed_limit);
+ if (lfsck && buf) {
+ rc = sprintf(buf, "%u\n",
+ lfsck->li_bookmark_ram.lb_speed_limit);
lfsck_instance_put(&env, lfsck);
} else {
rc = -ENXIO;
}
EXPORT_SYMBOL(lfsck_get_speed);
-int lfsck_set_speed(struct dt_device *key, int val)
+int lfsck_set_speed(struct dt_device *key, __u32 val)
{
struct lu_env env;
struct lfsck_instance *lfsck;
}
EXPORT_SYMBOL(lfsck_set_speed);
-int lfsck_get_windows(struct seq_file *m, struct dt_device *key)
+int lfsck_get_windows(char *buf, struct dt_device *key)
{
struct lu_env env;
struct lfsck_instance *lfsck;
lfsck = lfsck_instance_find(key, true, false);
if (likely(lfsck != NULL)) {
- seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_async_windows);
+ rc = sprintf(buf, "%u\n",
+ lfsck->li_bookmark_ram.lb_async_windows);
lfsck_instance_put(&env, lfsck);
} else {
rc = -ENXIO;
}
EXPORT_SYMBOL(lfsck_get_windows);
-int lfsck_set_windows(struct dt_device *key, int val)
+int lfsck_set_windows(struct dt_device *key, unsigned int val)
{
struct lu_env env;
struct lfsck_instance *lfsck;
if (likely(lfsck != NULL)) {
com = lfsck_component_find(lfsck, type);
if (likely(com != NULL)) {
- rc = com->lc_ops->lfsck_dump(&env, com, m);
+ com->lc_ops->lfsck_dump(&env, com, m);
lfsck_component_put(&env, com);
} else {
rc = -ENOTSUPP;
lr->lr_active = LFSCK_TYPES_ALL;
lr->lr_param = stop->ls_flags;
- laia->laia_com = NULL;
+ memset(laia, 0, sizeof(*laia));
laia->laia_ltds = ltds;
laia->laia_lr = lr;
- laia->laia_result = 0;
laia->laia_shared = 1;
down_read(<ds->ltd_rw_sem);
}
up_read(<ds->ltd_rw_sem);
- rc = ptlrpc_set_wait(set);
+ rc = ptlrpc_set_wait(env, set);
ptlrpc_set_destroy(set);
if (rc == 0)
struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
__u32 idx;
int rc = 0;
+ bool retry = false;
ENTRY;
LASSERT(start->ls_flags & LPF_BROADCAST);
- set = ptlrpc_prep_set();
- if (unlikely(set == NULL))
- RETURN(-ENOMEM);
-
memset(lr, 0, sizeof(*lr));
lr->lr_event = LE_START;
lr->lr_index = lfsck_dev_idx(lfsck);
LSV_ASYNC_WINDOWS | LSV_CREATE_OSTOBJ |
LSV_CREATE_MDTOBJ;
- laia->laia_com = NULL;
+ memset(laia, 0, sizeof(*laia));
laia->laia_ltds = ltds;
laia->laia_lr = lr;
- laia->laia_result = 0;
laia->laia_shared = 1;
+again:
+ set = ptlrpc_prep_set();
+ if (unlikely(!set))
+ RETURN(-ENOMEM);
+
down_read(<ds->ltd_rw_sem);
cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
ltd = lfsck_tgt_get(ltds, idx);
LASSERT(ltd != NULL);
+ if (retry && !ltd->ltd_retry_start) {
+ lfsck_tgt_put(ltd);
+ continue;
+ }
+
laia->laia_ltd = ltd;
+ ltd->ltd_retry_start = 0;
ltd->ltd_layout_done = 0;
ltd->ltd_namespace_done = 0;
ltd->ltd_synced_failures = 0;
RETURN(rc);
}
- rc = ptlrpc_set_wait(set);
+ rc = ptlrpc_set_wait(env, set);
ptlrpc_set_destroy(set);
if (rc == 0)
rc = laia->laia_result;
+ if (unlikely(rc == -EINPROGRESS)) {
+ retry = true;
+ set_current_state(TASK_INTERRUPTIBLE);
+ schedule_timeout(cfs_time_seconds(1));
+ set_current_state(TASK_RUNNING);
+ if (!signal_pending(current) &&
+ thread_is_running(&lfsck->li_thread))
+ goto again;
+
+ rc = -EINTR;
+ }
+
if (rc != 0) {
struct lfsck_stop *stop = &info->lti_stop;
struct l_wait_info lwi = { 0 };
struct lfsck_thread_args *lta;
struct task_struct *task;
+ struct lfsck_tgt_descs *ltds;
+ struct lfsck_tgt_desc *ltd;
+ __u32 idx;
int rc = 0;
__u16 valid = 0;
__u16 flags = 0;
__u16 type = 1;
ENTRY;
+ if (key->dd_rdonly)
+ RETURN(-EROFS);
+
lfsck = lfsck_instance_find(key, true, false);
if (unlikely(lfsck == NULL))
RETURN(-ENXIO);
+ if (unlikely(lfsck->li_stopping))
+ GOTO(put, rc = -ENXIO);
+
/* System is not ready, try again later. */
- if (unlikely(lfsck->li_namespace == NULL))
- GOTO(put, rc = -EAGAIN);
+ if (unlikely(lfsck->li_namespace == NULL ||
+ lfsck_dev_site(lfsck)->ss_server_fld == NULL))
+ GOTO(put, rc = -EINPROGRESS);
/* start == NULL means auto trigger paused LFSCK. */
- if ((start == NULL) &&
- (list_empty(&lfsck->li_list_scan) ||
- OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
- GOTO(put, rc = 0);
+ if (!start) {
+ if (list_empty(&lfsck->li_list_scan) ||
+ OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO))
+ GOTO(put, rc = 0);
+ } else if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
+ CERROR("%s: only allow to specify '-A | -o' via MDS\n",
+ lfsck_lfsck2name(lfsck));
+
+ GOTO(put, rc = -EPERM);
+ }
bk = &lfsck->li_bookmark_ram;
thread = &lfsck->li_thread;
mutex_lock(&lfsck->li_mutex);
spin_lock(&lfsck->li_lock);
+ if (unlikely(thread_is_stopping(thread))) {
+ /* Someone is stopping the LFSCK. */
+ spin_unlock(&lfsck->li_lock);
+ GOTO(out, rc = -EBUSY);
+ }
+
if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
rc = -EALREADY;
if (unlikely(start == NULL)) {
if (start == NULL)
goto trigger;
- if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
- CERROR("%s: only allow to specify '-A | -o' via MDS\n",
- lfsck_lfsck2name(lfsck));
-
- GOTO(out, rc = -EPERM);
- }
-
start->ls_version = bk->lb_version;
if (start->ls_active != 0) {
}
}
+ ltds = &lfsck->li_mdt_descs;
+ down_read(<ds->ltd_rw_sem);
+ cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
+ ltd = lfsck_ltd2tgt(ltds, idx);
+ LASSERT(ltd != NULL);
+
+ ltd->ltd_layout_done = 0;
+ ltd->ltd_namespace_done = 0;
+ ltd->ltd_synced_failures = 0;
+ lfsck_reset_ltd_status(ltd, LFSCK_TYPE_NAMESPACE);
+ lfsck_reset_ltd_status(ltd, LFSCK_TYPE_LAYOUT);
+ list_del_init(<d->ltd_layout_phase_list);
+ list_del_init(<d->ltd_layout_list);
+ list_del_init(<d->ltd_namespace_phase_list);
+ list_del_init(<d->ltd_namespace_list);
+ }
+ up_read(<ds->ltd_rw_sem);
+
+ ltds = &lfsck->li_ost_descs;
+ down_read(<ds->ltd_rw_sem);
+ cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
+ ltd = lfsck_ltd2tgt(ltds, idx);
+ LASSERT(ltd != NULL);
+
+ ltd->ltd_layout_done = 0;
+ ltd->ltd_synced_failures = 0;
+ lfsck_reset_ltd_status(ltd, LFSCK_TYPE_LAYOUT);
+ list_del_init(<d->ltd_layout_phase_list);
+ list_del_init(<d->ltd_layout_list);
+ }
+ up_read(<ds->ltd_rw_sem);
+
trigger:
lfsck->li_args_dir = LUDA_64BITHASH | LUDA_VERIFY | LUDA_TYPE;
if (bk->lb_param & LPF_DRYRUN)
flags |= DOIF_OUTUSED;
lfsck->li_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
- thread_set_flags(thread, 0);
lta = lfsck_thread_args_init(lfsck, NULL, lsp);
if (IS_ERR(lta))
GOTO(out, rc = PTR_ERR(lta));
__lfsck_set_speed(lfsck, bk->lb_speed_limit);
+ spin_lock(&lfsck->li_lock);
+ thread_set_flags(thread, SVC_STARTING);
+ spin_unlock(&lfsck->li_lock);
task = kthread_run(lfsck_master_engine, lta, "lfsck");
if (IS_ERR(task)) {
rc = PTR_ERR(task);
RETURN(-ENXIO);
thread = &lfsck->li_thread;
- /* release lfsck::li_mutex to avoid deadlock. */
- if (stop != NULL && stop->ls_flags & LPF_BROADCAST) {
- if (!lfsck->li_master) {
- CERROR("%s: only allow to specify '-A' via MDS\n",
- lfsck_lfsck2name(lfsck));
-
- GOTO(out, rc = -EPERM);
- }
-
- rc1 = lfsck_stop_all(env, lfsck, stop);
+ if (stop && stop->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
+ CERROR("%s: only allow to specify '-A' via MDS\n",
+ lfsck_lfsck2name(lfsck));
+ GOTO(put, rc = -EPERM);
}
- mutex_lock(&lfsck->li_mutex);
spin_lock(&lfsck->li_lock);
- /* no error if LFSCK is already stopped, or was never started */
- if (thread_is_init(thread) || thread_is_stopped(thread)) {
- spin_unlock(&lfsck->li_lock);
- GOTO(out, rc = 0);
- }
+ /* The target is umounted */
+ if (stop && stop->ls_status == LS_PAUSED)
+ lfsck->li_stopping = 1;
+
+ if (thread_is_init(thread) || thread_is_stopped(thread))
+ /* no error if LFSCK stopped already, or not started */
+ GOTO(unlock, rc = 0);
+
+ if (thread_is_stopping(thread))
+ /* Someone is stopping LFSCK. */
+ GOTO(unlock, rc = -EINPROGRESS);
- if (stop != NULL) {
+ if (stop) {
lfsck->li_status = stop->ls_status;
lfsck->li_flags = stop->ls_flags;
} else {
}
thread_set_flags(thread, SVC_STOPPING);
- spin_unlock(&lfsck->li_lock);
+
+ LASSERT(lfsck->li_task != NULL);
+ force_sig(SIGINT, lfsck->li_task);
+
+ if (lfsck->li_master) {
+ struct lfsck_component *com;
+ struct lfsck_assistant_data *lad;
+
+ list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
+ lad = com->lc_data;
+ spin_lock(&lad->lad_lock);
+ if (lad->lad_task != NULL)
+ force_sig(SIGINT, lad->lad_task);
+ spin_unlock(&lad->lad_lock);
+ }
+
+ list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
+ lad = com->lc_data;
+ spin_lock(&lad->lad_lock);
+ if (lad->lad_task != NULL)
+ force_sig(SIGINT, lad->lad_task);
+ spin_unlock(&lad->lad_lock);
+ }
+ }
wake_up_all(&thread->t_ctl_waitq);
+ spin_unlock(&lfsck->li_lock);
+ if (stop && stop->ls_flags & LPF_BROADCAST)
+ rc1 = lfsck_stop_all(env, lfsck, stop);
+
+ /* It was me set the status as 'stopping' just now, if it is not
+ * 'stopping' now, then either stopped, or re-started by race. */
l_wait_event(thread->t_ctl_waitq,
- thread_is_stopped(thread),
+ !thread_is_stopping(thread),
&lwi);
- GOTO(out, rc = 0);
+ GOTO(put, rc = 0);
-out:
- mutex_unlock(&lfsck->li_mutex);
+unlock:
+ spin_unlock(&lfsck->li_lock);
+put:
lfsck_instance_put(env, lfsck);
return rc != 0 ? rc : rc1;
}
EXPORT_SYMBOL(lfsck_stop);
+int lfsck_in_notify_local(const struct lu_env *env, struct dt_device *key,
+ struct lfsck_req_local *lrl, struct thandle *th)
+{
+ struct lfsck_instance *lfsck;
+ struct lfsck_component *com;
+ int rc = -EOPNOTSUPP;
+ ENTRY;
+
+ lfsck = lfsck_instance_find(key, true, false);
+ if (unlikely(!lfsck))
+ RETURN(-ENXIO);
+
+ com = lfsck_component_find(lfsck, lrl->lrl_active);
+ if (likely(com && com->lc_ops->lfsck_in_notify_local)) {
+ rc = com->lc_ops->lfsck_in_notify_local(env, com, lrl, th);
+ lfsck_component_put(env, com);
+ }
+
+ lfsck_instance_put(env, lfsck);
+
+ RETURN(rc);
+}
+EXPORT_SYMBOL(lfsck_in_notify_local);
+
int lfsck_in_notify(const struct lu_env *env, struct dt_device *key,
- struct lfsck_request *lr, struct thandle *th)
+ struct lfsck_request *lr)
{
int rc = -EOPNOTSUPP;
ENTRY;
}
case LE_PHASE1_DONE:
case LE_PHASE2_DONE:
- case LE_FID_ACCESSED:
case LE_PEER_EXIT:
case LE_CONDITIONAL_DESTROY:
- case LE_SKIP_NLINK_DECLARE:
- case LE_SKIP_NLINK:
case LE_SET_LMV_MASTER:
case LE_SET_LMV_SLAVE:
case LE_PAIRS_VERIFY: {
RETURN(-ENXIO);
com = lfsck_component_find(lfsck, lr->lr_active);
- if (likely(com != NULL)) {
- rc = com->lc_ops->lfsck_in_notify(env, com, lr, th);
+ if (likely(com)) {
+ rc = com->lc_ops->lfsck_in_notify(env, com, lr);
lfsck_component_put(env, com);
}
EXPORT_SYMBOL(lfsck_in_notify);
int lfsck_query(const struct lu_env *env, struct dt_device *key,
- struct lfsck_request *lr)
+ struct lfsck_request *req, struct lfsck_reply *rep,
+ struct lfsck_query *que)
{
struct lfsck_instance *lfsck;
struct lfsck_component *com;
- int rc;
+ int i;
+ int rc = 0;
+ __u16 type;
ENTRY;
lfsck = lfsck_instance_find(key, true, false);
if (unlikely(lfsck == NULL))
RETURN(-ENXIO);
- com = lfsck_component_find(lfsck, lr->lr_active);
- if (likely(com != NULL)) {
- rc = com->lc_ops->lfsck_query(env, com);
- lfsck_component_put(env, com);
+ if (que != NULL) {
+ if (que->lu_types == LFSCK_TYPES_ALL)
+ que->lu_types =
+ LFSCK_TYPES_SUPPORTED & ~LFSCK_TYPE_SCRUB;
+
+ if (que->lu_types & ~LFSCK_TYPES_SUPPORTED) {
+ que->lu_types &= ~LFSCK_TYPES_SUPPORTED;
+
+ GOTO(out, rc = -ENOTSUPP);
+ }
+
+ for (i = 0, type = 1 << i; i < LFSCK_TYPE_BITS;
+ i++, type = 1 << i) {
+ if (!(que->lu_types & type))
+ continue;
+
+again:
+ com = lfsck_component_find(lfsck, type);
+ if (unlikely(com == NULL))
+ GOTO(out, rc = -ENOTSUPP);
+
+ memset(que->lu_mdts_count[i], 0,
+ sizeof(__u32) * (LS_MAX + 1));
+ memset(que->lu_osts_count[i], 0,
+ sizeof(__u32) * (LS_MAX + 1));
+ que->lu_repaired[i] = 0;
+ rc = com->lc_ops->lfsck_query(env, com, req, rep,
+ que, i);
+ lfsck_component_put(env, com);
+ if (rc < 0)
+ GOTO(out, rc);
+ }
+
+ if (!(que->lu_flags & LPF_WAIT))
+ GOTO(out, rc);
+
+ for (i = 0, type = 1 << i; i < LFSCK_TYPE_BITS;
+ i++, type = 1 << i) {
+ if (!(que->lu_types & type))
+ continue;
+
+ if (que->lu_mdts_count[i][LS_SCANNING_PHASE1] != 0 ||
+ que->lu_mdts_count[i][LS_SCANNING_PHASE2] != 0 ||
+ que->lu_osts_count[i][LS_SCANNING_PHASE1] != 0 ||
+ que->lu_osts_count[i][LS_SCANNING_PHASE2] != 0) {
+ struct l_wait_info lwi;
+
+ /* If it is required to wait, then sleep
+ * 3 seconds and try to query again. */
+ lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(3),
+ NULL,
+ LWI_ON_SIGNAL_NOOP,
+ NULL);
+ rc = l_wait_event(lfsck->li_thread.t_ctl_waitq,
+ 0, &lwi);
+ if (rc == -ETIMEDOUT)
+ goto again;
+ }
+ }
} else {
- rc = -ENOTSUPP;
+ com = lfsck_component_find(lfsck, req->lr_active);
+ if (likely(com != NULL)) {
+ rc = com->lc_ops->lfsck_query(env, com, req, rep,
+ que, -1);
+ lfsck_component_put(env, com);
+ } else {
+ rc = -ENOTSUPP;
+ }
}
- lfsck_instance_put(env, lfsck);
+ GOTO(out, rc);
- RETURN(rc);
+out:
+ lfsck_instance_put(env, lfsck);
+ return rc;
}
+EXPORT_SYMBOL(lfsck_query);
int lfsck_register_namespace(const struct lu_env *env, struct dt_device *key,
struct ldlm_namespace *ns)
rc = dt_lookup(env, root,
(struct dt_rec *)(&lfsck->li_global_root_fid),
- (const struct dt_key *)"ROOT", BYPASS_CAPA);
+ (const struct dt_key *)"ROOT");
if (rc != 0)
GOTO(out, rc);
GOTO(out, rc = -ENOTDIR);
rc = dt_lookup(env, obj, (struct dt_rec *)fid,
- (const struct dt_key *)dotlustre, BYPASS_CAPA);
+ (const struct dt_key *)dotlustre);
if (rc != 0)
GOTO(out, rc);
*pfid = *fid;
rc = dt_lookup(env, obj, (struct dt_rec *)fid,
- (const struct dt_key *)lostfound,
- BYPASS_CAPA);
+ (const struct dt_key *)lostfound);
if (rc != 0)
GOTO(out, rc);
if (unlikely(index >= ltds->ltd_tgts_bitmap->size))
goto unlock;
- ltd = LTD_TGT(ltds, index);
+ ltd = lfsck_ltd2tgt(ltds, index);
if (unlikely(ltd == NULL))
goto unlock;
ltds->ltd_tgtnr--;
cfs_bitmap_clear(ltds->ltd_tgts_bitmap, index);
- LTD_TGT(ltds, index) = NULL;
+ lfsck_assign_tgt(ltds, NULL, index);
unlock:
if (ltd == NULL) {
INIT_LIST_HEAD(&lfsck_mdt_orphan_list);
lfsck_key_init_generic(&lfsck_thread_key, NULL);
rc = lu_context_key_register(&lfsck_thread_key);
- if (rc == 0) {
+ if (!rc) {
+ tgt_register_lfsck_in_notify_local(lfsck_in_notify_local);
tgt_register_lfsck_in_notify(lfsck_in_notify);
tgt_register_lfsck_query(lfsck_query);
}
lu_context_key_degister(&lfsck_thread_key);
}
-MODULE_AUTHOR("Intel Corporation <http://www.intel.com/>");
-MODULE_DESCRIPTION("LFSCK");
+MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
+MODULE_DESCRIPTION("Lustre File System Checker");
+MODULE_VERSION(LUSTRE_VERSION_STRING);
MODULE_LICENSE("GPL");
-cfs_module(lfsck, LUSTRE_VERSION_STRING, lfsck_init, lfsck_exit);
+module_init(lfsck_init);
+module_exit(lfsck_exit);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff