*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2015, Intel Corporation.
+ * Copyright (c) 2011, 2016, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
struct osd_inode_id *id)
{
+ int rc;
struct inode *inode = NULL;
/* if we look for an inode withing a running
LDISKFS_SB(osd_sb(dev))->s_es->s_volume_name, id->oii_ino);
iput(inode);
inode = ERR_PTR(-ENOENT);
+ } else if ((rc = osd_attach_jinode(inode))) {
+ iput(inode);
+ inode = ERR_PTR(rc);
} else {
ldiskfs_clear_inode_state(inode, LDISKFS_STATE_LUSTRE_DESTROY);
if (id->oii_gen == OSD_OII_NOGEN)
return inode;
}
-int osd_ldiskfs_add_entry(struct osd_thread_info *info,
+int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
handle_t *handle, struct dentry *child,
struct inode *inode, struct htree_lock *hlock)
{
rc = __ldiskfs_add_entry(handle, child, inode, hlock);
if (rc == -ENOBUFS || rc == -ENOSPC) {
- char fidbuf[FID_LEN + 1];
- struct lustre_mdt_attrs lma;
- struct lu_fid fid = { };
- char *errstr;
- struct dentry *p_dentry = child->d_parent;
-
- rc2 = osd_get_lma(info, p_dentry->d_inode, p_dentry,
- &lma);
+ struct lustre_mdt_attrs *lma = &info->oti_mdt_attrs;
+ struct inode *parent = child->d_parent->d_inode;
+ struct lu_fid *fid = NULL;
+
+ rc2 = osd_get_lma(info, parent, child->d_parent, lma);
if (rc2 == 0) {
- fid = lma.lma_self_fid;
- snprintf(fidbuf, sizeof(fidbuf), DFID, PFID(&fid));
+ fid = &lma->lma_self_fid;
} else if (rc2 == -ENODATA) {
- if (unlikely(p_dentry->d_inode ==
- inode->i_sb->s_root->d_inode))
- lu_local_obj_fid(&fid, OSD_FS_ROOT_OID);
- else if (info->oti_dev && !info->oti_dev->od_is_ost &&
- fid_seq_is_mdt0(fid_seq(&fid)))
- lu_igif_build(&fid, p_dentry->d_inode->i_ino,
- p_dentry->d_inode->i_generation);
- snprintf(fidbuf, sizeof(fidbuf), DFID, PFID(&fid));
- } else {
- snprintf(fidbuf, FID_LEN, "%s", "unknown");
+ if (unlikely(parent == inode->i_sb->s_root->d_inode)) {
+ fid = &info->oti_fid3;
+ lu_local_obj_fid(fid, OSD_FS_ROOT_OID);
+ } else if (!osd->od_is_ost && osd->od_index == 0) {
+ fid = &info->oti_fid3;
+ lu_igif_build(fid, parent->i_ino,
+ parent->i_generation);
+ }
}
- if (rc == -ENOSPC)
- errstr = "has reached";
+ if (fid != NULL)
+ CWARN("%s: directory (inode: %lu, FID: "DFID") %s "
+ "maximum entry limit\n",
+ osd_name(osd), parent->i_ino, PFID(fid),
+ rc == -ENOSPC ? "has reached" : "is approaching");
else
- errstr = "is approaching";
- CWARN("%.16s: directory (inode: %lu FID: %s) %s maximum entry limit\n",
- LDISKFS_SB(inode->i_sb)->s_es->s_volume_name,
- p_dentry->d_inode->i_ino, fidbuf, errstr);
+ CWARN("%s: directory (inode: %lu, FID: unknown) %s "
+ "maximum entry limit\n",
+ osd_name(osd), parent->i_ino,
+ rc == -ENOSPC ? "has reached" : "is approaching");
+
/* ignore such error now */
if (rc == -ENOBUFS)
rc = 0;
}
+
return rc;
}
(flags & SS_AUTO_PARTIAL || sf->sf_status == SS_SCANNING))
osd_check_lmv(info, dev, inode, oic);
+ result = osd_attach_jinode(inode);
+ if (result) {
+ obj->oo_inode = NULL;
+ iput(inode);
+ GOTO(out, result);
+ }
+
if (!ldiskfs_pdo)
GOTO(out, result = 0);
*/
static void osd_th_alloced(struct osd_thandle *oth)
{
- oth->oth_alloced = cfs_time_current();
+ oth->oth_alloced = ktime_get();
}
/**
*/
static void osd_th_started(struct osd_thandle *oth)
{
- oth->oth_started = cfs_time_current();
+ oth->oth_started = ktime_get();
}
/**
- * Helper function to convert time interval to microseconds packed in
- * long int.
+ * Check whether the we deal with this handle for too long.
*/
-static long interval_to_usec(cfs_time_t start, cfs_time_t end)
+static void __osd_th_check_slow(void *oth, struct osd_device *dev,
+ ktime_t alloced, ktime_t started,
+ ktime_t closed)
{
- struct timeval val;
+ ktime_t now = ktime_get();
- cfs_duration_usec(cfs_time_sub(end, start), &val);
- return val.tv_sec * 1000000 + val.tv_usec;
-}
+ LASSERT(dev != NULL);
-/**
- * Check whether the we deal with this handle for too long.
- */
-static void __osd_th_check_slow(void *oth, struct osd_device *dev,
- cfs_time_t alloced, cfs_time_t started,
- cfs_time_t closed)
-{
- cfs_time_t now = cfs_time_current();
-
- LASSERT(dev != NULL);
-
- lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_STARTING,
- interval_to_usec(alloced, started));
- lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_OPEN,
- interval_to_usec(started, closed));
- lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_CLOSING,
- interval_to_usec(closed, now));
-
- if (cfs_time_before(cfs_time_add(alloced, cfs_time_seconds(30)), now)) {
- CWARN("transaction handle %p was open for too long: "
- "now "CFS_TIME_T" ,"
- "alloced "CFS_TIME_T" ,"
- "started "CFS_TIME_T" ,"
- "closed "CFS_TIME_T"\n",
+ lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_STARTING,
+ ktime_us_delta(started, alloced));
+ lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_OPEN,
+ ktime_us_delta(closed, started));
+ lprocfs_counter_add(dev->od_stats, LPROC_OSD_THANDLE_CLOSING,
+ ktime_us_delta(now, closed));
+
+ if (ktime_before(ktime_add_ns(alloced, 30 * NSEC_PER_SEC), now)) {
+ CWARN("transaction handle %p was open for too long: now %lld, alloced %lld, started %lld, closed %lld\n",
oth, now, alloced, started, closed);
libcfs_debug_dumpstack(NULL);
}
}
-#define OSD_CHECK_SLOW_TH(oth, dev, expr) \
-{ \
- cfs_time_t __closed = cfs_time_current(); \
- cfs_time_t __alloced = oth->oth_alloced; \
- cfs_time_t __started = oth->oth_started; \
- \
- expr; \
- __osd_th_check_slow(oth, dev, __alloced, __started, __closed); \
+#define OSD_CHECK_SLOW_TH(oth, dev, expr) \
+{ \
+ ktime_t __closed = ktime_get(); \
+ ktime_t __alloced = oth->oth_alloced; \
+ ktime_t __started = oth->oth_started; \
+ \
+ expr; \
+ __osd_th_check_slow(oth, dev, __alloced, __started, __closed); \
}
#else /* OSD_THANDLE_STATS */
oh->ot_quota_trans = NULL;
if (oh->ot_handle != NULL) {
+ int rc2;
handle_t *hdl = oh->ot_handle;
/*
hdl->h_sync = th->th_sync;
oh->ot_handle = NULL;
- OSD_CHECK_SLOW_TH(oh, osd, rc = ldiskfs_journal_stop(hdl));
- if (rc != 0)
+ OSD_CHECK_SLOW_TH(oh, osd, rc2 = ldiskfs_journal_stop(hdl));
+ if (rc2 != 0)
CERROR("%s: failed to stop transaction: rc = %d\n",
- osd_name(osd), rc);
+ osd_name(osd), rc2);
+ if (!rc)
+ rc = rc2;
} else {
osd_trans_stop_cb(oh, th->th_result);
OBD_FREE_PTR(oh);
statfs_pack(sfs, ksfs);
if (unlikely(sb->s_flags & MS_RDONLY))
- sfs->os_state = OS_STATE_READONLY;
+ sfs->os_state |= OS_STATE_READONLY;
if (LDISKFS_HAS_INCOMPAT_FEATURE(sb,
LDISKFS_FEATURE_INCOMPAT_EXTENTS))
sfs->os_maxbytes = sb->s_maxbytes;
};
static int osd_mkdir(struct osd_thread_info *info, struct osd_object *obj,
- struct lu_attr *attr,
- struct dt_allocation_hint *hint,
- struct dt_object_format *dof,
- struct thandle *th)
+ struct lu_attr *attr,
+ struct dt_allocation_hint *hint,
+ struct dt_object_format *dof,
+ struct thandle *th)
{
- int result;
- struct osd_thandle *oth;
- __u32 mode = (attr->la_mode & (S_IFMT | S_IRWXUGO | S_ISVTX));
+ int result;
+ struct osd_thandle *oth;
+ __u32 mode = (attr->la_mode & (S_IFMT | S_IRWXUGO | S_ISVTX | S_ISGID));
- LASSERT(S_ISDIR(attr->la_mode));
+ LASSERT(S_ISDIR(attr->la_mode));
- oth = container_of(th, struct osd_thandle, ot_super);
- LASSERT(oth->ot_handle->h_transaction != NULL);
- result = osd_mkfile(info, obj, mode, hint, th);
+ oth = container_of(th, struct osd_thandle, ot_super);
+ LASSERT(oth->ot_handle->h_transaction != NULL);
+ result = osd_mkfile(info, obj, mode, hint, th);
- return result;
+ return result;
}
static int osd_mk_index(struct osd_thread_info *info, struct osd_object *obj,
/* it will check/delete the inode from remote parent,
* how to optimize it? unlink performance impaction XXX */
result = osd_delete_from_remote_parent(env, osd, obj, oh);
- if (result != 0 && result != -ENOENT) {
+ if (result != 0)
CERROR("%s: delete inode "DFID": rc = %d\n",
osd_name(osd), PFID(fid), result);
- }
+
spin_lock(&obj->oo_guard);
clear_nlink(inode);
spin_unlock(&obj->oo_guard);
RETURN(local);
}
+ /* restore i_gid in case S_ISGID is set, we will inherit S_ISGID and set
+ * correct gid on remote file, not agent here */
+ local->i_gid = current_fsgid();
ldiskfs_set_inode_state(local, LDISKFS_STATE_LUSTRE_NOSCRUB);
unlock_new_inode(local);
if (strcmp(name, XATTR_NAME_LMV) == 0) {
struct lustre_mdt_attrs *lma = &info->oti_mdt_attrs;
- int rc;
rc = osd_get_lma(info, inode, &info->oti_obj_dentry, lma);
if (rc != 0)
RETURN(rc);
}
- if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_LINKEA_OVERFLOW) &&
- strcmp(name, XATTR_NAME_LINK) == 0)
- return -ENOSPC;
-
rc = __osd_xattr_set(info, inode, name, buf->lb_buf, buf->lb_len,
fs_flags);
osd_trans_exec_check(env, handle, OSD_OT_XATTR_SET);
{
struct osd_thandle *oh;
struct inode *inode;
- int rc;
+ int rc, credits;
ENTRY;
LASSERT(!dt_object_remote(dt));
oh = container_of0(handle, struct osd_thandle, ot_super);
LASSERT(oh->ot_handle == NULL);
- /* due to DNE we may need to remove an agent inode */
- osd_trans_declare_op(env, oh, OSD_OT_DELETE,
- osd_dto_credits_noquota[DTO_INDEX_DELETE] +
- osd_dto_credits_noquota[DTO_OBJECT_DELETE]);
+ credits = osd_dto_credits_noquota[DTO_INDEX_DELETE];
+ if (key != NULL && unlikely(strcmp((char *)key, dotdot) == 0)) {
+ /* '..' to a remote object has a local representative */
+ credits += osd_dto_credits_noquota[DTO_INDEX_DELETE];
+ /* to reset LMAI_REMOTE_PARENT */
+ credits += 1;
+ }
+ osd_trans_declare_op(env, oh, OSD_OT_DELETE, credits);
inode = osd_dt_obj(dt)->oo_inode;
if (inode == NULL)
* /Agent directory, Check whether it needs to delete
* from agent directory */
if (unlikely(strcmp((char *)key, dotdot) == 0)) {
- rc = osd_delete_from_remote_parent(env, osd_obj2dev(obj), obj,
- oh);
- if (rc != 0 && rc != -ENOENT) {
- CERROR("%s: delete agent inode "DFID": rc = %d\n",
- osd_name(osd), PFID(fid), rc);
- }
-
- if (rc == -ENOENT)
- rc = 0;
-
- GOTO(out, rc);
+ int ret;
+
+ ret = osd_delete_from_remote_parent(env, osd_obj2dev(obj),
+ obj, oh);
+ if (ret != 0)
+ /* Sigh, the entry has been deleted, and
+ * it is not easy to revert it back, so
+ * let's keep this error private, and let
+ * LFSCK fix it. XXX */
+ CERROR("%s: delete remote parent "DFID": rc = %d\n",
+ osd_name(osd), PFID(fid), ret);
}
out:
-
LASSERT(osd_invariant(obj));
osd_trans_exec_check(env, handle, OSD_OT_DELETE);
RETURN(rc);
child = osd_child_dentry_get(info->oti_env, pobj, name, strlen(name));
child->d_fsdata = (void *)ldp;
ll_vfs_dq_init(pobj->oo_inode);
- rc = osd_ldiskfs_add_entry(info, oth->ot_handle, child,
- cinode, hlock);
+ rc = osd_ldiskfs_add_entry(info, osd_obj2dev(pobj), oth->ot_handle,
+ child, cinode, hlock);
if (rc == 0 && OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_TYPE)) {
struct ldiskfs_dir_entry_2 *de;
struct buffer_head *bh;
}
ldata.ld_buf = buf;
- rc = linkea_init(&ldata);
- if (rc == 0) {
+ rc = linkea_init_with_rec(&ldata);
+ if (!rc) {
linkea_first_entry(&ldata);
linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, NULL, fid);
}
RETURN(rc);
}
+static int osd_verify_ent_by_linkea(const struct lu_env *env,
+ struct inode *inode,
+ const struct lu_fid *pfid,
+ const char *name, const int namelen)
+{
+ struct osd_thread_info *oti = osd_oti_get(env);
+ struct lu_buf *buf = &oti->oti_big_buf;
+ struct dentry *dentry = &oti->oti_obj_dentry;
+ struct linkea_data ldata = { NULL };
+ struct lu_name cname = { .ln_name = name,
+ .ln_namelen = namelen };
+ int rc;
+ ENTRY;
+
+again:
+ rc = __osd_xattr_get(inode, dentry, XATTR_NAME_LINK,
+ buf->lb_buf, buf->lb_len);
+ if (rc == -ERANGE)
+ rc = __osd_xattr_get(inode, dentry, XATTR_NAME_LINK, NULL, 0);
+
+ if (rc < 0)
+ RETURN(rc);
+
+ if (unlikely(rc == 0))
+ RETURN(-ENODATA);
+
+ if (buf->lb_len < rc) {
+ lu_buf_realloc(buf, rc);
+ if (buf->lb_buf == NULL)
+ RETURN(-ENOMEM);
+
+ goto again;
+ }
+
+ ldata.ld_buf = buf;
+ rc = linkea_init_with_rec(&ldata);
+ if (!rc)
+ rc = linkea_links_find(&ldata, &cname, pfid);
+
+ RETURN(rc);
+}
+
/**
* Calls ->lookup() to find dentry. From dentry get inode and
* read inode's ea to get fid. This is required for interoperability
return it->oie_dirent->oied_namelen;
}
-static inline bool
-osd_dot_dotdot_has_space(struct ldiskfs_dir_entry_2 *de, int dot_dotdot)
+static inline bool osd_dotdot_has_space(struct ldiskfs_dir_entry_2 *de)
{
- LASSERTF(dot_dotdot == 1 || dot_dotdot == 2,
- "dot_dotdot = %d\n", dot_dotdot);
-
if (LDISKFS_DIR_REC_LEN(de) >=
- __LDISKFS_DIR_REC_LEN(dot_dotdot + 1 + sizeof(struct osd_fid_pack)))
+ __LDISKFS_DIR_REC_LEN(2 + 1 + sizeof(struct osd_fid_pack)))
return true;
return false;
static inline bool
osd_dirent_has_space(struct ldiskfs_dir_entry_2 *de, __u16 namelen,
- unsigned blocksize, int dot_dotdot)
+ unsigned blocksize, bool dotdot)
{
- if (dot_dotdot > 0)
- return osd_dot_dotdot_has_space(de, dot_dotdot);
+ if (dotdot)
+ return osd_dotdot_has_space(de);
if (ldiskfs_rec_len_from_disk(de->rec_len, blocksize) >=
__LDISKFS_DIR_REC_LEN(namelen + 1 + sizeof(struct osd_fid_pack)))
}
static int
-osd_dirent_reinsert(const struct lu_env *env, handle_t *jh,
- struct dentry *dentry, const struct lu_fid *fid,
- struct buffer_head *bh, struct ldiskfs_dir_entry_2 *de,
- struct htree_lock *hlock, int dot_dotdot)
+osd_dirent_reinsert(const struct lu_env *env, struct osd_device *dev,
+ handle_t *jh, struct dentry *dentry,
+ const struct lu_fid *fid, struct buffer_head *bh,
+ struct ldiskfs_dir_entry_2 *de, struct htree_lock *hlock,
+ bool dotdot)
{
struct inode *dir = dentry->d_parent->d_inode;
struct inode *inode = dentry->d_inode;
RETURN(0);
/* There is enough space to hold the FID-in-dirent. */
- if (osd_dirent_has_space(de, namelen, dir->i_sb->s_blocksize,
- dot_dotdot)) {
+ if (osd_dirent_has_space(de, namelen, dir->i_sb->s_blocksize, dotdot)) {
rc = ldiskfs_journal_get_write_access(jh, bh);
if (rc != 0)
RETURN(rc);
RETURN(rc);
}
- LASSERTF(dot_dotdot == 0, "dot_dotdot = %d\n", dot_dotdot);
+ LASSERT(!dotdot);
rc = ldiskfs_delete_entry(jh, dir, de, bh);
if (rc != 0)
osd_get_ldiskfs_dirent_param(ldp, fid);
dentry->d_fsdata = (void *)ldp;
ll_vfs_dq_init(dir);
- rc = osd_ldiskfs_add_entry(info, jh, dentry, inode, hlock);
+ rc = osd_ldiskfs_add_entry(info, dev, jh, dentry, inode, hlock);
/* It is too bad, we cannot reinsert the name entry back.
* That means we lose it! */
if (rc != 0)
struct lustre_mdt_attrs *lma = &info->oti_mdt_attrs;
struct osd_device *dev = osd_obj2dev(obj);
struct super_block *sb = osd_sb(dev);
- const char *devname =
- LDISKFS_SB(sb)->s_es->s_volume_name;
+ const char *devname = osd_name(dev);
struct osd_it_ea_dirent *ent = it->oie_dirent;
struct inode *dir = obj->oo_inode;
struct htree_lock *hlock = NULL;
struct ldiskfs_dir_entry_2 *de;
struct dentry *dentry;
struct inode *inode;
+ const struct lu_fid *pfid = lu_object_fid(&obj->oo_dt.do_lu);
int credits;
int rc;
- int dot_dotdot = 0;
+ bool dotdot = false;
bool dirty = false;
ENTRY;
+ if (ent->oied_name[0] == '.') {
+ if (ent->oied_namelen == 1)
+ RETURN(0);
+
+ if (ent->oied_namelen == 2 && ent->oied_name[1] == '.')
+ dotdot = true;
+ }
+
osd_id_gen(id, ent->oied_ino, OSD_OII_NOGEN);
inode = osd_iget(info, dev, id);
if (IS_ERR(inode)) {
rc = PTR_ERR(inode);
if (rc == -ENOENT || rc == -ESTALE) {
+ /* Maybe dangling name entry, or
+ * corrupted directory entry. */
*attr |= LUDA_UNKNOWN;
rc = 0;
} else {
- CDEBUG(D_LFSCK, "%.16s: fail to iget for dirent "
- "check_repair, dir = %lu/%u, name = %.*s: "
- "rc = %d\n",
+ CDEBUG(D_LFSCK, "%s: fail to iget() for dirent "
+ "check_repair, dir = %lu/%u, name = %.*s, "
+ "ino = %llu, rc = %d\n",
devname, dir->i_ino, dir->i_generation,
- ent->oied_namelen, ent->oied_name, rc);
+ ent->oied_namelen, ent->oied_name,
+ ent->oied_ino, rc);
}
RETURN(rc);
dentry = osd_child_dentry_by_inode(env, dir, ent->oied_name,
ent->oied_namelen);
rc = osd_get_lma(info, inode, dentry, lma);
- if (rc == -ENODATA)
+ if (rc == -ENODATA || !fid_is_sane(&lma->lma_self_fid))
lma = NULL;
else if (rc != 0)
GOTO(out, rc);
- if (ent->oied_name[0] == '.') {
- if (ent->oied_namelen == 1)
- dot_dotdot = 1;
- else if (ent->oied_namelen == 2 && ent->oied_name[1] == '.')
- dot_dotdot = 2;
- }
-
/* We need to ensure that the name entry is still valid.
* Because it may be removed or renamed by other already.
*
jh = osd_journal_start_sb(sb, LDISKFS_HT_MISC, credits);
if (IS_ERR(jh)) {
rc = PTR_ERR(jh);
- CDEBUG(D_LFSCK, "%.16s: fail to start trans for dirent "
+ CDEBUG(D_LFSCK, "%s: fail to start trans for dirent "
"check_repair, dir = %lu/%u, credits = %d, "
- "name = %.*s: rc = %d\n",
+ "name = %.*s, ino = %llu: rc = %d\n",
devname, dir->i_ino, dir->i_generation, credits,
- ent->oied_namelen, ent->oied_name, rc);
+ ent->oied_namelen, ent->oied_name,
+ ent->oied_ino, rc);
GOTO(out_inode, rc);
}
}
bh = osd_ldiskfs_find_entry(dir, &dentry->d_name, &de, NULL, hlock);
- /* For dot/dotdot entry, if there is not enough space to hold the
+ if (IS_ERR(bh) || le32_to_cpu(de->inode) != inode->i_ino) {
+ *attr |= LUDA_IGNORE;
+
+ GOTO(out, rc = 0);
+ }
+
+ /* For dotdot entry, if there is not enough space to hold the
* FID-in-dirent, just keep them there. It only happens when the
* device upgraded from 1.8 or restored from MDT file-level backup.
- * For the whole directory, only dot/dotdot entry have no FID-in-dirent
+ * For the whole directory, only dotdot entry have no FID-in-dirent
* and needs to get FID from LMA when readdir, it will not affect the
* performance much. */
- if (IS_ERR(bh) || (le32_to_cpu(de->inode) != inode->i_ino) ||
- (dot_dotdot != 0 && !osd_dot_dotdot_has_space(de, dot_dotdot))) {
- *attr |= LUDA_IGNORE;
+ if (dotdot && !osd_dotdot_has_space(de)) {
+ *attr |= LUDA_UNKNOWN;
GOTO(out, rc = 0);
}
if (lma != NULL) {
+ if (lu_fid_eq(fid, &lma->lma_self_fid))
+ GOTO(out, rc = 0);
+
if (unlikely(lma->lma_compat & LMAC_NOT_IN_OI)) {
struct lu_fid *tfid = &lma->lma_self_fid;
- *attr |= LUDA_IGNORE;
- /* It must be REMOTE_PARENT_DIR and as the
- * dotdot entry of remote directory */
- if (unlikely(dot_dotdot != 2 ||
- fid_seq(tfid) != FID_SEQ_LOCAL_FILE ||
- fid_oid(tfid) != REMOTE_PARENT_DIR_OID)) {
- CDEBUG(D_LFSCK, "%.16s: expect remote agent "
+ if (likely(dotdot &&
+ fid_seq(tfid) == FID_SEQ_LOCAL_FILE &&
+ fid_oid(tfid) == REMOTE_PARENT_DIR_OID)) {
+ /* It must be REMOTE_PARENT_DIR and as the
+ * 'dotdot' entry of remote directory */
+ *attr |= LUDA_IGNORE;
+ } else {
+ CDEBUG(D_LFSCK, "%s: expect remote agent "
"parent directory, but got %.*s under "
"dir = %lu/%u with the FID "DFID"\n",
devname, ent->oied_namelen,
ent->oied_name, dir->i_ino,
dir->i_generation, PFID(tfid));
- GOTO(out, rc = -EIO);
+ *attr |= LUDA_UNKNOWN;
}
GOTO(out, rc = 0);
}
+ }
- if (fid_is_sane(fid)) {
- /* FID-in-dirent is valid. */
- if (lu_fid_eq(fid, &lma->lma_self_fid))
- GOTO(out, rc = 0);
-
- /* Do not repair under dryrun mode. */
- if (*attr & LUDA_VERIFY_DRYRUN) {
- *attr |= LUDA_REPAIR;
+ if (!fid_is_zero(fid)) {
+ rc = osd_verify_ent_by_linkea(env, inode, pfid, ent->oied_name,
+ ent->oied_namelen);
+ if (rc == -ENOENT ||
+ (rc == -ENODATA &&
+ !(dev->od_scrub.os_file.sf_flags & SF_UPGRADE))) {
+ /* linkEA does not recognize the dirent entry,
+ * it may because the dirent entry corruption
+ * and points to other's inode. */
+ CDEBUG(D_LFSCK, "%s: the target inode does not "
+ "recognize the dirent, dir = %lu/%u, "
+ " name = %.*s, ino = %llu, "
+ DFID": rc = %d\n", devname, dir->i_ino,
+ dir->i_generation, ent->oied_namelen,
+ ent->oied_name, ent->oied_ino, PFID(fid), rc);
+ *attr |= LUDA_UNKNOWN;
- GOTO(out, rc = 0);
- }
+ GOTO(out, rc = 0);
+ }
- if (jh == NULL) {
- brelse(bh);
- dev->od_dirent_journal = 1;
- if (hlock != NULL) {
- ldiskfs_htree_unlock(hlock);
- hlock = NULL;
- } else {
- up_read(&obj->oo_ext_idx_sem);
- }
+ if (rc && rc != -ENODATA) {
+ CDEBUG(D_LFSCK, "%s: fail to verify FID in the dirent, "
+ "dir = %lu/%u, name = %.*s, ino = %llu, "
+ DFID": rc = %d\n", devname, dir->i_ino,
+ dir->i_generation, ent->oied_namelen,
+ ent->oied_name, ent->oied_ino, PFID(fid), rc);
+ *attr |= LUDA_UNKNOWN;
- goto again;
- }
+ GOTO(out, rc = 0);
+ }
+ }
+ if (lma != NULL) {
+ /* linkEA recognizes the dirent entry, the FID-in-LMA is
+ * valid, trusted, in spite of fid_is_sane(fid) or not. */
+ if (*attr & LUDA_VERIFY_DRYRUN) {
*fid = lma->lma_self_fid;
- dirty = true;
- /* Update the FID-in-dirent. */
- rc = osd_dirent_reinsert(env, jh, dentry, fid, bh, de,
- hlock, dot_dotdot);
- if (rc == 0)
- *attr |= LUDA_REPAIR;
- else
- CDEBUG(D_LFSCK, "%.16s: fail to update FID "
- "in the dirent, dir = %lu/%u, "
- "name = %.*s, "DFID": rc = %d\n",
- devname, dir->i_ino, dir->i_generation,
- ent->oied_namelen, ent->oied_name,
- PFID(fid), rc);
- } else {
- /* Do not repair under dryrun mode. */
- if (*attr & LUDA_VERIFY_DRYRUN) {
- *fid = lma->lma_self_fid;
- *attr |= LUDA_REPAIR;
+ *attr |= LUDA_REPAIR;
- GOTO(out, rc = 0);
- }
-
- if (jh == NULL) {
- brelse(bh);
- dev->od_dirent_journal = 1;
- if (hlock != NULL) {
- ldiskfs_htree_unlock(hlock);
- hlock = NULL;
- } else {
- up_read(&obj->oo_ext_idx_sem);
- }
+ GOTO(out, rc = 0);
+ }
- goto again;
+ if (jh == NULL) {
+ brelse(bh);
+ dev->od_dirent_journal = 1;
+ if (hlock != NULL) {
+ ldiskfs_htree_unlock(hlock);
+ hlock = NULL;
+ } else {
+ up_read(&obj->oo_ext_idx_sem);
}
- *fid = lma->lma_self_fid;
- dirty = true;
- /* Append the FID-in-dirent. */
- rc = osd_dirent_reinsert(env, jh, dentry, fid, bh, de,
- hlock, dot_dotdot);
- if (rc == 0)
- *attr |= LUDA_REPAIR;
- else
- CDEBUG(D_LFSCK, "%.16s: fail to append FID "
- "after the dirent, dir = %lu/%u, "
- "name = %.*s, "DFID": rc = %d\n",
- devname, dir->i_ino, dir->i_generation,
- ent->oied_namelen, ent->oied_name,
- PFID(fid), rc);
+ goto again;
}
+
+ *fid = lma->lma_self_fid;
+ dirty = true;
+ /* Update or append the FID-in-dirent. */
+ rc = osd_dirent_reinsert(env, dev, jh, dentry, fid,
+ bh, de, hlock, dotdot);
+ if (rc == 0)
+ *attr |= LUDA_REPAIR;
+ else
+ CDEBUG(D_LFSCK, "%s: fail to re-insert FID after "
+ "the dirent, dir = %lu/%u, name = %.*s, "
+ "ino = %llu, "DFID": rc = %d\n",
+ devname, dir->i_ino, dir->i_generation,
+ ent->oied_namelen, ent->oied_name,
+ ent->oied_ino, PFID(fid), rc);
} else {
- /* Do not repair under dryrun mode. */
+ /* lma is NULL, trust the FID-in-dirent if it is valid. */
if (*attr & LUDA_VERIFY_DRYRUN) {
if (fid_is_sane(fid)) {
*attr |= LUDA_REPAIR;
- } else {
+ } else if (dev->od_index == 0) {
lu_igif_build(fid, inode->i_ino,
inode->i_generation);
*attr |= LUDA_UPGRADE;
if (rc == 0)
*attr |= LUDA_REPAIR;
else
- CDEBUG(D_LFSCK, "%.16s: fail to set LMA for "
+ CDEBUG(D_LFSCK, "%s: fail to set LMA for "
"update dirent, dir = %lu/%u, "
- "name = %.*s, "DFID": rc = %d\n",
+ "name = %.*s, ino = %llu, "
+ DFID": rc = %d\n",
devname, dir->i_ino, dir->i_generation,
ent->oied_namelen, ent->oied_name,
- PFID(fid), rc);
- } else {
+ ent->oied_ino, PFID(fid), rc);
+ } else if (dev->od_index == 0) {
lu_igif_build(fid, inode->i_ino, inode->i_generation);
/* It is probably IGIF object. Only aappend the
* FID-in-dirent. OI scrub will process FID-in-LMA. */
- rc = osd_dirent_reinsert(env, jh, dentry, fid, bh, de,
- hlock, dot_dotdot);
+ rc = osd_dirent_reinsert(env, dev, jh, dentry, fid,
+ bh, de, hlock, dotdot);
if (rc == 0)
*attr |= LUDA_UPGRADE;
else
- CDEBUG(D_LFSCK, "%.16s: fail to append IGIF "
+ CDEBUG(D_LFSCK, "%s: fail to append IGIF "
"after the dirent, dir = %lu/%u, "
- "name = %.*s, "DFID": rc = %d\n",
+ "name = %.*s, ino = %llu, "
+ DFID": rc = %d\n",
devname, dir->i_ino, dir->i_generation,
ent->oied_namelen, ent->oied_name,
- PFID(fid), rc);
+ ent->oied_ino, PFID(fid), rc);
}
}
&attr);
}
- if (!fid_is_sane(fid)) {
- attr &= ~LUDA_IGNORE;
+ if (!fid_is_sane(fid))
attr |= LUDA_UNKNOWN;
- }
} else {
attr &= ~LU_DIRENT_ATTRS_MASK;
if (!fid_is_sane(fid)) {
struct osd_thread_info *info = osd_oti_get(env);
struct lu_fid *fid = &info->oti_fid;
struct inode *inode;
- int rc = 0, force_over_256tb = 0;
+ int rc = 0, force_over_512tb = 0;
ENTRY;
if (o->od_mnt != NULL)
RETURN(-EINVAL);
}
#endif
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 53, 0)
if (opts != NULL && strstr(opts, "force_over_128tb") != NULL) {
- CWARN("force_over_128tb option is depricated."
- "Filesystems less then 256TB can be created without any"
- "force options. Use force_over_256tb option for"
- "filesystems greather then 256TB.\n");
+ CWARN("force_over_128tb option is deprecated. "
+ "Filesystems less than 512TB can be created without any "
+ "force options. Use force_over_512tb option for "
+ "filesystems greater than 512TB.\n");
}
+#endif
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 1, 53, 0)
+ if (opts != NULL && strstr(opts, "force_over_256tb") != NULL) {
+ CWARN("force_over_256tb option is deprecated. "
+ "Filesystems less than 512TB can be created without any "
+ "force options. Use force_over_512tb option for "
+ "filesystems greater than 512TB.\n");
+ }
+#endif
- if (opts != NULL && strstr(opts, "force_over_256tb") != NULL)
- force_over_256tb = 1;
+ if (opts != NULL && strstr(opts, "force_over_512tb") != NULL)
+ force_over_512tb = 1;
__page = alloc_page(GFP_KERNEL);
if (__page == NULL)
"noextents",
/* strip out option we processed in osd */
"bigendian_extents",
-#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(3,0,53,0)
-#warning "remove force_over_128 option"
-#else
- "force_over_128tb (deprecated)",
-#endif
+ "force_over_128tb",
"force_over_256tb",
+ "force_over_512tb",
NULL
};
strcat(options, opts);
/* Glom up mount options */
if (*options != '\0')
strcat(options, ",");
- strlcat(options, "no_mbcache", PAGE_SIZE);
+ strlcat(options, "no_mbcache,nodelalloc", PAGE_SIZE);
type = get_fs_type("ldiskfs");
if (!type) {
GOTO(out, rc);
}
- if (ldiskfs_blocks_count(LDISKFS_SB(osd_sb(o))->s_es) > (64ULL << 30) &&
- force_over_256tb == 0) {
+ if (ldiskfs_blocks_count(LDISKFS_SB(osd_sb(o))->s_es) <<
+ osd_sb(o)->s_blocksize_bits > 512ULL << 40 &&
+ force_over_512tb == 0) {
CERROR("%s: device %s LDISKFS does not support filesystems "
- "greater than 256TB and can cause data corruption. "
- "Use \"force_over_256tb\" mount option to override.\n",
+ "greater than 512TB and can cause data corruption. "
+ "Use \"force_over_512tb\" mount option to override.\n",
name, dev);
GOTO(out, rc = -EINVAL);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff