/* Start / stop recording */
static int
mdd_changelog_on(const struct lu_env *env, struct mdd_device *mdd, int on)
{
int rc = 0;
if ((on == 1) && ((mdd->mdd_cl.mc_flags & CLM_ON) == 0)) {
LCONSOLE_INFO("%s: changelog on\n", mdd2obd_dev(mdd)->obd_name);
if (mdd->mdd_cl.mc_flags & CLM_ERR) {
CERROR("Changelogs cannot be enabled due to error "
"condition (see %s log).\n",
mdd2obd_dev(mdd)->obd_name);
rc = -ESRCH;
} else {
spin_lock(&mdd->mdd_cl.mc_lock);
mdd->mdd_cl.mc_flags |= CLM_ON;
spin_unlock(&mdd->mdd_cl.mc_lock);
rc = mdd_changelog_write_header(env, mdd, CLM_START);
}
} else if ((on == 0) && ((mdd->mdd_cl.mc_flags & CLM_ON) == CLM_ON)) {
LCONSOLE_INFO("%s: changelog off\n",mdd2obd_dev(mdd)->obd_name);
rc = mdd_changelog_write_header(env, mdd, CLM_FINI);
spin_lock(&mdd->mdd_cl.mc_lock);
mdd->mdd_cl.mc_flags &= ~CLM_ON;
spin_unlock(&mdd->mdd_cl.mc_lock);
}
return rc;
}
int
kgnilnd_tunables_init()
{
int rc = 0;
#if CONFIG_SYSCTL && !CFS_SYSFS_MODULE_PARM
kgnilnd_tunables.kgn_sysctl =
cfs_register_sysctl_table(kgnilnd_top_ctl_table, 0);
if (kgnilnd_tunables.kgn_sysctl == NULL)
CWARN("Can't setup /proc tunables\n");
#endif
switch (*kgnilnd_tunables.kgn_checksum) {
default:
CERROR("Invalid checksum module parameter: %d\n",
*kgnilnd_tunables.kgn_checksum);
rc = -EINVAL;
GOTO(out, rc);
case GNILND_CHECKSUM_OFF:
/* no checksumming */
break;
case GNILND_CHECKSUM_SMSG_HEADER:
LCONSOLE_INFO("SMSG header only checksumming enabled\n");
break;
case GNILND_CHECKSUM_SMSG:
LCONSOLE_INFO("SMSG checksumming enabled\n");
break;
case GNILND_CHECKSUM_SMSG_BTE:
LCONSOLE_INFO("SMSG + BTE checksumming enabled\n");
break;
}
if (*kgnilnd_tunables.kgn_max_immediate > GNILND_MAX_IMMEDIATE) {
LCONSOLE_ERROR("kgnilnd module parameter 'max_immediate' too large %d > %d\n",
*kgnilnd_tunables.kgn_max_immediate, GNILND_MAX_IMMEDIATE);
rc = -EINVAL;
GOTO(out, rc);
}
if (*kgnilnd_tunables.kgn_mbox_per_block < 1) {
*kgnilnd_tunables.kgn_mbox_per_block = 1;
}
if (*kgnilnd_tunables.kgn_concurrent_sends == 0) {
*kgnilnd_tunables.kgn_concurrent_sends = *kgnilnd_tunables.kgn_peer_credits;
} else if (*kgnilnd_tunables.kgn_concurrent_sends > *kgnilnd_tunables.kgn_peer_credits) {
LCONSOLE_ERROR("kgnilnd parameter 'concurrent_sends' too large: %d > %d (peer_credits)\n",
*kgnilnd_tunables.kgn_concurrent_sends, *kgnilnd_tunables.kgn_peer_credits);
rc = -EINVAL;
}
out:
return rc;
}
static int __seq_server_alloc_super(struct lu_server_seq *seq,
struct lu_seq_range *out,
const struct lu_env *env)
{
struct lu_seq_range *space = &seq->lss_space;
int rc;
ENTRY;
LASSERT(range_is_sane(space));
if (range_is_exhausted(space)) {
CERROR("%s: Sequences space is exhausted\n",
seq->lss_name);
RETURN(-ENOSPC);
} else {
range_alloc(out, space, seq->lss_width);
}
rc = seq_store_update(env, seq, out, 1 /* sync */);
LCONSOLE_INFO("%s: super-sequence allocation rc = %d " DRANGE"\n",
seq->lss_name, rc, PRANGE(out));
RETURN(rc);
}
static int
proc_toggle_rdmaq_override(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int old_val = kgnilnd_sysctl.ksd_rdmaq_override;
int rc = 0;
ENTRY;
rc = proc_dointvec(table, write, buffer, lenp, ppos);
if (!write) {
/* read */
RETURN(rc);
}
if (kgnilnd_data.kgn_init != GNILND_INIT_ALL) {
rc = -EINVAL;
RETURN(rc);
}
if (old_val != kgnilnd_sysctl.ksd_rdmaq_override) {
long new_mb = kgnilnd_sysctl.ksd_rdmaq_override * (long)(1024*1024);
LCONSOLE_INFO("changing RDMAQ override to %d mbytes/sec\n",
kgnilnd_sysctl.ksd_rdmaq_override);
/* override proc is mbytes, but we calc in bytes */
kgnilnd_data.kgn_rdmaq_override = new_mb;
smp_wmb();
}
RETURN(rc);
}
int seq_server_alloc_spec(struct lu_server_seq *seq,
struct lu_seq_range *spec,
const struct lu_env *env)
{
struct lu_seq_range *space = &seq->lss_space;
int rc = -ENOSPC;
ENTRY;
/*
* In some cases (like recovery after a disaster)
* we may need to allocate sequences manually
* Notice some sequences can be lost if requested
* range doesn't start at the beginning of current
* free space. Also notice it's not possible now
* to allocate sequences out of natural order.
*/
if (spec->lsr_start >= spec->lsr_end)
RETURN(-EINVAL);
if (spec->lsr_flags != LU_SEQ_RANGE_MDT &&
spec->lsr_flags != LU_SEQ_RANGE_OST)
RETURN(-EINVAL);
mutex_lock(&seq->lss_mutex);
if (spec->lsr_start >= space->lsr_start) {
space->lsr_start = spec->lsr_end;
rc = seq_store_update(env, seq, spec, 1 /* sync */);
LCONSOLE_INFO("%s: "DRANGE" sequences allocated: rc = %d \n",
seq->lss_name, PRANGE(spec), rc);
}
mutex_unlock(&seq->lss_mutex);
RETURN(rc);
}
static int mdd_fill_fldb(const struct lu_env *env, struct mdd_device *mdd)
{
struct seq_server_site *ss = mdd_seq_site(mdd);
struct lu_seq_range range;
int rc;
LASSERT(ss->ss_server_seq != NULL);
LASSERT(ss->ss_server_fld != NULL);
if (ss->ss_server_seq->lss_space.lsr_end == 0)
return 0;
memcpy(&range, &ss->ss_server_seq->lss_space, sizeof(range));
/* Pre-existing ZFS does not insert any entries to FLDB, we need
* to insert it to FLDB during convertion */
range.lsr_start = FID_SEQ_NORMAL;
fld_range_set_mdt(&range);
mutex_lock(&ss->ss_server_fld->lsf_lock);
rc = fld_insert_entry(env, ss->ss_server_fld, &range);
mutex_unlock(&ss->ss_server_fld->lsf_lock);
LCONSOLE_INFO("%s: insert missing range "DRANGE"\n",
mdd2obd_dev(mdd)->obd_name, PRANGE(&range));
return rc;
}
static int LL_PROC_PROTO(proc_toggle_rdmaq_override)
{
int old_val = kgnilnd_sysctl.ksd_rdmaq_override;
int rc = 0;
ENTRY;
rc = ll_proc_dointvec(table, write, filp, buffer, lenp, ppos);
if (!write) {
/* read */
RETURN(rc);
}
if (kgnilnd_data.kgn_init != GNILND_INIT_ALL) {
rc = -EINVAL;
RETURN(rc);
}
if (old_val != kgnilnd_sysctl.ksd_rdmaq_override) {
long new_mb = kgnilnd_sysctl.ksd_rdmaq_override * (long)(1024*1024);
LCONSOLE_INFO("changing RDMAQ override to %d mbytes/sec\n",
kgnilnd_sysctl.ksd_rdmaq_override);
/* override proc is mbytes, but we calc in bytes */
kgnilnd_data.kgn_rdmaq_override = new_mb;
smp_wmb();
}
RETURN(rc);
}
static ssize_t
ldiskfs_osd_index_in_idif_seq_write(struct file *file,
const char __user *buffer,
size_t count, loff_t *off)
{
struct lu_env env;
struct seq_file *m = file->private_data;
struct dt_device *dt = m->private;
struct osd_device *dev = osd_dt_dev(dt);
struct lu_target *tgt;
__s64 val;
int rc;
LASSERT(dev != NULL);
if (unlikely(dev->od_mnt == NULL))
return -EINPROGRESS;
rc = lprocfs_str_to_s64(buffer, count, &val);
if (rc != 0)
return rc;
if (dev->od_index_in_idif) {
if (val != 0)
return count;
LCONSOLE_WARN("%s: OST-index in IDIF has been enabled, "
"it cannot be reverted back.\n", osd_name(dev));
return -EPERM;
}
if (val == 0)
return count;
rc = lu_env_init(&env, LCT_DT_THREAD);
if (rc != 0)
return rc;
tgt = dev->od_dt_dev.dd_lu_dev.ld_site->ls_tgt;
tgt->lut_lsd.lsd_feature_rocompat |= OBD_ROCOMPAT_IDX_IN_IDIF;
rc = tgt_server_data_update(&env, tgt, 1);
lu_env_fini(&env);
if (rc < 0)
return rc;
LCONSOLE_INFO("%s: enable OST-index in IDIF successfully, "
"it cannot be reverted back.\n", osd_name(dev));
dev->od_index_in_idif = 1;
return count;
}
/**
* Performs cleanup procedures for passed \a obd given it is mgs obd.
*/
static int mgs_cleanup(struct obd_device *obd)
{
struct mgs_obd *mgs = &obd->u.mgs;
ENTRY;
if (mgs->mgs_sb == NULL)
RETURN(0);
mgs_fs_cleanup(obd);
server_put_mount(obd->obd_name, mgs->mgs_vfsmnt);
mgs->mgs_sb = NULL;
ldlm_namespace_free(obd->obd_namespace, NULL, 1);
obd->obd_namespace = NULL;
fsfilt_put_ops(obd->obd_fsops);
LCONSOLE_INFO("%s has stopped.\n", obd->obd_name);
RETURN(0);
}
int osp_precreate_rollover_new_seq(struct lu_env *env, struct osp_device *osp)
{
struct lu_fid *fid = &osp_env_info(env)->osi_fid;
struct lu_fid *last_fid = &osp->opd_last_used_fid;
int rc;
ENTRY;
rc = seq_client_get_seq(env, osp->opd_obd->u.cli.cl_seq, &fid->f_seq);
if (rc != 0) {
CERROR("%s: alloc fid error: rc = %d\n",
osp->opd_obd->obd_name, rc);
RETURN(rc);
}
fid->f_oid = 1;
fid->f_ver = 0;
LASSERTF(fid_seq(fid) != fid_seq(last_fid),
"fid "DFID", last_fid "DFID"\n", PFID(fid),
PFID(last_fid));
rc = osp_write_last_oid_seq_files(env, osp, fid, 1);
if (rc != 0) {
CERROR("%s: Can not update oid/seq file: rc = %d\n",
osp->opd_obd->obd_name, rc);
RETURN(rc);
}
LCONSOLE_INFO("%s: update sequence from "LPX64" to "LPX64"\n",
osp->opd_obd->obd_name, fid_seq(last_fid),
fid_seq(fid));
/* Update last_xxx to the new seq */
spin_lock(&osp->opd_pre_lock);
osp->opd_last_used_fid = *fid;
osp->opd_gap_start_fid = *fid;
osp->opd_pre_used_fid = *fid;
osp->opd_pre_last_created_fid = *fid;
spin_unlock(&osp->opd_pre_lock);
RETURN(rc);
}
int seq_server_init(struct lu_server_seq *seq,
struct dt_device *dev,
const char *prefix,
enum lu_mgr_type type,
struct seq_server_site *ss,
const struct lu_env *env)
{
int rc, is_srv = (type == LUSTRE_SEQ_SERVER);
ENTRY;
LASSERT(dev != NULL);
LASSERT(prefix != NULL);
LASSERT(ss != NULL);
LASSERT(ss->ss_lu != NULL);
seq->lss_cli = NULL;
seq->lss_type = type;
seq->lss_site = ss;
range_init(&seq->lss_space);
range_init(&seq->lss_lowater_set);
range_init(&seq->lss_hiwater_set);
seq->lss_set_width = LUSTRE_SEQ_BATCH_WIDTH;
mutex_init(&seq->lss_mutex);
seq->lss_width = is_srv ?
LUSTRE_SEQ_META_WIDTH : LUSTRE_SEQ_SUPER_WIDTH;
snprintf(seq->lss_name, sizeof(seq->lss_name),
"%s-%s", (is_srv ? "srv" : "ctl"), prefix);
rc = seq_store_init(seq, env, dev);
if (rc)
GOTO(out, rc);
/* Request backing store for saved sequence info. */
rc = seq_store_read(seq, env);
if (rc == -ENODATA) {
/* Nothing is read, init by default value. */
seq->lss_space = is_srv ?
LUSTRE_SEQ_ZERO_RANGE:
LUSTRE_SEQ_SPACE_RANGE;
LASSERT(ss != NULL);
seq->lss_space.lsr_index = ss->ss_node_id;
LCONSOLE_INFO("%s: No data found "
"on store. Initialize space\n",
seq->lss_name);
rc = seq_store_update(env, seq, NULL, 0);
if (rc) {
CERROR("%s: Can't write space data, "
"rc %d\n", seq->lss_name, rc);
}
} else if (rc) {
CERROR("%s: Can't read space data, rc %d\n",
seq->lss_name, rc);
GOTO(out, rc);
}
if (is_srv) {
LASSERT(range_is_sane(&seq->lss_space));
} else {
LASSERT(!range_is_zero(&seq->lss_space) &&
range_is_sane(&seq->lss_space));
}
rc = seq_server_proc_init(seq);
if (rc)
GOTO(out, rc);
EXIT;
out:
if (rc)
seq_server_fini(seq, env);
return rc;
}
static
int ll_getxattr_common(struct inode *inode, const char *name,
void *buffer, size_t size, __u64 valid)
{
struct ll_sb_info *sbi = ll_i2sbi(inode);
struct ptlrpc_request *req = NULL;
struct mdt_body *body;
int xattr_type, rc;
void *xdata;
struct obd_capa *oc;
struct rmtacl_ctl_entry *rce = NULL;
CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n",
inode->i_ino, inode->i_generation, inode);
/* listxattr have slightly different behavior from of ext3:
* without 'user_xattr' ext3 will list all xattr names but
* filtered out "^user..*"; we list them all for simplicity.
*/
if (!name) {
xattr_type = XATTR_OTHER_T;
goto do_getxattr;
}
xattr_type = get_xattr_type(name);
rc = xattr_type_filter(sbi, xattr_type);
if (rc)
return rc;
/* b15587: ignore security.capability xattr for now */
if ((xattr_type == XATTR_SECURITY_T &&
strcmp(name, "security.capability") == 0))
return -ENODATA;
/* LU-549: Disable security.selinux when selinux is disabled */
if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
strcmp(name, "security.selinux") == 0)
return -EOPNOTSUPP;
#ifdef CONFIG_FS_POSIX_ACL
if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
(xattr_type == XATTR_ACL_ACCESS_T ||
xattr_type == XATTR_ACL_DEFAULT_T)) {
rce = rct_search(&sbi->ll_rct, current_pid());
if (rce == NULL ||
(rce->rce_ops != RMT_LSETFACL &&
rce->rce_ops != RMT_LGETFACL &&
rce->rce_ops != RMT_RSETFACL &&
rce->rce_ops != RMT_RGETFACL))
return -EOPNOTSUPP;
}
/* posix acl is under protection of LOOKUP lock. when calling to this,
* we just have path resolution to the target inode, so we have great
* chance that cached ACL is uptodate.
*/
if (xattr_type == XATTR_ACL_ACCESS_T &&
!(sbi->ll_flags & LL_SBI_RMT_CLIENT)) {
struct ll_inode_info *lli = ll_i2info(inode);
struct posix_acl *acl;
spin_lock(&lli->lli_lock);
acl = posix_acl_dup(lli->lli_posix_acl);
spin_unlock(&lli->lli_lock);
if (!acl)
return -ENODATA;
rc = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
posix_acl_release(acl);
return rc;
}
if (xattr_type == XATTR_ACL_DEFAULT_T && !S_ISDIR(inode->i_mode))
return -ENODATA;
#endif
do_getxattr:
if (sbi->ll_xattr_cache_enabled && (rce == NULL ||
rce->rce_ops == RMT_LGETFACL ||
rce->rce_ops == RMT_LSETFACL)) {
rc = ll_xattr_cache_get(inode, name, buffer, size, valid);
if (rc < 0)
GOTO(out_xattr, rc);
} else {
oc = ll_mdscapa_get(inode);
rc = md_getxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
valid | (rce ? rce_ops2valid(rce->rce_ops) : 0),
name, NULL, 0, size, 0, &req);
capa_put(oc);
if (rc < 0)
GOTO(out_xattr, rc);
body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
LASSERT(body);
/* only detect the xattr size */
if (size == 0)
GOTO(out, rc = body->eadatasize);
if (size < body->eadatasize) {
CERROR("server bug: replied size %u > %u\n",
body->eadatasize, (int)size);
GOTO(out, rc = -ERANGE);
}
if (body->eadatasize == 0)
GOTO(out, rc = -ENODATA);
/* do not need swab xattr data */
xdata = req_capsule_server_sized_get(&req->rq_pill, &RMF_EADATA,
body->eadatasize);
if (!xdata)
GOTO(out, rc = -EFAULT);
memcpy(buffer, xdata, body->eadatasize);
rc = body->eadatasize;
}
#ifdef CONFIG_FS_POSIX_ACL
if (rce && rce->rce_ops == RMT_LSETFACL) {
ext_acl_xattr_header *acl;
acl = lustre_posix_acl_xattr_2ext(
(posix_acl_xattr_header *)buffer, rc);
if (IS_ERR(acl))
GOTO(out, rc = PTR_ERR(acl));
rc = ee_add(&sbi->ll_et, current_pid(), ll_inode2fid(inode),
xattr_type, acl);
if (unlikely(rc < 0)) {
lustre_ext_acl_xattr_free(acl);
GOTO(out, rc);
}
}
#endif
out_xattr:
if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
LCONSOLE_INFO(
"%s: disabling user_xattr feature because it is not supported on the server: rc = %d\n",
ll_get_fsname(inode->i_sb, NULL, 0), rc);
sbi->ll_flags &= ~LL_SBI_USER_XATTR;
}
out:
ptlrpc_req_finished(req);
return rc;
}
static
int ll_setxattr_common(struct inode *inode, const char *name,
const void *value, size_t size,
int flags, __u64 valid)
{
struct ll_sb_info *sbi = ll_i2sbi(inode);
struct ptlrpc_request *req = NULL;
int xattr_type, rc;
struct obd_capa *oc;
struct rmtacl_ctl_entry *rce = NULL;
#ifdef CONFIG_FS_POSIX_ACL
posix_acl_xattr_header *new_value = NULL;
ext_acl_xattr_header *acl = NULL;
#endif
const char *pv = value;
xattr_type = get_xattr_type(name);
rc = xattr_type_filter(sbi, xattr_type);
if (rc)
return rc;
/* b10667: ignore lustre special xattr for now */
if ((xattr_type == XATTR_TRUSTED_T && strcmp(name, "trusted.lov") == 0) ||
(xattr_type == XATTR_LUSTRE_T && strcmp(name, "lustre.lov") == 0))
return 0;
/* b15587: ignore security.capability xattr for now */
if ((xattr_type == XATTR_SECURITY_T &&
strcmp(name, "security.capability") == 0))
return 0;
/* LU-549: Disable security.selinux when selinux is disabled */
if (xattr_type == XATTR_SECURITY_T && !selinux_is_enabled() &&
strcmp(name, "security.selinux") == 0)
return -EOPNOTSUPP;
#ifdef CONFIG_FS_POSIX_ACL
if (sbi->ll_flags & LL_SBI_RMT_CLIENT &&
(xattr_type == XATTR_ACL_ACCESS_T ||
xattr_type == XATTR_ACL_DEFAULT_T)) {
rce = rct_search(&sbi->ll_rct, current_pid());
if (rce == NULL ||
(rce->rce_ops != RMT_LSETFACL &&
rce->rce_ops != RMT_RSETFACL))
return -EOPNOTSUPP;
if (rce->rce_ops == RMT_LSETFACL) {
struct eacl_entry *ee;
ee = et_search_del(&sbi->ll_et, current_pid(),
ll_inode2fid(inode), xattr_type);
LASSERT(ee != NULL);
if (valid & OBD_MD_FLXATTR) {
acl = lustre_acl_xattr_merge2ext(
(posix_acl_xattr_header *)value,
size, ee->ee_acl);
if (IS_ERR(acl)) {
ee_free(ee);
return PTR_ERR(acl);
}
size = CFS_ACL_XATTR_SIZE(\
le32_to_cpu(acl->a_count), \
ext_acl_xattr);
pv = (const char *)acl;
}
ee_free(ee);
} else if (rce->rce_ops == RMT_RSETFACL) {
size = lustre_posix_acl_xattr_filter(
(posix_acl_xattr_header *)value,
size, &new_value);
if (unlikely(size < 0))
return size;
pv = (const char *)new_value;
} else
return -EOPNOTSUPP;
valid |= rce_ops2valid(rce->rce_ops);
}
#endif
if (sbi->ll_xattr_cache_enabled &&
(rce == NULL || rce->rce_ops == RMT_LSETFACL)) {
rc = ll_xattr_cache_update(inode, name, pv, size, valid, flags);
} else {
oc = ll_mdscapa_get(inode);
rc = md_setxattr(sbi->ll_md_exp, ll_inode2fid(inode), oc,
valid, name, pv, size, 0, flags,
ll_i2suppgid(inode), &req);
capa_put(oc);
}
#ifdef CONFIG_FS_POSIX_ACL
if (new_value != NULL)
lustre_posix_acl_xattr_free(new_value, size);
if (acl != NULL)
lustre_ext_acl_xattr_free(acl);
#endif
if (rc) {
if (rc == -EOPNOTSUPP && xattr_type == XATTR_USER_T) {
LCONSOLE_INFO("Disabling user_xattr feature because "
"it is not supported on the server\n");
sbi->ll_flags &= ~LL_SBI_USER_XATTR;
}
return rc;
}
ptlrpc_req_finished(req);
return 0;
}
static int osp_precreate_thread(void *_arg)
{
struct osp_device *d = _arg;
struct ptlrpc_thread *thread = &d->opd_pre_thread;
struct l_wait_info lwi = { 0 };
struct lu_env env;
int rc;
ENTRY;
rc = lu_env_init(&env, d->opd_dt_dev.dd_lu_dev.ld_type->ldt_ctx_tags);
if (rc) {
CERROR("%s: init env error: rc = %d\n", d->opd_obd->obd_name,
rc);
RETURN(rc);
}
spin_lock(&d->opd_pre_lock);
thread->t_flags = SVC_RUNNING;
spin_unlock(&d->opd_pre_lock);
wake_up(&thread->t_ctl_waitq);
while (osp_precreate_running(d)) {
/*
* need to be connected to OST
*/
while (osp_precreate_running(d)) {
l_wait_event(d->opd_pre_waitq,
!osp_precreate_running(d) ||
d->opd_new_connection,
&lwi);
if (!d->opd_new_connection)
continue;
d->opd_new_connection = 0;
d->opd_got_disconnected = 0;
break;
}
if (!osp_precreate_running(d))
break;
LASSERT(d->opd_obd->u.cli.cl_seq != NULL);
if (d->opd_obd->u.cli.cl_seq->lcs_exp == NULL) {
/* Get new sequence for client first */
LASSERT(d->opd_exp != NULL);
d->opd_obd->u.cli.cl_seq->lcs_exp =
class_export_get(d->opd_exp);
rc = osp_init_pre_fid(d);
if (rc != 0) {
class_export_put(d->opd_exp);
d->opd_obd->u.cli.cl_seq->lcs_exp = NULL;
CERROR("%s: init pre fid error: rc = %d\n",
d->opd_obd->obd_name, rc);
continue;
}
}
osp_statfs_update(d);
/*
* Clean up orphans or recreate missing objects.
*/
rc = osp_precreate_cleanup_orphans(&env, d);
if (rc != 0)
continue;
/*
* connected, can handle precreates now
*/
while (osp_precreate_running(d)) {
l_wait_event(d->opd_pre_waitq,
!osp_precreate_running(d) ||
osp_precreate_near_empty(&env, d) ||
osp_statfs_need_update(d) ||
d->opd_got_disconnected, &lwi);
if (!osp_precreate_running(d))
break;
/* something happened to the connection
* have to start from the beginning */
if (d->opd_got_disconnected)
break;
if (osp_statfs_need_update(d))
osp_statfs_update(d);
/* To avoid handling different seq in precreate/orphan
* cleanup, it will hold precreate until current seq is
* used up. */
if (unlikely(osp_precreate_end_seq(&env, d) &&
!osp_create_end_seq(&env, d)))
continue;
if (unlikely(osp_precreate_end_seq(&env, d) &&
osp_create_end_seq(&env, d))) {
LCONSOLE_INFO("%s:"LPX64" is used up."
" Update to new seq\n",
d->opd_obd->obd_name,
fid_seq(&d->opd_pre_last_created_fid));
rc = osp_precreate_rollover_new_seq(&env, d);
if (rc)
continue;
}
if (osp_precreate_near_empty(&env, d)) {
rc = osp_precreate_send(&env, d);
/* osp_precreate_send() sets opd_pre_status
* in case of error, that prevent the using of
* failed device. */
if (rc < 0 && rc != -ENOSPC &&
rc != -ETIMEDOUT && rc != -ENOTCONN)
CERROR("%s: cannot precreate objects:"
" rc = %d\n",
d->opd_obd->obd_name, rc);
}
}
}
thread->t_flags = SVC_STOPPED;
lu_env_fini(&env);
wake_up(&thread->t_ctl_waitq);
RETURN(0);
}
int seq_server_init(const struct lu_env *env,
struct lu_server_seq *seq,
struct dt_device *dev,
const char *prefix,
enum lu_mgr_type type,
struct seq_server_site *ss)
{
int rc, is_srv = (type == LUSTRE_SEQ_SERVER);
ENTRY;
LASSERT(dev != NULL);
LASSERT(prefix != NULL);
LASSERT(ss != NULL);
LASSERT(ss->ss_lu != NULL);
/* A compile-time check for FIDs that used to be in lustre_idl.h
* but is moved here to remove CLASSERT/LASSERT in that header.
* Check all lu_fid fields are converted in fid_cpu_to_le() and friends
* and that there is no padding added by compiler to the struct. */
{
struct lu_fid tst;
CLASSERT(sizeof(tst) == sizeof(tst.f_seq) +
sizeof(tst.f_oid) + sizeof(tst.f_ver));
}
seq->lss_cli = NULL;
seq->lss_type = type;
seq->lss_site = ss;
lu_seq_range_init(&seq->lss_space);
lu_seq_range_init(&seq->lss_lowater_set);
lu_seq_range_init(&seq->lss_hiwater_set);
seq->lss_set_width = LUSTRE_SEQ_BATCH_WIDTH;
mutex_init(&seq->lss_mutex);
seq->lss_width = is_srv ?
LUSTRE_SEQ_META_WIDTH : LUSTRE_SEQ_SUPER_WIDTH;
snprintf(seq->lss_name, sizeof(seq->lss_name),
"%s-%s", (is_srv ? "srv" : "ctl"), prefix);
rc = seq_store_init(seq, env, dev);
if (rc)
GOTO(out, rc);
/* Request backing store for saved sequence info. */
rc = seq_store_read(seq, env);
if (rc == -ENODATA) {
/* Nothing is read, init by default value. */
seq->lss_space = is_srv ?
LUSTRE_SEQ_ZERO_RANGE:
LUSTRE_SEQ_SPACE_RANGE;
seq->lss_space.lsr_index = ss->ss_node_id;
LCONSOLE_INFO("%s: No data found "
"on store. Initialize space\n",
seq->lss_name);
rc = seq_store_update(env, seq, NULL, 0);
if (rc) {
CERROR("%s: Can't write space data, "
"rc %d\n", seq->lss_name, rc);
}
} else if (rc) {
CERROR("%s: Can't read space data, rc %d\n",
seq->lss_name, rc);
GOTO(out, rc);
}
if (is_srv) {
LASSERT(lu_seq_range_is_sane(&seq->lss_space));
} else {
LASSERT(!lu_seq_range_is_zero(&seq->lss_space) &&
lu_seq_range_is_sane(&seq->lss_space));
}
rc = seq_server_proc_init(seq);
if (rc)
GOTO(out, rc);
EXIT;
out:
if (rc)
seq_server_fini(seq, env);
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);
}
