int mdc_unlink(struct obd_export *exp, struct md_op_data *op_data,
struct ptlrpc_request **request)
{
CFS_LIST_HEAD(cancels);
struct obd_device *obd = class_exp2obd(exp);
struct ptlrpc_request *req = *request;
int count = 0, rc;
ENTRY;
LASSERT(req == NULL);
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)) &&
!OBD_FAIL_CHECK(OBD_FAIL_LDLM_BL_CALLBACK_NET))
count = mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
if ((op_data->op_flags & MF_MDC_CANCEL_FID3) &&
(fid_is_sane(&op_data->op_fid3)) &&
!OBD_FAIL_CHECK(OBD_FAIL_LDLM_BL_CALLBACK_NET))
count += mdc_resource_get_unused(exp, &op_data->op_fid3,
&cancels, LCK_EX,
MDS_INODELOCK_FULL);
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_MDS_REINT_UNLINK);
if (req == NULL) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
RETURN(-ENOMEM);
}
mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
op_data->op_namelen + 1);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
mdc_unlink_pack(req, op_data);
req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
obd->u.cli.cl_max_mds_easize);
req_capsule_set_size(&req->rq_pill, &RMF_LOGCOOKIES, RCL_SERVER,
obd->u.cli.cl_max_mds_cookiesize);
ptlrpc_request_set_replen(req);
*request = req;
rc = mdc_reint(req, obd->u.cli.cl_rpc_lock, LUSTRE_IMP_FULL);
if (rc == -ERESTARTSYS)
rc = 0;
RETURN(rc);
}
int mdc_link(struct obd_export *exp, struct md_op_data *op_data,
struct ptlrpc_request **request)
{
struct list_head cancels = LIST_HEAD_INIT(cancels);
struct ptlrpc_request *req;
int count = 0, rc;
ENTRY;
if ((op_data->op_flags & MF_MDC_CANCEL_FID2) &&
(fid_is_sane(&op_data->op_fid2)))
count = mdc_resource_get_unused(exp, &op_data->op_fid2,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)))
count += mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_REINT_LINK);
if (req == NULL) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
RETURN(-ENOMEM);
}
req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
op_data->op_namelen + 1);
/* get SELinux policy info if any */
rc = sptlrpc_get_sepol(req);
if (rc < 0) {
ptlrpc_request_free(req);
RETURN(rc);
}
req_capsule_set_size(&req->rq_pill, &RMF_SELINUX_POL, RCL_CLIENT,
strlen(req->rq_sepol) ?
strlen(req->rq_sepol) + 1 : 0);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
mdc_link_pack(req, op_data);
ptlrpc_request_set_replen(req);
rc = mdc_reint(req, LUSTRE_IMP_FULL);
*request = req;
if (rc == -ERESTARTSYS)
rc = 0;
RETURN(rc);
}
int mdc_unlink(struct obd_export *exp, struct md_op_data *op_data,
struct ptlrpc_request **request)
{
LIST_HEAD(cancels);
struct obd_device *obd = class_exp2obd(exp);
struct ptlrpc_request *req = *request;
int count = 0, rc;
LASSERT(!req);
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)))
count = mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
if ((op_data->op_flags & MF_MDC_CANCEL_FID3) &&
(fid_is_sane(&op_data->op_fid3)))
count += mdc_resource_get_unused(exp, &op_data->op_fid3,
&cancels, LCK_EX,
MDS_INODELOCK_FULL);
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_MDS_REINT_UNLINK);
if (!req) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
return -ENOMEM;
}
req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
op_data->op_namelen + 1);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
return rc;
}
mdc_unlink_pack(req, op_data);
req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
obd->u.cli.cl_default_mds_easize);
ptlrpc_request_set_replen(req);
*request = req;
rc = mdc_reint(req, LUSTRE_IMP_FULL);
if (rc == -ERESTARTSYS)
rc = 0;
return rc;
}
int lmv_fld_lookup(struct lmv_obd *lmv, const struct lu_fid *fid, u32 *mds)
{
struct obd_device *obd = lmv2obd_dev(lmv);
int rc;
/*
* FIXME: Currently ZFS still use local seq for ROOT unfortunately, and
* this fid_is_local check should be removed once LU-2240 is fixed
*/
if (!fid_is_sane(fid) || !(fid_seq_in_fldb(fid_seq(fid)) ||
fid_seq_is_local_file(fid_seq(fid)))) {
CERROR("%s: invalid FID " DFID "\n", obd->obd_name, PFID(fid));
return -EINVAL;
}
rc = fld_client_lookup(&lmv->lmv_fld, fid_seq(fid), mds,
LU_SEQ_RANGE_MDT, NULL);
if (rc) {
CERROR("Error while looking for mds number. Seq %#llx, err = %d\n",
fid_seq(fid), rc);
return rc;
}
CDEBUG(D_INODE, "FLD lookup got mds #%x for fid="DFID"\n",
*mds, PFID(fid));
if (*mds >= lmv->desc.ld_tgt_count) {
CERROR("FLD lookup got invalid mds #%x (max: %x) for fid=" DFID "\n", *mds, lmv->desc.ld_tgt_count,
PFID(fid));
rc = -EINVAL;
}
return rc;
}
static int out_destroy(struct tgt_session_info *tsi)
{
struct tgt_thread_info *tti = tgt_th_info(tsi->tsi_env);
struct update *update = tti->tti_u.update.tti_update;
struct dt_object *obj = tti->tti_u.update.tti_dt_object;
struct lu_fid *fid;
int rc;
ENTRY;
fid = &update->u_fid;
fid_le_to_cpu(fid, fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt), PFID(fid), -EPROTO);
RETURN(err_serious(-EPROTO));
}
if (!lu_object_exists(&obj->do_lu))
RETURN(-ENOENT);
rc = out_tx_destroy(tsi->tsi_env, obj, &tti->tti_tea,
tti->tti_u.update.tti_update_reply,
tti->tti_u.update.tti_update_reply_index);
RETURN(rc);
}
int lmv_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
void *lmm, int lmmsize, struct lookup_intent *it,
int flags, struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
int rc;
LASSERT(it != NULL);
LASSERT(fid_is_sane(&op_data->op_fid1));
CDEBUG(D_INODE, "INTENT LOCK '%s' for '%*s' on "DFID"\n",
LL_IT2STR(it), op_data->op_namelen, op_data->op_name,
PFID(&op_data->op_fid1));
rc = lmv_check_connect(obd);
if (rc)
return rc;
if (it->it_op & (IT_LOOKUP | IT_GETATTR | IT_LAYOUT))
rc = lmv_intent_lookup(exp, op_data, lmm, lmmsize, it,
flags, reqp, cb_blocking,
extra_lock_flags);
else if (it->it_op & IT_OPEN)
rc = lmv_intent_open(exp, op_data, lmm, lmmsize, it,
flags, reqp, cb_blocking,
extra_lock_flags);
else
LBUG();
return rc;
}
int lmv_fld_lookup(struct lmv_obd *lmv,
const struct lu_fid *fid,
mdsno_t *mds)
{
int rc;
/* FIXME: Currently ZFS still use local seq for ROOT unfortunately, and
* this fid_is_local check should be removed once LU-2240 is fixed */
LASSERTF((fid_seq_in_fldb(fid_seq(fid)) ||
fid_seq_is_local_file(fid_seq(fid))) &&
fid_is_sane(fid), DFID" is insane!\n", PFID(fid));
rc = fld_client_lookup(&lmv->lmv_fld, fid_seq(fid), mds,
LU_SEQ_RANGE_MDT, NULL);
if (rc) {
CERROR("Error while looking for mds number. Seq %#llx, err = %d\n",
fid_seq(fid), rc);
return rc;
}
CDEBUG(D_INODE, "FLD lookup got mds #%x for fid="DFID"\n",
*mds, PFID(fid));
if (*mds >= lmv->desc.ld_tgt_count) {
CERROR("FLD lookup got invalid mds #%x (max: %x) "
"for fid="DFID"\n", *mds, lmv->desc.ld_tgt_count,
PFID(fid));
rc = -EINVAL;
}
return rc;
}
/*
* This is is_subdir() variant, it is CMD if cmm forwards it to correct
* target. Source should not be ancestor of target dir. May be other rename
* checks can be moved here later.
*/
static int mdt_rename_sanity(struct mdt_thread_info *info, struct lu_fid *fid)
{
struct mdt_reint_record *rr = &info->mti_rr;
struct lu_fid dst_fid = *rr->rr_fid2;
struct mdt_object *dst;
int rc = 0;
ENTRY;
do {
LASSERT(fid_is_sane(&dst_fid));
dst = mdt_object_find(info->mti_env, info->mti_mdt, &dst_fid);
if (!IS_ERR(dst)) {
rc = mdo_is_subdir(info->mti_env,
mdt_object_child(dst), fid,
&dst_fid);
mdt_object_put(info->mti_env, dst);
if (rc != -EREMOTE && rc < 0) {
CERROR("Failed mdo_is_subdir(), rc %d\n", rc);
} else {
/* check the found fid */
if (lu_fid_eq(&dst_fid, fid))
rc = -EINVAL;
}
} else {
rc = PTR_ERR(dst);
}
} while (rc == -EREMOTE);
RETURN(rc);
}
int ofd_object_ff_load(const struct lu_env *env, struct ofd_object *fo)
{
struct ofd_thread_info *info = ofd_info(env);
struct filter_fid_old *ff = &info->fti_mds_fid_old;
struct lu_buf *buf = &info->fti_buf;
struct lu_fid *pfid = &fo->ofo_pfid;
int rc = 0;
if (fid_is_sane(pfid))
return 0;
buf->lb_buf = ff;
buf->lb_len = sizeof(*ff);
rc = dt_xattr_get(env, ofd_object_child(fo), buf, XATTR_NAME_FID,
BYPASS_CAPA);
if (rc < 0)
return rc;
if (rc < sizeof(struct lu_fid)) {
fid_zero(pfid);
return -ENODATA;
}
pfid->f_seq = le64_to_cpu(ff->ff_parent.f_seq);
pfid->f_oid = le32_to_cpu(ff->ff_parent.f_oid);
/* Currently, the filter_fid::ff_parent::f_ver is not the real parent
* MDT-object's FID::f_ver, instead it is the OST-object index in its
* parent MDT-object's layout EA. */
pfid->f_stripe_idx = le32_to_cpu(ff->ff_parent.f_stripe_idx);
return 0;
}
int mdc_link(struct obd_export *exp, struct md_op_data *op_data,
struct ptlrpc_request **request)
{
CFS_LIST_HEAD(cancels);
struct obd_device *obd = exp->exp_obd;
struct ptlrpc_request *req;
int count = 0, rc;
ENTRY;
if ((op_data->op_flags & MF_MDC_CANCEL_FID2) &&
(fid_is_sane(&op_data->op_fid2)))
count = mdc_resource_get_unused(exp, &op_data->op_fid2,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)))
count += mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX,
MDS_INODELOCK_UPDATE);
req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_REINT_LINK);
if (req == NULL) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
RETURN(-ENOMEM);
}
mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
mdc_set_capa_size(req, &RMF_CAPA2, op_data->op_capa2);
req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
op_data->op_namelen + 1);
rc = mdc_prep_elc_req(exp, req, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
mdc_link_pack(req, op_data);
ptlrpc_request_set_replen(req);
rc = mdc_reint(req, obd->u.cli.cl_rpc_lock, LUSTRE_IMP_FULL);
*request = req;
if (rc == -ERESTARTSYS)
rc = 0;
RETURN(rc);
}
static int out_index_insert(struct tgt_session_info *tsi)
{
struct tgt_thread_info *tti = tgt_th_info(tsi->tsi_env);
struct object_update *update = tti->tti_u.update.tti_update;
struct dt_object *obj = tti->tti_u.update.tti_dt_object;
struct dt_insert_rec *rec = &tti->tti_rec;
struct lu_fid *fid;
char *name;
__u32 *ptype;
int rc = 0;
size_t size;
ENTRY;
name = object_update_param_get(update, 0, NULL);
if (IS_ERR(name)) {
CERROR("%s: empty name for index insert: rc = %ld\n",
tgt_name(tsi->tsi_tgt), PTR_ERR(name));
RETURN(PTR_ERR(name));
}
fid = object_update_param_get(update, 1, &size);
if (IS_ERR(fid) || size != sizeof(*fid)) {
CERROR("%s: invalid fid: rc = %ld\n",
tgt_name(tsi->tsi_tgt), PTR_ERR(fid));
RETURN(PTR_ERR(fid));
}
if (ptlrpc_req_need_swab(tsi->tsi_pill->rc_req))
lustre_swab_lu_fid(fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt), PFID(fid), -EPROTO);
RETURN(-EPROTO);
}
ptype = object_update_param_get(update, 2, &size);
if (IS_ERR(ptype) || size != sizeof(*ptype)) {
CERROR("%s: invalid type for index insert: rc = %ld\n",
tgt_name(tsi->tsi_tgt), PTR_ERR(ptype));
RETURN(PTR_ERR(ptype));
}
if (ptlrpc_req_need_swab(tsi->tsi_pill->rc_req))
__swab32s(ptype);
rec->rec_fid = fid;
rec->rec_type = *ptype;
rc = out_tx_index_insert(tsi->tsi_env, obj, (const struct dt_rec *)rec,
(const struct dt_key *)name, &tti->tti_tea,
tti->tti_tea.ta_handle,
tti->tti_u.update.tti_update_reply,
tti->tti_u.update.tti_update_reply_index);
RETURN(rc);
}
int mdc_setattr(struct obd_export *exp, struct md_op_data *op_data,
void *ea, size_t ealen, struct ptlrpc_request **request)
{
struct list_head cancels = LIST_HEAD_INIT(cancels);
struct ptlrpc_request *req;
struct mdc_rpc_lock *rpc_lock;
struct obd_device *obd = exp->exp_obd;
int count = 0, rc;
__u64 bits;
ENTRY;
LASSERT(op_data != NULL);
bits = MDS_INODELOCK_UPDATE;
if (op_data->op_attr.ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID))
bits |= MDS_INODELOCK_LOOKUP;
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)))
count = mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX, bits);
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_MDS_REINT_SETATTR);
if (req == NULL) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
RETURN(-ENOMEM);
}
mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
req_capsule_set_size(&req->rq_pill, &RMF_MDT_EPOCH, RCL_CLIENT, 0);
req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT, ealen);
req_capsule_set_size(&req->rq_pill, &RMF_LOGCOOKIES, RCL_CLIENT, 0);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
rpc_lock = obd->u.cli.cl_rpc_lock;
if (op_data->op_attr.ia_valid & (ATTR_MTIME | ATTR_CTIME))
CDEBUG(D_INODE, "setting mtime "CFS_TIME_T
", ctime "CFS_TIME_T"\n",
LTIME_S(op_data->op_attr.ia_mtime),
LTIME_S(op_data->op_attr.ia_ctime));
mdc_setattr_pack(req, op_data, ea, ealen);
ptlrpc_request_set_replen(req);
rc = mdc_reint(req, rpc_lock, LUSTRE_IMP_FULL);
if (rc == -ERESTARTSYS)
rc = 0;
*request = req;
RETURN(rc);
}
static int out_create(struct tgt_session_info *tsi)
{
struct tgt_thread_info *tti = tgt_th_info(tsi->tsi_env);
struct object_update *update = tti->tti_u.update.tti_update;
struct dt_object *obj = tti->tti_u.update.tti_dt_object;
struct dt_object_format *dof = &tti->tti_u.update.tti_update_dof;
struct obdo *lobdo = &tti->tti_u.update.tti_obdo;
struct lu_attr *attr = &tti->tti_attr;
struct lu_fid *fid = NULL;
struct obdo *wobdo;
int size;
int rc;
ENTRY;
wobdo = object_update_param_get(update, 0, &size);
if (wobdo == NULL || size != sizeof(*wobdo)) {
CERROR("%s: obdo is NULL, invalid RPC: rc = %d\n",
tgt_name(tsi->tsi_tgt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
obdo_le_to_cpu(wobdo, wobdo);
lustre_get_wire_obdo(NULL, lobdo, wobdo);
la_from_obdo(attr, lobdo, lobdo->o_valid);
dof->dof_type = dt_mode_to_dft(attr->la_mode);
if (update->ou_params_count > 1) {
int size;
fid = object_update_param_get(update, 1, &size);
if (fid == NULL || size != sizeof(*fid)) {
CERROR("%s: invalid fid: rc = %d\n",
tgt_name(tsi->tsi_tgt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
if (ptlrpc_req_need_swab(tsi->tsi_pill->rc_req))
lustre_swab_lu_fid(fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid fid "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt), PFID(fid), -EPROTO);
RETURN(err_serious(-EPROTO));
}
}
if (lu_object_exists(&obj->do_lu))
RETURN(-EEXIST);
rc = out_tx_create(tsi->tsi_env, obj, attr, fid, dof,
&tti->tti_tea,
tti->tti_u.update.tti_update_reply,
tti->tti_u.update.tti_update_reply_index);
RETURN(rc);
}
int lmv_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
struct lookup_intent *it, struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
int rc;
ENTRY;
LASSERT(it != NULL);
LASSERT(fid_is_sane(&op_data->op_fid1));
CDEBUG(D_INODE, "INTENT LOCK '%s' for "DFID" '%.*s' on "DFID"\n",
LL_IT2STR(it), PFID(&op_data->op_fid2),
(int)op_data->op_namelen, op_data->op_name,
PFID(&op_data->op_fid1));
rc = lmv_check_connect(obd);
if (rc)
RETURN(rc);
if (it->it_op & (IT_LOOKUP | IT_GETATTR | IT_LAYOUT))
rc = lmv_intent_lookup(exp, op_data, it, reqp, cb_blocking,
extra_lock_flags);
else if (it->it_op & IT_OPEN)
rc = lmv_intent_open(exp, op_data, it, reqp, cb_blocking,
extra_lock_flags);
else
LBUG();
if (rc < 0) {
struct lustre_handle lock_handle;
if (it->d.lustre.it_lock_mode != 0) {
lock_handle.cookie = it->d.lustre.it_lock_handle;
ldlm_lock_decref(&lock_handle,
it->d.lustre.it_lock_mode);
}
it->d.lustre.it_lock_handle = 0;
it->d.lustre.it_lock_mode = 0;
if (it->d.lustre.it_remote_lock_mode != 0) {
lock_handle.cookie = it->d.lustre.it_remote_lock_handle;
ldlm_lock_decref(&lock_handle,
it->d.lustre.it_remote_lock_mode);
}
it->d.lustre.it_remote_lock_handle = 0;
it->d.lustre.it_remote_lock_mode = 0;
}
RETURN(rc);
}
static int out_create(struct mdt_thread_info *info)
{
struct update *update = info->mti_u.update.mti_update;
struct dt_object *obj = info->mti_u.update.mti_dt_object;
struct dt_object_format *dof = &info->mti_u.update.mti_update_dof;
struct obdo *lobdo = &info->mti_u.update.mti_obdo;
struct lu_attr *attr = &info->mti_attr.ma_attr;
struct lu_fid *fid = NULL;
struct obdo *wobdo;
int size;
int rc;
ENTRY;
wobdo = update_param_buf(update, 0, &size);
if (wobdo == NULL || size != sizeof(*wobdo)) {
CERROR("%s: obdo is NULL, invalid RPC: rc = %d\n",
mdt_obd_name(info->mti_mdt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
obdo_le_to_cpu(wobdo, wobdo);
lustre_get_wire_obdo(lobdo, wobdo);
la_from_obdo(attr, lobdo, lobdo->o_valid);
dof->dof_type = dt_mode_to_dft(attr->la_mode);
if (S_ISDIR(attr->la_mode)) {
int size;
fid = update_param_buf(update, 1, &size);
if (fid == NULL || size != sizeof(*fid)) {
CERROR("%s: invalid fid: rc = %d\n",
mdt_obd_name(info->mti_mdt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
fid_le_to_cpu(fid, fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid fid "DFID": rc = %d\n",
mdt_obd_name(info->mti_mdt),
PFID(fid), -EPROTO);
RETURN(err_serious(-EPROTO));
}
}
rc = out_tx_create(info, obj, attr, fid, dof, &info->mti_handle,
info->mti_u.update.mti_update_reply,
info->mti_u.update.mti_update_reply_index);
RETURN(rc);
}
/*
* Handler for: getattr, lookup and revalidate cases.
*/
int lmv_intent_lookup(struct obd_export *exp, struct md_op_data *op_data,
void *lmm, int lmmsize, struct lookup_intent *it,
int flags, struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *tgt = NULL;
struct mdt_body *body;
int rc = 0;
tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
if (IS_ERR(tgt))
return PTR_ERR(tgt);
if (!fid_is_sane(&op_data->op_fid2))
fid_zero(&op_data->op_fid2);
CDEBUG(D_INODE, "LOOKUP_INTENT with fid1="DFID", fid2="DFID
", name='%s' -> mds #%d\n", PFID(&op_data->op_fid1),
PFID(&op_data->op_fid2),
op_data->op_name ? op_data->op_name : "<NULL>",
tgt->ltd_idx);
op_data->op_bias &= ~MDS_CROSS_REF;
rc = md_intent_lock(tgt->ltd_exp, op_data, lmm, lmmsize, it,
flags, reqp, cb_blocking, extra_lock_flags);
if (rc < 0 || *reqp == NULL)
return rc;
/*
* MDS has returned success. Probably name has been resolved in
* remote inode. Let's check this.
*/
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
return -EPROTO;
/* Not cross-ref case, just get out of here. */
if (likely(!(body->valid & OBD_MD_MDS)))
return 0;
rc = lmv_intent_remote(exp, lmm, lmmsize, it, NULL, flags, reqp,
cb_blocking, extra_lock_flags);
return rc;
}
/** get (name+parent_id) for an entry
* \param linkno hardlink index
* \retval -ENODATA after last link
* \retval -ERANGE if namelen is too small
*/
int Lustre_GetNameParent(const char *path, int linkno,
lustre_fid *pfid, char *name,
int namelen)
{
int rc, i, len;
char buf[4096];
struct linkea_data ldata = { 0 };
struct lu_buf lb = { 0 };
rc = lgetxattr(path, XATTR_NAME_LINK, buf, sizeof(buf));
if (rc < 0)
return -errno;
lb.lb_buf = buf;
lb.lb_len = sizeof(buf);
ldata.ld_buf = &lb;
ldata.ld_leh = (struct link_ea_header *)buf;
ldata.ld_lee = LINKEA_FIRST_ENTRY(ldata);
ldata.ld_reclen = (ldata.ld_lee->lee_reclen[0] << 8)
| ldata.ld_lee->lee_reclen[1];
if (linkno >= ldata.ld_leh->leh_reccount)
/* beyond last link */
return -ENODATA;
for (i = 0; i < linkno; i++) {
ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
ldata.ld_reclen = (ldata.ld_lee->lee_reclen[0] << 8)
| ldata.ld_lee->lee_reclen[1];
}
memcpy(pfid, &ldata.ld_lee->lee_parent_fid, sizeof(*pfid));
fid_be_to_cpu(pfid, pfid);
if (!fid_is_sane(pfid))
{
DisplayLog(LVL_MAJOR, __func__, "insane fid: "DFID, PFID(pfid));
return -EPROTO;
}
len = ldata.ld_reclen - sizeof(struct link_ea_entry);
if (len >= namelen)
return -ERANGE;
strncpy(name, ldata.ld_lee->lee_name, len);
name[len] = '\0';
return 0;
}
int mdc_setattr(struct obd_export *exp, struct md_op_data *op_data,
void *ea, size_t ealen, struct ptlrpc_request **request)
{
LIST_HEAD(cancels);
struct ptlrpc_request *req;
int count = 0, rc;
__u64 bits;
bits = MDS_INODELOCK_UPDATE;
if (op_data->op_attr.ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
bits |= MDS_INODELOCK_LOOKUP;
if ((op_data->op_flags & MF_MDC_CANCEL_FID1) &&
(fid_is_sane(&op_data->op_fid1)))
count = mdc_resource_get_unused(exp, &op_data->op_fid1,
&cancels, LCK_EX, bits);
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_MDS_REINT_SETATTR);
if (!req) {
ldlm_lock_list_put(&cancels, l_bl_ast, count);
return -ENOMEM;
}
req_capsule_set_size(&req->rq_pill, &RMF_MDT_EPOCH, RCL_CLIENT, 0);
req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT, ealen);
req_capsule_set_size(&req->rq_pill, &RMF_LOGCOOKIES, RCL_CLIENT, 0);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
return rc;
}
if (op_data->op_attr.ia_valid & (ATTR_MTIME | ATTR_CTIME))
CDEBUG(D_INODE, "setting mtime %ld, ctime %ld\n",
LTIME_S(op_data->op_attr.ia_mtime),
LTIME_S(op_data->op_attr.ia_ctime));
mdc_setattr_pack(req, op_data, ea, ealen);
ptlrpc_request_set_replen(req);
rc = mdc_reint(req, LUSTRE_IMP_FULL);
if (rc == -ERESTARTSYS)
rc = 0;
*request = req;
return rc;
}
/**
* Lookup method for "fid" object. Only filenames with correct SEQ:OID format
* are valid. We also check if object with passed fid exists or not.
*/
static int obf_lookup(const struct lu_env *env, struct md_object *p,
const struct lu_name *lname, struct lu_fid *f,
struct md_op_spec *spec)
{
char *name = (char *)lname->ln_name;
struct mdd_device *mdd = mdo2mdd(p);
struct mdd_object *child;
int rc = 0;
while (*name == '[')
name++;
sscanf(name, SFID, RFID(f));
if (!fid_is_sane(f)) {
CWARN("%s: Trying to lookup invalid FID [%s] in %s/%s, FID "
"format should be "DFID"\n", mdd2obd_dev(mdd)->obd_name,
lname->ln_name, dot_lustre_name, mdd_obf_dir_name,
(__u64)FID_SEQ_NORMAL, 1, 0);
GOTO(out, rc = -EINVAL);
}
if (!fid_is_norm(f) && !fid_is_igif(f) && !fid_is_root(f) &&
!fid_seq_is_dot(f->f_seq)) {
CWARN("%s: Trying to lookup invalid FID "DFID" in %s/%s, "
"sequence should be >= "LPX64" or within ["LPX64","
""LPX64"].\n", mdd2obd_dev(mdd)->obd_name, PFID(f),
dot_lustre_name, mdd_obf_dir_name, (__u64)FID_SEQ_NORMAL,
(__u64)FID_SEQ_IGIF, (__u64)FID_SEQ_IGIF_MAX);
GOTO(out, rc = -EINVAL);
}
/* Check if object with this fid exists */
child = mdd_object_find(env, mdd, f);
if (child == NULL)
GOTO(out, rc = 0);
if (IS_ERR(child))
GOTO(out, rc = PTR_ERR(child));
if (mdd_object_exists(child) == 0)
rc = -ENOENT;
mdd_object_put(env, child);
out:
return rc;
}
static int out_index_insert(struct tgt_session_info *tsi)
{
struct tgt_thread_info *tti = tgt_th_info(tsi->tsi_env);
struct object_update *update = tti->tti_u.update.tti_update;
struct dt_object *obj = tti->tti_u.update.tti_dt_object;
struct lu_fid *fid;
char *name;
int rc = 0;
int size;
ENTRY;
name = object_update_param_get(update, 0, NULL);
if (name == NULL) {
CERROR("%s: empty name for index insert: rc = %d\n",
tgt_name(tsi->tsi_tgt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
fid = object_update_param_get(update, 1, &size);
if (fid == NULL || size != sizeof(*fid)) {
CERROR("%s: invalid fid: rc = %d\n",
tgt_name(tsi->tsi_tgt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
if (ptlrpc_req_need_swab(tsi->tsi_pill->rc_req))
lustre_swab_lu_fid(fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt), PFID(fid), -EPROTO);
RETURN(err_serious(-EPROTO));
}
rc = out_tx_index_insert(tsi->tsi_env, obj, name, fid,
&tti->tti_tea,
tti->tti_u.update.tti_update_reply,
tti->tti_u.update.tti_update_reply_index);
RETURN(rc);
}
static struct dentry *
ll_iget_for_nfs(struct super_block *sb, struct lu_fid *fid, struct lu_fid *parent)
{
struct inode *inode;
struct dentry *result;
CDEBUG(D_INFO, "Get dentry for fid: "DFID"\n", PFID(fid));
if (!fid_is_sane(fid))
return ERR_PTR(-ESTALE);
inode = search_inode_for_lustre(sb, fid);
if (IS_ERR(inode))
return ERR_CAST(inode);
if (is_bad_inode(inode)) {
/* we didn't find the right inode.. */
iput(inode);
return ERR_PTR(-ESTALE);
}
/**
* It is an anonymous dentry without OST objects created yet.
* We have to find the parent to tell MDS how to init lov objects.
*/
if (S_ISREG(inode->i_mode) && !ll_i2info(inode)->lli_has_smd &&
parent != NULL) {
struct ll_inode_info *lli = ll_i2info(inode);
spin_lock(&lli->lli_lock);
lli->lli_pfid = *parent;
spin_unlock(&lli->lli_lock);
}
result = d_obtain_alias(inode);
if (IS_ERR(result)) {
iput(inode);
return result;
}
return result;
}
/**
* Verify parent FID of an object.
*
* Check the parent FID is sane and start extended
* verification procedure otherwise.
*
* \param[in] env execution environment
* \param[in] fo OFD object
* \param[in] oa OBDO structure with PFID
*
* \retval 0 on successful verification
* \retval -EINPROGRESS if PFID is being repaired
* \retval -EPERM if PFID was verified but still insane
*/
int ofd_verify_ff(const struct lu_env *env, struct ofd_object *fo,
struct obdo *oa)
{
struct lu_fid *pfid = &fo->ofo_pfid;
int rc = 0;
ENTRY;
if (fid_is_sane(pfid)) {
if (likely(oa->o_parent_seq == pfid->f_seq &&
oa->o_parent_oid == pfid->f_oid &&
oa->o_stripe_idx == pfid->f_stripe_idx))
RETURN(0);
if (fo->ofo_pfid_verified)
RETURN(-EPERM);
}
/* The OST-object may be inconsistent, and we need further verification.
* To avoid block the RPC service thread, return -EINPROGRESS to client
* and make it retry later. */
if (fo->ofo_pfid_checking)
RETURN(-EINPROGRESS);
rc = ofd_object_ff_load(env, fo);
if (rc == -ENODATA)
RETURN(0);
if (rc < 0)
RETURN(rc);
if (likely(oa->o_parent_seq == pfid->f_seq &&
oa->o_parent_oid == pfid->f_oid &&
oa->o_stripe_idx == pfid->f_stripe_idx))
RETURN(0);
/* Push it to the dedicated thread for further verification. */
ofd_add_inconsistency_item(env, fo, oa);
RETURN(-EINPROGRESS);
}
static int out_index_insert(struct mdt_thread_info *info)
{
struct update *update = info->mti_u.update.mti_update;
struct dt_object *obj = info->mti_u.update.mti_dt_object;
struct lu_fid *fid;
char *name;
int rc = 0;
int size;
ENTRY;
name = (char *)update_param_buf(update, 0, NULL);
if (name == NULL) {
CERROR("%s: empty name for index insert: rc = %d\n",
mdt_obd_name(info->mti_mdt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
fid = (struct lu_fid *)update_param_buf(update, 1, &size);
if (fid == NULL || size != sizeof(*fid)) {
CERROR("%s: invalid fid: rc = %d\n",
mdt_obd_name(info->mti_mdt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
fid_le_to_cpu(fid, fid);
if (!fid_is_sane(fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
mdt_obd_name(info->mti_mdt), PFID(fid),
-EPROTO);
RETURN(err_serious(-EPROTO));
}
rc = out_tx_index_insert(info, obj, name, fid, &info->mti_handle,
info->mti_u.update.mti_update_reply,
info->mti_u.update.mti_update_reply_index);
RETURN(rc);
}
int lmv_fld_lookup(struct lmv_obd *lmv,
const struct lu_fid *fid,
mdsno_t *mds)
{
int rc;
ENTRY;
LASSERTF(fid_is_sane(fid), DFID" is insane!\n", PFID(fid));
/* FIXME: Because ZFS still use LOCAL fid sequence for root,
* and root will always be in MDT0, for local fid, it will
* return 0 directly. And it should be removed once the root
* FID has been assigned with special sequence */
if (fid_seq(fid) == FID_SEQ_LOCAL_FILE) {
*mds = 0;
RETURN(0);
}
rc = fld_client_lookup(&lmv->lmv_fld, fid_seq(fid), mds,
LU_SEQ_RANGE_MDT, NULL);
if (rc) {
CERROR("Error while looking for mds number. Seq "LPX64
", err = %d\n", fid_seq(fid), rc);
RETURN(rc);
}
CDEBUG(D_INODE, "FLD lookup got mds #%x for fid="DFID"\n",
*mds, PFID(fid));
if (*mds >= lmv->desc.ld_tgt_count) {
CERROR("FLD lookup got invalid mds #%x (max: %x) "
"for fid="DFID"\n", *mds, lmv->desc.ld_tgt_count,
PFID(fid));
rc = -EINVAL;
}
RETURN(rc);
}
static int osp_get_lastfid_from_ost(const struct lu_env *env,
struct osp_device *d)
{
struct ptlrpc_request *req = NULL;
struct obd_import *imp;
struct lu_fid *last_fid;
char *tmp;
int rc;
ENTRY;
imp = d->opd_obd->u.cli.cl_import;
LASSERT(imp);
req = ptlrpc_request_alloc(imp, &RQF_OST_GET_INFO_LAST_FID);
if (req == NULL)
RETURN(-ENOMEM);
req_capsule_set_size(&req->rq_pill, &RMF_GETINFO_KEY, RCL_CLIENT,
sizeof(KEY_LAST_FID));
rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
tmp = req_capsule_client_get(&req->rq_pill, &RMF_GETINFO_KEY);
memcpy(tmp, KEY_LAST_FID, sizeof(KEY_LAST_FID));
req->rq_no_delay = req->rq_no_resend = 1;
last_fid = req_capsule_client_get(&req->rq_pill, &RMF_FID);
fid_cpu_to_le(last_fid, &d->opd_last_used_fid);
ptlrpc_request_set_replen(req);
rc = ptlrpc_queue_wait(req);
if (rc) {
/* bad-bad OST.. let sysadm sort this out */
if (rc == -ENOTSUPP) {
CERROR("%s: server does not support FID: rc = %d\n",
d->opd_obd->obd_name, -ENOTSUPP);
}
ptlrpc_set_import_active(imp, 0);
GOTO(out, rc);
}
last_fid = req_capsule_server_get(&req->rq_pill, &RMF_FID);
if (last_fid == NULL) {
CERROR("%s: Got last_fid failed.\n", d->opd_obd->obd_name);
GOTO(out, rc = -EPROTO);
}
if (!fid_is_sane(last_fid)) {
CERROR("%s: Got insane last_fid "DFID"\n",
d->opd_obd->obd_name, PFID(last_fid));
GOTO(out, rc = -EPROTO);
}
/* Only update the last used fid, if the OST has objects for
* this sequence, i.e. fid_oid > 0 */
if (fid_oid(last_fid) > 0)
d->opd_last_used_fid = *last_fid;
CDEBUG(D_HA, "%s: Got last_fid "DFID"\n", d->opd_obd->obd_name,
PFID(last_fid));
out:
ptlrpc_req_finished(req);
RETURN(rc);
}
/**
* delete unreferenced files and directories in the PENDING directory
*
* Files that remain in PENDING after client->MDS recovery has completed
* have to be referenced (opened) by some client during recovery, or they
* will be deleted here (for clients that did not complete recovery).
*
* \param mdd MDD device finishing recovery
*
* \retval 0 success
* \retval -ve error
*/
static int orph_index_iterate(const struct lu_env *env,
struct mdd_device *mdd)
{
struct dt_object *dor = mdd->mdd_orphans;
struct lu_dirent *ent = &mdd_env_info(env)->mti_ent;
const struct dt_it_ops *iops;
struct dt_it *it;
struct lu_fid fid;
int key_sz = 0;
int rc;
__u64 cookie;
ENTRY;
/* In recovery phase, do not need for any lock here */
iops = &dor->do_index_ops->dio_it;
it = iops->init(env, dor, LUDA_64BITHASH, BYPASS_CAPA);
if (IS_ERR(it)) {
rc = PTR_ERR(it);
CERROR("%s: cannot clean PENDING: rc = %d\n",
mdd2obd_dev(mdd)->obd_name, rc);
GOTO(out, rc);
}
rc = iops->load(env, it, 0);
if (rc < 0)
GOTO(out_put, rc);
if (rc == 0) {
CERROR("%s: error loading iterator to clean PENDING\n",
mdd2obd_dev(mdd)->obd_name);
/* Index contains no zero key? */
GOTO(out_put, rc = -EIO);
}
do {
key_sz = iops->key_size(env, it);
/* filter out "." and ".." entries from PENDING dir. */
if (key_sz < 8)
goto next;
rc = iops->rec(env, it, (struct dt_rec *)ent, LUDA_64BITHASH);
if (rc != 0) {
CERROR("%s: fail to get FID for orphan it: rc = %d\n",
mdd2obd_dev(mdd)->obd_name, rc);
goto next;
}
fid_le_to_cpu(&fid, &ent->lde_fid);
if (!fid_is_sane(&fid)) {
CERROR("%s: bad FID "DFID" cleaning PENDING\n",
mdd2obd_dev(mdd)->obd_name, PFID(&fid));
goto next;
}
/* kill orphan object */
cookie = iops->store(env, it);
iops->put(env, it);
rc = orph_key_test_and_del(env, mdd, &fid,
(struct dt_key *)ent->lde_name);
/* after index delete reset iterator */
if (rc == 0)
rc = iops->get(env, it, (const void *)"");
else
rc = iops->load(env, it, cookie);
next:
rc = iops->next(env, it);
} while (rc == 0);
GOTO(out_put, rc = 0);
out_put:
iops->put(env, it);
iops->fini(env, it);
out:
return rc;
}
/**
* Object updates between Targets. Because all the updates has been
* dis-assemblied into object updates at sender side, so OUT will
* call OSD API directly to execute these updates.
*
* In DNE phase I all of the updates in the request need to be executed
* in one transaction, and the transaction has to be synchronously.
*
* Please refer to lustre/include/lustre/lustre_idl.h for req/reply
* format.
*/
int out_handle(struct tgt_session_info *tsi)
{
const struct lu_env *env = tsi->tsi_env;
struct tgt_thread_info *tti = tgt_th_info(env);
struct thandle_exec_args *ta = &tti->tti_tea;
struct req_capsule *pill = tsi->tsi_pill;
struct dt_device *dt = tsi->tsi_tgt->lut_bottom;
struct update_buf *ubuf;
struct update *update;
struct update_reply *update_reply;
int bufsize;
int count;
int old_batchid = -1;
unsigned off;
int i;
int rc = 0;
int rc1 = 0;
ENTRY;
req_capsule_set(pill, &RQF_UPDATE_OBJ);
bufsize = req_capsule_get_size(pill, &RMF_UPDATE, RCL_CLIENT);
if (bufsize != UPDATE_BUFFER_SIZE) {
CERROR("%s: invalid bufsize %d: rc = %d\n",
tgt_name(tsi->tsi_tgt), bufsize, -EPROTO);
RETURN(err_serious(-EPROTO));
}
ubuf = req_capsule_client_get(pill, &RMF_UPDATE);
if (ubuf == NULL) {
CERROR("%s: No buf!: rc = %d\n", tgt_name(tsi->tsi_tgt),
-EPROTO);
RETURN(err_serious(-EPROTO));
}
if (le32_to_cpu(ubuf->ub_magic) != UPDATE_BUFFER_MAGIC) {
CERROR("%s: invalid magic %x expect %x: rc = %d\n",
tgt_name(tsi->tsi_tgt), le32_to_cpu(ubuf->ub_magic),
UPDATE_BUFFER_MAGIC, -EPROTO);
RETURN(err_serious(-EPROTO));
}
count = le32_to_cpu(ubuf->ub_count);
if (count <= 0) {
CERROR("%s: No update!: rc = %d\n",
tgt_name(tsi->tsi_tgt), -EPROTO);
RETURN(err_serious(-EPROTO));
}
req_capsule_set_size(pill, &RMF_UPDATE_REPLY, RCL_SERVER,
UPDATE_BUFFER_SIZE);
rc = req_capsule_server_pack(pill);
if (rc != 0) {
CERROR("%s: Can't pack response: rc = %d\n",
tgt_name(tsi->tsi_tgt), rc);
RETURN(rc);
}
/* Prepare the update reply buffer */
update_reply = req_capsule_server_get(pill, &RMF_UPDATE_REPLY);
if (update_reply == NULL)
RETURN(err_serious(-EPROTO));
update_init_reply_buf(update_reply, count);
tti->tti_u.update.tti_update_reply = update_reply;
rc = out_tx_start(env, dt, ta);
if (rc != 0)
RETURN(rc);
tti->tti_mult_trans = !req_is_replay(tgt_ses_req(tsi));
/* Walk through updates in the request to execute them synchronously */
off = cfs_size_round(offsetof(struct update_buf, ub_bufs[0]));
for (i = 0; i < count; i++) {
struct tgt_handler *h;
struct dt_object *dt_obj;
update = (struct update *)((char *)ubuf + off);
if (old_batchid == -1) {
old_batchid = update->u_batchid;
} else if (old_batchid != update->u_batchid) {
/* Stop the current update transaction,
* create a new one */
rc = out_tx_end(env, ta);
if (rc != 0)
RETURN(rc);
rc = out_tx_start(env, dt, ta);
if (rc != 0)
RETURN(rc);
old_batchid = update->u_batchid;
}
fid_le_to_cpu(&update->u_fid, &update->u_fid);
if (!fid_is_sane(&update->u_fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt), PFID(&update->u_fid),
-EPROTO);
GOTO(out, rc = err_serious(-EPROTO));
}
dt_obj = dt_locate(env, dt, &update->u_fid);
if (IS_ERR(dt_obj))
GOTO(out, rc = PTR_ERR(dt_obj));
tti->tti_u.update.tti_dt_object = dt_obj;
tti->tti_u.update.tti_update = update;
tti->tti_u.update.tti_update_reply_index = i;
h = out_handler_find(update->u_type);
if (likely(h != NULL)) {
/* For real modification RPC, check if the update
* has been executed */
if (h->th_flags & MUTABOR) {
struct ptlrpc_request *req = tgt_ses_req(tsi);
if (out_check_resent(env, dt, dt_obj, req,
out_reconstruct,
update_reply, i))
GOTO(next, rc);
}
rc = h->th_act(tsi);
} else {
CERROR("%s: The unsupported opc: 0x%x\n",
tgt_name(tsi->tsi_tgt), update->u_type);
lu_object_put(env, &dt_obj->do_lu);
GOTO(out, rc = -ENOTSUPP);
}
next:
lu_object_put(env, &dt_obj->do_lu);
if (rc < 0)
GOTO(out, rc);
off += cfs_size_round(update_size(update));
}
out:
rc1 = out_tx_end(env, ta);
if (rc == 0)
rc = rc1;
RETURN(rc);
}
/*
* Handler for: getattr, lookup and revalidate cases.
*/
static int
lmv_intent_lookup(struct obd_export *exp, struct md_op_data *op_data,
struct lookup_intent *it, struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *tgt = NULL;
struct mdt_body *body;
struct lmv_stripe_md *lsm = op_data->op_mea1;
int rc = 0;
ENTRY;
/* If it returns ERR_PTR(-EBADFD) then it is an unknown hash type
* it will try all stripes to locate the object */
tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
if (IS_ERR(tgt) && (PTR_ERR(tgt) != -EBADFD))
RETURN(PTR_ERR(tgt));
/* Both migrating dir and unknown hash dir need to try
* all of sub-stripes */
if (lsm != NULL && !lmv_is_known_hash_type(lsm->lsm_md_hash_type)) {
struct lmv_oinfo *oinfo;
oinfo = &lsm->lsm_md_oinfo[0];
op_data->op_fid1 = oinfo->lmo_fid;
op_data->op_mds = oinfo->lmo_mds;
tgt = lmv_get_target(lmv, oinfo->lmo_mds, NULL);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
}
if (!fid_is_sane(&op_data->op_fid2))
fid_zero(&op_data->op_fid2);
CDEBUG(D_INODE, "LOOKUP_INTENT with fid1="DFID", fid2="DFID
", name='%s' -> mds #%u lsm=%p lsm_magic=%x\n",
PFID(&op_data->op_fid1), PFID(&op_data->op_fid2),
op_data->op_name ? op_data->op_name : "<NULL>",
tgt->ltd_idx, lsm, lsm == NULL ? -1 : lsm->lsm_md_magic);
op_data->op_bias &= ~MDS_CROSS_REF;
rc = md_intent_lock(tgt->ltd_exp, op_data, it, reqp, cb_blocking,
extra_lock_flags);
if (rc < 0)
RETURN(rc);
if (*reqp == NULL) {
/* If RPC happens, lsm information will be revalidated
* during update_inode process (see ll_update_lsm_md) */
if (op_data->op_mea2 != NULL) {
rc = lmv_revalidate_slaves(exp, op_data->op_mea2,
cb_blocking,
extra_lock_flags);
if (rc != 0)
RETURN(rc);
}
RETURN(rc);
} else if (it_disposition(it, DISP_LOOKUP_NEG) && lsm != NULL &&
lmv_need_try_all_stripes(lsm)) {
/* For migrating and unknown hash type directory, it will
* try to target the entry on other stripes */
int stripe_index;
for (stripe_index = 1;
stripe_index < lsm->lsm_md_stripe_count &&
it_disposition(it, DISP_LOOKUP_NEG); stripe_index++) {
struct lmv_oinfo *oinfo;
/* release the previous request */
ptlrpc_req_finished(*reqp);
it->d.lustre.it_data = NULL;
*reqp = NULL;
oinfo = &lsm->lsm_md_oinfo[stripe_index];
tgt = lmv_find_target(lmv, &oinfo->lmo_fid);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
CDEBUG(D_INODE, "Try other stripes " DFID"\n",
PFID(&oinfo->lmo_fid));
op_data->op_fid1 = oinfo->lmo_fid;
it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
rc = md_intent_lock(tgt->ltd_exp, op_data, it, reqp,
cb_blocking, extra_lock_flags);
if (rc != 0)
RETURN(rc);
}
}
/*
* MDS has returned success. Probably name has been resolved in
* remote inode. Let's check this.
*/
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
/* Not cross-ref case, just get out of here. */
if (unlikely((body->mbo_valid & OBD_MD_MDS))) {
rc = lmv_intent_remote(exp, it, NULL, reqp, cb_blocking,
extra_lock_flags);
if (rc != 0)
RETURN(rc);
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
}
RETURN(rc);
}
static int lmv_intent_remote(struct obd_export *exp, struct lookup_intent *it,
const struct lu_fid *parent_fid,
struct ptlrpc_request **reqp,
ldlm_blocking_callback cb_blocking,
__u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
struct ptlrpc_request *req = NULL;
struct lustre_handle plock;
struct md_op_data *op_data;
struct lmv_tgt_desc *tgt;
struct mdt_body *body;
int pmode;
int rc = 0;
ENTRY;
body = req_capsule_server_get(&(*reqp)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
LASSERT((body->mbo_valid & OBD_MD_MDS));
/*
* Unfortunately, we have to lie to MDC/MDS to retrieve
* attributes llite needs and provideproper locking.
*/
if (it->it_op & IT_LOOKUP)
it->it_op = IT_GETATTR;
/*
* We got LOOKUP lock, but we really need attrs.
*/
pmode = it->d.lustre.it_lock_mode;
if (pmode) {
plock.cookie = it->d.lustre.it_lock_handle;
it->d.lustre.it_lock_mode = 0;
it->d.lustre.it_data = NULL;
}
LASSERT(fid_is_sane(&body->mbo_fid1));
tgt = lmv_find_target(lmv, &body->mbo_fid1);
if (IS_ERR(tgt))
GOTO(out, rc = PTR_ERR(tgt));
OBD_ALLOC_PTR(op_data);
if (op_data == NULL)
GOTO(out, rc = -ENOMEM);
op_data->op_fid1 = body->mbo_fid1;
/* Sent the parent FID to the remote MDT */
if (parent_fid != NULL) {
/* The parent fid is only for remote open to
* check whether the open is from OBF,
* see mdt_cross_open */
LASSERT(it->it_op & IT_OPEN);
op_data->op_fid2 = *parent_fid;
}
op_data->op_bias = MDS_CROSS_REF;
CDEBUG(D_INODE, "REMOTE_INTENT with fid="DFID" -> mds #%u\n",
PFID(&body->mbo_fid1), tgt->ltd_idx);
rc = md_intent_lock(tgt->ltd_exp, op_data, it, &req, cb_blocking,
extra_lock_flags);
if (rc)
GOTO(out_free_op_data, rc);
/*
* LLite needs LOOKUP lock to track dentry revocation in order to
* maintain dcache consistency. Thus drop UPDATE|PERM lock here
* and put LOOKUP in request.
*/
if (it->d.lustre.it_lock_mode != 0) {
it->d.lustre.it_remote_lock_handle =
it->d.lustre.it_lock_handle;
it->d.lustre.it_remote_lock_mode = it->d.lustre.it_lock_mode;
}
if (pmode) {
it->d.lustre.it_lock_handle = plock.cookie;
it->d.lustre.it_lock_mode = pmode;
}
EXIT;
out_free_op_data:
OBD_FREE_PTR(op_data);
out:
if (rc && pmode)
ldlm_lock_decref(&plock, pmode);
ptlrpc_req_finished(*reqp);
*reqp = req;
return rc;
}
/**
* Object updates between Targets. Because all the updates has been
* dis-assemblied into object updates at sender side, so OUT will
* call OSD API directly to execute these updates.
*
* In DNE phase I all of the updates in the request need to be executed
* in one transaction, and the transaction has to be synchronously.
*
* Please refer to lustre/include/lustre/lustre_idl.h for req/reply
* format.
*/
int out_handle(struct tgt_session_info *tsi)
{
const struct lu_env *env = tsi->tsi_env;
struct tgt_thread_info *tti = tgt_th_info(env);
struct thandle_exec_args *ta = &tti->tti_tea;
struct req_capsule *pill = tsi->tsi_pill;
struct dt_device *dt = tsi->tsi_tgt->lut_bottom;
struct out_update_header *ouh;
struct out_update_buffer *oub = NULL;
struct object_update *update;
struct object_update_reply *reply;
struct ptlrpc_bulk_desc *desc = NULL;
struct l_wait_info lwi;
void **update_bufs;
int current_batchid = -1;
__u32 update_buf_count;
unsigned int i;
unsigned int reply_index = 0;
int rc = 0;
int rc1 = 0;
int ouh_size, reply_size;
int updates;
ENTRY;
req_capsule_set(pill, &RQF_OUT_UPDATE);
ouh_size = req_capsule_get_size(pill, &RMF_OUT_UPDATE_HEADER,
RCL_CLIENT);
if (ouh_size <= 0)
RETURN(err_serious(-EPROTO));
ouh = req_capsule_client_get(pill, &RMF_OUT_UPDATE_HEADER);
if (ouh == NULL)
RETURN(err_serious(-EPROTO));
if (ouh->ouh_magic != OUT_UPDATE_HEADER_MAGIC) {
CERROR("%s: invalid update buffer magic %x expect %x: "
"rc = %d\n", tgt_name(tsi->tsi_tgt), ouh->ouh_magic,
UPDATE_REQUEST_MAGIC, -EPROTO);
RETURN(err_serious(-EPROTO));
}
update_buf_count = ouh->ouh_count;
if (update_buf_count == 0)
RETURN(err_serious(-EPROTO));
OBD_ALLOC(update_bufs, sizeof(*update_bufs) * update_buf_count);
if (update_bufs == NULL)
RETURN(err_serious(-ENOMEM));
if (ouh->ouh_inline_length > 0) {
update_bufs[0] = ouh->ouh_inline_data;
} else {
struct out_update_buffer *tmp;
oub = req_capsule_client_get(pill, &RMF_OUT_UPDATE_BUF);
if (oub == NULL)
GOTO(out_free, rc = err_serious(-EPROTO));
desc = ptlrpc_prep_bulk_exp(pill->rc_req, update_buf_count,
PTLRPC_BULK_OPS_COUNT,
PTLRPC_BULK_GET_SINK |
PTLRPC_BULK_BUF_KVEC,
MDS_BULK_PORTAL,
&ptlrpc_bulk_kvec_ops);
if (desc == NULL)
GOTO(out_free, rc = err_serious(-ENOMEM));
tmp = oub;
for (i = 0; i < update_buf_count; i++, tmp++) {
if (tmp->oub_size >= OUT_MAXREQSIZE)
GOTO(out_free, rc = err_serious(-EPROTO));
OBD_ALLOC(update_bufs[i], tmp->oub_size);
if (update_bufs[i] == NULL)
GOTO(out_free, rc = err_serious(-ENOMEM));
desc->bd_frag_ops->add_iov_frag(desc, update_bufs[i],
tmp->oub_size);
}
pill->rc_req->rq_bulk_write = 1;
rc = sptlrpc_svc_prep_bulk(pill->rc_req, desc);
if (rc != 0)
GOTO(out_free, rc = err_serious(rc));
rc = target_bulk_io(pill->rc_req->rq_export, desc, &lwi);
if (rc < 0)
GOTO(out_free, rc = err_serious(rc));
}
/* validate the request and calculate the total update count and
* set it to reply */
reply_size = 0;
updates = 0;
for (i = 0; i < update_buf_count; i++) {
struct object_update_request *our;
int j;
our = update_bufs[i];
if (ptlrpc_req_need_swab(pill->rc_req))
lustre_swab_object_update_request(our);
if (our->ourq_magic != UPDATE_REQUEST_MAGIC) {
CERROR("%s: invalid update buffer magic %x"
" expect %x: rc = %d\n",
tgt_name(tsi->tsi_tgt), our->ourq_magic,
UPDATE_REQUEST_MAGIC, -EPROTO);
GOTO(out_free, rc = err_serious(-EPROTO));
}
updates += our->ourq_count;
/* need to calculate reply size */
for (j = 0; j < our->ourq_count; j++) {
update = object_update_request_get(our, j, NULL);
if (update == NULL)
GOTO(out, rc = err_serious(-EPROTO));
if (ptlrpc_req_need_swab(pill->rc_req))
lustre_swab_object_update(update);
if (!fid_is_sane(&update->ou_fid)) {
CERROR("%s: invalid FID "DFID": rc = %d\n",
tgt_name(tsi->tsi_tgt),
PFID(&update->ou_fid), -EPROTO);
GOTO(out, rc = err_serious(-EPROTO));
}
/* XXX: what ou_result_size can be considered safe? */
reply_size += sizeof(reply->ourp_lens[0]);
reply_size += sizeof(struct object_update_result);
reply_size += update->ou_result_size;
}
}
reply_size += sizeof(*reply);
if (unlikely(reply_size > ouh->ouh_reply_size)) {
CERROR("%s: too small reply buf %u for %u, need %u at least\n",
tgt_name(tsi->tsi_tgt), ouh->ouh_reply_size,
updates, reply_size);
GOTO(out_free, rc = err_serious(-EPROTO));
}
req_capsule_set_size(pill, &RMF_OUT_UPDATE_REPLY, RCL_SERVER,
ouh->ouh_reply_size);
rc = req_capsule_server_pack(pill);
if (rc != 0) {
CERROR("%s: Can't pack response: rc = %d\n",
tgt_name(tsi->tsi_tgt), rc);
GOTO(out_free, rc = err_serious(-EPROTO));
}
/* Prepare the update reply buffer */
reply = req_capsule_server_get(pill, &RMF_OUT_UPDATE_REPLY);
if (reply == NULL)
GOTO(out_free, rc = -EPROTO);
reply->ourp_magic = UPDATE_REPLY_MAGIC;
reply->ourp_count = updates;
tti->tti_u.update.tti_update_reply = reply;
tti->tti_mult_trans = !req_is_replay(tgt_ses_req(tsi));
/* Walk through updates in the request to execute them */
for (i = 0; i < update_buf_count; i++) {
struct tgt_handler *h;
struct dt_object *dt_obj;
int update_count;
struct object_update_request *our;
int j;
our = update_bufs[i];
update_count = our->ourq_count;
for (j = 0; j < update_count; j++) {
update = object_update_request_get(our, j, NULL);
dt_obj = dt_locate(env, dt, &update->ou_fid);
if (IS_ERR(dt_obj))
GOTO(out, rc = PTR_ERR(dt_obj));
if (dt->dd_record_fid_accessed) {
lfsck_pack_rfa(&tti->tti_lr,
lu_object_fid(&dt_obj->do_lu),
LE_FID_ACCESSED,
LFSCK_TYPE_LAYOUT);
tgt_lfsck_in_notify(env, dt, &tti->tti_lr,
NULL);
}
tti->tti_u.update.tti_dt_object = dt_obj;
tti->tti_u.update.tti_update = update;
tti->tti_u.update.tti_update_reply_index = reply_index;
h = out_handler_find(update->ou_type);
if (unlikely(h == NULL)) {
CERROR("%s: unsupported opc: 0x%x\n",
tgt_name(tsi->tsi_tgt), update->ou_type);
GOTO(next, rc = -ENOTSUPP);
}
/* Check resend case only for modifying RPC */
if (h->th_flags & MUTABOR) {
struct ptlrpc_request *req = tgt_ses_req(tsi);
if (out_check_resent(env, dt, dt_obj, req,
out_reconstruct, reply,
reply_index))
GOTO(next, rc = 0);
}
/* start transaction for modification RPC only */
if (h->th_flags & MUTABOR && current_batchid == -1) {
current_batchid = update->ou_batchid;
rc = out_tx_start(env, dt, ta, tsi->tsi_exp);
if (rc != 0)
GOTO(next, rc);
if (update->ou_flags & UPDATE_FL_SYNC)
ta->ta_handle->th_sync = 1;
}
/* Stop the current update transaction, if the update
* has different batchid, or read-only update */
if (((current_batchid != update->ou_batchid) ||
!(h->th_flags & MUTABOR)) &&
ta->ta_handle != NULL) {
rc = out_tx_end(env, ta, rc);
current_batchid = -1;
if (rc != 0)
GOTO(next, rc);
/* start a new transaction if needed */
if (h->th_flags & MUTABOR) {
rc = out_tx_start(env, dt, ta,
tsi->tsi_exp);
if (rc != 0)
GOTO(next, rc);
if (update->ou_flags & UPDATE_FL_SYNC)
ta->ta_handle->th_sync = 1;
current_batchid = update->ou_batchid;
}
}
rc = h->th_act(tsi);
next:
reply_index++;
lu_object_put(env, &dt_obj->do_lu);
if (rc < 0)
GOTO(out, rc);
}
}
out:
if (current_batchid != -1) {
rc1 = out_tx_end(env, ta, rc);
if (rc == 0)
rc = rc1;
}
out_free:
if (update_bufs != NULL) {
if (oub != NULL) {
for (i = 0; i < update_buf_count; i++, oub++) {
if (update_bufs[i] != NULL)
OBD_FREE(update_bufs[i], oub->oub_size);
}
}
OBD_FREE(update_bufs, sizeof(*update_bufs) * update_buf_count);
}
if (desc != NULL)
ptlrpc_free_bulk(desc);
RETURN(rc);
}
