/**
* Implements cl_io_operations::cio_prepare_write() method for osc layer.
*
* \retval -EIO transfer initiated against this osc will most likely fail
* \retval 0 transfer initiated against this osc will most likely succeed.
*
* The reason for this check is to immediately return an error to the caller
* in the case of a deactivated import. Note, that import can be deactivated
* later, while pages, dirtied by this IO, are still in the cache, but this is
* irrelevant, because that would still return an error to the application (if
* it does fsync), but many applications don't do fsync because of performance
* issues, and we wanted to return an -EIO at write time to notify the
* application.
*/
static int osc_io_prepare_write(const struct lu_env *env,
const struct cl_io_slice *ios,
const struct cl_page_slice *slice,
unsigned from, unsigned to)
{
struct osc_device *dev = lu2osc_dev(slice->cpl_obj->co_lu.lo_dev);
struct obd_import *imp = class_exp2cliimp(dev->od_exp);
struct osc_io *oio = cl2osc_io(env, ios);
int result = 0;
/*
* This implements OBD_BRW_CHECK logic from old client.
*/
if (imp == NULL || imp->imp_invalid)
result = -EIO;
if (result == 0 && oio->oi_lockless)
/* this page contains `invalid' data, but who cares?
* nobody can access the invalid data.
* in osc_io_commit_write(), we're going to write exact
* [from, to) bytes of this page to OST. -jay */
cl_page_export(env, slice->cpl_page, 1);
return result;
}
int osc_quotacheck(struct obd_device *unused, struct obd_export *exp,
struct obd_quotactl *oqctl)
{
struct client_obd *cli = &exp->exp_obd->u.cli;
struct ptlrpc_request *req;
struct obd_quotactl *body;
int rc;
ENTRY;
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
&RQF_OST_QUOTACHECK, LUSTRE_OST_VERSION,
OST_QUOTACHECK);
if (req == NULL)
RETURN(-ENOMEM);
body = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
*body = *oqctl;
ptlrpc_request_set_replen(req);
/* the next poll will find -ENODATA, that means quotacheck is
* going on */
cli->cl_qchk_stat = -ENODATA;
rc = ptlrpc_queue_wait(req);
if (rc)
cli->cl_qchk_stat = rc;
ptlrpc_req_finished(req);
RETURN(rc);
}
/**
* Check whether a qsd instance is all set to send quota request to master.
* This includes checking whether:
* - the connection to master is set up and usable,
* - the qsd isn't stopping
* - reintegration has been successfully completed and all indexes are
* up-to-date
*
* \param lqe - is the lquota entry for which we would like to send an quota
* request
* \param lockh - is the remote handle of the global lock returned on success
*
* \retval 0 on success, appropriate error on failure
*/
static int qsd_ready(struct lquota_entry *lqe, struct lustre_handle *lockh)
{
struct qsd_qtype_info *qqi = lqe2qqi(lqe);
struct qsd_instance *qsd = qqi->qqi_qsd;
struct obd_import *imp = NULL;
struct ldlm_lock *lock;
ENTRY;
read_lock(&qsd->qsd_lock);
/* is the qsd about to shut down? */
if (qsd->qsd_stopping) {
read_unlock(&qsd->qsd_lock);
LQUOTA_DEBUG(lqe, "dropping quota req since qsd is stopping");
/* Target is about to shut down, client will retry */
RETURN(-EINPROGRESS);
}
/* is the connection to the quota master ready? */
if (qsd->qsd_exp_valid)
imp = class_exp2cliimp(qsd->qsd_exp);
if (imp == NULL || imp->imp_invalid) {
read_unlock(&qsd->qsd_lock);
LQUOTA_DEBUG(lqe, "connection to master not ready");
RETURN(-ENOTCONN);
}
/* In most case, reintegration must have been triggered (when enable
* quota or on OST start), however, in rare race condition (enabling
* quota when starting OSTs), we might miss triggering reintegration
* for some qqi.
*
* If the previous reintegration failed for some reason, we'll
* re-trigger it here as well. */
if (!qqi->qqi_glb_uptodate || !qqi->qqi_slv_uptodate) {
read_unlock(&qsd->qsd_lock);
LQUOTA_DEBUG(lqe, "not up-to-date, dropping request and "
"kicking off reintegration");
qsd_start_reint_thread(qqi);
RETURN(-EINPROGRESS);
}
/* Fill the remote global lock handle, master will check this handle
* to see if the slave is sending request with stale lock */
lustre_handle_copy(lockh, &qqi->qqi_lockh);
read_unlock(&qsd->qsd_lock);
if (!lustre_handle_is_used(lockh))
RETURN(-ENOLCK);
lock = ldlm_handle2lock(lockh);
if (lock == NULL)
RETURN(-ENOLCK);
/* return remote lock handle to be packed in quota request */
lustre_handle_copy(lockh, &lock->l_remote_handle);
LDLM_LOCK_PUT(lock);
RETURN(0);
}
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);
}
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)
{
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);
}
/*
* Send non-intent quota request to master.
*
* \param env - the environment passed by the caller
* \param exp - is the export to use to send the acquire RPC
* \param qbody - quota body to be packed in request
* \param sync - synchronous or asynchronous
* \param completion - completion callback
* \param qqi - is the qsd_qtype_info structure to pass to the completion
* function
* \param lqe - is the qid entry to be processed
*
* \retval 0 - success
* \retval -ve - appropriate errors
*/
int qsd_send_dqacq(const struct lu_env *env, struct obd_export *exp,
struct quota_body *qbody, bool sync,
qsd_req_completion_t completion, struct qsd_qtype_info *qqi,
struct lustre_handle *lockh, struct lquota_entry *lqe)
{
struct ptlrpc_request *req;
struct quota_body *req_qbody;
struct qsd_async_args *aa;
int rc;
ENTRY;
LASSERT(exp);
req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_QUOTA_DQACQ);
if (req == NULL)
GOTO(out, rc = -ENOMEM);
req->rq_no_resend = req->rq_no_delay = 1;
req->rq_no_retry_einprogress = 1;
rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, QUOTA_DQACQ);
if (rc) {
ptlrpc_request_free(req);
GOTO(out, rc);
}
req->rq_request_portal = MDS_READPAGE_PORTAL;
req_qbody = req_capsule_client_get(&req->rq_pill, &RMF_QUOTA_BODY);
*req_qbody = *qbody;
ptlrpc_request_set_replen(req);
CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
aa = ptlrpc_req_async_args(req);
aa->aa_exp = exp;
aa->aa_qqi = qqi;
aa->aa_arg = (void *)lqe;
aa->aa_completion = completion;
lustre_handle_copy(&aa->aa_lockh, lockh);
if (sync) {
rc = ptlrpc_queue_wait(req);
rc = qsd_dqacq_interpret(env, req, aa, rc);
ptlrpc_req_finished(req);
} else {
req->rq_interpret_reply = qsd_dqacq_interpret;
ptlrpcd_add_req(req);
}
RETURN(rc);
out:
completion(env, qqi, qbody, NULL, lockh, NULL, lqe, rc);
return rc;
}
static int osc_io_data_version_start(const struct lu_env *env,
const struct cl_io_slice *slice)
{
struct cl_data_version_io *dv = &slice->cis_io->u.ci_data_version;
struct osc_io *oio = cl2osc_io(env, slice);
struct obdo *oa = &oio->oi_oa;
struct osc_async_cbargs *cbargs = &oio->oi_cbarg;
struct osc_object *obj = cl2osc(slice->cis_obj);
struct lov_oinfo *loi = obj->oo_oinfo;
struct obd_export *exp = osc_export(obj);
struct ptlrpc_request *req;
struct ost_body *body;
struct osc_data_version_args *dva;
int rc;
ENTRY;
memset(oa, 0, sizeof(*oa));
oa->o_oi = loi->loi_oi;
oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
if (dv->dv_flags & (LL_DV_RD_FLUSH | LL_DV_WR_FLUSH)) {
oa->o_valid |= OBD_MD_FLFLAGS;
oa->o_flags |= OBD_FL_SRVLOCK;
if (dv->dv_flags & LL_DV_WR_FLUSH)
oa->o_flags |= OBD_FL_FLUSH;
}
init_completion(&cbargs->opc_sync);
req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
if (req == NULL)
RETURN(-ENOMEM);
rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
if (rc < 0) {
ptlrpc_request_free(req);
RETURN(rc);
}
body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
ptlrpc_request_set_replen(req);
req->rq_interpret_reply = osc_data_version_interpret;
CLASSERT(sizeof(*dva) <= sizeof(req->rq_async_args));
dva = ptlrpc_req_async_args(req);
dva->dva_oio = oio;
ptlrpcd_add_req(req);
RETURN(0);
}
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 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;
}
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);
}
int client_quota_check(struct obd_device *unused, struct obd_export *exp,
struct obd_quotactl *oqctl)
{
struct client_obd *cli = &exp->exp_obd->u.cli;
struct ptlrpc_request *req;
struct obd_quotactl *body;
const struct req_format *rf;
int ver, opc, rc;
ENTRY;
if (!strcmp(exp->exp_obd->obd_type->typ_name, LUSTRE_MDC_NAME)) {
rf = &RQF_MDS_QUOTACHECK;
ver = LUSTRE_MDS_VERSION;
opc = MDS_QUOTACHECK;
} else if (!strcmp(exp->exp_obd->obd_type->typ_name, LUSTRE_OSC_NAME)) {
rf = &RQF_OST_QUOTACHECK;
ver = LUSTRE_OST_VERSION;
opc = OST_QUOTACHECK;
} else {
RETURN(-EINVAL);
}
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), rf, ver, opc);
if (req == NULL)
RETURN(-ENOMEM);
body = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
*body = *oqctl;
ptlrpc_request_set_replen(req);
/* the next poll will find -ENODATA, that means quotacheck is
* going on */
cli->cl_qchk_stat = -ENODATA;
rc = ptlrpc_queue_wait(req);
if (rc)
cli->cl_qchk_stat = rc;
ptlrpc_req_finished(req);
RETURN(rc);
}
int osc_quotactl(struct obd_device *unused, struct obd_export *exp,
struct obd_quotactl *oqctl)
{
struct ptlrpc_request *req;
struct obd_quotactl *oqc;
int rc;
ENTRY;
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
&RQF_OST_QUOTACTL, LUSTRE_OST_VERSION,
OST_QUOTACTL);
if (req == NULL)
RETURN(-ENOMEM);
oqc = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
*oqc = *oqctl;
ptlrpc_request_set_replen(req);
ptlrpc_at_set_req_timeout(req);
req->rq_no_resend = 1;
rc = ptlrpc_queue_wait(req);
if (rc)
CERROR("ptlrpc_queue_wait failed, rc: %d\n", rc);
if (req->rq_repmsg &&
(oqc = req_capsule_server_get(&req->rq_pill, &RMF_OBD_QUOTACTL))) {
*oqctl = *oqc;
} else if (!rc) {
CERROR ("Can't unpack obd_quotactl\n");
rc = -EPROTO;
}
ptlrpc_req_finished(req);
RETURN(rc);
}
int client_quota_adjust_qunit(struct obd_export *exp,
struct quota_adjust_qunit *oqaq,
struct lustre_quota_ctxt *qctxt,
struct ptlrpc_request_set *rqset)
{
struct ptlrpc_request *req;
struct quota_adjust_qunit *oqa;
int rc = 0;
ENTRY;
/* client don't support this kind of operation, abort it */
if (!(exp->exp_connect_flags & OBD_CONNECT_CHANGE_QS)) {
CDEBUG(D_QUOTA, "osc: %s don't support change qunit size\n",
exp->exp_obd->obd_name);
RETURN(rc);
}
if (strcmp(exp->exp_obd->obd_type->typ_name, LUSTRE_OSC_NAME))
RETURN(-EINVAL);
LASSERT(rqset);
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
&RQF_OST_QUOTA_ADJUST_QUNIT,
LUSTRE_OST_VERSION,
OST_QUOTA_ADJUST_QUNIT);
if (req == NULL)
RETURN(-ENOMEM);
oqa = req_capsule_client_get(&req->rq_pill, &RMF_QUOTA_ADJUST_QUNIT);
*oqa = *oqaq;
ptlrpc_request_set_replen(req);
ptlrpc_set_add_req(rqset, req);
RETURN (rc);
}
static int seq_client_rpc(struct lu_client_seq *seq,
struct lu_seq_range *output, __u32 opc,
const char *opcname)
{
struct obd_export *exp = seq->lcs_exp;
struct ptlrpc_request *req;
struct lu_seq_range *out, *in;
__u32 *op;
unsigned int debug_mask;
int rc;
ENTRY;
LASSERT(exp != NULL && !IS_ERR(exp));
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_SEQ_QUERY,
LUSTRE_MDS_VERSION, SEQ_QUERY);
if (req == NULL)
RETURN(-ENOMEM);
/* Init operation code */
op = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_OPC);
*op = opc;
/* Zero out input range, this is not recovery yet. */
in = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_RANGE);
lu_seq_range_init(in);
ptlrpc_request_set_replen(req);
in->lsr_index = seq->lcs_space.lsr_index;
if (seq->lcs_type == LUSTRE_SEQ_METADATA)
fld_range_set_mdt(in);
else
fld_range_set_ost(in);
if (opc == SEQ_ALLOC_SUPER) {
req->rq_request_portal = SEQ_CONTROLLER_PORTAL;
req->rq_reply_portal = MDC_REPLY_PORTAL;
/* During allocating super sequence for data object,
* the current thread might hold the export of MDT0(MDT0
* precreating objects on this OST), and it will send the
* request to MDT0 here, so we can not keep resending the
* request here, otherwise if MDT0 is failed(umounted),
* it can not release the export of MDT0 */
if (seq->lcs_type == LUSTRE_SEQ_DATA)
req->rq_no_delay = req->rq_no_resend = 1;
debug_mask = D_CONSOLE;
} else {
if (seq->lcs_type == LUSTRE_SEQ_METADATA) {
req->rq_reply_portal = MDC_REPLY_PORTAL;
req->rq_request_portal = SEQ_METADATA_PORTAL;
} else {
req->rq_reply_portal = OSC_REPLY_PORTAL;
req->rq_request_portal = SEQ_DATA_PORTAL;
}
debug_mask = D_INFO;
}
/* Allow seq client RPC during recovery time. */
req->rq_allow_replay = 1;
ptlrpc_at_set_req_timeout(req);
rc = ptlrpc_queue_wait(req);
if (rc)
GOTO(out_req, rc);
out = req_capsule_server_get(&req->rq_pill, &RMF_SEQ_RANGE);
*output = *out;
if (!lu_seq_range_is_sane(output)) {
CERROR("%s: Invalid range received from server: "
DRANGE"\n", seq->lcs_name, PRANGE(output));
GOTO(out_req, rc = -EINVAL);
}
if (lu_seq_range_is_exhausted(output)) {
CERROR("%s: Range received from server is exhausted: "
DRANGE"]\n", seq->lcs_name, PRANGE(output));
GOTO(out_req, rc = -EINVAL);
}
CDEBUG_LIMIT(debug_mask, "%s: Allocated %s-sequence "DRANGE"]\n",
seq->lcs_name, opcname, PRANGE(output));
EXIT;
out_req:
ptlrpc_req_finished(req);
return rc;
}
static int osc_object_fiemap(const struct lu_env *env, struct cl_object *obj,
struct ll_fiemap_info_key *fmkey,
struct fiemap *fiemap, size_t *buflen)
{
struct obd_export *exp = osc_export(cl2osc(obj));
struct ldlm_res_id resid;
union ldlm_policy_data policy;
struct lustre_handle lockh;
enum ldlm_mode mode = LCK_MINMODE;
struct ptlrpc_request *req;
struct fiemap *reply;
char *tmp;
int rc;
ENTRY;
fmkey->lfik_oa.o_oi = cl2osc(obj)->oo_oinfo->loi_oi;
if (!(fmkey->lfik_fiemap.fm_flags & FIEMAP_FLAG_SYNC))
goto skip_locking;
policy.l_extent.start = fmkey->lfik_fiemap.fm_start & PAGE_MASK;
if (OBD_OBJECT_EOF - fmkey->lfik_fiemap.fm_length <=
fmkey->lfik_fiemap.fm_start + PAGE_SIZE - 1)
policy.l_extent.end = OBD_OBJECT_EOF;
else
policy.l_extent.end = (fmkey->lfik_fiemap.fm_start +
fmkey->lfik_fiemap.fm_length +
PAGE_SIZE - 1) & PAGE_MASK;
ostid_build_res_name(&fmkey->lfik_oa.o_oi, &resid);
mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
LDLM_FL_BLOCK_GRANTED | LDLM_FL_LVB_READY,
&resid, LDLM_EXTENT, &policy,
LCK_PR | LCK_PW, &lockh, 0);
if (mode) { /* lock is cached on client */
if (mode != LCK_PR) {
ldlm_lock_addref(&lockh, LCK_PR);
ldlm_lock_decref(&lockh, LCK_PW);
}
} else { /* no cached lock, needs acquire lock on server side */
fmkey->lfik_oa.o_valid |= OBD_MD_FLFLAGS;
fmkey->lfik_oa.o_flags |= OBD_FL_SRVLOCK;
}
skip_locking:
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_OST_GET_INFO_FIEMAP);
if (req == NULL)
GOTO(drop_lock, rc = -ENOMEM);
req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY, RCL_CLIENT,
sizeof(*fmkey));
req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL, RCL_CLIENT,
*buflen);
req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL, RCL_SERVER,
*buflen);
rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
if (rc != 0) {
ptlrpc_request_free(req);
GOTO(drop_lock, rc);
}
tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
memcpy(tmp, fmkey, sizeof(*fmkey));
tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
memcpy(tmp, fiemap, *buflen);
ptlrpc_request_set_replen(req);
rc = ptlrpc_queue_wait(req);
if (rc != 0)
GOTO(fini_req, rc);
reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
if (reply == NULL)
GOTO(fini_req, rc = -EPROTO);
memcpy(fiemap, reply, *buflen);
fini_req:
ptlrpc_req_finished(req);
drop_lock:
if (mode)
ldlm_lock_decref(&lockh, LCK_PR);
RETURN(rc);
}
int mdc_create(struct obd_export *exp, struct md_op_data *op_data,
const void *data, int datalen, int mode, __u32 uid, __u32 gid,
cfs_cap_t cap_effective, __u64 rdev,
struct ptlrpc_request **request)
{
struct ptlrpc_request *req;
int level, rc;
int count, resends = 0;
struct obd_import *import = exp->exp_obd->u.cli.cl_import;
int generation = import->imp_generation;
CFS_LIST_HEAD(cancels);
ENTRY;
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2)) {
/*
* mdc_fid_alloc() may return errno 1 in case of switch to new
* sequence, handle this.
*/
rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
if (rc < 0) {
CERROR("Can't alloc new fid, rc %d\n", rc);
RETURN(rc);
}
}
rebuild:
count = 0;
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_CREATE_RMT_ACL);
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);
req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
data && datalen ? datalen : 0);
rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
/*
* mdc_create_pack() fills msg->bufs[1] with name and msg->bufs[2] with
* tgt, for symlinks or lov MD data.
*/
mdc_create_pack(req, op_data, data, datalen, mode, uid,
gid, cap_effective, rdev);
ptlrpc_request_set_replen(req);
/* ask ptlrpc not to resend on EINPROGRESS since we have our own retry
* logic here */
req->rq_no_retry_einprogress = 1;
if (resends) {
req->rq_generation_set = 1;
req->rq_import_generation = generation;
req->rq_sent = cfs_time_current_sec() + resends;
}
level = LUSTRE_IMP_FULL;
resend:
rc = mdc_reint(req, exp->exp_obd->u.cli.cl_rpc_lock, level);
/* Resend if we were told to. */
if (rc == -ERESTARTSYS) {
level = LUSTRE_IMP_RECOVER;
goto resend;
} else if (rc == -EINPROGRESS) {
/* Retry create infinitely until succeed or get other
* error code. */
ptlrpc_req_finished(req);
resends++;
CDEBUG(D_HA, "%s: resend:%d create on "DFID"/"DFID"\n",
exp->exp_obd->obd_name, resends,
PFID(&op_data->op_fid1), PFID(&op_data->op_fid2));
if (generation == import->imp_generation) {
goto rebuild;
} else {
CDEBUG(D_HA, "resend cross eviction\n");
RETURN(-EIO);
}
} else if (rc == 0) {
struct mdt_body *body;
struct lustre_capa *capa;
body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
LASSERT(body);
if (body->valid & OBD_MD_FLMDSCAPA) {
capa = req_capsule_server_get(&req->rq_pill,
&RMF_CAPA1);
if (capa == NULL)
rc = -EPROTO;
}
}
*request = req;
RETURN(rc);
}
/*
* Fetch a global or slave index from the QMT.
*
* \param env - the environment passed by the caller
* \param exp - is the export to use to issue the OBD_IDX_READ RPC
* \param ii - is the index information to be packed in the request
* on success, the index information returned by the server
* is copied there.
* \param npages - is the number of pages in the pages array
* \param pages - is an array of @npages pages
*
* \retval 0 - success
* \retval -ve - appropriate errors
*/
int qsd_fetch_index(const struct lu_env *env, struct obd_export *exp,
struct idx_info *ii, unsigned int npages,
struct page **pages, bool *need_swab)
{
struct ptlrpc_request *req;
struct idx_info *req_ii;
struct ptlrpc_bulk_desc *desc;
int rc, i;
ENTRY;
LASSERT(exp);
req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OBD_IDX_READ);
if (req == NULL)
RETURN(-ENOMEM);
rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, OBD_IDX_READ);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
req->rq_request_portal = MDS_READPAGE_PORTAL;
ptlrpc_at_set_req_timeout(req);
/* allocate bulk descriptor */
desc = ptlrpc_prep_bulk_imp(req, npages, 1, BULK_PUT_SINK,
MDS_BULK_PORTAL);
if (desc == NULL) {
ptlrpc_request_free(req);
RETURN(-ENOMEM);
}
/* req now owns desc and will free it when it gets freed */
for (i = 0; i < npages; i++)
ptlrpc_prep_bulk_page_pin(desc, pages[i], 0, PAGE_CACHE_SIZE);
/* pack index information in request */
req_ii = req_capsule_client_get(&req->rq_pill, &RMF_IDX_INFO);
*req_ii = *ii;
ptlrpc_request_set_replen(req);
/* send request to master and wait for RPC to complete */
rc = ptlrpc_queue_wait(req);
if (rc)
GOTO(out, rc);
rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk,
req->rq_bulk->bd_nob_transferred);
if (rc < 0)
GOTO(out, rc);
else
/* sptlrpc_cli_unwrap_bulk_read() returns the number of bytes
* transferred*/
rc = 0;
req_ii = req_capsule_server_get(&req->rq_pill, &RMF_IDX_INFO);
*ii = *req_ii;
*need_swab = ptlrpc_rep_need_swab(req);
EXIT;
out:
ptlrpc_req_finished(req);
return rc;
}
/*
* Get intent per-ID lock or global-index lock from master.
*
* \param env - the environment passed by the caller
* \param exp - is the export to use to send the intent RPC
* \param qbody - quota body to be packed in request
* \param sync - synchronous or asynchronous (pre-acquire)
* \param it_op - IT_QUOTA_DQACQ or IT_QUOTA_CONN
* \param completion - completion callback
* \param qqi - is the qsd_qtype_info structure to pass to the completion
* function
* \param lvb - is the lvb associated with the lock and returned by the
* server
* \param arg - is an opaq argument passed to the completion callback
*
* \retval 0 - success
* \retval -ve - appropriate errors
*/
int qsd_intent_lock(const struct lu_env *env, struct obd_export *exp,
struct quota_body *qbody, bool sync, int it_op,
qsd_req_completion_t completion, struct qsd_qtype_info *qqi,
struct lquota_lvb *lvb, void *arg)
{
struct qsd_thread_info *qti = qsd_info(env);
struct ptlrpc_request *req;
struct qsd_async_args *aa = NULL;
struct ldlm_intent *lit;
struct quota_body *req_qbody;
__u64 flags = LDLM_FL_HAS_INTENT;
int rc;
ENTRY;
LASSERT(exp != NULL);
LASSERT(!lustre_handle_is_used(&qbody->qb_lockh));
memset(&qti->qti_lockh, 0, sizeof(qti->qti_lockh));
req = ptlrpc_request_alloc(class_exp2cliimp(exp),
&RQF_LDLM_INTENT_QUOTA);
if (req == NULL)
GOTO(out, rc = -ENOMEM);
req->rq_no_retry_einprogress = 1;
rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
if (rc) {
ptlrpc_request_free(req);
GOTO(out, rc);
}
lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
lit->opc = (__u64)it_op;
req_qbody = req_capsule_client_get(&req->rq_pill, &RMF_QUOTA_BODY);
*req_qbody = *qbody;
req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
sizeof(*lvb));
ptlrpc_request_set_replen(req);
switch(it_op) {
case IT_QUOTA_CONN:
/* build resource name associated with global index */
fid_build_reg_res_name(&qbody->qb_fid, &qti->qti_resid);
/* copy einfo template and fill ei_cbdata with qqi pointer */
memcpy(&qti->qti_einfo, &qsd_glb_einfo, sizeof(qti->qti_einfo));
qti->qti_einfo.ei_cbdata = qqi;
/* don't cancel global lock on memory pressure */
flags |= LDLM_FL_NO_LRU;
break;
case IT_QUOTA_DQACQ:
/* build resource name associated for per-ID quota lock */
fid_build_quota_res_name(&qbody->qb_fid, &qbody->qb_id,
&qti->qti_resid);
/* copy einfo template and fill ei_cbdata with lqe pointer */
memcpy(&qti->qti_einfo, &qsd_id_einfo, sizeof(qti->qti_einfo));
qti->qti_einfo.ei_cbdata = arg;
break;
default:
LASSERTF(0, "invalid it_op %d", it_op);
}
/* build lock enqueue request */
rc = ldlm_cli_enqueue(exp, &req, &qti->qti_einfo, &qti->qti_resid, NULL,
&flags, (void *)lvb, sizeof(*lvb), LVB_T_LQUOTA,
&qti->qti_lockh, 1);
if (rc < 0) {
ptlrpc_req_finished(req);
GOTO(out, rc);
}
/* grab reference on backend structure for the new lock */
switch(it_op) {
case IT_QUOTA_CONN:
/* grab reference on qqi for new lock */
#ifdef USE_LU_REF
{
struct ldlm_lock *lock;
lock = ldlm_handle2lock(&qti->qti_lockh);
if (lock == NULL) {
ptlrpc_req_finished(req);
GOTO(out, rc = -ENOLCK);
}
lu_ref_add(&qqi->qqi_reference, "glb_lock", lock);
LDLM_LOCK_PUT(lock);
}
#endif
qqi_getref(qqi);
break;
case IT_QUOTA_DQACQ:
/* grab reference on lqe for new lock */
lqe_getref((struct lquota_entry *)arg);
/* all acquire/release request are sent with no_resend and
* no_delay flag */
req->rq_no_resend = req->rq_no_delay = 1;
break;
default:
break;
}
CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
aa = ptlrpc_req_async_args(req);
aa->aa_exp = exp;
aa->aa_qqi = qqi;
aa->aa_arg = arg;
aa->aa_lvb = lvb;
aa->aa_completion = completion;
lustre_handle_copy(&aa->aa_lockh, &qti->qti_lockh);
if (sync) {
/* send lock enqueue request and wait for completion */
rc = ptlrpc_queue_wait(req);
rc = qsd_intent_interpret(env, req, aa, rc);
ptlrpc_req_finished(req);
} else {
/* queue lock request and return */
req->rq_interpret_reply = qsd_intent_interpret;
ptlrpcd_add_req(req, PDL_POLICY_LOCAL, -1);
}
RETURN(rc);
out:
completion(env, qqi, qbody, NULL, &qti->qti_lockh, lvb, arg, rc);
return rc;
}
int mdc_create(struct obd_export *exp, struct md_op_data *op_data,
const void *data, size_t datalen,
umode_t mode, uid_t uid, gid_t gid,
cfs_cap_t cap_effective, __u64 rdev,
struct ptlrpc_request **request)
{
struct ptlrpc_request *req;
int level, rc;
int count, resends = 0;
struct obd_import *import = exp->exp_obd->u.cli.cl_import;
int generation = import->imp_generation;
struct list_head cancels = LIST_HEAD_INIT(cancels);
ENTRY;
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2)) {
/*
* mdc_fid_alloc() may return errno 1 in case of switch to new
* sequence, handle this.
*/
rc = mdc_fid_alloc(NULL, exp, &op_data->op_fid2, op_data);
if (rc < 0)
RETURN(rc);
}
rebuild:
count = 0;
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_CREATE_ACL);
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);
req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
data && datalen ? datalen : 0);
req_capsule_set_size(&req->rq_pill, &RMF_FILE_SECCTX_NAME,
RCL_CLIENT, op_data->op_file_secctx_name != NULL ?
strlen(op_data->op_file_secctx_name) + 1 : 0);
req_capsule_set_size(&req->rq_pill, &RMF_FILE_SECCTX, RCL_CLIENT,
op_data->op_file_secctx_size);
/* 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_create_pack() fills msg->bufs[1] with name and msg->bufs[2] with
* tgt, for symlinks or lov MD data.
*/
mdc_create_pack(req, op_data, data, datalen, mode, uid,
gid, cap_effective, rdev);
ptlrpc_request_set_replen(req);
/* ask ptlrpc not to resend on EINPROGRESS since we have our own retry
* logic here */
req->rq_no_retry_einprogress = 1;
if (resends) {
req->rq_generation_set = 1;
req->rq_import_generation = generation;
req->rq_sent = ktime_get_real_seconds() + resends;
}
level = LUSTRE_IMP_FULL;
resend:
rc = mdc_reint(req, level);
/* Resend if we were told to. */
if (rc == -ERESTARTSYS) {
level = LUSTRE_IMP_RECOVER;
goto resend;
} else if (rc == -EINPROGRESS) {
/* Retry create infinitely until succeed or get other
* error code or interrupted. */
ptlrpc_req_finished(req);
if (generation == import->imp_generation) {
if (signal_pending(current))
RETURN(-EINTR);
resends++;
CDEBUG(D_HA, "%s: resend:%d create on "DFID"/"DFID"\n",
exp->exp_obd->obd_name, resends,
PFID(&op_data->op_fid1),
PFID(&op_data->op_fid2));
goto rebuild;
} else {
CDEBUG(D_HA, "resend cross eviction\n");
RETURN(-EIO);
}
}
*request = req;
RETURN(rc);
}
int mdc_create(struct obd_export *exp, struct md_op_data *op_data,
const void *data, int datalen, int mode, __u32 uid, __u32 gid,
cfs_cap_t cap_effective, __u64 rdev,
struct ptlrpc_request **request)
{
struct ptlrpc_request *req;
int level, rc;
int count = 0;
CFS_LIST_HEAD(cancels);
ENTRY;
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2)) {
/*
* mdc_fid_alloc() may return errno 1 in case of switch to new
* sequence, handle this.
*/
rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
if (rc < 0) {
CERROR("Can't alloc new fid, rc %d\n", rc);
RETURN(rc);
}
}
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_CREATE_RMT_ACL);
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);
req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
data && datalen ? datalen : 0);
rc = mdc_prep_elc_req(exp, req, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
/*
* mdc_create_pack() fills msg->bufs[1] with name and msg->bufs[2] with
* tgt, for symlinks or lov MD data.
*/
mdc_create_pack(req, op_data, data, datalen, mode, uid,
gid, cap_effective, rdev);
ptlrpc_request_set_replen(req);
level = LUSTRE_IMP_FULL;
resend:
rc = mdc_reint(req, exp->exp_obd->u.cli.cl_rpc_lock, level);
/* Resend if we were told to. */
if (rc == -ERESTARTSYS) {
level = LUSTRE_IMP_RECOVER;
goto resend;
} else if (rc == 0) {
struct mdt_body *body;
struct lustre_capa *capa;
body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
LASSERT(body);
if (body->valid & OBD_MD_FLMDSCAPA) {
capa = req_capsule_server_get(&req->rq_pill,
&RMF_CAPA1);
if (capa == NULL)
rc = -EPROTO;
}
}
*request = req;
RETURN(rc);
}
/* If mdc_setattr is called with an 'iattr', then it is a normal RPC that
* should take the normal semaphore and go to the normal portal.
*
* If it is called with iattr->ia_valid & ATTR_FROM_OPEN, then it is a
* magic open-path setattr that should take the setattr semaphore and
* go to the setattr portal. */
int mdc_setattr(struct obd_export *exp, struct md_op_data *op_data,
void *ea, int ealen, void *ea2, int ea2len,
struct ptlrpc_request **request, struct md_open_data **mod)
{
CFS_LIST_HEAD(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);
if ((op_data->op_flags & (MF_SOM_CHANGE | MF_EPOCH_OPEN)) == 0)
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,
ea2len);
rc = mdc_prep_elc_req(exp, req, &cancels, count);
if (rc) {
ptlrpc_request_free(req);
RETURN(rc);
}
if (op_data->op_attr.ia_valid & ATTR_FROM_OPEN) {
req->rq_request_portal = MDS_SETATTR_PORTAL;
ptlrpc_at_set_req_timeout(req);
rpc_lock = obd->u.cli.cl_setattr_lock;
} else {
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, ea2, ea2len);
ptlrpc_request_set_replen(req);
if (mod && (op_data->op_flags & MF_EPOCH_OPEN) &&
req->rq_import->imp_replayable)
{
LASSERT(*mod == NULL);
*mod = obd_mod_alloc();
if (*mod == NULL) {
DEBUG_REQ(D_ERROR, req, "Can't allocate "
"md_open_data");
} else {
req->rq_replay = 1;
req->rq_cb_data = *mod;
(*mod)->mod_open_req = req;
req->rq_commit_cb = mdc_commit_open;
/**
* Take an extra reference on \var mod, it protects \var
* mod from being freed on eviction (commit callback is
* called despite rq_replay flag).
* Will be put on mdc_done_writing().
*/
obd_mod_get(*mod);
}
}
rc = mdc_reint(req, rpc_lock, LUSTRE_IMP_FULL);
/* Save the obtained info in the original RPC for the replay case. */
if (rc == 0 && (op_data->op_flags & MF_EPOCH_OPEN)) {
struct mdt_ioepoch *epoch;
struct mdt_body *body;
epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
LASSERT(epoch != NULL);
LASSERT(body != NULL);
epoch->handle = body->handle;
epoch->ioepoch = body->ioepoch;
req->rq_replay_cb = mdc_replay_open;
/** bug 3633, open may be committed and estale answer is not error */
} else if (rc == -ESTALE && (op_data->op_flags & MF_SOM_CHANGE)) {
rc = 0;
} else if (rc == -ERESTARTSYS) {
rc = 0;
}
*request = req;
if (rc && req->rq_commit_cb) {
/* Put an extra reference on \var mod on error case. */
obd_mod_put(*mod);
req->rq_commit_cb(req);
}
RETURN(rc);
}
static int seq_client_rpc(struct lu_client_seq *seq,
struct lu_seq_range *output, __u32 opc,
const char *opcname)
{
struct obd_export *exp = seq->lcs_exp;
struct ptlrpc_request *req;
struct lu_seq_range *out, *in;
__u32 *op;
int rc;
ENTRY;
req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_SEQ_QUERY,
LUSTRE_MDS_VERSION, SEQ_QUERY);
if (req == NULL)
RETURN(-ENOMEM);
/* Init operation code */
op = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_OPC);
*op = opc;
/* Zero out input range, this is not recovery yet. */
in = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_RANGE);
range_init(in);
ptlrpc_request_set_replen(req);
if (seq->lcs_type == LUSTRE_SEQ_METADATA) {
req->rq_request_portal = SEQ_METADATA_PORTAL;
in->lsr_flags = LU_SEQ_RANGE_MDT;
} else {
LASSERTF(seq->lcs_type == LUSTRE_SEQ_DATA,
"unknown lcs_type %u\n", seq->lcs_type);
req->rq_request_portal = SEQ_DATA_PORTAL;
in->lsr_flags = LU_SEQ_RANGE_OST;
}
if (opc == SEQ_ALLOC_SUPER) {
/* Update index field of *in, it is required for
* FLD update on super sequence allocator node. */
in->lsr_index = seq->lcs_space.lsr_index;
req->rq_request_portal = SEQ_CONTROLLER_PORTAL;
} else {
LASSERTF(opc == SEQ_ALLOC_META,
"unknown opcode %u\n, opc", opc);
}
ptlrpc_at_set_req_timeout(req);
mdc_get_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
rc = ptlrpc_queue_wait(req);
mdc_put_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
if (rc)
GOTO(out_req, rc);
out = req_capsule_server_get(&req->rq_pill, &RMF_SEQ_RANGE);
*output = *out;
if (!range_is_sane(output)) {
CERROR("%s: Invalid range received from server: "
DRANGE"\n", seq->lcs_name, PRANGE(output));
GOTO(out_req, rc = -EINVAL);
}
if (range_is_exhausted(output)) {
CERROR("%s: Range received from server is exhausted: "
DRANGE"]\n", seq->lcs_name, PRANGE(output));
GOTO(out_req, rc = -EINVAL);
}
CDEBUG(D_INFO, "%s: Allocated %s-sequence "DRANGE"]\n",
seq->lcs_name, opcname, PRANGE(output));
EXIT;
out_req:
ptlrpc_req_finished(req);
return rc;
}
