* Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2012, 2014 Intel Corporation.
+ * Copyright (c) 2012, 2015, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include <lustre/lustre_idl.h>
#include <lustre_dlm.h>
#include <lustre_quota.h>
+#include <lustre_nodemap.h>
#include "ofd_internal.h"
struct lu_device *d;
struct ofd_thread_info *info = ofd_info(env);
struct lustre_mount_info *lmi;
+ struct lustre_mount_data *lmd;
int rc;
char *osdname;
RETURN(-ENODEV);
}
+ lmd = s2lsi(lmi->lmi_sb)->lsi_lmd;
+ if (lmd != NULL && lmd->lmd_flags & LMD_FLG_SKIP_LFSCK)
+ m->ofd_skip_lfsck = 1;
+
/* find bottom osd */
OBD_ALLOC(osdname, MTI_NAME_MAXLEN);
if (osdname == NULL)
strcat(flags, "A");
lustre_cfg_bufs_set_string(&bufs, 1, flags);
lcfg = lustre_cfg_new(LCFG_CLEANUP, &bufs);
- if (!lcfg) {
- CERROR("Cannot alloc lcfg!\n");
+ if (lcfg == NULL)
RETURN_EXIT;
- }
LASSERT(top);
top->ld_ops->ldo_process_config(env, top, lcfg);
return (*p)(env, cookie, LUSTRE_OST_NAME"-object@%p", o);
}
-struct lu_object_operations ofd_obj_ops = {
+static struct lu_object_operations ofd_obj_ops = {
.loo_object_init = ofd_object_init,
.loo_object_free = ofd_object_free,
.loo_object_print = ofd_object_print
obd->obd_uses_nid_stats = 1;
- entry = lprocfs_seq_register("exports", obd->obd_proc_entry, NULL,
- NULL);
+ entry = lprocfs_register("exports", obd->obd_proc_entry, NULL, NULL);
if (IS_ERR(entry)) {
rc = PTR_ERR(entry);
CERROR("%s: error %d setting up lprocfs for %s\n",
* \retval 0 if successful
* \retval negative value on error
*/
-int ofd_set_info_hdl(struct tgt_session_info *tsi)
+static int ofd_set_info_hdl(struct tgt_session_info *tsi)
{
struct ptlrpc_request *req = tgt_ses_req(tsi);
struct ost_body *body = NULL, *repbody;
void *key, *val = NULL;
int keylen, vallen, rc = 0;
bool is_grant_shrink;
- struct ofd_device *ofd = ofd_exp(tsi->tsi_exp);
ENTRY;
if (vallen > 0)
obd_export_evict_by_nid(tsi->tsi_exp->exp_obd, val);
rc = 0;
- } else if (KEY_IS(KEY_CAPA_KEY)) {
- rc = ofd_update_capa_key(ofd, val);
} else if (KEY_IS(KEY_SPTLRPC_CONF)) {
rc = tgt_adapt_sptlrpc_conf(tsi->tsi_tgt, 0);
} else {
* \retval negative value on error
*/
int ofd_fiemap_get(const struct lu_env *env, struct ofd_device *ofd,
- struct lu_fid *fid, struct ll_user_fiemap *fiemap)
+ struct lu_fid *fid, struct fiemap *fiemap)
{
struct ofd_object *fo;
int rc;
*/
static int lock_zero_regions(struct ldlm_namespace *ns,
struct ldlm_res_id *res_id,
- struct ll_user_fiemap *fiemap,
+ struct fiemap *fiemap,
struct list_head *locked)
{
__u64 begin = fiemap->fm_start;
unsigned int i;
int rc = 0;
- struct ll_fiemap_extent *fiemap_start = fiemap->fm_extents;
+ struct fiemap_extent *fiemap_start = fiemap->fm_extents;
ENTRY;
* \retval 0 if successful
* \retval negative value on error
*/
-int ofd_get_info_hdl(struct tgt_session_info *tsi)
+static int ofd_get_info_hdl(struct tgt_session_info *tsi)
{
struct obd_export *exp = tsi->tsi_exp;
struct ofd_device *ofd = ofd_exp(exp);
RCL_CLIENT);
if (KEY_IS(KEY_LAST_ID)) {
- obd_id *last_id;
+ u64 *last_id;
struct ofd_seq *oseq;
req_capsule_extend(tsi->tsi_pill, &RQF_OST_GET_INFO_LAST_ID);
last_id = req_capsule_server_get(tsi->tsi_pill, &RMF_OBD_ID);
oseq = ofd_seq_load(tsi->tsi_env, ofd,
- (obd_seq)exp->exp_filter_data.fed_group);
+ (u64)exp->exp_filter_data.fed_group);
if (IS_ERR(oseq))
rc = -EFAULT;
else
ofd_seq_put(tsi->tsi_env, oseq);
} else if (KEY_IS(KEY_FIEMAP)) {
struct ll_fiemap_info_key *fm_key;
- struct ll_user_fiemap *fiemap;
+ struct fiemap *fiemap;
struct lu_fid *fid;
req_capsule_extend(tsi->tsi_pill, &RQF_OST_GET_INFO_FIEMAP);
fm_key = req_capsule_client_get(tsi->tsi_pill, &RMF_FIEMAP_KEY);
- rc = tgt_validate_obdo(tsi, &fm_key->oa);
+ rc = tgt_validate_obdo(tsi, &fm_key->lfik_oa);
if (rc)
RETURN(err_serious(rc));
- fid = &fm_key->oa.o_oi.oi_fid;
+ fid = &fm_key->lfik_oa.o_oi.oi_fid;
CDEBUG(D_INODE, "get FIEMAP of object "DFID"\n", PFID(fid));
- replylen = fiemap_count_to_size(fm_key->fiemap.fm_extent_count);
+ replylen = fiemap_count_to_size(
+ fm_key->lfik_fiemap.fm_extent_count);
req_capsule_set_size(tsi->tsi_pill, &RMF_FIEMAP_VAL,
RCL_SERVER, replylen);
if (fiemap == NULL)
RETURN(-ENOMEM);
- *fiemap = fm_key->fiemap;
+ *fiemap = fm_key->lfik_fiemap;
rc = ofd_fiemap_get(tsi->tsi_env, ofd, fid, fiemap);
/* LU-3219: Lock the sparse areas to make sure dirty
* flushed back from client, then call fiemap again. */
- if (fm_key->oa.o_valid & OBD_MD_FLFLAGS &&
- fm_key->oa.o_flags & OBD_FL_SRVLOCK) {
+ if (fm_key->lfik_oa.o_valid & OBD_MD_FLFLAGS &&
+ fm_key->lfik_oa.o_flags & OBD_FL_SRVLOCK) {
struct list_head locked;
INIT_LIST_HEAD(&locked);
struct lustre_handle lh = { 0 };
struct ofd_object *fo;
__u64 flags = 0;
- ldlm_mode_t lock_mode = LCK_PR;
+ enum ldlm_mode lock_mode = LCK_PR;
bool srvlock;
int rc;
ENTRY;
struct lu_fid *fid = &info->fti_fid;
struct ost_id *oi = &oa->o_oi;
struct ofd_seq *oseq;
- obd_seq seq = ostid_seq(oi);
- obd_id end_id = ostid_id(oi);
- obd_id last;
- obd_id oid;
+ u64 seq = ostid_seq(oi);
+ u64 end_id = ostid_id(oi);
+ u64 last;
+ u64 oid;
int skip_orphan;
int rc = 0;
struct obdo *rep_oa;
struct obd_export *exp = tsi->tsi_exp;
struct ofd_device *ofd = ofd_exp(exp);
- obd_seq seq = ostid_seq(&oa->o_oi);
- obd_id oid = ostid_id(&oa->o_oi);
+ u64 seq = ostid_seq(&oa->o_oi);
+ u64 oid = ostid_id(&oa->o_oi);
struct ofd_seq *oseq;
int rc = 0, diff;
int sync_trans = 0;
+ long granted = 0;
ENTRY;
CDEBUG(D_HA, "ofd_last_id() = "LPU64" -> diff = %d\n",
ofd_seq_last_oid(oseq), diff);
if (-diff > OST_MAX_PRECREATE) {
- /* FIXME: should reset precreate_next_id on MDS */
+ /* Let MDS know that we are so far ahead. */
+ ostid_set_id(&rep_oa->o_oi, ofd_seq_last_oid(oseq) + 1);
rc = 0;
} else if (diff < 0) {
rc = ofd_orphans_destroy(tsi->tsi_env, exp,
ofd_name(ofd), POSTID(&oa->o_oi));
GOTO(out, rc = -EINVAL);
}
+
+ if (diff < 0) {
+ /* LU-5648 */
+ CERROR("%s: invalid precreate request for "
+ DOSTID", last_id " LPU64 ". "
+ "Likely MDS last_id corruption\n",
+ ofd_name(ofd), POSTID(&oa->o_oi),
+ ofd_seq_last_oid(oseq));
+ GOTO(out, rc = -EINVAL);
+ }
}
}
if (diff > 0) {
cfs_time_t enough_time = cfs_time_shift(DISK_TIMEOUT);
- obd_id next_id;
+ u64 next_id;
int created = 0;
int count;
if (!(oa->o_valid & OBD_MD_FLFLAGS) ||
!(oa->o_flags & OBD_FL_DELORPHAN)) {
/* don't enforce grant during orphan recovery */
- rc = ofd_grant_create(tsi->tsi_env,
- ofd_obd(ofd)->obd_self_export,
- &diff);
- if (rc) {
+ granted = ofd_grant_create(tsi->tsi_env,
+ ofd_obd(ofd)->obd_self_export,
+ &diff);
+ if (granted < 0) {
+ rc = granted;
+ granted = 0;
CDEBUG(D_HA, "%s: failed to acquire grant "
"space for precreate (%d): rc = %d\n",
ofd_name(ofd), diff, rc);
* LFSCK will eventually clean up any orphans. LU-14 */
if (diff > 5 * OST_MAX_PRECREATE) {
diff = OST_MAX_PRECREATE / 2;
- LCONSOLE_WARN("%s: precreate FID "DOSTID" is over %u "
- "larger than the LAST_ID "DOSTID", only "
- "precreating the last %u objects.\n",
- ofd_name(ofd), POSTID(&oa->o_oi),
- 5 * OST_MAX_PRECREATE,
- POSTID(&oseq->os_oi), diff);
+ LCONSOLE_WARN("%s: Too many FIDs to precreate "
+ "OST replaced or reformatted: "
+ "LFSCK will clean up",
+ ofd_name(ofd));
+
+ CDEBUG(D_HA, "%s: precreate FID "DOSTID" is over "
+ "%u larger than the LAST_ID "DOSTID", only "
+ "precreating the last %u objects.\n",
+ ofd_name(ofd), POSTID(&oa->o_oi),
+ 5 * OST_MAX_PRECREATE,
+ POSTID(&oseq->os_oi), diff);
ofd_seq_last_oid_set(oseq, ostid_id(&oa->o_oi) - diff);
}
ofd_name(ofd), rc);
if (!(oa->o_valid & OBD_MD_FLFLAGS) ||
- !(oa->o_flags & OBD_FL_DELORPHAN))
- ofd_grant_commit(tsi->tsi_env,
- ofd_obd(ofd)->obd_self_export, rc);
+ !(oa->o_flags & OBD_FL_DELORPHAN)) {
+ ofd_grant_commit(ofd_obd(ofd)->obd_self_export, granted,
+ rc);
+ granted = 0;
+ }
ostid_set_id(&rep_oa->o_oi, ofd_seq_last_oid(oseq));
}
struct ofd_device *ofd = ofd_exp(tsi->tsi_exp);
struct ofd_thread_info *fti = tsi2ofd_info(tsi);
struct lu_fid *fid = &fti->fti_fid;
- obd_id oid;
- obd_count count;
+ u64 oid;
+ u32 count;
int rc = 0;
ENTRY;
dlm = req_capsule_client_get(tsi->tsi_pill, &RMF_DLM_REQ);
if (dlm == NULL)
RETURN(-EFAULT);
- ldlm_request_cancel(tgt_ses_req(tsi), dlm, 0);
+ ldlm_request_cancel(tgt_ses_req(tsi), dlm, 0, LATF_SKIP);
}
*fid = body->oa.o_oi.oi_fid;
static int ofd_quotactl(struct tgt_session_info *tsi)
{
struct obd_quotactl *oqctl, *repoqc;
+ struct lu_nodemap *nodemap =
+ tsi->tsi_exp->exp_target_data.ted_nodemap;
+ int id;
int rc;
ENTRY;
if (repoqc == NULL)
RETURN(err_serious(-ENOMEM));
- /* report success for quota on/off for interoperability with current MDT
- * stack */
- if (oqctl->qc_cmd == Q_QUOTAON || oqctl->qc_cmd == Q_QUOTAOFF)
- RETURN(0);
-
*repoqc = *oqctl;
+
+ id = repoqc->qc_id;
+ if (oqctl->qc_type == USRQUOTA)
+ id = nodemap_map_id(nodemap, NODEMAP_UID,
+ NODEMAP_CLIENT_TO_FS,
+ repoqc->qc_id);
+ else if (oqctl->qc_type == GRPQUOTA)
+ id = nodemap_map_id(nodemap, NODEMAP_GID,
+ NODEMAP_CLIENT_TO_FS,
+ repoqc->qc_id);
+
+ if (repoqc->qc_id != id)
+ swap(repoqc->qc_id, id);
+
rc = lquotactl_slv(tsi->tsi_env, tsi->tsi_tgt->lut_bottom, repoqc);
ofd_counter_incr(tsi->tsi_exp, LPROC_OFD_STATS_QUOTACTL,
tsi->tsi_jobid, 1);
+ if (repoqc->qc_id != id)
+ swap(repoqc->qc_id, id);
+
RETURN(rc);
}
*
* \retval amount of time to extend the timeout with
*/
-static inline int prolong_timeout(struct ptlrpc_request *req)
+static inline int prolong_timeout(struct ptlrpc_request *req,
+ struct ldlm_lock *lock)
{
struct ptlrpc_service_part *svcpt = req->rq_rqbd->rqbd_svcpt;
if (AT_OFF)
return obd_timeout / 2;
- return max(at_est2timeout(at_get(&svcpt->scp_at_estimate)),
- ldlm_timeout);
+ /* We are in the middle of the process - BL AST is sent, CANCEL
+ is ahead. Take half of AT + IO process time. */
+ return at_est2timeout(at_get(&svcpt->scp_at_estimate)) +
+ (ldlm_bl_timeout(lock) >> 1);
}
/**
*/
static int ofd_prolong_one_lock(struct tgt_session_info *tsi,
struct ldlm_lock *lock,
- struct ldlm_extent *extent, int timeout)
+ struct ldlm_extent *extent)
{
+ int timeout = prolong_timeout(tgt_ses_req(tsi), lock);
if (lock->l_flags & LDLM_FL_DESTROYED) /* lock already cancelled */
return 0;
.end = end
};
struct ldlm_lock *lock;
- int timeout = prolong_timeout(tgt_ses_req(tsi));
int lock_count = 0;
ENTRY;
/* bingo */
LASSERT(lock->l_export == exp);
lock_count = ofd_prolong_one_lock(tsi, lock,
- &extent, timeout);
+ &extent);
LDLM_LOCK_PUT(lock);
RETURN(lock_count);
}
+ lock->l_last_used = cfs_time_current();
LDLM_LOCK_PUT(lock);
}
}
LASSERT(lock->l_flags & LDLM_FL_AST_SENT);
LASSERT(lock->l_resource->lr_type == LDLM_EXTENT);
+ /* ignore waiting locks, no more granted locks in the list */
+ if (lock->l_granted_mode != lock->l_req_mode)
+ break;
+
if (!ldlm_res_eq(&tsi->tsi_resid, &lock->l_resource->lr_name))
continue;
&extent))
continue;
- lock_count += ofd_prolong_one_lock(tsi, lock, &extent, timeout);
+ lock_count += ofd_prolong_one_lock(tsi, lock, &extent);
}
spin_unlock_bh(&exp->exp_bl_list_lock);
static int ofd_rw_hpreq_lock_match(struct ptlrpc_request *req,
struct ldlm_lock *lock)
{
- struct niobuf_remote *rnb;
- struct obd_ioobj *ioo;
- ldlm_mode_t mode;
- struct ldlm_extent ext;
- __u32 opc = lustre_msg_get_opc(req->rq_reqmsg);
+ struct niobuf_remote *rnb;
+ struct obd_ioobj *ioo;
+ enum ldlm_mode mode;
+ struct ldlm_extent ext;
+ __u32 opc = lustre_msg_get_opc(req->rq_reqmsg);
ENTRY;
ofd_punch_hpreq_check(req);
}
-struct ptlrpc_hpreq_ops ofd_hpreq_rw = {
+static struct ptlrpc_hpreq_ops ofd_hpreq_rw = {
.hpreq_lock_match = ofd_rw_hpreq_lock_match,
.hpreq_check = ofd_rw_hpreq_check,
.hpreq_fini = ofd_rw_hpreq_fini
};
-struct ptlrpc_hpreq_ops ofd_hpreq_punch = {
+static struct ptlrpc_hpreq_ops ofd_hpreq_punch = {
.hpreq_lock_match = ofd_punch_hpreq_lock_match,
.hpreq_check = ofd_punch_hpreq_check,
.hpreq_fini = ofd_punch_hpreq_fini
.tos_hs = tgt_lfsck_handlers
},
{
+ .tos_opc_start = SEC_FIRST_OPC,
+ .tos_opc_end = SEC_LAST_OPC,
+ .tos_hs = tgt_sec_ctx_handlers
+ },
+ {
.tos_hs = NULL
}
};
spin_lock_init(&m->ofd_batch_lock);
init_rwsem(&m->ofd_lastid_rwsem);
- obd->u.filter.fo_fl_oss_capa = 0;
- INIT_LIST_HEAD(&obd->u.filter.fo_capa_keys);
- obd->u.filter.fo_capa_hash = init_capa_hash();
- if (obd->u.filter.fo_capa_hash == NULL)
- RETURN(-ENOMEM);
-
m->ofd_dt_dev.dd_lu_dev.ld_ops = &ofd_lu_ops;
m->ofd_dt_dev.dd_lu_dev.ld_obd = obd;
/* set this lu_device to obd, because error handling need it */
info = ofd_info_init(env, NULL);
if (info == NULL)
- RETURN(-EFAULT);
+ GOTO(err_fini_proc, rc = -EFAULT);
rc = ofd_stack_init(env, m, cfg);
if (rc) {
if (rc != 0)
GOTO(err_fini_fs, rc);
+ tgt_adapt_sptlrpc_conf(&m->ofd_lut, 1);
+
RETURN(0);
err_fini_fs:
err_fini_lut:
tgt_fini(env, &m->ofd_lut);
err_free_ns:
- ldlm_namespace_free(m->ofd_namespace, 0, obd->obd_force);
+ ldlm_namespace_free(m->ofd_namespace, NULL, obd->obd_force);
obd->obd_namespace = m->ofd_namespace = NULL;
err_fini_stack:
ofd_stack_fini(env, m, &m->ofd_osd->dd_lu_dev);
stop.ls_flags = 0;
lfsck_stop(env, m->ofd_osd, &stop);
target_recovery_fini(obd);
+ if (m->ofd_namespace != NULL)
+ ldlm_namespace_free_prior(m->ofd_namespace, NULL,
+ d->ld_obd->obd_force);
+
obd_exports_barrier(obd);
obd_zombie_barrier();
lfsck_degister(env, m->ofd_osd);
ofd_fs_cleanup(env, m);
- ofd_free_capa_keys(m);
- cleanup_capa_hash(obd->u.filter.fo_capa_hash);
-
if (m->ofd_namespace != NULL) {
- ldlm_namespace_free(m->ofd_namespace, NULL,
- d->ld_obd->obd_force);
+ ldlm_namespace_free_post(m->ofd_namespace);
d->ld_obd->obd_namespace = m->ofd_namespace = NULL;
}
* \retval 0 if successful
* \retval negative value on error
*/
-int __init ofd_init(void)
+static int __init ofd_init(void)
{
int rc;
* This function is called upon OFD module unloading.
* It frees all related structures and unregisters OFD device type.
*/
-void __exit ofd_exit(void)
+static void __exit ofd_exit(void)
{
ofd_fmd_exit();
lu_kmem_fini(ofd_caches);
class_unregister_type(LUSTRE_OST_NAME);
}
-MODULE_AUTHOR("Whamcloud, Inc. <http://www.whamcloud.com/>");
+MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
MODULE_DESCRIPTION("Lustre Object Filtering Device");
+MODULE_VERSION(LUSTRE_VERSION_STRING);
MODULE_LICENSE("GPL");
module_init(ofd_init);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff