* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include <lustre_lib.h>
#include <lustre_net.h>
#include <obd_class.h>
+#include <lustre_lmv.h>
#include <lprocfs_status.h>
+#include <cl_object.h>
+#include <lclient.h>
#include <lustre_lite.h>
#include <lustre_fid.h>
#include "lmv_internal.h"
+/* This hash is only for testing purpose */
+static inline unsigned int
+lmv_hash_all_chars(unsigned int count, const char *name, int namelen)
+{
+ unsigned int c = 0;
+ const unsigned char *p = (const unsigned char *)name;
+
+ while (--namelen >= 0)
+ c += p[namelen];
+
+ c = c % count;
+
+ return c;
+}
+
+static inline unsigned int
+lmv_hash_fnv1a(unsigned int count, const char *name, int namelen)
+{
+ __u64 hash;
+
+ hash = lustre_hash_fnv_1a_64(name, namelen);
+
+ hash = hash % count;
+
+ return hash;
+}
+
+int lmv_name_to_stripe_index(enum lmv_hash_type hashtype,
+ unsigned int max_mdt_index,
+ const char *name, int namelen)
+{
+ int idx;
+
+ LASSERT(namelen > 0);
+ if (max_mdt_index <= 1)
+ return 0;
+
+ switch (hashtype) {
+ case LMV_HASH_TYPE_ALL_CHARS:
+ idx = lmv_hash_all_chars(max_mdt_index, name, namelen);
+ break;
+ case LMV_HASH_TYPE_FNV_1A_64:
+ idx = lmv_hash_fnv1a(max_mdt_index, name, namelen);
+ break;
+ /* LMV_HASH_TYPE_MIGRATION means the file is being migrated,
+ * and the file should be accessed by client, except for
+ * lookup(see lmv_intent_lookup), return -EACCES here */
+ case LMV_HASH_TYPE_MIGRATION:
+ CERROR("%.*s is being migrated: rc = %d\n", namelen,
+ name, -EACCES);
+ return -EACCES;
+ default:
+ CERROR("Unknown hash type 0x%x\n", hashtype);
+ return -EINVAL;
+ }
+
+ CDEBUG(D_INFO, "name %.*s hash_type %d idx %d\n", namelen, name,
+ hashtype, idx);
+
+ LASSERT(idx < max_mdt_index);
+ return idx;
+}
+
static void lmv_activate_target(struct lmv_obd *lmv,
struct lmv_tgt_desc *tgt,
int activate)
* -ENOTCONN: The UUID is found, but the target connection is bad (!)
* -EBADF : The UUID is found, but the OBD of the wrong type (!)
*/
-static int lmv_set_mdc_active(struct lmv_obd *lmv, struct obd_uuid *uuid,
- int activate)
-{
- struct lmv_tgt_desc *tgt;
- struct obd_device *obd;
- int i;
- int rc = 0;
- ENTRY;
+static int lmv_set_mdc_active(struct lmv_obd *lmv,
+ const struct obd_uuid *uuid,
+ int activate)
+{
+ struct lmv_tgt_desc *tgt = NULL;
+ struct obd_device *obd;
+ __u32 i;
+ int rc = 0;
+ ENTRY;
- CDEBUG(D_INFO, "Searching in lmv %p for uuid %s (activate=%d)\n",
- lmv, uuid->uuid, activate);
+ CDEBUG(D_INFO, "Searching in lmv %p for uuid %s (activate=%d)\n",
+ lmv, uuid->uuid, activate);
spin_lock(&lmv->lmv_lock);
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
struct obd_uuid *lmv_get_uuid(struct obd_export *exp)
{
- struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_tgt_desc *tgt = lmv->tgts[0];
- return obd_get_uuid(lmv->tgts[0]->ltd_exp);
+ return (tgt == NULL) ? NULL : obd_get_uuid(tgt->ltd_exp);
}
static int lmv_notify(struct obd_device *obd, struct obd_device *watched,
lmv->conn_data = *data;
#ifdef __KERNEL__
- lmv_proc_dir = lprocfs_register("target_obds", obd->obd_proc_entry,
- NULL, NULL);
- if (IS_ERR(lmv_proc_dir)) {
- CERROR("could not register /proc/fs/lustre/%s/%s/target_obds.",
- obd->obd_type->typ_name, obd->obd_name);
- lmv_proc_dir = NULL;
- }
+ if (obd->obd_proc_private != NULL) {
+ lmv_proc_dir = obd->obd_proc_private;
+ } else {
+ lmv_proc_dir = lprocfs_seq_register("target_obds",
+ obd->obd_proc_entry,
+ NULL, NULL);
+ if (IS_ERR(lmv_proc_dir)) {
+ CERROR("could not register /proc/fs/lustre/%s/%s/target_obds.",
+ obd->obd_type->typ_name, obd->obd_name);
+ lmv_proc_dir = NULL;
+ }
+ obd->obd_proc_private = lmv_proc_dir;
+ }
#endif
/*
* and MDC stuff will be called directly, for instance while reading
* ../mdc/../kbytesfree procfs file, etc.
*/
- if (data->ocd_connect_flags & OBD_CONNECT_REAL)
+ if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_REAL))
rc = lmv_check_connect(obd);
#ifdef __KERNEL__
- if (rc) {
- if (lmv_proc_dir)
- lprocfs_remove(&lmv_proc_dir);
- }
+ if (rc && lmv_proc_dir) {
+ lprocfs_remove(&lmv_proc_dir);
+ obd->obd_proc_private = NULL;
+ }
#endif
-
RETURN(rc);
}
static void lmv_set_timeouts(struct obd_device *obd)
{
- struct lmv_tgt_desc *tgt;
- struct lmv_obd *lmv;
- int i;
+ struct lmv_obd *lmv;
+ __u32 i;
lmv = &obd->u.lmv;
if (lmv->server_timeout == 0)
return;
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- tgt = lmv->tgts[i];
- if (tgt == NULL || tgt->ltd_exp == NULL || tgt->ltd_active == 0)
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
continue;
obd_set_info_async(NULL, tgt->ltd_exp, sizeof(KEY_INTERMDS),
static int lmv_init_ea_size(struct obd_export *exp, int easize,
int def_easize, int cookiesize)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- int i;
- int rc = 0;
- int change = 0;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ __u32 i;
+ int rc = 0;
+ int change = 0;
+ ENTRY;
if (lmv->max_easize < easize) {
lmv->max_easize = easize;
RETURN(0);
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- if (lmv->tgts[i] == NULL ||
- lmv->tgts[i]->ltd_exp == NULL ||
- lmv->tgts[i]->ltd_active == 0) {
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active) {
CWARN("%s: NULL export for %d\n", obd->obd_name, i);
continue;
}
- rc = md_init_ea_size(lmv->tgts[i]->ltd_exp, easize, def_easize,
+ rc = md_init_ea_size(tgt->ltd_exp, easize, def_easize,
cookiesize);
if (rc) {
CERROR("%s: obd_init_ea_size() failed on MDT target %d:"
#endif
struct lmv_obd *lmv = &obd->u.lmv;
struct obd_uuid *cluuid = &lmv->cluuid;
- struct obd_connect_data *mdc_data = NULL;
struct obd_uuid lmv_mdc_uuid = { "LMV_MDC_UUID" };
struct obd_device *mdc_obd;
struct obd_export *mdc_exp;
fld_client_add_target(&lmv->lmv_fld, &target);
- mdc_data = &class_exp2cliimp(mdc_exp)->imp_connect_data;
-
rc = obd_register_observer(mdc_obd, obd);
if (rc) {
obd_disconnect(mdc_exp);
cfs_atomic_read(&obd->obd_refcount));
#ifdef __KERNEL__
- lmv_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
- if (lmv_proc_dir) {
- struct proc_dir_entry *mdc_symlink;
-
- LASSERT(mdc_obd->obd_type != NULL);
- LASSERT(mdc_obd->obd_type->typ_name != NULL);
- mdc_symlink = lprocfs_add_symlink(mdc_obd->obd_name,
- lmv_proc_dir,
- "../../../%s/%s",
- mdc_obd->obd_type->typ_name,
- mdc_obd->obd_name);
- if (mdc_symlink == NULL) {
- CERROR("Could not register LMV target "
- "/proc/fs/lustre/%s/%s/target_obds/%s.",
- obd->obd_type->typ_name, obd->obd_name,
- mdc_obd->obd_name);
- lprocfs_remove(&lmv_proc_dir);
- lmv_proc_dir = NULL;
- }
- }
+ lmv_proc_dir = obd->obd_proc_private;
+ if (lmv_proc_dir) {
+ struct proc_dir_entry *mdc_symlink;
+
+ LASSERT(mdc_obd->obd_type != NULL);
+ LASSERT(mdc_obd->obd_type->typ_name != NULL);
+ mdc_symlink = lprocfs_add_symlink(mdc_obd->obd_name,
+ lmv_proc_dir,
+ "../../../%s/%s",
+ mdc_obd->obd_type->typ_name,
+ mdc_obd->obd_name);
+ if (mdc_symlink == NULL) {
+ CERROR("Could not register LMV target "
+ "/proc/fs/lustre/%s/%s/target_obds/%s.",
+ obd->obd_type->typ_name, obd->obd_name,
+ mdc_obd->obd_name);
+ lprocfs_remove(&lmv_proc_dir);
+ obd->obd_proc_private = NULL;
+ }
+ }
#endif
- RETURN(0);
+ RETURN(0);
}
static void lmv_del_target(struct lmv_obd *lmv, int index)
int lmv_check_connect(struct obd_device *obd)
{
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt;
- int i;
- int rc;
- int easize;
- ENTRY;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt;
+ __u32 i;
+ int rc;
+ int easize;
+ ENTRY;
if (lmv->connected)
RETURN(0);
RETURN(-EINVAL);
}
- CDEBUG(D_CONFIG, "Time to connect %s to %s\n",
- lmv->cluuid.uuid, obd->obd_name);
-
LASSERT(lmv->tgts != NULL);
+ if (lmv->tgts[0] == NULL) {
+ lmv_init_unlock(lmv);
+ CERROR("%s: no target configured for index 0.\n",
+ obd->obd_name);
+ RETURN(-EINVAL);
+ }
+
+ CDEBUG(D_CONFIG, "Time to connect %s to %s\n",
+ lmv->cluuid.uuid, obd->obd_name);
+
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
tgt = lmv->tgts[i];
if (tgt == NULL)
GOTO(out_disc, rc);
}
- lmv_set_timeouts(obd);
- class_export_put(lmv->exp);
- lmv->connected = 1;
- easize = lmv_get_easize(lmv);
- lmv_init_ea_size(obd->obd_self_export, easize, 0, 0);
- lmv_init_unlock(lmv);
- RETURN(0);
+ lmv_set_timeouts(obd);
+ class_export_put(lmv->exp);
+ lmv->connected = 1;
+ easize = lmv_mds_md_size(lmv->desc.ld_tgt_count, LMV_MAGIC);
+ lmv_init_ea_size(obd->obd_self_export, easize, 0, 0);
+ lmv_init_unlock(lmv);
+ RETURN(0);
out_disc:
while (i-- > 0) {
}
#ifdef __KERNEL__
- lmv_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
- if (lmv_proc_dir) {
- struct proc_dir_entry *mdc_symlink;
-
- mdc_symlink = lprocfs_srch(lmv_proc_dir, mdc_obd->obd_name);
- if (mdc_symlink) {
- lprocfs_remove(&mdc_symlink);
- } else {
- CERROR("/proc/fs/lustre/%s/%s/target_obds/%s missing\n",
- obd->obd_type->typ_name, obd->obd_name,
- mdc_obd->obd_name);
- }
- }
+ lmv_proc_dir = obd->obd_proc_private;
+ if (lmv_proc_dir)
+ lprocfs_remove_proc_entry(mdc_obd->obd_name, lmv_proc_dir);
#endif
rc = obd_fid_fini(tgt->ltd_exp->exp_obd);
if (rc)
static int lmv_disconnect(struct obd_export *exp)
{
- struct obd_device *obd = class_exp2obd(exp);
-#ifdef __KERNEL__
- struct proc_dir_entry *lmv_proc_dir;
-#endif
- struct lmv_obd *lmv = &obd->u.lmv;
- int rc;
- int i;
- ENTRY;
+ struct obd_device *obd = class_exp2obd(exp);
+ struct lmv_obd *lmv = &obd->u.lmv;
+ int rc;
+ __u32 i;
+ ENTRY;
if (!lmv->tgts)
goto out_local;
}
#ifdef __KERNEL__
- lmv_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
- if (lmv_proc_dir) {
- lprocfs_remove(&lmv_proc_dir);
- } else {
- CERROR("/proc/fs/lustre/%s/%s/target_obds missing\n",
- obd->obd_type->typ_name, obd->obd_name);
- }
+ if (obd->obd_proc_private)
+ lprocfs_remove((struct proc_dir_entry **)&obd->obd_proc_private);
+ else
+ CERROR("/proc/fs/lustre/%s/%s/target_obds missing\n",
+ obd->obd_type->typ_name, obd->obd_name);
#endif
out_local:
RETURN(rc);
}
+static int lmv_fid2path(struct obd_export *exp, int len, void *karg, void *uarg)
+{
+ struct obd_device *obddev = class_exp2obd(exp);
+ struct lmv_obd *lmv = &obddev->u.lmv;
+ struct getinfo_fid2path *gf;
+ struct lmv_tgt_desc *tgt;
+ struct getinfo_fid2path *remote_gf = NULL;
+ int remote_gf_size = 0;
+ int rc;
+
+ gf = (struct getinfo_fid2path *)karg;
+ tgt = lmv_find_target(lmv, &gf->gf_fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+repeat_fid2path:
+ rc = obd_iocontrol(OBD_IOC_FID2PATH, tgt->ltd_exp, len, gf, uarg);
+ if (rc != 0 && rc != -EREMOTE)
+ GOTO(out_fid2path, rc);
+
+ /* If remote_gf != NULL, it means just building the
+ * path on the remote MDT, copy this path segement to gf */
+ if (remote_gf != NULL) {
+ struct getinfo_fid2path *ori_gf;
+ char *ptr;
+
+ ori_gf = (struct getinfo_fid2path *)karg;
+ if (strlen(ori_gf->gf_path) +
+ strlen(gf->gf_path) > ori_gf->gf_pathlen)
+ GOTO(out_fid2path, rc = -EOVERFLOW);
+
+ ptr = ori_gf->gf_path;
+
+ memmove(ptr + strlen(gf->gf_path) + 1, ptr,
+ strlen(ori_gf->gf_path));
+
+ strncpy(ptr, gf->gf_path, strlen(gf->gf_path));
+ ptr += strlen(gf->gf_path);
+ *ptr = '/';
+ }
+
+ CDEBUG(D_INFO, "%s: get path %s "DFID" rec: "LPU64" ln: %u\n",
+ tgt->ltd_exp->exp_obd->obd_name,
+ gf->gf_path, PFID(&gf->gf_fid), gf->gf_recno,
+ gf->gf_linkno);
+
+ if (rc == 0)
+ GOTO(out_fid2path, rc);
+
+ /* sigh, has to go to another MDT to do path building further */
+ if (remote_gf == NULL) {
+ remote_gf_size = sizeof(*remote_gf) + PATH_MAX;
+ OBD_ALLOC(remote_gf, remote_gf_size);
+ if (remote_gf == NULL)
+ GOTO(out_fid2path, rc = -ENOMEM);
+ remote_gf->gf_pathlen = PATH_MAX;
+ }
+
+ if (!fid_is_sane(&gf->gf_fid)) {
+ CERROR("%s: invalid FID "DFID": rc = %d\n",
+ tgt->ltd_exp->exp_obd->obd_name,
+ PFID(&gf->gf_fid), -EINVAL);
+ GOTO(out_fid2path, rc = -EINVAL);
+ }
+
+ tgt = lmv_find_target(lmv, &gf->gf_fid);
+ if (IS_ERR(tgt))
+ GOTO(out_fid2path, rc = -EINVAL);
+
+ remote_gf->gf_fid = gf->gf_fid;
+ remote_gf->gf_recno = -1;
+ remote_gf->gf_linkno = -1;
+ memset(remote_gf->gf_path, 0, remote_gf->gf_pathlen);
+ gf = remote_gf;
+ goto repeat_fid2path;
+
+out_fid2path:
+ if (remote_gf != NULL)
+ OBD_FREE(remote_gf, remote_gf_size);
+ RETURN(rc);
+}
+
+static int lmv_hsm_req_count(struct lmv_obd *lmv,
+ const struct hsm_user_request *hur,
+ const struct lmv_tgt_desc *tgt_mds)
+{
+ __u32 i;
+ int nr = 0;
+ struct lmv_tgt_desc *curr_tgt;
+
+ /* count how many requests must be sent to the given target */
+ for (i = 0; i < hur->hur_request.hr_itemcount; i++) {
+ curr_tgt = lmv_find_target(lmv, &hur->hur_user_item[i].hui_fid);
+ if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid))
+ nr++;
+ }
+ return nr;
+}
+
+static void lmv_hsm_req_build(struct lmv_obd *lmv,
+ struct hsm_user_request *hur_in,
+ const struct lmv_tgt_desc *tgt_mds,
+ struct hsm_user_request *hur_out)
+{
+ __u32 i, nr_out;
+ struct lmv_tgt_desc *curr_tgt;
+
+ /* build the hsm_user_request for the given target */
+ hur_out->hur_request = hur_in->hur_request;
+ nr_out = 0;
+ for (i = 0; i < hur_in->hur_request.hr_itemcount; i++) {
+ curr_tgt = lmv_find_target(lmv,
+ &hur_in->hur_user_item[i].hui_fid);
+ if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid)) {
+ hur_out->hur_user_item[nr_out] =
+ hur_in->hur_user_item[i];
+ nr_out++;
+ }
+ }
+ hur_out->hur_request.hr_itemcount = nr_out;
+ memcpy(hur_data(hur_out), hur_data(hur_in),
+ hur_in->hur_request.hr_data_len);
+}
+
+static int lmv_hsm_ct_unregister(struct lmv_obd *lmv, unsigned int cmd, int len,
+ struct lustre_kernelcomm *lk, void *uarg)
+{
+ __u32 i;
+ int rc;
+ struct kkuc_ct_data *kcd = NULL;
+ ENTRY;
+
+ /* unregister request (call from llapi_hsm_copytool_fini) */
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ continue;
+ /* best effort: try to clean as much as possible
+ * (continue on error) */
+ obd_iocontrol(cmd, tgt->ltd_exp, len, lk, uarg);
+ }
+
+ /* Whatever the result, remove copytool from kuc groups.
+ * Unreached coordinators will get EPIPE on next requests
+ * and will unregister automatically.
+ */
+ rc = libcfs_kkuc_group_rem(lk->lk_uid, lk->lk_group, (void **)&kcd);
+ if (kcd != NULL)
+ OBD_FREE_PTR(kcd);
+
+ RETURN(rc);
+}
+
+static int lmv_hsm_ct_register(struct lmv_obd *lmv, unsigned int cmd, int len,
+ struct lustre_kernelcomm *lk, void *uarg)
+{
+ struct file *filp;
+ __u32 i, j;
+ int err, rc;
+ bool any_set = false;
+ struct kkuc_ct_data *kcd;
+ ENTRY;
+
+ /* All or nothing: try to register to all MDS.
+ * In case of failure, unregister from previous MDS,
+ * except if it because of inactive target. */
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ continue;
+ err = obd_iocontrol(cmd, tgt->ltd_exp, len, lk, uarg);
+ if (err) {
+ if (tgt->ltd_active) {
+ /* permanent error */
+ CERROR("%s: iocontrol MDC %s on MDT"
+ " idx %d cmd %x: err = %d\n",
+ class_exp2obd(lmv->exp)->obd_name,
+ tgt->ltd_uuid.uuid, i, cmd, err);
+ rc = err;
+ lk->lk_flags |= LK_FLG_STOP;
+ /* unregister from previous MDS */
+ for (j = 0; j < i; j++) {
+ tgt = lmv->tgts[j];
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ continue;
+ obd_iocontrol(cmd, tgt->ltd_exp, len,
+ lk, uarg);
+ }
+ RETURN(rc);
+ }
+ /* else: transient error.
+ * kuc will register to the missing MDT
+ * when it is back */
+ } else {
+ any_set = true;
+ }
+ }
+
+ if (!any_set)
+ /* no registration done: return error */
+ RETURN(-ENOTCONN);
+
+ /* at least one registration done, with no failure */
+ filp = fget(lk->lk_wfd);
+ if (filp == NULL)
+ RETURN(-EBADF);
+
+ OBD_ALLOC_PTR(kcd);
+ if (kcd == NULL) {
+ fput(filp);
+ RETURN(-ENOMEM);
+ }
+ kcd->kcd_magic = KKUC_CT_DATA_MAGIC;
+ kcd->kcd_uuid = lmv->cluuid;
+ kcd->kcd_archive = lk->lk_data;
+
+ rc = libcfs_kkuc_group_add(filp, lk->lk_uid, lk->lk_group, kcd);
+ if (rc != 0) {
+ if (filp != NULL)
+ fput(filp);
+ OBD_FREE_PTR(kcd);
+ }
+
+ RETURN(rc);
+}
+
+
+
+
static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
int len, void *karg, void *uarg)
{
- struct obd_device *obddev = class_exp2obd(exp);
- struct lmv_obd *lmv = &obddev->u.lmv;
- int i = 0;
- int rc = 0;
- int set = 0;
- int count = lmv->desc.ld_tgt_count;
- ENTRY;
+ struct obd_device *obddev = class_exp2obd(exp);
+ struct lmv_obd *lmv = &obddev->u.lmv;
+ struct lmv_tgt_desc *tgt = NULL;
+ __u32 i = 0;
+ int rc = 0;
+ int set = 0;
+ __u32 count = lmv->desc.ld_tgt_count;
+ ENTRY;
if (count == 0)
RETURN(-ENOTTY);
if ((index >= count))
RETURN(-ENODEV);
- if (lmv->tgts[index] == NULL ||
- lmv->tgts[index]->ltd_active == 0)
+ tgt = lmv->tgts[index];
+ if (tgt == NULL || !tgt->ltd_active)
RETURN(-ENODATA);
- mdc_obd = class_exp2obd(lmv->tgts[index]->ltd_exp);
+ mdc_obd = class_exp2obd(tgt->ltd_exp);
if (!mdc_obd)
RETURN(-EINVAL);
/* copy UUID */
- if (cfs_copy_to_user(data->ioc_pbuf2, obd2cli_tgt(mdc_obd),
- min((int) data->ioc_plen2,
- (int) sizeof(struct obd_uuid))))
+ if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(mdc_obd),
+ min((int) data->ioc_plen2,
+ (int) sizeof(struct obd_uuid))))
RETURN(-EFAULT);
- rc = obd_statfs(NULL, lmv->tgts[index]->ltd_exp, &stat_buf,
+ rc = obd_statfs(NULL, tgt->ltd_exp, &stat_buf,
cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
0);
if (rc)
RETURN(rc);
- if (cfs_copy_to_user(data->ioc_pbuf1, &stat_buf,
- min((int) data->ioc_plen1,
- (int) sizeof(stat_buf))))
+ if (copy_to_user(data->ioc_pbuf1, &stat_buf,
+ min((int) data->ioc_plen1,
+ (int) sizeof(stat_buf))))
RETURN(-EFAULT);
break;
}
case OBD_IOC_QUOTACTL: {
struct if_quotactl *qctl = karg;
- struct lmv_tgt_desc *tgt = NULL;
struct obd_quotactl *oqctl;
- if (qctl->qc_valid == QC_MDTIDX) {
- if (qctl->qc_idx < 0 || count <= qctl->qc_idx)
- RETURN(-EINVAL);
+ if (qctl->qc_valid == QC_MDTIDX) {
+ if (count <= qctl->qc_idx)
+ RETURN(-EINVAL);
tgt = lmv->tgts[qctl->qc_idx];
if (tgt == NULL || tgt->ltd_exp == NULL)
if (i >= count)
RETURN(-EAGAIN);
- LASSERT(tgt && tgt->ltd_exp);
+ LASSERT(tgt != NULL && tgt->ltd_exp != NULL);
OBD_ALLOC_PTR(oqctl);
if (!oqctl)
RETURN(-ENOMEM);
if (icc->icc_mdtindex >= count)
RETURN(-ENODEV);
- if (lmv->tgts[icc->icc_mdtindex] == NULL ||
- lmv->tgts[icc->icc_mdtindex]->ltd_exp == NULL ||
- lmv->tgts[icc->icc_mdtindex]->ltd_active == 0)
+ tgt = lmv->tgts[icc->icc_mdtindex];
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
RETURN(-ENODEV);
- rc = obd_iocontrol(cmd, lmv->tgts[icc->icc_mdtindex]->ltd_exp,
- sizeof(*icc), icc, NULL);
+ rc = obd_iocontrol(cmd, tgt->ltd_exp, sizeof(*icc), icc, NULL);
break;
}
case LL_IOC_GET_CONNECT_FLAGS: {
- if (lmv->tgts[0] == NULL)
+ tgt = lmv->tgts[0];
+ if (tgt == NULL || tgt->ltd_exp == NULL)
RETURN(-ENODATA);
- rc = obd_iocontrol(cmd, lmv->tgts[0]->ltd_exp, len, karg, uarg);
+ rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
break;
}
case OBD_IOC_FID2PATH: {
- struct getinfo_fid2path *gf;
- struct lmv_tgt_desc *tgt;
+ rc = lmv_fid2path(exp, len, karg, uarg);
+ break;
+ }
+ case LL_IOC_HSM_STATE_GET:
+ case LL_IOC_HSM_STATE_SET:
+ case LL_IOC_HSM_ACTION: {
+ struct md_op_data *op_data = karg;
+
+ tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ if (tgt->ltd_exp == NULL)
+ RETURN(-EINVAL);
+
+ rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
+ break;
+ }
+ case LL_IOC_HSM_PROGRESS: {
+ const struct hsm_progress_kernel *hpk = karg;
- gf = (struct getinfo_fid2path *)karg;
- tgt = lmv_find_target(lmv, &gf->gf_fid);
+ tgt = lmv_find_target(lmv, &hpk->hpk_fid);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
break;
}
- case LL_IOC_HSM_STATE_GET:
- case LL_IOC_HSM_STATE_SET:
- case LL_IOC_HSM_ACTION:
+ case LL_IOC_HSM_REQUEST: {
+ struct hsm_user_request *hur = karg;
+ unsigned int reqcount = hur->hur_request.hr_itemcount;
+
+ if (reqcount == 0)
+ RETURN(0);
+
+ /* if the request is about a single fid
+ * or if there is a single MDS, no need to split
+ * the request. */
+ if (reqcount == 1 || count == 1) {
+ tgt = lmv_find_target(lmv,
+ &hur->hur_user_item[0].hui_fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+ rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
+ } else {
+ /* split fid list to their respective MDS */
+ for (i = 0; i < count; i++) {
+ unsigned int nr, reqlen;
+ int rc1;
+ struct hsm_user_request *req;
+
+ tgt = lmv->tgts[i];
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ continue;
+
+ nr = lmv_hsm_req_count(lmv, hur, tgt);
+ if (nr == 0) /* nothing for this MDS */
+ continue;
+
+ /* build a request with fids for this MDS */
+ reqlen = offsetof(typeof(*hur),
+ hur_user_item[nr])
+ + hur->hur_request.hr_data_len;
+ OBD_ALLOC_LARGE(req, reqlen);
+ if (req == NULL)
+ RETURN(-ENOMEM);
+
+ lmv_hsm_req_build(lmv, hur, tgt, req);
+
+ rc1 = obd_iocontrol(cmd, tgt->ltd_exp, reqlen,
+ req, uarg);
+ if (rc1 != 0 && rc == 0)
+ rc = rc1;
+ OBD_FREE_LARGE(req, reqlen);
+ }
+ }
+ break;
+ }
case LL_IOC_LOV_SWAP_LAYOUTS: {
struct md_op_data *op_data = karg;
- struct lmv_tgt_desc *tgt;
+ struct lmv_tgt_desc *tgt1, *tgt2;
- tgt = lmv_find_target(lmv, &op_data->op_fid1);
- if (!tgt->ltd_exp)
+ tgt1 = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(tgt1))
+ RETURN(PTR_ERR(tgt1));
+
+ tgt2 = lmv_find_target(lmv, &op_data->op_fid2);
+ if (IS_ERR(tgt2))
+ RETURN(PTR_ERR(tgt2));
+
+ if ((tgt1->ltd_exp == NULL) || (tgt2->ltd_exp == NULL))
RETURN(-EINVAL);
- rc = obd_iocontrol(cmd, lmv->tgts[0]->ltd_exp, len, karg, uarg);
+ /* only files on same MDT can have their layouts swapped */
+ if (tgt1->ltd_idx != tgt2->ltd_idx)
+ RETURN(-EPERM);
+
+ rc = obd_iocontrol(cmd, tgt1->ltd_exp, len, karg, uarg);
+ break;
+ }
+ case LL_IOC_HSM_CT_START: {
+ struct lustre_kernelcomm *lk = karg;
+ if (lk->lk_flags & LK_FLG_STOP)
+ rc = lmv_hsm_ct_unregister(lmv, cmd, len, lk, uarg);
+ else
+ rc = lmv_hsm_ct_register(lmv, cmd, len, lk, uarg);
break;
}
default:
struct obd_device *mdc_obd;
int err;
- if (lmv->tgts[i] == NULL ||
- lmv->tgts[i]->ltd_exp == NULL)
+ tgt = lmv->tgts[i];
+ if (tgt == NULL || tgt->ltd_exp == NULL)
continue;
/* ll_umount_begin() sets force flag but for lmv, not
* mdc. Let's pass it through */
- mdc_obd = class_exp2obd(lmv->tgts[i]->ltd_exp);
+ mdc_obd = class_exp2obd(tgt->ltd_exp);
mdc_obd->obd_force = obddev->obd_force;
- err = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len,
- karg, uarg);
+ err = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) {
RETURN(err);
} else if (err) {
- if (lmv->tgts[i]->ltd_active) {
+ if (tgt->ltd_active) {
CERROR("error: iocontrol MDC %s on MDT"
- "idx %d cmd %x: err = %d\n",
- lmv->tgts[i]->ltd_uuid.uuid,
- i, cmd, err);
+ " idx %d cmd %x: err = %d\n",
+ tgt->ltd_uuid.uuid, i, cmd, err);
if (!rc)
rc = err;
}
* If stripe_offset is provided during setdirstripe
* (setdirstripe -i xx), xx MDS will be choosen.
*/
- if (op_data->op_cli_flags & CLI_SET_MEA) {
+ if (op_data->op_cli_flags & CLI_SET_MEA && op_data->op_data != NULL) {
struct lmv_user_md *lum;
- lum = (struct lmv_user_md *)op_data->op_data;
- if (lum->lum_type == LMV_STRIPE_TYPE &&
- lum->lum_stripe_offset != -1) {
- if (lum->lum_stripe_offset >= lmv->desc.ld_tgt_count) {
- CERROR("%s: Stripe_offset %d > MDT count %d:"
- " rc = %d\n", obd->obd_name,
- lum->lum_stripe_offset,
- lmv->desc.ld_tgt_count, -ERANGE);
- RETURN(-ERANGE);
- }
+ lum = op_data->op_data;
+
+ if (lum->lum_stripe_offset != (__u32)-1) {
*mds = lum->lum_stripe_offset;
- RETURN(0);
+ } else {
+ /* -1 means default, which will be in the same MDT with
+ * the stripe */
+ *mds = op_data->op_mds;
+ lum->lum_stripe_offset = op_data->op_mds;
}
+ } else {
+ /* Allocate new fid on target according to operation type and
+ * parent home mds. */
+ *mds = op_data->op_mds;
}
- /* Allocate new fid on target according to operation type and parent
- * home mds. */
- *mds = op_data->op_mds;
RETURN(0);
}
int __lmv_fid_alloc(struct lmv_obd *lmv, struct lu_fid *fid,
- mdsno_t mds)
+ mdsno_t mds)
{
- struct lmv_tgt_desc *tgt;
- int rc;
- ENTRY;
+ struct lmv_tgt_desc *tgt;
+ int rc;
+ ENTRY;
- tgt = lmv_get_target(lmv, mds);
+ tgt = lmv_get_target(lmv, mds);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
- /*
- * New seq alloc and FLD setup should be atomic. Otherwise we may find
- * on server that seq in new allocated fid is not yet known.
- */
+ /*
+ * New seq alloc and FLD setup should be atomic. Otherwise we may find
+ * on server that seq in new allocated fid is not yet known.
+ */
mutex_lock(&tgt->ltd_fid_mutex);
- if (tgt == NULL || tgt->ltd_active == 0 || tgt->ltd_exp == NULL)
+ if (tgt->ltd_active == 0 || tgt->ltd_exp == NULL)
GOTO(out, rc = -ENODEV);
/*
static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
{
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lprocfs_static_vars lvars;
- struct lmv_desc *desc;
- int rc;
- ENTRY;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_desc *desc;
+ int rc;
+ ENTRY;
if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
CERROR("LMV setup requires a descriptor\n");
spin_lock_init(&lmv->lmv_lock);
mutex_init(&lmv->init_mutex);
- lprocfs_lmv_init_vars(&lvars);
-
- lprocfs_obd_setup(obd, lvars.obd_vars);
#ifdef LPROCFS
- {
- rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd",
- 0444, &lmv_proc_target_fops, obd);
- if (rc)
- CWARN("%s: error adding LMV target_obd file: rc = %d\n",
- obd->obd_name, rc);
- }
+ obd->obd_vars = lprocfs_lmv_obd_vars;
+ lprocfs_seq_obd_setup(obd);
+ lprocfs_alloc_md_stats(obd, 0);
+ rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd",
+ 0444, &lmv_proc_target_fops, obd);
+ if (rc)
+ CWARN("%s: error adding LMV target_obd file: rc = %d\n",
+ obd->obd_name, rc);
#endif
rc = fld_client_init(&lmv->lmv_fld, obd->obd_name,
LUSTRE_CLI_FLD_HASH_DHT);
obd_str2uuid(&obd_uuid, lustre_cfg_buf(lcfg, 1));
- if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1)
+ if (sscanf(lustre_cfg_buf(lcfg, 2), "%u", &index) != 1)
GOTO(out, rc = -EINVAL);
if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1)
GOTO(out, rc = -EINVAL);
static int lmv_statfs(const struct lu_env *env, struct obd_export *exp,
struct obd_statfs *osfs, __u64 max_age, __u32 flags)
{
- struct obd_device *obd = class_exp2obd(exp);
- struct lmv_obd *lmv = &obd->u.lmv;
- struct obd_statfs *temp;
- int rc = 0;
- int i;
- ENTRY;
+ struct obd_device *obd = class_exp2obd(exp);
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct obd_statfs *temp;
+ int rc = 0;
+ __u32 i;
+ ENTRY;
rc = lmv_check_connect(obd);
if (rc)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
- int i;
+ __u32 i;
int rc;
ENTRY;
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
- int i;
+ __u32 i;
int rc;
ENTRY;
RETURN(rc);
}
+/**
+ * Choosing the MDT by name or FID in @op_data.
+ * For non-striped directory, it will locate MDT by fid.
+ * For striped-directory, it will locate MDT by name. And also
+ * it will reset op_fid1 with the FID of the choosen stripe.
+ **/
+struct lmv_tgt_desc *
+lmv_locate_target_for_name(struct lmv_obd *lmv, struct lmv_stripe_md *lsm,
+ const char *name, int namelen, struct lu_fid *fid,
+ mdsno_t *mds)
+{
+ struct lmv_tgt_desc *tgt;
+ const struct lmv_oinfo *oinfo;
+
+ oinfo = lsm_name_to_stripe_info(lsm, name, namelen);
+ if (IS_ERR(oinfo))
+ RETURN((void *)oinfo);
+ *fid = oinfo->lmo_fid;
+ *mds = oinfo->lmo_mds;
+ tgt = lmv_get_target(lmv, *mds);
+
+ CDEBUG(D_INFO, "locate on mds %u "DFID"\n", *mds, PFID(fid));
+ return tgt;
+}
+
struct lmv_tgt_desc
*lmv_locate_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
struct lu_fid *fid)
{
- struct lmv_tgt_desc *tgt;
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
+ struct lmv_tgt_desc *tgt;
- tgt = lmv_find_target(lmv, fid);
- op_data->op_mds = tgt->ltd_idx;
+ if (lsm == NULL || lsm->lsm_md_stripe_count <= 1 ||
+ op_data->op_namelen == 0 ||
+ lsm->lsm_md_magic == LMV_MAGIC_MIGRATE) {
+ tgt = lmv_find_target(lmv, fid);
+ if (IS_ERR(tgt))
+ return tgt;
- return tgt;
+ op_data->op_mds = tgt->ltd_idx;
+ return tgt;
+ }
+
+ return lmv_locate_target_for_name(lmv, lsm, op_data->op_name,
+ op_data->op_namelen, fid,
+ &op_data->op_mds);
}
int lmv_create(struct obd_export *exp, struct md_op_data *op_data,
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
+ CDEBUG(D_INODE, "CREATE name '%.*s' on "DFID" -> mds #%x\n",
+ op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
+ op_data->op_mds);
+
rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
if (rc)
RETURN(rc);
- CDEBUG(D_INODE, "CREATE '%*s' on "DFID" -> mds #%x\n",
- op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
- op_data->op_mds);
+ /* Send the create request to the MDT where the object
+ * will be located */
+ tgt = lmv_find_target(lmv, &op_data->op_fid2);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ op_data->op_mds = tgt->ltd_idx;
+
+ CDEBUG(D_INODE, "CREATE obj "DFID" -> mds #%x\n",
+ PFID(&op_data->op_fid2), op_data->op_mds);
op_data->op_flags |= MF_MDC_CANCEL_FID1;
rc = md_create(tgt->ltd_exp, op_data, data, datalen, mode, uid, gid,
cap_effective, rdev, request);
-
if (rc == 0) {
if (*request == NULL)
RETURN(rc);
static int
lmv_enqueue_remote(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
- struct lookup_intent *it, struct md_op_data *op_data,
- struct lustre_handle *lockh, void *lmm, int lmmsize,
- int extra_lock_flags)
+ struct lookup_intent *it, struct md_op_data *op_data,
+ struct lustre_handle *lockh, void *lmm, int lmmsize,
+ __u64 extra_lock_flags)
{
struct ptlrpc_request *req = it->d.lustre.it_data;
struct obd_device *obd = exp->exp_obd;
it->d.lustre.it_data = NULL;
fid1 = body->fid1;
- it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
ptlrpc_req_finished(req);
tgt = lmv_find_target(lmv, &fid1);
fl == MF_MDC_CANCEL_FID4 ? &op_data->op_fid4 : \
NULL)
-static int lmv_early_cancel(struct obd_export *exp, struct md_op_data *op_data,
- int op_tgt, ldlm_mode_t mode, int bits, int flag)
+static int lmv_early_cancel(struct obd_export *exp, struct lmv_tgt_desc *tgt,
+ struct md_op_data *op_data,
+ int op_tgt, ldlm_mode_t mode, int bits, int flag)
{
- struct lu_fid *fid = md_op_data_fid(op_data, flag);
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt;
- ldlm_policy_data_t policy = {{0}};
- int rc = 0;
- ENTRY;
+ struct lu_fid *fid = md_op_data_fid(op_data, flag);
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ ldlm_policy_data_t policy = {{ 0 }};
+ int rc = 0;
+ ENTRY;
- if (!fid_is_sane(fid))
- RETURN(0);
+ if (!fid_is_sane(fid))
+ RETURN(0);
- tgt = lmv_find_target(lmv, fid);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ if (tgt == NULL) {
+ tgt = lmv_find_target(lmv, fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+ }
if (tgt->ltd_idx != op_tgt) {
CDEBUG(D_INODE, "EARLY_CANCEL on "DFID"\n", PFID(fid));
PFID(&op_data->op_fid2), op_data->op_namelen,
op_data->op_name, PFID(&op_data->op_fid1));
- op_data->op_fsuid = cfs_curproc_fsuid();
- op_data->op_fsgid = cfs_curproc_fsgid();
+ op_data->op_fsuid = current_fsuid();
+ op_data->op_fsgid = current_fsgid();
op_data->op_cap = cfs_curproc_cap_pack();
- tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
- if (IS_ERR(tgt))
+ if (op_data->op_mea2 != NULL) {
+ struct lmv_stripe_md *lsm = op_data->op_mea2;
+ const struct lmv_oinfo *oinfo;
+
+ oinfo = lsm_name_to_stripe_info(lsm, op_data->op_name,
+ op_data->op_namelen);
+ if (IS_ERR(oinfo))
+ RETURN(PTR_ERR(oinfo));
+
+ op_data->op_fid2 = oinfo->lmo_fid;
+ }
+
+ tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
+ if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
/*
* Cancel UPDATE lock on child (fid1).
*/
op_data->op_flags |= MF_MDC_CANCEL_FID2;
- rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
+ rc = lmv_early_cancel(exp, NULL, op_data, tgt->ltd_idx, LCK_EX,
MDS_INODELOCK_UPDATE, MF_MDC_CANCEL_FID1);
if (rc != 0)
RETURN(rc);
const char *old, int oldlen, const char *new, int newlen,
struct ptlrpc_request **request)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *src_tgt;
- struct lmv_tgt_desc *tgt_tgt;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *src_tgt;
int rc;
ENTRY;
- LASSERT(oldlen != 0);
+ LASSERT(oldlen != 0);
- CDEBUG(D_INODE, "RENAME %*s in "DFID" to %*s in "DFID"\n",
- oldlen, old, PFID(&op_data->op_fid1),
- newlen, new, PFID(&op_data->op_fid2));
+ CDEBUG(D_INODE, "RENAME %.*s in "DFID":%d to %.*s in "DFID":%d\n",
+ oldlen, old, PFID(&op_data->op_fid1),
+ op_data->op_mea1 ? op_data->op_mea1->lsm_md_stripe_count : 0,
+ newlen, new, PFID(&op_data->op_fid2),
+ op_data->op_mea2 ? op_data->op_mea2->lsm_md_stripe_count : 0);
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
+ rc = lmv_check_connect(obd);
+ if (rc)
+ RETURN(rc);
- op_data->op_fsuid = cfs_curproc_fsuid();
- op_data->op_fsgid = cfs_curproc_fsgid();
+ op_data->op_fsuid = current_fsuid();
+ op_data->op_fsgid = current_fsgid();
op_data->op_cap = cfs_curproc_cap_pack();
- src_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
+ if (op_data->op_cli_flags & CLI_MIGRATE) {
+ LASSERTF(fid_is_sane(&op_data->op_fid3), "invalid FID "DFID"\n",
+ PFID(&op_data->op_fid3));
+ rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
+ if (rc)
+ RETURN(rc);
+ src_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid3);
+ } else {
+ if (op_data->op_mea1 != NULL) {
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
+
+ src_tgt = lmv_locate_target_for_name(lmv, lsm, old,
+ oldlen,
+ &op_data->op_fid1,
+ &op_data->op_mds);
+ if (IS_ERR(src_tgt))
+ RETURN(PTR_ERR(src_tgt));
+ } else {
+ src_tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(src_tgt))
+ RETURN(PTR_ERR(src_tgt));
+
+ op_data->op_mds = src_tgt->ltd_idx;
+ }
+
+ if (op_data->op_mea2) {
+ struct lmv_stripe_md *lsm = op_data->op_mea2;
+ const struct lmv_oinfo *oinfo;
+
+ oinfo = lsm_name_to_stripe_info(lsm, new, newlen);
+ if (IS_ERR(oinfo))
+ RETURN(PTR_ERR(oinfo));
+
+ op_data->op_fid2 = oinfo->lmo_fid;
+ }
+ }
if (IS_ERR(src_tgt))
RETURN(PTR_ERR(src_tgt));
- tgt_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
- if (IS_ERR(tgt_tgt))
- RETURN(PTR_ERR(tgt_tgt));
/*
* LOOKUP lock on src child (fid3) should also be cancelled for
* src_tgt in mdc_rename.
* Cancel UPDATE locks on tgt parent (fid2), tgt_tgt is its
* own target.
*/
- rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
+ rc = lmv_early_cancel(exp, NULL, op_data, src_tgt->ltd_idx,
LCK_EX, MDS_INODELOCK_UPDATE,
MF_MDC_CANCEL_FID2);
+ if (rc != 0)
+ RETURN(rc);
/*
- * Cancel LOOKUP locks on tgt child (fid4) for parent tgt_tgt.
+ * Cancel LOOKUP locks on source child (fid3) for parent tgt_tgt.
*/
- if (rc == 0) {
- rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
+ if (fid_is_sane(&op_data->op_fid3)) {
+ struct lmv_tgt_desc *tgt;
+
+ tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ /* Cancel LOOKUP lock on its parent */
+ rc = lmv_early_cancel(exp, tgt, op_data, src_tgt->ltd_idx,
LCK_EX, MDS_INODELOCK_LOOKUP,
- MF_MDC_CANCEL_FID4);
+ MF_MDC_CANCEL_FID3);
+ if (rc != 0)
+ RETURN(rc);
+
+ rc = lmv_early_cancel(exp, NULL, op_data, src_tgt->ltd_idx,
+ LCK_EX, MDS_INODELOCK_FULL,
+ MF_MDC_CANCEL_FID3);
+ if (rc != 0)
+ RETURN(rc);
}
/*
* Cancel all the locks on tgt child (fid4).
*/
- if (rc == 0)
- rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
+ if (fid_is_sane(&op_data->op_fid4))
+ rc = lmv_early_cancel(exp, NULL, op_data, src_tgt->ltd_idx,
LCK_EX, MDS_INODELOCK_FULL,
MF_MDC_CANCEL_FID4);
- if (rc == 0)
- rc = md_rename(src_tgt->ltd_exp, op_data, old, oldlen,
- new, newlen, request);
+ CDEBUG(D_INODE, DFID":m%d to "DFID"\n", PFID(&op_data->op_fid1),
+ op_data->op_mds, PFID(&op_data->op_fid2));
+
+ rc = md_rename(src_tgt->ltd_exp, op_data, old, oldlen, new, newlen,
+ request);
+
RETURN(rc);
}
RETURN(rc);
}
-static int lmv_sync(struct obd_export *exp, const struct lu_fid *fid,
- struct obd_capa *oc, struct ptlrpc_request **request)
+static int lmv_fsync(struct obd_export *exp, const struct lu_fid *fid,
+ struct obd_capa *oc, struct ptlrpc_request **request)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt;
- int rc;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt;
+ int rc;
+ ENTRY;
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
+ rc = lmv_check_connect(obd);
+ if (rc != 0)
+ RETURN(rc);
- tgt = lmv_find_target(lmv, fid);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ tgt = lmv_find_target(lmv, fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
- rc = md_sync(tgt->ltd_exp, fid, oc, request);
- RETURN(rc);
+ rc = md_fsync(tgt->ltd_exp, fid, oc, request);
+ RETURN(rc);
}
-static int lmv_readpage(struct obd_export *exp, struct md_op_data *op_data,
- struct page **pages, struct ptlrpc_request **request)
+/*
+ * Adjust a set of pages, each page containing an array of lu_dirpages,
+ * so that each page can be used as a single logical lu_dirpage.
+ *
+ * A lu_dirpage is laid out as follows, where s = ldp_hash_start,
+ * e = ldp_hash_end, f = ldp_flags, p = padding, and each "ent" is a
+ * struct lu_dirent. It has size up to LU_PAGE_SIZE. The ldp_hash_end
+ * value is used as a cookie to request the next lu_dirpage in a
+ * directory listing that spans multiple pages (two in this example):
+ * ________
+ * | |
+ * .|--------v------- -----.
+ * |s|e|f|p|ent|ent| ... |ent|
+ * '--|-------------- -----' Each CFS_PAGE contains a single
+ * '------. lu_dirpage.
+ * .---------v------- -----.
+ * |s|e|f|p|ent| 0 | ... | 0 |
+ * '----------------- -----'
+ *
+ * However, on hosts where the native VM page size (PAGE_CACHE_SIZE) is
+ * larger than LU_PAGE_SIZE, a single host page may contain multiple
+ * lu_dirpages. After reading the lu_dirpages from the MDS, the
+ * ldp_hash_end of the first lu_dirpage refers to the one immediately
+ * after it in the same CFS_PAGE (arrows simplified for brevity, but
+ * in general e0==s1, e1==s2, etc.):
+ *
+ * .-------------------- -----.
+ * |s0|e0|f0|p|ent|ent| ... |ent|
+ * |---v---------------- -----|
+ * |s1|e1|f1|p|ent|ent| ... |ent|
+ * |---v---------------- -----| Here, each CFS_PAGE contains
+ * ... multiple lu_dirpages.
+ * |---v---------------- -----|
+ * |s'|e'|f'|p|ent|ent| ... |ent|
+ * '---|---------------- -----'
+ * v
+ * .----------------------------.
+ * | next CFS_PAGE |
+ *
+ * This structure is transformed into a single logical lu_dirpage as follows:
+ *
+ * - Replace e0 with e' so the request for the next lu_dirpage gets the page
+ * labeled 'next CFS_PAGE'.
+ *
+ * - Copy the LDF_COLLIDE flag from f' to f0 to correctly reflect whether
+ * a hash collision with the next page exists.
+ *
+ * - Adjust the lde_reclen of the ending entry of each lu_dirpage to span
+ * to the first entry of the next lu_dirpage.
+ */
+#if PAGE_CACHE_SIZE > LU_PAGE_SIZE
+static void lmv_adjust_dirpages(struct page **pages, int ncfspgs, int nlupgs)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- __u64 offset = op_data->op_offset;
- int rc;
- int i;
- /* number of pages read, in CFS_PAGE_SIZE */
- int nrdpgs;
- /* number of pages transferred in LU_PAGE_SIZE */
- int nlupgs;
- struct lmv_tgt_desc *tgt;
- struct lu_dirpage *dp;
- struct lu_dirent *ent;
- ENTRY;
-
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
-
- CDEBUG(D_INODE, "READPAGE at "LPX64" from "DFID"\n",
- offset, PFID(&op_data->op_fid1));
-
- /*
- * This case handle directory lookup in clustered metadata case (i.e.
- * split directory is located on multiple md servers.)
- * each server keeps directory entries for certain range of hashes.
- * E.g. we have N server and suppose hash range is 0 to MAX_HASH.
- * first server will keep records with hashes [ 0 ... MAX_HASH /N - 1],
- * second one with hashes [MAX_HASH / N ... 2 * MAX_HASH / N] and
- * so on....
- * readdir can simply start reading entries from 0 - N server in
- * order but that will not scale well as all client will request dir in
- * to server in same order.
- * Following algorithm does optimization:
- * Instead of doing readdir in 1, 2, ...., N order, client with a
- * rank R does readdir in R, R + 1, ..., N, 1, ... R - 1 order.
- * (every client has rank R)
- * But ll_readdir() expect offset range [0 to MAX_HASH/N) but
- * since client ask dir from MDS{R} client has pages with offsets
- * [R*MAX_HASH/N ... (R + 1)*MAX_HASH/N] there for we do hash_adj
- * on hash values that we get.
- * Since these codes might be still useful for sharded directory, so
- * Keeping this code for further reference
- if (0) {
- LASSERT(nr > 0);
- seg_size = MAX_HASH_SIZE;
- do_div(seg_size, nr);
- los = obj->lo_stripes;
- tgt = lmv_get_target(lmv, los[0].ls_mds);
- rank = lmv_node_rank(tgt->ltd_exp, fid) % nr;
- tgt_tmp = offset;
- do_div(tgt_tmp, seg_size);
- tgt0_idx = do_div(tgt_tmp, nr);
- tgt_idx = (tgt0_idx + rank) % nr;
-
- if (tgt_idx < tgt0_idx)
- * Wrap around.
- *
- * Last segment has unusual length due to division
- * rounding.
- hash_adj = MAX_HASH_SIZE - seg_size * nr;
- else
- hash_adj = 0;
+ int i;
- hash_adj += rank * seg_size;
+ for (i = 0; i < ncfspgs; i++) {
+ struct lu_dirpage *dp = kmap(pages[i]);
+ struct lu_dirpage *first = dp;
+ struct lu_dirent *end_dirent = NULL;
+ struct lu_dirent *ent;
+ __u64 hash_end = dp->ldp_hash_end;
+ __u32 flags = dp->ldp_flags;
- CDEBUG(D_INODE, "Readpage hash adjustment: %x "LPX64" "
- LPX64"/%x -> "LPX64"/%x\n", rank, hash_adj,
- offset, tgt0_idx, offset + hash_adj, tgt_idx);
+ while (--nlupgs > 0) {
+ ent = lu_dirent_start(dp);
+ for (end_dirent = ent; ent != NULL;
+ end_dirent = ent, ent = lu_dirent_next(ent));
+
+ /* Advance dp to next lu_dirpage. */
+ dp = (struct lu_dirpage *)((char *)dp + LU_PAGE_SIZE);
+
+ /* Check if we've reached the end of the CFS_PAGE. */
+ if (!((unsigned long)dp & ~CFS_PAGE_MASK))
+ break;
+
+ /* Save the hash and flags of this lu_dirpage. */
+ hash_end = dp->ldp_hash_end;
+ flags = dp->ldp_flags;
+
+ /* Check if lu_dirpage contains no entries. */
+ if (!end_dirent)
+ break;
+
+ /* Enlarge the end entry lde_reclen from 0 to
+ * first entry of next lu_dirpage. */
+ LASSERT(le16_to_cpu(end_dirent->lde_reclen) == 0);
+ end_dirent->lde_reclen =
+ cpu_to_le16((char *)(dp->ldp_entries) -
+ (char *)end_dirent);
+ }
- offset = (offset + hash_adj) & MAX_HASH_SIZE;
- rid = lsm->mea_oinfo[tgt_idx].lmo_fid;
- tgt = lmv_get_target(lmv, lsm->mea_oinfo[tgt_idx].lmo_mds);
+ first->ldp_hash_end = hash_end;
+ first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
+ first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
- CDEBUG(D_INODE, "Forward to "DFID" with offset %lu i %d\n",
- PFID(&rid), (unsigned long)offset, tgt_idx);
+ kunmap(pages[i]);
}
- */
- tgt = lmv_find_target(lmv, &op_data->op_fid1);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ LASSERTF(nlupgs == 0, "left = %d", nlupgs);
+}
+#else
+#define lmv_adjust_dirpages(pages, ncfspgs, nlupgs) do {} while (0)
+#endif /* PAGE_CACHE_SIZE > LU_PAGE_SIZE */
+
+#define NORMAL_MAX_STRIPES 4
+int lmv_read_entry(struct obd_export *exp, struct md_op_data *op_data,
+ struct md_callback *cb_op, struct lu_dirent **ldp,
+ struct page **ppage)
+{
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
+ struct lu_dirent *tmp_ents[NORMAL_MAX_STRIPES];
+ struct lu_dirent **ents = NULL;
+ int stripe_count;
+ __u64 min_hash;
+ int min_idx = 0;
+ struct page *min_page = NULL;
+ int i;
+ int rc;
+ ENTRY;
- rc = md_readpage(tgt->ltd_exp, op_data, pages, request);
- if (rc != 0)
+ rc = lmv_check_connect(obd);
+ if (rc)
RETURN(rc);
- nrdpgs = ((*request)->rq_bulk->bd_nob_transferred + CFS_PAGE_SIZE - 1)
- >> CFS_PAGE_SHIFT;
- nlupgs = (*request)->rq_bulk->bd_nob_transferred >> LU_PAGE_SHIFT;
- LASSERT(!((*request)->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK));
- LASSERT(nrdpgs > 0 && nrdpgs <= op_data->op_npages);
-
- CDEBUG(D_INODE, "read %d(%d)/%d pages\n", nrdpgs, nlupgs,
- op_data->op_npages);
+ if (lsm == NULL)
+ stripe_count = 1;
+ else
+ stripe_count = lsm->lsm_md_stripe_count;
- for (i = 0; i < nrdpgs; i++) {
-#if CFS_PAGE_SIZE > LU_PAGE_SIZE
- struct lu_dirpage *first;
- __u64 hash_end = 0;
- __u32 flags = 0;
-#endif
- struct lu_dirent *tmp = NULL;
-
- dp = cfs_kmap(pages[i]);
- ent = lu_dirent_start(dp);
-#if CFS_PAGE_SIZE > LU_PAGE_SIZE
- first = dp;
- hash_end = dp->ldp_hash_end;
-repeat:
-#endif
- nlupgs--;
+ if (stripe_count > NORMAL_MAX_STRIPES) {
+ OBD_ALLOC(ents, sizeof(ents[0]) * stripe_count);
+ if (ents == NULL)
+ GOTO(out, rc = -ENOMEM);
+ } else {
+ ents = tmp_ents;
+ memset(ents, 0, sizeof(ents[0]) * stripe_count);
+ }
- for (tmp = ent; ent != NULL;
- tmp = ent, ent = lu_dirent_next(ent));
-#if CFS_PAGE_SIZE > LU_PAGE_SIZE
- dp = (struct lu_dirpage *)((char *)dp + LU_PAGE_SIZE);
- if (((unsigned long)dp & ~CFS_PAGE_MASK) && nlupgs > 0) {
- ent = lu_dirent_start(dp);
+ min_hash = MDS_DIR_END_OFF;
+ for (i = 0; i < stripe_count; i++) {
+ struct lmv_tgt_desc *tgt;
+ struct page *page = NULL;
+
+ if (likely(lsm == NULL)) {
+ tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(tgt))
+ GOTO(out, rc = PTR_ERR(tgt));
+ LASSERT(op_data->op_data != NULL);
+ } else {
+ tgt = lmv_get_target(lmv, lsm->lsm_md_oinfo[i].lmo_mds);
+ if (IS_ERR(tgt))
+ GOTO(out, rc = PTR_ERR(tgt));
+ op_data->op_fid1 = lsm->lsm_md_oinfo[i].lmo_fid;
+ op_data->op_fid2 = lsm->lsm_md_oinfo[i].lmo_fid;
+ op_data->op_stripe_offset = i;
+ }
- if (tmp) {
- /* enlarge the end entry lde_reclen from 0 to
- * first entry of next lu_dirpage, in this way
- * several lu_dirpages can be stored into one
- * client page on client. */
- tmp = ((void *)tmp) +
- le16_to_cpu(tmp->lde_reclen);
- tmp->lde_reclen =
- cpu_to_le16((char *)(dp->ldp_entries) -
- (char *)tmp);
- goto repeat;
- }
+ rc = md_read_entry(tgt->ltd_exp, op_data, cb_op, &ents[i],
+ &page);
+ if (rc != 0)
+ GOTO(out, rc);
+
+ if (ents[i] != NULL &&
+ le64_to_cpu(ents[i]->lde_hash) <= min_hash) {
+ if (min_page != NULL)
+ page_cache_release(min_page);
+ min_page = page;
+ min_hash = le64_to_cpu(ents[i]->lde_hash);
+ min_idx = i;
}
- first->ldp_hash_end = hash_end;
- first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
- first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
-#else
- SET_BUT_UNUSED(tmp);
-#endif
- cfs_kunmap(pages[i]);
}
+
+ if (min_hash != MDS_DIR_END_OFF)
+ *ldp = ents[min_idx];
+ else
+ *ldp = NULL;
+out:
+ if (stripe_count > NORMAL_MAX_STRIPES && ents != NULL)
+ OBD_FREE(ents, sizeof(ents[0]) * stripe_count);
+
+ if (rc != 0 && min_page != NULL) {
+ kunmap(min_page);
+ page_cache_release(min_page);
+ } else {
+ *ppage = min_page;
+ }
+
RETURN(rc);
}
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *tgt = NULL;
+ struct lmv_tgt_desc *parent_tgt = NULL;
struct mdt_body *body;
int rc;
ENTRY;
RETURN(rc);
retry:
/* Send unlink requests to the MDT where the child is located */
- if (likely(!fid_is_zero(&op_data->op_fid2)))
- tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
- else
+ if (likely(!fid_is_zero(&op_data->op_fid2))) {
+ tgt = lmv_find_target(lmv, &op_data->op_fid2);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+
+ /* For striped dir, we need to locate the parent as well */
+ if (op_data->op_mea1 != NULL &&
+ op_data->op_mea1->lsm_md_stripe_count > 1) {
+ struct lmv_tgt_desc *tmp;
+
+ LASSERT(op_data->op_name != NULL &&
+ op_data->op_namelen != 0);
+ tmp = lmv_locate_target_for_name(lmv,
+ op_data->op_mea1,
+ op_data->op_name,
+ op_data->op_namelen,
+ &op_data->op_fid1,
+ &op_data->op_mds);
+ if (IS_ERR(tmp))
+ RETURN(PTR_ERR(tmp));
+ }
+ } else {
tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
+ }
- op_data->op_fsuid = cfs_curproc_fsuid();
- op_data->op_fsgid = cfs_curproc_fsgid();
+ op_data->op_fsuid = current_fsuid();
+ op_data->op_fsgid = current_fsgid();
op_data->op_cap = cfs_curproc_cap_pack();
/*
/*
* Cancel FULL locks on child (fid3).
*/
- rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
- MDS_INODELOCK_FULL, MF_MDC_CANCEL_FID3);
+ parent_tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ if (IS_ERR(parent_tgt))
+ RETURN(PTR_ERR(parent_tgt));
+ if (parent_tgt != tgt) {
+ rc = lmv_early_cancel(exp, parent_tgt, op_data, tgt->ltd_idx,
+ LCK_EX, MDS_INODELOCK_LOOKUP,
+ MF_MDC_CANCEL_FID3);
+ }
+
+ rc = lmv_early_cancel(exp, NULL, op_data, tgt->ltd_idx, LCK_EX,
+ MDS_INODELOCK_FULL, MF_MDC_CANCEL_FID3);
if (rc != 0)
RETURN(rc);
case OBD_CLEANUP_EXPORTS:
fld_client_proc_fini(&lmv->lmv_fld);
lprocfs_obd_cleanup(obd);
+ lprocfs_free_md_stats(obd);
break;
default:
break;
lmv = &obd->u.lmv;
if (keylen >= strlen("remote_flag") && !strcmp(key, "remote_flag")) {
- struct lmv_tgt_desc *tgt;
int i;
rc = lmv_check_connect(obd);
LASSERT(*vallen == sizeof(__u32));
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- tgt = lmv->tgts[i];
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
/*
* All tgts should be connected when this gets called.
*/
obd_count keylen, void *key, obd_count vallen,
void *val, struct ptlrpc_request_set *set)
{
- struct lmv_tgt_desc *tgt;
+ struct lmv_tgt_desc *tgt = NULL;
struct obd_device *obd;
struct lmv_obd *lmv;
int rc = 0;
RETURN(-EINVAL);
}
-int lmv_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
- struct lov_stripe_md *lsm)
+static int lmv_pack_md_v1(const struct lmv_stripe_md *lsm,
+ struct lmv_mds_md_v1 *lmm1)
{
- struct obd_device *obd = class_exp2obd(exp);
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_stripe_md *meap;
- struct lmv_stripe_md *lsmp;
- int mea_size;
- int i;
- ENTRY;
+ int cplen;
+ int i;
- mea_size = lmv_get_easize(lmv);
- if (!lmmp)
- RETURN(mea_size);
+ lmm1->lmv_magic = cpu_to_le32(lsm->lsm_md_magic);
+ lmm1->lmv_stripe_count = cpu_to_le32(lsm->lsm_md_stripe_count);
+ lmm1->lmv_master_mdt_index = cpu_to_le32(lsm->lsm_md_master_mdt_index);
+ lmm1->lmv_hash_type = cpu_to_le32(lsm->lsm_md_hash_type);
+ cplen = strlcpy(lmm1->lmv_pool_name, lsm->lsm_md_pool_name,
+ sizeof(lmm1->lmv_pool_name));
+ if (cplen >= sizeof(lmm1->lmv_pool_name))
+ return -E2BIG;
- if (*lmmp && !lsm) {
- OBD_FREE_LARGE(*lmmp, mea_size);
- *lmmp = NULL;
- RETURN(0);
- }
+ for (i = 0; i < lsm->lsm_md_stripe_count; i++)
+ fid_cpu_to_le(&lmm1->lmv_stripe_fids[i],
+ &lsm->lsm_md_oinfo[i].lmo_fid);
+ return 0;
+}
- if (*lmmp == NULL) {
- OBD_ALLOC_LARGE(*lmmp, mea_size);
- if (*lmmp == NULL)
- RETURN(-ENOMEM);
- }
+int lmv_pack_md(union lmv_mds_md **lmmp, const struct lmv_stripe_md *lsm,
+ int stripe_count)
+{
+ int lmm_size = 0;
+ bool allocated = false;
+ int rc = 0;
+ ENTRY;
- if (!lsm)
- RETURN(mea_size);
+ LASSERT(lmmp != NULL);
+ /* Free lmm */
+ if (*lmmp != NULL && lsm == NULL) {
+ int stripe_count;
- lsmp = (struct lmv_stripe_md *)lsm;
- meap = (struct lmv_stripe_md *)*lmmp;
+ stripe_count = lmv_mds_md_stripe_count_get(*lmmp);
+ lmm_size = lmv_mds_md_size(stripe_count,
+ le32_to_cpu((*lmmp)->lmv_magic));
+ if (lmm_size == 0)
+ RETURN(-EINVAL);
+ OBD_FREE(*lmmp, lmm_size);
+ *lmmp = NULL;
+ RETURN(0);
+ }
- if (lsmp->mea_magic != MEA_MAGIC_LAST_CHAR &&
- lsmp->mea_magic != MEA_MAGIC_ALL_CHARS)
- RETURN(-EINVAL);
+ /* Alloc lmm */
+ if (*lmmp == NULL && lsm == NULL) {
+ lmm_size = lmv_mds_md_size(stripe_count, LMV_MAGIC);
+ LASSERT(lmm_size > 0);
+ OBD_ALLOC(*lmmp, lmm_size);
+ if (*lmmp == NULL)
+ RETURN(-ENOMEM);
+ lmv_mds_md_stripe_count_set(*lmmp, stripe_count);
+ (*lmmp)->lmv_magic = cpu_to_le32(LMV_MAGIC);
+ RETURN(lmm_size);
+ }
- meap->mea_magic = cpu_to_le32(lsmp->mea_magic);
- meap->mea_count = cpu_to_le32(lsmp->mea_count);
- meap->mea_master = cpu_to_le32(lsmp->mea_master);
+ /* pack lmm */
+ LASSERT(lsm != NULL);
+ lmm_size = lmv_mds_md_size(lsm->lsm_md_stripe_count, lsm->lsm_md_magic);
+ if (*lmmp == NULL) {
+ OBD_ALLOC(*lmmp, lmm_size);
+ if (*lmmp == NULL)
+ RETURN(-ENOMEM);
+ allocated = true;
+ }
- for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- meap->mea_ids[i] = meap->mea_ids[i];
- fid_cpu_to_le(&meap->mea_ids[i], &meap->mea_ids[i]);
- }
+ switch (lsm->lsm_md_magic) {
+ case LMV_MAGIC_V1:
+ rc = lmv_pack_md_v1(lsm, &(*lmmp)->lmv_md_v1);
+ break;
+ default:
+ rc = -EINVAL;
+ break;
+ }
+
+ if (rc != 0 && allocated) {
+ OBD_FREE(*lmmp, lmm_size);
+ *lmmp = NULL;
+ }
- RETURN(mea_size);
+ RETURN(lmm_size);
}
+EXPORT_SYMBOL(lmv_pack_md);
-int lmv_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
- struct lov_mds_md *lmm, int lmm_size)
+static int lmv_unpack_md_v1(struct obd_export *exp, struct lmv_stripe_md *lsm,
+ const struct lmv_mds_md_v1 *lmm1)
{
- struct obd_device *obd = class_exp2obd(exp);
- struct lmv_stripe_md **tmea = (struct lmv_stripe_md **)lsmp;
- struct lmv_stripe_md *mea = (struct lmv_stripe_md *)lmm;
- struct lmv_obd *lmv = &obd->u.lmv;
- int mea_size;
- int i;
- __u32 magic;
- ENTRY;
+ struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ int stripe_count;
+ int cplen;
+ int i;
+ int rc = 0;
+ ENTRY;
- mea_size = lmv_get_easize(lmv);
- if (lsmp == NULL)
- return mea_size;
+ lsm->lsm_md_magic = le32_to_cpu(lmm1->lmv_magic);
+ lsm->lsm_md_stripe_count = le32_to_cpu(lmm1->lmv_stripe_count);
+ lsm->lsm_md_master_mdt_index = le32_to_cpu(lmm1->lmv_master_mdt_index);
+ lsm->lsm_md_hash_type = le32_to_cpu(lmm1->lmv_hash_type);
+ lsm->lsm_md_layout_version = le32_to_cpu(lmm1->lmv_layout_version);
+ cplen = strlcpy(lsm->lsm_md_pool_name, lmm1->lmv_pool_name,
+ sizeof(lsm->lsm_md_pool_name));
+
+ if (cplen >= sizeof(lsm->lsm_md_pool_name))
+ RETURN(-E2BIG);
+
+ CDEBUG(D_INFO, "unpack lsm count %d, master %d hash_type %d"
+ "layout_version %d\n", lsm->lsm_md_stripe_count,
+ lsm->lsm_md_master_mdt_index, lsm->lsm_md_hash_type,
+ lsm->lsm_md_layout_version);
+
+ stripe_count = le32_to_cpu(lmm1->lmv_stripe_count);
+ for (i = 0; i < le32_to_cpu(stripe_count); i++) {
+ fid_le_to_cpu(&lsm->lsm_md_oinfo[i].lmo_fid,
+ &lmm1->lmv_stripe_fids[i]);
+ rc = lmv_fld_lookup(lmv, &lsm->lsm_md_oinfo[i].lmo_fid,
+ &lsm->lsm_md_oinfo[i].lmo_mds);
+ if (rc != 0)
+ RETURN(rc);
+ CDEBUG(D_INFO, "unpack fid #%d "DFID"\n", i,
+ PFID(&lsm->lsm_md_oinfo[i].lmo_fid));
+ }
- if (*lsmp != NULL && lmm == NULL) {
- OBD_FREE_LARGE(*tmea, mea_size);
- *lsmp = NULL;
- RETURN(0);
- }
+ RETURN(rc);
+}
+
+int lmv_unpack_md(struct obd_export *exp, struct lmv_stripe_md **lsmp,
+ const union lmv_mds_md *lmm, int stripe_count)
+{
+ struct lmv_stripe_md *lsm;
+ int lsm_size;
+ int rc;
+ bool allocated = false;
+ ENTRY;
- LASSERT(mea_size == lmm_size);
+ LASSERT(lsmp != NULL);
- OBD_ALLOC_LARGE(*tmea, mea_size);
- if (*tmea == NULL)
- RETURN(-ENOMEM);
+ lsm = *lsmp;
+ /* Free memmd */
+ if (lsm != NULL && lmm == NULL) {
+#ifdef __KERNEL__
+ int i;
+ for (i = 1; i < lsm->lsm_md_stripe_count; i++) {
+ if (lsm->lsm_md_oinfo[i].lmo_root != NULL)
+ iput(lsm->lsm_md_oinfo[i].lmo_root);
+ }
+#endif
+ lsm_size = lmv_stripe_md_size(lsm->lsm_md_stripe_count);
+ OBD_FREE(lsm, lsm_size);
+ *lsmp = NULL;
+ RETURN(0);
+ }
- if (!lmm)
- RETURN(mea_size);
+ /* Alloc memmd */
+ if (lsm == NULL && lmm == NULL) {
+ lsm_size = lmv_stripe_md_size(stripe_count);
+ OBD_ALLOC(lsm, lsm_size);
+ if (lsm == NULL)
+ RETURN(-ENOMEM);
+ lsm->lsm_md_stripe_count = stripe_count;
+ *lsmp = lsm;
+ RETURN(0);
+ }
- if (mea->mea_magic == MEA_MAGIC_LAST_CHAR ||
- mea->mea_magic == MEA_MAGIC_ALL_CHARS ||
- mea->mea_magic == MEA_MAGIC_HASH_SEGMENT)
- {
- magic = le32_to_cpu(mea->mea_magic);
- } else {
- /*
- * Old mea is not handled here.
- */
- CERROR("Old not supportable EA is found\n");
- LBUG();
- }
+ /* Unpack memmd */
+ if (le32_to_cpu(lmm->lmv_magic) != LMV_MAGIC_V1 &&
+ le32_to_cpu(lmm->lmv_magic) != LMV_MAGIC_MIGRATE &&
+ le32_to_cpu(lmm->lmv_magic) != LMV_USER_MAGIC) {
+ CERROR("%s: invalid lmv magic %x: rc = %d\n",
+ exp->exp_obd->obd_name, le32_to_cpu(lmm->lmv_magic),
+ -EIO);
+ RETURN(-EIO);
+ }
- (*tmea)->mea_magic = magic;
- (*tmea)->mea_count = le32_to_cpu(mea->mea_count);
- (*tmea)->mea_master = le32_to_cpu(mea->mea_master);
+ if (le32_to_cpu(lmm->lmv_magic) == LMV_MAGIC_V1 ||
+ le32_to_cpu(lmm->lmv_magic) == LMV_MAGIC_MIGRATE)
+ lsm_size = lmv_stripe_md_size(lmv_mds_md_stripe_count_get(lmm));
+ else
+ /**
+ * Unpack default dirstripe(lmv_user_md) to lmv_stripe_md,
+ * stripecount should be 0 then.
+ */
+ lsm_size = lmv_stripe_md_size(0);
- for (i = 0; i < (*tmea)->mea_count; i++) {
- (*tmea)->mea_ids[i] = mea->mea_ids[i];
- fid_le_to_cpu(&(*tmea)->mea_ids[i], &(*tmea)->mea_ids[i]);
- }
- RETURN(mea_size);
+ lsm_size = lmv_stripe_md_size(lmv_mds_md_stripe_count_get(lmm));
+ if (lsm == NULL) {
+ OBD_ALLOC(lsm, lsm_size);
+ if (lsm == NULL)
+ RETURN(-ENOMEM);
+ allocated = true;
+ *lsmp = lsm;
+ }
+
+ switch (le32_to_cpu(lmm->lmv_magic)) {
+ case LMV_MAGIC_V1:
+ case LMV_MAGIC_MIGRATE:
+ rc = lmv_unpack_md_v1(exp, lsm, &lmm->lmv_md_v1);
+ break;
+ default:
+ CERROR("%s: unrecognized magic %x\n", exp->exp_obd->obd_name,
+ le32_to_cpu(lmm->lmv_magic));
+ rc = -EINVAL;
+ break;
+ }
+
+ if (rc != 0 && allocated) {
+ OBD_FREE(lsm, lsm_size);
+ *lsmp = NULL;
+ lsm_size = rc;
+ }
+ RETURN(lsm_size);
+}
+
+int lmv_alloc_memmd(struct lmv_stripe_md **lsmp, int stripes)
+{
+ return lmv_unpack_md(NULL, lsmp, NULL, stripes);
+}
+EXPORT_SYMBOL(lmv_alloc_memmd);
+
+void lmv_free_memmd(struct lmv_stripe_md *lsm)
+{
+ lmv_unpack_md(NULL, &lsm, NULL, 0);
+}
+EXPORT_SYMBOL(lmv_free_memmd);
+
+int lmv_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
+ struct lov_mds_md *lmm, int disk_len)
+{
+ return lmv_unpack_md(exp, (struct lmv_stripe_md **)lsmp,
+ (union lmv_mds_md *)lmm, disk_len);
+}
+
+int lmv_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
+ struct lov_stripe_md *lsm)
+{
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv_obd = &obd->u.lmv;
+ const struct lmv_stripe_md *lmv = (struct lmv_stripe_md *)lsm;
+ int stripe_count;
+
+ if (lmmp == NULL) {
+ if (lsm != NULL)
+ stripe_count = lmv->lsm_md_stripe_count;
+ else
+ stripe_count = lmv_obd->desc.ld_tgt_count;
+
+ return lmv_mds_md_size(stripe_count, LMV_MAGIC_V1);
+ }
+
+ return lmv_pack_md((union lmv_mds_md **)lmmp, lmv, 0);
}
static int lmv_cancel_unused(struct obd_export *exp, const struct lu_fid *fid,
struct lmv_obd *lmv = &obd->u.lmv;
int rc = 0;
int err;
- int i;
+ __u32 i;
ENTRY;
LASSERT(fid != NULL);
- for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL ||
- lmv->tgts[i]->ltd_active == 0)
+ for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
continue;
- err = md_cancel_unused(lmv->tgts[i]->ltd_exp, fid,
- policy, mode, flags, opaque);
+ err = md_cancel_unused(tgt->ltd_exp, fid, policy, mode, flags,
+ opaque);
if (!rc)
rc = err;
- }
+ }
RETURN(rc);
}
int lmv_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
__u64 *bits)
{
- struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
- int rc;
+ struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_tgt_desc *tgt = lmv->tgts[0];
+ int rc;
ENTRY;
- rc = md_set_lock_data(lmv->tgts[0]->ltd_exp, lockh, data, bits);
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ RETURN(-EINVAL);
+ rc = md_set_lock_data(tgt->ltd_exp, lockh, data, bits);
RETURN(rc);
}
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
ldlm_mode_t rc;
- int i;
+ __u32 i;
ENTRY;
CDEBUG(D_INODE, "Lock match for "DFID"\n", PFID(fid));
* one fid was created in.
*/
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- if (lmv->tgts[i] == NULL ||
- lmv->tgts[i]->ltd_exp == NULL ||
- lmv->tgts[i]->ltd_active == 0)
+ struct lmv_tgt_desc *tgt = lmv->tgts[i];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
continue;
- rc = md_lock_match(lmv->tgts[i]->ltd_exp, flags, fid,
- type, policy, mode, lockh);
+ rc = md_lock_match(tgt->ltd_exp, flags, fid, type, policy, mode,
+ lockh);
if (rc)
RETURN(rc);
}
struct lustre_md *md)
{
struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_tgt_desc *tgt = lmv->tgts[0];
+
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ RETURN(-EINVAL);
return md_get_lustre_md(lmv->tgts[0]->ltd_exp, req, dt_exp, md_exp, md);
}
int lmv_free_lustre_md(struct obd_export *exp, struct lustre_md *md)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt = lmv->tgts[0];
ENTRY;
- if (md->mea)
- obd_free_memmd(exp, (void *)&md->mea);
+ if (md->lmv != NULL) {
+ lmv_free_memmd(md->lmv);
+ md->lmv = NULL;
+ }
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ RETURN(-EINVAL);
RETURN(md_free_lustre_md(lmv->tgts[0]->ltd_exp, md));
}
int lmv_set_open_replay_data(struct obd_export *exp,
- struct obd_client_handle *och,
- struct ptlrpc_request *open_req)
+ struct obd_client_handle *och,
+ struct lookup_intent *it)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt;
+ ENTRY;
- tgt = lmv_find_target(lmv, &och->och_fid);
- if (IS_ERR(tgt))
- RETURN(PTR_ERR(tgt));
+ tgt = lmv_find_target(lmv, &och->och_fid);
+ if (IS_ERR(tgt))
+ RETURN(PTR_ERR(tgt));
- RETURN(md_set_open_replay_data(tgt->ltd_exp, och, open_req));
+ RETURN(md_set_open_replay_data(tgt->ltd_exp, och, it));
}
int lmv_clear_open_replay_data(struct obd_export *exp,
int lmv_unpack_capa(struct obd_export *exp, struct ptlrpc_request *req,
const struct req_msg_field *field, struct obd_capa **oc)
{
- struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
+ struct lmv_tgt_desc *tgt = lmv->tgts[0];
- return md_unpack_capa(lmv->tgts[0]->ltd_exp, req, field, oc);
+ if (tgt == NULL || tgt->ltd_exp == NULL)
+ RETURN(-EINVAL);
+ return md_unpack_capa(tgt->ltd_exp, req, field, oc);
}
int lmv_intent_getattr_async(struct obd_export *exp,
if (rc)
RETURN(rc);
- tgt = lmv_find_target(lmv, &op_data->op_fid1);
+ tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
struct obd_device *obd = class_exp2obd(exp);
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *tgt = lmv->tgts[0];
- int rc = 0, i;
+ int rc = 0;
+ __u32 i;
__u64 curspace, curinodes;
ENTRY;
- if (!lmv->desc.ld_tgt_count || !tgt->ltd_active) {
- CERROR("master lmv inactive\n");
- RETURN(-EIO);
- }
+ if (tgt == NULL ||
+ tgt->ltd_exp == NULL ||
+ !tgt->ltd_active ||
+ lmv->desc.ld_tgt_count == 0) {
+ CERROR("master lmv inactive\n");
+ RETURN(-EIO);
+ }
if (oqctl->qc_cmd != Q_GETOQUOTA) {
rc = obd_quotactl(tgt->ltd_exp, oqctl);
int err;
tgt = lmv->tgts[i];
- if (tgt == NULL || tgt->ltd_exp == NULL || tgt->ltd_active == 0)
+ if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
continue;
- if (!tgt->ltd_active) {
- CDEBUG(D_HA, "mdt %d is inactive.\n", i);
- continue;
- }
err = obd_quotactl(tgt->ltd_exp, oqctl);
if (err) {
int lmv_quotacheck(struct obd_device *unused, struct obd_export *exp,
struct obd_quotactl *oqctl)
{
- struct obd_device *obd = class_exp2obd(exp);
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lmv_tgt_desc *tgt;
- int i, rc = 0;
- ENTRY;
+ struct obd_device *obd = class_exp2obd(exp);
+ struct lmv_obd *lmv = &obd->u.lmv;
+ struct lmv_tgt_desc *tgt;
+ __u32 i;
+ int rc = 0;
+ ENTRY;
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- int err;
+ int err;
tgt = lmv->tgts[i];
if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active) {
CERROR("lmv idx %d inactive\n", i);
RETURN(rc);
}
+int lmv_update_lsm_md(struct obd_export *exp, struct lmv_stripe_md *lsm,
+ struct mdt_body *body, ldlm_blocking_callback cb_blocking)
+{
+ if (lsm->lsm_md_stripe_count <= 1)
+ return 0;
+
+ return lmv_revalidate_slaves(exp, body, lsm, cb_blocking, 0);
+}
+
+int lmv_merge_attr(struct obd_export *exp, const struct lmv_stripe_md *lsm,
+ struct cl_attr *attr)
+{
+#ifdef __KERNEL__
+ int i;
+
+ for (i = 0; i < lsm->lsm_md_stripe_count; i++) {
+ struct inode *inode = lsm->lsm_md_oinfo[i].lmo_root;
+
+ CDEBUG(D_INFO, ""DFID" size %llu, nlink %u, atime %lu ctime"
+ "%lu, mtime %lu.\n", PFID(&lsm->lsm_md_oinfo[i].lmo_fid),
+ i_size_read(inode), inode->i_nlink,
+ LTIME_S(inode->i_atime), LTIME_S(inode->i_ctime),
+ LTIME_S(inode->i_mtime));
+
+ /* for slave stripe, it needs to subtract nlink for . and .. */
+ if (i != 0)
+ attr->cat_nlink += inode->i_nlink - 2;
+ else
+ attr->cat_nlink = inode->i_nlink;
+
+ attr->cat_size += i_size_read(inode);
+
+ if (attr->cat_atime < LTIME_S(inode->i_atime))
+ attr->cat_atime = LTIME_S(inode->i_atime);
+
+ if (attr->cat_ctime < LTIME_S(inode->i_ctime))
+ attr->cat_ctime = LTIME_S(inode->i_ctime);
+
+ if (attr->cat_mtime < LTIME_S(inode->i_mtime))
+ attr->cat_mtime = LTIME_S(inode->i_mtime);
+ }
+#endif
+ return 0;
+}
+
struct obd_ops lmv_obd_ops = {
.o_owner = THIS_MODULE,
.o_setup = lmv_setup,
.m_rename = lmv_rename,
.m_setattr = lmv_setattr,
.m_setxattr = lmv_setxattr,
- .m_sync = lmv_sync,
- .m_readpage = lmv_readpage,
+ .m_fsync = lmv_fsync,
+ .m_read_entry = lmv_read_entry,
.m_unlink = lmv_unlink,
.m_init_ea_size = lmv_init_ea_size,
.m_cancel_unused = lmv_cancel_unused,
.m_set_lock_data = lmv_set_lock_data,
.m_lock_match = lmv_lock_match,
- .m_get_lustre_md = lmv_get_lustre_md,
- .m_free_lustre_md = lmv_free_lustre_md,
+ .m_get_lustre_md = lmv_get_lustre_md,
+ .m_free_lustre_md = lmv_free_lustre_md,
+ .m_update_lsm_md = lmv_update_lsm_md,
+ .m_merge_attr = lmv_merge_attr,
.m_set_open_replay_data = lmv_set_open_replay_data,
.m_clear_open_replay_data = lmv_clear_open_replay_data,
.m_renew_capa = lmv_renew_capa,
int __init lmv_init(void)
{
- struct lprocfs_static_vars lvars;
- int rc;
-
- lprocfs_lmv_init_vars(&lvars);
-
- rc = class_register_type(&lmv_obd_ops, &lmv_md_ops,
- lvars.module_vars, LUSTRE_LMV_NAME, NULL);
- return rc;
+ return class_register_type(&lmv_obd_ops, &lmv_md_ops, NULL,
+#ifndef HAVE_ONLY_PROCFS_SEQ
+ NULL,
+#endif
+ LUSTRE_LMV_NAME, NULL);
}
#ifdef __KERNEL__