*/
#define DEBUG_SUBSYSTEM S_LMV
-#ifdef __KERNEL__
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/mm.h>
-#include <asm/div64.h>
+#include <linux/math64.h>
#include <linux/seq_file.h>
#include <linux/namei.h>
-#else
-#include <liblustre.h>
-#endif
#include <lustre/lustre_idl.h>
#include <obd_support.h>
#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 <lustre_ioctl.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(__u32 lmv_hash_type, unsigned int stripe_count,
+ const char *name, int namelen)
+{
+ int idx;
+ __u32 hash_type = lmv_hash_type & LMV_HASH_TYPE_MASK;
+
+ LASSERT(namelen > 0);
+ if (stripe_count <= 1)
+ return 0;
+
+ /* for migrating object, always start from 0 stripe */
+ if (lmv_hash_type & LMV_HASH_FLAG_MIGRATION)
+ return 0;
+
+ switch (hash_type) {
+ case LMV_HASH_TYPE_ALL_CHARS:
+ idx = lmv_hash_all_chars(stripe_count, name, namelen);
+ break;
+ case LMV_HASH_TYPE_FNV_1A_64:
+ idx = lmv_hash_fnv1a(stripe_count, name, namelen);
+ break;
+ default:
+ idx = -EBADFD;
+ break;
+ }
+
+ CDEBUG(D_INFO, "name %.*s hash_type %d idx %d\n", namelen, name,
+ hash_type, idx);
+
+ return idx;
+}
+
static void lmv_activate_target(struct lmv_obd *lmv,
struct lmv_tgt_desc *tgt,
int activate)
struct obd_uuid *cluuid, struct obd_connect_data *data,
void *localdata)
{
-#ifdef __KERNEL__
- struct proc_dir_entry *lmv_proc_dir;
-#endif
struct lmv_obd *lmv = &obd->u.lmv;
struct lustre_handle conn = { 0 };
int rc = 0;
if (data)
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;
- }
-#endif
+ if (lmv->targets_proc_entry == NULL) {
+ lmv->targets_proc_entry = lprocfs_seq_register("target_obds",
+ obd->obd_proc_entry,
+ NULL, NULL);
+ if (IS_ERR(lmv->targets_proc_entry)) {
+ CERROR("%s: cannot register "
+ "/proc/fs/lustre/%s/%s/target_obds\n",
+ obd->obd_name, obd->obd_type->typ_name,
+ obd->obd_name);
+ lmv->targets_proc_entry = NULL;
+ }
+ }
/*
* All real clients should perform actual connection right away, because
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);
- }
-#endif
-
- RETURN(rc);
+ if (rc && lmv->targets_proc_entry != NULL)
+ lprocfs_remove(&lmv->targets_proc_entry);
+ RETURN(rc);
}
static void lmv_set_timeouts(struct obd_device *obd)
}
static int lmv_init_ea_size(struct obd_export *exp, int easize,
- int def_easize, int cookiesize)
+ int def_easize, int cookiesize, int def_cookiesize)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
lmv->max_cookiesize = cookiesize;
change = 1;
}
- if (change == 0)
- RETURN(0);
+ if (lmv->max_def_cookiesize < def_cookiesize) {
+ lmv->max_def_cookiesize = def_cookiesize;
+ change = 1;
+ }
+ if (change == 0)
+ RETURN(0);
- if (lmv->connected == 0)
- RETURN(0);
+ if (lmv->connected == 0)
+ RETURN(0);
for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
struct lmv_tgt_desc *tgt = lmv->tgts[i];
}
rc = md_init_ea_size(tgt->ltd_exp, easize, def_easize,
- cookiesize);
+ cookiesize, def_cookiesize);
if (rc) {
CERROR("%s: obd_init_ea_size() failed on MDT target %d:"
- " rc = %d.\n", obd->obd_name, i, rc);
+ " rc = %d\n", obd->obd_name, i, rc);
break;
}
}
int lmv_connect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
{
-#ifdef __KERNEL__
- struct proc_dir_entry *lmv_proc_dir;
-#endif
struct lmv_obd *lmv = &obd->u.lmv;
struct obd_uuid *cluuid = &lmv->cluuid;
struct obd_uuid lmv_mdc_uuid = { "LMV_MDC_UUID" };
}
}
- tgt->ltd_active = 1;
- tgt->ltd_exp = mdc_exp;
- lmv->desc.ld_active_tgt_count++;
-
- md_init_ea_size(tgt->ltd_exp, lmv->max_easize,
- lmv->max_def_easize, lmv->max_cookiesize);
-
- CDEBUG(D_CONFIG, "Connected to %s(%s) successfully (%d)\n",
- mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
- 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;
- }
- }
-#endif
- RETURN(0);
+ tgt->ltd_active = 1;
+ tgt->ltd_exp = mdc_exp;
+ lmv->desc.ld_active_tgt_count++;
+
+ md_init_ea_size(tgt->ltd_exp, lmv->max_easize, lmv->max_def_easize,
+ lmv->max_cookiesize, lmv->max_def_cookiesize);
+
+ CDEBUG(D_CONFIG, "Connected to %s(%s) successfully (%d)\n",
+ mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
+ atomic_read(&obd->obd_refcount));
+
+ if (lmv->targets_proc_entry != NULL) {
+ 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->targets_proc_entry,
+ "../../../%s/%s",
+ mdc_obd->obd_type->typ_name,
+ mdc_obd->obd_name);
+ if (mdc_symlink == NULL) {
+ CERROR("cannot register LMV target "
+ "/proc/fs/lustre/%s/%s/target_obds/%s\n",
+ obd->obd_type->typ_name, obd->obd_name,
+ mdc_obd->obd_name);
+ }
+ }
+ RETURN(0);
}
static void lmv_del_target(struct lmv_obd *lmv, int index)
{
struct lmv_obd *lmv = &obd->u.lmv;
struct lmv_tgt_desc *tgt;
+ int orig_tgt_count = 0;
int rc = 0;
ENTRY;
tgt->ltd_uuid = *uuidp;
tgt->ltd_active = 0;
lmv->tgts[index] = tgt;
- if (index >= lmv->desc.ld_tgt_count)
+ if (index >= lmv->desc.ld_tgt_count) {
+ orig_tgt_count = lmv->desc.ld_tgt_count;
lmv->desc.ld_tgt_count = index + 1;
+ }
if (lmv->connected) {
rc = lmv_connect_mdc(obd, tgt);
- if (rc) {
+ if (rc != 0) {
spin_lock(&lmv->lmv_lock);
- lmv->desc.ld_tgt_count--;
+ if (lmv->desc.ld_tgt_count == index + 1)
+ lmv->desc.ld_tgt_count = orig_tgt_count;
memset(tgt, 0, sizeof(*tgt));
spin_unlock(&lmv->lmv_lock);
- } else {
- int easize = sizeof(struct lmv_stripe_md) +
- lmv->desc.ld_tgt_count *
- sizeof(struct lu_fid);
- lmv_init_ea_size(obd->obd_self_export, easize, 0, 0);
- }
- }
+ } else {
+ int easize = sizeof(struct lmv_stripe_md) +
+ lmv->desc.ld_tgt_count * sizeof(struct lu_fid);
+ lmv_init_ea_size(obd->obd_self_export, easize, 0, 0, 0);
+ }
+ }
- lmv_init_unlock(lmv);
- RETURN(rc);
+ lmv_init_unlock(lmv);
+ RETURN(rc);
}
int lmv_check_connect(struct obd_device *obd)
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, 0);
+ lmv_init_unlock(lmv);
+ RETURN(0);
out_disc:
while (i-- > 0) {
static int lmv_disconnect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
{
-#ifdef __KERNEL__
- struct proc_dir_entry *lmv_proc_dir;
-#endif
struct lmv_obd *lmv = &obd->u.lmv;
struct obd_device *mdc_obd;
int rc;
mdc_obd->obd_no_recov = obd->obd_no_recov;
}
-#ifdef __KERNEL__
- lmv_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
- if (lmv_proc_dir) {
- struct proc_dir_entry *mdc_symlink;
+ if (lmv->targets_proc_entry != NULL)
+ lprocfs_remove_proc_entry(mdc_obd->obd_name,
+ lmv->targets_proc_entry);
- 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);
- }
- }
-#endif
rc = obd_fid_fini(tgt->ltd_exp->exp_obd);
if (rc)
CERROR("Can't finanize fids factory\n");
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;
__u32 i;
lmv_disconnect_mdc(obd, lmv->tgts[i]);
}
-#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);
- }
-#endif
+ if (lmv->targets_proc_entry != NULL)
+ lprocfs_remove(&lmv->targets_proc_entry);
+ else
+ CERROR("/proc/fs/lustre/%s/%s/target_obds missing\n",
+ obd->obd_type->typ_name, obd->obd_name);
out_local:
/*
* 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);
- }
- *mds = lum->lum_stripe_offset;
- RETURN(0);
+ lum = op_data->op_data;
+
+ if (le32_to_cpu(lum->lum_stripe_offset) != (__u32)-1) {
+ *mds = le32_to_cpu(lum->lum_stripe_offset);
+ } else {
+ /* -1 means default, which will be in the same MDT with
+ * the stripe */
+ *mds = op_data->op_mds;
+ lum->lum_stripe_offset = cpu_to_le32(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 rc;
ENTRY;
- tgt = lmv_get_target(lmv, mds);
+ tgt = lmv_get_target(lmv, mds, NULL);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
if (tgt->ltd_active == 0 || tgt->ltd_exp == NULL)
GOTO(out, rc = -ENODEV);
- /*
- * Asking underlaying tgt layer to allocate new fid.
- */
- rc = obd_fid_alloc(tgt->ltd_exp, fid, NULL);
- if (rc > 0) {
- LASSERT(fid_is_sane(fid));
- rc = 0;
- }
+ /*
+ * Asking underlying tgt layer to allocate new fid.
+ */
+ rc = obd_fid_alloc(NULL, tgt->ltd_exp, fid, NULL);
+ if (rc > 0) {
+ LASSERT(fid_is_sane(fid));
+ rc = 0;
+ }
EXIT;
out:
return rc;
}
-int lmv_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
- struct md_op_data *op_data)
+int lmv_fid_alloc(const struct lu_env *env, struct obd_export *exp,
+ struct lu_fid *fid, struct md_op_data *op_data)
{
struct obd_device *obd = class_exp2obd(exp);
struct lmv_obd *lmv = &obd->u.lmv;
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);
- lprocfs_alloc_md_stats(obd, 0);
#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_find_cbdata(struct obd_export *exp, const struct lu_fid *fid,
ldlm_iterator_t it, void *data)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- __u32 i;
- int rc;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ int i;
+ int tgt;
+ int rc;
+ ENTRY;
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
+ rc = lmv_check_connect(obd);
+ if (rc)
+ RETURN(rc);
- CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
+ CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
/*
* With DNE every object can have two locks in different namespaces:
* lookup lock in space of MDT storing direntry and update/open lock in
- * space of MDT storing inode.
+ * space of MDT storing inode. Try the MDT that the FID maps to first,
+ * since this can be easily found, and only try others if that fails.
*/
- for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL)
+ for (i = 0, tgt = lmv_find_target_index(lmv, fid);
+ i < lmv->desc.ld_tgt_count;
+ i++, tgt = (tgt + 1) % lmv->desc.ld_tgt_count) {
+ if (tgt < 0) {
+ CDEBUG(D_HA, "%s: "DFID" is inaccessible: rc = %d\n",
+ obd->obd_name, PFID(fid), tgt);
+ tgt = 0;
+ }
+
+ if (lmv->tgts[tgt] == NULL ||
+ lmv->tgts[tgt]->ltd_exp == NULL)
continue;
- rc = md_find_cbdata(lmv->tgts[i]->ltd_exp, fid, it, data);
+
+ rc = md_find_cbdata(lmv->tgts[tgt]->ltd_exp, fid, it, data);
if (rc)
RETURN(rc);
}
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(ERR_CAST(oinfo));
+ *fid = oinfo->lmo_fid;
+ *mds = oinfo->lmo_mds;
+ tgt = lmv_get_target(lmv, *mds, NULL);
+
+ CDEBUG(D_INFO, "locate on mds %u "DFID"\n", *mds, PFID(fid));
+ return tgt;
+}
+
+/**
+ * Locate mds by fid or name
+ *
+ * For striped directory (lsm != NULL), it will locate the stripe
+ * by name hash (see lsm_name_to_stripe_info()). Note: if the hash_type
+ * is unknown, it will return -EBADFD, and lmv_intent_lookup might need
+ * walk through all of stripes to locate the entry.
+ *
+ * For normal direcotry, it will locate MDS by FID directly.
+ * \param[in] lmv LMV device
+ * \param[in] op_data client MD stack parameters, name, namelen
+ * mds_num etc.
+ * \param[in] fid object FID used to locate MDS.
+ *
+ * retval pointer to the lmv_tgt_desc if succeed.
+ * ERR_PTR(errno) if failed.
+ */
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);
- if (IS_ERR(tgt))
- return tgt;
+ if (lsm == NULL || op_data->op_namelen == 0) {
+ tgt = lmv_find_target(lmv, fid);
+ if (IS_ERR(tgt))
+ return tgt;
- op_data->op_mds = tgt->ltd_idx;
+ op_data->op_mds = tgt->ltd_idx;
+ return tgt;
+ }
- 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));
- rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
+ 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(NULL, 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,
- __u64 extra_lock_flags)
-{
- struct ptlrpc_request *req = it->d.lustre.it_data;
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- struct lustre_handle plock;
- struct lmv_tgt_desc *tgt;
- struct md_op_data *rdata;
- struct lu_fid fid1;
- struct mdt_body *body;
- int rc = 0;
- int pmode;
- ENTRY;
-
- body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
- LASSERT(body != NULL);
-
- if (!(body->valid & OBD_MD_MDS))
- RETURN(0);
-
- CDEBUG(D_INODE, "REMOTE_ENQUEUE '%s' on "DFID" -> "DFID"\n",
- LL_IT2STR(it), PFID(&op_data->op_fid1), PFID(&body->fid1));
-
- /*
- * We got LOOKUP lock, but we really need attrs.
- */
- pmode = it->d.lustre.it_lock_mode;
- LASSERT(pmode != 0);
- memcpy(&plock, lockh, sizeof(plock));
- it->d.lustre.it_lock_mode = 0;
- 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);
- if (IS_ERR(tgt))
- GOTO(out, rc = PTR_ERR(tgt));
-
- OBD_ALLOC_PTR(rdata);
- if (rdata == NULL)
- GOTO(out, rc = -ENOMEM);
-
- rdata->op_fid1 = fid1;
- rdata->op_bias = MDS_CROSS_REF;
-
- rc = md_enqueue(tgt->ltd_exp, einfo, it, rdata, lockh,
- lmm, lmmsize, NULL, extra_lock_flags);
- OBD_FREE_PTR(rdata);
- EXIT;
-out:
- ldlm_lock_decref(&plock, pmode);
- return rc;
-}
-
-static int
lmv_enqueue(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,
- struct ptlrpc_request **req, __u64 extra_lock_flags)
+ const union ldlm_policy_data *policy,
+ struct lookup_intent *it, struct md_op_data *op_data,
+ struct lustre_handle *lockh, __u64 extra_lock_flags)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
CDEBUG(D_INODE, "ENQUEUE '%s' on "DFID" -> mds #%d\n",
LL_IT2STR(it), PFID(&op_data->op_fid1), tgt->ltd_idx);
- rc = md_enqueue(tgt->ltd_exp, einfo, it, op_data, lockh,
- lmm, lmmsize, req, extra_lock_flags);
+ rc = md_enqueue(tgt->ltd_exp, einfo, policy, it, op_data, lockh,
+ extra_lock_flags);
- if (rc == 0 && it && it->it_op == IT_OPEN) {
- rc = lmv_enqueue_remote(exp, einfo, it, op_data, lockh,
- lmm, lmmsize, extra_lock_flags);
- }
RETURN(rc);
}
static int
lmv_getattr_name(struct obd_export *exp,struct md_op_data *op_data,
- struct ptlrpc_request **request)
+ struct ptlrpc_request **preq)
{
struct ptlrpc_request *req = NULL;
struct obd_device *obd = exp->exp_obd;
op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
tgt->ltd_idx);
- rc = md_getattr_name(tgt->ltd_exp, op_data, request);
+ rc = md_getattr_name(tgt->ltd_exp, op_data, preq);
if (rc != 0)
RETURN(rc);
- body = req_capsule_server_get(&(*request)->rq_pill,
- &RMF_MDT_BODY);
+ body = req_capsule_server_get(&(*preq)->rq_pill, &RMF_MDT_BODY);
LASSERT(body != NULL);
- if (body->valid & OBD_MD_MDS) {
- struct lu_fid rid = body->fid1;
+ if (body->mbo_valid & OBD_MD_MDS) {
+ struct lu_fid rid = body->mbo_fid1;
CDEBUG(D_INODE, "Request attrs for "DFID"\n",
PFID(&rid));
tgt = lmv_find_target(lmv, &rid);
if (IS_ERR(tgt)) {
- ptlrpc_req_finished(*request);
+ ptlrpc_req_finished(*preq);
+ preq = NULL;
RETURN(PTR_ERR(tgt));
}
op_data->op_namelen = 0;
op_data->op_name = NULL;
rc = md_getattr_name(tgt->ltd_exp, op_data, &req);
- ptlrpc_req_finished(*request);
- *request = req;
+ ptlrpc_req_finished(*preq);
+ *preq = req;
}
RETURN(rc);
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 = current_fsuid();
- op_data->op_fsgid = current_fsgid();
+ op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
+ op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
op_data->op_cap = cfs_curproc_cap_pack();
+ 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 = current_fsuid();
- op_data->op_fsgid = current_fsgid();
+ op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
+ op_data->op_fsgid = from_kgid(&init_user_ns, 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(NULL, 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);
}
-/*
- * 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.
+/**
+ * Get current minimum entry from striped directory
*
- * 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 |
- * '----------------- -----'
+ * This function will search the dir entry, whose hash value is the
+ * closest(>=) to @hash_offset, from all of sub-stripes, and it is
+ * only being called for striped directory.
*
- * 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.):
+ * \param[in] exp export of LMV
+ * \param[in] op_data parameters transferred beween client MD stack
+ * stripe_information will be included in this
+ * parameter
+ * \param[in] cb_op ldlm callback being used in enqueue in
+ * mdc_read_page
+ * \param[in] hash_offset the hash value, which is used to locate
+ * minum(closet) dir entry
+ * \param[in|out] stripe_offset the caller use this to indicate the stripe
+ * index of last entry, so to avoid hash conflict
+ * between stripes. It will also be used to
+ * return the stripe index of current dir entry.
+ * \param[in|out] entp the minum entry and it also is being used
+ * to input the last dir entry to resolve the
+ * hash conflict
*
- * .-------------------- -----.
- * |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 |
+ * \param[out] ppage the page which holds the minum entry
*
- * This structure is transformed into a single logical lu_dirpage as follows:
+ * \retval = 0 get the entry successfully
+ * negative errno (< 0) does not get the entry
+ */
+static int lmv_get_min_striped_entry(struct obd_export *exp,
+ struct md_op_data *op_data,
+ struct md_callback *cb_op,
+ __u64 hash_offset, int *stripe_offset,
+ struct lu_dirent **entp,
+ 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 lmv_tgt_desc *tgt;
+ int stripe_count;
+ struct lu_dirent *min_ent = NULL;
+ struct page *min_page = NULL;
+ int min_idx = 0;
+ int i;
+ int rc = 0;
+ ENTRY;
+
+ stripe_count = lsm->lsm_md_stripe_count;
+ for (i = 0; i < stripe_count; i++) {
+ struct lu_dirent *ent = NULL;
+ struct page *page = NULL;
+ struct lu_dirpage *dp;
+ __u64 stripe_hash = hash_offset;
+
+ tgt = lmv_get_target(lmv, lsm->lsm_md_oinfo[i].lmo_mds, NULL);
+ if (IS_ERR(tgt))
+ GOTO(out, rc = PTR_ERR(tgt));
+
+ /* op_data will be shared by each stripe, so we need
+ * reset these value for each stripe */
+ op_data->op_stripe_offset = i;
+ 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_data = lsm->lsm_md_oinfo[i].lmo_root;
+next:
+ rc = md_read_page(tgt->ltd_exp, op_data, cb_op, stripe_hash,
+ &page);
+ if (rc != 0)
+ GOTO(out, rc);
+
+ dp = page_address(page);
+ for (ent = lu_dirent_start(dp); ent != NULL;
+ ent = lu_dirent_next(ent)) {
+ /* Skip dummy entry */
+ if (le16_to_cpu(ent->lde_namelen) == 0)
+ continue;
+
+ if (le64_to_cpu(ent->lde_hash) < hash_offset)
+ continue;
+
+ if (le64_to_cpu(ent->lde_hash) == hash_offset &&
+ (*entp == ent || i < *stripe_offset))
+ continue;
+
+ /* skip . and .. for other stripes */
+ if (i != 0 &&
+ (strncmp(ent->lde_name, ".",
+ le16_to_cpu(ent->lde_namelen)) == 0 ||
+ strncmp(ent->lde_name, "..",
+ le16_to_cpu(ent->lde_namelen)) == 0))
+ continue;
+ break;
+ }
+
+ if (ent == NULL) {
+ stripe_hash = le64_to_cpu(dp->ldp_hash_end);
+
+ kunmap(page);
+ page_cache_release(page);
+ page = NULL;
+
+ /* reach the end of current stripe, go to next stripe */
+ if (stripe_hash == MDS_DIR_END_OFF)
+ continue;
+ else
+ goto next;
+ }
+
+ if (min_ent != NULL) {
+ if (le64_to_cpu(min_ent->lde_hash) >
+ le64_to_cpu(ent->lde_hash)) {
+ min_ent = ent;
+ kunmap(min_page);
+ page_cache_release(min_page);
+ min_idx = i;
+ min_page = page;
+ } else {
+ kunmap(page);
+ page_cache_release(page);
+ page = NULL;
+ }
+ } else {
+ min_ent = ent;
+ min_page = page;
+ min_idx = i;
+ }
+ }
+
+out:
+ if (*ppage != NULL) {
+ kunmap(*ppage);
+ page_cache_release(*ppage);
+ }
+ *stripe_offset = min_idx;
+ *entp = min_ent;
+ *ppage = min_page;
+ RETURN(rc);
+}
+
+/**
+ * Build dir entry page from a striped directory
*
- * - Replace e0 with e' so the request for the next lu_dirpage gets the page
- * labeled 'next CFS_PAGE'.
+ * This function gets one entry by @offset from a striped directory. It will
+ * read entries from all of stripes, and choose one closest to the required
+ * offset(&offset). A few notes
+ * 1. skip . and .. for non-zero stripes, because there can only have one .
+ * and .. in a directory.
+ * 2. op_data will be shared by all of stripes, instead of allocating new
+ * one, so need to restore before reusing.
+ * 3. release the entry page if that is not being chosen.
*
- * - Copy the LDF_COLLIDE flag from f' to f0 to correctly reflect whether
- * a hash collision with the next page exists.
+ * \param[in] exp obd export refer to LMV
+ * \param[in] op_data hold those MD parameters of read_entry
+ * \param[in] cb_op ldlm callback being used in enqueue in mdc_read_entry
+ * \param[out] ldp the entry being read
+ * \param[out] ppage the page holding the entry. Note: because the entry
+ * will be accessed in upper layer, so we need hold the
+ * page until the usages of entry is finished, see
+ * ll_dir_entry_next.
*
- * - Adjust the lde_reclen of the ending entry of each lu_dirpage to span
- * to the first entry of the next lu_dirpage.
+ * retval =0 if get entry successfully
+ * <0 cannot get entry
*/
-#if PAGE_CACHE_SIZE > LU_PAGE_SIZE
-static void lmv_adjust_dirpages(struct page **pages, int ncfspgs, int nlupgs)
+static int lmv_read_striped_page(struct obd_export *exp,
+ struct md_op_data *op_data,
+ struct md_callback *cb_op,
+ __u64 offset, struct page **ppage)
{
- int i;
+ struct obd_device *obd = exp->exp_obd;
+ struct lu_fid master_fid = op_data->op_fid1;
+ struct inode *master_inode = op_data->op_data;
+ __u64 hash_offset = offset;
+ struct lu_dirpage *dp;
+ struct page *min_ent_page = NULL;
+ struct page *ent_page = NULL;
+ struct lu_dirent *ent;
+ void *area;
+ int ent_idx = 0;
+ struct lu_dirent *min_ent = NULL;
+ struct lu_dirent *last_ent;
+ int left_bytes;
+ int rc;
+ ENTRY;
+
+ rc = lmv_check_connect(obd);
+ if (rc)
+ RETURN(rc);
- 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;
-
- 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);
+ /* Allocate a page and read entries from all of stripes and fill
+ * the page by hash order */
+ ent_page = alloc_page(GFP_KERNEL);
+ if (ent_page == NULL)
+ RETURN(-ENOMEM);
+
+ /* Initialize the entry page */
+ dp = kmap(ent_page);
+ memset(dp, 0, sizeof(*dp));
+ dp->ldp_hash_start = cpu_to_le64(offset);
+ dp->ldp_flags |= LDF_COLLIDE;
+
+ area = dp + 1;
+ left_bytes = PAGE_CACHE_SIZE - sizeof(*dp);
+ ent = area;
+ last_ent = ent;
+ do {
+ __u16 ent_size;
+
+ /* Find the minum entry from all sub-stripes */
+ rc = lmv_get_min_striped_entry(exp, op_data, cb_op, hash_offset,
+ &ent_idx, &min_ent,
+ &min_ent_page);
+ if (rc != 0)
+ GOTO(out, rc);
+
+ /* If it can not get minum entry, it means it already reaches
+ * the end of this directory */
+ if (min_ent == NULL) {
+ last_ent->lde_reclen = 0;
+ hash_offset = MDS_DIR_END_OFF;
+ GOTO(out, rc);
}
- first->ldp_hash_end = hash_end;
- first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
- first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
+ ent_size = le16_to_cpu(min_ent->lde_reclen);
+
+ /* the last entry lde_reclen is 0, but it might not
+ * the end of this entry of this temporay entry */
+ if (ent_size == 0)
+ ent_size = lu_dirent_calc_size(
+ le16_to_cpu(min_ent->lde_namelen),
+ le32_to_cpu(min_ent->lde_attrs));
+ if (ent_size > left_bytes) {
+ last_ent->lde_reclen = cpu_to_le16(0);
+ hash_offset = le64_to_cpu(min_ent->lde_hash);
+ GOTO(out, rc);
+ }
- kunmap(pages[i]);
+ memcpy(ent, min_ent, ent_size);
+
+ /* Replace . with master FID and Replace .. with the parent FID
+ * of master object */
+ if (strncmp(ent->lde_name, ".",
+ le16_to_cpu(ent->lde_namelen)) == 0 &&
+ le16_to_cpu(ent->lde_namelen) == 1)
+ fid_cpu_to_le(&ent->lde_fid, &master_fid);
+ else if (strncmp(ent->lde_name, "..",
+ le16_to_cpu(ent->lde_namelen)) == 0 &&
+ le16_to_cpu(ent->lde_namelen) == 2)
+ fid_cpu_to_le(&ent->lde_fid, &op_data->op_fid3);
+
+ left_bytes -= ent_size;
+ ent->lde_reclen = cpu_to_le16(ent_size);
+ last_ent = ent;
+ ent = (void *)ent + ent_size;
+ hash_offset = le64_to_cpu(min_ent->lde_hash);
+ if (hash_offset == MDS_DIR_END_OFF) {
+ last_ent->lde_reclen = 0;
+ break;
+ }
+ } while (1);
+out:
+ if (min_ent_page != NULL) {
+ kunmap(min_ent_page);
+ page_cache_release(min_ent_page);
}
- LASSERTF(nlupgs == 0, "left = %d", nlupgs);
+
+ if (unlikely(rc != 0)) {
+ __free_page(ent_page);
+ ent_page = NULL;
+ } else {
+ if (ent == area)
+ dp->ldp_flags |= LDF_EMPTY;
+ dp->ldp_flags = cpu_to_le32(dp->ldp_flags);
+ dp->ldp_hash_end = cpu_to_le64(hash_offset);
+ }
+
+ /* We do not want to allocate md_op_data during each
+ * dir entry reading, so op_data will be shared by every stripe,
+ * then we need to restore it back to original value before
+ * return to the upper layer */
+ op_data->op_fid1 = master_fid;
+ op_data->op_fid2 = master_fid;
+ op_data->op_data = master_inode;
+
+ *ppage = ent_page;
+
+ RETURN(rc);
}
-#else
-#define lmv_adjust_dirpages(pages, ncfspgs, nlupgs) do {} while (0)
-#endif /* PAGE_CACHE_SIZE > LU_PAGE_SIZE */
-static int lmv_readpage(struct obd_export *exp, struct md_op_data *op_data,
- struct page **pages, struct ptlrpc_request **request)
+int lmv_read_page(struct obd_export *exp, struct md_op_data *op_data,
+ struct md_callback *cb_op, __u64 offset,
+ struct page **ppage)
{
struct obd_device *obd = exp->exp_obd;
struct lmv_obd *lmv = &obd->u.lmv;
- __u64 offset = op_data->op_offset;
- int rc;
- int ncfspgs; /* pages read in PAGE_CACHE_SIZE */
- int nlupgs; /* pages read in LU_PAGE_SIZE */
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
struct lmv_tgt_desc *tgt;
+ int rc;
ENTRY;
rc = lmv_check_connect(obd);
- if (rc)
+ if (rc != 0)
RETURN(rc);
- CDEBUG(D_INODE, "READPAGE at "LPX64" from "DFID"\n",
- offset, PFID(&op_data->op_fid1));
+ if (unlikely(lsm != NULL)) {
+ rc = lmv_read_striped_page(exp, op_data, cb_op, offset, ppage);
+ RETURN(rc);
+ }
tgt = lmv_find_target(lmv, &op_data->op_fid1);
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
- rc = md_readpage(tgt->ltd_exp, op_data, pages, request);
- if (rc != 0)
- RETURN(rc);
-
- ncfspgs = ((*request)->rq_bulk->bd_nob_transferred +
- PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
- nlupgs = (*request)->rq_bulk->bd_nob_transferred >> LU_PAGE_SHIFT;
- LASSERT(!((*request)->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK));
- LASSERT(ncfspgs > 0 && ncfspgs <= op_data->op_npages);
-
- CDEBUG(D_INODE, "read %d(%d)/%d pages\n", ncfspgs, nlupgs,
- op_data->op_npages);
-
- lmv_adjust_dirpages(pages, ncfspgs, nlupgs);
+ rc = md_read_page(tgt->ltd_exp, op_data, cb_op, offset, ppage);
RETURN(rc);
}
+/**
+ * Unlink a file/directory
+ *
+ * Unlink a file or directory under the parent dir. The unlink request
+ * usually will be sent to the MDT where the child is located, but if
+ * the client does not have the child FID then request will be sent to the
+ * MDT where the parent is located.
+ *
+ * If the parent is a striped directory then it also needs to locate which
+ * stripe the name of the child is located, and replace the parent FID
+ * (@op->op_fid1) with the stripe FID. Note: if the stripe is unknown,
+ * it will walk through all of sub-stripes until the child is being
+ * unlinked finally.
+ *
+ * \param[in] exp export refer to LMV
+ * \param[in] op_data different parameters transferred beween client
+ * MD stacks, name, namelen, FIDs etc.
+ * op_fid1 is the parent FID, op_fid2 is the child
+ * FID.
+ * \param[out] request point to the request of unlink.
+ *
+ * retval 0 if succeed
+ * negative errno if failed.
+ */
static int lmv_unlink(struct obd_export *exp, struct md_op_data *op_data,
struct ptlrpc_request **request)
{
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;
+ int stripe_index = 0;
+ struct lmv_stripe_md *lsm = op_data->op_mea1;
ENTRY;
rc = lmv_check_connect(obd);
if (rc)
RETURN(rc);
-retry:
+retry_unlink:
+ /* For striped dir, we need to locate the parent as well */
+ if (lsm != NULL) {
+ struct lmv_tgt_desc *tmp;
+
+ LASSERT(op_data->op_name != NULL &&
+ op_data->op_namelen != 0);
+
+ tmp = lmv_locate_target_for_name(lmv, lsm,
+ op_data->op_name,
+ op_data->op_namelen,
+ &op_data->op_fid1,
+ &op_data->op_mds);
+
+ /* return -EBADFD means unknown hash type, might
+ * need try all sub-stripe here */
+ if (IS_ERR(tmp) && PTR_ERR(tmp) != -EBADFD)
+ RETURN(PTR_ERR(tmp));
+
+ /* Note: both migrating dir and unknown hash dir need to
+ * try all of sub-stripes, so we need start search the
+ * name from stripe 0, but migrating dir is already handled
+ * inside lmv_locate_target_for_name(), so we only check
+ * unknown hash type directory here */
+ if (!lmv_is_known_hash_type(lsm)) {
+ struct lmv_oinfo *oinfo;
+
+ oinfo = &lsm->lsm_md_oinfo[stripe_index];
+
+ op_data->op_fid1 = oinfo->lmo_fid;
+ op_data->op_mds = oinfo->lmo_mds;
+ }
+ }
+
+try_next_stripe:
/* 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);
+ tgt = lmv_find_target(lmv, &op_data->op_fid2);
+ else if (lsm != NULL)
+ tgt = lmv_get_target(lmv, op_data->op_mds, NULL);
else
tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
+
if (IS_ERR(tgt))
RETURN(PTR_ERR(tgt));
- op_data->op_fsuid = current_fsuid();
- op_data->op_fsgid = current_fsgid();
+ op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
+ op_data->op_fsgid = from_kgid(&init_user_ns, 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);
PFID(&op_data->op_fid1), PFID(&op_data->op_fid2), tgt->ltd_idx);
rc = md_unlink(tgt->ltd_exp, op_data, request);
- if (rc != 0 && rc != -EREMOTE)
+ if (rc != 0 && rc != -EREMOTE && rc != -ENOENT)
RETURN(rc);
+ /* Try next stripe if it is needed. */
+ if (rc == -ENOENT && lsm != NULL && lmv_need_try_all_stripes(lsm)) {
+ struct lmv_oinfo *oinfo;
+
+ stripe_index++;
+ if (stripe_index >= lsm->lsm_md_stripe_count)
+ RETURN(rc);
+
+ oinfo = &lsm->lsm_md_oinfo[stripe_index];
+
+ op_data->op_fid1 = oinfo->lmo_fid;
+ op_data->op_mds = oinfo->lmo_mds;
+
+ ptlrpc_req_finished(*request);
+ *request = NULL;
+
+ goto try_next_stripe;
+ }
+
body = req_capsule_server_get(&(*request)->rq_pill, &RMF_MDT_BODY);
if (body == NULL)
RETURN(-EPROTO);
/* Not cross-ref case, just get out of here. */
- if (likely(!(body->valid & OBD_MD_MDS)))
+ if (likely(!(body->mbo_valid & OBD_MD_MDS)))
RETURN(0);
CDEBUG(D_INODE, "%s: try unlink to another MDT for "DFID"\n",
- exp->exp_obd->obd_name, PFID(&body->fid1));
+ exp->exp_obd->obd_name, PFID(&body->mbo_fid1));
/* This is a remote object, try remote MDT, Note: it may
* try more than 1 time here, Considering following case
*
* In theory, it might try unlimited time here, but it should
* be very rare case. */
- op_data->op_fid2 = body->fid1;
+ op_data->op_fid2 = body->mbo_fid1;
ptlrpc_req_finished(*request);
*request = NULL;
- goto retry;
+ goto retry_unlink;
}
static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
if (!obd_get_info(env, tgt->ltd_exp, keylen, key,
vallen, val, NULL))
RETURN(0);
- }
- RETURN(-EINVAL);
- } else if (KEY_IS(KEY_MAX_EASIZE) || KEY_IS(KEY_CONN_DATA)) {
- rc = lmv_check_connect(obd);
- if (rc)
- RETURN(rc);
+ }
+ RETURN(-EINVAL);
+ } else if (KEY_IS(KEY_MAX_EASIZE) ||
+ KEY_IS(KEY_DEFAULT_EASIZE) ||
+ KEY_IS(KEY_MAX_COOKIESIZE) ||
+ KEY_IS(KEY_DEFAULT_COOKIESIZE) ||
+ KEY_IS(KEY_CONN_DATA)) {
+ rc = lmv_check_connect(obd);
+ if (rc)
+ RETURN(rc);
/*
* Forwarding this request to first MDS, it should know LOV
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;
- __u32 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);
+ }
+
+ /* 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;
+ }
- 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);
+ switch (lsm->lsm_md_magic) {
+ case LMV_MAGIC_V1:
+ rc = lmv_pack_md_v1(lsm, &(*lmmp)->lmv_md_v1);
+ break;
+ default:
+ rc = -EINVAL;
+ break;
+ }
- for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- meap->mea_ids[i] = lsmp->mea_ids[i];
- fid_cpu_to_le(&meap->mea_ids[i], &lsmp->mea_ids[i]);
+ 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)
-{
- 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;
- __u32 i;
- __u32 magic;
- ENTRY;
+static int lmv_unpack_md_v1(struct obd_export *exp, struct lmv_stripe_md *lsm,
+ const struct lmv_mds_md_v1 *lmm1)
+{
+ 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);
+ if (OBD_FAIL_CHECK(OBD_FAIL_UNKNOWN_LMV_STRIPE))
+ lsm->lsm_md_hash_type = LMV_HASH_TYPE_UNKNOWN;
+ else
+ 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);
+}
- LASSERT(mea_size == lmm_size);
+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;
- OBD_ALLOC_LARGE(*tmea, mea_size);
- if (*tmea == NULL)
- RETURN(-ENOMEM);
+ LASSERT(lsmp != NULL);
- if (!lmm)
- RETURN(mea_size);
+ lsm = *lsmp;
+ /* Free memmd */
+ if (lsm != NULL && lmm == NULL) {
+ int i;
+ for (i = 0; i < lsm->lsm_md_stripe_count; i++) {
+ /* For migrating inode, the master stripe and master
+ * object will be the same, so do not need iput, see
+ * ll_update_lsm_md */
+ if (!(lsm->lsm_md_hash_type & LMV_HASH_FLAG_MIGRATION &&
+ i == 0) && lsm->lsm_md_oinfo[i].lmo_root != NULL)
+ iput(lsm->lsm_md_oinfo[i].lmo_root);
+ }
+ lsm_size = lmv_stripe_md_size(lsm->lsm_md_stripe_count);
+ OBD_FREE(lsm, lsm_size);
+ *lsmp = NULL;
+ 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();
- }
+ /* 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);
+ }
- (*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_STRIPE)
+ RETURN(-EPERM);
- 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);
+ /* Unpack memmd */
+ if (le32_to_cpu(lmm->lmv_magic) != LMV_MAGIC_V1 &&
+ 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);
+ }
+
+ if (le32_to_cpu(lmm->lmv_magic) == LMV_MAGIC_V1)
+ 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);
+
+ 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:
+ 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,
ldlm_policy_data_t *policy, ldlm_mode_t mode,
struct lustre_handle *lockh)
{
- struct obd_device *obd = exp->exp_obd;
- struct lmv_obd *lmv = &obd->u.lmv;
- ldlm_mode_t rc;
- __u32 i;
- ENTRY;
+ struct obd_device *obd = exp->exp_obd;
+ struct lmv_obd *lmv = &obd->u.lmv;
+ ldlm_mode_t rc;
+ int tgt;
+ int i;
+ ENTRY;
- CDEBUG(D_INODE, "Lock match for "DFID"\n", PFID(fid));
+ CDEBUG(D_INODE, "Lock match for "DFID"\n", PFID(fid));
/*
- * With CMD every object can have two locks in different namespaces:
- * lookup lock in space of mds storing direntry and update/open lock in
- * space of mds storing inode. Thus we check all targets, not only that
- * one fid was created in.
- */
- for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
- struct lmv_tgt_desc *tgt = lmv->tgts[i];
+ * With DNE every object can have two locks in different namespaces:
+ * lookup lock in space of MDT storing direntry and update/open lock in
+ * space of MDT storing inode. Try the MDT that the FID maps to first,
+ * since this can be easily found, and only try others if that fails.
+ */
+ for (i = 0, tgt = lmv_find_target_index(lmv, fid);
+ i < lmv->desc.ld_tgt_count;
+ i++, tgt = (tgt + 1) % lmv->desc.ld_tgt_count) {
+ if (tgt < 0) {
+ CDEBUG(D_HA, "%s: "DFID" is inaccessible: rc = %d\n",
+ obd->obd_name, PFID(fid), tgt);
+ tgt = 0;
+ }
- if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active)
+ if (lmv->tgts[tgt] == NULL ||
+ lmv->tgts[tgt]->ltd_exp == NULL ||
+ lmv->tgts[tgt]->ltd_active == 0)
continue;
- rc = md_lock_match(tgt->ltd_exp, flags, fid, type, policy, mode,
- lockh);
- if (rc)
- RETURN(rc);
- }
+ rc = md_lock_match(lmv->tgts[tgt]->ltd_exp, flags, fid,
+ type, policy, mode, lockh);
+ if (rc)
+ RETURN(rc);
+ }
- RETURN(0);
+ RETURN(0);
}
int lmv_get_lustre_md(struct obd_export *exp, struct ptlrpc_request *req,
struct obd_export *dt_exp, struct obd_export *md_exp,
struct lustre_md *md)
{
- 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];
if (tgt == NULL || tgt->ltd_exp == NULL)
RETURN(-EINVAL);
- return md_get_lustre_md(tgt->ltd_exp, req, dt_exp, md_exp, md);
+
+ 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 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(tgt->ltd_exp, md));
+ 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,
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));
RETURN(rc);
}
+int lmv_get_fid_from_lsm(struct obd_export *exp,
+ const struct lmv_stripe_md *lsm,
+ const char *name, int namelen, struct lu_fid *fid)
+{
+ const struct lmv_oinfo *oinfo;
+
+ LASSERT(lsm != NULL);
+ oinfo = lsm_name_to_stripe_info(lsm, name, namelen);
+ if (IS_ERR(oinfo))
+ return PTR_ERR(oinfo);
+
+ *fid = oinfo->lmo_fid;
+
+ RETURN(0);
+}
+
/**
* For lmv, only need to send request to master MDT, and the master MDT will
* process with other slave MDTs. The only exception is Q_GETOQUOTA for which
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)
+{
+ 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)
+{
+ 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);
+ }
+ return 0;
+}
+
struct obd_ops lmv_obd_ops = {
.o_owner = THIS_MODULE,
.o_setup = lmv_setup,
.m_setattr = lmv_setattr,
.m_setxattr = lmv_setxattr,
.m_fsync = lmv_fsync,
- .m_readpage = lmv_readpage,
+ .m_read_page = lmv_read_page,
.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,
.m_unpack_capa = lmv_unpack_capa,
.m_get_remote_perm = lmv_get_remote_perm,
.m_intent_getattr_async = lmv_intent_getattr_async,
- .m_revalidate_lock = lmv_revalidate_lock
+ .m_revalidate_lock = lmv_revalidate_lock,
+ .m_get_fid_from_lsm = lmv_get_fid_from_lsm,
};
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, true, NULL,
+#ifndef HAVE_ONLY_PROCFS_SEQ
+ NULL,
+#endif
+ LUSTRE_LMV_NAME, NULL);
}
-#ifdef __KERNEL__
static void lmv_exit(void)
{
class_unregister_type(LUSTRE_LMV_NAME);
module_init(lmv_init);
module_exit(lmv_exit);
-#endif
git://git.whamcloud.com / fs / lustre-release.git / blobdiff