/*
* lustre/lod/lod_lov.c
*
- * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
+ * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
*/
#define DEBUG_SUBSYSTEM S_MDS
#include <obd_class.h>
-#include <obd_lov.h>
#include <lustre_lfsck.h>
+#include <lustre_lmv.h>
#include "lod_internal.h"
ltd->ltd_refcount--;
if (ltd->ltd_refcount == 0 && ltd->ltd_death_row) {
struct lod_tgt_desc *tgt_desc, *tmp;
- int idx;
- CFS_LIST_HEAD(kill);
+ struct list_head kill;
+ unsigned int idx;
CDEBUG(D_CONFIG, "destroying %d ltd desc\n",
ltd->ltd_death_row);
+ INIT_LIST_HEAD(&kill);
+
cfs_foreach_bit(ltd->ltd_tgt_bitmap, idx) {
tgt_desc = LTD_TGT(ltd, idx);
LASSERT(tgt_desc);
if (!tgt_desc->ltd_reap)
continue;
- cfs_list_add(&tgt_desc->ltd_kill, &kill);
+ list_add(&tgt_desc->ltd_kill, &kill);
LTD_TGT(ltd, idx) = NULL;
/*FIXME: only support ost pool for now */
if (ltd == &lod->lod_ost_descs) {
mutex_unlock(<d->ltd_mutex);
up_read(<d->ltd_rw_sem);
- cfs_list_for_each_entry_safe(tgt_desc, tmp, &kill, ltd_kill) {
+ list_for_each_entry_safe(tgt_desc, tmp, &kill, ltd_kill) {
int rc;
- cfs_list_del(&tgt_desc->ltd_kill);
+ list_del(&tgt_desc->ltd_kill);
if (ltd == &lod->lod_ost_descs) {
/* remove from QoS structures */
rc = qos_del_tgt(lod, tgt_desc);
int lod_fini_tgt(const struct lu_env *env, struct lod_device *lod,
struct lod_tgt_descs *ltd, bool for_ost)
{
- int idx;
+ unsigned int idx;
if (ltd->ltd_tgts_size <= 0)
return 0;
return(rc);
}
-int lod_ea_store_resize(struct lod_thread_info *info, int size)
+int lod_ea_store_resize(struct lod_thread_info *info, size_t size)
{
- int round = size_roundup_power2(size);
+ __u32 round = size_roundup_power2(size);
LASSERT(round <=
lov_mds_md_size(LOV_MAX_STRIPE_COUNT, LOV_MAGIC_V3));
struct lov_mds_md_v1 *lmm;
struct lov_ost_data_v1 *objs;
__u32 magic;
- int i, rc, lmm_size;
- int cplen = 0;
+ int i, rc;
+ size_t lmm_size;
ENTRY;
LASSERT(lo);
objs = &lmm->lmm_objects[0];
} else {
struct lov_mds_md_v3 *v3 = (struct lov_mds_md_v3 *) lmm;
- cplen = strlcpy(v3->lmm_pool_name, lo->ldo_pool,
+ size_t cplen = strlcpy(v3->lmm_pool_name, lo->ldo_pool,
sizeof(v3->lmm_pool_name));
if (cplen >= sizeof(v3->lmm_pool_name))
RETURN(-E2BIG);
info->lti_buf.lb_len = lmm_size;
rc = dt_xattr_set(env, next, &info->lti_buf, XATTR_NAME_LOV, 0,
th, BYPASS_CAPA);
+ if (rc < 0)
+ lod_object_free_striping(env, lo);
RETURN(rc);
}
v3->lmm_stripe_count = cpu_to_le16(lo->ldo_def_stripenr);
v3->lmm_stripe_offset = cpu_to_le16(lo->ldo_def_stripe_offset);
v3->lmm_stripe_size = cpu_to_le32(lo->ldo_def_stripe_size);
- if (lo->ldo_pool)
- strncpy(v3->lmm_pool_name, lo->ldo_pool,
+ if (lo->ldo_pool != NULL)
+ strlcpy(v3->lmm_pool_name, lo->ldo_pool,
sizeof(v3->lmm_pool_name));
info->lti_buf.lb_buf = v3;
info->lti_buf.lb_len = sizeof(*v3);
RETURN(rc);
}
-static int validate_lod_and_idx(struct lod_device *md, int idx)
+static int validate_lod_and_idx(struct lod_device *md, __u32 idx)
{
if (unlikely(idx >= md->lod_ost_descs.ltd_tgts_size ||
!cfs_bitmap_check(md->lod_ost_bitmap, idx))) {
struct lu_device *nd;
struct dt_object **stripe;
int stripe_len;
- int i, idx, rc = 0;
+ int i, rc = 0;
+ __u32 idx;
ENTRY;
LASSERT(lo != NULL);
RETURN(-ENOMEM);
for (i = 0; i < lo->ldo_stripenr; i++) {
+ if (unlikely(lovea_slot_is_dummy(&objs[i])))
+ continue;
+
ostid_le_to_cpu(&objs[i].l_ost_oi, &info->lti_ostid);
idx = le32_to_cpu(objs[i].l_ost_idx);
rc = ostid_to_fid(&info->lti_fid, &info->lti_ostid, idx);
lu_object_put(env, &stripe[i]->do_lu);
OBD_FREE(stripe, sizeof(stripe[0]) * stripe_len);
+ lo->ldo_stripenr = 0;
} else {
lo->ldo_stripe = stripe;
lo->ldo_stripes_allocated = stripe_len;
RETURN(rc);
}
-/*
- * Load and parse striping information, create in-core representation for the
- * stripes
- */
-int lod_load_striping(const struct lu_env *env, struct lod_object *lo)
+int lod_load_striping_locked(const struct lu_env *env, struct lod_object *lo)
{
struct lod_thread_info *info = lod_env_info(env);
+ struct lu_buf *buf = &info->lti_buf;
struct dt_object *next = dt_object_child(&lo->ldo_obj);
int rc = 0;
ENTRY;
- /*
- * currently this code is supposed to be called from declaration
- * phase only, thus the object is not expected to be locked by caller
- */
- dt_write_lock(env, next, 0);
/* already initialized? */
if (lo->ldo_stripe != NULL)
GOTO(out, rc = 0);
+ if (!dt_object_exists(next))
+ GOTO(out, rc = 0);
+
/* Do not load stripe for slaves of striped dir */
- if (!dt_object_exists(next) || lo->ldo_dir_slave_stripe)
+ if (lo->ldo_dir_slave_stripe)
GOTO(out, rc = 0);
- /* only regular files can be striped */
- if (lu_object_attr(lod2lu_obj(lo)) & S_IFREG) {
+ if (S_ISREG(lu_object_attr(lod2lu_obj(lo)))) {
rc = lod_get_lov_ea(env, lo);
if (rc <= 0)
GOTO(out, rc);
* there is LOV EA (striping information) in this object
* let's parse it and create in-core objects for the stripes
*/
- info->lti_buf.lb_buf = info->lti_ea_store;
- info->lti_buf.lb_len = info->lti_ea_store_size;
- rc = lod_parse_striping(env, lo, &info->lti_buf);
- } else if (lu_object_attr(lod2lu_obj(lo)) & S_IFDIR) {
+ buf->lb_buf = info->lti_ea_store;
+ buf->lb_len = info->lti_ea_store_size;
+ rc = lod_parse_striping(env, lo, buf);
+ } else if (S_ISDIR(lu_object_attr(lod2lu_obj(lo)))) {
rc = lod_get_lmv_ea(env, lo);
- if (rc <= 0)
- GOTO(out, rc);
+ if (rc < (typeof(rc))sizeof(struct lmv_mds_md_v1))
+ GOTO(out, rc = rc > 0 ? -EINVAL : rc);
+
+ buf->lb_buf = info->lti_ea_store;
+ buf->lb_len = info->lti_ea_store_size;
+ if (rc == sizeof(struct lmv_mds_md_v1)) {
+ rc = lod_load_lmv_shards(env, lo, buf, true);
+ if (buf->lb_buf != info->lti_ea_store) {
+ OBD_FREE_LARGE(info->lti_ea_store,
+ info->lti_ea_store_size);
+ info->lti_ea_store = buf->lb_buf;
+ info->lti_ea_store_size = buf->lb_len;
+ }
+
+ if (rc < 0)
+ GOTO(out, rc);
+ }
+
/*
* there is LOV EA (striping information) in this object
* let's parse it and create in-core objects for the stripes
*/
- info->lti_buf.lb_buf = info->lti_ea_store;
- info->lti_buf.lb_len = info->lti_ea_store_size;
- rc = lod_parse_dir_striping(env, lo, &info->lti_buf);
+ rc = lod_parse_dir_striping(env, lo, buf);
}
out:
- dt_write_unlock(env, next);
RETURN(rc);
}
+/**
+ * Load and parse striping information, create in-core representation for the
+ * stripes
+ **/
+int lod_load_striping(const struct lu_env *env, struct lod_object *lo)
+{
+ struct dt_object *next = dt_object_child(&lo->ldo_obj);
+ int rc = 0;
+
+ /* currently this code is supposed to be called from declaration
+ * phase only, thus the object is not expected to be locked by caller */
+ dt_write_lock(env, next, 0);
+ rc = lod_load_striping_locked(env, lo);
+ dt_write_unlock(env, next);
+ return rc;
+}
+
+/* verify the striping information for directory */
int lod_verify_striping(struct lod_device *d, const struct lu_buf *buf,
- int specific)
+ bool is_from_disk)
{
struct lov_user_md_v1 *lum;
struct lov_user_md_v3 *lum3;
GOTO(out, rc = -EINVAL);
}
- if ((specific && le32_to_cpu(lum->lmm_pattern) != LOV_PATTERN_RAID0) ||
- (!specific && lum->lmm_pattern != 0)) {
+ /* the user uses "0" for default stripe pattern normally. */
+ if (!is_from_disk && lum->lmm_pattern == 0)
+ lum->lmm_pattern = cpu_to_le32(LOV_PATTERN_RAID0);
+
+ if (le32_to_cpu(lum->lmm_pattern) != LOV_PATTERN_RAID0) {
CDEBUG(D_IOCTL, "bad userland stripe pattern: %#x\n",
le32_to_cpu(lum->lmm_pattern));
GOTO(out, rc = -EINVAL);
}
}
- stripe_count = le16_to_cpu(lum->lmm_stripe_count);
if (magic == LOV_USER_MAGIC_V1 || magic == LOV_MAGIC_V1_DEF)
lum_size = offsetof(struct lov_user_md_v1,
- lmm_objects[stripe_count]);
+ lmm_objects[0]);
else if (magic == LOV_USER_MAGIC_V3 || magic == LOV_MAGIC_V3_DEF)
lum_size = offsetof(struct lov_user_md_v3,
- lmm_objects[stripe_count]);
+ lmm_objects[0]);
else
- LBUG();
+ GOTO(out, rc = -EINVAL);
- if (specific && buf->lb_len != lum_size) {
+ stripe_count = le16_to_cpu(lum->lmm_stripe_count);
+ if (buf->lb_len != lum_size) {
CDEBUG(D_IOCTL, "invalid buf len %zd for lov_user_md with "
"magic %#x and stripe_count %u\n",
buf->lb_len, magic, stripe_count);
GOTO(out, rc = -EINVAL);
}
- if (specific && stripe_count > pool_tgt_count(pool)) {
+ if (is_from_disk && stripe_count > pool_tgt_count(pool)) {
CDEBUG(D_IOCTL,
"stripe count %u > # OSTs %u in the pool\n",
stripe_count, pool_tgt_count(pool));
if (*val != 0)
LCONSOLE_INFO("Increasing default stripe size to "
"minimum value %u\n",
- LOV_DEFAULT_STRIPE_SIZE);
- *val = LOV_DEFAULT_STRIPE_SIZE;
+ LOV_DESC_STRIPE_SIZE_DEFAULT);
+ *val = LOV_DESC_STRIPE_SIZE_DEFAULT;
} else if (*val & (LOV_MIN_STRIPE_SIZE - 1)) {
*val &= ~(LOV_MIN_STRIPE_SIZE - 1);
LCONSOLE_WARN("Changing default stripe size to "LPU64" (a "
{
/* fix qos_maxage */
if (*val == 0)
- *val = QOS_DEFAULT_MAXAGE;
+ *val = LOV_DESC_QOS_MAXAGE_DEFAULT;
}
void lod_fix_desc(struct lov_desc *desc)
lod->lod_sp_me = LUSTRE_SP_CLI;
/* Set up allocation policy (QoS and RR) */
- CFS_INIT_LIST_HEAD(&lod->lod_qos.lq_oss_list);
+ INIT_LIST_HEAD(&lod->lod_qos.lq_oss_list);
init_rwsem(&lod->lod_qos.lq_rw_sem);
lod->lod_qos.lq_dirty = 1;
lod->lod_qos.lq_rr.lqr_dirty = 1;
if (lod->lod_pools_hash_body == NULL)
RETURN(-ENOMEM);
- CFS_INIT_LIST_HEAD(&lod->lod_pool_list);
+ INIT_LIST_HEAD(&lod->lod_pool_list);
lod->lod_pool_count = 0;
rc = lod_ost_pool_init(&lod->lod_pool_info, 0);
if (rc)
int lod_pools_fini(struct lod_device *lod)
{
struct obd_device *obd = lod2obd(lod);
- cfs_list_t *pos, *tmp;
- struct pool_desc *pool;
+ struct pool_desc *pool, *tmp;
ENTRY;
- cfs_list_for_each_safe(pos, tmp, &lod->lod_pool_list) {
- pool = cfs_list_entry(pos, struct pool_desc, pool_list);
+ list_for_each_entry_safe(pool, tmp, &lod->lod_pool_list, pool_list) {
/* free pool structs */
CDEBUG(D_INFO, "delete pool %p\n", pool);
/* In the function below, .hs_keycmp resolves to
git://git.whamcloud.com / fs / lustre-release.git / blobdiff