LU-2430 mdd: add lfs mv to migrate inode.

[fs/lustre-release.git] / lustre / lod / lod_qos.c

diff --git a/lustre/lod/lod_qos.c b/lustre/lod/lod_qos.c
index 29c582a..b4b8d93 100644 (file)
--- a/lustre/lod/lod_qos.c
+++ b/lustre/lod/lod_qos.c
@@ -23,7 +23,7 @@
  * Copyright  2009 Sun Microsystems, Inc. All rights reserved
  * Use is subject to license terms.
  *
- * Copyright (c) 2012, Intel Corporation.
+ * Copyright (c) 2012, 2013, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -61,7 +61,7 @@
 
 int qos_add_tgt(struct lod_device *lod, struct lod_tgt_desc *ost_desc)
 {
-       struct lov_qos_oss *oss = NULL, *temposs;
+       struct lod_qos_oss *oss = NULL, *temposs;
        struct obd_export  *exp = ost_desc->ltd_exp;
        int                 rc = 0, found = 0;
        cfs_list_t         *list;
@@ -120,7 +120,7 @@ out:
 
 int qos_del_tgt(struct lod_device *lod, struct lod_tgt_desc *ost_desc)
 {
-       struct lov_qos_oss *oss;
+       struct lod_qos_oss *oss;
        int                 rc = 0;
        ENTRY;
 
@@ -159,6 +159,10 @@ static int lod_statfs_and_check(const struct lu_env *env, struct lod_device *d,
        if (rc && rc != -ENOTCONN)
                CERROR("%s: statfs: rc = %d\n", lod2obd(d)->obd_name, rc);
 
+       /* If the OST is readonly then we can't allocate objects there */
+       if (sfs->os_state & OS_STATE_READONLY)
+               rc = -EROFS;
+
        /* check whether device has changed state (active, inactive) */
        if (rc != 0 && ost->ltd_active) {
                /* turned inactive? */
@@ -234,7 +238,7 @@ out:
    depends on size of each ost in an oss */
 static int lod_qos_calc_ppo(struct lod_device *lod)
 {
-       struct lov_qos_oss *oss;
+       struct lod_qos_oss *oss;
        __u64               ba_max, ba_min, temp;
        __u32               num_active;
        int                 rc, i, prio_wide;
@@ -367,7 +371,7 @@ static int lod_qos_used(struct lod_device *lod, struct ost_pool *osts,
                        __u32 index, __u64 *total_wt)
 {
        struct lod_tgt_desc *ost;
-       struct lov_qos_oss  *oss;
+       struct lod_qos_oss  *oss;
        int j;
        ENTRY;
 
@@ -443,9 +447,9 @@ static int lod_qos_used(struct lod_device *lod, struct ost_pool *osts,
 #define LOV_QOS_EMPTY ((__u32)-1)
 /* compute optimal round-robin order, based on OSTs per OSS */
 static int lod_qos_calc_rr(struct lod_device *lod, struct ost_pool *src_pool,
-                          struct lov_qos_rr *lqr)
+                          struct lod_qos_rr *lqr)
 {
-       struct lov_qos_oss  *oss;
+       struct lod_qos_oss  *oss;
        struct lod_tgt_desc *ost;
        unsigned placed, real_count;
        int i, rc;
@@ -620,8 +624,6 @@ static int inline lod_qos_dev_is_full(struct obd_statfs *msfs)
        return (msfs->os_bavail < used);
 }
 
-int lod_ea_store_resize(struct lod_thread_info *info, int size);
-
 static inline int lod_qos_ost_in_use_clear(const struct lu_env *env, int stripes)
 {
        struct lod_thread_info *info = lod_env_info(env);
@@ -632,7 +634,7 @@ static inline int lod_qos_ost_in_use_clear(const struct lu_env *env, int stripes
                CERROR("can't allocate memory for ost-in-use array\n");
                return -ENOMEM;
        }
-       memset(info->lti_ea_store, 0, sizeof(int) * stripes);
+       memset(info->lti_ea_store, -1, sizeof(int) * stripes);
        return 0;
 }
 
@@ -667,7 +669,7 @@ static int lod_alloc_rr(const struct lu_env *env, struct lod_object *lo,
        struct obd_statfs *sfs = &lod_env_info(env)->lti_osfs;
        struct pool_desc  *pool = NULL;
        struct ost_pool   *osts;
-       struct lov_qos_rr *lqr;
+       struct lod_qos_rr *lqr;
        struct dt_object  *o;
        unsigned           array_idx;
        int                i, rc;
@@ -771,7 +773,7 @@ repeat_find:
                /*
                 * try to use another OSP if this one is degraded
                 */
-               if (sfs->os_state == OS_STATE_DEGRADED && speed < 2) {
+               if (sfs->os_state & OS_STATE_DEGRADED && speed < 2) {
                        QOS_DEBUG("#%d: degraded\n", ost_idx);
                        continue;
                }
@@ -921,6 +923,7 @@ repeat_find:
                /*
                 * We've successfuly declared (reserved) an object
                 */
+               lod_qos_ost_in_use(env, stripe_num, ost_idx);
                stripe[stripe_num] = o;
                stripe_num++;
 
@@ -1210,41 +1213,40 @@ static int lod_use_defined_striping(const struct lu_env *env,
                                    struct lod_object *mo,
                                    const struct lu_buf *buf)
 {
-       struct lod_device      *d = lu2lod_dev(lod2lu_obj(mo)->lo_dev);
        struct lov_mds_md_v1   *v1 = buf->lb_buf;
        struct lov_mds_md_v3   *v3 = buf->lb_buf;
        struct lov_ost_data_v1 *objs;
        __u32                   magic;
-       int                     rc;
+       int                     rc = 0;
        ENTRY;
 
-       rc = lod_verify_striping(d, buf, 1);
-       if (rc)
-               RETURN(rc);
-
        magic = le32_to_cpu(v1->lmm_magic);
        if (magic == LOV_MAGIC_V1_DEF) {
+               magic = LOV_MAGIC_V1;
                objs = &v1->lmm_objects[0];
        } else if (magic == LOV_MAGIC_V3_DEF) {
+               magic = LOV_MAGIC_V3;
                objs = &v3->lmm_objects[0];
                lod_object_set_pool(mo, v3->lmm_pool_name);
        } else {
                GOTO(out, rc = -EINVAL);
        }
 
-       /*
-        * LOD shouldn't be aware of recovery at all,
-        * but this track recovery status (to some extent)
-        * to be do additional checks like this one
-        */
-       LASSERT(d->lod_recovery_completed == 0);
-
+       mo->ldo_pattern = le32_to_cpu(v1->lmm_pattern);
        mo->ldo_stripe_size = le32_to_cpu(v1->lmm_stripe_size);
        mo->ldo_stripenr = le16_to_cpu(v1->lmm_stripe_count);
        mo->ldo_layout_gen = le16_to_cpu(v1->lmm_layout_gen);
+
+       /* fixup for released file before object initialization */
+       if (mo->ldo_pattern & LOV_PATTERN_F_RELEASED) {
+               mo->ldo_released_stripenr = mo->ldo_stripenr;
+               mo->ldo_stripenr = 0;
+       }
+
        LASSERT(buf->lb_len >= lov_mds_md_size(mo->ldo_stripenr, magic));
 
-       rc = lod_initialize_objects(env, mo, objs);
+       if (mo->ldo_stripenr > 0)
+               rc = lod_initialize_objects(env, mo, objs);
 
 out:
        RETURN(rc);
@@ -1288,10 +1290,14 @@ static int lod_qos_parse_config(const struct lu_env *env,
                RETURN(-EINVAL);
        }
 
-       if (v1->lmm_pattern != 0 && v1->lmm_pattern != LOV_PATTERN_RAID0) {
+       v1->lmm_magic = magic;
+       if (v1->lmm_pattern == 0)
+               v1->lmm_pattern = LOV_PATTERN_RAID0;
+       if (lov_pattern(v1->lmm_pattern) != LOV_PATTERN_RAID0) {
                CERROR("invalid pattern: %x\n", v1->lmm_pattern);
                RETURN(-EINVAL);
        }
+       lo->ldo_pattern = v1->lmm_pattern;
 
        if (v1->lmm_stripe_size)
                lo->ldo_stripe_size = v1->lmm_stripe_size;
@@ -1345,6 +1351,12 @@ static int lod_qos_parse_config(const struct lu_env *env,
        } else
                lod_object_set_pool(lo, NULL);
 
+       /* fixup for released file */
+       if (lo->ldo_pattern & LOV_PATTERN_F_RELEASED) {
+               lo->ldo_released_stripenr = lo->ldo_stripenr;
+               lo->ldo_stripenr = 0;
+       }
+
        RETURN(0);
 }
 
@@ -1382,6 +1394,10 @@ int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo,
        if (rc)
                GOTO(out, rc);
 
+       /* A released file is being created */
+       if (lo->ldo_stripenr == 0)
+               GOTO(out, rc = 0);
+
        if (likely(lo->ldo_stripe == NULL)) {
                /*
                 * no striping has been created so far