Merge branch 'for-linus' of git://github.com/NewDreamNetwork/ceph-client

* 'for-linus' of git://github.com/NewDreamNetwork/ceph-client:
  libceph: fix pg_temp mapping update
  libceph: fix pg_temp mapping calculation
  libceph: fix linger request requeuing
  libceph: fix parse options memory leak
  libceph: initialize ack_stamp to avoid unnecessary connection reset

4 files changed, 48 insertions, 42 deletions

diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c
index 132963abc266..2883ea01e680 100644
--- a/net/ceph/ceph_common.c
+++ b/net/ceph/ceph_common.c
@@ -232,6 +232,7 @@ void ceph_destroy_options(struct ceph_options *opt)
                 ceph_crypto_key_destroy(opt->key);
                 kfree(opt->key);
         }
+        kfree(opt->mon_addr);
         kfree(opt);
 }
 EXPORT_SYMBOL(ceph_destroy_options);
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index c340e2e0765b..9918e9eb276e 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -2307,6 +2307,7 @@ struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags)
         m->front_max = front_len;
         m->front_is_vmalloc = false;
         m->more_to_follow = false;
+        m->ack_stamp = 0;
         m->pool = NULL;
 
         /* middle */
diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index 16836a7df7a6..88ad8a2501b5 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -217,6 +217,7 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
         INIT_LIST_HEAD(&req->r_unsafe_item);
         INIT_LIST_HEAD(&req->r_linger_item);
         INIT_LIST_HEAD(&req->r_linger_osd);
+        INIT_LIST_HEAD(&req->r_req_lru_item);
         req->r_flags = flags;
 
         WARN_ON((flags & (CEPH_OSD_FLAG_READ|CEPH_OSD_FLAG_WRITE)) == 0);
@@ -816,13 +817,10 @@ static void __register_request(struct ceph_osd_client *osdc,
 {
         req->r_tid = ++osdc->last_tid;
         req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
-        INIT_LIST_HEAD(&req->r_req_lru_item);
-
         dout("__register_request %p tid %lld\n", req, req->r_tid);
         __insert_request(osdc, req);
         ceph_osdc_get_request(req);
         osdc->num_requests++;
-
         if (osdc->num_requests == 1) {
                 dout(" first request, scheduling timeout\n");
                 __schedule_osd_timeout(osdc);
diff --git a/net/ceph/osdmap.c b/net/ceph/osdmap.c
index e97c3588c3ec..fd863fe76934 100644
--- a/net/ceph/osdmap.c
+++ b/net/ceph/osdmap.c
@@ -339,6 +339,7 @@ static int __insert_pg_mapping(struct ceph_pg_mapping *new,
         struct ceph_pg_mapping *pg = NULL;
         int c;
 
+        dout("__insert_pg_mapping %llx %p\n", *(u64 *)&new->pgid, new);
         while (*p) {
                 parent = *p;
                 pg = rb_entry(parent, struct ceph_pg_mapping, node);
@@ -366,16 +367,33 @@ static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
         while (n) {
                 pg = rb_entry(n, struct ceph_pg_mapping, node);
                 c = pgid_cmp(pgid, pg->pgid);
-                if (c < 0)
+                if (c < 0) {
                         n = n->rb_left;
-                else if (c > 0)
+                } else if (c > 0) {
                         n = n->rb_right;
-                else
+                } else {
+                        dout("__lookup_pg_mapping %llx got %p\n",
+                             *(u64 *)&pgid, pg);
                         return pg;
+                }
         }
         return NULL;
 }
 
+static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg pgid)
+{
+        struct ceph_pg_mapping *pg = __lookup_pg_mapping(root, pgid);
+
+        if (pg) {
+                dout("__remove_pg_mapping %llx %p\n", *(u64 *)&pgid, pg);
+                rb_erase(&pg->node, root);
+                kfree(pg);
+                return 0;
+        }
+        dout("__remove_pg_mapping %llx dne\n", *(u64 *)&pgid);
+        return -ENOENT;
+}
+
 /*
  * rbtree of pg pool info
  */
@@ -711,7 +729,6 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
         void *start = *p;
         int err = -EINVAL;
         u16 version;
-        struct rb_node *rbp;
 
         ceph_decode_16_safe(p, end, version, bad);
         if (version > CEPH_OSDMAP_INC_VERSION) {
@@ -861,7 +878,6 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
         }
 
         /* new_pg_temp */
-        rbp = rb_first(&map->pg_temp);
         ceph_decode_32_safe(p, end, len, bad);
         while (len--) {
                 struct ceph_pg_mapping *pg;
@@ -872,18 +888,6 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
                 ceph_decode_copy(p, &pgid, sizeof(pgid));
                 pglen = ceph_decode_32(p);
 
-                /* remove any? */
-                while (rbp && pgid_cmp(rb_entry(rbp, struct ceph_pg_mapping,
-                                                node)->pgid, pgid) <= 0) {
-                        struct ceph_pg_mapping *cur =
-                                rb_entry(rbp, struct ceph_pg_mapping, node);
-
-                        rbp = rb_next(rbp);
-                        dout(" removed pg_temp %llx\n", *(u64 *)&cur->pgid);
-                        rb_erase(&cur->node, &map->pg_temp);
-                        kfree(cur);
-                }
-
                 if (pglen) {
                         /* insert */
                         ceph_decode_need(p, end, pglen*sizeof(u32), bad);
@@ -903,17 +907,11 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
                         }
                         dout(" added pg_temp %llx len %d\n", *(u64 *)&pgid,
                              pglen);
+                } else {
+                        /* remove */
+                        __remove_pg_mapping(&map->pg_temp, pgid);
                 }
         }
-        while (rbp) {
-                struct ceph_pg_mapping *cur =
-                        rb_entry(rbp, struct ceph_pg_mapping, node);
-
-                rbp = rb_next(rbp);
-                dout(" removed pg_temp %llx\n", *(u64 *)&cur->pgid);
-                rb_erase(&cur->node, &map->pg_temp);
-                kfree(cur);
-        }
 
         /* ignore the rest */
         *p = end;
@@ -1046,10 +1044,25 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
         struct ceph_pg_mapping *pg;
         struct ceph_pg_pool_info *pool;
         int ruleno;
-        unsigned poolid, ps, pps;
+        unsigned poolid, ps, pps, t;
         int preferred;
 
+        poolid = le32_to_cpu(pgid.pool);
+        ps = le16_to_cpu(pgid.ps);
+        preferred = (s16)le16_to_cpu(pgid.preferred);
+
+        pool = __lookup_pg_pool(&osdmap->pg_pools, poolid);
+        if (!pool)
+                return NULL;
+
         /* pg_temp? */
+        if (preferred >= 0)
+                t = ceph_stable_mod(ps, le32_to_cpu(pool->v.lpg_num),
+                                    pool->lpgp_num_mask);
+        else
+                t = ceph_stable_mod(ps, le32_to_cpu(pool->v.pg_num),
+                                    pool->pgp_num_mask);
+        pgid.ps = cpu_to_le16(t);
         pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid);
         if (pg) {
                 *num = pg->len;
@@ -1057,18 +1070,6 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
         }
 
         /* crush */
-        poolid = le32_to_cpu(pgid.pool);
-        ps = le16_to_cpu(pgid.ps);
-        preferred = (s16)le16_to_cpu(pgid.preferred);
-
-        /* don't forcefeed bad device ids to crush */
-        if (preferred >= osdmap->max_osd ||
-            preferred >= osdmap->crush->max_devices)
-                preferred = -1;
-
-        pool = __lookup_pg_pool(&osdmap->pg_pools, poolid);
-        if (!pool)
-                return NULL;
         ruleno = crush_find_rule(osdmap->crush, pool->v.crush_ruleset,
                                  pool->v.type, pool->v.size);
         if (ruleno < 0) {
@@ -1078,6 +1079,11 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
                 return NULL;
         }
 
+        /* don't forcefeed bad device ids to crush */
+        if (preferred >= osdmap->max_osd ||
+            preferred >= osdmap->crush->max_devices)
+                preferred = -1;
+
         if (preferred >= 0)
                 pps = ceph_stable_mod(ps,
                                       le32_to_cpu(pool->v.lpgp_num),