libceph: decode into cpu-native ceph_pg type
Sage Weil [Sat, 23 Feb 2013 18:38:16 +0000 (10:38 -0800)]
Always decode data into our cpu-native ceph_pg type that has the correct
field widths.  Limit any remaining uses of ceph_pg_v1 to dealing with the
legacy protocol.

Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

fs/ceph/ioctl.c
include/linux/ceph/osd_client.h
include/linux/ceph/osdmap.h
net/ceph/debugfs.c
net/ceph/osd_client.c
net/ceph/osdmap.c

index e831436..fb036ed 100644 (file)
@@ -186,7 +186,7 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg)
        u64 len = 1, olen;
        u64 tmp;
        struct ceph_object_layout ol;
-       struct ceph_pg_v1 pgid;
+       struct ceph_pg pgid;
        int r;
 
        /* copy and validate */
@@ -212,7 +212,8 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg)
        ceph_calc_object_layout(&ol, dl.object_name, &ci->i_layout,
                                osdc->osdmap);
 
-       pgid = ol.ol_pgid;
+       pgid.pool = le32_to_cpu(ol.ol_pgid.pool);
+       pgid.seed = le16_to_cpu(ol.ol_pgid.ps);
        dl.osd = ceph_calc_pg_primary(osdc->osdmap, pgid);
        if (dl.osd >= 0) {
                struct ceph_entity_addr *a =
index be28673..388158f 100644 (file)
@@ -56,7 +56,7 @@ struct ceph_osd_request {
        struct list_head r_linger_item;
        struct list_head r_linger_osd;
        struct ceph_osd *r_osd;
-       struct ceph_pg_v1 r_pgid;
+       struct ceph_pg   r_pgid;
        int              r_pg_osds[CEPH_PG_MAX_SIZE];
        int              r_num_pg_osds;
 
index eb4989a..8a612df 100644 (file)
  * The map can be updated either via an incremental map (diff) describing
  * the change between two successive epochs, or as a fully encoded map.
  */
+struct ceph_pg {
+       uint64_t pool;
+       uint32_t seed;
+};
+
 struct ceph_pg_pool_info {
        struct rb_node node;
        int id;
@@ -28,7 +33,7 @@ struct ceph_pg_pool_info {
 
 struct ceph_pg_mapping {
        struct rb_node node;
-       struct ceph_pg_v1 pgid;
+       struct ceph_pg pgid;
        int len;
        int osds[];
 };
@@ -119,10 +124,10 @@ extern int ceph_calc_object_layout(struct ceph_object_layout *ol,
                                   struct ceph_file_layout *fl,
                                   struct ceph_osdmap *osdmap);
 extern int ceph_calc_pg_acting(struct ceph_osdmap *osdmap,
-                              struct ceph_pg_v1 pgid,
+                              struct ceph_pg pgid,
                               int *acting);
 extern int ceph_calc_pg_primary(struct ceph_osdmap *osdmap,
-                               struct ceph_pg_v1 pgid);
+                               struct ceph_pg pgid);
 
 extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
 extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);
index 38b5dc1..61a9af6 100644 (file)
@@ -131,10 +131,9 @@ static int osdc_show(struct seq_file *s, void *pp)
 
                req = rb_entry(p, struct ceph_osd_request, r_node);
 
-               seq_printf(s, "%lld\tosd%d\t%d.%x\t", req->r_tid,
+               seq_printf(s, "%lld\tosd%d\t%lld.%x\t", req->r_tid,
                           req->r_osd ? req->r_osd->o_osd : -1,
-                          le32_to_cpu(req->r_pgid.pool),
-                          le16_to_cpu(req->r_pgid.ps));
+                          req->r_pgid.pool, req->r_pgid.seed);
 
                head = req->r_request->front.iov_base;
                op = (void *)(head + 1);
index e3ab8d6..1990834 100644 (file)
@@ -914,7 +914,7 @@ static int __map_request(struct ceph_osd_client *osdc,
                         struct ceph_osd_request *req, int force_resend)
 {
        struct ceph_osd_request_head *reqhead = req->r_request->front.iov_base;
-       struct ceph_pg_v1 pgid;
+       struct ceph_pg pgid;
        int acting[CEPH_PG_MAX_SIZE];
        int o = -1, num = 0;
        int err;
@@ -926,7 +926,8 @@ static int __map_request(struct ceph_osd_client *osdc,
                list_move(&req->r_req_lru_item, &osdc->req_notarget);
                return err;
        }
-       pgid = reqhead->layout.ol_pgid;
+       pgid.pool = le32_to_cpu(reqhead->layout.ol_pgid.pool);
+       pgid.seed = le16_to_cpu(reqhead->layout.ol_pgid.ps);
        req->r_pgid = pgid;
 
        err = ceph_calc_pg_acting(osdc->osdmap, pgid, acting);
@@ -943,8 +944,8 @@ static int __map_request(struct ceph_osd_client *osdc,
            (req->r_osd == NULL && o == -1))
                return 0;  /* no change */
 
-       dout("map_request tid %llu pgid %d.%x osd%d (was osd%d)\n",
-            req->r_tid, le32_to_cpu(pgid.pool), le16_to_cpu(pgid.ps), o,
+       dout("map_request tid %llu pgid %lld.%x osd%d (was osd%d)\n",
+            req->r_tid, pgid.pool, pgid.seed, o,
             req->r_osd ? req->r_osd->o_osd : -1);
 
        /* record full pg acting set */
index 8c89ac2..81118db 100644 (file)
@@ -350,14 +350,15 @@ bad:
  * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid
  * to a set of osds)
  */
-static int pgid_cmp(struct ceph_pg_v1 l, struct ceph_pg_v1 r)
+static int pgid_cmp(struct ceph_pg l, struct ceph_pg r)
 {
-       u64 a = *(u64 *)&l;
-       u64 b = *(u64 *)&r;
-
-       if (a < b)
+       if (l.pool < r.pool)
+               return -1;
+       if (l.pool > r.pool)
+               return 1;
+       if (l.seed < r.seed)
                return -1;
-       if (a > b)
+       if (l.seed > r.seed)
                return 1;
        return 0;
 }
@@ -389,7 +390,7 @@ static int __insert_pg_mapping(struct ceph_pg_mapping *new,
 }
 
 static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
-                                                  struct ceph_pg_v1 pgid)
+                                                  struct ceph_pg pgid)
 {
        struct rb_node *n = root->rb_node;
        struct ceph_pg_mapping *pg;
@@ -403,25 +404,26 @@ static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
                } else if (c > 0) {
                        n = n->rb_right;
                } else {
-                       dout("__lookup_pg_mapping %llx got %p\n",
-                            *(u64 *)&pgid, pg);
+                       dout("__lookup_pg_mapping %lld.%x got %p\n",
+                            pgid.pool, pgid.seed, pg);
                        return pg;
                }
        }
        return NULL;
 }
 
-static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg_v1 pgid)
+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);
+               dout("__remove_pg_mapping %lld.%x %p\n", pgid.pool, pgid.seed,
+                    pg);
                rb_erase(&pg->node, root);
                kfree(pg);
                return 0;
        }
-       dout("__remove_pg_mapping %llx dne\n", *(u64 *)&pgid);
+       dout("__remove_pg_mapping %lld.%x dne\n", pgid.pool, pgid.seed);
        return -ENOENT;
 }
 
@@ -721,11 +723,14 @@ struct ceph_osdmap *osdmap_decode(void **p, void *end)
        ceph_decode_32_safe(p, end, len, bad);
        for (i = 0; i < len; i++) {
                int n, j;
-               struct ceph_pg_v1 pgid;
+               struct ceph_pg pgid;
+               struct ceph_pg_v1 pgid_v1;
                struct ceph_pg_mapping *pg;
 
                ceph_decode_need(p, end, sizeof(u32) + sizeof(u64), bad);
-               ceph_decode_copy(p, &pgid, sizeof(pgid));
+               ceph_decode_copy(p, &pgid_v1, sizeof(pgid_v1));
+               pgid.pool = le32_to_cpu(pgid_v1.pool);
+               pgid.seed = le16_to_cpu(pgid_v1.ps);
                n = ceph_decode_32(p);
                err = -EINVAL;
                if (n > (UINT_MAX - sizeof(*pg)) / sizeof(u32))
@@ -743,7 +748,8 @@ struct ceph_osdmap *osdmap_decode(void **p, void *end)
                err = __insert_pg_mapping(pg, &map->pg_temp);
                if (err)
                        goto bad;
-               dout(" added pg_temp %llx len %d\n", *(u64 *)&pgid, len);
+               dout(" added pg_temp %lld.%x len %d\n", pgid.pool, pgid.seed,
+                    len);
        }
 
        /* crush */
@@ -944,10 +950,13 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
        while (len--) {
                struct ceph_pg_mapping *pg;
                int j;
-               struct ceph_pg_v1 pgid;
+               struct ceph_pg_v1 pgid_v1;
+               struct ceph_pg pgid;
                u32 pglen;
                ceph_decode_need(p, end, sizeof(u64) + sizeof(u32), bad);
-               ceph_decode_copy(p, &pgid, sizeof(pgid));
+               ceph_decode_copy(p, &pgid_v1, sizeof(pgid_v1));
+               pgid.pool = le32_to_cpu(pgid_v1.pool);
+               pgid.seed = le16_to_cpu(pgid_v1.ps);
                pglen = ceph_decode_32(p);
 
                if (pglen) {
@@ -973,8 +982,8 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
                                kfree(pg);
                                goto bad;
                        }
-                       dout(" added pg_temp %llx len %d\n", *(u64 *)&pgid,
-                            pglen);
+                       dout(" added pg_temp %lld.%x len %d\n", pgid.pool,
+                            pgid.seed, pglen);
                } else {
                        /* remove */
                        __remove_pg_mapping(&map->pg_temp, pgid);
@@ -1079,26 +1088,25 @@ int ceph_calc_object_layout(struct ceph_object_layout *ol,
                            struct ceph_osdmap *osdmap)
 {
        unsigned int num, num_mask;
-       struct ceph_pg_v1 pgid;
-       int poolid = le32_to_cpu(fl->fl_pg_pool);
+       struct ceph_pg pgid;
        struct ceph_pg_pool_info *pool;
-       unsigned int ps;
 
        BUG_ON(!osdmap);
 
-       pool = __lookup_pg_pool(&osdmap->pg_pools, poolid);
+       pgid.pool = le32_to_cpu(fl->fl_pg_pool);
+       pool = __lookup_pg_pool(&osdmap->pg_pools, pgid.pool);
        if (!pool)
                return -EIO;
-       ps = ceph_str_hash(pool->v.object_hash, oid, strlen(oid));
+       pgid.seed = ceph_str_hash(pool->v.object_hash, oid, strlen(oid));
        num = le32_to_cpu(pool->v.pg_num);
        num_mask = pool->pg_num_mask;
 
-       pgid.ps = cpu_to_le16(ps);
-       pgid.preferred = cpu_to_le16(-1);
-       pgid.pool = fl->fl_pg_pool;
-       dout("calc_object_layout '%s' pgid %d.%x\n", oid, poolid, ps);
+       dout("calc_object_layout '%s' pgid %lld.%x\n", oid, pgid.pool,
+            pgid.seed);
 
-       ol->ol_pgid = pgid;
+       ol->ol_pgid.ps = cpu_to_le16(pgid.seed);
+       ol->ol_pgid.pool = fl->fl_pg_pool;
+       ol->ol_pgid.preferred = cpu_to_le16(-1);
        ol->ol_stripe_unit = fl->fl_object_stripe_unit;
        return 0;
 }
@@ -1108,7 +1116,7 @@ EXPORT_SYMBOL(ceph_calc_object_layout);
  * Calculate raw osd vector for the given pgid.  Return pointer to osd
  * array, or NULL on failure.
  */
-static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
+static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
                        int *osds, int *num)
 {
        struct ceph_pg_mapping *pg;
@@ -1116,8 +1124,8 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
        int ruleno;
        unsigned int poolid, ps, pps, t, r;
 
-       poolid = le32_to_cpu(pgid.pool);
-       ps = le16_to_cpu(pgid.ps);
+       poolid = pgid.pool;
+       ps = pgid.seed;
 
        pool = __lookup_pg_pool(&osdmap->pg_pools, poolid);
        if (!pool)
@@ -1126,7 +1134,7 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
        /* pg_temp? */
        t = ceph_stable_mod(ps, le32_to_cpu(pool->v.pg_num),
                            pool->pgp_num_mask);
-       pgid.ps = cpu_to_le16(t);
+       pgid.seed = t;
        pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid);
        if (pg) {
                *num = pg->len;
@@ -1163,7 +1171,7 @@ static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
 /*
  * Return acting set for given pgid.
  */
-int ceph_calc_pg_acting(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
+int ceph_calc_pg_acting(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
                        int *acting)
 {
        int rawosds[CEPH_PG_MAX_SIZE], *osds;
@@ -1184,7 +1192,7 @@ int ceph_calc_pg_acting(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid,
 /*
  * Return primary osd for given pgid, or -1 if none.
  */
-int ceph_calc_pg_primary(struct ceph_osdmap *osdmap, struct ceph_pg_v1 pgid)
+int ceph_calc_pg_primary(struct ceph_osdmap *osdmap, struct ceph_pg pgid)
 {
        int rawosds[CEPH_PG_MAX_SIZE], *osds;
        int i, num = CEPH_PG_MAX_SIZE;