libceph: introduce ceph_spg, ceph_pg_to_primary_shard()

Store both raw pgid and actual spgid in ceph_osd_request_target.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

5 files changed, 60 insertions, 4 deletions

diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h
index ef630ebd1169..6114f7b02135 100644
--- a/include/linux/ceph/osd_client.h
+++ b/include/linux/ceph/osd_client.h
@@ -136,7 +136,8 @@ struct ceph_osd_request_target {
 	struct ceph_object_id target_oid;
 	struct ceph_object_locator target_oloc;
 
-	struct ceph_pg pgid;
+	struct ceph_pg pgid;               /* last raw pg we mapped to */
+	struct ceph_spg spgid;             /* last actual spg we mapped to */
 	u32 pg_num;
 	u32 pg_num_mask;
 	struct ceph_osds acting;
diff --git a/include/linux/ceph/osdmap.h b/include/linux/ceph/osdmap.h
index 938656f70807..7ae5b416b4b6 100644
--- a/include/linux/ceph/osdmap.h
+++ b/include/linux/ceph/osdmap.h
@@ -24,6 +24,13 @@ struct ceph_pg {
 	uint32_t seed;
 };
 
+#define CEPH_SPG_NOSHARD	-1
+
+struct ceph_spg {
+	struct ceph_pg pgid;
+	s8 shard;
+};
+
 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs);
 
 #define CEPH_POOL_FLAG_HASHPSPOOL	(1ULL << 0) /* hash pg seed and pool id
@@ -271,6 +278,9 @@ void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap,
 			       const struct ceph_pg *raw_pgid,
 			       struct ceph_osds *up,
 			       struct ceph_osds *acting);
+bool ceph_pg_to_primary_shard(struct ceph_osdmap *osdmap,
+			      const struct ceph_pg *raw_pgid,
+			      struct ceph_spg *spgid);
 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
 			      const struct ceph_pg *raw_pgid);
 
diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c
index 71ba13927b3d..50ab1bdb16e2 100644
--- a/net/ceph/debugfs.c
+++ b/net/ceph/debugfs.c
@@ -147,11 +147,20 @@ static int monc_show(struct seq_file *s, void *p)
 	return 0;
 }
 
+static void dump_spgid(struct seq_file *s, const struct ceph_spg *spgid)
+{
+	seq_printf(s, "%llu.%x", spgid->pgid.pool, spgid->pgid.seed);
+	if (spgid->shard != CEPH_SPG_NOSHARD)
+		seq_printf(s, "s%d", spgid->shard);
+}
+
 static void dump_target(struct seq_file *s, struct ceph_osd_request_target *t)
 {
 	int i;
 
-	seq_printf(s, "osd%d\t%llu.%x\t[", t->osd, t->pgid.pool, t->pgid.seed);
+	seq_printf(s, "osd%d\t%llu.%x\t", t->osd, t->pgid.pool, t->pgid.seed);
+	dump_spgid(s, &t->spgid);
+	seq_puts(s, "\t[");
 	for (i = 0; i < t->up.size; i++)
 		seq_printf(s, "%s%d", (!i ? "" : ","), t->up.osds[i]);
 	seq_printf(s, "]/%d\t[", t->up.primary);
diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index aca6319b99b6..66509414f4c8 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -373,6 +373,7 @@ static void target_copy(struct ceph_osd_request_target *dest,
 	ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
 
 	dest->pgid = src->pgid; /* struct */
+	dest->spgid = src->spgid; /* struct */
 	dest->pg_num = src->pg_num;
 	dest->pg_num_mask = src->pg_num_mask;
 	ceph_osds_copy(&dest->acting, &src->acting);
@@ -1394,6 +1395,7 @@ static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 	    ceph_osds_changed(&t->acting, &acting, any_change) ||
 	    force_resend) {
 		t->pgid = pgid; /* struct */
+		ceph_pg_to_primary_shard(osdc->osdmap, &pgid, &t->spgid);
 		ceph_osds_copy(&t->acting, &acting);
 		ceph_osds_copy(&t->up, &up);
 		t->size = pi->size;
@@ -1595,9 +1597,10 @@ static void send_request(struct ceph_osd_request *req)
 
 	encode_request(req, req->r_request);
 
-	dout("%s req %p tid %llu to pg %llu.%x osd%d flags 0x%x attempt %d\n",
+	dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d flags 0x%x attempt %d\n",
 	     __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
-	     req->r_t.osd, req->r_flags, req->r_attempts);
+	     req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
+	     req->r_t.spgid.shard, osd->o_osd, req->r_flags, req->r_attempts);
 
 	req->r_t.paused = false;
 	req->r_stamp = jiffies;
diff --git a/net/ceph/osdmap.c b/net/ceph/osdmap.c
index 710ee3dc01b9..a4155620eace 100644
--- a/net/ceph/osdmap.c
+++ b/net/ceph/osdmap.c
@@ -2265,6 +2265,39 @@ out:
 	WARN_ON(!osds_valid(up) || !osds_valid(acting));
 }
 
+bool ceph_pg_to_primary_shard(struct ceph_osdmap *osdmap,
+			      const struct ceph_pg *raw_pgid,
+			      struct ceph_spg *spgid)
+{
+	struct ceph_pg_pool_info *pi;
+	struct ceph_pg pgid;
+	struct ceph_osds up, acting;
+	int i;
+
+	pi = ceph_pg_pool_by_id(osdmap, raw_pgid->pool);
+	if (!pi)
+		return false;
+
+	raw_pg_to_pg(pi, raw_pgid, &pgid);
+
+	if (ceph_can_shift_osds(pi)) {
+		spgid->pgid = pgid; /* struct */
+		spgid->shard = CEPH_SPG_NOSHARD;
+		return true;
+	}
+
+	ceph_pg_to_up_acting_osds(osdmap, &pgid, &up, &acting);
+	for (i = 0; i < acting.size; i++) {
+		if (acting.osds[i] == acting.primary) {
+			spgid->pgid = pgid; /* struct */
+			spgid->shard = i;
+			return true;
+		}
+	}
+
+	return false;
+}
+
 /*
  * Return acting primary for given PG, or -1 if none.
  */