1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
|
--- a/dbinc/repmgr.h
+++ b/dbinc/repmgr.h
@@ -374,6 +374,7 @@ typedef struct {
#define SITE_FROM_EID(eid) (&db_rep->sites[eid])
#define EID_FROM_SITE(s) ((int)((s) - (&db_rep->sites[0])))
#define IS_VALID_EID(e) ((e) >= 0)
+#define IS_KNOWN_REMOTE_SITE(e) ((e) >= 0 && ((u_int)(e)) < db_rep->site_cnt)
#define SELF_EID INT_MAX
#define IS_PEER_POLICY(p) ((p) == DB_REPMGR_ACKS_ALL_PEERS || \
--- a/rep/rep_elect.c
+++ b/rep/rep_elect.c
@@ -33,7 +33,7 @@ static int __rep_elect_init
static int __rep_fire_elected __P((ENV *, REP *, u_int32_t));
static void __rep_elect_master __P((ENV *, REP *));
static int __rep_tally __P((ENV *, REP *, int, u_int32_t *, u_int32_t, roff_t));
-static int __rep_wait __P((ENV *, db_timeout_t *, int *, int, u_int32_t));
+static int __rep_wait __P((ENV *, db_timeout_t *, int, u_int32_t));
/*
* __rep_elect --
@@ -55,7 +55,7 @@ __rep_elect(dbenv, given_nsites, nvotes,
ENV *env;
LOG *lp;
REP *rep;
- int done, eid, elected, full_elect, locked, in_progress, need_req;
+ int done, elected, full_elect, locked, in_progress, need_req;
int ret, send_vote, t_ret;
u_int32_t ack, ctlflags, egen, nsites, orig_tally, priority, realpri;
u_int32_t tiebreaker;
@@ -181,8 +181,7 @@ __rep_elect(dbenv, given_nsites, nvotes,
REP_SYSTEM_UNLOCK(env);
(void)__rep_send_message(env, DB_EID_BROADCAST,
REP_MASTER_REQ, NULL, NULL, 0, 0);
- ret = __rep_wait(env, &to, &eid,
- 0, REP_F_EPHASE0);
+ ret = __rep_wait(env, &to, 0, REP_F_EPHASE0);
REP_SYSTEM_LOCK(env);
F_CLR(rep, REP_F_EPHASE0);
switch (ret) {
@@ -286,11 +285,11 @@ restart:
REP_SYSTEM_LOCK(env);
goto vote;
}
- ret = __rep_wait(env, &to, &eid, full_elect, REP_F_EPHASE1);
+ ret = __rep_wait(env, &to, full_elect, REP_F_EPHASE1);
switch (ret) {
case 0:
/* Check if election complete or phase complete. */
- if (eid != DB_EID_INVALID && !IN_ELECTION(rep)) {
+ if (!IN_ELECTION(rep)) {
RPRINT(env, DB_VERB_REP_ELECT,
(env, "Ended election phase 1"));
goto edone;
@@ -398,15 +397,12 @@ phase2:
REP_SYSTEM_LOCK(env);
goto i_won;
}
- ret = __rep_wait(env, &to, &eid, full_elect, REP_F_EPHASE2);
+ ret = __rep_wait(env, &to, full_elect, REP_F_EPHASE2);
RPRINT(env, DB_VERB_REP_ELECT,
(env, "Ended election phase 2 %d", ret));
switch (ret) {
case 0:
- if (eid != DB_EID_INVALID)
- goto edone;
- ret = DB_REP_UNAVAIL;
- break;
+ goto edone;
case DB_REP_EGENCHG:
if (to > timeout)
to = timeout;
@@ -1050,13 +1046,6 @@ __rep_elect_master(env, rep)
ENV *env;
REP *rep;
{
- /*
- * We often come through here twice, sometimes even more. We mustn't
- * let the redundant calls affect stats counting. But rep_elect relies
- * on this first part for setting eidp.
- */
- rep->master_id = rep->eid;
-
if (F_ISSET(rep, REP_F_MASTERELECT | REP_F_MASTER)) {
/* We've been through here already; avoid double counting. */
return;
@@ -1093,10 +1082,10 @@ __rep_fire_elected(env, rep, egen)
(timeout > 5000000) ? 500000 : ((timeout >= 10) ? timeout / 10 : 1);
static int
-__rep_wait(env, timeoutp, eidp, full_elect, flags)
+__rep_wait(env, timeoutp, full_elect, flags)
ENV *env;
db_timeout_t *timeoutp;
- int *eidp, full_elect;
+ int full_elect;
u_int32_t flags;
{
DB_REP *db_rep;
@@ -1174,7 +1163,6 @@ __rep_wait(env, timeoutp, eidp, full_ele
F_CLR(rep, REP_F_EGENUPDATE);
ret = DB_REP_EGENCHG;
} else if (phase_over) {
- *eidp = rep->master_id;
done = 1;
ret = 0;
}
--- a/repmgr/repmgr_net.c
+++ b/repmgr/repmgr_net.c
@@ -100,6 +100,8 @@ __repmgr_send(dbenv, control, rec, lsnp,
control, rec, &nsites_sent, &npeers_sent)) != 0)
goto out;
} else {
+ DB_ASSERT(env, IS_KNOWN_REMOTE_SITE(eid));
+
/*
* If this is a request that can be sent anywhere, then see if
* we can send it to our peer (to save load on the master), but
|