Unfortunately I have "still no cigar". But here is some more info that
might help you:
* I patched afs_BlackListOnce() and its invocation a little bit more
because I think as is it allways returns 1. Se attached patch.
* If I turn on fstrace, I can not repeat the timeout any more,
probably the client is to slow to trigger the bug then.
Other guess: How dangerous is the new goto out in afs_Analyze() as it
bypasses a lot of code?
Back to logfile debugging after lunch,
Harald.
--- openafs-1.4.8pre2/src/afs/afs_analyze.c.orig 2008-08-22
20:53:16.000000000 +0200
+++ openafs-1.4.8pre2/src/afs/afs_analyze.c 2008-10-15 10:19:08.000000000
+0200
@@ -455,7 +455,7 @@
* blacklist.
*
* Returns:
- * Non-zero value if further servers are available to try,
+ * The number of further servers available to try,
* zero otherwise.
*
* Environment:
@@ -485,7 +485,7 @@
areq->skipserver[i] = 1;
}
if (tvp->serverHost[i] &&
- !(tvp->serverHost[i]->addr->sa_flags &
+ (tvp->serverHost[i]->addr->sa_flags &
SRVR_ISDOWN)) {
areq->skipserver[i] = 1;
}
@@ -494,9 +494,8 @@
}
}
for (i = 0; i < MAXHOSTS; i++) {
- if (areq->skipserver[i] == 0) {
- serversleft = 1;
- break;
+ if (tvp->serverHost[i] && areq->skipserver[i] == 0) {
+ serversleft++;
}
}
return serversleft;
@@ -545,7 +544,6 @@
struct server *tsp;
struct volume *tvp;
afs_int32 shouldRetry = 0;
- afs_int32 serversleft = 1;
struct afs_stats_RPCErrors *aerrP;
afs_int32 markeddown;
@@ -672,9 +670,8 @@
#endif /* AFS_64BIT_CLIENT */
if ((acode < 0) && (acode != VRESTARTING)) {
if (acode == RX_CALL_TIMEOUT) {
- serversleft = afs_BlackListOnce(areq, afid, tsp);
areq->idleError++;
- if (serversleft) {
+ if (afs_BlackListOnce(areq, afid, tsp)) {
shouldRetry = 1;
} else {
shouldRetry = 0;
@@ -746,10 +743,8 @@
("afs: Tokens for user of AFS id %d for cell %s have
expired\n",
tu->vid, aconn->srvr->server->cell->cellName);
} else {
- serversleft = afs_BlackListOnce(areq, afid, tsp);
areq->tokenError++;
-
- if (serversleft) {
+ if (afs_BlackListOnce(areq, afid, tsp)) {
afs_warnuser
("afs: Tokens for user of AFS id %d for cell %s: rxkad
error=%d\n",
tu->vid, aconn->srvr->server->cell->cellName, acode);
