Lets try to finish work by stopping all syncs and fall back to what we
have in cache after 7/8 of the timeout (timeout - 1/2 repo_timeout).
This way we still have 1/8 of time to finish the calculation and produce
output.
Tested this diff by setting the deadline to fire after 60sec.
--
:wq Claudio
Index: main.c
===================================================================
RCS file: /cvs/src/usr.sbin/rpki-client/main.c,v
retrieving revision 1.217
diff -u -p -r1.217 main.c
--- main.c 2 Sep 2022 19:14:04 -0000 1.217
+++ main.c 2 Sep 2022 19:58:19 -0000
@@ -66,6 +66,7 @@ int noop;
int filemode;
int rrdpon = 1;
int repo_timeout;
+time_t deadline;
struct skiplist skiplist = LIST_HEAD_INITIALIZER(skiplist);
@@ -1044,6 +1045,9 @@ main(int argc, char *argv[])
*/
alarm(timeout);
signal(SIGALRM, suicide);
+
+ /* give up a bit before the hard timeout and try to finish up */
+ deadline = getmonotime() + timeout - repo_timeout / 2;
}
if (pledge("stdio rpath wpath cpath fattr sendfd unveil", NULL) == -1)
Index: repo.c
===================================================================
RCS file: /cvs/src/usr.sbin/rpki-client/repo.c,v
retrieving revision 1.38
diff -u -p -r1.38 repo.c
--- repo.c 2 Sep 2022 19:10:37 -0000 1.38
+++ repo.c 2 Sep 2022 19:15:41 -0000
@@ -41,6 +41,8 @@ extern struct stats stats;
extern int noop;
extern int rrdpon;
extern int repo_timeout;
+extern time_t deadline;
+int nofetch;
enum repo_state {
REPO_LOADING = 0,
@@ -288,7 +290,7 @@ repo_done(const void *vp, int ok)
if (vp == rp->rsync)
entityq_flush(&rp->queue, rp);
if (vp == rp->rrdp) {
- if (!ok) {
+ if (!ok && !nofetch) {
/* try to fall back to rsync */
rp->rrdp = NULL;
rp->rsync = rsync_get(rp->repouri,
@@ -937,8 +939,8 @@ rrdp_finish(unsigned int id, int ok)
stats.rrdp_repos++;
rr->state = REPO_DONE;
} else {
- warnx("%s: load from network failed, fallback to rsync",
- rr->notifyuri);
+ warnx("%s: load from network failed, fallback to %s",
+ rr->notifyuri, nofetch ? "cache" : "rsync");
stats.rrdp_fails++;
rr->state = REPO_FAILED;
/* clear the RRDP repo since it failed */
@@ -1044,7 +1046,6 @@ repo_lookup(int talid, const char *uri,
{
struct repo *rp;
char *repouri;
- int nofetch = 0;
if ((repouri = rsync_base_uri(uri)) == NULL)
errx(1, "bad caRepository URI: %s", uri);
@@ -1223,8 +1224,26 @@ repo_check_timeout(int timeout)
{
struct repo *rp;
time_t now;
+ int diff;
now = getmonotime();
+
+ /* check against our runtime deadline first */
+ if (deadline != 0) {
+ if (deadline <= now) {
+ warnx("deadline reached, giving up on repository sync");
+ nofetch = 1;
+ /* clear deadline since nofetch is set */
+ deadline = 0;
+ /* increase now enough so that all pending repos fail */
+ now += repo_timeout;
+ } else {
+ diff = deadline - now;
+ diff *= 1000;
+ if (timeout == INFTIM || diff < timeout)
+ timeout = diff;
+ }
+ }
/* Look up in repository table. (Lookup should actually fail here) */
SLIST_FOREACH(rp, &repos, entry) {
if (repo_state(rp) == REPO_LOADING) {
@@ -1233,7 +1252,7 @@ repo_check_timeout(int timeout)
rp->repouri);
repo_abort(rp);
} else {
- int diff = rp->alarm - now;
+ diff = rp->alarm - now;
diff *= 1000;
if (timeout == INFTIM || diff < timeout)
timeout = diff;
