implement result address sorting in the resolver (getaddrinfo, etc.)

2 files changed, 136 insertions, 0 deletions

diff --git a/src/network/lookup.h b/src/network/lookup.h
index 19c9e488..4e45d869 100644
--- a/src/network/lookup.h
+++ b/src/network/lookup.h
@@ -7,6 +7,7 @@ struct address {
 	int family;
 	unsigned scopeid;
 	uint8_t addr[16];
+	int sortkey;
 };
 
 struct service {
diff --git a/src/network/lookup_name.c b/src/network/lookup_name.c
index 743aa082..0225a934 100644
--- a/src/network/lookup_name.c
+++ b/src/network/lookup_name.c
@@ -7,6 +7,8 @@
 #include <stdlib.h>
 #include <string.h>
 #include <fcntl.h>
+#include <unistd.h>
+#include <pthread.h>
 #include "lookup.h"
 #include "stdio_impl.h"
 #include "syscall.h"
@@ -146,6 +148,80 @@ static int name_from_dns(struct address buf[static MAXADDRS], char canon[static
 	return EAI_FAIL;
 }
 
+static const struct policy {
+	unsigned char addr[16];
+	unsigned char len, mask;
+	unsigned char prec, label;
+} defpolicy[] = {
+	{ "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\1", 15, 0xff, 50, 0 },
+	{ "\0\0\0\0\0\0\0\0\0\0\xff\xff", 11, 0xff, 35, 4 },
+	{ "\x20\2", 1, 0xff, 30, 2 },
+	{ "\x20\1", 3, 0xff, 5, 5 },
+	{ "\xfc", 0, 0xfe, 3, 13 },
+#if 0
+	/* These are deprecated and/or returned to the address
+	 * pool, so despite the RFC, treating them as special
+	 * is probably wrong. */
+	{ "", 11, 0xff, 1, 3 },
+	{ "\xfe\xc0", 1, 0xc0, 1, 11 },
+	{ "\x3f\xfe", 1, 0xff, 1, 12 },
+#endif
+	/* Last rule must match all addresses to stop loop. */
+	{ "", 0, 0, 40, 1 },
+};
+
+static const struct policy *policyof(const struct in6_addr *a)
+{
+	int i;
+	for (i=0; ; i++) {
+		if (memcmp(a->s6_addr, defpolicy[i].addr, defpolicy[i].len))
+			continue;
+		if ((a->s6_addr[defpolicy[i].len] & defpolicy[i].mask)
+		    != defpolicy[i].addr[defpolicy[i].len])
+			continue;
+		return defpolicy+i;
+	}
+}
+
+static int labelof(const struct in6_addr *a)
+{
+	return policyof(a)->label;
+}
+
+static int scopeof(const struct in6_addr *a)
+{
+	if (IN6_IS_ADDR_MULTICAST(a)) return a->s6_addr[1] & 15;
+	if (IN6_IS_ADDR_LINKLOCAL(a)) return 2;
+	if (IN6_IS_ADDR_LOOPBACK(a)) return 2;
+	if (IN6_IS_ADDR_SITELOCAL(a)) return 5;
+	return 14;
+}
+
+static int prefixmatch(const struct in6_addr *s, const struct in6_addr *d)
+{
+	/* FIXME: The common prefix length should be limited to no greater
+	 * than the nominal length of the prefix portion of the source
+	 * address. However the definition of the source prefix length is
+	 * not clear and thus this limiting is not yet implemented. */
+	unsigned i;
+	for (i=0; i<128 && !((s->s6_addr[i/8]^d->s6_addr[i/8])&(128>>(i%8))); i++);
+	return i;
+}
+
+#define DAS_USABLE              0x40000000
+#define DAS_MATCHINGSCOPE       0x20000000
+#define DAS_MATCHINGLABEL       0x10000000
+#define DAS_PREC_SHIFT          20
+#define DAS_SCOPE_SHIFT         16
+#define DAS_PREFIX_SHIFT        8
+#define DAS_ORDER_SHIFT         0
+
+static int addrcmp(const void *_a, const void *_b)
+{
+	const struct address *a = _a, *b = _b;
+	return b->sortkey - a->sortkey;
+}
+
 int __lookup_name(struct address buf[static MAXADDRS], char canon[static 256], const char *name, int family, int flags)
 {
 	int cnt = 0, i, j;
@@ -198,5 +274,64 @@ int __lookup_name(struct address buf[static MAXADDRS], char canon[static 256], c
 		}
 	}
 
+	/* No further processing is needed if there are fewer than 2
+	 * results or if there are only IPv4 results. */
+	if (cnt<2 || family==AF_INET) return cnt;
+	for (i=0; buf[i].family == AF_INET; i++)
+		if (i==cnt) return cnt;
+
+	int cs;
+	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cs);
+
+	/* The following implements a subset of RFC 3484/6724 destination
+	 * address selection by generating a single 31-bit sort key for
+	 * each address. Rules 3, 4, and 7 are omitted for having
+	 * excessive runtime and code size cost and dubious benefit.
+	 * So far the label/precedence table cannot be customized. */
+	for (i=0; i<cnt; i++) {
+		int key = 0;
+		struct sockaddr_in6 sa, da = {
+			.sin6_family = AF_INET6,
+			.sin6_scope_id = buf[i].scopeid,
+			.sin6_port = 65535
+		};
+		if (buf[i].family == AF_INET6) {
+			memcpy(da.sin6_addr.s6_addr, buf[i].addr, 16);
+		} else {
+			memcpy(da.sin6_addr.s6_addr,
+				"\0\0\0\0\0\0\0\0\0\0\xff\xff", 12);
+			memcpy(da.sin6_addr.s6_addr+12, buf[i].addr, 4);
+		}
+		const struct policy *dpolicy = policyof(&da.sin6_addr);
+		int dscope = scopeof(&da.sin6_addr);
+		int dlabel = dpolicy->label;
+		int dprec = dpolicy->prec;
+		int prefixlen = 0;
+		int fd = socket(AF_INET6, SOCK_DGRAM|SOCK_CLOEXEC, IPPROTO_UDP);
+		if (fd >= 0) {
+			if (!connect(fd, (void *)&da, sizeof da)) {
+				key |= DAS_USABLE;
+				if (!getsockname(fd, (void *)&sa,
+				    &(socklen_t){sizeof sa})) {
+					if (dscope == scopeof(&sa.sin6_addr))
+						key |= DAS_MATCHINGSCOPE;
+					if (dlabel == labelof(&sa.sin6_addr))
+						key |= DAS_MATCHINGLABEL;
+					prefixlen = prefixmatch(&sa.sin6_addr,
+						&da.sin6_addr);
+				}
+			}
+			close(fd);
+		}
+		key |= dprec << DAS_PREC_SHIFT;
+		key |= (15-dscope) << DAS_SCOPE_SHIFT;
+		key |= prefixlen << DAS_PREFIX_SHIFT;
+		key |= (MAXADDRS-i) << DAS_ORDER_SHIFT;
+		buf[i].sortkey = key;
+	}
+	qsort(buf, cnt, sizeof *buf, addrcmp);
+
+	pthread_setcancelstate(cs, 0);
+
 	return cnt;
 }