Change the NFSv4 nfsuserd(8) daemon so that it doesn't preload the

uid<->username mapping cache with an entry when another entry
for that uid is already loaded. This fixes a case where the
mapping of "toor" would replace "root" when the daemon was started,
resulting in no mapping for "root" until the cache entry for "toor"
timed out.
The algorithm is inefficient, but since it is only done once when
the daemon is started up, I don't think that's an issue.

MFC after:	2 weeks

1 files changed, 25 insertions, 4 deletions

diff --git a/usr.sbin/nfsuserd/nfsuserd.c b/usr.sbin/nfsuserd/nfsuserd.c
index 9dd6396..35e2849 100644
--- a/usr.sbin/nfsuserd/nfsuserd.c
+++ b/usr.sbin/nfsuserd/nfsuserd.c
@@ -76,6 +76,8 @@ static bool_t	xdr_retval(XDR *, caddr_t);
 #define	MAXNAME		1024
 #define	MAXNFSUSERD	20
 #define	DEFNFSUSERD	4
+#define	MAXUSERMAX	100000
+#define	MINUSERMAX	10
 #define	DEFUSERMAX	200
 #define	DEFUSERTIMEOUT	(1 * 60)
 struct info {
@@ -96,8 +98,8 @@ pid_t slaves[MAXNFSUSERD];
 int
 main(int argc, char *argv[])
 {
-	int i;
-	int error, len, mustfreeai = 0;
+	int i, j;
+	int error, fnd_dup, len, mustfreeai = 0, start_uidpos;
 	struct nfsd_idargs nid;
 	struct passwd *pwd;
 	struct group *grp;
@@ -107,6 +109,7 @@ main(int argc, char *argv[])
 	sigset_t signew;
 	char hostname[MAXHOSTNAMELEN + 1], *cp;
 	struct addrinfo *aip, hints;
+	static uid_t check_dups[MAXUSERMAX];
 
 	if (modfind("nfscommon") < 0) {
 		/* Not present in kernel, try loading it */
@@ -163,9 +166,10 @@ main(int argc, char *argv[])
 			argc--;
 			argv++;
 			i = atoi(*argv);
-			if (i < 10 || i > 100000) {
+			if (i < MINUSERMAX || i > MAXUSERMAX) {
 				fprintf(stderr,
-				    "usermax %d out of range 10<->100000\n", i);
+				    "usermax %d out of range %d<->%d\n", i,
+				    MINUSERMAX, MAXUSERMAX);
 				usage();
 			}
 			nid.nid_usermax = i;
@@ -326,8 +330,25 @@ main(int argc, char *argv[])
 	/*
 	 * Loop around adding all users.
 	 */
+	start_uidpos = i;
 	setpwent();
 	while (i < nid.nid_usermax && (pwd = getpwent())) {
+		fnd_dup = 0;
+		/*
+		 * Yes, this is inefficient, but it is only done once when
+		 * the daemon is started and will run in a fraction of a second
+		 * for nid_usermax at 10000. If nid_usermax is cranked up to
+		 * 100000, it will take several seconds, depending on the CPU.
+		 */
+		for (j = 0; j < (i - start_uidpos); j++)
+			if (check_dups[j] == pwd->pw_uid) {
+				/* Found another entry for uid, so skip it */
+				fnd_dup = 1;
+				break;
+			}
+		if (fnd_dup != 0)
+			continue;
+		check_dups[i - start_uidpos] = pwd->pw_uid;
 		nid.nid_uid = pwd->pw_uid;
 		nid.nid_name = pwd->pw_name;
 		nid.nid_namelen = strlen(pwd->pw_name);