From: Antti Kantee <pooka@iki.fi>
Date: Wed, 26 Nov 2014 16:57:11 +0000 (+0000)
Subject: put the "standard" demo routines in a subdirectory
X-Git-Url: http://xenbits.xensource.com/gitweb?a=commitdiff_plain;h=56f86ecd280b276e1119926a582b540e57d03189;p=rumpuser-xen.git

put the "standard" demo routines in a subdirectory
---

diff --git a/Makefile b/Makefile
index fa7ef18..ead2f7d 100644
--- a/Makefile
+++ b/Makefile
@@ -176,8 +176,9 @@ tests:
 httpd:
 	$(APP_TOOLS_MAKE) -C httpd -f Makefile.boot
 
-rump-kernel: rumpkern_demo.c pthread_test.c httpd
-	app-tools/rumpapp-xen-cc -o $@ rumpkern_demo.c pthread_test.c httpd/*.o
+STDTESTS=tests/libstdtests/rumpkern_demo.c tests/libstdtests/pthread_test.c
+rump-kernel: ${STDTESTS} httpd
+	app-tools/rumpapp-xen-cc -o $@ ${STDTESTS} httpd/*.o
 
 .PHONY: clean arch_clean app-tools_clean
 
diff --git a/pthread_test.c b/pthread_test.c
deleted file mode 100644
index 023be19..0000000
--- a/pthread_test.c
+++ /dev/null
@@ -1,133 +0,0 @@
-/*
- * This is a demonstration program to test pthreads on rumprun.
- * It's not very complete ...
- */
-
-#include <sys/types.h>
-
-#include <err.h>
-#include <errno.h>
-#include <pthread.h>
-#include <unistd.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-
-static pthread_mutex_t mtx;
-static pthread_cond_t cv, cv2;
-
-static int nthreads = 4;
-
-static void
-threxit(void *arg)
-{
-
-	pthread_mutex_lock(&mtx);
-	if (--nthreads == 0) {
-		printf("signalling\n");
-		pthread_cond_signal(&cv2);
-	}
-	pthread_mutex_unlock(&mtx);
-
-	printf("thread %p EXIT %d\n", arg, nthreads);
-}
-
-static void *
-mythread(void *arg)
-{
-
-	printf("thread %p\n", arg);
-
-	pthread_mutex_lock(&mtx);
-	printf("got lock %p\n", arg);
-	sched_yield();
-	pthread_mutex_unlock(&mtx);
-	printf("unlocked lock %p\n", arg);
-	sched_yield();
-
-	threxit(arg);
-
-	return NULL;
-}
-
-static int predicate;
-
-static void *
-waitthread(void *arg)
-{
-
-	printf("thread %p\n", arg);
-	pthread_mutex_lock(&mtx);
-	while (!predicate) {
-		printf("no good, need to wait %p\n", arg);
-		pthread_cond_wait(&cv, &mtx);
-	}
-	pthread_mutex_unlock(&mtx);
-	printf("condvar complete %p!\n", arg);
-
-	threxit(arg);
-
-	return NULL;
-}
-
-static void *
-wakeupthread(void *arg)
-{
-
-	printf("thread %p\n", arg);
-	pthread_mutex_lock(&mtx);
-	predicate = 1;
-	printf("rise and shine %p!\n", arg);
-	pthread_cond_signal(&cv);
-	pthread_mutex_unlock(&mtx);
-
-	threxit(arg);
-
-	return NULL;
-}
-
-void
-test_pthread(void)
-{
-	struct timespec ts;
-	pthread_t pt;
-
-	pthread_mutex_init(&mtx, NULL);
-	pthread_cond_init(&cv, NULL);
-	pthread_cond_init(&cv2, NULL);
-
-	if (pthread_create(&pt, NULL, mythread, (void *)0x01) != 0)
-		errx(1, "pthread_create()");
-
-	clock_gettime(CLOCK_REALTIME, &ts);
-	ts.tv_nsec += 1000*1000;
-	pthread_mutex_lock(&mtx);
-	if (pthread_cond_timedwait(&cv2, &mtx, &ts) != ETIMEDOUT) {
-		printf("cond_timedwait fail\n");
-		abort();
-	}
-	pthread_mutex_unlock(&mtx);
-
-	if (pthread_create(&pt, NULL, mythread, (void *)0x02) != 0)
-		errx(1, "pthread_create()");
-	if (pthread_create(&pt, NULL, waitthread, (void *)0x03) != 0)
-		errx(1, "pthread_create()");
-	if (pthread_create(&pt, NULL, wakeupthread, (void *)0x04) != 0)
-		errx(1, "pthread_create()");
-
-	pthread_mutex_lock(&mtx);
-	/* get time after locking => ensure loop runs before threads finish */
-	clock_gettime(CLOCK_REALTIME, &ts);
-	ts.tv_sec++;
-	while (nthreads) {
-		int rv;
-
-		printf("mainthread condwaiting\n");
-		if ((rv = pthread_cond_timedwait(&cv2, &mtx, &ts)) != 0) {
-			printf("drain condwait fail %d %d\n", rv, nthreads);
-		}
-	}
-	pthread_mutex_unlock(&mtx);
-
-	printf("main thread done\n");
-}
diff --git a/rumpkern_demo.c b/rumpkern_demo.c
deleted file mode 100644
index c85557e..0000000
--- a/rumpkern_demo.c
+++ /dev/null
@@ -1,402 +0,0 @@
-/* Copyright (c) 2013 Antti Kantee.  See COPYING */
-
-#include <mini-os/console.h>
-#include <mini-os/netfront.h>
-
-#include <sys/cdefs.h>
-#include <sys/types.h>
-
-#include <netinet/in.h>
-
-#include <ufs/ufs/ufsmount.h>
-
-#include <dirent.h>
-#include <err.h>
-#include <errno.h>
-#include <fcntl.h>
-#include <poll.h>
-#include <pthread.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-
-#include <rump/rump.h>
-#include <rump/netconfig.h>
-
-void demo_thread(void *);
-
-#define BLKDEV(num) "/BLK" __STRING(num)
-#define BUFSIZE (64*1024)
-
-static void
-dofs(void)
-{
-	struct ufs_args ua;
-	struct dirent *dent;
-	DIR *dp;
-	void *buf;
-	int8_t *p;
-	int fd;
-	int rv;
-
-	if (open("/not_there", O_RDWR) != -1 || errno != ENOENT)
-		errx(1, "errno test");
-
-	if ((rv = rump_pub_etfs_register(BLKDEV(0), "blk0",
-	    RUMP_ETFS_BLK)) != 0)
-		errx(1, "etfs %s", strerror(rv));
-
-	mkdir("/mnt", 0777);
-	ua.fspec = BLKDEV(0);
-	if (mount(MOUNT_FFS, "/mnt", 0, &ua, sizeof(ua)) != 0)
-		err(1, "mount");
-
-	buf = malloc(BUFSIZE);
-	chdir("/mnt");
-
-	dp = opendir(".");
-	if (dp == NULL)
-		err(1, "opendir");
-
-	printf("\treading directory contents\n");
-	while ((dent = readdir(dp)) != NULL) {
-		printf("%d %"PRIu64" %s\n",
-		    dent->d_type, dent->d_fileno, dent->d_name);
-	}
-	closedir(dp);
-
-	/* assume README.md exists, open it, and display first line */
-	if ((fd = open("README.md", O_RDWR)) == -1)
-		err(1, "open README");
-
-	memset(buf, 0, sizeof(buf));
-	if (read(fd, buf, 200) < 200)
-		err(1, "read");
-
-	if ((p = (void *)strchr(buf, '\n')) == NULL)
-		errx(1, "strchr");
-	*p = '\0';
-	printf("Reading first line of README.md:\n\t===\n%s\n\t===\n\n", (char *)buf);
-
-#define GARBAGE "   TRASHED!   "
-	/* write some garbage in there.  spot the difference the next time */
-	pwrite(fd, GARBAGE, sizeof(GARBAGE)-1, 20);
-	close(fd);
-
-	/*
-	 * FS will be unmounted when you type "reboot" into the
-	 * net demo, just like in any regular OS
-	 */
-	sync(); /* but just to be safe */
-}
-
-#define MAXCONN 64
-
-struct conn {
-	int c_bpos;
-	int c_cnt;
-	char c_buf[0xe0];
-};
-
-static struct pollfd pfds[MAXCONN];
-static struct conn conns[MAXCONN];
-int maxfd, masterfd;
-
-static void
-acceptconn(void)
-{
-	struct sockaddr_in sin;
-	socklen_t slen = sizeof(sin);
-	int s;
-
-	if ((s = accept(masterfd, (struct sockaddr *)&sin, &slen)) == -1)
-		return;
-
-	/* drop */
-	if (s >= MAXCONN) {
-		close(s);
-		return;
-	}
-
-	/* init */
-	pfds[s].fd = s;
-	memset(&conns[s], 0, sizeof(conns[s]));
-
-	/* XXX: not g/c'd */
-	if (s+1 > maxfd)
-		maxfd = s+1;
-
-#define PROMPT "LOGON: "
-	/* just assume this will go into the socket without blocking */
-	write(s, PROMPT, sizeof(PROMPT)-1);
-#undef PROMPT
-}
-
-static void
-readconn(int i)
-{
-	struct conn *c = &conns[i];
-	char *p;
-	ssize_t nn;
-
-	nn = read(i, c->c_buf+c->c_bpos, sizeof(c->c_buf)-c->c_bpos);
-	/* treat errors and EOF the same way, we shouldn't get EAGAIN */
-	if (nn <= 0) {
-		close(i);
-		pfds[i].fd = -1;
-		c->c_cnt = -1;
-		return;
-	}
-
-	if ((p = strchr(c->c_buf, '\n')) == NULL) {
-		c->c_bpos += nn;
-		return;
-	}
-
-	*p = '\0';
-#define GREET "GREETINGS PROFESSOR FALKEN.\n"
-#define NOPE "LOGIN INCORRECT\n"
-	/* multiple holes here, some more microsofty than others */
-	if (strncmp(c->c_buf, "Joshua", 6) == 0) {
-		write(i, GREET, sizeof(GREET)-1);
-	} else if (strncmp(c->c_buf, "reboot", 6) == 0) {
-		reboot(0, 0);
-	} else {
-		write(i, NOPE, sizeof(NOPE)-1);
-	}
-	pfds[i].fd = -1;
-#undef GREET
-#undef NOPE
-}
-
-static void
-processzombies(void)
-{
-	int i;
-
-	/*
-	 * Let each connection live ~10s regardless of whether it's
-	 * completed or not.
-	 */
-	for (i = 1; i < MAXCONN; i++) {
-		if (conns[i].c_cnt != -1 && ++conns[i].c_cnt > 10
-                    && pfds[i].fd != -1) {
-			close(i);
-		}
-	}
-}
-
-static void
-setupnet(void)
-{
-	int rv;
-
-	if ((rv = rump_pub_netconfig_ifcreate("xenif0")) != 0) {
-		printf("creating xenif0 failed: %d\n", rv);
-		return;
-	}
-
-	/*
-	 * Configure the interface using DHCP.  DHCP support is a bit
-	 * flimsy, so if this doesn't work properly, you can also use
-	 * the manual interface configuration options.
-	 */
-	if ((rv = rump_pub_netconfig_dhcp_ipv4_oneshot("xenif0")) != 0) {
-		printf("getting IP for xenif0 via DHCP failed: %d\n", rv);
-		return;
-	}
-}
-
-static int
-sucketonport(uint16_t port)
-{
-	struct sockaddr_in sin;
-	int s;
-
-	s = socket(PF_INET, SOCK_STREAM, 0);
-	if (s == -1)
-		err(1, "socket");
-
-	memset(&sin, 0, sizeof(sin));
-	sin.sin_len = sizeof(sin);
-	sin.sin_family = AF_INET;
-	sin.sin_port = htons(port);
-	sin.sin_addr.s_addr = INADDR_ANY;
-	if (bind(s, (struct sockaddr *)&sin, sizeof(sin)) == -1)
-		err(1, "bind");
-
-	if (listen(s, 10) == -1)
-		err(1, "unix man, please listen");
-
-	return s;
-}
-
-static void
-donet(void)
-{
-	uint64_t zombietime;
-	int rv, i;
-
-	setupnet();
-	masterfd = sucketonport(4096);
-	ASSERT(masterfd < MAXCONN);
-
-	for (i = 0; i < MAXCONN; i++) {
-		pfds[i].fd = -1;
-		pfds[i].events = POLLIN;
-		conns[i].c_cnt = -1;
-	}
-	pfds[masterfd].fd = masterfd;
-	maxfd = masterfd+1;
-
-	printf("WOPR reporting for duty on port 4096\n");
-
-	zombietime = NOW();
-	for (;;) {
-		if (NOW() - zombietime >= SECONDS(1)) {
-			processzombies();
-			zombietime = NOW();
-		}
-
-		rv = poll(pfds, maxfd, 1000);
-		if (rv == 0) {
-			printf("still waiting ... %"PRId64"d\n", NOW());
-			continue;
-		}
-
-		if (rv == -1) {
-			printf("fail poll %d\n", errno);
-			reboot(0, 0);
-		}
-
-		if (pfds[masterfd].revents & POLLIN) {
-			acceptconn();
-			rv--;
-		}
-
-		for (i = 0; i < MAXCONN && rv; i++) {
-			if (i == masterfd)
-				continue;
-			if (pfds[i].fd != -1 && pfds[i].revents & POLLIN) {
-				readconn(i);
-				rv--;
-			}
-		}
-		ASSERT(rv == 0);
-	}
-}
-
-int main(int, char **);
-void *
-wwwbozo(void *arg)
-{
-	char *argv[] = { "bozo", "-X", "/etc" };
-
-	rump_pub_lwproc_switch(arg);
-
-	/*
-	 * Call program main.  Since we don't have a new vm space,
-	 * ensure that options will be re-parsed.
-	 */
-	optind = 1;
-	optreset = 1;
-	main(sizeof(argv)/sizeof(argv[0]), argv);
-
-	/* among other things, will close fd's */
-	rump_pub_lwproc_releaselwp();
-
-	return NULL;
-}
-
-static void
-dohttpd(void)
-{
-	struct ufs_args ua;
-	struct sockaddr_in sin;
-	socklen_t slen;
-	int rv, s, sa, count;
-	struct lwp *l;
-	pthread_t pt;
-
-	if ((rv = rump_pub_etfs_register(BLKDEV(1),
-	    "blk1", RUMP_ETFS_BLK)) != 0)
-		errx(1, "etfs %s", strerror(rv));
-
-	mkdir("/etc", 0777);
-	ua.fspec = BLKDEV(1);
-	if (mount(MOUNT_FFS, "/etc", MNT_RDONLY, &ua, sizeof(ua)) == -1)
-		err(1, "mount");
-	setupnet();
-
-	/* create a decicated process which does the work */
-	rump_pub_lwproc_rfork(RUMP_RFCFDG);
-	l = rump_pub_lwproc_curlwp();
-
-	/*
-	 * ok, now.  we run bozohttpd in inetd mode to gain control
-	 * of the worker model is uses.  This means that we have to:
-	 *  1: open the listening socket and accept connections
-	 *  2: rfork with descriptor inheritance
-	 *  3: dup2 the accepted fd's to {0,1,2}
-	 *  4: create a mini os thread to handle the request(s)
-	 */
-	s = sucketonport(80);
-	for (count = 0;; count++) {
-		/* 1: accept */
-		slen = sizeof(sin);
-		sa = accept(s, (struct sockaddr *)&sin, &slen);
-		if (sa == -1)
-			err(1, "accept round %d", count);
-
-		/* 2: rfork */
-		rv = rump_pub_lwproc_rfork(RUMP_RFFDG);
-		if (rv != 0)
-			errx(1, "fork failed: %s", strerror(rv));
-
-		/* 3: setup fd's */
-		dup2(sa, 0);
-		dup2(sa, 1);
-		dup2(sa, 2);
-		fcntl(3, F_CLOSEM);
-
-		/* 4: create thread */
-		pthread_create(&pt, NULL, wwwbozo, rump_pub_lwproc_curlwp());
-
-		/*
-		 * back to handling proc.  yea, this a slightly gray
-		 * area of semantics, and we could do a lot better in
-		 * preventing a race.  that ssaid, we should be fine due
-		 * to cooperative scheduling.
-		 *
-		 * then, release the reference to the socket fd
-		 */
-		rump_pub_lwproc_switch(l);
-		close(sa);
-	}
-}
-
-void test_pthread(void);
-
-int
-app_main(start_info_t *si)
-{
-	long tests;
-
-	printf("running demos, command line: %s\n", si->cmd_line);
-
-	if (si->cmd_line[0]) {
-		tests = strtol((const char *)si->cmd_line, NULL, 16);
-	}
-
-	if (tests & 0x1)
-		dofs();
-	if (tests & 0x8)
-		test_pthread();
-	if (tests & 0x2)
-		donet();
-	if (tests & 0x4)
-		dohttpd();
-
-	return 0;
-}
diff --git a/tests/libstdtests/pthread_test.c b/tests/libstdtests/pthread_test.c
new file mode 100644
index 0000000..023be19
--- /dev/null
+++ b/tests/libstdtests/pthread_test.c
@@ -0,0 +1,133 @@
+/*
+ * This is a demonstration program to test pthreads on rumprun.
+ * It's not very complete ...
+ */
+
+#include <sys/types.h>
+
+#include <err.h>
+#include <errno.h>
+#include <pthread.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+
+static pthread_mutex_t mtx;
+static pthread_cond_t cv, cv2;
+
+static int nthreads = 4;
+
+static void
+threxit(void *arg)
+{
+
+	pthread_mutex_lock(&mtx);
+	if (--nthreads == 0) {
+		printf("signalling\n");
+		pthread_cond_signal(&cv2);
+	}
+	pthread_mutex_unlock(&mtx);
+
+	printf("thread %p EXIT %d\n", arg, nthreads);
+}
+
+static void *
+mythread(void *arg)
+{
+
+	printf("thread %p\n", arg);
+
+	pthread_mutex_lock(&mtx);
+	printf("got lock %p\n", arg);
+	sched_yield();
+	pthread_mutex_unlock(&mtx);
+	printf("unlocked lock %p\n", arg);
+	sched_yield();
+
+	threxit(arg);
+
+	return NULL;
+}
+
+static int predicate;
+
+static void *
+waitthread(void *arg)
+{
+
+	printf("thread %p\n", arg);
+	pthread_mutex_lock(&mtx);
+	while (!predicate) {
+		printf("no good, need to wait %p\n", arg);
+		pthread_cond_wait(&cv, &mtx);
+	}
+	pthread_mutex_unlock(&mtx);
+	printf("condvar complete %p!\n", arg);
+
+	threxit(arg);
+
+	return NULL;
+}
+
+static void *
+wakeupthread(void *arg)
+{
+
+	printf("thread %p\n", arg);
+	pthread_mutex_lock(&mtx);
+	predicate = 1;
+	printf("rise and shine %p!\n", arg);
+	pthread_cond_signal(&cv);
+	pthread_mutex_unlock(&mtx);
+
+	threxit(arg);
+
+	return NULL;
+}
+
+void
+test_pthread(void)
+{
+	struct timespec ts;
+	pthread_t pt;
+
+	pthread_mutex_init(&mtx, NULL);
+	pthread_cond_init(&cv, NULL);
+	pthread_cond_init(&cv2, NULL);
+
+	if (pthread_create(&pt, NULL, mythread, (void *)0x01) != 0)
+		errx(1, "pthread_create()");
+
+	clock_gettime(CLOCK_REALTIME, &ts);
+	ts.tv_nsec += 1000*1000;
+	pthread_mutex_lock(&mtx);
+	if (pthread_cond_timedwait(&cv2, &mtx, &ts) != ETIMEDOUT) {
+		printf("cond_timedwait fail\n");
+		abort();
+	}
+	pthread_mutex_unlock(&mtx);
+
+	if (pthread_create(&pt, NULL, mythread, (void *)0x02) != 0)
+		errx(1, "pthread_create()");
+	if (pthread_create(&pt, NULL, waitthread, (void *)0x03) != 0)
+		errx(1, "pthread_create()");
+	if (pthread_create(&pt, NULL, wakeupthread, (void *)0x04) != 0)
+		errx(1, "pthread_create()");
+
+	pthread_mutex_lock(&mtx);
+	/* get time after locking => ensure loop runs before threads finish */
+	clock_gettime(CLOCK_REALTIME, &ts);
+	ts.tv_sec++;
+	while (nthreads) {
+		int rv;
+
+		printf("mainthread condwaiting\n");
+		if ((rv = pthread_cond_timedwait(&cv2, &mtx, &ts)) != 0) {
+			printf("drain condwait fail %d %d\n", rv, nthreads);
+		}
+	}
+	pthread_mutex_unlock(&mtx);
+
+	printf("main thread done\n");
+}
diff --git a/tests/libstdtests/rumpkern_demo.c b/tests/libstdtests/rumpkern_demo.c
new file mode 100644
index 0000000..c85557e
--- /dev/null
+++ b/tests/libstdtests/rumpkern_demo.c
@@ -0,0 +1,402 @@
+/* Copyright (c) 2013 Antti Kantee.  See COPYING */
+
+#include <mini-os/console.h>
+#include <mini-os/netfront.h>
+
+#include <sys/cdefs.h>
+#include <sys/types.h>
+
+#include <netinet/in.h>
+
+#include <ufs/ufs/ufsmount.h>
+
+#include <dirent.h>
+#include <err.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#include <rump/rump.h>
+#include <rump/netconfig.h>
+
+void demo_thread(void *);
+
+#define BLKDEV(num) "/BLK" __STRING(num)
+#define BUFSIZE (64*1024)
+
+static void
+dofs(void)
+{
+	struct ufs_args ua;
+	struct dirent *dent;
+	DIR *dp;
+	void *buf;
+	int8_t *p;
+	int fd;
+	int rv;
+
+	if (open("/not_there", O_RDWR) != -1 || errno != ENOENT)
+		errx(1, "errno test");
+
+	if ((rv = rump_pub_etfs_register(BLKDEV(0), "blk0",
+	    RUMP_ETFS_BLK)) != 0)
+		errx(1, "etfs %s", strerror(rv));
+
+	mkdir("/mnt", 0777);
+	ua.fspec = BLKDEV(0);
+	if (mount(MOUNT_FFS, "/mnt", 0, &ua, sizeof(ua)) != 0)
+		err(1, "mount");
+
+	buf = malloc(BUFSIZE);
+	chdir("/mnt");
+
+	dp = opendir(".");
+	if (dp == NULL)
+		err(1, "opendir");
+
+	printf("\treading directory contents\n");
+	while ((dent = readdir(dp)) != NULL) {
+		printf("%d %"PRIu64" %s\n",
+		    dent->d_type, dent->d_fileno, dent->d_name);
+	}
+	closedir(dp);
+
+	/* assume README.md exists, open it, and display first line */
+	if ((fd = open("README.md", O_RDWR)) == -1)
+		err(1, "open README");
+
+	memset(buf, 0, sizeof(buf));
+	if (read(fd, buf, 200) < 200)
+		err(1, "read");
+
+	if ((p = (void *)strchr(buf, '\n')) == NULL)
+		errx(1, "strchr");
+	*p = '\0';
+	printf("Reading first line of README.md:\n\t===\n%s\n\t===\n\n", (char *)buf);
+
+#define GARBAGE "   TRASHED!   "
+	/* write some garbage in there.  spot the difference the next time */
+	pwrite(fd, GARBAGE, sizeof(GARBAGE)-1, 20);
+	close(fd);
+
+	/*
+	 * FS will be unmounted when you type "reboot" into the
+	 * net demo, just like in any regular OS
+	 */
+	sync(); /* but just to be safe */
+}
+
+#define MAXCONN 64
+
+struct conn {
+	int c_bpos;
+	int c_cnt;
+	char c_buf[0xe0];
+};
+
+static struct pollfd pfds[MAXCONN];
+static struct conn conns[MAXCONN];
+int maxfd, masterfd;
+
+static void
+acceptconn(void)
+{
+	struct sockaddr_in sin;
+	socklen_t slen = sizeof(sin);
+	int s;
+
+	if ((s = accept(masterfd, (struct sockaddr *)&sin, &slen)) == -1)
+		return;
+
+	/* drop */
+	if (s >= MAXCONN) {
+		close(s);
+		return;
+	}
+
+	/* init */
+	pfds[s].fd = s;
+	memset(&conns[s], 0, sizeof(conns[s]));
+
+	/* XXX: not g/c'd */
+	if (s+1 > maxfd)
+		maxfd = s+1;
+
+#define PROMPT "LOGON: "
+	/* just assume this will go into the socket without blocking */
+	write(s, PROMPT, sizeof(PROMPT)-1);
+#undef PROMPT
+}
+
+static void
+readconn(int i)
+{
+	struct conn *c = &conns[i];
+	char *p;
+	ssize_t nn;
+
+	nn = read(i, c->c_buf+c->c_bpos, sizeof(c->c_buf)-c->c_bpos);
+	/* treat errors and EOF the same way, we shouldn't get EAGAIN */
+	if (nn <= 0) {
+		close(i);
+		pfds[i].fd = -1;
+		c->c_cnt = -1;
+		return;
+	}
+
+	if ((p = strchr(c->c_buf, '\n')) == NULL) {
+		c->c_bpos += nn;
+		return;
+	}
+
+	*p = '\0';
+#define GREET "GREETINGS PROFESSOR FALKEN.\n"
+#define NOPE "LOGIN INCORRECT\n"
+	/* multiple holes here, some more microsofty than others */
+	if (strncmp(c->c_buf, "Joshua", 6) == 0) {
+		write(i, GREET, sizeof(GREET)-1);
+	} else if (strncmp(c->c_buf, "reboot", 6) == 0) {
+		reboot(0, 0);
+	} else {
+		write(i, NOPE, sizeof(NOPE)-1);
+	}
+	pfds[i].fd = -1;
+#undef GREET
+#undef NOPE
+}
+
+static void
+processzombies(void)
+{
+	int i;
+
+	/*
+	 * Let each connection live ~10s regardless of whether it's
+	 * completed or not.
+	 */
+	for (i = 1; i < MAXCONN; i++) {
+		if (conns[i].c_cnt != -1 && ++conns[i].c_cnt > 10
+                    && pfds[i].fd != -1) {
+			close(i);
+		}
+	}
+}
+
+static void
+setupnet(void)
+{
+	int rv;
+
+	if ((rv = rump_pub_netconfig_ifcreate("xenif0")) != 0) {
+		printf("creating xenif0 failed: %d\n", rv);
+		return;
+	}
+
+	/*
+	 * Configure the interface using DHCP.  DHCP support is a bit
+	 * flimsy, so if this doesn't work properly, you can also use
+	 * the manual interface configuration options.
+	 */
+	if ((rv = rump_pub_netconfig_dhcp_ipv4_oneshot("xenif0")) != 0) {
+		printf("getting IP for xenif0 via DHCP failed: %d\n", rv);
+		return;
+	}
+}
+
+static int
+sucketonport(uint16_t port)
+{
+	struct sockaddr_in sin;
+	int s;
+
+	s = socket(PF_INET, SOCK_STREAM, 0);
+	if (s == -1)
+		err(1, "socket");
+
+	memset(&sin, 0, sizeof(sin));
+	sin.sin_len = sizeof(sin);
+	sin.sin_family = AF_INET;
+	sin.sin_port = htons(port);
+	sin.sin_addr.s_addr = INADDR_ANY;
+	if (bind(s, (struct sockaddr *)&sin, sizeof(sin)) == -1)
+		err(1, "bind");
+
+	if (listen(s, 10) == -1)
+		err(1, "unix man, please listen");
+
+	return s;
+}
+
+static void
+donet(void)
+{
+	uint64_t zombietime;
+	int rv, i;
+
+	setupnet();
+	masterfd = sucketonport(4096);
+	ASSERT(masterfd < MAXCONN);
+
+	for (i = 0; i < MAXCONN; i++) {
+		pfds[i].fd = -1;
+		pfds[i].events = POLLIN;
+		conns[i].c_cnt = -1;
+	}
+	pfds[masterfd].fd = masterfd;
+	maxfd = masterfd+1;
+
+	printf("WOPR reporting for duty on port 4096\n");
+
+	zombietime = NOW();
+	for (;;) {
+		if (NOW() - zombietime >= SECONDS(1)) {
+			processzombies();
+			zombietime = NOW();
+		}
+
+		rv = poll(pfds, maxfd, 1000);
+		if (rv == 0) {
+			printf("still waiting ... %"PRId64"d\n", NOW());
+			continue;
+		}
+
+		if (rv == -1) {
+			printf("fail poll %d\n", errno);
+			reboot(0, 0);
+		}
+
+		if (pfds[masterfd].revents & POLLIN) {
+			acceptconn();
+			rv--;
+		}
+
+		for (i = 0; i < MAXCONN && rv; i++) {
+			if (i == masterfd)
+				continue;
+			if (pfds[i].fd != -1 && pfds[i].revents & POLLIN) {
+				readconn(i);
+				rv--;
+			}
+		}
+		ASSERT(rv == 0);
+	}
+}
+
+int main(int, char **);
+void *
+wwwbozo(void *arg)
+{
+	char *argv[] = { "bozo", "-X", "/etc" };
+
+	rump_pub_lwproc_switch(arg);
+
+	/*
+	 * Call program main.  Since we don't have a new vm space,
+	 * ensure that options will be re-parsed.
+	 */
+	optind = 1;
+	optreset = 1;
+	main(sizeof(argv)/sizeof(argv[0]), argv);
+
+	/* among other things, will close fd's */
+	rump_pub_lwproc_releaselwp();
+
+	return NULL;
+}
+
+static void
+dohttpd(void)
+{
+	struct ufs_args ua;
+	struct sockaddr_in sin;
+	socklen_t slen;
+	int rv, s, sa, count;
+	struct lwp *l;
+	pthread_t pt;
+
+	if ((rv = rump_pub_etfs_register(BLKDEV(1),
+	    "blk1", RUMP_ETFS_BLK)) != 0)
+		errx(1, "etfs %s", strerror(rv));
+
+	mkdir("/etc", 0777);
+	ua.fspec = BLKDEV(1);
+	if (mount(MOUNT_FFS, "/etc", MNT_RDONLY, &ua, sizeof(ua)) == -1)
+		err(1, "mount");
+	setupnet();
+
+	/* create a decicated process which does the work */
+	rump_pub_lwproc_rfork(RUMP_RFCFDG);
+	l = rump_pub_lwproc_curlwp();
+
+	/*
+	 * ok, now.  we run bozohttpd in inetd mode to gain control
+	 * of the worker model is uses.  This means that we have to:
+	 *  1: open the listening socket and accept connections
+	 *  2: rfork with descriptor inheritance
+	 *  3: dup2 the accepted fd's to {0,1,2}
+	 *  4: create a mini os thread to handle the request(s)
+	 */
+	s = sucketonport(80);
+	for (count = 0;; count++) {
+		/* 1: accept */
+		slen = sizeof(sin);
+		sa = accept(s, (struct sockaddr *)&sin, &slen);
+		if (sa == -1)
+			err(1, "accept round %d", count);
+
+		/* 2: rfork */
+		rv = rump_pub_lwproc_rfork(RUMP_RFFDG);
+		if (rv != 0)
+			errx(1, "fork failed: %s", strerror(rv));
+
+		/* 3: setup fd's */
+		dup2(sa, 0);
+		dup2(sa, 1);
+		dup2(sa, 2);
+		fcntl(3, F_CLOSEM);
+
+		/* 4: create thread */
+		pthread_create(&pt, NULL, wwwbozo, rump_pub_lwproc_curlwp());
+
+		/*
+		 * back to handling proc.  yea, this a slightly gray
+		 * area of semantics, and we could do a lot better in
+		 * preventing a race.  that ssaid, we should be fine due
+		 * to cooperative scheduling.
+		 *
+		 * then, release the reference to the socket fd
+		 */
+		rump_pub_lwproc_switch(l);
+		close(sa);
+	}
+}
+
+void test_pthread(void);
+
+int
+app_main(start_info_t *si)
+{
+	long tests;
+
+	printf("running demos, command line: %s\n", si->cmd_line);
+
+	if (si->cmd_line[0]) {
+		tests = strtol((const char *)si->cmd_line, NULL, 16);
+	}
+
+	if (tests & 0x1)
+		dofs();
+	if (tests & 0x8)
+		test_pthread();
+	if (tests & 0x2)
+		donet();
+	if (tests & 0x4)
+		dohttpd();
+
+	return 0;
+}