From 53cc4303b98e85bc0144d9ddaff1c116c7e757a3 Mon Sep 17 00:00:00 2001 From: ngie Date: Fri, 10 Feb 2017 02:21:57 +0000 Subject: MFC r304797,r305467: r304797 (by jmmv): Make use of Kyua's work directories. Change the vnode tests to use the current directory when creating temporary files, which we can assume is a volatile work directory, and then make the kqueue_test.sh driver _not_ abandon the directory created by Kyua. This makes the various kqueue tests independent of each other, and ensures the temporary file is cleaned up on failure. Problem spotted by asomers@ when reviewing D4254. r305467: Move tests/sys/kqueue/... to tests/sys/kqueue/libkqueue/... This is being done to clearly distinguish the libkqueue tests from the (soon to be imported) NetBSD tests. --- tests/sys/kqueue/libkqueue/read.c | 326 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 326 insertions(+) create mode 100644 tests/sys/kqueue/libkqueue/read.c (limited to 'tests/sys/kqueue/libkqueue/read.c') diff --git a/tests/sys/kqueue/libkqueue/read.c b/tests/sys/kqueue/libkqueue/read.c new file mode 100644 index 0000000..2837144 --- /dev/null +++ b/tests/sys/kqueue/libkqueue/read.c @@ -0,0 +1,326 @@ +/* + * Copyright (c) 2009 Mark Heily + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + * + * $FreeBSD$ + */ + +#include "common.h" + +int kqfd; +int sockfd[2]; + +static void +kevent_socket_drain(void) +{ + char buf[1]; + + /* Drain the read buffer, then make sure there are no more events. */ + puts("draining the read buffer"); + if (read(sockfd[0], &buf[0], 1) < 1) + err(1, "read(2)"); +} + +static void +kevent_socket_fill(void) +{ + puts("filling the read buffer"); + if (write(sockfd[1], ".", 1) < 1) + err(1, "write(2)"); +} + + +void +test_kevent_socket_add(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_ADD)"; + struct kevent kev; + + test_begin(test_id); + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + success(); +} + +void +test_kevent_socket_get(void) +{ + const char *test_id = "kevent(EVFILT_READ) wait"; + struct kevent kev; + + test_begin(test_id); + + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + kevent_socket_fill(); + + kev.data = 1; + kevent_cmp(&kev, kevent_get(kqfd)); + + kevent_socket_drain(); + test_no_kevents(); + + kev.flags = EV_DELETE; + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + success(); +} + +void +test_kevent_socket_clear(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_CLEAR)"; + struct kevent kev; + + test_begin(test_id); + + test_no_kevents(); + + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD | EV_CLEAR, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + kevent_socket_fill(); + kevent_socket_fill(); + + kev.data = 2; + kevent_cmp(&kev, kevent_get(kqfd)); + + /* We filled twice, but drain once. Edge-triggered would not generate + additional events. + */ + kevent_socket_drain(); + test_no_kevents(); + + kevent_socket_drain(); + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_DELETE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + success(); +} + +void +test_kevent_socket_disable_and_enable(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_DISABLE)"; + struct kevent kev; + + test_begin(test_id); + + /* + * Write to the socket before adding the event. This way we can verify that + * enabling a triggered kevent causes the event to be returned immediately. + */ + kevent_socket_fill(); + + /* Add a disabled event. */ + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD | EV_DISABLE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + test_no_kevents(); + + /* Re-enable the knote, then see if an event is generated */ + kev.flags = EV_ENABLE; + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + kev.flags = EV_ADD; + kev.data = 1; + kevent_cmp(&kev, kevent_get(kqfd)); + + kevent_socket_drain(); + + kev.flags = EV_DELETE; + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + success(); +} + +void +test_kevent_socket_del(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_DELETE)"; + struct kevent kev; + + test_begin(test_id); + + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_DELETE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + kevent_socket_fill(); + test_no_kevents(); + kevent_socket_drain(); + + success(); +} + +void +test_kevent_socket_oneshot(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_ONESHOT)"; + struct kevent kev; + + test_begin(test_id); + + /* Re-add the watch and make sure no events are pending */ + puts("-- re-adding knote"); + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD | EV_ONESHOT, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + test_no_kevents(); + + puts("-- getting one event"); + kevent_socket_fill(); + kev.data = 1; + kevent_cmp(&kev, kevent_get(kqfd)); + + puts("-- checking knote disabled"); + test_no_kevents(); + + /* Try to delete the knote, it should already be deleted */ + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_DELETE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) == 0) + err(1, "%s", test_id); + + kevent_socket_drain(); + + success(); +} + + +#if HAVE_EV_DISPATCH +void +test_kevent_socket_dispatch(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_DISPATCH)"; + + test_begin(test_id); + + struct kevent kev; + + /* Re-add the watch and make sure no events are pending */ + puts("-- re-adding knote"); + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD | EV_DISPATCH, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + test_no_kevents(); + + /* The event will occur only once, even though EV_CLEAR is not + specified. */ + kevent_socket_fill(); + kev.data = 1; + kevent_cmp(&kev, kevent_get(kqfd)); + test_no_kevents(); + + /* Since the knote is disabled, the EV_DELETE operation succeeds. */ + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_DELETE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + kevent_socket_drain(); + + success(); +} +#endif /* HAVE_EV_DISPATCH */ + +#if BROKEN +void +test_kevent_socket_lowat(void) +{ + const char *test_id = "kevent(EVFILT_READ, NOTE_LOWAT)"; + struct kevent kev; + + test_begin(test_id); + + /* Re-add the watch and make sure no events are pending */ + puts("-- re-adding knote, setting low watermark to 2 bytes"); + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD | EV_ONESHOT, NOTE_LOWAT, 2, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + test_no_kevents(); + + puts("-- checking that one byte does not trigger an event.."); + kevent_socket_fill(); + test_no_kevents(); + + puts("-- checking that two bytes triggers an event.."); + kevent_socket_fill(); + if (kevent(kqfd, NULL, 0, &kev, 1, NULL) != 1) + err(1, "%s", test_id); + KEV_CMP(kev, sockfd[0], EVFILT_READ, 0); + test_no_kevents(); + + kevent_socket_drain(); + kevent_socket_drain(); + + success(); +} +#endif + +void +test_kevent_socket_eof(void) +{ + const char *test_id = "kevent(EVFILT_READ, EV_EOF)"; + struct kevent kev; + + test_begin(test_id); + + /* Re-add the watch and make sure no events are pending */ + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_ADD, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + test_no_kevents(); + + if (close(sockfd[1]) < 0) + err(1, "close(2)"); + + kev.flags |= EV_EOF; + kevent_cmp(&kev, kevent_get(kqfd)); + + /* Delete the watch */ + EV_SET(&kev, sockfd[0], EVFILT_READ, EV_DELETE, 0, 0, &sockfd[0]); + if (kevent(kqfd, &kev, 1, NULL, 0, NULL) < 0) + err(1, "%s", test_id); + + success(); +} + +void +test_evfilt_read() +{ + /* Create a connected pair of full-duplex sockets for testing socket events */ + if (socketpair(AF_UNIX, SOCK_STREAM, 0, sockfd) < 0) + abort(); + + kqfd = kqueue(); + test_kevent_socket_add(); + test_kevent_socket_del(); + test_kevent_socket_get(); + test_kevent_socket_disable_and_enable(); + test_kevent_socket_oneshot(); + test_kevent_socket_clear(); +#if HAVE_EV_DISPATCH + test_kevent_socket_dispatch(); +#endif + test_kevent_socket_eof(); + close(kqfd); +} -- cgit v1.1