From 3a531260a14631ae8d231279b8738884bf808e7b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 20 Oct 2012 12:39:03 -0300 Subject: perf trace: Validate syscall id before growing syscall table In some cases the ID for a syscall read thru the raw_syscalls tracepoint is bogus, still needs to be investigated why, but to make the tool more robust first try to resolve the ID to a name via libaudit and if it fails, don't grow the table. Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-0lsokw3xor7c4ijo45u6bauh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-trace.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'tools') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index dec8ced61fb0..83c6515e425d 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -56,6 +56,10 @@ static int trace__read_syscall_info(struct trace *trace, int id) { char tp_name[128]; struct syscall *sc; + const char *name = audit_syscall_to_name(id, trace->audit_machine); + + if (name == NULL) + return -1; if (id > trace->syscalls.max) { struct syscall *nsyscalls = realloc(trace->syscalls.table, (id + 1) * sizeof(*sc)); @@ -75,11 +79,8 @@ static int trace__read_syscall_info(struct trace *trace, int id) } sc = trace->syscalls.table + id; - sc->name = audit_syscall_to_name(id, trace->audit_machine); - if (sc->name == NULL) - return -1; - - sc->fmt = syscall_fmt__find(sc->name); + sc->name = name; + sc->fmt = syscall_fmt__find(sc->name); snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->name); sc->tp_format = event_format__new("syscalls", tp_name); -- cgit v1.3-18-gd494 From fc551f8d4427150da1ee7cbb0f53525c49ef4bca Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 20 Oct 2012 13:08:46 -0300 Subject: perf trace: Check if sample raw_data field is set Sometimes we're segfaulting because we were expecting that the perf_sample.raw_data field was set as requested, but in some cases that needs further investigation, that field can be NULL, leading to segfaults. Make the tool more robust by checking that before calling any per event handlers that may try to use that field. Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-g1fmodl6ys4lq8honbj1igoi@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-trace.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'tools') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 83c6515e425d..7aaee39f6774 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -268,6 +268,13 @@ again: if (evlist->threads->map[0] == -1 || evlist->threads->nr > 1) printf("%d ", sample.tid); + if (sample.raw_data == NULL) { + printf("%s sample with no payload for tid: %d, cpu %d, raw_size=%d, skipping...\n", + perf_evsel__name(evsel), sample.tid, + sample.cpu, sample.raw_size); + continue; + } + handler = evsel->handler.func; handler(trace, evsel, &sample); } -- cgit v1.3-18-gd494 From 670ab5d21c7e168c89a36fdd2c69fb7af63d35a1 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Mon, 22 Oct 2012 16:12:23 +0900 Subject: perf help: Fix --help for builtins It seems that commit cc5848213329 ("perf help: Remove use of die and handle errors") caused the problem - it changed the initial value of 'help_format' from HELP_FORMAT_MAN to HELP_FORMAT_NONE. This broke the --help option for all builtins, that would produce no output, while 'man perf-top' would work it MANPATH is properly setup. Reported-by: Stephane Eranian Signed-off-by: Namhyung Kim Tested-by: David Ahern Acked-by: David Ahern Cc: David Ahern Cc: Jiri Olsa Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/87r4orj7zc.fsf@sejong.aot.lge.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-help.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools') diff --git a/tools/perf/builtin-help.c b/tools/perf/builtin-help.c index 411ee5664e98..178b88ae3d2f 100644 --- a/tools/perf/builtin-help.c +++ b/tools/perf/builtin-help.c @@ -414,7 +414,7 @@ static int show_html_page(const char *perf_cmd) int cmd_help(int argc, const char **argv, const char *prefix __maybe_unused) { bool show_all = false; - enum help_format help_format = HELP_FORMAT_NONE; + enum help_format help_format = HELP_FORMAT_MAN; struct option builtin_help_options[] = { OPT_BOOLEAN('a', "all", &show_all, "print all available commands"), OPT_SET_UINT('m', "man", &help_format, "show man page", HELP_FORMAT_MAN), -- cgit v1.3-18-gd494 From 9fdbf671ba7e8adb2cbb93d716232ebcab55f6bd Mon Sep 17 00:00:00 2001 From: Luigi Semenzato Date: Tue, 21 Aug 2012 14:52:20 -0700 Subject: perf tools: do not flush maps on COMM for perf report This fixes a long-standing bug caused by the lack of separate COMM and EXEC record types, which makes "perf report" lose track of symbols when a process renames itself. With this fix (suggested by Stephane Eranian), a COMM (rename) no longer flushes the maps, which is the correct behavior. An EXEC also no longer flushes the maps, but this doesn't matter because as new mappings are created (for the executable and the libraries) the old mappings are automatically removed. This is not by accident: the functionality is necessary because DLLs can be explicitly loaded at any time with dlopen(), possibly on top of existing text, so "perf report" handles correctly the clobbering of new mappings on top of old ones. An alternative patch (which I proposed earlier) would be to introduce a separate PERF_RECORD_EXEC type, but it is a much larger change (about 300 lines) and is not necessary. Signed-off-by: Luigi Semenzato Tested-by: Stephane Eranian Acked-by: Stephane Eranian Cc: "Eric W. Biederman" Cc: "Rafael J. Wysocki" Cc: Alexander Viro Cc: Andi Kleen Cc: Andrew Morton Cc: David Ahern Cc: Frederic Weisbecker Cc: Greg Kroah-Hartman Cc: Ingo Molnar Cc: Lucas De Marchi Cc: Namhyung Kim Cc: Oleg Nesterov Cc: Olof Johansson Cc: Paul Gortmaker Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Robert Richter Cc: Sonny Rao Cc: Stephane Eranian Cc: Stephen Wilson Cc: Tejun Heo Cc: Vasiliy Kulikov Link: http://lkml.kernel.org/r/1345585940-6497-1-git-send-email-semenzato@chromium.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/thread.c | 1 - 1 file changed, 1 deletion(-) (limited to 'tools') diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index fb4b7ea6752f..8b3e5939afb6 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -39,7 +39,6 @@ int thread__set_comm(struct thread *self, const char *comm) err = self->comm == NULL ? -ENOMEM : 0; if (!err) { self->comm_set = true; - map_groups__flush(&self->mg); } return err; } -- cgit v1.3-18-gd494 From 42be73989c18931a394515b83aeae9246aa8b7b9 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sat, 20 Oct 2012 18:29:34 +0200 Subject: perf test: Fix exclude_guest parse events tests Event parsing tests are broken by following commit: perf tool: Precise mode requires exclude_guest commit 1342798cc13e3b48d9b5738f0c8fa812ccea8101 Author: David Ahern Date: Thu Sep 13 14:59:13 2012 -0600 which enables 'exclude_guest' modifier any time the 'precise' modifier is detected. Fixing related tests and adding special comment. Reported-by: Arnaldo Carvalho de Melo Signed-off-by: Jiri Olsa Tested-by: David Ahern Acked-by: David Ahern Cc: David Ahern Cc: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/parse-events-test.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'tools') diff --git a/tools/perf/util/parse-events-test.c b/tools/perf/util/parse-events-test.c index 28c18d1d52c3..516ecd9ddd6e 100644 --- a/tools/perf/util/parse-events-test.c +++ b/tools/perf/util/parse-events-test.c @@ -513,7 +513,8 @@ static int test__group1(struct perf_evlist *evlist) TEST_ASSERT_VAL("wrong exclude_user", !evsel->attr.exclude_user); TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); - TEST_ASSERT_VAL("wrong exclude guest", !evsel->attr.exclude_guest); + /* use of precise requires exclude_guest */ + TEST_ASSERT_VAL("wrong exclude guest", evsel->attr.exclude_guest); TEST_ASSERT_VAL("wrong exclude host", !evsel->attr.exclude_host); TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip == 2); TEST_ASSERT_VAL("wrong leader", evsel->leader == leader); @@ -599,7 +600,8 @@ static int test__group3(struct perf_evlist *evlist __maybe_unused) TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); - TEST_ASSERT_VAL("wrong exclude guest", !evsel->attr.exclude_guest); + /* use of precise requires exclude_guest */ + TEST_ASSERT_VAL("wrong exclude guest", evsel->attr.exclude_guest); TEST_ASSERT_VAL("wrong exclude host", !evsel->attr.exclude_host); TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip == 3); TEST_ASSERT_VAL("wrong leader", evsel->leader == leader); @@ -662,7 +664,8 @@ static int test__group4(struct perf_evlist *evlist __maybe_unused) TEST_ASSERT_VAL("wrong exclude_user", !evsel->attr.exclude_user); TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); - TEST_ASSERT_VAL("wrong exclude guest", !evsel->attr.exclude_guest); + /* use of precise requires exclude_guest */ + TEST_ASSERT_VAL("wrong exclude guest", evsel->attr.exclude_guest); TEST_ASSERT_VAL("wrong exclude host", !evsel->attr.exclude_host); TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip == 1); TEST_ASSERT_VAL("wrong group name", !evsel->group_name); @@ -676,7 +679,8 @@ static int test__group4(struct perf_evlist *evlist __maybe_unused) TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); - TEST_ASSERT_VAL("wrong exclude guest", !evsel->attr.exclude_guest); + /* use of precise requires exclude_guest */ + TEST_ASSERT_VAL("wrong exclude guest", evsel->attr.exclude_guest); TEST_ASSERT_VAL("wrong exclude host", !evsel->attr.exclude_host); TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip == 2); TEST_ASSERT_VAL("wrong leader", evsel->leader == leader); -- cgit v1.3-18-gd494 From 59ce8764bdfe8f3c6c02d3215741584dbb43409d Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 25 Oct 2012 13:37:57 -0700 Subject: UAPI: fix tools/vm/page-types.c Fix tools/vm/page-types.c to use the UAPI variant of linux/kernel-page-flags.h lest the following error appear: In file included from page-types.c:38:0: ../../include/linux/kernel-page-flags.h:4:42: fatal error: uapi/linux/kernel-page-flags.h: No such file or directory Reported-by: Daniel Hazelton Signed-off-by: David Howells Reviewed-by: Fengguang Wu Tested-by: Daniel Hazelton Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- tools/vm/page-types.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools') diff --git a/tools/vm/page-types.c b/tools/vm/page-types.c index cd1b03e80899..b76edf2f8333 100644 --- a/tools/vm/page-types.c +++ b/tools/vm/page-types.c @@ -35,7 +35,7 @@ #include #include #include "../../include/uapi/linux/magic.h" -#include "../../include/linux/kernel-page-flags.h" +#include "../../include/uapi/linux/kernel-page-flags.h" #ifndef MAX_PATH -- cgit v1.3-18-gd494 From fc314d0a4a933603f521de343634910a4ed9b37b Mon Sep 17 00:00:00 2001 From: Daniel Hazelton Date: Thu, 25 Oct 2012 13:37:59 -0700 Subject: tools/testing/selftests/epoll/test_epoll.c: fix build Latest Linus head run of "make selftests" in the tools directory failed with references to undefined variables. Reference was to 'write_thread_data' which is the name of a struct that is being used, not the variable itself. Change reference so it points to the variable. Signed-off-by: Daniel Hazelton Cc: "Paton J. Lewis" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- tools/testing/selftests/epoll/test_epoll.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools') diff --git a/tools/testing/selftests/epoll/test_epoll.c b/tools/testing/selftests/epoll/test_epoll.c index e0fcff1e8331..f7525392ce84 100644 --- a/tools/testing/selftests/epoll/test_epoll.c +++ b/tools/testing/selftests/epoll/test_epoll.c @@ -162,14 +162,14 @@ void *write_thread_function(void *function_data) int index; struct write_thread_data *thread_data = (struct write_thread_data *)function_data; - while (!write_thread_data->stop) + while (!thread_data->stop) for (index = 0; !thread_data->stop && (index < thread_data->n_fds); ++index) if ((write(thread_data->fds[index], &data, 1) < 1) && (errno != EAGAIN) && (errno != EWOULDBLOCK)) { - write_thread_data->status = errno; + thread_data->status = errno; return; } } -- cgit v1.3-18-gd494 From 8bc5e4ea3ea0e24142db2dc941233eab2a223ed4 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Fri, 26 Oct 2012 00:10:32 -0400 Subject: ktest: Fix ktest confusion with CONFIG_MODULES_USE_ELF_RELA In order to decide if ktest should bother installing modules on the target box, it checks if the config file has CONFIG_MODULES=y. But it also checks if the '=y' part exists. It only will install modules if the config exists and is set with '=y'. But as the regex that was used tests: /^CONFIG_MODULES(=y)?/ this will also match: CONFIG_MODULES_USE_ELF_RELA as the '=y' part was optional and it did not test the rest of the line. When this happens, ktest will stop checking the rest of the configs but it will also think that no modules are needed to be installed. What it should do is only jump out of the loop if it actually found a CONFIG_MODULES that is set to true. Otherwise, ktest wont install the necessary modules needed for proper booting of the test target. Signed-off-by: Steven Rostedt --- tools/testing/ktest/ktest.pl | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'tools') diff --git a/tools/testing/ktest/ktest.pl b/tools/testing/ktest/ktest.pl index b51d787176d3..c7ba7614061b 100755 --- a/tools/testing/ktest/ktest.pl +++ b/tools/testing/ktest/ktest.pl @@ -1740,8 +1740,10 @@ sub install { open(IN, "$output_config") or dodie("Can't read config file"); while () { if (/CONFIG_MODULES(=y)?/) { - $install_mods = 1 if (defined($1)); - last; + if (defined($1)) { + $install_mods = 1; + last; + } } } close(IN); -- cgit v1.3-18-gd494 From a80a6b85b428e6ce12a8363bb1f08d44c50f3252 Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Thu, 8 Nov 2012 15:53:35 -0800 Subject: revert "epoll: support for disabling items, and a self-test app" Revert commit 03a7beb55b9f ("epoll: support for disabling items, and a self-test app") pending resolution of the issues identified by Michael Kerrisk, copied below. We'll revisit this for 3.8. : I've taken a look at this patch as it currently stands in 3.7-rc1, and : done a bit of testing. (By the way, the test program : tools/testing/selftests/epoll/test_epoll.c does not compile...) : : There are one or two places where the behavior seems a little strange, : so I have a question or two at the end of this mail. But other than : that, I want to check my understanding so that the interface can be : correctly documented. : : Just to go though my understanding, the problem is the following : scenario in a multithreaded application: : : 1. Multiple threads are performing epoll_wait() operations, : and maintaining a user-space cache that contains information : corresponding to each file descriptor being monitored by : epoll_wait(). : : 2. At some point, a thread wants to delete (EPOLL_CTL_DEL) : a file descriptor from the epoll interest list, and : delete the corresponding record from the user-space cache. : : 3. The problem with (2) is that some other thread may have : previously done an epoll_wait() that retrieved information : about the fd in question, and may be in the middle of using : information in the cache that relates to that fd. Thus, : there is a potential race. : : 4. The race can't solved purely in user space, because doing : so would require applying a mutex across the epoll_wait() : call, which would of course blow thread concurrency. : : Right? : : Your solution is the EPOLL_CTL_DISABLE operation. I want to : confirm my understanding about how to use this flag, since : the description that has accompanied the patches so far : has been a bit sparse : : 0. In the scenario you're concerned about, deleting a file : descriptor means (safely) doing the following: : (a) Deleting the file descriptor from the epoll interest list : using EPOLL_CTL_DEL : (b) Deleting the corresponding record in the user-space cache : : 1. It's only meaningful to use this EPOLL_CTL_DISABLE in : conjunction with EPOLLONESHOT. : : 2. Using EPOLL_CTL_DISABLE without using EPOLLONESHOT in : conjunction is a logical error. : : 3. The correct way to code multithreaded applications using : EPOLL_CTL_DISABLE and EPOLLONESHOT is as follows: : : a. All EPOLL_CTL_ADD and EPOLL_CTL_MOD operations should : should EPOLLONESHOT. : : b. When a thread wants to delete a file descriptor, it : should do the following: : : [1] Call epoll_ctl(EPOLL_CTL_DISABLE) : [2] If the return status from epoll_ctl(EPOLL_CTL_DISABLE) : was zero, then the file descriptor can be safely : deleted by the thread that made this call. : [3] If the epoll_ctl(EPOLL_CTL_DISABLE) fails with EBUSY, : then the descriptor is in use. In this case, the calling : thread should set a flag in the user-space cache to : indicate that the thread that is using the descriptor : should perform the deletion operation. : : Is all of the above correct? : : The implementation depends on checking on whether : (events & ~EP_PRIVATE_BITS) == 0 : This replies on the fact that EPOLL_CTL_AD and EPOLL_CTL_MOD always : set EPOLLHUP and EPOLLERR in the 'events' mask, and EPOLLONESHOT : causes those flags (as well as all others in ~EP_PRIVATE_BITS) to be : cleared. : : A corollary to the previous paragraph is that using EPOLL_CTL_DISABLE : is only useful in conjunction with EPOLLONESHOT. However, as things : stand, one can use EPOLL_CTL_DISABLE on a file descriptor that does : not have EPOLLONESHOT set in 'events' This results in the following : (slightly surprising) behavior: : : (a) The first call to epoll_ctl(EPOLL_CTL_DISABLE) returns 0 : (the indicator that the file descriptor can be safely deleted). : (b) The next call to epoll_ctl(EPOLL_CTL_DISABLE) fails with EBUSY. : : This doesn't seem particularly useful, and in fact is probably an : indication that the user made a logic error: they should only be using : epoll_ctl(EPOLL_CTL_DISABLE) on a file descriptor for which : EPOLLONESHOT was set in 'events'. If that is correct, then would it : not make sense to return an error to user space for this case? Cc: Michael Kerrisk Cc: "Paton J. Lewis" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/eventpoll.c | 38 +--- include/uapi/linux/eventpoll.h | 1 - tools/testing/selftests/Makefile | 2 +- tools/testing/selftests/epoll/Makefile | 11 - tools/testing/selftests/epoll/test_epoll.c | 344 ----------------------------- 5 files changed, 4 insertions(+), 392 deletions(-) delete mode 100644 tools/testing/selftests/epoll/Makefile delete mode 100644 tools/testing/selftests/epoll/test_epoll.c (limited to 'tools') diff --git a/fs/eventpoll.c b/fs/eventpoll.c index da72250ddc1c..cd96649bfe62 100644 --- a/fs/eventpoll.c +++ b/fs/eventpoll.c @@ -346,7 +346,7 @@ static inline struct epitem *ep_item_from_epqueue(poll_table *p) /* Tells if the epoll_ctl(2) operation needs an event copy from userspace */ static inline int ep_op_has_event(int op) { - return op == EPOLL_CTL_ADD || op == EPOLL_CTL_MOD; + return op != EPOLL_CTL_DEL; } /* Initialize the poll safe wake up structure */ @@ -676,34 +676,6 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi) return 0; } -/* - * Disables a "struct epitem" in the eventpoll set. Returns -EBUSY if the item - * had no event flags set, indicating that another thread may be currently - * handling that item's events (in the case that EPOLLONESHOT was being - * used). Otherwise a zero result indicates that the item has been disabled - * from receiving events. A disabled item may be re-enabled via - * EPOLL_CTL_MOD. Must be called with "mtx" held. - */ -static int ep_disable(struct eventpoll *ep, struct epitem *epi) -{ - int result = 0; - unsigned long flags; - - spin_lock_irqsave(&ep->lock, flags); - if (epi->event.events & ~EP_PRIVATE_BITS) { - if (ep_is_linked(&epi->rdllink)) - list_del_init(&epi->rdllink); - /* Ensure ep_poll_callback will not add epi back onto ready - list: */ - epi->event.events &= EP_PRIVATE_BITS; - } - else - result = -EBUSY; - spin_unlock_irqrestore(&ep->lock, flags); - - return result; -} - static void ep_free(struct eventpoll *ep) { struct rb_node *rbp; @@ -1048,6 +1020,8 @@ static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi) rb_insert_color(&epi->rbn, &ep->rbr); } + + #define PATH_ARR_SIZE 5 /* * These are the number paths of length 1 to 5, that we are allowing to emanate @@ -1813,12 +1787,6 @@ SYSCALL_DEFINE4(epoll_ctl, int, epfd, int, op, int, fd, } else error = -ENOENT; break; - case EPOLL_CTL_DISABLE: - if (epi) - error = ep_disable(ep, epi); - else - error = -ENOENT; - break; } mutex_unlock(&ep->mtx); diff --git a/include/uapi/linux/eventpoll.h b/include/uapi/linux/eventpoll.h index 8c99ce7202c5..2c267bcbb85c 100644 --- a/include/uapi/linux/eventpoll.h +++ b/include/uapi/linux/eventpoll.h @@ -25,7 +25,6 @@ #define EPOLL_CTL_ADD 1 #define EPOLL_CTL_DEL 2 #define EPOLL_CTL_MOD 3 -#define EPOLL_CTL_DISABLE 4 /* * Request the handling of system wakeup events so as to prevent system suspends diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile index 43480149119e..85baf11e2acd 100644 --- a/tools/testing/selftests/Makefile +++ b/tools/testing/selftests/Makefile @@ -1,4 +1,4 @@ -TARGETS = breakpoints kcmp mqueue vm cpu-hotplug memory-hotplug epoll +TARGETS = breakpoints kcmp mqueue vm cpu-hotplug memory-hotplug all: for TARGET in $(TARGETS); do \ diff --git a/tools/testing/selftests/epoll/Makefile b/tools/testing/selftests/epoll/Makefile deleted file mode 100644 index 19806ed62f50..000000000000 --- a/tools/testing/selftests/epoll/Makefile +++ /dev/null @@ -1,11 +0,0 @@ -# Makefile for epoll selftests - -all: test_epoll -%: %.c - gcc -pthread -g -o $@ $^ - -run_tests: all - ./test_epoll - -clean: - $(RM) test_epoll diff --git a/tools/testing/selftests/epoll/test_epoll.c b/tools/testing/selftests/epoll/test_epoll.c deleted file mode 100644 index f7525392ce84..000000000000 --- a/tools/testing/selftests/epoll/test_epoll.c +++ /dev/null @@ -1,344 +0,0 @@ -/* - * tools/testing/selftests/epoll/test_epoll.c - * - * Copyright 2012 Adobe Systems Incorporated - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * Paton J. Lewis - * - */ - -#include -#include -#include -#include -#include -#include -#include -#include - -/* - * A pointer to an epoll_item_private structure will be stored in the epoll - * item's event structure so that we can get access to the epoll_item_private - * data after calling epoll_wait: - */ -struct epoll_item_private { - int index; /* Position of this struct within the epoll_items array. */ - int fd; - uint32_t events; - pthread_mutex_t mutex; /* Guards the following variables... */ - int stop; - int status; /* Stores any error encountered while handling item. */ - /* The following variable allows us to test whether we have encountered - a problem while attempting to cancel and delete the associated - event. When the test program exits, 'deleted' should be exactly - one. If it is greater than one, then the failed test reflects a real - world situation where we would have tried to access the epoll item's - private data after deleting it: */ - int deleted; -}; - -struct epoll_item_private *epoll_items; - -/* - * Delete the specified item from the epoll set. In a real-world secneario this - * is where we would free the associated data structure, but in this testing - * environment we retain the structure so that we can test for double-deletion: - */ -void delete_item(int index) -{ - __sync_fetch_and_add(&epoll_items[index].deleted, 1); -} - -/* - * A pointer to a read_thread_data structure will be passed as the argument to - * each read thread: - */ -struct read_thread_data { - int stop; - int status; /* Indicates any error encountered by the read thread. */ - int epoll_set; -}; - -/* - * The function executed by the read threads: - */ -void *read_thread_function(void *function_data) -{ - struct read_thread_data *thread_data = - (struct read_thread_data *)function_data; - struct epoll_event event_data; - struct epoll_item_private *item_data; - char socket_data; - - /* Handle events until we encounter an error or this thread's 'stop' - condition is set: */ - while (1) { - int result = epoll_wait(thread_data->epoll_set, - &event_data, - 1, /* Number of desired events */ - 1000); /* Timeout in ms */ - if (result < 0) { - /* Breakpoints signal all threads. Ignore that while - debugging: */ - if (errno == EINTR) - continue; - thread_data->status = errno; - return 0; - } else if (thread_data->stop) - return 0; - else if (result == 0) /* Timeout */ - continue; - - /* We need the mutex here because checking for the stop - condition and re-enabling the epoll item need to be done - together as one atomic operation when EPOLL_CTL_DISABLE is - available: */ - item_data = (struct epoll_item_private *)event_data.data.ptr; - pthread_mutex_lock(&item_data->mutex); - - /* Remove the item from the epoll set if we want to stop - handling that event: */ - if (item_data->stop) - delete_item(item_data->index); - else { - /* Clear the data that was written to the other end of - our non-blocking socket: */ - do { - if (read(item_data->fd, &socket_data, 1) < 1) { - if ((errno == EAGAIN) || - (errno == EWOULDBLOCK)) - break; - else - goto error_unlock; - } - } while (item_data->events & EPOLLET); - - /* The item was one-shot, so re-enable it: */ - event_data.events = item_data->events; - if (epoll_ctl(thread_data->epoll_set, - EPOLL_CTL_MOD, - item_data->fd, - &event_data) < 0) - goto error_unlock; - } - - pthread_mutex_unlock(&item_data->mutex); - } - -error_unlock: - thread_data->status = item_data->status = errno; - pthread_mutex_unlock(&item_data->mutex); - return 0; -} - -/* - * A pointer to a write_thread_data structure will be passed as the argument to - * the write thread: - */ -struct write_thread_data { - int stop; - int status; /* Indicates any error encountered by the write thread. */ - int n_fds; - int *fds; -}; - -/* - * The function executed by the write thread. It writes a single byte to each - * socket in turn until the stop condition for this thread is set. If writing to - * a socket would block (i.e. errno was EAGAIN), we leave that socket alone for - * the moment and just move on to the next socket in the list. We don't care - * about the order in which we deliver events to the epoll set. In fact we don't - * care about the data we're writing to the pipes at all; we just want to - * trigger epoll events: - */ -void *write_thread_function(void *function_data) -{ - const char data = 'X'; - int index; - struct write_thread_data *thread_data = - (struct write_thread_data *)function_data; - while (!thread_data->stop) - for (index = 0; - !thread_data->stop && (index < thread_data->n_fds); - ++index) - if ((write(thread_data->fds[index], &data, 1) < 1) && - (errno != EAGAIN) && - (errno != EWOULDBLOCK)) { - thread_data->status = errno; - return; - } -} - -/* - * Arguments are currently ignored: - */ -int main(int argc, char **argv) -{ - const int n_read_threads = 100; - const int n_epoll_items = 500; - int index; - int epoll_set = epoll_create1(0); - struct write_thread_data write_thread_data = { - 0, 0, n_epoll_items, malloc(n_epoll_items * sizeof(int)) - }; - struct read_thread_data *read_thread_data = - malloc(n_read_threads * sizeof(struct read_thread_data)); - pthread_t *read_threads = malloc(n_read_threads * sizeof(pthread_t)); - pthread_t write_thread; - - printf("-----------------\n"); - printf("Runing test_epoll\n"); - printf("-----------------\n"); - - epoll_items = malloc(n_epoll_items * sizeof(struct epoll_item_private)); - - if (epoll_set < 0 || epoll_items == 0 || write_thread_data.fds == 0 || - read_thread_data == 0 || read_threads == 0) - goto error; - - if (sysconf(_SC_NPROCESSORS_ONLN) < 2) { - printf("Error: please run this test on a multi-core system.\n"); - goto error; - } - - /* Create the socket pairs and epoll items: */ - for (index = 0; index < n_epoll_items; ++index) { - int socket_pair[2]; - struct epoll_event event_data; - if (socketpair(AF_UNIX, - SOCK_STREAM | SOCK_NONBLOCK, - 0, - socket_pair) < 0) - goto error; - write_thread_data.fds[index] = socket_pair[0]; - epoll_items[index].index = index; - epoll_items[index].fd = socket_pair[1]; - if (pthread_mutex_init(&epoll_items[index].mutex, NULL) != 0) - goto error; - /* We always use EPOLLONESHOT because this test is currently - structured to demonstrate the need for EPOLL_CTL_DISABLE, - which only produces useful information in the EPOLLONESHOT - case (without EPOLLONESHOT, calling epoll_ctl with - EPOLL_CTL_DISABLE will never return EBUSY). If support for - testing events without EPOLLONESHOT is desired, it should - probably be implemented in a separate unit test. */ - epoll_items[index].events = EPOLLIN | EPOLLONESHOT; - if (index < n_epoll_items / 2) - epoll_items[index].events |= EPOLLET; - epoll_items[index].stop = 0; - epoll_items[index].status = 0; - epoll_items[index].deleted = 0; - event_data.events = epoll_items[index].events; - event_data.data.ptr = &epoll_items[index]; - if (epoll_ctl(epoll_set, - EPOLL_CTL_ADD, - epoll_items[index].fd, - &event_data) < 0) - goto error; - } - - /* Create and start the read threads: */ - for (index = 0; index < n_read_threads; ++index) { - read_thread_data[index].stop = 0; - read_thread_data[index].status = 0; - read_thread_data[index].epoll_set = epoll_set; - if (pthread_create(&read_threads[index], - NULL, - read_thread_function, - &read_thread_data[index]) != 0) - goto error; - } - - if (pthread_create(&write_thread, - NULL, - write_thread_function, - &write_thread_data) != 0) - goto error; - - /* Cancel all event pollers: */ -#ifdef EPOLL_CTL_DISABLE - for (index = 0; index < n_epoll_items; ++index) { - pthread_mutex_lock(&epoll_items[index].mutex); - ++epoll_items[index].stop; - if (epoll_ctl(epoll_set, - EPOLL_CTL_DISABLE, - epoll_items[index].fd, - NULL) == 0) - delete_item(index); - else if (errno != EBUSY) { - pthread_mutex_unlock(&epoll_items[index].mutex); - goto error; - } - /* EBUSY means events were being handled; allow the other thread - to delete the item. */ - pthread_mutex_unlock(&epoll_items[index].mutex); - } -#else - for (index = 0; index < n_epoll_items; ++index) { - pthread_mutex_lock(&epoll_items[index].mutex); - ++epoll_items[index].stop; - pthread_mutex_unlock(&epoll_items[index].mutex); - /* Wait in case a thread running read_thread_function is - currently executing code between epoll_wait and - pthread_mutex_lock with this item. Note that a longer delay - would make double-deletion less likely (at the expense of - performance), but there is no guarantee that any delay would - ever be sufficient. Note also that we delete all event - pollers at once for testing purposes, but in a real-world - environment we are likely to want to be able to cancel event - pollers at arbitrary times. Therefore we can't improve this - situation by just splitting this loop into two loops - (i.e. signal 'stop' for all items, sleep, and then delete all - items). We also can't fix the problem via EPOLL_CTL_DEL - because that command can't prevent the case where some other - thread is executing read_thread_function within the region - mentioned above: */ - usleep(1); - pthread_mutex_lock(&epoll_items[index].mutex); - if (!epoll_items[index].deleted) - delete_item(index); - pthread_mutex_unlock(&epoll_items[index].mutex); - } -#endif - - /* Shut down the read threads: */ - for (index = 0; index < n_read_threads; ++index) - __sync_fetch_and_add(&read_thread_data[index].stop, 1); - for (index = 0; index < n_read_threads; ++index) { - if (pthread_join(read_threads[index], NULL) != 0) - goto error; - if (read_thread_data[index].status) - goto error; - } - - /* Shut down the write thread: */ - __sync_fetch_and_add(&write_thread_data.stop, 1); - if ((pthread_join(write_thread, NULL) != 0) || write_thread_data.status) - goto error; - - /* Check for final error conditions: */ - for (index = 0; index < n_epoll_items; ++index) { - if (epoll_items[index].status != 0) - goto error; - if (pthread_mutex_destroy(&epoll_items[index].mutex) < 0) - goto error; - } - for (index = 0; index < n_epoll_items; ++index) - if (epoll_items[index].deleted != 1) { - printf("Error: item data deleted %1d times.\n", - epoll_items[index].deleted); - goto error; - } - - printf("[PASS]\n"); - return 0; - - error: - printf("[FAIL]\n"); - return errno; -} -- cgit v1.3-18-gd494