From 8feae5adec17ec190118ad84da84a6ad9fa3e30b Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Tue, 7 Mar 2023 23:32:00 -0800 Subject: selftest/seccomp: add a new test for the sync mode of seccomp_user_notify Test output: # RUN global.user_notification_sync ... # OK global.user_notification_sync ok 51 global.user_notification_sync Signed-off-by: Andrei Vagin Acked-by: "Peter Zijlstra (Intel)" Link: https://lore.kernel.org/r/20230308073201.3102738-6-avagin@google.com Signed-off-by: Kees Cook --- tools/testing/selftests/seccomp/seccomp_bpf.c | 55 +++++++++++++++++++++++++++ 1 file changed, 55 insertions(+) (limited to 'tools') diff --git a/tools/testing/selftests/seccomp/seccomp_bpf.c b/tools/testing/selftests/seccomp/seccomp_bpf.c index 43ec36b179dc..f6a04d88e02f 100644 --- a/tools/testing/selftests/seccomp/seccomp_bpf.c +++ b/tools/testing/selftests/seccomp/seccomp_bpf.c @@ -4255,6 +4255,61 @@ TEST(user_notification_addfd_rlimit) close(memfd); } +#ifndef SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP +#define SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP (1UL << 0) +#define SECCOMP_IOCTL_NOTIF_SET_FLAGS SECCOMP_IOW(4, __u64) +#endif + +TEST(user_notification_sync) +{ + struct seccomp_notif req = {}; + struct seccomp_notif_resp resp = {}; + int status, listener; + pid_t pid; + long ret; + + ret = prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0); + ASSERT_EQ(0, ret) { + TH_LOG("Kernel does not support PR_SET_NO_NEW_PRIVS!"); + } + + listener = user_notif_syscall(__NR_getppid, + SECCOMP_FILTER_FLAG_NEW_LISTENER); + ASSERT_GE(listener, 0); + + /* Try to set invalid flags. */ + EXPECT_SYSCALL_RETURN(-EINVAL, + ioctl(listener, SECCOMP_IOCTL_NOTIF_SET_FLAGS, 0xffffffff, 0)); + + ASSERT_EQ(ioctl(listener, SECCOMP_IOCTL_NOTIF_SET_FLAGS, + SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP, 0), 0); + + pid = fork(); + ASSERT_GE(pid, 0); + if (pid == 0) { + ret = syscall(__NR_getppid); + ASSERT_EQ(ret, USER_NOTIF_MAGIC) { + _exit(1); + } + _exit(0); + } + + req.pid = 0; + ASSERT_EQ(ioctl(listener, SECCOMP_IOCTL_NOTIF_RECV, &req), 0); + + ASSERT_EQ(req.data.nr, __NR_getppid); + + resp.id = req.id; + resp.error = 0; + resp.val = USER_NOTIF_MAGIC; + resp.flags = 0; + ASSERT_EQ(ioctl(listener, SECCOMP_IOCTL_NOTIF_SEND, &resp), 0); + + ASSERT_EQ(waitpid(pid, &status, 0), pid); + ASSERT_EQ(status, 0); +} + + /* Make sure PTRACE_O_SUSPEND_SECCOMP requires CAP_SYS_ADMIN. */ FIXTURE(O_SUSPEND_SECCOMP) { pid_t pid; -- cgit v1.2.3 From 7d5cb68af638fc370e89df4c2d8d8c4708600e67 Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Tue, 7 Mar 2023 23:32:01 -0800 Subject: perf/benchmark: add a new benchmark for seccom_unotify The benchmark is similar to the pipe benchmark. It creates two processes, one is calling syscalls, and another process is handling them via seccomp user notifications. It measures the time required to run a specified number of interations. $ ./perf bench sched seccomp-notify --sync-mode --loop 1000000 # Running 'sched/seccomp-notify' benchmark: # Executed 1000000 system calls Total time: 2.769 [sec] 2.769629 usecs/op 361059 ops/sec $ ./perf bench sched seccomp-notify # Running 'sched/seccomp-notify' benchmark: # Executed 1000000 system calls Total time: 8.571 [sec] 8.571119 usecs/op 116670 ops/sec Signed-off-by: Andrei Vagin Acked-by: "Peter Zijlstra (Intel)" Link: https://lore.kernel.org/r/20230308073201.3102738-7-avagin@google.com Link: https://lore.kernel.org/r/20230630051953.454638-1-avagin@gmail.com [kees: Added PRIu64 format string] Signed-off-by: Kees Cook --- tools/arch/x86/include/uapi/asm/unistd_32.h | 3 + tools/arch/x86/include/uapi/asm/unistd_64.h | 3 + tools/perf/bench/Build | 1 + tools/perf/bench/bench.h | 1 + tools/perf/bench/sched-seccomp-notify.c | 178 ++++++++++++++++++++++++++++ tools/perf/builtin-bench.c | 1 + 6 files changed, 187 insertions(+) create mode 100644 tools/perf/bench/sched-seccomp-notify.c (limited to 'tools') diff --git a/tools/arch/x86/include/uapi/asm/unistd_32.h b/tools/arch/x86/include/uapi/asm/unistd_32.h index bc48a4dabe5d..4798f9d18fe8 100644 --- a/tools/arch/x86/include/uapi/asm/unistd_32.h +++ b/tools/arch/x86/include/uapi/asm/unistd_32.h @@ -26,3 +26,6 @@ #ifndef __NR_setns #define __NR_setns 346 #endif +#ifdef __NR_seccomp +#define __NR_seccomp 354 +#endif diff --git a/tools/arch/x86/include/uapi/asm/unistd_64.h b/tools/arch/x86/include/uapi/asm/unistd_64.h index f70d2cada256..d0f2043d7132 100644 --- a/tools/arch/x86/include/uapi/asm/unistd_64.h +++ b/tools/arch/x86/include/uapi/asm/unistd_64.h @@ -26,3 +26,6 @@ #ifndef __NR_getcpu #define __NR_getcpu 309 #endif +#ifndef __NR_seccomp +#define __NR_seccomp 317 +#endif diff --git a/tools/perf/bench/Build b/tools/perf/bench/Build index 0f158dc8139b..07bbc449329e 100644 --- a/tools/perf/bench/Build +++ b/tools/perf/bench/Build @@ -1,5 +1,6 @@ perf-y += sched-messaging.o perf-y += sched-pipe.o +perf-y += sched-seccomp-notify.o perf-y += syscall.o perf-y += mem-functions.o perf-y += futex-hash.o diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index 0d2b65976212..a0625c77bea3 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -21,6 +21,7 @@ extern struct timeval bench__start, bench__end, bench__runtime; int bench_numa(int argc, const char **argv); int bench_sched_messaging(int argc, const char **argv); int bench_sched_pipe(int argc, const char **argv); +int bench_sched_seccomp_notify(int argc, const char **argv); int bench_syscall_basic(int argc, const char **argv); int bench_syscall_getpgid(int argc, const char **argv); int bench_syscall_fork(int argc, const char **argv); diff --git a/tools/perf/bench/sched-seccomp-notify.c b/tools/perf/bench/sched-seccomp-notify.c new file mode 100644 index 000000000000..b04ebcde4036 --- /dev/null +++ b/tools/perf/bench/sched-seccomp-notify.c @@ -0,0 +1,178 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include "bench.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define LOOPS_DEFAULT 1000000UL +static uint64_t loops = LOOPS_DEFAULT; +static bool sync_mode; + +static const struct option options[] = { + OPT_U64('l', "loop", &loops, "Specify number of loops"), + OPT_BOOLEAN('s', "sync-mode", &sync_mode, + "Enable the synchronious mode for seccomp notifications"), + OPT_END() +}; + +static const char * const bench_seccomp_usage[] = { + "perf bench sched secccomp-notify ", + NULL +}; + +static int seccomp(unsigned int op, unsigned int flags, void *args) +{ + return syscall(__NR_seccomp, op, flags, args); +} + +static int user_notif_syscall(int nr, unsigned int flags) +{ + struct sock_filter filter[] = { + BPF_STMT(BPF_LD|BPF_W|BPF_ABS, + offsetof(struct seccomp_data, nr)), + BPF_JUMP(BPF_JMP|BPF_JEQ|BPF_K, nr, 0, 1), + BPF_STMT(BPF_RET|BPF_K, SECCOMP_RET_USER_NOTIF), + BPF_STMT(BPF_RET|BPF_K, SECCOMP_RET_ALLOW), + }; + + struct sock_fprog prog = { + .len = (unsigned short)ARRAY_SIZE(filter), + .filter = filter, + }; + + return seccomp(SECCOMP_SET_MODE_FILTER, flags, &prog); +} + +#define USER_NOTIF_MAGIC INT_MAX +static void user_notification_sync_loop(int listener) +{ + struct seccomp_notif_resp resp; + struct seccomp_notif req; + uint64_t nr; + + for (nr = 0; nr < loops; nr++) { + memset(&req, 0, sizeof(req)); + if (ioctl(listener, SECCOMP_IOCTL_NOTIF_RECV, &req)) + err(EXIT_FAILURE, "SECCOMP_IOCTL_NOTIF_RECV failed"); + + if (req.data.nr != __NR_gettid) + errx(EXIT_FAILURE, "unexpected syscall: %d", req.data.nr); + + resp.id = req.id; + resp.error = 0; + resp.val = USER_NOTIF_MAGIC; + resp.flags = 0; + if (ioctl(listener, SECCOMP_IOCTL_NOTIF_SEND, &resp)) + err(EXIT_FAILURE, "SECCOMP_IOCTL_NOTIF_SEND failed"); + } +} + +#ifndef SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP +#define SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP (1UL << 0) +#define SECCOMP_IOCTL_NOTIF_SET_FLAGS SECCOMP_IOW(4, __u64) +#endif +int bench_sched_seccomp_notify(int argc, const char **argv) +{ + struct timeval start, stop, diff; + unsigned long long result_usec = 0; + int status, listener; + pid_t pid; + long ret; + + argc = parse_options(argc, argv, options, bench_seccomp_usage, 0); + + gettimeofday(&start, NULL); + + prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0); + listener = user_notif_syscall(__NR_gettid, + SECCOMP_FILTER_FLAG_NEW_LISTENER); + if (listener < 0) + err(EXIT_FAILURE, "can't create a notification descriptor"); + + pid = fork(); + if (pid < 0) + err(EXIT_FAILURE, "fork"); + if (pid == 0) { + if (prctl(PR_SET_PDEATHSIG, SIGKILL, 0, 0, 0)) + err(EXIT_FAILURE, "can't set the parent death signal"); + while (1) { + ret = syscall(__NR_gettid); + if (ret == USER_NOTIF_MAGIC) + continue; + break; + } + _exit(1); + } + + if (sync_mode) { + if (ioctl(listener, SECCOMP_IOCTL_NOTIF_SET_FLAGS, + SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP, 0)) + err(EXIT_FAILURE, + "can't set SECCOMP_USER_NOTIF_FD_SYNC_WAKE_UP"); + } + user_notification_sync_loop(listener); + + kill(pid, SIGKILL); + if (waitpid(pid, &status, 0) != pid) + err(EXIT_FAILURE, "waitpid(%d) failed", pid); + if (!WIFSIGNALED(status) || WTERMSIG(status) != SIGKILL) + errx(EXIT_FAILURE, "unexpected exit code: %d", status); + + gettimeofday(&stop, NULL); + timersub(&stop, &start, &diff); + + switch (bench_format) { + case BENCH_FORMAT_DEFAULT: + printf("# Executed %" PRIu64 " system calls\n\n", + loops); + + result_usec = diff.tv_sec * USEC_PER_SEC; + result_usec += diff.tv_usec; + + printf(" %14s: %lu.%03lu [sec]\n\n", "Total time", + (unsigned long) diff.tv_sec, + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); + + printf(" %14lf usecs/op\n", + (double)result_usec / (double)loops); + printf(" %14d ops/sec\n", + (int)((double)loops / + ((double)result_usec / (double)USEC_PER_SEC))); + break; + + case BENCH_FORMAT_SIMPLE: + printf("%lu.%03lu\n", + (unsigned long) diff.tv_sec, + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); + break; + + default: + /* reaching here is something disaster */ + fprintf(stderr, "Unknown format:%d\n", bench_format); + exit(1); + break; + } + + return 0; +} diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index db435b791a09..5033e8bab276 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -47,6 +47,7 @@ static struct bench numa_benchmarks[] = { static struct bench sched_benchmarks[] = { { "messaging", "Benchmark for scheduling and IPC", bench_sched_messaging }, { "pipe", "Benchmark for pipe() between two processes", bench_sched_pipe }, + { "seccomp-notify", "Benchmark for seccomp user notify", bench_sched_seccomp_notify}, { "all", "Run all scheduler benchmarks", NULL }, { NULL, NULL, NULL } }; -- cgit v1.2.3 From fbc5d382407eb5d6a2eeef245cc2ca278d590645 Mon Sep 17 00:00:00 2001 From: Kees Cook Date: Wed, 2 Aug 2023 21:04:32 -0700 Subject: selftests/seccomp: Handle arm32 corner cases better It turns out arm32 doesn't handle syscall -1 gracefully, so skip testing for that. Additionally skip tests that depend on clone3 when it is not available (for example when building the seccomp selftests on an old arm image without clone3 headers). And improve error reporting for when nanosleep fails, as seen on arm32 since v5.15. Cc: Lecopzer Chen Signed-off-by: Kees Cook --- tools/testing/selftests/seccomp/seccomp_bpf.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'tools') diff --git a/tools/testing/selftests/seccomp/seccomp_bpf.c b/tools/testing/selftests/seccomp/seccomp_bpf.c index f6a04d88e02f..38f651469968 100644 --- a/tools/testing/selftests/seccomp/seccomp_bpf.c +++ b/tools/testing/selftests/seccomp/seccomp_bpf.c @@ -2184,6 +2184,9 @@ FIXTURE_TEARDOWN(TRACE_syscall) TEST(negative_ENOSYS) { +#if defined(__arm__) + SKIP(return, "arm32 does not support calling syscall -1"); +#endif /* * There should be no difference between an "internal" skip * and userspace asking for syscall "-1". @@ -3072,7 +3075,8 @@ TEST(syscall_restart) timeout.tv_sec = 1; errno = 0; EXPECT_EQ(0, nanosleep(&timeout, NULL)) { - TH_LOG("Call to nanosleep() failed (errno %d)", errno); + TH_LOG("Call to nanosleep() failed (errno %d: %s)", + errno, strerror(errno)); } /* Read final sync from parent. */ @@ -3908,6 +3912,9 @@ TEST(user_notification_filter_empty) TH_LOG("Kernel does not support PR_SET_NO_NEW_PRIVS!"); } + if (__NR_clone3 < 0) + SKIP(return, "Test not built with clone3 support"); + pid = sys_clone3(&args, sizeof(args)); ASSERT_GE(pid, 0); @@ -3962,6 +3969,9 @@ TEST(user_notification_filter_empty_threaded) TH_LOG("Kernel does not support PR_SET_NO_NEW_PRIVS!"); } + if (__NR_clone3 < 0) + SKIP(return, "Test not built with clone3 support"); + pid = sys_clone3(&args, sizeof(args)); ASSERT_GE(pid, 0); -- cgit v1.2.3