tools/testing/selftests/bpf: Add self-tests for new helper bpf_get_ns_current_pid_tgid.
authorCarlos Neira <cneirabustos@gmail.com>
Wed, 4 Mar 2020 20:41:57 +0000 (17:41 -0300)
committerAlexei Starovoitov <ast@kernel.org>
Fri, 13 Mar 2020 00:40:47 +0000 (17:40 -0700)
Self tests added for new helper bpf_get_ns_current_pid_tgid

Signed-off-by: Carlos Neira <cneirabustos@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200304204157.58695-4-cneirabustos@gmail.com
tools/testing/selftests/bpf/.gitignore
tools/testing/selftests/bpf/Makefile
tools/testing/selftests/bpf/prog_tests/ns_current_pid_tgid.c [new file with mode: 0644]
tools/testing/selftests/bpf/progs/test_ns_current_pid_tgid.c [new file with mode: 0644]
tools/testing/selftests/bpf/test_current_pid_tgid_new_ns.c [new file with mode: 0644]

index ec46485..2198cd8 100644 (file)
@@ -31,6 +31,7 @@ test_tcp_check_syncookie_user
 test_sysctl
 test_hashmap
 test_btf_dump
+test_current_pid_tgid_new_ns
 xdping
 test_cpp
 *.skel.h
index ee4ad34..da4389d 100644 (file)
@@ -32,7 +32,8 @@ TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test
        test_sock test_btf test_sockmap get_cgroup_id_user test_socket_cookie \
        test_cgroup_storage \
        test_netcnt test_tcpnotify_user test_sock_fields test_sysctl test_hashmap \
-       test_progs-no_alu32
+       test_progs-no_alu32 \
+       test_current_pid_tgid_new_ns
 
 # Also test bpf-gcc, if present
 ifneq ($(BPF_GCC),)
diff --git a/tools/testing/selftests/bpf/prog_tests/ns_current_pid_tgid.c b/tools/testing/selftests/bpf/prog_tests/ns_current_pid_tgid.c
new file mode 100644 (file)
index 0000000..542240e
--- /dev/null
@@ -0,0 +1,88 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Carlos Neira cneirabustos@gmail.com */
+#include <test_progs.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+
+struct bss {
+       __u64 dev;
+       __u64 ino;
+       __u64 pid_tgid;
+       __u64 user_pid_tgid;
+};
+
+void test_ns_current_pid_tgid(void)
+{
+       const char *probe_name = "raw_tracepoint/sys_enter";
+       const char *file = "test_ns_current_pid_tgid.o";
+       int err, key = 0, duration = 0;
+       struct bpf_link *link = NULL;
+       struct bpf_program *prog;
+       struct bpf_map *bss_map;
+       struct bpf_object *obj;
+       struct bss bss;
+       struct stat st;
+       __u64 id;
+
+       obj = bpf_object__open_file(file, NULL);
+       if (CHECK(IS_ERR(obj), "obj_open", "err %ld\n", PTR_ERR(obj)))
+               return;
+
+       err = bpf_object__load(obj);
+       if (CHECK(err, "obj_load", "err %d errno %d\n", err, errno))
+               goto cleanup;
+
+       bss_map = bpf_object__find_map_by_name(obj, "test_ns_.bss");
+       if (CHECK(!bss_map, "find_bss_map", "failed\n"))
+               goto cleanup;
+
+       prog = bpf_object__find_program_by_title(obj, probe_name);
+       if (CHECK(!prog, "find_prog", "prog '%s' not found\n",
+                 probe_name))
+               goto cleanup;
+
+       memset(&bss, 0, sizeof(bss));
+       pid_t tid = syscall(SYS_gettid);
+       pid_t pid = getpid();
+
+       id = (__u64) tid << 32 | pid;
+       bss.user_pid_tgid = id;
+
+       if (CHECK_FAIL(stat("/proc/self/ns/pid", &st))) {
+               perror("Failed to stat /proc/self/ns/pid");
+               goto cleanup;
+       }
+
+       bss.dev = st.st_dev;
+       bss.ino = st.st_ino;
+
+       err = bpf_map_update_elem(bpf_map__fd(bss_map), &key, &bss, 0);
+       if (CHECK(err, "setting_bss", "failed to set bss : %d\n", err))
+               goto cleanup;
+
+       link = bpf_program__attach_raw_tracepoint(prog, "sys_enter");
+       if (CHECK(IS_ERR(link), "attach_raw_tp", "err %ld\n",
+                 PTR_ERR(link))) {
+               link = NULL;
+               goto cleanup;
+       }
+
+       /* trigger some syscalls */
+       usleep(1);
+
+       err = bpf_map_lookup_elem(bpf_map__fd(bss_map), &key, &bss);
+       if (CHECK(err, "set_bss", "failed to get bss : %d\n", err))
+               goto cleanup;
+
+       if (CHECK(id != bss.pid_tgid, "Compare user pid/tgid vs. bpf pid/tgid",
+                 "User pid/tgid %llu BPF pid/tgid %llu\n", id, bss.pid_tgid))
+               goto cleanup;
+cleanup:
+       if (!link) {
+               bpf_link__destroy(link);
+               link = NULL;
+       }
+       bpf_object__close(obj);
+}
diff --git a/tools/testing/selftests/bpf/progs/test_ns_current_pid_tgid.c b/tools/testing/selftests/bpf/progs/test_ns_current_pid_tgid.c
new file mode 100644 (file)
index 0000000..1dca70a
--- /dev/null
@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019 Carlos Neira cneirabustos@gmail.com */
+
+#include <linux/bpf.h>
+#include <stdint.h>
+#include <bpf/bpf_helpers.h>
+
+static volatile struct {
+       __u64 dev;
+       __u64 ino;
+       __u64 pid_tgid;
+       __u64 user_pid_tgid;
+} res;
+
+SEC("raw_tracepoint/sys_enter")
+int trace(void *ctx)
+{
+       __u64  ns_pid_tgid, expected_pid;
+       struct bpf_pidns_info nsdata;
+       __u32 key = 0;
+
+       if (bpf_get_ns_current_pid_tgid(res.dev, res.ino, &nsdata,
+                  sizeof(struct bpf_pidns_info)))
+               return 0;
+
+       ns_pid_tgid = (__u64)nsdata.tgid << 32 | nsdata.pid;
+       expected_pid = res.user_pid_tgid;
+
+       if (expected_pid != ns_pid_tgid)
+               return 0;
+
+       res.pid_tgid = ns_pid_tgid;
+
+       return 0;
+}
+
+char _license[] SEC("license") = "GPL";
diff --git a/tools/testing/selftests/bpf/test_current_pid_tgid_new_ns.c b/tools/testing/selftests/bpf/test_current_pid_tgid_new_ns.c
new file mode 100644 (file)
index 0000000..ed253f2
--- /dev/null
@@ -0,0 +1,159 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Carlos Neira cneirabustos@gmail.com */
+#define _GNU_SOURCE
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <sched.h>
+#include <sys/wait.h>
+#include <sys/mount.h>
+#include "test_progs.h"
+
+#define CHECK_NEWNS(condition, tag, format...) ({              \
+       int __ret = !!(condition);                      \
+       if (__ret) {                                    \
+               printf("%s:FAIL:%s ", __func__, tag);   \
+               printf(format);                         \
+       } else {                                        \
+               printf("%s:PASS:%s\n", __func__, tag);  \
+       }                                               \
+       __ret;                                          \
+})
+
+struct bss {
+       __u64 dev;
+       __u64 ino;
+       __u64 pid_tgid;
+       __u64 user_pid_tgid;
+};
+
+int main(int argc, char **argv)
+{
+       pid_t pid;
+       int exit_code = 1;
+       struct stat st;
+
+       printf("Testing bpf_get_ns_current_pid_tgid helper in new ns\n");
+
+       if (stat("/proc/self/ns/pid", &st)) {
+               perror("stat failed on /proc/self/ns/pid ns\n");
+               printf("%s:FAILED\n", argv[0]);
+               return exit_code;
+       }
+
+       if (CHECK_NEWNS(unshare(CLONE_NEWPID | CLONE_NEWNS),
+                       "unshare CLONE_NEWPID | CLONE_NEWNS", "error errno=%d\n", errno))
+               return exit_code;
+
+       pid = fork();
+       if (pid == -1) {
+               perror("Fork() failed\n");
+               printf("%s:FAILED\n", argv[0]);
+               return exit_code;
+       }
+
+       if (pid > 0) {
+               int status;
+
+               usleep(5);
+               waitpid(pid, &status, 0);
+               return 0;
+       } else {
+
+               pid = fork();
+               if (pid == -1) {
+                       perror("Fork() failed\n");
+                       printf("%s:FAILED\n", argv[0]);
+                       return exit_code;
+               }
+
+               if (pid > 0) {
+                       int status;
+                       waitpid(pid, &status, 0);
+                       return 0;
+               } else {
+                       if (CHECK_NEWNS(mount("none", "/proc", NULL, MS_PRIVATE|MS_REC, NULL),
+                               "Unmounting proc", "Cannot umount proc! errno=%d\n", errno))
+                               return exit_code;
+
+                       if (CHECK_NEWNS(mount("proc", "/proc", "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL),
+                               "Mounting proc", "Cannot mount proc! errno=%d\n", errno))
+                               return exit_code;
+
+                       const char *probe_name = "raw_tracepoint/sys_enter";
+                       const char *file = "test_ns_current_pid_tgid.o";
+                       struct bpf_link *link = NULL;
+                       struct bpf_program *prog;
+                       struct bpf_map *bss_map;
+                       struct bpf_object *obj;
+                       int exit_code = 1;
+                       int err, key = 0;
+                       struct bss bss;
+                       struct stat st;
+                       __u64 id;
+
+                       obj = bpf_object__open_file(file, NULL);
+                       if (CHECK_NEWNS(IS_ERR(obj), "obj_open", "err %ld\n", PTR_ERR(obj)))
+                               return exit_code;
+
+                       err = bpf_object__load(obj);
+                       if (CHECK_NEWNS(err, "obj_load", "err %d errno %d\n", err, errno))
+                               goto cleanup;
+
+                       bss_map = bpf_object__find_map_by_name(obj, "test_ns_.bss");
+                       if (CHECK_NEWNS(!bss_map, "find_bss_map", "failed\n"))
+                               goto cleanup;
+
+                       prog = bpf_object__find_program_by_title(obj, probe_name);
+                       if (CHECK_NEWNS(!prog, "find_prog", "prog '%s' not found\n",
+                                               probe_name))
+                               goto cleanup;
+
+                       memset(&bss, 0, sizeof(bss));
+                       pid_t tid = syscall(SYS_gettid);
+                       pid_t pid = getpid();
+
+                       id = (__u64) tid << 32 | pid;
+                       bss.user_pid_tgid = id;
+
+                       if (CHECK_NEWNS(stat("/proc/self/ns/pid", &st),
+                               "stat new ns", "Failed to stat /proc/self/ns/pid errno=%d\n", errno))
+                               goto cleanup;
+
+                       bss.dev = st.st_dev;
+                       bss.ino = st.st_ino;
+
+                       err = bpf_map_update_elem(bpf_map__fd(bss_map), &key, &bss, 0);
+                       if (CHECK_NEWNS(err, "setting_bss", "failed to set bss : %d\n", err))
+                               goto cleanup;
+
+                       link = bpf_program__attach_raw_tracepoint(prog, "sys_enter");
+                       if (CHECK_NEWNS(IS_ERR(link), "attach_raw_tp", "err %ld\n",
+                                               PTR_ERR(link))) {
+                               link = NULL;
+                               goto cleanup;
+                       }
+
+                       /* trigger some syscalls */
+                       usleep(1);
+
+                       err = bpf_map_lookup_elem(bpf_map__fd(bss_map), &key, &bss);
+                       if (CHECK_NEWNS(err, "set_bss", "failed to get bss : %d\n", err))
+                               goto cleanup;
+
+                       if (CHECK_NEWNS(id != bss.pid_tgid, "Compare user pid/tgid vs. bpf pid/tgid",
+                                               "User pid/tgid %llu BPF pid/tgid %llu\n", id, bss.pid_tgid))
+                               goto cleanup;
+
+                       exit_code = 0;
+                       printf("%s:PASS\n", argv[0]);
+cleanup:
+                       if (!link) {
+                               bpf_link__destroy(link);
+                               link = NULL;
+                       }
+                       bpf_object__close(obj);
+               }
+       }
+}