Merge branch 'bpf: fix the crash caused by task iterators over vma'

Kui-Feng Lee says:

====================

This issue is related to task iterators over vma. A system crash can
occur when a task iterator travels through vma of tasks as the death
of a task will clear the pointer to its mm, even though the
task_struct is still held. As a result, an unexpected crash happens
due to a null pointer. To address this problem, a reference to mm is
kept on the iterator to make sure that the pointer is always
valid. This patch set provides a solution for this crash by properly
referencing mm on task iterators over vma.

The major changes from v1 are:

 - Fix commit logs of the test case.

 - Use reverse Christmas tree coding style.

 - Remove unnecessary error handling for time().

v1: https://lore.kernel.org/bpf/20221216015912.991616-1-kuifeng@meta.com/
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

1 files changed, 73 insertions, 0 deletions

diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_iter.c b/tools/testing/selftests/bpf/prog_tests/bpf_iter.c
index 6f8ed61fc4b4..3af6450763e9 100644
--- a/tools/testing/selftests/bpf/prog_tests/bpf_iter.c
+++ b/tools/testing/selftests/bpf/prog_tests/bpf_iter.c
@@ -1465,6 +1465,77 @@ out:
 	bpf_iter_task_vma__destroy(skel);
 }
 
+static void test_task_vma_dead_task(void)
+{
+	struct bpf_iter_task_vma *skel;
+	int wstatus, child_pid = -1;
+	time_t start_tm, cur_tm;
+	int err, iter_fd = -1;
+	int wait_sec = 3;
+
+	skel = bpf_iter_task_vma__open();
+	if (!ASSERT_OK_PTR(skel, "bpf_iter_task_vma__open"))
+		return;
+
+	skel->bss->pid = getpid();
+
+	err = bpf_iter_task_vma__load(skel);
+	if (!ASSERT_OK(err, "bpf_iter_task_vma__load"))
+		goto out;
+
+	skel->links.proc_maps = bpf_program__attach_iter(
+		skel->progs.proc_maps, NULL);
+
+	if (!ASSERT_OK_PTR(skel->links.proc_maps, "bpf_program__attach_iter")) {
+		skel->links.proc_maps = NULL;
+		goto out;
+	}
+
+	start_tm = time(NULL);
+	cur_tm = start_tm;
+
+	child_pid = fork();
+	if (child_pid == 0) {
+		/* Fork short-lived processes in the background. */
+		while (cur_tm < start_tm + wait_sec) {
+			system("echo > /dev/null");
+			cur_tm = time(NULL);
+		}
+		exit(0);
+	}
+
+	if (!ASSERT_GE(child_pid, 0, "fork_child"))
+		goto out;
+
+	while (cur_tm < start_tm + wait_sec) {
+		iter_fd = bpf_iter_create(bpf_link__fd(skel->links.proc_maps));
+		if (!ASSERT_GE(iter_fd, 0, "create_iter"))
+			goto out;
+
+		/* Drain all data from iter_fd. */
+		while (cur_tm < start_tm + wait_sec) {
+			err = read_fd_into_buffer(iter_fd, task_vma_output, CMP_BUFFER_SIZE);
+			if (!ASSERT_GE(err, 0, "read_iter_fd"))
+				goto out;
+
+			cur_tm = time(NULL);
+
+			if (err == 0)
+				break;
+		}
+
+		close(iter_fd);
+		iter_fd = -1;
+	}
+
+	check_bpf_link_info(skel->progs.proc_maps);
+
+out:
+	waitpid(child_pid, &wstatus, 0);
+	close(iter_fd);
+	bpf_iter_task_vma__destroy(skel);
+}
+
 void test_bpf_sockmap_map_iter_fd(void)
 {
 	struct bpf_iter_sockmap *skel;
@@ -1586,6 +1657,8 @@ void test_bpf_iter(void)
 		test_task_file();
 	if (test__start_subtest("task_vma"))
 		test_task_vma();
+	if (test__start_subtest("task_vma_dead_task"))
+		test_task_vma_dead_task();
 	if (test__start_subtest("task_btf"))
 		test_task_btf();
 	if (test__start_subtest("tcp4"))