bpf-for-netdev

-----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCAAdFiEE+soXsSLHKoYyzcli6rmadz2vbToFAmO17/QACgkQ6rmadz2v
 bTpFWBAAk85+qkRYyagEoAywkxPvRye6oYu321v69YjS7PdxqQ+sKSMyLR5brEaW
 QiQSQP/elwcBRyQAf4L/Fe++TZndpvuZ6CB+KUidOowtarxXPllc/TgWeNgF7A5l
 TuX3eQSLNdF5itdzoJzmiEQuBCHHXPce2L6UfNlP897yaWgxhawlkVEW+w8iB/UQ
 ZQCNGoDpga917DheLLx2kId046v70CUgwm8CU1rcAeA7Zi7B518OCzLm9z37l2em
 UWxPLOuQVA5u4O635dTRVUyluRPOvF1pjDei2gCmvC2JJS0X5eZbEEXCsZ04UbLd
 pd3y0+2WVOjiJ6kaSg6/U47G0gAjHfUHTLAoPzDw4LvlaXzyWZxK8Kxr1wHZRiYX
 q/iW1rUaWhsUI8XqROUnL3TA9UbFoKvSQIIQ8JEQN9jcd6enfMuBkwEDmZZIL9rR
 rZp3qZj+sxqGcKRpxorW5buM4vslYuoUD8/+9Lc+hkQ7QMCeDqdNzKEWtSCo3qjT
 ktbmZpNjCNnDhjyFZGuPsPxxK8I4QnuxGGeJvZeLZFlQq0Ft82APqFKzctdNAwFI
 NER4ihF0ohXdACrdL4PwaNXY10MCboHrQwpUy7vd2PUOAH+VI8nudSuweVXmeYPf
 ldH3w2N0G/wEy0QOvEo7SUEx3DiGTUNzHvu/ts4fDwUBqrBrzEM=
 =vD3D
 -----END PGP SIGNATURE-----

Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Alexei Starovoitov says:

====================
bpf 2023-01-04

We've added 5 non-merge commits during the last 8 day(s) which contain
a total of 5 files changed, 112 insertions(+), 18 deletions(-).

The main changes are:

1) Always use maximal size for copy_array in the verifier to fix
   KASAN tracking, from Kees.

2) Fix bpf task iterator walking through dead tasks, from Kui-Feng.

3) Make sure livepatch and bpf fexit can coexist, from Chuang.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  bpf: Always use maximal size for copy_array()
  selftests/bpf: add a test for iter/task_vma for short-lived processes
  bpf: keep a reference to the mm, in case the task is dead.
  selftests/bpf: Temporarily disable part of btf_dump:var_data test.
  bpf: Fix panic due to wrong pageattr of im->image
====================

Link: https://lore.kernel.org/r/20230104215500.79435-1-alexei.starovoitov@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

kernel/bpf/task_iter.c

@@ -438,6 +438,7 @@ struct bpf_iter_seq_task_vma_info {
 	 */
 	struct bpf_iter_seq_task_common common;
 	struct task_struct *task;
+	struct mm_struct *mm;
 	struct vm_area_struct *vma;
 	u32 tid;
 	unsigned long prev_vm_start;
@@ -456,16 +457,19 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
 	enum bpf_task_vma_iter_find_op op;
 	struct vm_area_struct *curr_vma;
 	struct task_struct *curr_task;
+	struct mm_struct *curr_mm;
 	u32 saved_tid = info->tid;
 
 	/* If this function returns a non-NULL vma, it holds a reference to
-	 * the task_struct, and holds read lock on vma->mm->mmap_lock.
+	 * the task_struct, holds a refcount on mm->mm_users, and holds
+	 * read lock on vma->mm->mmap_lock.
 	 * If this function returns NULL, it does not hold any reference or
 	 * lock.
 	 */
 	if (info->task) {
 		curr_task = info->task;
 		curr_vma = info->vma;
+		curr_mm = info->mm;
 		/* In case of lock contention, drop mmap_lock to unblock
 		 * the writer.
 		 *
@@ -504,13 +508,15 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
 		 *    4.2) VMA2 and VMA2' covers different ranges, process
 		 *         VMA2'.
 		 */
-		if (mmap_lock_is_contended(curr_task->mm)) {
+		if (mmap_lock_is_contended(curr_mm)) {
 			info->prev_vm_start = curr_vma->vm_start;
 			info->prev_vm_end = curr_vma->vm_end;
 			op = task_vma_iter_find_vma;
-			mmap_read_unlock(curr_task->mm);
-			if (mmap_read_lock_killable(curr_task->mm))
+			mmap_read_unlock(curr_mm);
+			if (mmap_read_lock_killable(curr_mm)) {
+				mmput(curr_mm);
 				goto finish;
+			}
 		} else {
 			op = task_vma_iter_next_vma;
 		}
@@ -535,42 +541,47 @@ again:
 			op = task_vma_iter_find_vma;
 		}
 
-		if (!curr_task->mm)
+		curr_mm = get_task_mm(curr_task);
+		if (!curr_mm)
 			goto next_task;
 
-		if (mmap_read_lock_killable(curr_task->mm))
+		if (mmap_read_lock_killable(curr_mm)) {
+			mmput(curr_mm);
 			goto finish;
+		}
 	}
 
 	switch (op) {
 	case task_vma_iter_first_vma:
-		curr_vma = find_vma(curr_task->mm, 0);
+		curr_vma = find_vma(curr_mm, 0);
 		break;
 	case task_vma_iter_next_vma:
-		curr_vma = find_vma(curr_task->mm, curr_vma->vm_end);
+		curr_vma = find_vma(curr_mm, curr_vma->vm_end);
 		break;
 	case task_vma_iter_find_vma:
 		/* We dropped mmap_lock so it is necessary to use find_vma
 		 * to find the next vma. This is similar to the  mechanism
 		 * in show_smaps_rollup().
 		 */
-		curr_vma = find_vma(curr_task->mm, info->prev_vm_end - 1);
+		curr_vma = find_vma(curr_mm, info->prev_vm_end - 1);
 		/* case 1) and 4.2) above just use curr_vma */
 
 		/* check for case 2) or case 4.1) above */
 		if (curr_vma &&
 		    curr_vma->vm_start == info->prev_vm_start &&
 		    curr_vma->vm_end == info->prev_vm_end)
-			curr_vma = find_vma(curr_task->mm, curr_vma->vm_end);
+			curr_vma = find_vma(curr_mm, curr_vma->vm_end);
 		break;
 	}
 	if (!curr_vma) {
 		/* case 3) above, or case 2) 4.1) with vma->next == NULL */
-		mmap_read_unlock(curr_task->mm);
+		mmap_read_unlock(curr_mm);
+		mmput(curr_mm);
 		goto next_task;
 	}
 	info->task = curr_task;
 	info->vma = curr_vma;
+	info->mm = curr_mm;
 	return curr_vma;
 
 next_task:
@@ -579,6 +590,7 @@ next_task:
 
 	put_task_struct(curr_task);
 	info->task = NULL;
+	info->mm = NULL;
 	info->tid++;
 	goto again;
 
@@ -587,6 +599,7 @@ finish:
 		put_task_struct(curr_task);
 	info->task = NULL;
 	info->vma = NULL;
+	info->mm = NULL;
 	return NULL;
 }
 
@@ -658,7 +671,9 @@ static void task_vma_seq_stop(struct seq_file *seq, void *v)
 		 */
 		info->prev_vm_start = ~0UL;
 		info->prev_vm_end = info->vma->vm_end;
-		mmap_read_unlock(info->task->mm);
+		mmap_read_unlock(info->mm);
+		mmput(info->mm);
+		info->mm = NULL;
 		put_task_struct(info->task);
 		info->task = NULL;
 	}

kernel/bpf/trampoline.c

@@ -488,6 +488,10 @@ again:
 		/* reset fops->func and fops->trampoline for re-register */
 		tr->fops->func = NULL;
 		tr->fops->trampoline = 0;
+
+		/* reset im->image memory attr for arch_prepare_bpf_trampoline */
+		set_memory_nx((long)im->image, 1);
+		set_memory_rw((long)im->image, 1);
 		goto again;
 	}
 #endif

kernel/bpf/verifier.c

@@ -1054,6 +1054,8 @@ static void print_insn_state(struct bpf_verifier_env *env,
  */
 static void *copy_array(void *dst, const void *src, size_t n, size_t size, gfp_t flags)
 {
+	size_t alloc_bytes;
+	void *orig = dst;
 	size_t bytes;
 
 	if (ZERO_OR_NULL_PTR(src))
@@ -1062,11 +1064,11 @@ static void *copy_array(void *dst, const void *src, size_t n, size_t size, gfp_t
 	if (unlikely(check_mul_overflow(n, size, &bytes)))
 		return NULL;
 
-	if (ksize(dst) < ksize(src)) {
-		kfree(dst);
-		dst = kmalloc_track_caller(kmalloc_size_roundup(bytes), flags);
-		if (!dst)
-			return NULL;
+	alloc_bytes = max(ksize(orig), kmalloc_size_roundup(bytes));
+	dst = krealloc(orig, alloc_bytes, flags);
+	if (!dst) {
+		kfree(orig);
+		return NULL;
 	}
 
 	memcpy(dst, src, bytes);

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"))

tools/testing/selftests/bpf/prog_tests/btf_dump.c

@@ -801,7 +801,7 @@ static void test_btf_dump_struct_data(struct btf *btf, struct btf_dump *d,
 static void test_btf_dump_var_data(struct btf *btf, struct btf_dump *d,
 				   char *str)
 {
-#if defined(__i386__) || defined(__x86_64__) || defined(__aarch64__)
+#if 0
 	TEST_BTF_DUMP_VAR(btf, d, NULL, str, "cpu_number", int, BTF_F_COMPACT,
 			  "int cpu_number = (int)100", 100);
 #endif