On Mon, Aug 21, 2023 at 10:06 PM Dave Marchevsky <davemarchevsky@xxxxxx> wrote:
>
> The open-coded task_vma iter added earlier in this series allows for
> natural iteration over a task's vmas using existing open-coded iter
> infrastructure, specifically bpf_for_each.
>
> This patch adds a test demonstrating this pattern and validating
> correctness. The vma->vm_start and vma->vm_end addresses of the first
> 1000 vmas are recorded and compared to /proc/PID/maps output. As
> expected, both see the same vmas and addresses - with the exception of
> the [vsyscall] vma - which is explained in a comment in the prog_tests
> program.
>
> Signed-off-by: Dave Marchevsky <davemarchevsky@xxxxxx>
> ---
> .../testing/selftests/bpf/prog_tests/iters.c | 71 +++++++++++++++++++
> .../selftests/bpf/progs/iters_task_vma.c | 56 +++++++++++++++
> 2 files changed, 127 insertions(+)
> create mode 100644 tools/testing/selftests/bpf/progs/iters_task_vma.c
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/iters.c b/tools/testing/selftests/bpf/prog_tests/iters.c
> index 10804ae5ae97..f91f4a49066a 100644
> --- a/tools/testing/selftests/bpf/prog_tests/iters.c
> +++ b/tools/testing/selftests/bpf/prog_tests/iters.c
> @@ -8,6 +8,7 @@
> #include "iters_looping.skel.h"
> #include "iters_num.skel.h"
> #include "iters_testmod_seq.skel.h"
> +#include "iters_task_vma.skel.h"
>
> static void subtest_num_iters(void)
> {
> @@ -90,6 +91,74 @@ static void subtest_testmod_seq_iters(void)
> iters_testmod_seq__destroy(skel);
> }
>
> +static void subtest_task_vma_iters(void)
> +{
> + unsigned long start, end, bpf_iter_start, bpf_iter_end;
> + struct iters_task_vma *skel;
> + char rest_of_line[1000];
> + unsigned int seen;
> + int err;
> + FILE *f;
> +
> + skel = iters_task_vma__open();
> + if (!ASSERT_OK_PTR(skel, "skel_open"))
> + return;
> +
> + bpf_program__set_autoload(skel->progs.iter_task_vma_for_each, true);
why marking prog with SEC("?") just to explicitly enable autoload for
it? Just drop "?" in the first place?
> +
> + err = iters_task_vma__load(skel);
> + if (!ASSERT_OK(err, "skel_load"))
> + goto cleanup;
> +
> + skel->bss->target_pid = getpid();
> +
> + err = iters_task_vma__attach(skel);
> + if (!ASSERT_OK(err, "skel_attach"))
> + goto cleanup;
> +
> + iters_task_vma__detach(skel);
> + getpgid(skel->bss->target_pid);
Is this a trigger for the program? If yes, it's confusing that it
happens after skeleton detach. Is this intentional?
> +
> + if (!ASSERT_GT(skel->bss->vmas_seen, 0, "vmas_seen_gt_zero"))
> + goto cleanup;
> +
> + f = fopen("/proc/self/maps", "r");
> + if (!ASSERT_OK_PTR(f, "proc_maps_fopen"))
> + goto cleanup;
> +
> + seen = 0;
> + while (fscanf(f, "%lx-%lx %[^\n]\n", &start, &end, rest_of_line) == 3) {
> + /* [vsyscall] vma isn't _really_ part of task->mm vmas.
> + * /proc/PID/maps returns it when out of vmas - see get_gate_vma
> + * calls in fs/proc/task_mmu.c
> + */
> + if (strstr(rest_of_line, "[vsyscall]"))
> + continue;
> +
> + err = bpf_map_lookup_elem(bpf_map__fd(skel->maps.vm_start),
> + &seen, &bpf_iter_start);
> + if (!ASSERT_OK(err, "vm_start map_lookup_elem"))
> + goto cleanup;
> +
> + err = bpf_map_lookup_elem(bpf_map__fd(skel->maps.vm_end),
> + &seen, &bpf_iter_end);
> + if (!ASSERT_OK(err, "vm_end map_lookup_elem"))
> + goto cleanup;
> +
> + ASSERT_EQ(bpf_iter_start, start, "vma->vm_start match");
> + ASSERT_EQ(bpf_iter_end, end, "vma->vm_end match");
> + seen++;
> + }
> +
> + fclose(f);
> +
> + if (!ASSERT_EQ(skel->bss->vmas_seen, seen, "vmas_seen_eq"))
> + goto cleanup;
> +
> +cleanup:
> + iters_task_vma__destroy(skel);
> +}
> +
> void test_iters(void)
> {
> RUN_TESTS(iters_state_safety);
> @@ -103,4 +172,6 @@ void test_iters(void)
> subtest_num_iters();
> if (test__start_subtest("testmod_seq"))
> subtest_testmod_seq_iters();
> + if (test__start_subtest("task_vma"))
> + subtest_task_vma_iters();
> }
> diff --git a/tools/testing/selftests/bpf/progs/iters_task_vma.c b/tools/testing/selftests/bpf/progs/iters_task_vma.c
> new file mode 100644
> index 000000000000..b961d0a12223
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/iters_task_vma.c
> @@ -0,0 +1,56 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Copyright (c) 2023 Meta Platforms, Inc. and affiliates. */
> +
> +#include <limits.h>
> +#include <linux/errno.h>
> +#include "vmlinux.h"
> +#include <bpf/bpf_helpers.h>
> +#include "bpf_misc.h"
> +
> +pid_t target_pid = 0;
> +unsigned int vmas_seen = 0;
> +
> +struct {
> + __uint(type, BPF_MAP_TYPE_ARRAY);
> + __uint(max_entries, 1000);
> + __type(key, int);
> + __type(value, unsigned long);
> +} vm_start SEC(".maps");
> +
> +struct {
> + __uint(type, BPF_MAP_TYPE_ARRAY);
> + __uint(max_entries, 1000);
> + __type(key, int);
> + __type(value, unsigned long);
> +} vm_end SEC(".maps");
> +
> +SEC("?raw_tp/sys_enter")
> +int iter_task_vma_for_each(const void *ctx)
> +{
> + struct task_struct *task = bpf_get_current_task_btf();
> + struct vm_area_struct *vma;
> + unsigned long *start, *end;
> +
> + if (task->pid != target_pid)
> + return 0;
> +
> + bpf_for_each(task_vma, vma, task, 0) {
> + if (vmas_seen >= 1000)
> + break;
> +
this test is not idempotent, if there will be more than one syscall,
you'll keep incrementing vmas_seen. This might result in flaky test,
so I'd suggest to recalculate vmas_seen each time, as a local counter,
and then at the end writing it to global var:
int zero;
int vmas_seen;
...
int seen = zero;
... all the same logic but using seen as a key ...
vmas_seen = seen.
> + start = bpf_map_lookup_elem(&vm_start, &vmas_seen);
> + if (!start)
> + break;
> + *start = vma->vm_start;
> +
> + end = bpf_map_lookup_elem(&vm_end, &vmas_seen);
> + if (!end)
> + break;
> + *end = vma->vm_end;
> +
> + vmas_seen++;
> + }
> + return 0;
> +}
> +
> +char _license[] SEC("license") = "GPL";
> --
> 2.34.1
>
