Re: [PATCH bpf-next v11 4/7] selftests/bpf: add test for accessing ctx from syscall program type

On Tue, 6 Sept 2022 at 17:13, Benjamin Tissoires benjamin.tissoires@redhat.com wrote:

We need to also export the kfunc set to the syscall program type, and then add a couple of eBPF programs that are testing those calls.

The first one checks for valid access, and the second one is OK from a static analysis point of view but fails at run time because we are trying to access outside of the allocated memory.

Signed-off-by: Benjamin Tissoires benjamin.tissoires@redhat.com

---

CI is failing for test_progs-no_alu32: https://github.com/kernel-patches/bpf/runs/8220916615?check_suite_focus=true

changes in v11:

• use new way of declaring tests

changes in v10:

• use new definitions for tests in an array
• add a new kfunc syscall_test_null_fail test

no changes in v9

no changes in v8

changes in v7:

• add 1 more case to ensure we can read the entire sizeof(ctx)
• add a test case for when the context is NULL

new in v6

net/bpf/test_run.c | 1 + .../selftests/bpf/prog_tests/kfunc_call.c | 143 +++++++++++++++++- .../selftests/bpf/progs/kfunc_call_fail.c | 39 +++++ .../selftests/bpf/progs/kfunc_call_test.c | 38 +++++ 4 files changed, 214 insertions(+), 7 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/kfunc_call_fail.c

diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c index 25d8ecf105aa..f16baf977a21 100644 --- a/net/bpf/test_run.c +++ b/net/bpf/test_run.c @@ -1634,6 +1634,7 @@ static int __init bpf_prog_test_run_init(void)

    ret = register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS, &bpf_prog_test_kfunc_set);
    ret = ret ?: register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &bpf_prog_test_kfunc_set);

•   ret = ret ?: register_btf_kfunc_id_set(BPF_PROG_TYPE_SYSCALL, &bpf_prog_test_kfunc_set);
    return ret ?: register_btf_id_dtor_kfuncs(bpf_prog_test_dtor_kfunc,
                                              ARRAY_SIZE(bpf_prog_test_dtor_kfunc),
                                              THIS_MODULE);
  

diff --git a/tools/testing/selftests/bpf/prog_tests/kfunc_call.c b/tools/testing/selftests/bpf/prog_tests/kfunc_call.c index 9dfbe5355a2d..d5881c3331a8 100644 --- a/tools/testing/selftests/bpf/prog_tests/kfunc_call.c +++ b/tools/testing/selftests/bpf/prog_tests/kfunc_call.c @@ -2,6 +2,7 @@ /* Copyright (c) 2021 Facebook */ #include <test_progs.h> #include <network_helpers.h> +#include "kfunc_call_fail.skel.h" #include "kfunc_call_test.skel.h" #include "kfunc_call_test.lskel.h" #include "kfunc_call_test_subprog.skel.h" @@ -10,37 +11,96 @@

#include "cap_helpers.h"

+static size_t log_buf_sz = 1048576; /* 1 MB */ +static char obj_log_buf[1048576];

+enum kfunc_test_type {

•   tc_test = 0,
  

•   syscall_test,
  

•   syscall_null_ctx_test,
  

+};

struct kfunc_test_params { const char *prog_name; unsigned long lskel_prog_desc_offset; int retval;

•   enum kfunc_test_type test_type;
  

•   const char *expected_err_msg;
  

};

-#define TC_TEST(name, __retval) \ +#define __BPF_TEST_SUCCESS(name, __retval, type) \ { \ .prog_name = #name, \ .lskel_prog_desc_offset = offsetof(struct kfunc_call_test_lskel, progs.name), \ .retval = __retval, \

•     .test_type = type, \
  

•     .expected_err_msg = NULL, \
  

•   }
  

+#define __BPF_TEST_FAIL(name, __retval, type, error_msg) \

•   { \
  

•     .prog_name = #name, \
  

•     .lskel_prog_desc_offset = 0 /* unused when test is failing */, \
  

•     .retval = __retval, \
  

•     .test_type = type, \
  

•     .expected_err_msg = error_msg, \
    }
  
  

+#define TC_TEST(name, retval) __BPF_TEST_SUCCESS(name, retval, tc_test) +#define SYSCALL_TEST(name, retval) __BPF_TEST_SUCCESS(name, retval, syscall_test) +#define SYSCALL_NULL_CTX_TEST(name, retval) __BPF_TEST_SUCCESS(name, retval, syscall_null_ctx_test)

+#define SYSCALL_NULL_CTX_FAIL(name, retval, error_msg) \

•   __BPF_TEST_FAIL(name, retval, syscall_null_ctx_test, error_msg)
  

static struct kfunc_test_params kfunc_tests[] = {

•   /* failure cases:
  

•    * if retval is 0 -> the program will fail to load and the error message is an error
  

•    * if retval is not 0 -> the program can be loaded but running it will gives the
  

•    *                       provided return value. The error message is thus the one
  

•    *                       from a successful load
  

•    */
  

•   SYSCALL_NULL_CTX_FAIL(kfunc_syscall_test_fail, -EINVAL, "processed 4 insns"),
  

•   SYSCALL_NULL_CTX_FAIL(kfunc_syscall_test_null_fail, -EINVAL, "processed 4 insns"),
  

•   /* success cases */
    TC_TEST(kfunc_call_test1, 12),
    TC_TEST(kfunc_call_test2, 3),
    TC_TEST(kfunc_call_test_ref_btf_id, 0),
  

•   SYSCALL_TEST(kfunc_syscall_test, 0),
  

•   SYSCALL_NULL_CTX_TEST(kfunc_syscall_test_null, 0),
  

+};

+struct syscall_test_args {

•   __u8 data[16];
  

•   size_t size;
  

};

static void verify_success(struct kfunc_test_params *param) { struct kfunc_call_test_lskel *lskel = NULL;

•   LIBBPF_OPTS(bpf_test_run_opts, topts);
    struct bpf_prog_desc *lskel_prog;
    struct kfunc_call_test *skel;
    struct bpf_program *prog;
    int prog_fd, err;
  

•   LIBBPF_OPTS(bpf_test_run_opts, topts,
  

•           .data_in = &pkt_v4,
  

•           .data_size_in = sizeof(pkt_v4),
  

•           .repeat = 1,
  

•   );
  

•   struct syscall_test_args args = {
  

•           .size = 10,
  

•   };
  

•   switch (param->test_type) {
  

•   case syscall_test:
  

•           topts.ctx_in = &args;
  

•           topts.ctx_size_in = sizeof(args);
  

•           /* fallthrough */
  

•   case syscall_null_ctx_test:
  

•           break;
  

•   case tc_test:
  

•           topts.data_in = &pkt_v4;
  

•           topts.data_size_in = sizeof(pkt_v4);
  

•           topts.repeat = 1;
  

•           break;
  

•   }
  
    /* first test with normal libbpf */
    skel = kfunc_call_test__open_and_load();
  

@@ -79,6 +139,72 @@ static void verify_success(struct kfunc_test_params *param) kfunc_call_test_lskel__destroy(lskel); }

+static void verify_fail(struct kfunc_test_params *param) +{

•   LIBBPF_OPTS(bpf_object_open_opts, opts);
  

•   LIBBPF_OPTS(bpf_test_run_opts, topts);
  

•   struct bpf_program *prog;
  

•   struct kfunc_call_fail *skel;
  

•   int prog_fd, err;
  

•   struct syscall_test_args args = {
  

•           .size = 10,
  

•   };
  

•   opts.kernel_log_buf = obj_log_buf;
  

•   opts.kernel_log_size = log_buf_sz;
  

•   opts.kernel_log_level = 1;
  

•   switch (param->test_type) {
  

•   case syscall_test:
  

•           topts.ctx_in = &args;
  

•           topts.ctx_size_in = sizeof(args);
  

•           /* fallthrough */
  

•   case syscall_null_ctx_test:
  

•           break;
  

•   case tc_test:
  

•           topts.data_in = &pkt_v4;
  

•           topts.data_size_in = sizeof(pkt_v4);
  

•           break;
  

•           topts.repeat = 1;
  

•   }
  

•   skel = kfunc_call_fail__open_opts(&opts);
  

•   if (!ASSERT_OK_PTR(skel, "kfunc_call_fail__open_opts"))
  

•           goto cleanup;
  

•   prog = bpf_object__find_program_by_name(skel->obj, param->prog_name);
  

•   if (!ASSERT_OK_PTR(prog, "bpf_object__find_program_by_name"))
  

•           goto cleanup;
  

•   bpf_program__set_autoload(prog, true);
  

•   err = kfunc_call_fail__load(skel);
  

•   if (!param->retval) {
  

•           /* the verifier is supposed to complain and refuses to load */
  

•           if (!ASSERT_ERR(err, "unexpected load success"))
  

•                   goto out_err;
  

•   } else {
  

•           /* the program is loaded but must dynamically fail */
  

•           if (!ASSERT_OK(err, "unexpected load error"))
  

•                   goto out_err;
  

•           prog_fd = bpf_program__fd(prog);
  

•           err = bpf_prog_test_run_opts(prog_fd, &topts);
  

•           if (!ASSERT_EQ(err, param->retval, param->prog_name))
  

•                   goto out_err;
  

•   }
  

+out_err:

•   if (!ASSERT_OK_PTR(strstr(obj_log_buf, param->expected_err_msg), "expected_err_msg")) {
  

•           fprintf(stderr, "Expected err_msg: %s\n", param->expected_err_msg);
  

•           fprintf(stderr, "Verifier output: %s\n", obj_log_buf);
  

•   }
  

+cleanup:

•   kfunc_call_fail__destroy(skel);
  

+}

static void test_main(void) { int i; @@ -87,7 +213,10 @@ static void test_main(void) if (!test__start_subtest(kfunc_tests[i].prog_name)) continue;

•           verify_success(&kfunc_tests[i]);
  

•           if (!kfunc_tests[i].expected_err_msg)
  

•                   verify_success(&kfunc_tests[i]);
  

•           else
  

•                   verify_fail(&kfunc_tests[i]);
    }
  

}

diff --git a/tools/testing/selftests/bpf/progs/kfunc_call_fail.c b/tools/testing/selftests/bpf/progs/kfunc_call_fail.c new file mode 100644 index 000000000000..4168027f2ab1 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/kfunc_call_fail.c @@ -0,0 +1,39 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2021 Facebook */ +#include <vmlinux.h> +#include <bpf/bpf_helpers.h>

+extern void bpf_kfunc_call_test_mem_len_pass1(void *mem, int len) __ksym;

+struct syscall_test_args {

•   __u8 data[16];
  

•   size_t size;
  

+};

+SEC("?syscall") +int kfunc_syscall_test_fail(struct syscall_test_args *args) +{

•   bpf_kfunc_call_test_mem_len_pass1(&args->data, sizeof(*args) + 1);
  

•   return 0;
  

+}

+SEC("?syscall") +int kfunc_syscall_test_null_fail(struct syscall_test_args *args) +{

•   /* Must be called with args as a NULL pointer
  

•    * we do not check for it to have the verifier consider that
  

•    * the pointer might not be null, and so we can load it.
  

•    *
  

•    * So the following can not be added:
  

•    *
  

•    * if (args)
  

•    *      return -22;
  

•    */
  

•   bpf_kfunc_call_test_mem_len_pass1(args, sizeof(*args));
  

•   return 0;
  

+}

+char _license[] SEC("license") = "GPL"; diff --git a/tools/testing/selftests/bpf/progs/kfunc_call_test.c b/tools/testing/selftests/bpf/progs/kfunc_call_test.c index 5aecbb9fdc68..94c05267e5e7 100644 --- a/tools/testing/selftests/bpf/progs/kfunc_call_test.c +++ b/tools/testing/selftests/bpf/progs/kfunc_call_test.c @@ -92,4 +92,42 @@ int kfunc_call_test_pass(struct __sk_buff *skb) return 0; }

+struct syscall_test_args {

•   __u8 data[16];
  

•   size_t size;
  

+};

+SEC("syscall") +int kfunc_syscall_test(struct syscall_test_args *args) +{

•   const int size = args->size;
  

•   if (size > sizeof(args->data))
  

•           return -7; /* -E2BIG */
  

Looks like it is due to this. Verifier is confused because: r7 = args->data; r1 = r7;

then it does r1 <<= 32; r1 >>=32; clearing upper 32 bits, so both r1 and r7 lose the id association which propagates the bounds of r1 learnt from comparison of it with sizeof(args->data);

•   bpf_kfunc_call_test_mem_len_pass1(&args->data, sizeof(args->data));
  

•   bpf_kfunc_call_test_mem_len_pass1(&args->data, sizeof(*args));
  

Later llvm assigns r7 to r2 for this call's 2nd arg. At this point the verifier still thinks r7 is unbounded, while to make a call with mem, len pair you need non-negative min value.

Easiest way might be to just do args->size & sizeof(args->data), as the verifier log says. You might still keep the error above. Others may have better ideas/insights.

•   bpf_kfunc_call_test_mem_len_pass1(&args->data, size);
  

•   return 0;
  

+}

+SEC("syscall") +int kfunc_syscall_test_null(struct syscall_test_args *args) +{

•   /* Must be called with args as a NULL pointer
  

•    * we do not check for it to have the verifier consider that
  

•    * the pointer might not be null, and so we can load it.
  

•    *
  

•    * So the following can not be added:
  

•    *
  

•    * if (args)
  

•    *      return -22;
  

•    */
  

•   bpf_kfunc_call_test_mem_len_pass1(args, 0);
  

•   return 0;
  

+}

char _license[] SEC("license") = "GPL";

2.36.1