bpf: Refactor RCU enforcement in the verifier.

bpf_rcu_read_lock/unlock() are only available in clang compiled kernels. Lack
of such key mechanism makes it impossible for sleepable bpf programs to use RCU
pointers.

Allow bpf_rcu_read_lock/unlock() in GCC compiled kernels (though GCC doesn't
support btf_type_tag yet) and allowlist certain field dereferences in important
data structures like tast_struct, cgroup, socket that are used by sleepable
programs either as RCU pointer or full trusted pointer (which is valid outside
of RCU CS). Use BTF_TYPE_SAFE_RCU and BTF_TYPE_SAFE_TRUSTED macros for such
tagging. They will be removed once GCC supports btf_type_tag.

With that refactor check_ptr_to_btf_access(). Make it strict in enforcing
PTR_TRUSTED and PTR_UNTRUSTED while deprecating old PTR_TO_BTF_ID without
modifier flags. There is a chance that this strict enforcement might break
existing programs (especially on GCC compiled kernels), but this cleanup has to
start sooner than later. Note PTR_TO_CTX access still yields old deprecated
PTR_TO_BTF_ID. Once it's converted to strict PTR_TRUSTED or PTR_UNTRUSTED the
kfuncs and helpers will be able to default to KF_TRUSTED_ARGS. KF_RCU will
remain as a weaker version of KF_TRUSTED_ARGS where obj refcnt could be 0.

Adjust rcu_read_lock selftest to run on gcc and clang compiled kernels.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: David Vernet <void@manifault.com>
Link: https://lore.kernel.org/bpf/20230303041446.3630-7-alexei.starovoitov@gmail.com

7 files changed, 18 insertions, 28 deletions

diff --git a/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c b/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
index 2cc759956e3b..63e776f4176e 100644
--- a/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
+++ b/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
@@ -193,7 +193,7 @@ out:
 	cgrp_ls_sleepable__destroy(skel);
 }
 
-static void test_no_rcu_lock(__u64 cgroup_id)
+static void test_yes_rcu_lock(__u64 cgroup_id)
 {
 	struct cgrp_ls_sleepable *skel;
 	int err;
@@ -204,7 +204,7 @@ static void test_no_rcu_lock(__u64 cgroup_id)
 
 	skel->bss->target_pid = syscall(SYS_gettid);
 
-	bpf_program__set_autoload(skel->progs.no_rcu_lock, true);
+	bpf_program__set_autoload(skel->progs.yes_rcu_lock, true);
 	err = cgrp_ls_sleepable__load(skel);
 	if (!ASSERT_OK(err, "skel_load"))
 		goto out;
@@ -220,7 +220,7 @@ out:
 	cgrp_ls_sleepable__destroy(skel);
 }
 
-static void test_rcu_lock(void)
+static void test_no_rcu_lock(void)
 {
 	struct cgrp_ls_sleepable *skel;
 	int err;
@@ -229,7 +229,7 @@ static void test_rcu_lock(void)
 	if (!ASSERT_OK_PTR(skel, "skel_open"))
 		return;
 
-	bpf_program__set_autoload(skel->progs.yes_rcu_lock, true);
+	bpf_program__set_autoload(skel->progs.no_rcu_lock, true);
 	err = cgrp_ls_sleepable__load(skel);
 	ASSERT_ERR(err, "skel_load");
 
@@ -256,10 +256,10 @@ void test_cgrp_local_storage(void)
 		test_negative();
 	if (test__start_subtest("cgroup_iter_sleepable"))
 		test_cgroup_iter_sleepable(cgroup_fd, cgroup_id);
+	if (test__start_subtest("yes_rcu_lock"))
+		test_yes_rcu_lock(cgroup_id);
 	if (test__start_subtest("no_rcu_lock"))
-		test_no_rcu_lock(cgroup_id);
-	if (test__start_subtest("rcu_lock"))
-		test_rcu_lock();
+		test_no_rcu_lock();
 
 	close(cgroup_fd);
 }
diff --git a/tools/testing/selftests/bpf/prog_tests/rcu_read_lock.c b/tools/testing/selftests/bpf/prog_tests/rcu_read_lock.c
index 447d8560ecb6..3f1f58d3a729 100644
--- a/tools/testing/selftests/bpf/prog_tests/rcu_read_lock.c
+++ b/tools/testing/selftests/bpf/prog_tests/rcu_read_lock.c
@@ -25,10 +25,10 @@ static void test_success(void)
 
 	bpf_program__set_autoload(skel->progs.get_cgroup_id, true);
 	bpf_program__set_autoload(skel->progs.task_succ, true);
-	bpf_program__set_autoload(skel->progs.no_lock, true);
 	bpf_program__set_autoload(skel->progs.two_regions, true);
 	bpf_program__set_autoload(skel->progs.non_sleepable_1, true);
 	bpf_program__set_autoload(skel->progs.non_sleepable_2, true);
+	bpf_program__set_autoload(skel->progs.task_trusted_non_rcuptr, true);
 	err = rcu_read_lock__load(skel);
 	if (!ASSERT_OK(err, "skel_load"))
 		goto out;
@@ -69,6 +69,7 @@ out:
 
 static const char * const inproper_region_tests[] = {
 	"miss_lock",
+	"no_lock",
 	"miss_unlock",
 	"non_sleepable_rcu_mismatch",
 	"inproper_sleepable_helper",
@@ -99,7 +100,6 @@ out:
 }
 
 static const char * const rcuptr_misuse_tests[] = {
-	"task_untrusted_non_rcuptr",
 	"task_untrusted_rcuptr",
 	"cross_rcu_region",
 };
@@ -128,17 +128,8 @@ out:
 
 void test_rcu_read_lock(void)
 {
-	struct btf *vmlinux_btf;
 	int cgroup_fd;
 
-	vmlinux_btf = btf__load_vmlinux_btf();
-	if (!ASSERT_OK_PTR(vmlinux_btf, "could not load vmlinux BTF"))
-		return;
-	if (btf__find_by_name_kind(vmlinux_btf, "rcu", BTF_KIND_TYPE_TAG) < 0) {
-		test__skip();
-		goto out;
-	}
-
 	cgroup_fd = test__join_cgroup("/rcu_read_lock");
 	if (!ASSERT_GE(cgroup_fd, 0, "join_cgroup /rcu_read_lock"))
 		goto out;
@@ -153,6 +144,5 @@ void test_rcu_read_lock(void)
 	if (test__start_subtest("negative_tests_rcuptr_misuse"))
 		test_rcuptr_misuse();
 	close(cgroup_fd);
-out:
-	btf__free(vmlinux_btf);
+out:;
 }
diff --git a/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c b/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
index 2d11ed528b6f..7615dc23d301 100644
--- a/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
+++ b/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
@@ -49,7 +49,7 @@ int no_rcu_lock(void *ctx)
 	if (task->pid != target_pid)
 		return 0;
 
-	/* ptr_to_btf_id semantics. should work. */
+	/* task->cgroups is untrusted in sleepable prog outside of RCU CS */
 	cgrp = task->cgroups->dfl_cgrp;
 	ptr = bpf_cgrp_storage_get(&map_a, cgrp, 0,
 				   BPF_LOCAL_STORAGE_GET_F_CREATE);
@@ -71,7 +71,7 @@ int yes_rcu_lock(void *ctx)
 
 	bpf_rcu_read_lock();
 	cgrp = task->cgroups->dfl_cgrp;
-	/* cgrp is untrusted and cannot pass to bpf_cgrp_storage_get() helper. */
+	/* cgrp is trusted under RCU CS */
 	ptr = bpf_cgrp_storage_get(&map_a, cgrp, 0, BPF_LOCAL_STORAGE_GET_F_CREATE);
 	if (ptr)
 		cgroup_id = cgrp->kn->id;
diff --git a/tools/testing/selftests/bpf/progs/cpumask_failure.c b/tools/testing/selftests/bpf/progs/cpumask_failure.c
index 33e8e86dd090..c16f7563b84e 100644
--- a/tools/testing/selftests/bpf/progs/cpumask_failure.c
+++ b/tools/testing/selftests/bpf/progs/cpumask_failure.c
@@ -44,7 +44,7 @@ int BPF_PROG(test_alloc_double_release, struct task_struct *task, u64 clone_flag
 }
 
 SEC("tp_btf/task_newtask")
-__failure __msg("bpf_cpumask_acquire args#0 expected pointer to STRUCT bpf_cpumask")
+__failure __msg("must be referenced")
 int BPF_PROG(test_acquire_wrong_cpumask, struct task_struct *task, u64 clone_flags)
 {
 	struct bpf_cpumask *cpumask;
diff --git a/tools/testing/selftests/bpf/progs/nested_trust_failure.c b/tools/testing/selftests/bpf/progs/nested_trust_failure.c
index 14aff7676436..0d1aa6bbace4 100644
--- a/tools/testing/selftests/bpf/progs/nested_trust_failure.c
+++ b/tools/testing/selftests/bpf/progs/nested_trust_failure.c
@@ -17,7 +17,7 @@ char _license[] SEC("license") = "GPL";
  */
 
 SEC("tp_btf/task_newtask")
-__failure __msg("R2 must be referenced or trusted")
+__failure __msg("R2 must be")
 int BPF_PROG(test_invalid_nested_user_cpus, struct task_struct *task, u64 clone_flags)
 {
 	bpf_cpumask_test_cpu(0, task->user_cpus_ptr);
diff --git a/tools/testing/selftests/bpf/progs/rcu_read_lock.c b/tools/testing/selftests/bpf/progs/rcu_read_lock.c
index 5cecbdbbb16e..7250bb76d18a 100644
--- a/tools/testing/selftests/bpf/progs/rcu_read_lock.c
+++ b/tools/testing/selftests/bpf/progs/rcu_read_lock.c
@@ -81,7 +81,7 @@ int no_lock(void *ctx)
 {
 	struct task_struct *task, *real_parent;
 
-	/* no bpf_rcu_read_lock(), old code still works */
+	/* old style ptr_to_btf_id is not allowed in sleepable */
 	task = bpf_get_current_task_btf();
 	real_parent = task->real_parent;
 	(void)bpf_task_storage_get(&map_a, real_parent, 0, 0);
@@ -286,13 +286,13 @@ out:
 }
 
 SEC("?fentry.s/" SYS_PREFIX "sys_getpgid")
-int task_untrusted_non_rcuptr(void *ctx)
+int task_trusted_non_rcuptr(void *ctx)
 {
 	struct task_struct *task, *group_leader;
 
 	task = bpf_get_current_task_btf();
 	bpf_rcu_read_lock();
-	/* the pointer group_leader marked as untrusted */
+	/* the pointer group_leader is explicitly marked as trusted */
 	group_leader = task->real_parent->group_leader;
 	(void)bpf_task_storage_get(&map_a, group_leader, 0, 0);
 	bpf_rcu_read_unlock();
diff --git a/tools/testing/selftests/bpf/verifier/calls.c b/tools/testing/selftests/bpf/verifier/calls.c
index 9a326a800e5c..5702fc9761ef 100644
--- a/tools/testing/selftests/bpf/verifier/calls.c
+++ b/tools/testing/selftests/bpf/verifier/calls.c
@@ -181,7 +181,7 @@
 	},
 	.result_unpriv = REJECT,
 	.result = REJECT,
-	.errstr = "negative offset ptr_ ptr R1 off=-4 disallowed",
+	.errstr = "ptr R1 off=-4 disallowed",
 },
 {
 	"calls: invalid kfunc call: PTR_TO_BTF_ID with variable offset",