Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Alexei Starovoitov says:

====================
pull-request: bpf 2021-03-20

The following pull-request contains BPF updates for your *net* tree.

We've added 5 non-merge commits during the last 3 day(s) which contain
a total of 8 files changed, 155 insertions(+), 12 deletions(-).

The main changes are:

1) Use correct nops in fexit trampoline, from Stanislav.

2) Fix BTF dump, from Jean-Philippe.

3) Fix umd memory leak, from Zqiang.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

arch/x86/net/bpf_jit_comp.c

@@ -2012,7 +2012,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
 		/* remember return value in a stack for bpf prog to access */
 		emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_0, -8);
 		im->ip_after_call = prog;
-		emit_nops(&prog, 5);
+		memcpy(prog, ideal_nops[NOP_ATOMIC5], X86_PATCH_SIZE);
+		prog += X86_PATCH_SIZE;
 	}
 
 	if (fmod_ret->nr_progs) {

include/linux/usermode_driver.h

@@ -14,5 +14,6 @@ struct umd_info {
 int umd_load_blob(struct umd_info *info, const void *data, size_t len);
 int umd_unload_blob(struct umd_info *info);
 int fork_usermode_driver(struct umd_info *info);
+void umd_cleanup_helper(struct umd_info *info);
 
 #endif /* __LINUX_USERMODE_DRIVER_H__ */

kernel/bpf/preload/bpf_preload_kern.c

@@ -60,9 +60,12 @@ static int finish(void)
 			 &magic, sizeof(magic), &pos);
 	if (n != sizeof(magic))
 		return -EPIPE;
+
 	tgid = umd_ops.info.tgid;
-	wait_event(tgid->wait_pidfd, thread_group_exited(tgid));
-	umd_ops.info.tgid = NULL;
+	if (tgid) {
+		wait_event(tgid->wait_pidfd, thread_group_exited(tgid));
+		umd_cleanup_helper(&umd_ops.info);
+	}
 	return 0;
 }
 
@@ -80,10 +83,18 @@ static int __init load_umd(void)
 
 static void __exit fini_umd(void)
 {
+	struct pid *tgid;
+
 	bpf_preload_ops = NULL;
+
 	/* kill UMD in case it's still there due to earlier error */
-	kill_pid(umd_ops.info.tgid, SIGKILL, 1);
-	umd_ops.info.tgid = NULL;
+	tgid = umd_ops.info.tgid;
+	if (tgid) {
+		kill_pid(tgid, SIGKILL, 1);
+
+		wait_event(tgid->wait_pidfd, thread_group_exited(tgid));
+		umd_cleanup_helper(&umd_ops.info);
+	}
 	umd_unload_blob(&umd_ops.info);
 }
 late_initcall(load_umd);

kernel/usermode_driver.c

@@ -139,13 +139,22 @@ static void umd_cleanup(struct subprocess_info *info)
 	struct umd_info *umd_info = info->data;
 
 	/* cleanup if umh_setup() was successful but exec failed */
-	if (info->retval) {
-		fput(umd_info->pipe_to_umh);
-		fput(umd_info->pipe_from_umh);
-		put_pid(umd_info->tgid);
-		umd_info->tgid = NULL;
-	}
+	if (info->retval)
+		umd_cleanup_helper(umd_info);
+}
+
+/**
+ * umd_cleanup_helper - release the resources which were allocated in umd_setup
+ * @info: information about usermode driver
+ */
+void umd_cleanup_helper(struct umd_info *info)
+{
+	fput(info->pipe_to_umh);
+	fput(info->pipe_from_umh);
+	put_pid(info->tgid);
+	info->tgid = NULL;
 }
+EXPORT_SYMBOL_GPL(umd_cleanup_helper);
 
 /**
  * fork_usermode_driver - fork a usermode driver

tools/lib/bpf/btf_dump.c

@@ -462,7 +462,7 @@ static int btf_dump_order_type(struct btf_dump *d, __u32 id, bool through_ptr)
 		return err;
 
 	case BTF_KIND_ARRAY:
-		return btf_dump_order_type(d, btf_array(t)->type, through_ptr);
+		return btf_dump_order_type(d, btf_array(t)->type, false);
 
 	case BTF_KIND_STRUCT:
 	case BTF_KIND_UNION: {

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

@@ -0,0 +1,82 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+#define _GNU_SOURCE
+#include <sched.h>
+#include <test_progs.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+#include "fexit_sleep.skel.h"
+
+static int do_sleep(void *skel)
+{
+	struct fexit_sleep *fexit_skel = skel;
+	struct timespec ts1 = { .tv_nsec = 1 };
+	struct timespec ts2 = { .tv_sec = 10 };
+
+	fexit_skel->bss->pid = getpid();
+	(void)syscall(__NR_nanosleep, &ts1, NULL);
+	(void)syscall(__NR_nanosleep, &ts2, NULL);
+	return 0;
+}
+
+#define STACK_SIZE (1024 * 1024)
+static char child_stack[STACK_SIZE];
+
+void test_fexit_sleep(void)
+{
+	struct fexit_sleep *fexit_skel = NULL;
+	int wstatus, duration = 0;
+	pid_t cpid;
+	int err, fexit_cnt;
+
+	fexit_skel = fexit_sleep__open_and_load();
+	if (CHECK(!fexit_skel, "fexit_skel_load", "fexit skeleton failed\n"))
+		goto cleanup;
+
+	err = fexit_sleep__attach(fexit_skel);
+	if (CHECK(err, "fexit_attach", "fexit attach failed: %d\n", err))
+		goto cleanup;
+
+	cpid = clone(do_sleep, child_stack + STACK_SIZE, CLONE_FILES | SIGCHLD, fexit_skel);
+	if (CHECK(cpid == -1, "clone", strerror(errno)))
+		goto cleanup;
+
+	/* wait until first sys_nanosleep ends and second sys_nanosleep starts */
+	while (READ_ONCE(fexit_skel->bss->fentry_cnt) != 2);
+	fexit_cnt = READ_ONCE(fexit_skel->bss->fexit_cnt);
+	if (CHECK(fexit_cnt != 1, "fexit_cnt", "%d", fexit_cnt))
+		goto cleanup;
+
+	/* close progs and detach them. That will trigger two nop5->jmp5 rewrites
+	 * in the trampolines to skip nanosleep_fexit prog.
+	 * The nanosleep_fentry prog will get detached first.
+	 * The nanosleep_fexit prog will get detached second.
+	 * Detaching will trigger freeing of both progs JITed images.
+	 * There will be two dying bpf_tramp_image-s, but only the initial
+	 * bpf_tramp_image (with both _fentry and _fexit progs will be stuck
+	 * waiting for percpu_ref_kill to confirm). The other one
+	 * will be freed quickly.
+	 */
+	close(bpf_program__fd(fexit_skel->progs.nanosleep_fentry));
+	close(bpf_program__fd(fexit_skel->progs.nanosleep_fexit));
+	fexit_sleep__detach(fexit_skel);
+
+	/* kill the thread to unwind sys_nanosleep stack through the trampoline */
+	kill(cpid, 9);
+
+	if (CHECK(waitpid(cpid, &wstatus, 0) == -1, "waitpid", strerror(errno)))
+		goto cleanup;
+	if (CHECK(WEXITSTATUS(wstatus) != 0, "exitstatus", "failed"))
+		goto cleanup;
+
+	/* The bypassed nanosleep_fexit prog shouldn't have executed.
+	 * Unlike progs the maps were not freed and directly accessible.
+	 */
+	fexit_cnt = READ_ONCE(fexit_skel->bss->fexit_cnt);
+	if (CHECK(fexit_cnt != 1, "fexit_cnt", "%d", fexit_cnt))
+		goto cleanup;
+
+cleanup:
+	fexit_sleep__destroy(fexit_skel);
+}

tools/testing/selftests/bpf/progs/btf_dump_test_case_syntax.c

@@ -174,6 +174,12 @@ struct struct_in_struct {
 	};
 };
 
+struct struct_in_array {};
+
+struct struct_in_array_typed {};
+
+typedef struct struct_in_array_typed struct_in_array_t[2];
+
 struct struct_with_embedded_stuff {
 	int a;
 	struct {
@@ -203,6 +209,8 @@ struct struct_with_embedded_stuff {
 	} r[5];
 	struct struct_in_struct s[10];
 	int t[11];
+	struct struct_in_array (*u)[2];
+	struct_in_array_t *v;
 };
 
 struct root_struct {

tools/testing/selftests/bpf/progs/fexit_sleep.c

@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+
+char LICENSE[] SEC("license") = "GPL";
+
+int pid = 0;
+int fentry_cnt = 0;
+int fexit_cnt = 0;
+
+SEC("fentry/__x64_sys_nanosleep")
+int BPF_PROG(nanosleep_fentry, const struct pt_regs *regs)
+{
+	if ((int)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	fentry_cnt++;
+	return 0;
+}
+
+SEC("fexit/__x64_sys_nanosleep")
+int BPF_PROG(nanosleep_fexit, const struct pt_regs *regs, int ret)
+{
+	if ((int)bpf_get_current_pid_tgid() != pid)
+		return 0;
+
+	fexit_cnt++;
+	return 0;
+}