Merge branch 'libbpf: allow users to set kprobe/uprobe attach mode'

Menglong Dong says:

====================

From: Menglong Dong <imagedong@tencent.com>

By default, libbpf will attach the kprobe/uprobe BPF program in the
latest mode that supported by kernel. In this series, we add the support
to let users manually attach kprobe/uprobe in legacy/perf/link mode in
the 1th patch.

And in the 2th patch, we split the testing 'attach_probe' into multi
subtests, as Andrii suggested.

In the 3th patch, we add the testings for loading kprobe/uprobe in
different mode.

Changes since v3:
- rename eBPF to BPF in the doc
- use OPTS_GET() to get the value of 'force_ioctl_attach'
- error out on attach mode is not supported
- use test_attach_probe_manual__open_and_load() directly

Changes since v2:
- fix the typo in the 2th patch

Changes since v1:
- some small changes in the 1th patch, as Andrii suggested
- split 'attach_probe' into multi subtests
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>

tools/lib/bpf/libbpf.c

@@ -9724,6 +9724,7 @@ struct bpf_link *bpf_program__attach_perf_event_opts(const struct bpf_program *p
 	char errmsg[STRERR_BUFSIZE];
 	struct bpf_link_perf *link;
 	int prog_fd, link_fd = -1, err;
+	bool force_ioctl_attach;
 
 	if (!OPTS_VALID(opts, bpf_perf_event_opts))
 		return libbpf_err_ptr(-EINVAL);
@@ -9747,7 +9748,8 @@ struct bpf_link *bpf_program__attach_perf_event_opts(const struct bpf_program *p
 	link->link.dealloc = &bpf_link_perf_dealloc;
 	link->perf_event_fd = pfd;
 
-	if (kernel_supports(prog->obj, FEAT_PERF_LINK)) {
+	force_ioctl_attach = OPTS_GET(opts, force_ioctl_attach, false);
+	if (kernel_supports(prog->obj, FEAT_PERF_LINK) && !force_ioctl_attach) {
 		DECLARE_LIBBPF_OPTS(bpf_link_create_opts, link_opts,
 			.perf_event.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0));
 
@@ -10106,6 +10108,7 @@ bpf_program__attach_kprobe_opts(const struct bpf_program *prog,
 				const struct bpf_kprobe_opts *opts)
 {
 	DECLARE_LIBBPF_OPTS(bpf_perf_event_opts, pe_opts);
+	enum probe_attach_mode attach_mode;
 	char errmsg[STRERR_BUFSIZE];
 	char *legacy_probe = NULL;
 	struct bpf_link *link;
@@ -10116,11 +10119,32 @@ bpf_program__attach_kprobe_opts(const struct bpf_program *prog,
 	if (!OPTS_VALID(opts, bpf_kprobe_opts))
 		return libbpf_err_ptr(-EINVAL);
 
+	attach_mode = OPTS_GET(opts, attach_mode, PROBE_ATTACH_MODE_DEFAULT);
 	retprobe = OPTS_GET(opts, retprobe, false);
 	offset = OPTS_GET(opts, offset, 0);
 	pe_opts.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0);
 
 	legacy = determine_kprobe_perf_type() < 0;
+	switch (attach_mode) {
+	case PROBE_ATTACH_MODE_LEGACY:
+		legacy = true;
+		pe_opts.force_ioctl_attach = true;
+		break;
+	case PROBE_ATTACH_MODE_PERF:
+		if (legacy)
+			return libbpf_err_ptr(-ENOTSUP);
+		pe_opts.force_ioctl_attach = true;
+		break;
+	case PROBE_ATTACH_MODE_LINK:
+		if (legacy || !kernel_supports(prog->obj, FEAT_PERF_LINK))
+			return libbpf_err_ptr(-ENOTSUP);
+		break;
+	case PROBE_ATTACH_MODE_DEFAULT:
+		break;
+	default:
+		return libbpf_err_ptr(-EINVAL);
+	}
+
 	if (!legacy) {
 		pfd = perf_event_open_probe(false /* uprobe */, retprobe,
 					    func_name, offset,
@@ -10852,6 +10876,7 @@ bpf_program__attach_uprobe_opts(const struct bpf_program *prog, pid_t pid,
 	const char *archive_path = NULL, *archive_sep = NULL;
 	char errmsg[STRERR_BUFSIZE], *legacy_probe = NULL;
 	DECLARE_LIBBPF_OPTS(bpf_perf_event_opts, pe_opts);
+	enum probe_attach_mode attach_mode;
 	char full_path[PATH_MAX];
 	struct bpf_link *link;
 	size_t ref_ctr_off;
@@ -10862,6 +10887,7 @@ bpf_program__attach_uprobe_opts(const struct bpf_program *prog, pid_t pid,
 	if (!OPTS_VALID(opts, bpf_uprobe_opts))
 		return libbpf_err_ptr(-EINVAL);
 
+	attach_mode = OPTS_GET(opts, attach_mode, PROBE_ATTACH_MODE_DEFAULT);
 	retprobe = OPTS_GET(opts, retprobe, false);
 	ref_ctr_off = OPTS_GET(opts, ref_ctr_offset, 0);
 	pe_opts.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0);
@@ -10903,6 +10929,26 @@ bpf_program__attach_uprobe_opts(const struct bpf_program *prog, pid_t pid,
 	}
 
 	legacy = determine_uprobe_perf_type() < 0;
+	switch (attach_mode) {
+	case PROBE_ATTACH_MODE_LEGACY:
+		legacy = true;
+		pe_opts.force_ioctl_attach = true;
+		break;
+	case PROBE_ATTACH_MODE_PERF:
+		if (legacy)
+			return libbpf_err_ptr(-ENOTSUP);
+		pe_opts.force_ioctl_attach = true;
+		break;
+	case PROBE_ATTACH_MODE_LINK:
+		if (legacy || !kernel_supports(prog->obj, FEAT_PERF_LINK))
+			return libbpf_err_ptr(-ENOTSUP);
+		break;
+	case PROBE_ATTACH_MODE_DEFAULT:
+		break;
+	default:
+		return libbpf_err_ptr(-EINVAL);
+	}
+
 	if (!legacy) {
 		pfd = perf_event_open_probe(true /* uprobe */, retprobe, binary_path,
 					    func_offset, pid, ref_ctr_off);

tools/lib/bpf/libbpf.h

@@ -447,12 +447,15 @@ LIBBPF_API struct bpf_link *
 bpf_program__attach(const struct bpf_program *prog);
 
 struct bpf_perf_event_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* custom user-provided value fetchable through bpf_get_attach_cookie() */
 	__u64 bpf_cookie;
+	/* don't use BPF link when attach BPF program */
+	bool force_ioctl_attach;
+	size_t :0;
 };
-#define bpf_perf_event_opts__last_field bpf_cookie
+#define bpf_perf_event_opts__last_field force_ioctl_attach
 
 LIBBPF_API struct bpf_link *
 bpf_program__attach_perf_event(const struct bpf_program *prog, int pfd);
@@ -461,8 +464,25 @@ LIBBPF_API struct bpf_link *
 bpf_program__attach_perf_event_opts(const struct bpf_program *prog, int pfd,
 				    const struct bpf_perf_event_opts *opts);
 
+/**
+ * enum probe_attach_mode - the mode to attach kprobe/uprobe
+ *
+ * force libbpf to attach kprobe/uprobe in specific mode, -ENOTSUP will
+ * be returned if it is not supported by the kernel.
+ */
+enum probe_attach_mode {
+	/* attach probe in latest supported mode by kernel */
+	PROBE_ATTACH_MODE_DEFAULT = 0,
+	/* attach probe in legacy mode, using debugfs/tracefs */
+	PROBE_ATTACH_MODE_LEGACY,
+	/* create perf event with perf_event_open() syscall */
+	PROBE_ATTACH_MODE_PERF,
+	/* attach probe with BPF link */
+	PROBE_ATTACH_MODE_LINK,
+};
+
 struct bpf_kprobe_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* custom user-provided value fetchable through bpf_get_attach_cookie() */
 	__u64 bpf_cookie;
@@ -470,9 +490,11 @@ struct bpf_kprobe_opts {
 	size_t offset;
 	/* kprobe is return probe */
 	bool retprobe;
+	/* kprobe attach mode */
+	enum probe_attach_mode attach_mode;
 	size_t :0;
 };
-#define bpf_kprobe_opts__last_field retprobe
+#define bpf_kprobe_opts__last_field attach_mode
 
 LIBBPF_API struct bpf_link *
 bpf_program__attach_kprobe(const struct bpf_program *prog, bool retprobe,
@@ -506,7 +528,7 @@ bpf_program__attach_kprobe_multi_opts(const struct bpf_program *prog,
 				      const struct bpf_kprobe_multi_opts *opts);
 
 struct bpf_ksyscall_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* custom user-provided value fetchable through bpf_get_attach_cookie() */
 	__u64 bpf_cookie;
@@ -552,7 +574,7 @@ bpf_program__attach_ksyscall(const struct bpf_program *prog,
 			     const struct bpf_ksyscall_opts *opts);
 
 struct bpf_uprobe_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* offset of kernel reference counted USDT semaphore, added in
 	 * a6ca88b241d5 ("trace_uprobe: support reference counter in fd-based uprobe")
@@ -570,9 +592,11 @@ struct bpf_uprobe_opts {
 	 * binary_path.
 	 */
 	const char *func_name;
+	/* uprobe attach mode */
+	enum probe_attach_mode attach_mode;
 	size_t :0;
 };
-#define bpf_uprobe_opts__last_field func_name
+#define bpf_uprobe_opts__last_field attach_mode
 
 /**
  * @brief **bpf_program__attach_uprobe()** attaches a BPF program
@@ -646,7 +670,7 @@ bpf_program__attach_usdt(const struct bpf_program *prog,
 			 const struct bpf_usdt_opts *opts);
 
 struct bpf_tracepoint_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* custom user-provided value fetchable through bpf_get_attach_cookie() */
 	__u64 bpf_cookie;
@@ -1110,7 +1134,7 @@ struct user_ring_buffer;
 typedef int (*ring_buffer_sample_fn)(void *ctx, void *data, size_t size);
 
 struct ring_buffer_opts {
-	size_t sz; /* size of this struct, for forward/backward compatiblity */
+	size_t sz; /* size of this struct, for forward/backward compatibility */
 };
 
 #define ring_buffer_opts__last_field sz
@@ -1475,7 +1499,7 @@ LIBBPF_API void
 bpf_object__destroy_subskeleton(struct bpf_object_subskeleton *s);
 
 struct gen_loader_opts {
-	size_t sz; /* size of this struct, for forward/backward compatiblity */
+	size_t sz; /* size of this struct, for forward/backward compatibility */
 	const char *data;
 	const char *insns;
 	__u32 data_sz;
@@ -1493,13 +1517,13 @@ enum libbpf_tristate {
 };
 
 struct bpf_linker_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 };
 #define bpf_linker_opts__last_field sz
 
 struct bpf_linker_file_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 };
 #define bpf_linker_file_opts__last_field sz
@@ -1542,7 +1566,7 @@ typedef int (*libbpf_prog_attach_fn_t)(const struct bpf_program *prog, long cook
 				       struct bpf_link **link);
 
 struct libbpf_prog_handler_opts {
-	/* size of this struct, for forward/backward compatiblity */
+	/* size of this struct, for forward/backward compatibility */
 	size_t sz;
 	/* User-provided value that is passed to prog_setup_fn,
 	 * prog_prepare_load_fn, and prog_attach_fn callbacks. Allows user to

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

+// SPDX-License-Identifier: GPL-2.0
+// Copyright (c) 2017 Facebook
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+#include "bpf_misc.h"
+
+int kprobe_res = 0;
+
+/**
+ * This program will be manually made sleepable on the userspace side
+ * and should thus be unattachable.
+ */
+SEC("kprobe/" SYS_PREFIX "sys_nanosleep")
+int handle_kprobe_sleepable(struct pt_regs *ctx)
+{
+	kprobe_res = 1;
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";

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

@@ -7,12 +7,8 @@
 #include <bpf/bpf_core_read.h>
 #include "bpf_misc.h"
 
-int kprobe_res = 0;
 int kprobe2_res = 0;
-int kretprobe_res = 0;
 int kretprobe2_res = 0;
-int uprobe_res = 0;
-int uretprobe_res = 0;
 int uprobe_byname_res = 0;
 int uretprobe_byname_res = 0;
 int uprobe_byname2_res = 0;
@@ -23,13 +19,6 @@ int uretprobe_byname3_sleepable_res = 0;
 int uretprobe_byname3_res = 0;
 void *user_ptr = 0;
 
-SEC("kprobe")
-int handle_kprobe(struct pt_regs *ctx)
-{
-	kprobe_res = 1;
-	return 0;
-}
-
 SEC("ksyscall/nanosleep")
 int BPF_KSYSCALL(handle_kprobe_auto, struct __kernel_timespec *req, struct __kernel_timespec *rem)
 {
@@ -37,24 +26,6 @@ int BPF_KSYSCALL(handle_kprobe_auto, struct __kernel_timespec *req, struct __ker
 	return 0;
 }
 
-/**
- * This program will be manually made sleepable on the userspace side
- * and should thus be unattachable.
- */
-SEC("kprobe/" SYS_PREFIX "sys_nanosleep")
-int handle_kprobe_sleepable(struct pt_regs *ctx)
-{
-	kprobe_res = 2;
-	return 0;
-}
-
-SEC("kretprobe")
-int handle_kretprobe(struct pt_regs *ctx)
-{
-	kretprobe_res = 2;
-	return 0;
-}
-
 SEC("kretsyscall/nanosleep")
 int BPF_KRETPROBE(handle_kretprobe_auto, int ret)
 {
@@ -63,16 +34,14 @@ int BPF_KRETPROBE(handle_kretprobe_auto, int ret)
 }
 
 SEC("uprobe")
-int handle_uprobe(struct pt_regs *ctx)
+int handle_uprobe_ref_ctr(struct pt_regs *ctx)
 {
-	uprobe_res = 3;
 	return 0;
 }
 
 SEC("uretprobe")
-int handle_uretprobe(struct pt_regs *ctx)
+int handle_uretprobe_ref_ctr(struct pt_regs *ctx)
 {
-	uretprobe_res = 4;
 	return 0;
 }
 

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

+// SPDX-License-Identifier: GPL-2.0
+// Copyright (c) 2017 Facebook
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+#include "bpf_misc.h"
+
+int kprobe_res = 0;
+int kretprobe_res = 0;
+int uprobe_res = 0;
+int uretprobe_res = 0;
+int uprobe_byname_res = 0;
+void *user_ptr = 0;
+
+SEC("kprobe")
+int handle_kprobe(struct pt_regs *ctx)
+{
+	kprobe_res = 1;
+	return 0;
+}
+
+SEC("kretprobe")
+int handle_kretprobe(struct pt_regs *ctx)
+{
+	kretprobe_res = 2;
+	return 0;
+}
+
+SEC("uprobe")
+int handle_uprobe(struct pt_regs *ctx)
+{
+	uprobe_res = 3;
+	return 0;
+}
+
+SEC("uretprobe")
+int handle_uretprobe(struct pt_regs *ctx)
+{
+	uretprobe_res = 4;
+	return 0;
+}
+
+SEC("uprobe")
+int handle_uprobe_byname(struct pt_regs *ctx)
+{
+	uprobe_byname_res = 5;
+	return 0;
+}
+
+
+char _license[] SEC("license") = "GPL";