[PATCH v3 bpf-next 2/4] libbpf: add auto-attach for uprobes based on section name

[Alan Maguire <alan.maguire@oracle.com>]

Now that u[ret]probes can use name-based specification, it makes
sense to add support for auto-attach based on SEC() definition.
The format proposed is

	SEC("u[ret]probe//path/to/prog:[raw_offset|[function_name[+offset]]")

For example, to trace malloc() in libc:

        SEC("uprobe//usr/lib64/libc.so.6:malloc")

Auto-attach is done for all tasks (pid -1).

Note that there is a backwards-compatibility issue here.  Consider a BPF
object consisting of a set of BPF programs, including a uprobe program.
Because uprobes did not previously support auto-attach, it's possible that
because the uprobe section name is not in auto-attachable form, overall
BPF skeleton attach would now fail due to the failure of the uprobe program
to auto-attach.  So we need to handle the case of auto-attach failure where
the form of the section name is not suitable for auto-attach without
a complete attach failure.  On surveying the code, bpf_program__attach()
already returns -ESRCH in cases where no auto-attach function is
supplied, so for consistency with that - and because that return value
is less likely to collide with actual attach failures than -EOPNOTSUPP -
it is used as the attach function return value signalling auto-attach
is not possible.

Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
---
 tools/lib/bpf/libbpf.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 74 insertions(+), 3 deletions(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index eb95629..e2b4415 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -8581,6 +8581,7 @@ int bpf_program__set_log_buf(struct bpf_program *prog, char *log_buf, size_t log
 }
 
 static struct bpf_link *attach_kprobe(const struct bpf_program *prog, long cookie);
+static struct bpf_link *attach_uprobe(const struct bpf_program *prog, long cookie);
 static struct bpf_link *attach_tp(const struct bpf_program *prog, long cookie);
 static struct bpf_link *attach_raw_tp(const struct bpf_program *prog, long cookie);
 static struct bpf_link *attach_trace(const struct bpf_program *prog, long cookie);
@@ -8592,9 +8593,9 @@ int bpf_program__set_log_buf(struct bpf_program *prog, char *log_buf, size_t log
 	SEC_DEF("sk_reuseport/migrate",	SK_REUSEPORT, BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, SEC_ATTACHABLE | SEC_SLOPPY_PFX),
 	SEC_DEF("sk_reuseport",		SK_REUSEPORT, BPF_SK_REUSEPORT_SELECT, SEC_ATTACHABLE | SEC_SLOPPY_PFX),
 	SEC_DEF("kprobe/",		KPROBE,	0, SEC_NONE, attach_kprobe),
-	SEC_DEF("uprobe/",		KPROBE,	0, SEC_NONE),
+	SEC_DEF("uprobe/",		KPROBE, 0, SEC_NONE, attach_uprobe),
 	SEC_DEF("kretprobe/",		KPROBE, 0, SEC_NONE, attach_kprobe),
-	SEC_DEF("uretprobe/",		KPROBE, 0, SEC_NONE),
+	SEC_DEF("uretprobe/",		KPROBE, 0, SEC_NONE, attach_uprobe),
 	SEC_DEF("tc",			SCHED_CLS, 0, SEC_NONE),
 	SEC_DEF("classifier",		SCHED_CLS, 0, SEC_NONE | SEC_SLOPPY_PFX),
 	SEC_DEF("action",		SCHED_ACT, 0, SEC_NONE | SEC_SLOPPY_PFX),
@@ -10525,6 +10526,64 @@ static long elf_find_func_offset(const char *binary_path, const char *name)
 
 }
 
+/* Format of u[ret]probe section definition supporting auto-attach:
+ * u[ret]probe//path/to/prog:function[+offset]
+ *
+ * Many uprobe programs do not avail of auto-attach, so we need to handle the
+ * case where the format is uprobe/myfunc by returning NULL rather than an
+ * error.
+ */
+static struct bpf_link *attach_uprobe(const struct bpf_program *prog, long cookie)
+{
+	DECLARE_LIBBPF_OPTS(bpf_uprobe_opts, opts);
+	char *func_name, binary_path[512];
+	unsigned int raw_offset;
+	char *func, *probe_name;
+	struct bpf_link *link;
+	size_t offset = 0;
+	int n, err;
+
+	opts.retprobe = str_has_pfx(prog->sec_name, "uretprobe/");
+	if (opts.retprobe)
+		probe_name = prog->sec_name + sizeof("uretprobe/") - 1;
+	else
+		probe_name = prog->sec_name + sizeof("uprobe/") - 1;
+
+	snprintf(binary_path, sizeof(binary_path), "%s", probe_name);
+	/* ':' should be prior to function+offset */
+	func_name = strrchr(binary_path, ':');
+	if (!func_name) {
+		pr_debug("section '%s' is old-style u[ret]probe/function, cannot auto-attach\n",
+			 prog->sec_name);
+		return libbpf_err_ptr(-ESRCH);
+	}
+	func_name[0] = '\0';
+	func_name++;
+	n = sscanf(func_name, "%m[a-zA-Z0-9_.]+%li", &func, &offset);
+	if (n < 1) {
+		err = -EINVAL;
+		pr_warn("uprobe name is invalid: %s\n", func_name);
+		return libbpf_err_ptr(err);
+	}
+	/* Is func a raw address? */
+	if (n == 1 && sscanf(func, "%x", &raw_offset) == 1) {
+		free(func);
+		func = NULL;
+		offset = (size_t)raw_offset;
+	}
+	if (opts.retprobe && offset != 0) {
+		free(func);
+		err = -EINVAL;
+		pr_warn("uretprobes do not support offset specification\n");
+		return libbpf_err_ptr(err);
+	}
+	opts.func_name = func;
+
+	link = bpf_program__attach_uprobe_opts(prog, -1, binary_path, offset, &opts);
+	free(func);
+	return link;
+}
+
 struct bpf_link *bpf_program__attach_uprobe(const struct bpf_program *prog,
 					    bool retprobe, pid_t pid,
 					    const char *binary_path,
@@ -12041,7 +12100,19 @@ int bpf_object__attach_skeleton(struct bpf_object_skeleton *s)
 
 		*link = bpf_program__attach(prog);
 		err = libbpf_get_error(*link);
-		if (err) {
+		switch (err) {
+		case 0:
+			break;
+		case -ESRCH:
+			/* -ESRCH is used as it is less likely to collide with other error
+			 * cases in program attach while being consistent with the value
+			 * returned by bpf_program__attach() where no auto-attach function
+			 * is provided.
+			 */
+			pr_warn("auto-attach not supported for program '%s'\n",
+				bpf_program__name(prog));
+			break;
+		default:
 			pr_warn("failed to auto-attach program '%s': %d\n",
 				bpf_program__name(prog), err);
 			return libbpf_err(err);
-- 
1.8.3.1