Re: [PATCH bpf-next 1/3] bpftool: match several programs with same tag

On Tue, Dec 10, 2019 at 05:29:18PM +0000, Quentin Monnet wrote:

Hi Paul,

2019-12-10 17:06 UTC+0100 ~ Paul Chaignon [off-list ref]

quoted

When several BPF programs have the same tag, bpftool matches only the
first (in ID order).  This patch changes that behavior such that dump and
show commands return all matched programs.  Commands that require a single
program (e.g., pin and attach) will error out if given a tag that matches
several.  bpftool prog dump will also error out if file or visual are
given and several programs have the given tag.

In the case of the dump command, a program header is added before each
dump only if the tag matches several programs; this patch doesn't change
the output if a single program matches.

Signed-off-by: Paul Chaignon <redacted>
---
 .../bpftool/Documentation/bpftool-prog.rst    |  16 +-
 tools/bpf/bpftool/prog.c                      | 371 ++++++++++++------
 2 files changed, 272 insertions(+), 115 deletions(-)

[...]

quoted

@@ -351,21 +421,43 @@ static int show_prog(int fd)
 
 static int do_show(int argc, char **argv)
 {
+	int fd, nb_fds, i;
+	int *fds = NULL;
 	__u32 id = 0;
 	int err;
-	int fd;
 
 	if (show_pinned)
 		build_pinned_obj_table(&prog_table, BPF_OBJ_PROG);
 
 	if (argc == 2) {
-		fd = prog_parse_fd(&argc, &argv);
-		if (fd < 0)
+		fds = malloc(sizeof(int));
+		if (!fds) {
+			p_err("mem alloc failed");
 			return -1;
+		}
+		nb_fds = prog_parse_fds(&argc, &argv, fds);
+		if (nb_fds < 1)
+			goto err_free;
 
-		err = show_prog(fd);
-		close(fd);
-		return err;
+		if (json_output && nb_fds > 1)
+			jsonw_start_array(json_wtr);	/* root array */
+		for (i = 0; i < nb_fds; i++) {
+			err = show_prog(fds[i]);
+			close(fds[i]);
+			if (err) {
+				for (i++; i < nb_fds; i++)
+					close(fds[i]);
+				goto err_free;

Alternatively, we could keep trying to list the remaining programs. For
example, if the system has a long list of BPF programs running and one
of them is removed while printing the list, we would still have the rest
of the list.

As discussed off the list, since the kernel keeps a refcount for programs,
a program won't be removed while printing the list, not as long as we hold
an fd to it.  Unless there's another common case of failure, I'm going to
keep the current behavior in the v2 and break out of the loop on errors
(though, with the appropriate JSON array closing).

If we went this way, maybe just set err to non-zero if no program at all
could be printed?

quoted

+			}
+		}
+		if (json_output && nb_fds > 1)
+			jsonw_end_array(json_wtr);	/* root array */
+
+		return 0;
+
+err_free:
+		free(fds);
+		return -1;
 	}
 
 	if (argc)