On Wed, Mar 30, 2022 at 8:27 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote:
>
> bpf_program__attach_uprobe_opts() requires a binary_path argument
> specifying binary to instrument. Supporting simply specifying
> "libc.so.6" or "foo" should be possible too.
>
> Library search checks LD_LIBRARY_PATH, then /usr/lib64, /usr/lib.
> This allows users to run BPF programs prefixed with
> LD_LIBRARY_PATH=/path2/lib while still searching standard locations.
> Similarly for non .so files, we check PATH and /usr/bin, /usr/sbin.
>
> Path determination will be useful for auto-attach of BPF uprobe programs
> using SEC() definition.
>
> Signed-off-by: Alan Maguire <alan.maguire@xxxxxxxxxx>
> ---
This patch's subject was very long, I took the liberty to shorten it as:
libbpf: auto-resolve programs/libraries when necessary for uprobes
Please yell if you strongly object.
> tools/lib/bpf/libbpf.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 52 insertions(+), 1 deletion(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 809fe20..a7d5954 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -10517,6 +10517,46 @@ static int perf_event_uprobe_open_legacy(const char *probe_name, bool retprobe,
> return pfd;
> }
>
> +/* Get full path to program/shared library. */
> +static int resolve_full_path(const char *file, char *result, size_t result_sz)
> +{
> + const char *search_paths[2];
> + int i;
> +
> + if (strstr(file, ".so")) {
this check can be too brittle. It will think that binary called
"i.am.sorry" will be re assumed a shared library. While you can always
subvert this logic, maybe checking for ".so" suffix (not a substring)
and, if no such suffix found, looking for ".so." substring as a second
check? "i.am.so.sorry" still would fool it, but that looks like a very
artificial stretch.
let's improve this in a follow up
> + search_paths[0] = getenv("LD_LIBRARY_PATH");
> + search_paths[1] = "/usr/lib64:/usr/lib";
[...]
> {
> DECLARE_LIBBPF_OPTS(bpf_perf_event_opts, pe_opts);
> char errmsg[STRERR_BUFSIZE], *legacy_probe = NULL;
> + char full_binary_path[PATH_MAX];
> struct bpf_link *link;
> size_t ref_ctr_off;
> int pfd, err;
> @@ -10536,12 +10577,22 @@ static int perf_event_uprobe_open_legacy(const char *probe_name, bool retprobe,
> ref_ctr_off = OPTS_GET(opts, ref_ctr_offset, 0);
> pe_opts.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0);
>
> + if (binary_path && !strchr(binary_path, '/')) {
> + err = resolve_full_path(binary_path, full_binary_path,
> + sizeof(full_binary_path));
> + if (err) {
> + pr_warn("failed to resolve full path for '%s'\n", binary_path);
we use "prog '%s': " prefix for error messages within
bpf_program__attach_xxx() wherever possible. Fixed up while applying.
> + return libbpf_err_ptr(err);
> + }
> + binary_path = full_binary_path;
> + }
> +
> legacy = determine_uprobe_perf_type() < 0;
> if (!legacy) {
> pfd = perf_event_open_probe(true /* uprobe */, retprobe, binary_path,
> func_offset, pid, ref_ctr_off);
> } else {
> - char probe_name[512];
> + char probe_name[PATH_MAX + 64];
>
> if (ref_ctr_off)
> return libbpf_err_ptr(-EINVAL);
> --
> 1.8.3.1
>
