Re: [PATCH v3] libbpf: Add some details for BTF parsing failures

From: Andrii Nakryiko
Date: Mon Jan 29 2024 - 19:44:08 EST


On Thu, Jan 25, 2024 at 3:18 PM Ian Rogers <irogers@xxxxxxxxxx> wrote:
>
> As CONFIG_DEBUG_INFO_BTF is default off the existing "failed to find
> valid kernel BTF" message makes diagnosing the kernel build issue some
> what cryptic. Add a little more detail with the hope of helping users.
>
> Before:
> ```
> libbpf: failed to find valid kernel BTF
> libbpf: Error loading vmlinux BTF: -3
> ```
>
> After not accessible:
> ```
> libbpf: access to canonical vmlinux (/sys/kernel/btf/vmlinux) to load BTF failed: No such file or directory
> libbpf: was CONFIG_DEBUG_INFO_BTF enabled?
> libbpf: failed to find valid kernel BTF
> libbpf: Error loading vmlinux BTF: -3
> ```
>
> After not readable:
> ```
> libbpf: unable to read canonical vmlinux (/sys/kernel/btf/vmlinux): Permission denied
> libbpf: failed to find valid kernel BTF
> libbpf: Error loading vmlinux BTF: -3
> ```
>
> Closes: https://lore.kernel.org/bpf/CAP-5=fU+DN_+Y=Y4gtELUsJxKNDDCOvJzPHvjUVaUoeFAzNnig@xxxxxxxxxxxxxx/
> Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
>
> ---
> v3. Try to address review comments from Andrii Nakryiko.

I did some further simplifications and clean ups while applying.

I dropped an extra faccessat(R_OK) check for /sys/kernel/btf/vmlinux
and instead if F_OK passes, just go ahead and try to parse
/sys/kernel/btf/vmlinux. If we have no access, we should get -EPERM or
-EACCESS (I didn't check which), otherwise we'll either parse or won't
find any BTF, both are errors. If /sys/kernel/btf/vmlinux exists,
there seems to be little point nowadays to try fallback locations,
kernel clearly is modern enough to generate /sys/kernel/btf/vmlinux,
so we just bail out with error.

Please check the landed commit in bpf-next and let me know if it
doesn't cover your use case properly.

> ---
> tools/lib/bpf/btf.c | 35 +++++++++++++++++++++++++++--------
> 1 file changed, 27 insertions(+), 8 deletions(-)
>
> diff --git a/tools/lib/bpf/btf.c b/tools/lib/bpf/btf.c
> index ec92b87cae01..45983f42aba9 100644
> --- a/tools/lib/bpf/btf.c
> +++ b/tools/lib/bpf/btf.c
> @@ -4932,10 +4932,9 @@ static int btf_dedup_remap_types(struct btf_dedup *d)
> */
> struct btf *btf__load_vmlinux_btf(void)
> {
> + const char *canonical_vmlinux = "/sys/kernel/btf/vmlinux";
> + /* fall back locations, trying to find vmlinux on disk */
> const char *locations[] = {
> - /* try canonical vmlinux BTF through sysfs first */
> - "/sys/kernel/btf/vmlinux",
> - /* fall back to trying to find vmlinux on disk otherwise */
> "/boot/vmlinux-%1$s",
> "/lib/modules/%1$s/vmlinux-%1$s",
> "/lib/modules/%1$s/build/vmlinux",
> @@ -4946,14 +4945,34 @@ struct btf *btf__load_vmlinux_btf(void)
> };
> char path[PATH_MAX + 1];
> struct utsname buf;
> - struct btf *btf;
> + struct btf *btf = NULL;
> int i, err;
>
> - uname(&buf);
> + /* is canonical sysfs location accessible? */
> + err = faccessat(AT_FDCWD, canonical_vmlinux, F_OK, AT_EACCESS);
> + if (err) {
> + pr_warn("access to canonical vmlinux (%s) to load BTF failed: %s\n",
> + canonical_vmlinux, strerror(errno));
> + pr_warn("was CONFIG_DEBUG_INFO_BTF enabled?\n");
> + } else {
> + err = faccessat(AT_FDCWD, canonical_vmlinux, R_OK, AT_EACCESS);
> + if (err) {
> + pr_warn("unable to read canonical vmlinux (%s): %s\n",
> + canonical_vmlinux, strerror(errno));
> + }
> + }
> + if (!err) {
> + /* load canonical and return any parsing failures */
> + btf = btf__parse(canonical_vmlinux, NULL);
> + err = libbpf_get_error(btf);
> + pr_debug("loading kernel BTF '%s': %d\n", canonical_vmlinux, err);
> + return btf;
> + }
>
> + /* try fallback locations */
> + uname(&buf);
> for (i = 0; i < ARRAY_SIZE(locations); i++) {
> snprintf(path, PATH_MAX, locations[i], buf.release);
> -
> if (faccessat(AT_FDCWD, path, R_OK, AT_EACCESS))
> continue;
>
> @@ -4965,9 +4984,9 @@ struct btf *btf__load_vmlinux_btf(void)
>
> return btf;
> }
> -
> pr_warn("failed to find valid kernel BTF\n");
> - return libbpf_err_ptr(-ESRCH);
> + /* return the last error or ESRCH if no fallback locations were found */
> + return btf ?: libbpf_err_ptr(-ESRCH);
> }
>
> struct btf *libbpf_find_kernel_btf(void) __attribute__((alias("btf__load_vmlinux_btf")));
> --
> 2.43.0.429.g432eaa2c6b-goog
>