Re: [PATCH bpf-next v3 1/2] libbpf: Introduce 'btf_custom_path' to 'bpf_obj_open_opts'

From: Andrii Nakryiko
Date: Mon Jul 12 2021 - 19:53:37 EST


On Thu, Jul 8, 2021 at 7:48 PM Shuyi Cheng <chengshuyi@xxxxxxxxxxxxxxxxx> wrote:
>
> btf_custom_path allows developers to load custom BTF, and subsequent
> CO-RE will use custom BTF for relocation.
>
> Learn from Andrii's comments in [0], add the btf_custom_path parameter
> to bpf_obj_open_opts, you can directly use the skeleton's
> <objname>_bpf__open_opts function to pass in the btf_custom_path
> parameter.
>
> Prior to this, there was also a developer who provided a patch with
> similar functions. It is a pity that the follow-up did not continue to
> advance. See [1].
>
> [0]https://lore.kernel.org/bpf/CAEf4BzbJZLjNoiK8_VfeVg_Vrg=9iYFv+po-38SMe=UzwDKJ=Q@xxxxxxxxxxxxxx/#t
> [1]https://yhbt.net/lore/all/CAEf4Bzbgw49w2PtowsrzKQNcxD4fZRE6AKByX-5-dMo-+oWHHA@xxxxxxxxxxxxxx/
>
> Signed-off-by: Shuyi Cheng <chengshuyi@xxxxxxxxxxxxxxxxx>
> ---
> tools/lib/bpf/libbpf.c | 46 ++++++++++++++++++++++++++++++++++++++++++++--
> tools/lib/bpf/libbpf.h | 6 +++++-
> 2 files changed, 49 insertions(+), 3 deletions(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 1e04ce7..6702b7f 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -498,6 +498,10 @@ struct bpf_object {
> * it at load time.
> */
> struct btf *btf_vmlinux;
> + /* custom BTF is in addition to vmlinux BTF (i.e., Use the CO-RE
> + * feature in the old kernel).
> + */

nit: "path to custom BTF used for CO-RE relocations" as a description?

> + char *btf_custom_path;
> /* vmlinux BTF override for CO-RE relocations */
> struct btf *btf_vmlinux_override;
> /* Lazily initialized kernel module BTFs */
> @@ -2668,6 +2672,27 @@ static bool obj_needs_vmlinux_btf(const struct bpf_object *obj)
> return false;
> }
>
> +static int bpf_object__load_override_btf(struct bpf_object *obj)
> +{
> + int err;
> +
> + if (obj->btf_vmlinux_override)
> + return 0;
> +
> + if (!obj->btf_custom_path)
> + return 0;
> +
> + obj->btf_vmlinux_override = btf__parse(obj->btf_custom_path, NULL);
> + err = libbpf_get_error(obj->btf_vmlinux_override);
> + pr_debug("loading custom BTF '%s': %d\n", obj->btf_custom_path, err);
> + if (err) {
> + pr_warn("failed to parse custom BTF\n");
> + obj->btf_vmlinux_override = NULL;
> + }
> +
> + return err;
> +}

see below, I don't think we need this function

> +
> static int bpf_object__load_vmlinux_btf(struct bpf_object *obj, bool force)
> {
> int err;
> @@ -7554,7 +7579,7 @@ int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
> __bpf_object__open(const char *path, const void *obj_buf, size_t obj_buf_sz,
> const struct bpf_object_open_opts *opts)
> {
> - const char *obj_name, *kconfig;
> + const char *obj_name, *kconfig, *btf_tmp_path;
> struct bpf_program *prog;
> struct bpf_object *obj;
> char tmp_name[64];
> @@ -7584,6 +7609,19 @@ int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
> obj = bpf_object__new(path, obj_buf, obj_buf_sz, obj_name);
> if (IS_ERR(obj))
> return obj;
> +
> + btf_tmp_path = OPTS_GET(opts, btf_custom_path, NULL);
> + if (btf_tmp_path) {
> + if (strlen(btf_tmp_path) >= PATH_MAX) {
> + err = -ENAMETOOLONG;
> + goto out;
> + }
> + obj->btf_custom_path = strdup(btf_tmp_path);
> + if (!obj->btf_custom_path) {
> + err = -ENOMEM;
> + goto out;
> + }
> + }

this part looks good

>
> kconfig = OPTS_GET(opts, kconfig, NULL);
> if (kconfig) {
> @@ -8049,6 +8087,7 @@ int bpf_object__load_xattr(struct bpf_object_load_attr *attr)
> bpf_gen__init(obj->gen_loader, attr->log_level);
>
> err = bpf_object__probe_loading(obj);
> + err = err ? : bpf_object__load_override_btf(obj);

btf_vmlinux_override is already working properly, so all you need to
do here (once you remembered btf_custom_path) is to pass it to
bpf_object__relocate. All the rest is taken care of, so you don't need
to add extra cleanup (except for zfree(btf_custom_path)).

> err = err ? : bpf_object__load_vmlinux_btf(obj, false);
> err = err ? : bpf_object__resolve_externs(obj, obj->kconfig);
> err = err ? : bpf_object__sanitize_and_load_btf(obj);

so few lines below this code, after bpf_object__create_maps(obj):

err = err ? : bpf_object__relocate(obj, obj->btf_custom_path ?:
attr->target_btf_path);

This way open_opts->btf_custom_path serves as an override for
load_attr's target_btf_path (which we are going to deprecate anyways).

The only remaining thing is to make sure that
bpf_object__load_vmlinux_btf() won't attempt to load real vmlinux BTF
*just for CO-RE*. So in obj_needs_vmlinux_btf() make sure to not
return true for CO-RE relocations if custom_btf_path is specified (see
Vamsi's patch which has this logic already).

> @@ -8075,9 +8114,11 @@ int bpf_object__load_xattr(struct bpf_object_load_attr *attr)
> }
> free(obj->btf_modules);
>
> - /* clean up vmlinux BTF */
> + /* clean up vmlinux BTF and custom BTF*/
> btf__free(obj->btf_vmlinux);
> obj->btf_vmlinux = NULL;
> + btf__free(obj->btf_vmlinux_override);
> + obj->btf_vmlinux_override = NULL;

this shouldn't be necessary, bpf_object__relocate_core() handled this

>
> obj->loaded = true; /* doesn't matter if successfully or not */
>
> @@ -8702,6 +8743,7 @@ void bpf_object__close(struct bpf_object *obj)
> for (i = 0; i < obj->nr_maps; i++)
> bpf_map__destroy(&obj->maps[i]);
>
> + zfree(&obj->btf_custom_path);
> zfree(&obj->kconfig);
> zfree(&obj->externs);
> obj->nr_extern = 0;
> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> index 6e61342..5002d1f 100644
> --- a/tools/lib/bpf/libbpf.h
> +++ b/tools/lib/bpf/libbpf.h
> @@ -94,8 +94,12 @@ struct bpf_object_open_opts {
> * system Kconfig for CONFIG_xxx externs.
> */
> const char *kconfig;
> + /* custom BTF is in addition to vmlinux BTF (i.e., Use the CO-RE
> + * feature in the old kernel).

Instead of "Use the CO-RE feature in the old kernel", let's point out
explicitly that this custom BTF is used *only* for CO-RE, and any
other feature that relies on kernel BTF will still need actual vmlinux
BTF. Something like this:

/* Path to the custom BTF to be used for BPF CO-RE relocations.
* This custom BTF completely replaces the use of vmlinux BTF
* for the purpose of CO-RE relocations.
* NOTE: any other BPF feature (e.g., fentry/fexit programs,
* struct_ops, etc) will need actual kernel BTF at /sys/kernel/btf/vmlinux.
*/

I think this will make it much clearer what's the intended purpose here.

> + */
> + char *btf_custom_path;
> };
> -#define bpf_object_open_opts__last_field kconfig
> +#define bpf_object_open_opts__last_field btf_custom_path
>
> LIBBPF_API struct bpf_object *bpf_object__open(const char *path);
> LIBBPF_API struct bpf_object *
> --
> 1.8.3.1
>