On Thu, Nov 11, 2021 at 9:02 PM Alexei Starovoitov
<alexei.starovoitov@xxxxxxxxx> wrote:
>
> From: Alexei Starovoitov <ast@xxxxxxxxxx>
>
> Given BPF program's BTF perform a linear search through kernel BTFs for
> a possible candidate.
> Then wire the result into bpf_core_apply_relo_insn().
>
> Signed-off-by: Alexei Starovoitov <ast@xxxxxxxxxx>
> ---
> kernel/bpf/btf.c | 138 ++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 137 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index efb7fa2f81a2..aeb591579282 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -25,6 +25,7 @@
> #include <linux/kobject.h>
> #include <linux/sysfs.h>
> #include <net/sock.h>
> +#include "../tools/lib/bpf/relo_core.h"
>
> /* BTF (BPF Type Format) is the meta data format which describes
> * the data types of BPF program/map. Hence, it basically focus
> @@ -6440,9 +6441,144 @@ size_t bpf_core_essential_name_len(const char *name)
> return n;
> }
>
> +static void bpf_core_free_cands(struct bpf_core_cand_list *cands)
> +{
> + if (!cands)
> + return;
> + kfree(cands->cands);
> + kfree(cands);
indentation is off?
> +}
> +
> +static int bpf_core_add_cands(struct bpf_verifier_log *log,
> + struct bpf_core_cand *local_cand,
and here?
> + size_t local_essent_len,
> + const struct btf *targ_btf,
> + int targ_start_id,
> + struct bpf_core_cand_list *cands)
> +{
> + struct bpf_core_cand *new_cands, *cand;
> + const struct btf_type *t;
> + const char *targ_name;
> + size_t targ_essent_len;
> + int n, i;
> +
[...]
> int bpf_core_relo_apply(struct bpf_verifier_log *log, const struct btf *btf,
> const struct bpf_core_relo *relo, int relo_idx,
> void *insn)
> {
> - return -EOPNOTSUPP;
> + struct bpf_core_cand_list *cands = NULL;
> + int err;
> +
> + if (relo->kind != BPF_CORE_TYPE_ID_LOCAL) {
> + cands = bpf_core_find_cands(log, btf, relo->type_id);
> + if (IS_ERR(cands)) {
> + bpf_log(log, "target candidate search failed for %d\n",
> + relo->type_id);
> + return PTR_ERR(cands);
some indentation issues here as well
> + }
> + }
> + err = bpf_core_apply_relo_insn((void *)log, insn, relo->insn_off / 8,
> + relo, relo_idx, btf, cands);
> + bpf_core_free_cands(cands);
Why did you decide to not persist the candidate list? It is a
significant slowdown even on moderately large BPF programs, as you are
linearly re-searching vmlinux BTF multiple times for the same root
type.
> + return err;
> }
> --
> 2.30.2
>
