Re: [RFC PATCH v9 05/11] bpf: net_sched: Support implementation of Qdisc_ops in bpf

On Thu, Jul 25, 2024 at 2:25 PM Martin KaFai Lau [off-list ref] wrote:

On 7/14/24 10:51 AM, Amery Hung wrote:

quoted

+static const struct bpf_func_proto *
+bpf_qdisc_get_func_proto(enum bpf_func_id func_id,
+                      const struct bpf_prog *prog)
+{
+     switch (func_id) {

Instead of an empty switch, it should be useful to provide the skb->data related
helper. It can start with read only dynptr first, the BPF_FUNC_dynptr_read
helper here.

Also, the kfuncs: bpf_dynptr_slice and bpf_dynptr_from_skb_rdonly.

I will add the helper and kfuncs and try them out.

quoted

+     default:
+             return bpf_base_func_proto(func_id, prog);

[ ... ]

quoted

+     }
+}
+
+BTF_ID_LIST_SINGLE(bpf_sk_buff_ids, struct, sk_buff)
+BTF_ID_LIST_SINGLE(bpf_sk_buff_ptr_ids, struct, bpf_sk_buff_ptr)
+
+static bool bpf_qdisc_is_valid_access(int off, int size,
+                                   enum bpf_access_type type,
+                                   const struct bpf_prog *prog,
+                                   struct bpf_insn_access_aux *info)
+{
+     struct btf *btf = prog->aux->attach_btf;
+     u32 arg;
+
+     arg = get_ctx_arg_idx(btf, prog->aux->attach_func_proto, off);
+     if (!strcmp(prog->aux->attach_func_name, "enqueue")) {
+             if (arg == 2) {
+                     info->reg_type = PTR_TO_BTF_ID | PTR_TRUSTED;
+                     info->btf = btf;
+                     info->btf_id = bpf_sk_buff_ptr_ids[0];
+                     return true;

This will allow type == BPF_WRITE to ctx which should be rejected. The below
bpf_tracing_btf_ctx_access() could have rejected it.

Right. I will check the access type of the "to_free" argument in .enqueue.

quoted

+             }
+     }
+
+     return bpf_tracing_btf_ctx_access(off, size, type, prog, info);
+}
+

[ ... ]

quoted

+
+static bool is_unsupported(u32 member_offset)
+{
+     unsigned int i;
+
+     for (i = 0; i < ARRAY_SIZE(unsupported_ops); i++) {
+             if (member_offset == unsupported_ops[i])
+                     return true;
+     }
+
+     return false;
+}
+
+static int bpf_qdisc_check_member(const struct btf_type *t,
+                               const struct btf_member *member,
+                               const struct bpf_prog *prog)
+{
+     if (is_unsupported(__btf_member_bit_offset(t, member) / 8))

Note that the ".check_member" and the "is_unsupported" can be removed as you
also noticed on the recent unsupported ops cleanup patches.

Thanks for looping me in. I removed them when testing the series.

quoted

+             return -ENOTSUPP;
+     return 0;
+}

[ ... ]

quoted

+static struct Qdisc_ops __bpf_ops_qdisc_ops = {
+     .enqueue = Qdisc_ops__enqueue,
+     .dequeue = Qdisc_ops__dequeue,
+     .peek = Qdisc_ops__peek,
+     .init = Qdisc_ops__init,
+     .reset = Qdisc_ops__reset,
+     .destroy = Qdisc_ops__destroy,
+     .change = Qdisc_ops__change,
+     .attach = Qdisc_ops__attach,
+     .change_tx_queue_len = Qdisc_ops__change_tx_queue_len,
+     .change_real_num_tx = Qdisc_ops__change_real_num_tx,
+     .dump = Qdisc_ops__dump,
+     .dump_stats = Qdisc_ops__dump_stats,

Similar to the above is_unsupported comment. The unsupported ops should be
removed from the cfi_stubs.

quoted

+     .ingress_block_set = Qdisc_ops__ingress_block_set,
+     .egress_block_set = Qdisc_ops__egress_block_set,
+     .ingress_block_get = Qdisc_ops__ingress_block_get,
+     .egress_block_get = Qdisc_ops__egress_block_get,
+};
+
+static struct bpf_struct_ops bpf_Qdisc_ops = {
+     .verifier_ops = &bpf_qdisc_verifier_ops,
+     .reg = bpf_qdisc_reg,
+     .unreg = bpf_qdisc_unreg,
+     .check_member = bpf_qdisc_check_member,
+     .init_member = bpf_qdisc_init_member,
+     .init = bpf_qdisc_init,
+     .validate = bpf_qdisc_validate,

".validate" is optional. The empty "bpf_qdisc_validate" can be removed.

Got it.

quoted

+     .name = "Qdisc_ops",
+     .cfi_stubs = &__bpf_ops_qdisc_ops,
+     .owner = THIS_MODULE,
+};