[bpf-next,1/2] bpf: Fix context type resolving for extension programs

Eelco reported we can't properly access arguments if the tracing
program is attached to extension program.

Having following program:

  SEC("classifier/test_pkt_md_access")
  int test_pkt_md_access(struct __sk_buff *skb)

with its extension:

  SEC("freplace/test_pkt_md_access")
  int test_pkt_md_access_new(struct __sk_buff *skb)

and tracing that extension with:

  SEC("fentry/test_pkt_md_access_new")
  int BPF_PROG(fentry, struct sk_buff *skb)

It's not possible to access skb argument in the fentry program,
with following error from verifier:

  ; int BPF_PROG(fentry, struct sk_buff *skb)
  0: (79) r1 = *(u64 *)(r1 +0)
  invalid bpf_context access off=0 size=8

The problem is that btf_ctx_access gets the context type for the
traced program, which is in this case the extension.

But when we trace extension program, we want to get the context
type of the program that the extension is attached to, so we can
access the argument properly in the trace program.

Reported-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 kernel/bpf/btf.c | 8 ++++++++
 1 file changed, 8 insertions(+)

Jiri Olsa <jolsa@kernel.org> writes:

> Eelco reported we can't properly access arguments if the tracing
> program is attached to extension program.
>
> Having following program:
>
>   SEC("classifier/test_pkt_md_access")
>   int test_pkt_md_access(struct __sk_buff *skb)
>
> with its extension:
>
>   SEC("freplace/test_pkt_md_access")
>   int test_pkt_md_access_new(struct __sk_buff *skb)
>
> and tracing that extension with:
>
>   SEC("fentry/test_pkt_md_access_new")
>   int BPF_PROG(fentry, struct sk_buff *skb)
>
> It's not possible to access skb argument in the fentry program,
> with following error from verifier:
>
>   ; int BPF_PROG(fentry, struct sk_buff *skb)
>   0: (79) r1 = *(u64 *)(r1 +0)
>   invalid bpf_context access off=0 size=8
>
> The problem is that btf_ctx_access gets the context type for the
> traced program, which is in this case the extension.
>
> But when we trace extension program, we want to get the context
> type of the program that the extension is attached to, so we can
> access the argument properly in the trace program.
>
> Reported-by: Eelco Chaudron <echaudro@redhat.com>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  kernel/bpf/btf.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
> index f9ac6935ab3c..37ad01c32e5a 100644
> --- a/kernel/bpf/btf.c
> +++ b/kernel/bpf/btf.c
> @@ -3859,6 +3859,14 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
>  	}
>  
>  	info->reg_type = PTR_TO_BTF_ID;
> +
> +	/* When we trace extension program, we want to get the context
> +	 * type of the program that the extension is attached to, so
> +	 * we can access the argument properly in the trace program.
> +	 */
> +	if (tgt_prog && tgt_prog->type == BPF_PROG_TYPE_EXT)
> +		tgt_prog = tgt_prog->aux->linked_prog;
> +

In the discussion about multi-attach for freplace we kinda concluded[0]
that this linked_prog pointer was going away after attach. I have this
basically working, but need to test a bit more before posting it (see
[1] for current status).

But with this I guess we'll need to either do something different? Maybe
go chase down the target via the bpf_link or something?

-Toke

[0] https://lore.kernel.org/bpf/20200722002918.574pruibvlxfblyq@ast-mbp.dhcp.thefacebook.com/
[1] https://git.kernel.org/pub/scm/linux/kernel/git/toke/linux.git/log/?h=bpf-freplace-multi-attach-alt-03

Alexei Starovoitov <alexei.starovoitov@gmail.com> writes:

> On Wed, Sep 9, 2020 at 8:11 AM Jiri Olsa <jolsa@kernel.org> wrote:
>>
>> Eelco reported we can't properly access arguments if the tracing
>> program is attached to extension program.
>>
>> Having following program:
>>
>>   SEC("classifier/test_pkt_md_access")
>>   int test_pkt_md_access(struct __sk_buff *skb)
>>
>> with its extension:
>>
>>   SEC("freplace/test_pkt_md_access")
>>   int test_pkt_md_access_new(struct __sk_buff *skb)
>>
>> and tracing that extension with:
>>
>>   SEC("fentry/test_pkt_md_access_new")
>>   int BPF_PROG(fentry, struct sk_buff *skb)
>>
>> It's not possible to access skb argument in the fentry program,
>> with following error from verifier:
>>
>>   ; int BPF_PROG(fentry, struct sk_buff *skb)
>>   0: (79) r1 = *(u64 *)(r1 +0)
>>   invalid bpf_context access off=0 size=8
>>
>> The problem is that btf_ctx_access gets the context type for the
>> traced program, which is in this case the extension.
>>
>> But when we trace extension program, we want to get the context
>> type of the program that the extension is attached to, so we can
>> access the argument properly in the trace program.
>>
>> Reported-by: Eelco Chaudron <echaudro@redhat.com>
>> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
>> ---
>>  kernel/bpf/btf.c | 8 ++++++++
>>  1 file changed, 8 insertions(+)
>>
>> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
>> index f9ac6935ab3c..37ad01c32e5a 100644
>> --- a/kernel/bpf/btf.c
>> +++ b/kernel/bpf/btf.c
>> @@ -3859,6 +3859,14 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
>>         }
>>
>>         info->reg_type = PTR_TO_BTF_ID;
>> +
>> +       /* When we trace extension program, we want to get the context
>> +        * type of the program that the extension is attached to, so
>> +        * we can access the argument properly in the trace program.
>> +        */
>> +       if (tgt_prog && tgt_prog->type == BPF_PROG_TYPE_EXT)
>> +               tgt_prog = tgt_prog->aux->linked_prog;
>> +
>>         if (tgt_prog) {
>>                 ret = btf_translate_to_vmlinux(log, btf, t, tgt_prog->type, arg);
>
> I think it would be cleaner to move resolve_prog_type() from verifier.c
> and use that helper function here.

FYI, I've added a different version of this patch to my freplace
multi-attach series (since the approach here was incompatible with
that).

-Toke