Re: eBPF verifier thoughts (Re: [PATCH v15 net-next 00/11] eBPF syscall, verifier, testsuite)

From: Andy Lutomirski
Date: Fri Sep 26 2014 - 17:49:24 EST

On Fri, Sep 26, 2014 at 2:46 PM, Alexei Starovoitov <ast@xxxxxxxxxxxx> wrote:
> On Fri, Sep 26, 2014 at 1:42 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
>> On Fri, Sep 26, 2014 at 1:09 PM, Alexei Starovoitov <ast@xxxxxxxxxxxx> wrote:
>>> On Fri, Sep 26, 2014 at 12:51 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
>>>> On Fri, Sep 26, 2014 at 12:34 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
>>>> To add one more point:
>>>> With the current verifier design, it's impossible to write a userspace
>>>> tool that can take an eBPF program and check it. The verification is
>>>> far too context-dependent for that to be possible. I won't go so far
>>>> as to say that a userspace tool needs to *exist*, but I strongly
>>>> object to exposing a verification algorithm that *precludes* writing
>>>> such a tool.
>>> that's just not true.
>>> why is it not possible?
>> Because the types of referenced objects aren't encoded in the blob
>> that a user program loads, unless I'm missing something.
> patch #8 'handle pseudo BPF_LD_IMM64 insn' of this set
> handles first type == map. Other types will be added in the future.
> The same verification can be done in user space.
> It's pretty much copy paste for everything from the kernel.
> I don't understand yet why you really must do it in in userspace
> in addition to doing it in kernel. It's definitely doable.
> Instead of asking kernel to create a map, user space
> can just remember map attributes (key_size, value_size)
> and continues verification in userspace.
>> But the eBPF binary doesn't encode this information. In fact, the
>> caller of an ebpf syscall may not even have access to this
>> information.
> I don't follow. What info are you talking about?
> are you saying that program only that references maps via fds
> is not verifiable unless one knows what this fds refer to?
> yeah, but we're talking user space verification here.
> user knows what maps it creates with what attributes.
> Also we can add a command to this syscall to fetch map
> attributes. That would be trivial _incremental_ addition, right?

That would also work, I suppose.

>> I think this is addressable as a smallish change on top of your code.
>> Rather than looking up a map when you need to learn its key and value
>> size, I think that all you need to do is to look in a program section
>> for the key and value size (and the fact that it's a map) and confirm
>> that the referenced map *matches* the stored values.
> we can add extra info to the program that will encode
> program assumptions about maps. Sure. Though I think
> it's extra info that kernel doesn't really need, since it can
> only check that program assumptions match to what
> kernel already knows. Kernel cannot rely on them.
> So I'm not sure what this extra check really buys.
> Anyway, if you think it's a smallish change, we can do it
> incrementally on top of existing stuff, right?
> Why this arguing then?
> Sounds like you want to help with the development?
> This is great! I'm all for it :)

Andy Lutomirski
AMA Capital Management, LLC
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at