Re: [PATCH net-next 1/4] bpf: allow bpf programs to tail-call other bpf programs

From: Andy Lutomirski <luto@amacapital.net>
To: Alexei Starovoitov <ast@plumgrid.com>
Cc: "David S. Miller" <davem@davemloft.net>,
	Ingo Molnar <mingo@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Michael Holzheu <holzheu@linux.vnet.ibm.com>,
	Zi Shen Lim <zlim.lnx@gmail.com>,
	Linux API <linux-api@vger.kernel.org>,
	Network Development <netdev@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH net-next 1/4] bpf: allow bpf programs to tail-call other bpf programs
Date: Thu, 21 May 2015 09:20:46 -0700	[thread overview]
Message-ID: <CALCETrVNwOVRJ4KH6mEo4hPq_QjTCb8mcwRf1sy5kd1nmnSZ+A@mail.gmail.com> (raw)
In-Reply-To: <555BD2E4.5050608@plumgrid.com>

On Tue, May 19, 2015 at 5:18 PM, Alexei Starovoitov <ast@plumgrid.com> wrote:
> On 5/19/15 5:13 PM, Andy Lutomirski wrote:
>>
>>
>> IMO this is starting to get a bit ugly.  Would it be possible to have
>> the program dereference the subprogram reference itself from the jump
>> table?  There would have to be a verifier type that represents a
>> reference to a program tail-call entry point, but that seems better
>> than having this weird indirection.
>
>
> Which part? I don't think you've looked at examples yet.
> network parser has to call itself. Otherwise we cannot parse 10 mpls
> labels or TLVs.
> Indirection via jump_table also has to be there.
> We need to dynamically add and remove programs form this jump table.
> It cannot be all static.
>

What I mean is: why do we need the interface to be "look up this index
in an array and just to what it references" as a single atomic
instruction?  Can't we break it down into first "look up this index in
an array" and then "do this tail call"?

I don't see why everything needs to be a map.

--Andy