netdev - Re: [PATCH bpf-next 1/2] net: xdp: refactor XDP_QUERY_PROG{,

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <87pno935yb.fsf@toke.dk>
Date:   Thu, 23 May 2019 10:55:56 +0200
From:   Toke Høiland-Jørgensen <toke@...hat.com>
To:     Jakub Kicinski <jakub.kicinski@...ronome.com>,
        Björn Töpel 
        <bjorn.topel@...il.com>
Cc:     Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Netdev <netdev@...r.kernel.org>,
        Björn Töpel 
        <bjorn.topel@...el.com>,
        "Karlsson\, Magnus" <magnus.karlsson@...el.com>,
        Jesper Dangaard Brouer <brouer@...hat.com>,
        bpf <bpf@...r.kernel.org>
Subject: Re: [PATCH bpf-next 1/2] net: xdp: refactor XDP_QUERY_PROG{,_HW} to netdev

Jakub Kicinski <jakub.kicinski@...ronome.com> writes:

> On Wed, 22 May 2019 22:54:44 +0200, Björn Töpel wrote:
>> > > Now, the same commands give:
>> > >
>> > >   # ip link set dev eth0 xdp obj foo.o sec main
>> > >   # ip link set dev eth0 xdpgeneric off
>> > >   Error: native and generic XDP can't be active at the same time.  
>> >
>> > I'm not clear why this change is necessary? It is a change in
>> > behaviour, and if anything returning ENOENT would seem cleaner
>> > in this case.
>> 
>> To me, the existing behavior was non-intuitive. If most people *don't*
>> agree, I'll remove this change. So, what do people think about this?
>> :-)
>
> Having things start to fail after they were successful/ignored
> is one of those ABI breakage types Linux and netdev usually takes
> pretty seriously, unfortunately.  Especially when motivation is 
> "it's more intuitive" :)
>
> If nobody chimes in please break out this behaviour change into 
> a commit of its own.

Björn and I already had this discussion off-list. I think we ended up
concluding that since it's not changing kernel *behaviour*, but just
making an existing error explicit, it might be acceptable from an ABI
breakage PoV. And I'd generally prefer explicit errors over silent
failures.

But yeah, can totally see why it could also be considered a breaking
change...

-Toke