Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <0000000000000a9f3605cb1d2455@google.com> <20210904042010.5henc6pn5zotdq6v@treble>
In-Reply-To: <20210904042010.5henc6pn5zotdq6v@treble>
From:   Jussi Maki <joamaki@gmail.com>
Date:   Sun, 5 Sep 2021 12:08:05 +0200
Message-ID: <CAHn8xcnF5FT+i5jwsZGswyYt9Q9Amnj8N07wK08Fe_doGsBBaA@mail.gmail.com>
Subject: Re: [syzbot] WARNING in __static_key_slow_dec
To:     Josh Poimboeuf <jpoimboe@redhat.com>
Cc:     syzbot <syzbot+30622fb04ddd72a4d167@syzkaller.appspotmail.com>,
        ardb@kernel.org, jbaron@akamai.com, linux-kernel@vger.kernel.org,
        peterz@infradead.org, rostedt@goodmis.org,
        syzkaller-bugs@googlegroups.com,
        Daniel Borkmann <daniel@iogearbox.net>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sat, Sep 4, 2021 at 6:20 AM Josh Poimboeuf <jpoimboe@redhat.com> wrote:
>
> Looks like a jump label usage bug ("jump label: negative count!") may
> have been introduced, presumably with the following commit:
>
>   9e2ee5c7e7c3 ("net, bonding: Add XDP support to the bonding driver")
>
> Jussi, any idea?

Hey, thanks for looping me in.

The code in bond_xdp_set is unsafely assuming that it won't get called
multiple times with prog=NULL and that's likely the issue here. The
assumption seems to mostly hold as I haven't been able to reproduce
this locally yet, even with nested bonding or with non-XDP slaves in
the mix. I'll look into this more next week and try to decipher what
steps syzbot is doing. As soon as I have a repro + fix I'll send it to
the net tree.