Re: [PATCH] bpf: cgroup: remove WARN_ON at bpf_cgroup_link_release

From: Dongliang Mu <mudongliangabcd@gmail.com>
To: Daniel Borkmann <daniel@iogearbox.net>
Cc: Dongliang Mu <dzm91@hust.edu.cn>,
	Alexei Starovoitov <ast@kernel.org>,
	Andrii Nakryiko <andrii@kernel.org>,
	Martin KaFai Lau <kafai@fb.com>, Song Liu <songliubraving@fb.com>,
	Yonghong Song <yhs@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>,
	syzkaller <syzkaller@googlegroups.com>,
	netdev@vger.kernel.org, bpf@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] bpf: cgroup: remove WARN_ON at bpf_cgroup_link_release
Date: Wed, 2 Mar 2022 13:46:04 +0800	[thread overview]
Message-ID: <CAD-N9QUptT_h4FtjPTNyc72jrUHW0R_-Ggf3R1V3Fsz4hoyuAw@mail.gmail.com> (raw)
In-Reply-To: <d4bed569-d448-8b59-0774-c036e4c9abe9@iogearbox.net>

On Wed, Mar 2, 2022 at 12:28 AM Daniel Borkmann <daniel@iogearbox.net> wrote:
>
> On 2/27/22 2:40 PM, Dongliang Mu wrote:
> > From: Dongliang Mu <mudongliangabcd@gmail.com>
> >
> > When syzkaller injects fault into memory allocation at
> > bpf_prog_array_alloc, the kernel encounters a memory failure and
> > returns non-zero, thus leading to one WARN_ON at
> > bpf_cgroup_link_release. The stack trace is as follows:
> >
> >   __kmalloc+0x7e/0x3d0
> >   bpf_prog_array_alloc+0x4f/0x60
> >   compute_effective_progs+0x132/0x580
> >   ? __sanitizer_cov_trace_pc+0x1a/0x40
> >   update_effective_progs+0x5e/0x260
> >   __cgroup_bpf_detach+0x293/0x760
> >   bpf_cgroup_link_release+0xad/0x400
> >   bpf_link_free+0xca/0x190
> >   bpf_link_put+0x161/0x1b0
> >   bpf_link_release+0x33/0x40
> >   __fput+0x286/0x9f0
> >
> > Fix this by removing the WARN_ON for __cgroup_bpf_detach.
> >
> > Reported-by: syzkaller <syzkaller@googlegroups.com>
> > Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
> > ---
> >   kernel/bpf/cgroup.c | 4 ++--
> >   1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
> > index 514b4681a90a..fdbdcee6c9fa 100644
> > --- a/kernel/bpf/cgroup.c
> > +++ b/kernel/bpf/cgroup.c
> > @@ -896,8 +896,8 @@ static void bpf_cgroup_link_release(struct bpf_link *link)
> >               return;
> >       }
> >
> > -     WARN_ON(__cgroup_bpf_detach(cg_link->cgroup, NULL, cg_link,
> > -                                 cg_link->type));
> > +     __cgroup_bpf_detach(cg_link->cgroup, NULL, cg_link,
> > +                                 cg_link->type);
>
> "Fixing" by removing WARN_ON is just papering over the issue which in this case as
> mentioned is allocation failure on detach/teardown when allocating and recomputing
> effective prog arrays..

Hi Daniel,

you're right. This is not a good fix, any idea to fix the underlying
bug perfectly?

>
> >       cg = cg_link->cgroup;
> >       cg_link->cgroup = NULL;
> >
>