linux-kernel - Re: [PATCH bpf-next v3 1/7] bpf: Refactor trampoline update code

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:   Wed, 4 Mar 2020 19:44:41 +0100
From:   KP Singh <kpsingh@...omium.org>
To:     Daniel Borkmann <daniel@...earbox.net>
Cc:     linux-security-module@...r.kernel.org,
        linux-kernel@...r.kernel.org, bpf@...r.kernel.org,
        Andrii Nakryiko <andriin@...com>,
        Alexei Starovoitov <ast@...nel.org>,
        Paul Turner <pjt@...gle.com>, Jann Horn <jannh@...gle.com>,
        Florent Revest <revest@...omium.org>,
        Brendan Jackman <jackmanb@...omium.org>
Subject: Re: [PATCH bpf-next v3 1/7] bpf: Refactor trampoline update code

On 04-Mär 19:37, Daniel Borkmann wrote:
> On 3/4/20 4:47 PM, KP Singh wrote:
> > From: KP Singh <kpsingh@...gle.com>
> > 
> > As we need to introduce a third type of attachment for trampolines, the
> > flattened signature of arch_prepare_bpf_trampoline gets even more
> > complicated.
> > 
> > Refactor the prog and count argument to arch_prepare_bpf_trampoline to
> > use bpf_tramp_progs to simplify the addition and accounting for new
> > attachment types.
> > 
> > Signed-off-by: KP Singh <kpsingh@...gle.com>
> > Acked-by: Andrii Nakryiko <andriin@...com>
> 
> [...]
> > diff --git a/kernel/bpf/bpf_struct_ops.c b/kernel/bpf/bpf_struct_ops.c
> > index c498f0fffb40..9f7e0328a644 100644
> > --- a/kernel/bpf/bpf_struct_ops.c
> > +++ b/kernel/bpf/bpf_struct_ops.c
> > @@ -320,6 +320,7 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> >   	struct bpf_struct_ops_value *uvalue, *kvalue;
> >   	const struct btf_member *member;
> >   	const struct btf_type *t = st_ops->type;
> > +	struct bpf_tramp_progs *tprogs = NULL;
> >   	void *udata, *kdata;
> >   	int prog_fd, err = 0;
> >   	void *image;
> > @@ -425,10 +426,18 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> >   			goto reset_unlock;
> >   		}
> > +		tprogs = kcalloc(BPF_TRAMP_MAX, sizeof(*tprogs), GFP_KERNEL);
> > +		if (!tprogs) {
> > +			err = -ENOMEM;
> > +			goto reset_unlock;
> > +		}
> > +
> 
> Looking over the code again, I'm quite certain that here's a memleak
> since the kcalloc() is done in the for_each_member() loop in the ops
> update but then going out of scope and in the exit path we only kfree
> the last tprogs.

You're right, nice catch. Fixing it.

- KP

> 
> > +		tprogs[BPF_TRAMP_FENTRY].progs[0] = prog;
> > +		tprogs[BPF_TRAMP_FENTRY].nr_progs = 1;
> >   		err = arch_prepare_bpf_trampoline(image,
> >   						  st_map->image + PAGE_SIZE,
> >   						  &st_ops->func_models[i], 0,
> > -						  &prog, 1, NULL, 0, NULL);
> > +						  tprogs, NULL);
> >   		if (err < 0)
> >   			goto reset_unlock;
> > @@ -469,6 +478,7 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> >   	memset(uvalue, 0, map->value_size);
> >   	memset(kvalue, 0, map->value_size);
> >   unlock:
> > +	kfree(tprogs);
> >   	mutex_unlock(&st_map->lock);
> >   	return err;
> >   }