linux-kernel - Re: [PATCH -next] bpf, test_run: fix alignment problem in bpf_prog_test_run

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]

Message-ID: <20221101210542.724e3442@kernel.org>
Date:   Tue, 1 Nov 2022 21:05:42 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     zhongbaisong <zhongbaisong@...wei.com>
Cc:     Daniel Borkmann <daniel@...earbox.net>, <edumazet@...gle.com>,
        <davem@...emloft.net>, <pabeni@...hat.com>,
        <linux-kernel@...r.kernel.org>, <bpf@...r.kernel.org>,
        <netdev@...r.kernel.org>, <ast@...nel.org>, <song@...nel.org>,
        <yhs@...com>, <haoluo@...gle.com>,
        Alexander Potapenko <glider@...gle.com>,
        Marco Elver <elver@...gle.com>,
        Dmitry Vyukov <dvyukov@...gle.com>,
        Linux MM <linux-mm@...ck.org>, <kasan-dev@...glegroups.com>,
        Kees Cook <keescook@...omium.org>
Subject: Re: [PATCH -next] bpf, test_run: fix alignment problem in
 bpf_prog_test_run_skb()

On Wed, 2 Nov 2022 10:59:44 +0800 zhongbaisong wrote:
> On 2022/11/2 0:45, Daniel Borkmann wrote:
> > [ +kfence folks ]  
> 
> + cc: Alexander Potapenko, Marco Elver, Dmitry Vyukov
> 
> Do you have any suggestions about this problem?

+ Kees who has been sending similar patches for drivers

> > On 11/1/22 5:04 AM, Baisong Zhong wrote:  
> >> Recently, we got a syzkaller problem because of aarch64
> >> alignment fault if KFENCE enabled.
> >>
> >> When the size from user bpf program is an odd number, like
> >> 399, 407, etc, it will cause skb shard info's alignment access,
> >> as seen below:
> >>
> >> BUG: KFENCE: use-after-free read in __skb_clone+0x23c/0x2a0 
> >> net/core/skbuff.c:1032
> >>
> >> Use-after-free read at 0xffff6254fffac077 (in kfence-#213):
> >>   __lse_atomic_add arch/arm64/include/asm/atomic_lse.h:26 [inline]
> >>   arch_atomic_add arch/arm64/include/asm/atomic.h:28 [inline]
> >>   arch_atomic_inc include/linux/atomic-arch-fallback.h:270 [inline]
> >>   atomic_inc include/asm-generic/atomic-instrumented.h:241 [inline]
> >>   __skb_clone+0x23c/0x2a0 net/core/skbuff.c:1032
> >>   skb_clone+0xf4/0x214 net/core/skbuff.c:1481
> >>   ____bpf_clone_redirect net/core/filter.c:2433 [inline]
> >>   bpf_clone_redirect+0x78/0x1c0 net/core/filter.c:2420
> >>   bpf_prog_d3839dd9068ceb51+0x80/0x330
> >>   bpf_dispatcher_nop_func include/linux/bpf.h:728 [inline]
> >>   bpf_test_run+0x3c0/0x6c0 net/bpf/test_run.c:53
> >>   bpf_prog_test_run_skb+0x638/0xa7c net/bpf/test_run.c:594
> >>   bpf_prog_test_run kernel/bpf/syscall.c:3148 [inline]
> >>   __do_sys_bpf kernel/bpf/syscall.c:4441 [inline]
> >>   __se_sys_bpf+0xad0/0x1634 kernel/bpf/syscall.c:4381
> >>
> >> kfence-#213: 0xffff6254fffac000-0xffff6254fffac196, size=407, 
> >> cache=kmalloc-512
> >>
> >> allocated by task 15074 on cpu 0 at 1342.585390s:
> >>   kmalloc include/linux/slab.h:568 [inline]
> >>   kzalloc include/linux/slab.h:675 [inline]
> >>   bpf_test_init.isra.0+0xac/0x290 net/bpf/test_run.c:191
> >>   bpf_prog_test_run_skb+0x11c/0xa7c net/bpf/test_run.c:512
> >>   bpf_prog_test_run kernel/bpf/syscall.c:3148 [inline]
> >>   __do_sys_bpf kernel/bpf/syscall.c:4441 [inline]
> >>   __se_sys_bpf+0xad0/0x1634 kernel/bpf/syscall.c:4381
> >>   __arm64_sys_bpf+0x50/0x60 kernel/bpf/syscall.c:4381
> >>
> >> To fix the problem, we round up allocations with kmalloc_size_roundup()
> >> so that build_skb()'s use of kize() is always alignment and no special
> >> handling of the memory is needed by KFENCE.
> >>
> >> Fixes: 1cf1cae963c2 ("bpf: introduce BPF_PROG_TEST_RUN command")
> >> Signed-off-by: Baisong Zhong <zhongbaisong@...wei.com>
> >> ---
> >>   net/bpf/test_run.c | 1 +
> >>   1 file changed, 1 insertion(+)
> >>
> >> diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c
> >> index 13d578ce2a09..058b67108873 100644
> >> --- a/net/bpf/test_run.c
> >> +++ b/net/bpf/test_run.c
> >> @@ -774,6 +774,7 @@ static void *bpf_test_init(const union bpf_attr 
> >> *kattr, u32 user_size,
> >>       if (user_size > size)
> >>           return ERR_PTR(-EMSGSIZE);
> >> +    size = kmalloc_size_roundup(size);
> >>       data = kzalloc(size + headroom + tailroom, GFP_USER);  
> > 
> > The fact that you need to do this roundup on call sites feels broken, no?
> > Was there some discussion / consensus that now all k*alloc() call sites
> > would need to be fixed up? Couldn't this be done transparently in k*alloc()
> > when KFENCE is enabled? I presume there may be lots of other such occasions
> > in the kernel where similar issue triggers, fixing up all call-sites feels
> > like ton of churn compared to api-internal, generic fix.
> >   
> >>       if (!data)
> >>           return ERR_PTR(-ENOMEM);
> >>  
> > 
> > Thanks,
> > Daniel
> >  
> 
>