Re: [PATCH net] ipv4: Fix uninit-value access in __ip_make_skb()

[Shigeru Yoshida]

On Mon, 25 Mar 2024 20:56:43 +0100, Eric Dumazet wrote:
> On Mon, Mar 25, 2024 at 6:46 PM Shigeru Yoshida <syoshida@xxxxxxxxxx> wrote:
>>
>> On Mon, 25 Mar 2024 11:05:33 +0100, Eric Dumazet wrote:
>> > On Mon, Mar 25, 2024 at 10:38 AM Shigeru Yoshida <syoshida@xxxxxxxxxx> wrote:
>> >>
>> >> On Mon, 25 Mar 2024 10:01:25 +0100, Eric Dumazet wrote:
>> >> > On Sun, Mar 24, 2024 at 6:06 AM Shigeru Yoshida <syoshida@xxxxxxxxxx> wrote:
>> >> >>
>> >> >> KMSAN reported uninit-value access in __ip_make_skb() [1].  __ip_make_skb()
>> >> >> tests HDRINCL to know if the skb has icmphdr. However, HDRINCL can cause a
>> >> >> race condition. If calling setsockopt(2) with IP_HDRINCL changes HDRINCL
>> >> >> while __ip_make_skb() is running, the function will access icmphdr in the
>> >> >> skb even if it is not included. This causes the issue reported by KMSAN.
>> >> >>
>> >> >> Check FLOWI_FLAG_KNOWN_NH on fl4->flowi4_flags instead of testing HDRINCL
>> >> >> on the socket.
>> >> >>
>> >> >> [1]
>> >> >
>> >> > What is the kernel version for this trace ?
>> >>
>> >> Sorry, I used the following version:
>> >>
>> >> CPU: 1 PID: 15709 Comm: syz-executor.7 Not tainted 6.8.0-11567-gb3603fcb79b1 #25
>> >> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-1.fc39 04/01/2014
>> >>
>> >> >> BUG: KMSAN: uninit-value in __ip_make_skb+0x2b74/0x2d20 net/ipv4/ip_output.c:1481
>> >> >>  __ip_make_skb+0x2b74/0x2d20 net/ipv4/ip_output.c:1481
>> >> >>  ip_finish_skb include/net/ip.h:243 [inline]
>> >> >>  ip_push_pending_frames+0x4c/0x5c0 net/ipv4/ip_output.c:1508
>> >> >>  raw_sendmsg+0x2381/0x2690 net/ipv4/raw.c:654
>> >> >>  inet_sendmsg+0x27b/0x2a0 net/ipv4/af_inet.c:851
>> >> >>  sock_sendmsg_nosec net/socket.c:730 [inline]
>> >> >>  __sock_sendmsg+0x274/0x3c0 net/socket.c:745
>> >> >>  __sys_sendto+0x62c/0x7b0 net/socket.c:2191
>> >> >>  __do_sys_sendto net/socket.c:2203 [inline]
>> >> >>  __se_sys_sendto net/socket.c:2199 [inline]
>> >> >>  __x64_sys_sendto+0x130/0x200 net/socket.c:2199
>> >> >>  do_syscall_64+0xd8/0x1f0 arch/x86/entry/common.c:83
>> >> >>  entry_SYSCALL_64_after_hwframe+0x6d/0x75
>> >> >>
>> >> >> Uninit was created at:
>> >> >>  slab_post_alloc_hook mm/slub.c:3804 [inline]
>> >> >>  slab_alloc_node mm/slub.c:3845 [inline]
>> >> >>  kmem_cache_alloc_node+0x5f6/0xc50 mm/slub.c:3888
>> >> >>  kmalloc_reserve+0x13c/0x4a0 net/core/skbuff.c:577
>> >> >>  __alloc_skb+0x35a/0x7c0 net/core/skbuff.c:668
>> >> >>  alloc_skb include/linux/skbuff.h:1318 [inline]
>> >> >>  __ip_append_data+0x49ab/0x68c0 net/ipv4/ip_output.c:1128
>> >> >>  ip_append_data+0x1e7/0x260 net/ipv4/ip_output.c:1365
>> >> >>  raw_sendmsg+0x22b1/0x2690 net/ipv4/raw.c:648
>> >> >>  inet_sendmsg+0x27b/0x2a0 net/ipv4/af_inet.c:851
>> >> >>  sock_sendmsg_nosec net/socket.c:730 [inline]
>> >> >>  __sock_sendmsg+0x274/0x3c0 net/socket.c:745
>> >> >>  __sys_sendto+0x62c/0x7b0 net/socket.c:2191
>> >> >>  __do_sys_sendto net/socket.c:2203 [inline]
>> >> >>  __se_sys_sendto net/socket.c:2199 [inline]
>> >> >>  __x64_sys_sendto+0x130/0x200 net/socket.c:2199
>> >> >>  do_syscall_64+0xd8/0x1f0 arch/x86/entry/common.c:83
>> >> >>  entry_SYSCALL_64_after_hwframe+0x6d/0x75
>> >> >>
>> >> >> Fixes: 99e5acae193e ("ipv4: Fix potential uninit variable access bug in __ip_make_skb()")
>> >> >> Reported-by: syzkaller <syzkaller@xxxxxxxxxxxxxxxx>
>> >> >> Signed-off-by: Shigeru Yoshida <syoshida@xxxxxxxxxx>
>> >> >> ---
>> >> >> I think IPv6 has a similar issue. If this patch is accepted, I will send
>> >> >> a patch for IPv6.
>> >> >> ---
>> >> >>  net/ipv4/ip_output.c | 2 +-
>> >> >>  1 file changed, 1 insertion(+), 1 deletion(-)
>> >> >>
>> >> >> diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
>> >> >> index 1fe794967211..39229fd0601a 100644
>> >> >> --- a/net/ipv4/ip_output.c
>> >> >> +++ b/net/ipv4/ip_output.c
>> >> >> @@ -1473,7 +1473,7 @@ struct sk_buff *__ip_make_skb(struct sock *sk,
>> >> >>                  * by icmp_hdr(skb)->type.
>> >> >>                  */
>> >> >>                 if (sk->sk_type == SOCK_RAW &&
>> >> >> -                   !inet_test_bit(HDRINCL, sk))
>> >> >> +                   !(fl4->flowi4_flags & FLOWI_FLAG_KNOWN_NH))
>> >> >>                         icmp_type = fl4->fl4_icmp_type;
>> >> >>                 else
>> >> >>                         icmp_type = icmp_hdr(skb)->type;
>> >> >> --
>> >> >> 2.44.0
>> >> >>
>> >> >
>> >> > Thanks for your patch.
>> >> >
>> >> > I do not think this is enough, as far as syzkaller is concerned.
>> >> >
>> >> > raw_probe_proto_opt() can leave garbage in fl4_icmp_type (and fl4_icmp_code)
>> >>
>> >> Thank you for your comment. But I don't understand it clearly. What
>> >> exactly do you mean by "garbage"?
>> >>
>> >> raw_probe_proto_opt() immediately returns 0 if fl4->flowi4_proto is
>> >> not IPPROTO_ICMP:
>> >>
>> >> static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
>> >> {
>> >>         int err;
>> >>
>> >>         if (fl4->flowi4_proto != IPPROTO_ICMP)
>> >>                 return 0;
>> >>
>> >> In this case, the function doesn't set fl4_icmp_type. Do you mean this
>> >> case?
>> >
>> > There are multiple ways to return early from this function.
>> >
>> > In all of them, fl4->fl4_icmp_type is left uninitialized, so syzbot
>> > will find ways to trigger a related bug,
>> > if you assume later that fl4->fl4_icmp_type contains valid (initialized) data.
>>
>> Thank you for your reply. I see your point.
>>
>> fl4->fl4_icmp_type is part of flowi_uli union in flowi4 structure, and
>> flowi4_init_output() initializes fl4_dport and fl4_sport to zero.
>>
>> I thought this also initializes fl4_icmp_type and fl4_icmp_code. Do
>> you think we should initialize fl4_icmp_type and fl4_icmp_code
>> explicitly, otherwise am I misunderstanding?
> 
> Yes, I am precisely saying this : do not rely on some union layout,
> without mentioning it in the changelog
> or even better in a comment.
> 
> If you want to avoid clearing these fields, please add a
> BUILD_BUG_ON() to make sure
> a unrelated future change in include/net/flow.h does not break a
> hidden assumption.
> 
> (ie : clearing fl4_dport and fl4_sport also clears fl4_icmp_type and
> fl4_icmp_code.)

Sorry for the late response.

Thank you Eric! I got it. I will try to make v2 patch this weekend.

Thanks,
Shigeru