Re: [PATCH net] net/smc: fix general protection fault in __smc_diag_dump

[Wang Liang <wangliang74@huawei.com>]

在 2025/4/3 19:55, Wenjia Zhang 写道:
>
>
> On 31.03.25 10:10, Wang Liang wrote:
>> Syzbot reported a general protection fault:
>>
>>    CPU: 0 UID: 0 PID: 5830 Comm: syz-executor600 Not tainted 
>> 6.14.0-rc4-syzkaller-00090-gdd83757f6e68 #0
>>    RIP: 0010:smc_diag_msg_common_fill net/smc/smc_diag.c:44 [inline]
>>    RIP: 0010:__smc_diag_dump.constprop.0+0x3de/0x23d0 
>> net/smc/smc_diag.c:89
>>    Call Trace:
>>     <TASK>
>>     smc_diag_dump_proto+0x26d/0x420 net/smc/smc_diag.c:217
>>     smc_diag_dump+0x84/0x90 net/smc/smc_diag.c:236
>>     netlink_dump+0x53c/0xd00 net/netlink/af_netlink.c:2318
>>     __netlink_dump_start+0x6ca/0x970 net/netlink/af_netlink.c:2433
>>     netlink_dump_start include/linux/netlink.h:340 [inline]
>>     smc_diag_handler_dump+0x1fb/0x240 net/smc/smc_diag.c:251
>>     __sock_diag_cmd net/core/sock_diag.c:249 [inline]
>>     sock_diag_rcv_msg+0x437/0x790 net/core/sock_diag.c:287
>>     netlink_rcv_skb+0x16b/0x440 net/netlink/af_netlink.c:2543
>>     netlink_unicast_kernel net/netlink/af_netlink.c:1322 [inline]
>>     netlink_unicast+0x53c/0x7f0 net/netlink/af_netlink.c:1348
>>     netlink_sendmsg+0x8b8/0xd70 net/netlink/af_netlink.c:1892
>>     sock_sendmsg_nosec net/socket.c:718 [inline]
>>     __sock_sendmsg net/socket.c:733 [inline]
>>     ____sys_sendmsg+0xaaf/0xc90 net/socket.c:2573
>>     ___sys_sendmsg+0x135/0x1e0 net/socket.c:2627
>>     __sys_sendmsg+0x16e/0x220 net/socket.c:2659
>>     do_syscall_x64 arch/x86/entry/common.c:52 [inline]
>>     do_syscall_64+0xcd/0x250 arch/x86/entry/common.c:83
>>     entry_SYSCALL_64_after_hwframe+0x77/0x7f
>>     </TASK>
>>
>> When create smc socket, smc_inet_init_sock() first add sk to the 
>> smc_hash
>> by smc_hash_sk(), then create smc->clcsock. it is possible that, after
>> smc_diag_dump_proto() traverses the smc_hash, smc->clcsock is not 
>> created
>> when the function visit it.
>>
>> The process like this:
>>
>>    (CPU1)                         | (CPU2)
>>    inet6_create()                 |
>>      smc_inet_init_sock()         |
>>        smc_sk_init()              |
>>          smc_hash_sk()            |
>>            head = &smc_hash->ht;  |
>>            sk_add_node(sk, head); |
>>                                   | smc_diag_dump_proto
>>                                   |   head = &smc_hash->ht;
>>                                   |     sk_for_each(sk, head)
>>                                   |       __smc_diag_dump()
>>                                   |         visit smc->clcsock
>>        smc_create_clcsk()         |
>>            set smc->clcsock       |
>>
>> Fix this by initialize smc->clcsock to NULL before add sk to smc_hash in
>> smc_sk_init().
>>
>> Reported-by: syzbot+271fed3ed6f24600c364@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=271fed3ed6f24600c364
>> Fixes: f16a7dd5cf27 ("smc: netlink interface for SMC sockets")
>> Signed-off-by: Wang Liang <wangliang74@huawei.com>
>> ---
>>   net/smc/af_smc.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
>> index 3e6cb35baf25..454801188514 100644
>> --- a/net/smc/af_smc.c
>> +++ b/net/smc/af_smc.c
>> @@ -371,6 +371,7 @@ void smc_sk_init(struct net *net, struct sock 
>> *sk, int protocol)
>>       sk->sk_protocol = protocol;
>>       WRITE_ONCE(sk->sk_sndbuf, 2 * READ_ONCE(net->smc.sysctl_wmem));
>>       WRITE_ONCE(sk->sk_rcvbuf, 2 * READ_ONCE(net->smc.sysctl_rmem));
>> +    smc->clcsock = NULL;
>>       INIT_WORK(&smc->tcp_listen_work, smc_tcp_listen_work);
>>       INIT_WORK(&smc->connect_work, smc_connect_work);
>>       INIT_DELAYED_WORK(&smc->conn.tx_work, smc_tx_work);
>
> I have to agree with this workaround, even though I see that is not 
> the best solution. Thus, I'd like to give my R-b:
>
> Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
>
> Btw. @D. Wythe, would you mind sending me the link of your proposal 
> you mentioned please? Let me have a look. It seems like I missed it.
>
> Thanks,
> Wenjia
>

Hello, is this patch rejected?
If there are some new fix patchs, please let me know.
Thanks.

>
>