Re: [PATCH net] net/smc: fix illegal rmb_desc access in SMC-D connection dump

[Wen Gu <guwen@linux.alibaba.com>]

On 2024/1/18 09:50, Dust Li wrote:
> On Wed, Jan 17, 2024 at 08:27:49PM +0800, Wen Gu wrote:
>> A crash was found when dumping SMC-D connections. It can be reproduced
>> by following steps:
>>
>> - run nginx/wrk test:
>>   smc_run nginx
>>   smc_run wrk -t 16 -c 1000 -d <duration> -H 'Connection: Close' <URL>
>>
>> - continuously dump SMC-D connections in parallel:
>>   watch -n 1 'smcss -D'
>>
>> BUG: kernel NULL pointer dereference, address: 0000000000000030
>> CPU: 2 PID: 7204 Comm: smcss Kdump: loaded Tainted: G	E      6.7.0+ #55
>> RIP: 0010:__smc_diag_dump.constprop.0+0x5e5/0x620 [smc_diag]
>> Call Trace:
>>   <TASK>
>>   ? __die+0x24/0x70
>>   ? page_fault_oops+0x66/0x150
>>   ? exc_page_fault+0x69/0x140
>>   ? asm_exc_page_fault+0x26/0x30
>>   ? __smc_diag_dump.constprop.0+0x5e5/0x620 [smc_diag]
>>   ? __kmalloc_node_track_caller+0x35d/0x430
>>   ? __alloc_skb+0x77/0x170
>>   smc_diag_dump_proto+0xd0/0xf0 [smc_diag]
>>   smc_diag_dump+0x26/0x60 [smc_diag]
>>   netlink_dump+0x19f/0x320
>>   __netlink_dump_start+0x1dc/0x300
>>   smc_diag_handler_dump+0x6a/0x80 [smc_diag]
>>   ? __pfx_smc_diag_dump+0x10/0x10 [smc_diag]
>>   sock_diag_rcv_msg+0x121/0x140
>>   ? __pfx_sock_diag_rcv_msg+0x10/0x10
>>   netlink_rcv_skb+0x5a/0x110
>>   sock_diag_rcv+0x28/0x40
>>   netlink_unicast+0x22a/0x330
>>   netlink_sendmsg+0x1f8/0x420
>>   __sock_sendmsg+0xb0/0xc0
>>   ____sys_sendmsg+0x24e/0x300
>>   ? copy_msghdr_from_user+0x62/0x80
>>   ___sys_sendmsg+0x7c/0xd0
>>   ? __do_fault+0x34/0x160
>>   ? do_read_fault+0x5f/0x100
>>   ? do_fault+0xb0/0x110
>>   ? __handle_mm_fault+0x2b0/0x6c0
>>   __sys_sendmsg+0x4d/0x80
>>   do_syscall_64+0x69/0x180
>>   entry_SYSCALL_64_after_hwframe+0x6e/0x76
>>
>> It is possible that the connection is in process of being established
>> when we dump it. Assumed that the connection has been registered in a
>> link group by smc_conn_create() but the rmb_desc has not yet been
>> initialized by smc_buf_create(), thus causing the illegal access to
>> conn->rmb_desc. So fix it by checking before dump.
>>
>> Fixes: ce51f63e63c5 ("net/smc: Prevent kernel-infoleak in __smc_diag_dump()")
> 
> ce51f63e63c5 ("net/smc: Prevent kernel-infoleak in __smc_diag_dump()")
> only add a memset() of 'struct smcd_diag_dmbinfo dinfo', which I don't
> think is not the real cause of the bug.
> 

After re-checking the definition of Fixes tag, I agree that
ce51f63e63c5 is not appropriate and
4b1b7d3b30a6 ("net/smc: add SMC-D diag support") should be used.

Thank you!

>> Signed-off-by: Wen Gu <guwen@linux.alibaba.com>
>> ---
>> net/smc/smc_diag.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/net/smc/smc_diag.c b/net/smc/smc_diag.c
>> index 52f7c4f1e767..5a33908015f3 100644
>> --- a/net/smc/smc_diag.c
>> +++ b/net/smc/smc_diag.c
>> @@ -164,7 +164,7 @@ static int __smc_diag_dump(struct sock *sk, struct sk_buff *skb,
>> 	}
>> 	if (smc_conn_lgr_valid(&smc->conn) && smc->conn.lgr->is_smcd &&
>> 	    (req->diag_ext & (1 << (SMC_DIAG_DMBINFO - 1))) &&
>> -	    !list_empty(&smc->conn.lgr->list)) {
>> +	    !list_empty(&smc->conn.lgr->list) && smc->conn.rmb_desc) {
>> 		struct smc_connection *conn = &smc->conn;
>> 		struct smcd_diag_dmbinfo dinfo;
>> 		struct smcd_dev *smcd = conn->lgr->smcd;
>> -- 
>> 2.43.0