Re: [PATCH net] bpf, sockmap: fix the sk->sk_forward_alloc warning of sk_stream_kill_queues()

[wangyufen <wangyufen@huawei.com>]

在 2022/10/29 6:17, John Fastabend 写道:
> Wang Yufen wrote:
>> When running `test_sockmap` selftests, got the following warning:
>>
>> WARNING: CPU: 2 PID: 197 at net/core/stream.c:205 sk_stream_kill_queues+0xd3/0xf0
>> Call Trace:
>>    <TASK>
>>    inet_csk_destroy_sock+0x55/0x110
>>    tcp_rcv_state_process+0xd28/0x1380
>>    ? tcp_v4_do_rcv+0x77/0x2c0
>>    tcp_v4_do_rcv+0x77/0x2c0
>>    __release_sock+0x106/0x130
>>    __tcp_close+0x1a7/0x4e0
>>    tcp_close+0x20/0x70
>>    inet_release+0x3c/0x80
>>    __sock_release+0x3a/0xb0
>>    sock_close+0x14/0x20
>>    __fput+0xa3/0x260
>>    task_work_run+0x59/0xb0
>>    exit_to_user_mode_prepare+0x1b3/0x1c0
>>    syscall_exit_to_user_mode+0x19/0x50
>>    do_syscall_64+0x48/0x90
>>    entry_SYSCALL_64_after_hwframe+0x44/0xae
>>
>> The root case is: In commit 84472b436e76 ("bpf, sockmap: Fix more
>> uncharged while msg has more_data") , I used msg->sg.size replace
>> tosend rudely, which break the
>>     if (msg->apply_bytes && msg->apply_bytes < send)
>> scene.
> Ah nice catch. Feel free to add my ACK on a v2 with small typo fixup.
>
> Acked-by: John Fastabend <john.fastabend@gmail.com>
>
>> Fixes: 84472b436e76 ("bpf, sockmap: Fix more uncharged while msg has more_data")
>> Reported-by: Jakub Sitnicki <jakub@cloudflare.com>
>> Signed-off-by: Wang Yufen <wangyufen@huawei.com>
>> ---
>>   net/ipv4/tcp_bpf.c | 10 ++++++----
>>   1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
>> index a1626af..38d4735 100644
>> --- a/net/ipv4/tcp_bpf.c
>> +++ b/net/ipv4/tcp_bpf.c
>> @@ -278,7 +278,7 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>>   {
>>   	bool cork = false, enospc = sk_msg_full(msg);
>>   	struct sock *sk_redir;
>> -	u32 tosend, delta = 0;
>> +	u32 tosend, orgsize, sended, delta = 0;
>>   	u32 eval = __SK_NONE;
>>   	int ret;
>>   
>> @@ -333,10 +333,12 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>>   			cork = true;
>>   			psock->cork = NULL;
>>   		}
>> -		sk_msg_return(sk, msg, msg->sg.size);
>> +		sk_msg_return(sk, msg, tosend);
>>   		release_sock(sk);
>>   
>> +		orgsize = msg->sg.size;
>>   		ret = tcp_bpf_sendmsg_redir(sk_redir, msg, tosend, flags);
>> +		sended = orgsize - msg->sg.size;
> Small english nitpick. Past tense of send is sent so could we make this,
>
>                  sent = orgsize - msg->sg.size;

I got it. Thanks.

>
>>   
>>   		if (eval == __SK_REDIRECT)
>>   			sock_put(sk_redir);
>> @@ -374,8 +376,8 @@ static int tcp_bpf_send_verdict(struct sock *sk, struct sk_psock *psock,
>>   		if (msg &&
>>   		    msg->sg.data[msg->sg.start].page_link &&
>>   		    msg->sg.data[msg->sg.start].length) {
>> -			if (eval == __SK_REDIRECT)
>> -				sk_mem_charge(sk, msg->sg.size);
>> +			if (eval == __SK_REDIRECT && tosend > sended)
> Other nit, you could probably omit the 'tosend > sended' check here. Because
> otherwise tosend == sended and the mem_charge of zer is a nop. But OTOH
> its probably ok to keep the check to avoid some extra work.
>
>> +				sk_mem_charge(sk, tosend - sended);
>>   			goto more_data;
>>   		}
>>   	}
>> -- 
>> 1.8.3.1