It is indeed okay to use system_wq at present. Dues to the load
balancing issues we found, queue_work() always submits tasks to the
worker on the current CPU. tcp_listen_work() execution once may submit a
large number of tasks to the worker of the current CPU, causing
unnecessary pending, even though worker on other CPU are totaly free. I
was plan to make tcp_listen_work() blocked wait on worker of every CPU,
so I create a new workqueue, and that's the only reason for it. But this
problem is not very urgent, and I don't have strong opinion too...
在 2022/2/9 上午1:06, Karsten Graul 写道:
On 08/02/2022 13:53, D. Wythe wrote:
+static struct workqueue_struct *smc_tcp_ls_wq; /* wq for tcp listen work */
struct workqueue_struct *smc_hs_wq; /* wq for handshake work */
struct workqueue_struct *smc_close_wq; /* wq for close work */
@@ -2227,7 +2228,7 @@ static void smc_clcsock_data_ready(struct sock *listen_clcsock)
lsmc->clcsk_data_ready(listen_clcsock);
if (lsmc->sk.sk_state == SMC_LISTEN) {
sock_hold(&lsmc->sk); /* sock_put in smc_tcp_listen_work() */
- if (!queue_work(smc_hs_wq, &lsmc->tcp_listen_work))
+ if (!queue_work(smc_tcp_ls_wq, &lsmc->tcp_listen_work))
sock_put(&lsmc->sk);
It works well this way, but given the fact that there is one tcp_listen worker per
listen socket and these workers finish relatively quickly, wouldn't it be okay to
use the system_wq instead of using an own queue? But I have no strong opinion about that...
