| Message ID | 20211120075451.16764-1-guodaxing@huawei.com (mailing list archive) |
|---|---|
| State | Awaiting Upstream |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | net/smc: loop in smc_listen | expand |
On Sat, Nov 20, 2021 at 03:54:51PM +0800, Daxing Guo wrote: > From: Guo DaXing <guodaxing@huawei.com> > > The kernel_listen function in smc_listen will fail when all the available > ports are occupied. At this point smc->clcsock->sk->sk_data_ready has > been changed to smc_clcsock_data_ready. When we call smc_listen again, > now both smc->clcsock->sk->sk_data_ready and smc->clcsk_data_ready point > to the smc_clcsock_data_ready function. > > The smc_clcsock_data_ready() function calls lsmc->clcsk_data_ready which > now points to itself resulting in an infinite loop. > > This patch restores smc->clcsock->sk->sk_data_ready with the old value. Hi Guo, This indeed seems to be an issue. When listen fails, the original clcsock's sk_data_ready overwrites by smc_clcsock_data_ready and can't be recovered. I will also test it in my environment, thanks. Cheers, Tony Lu > > Signed-off-by: Guo DaXing <guodaxing@huawei.com> > --- > net/smc/af_smc.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c > index 59284da9116d..078f5edf6d4d 100644 > --- a/net/smc/af_smc.c > +++ b/net/smc/af_smc.c > @@ -2120,8 +2120,10 @@ static int smc_listen(struct socket *sock, int backlog) > smc->clcsock->sk->sk_user_data = > (void *)((uintptr_t)smc | SK_USER_DATA_NOCOPY); > rc = kernel_listen(smc->clcsock, backlog); > - if (rc) > + if (rc) { > + smc->clcsock->sk->sk_data_ready = smc->clcsk_data_ready; > goto out; > + } > sk->sk_max_ack_backlog = backlog; > sk->sk_ack_backlog = 0; > sk->sk_state = SMC_LISTEN; > -- > 2.20.1
On 20/11/2021 08:54, Daxing Guo wrote: > From: Guo DaXing <guodaxing@huawei.com> > > The kernel_listen function in smc_listen will fail when all the available > ports are occupied. At this point smc->clcsock->sk->sk_data_ready has > been changed to smc_clcsock_data_ready. When we call smc_listen again, > now both smc->clcsock->sk->sk_data_ready and smc->clcsk_data_ready point > to the smc_clcsock_data_ready function. > > The smc_clcsock_data_ready() function calls lsmc->clcsk_data_ready which > now points to itself resulting in an infinite loop. > > This patch restores smc->clcsock->sk->sk_data_ready with the old value. > > Signed-off-by: Guo DaXing <guodaxing@huawei.com> > --- Thanks for your patch, I will pick it up and submit it to the net tree.
On Sat, Nov 20, 2021 at 03:54:51PM +0800, Daxing Guo wrote: > From: Guo DaXing <guodaxing@huawei.com> > > The kernel_listen function in smc_listen will fail when all the available > ports are occupied. At this point smc->clcsock->sk->sk_data_ready has > been changed to smc_clcsock_data_ready. When we call smc_listen again, > now both smc->clcsock->sk->sk_data_ready and smc->clcsk_data_ready point > to the smc_clcsock_data_ready function. > > The smc_clcsock_data_ready() function calls lsmc->clcsk_data_ready which > now points to itself resulting in an infinite loop. > > This patch restores smc->clcsock->sk->sk_data_ready with the old value. > > Signed-off-by: Guo DaXing <guodaxing@huawei.com> It works in my testing environment, thanks. Acked-by: Tony Lu <tonylu@linux.alibaba.com>
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index 59284da9116d..078f5edf6d4d 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -2120,8 +2120,10 @@ static int smc_listen(struct socket *sock, int backlog) smc->clcsock->sk->sk_user_data = (void *)((uintptr_t)smc | SK_USER_DATA_NOCOPY); rc = kernel_listen(smc->clcsock, backlog); - if (rc) + if (rc) { + smc->clcsock->sk->sk_data_ready = smc->clcsk_data_ready; goto out; + } sk->sk_max_ack_backlog = backlog; sk->sk_ack_backlog = 0; sk->sk_state = SMC_LISTEN;