On Tue, Aug 30, 2022 at 11:23:14PM +0800, liuyacan@xxxxxxxxxxxxxxxx wrote:
> From: Yacan Liu <liuyacan@xxxxxxxxxxxxxxxx>
>
> For passive connections, the refcount increment has been done in
> smc_clcsock_accept()-->smc_sock_alloc().
>
> Fixes: 3b2dec2603d5 ("net/smc: restructure client and server code in af_smc")
> Signed-off-by: Yacan Liu <liuyacan@xxxxxxxxxxxxxxxx>
>
> ---
> Change in v2:
> -- Tune commit message
> ---
> net/smc/af_smc.c | 1 -
> 1 file changed, 1 deletion(-)
>
> diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
> index 79c1318af..0939cc3b9 100644
> --- a/net/smc/af_smc.c
> +++ b/net/smc/af_smc.c
> @@ -1855,7 +1855,6 @@ static void smc_listen_out_connected(struct smc_sock *new_smc)
> {
> struct sock *newsmcsk = &new_smc->sk;
>
> - sk_refcnt_debug_inc(newsmcsk);
> if (newsmcsk->sk_state == SMC_INIT)
> newsmcsk->sk_state = SMC_ACTIVE;
>
Thanks for this fixes. I dig into this sk_refcnt_debug_* facility. It
seems this is a very old debug methods and doesn't help a lot for sock
leak issue. Maybe there is another method to help track this issue?
For this patch, It looks good for me and tested in our environment.
Reviewed-by: Tony Lu <tonylu@xxxxxxxxxxxxxxxxx>
Cheers,
Tony Lu
