Hi,
* Dan Carpenter <dan.carpenter@xxxxxxxxxx> [220523 06:57]:
> Hello Tony Lindgren,
>
> The patch 925ea0fa5277: "tty: n_gsm: Fix packet data hex dump output"
> from May 12, 2022, leads to the following Smatch static checker
> warning:
>
> drivers/tty/n_gsm.c:720 gsm_data_kick()
> warn: sleeping in atomic context
>
> drivers/tty/n_gsm.c
> 698 static void gsm_data_kick(struct gsm_mux *gsm, struct gsm_dlci *dlci)
> 699 {
> 700 struct gsm_msg *msg, *nmsg;
> 701 int len;
> 702
> 703 list_for_each_entry_safe(msg, nmsg, &gsm->tx_list, list) {
> 704 if (gsm->constipated && msg->addr)
> 705 continue;
> 706 if (gsm->encoding != 0) {
> 707 gsm->txframe[0] = GSM1_SOF;
> 708 len = gsm_stuff_frame(msg->data,
> 709 gsm->txframe + 1, msg->len);
> 710 gsm->txframe[len + 1] = GSM1_SOF;
> 711 len += 2;
> 712 } else {
> 713 gsm->txframe[0] = GSM0_SOF;
> 714 memcpy(gsm->txframe + 1 , msg->data, msg->len);
> 715 gsm->txframe[msg->len + 1] = GSM0_SOF;
> 716 len = msg->len + 2;
> 717 }
> 718
> 719 if (debug & 4)
> --> 720 gsm_hex_dump_bytes(__func__, gsm->txframe, len);
>
> gsm_control_message() is holding a spin lock so gsm_hex_dump_bytes()
> might need to use GFP_ATOMIC instead of GFP_KERNEL.
OK thanks for the report. I'll send a patch for GFP_ATOMIC later on today.
Regards,
Tony
