Hello Dan Carpenter,
Thanks for the feedback.
> -----Original Message-----
> From: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> Sent: Tuesday, July 4, 2023 2:20 PM
> To: Biju Das <biju.das.jz@xxxxxxxxxxxxxx>
> Cc: linux-serial@xxxxxxxxxxxxxxx; linux-renesas-soc@xxxxxxxxxxxxxxx
> Subject: [bug report] tty: serial: sh-sci: Add RZ/G2L SCIFA DMA tx
> support
>
> Hello Biju Das,
>
> The patch 8749061be196: "tty: serial: sh-sci: Add RZ/G2L SCIFA DMA tx
> support" from Apr 12, 2023, leads to the following Smatch static checker
> warning:
>
> drivers/tty/serial/sh-sci.c:593 sci_start_tx()
> warn: sleeping in atomic context
>
> drivers/tty/serial/sh-sci.c
> 573 static void sci_start_tx(struct uart_port *port)
> 574 {
> 575 struct sci_port *s = to_sci_port(port);
> 576 unsigned short ctrl;
> 577
> 578 #ifdef CONFIG_SERIAL_SH_SCI_DMA
> 579 if (port->type == PORT_SCIFA || port->type ==
> PORT_SCIFB) {
> 580 u16 new, scr = serial_port_in(port, SCSCR);
> 581 if (s->chan_tx)
> 582 new = scr | SCSCR_TDRQE;
> 583 else
> 584 new = scr & ~SCSCR_TDRQE;
> 585 if (new != scr)
> 586 serial_port_out(port, SCSCR, new);
> 587 }
> 588
> 589 if (s->chan_tx && !uart_circ_empty(&s->port.state->xmit)
> &&
> 590 dma_submit_error(s->cookie_tx)) {
> 591 if (s->cfg->regtype == SCIx_RZ_SCIFA_REGTYPE)
> 592 /* Switch irq from SCIF to DMA */
> --> 593 disable_irq(s->irqs[SCIx_TXI_IRQ]);
>
> disable_irq() is a might_sleep() function so it can't be called while
> holding a spinlock, but the sci_dma_tx_work_fn() function is holding a
> spinlock when it calls sci_start_tx().
Ok, will change disable_irq->disable_irq_nosync, so that it can be called from atomic context.
Cheers,
Biju
>
> 594
> 595 s->cookie_tx = 0;
> 596 schedule_work(&s->work_tx);
> 597 }
> 598 #endif
