serial: sh-sci: prevent lockup on full TTY buffers
authorUlrich Hecht <ulrich.hecht+renesas@gmail.com>
Thu, 15 Feb 2018 12:02:27 +0000 (13:02 +0100)
committerBen Hutchings <ben@decadent.org.uk>
Thu, 31 May 2018 23:30:19 +0000 (00:30 +0100)
commit 7842055bfce4bf0170d0f61df8b2add8399697be upstream.

When the TTY buffers fill up to the configured maximum, a system lockup
occurs:

[  598.820128] INFO: rcu_preempt detected stalls on CPUs/tasks:
[  598.825796]  0-...!: (1 GPs behind) idle=5a6/2/0 softirq=1974/1974 fqs=1
[  598.832577]  (detected by 3, t=62517 jiffies, g=296, c=295, q=126)
[  598.838755] Task dump for CPU 0:
[  598.841977] swapper/0       R  running task        0     0      0 0x00000022
[  598.849023] Call trace:
[  598.851476]  __switch_to+0x98/0xb0
[  598.854870]            (null)

This can be prevented by doing a dummy read of the RX data register.

This issue affects both HSCIF and SCIF ports. Reported for R-Car H3 ES2.0;
reproduced and fixed on H3 ES1.1. Probably affects other R-Car platforms
as well.

Reported-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Ulrich Hecht <ulrich.hecht+renesas@gmail.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Nguyen Viet Dung <dung.nguyen.aj@renesas.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2:
 - Use sci_in()
 - Adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
drivers/tty/serial/sh-sci.c

index 60f71ec..4b73525 100644 (file)
@@ -668,6 +668,8 @@ static void sci_receive_chars(struct uart_port *port)
                /* Tell the rest of the system the news. New characters! */
                tty_flip_buffer_push(tty);
        } else {
+               /* TTY buffers full; read from RX reg to prevent lockup */
+               sci_in(port, SCxRDR);
                sci_in(port, SCxSR); /* dummy read */
                sci_out(port, SCxSR, SCxSR_RDxF_CLEAR(port));
        }