| Message ID | 20181009174158.4476-1-geert+renesas@glider.be (mailing list archive) |
|---|---|
| State | Accepted |
| Delegated to: | Geert Uytterhoeven |
| Headers | show |
| Series | serial: sh-sci: Fix receive on SCIFA/SCIFB variants with DMA | expand |
On Tue, Oct 09, 2018 at 07:41:58PM +0200, Geert Uytterhoeven wrote: > On SCIFA and SCIFB serial ports with DMA support (i.e. some ports on > R-Car Gen2 and RZ/G1 SoCs), receive DMA operations are submitted before > the DMA channel pointer is initialized. Hence this fails, and the > driver tries to fall back to PIO. However, at this early phase in the > initialization sequence, fallback to PIO does not work, leading to a > serial port that cannot receive any data. > > Fix this by calling sci_submit_rx() after initialization of the DMA > channel pointer. > > Reported-by: Jinzai Solution and RVC Test Teams via Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Brainstorming: keep Shimoda-san's address as is and mention the test teams in the commit message? > Fixes: 2c4ee23530ffc022 ("serial: sh-sci: Postpone DMA release when falling back to PIO") > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com> > --- > Reported on r8a7790/lager using SCIFA1 on Debug Serial 1. I hope to test this on Lager this weekend. I am quite sure it will work now, but then we have this formally checked, too. > Tested on r8a7791/koelsch using SCIFA3 on EXIO-B.
> > Fixes: 2c4ee23530ffc022 ("serial: sh-sci: Postpone DMA release when falling back to PIO") > > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> > > Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com> > > > --- > > Reported on r8a7790/lager using SCIFA1 on Debug Serial 1. > > I hope to test this on Lager this weekend. I am quite sure it will work > now, but then we have this formally checked, too. Done now: Tested-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c index b73b27dc45e60ffb..7e98a4f3ec77abbd 100644 --- a/drivers/tty/serial/sh-sci.c +++ b/drivers/tty/serial/sh-sci.c @@ -1614,10 +1614,10 @@ static void sci_request_dma(struct uart_port *port) hrtimer_init(&s->rx_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); s->rx_timer.function = rx_timer_fn; + s->chan_rx_saved = s->chan_rx = chan; + if (port->type == PORT_SCIFA || port->type == PORT_SCIFB) sci_submit_rx(s); - - s->chan_rx_saved = s->chan_rx = chan; } }
On SCIFA and SCIFB serial ports with DMA support (i.e. some ports on R-Car Gen2 and RZ/G1 SoCs), receive DMA operations are submitted before the DMA channel pointer is initialized. Hence this fails, and the driver tries to fall back to PIO. However, at this early phase in the initialization sequence, fallback to PIO does not work, leading to a serial port that cannot receive any data. Fix this by calling sci_submit_rx() after initialization of the DMA channel pointer. Reported-by: Jinzai Solution and RVC Test Teams via Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Fixes: 2c4ee23530ffc022 ("serial: sh-sci: Postpone DMA release when falling back to PIO") Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> --- Reported on r8a7790/lager using SCIFA1 on Debug Serial 1. Tested on r8a7791/koelsch using SCIFA3 on EXIO-B. This fixes a regression introduced in v4.19-rc1, so please queue for v4.19. Thanks! --- drivers/tty/serial/sh-sci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)