Hi Giulio,
Am Donnerstag, 26. März 2020, 03:02:39 CEST schrieb Giulio Benetti:
> Il 26/03/2020 01:05, Heiko Stübner ha scritto:
> > Am Donnerstag, 26. März 2020, 00:47:38 CET schrieb Giulio Benetti:
> >> very cleaner way to handle TEMT as a capability!
> >> And I've found one thing...
> >>
> >> Il 26/03/2020 00:14, Heiko Stuebner ha scritto:
> >>> From: Giulio Benetti <giulio.benetti@xxxxxxxxxxxxxxxx>
> >>>
> >>> Some 8250 ports have a TEMT interrupt but it's not a part of the 8250
> >>> standard, instead only available on some implementations.
> >>>
> >>> The current em485 implementation does not work on ports without it.
> >>> The only chance to make it work is to loop-read on LSR register.
> >>>
> >>> So add UART_CAP_TEMT to mark 8250 uarts having this interrupt,
> >>> update all current em485 users with that capability and make
> >>> the stop_tx function loop-read on uarts not having it.
> >>>
> >>> Signed-off-by: Giulio Benetti <giulio.benetti@xxxxxxxxxxxxxxxx>
> >>> [moved to use added UART_CAP_TEMT, use readx_poll_timeout]
> >>> Signed-off-by: Heiko Stuebner <heiko.stuebner@xxxxxxxxxxxxxxxxxxxxx>
> >>> ---
> >>> drivers/tty/serial/8250/8250.h | 1 +
> >>> drivers/tty/serial/8250/8250_bcm2835aux.c | 2 +-
> >>> drivers/tty/serial/8250/8250_of.c | 2 ++
> >>> drivers/tty/serial/8250/8250_omap.c | 2 +-
> >>> drivers/tty/serial/8250/8250_port.c | 25 +++++++++++++++++++----
> >>> 5 files changed, 26 insertions(+), 6 deletions(-)
> >>>
> >>> diff --git a/drivers/tty/serial/8250/8250.h b/drivers/tty/serial/8250/8250.h
> >>> index 52bb21205bb6..770eb00db497 100644
> >>> --- a/drivers/tty/serial/8250/8250.h
> >>> +++ b/drivers/tty/serial/8250/8250.h
> >>> @@ -82,6 +82,7 @@ struct serial8250_config {
> >>> #define UART_CAP_MINI (1 << 17) /* Mini UART on BCM283X family lacks:
> >>> * STOP PARITY EPAR SPAR WLEN5 WLEN6
> >>> */
> >>> +#define UART_CAP_TEMT (1 << 18) /* UART has TEMT interrupt */
> >>>
> >>> #define UART_BUG_QUOT (1 << 0) /* UART has buggy quot LSB */
> >>> #define UART_BUG_TXEN (1 << 1) /* UART has buggy TX IIR status */
> >>> diff --git a/drivers/tty/serial/8250/8250_bcm2835aux.c b/drivers/tty/serial/8250/8250_bcm2835aux.c
> >>> index 12d03e678295..3881242424ca 100644
> >>> --- a/drivers/tty/serial/8250/8250_bcm2835aux.c
> >>> +++ b/drivers/tty/serial/8250/8250_bcm2835aux.c
> >>> @@ -91,7 +91,7 @@ static int bcm2835aux_serial_probe(struct platform_device *pdev)
> >>> return -ENOMEM;
> >>>
> >>> /* initialize data */
> >>> - up.capabilities = UART_CAP_FIFO | UART_CAP_MINI;
> >>> + up.capabilities = UART_CAP_FIFO | UART_CAP_MINI | UART_CAP_TEMT;
> >>> up.port.dev = &pdev->dev;
> >>> up.port.regshift = 2;
> >>> up.port.type = PORT_16550;
> >>> diff --git a/drivers/tty/serial/8250/8250_of.c b/drivers/tty/serial/8250/8250_of.c
> >>> index 65e9045dafe6..841f6fcb2878 100644
> >>> --- a/drivers/tty/serial/8250/8250_of.c
> >>> +++ b/drivers/tty/serial/8250/8250_of.c
> >>> @@ -225,6 +225,8 @@ static int of_platform_serial_probe(struct platform_device *ofdev)
> >>> &port8250.overrun_backoff_time_ms) != 0)
> >>> port8250.overrun_backoff_time_ms = 0;
> >>>
> >>> + port8250.capabilities |= UART_CAP_TEMT;
> >>> +
> >>
> >> Shouldn't this be NOT UART_CAP_TEMT set by default? On all other
> >> vendor specific files you enable it, I think here you shouldn't enable
> >> it too by default. Right?
> >
> > 8250_of does use the em485 emulation - see of_platform_serial_setup()
> > So I did go by the lazy assumption that any 8250 driver using rs485
> > before my series always used the interrupt driver code path, so
> > implicitly required to have the TEMT interrupt.
> >
> > Of course, you're right that with the 8250_of maybe not all variants
> > actually do have this interrupt, so falling back to the polling here might
> > be safer.
>
> Probably here it's worth introducing a dt boolean property like
> "temt-capability", then you set or not UART_CAP_TEMT according to its
> presence in dts. This way all cases are covered and we can act
> completely through dts files.
>
> What about that?
Sorry that this was sitting around for over a month.
I think there are two problems with this:
(1) this would break backwards compatibility ... right now the whole code
just assumes that everyone does support the TEMT interrupt, so adding
a property to keep it working would break old DTs, which is something that
should not happen ... I guess one option would be to use the inverse
no-temt-interrupt
(2) uarts handled by 8250_of are still identified by their compatible
though and there is no generic 8250-of compatible, so the
presence / absence of the temt capability should actually just be
bound to the relevant compatible.
So my "gut feeling" is to just keep the current way
(was expecting temt-capability before anyway) until an uart
variant without temt comes along
Heiko
