Re: [PATCH 4/8] serial: 8250: Handle case port doesn't have TEMT interrupt using em485.

[Andy Shevchenko]

On Mon, 2018-06-04 at 13:50 +0200, Giulio Benetti wrote:
> Hi,
> 
> Il 04/06/2018 13:38, Andy Shevchenko ha scritto:
> > On Mon, 2018-06-04 at 12:50 +0200, Giulio Benetti wrote:
> > > Hi,
> > > 
> > > Il 04/06/2018 12:17, Andy Shevchenko ha scritto:
> > > > On Fri, 2018-06-01 at 14:40 +0200, Giulio Benetti wrote:
> > > > > Some 8250 ports only have TEMT interrupt, so current
> > > > > implementation
> > > > > can't work for ports without it. The only chance to make it
> > > > > work
> > > > > is to
> > > > > loop-read on LSR register.
> > > > > 
> > > > > With NO TEMT interrupt check if both TEMT and THRE are set
> > > > > looping
> > > > > on
> > > > > LSR register.
> > > > > --- a/drivers/tty/serial/8250/8250_dw.c
> > > > > +++ b/drivers/tty/serial/8250/8250_dw.c
> > > > > -		int ret = serial8250_em485_init(up);
> > > > > +		int ret = serial8250_em485_init(up, false);
> > > > 
> > > > Is true for all possible DW configured types? Or it's your
> > > > particular
> > > > case?
> > > > 
> > > 
> > > I've checked on Synopsis Designware 8250 datasheet and it's not
> > > supported.
> > > Here is datasheet I went through:
> > > https://linux-sunxi.org/images/d/d2/Dw_apb_uart_db.pdf
> > > 
> > > There seems not to be TEMT interrupt, I use it under sunxi SoC and
> > > on
> > > their datasheet(A20 for example), they don't report that interrupt
> > > too.
> > > So it seems to be valid for all DW configured types, anyway I
> > > don't
> > > know
> > > how many IP reviews there could be of that peripheral.
> > 
> > This is an excerpt from the document you referred to:
> > 
> > --- 8< --- 8< ---
> > 
> > 6 TEMT R Transmitter Empty bit. If in FIFO mode (FIFO_MODE != NONE)
> > and
> > FIFOs enabled (FCR[0] set to one), this bit is set whenever the
> > Transmitter Shift Register and the FIFO are both empty. If in non-
> > FIFO
> > mode or FIFOs are disabled, this bit is set whenever the Transmitter
> > Holding Register and the Transmitter Shift Register are both empty.
> > 
> > Reset Value: 0x1
> > 
> > --- 8< --- 8< ---
> > 
> > 
> > If I'm reading this correctly the support is there. Or otherwise,
> > care
> > to point exact paragraph needs to be read and checked?
> 
> In the beginning I thought the same as you but
> unfortunately LSR is only a status register and IER doesn't have 
> corresponding TEMT bit to enable an interrupt on TEMT triggering.
> On OMAP instead there is a specific interrupt bound to TEMT LSR flag.
> And THRE interrupt is not enough because shift register won't be
> empty 
> when it triggers, so you would loose some bit of last byte to be 
> transmitted.

Hmm... Okay, it's something you and Matwey better to discuss.

P.S. Latest version of document I have does describe RS485 HW supported
mode. I don't know if it was added recently to the IP itself, or just
missed documentation. That's what you need to clarify with Synopsys.

-- 
Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Intel Finland Oy