Hi Mark,

This driver is still evolving, there's newer code on my lora-next branch
already: https://github.com/afaerber/linux/commits/lora-next

The reason you're in CC on this RFC is two-fold:

1) You applied Ben's patch to associate "semtech,sx1301" with spidev,
whereas I am now preparing a new driver for the same compatible.

2) This SPI device is in turn exposing the two SPI masters that you
already found below, and I didn't see a sane way to split that code out
into drivers/spi/, so it's in drivers/net/lora/ here - has there been
any precedence either way?

More inline ...

Am 02.07.2018 um 18:12 schrieb Mark Brown:
> On Sun, Jul 01, 2018 at 01:08:04PM +0200, Andreas Färber wrote:
> 
>> +static void sx1301_radio_spi_set_cs(struct spi_device *spi, bool enable)
>> +{
>> +	int ret;
>> +
>> +	dev_dbg(&spi->dev, "setting SPI CS to %s\n", enable ? "1" : "0");
>> +
>> +	if (enable)
>> +		return;
>> +
>> +	ret = sx1301_radio_set_cs(spi->controller, enable);
>> +	if (ret)
>> +		dev_warn(&spi->dev, "failed to write CS (%d)\n", ret);
>> +}
> 
> So we never disable chip select?

Not here, I instead did that in transfer_one below.

Unfortunately there seems to be no documentation, only reference code:

https://github.com/Lora-net/lora_gateway/blob/master/libloragw/src/loragw_radio.c#L121
https://github.com/Lora-net/lora_gateway/blob/master/libloragw/src/loragw_radio.c#L165

It sets CS to 0 before writing to address and data registers, then
immediately sets CS to 1 and back to 0 before reading or ending the
write transaction. I've tried to force the same behavior in this driver.
My guess was that CS is high-active during the short 1-0 cycle, because
if it's low-active during the register writes then why the heck is it
set to 0 again in the end instead of keeping at 1... confusing.

Maybe the Semtech folks CC'ed can comment how these registers work?

>> +	if (tx_buf) {
>> +		ret = sx1301_write(ssx->parent, ssx->regs + REG_RADIO_X_ADDR, tx_buf ? tx_buf[0] : 0);
> 
> This looks confused.  We're in an if (tx_buf) block but there's a use of
> the ternery operator that appears to be checking if we have a tx_buf?

Yeah, as mentioned this RFC is not ready for merging - checkpatch.pl
will complain about lines too long, and TODOs are sprinkled all over or
not even mentioned. It's a Proof of Concept that a net_device could work
for a wide range of spi and serdev based drivers, and on top this device
has more than one channel, which may influence network-level design
discussions.

That said, I'll happily drop the second check. Thanks for spotting!

>> +		if (ret) {
>> +			dev_err(&spi->dev, "SPI radio address write failed\n");
>> +			return ret;
>> +		}
>> +
>> +		ret = sx1301_write(ssx->parent, ssx->regs + REG_RADIO_X_DATA, (tx_buf && xfr->len >= 2) ? tx_buf[1] : 0);
>> +		if (ret) {
>> +			dev_err(&spi->dev, "SPI radio data write failed\n");
>> +			return ret;
>> +		}
> 
> This looks awfully like you're coming in at the wrong abstraction layer
> and the hardware actually implements a register abstraction rather than
> a SPI one so you should be using regmap as the abstraction.

I don't understand. Ben has suggested using regmap for the SPI _device_
that we're talking to, which may be a good idea. But this SX1301 device
in turn has two SPI _masters_ talking to an SX125x slave each. I don't
see how using regmap instead of my wrappers avoids this spi_controller?
The whole point of this spi_controller is to abstract and separate the
SX1255 vs. SX1257 vs. whatever-radio-attached into a separate driver,
instead of mixing it into the SX1301 driver - to me that looks cleaner
and more extensible. It also has the side-effect that we could configure
the two radios via DT (frequencies, clk output, etc.).

You will find a datasheet with some diagrams mentioning "SPI" at:
https://www.semtech.com/products/wireless-rf/lora-gateways/SX1301

>> +	if (rx_buf) {
>> +		ret = sx1301_read(ssx->parent, ssx->regs + REG_RADIO_X_DATA_READBACK, &rx_buf[xfr->len - 1]);
>> +		if (ret) {
>> +			dev_err(&spi->dev, "SPI radio data read failed\n");
>> +			return ret;
>> +		}
>> +	}
> 
> For a read we never set an address?

To read, you first write the address via tx_buf, then either in the same
transfer in the third byte or in a subsequent one-byte transfer as first
byte you get the data.

If you have better ideas how to structure this, do let me know.

>> +static void sx1301_radio_setup(struct spi_controller *ctrl)
>> +{
>> +	ctrl->mode_bits = SPI_CS_HIGH | SPI_NO_CS;
> 
> This controller has no chip select but we provided a set_cs operation?

Oops, I played around with those two options and was hoping SPI_NO_CS
would avoid the undesired set_cs invocations, but it didn't work as
expected and so I added the "if (enabled)" check above.

Thanks for your review,

Andreas

-- 
SUSE Linux GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)