On Mon, 2016-06-27 at 14:44 +0200, Alexandre Belloni wrote: > On 27/06/2016 at 14:50:49 +0300, Sergei Ianovich wrote : > > I don't think this is the right way. DS-1302 is an SPI device, not > > a > > GPIO one. It can be connected to a hardware SPI controller or a > > software one (on top of GPIO or memory). > > > > Your patch re-adds Microwire SPI control logic to RTC subsystem, which > > was cleared by my rewrite of drivers/rtc/rtc-ds1302.c. The logic is > > already present in bitbang_txrx_be_cpha0_lsb() in drivers/spi/spi- > > lp8841-rtc.c. > > > > I still think you need to implement spi-gpio-3wire with LSB-first > > support in SPI subsystem instead. It wasn't done when I was adding > > LP8841 support, because LP8841 was the only use case of Microwire SPI > > control logic. If you add it, drivers/spi/spi-lp8841-rtc.c can be > > removed and replaced by a GPIO driver to host a new spi-gpio-3wire > > device. > > Well, back in April, we concluded it was not easily doable after > discussing with Mark and there was still issues after implementing it in > spi-gpio. > My understanding is that while microwire seems compatible with SPI mode > 0, it actually isn't and this should be treated as a different mode. > If we want to do something generic, I think we should have a > microwire-gpio driver. Maybe in the SPI subsystem? I've seen Akinobu Mita report that he added support for 3wire to spi- gpio, and it didn't work. That's not a big difference from where we are now, when there is just no support for 3wire. Adding a working support for 3wire to spi-gpio needs 2^4 - 2^2 = 12 new bitbang functions to handle 3wire and lsb-first modes. It is a bit difficult to test all of them. I don't have enough hardware for example. In addition, it is unlikely that a 3wire GPIO SPI master would host more than a single device. That's why I think it is easier to add a new spi-gpio-3wire (or spi-gpio-microwire) driver. > How do yo currently select microwire mode for PX270? I didn't say I use PXA270 to drive this RTC. I just say it is possible. The driver needs to set bits 5:4 of SSCR0_1/2/3 register to 0b10. Microwire mode will be selected for a built-in SPI port. All bitbanging will be done by the chip, the driver will just need to set up DMA transfer. This is an example of a pretty sophisticated hardware controller. I actually use a simple software controller in LP8841. The driver is in drivers/spi/spi-lp8841-rtc.c. -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
