18:15 <kettenis> with the power domans I need to add a "u-boot,dm-pre-reloc" property to the power-controller nodes that correspond to the power domains used by the serial console

18:49 <jannau> j_ey: the problem is probably applespi. your driver (modified) works if I remove the cs_change spi transfers which the corellium spi driver didn't handle correctly to my understanding

19:01 <j_ey> jannau: can you paste the diff please? good to know it's cs related

19:02 <jannau> https://gist.github.com/jannau/dab07c6c5bb25ad7f8d928f9825c50eb

19:03 <j_ey> ty

19:05 <jannau> still possible to break the ouchpad via capslock though

19:06 <jannau> and I have local commit to program the clock in (*prepare_message)

19:13 <jannau> touchpad is broken in general

19:17 <jannau> j_ey: touchpad works with https://gist.github.com/jannau/64fcad7e89a52b46cb312a7135b10b8d feel free to squash

19:17 <jannau> caps lock led toggling still commented out

19:20 <jannau> caps lock led toggling breaks touchpad

19:21 <j_ey> jannau: thanks, will squash that in

19:23 <j_ey> jannau: can you figure out if youre still getting packets (print something in applespi_got_data i guess)

19:26 <sven> Hmm… that prepare message thing smells like a race condition fwiw

19:28 <sven> hrm, or maybe something breaks if the clock is re-setup between each transfer

19:28 <jannau> it prints "applespi_got_data: 27 callbacks suppressed" and 10 x "Received corrupt packet"

19:28 <sven> would be good to take a look at what’s going on with the hw signals

19:29 <jannau> my suspicion was that (re-)programming the clock while cs is asserted breaks things

19:31 <jannau> still strange that it only breaks the touchpad and not the keyboard

19:31 <sven> So usually you’d first set CS and then enable the clock, but the clock should only be enabled anyway when there’s something in the FIFO

19:32 <sven> it might be that the (internal) clock takes a bit to be stable

19:33 <sven> can’t really tell without marcan hooking up a logic analyzer or oscilloscope to the spi lines

19:34 <sven> but this smells like a timing issue

19:42 <jannau> udelay(150) after apple_spimc_clock fixes the touchpad as well. 100 is not sufficient

19:47 <sven> interesting. I wonder if it’s the device that needs time to prepare or if it’s an issue with the controller

19:47 <sven> the clock should actually be stable almost immediately since it just needs to divide the input anyway

20:07 <jannau> the delay is needed between apple_spimc_clock and writing REG_TXCNT/REG_RXCNT

20:08 <jannau> breaks somewhere between 106 and 103

20:12 <jannau> or not, the delay has to be longer if it is after REG_TXCNT/REG_RXCNT

20:19 <jannau> looks like it could be the target. works without delay and apple_spimc_drain_rx in apple_spimc_transfer_one commented out

20:20 <j_ey> the target?

20:25 <jannau> the device (keyboard, ...)

20:29 <sven> or the controller does something strange that only a single device doesn't like

20:31 <sven> we'll probably need to know what's going on with the actual hardware lines to figure this out

20:35 <jannau> the caps lock led in applespi still break things the touchpad