**** BEGIN LOGGING AT Wed Jul 10 03:01:22 2019 Jul 10 04:54:41 anyone of a soft McASP for the 'bone? Just a PRU based I2S xceiver Jul 10 08:15:21 zmatt: yes, precisely :) Jul 10 08:16:04 that was my confusion .. mainline seems to have USE figured, but not CONFIG of the controllers,etc Jul 10 08:16:48 qed, my comment about 'xyz is custom' [to the board with the gpio/controller] Jul 10 09:31:48 uhh, no? other way around? Jul 10 09:32:15 gpio controllers are handled just fine by mainline, gpios aren't Jul 10 10:11:22 oh lol OK Jul 10 10:11:35 I keep dibbin in and out, context switch kills :p Jul 10 10:22:24 and the new gpio chardev is a step in the wrong direction, since it is attached to the gpio controller, even though its existence is, or should be, irrelevant for most applications Jul 10 10:29:57 m Jul 10 10:32:55 oh? Jul 10 10:33:22 so gpios themselves are nothing to do withthe gpio controller? that seems conceptually weird? Jul 10 10:33:35 or is it just that kernel devs don't understand embedded computing at all ... Jul 10 10:33:55 Jul 10 11:14:29 veremitz: I mean applications don't need to care Jul 10 11:14:36 you don't use a gpio controller, you use a gpio Jul 10 11:16:04 the fact that gpios are bundled together into gpio controllers is mostly an implementation detail that the kernel needs to know about Jul 10 11:17:51 ok so its an abstraction failure Jul 10 11:17:54 (surprise) Jul 10 13:50:37 I say "mostly" because there is a performance difference: a group of gpios can be read or changed more quickly if they're all on the same controller (although the difference is probably negligible compared to the time it takes to perform a system call in the first place) Jul 10 14:02:54 do you get more than one controller in an average SoC? Jul 10 14:09:19 can't have too many jkridner's ;) Jul 10 14:09:32 one for each day of the week LOL Jul 10 15:10:49 TI's gpio controller supports up to 32 gpios, the AM335x therefore has 4 instances, the AM572x has 8 instances Jul 10 15:13:12 and there's a 9th gpio controller on the bbx15: the pmic has a few gpios Jul 10 16:27:20 rcn-ee[m]: I don't know when they added it, but I just discovered that you can add the x-systemd.growfs option to a filesystem in fstab to have systemd automatically grow the filesystem to span the space available on the block device Jul 10 16:36:27 rcn-ee[m]: also useful, you can add ConditionFirstBoot=yes to the [Unit] section of services to make them run only on first boot (which is determined by the absence of /etc/machine-id at boot) Jul 10 19:58:03 ah, interesting Jul 10 20:11:30 veremitz: that's why gpio numbers in my pins spreadsheet (and in output of my show-pins utility) are written in the format N.NN Jul 10 20:14:15 cool Jul 10 20:17:29 you'll often also see people use the kernel's global gpio numbers, e.g. gpio97 for what I'd call gpio 3.01. the thing that makes that slightly awkward is that the kernel just assigns a sequential block of gpio numbers to each gpio controller (based on how many gpios it declares to have) in whatever order those gpio controllers happen to get initialized Jul 10 20:17:53 /18/18 Jul 10 20:18:10 in practice, for gpio controllers that are platform devices the order does seem to be deterministic, but I don't think that's actually guaranteed by the kernel Jul 10 20:21:46 sounds abit like udev netdevs .. *chuckle* which probably -are- deterministic Somehow .. Jul 10 20:21:57 I should create aliases tbf .. Jul 10 20:22:08 [generally, and/or more often]* Jul 10 20:36:31 veremitz: well for gpios that's why I use gpio-of-helper. for example the DT fragments and udev rule in the gpio-demo.dtsi I showed earlier results in: https://pastebin.com/raw/YKW7Wcqu Jul 10 20:41:34 zmatt: oh I don't doubt your methodology or tooling to be anything less than Excellent .. true understanding and competency are rarely found, more so in combination .. Jul 10 20:41:53 I try Jul 10 20:42:05 ^ that goes farther than most :) Jul 10 20:42:12 now if only I scored equally well on productivity -.- Jul 10 20:42:23 eh, preaching to the choir there .. Jul 10 20:47:14 I'm not sure why setting up gpios like this is exception rather than rule though, especially considering how easy it is to do. am I the only one who cares about things like avoiding the need for root privileges and making applications not hardcode physical gpio numbers? Jul 10 20:51:25 ah, but you can do *Anything* as root ... Jul 10 20:51:28 ... Ooops ... Jul 10 20:51:59 too few people know what (a) different ways are (b) good ways from bad .. Jul 10 20:55:47 -.- Jul 10 21:06:07 yup Jul 10 21:28:04 heh, forensic archeology in an ancient piece of hardware... https://www.youtube.com/watch?v=I7VLIjsKxkU Jul 11 02:50:36 zmatt you just want to rip that apart don't you? Still rope memory that has to be one of the weirder things I have had to think about. It makes sense considering the technology of the time but still. **** ENDING LOGGING AT Thu Jul 11 03:01:20 2019