**** BEGIN LOGGING AT Wed Nov 07 02:59:59 2018 Nov 07 09:54:50 greguu, so pxa was again broken by dt advancements Nov 07 09:56:56 yeah, looks like it. I am waiting for roberts feedback Nov 07 09:59:39 ah, found Nov 07 10:00:05 check this old link, anarsoul /Vasily) did port a (pxa?) device to dt Nov 07 10:00:08 https://lore.kernel.org/patchwork/patch/542530/ Nov 07 10:00:19 this tree https://github.com/anarsoul/linux-2.6/commits/v3.17-pxa-dmaengine-wip Nov 07 10:00:42 we can ask him, maybe anarsoul has no new toys atm.. Nov 07 10:01:55 i got a zipit now Nov 07 10:02:02 ut got dt too Nov 07 10:02:43 anarsoul is busy with pine64 stuff Nov 07 10:02:54 anarsoul told me to 'just' repeat his steps Nov 07 10:03:17 we can try Nov 07 10:03:28 pine64, I see Nov 07 10:03:37 not the scoop stuff the zaurus has is specific Nov 07 10:04:17 I remember there is something not obvious to do Nov 07 10:04:48 really needing a kernel hacker Nov 07 10:06:43 or we patch ur way through until they drop armv5 Nov 07 10:06:50 haha Nov 07 10:14:22 https://patchwork.kernel.org/patch/9153981/ Nov 07 10:14:48 anaroul is logged in but I can't find him in any channel, maybe he's invisible Nov 07 10:15:01 so its not straighforward for zaurus Nov 07 10:15:09 z Nov 07 10:15:16 #zipit Nov 07 10:15:53 he is there Nov 07 10:16:53 or nick anarsoul|2 Nov 07 10:17:07 I just did a whois Nov 07 10:17:32 he mus have the -i flag Nov 07 10:17:43 ok Nov 07 10:17:59 he was in Vancopuver las time Nov 07 10:18:04 night there Nov 07 10:18:23 (but ininfluent for an hacker) Nov 07 10:18:31 pretty dark here too Nov 07 10:20:11 I've pm him and asked to log here Nov 07 10:20:57 I remember the above 2016 patch, in my mind things were fixed... Nov 07 10:23:41 greguu, and about pcmcia issue? Did you dare to write to Russell? Nov 07 10:23:46 all that spitz scoop gpio stuff may be a headache Nov 07 10:24:01 heh.. chicken-egg :) Nov 07 10:24:09 no this went to robert too, I CC you Nov 07 12:00:35 greguu, I see the cpufreq email now, gmail tagged it linux-arm-kernel so it wasn't in inbox.. Nov 07 22:17:13 anarsoul, hi there Nov 07 22:17:18 thx for coming Nov 07 22:17:25 hi ant_home Nov 07 22:17:38 so what's up? Nov 07 22:17:57 we dostill play with zaurus kernels Nov 07 22:18:21 greguu is fixing some bugs introduced by pxa dt transition Nov 07 22:18:33 I remembered you did that for zipit Nov 07 22:18:55 but I don't remember how hard this could be... Nov 07 22:19:21 should be relatively easy Nov 07 22:19:43 you'll have hybrid type, since I doubt it's possible to do full dt conversion Nov 07 22:20:19 i.e. some devices will be created in board file, rest described in dts Nov 07 22:21:03 now we have the issue about gpio and pinctrl Nov 07 22:21:15 https://github.com/anarsoul/linux-2.6/commits/v4.4-z2 Nov 07 22:21:32 27 patches, I guess 3rd of them aren't related to dt conversion Nov 07 22:22:41 there should be less patches for 4.20-rc1 Nov 07 22:22:59 great, I only saw 3.18 :) Nov 07 22:23:06 this is one prob: Nov 07 22:23:06 https://www.spinics.net/lists/linux-gpio/msg33963.html Nov 07 22:23:10 I don't really follow pxa progress nowadays Nov 07 22:23:24 hh, pine64 Nov 07 22:23:46 and rock64 :) so allwinner a64, rk3328 and rk3399 Nov 07 22:24:14 (rk is rockchip) Nov 07 22:24:16 I have fixed the last issues and klibc is ok for arm64 Nov 07 22:24:32 kexec and friends Nov 07 22:25:44 ant_home: anyway, I'd suggest trying to boot minimal dt-enabled kernel with dts for zaurus Nov 07 22:25:56 it should be easy if you have serial console Nov 07 22:26:25 we added dt support to kexecboot, I can test very easily Nov 07 22:27:08 all you need is to create dts for zaurus :) Nov 07 22:27:19 ah, by the way, do you even happen to have written a loop checking for the insertion of SD/CF/.. block devices? Nov 07 22:27:48 we do have a fix delay but with deferred probing is sometimes insufficient, mostly on spitz Nov 07 22:27:56 nope Nov 07 22:28:13 (there is a second, separate bug about pcmcia on spitz btw) Nov 07 22:28:33 ok, we have to start with a simple dts Nov 07 22:28:48 the issue are the gpio expanders and scoop I fear Nov 07 22:29:32 ant_home: what's wrong with gpio expanders? Nov 07 22:29:58 it's pretty easy to adapt a driver for device tree Nov 07 22:30:12 just add array of compatibles and that's it Nov 07 22:30:27 the old issue: https://patchwork.kernel.org/patch/9153981/ Nov 07 22:30:34 if it requests resources using standard functions it will be able to pick them up from dts Nov 07 22:31:12 Now the true chicken and egg problem is than machine files, Nov 07 22:31:12 ie. arch/arm/mach-pxa/xxx.c are using gpio before the drivers are Nov 07 22:31:12 probed, in the init_machine() function, and that's why pinctrl/gpio for Nov 07 22:31:12 legacy machine files is a bit difficult. Nov 07 22:32:25 which machine is it? Nov 07 22:32:32 zaurus Nov 07 22:32:42 what board file? Nov 07 22:32:50 there's no zaurus.c in arch/arm/mach-pxa Nov 07 22:33:05 ah, spitz.c and corgi.c i.e. Nov 07 22:33:26 corgi akita spitz tosa poodle are the pxa Nov 07 22:34:36 poodle has a locomo chip as well, as on the older collie but with different purposes Nov 07 22:35:09 for this locomo, lumag wrote a driver for 4.4 that is now bitrotting... Nov 07 22:35:24 sigh, he had no time to upstream all that Nov 07 22:36:02 I keep all in meta-handheld layer Nov 07 22:37:18 well Nov 07 22:37:22 4.4 isn't that old Nov 07 22:37:30 port it to 4.20 Nov 07 22:37:42 right but gpio and spi changes made the patch obsolete Nov 07 22:37:58 well, anyway, start with small steps Nov 07 22:38:10 enable CONFIG_MACH_PXA27X_DT Nov 07 22:38:29 wait, we are on 4.19 with these boards Nov 07 22:38:44 even -next if you say Nov 07 22:39:19 where to start? Nov 07 22:39:45 ant_home: I'd suggest to use Linus' git master, which is at 4.20-rc1 currently Nov 07 22:39:55 see https://github.com/anarsoul/linux-2.6/commit/f9d2b51dfed51e502a5081131cfe96d6c3751df3 Nov 07 22:40:01 ah, ok, thanks Nov 07 22:40:11 greguu, ^^ Nov 07 22:40:25 that's the patch that converts z2 to dt init Nov 07 22:40:40 you have to make something similar for your boards Nov 07 22:41:58 I'm not sure about pxa2xx_mfp_config() though Nov 07 22:42:08 4.4 didn't have pinctrl driver for pxa2xx Nov 07 22:42:19 see, can you please shed light on regulators on these boards? Nov 07 22:43:05 declare them in dts? I believe they're on i2c? Nov 07 22:43:26 once lumag said there are no regulators Nov 07 22:43:34 meaning these are fixed Nov 07 22:43:45 see https://github.com/anarsoul/linux-2.6/commit/f9d2b51dfed51e502a5081131cfe96d6c3751df3#diff-1b39f5e1954bc6b9d755c5a83c790130R56 Nov 07 22:44:32 well Nov 07 22:44:35 spitz declares isl6271a Nov 07 22:44:43 which controls cpu voltage Nov 07 22:45:37 ah, full contraints was aded Nov 07 22:46:02 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/arm/mach-pxa/corgi.c?h=v4.20-rc1&id=271e80176aae4e5b481f4bb92df9768c6075bbca Nov 07 22:47:15 ant_home: anyway, try to booting your device with minimal dts first Nov 07 22:47:31 and then add rest of devices, one by one Nov 07 22:47:37 well Nov 07 22:47:44 so boot from nand is the easiest Nov 07 22:47:56 wording is bad Nov 07 22:48:12 try to booting your board with minimal dts and then add rest of devices (e.g. lcd, usb, etc) Nov 07 22:48:26 ok, just serial Nov 07 22:48:33 yeah Nov 07 22:48:47 you don't even need support for loading dtb from your bootloader Nov 07 22:49:31 use CONFIG_ARM_APPENDED_DTB Nov 07 22:50:48 I expect it to boot like any other dt-converted pxa board :) Nov 07 22:51:04 the problem will be adding all the hardware around the cpu ;) Nov 07 22:51:14 ant_home: not really Nov 07 22:51:31 you may need to convert zaurus-specific drivers to use dt Nov 07 22:51:47 but most of hardware in soc is already dt-enabled Nov 07 22:51:55 well, it was in 4.4, it's even better in 4.20 Nov 07 22:51:56 ok, I meant the drivers for i2c, spi, ssp, ... Nov 07 22:52:15 ant_home: spi is ssp, isn't it? :) Nov 07 22:52:24 i2c is already dt-enabled Nov 07 22:52:54 so is spi aka ssp Nov 07 22:53:10 trust me, if nobody did it it is NOT a low hanging fruit... Nov 07 22:53:12 see Documentation/devicetree/bindings for information on bindings Nov 07 22:53:42 ant_home: you have qemu support for your hardware Nov 07 22:53:44 iirc pcmcia is specially hostile Nov 07 22:53:45 you can debug anything. Nov 07 22:53:54 well, no :) Nov 07 22:54:12 I had two cases of gcc breaking boot on device but not on qemuarm Nov 07 22:54:18 qemo know better ;) Nov 07 22:54:41 OK, you need sane compiler for sure Nov 07 22:55:06 qemu isn't as strict as real hardware Nov 07 22:55:07 heh, it happened with gcc7 last year and gcc8 this summer Nov 07 22:55:23 but it allows you to debug crashes in early init Nov 07 22:55:38 ant_home: no one uses armv5 nowadays :( Nov 07 22:55:44 normally yes Nov 07 22:57:08 > > PXA is one of those platforms that have so very many users so Nov 07 22:57:08 > > I do think we should put in effort to modernize it if we can. Nov 07 22:57:14 LinusW ? Nov 07 22:57:42 ant_home: anyway, just start working on it Nov 07 22:57:49 if you hit any road blocks - ping me Nov 07 22:57:53 we just need a good supervisor :D Nov 07 22:58:00 many thanks Nov 07 22:58:10 we have home-works to do Nov 07 22:58:34 we'll summon you again... Nov 07 22:58:42 :) Nov 07 22:58:58 I added this channel to autojoin list Nov 07 23:00:08 ant_home: do you have serial port on zaurus? Nov 07 23:02:24 yes Nov 07 23:02:31 cool Nov 07 23:02:38 it should make development easier Nov 07 23:02:51 friends here have as well, al models covered Nov 07 23:04:24 the issue is when the kernel decompressor is miscompiled...no way to get anything heh Nov 07 23:04:32 2 times..sigh Nov 07 23:06:13 qemu does fix the alignment so you never know Nov 07 23:07:19 even this summer when musl libc was miscompiled and kernel was full of align faults qemu did swallow all Nov 07 23:07:41 bug or feature? Nov 07 23:08:18 ant_home: probably a bug Nov 07 23:08:24 but I doubt they'll fix it Nov 07 23:08:35 wellm for revenge I broke it ;) Nov 07 23:30:39 and I'm not digging into qemu tcg again :) Nov 07 23:30:39 to boot zaurus on qemu you must add mtdparts= Nov 07 23:30:39 because kernel now reads from nand address Nov 07 23:30:39 the partition table Nov 07 23:30:39 ant_home: well, that's actually easy to implement in qemu Nov 07 23:30:39 good, if anyone cares.... Nov 07 23:30:39 see hw/arm/spitz.c **** BEGIN LOGGING AT Wed Nov 07 23:30:56 2018 **** ENDING LOGGING AT Thu Nov 08 02:59:59 2018