**** BEGIN LOGGING AT Fri Nov 02 02:59:58 2018 **** BEGIN LOGGING AT Fri Nov 02 04:09:02 2018 Nov 02 08:46:04 hi ant_work Nov 02 08:47:41 hello Nov 02 08:47:59 Hi, I did some work on these two issues in mainline Nov 02 08:48:17 first, the keyboard issue (gpio-pxa.c) Nov 02 08:49:17 it is related to PINCTRL. Even if PINCTRL is disabled in .config, we need to declare PXA27X_CPIO to have no PINCTRL support in gpio-pxa.c Nov 02 08:49:34 ok, here it is easy then, we can blame LinusW Nov 02 08:49:50 he is pinctrl man Nov 02 08:50:11 well, I am not sure if PINCTRL for PXA27x is complete yet. Nov 02 08:50:20 we can try with a private msg to him and Robert Nov 02 08:50:23 https://github.com/greguu/linux-4.19-c3x00/commit/603d43c2fbaaaf97639d2458a72b285cbc539ecc Nov 02 08:50:29 simple fix, no revert needed Nov 02 08:50:31 last option is to write on kernel ML Nov 02 08:50:51 second, is the PCMCIA issue Nov 02 08:51:06 hm, I see, I think they hae 'forgotten' the non-dt boards Nov 02 08:51:26 yes, we need to adopt dt too Nov 02 08:52:02 there were some problematic drivers iirc for Z Nov 02 08:52:26 ok, the PCMCIA issue actually is related to CPUFREQ and not GPIO Nov 02 08:52:45 it took me dozens of builds to figure this one out Nov 02 08:52:47 ok, did you enable it in .config? Nov 02 08:52:54 yes, I have it enabled Nov 02 08:53:22 Russel introduce a patch in 2016 Nov 02 08:53:49 pcmcia: soc_common: switch to a per-socket cpufreq notifier Nov 02 08:54:06 this one turns the first socket:0 off at boot if no card is found Nov 02 08:54:29 socket:0 is the spare one on Borzoi/Spitz Nov 02 08:54:48 argh Nov 02 08:54:53 interesting is also that if you insert a card a boot it works, even if you swap the card later Nov 02 08:55:17 current fix is to revert that patch from Russel https://github.com/greguu/linux-4.19-c3x00/commit/4a8712b1c087c45ac9af5cb95c778a201280adb6 Nov 02 08:56:08 I will dig deeper to find the flaw in russels patch, but I think is related to the removed mutex_unlock() Nov 02 08:56:40 we should approach RMK carefully :) Nov 02 08:57:37 hehe, yes, I will try first to identify the issue in more detail. It may be we need to adopt spitz.c for this to work Nov 02 08:57:58 I think both patches could go through Robert's pxa-tree Nov 02 08:58:25 well, not quite ready to submit, these are just workarounds Nov 02 08:58:56 not sure what the implications are on the first patch for other pxa27x platforms Nov 02 08:59:08 the second one is just a revert.. so more work to be done Nov 02 08:59:22 good is that 4.19 works quite well so far Nov 02 08:59:37 yes, a fix will pass, a revert is very hard Nov 02 09:00:35 there is also a new FBDEV kernel option in 4.19 Nov 02 09:00:57 btw I am pretty sure I did not configure cpufreq...must check Nov 02 09:02:47 maybe even if CPUFREQ is disabled it will impact PCMCIA due to the error handling removed in Russel's patch Nov 02 09:03:05 so far I can confirm reverting his patch solved the issue Nov 02 09:03:31 the error handling is outside of #ifdef CONFIG_CPU_FREQ Nov 02 09:03:40 maybe thats where the flaw is Nov 02 09:08:33 well, there is surely one Nov 02 09:09:35 it was a pain to figure out as I was focused on GPIO and the regulator patches, not CPUFREQ Nov 02 09:14:59 the new FBDEV option is Nov 02 09:15:02 CONFIG_FRAMEBUFFER_CONSOLE_DEFERRED_TAKEOVER Nov 02 09:15:34 unfortunately this option will need to be enable in the kexecboot kernel too to make any sense Nov 02 09:16:09 if not, the main kernel will show briefly a distorted screen with random bits displayed on the framebuffer Nov 02 09:16:16 so I left it disabled Nov 02 09:17:20 well, we did mostly disable console output Nov 02 09:18:03 just a blinking char Nov 02 09:18:07 yes, this is just a new tweak to make it more quiet if you have some logo displayed and are not passing "quiet" Nov 02 09:18:10 yes Nov 02 09:18:10 and btw I have to try vt.global_cursor_default=0 in the kernel cmdline got rid of the cursor. Nov 02 09:18:25 well, this is what the new patch may do Nov 02 09:18:33 the new option I mean Nov 02 09:18:44 get rid of that cursor Nov 02 09:18:47 if we can simplify kexecboot spash, why not Nov 02 09:19:06 well, 4.19 too big to fit in NAND Nov 02 09:19:21 we may be stuck with 4.14, or 4.4 if you need more Nov 02 09:19:22 even disabling net? Nov 02 09:19:43 4.14 just fits, see my voidz kexecboot, _just_ fits Nov 02 09:19:46 iirc I compiled a 4.16, didn't go further Nov 02 09:20:11 well, the new kexec binary is also bigger Nov 02 09:20:22 I did not try 4.16, I think with 4.14 I had to remove some stuff like JIFFS2 Nov 02 09:21:55 I must rewrite the detection code in devicescan so it will add some kb Nov 02 09:22:20 we discussed long ago with Jay7 about optimizing the headers Nov 02 09:22:41 i.e. sahring klibc headers btw kexec and kexecboot Nov 02 09:23:09 to save a few kb... Nov 02 09:26:47 https://pastebin.com/hY3q51aV Nov 02 09:27:07 this is my 4.14 kexecboot config. Maybe something to improve ? Nov 02 09:29:26 initramfs.cpio.xz is 69168 B or just 68K Nov 02 09:31:47 the zImage is 1201624 B Nov 02 09:31:52 so just fits Nov 02 09:33:46 I think I had more headroom Nov 02 09:34:12 well, try and share your config. Its getting pretty close for sure Nov 02 09:35:00 I should have kept the 4.16 at home Nov 02 09:35:24 little changes btw 4.14 and 4.16 Nov 02 09:35:26 4.14.18 still has JIFFS2 enabled, I see Nov 02 09:35:44 yes, 4.16 may just fit, 4.19 was a bit bigger already Nov 02 09:35:52 I did not change these yet Nov 02 09:35:53 http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux Nov 02 09:36:28 last pushed is the 4.4 Nov 02 09:36:54 yeah, but that one can not mount newer ext4, thats why I went 4.14 Nov 02 09:37:18 I think I used your defconfig for my kernel as a base Nov 02 09:37:52 maybe we must just let the 3.x with fat/jffs2 and the 4.x with f2fs and new ext4 Nov 02 09:38:26 no ubi ? Nov 02 09:38:35 ubi for both ofc Nov 02 09:39:34 anyway, great findings Nov 02 09:40:40 now, pls put it all together and send to Robert or Lionus Walleji Nov 02 09:41:50 ok, will do. I do know Linus Walleji as I had contact with him in the past about some router stuff. RTL8366SR Nov 02 09:42:29 but with Russel I will make sure first to pin point the issue Nov 02 09:42:30 haha Nov 02 09:43:35 what Robert ? Robert Jarzmik ? Nov 02 09:45:12 yes Nov 02 09:45:39 ok, I will get it together this weekend an maybe can make it for the next merge window Nov 02 09:45:42 the pinctr patch could go through LinusW or Robert's tree Nov 02 09:46:12 LinusW is normally quick, Robert slower but he is the maintainer... Nov 02 09:46:24 RMK above both Nov 02 09:47:05 I'll test your patches on c7x0 Nov 02 09:52:49 ok great Nov 02 09:53:28 suspend via pwr button works fine under void linux Nov 02 09:53:34 with 4.19 Nov 02 10:03:21 and resume? :D Nov 02 10:03:37 works Nov 02 10:03:53 what demon are you using? zaurusd? Nov 02 10:04:27 no, I have not ported zaurusd yet. Nov 02 10:04:48 so just lacks rotation Nov 02 10:05:01 does alsa work? Nov 02 10:05:04 yes, there is no kdrive yet too Nov 02 10:05:09 yes alsa works Nov 02 10:05:18 wifi works Nov 02 10:05:25 ok, I tested irda and bt with 4.14 months ago.. Nov 02 10:05:38 wifi was tested by Varti iirc Nov 02 10:05:42 irda is nice, I used it with my old nokia to get gprs data Nov 02 10:05:52 I may try with void Nov 02 10:05:55 is now deprecated :/ Nov 02 10:06:04 ah, but should still work ? Nov 02 10:06:11 driver are now in staging Nov 02 10:06:18 hm.. Nov 02 10:06:36 at one point they disappear from my config so I lloked at that Nov 02 10:06:51 iirc after 4.14 Nov 02 10:06:52 it was slow anyway, but a nice feature.. Nov 02 10:06:56 back then.. Nov 02 10:07:12 heh.. nokia gprs and zaurus...2004 iirc Nov 02 10:07:50 I still used it in 2010/2011 Nov 02 10:08:28 was ok to get email and do IRC Nov 02 10:08:41 well, those socket bt cf card are nice and fit well but they drain the battery. wifi-cf is even worse Nov 02 10:08:54 I still got the phone, so I may try again Nov 02 10:08:57 power savvy was only the 10/100 cf Nov 02 10:09:09 hmm no 100, just 10mb/s Nov 02 10:09:28 yes, my CF card is 10 Mbit only Nov 02 10:09:40 th eethernet I mean Nov 02 10:09:46 yes, ethernet Nov 02 10:10:00 the WiFi card is 11g I think Nov 02 10:10:09 yes, b/g Nov 02 10:10:53 say, there isn't any official 4.20-rcx tree? Is it already renamed to 5 ? Nov 02 10:11:06 I use a USB to ethernet adapter most of the times, it does 100 Mbit, but I get usually only 10-12mbit (due to USB1.1 limit) Nov 02 10:11:29 not sure if it will be 4.20 or 5 Nov 02 10:11:36 surprise Nov 02 10:11:56 it depends on the code line count afaik Nov 02 10:12:36 https://wiki.linuxfoundation.org/realtime/rtl/blog Nov 02 10:13:03 lumag once tested the rt tree Nov 02 10:14:39 hm,I guess linus t will push 5 once the code line threshold is reached Nov 02 10:14:46 https://www.kernel.org/category/releases.html Nov 02 10:14:53 apparently 4.19 is LTS Nov 02 10:15:15 then wer stay on it for a while Nov 02 10:15:59 yes, 4.19 is LTS, thats why I thought its worth messing with it again Nov 02 10:17:13 I think there will be a 4.20... Nov 02 10:18:11 ok, and last 3 is 3.18.124 [EOL] Nov 02 10:22:09 yeah, that is sad, as its the current kernel 3.18.41 is used in the Gemini PDA Nov 02 10:22:38 I think I bought a paperweight knowing the half life of 2 years ahead Nov 02 10:22:47 lol. Nov 02 10:24:01 despite being EOL, 3.18 was is getting still some updates Nov 02 12:39:47 > I think I bought a paperweight knowing the half life of 2 years ahead Nov 02 12:39:48 I think the Gemini PDA will be upgraded from 3.18 to 4.x Nov 02 15:30:17 greguu, about size :/ Nov 02 15:30:18 [PATCH] ubifs: auth: add CONFIG_KEYS dependency Nov 02 15:30:41 config UBIFS_FS_AUTHENTICATION Nov 02 15:30:41 bool "UBIFS authentication support" Nov 02 15:30:41 + depends on KEYS Nov 02 23:18:28 omegamoon[m]: that would be great. Is there any indicator this will happen ? Oreo works also with 3.18 **** ENDING LOGGING AT Sat Nov 03 02:59:58 2018