2020-10-31

<nekomancer[m]> chewitt: returning to N2 watchdog patch. Can your patch be added to armbian mainline?
<lanefu> yeah chewitt thanks for the background
<chewitt> once in a while there's a 'special' person, but those are normally a lost cause in many ways .. missing attribution is usually the least concern :)
<chewitt> I find most peopled do things out of inexperience and ignorance, and with polite education are happy to fix things
<chewitt> Oleg's expectations appear to be different
<chewitt> speaking personally.. I expect a percentage of people to copy stuff without attribution .. it's inevitable
<chewitt> he also persists in talking about "his code" in a very possessive way (even in Russian) which isn't realistic when it carries at GPLv2 license which encourages people to take the stuff you publicly post to GitHub
<chewitt> but he's been wound up by idiot users and comes over a bit paranoid .. which adds confusion and emotion to any discussion
<chewitt> (when you do stuff that vendor u-boot doesn't handle)
<chewitt> I think the "damage boxes" is a misinterpretation/mistranslation of his desire to not-support the tweaked bootloader that CE uses, which can lead to boxes that don't boot
<chewitt> 'tis a cursed platform :)
<chewitt> and for some reason Amlogic users often turn out to be particularly whiny shits to deal with
<chewitt> so he's been hounded a bit .. and over time the original issues have been obscured by new ones
<chewitt> the same forum people followed him to LE forums, and Khadas forums
<chewitt> yeah, the translations don't help
<chewitt> I have tried (and apparently failed) at teaching him 'git rebase'
<chewitt> one weekly commit with 200,000+ files touched .. no fcuking way I am wasting time trying to spot what he changed
<chewitt> also because his commit style makes it impossible to collaborate with him
<chewitt> he hasn't been so active around LE recently, partly because of that
<chewitt> but some of their forum users bait Oleg .. and he does dig holes and jump-in sometimes with odd statements and accusations which fuel fires
<chewitt> anyway.. old news now and we normalised senior dev relations between teams a while ago
<chewitt> after that episode I think they realised we/me might have been telling the truth about him :)
<chewitt> for a long time I'd been portrayed as 'the great satan' for kicking him out of LE (although he resigned rather than we kicked, tho wish we had)
<chewitt> the senior folks in their team kicked him out, then he absconded with the paypal donations and trashed all their git repos (deleted etc.)
<chewitt> he's no longer around, which is good
<chewitt> at that point in time, the project lead being a narcissistic arsehole of epic stature, who has persistently ignored copyrights/attribution
<chewitt> but the underlying issue is related to the folks at CoreELEC who took some of his scripts, converted them to binary code, and then included that in their images
<chewitt> I do make him post in Russian sometimes so that I can read it, or get Mrs Hewitt to read it back to me (native speaker)
<chewitt> so something where he's mildly annoyed gets translated to being thoroughly pissed off
<chewitt> that's also my specific experience with Oleg (who also contributes to LE)
<chewitt> I do speak/read/write a little Russian and Google Translate is reasonable but always seems to choose strong/firm words when going RU > EN
<nekomancer[m]> chewitt, I build N2 with your patch and now I have watchdog device! Thank you!

2020-10-30

<nekomancer[m]> chewitt: I will test this https://patchwork.kernel.org/project/linux-amlogic/patch/20201030180057.23886-1-christianshewitt@gmail.com/ with N2 -current and -dev nearest night
<chewitt> send emOne my way if you see him .. ISTR he has an 'air' box with the card
<chewitt> haven't tested it yet as I don't have hardware, but i'm sure someone will pop up
<chewitt> I got an RTL8822CS driver to build earlier too .. which is appearing in S905X3 boxes
<chewitt> typo :)
<Tonymac32> awesome chewitt
<chewitt> node was moved from wrong bus to right bus
<chewitt> nekomancer[m] Tonymac32 revised patch, confirmed working by narmstrong https://github.com/chewitt/linux/commit/1c7ff5285878ee8a2d46a5f85c8fa3b9e7dafd05
<chewitt> test it, then pm me a real name and email for 'Tested-by: blah <blah@blah.com>' and I'll send it upstream
<chewitt> ^ looks present
<chewitt> /sys/module/watchdog
<chewitt> /sys/module/stmmac/parameters/watchdog
<chewitt> /sys/devices/platform/soc/ff800000.bus/ff80f0d0.wtd/watchdog
<chewitt> /sys/devices/platform/soc/ff800000.bus/ff80f0d0.wtd/misc/watchdog
<chewitt> /sys/class/watchdog
<chewitt> /sys/class/misc/watchdog
<chewitt> /proc/sys/kernel/watchdog
<chewitt> /dev/watchdog
<chewitt> C4:/ # find / -name watchdog
<chewitt> did you enable the module in defconfig?
<chewitt> 5.9.1 but driver is upstream for a long time so should be fine
<chewitt> C4 here shows /dev/watchdog ..
<chewitt> I have no idea what a watchdog is, let alone how to test it
<chewitt> it compiles .. no idea whether it works .. let me know :)
<chewitt> the forum post with snippet from vendor kernel shows "watchdog@0xffd0f0d0" so looks like the same address
<chewitt> as axg and g12 have a lot in common
<chewitt> so you can probably copy it to meson-g12-common.dtsi
<chewitt> but meson-axg.dtsi references ^ that code
<chewitt> looks like gx devices inherit from meson-gx.dtsi, but g12 doesn't
<chewitt> ^ prob needs something like that
<chewitt> };
<chewitt> clocks = <&xtal>;
<chewitt> compatible = "amlogic,meson-gxbb-wdt";
<chewitt> reg = <0x0 0xf0d0 0x0 0x10>;
<chewitt> watchdog@f0d0 {
<Tonymac32> chewitt G12b is somehow inheriting GX stuff? I didn't see that in the device tree, but I haven't exhaustively searched either...
<chewitt> CONFIG_MESON_GXBB_WATCHDOG=y or CONFIG_MESON_WATCHDOG=y in defconfig
<c0rnelius> chewitt: hey. you got a n2+ uboot bin laying around?
<chewitt> it should be inherited from here
<nekomancer[m]> chewitt, possible you have an idea, how to solve? https://forum.armbian.com/topic/15701-no-watchdog-on-n2-with-current-58-and-dev-59-kernels/
<chewitt> and comrade oleg is also a partial contributor to my mess too
<chewitt> genuine ask - as an irregular visitor to the forums I've not followed the drama in detail
<chewitt> btw, how did the tvbox thing end up?
<Tonymac32> chewitt that would be good
<chewitt> there's some discussion (collusion) to perhaps smuggle them as 'fixes' to get them in, since LTS
<chewitt> 5.10 is missing some panfrost bits someone forgot/overlooked to merge
<Tonymac32> chewitt XD easy now
<[TheBug]> chewitt: yeah but what does H616 have that actually makes it appealing?
<chewitt> H616 exists .. in TV boxes (which are popular around here I gather)
<chewitt> sadly they are 'Allwinner' people
<chewitt> there are folks in LE with the skills/aptitude to do that
<chewitt> or asking Amlogic for the sources; which I will do (and they will say no to) .. but I'll ask anyway
<chewitt> this requires reverse engineering the firmware blobs
<chewitt> the best but slowest approach would be to de-blob the decoder, and move to a stateless approach where all the decoding cleverness is done in visible software
<chewitt> and 4k/10-bit requires lots
<chewitt> so everything needs to be done 'with cma'
<chewitt> and apparently it's not possible to have cma and non-cma approaches in the same driver
<chewitt> some support compression (afbc) and others don't
<chewitt> the CMA silliness is due to inconsistencies in decoding approach between Amlogic codecss
<chewitt> the default should probably be lower at 128MB to handle the 512MB version of LaFrite
<chewitt> I'm also trying to get my head around why such insane CMA reservations are needed
<chewitt> 10-bit deadlocks the board .. more understanding to be done
<chewitt> HEVC 8-bit seems to work well
<chewitt> Tonymac32 .. waprme (minimyth2) and myself have been poking HEVC support a little, trying to revive the work Maxime was doing
<chewitt> non, je suis Anglais
<chewitt> english turn-of-phrase .. s/merchant/specialist
<chewitt> I'm a highly-evolved copy/paste merchant
<chewitt> plus, nobody should mistake me for a developer .. I don't do code beyond bash script hacking
<chewitt> but I do loiter here and answer Q's on the huge patchset that I accumulate and help spread some FOSS love :)
<chewitt> Hi Tonymac32
<chewitt> and I don't 'contribute' to Armbian because running LibreELEC and doing behind-the-scenes stuff for Kodi already gets me in trouble with time/family/work
<Tonymac32> hi chewitt
<chewitt> and the media stuff for Amlogic is very much 'WIP'
<chewitt> nekomancer[m] stuff that is well tested and has a decent chance of being accepted goes upstream, but otherwise I have to hoard it locally in perpetuity

2020-10-29

<Tonymac32> nekomancer[m] chewitt does upstream some stuff, but remember others have to accept it, which doesn't always happen
<nekomancer[m]> chewitt holds a lot of patches to linux on github. did somebody know why he not upstream it into linux nor armbian?

2020-10-26

<IgorPec> chewitt: tbh i have no idea atm. we only have estimation for x86, arm should be 20-30% less
<chewitt> IgorPec: being curious, what's the rough cost of that ARM server?

2020-10-24

<[TheBug]> Chewitt hits it right on the nose there pretty much
<IgorPec> is a good source of information for that pamflet, starting with chewitt

2020-10-19

<nekomancer[m]> chewitt have that. but they in her repo.

2020-10-11

<nekomancer[m]> early I use dts patch from chewitt's repo to work RTC o n N2. but now there all changed and my trials to port it was fail/
<c0rnelius> I also made one for the N2 against chewitt's repo, but I don't own the board so I can't test it.
<c0rnelius> I have the same prob on the C4 and Le Potato. I'm assuming there are patches out there. chewitt would know I'm sure.

2020-09-24

<nekomancer[m]> .seen chewitt

2020-09-20

<nekomancer[m]> .seen chewitt

2020-09-05

<nekomancer[m]> .seen chewitt

2020-08-09

<archetech> chewitt Original N2 u-boot files will not boot N2+

2020-07-21

<chewitt> at least, not for the one-liner device-tree kind of changes
<chewitt> upstreaming is really not that difficult or scary
<chewitt> if it does .. figure out the fix and I'll be happy to either upstream it with attribution or coach someone on how to upstream stuff
<chewitt> VIM(s) and N2 have different power plumbing so no guarantee the untested opinion of Khadas staff has relevance to N2
<chewitt> ergo.. problems don't exist
<chewitt> and Khadas don't send emails to the linux-amlogic list to report problems
<chewitt> I don't ever run "ifconfig eth0 down && ifconfig eth0 up" so I never see issues

2020-07-20

<nekomancer[m]> as it was on chewitt
<nekomancer[m]> chewitt: did you have a new set of N2 dtb with RTC, that buildable?
<nekomancer[m]> chewitt:
<Tonymac32> chewitt confirmed on the linux-amlogic feed that the green line on the lefthand side of the screen exists, so I'm not seeing things. (Just verified it's on my La Frite as well)
<chewitt> so each device having a unique name is here to stay
<chewitt> nb: I did attempt to upstream something that simplified the model names in device-tree, but that was shot down in flames
<chewitt> but since the routings are visible in device-tree it should be possible to query device-tree and then use some rules to configure stuff semi-automatically
<chewitt> mixer setttings in userspace is a slightly bigger challenge
<chewitt> Tonymac32 so far the alias approach in alsa-lib allows everything to come down to four confs which is more manageable
<Tonymac32> So I will adjust all to be like chewitt's for our collective sanity
<Tonymac32> chewitt likes to make them all the same (makes sense), mainline makes every single one different

2020-07-19

<chewitt> just another board (tm)
<chewitt> not sure it will get much airtime now I have a new shiny to play with, but I didn't find any issues with it
<chewitt> it works fine
<chewitt> the N2 board support also stops the MAC changing on each boot which is annoying
<chewitt> C4 doesn't boot reliably without the 24MHz hack .. still to be investigated
<chewitt> @Tonymac32 this is my u-boot branch .. it has some extras https://github.com/chewitt/u-boot/commits/2020.07-patches
<Tonymac32> chewitt thanks for the N2+ info/stuff, I had N2 booting mainline u-boot but the kernel was locking immediately and I was at the time setting up sound on 5.6. I need to circle back
<chewitt> still has the all-time boot record for boards I've fiddled with :)
<chewitt> dmesg http://ix.io/2s4L
<lanefu> chewitt: that definiately makes it leaner...
<chewitt> this is running LE defconfig which avoids 50% of the drivers that Armbian probably packages
<chewitt> ^ faster than a greased weasel
<chewitt> Startup finished in 1.767s (kernel) + 1.645s (userspace) = 3.413s
<chewitt> kodi.target reached after 1.645s in userspace
<chewitt> N2PLUS:/ # systemd-analyze
<chewitt> I figured providing an obviously named folder will stop emails asking if the N2 sources also work with N2+ :)
<chewitt> the LE sources for odroid-n2/odroid-n2-plus now contain the same data
<chewitt> older (original) N2 sources only boot N2, will not boot N2+
<chewitt> same (new/updated) u-boot config and sources works for N2 and N2+
<nekomancer[m]> chewitt: `./compile.sh docker CLEAN_LEVEL=make,debs,sources,images FORCE_CHECKOUT=no BOARD=odroidn2 BRANCH=dev RELEASE=focal BUILD_DESKTOP=no KERNEL_ONLY=no KERNEL_CONFIGURE=yes EXPERT=yes BUILD_MINIMAL=no ROOTFS_TYPE=btrfs`
<chewitt> one of these days I should learn how to build Armbian :)
<chewitt> (his commit style obscures that kind of thing)
<chewitt> it's how oleg builds images (from his own branch) but I don't know what hacks he added to the armbian build system for that
<chewitt> since that's a cohesive set of patches
<chewitt> it would be easier to take my whole branch
<chewitt> for Amlogic stuff, I'd be happy if you flagged anything that I don't have in my branch so I can include it and then maybe upstream
<Tonymac32> any patch that is from chewitt will probably fail
<chewitt> eww..
<Tonymac32> yeah that's expected, we aren't patching chewitt's kernel. ;)
<chewitt> might be easiest to add ".patch" on the URL and then add a patch file into the laundry list of patches Armbian hoards :)
<nekomancer[m]> chewitt: I don't understand how to get right, I only know hash you tell me some day sago
<chewitt> and I rebase that particular branch constantly so hashes might have short TTL
<chewitt> nope
<nekomancer[m]> chewitt: does your branch have name or tag instead hash?
<chewitt> which is currently fc32c3cbaf2d43b10cd1883db4536dfbb7089b05
<chewitt> best to pull from the HEAD commit of the branch
<chewitt> most build systems will request a snapshot from github
<chewitt> ahh.. no
<chewitt> if you cloned the entire repo, you start in the master branch
<nekomancer[m]> If I want to build armbian image with chewitt patch — I have to add to compile.sh parameters `KERNELSOURCE=https://github.com/chewitt/linux.git KERNELBRANCH=commit:25db1cd3522903f4b7bb62cba2c0da60e5c76064`? is it correct?
* nekomancer[m] clone chewitt repo, but can't get commit https://github.com/chewitt/linux/commit/25db1cd3522903f4b7bb62cba2c0da60e5c76064 with checkout.

2020-07-18

<nekomancer[m]> <Tonymac32 "nekomancer what's up?"> no-no, just stupid question on chewitt's github repo
<nekomancer[m]> chewitt ping
<chewitt> because it works for me :)
<chewitt> @nekomancer[m] are you building from my branch? .. did you enable CONFIG_RTC_DRV_PCF8563=m in defconfig?

2020-07-17

<nekomancer[m]> chewitt: I set to my N2 `Linux droid 5.7.9-meson64 #trunk.40 SMP PREEMPT Fri Jul 17 18:12:38 UTC 2020` from nightly. but seems /dev/rtc stull point to wakeup timer. What's wrong? Shouldn't I get a patch you merged https://github.com/chewitt/linux/commit/25db1cd3522903f4b7bb62cba2c0da60e5c76064 ?

2020-07-16

<chewitt> unless you mean the vrtc thing
<chewitt> if you prefer to read code on the N2+

2020-07-12

<chewitt> something is still missing on the AIU driver for GXBB > GXM devices
<Tonymac_32> chewitt thank you!
<chewitt> (on G12A+)
<chewitt> gives working pass-through audio :)
<chewitt> plus device-tree content in my amlogic branch

2020-07-08

<chewitt> I see this habit of bundling bin files in lots of the retro-gaming distros and I don't see why it's needed
<chewitt> DRM should read the properties from HDMI and configure itself automatically
<chewitt> but I'm wondering why people need to use edid.bin files
<chewitt> anything involving Xorg is a mystery to me
<lanefu> chewitt: does this problem make sense to you? basically drm driver identifies output connector as `HDMI-A-1` but Xorg doens't see/ have that port configured.... https://forum.armbian.com/topic/14167-4k-hdmi-output/?do=findComment&comment=105039

2020-07-04

<Tonymac32> chewitt insane question, when I was patching 5.6, gxbb audio came out sounding like alvin and the chipmunks, I'm going to double check the chosen clocks, but any notes you might have would be welcome
<ArmbianHelper> lanefu thinks [TheBug] meant to say: jock: thanks for coming and I look forward to reaching out to you and chewitt and some othersheresoon about TVBox stuff, so be looking out :)
<ArmbianHelper> [TheBug] meant to say: jock: thanks for coming and I look forward to reaching out to you and chewitt and some otheres her soon about TVBox stuff, so be looking out :)
<[TheBug]> jock: thanks for coming and I look forward to reaching out to you and chewitt and some others her soon about TVBox stuff, so be looking out :)
<chewitt> LE experience is that users don't distinguish between an official release and a community release .. everything is "LE"
<chewitt> support anything that has an upstream device-tree .. and ensure people who want to support boxes upstream the stuff they work on
<chewitt> LE solution was the rename both firmware, detect the board compatible at boot-time and use this to symlink the appropriate firmware
<chewitt> VIM1 basic has this problem too .. chip IDs as the wrong broadcom device and there are other devices using the other (correct) ID
<chewitt> i'd be very happy to collaborate on patches and sending anything that armbian users (and you lot) find upstream
<chewitt> branch that gets rebased more frequently here https://github.com/chewitt/linux/commits/amlogic
<chewitt> I am mostly tracking whatever isn't highly experimental here https://github.com/chewitt/linux/commits/amlogic-5.7-integ
<chewitt> :)
<Tonymac32> chewitt time is key, I've put maybe 25 minutes in because of other topics
<chewitt> (hello btw)
<chewitt> I have working audio on all the board/boxes I have .. so there's a solution, even if armbian hasn't figured out the patches yet
<chewitt> still haven't been able to have that probe right
<chewitt> I will send once I figure out what's needed for rtc on C4
<chewitt> IgorPec I have some bits for enabling rtc which I plan to upstream too
<IgorPec> chewitt: tnx, we will double check if we are missing anything
<chewitt> https://github.com/chewitt/linux/commits/amlogic has everything you might want
<ArmbianHelper> chewitt meant to say: the userspace stuff is still "guesswork in progress"
<chewitt> s/guess/guesswork
<chewitt> the userspace stuff is still "guess in progress"
<chewitt> I'm still working up enthusiasm to submit minimal audio stuff in device-trees (and then be told off for doing it all wrong) by the audio maintainer
<chewitt> Tonymac32 all the audio stuff is upstream in 5.8 apart from adding support to device-trees

2020-06-30

<Tonymac32> thanks chewitt I'll take a look later today (day job ATM)
<chewitt> I believe a fix will come, but until then the revert is fine
<chewitt> 5.7.5 and earlier are okay
<chewitt> commit is 1bb707fbfd5c246028d76b8f11a19dfd118d6306
<chewitt> 5.7.6 has an accidental backport from 5.8 included that needs to be reverted
<chewitt> I spotted someone mentioning that audio was broken in 5.7.x for Amlogic the other day (or night)

2020-06-27

<lanefu> chewitt: 5.7.6
<chewitt> 4K timing exist for Amlogic for ~ 2 years now
<chewitt> lanefu: which kernel?

2020-06-19

<nekomancer[m]> chewitt: did they accepted your patch?
<chewitt> nekomancer[m] it seems moving vrtc to /dev/rtc1 is as simple as setting an alias in dts

2020-06-15

2020-06-14

<chewitt> the other 50% want (or need) to be on the bleeding edge, which is, err.. bloody
<chewitt> I see about 50% are happy with that kind of "board support package" approach
<chewitt> vendors like HK that dare to offer something with 5.4 LTS kernels is already a bold move
<chewitt> if you consider that Amlogic itself is blind to the existence of Linux .. everything is focussed on Android
<chewitt> some udev fu might be possible
<chewitt> but, now it's been mentioned someone will sleep on it and likely have an idea
<chewitt> "It can be implemented but not sure it will be accepted upstream"
<chewitt> I've no idea how that might be done, but I'll start a convo with the folks who added it
<nekomancer[m]> okay. rtc1 much better than nothing. chewitt , please, push it at least as rtc1
<chewitt> as long as they continue to bankroll upstream development I'll not whinge too much :)
<chewitt> ^ better description
<chewitt> this was done for the next-gen Google device, so there was a reason behind it
<chewitt> disabling vrtc ensures the on-board will be /dev/rtc0, but also ensures the device cannot self-wake on timer, as (reading the commits) this requires vrtc
<chewitt> I've seen khadas adding a &vrtc; node in device-tree to disable vrtc
<chewitt> i'm not sure it's been thought through by the maintainers
<chewitt> no idea
<chewitt> numbering is prob. dependent on probe order so vrtc (needed for wake) is /dev/rtc0 and on-board is prob. /dev/rtc1
<chewitt> Armbian might be, but we expect a TV to be connected
<chewitt> LE is not a headless distro :)
<chewitt> or you can add "ssh" to kernel boot params in extlinux.conf to force the daemon to start
<chewitt> first-run wizard has a screen for "enable SSH"
<chewitt> sad, or secure?
<nekomancer[m]> chewitt: is renumbering possible on dtb level?
<chewitt> looking at the collection of stuff in the Armbian repo the other day, some upstreaming is needed :)
<chewitt> and if you confirm, I'll send it upstream so we (LE/Armbian) can later drop the patch
<chewitt> test the image, if it works you can nag one of the Armbian maintainers to pick the same patch to their kernel patch-set
<chewitt> indeed
<chewitt> write to a spare SD card and have the boot switch to the right
<nekomancer[m]> chewitt: LE image? where to get it?
<chewitt> I can also point you to an LE image that you can put on an SD card to validate that the RTC bits are working
<chewitt> I guess the second one points to a URL somewhere
<chewitt> my N2 is being remotely used by someone for testing at the moment so I can't check myself
<chewitt> let me know if the rtc chip works
<chewitt> nekomancer[m] if you build from https://github.com/chewitt/linux/commits/amlogic there is a patch from the HK forums in the branch
<chewitt> if someone can point me to the device-tree content required i'll happily add it, and if someone then confirms it works I will send it upstream
<chewitt> doesn't look the N2 has any i2c stuff defined
<chewitt> and I don't track the HK forum much as it's always dog slow for me
<chewitt> nekomancer[m] not aware that I did anything specific for either of them, N2 is relatively untouched (and upstream)

2020-06-13

<nekomancer[m]> chewitt: do you have watchdog and hwclock for odroid-N4 in your device tree?

2020-06-12

<chewitt> Martin B was looking at the bsp sources to see what might be missing
<chewitt> there's some juju needed in the driver to fix this I think
<chewitt> ahh, reading an Armbian forum thread, the ethfix dtb is with 100Mb speed forced, so not a 'fix' and no further in the investigation than I got :)
<chewitt> but being good neighbours with Armbian would be great
<chewitt> not sure I can volunteer for anything .. one distro is enough :)
<chewitt> I'm already maintaining TV boxes so happy to share my experiences of what works, what to avoid
<[TheBug]> chewitt: yes balbes150 and a bunch of others are also participating -- the conversation is more so what the long term support of Armbian for TV boxes will look like, if Armbian will take a vested interest in keeping things up to date, etc. We are in need of both input on how that can be done as well as people who have some time to vounteer and help with those processes -- if you do think
<chewitt> I'm 'chewitt' on the Armbian forum
<chewitt> we're all chasing the same kernel bits
<chewitt> Oleg (balbes150) has been using my kernels to push Armbian on TV boxes for a while
<chewitt> [TheBug] I have no direct interest in working on Armbian, but there is large overlap with LibreELEC (which I work on) so happy to participate
<chewitt> I don't have A95X-F3 or X96-air but they appear to be working okay .. or with the same common faults