libv changed the topic of #linux-sunxi to: Allwinner/sunxi development - Did you try looking at our wiki? https://linux-sunxi.org - Don't ask to ask. Just ask and wait for an answer! - This channel is logged at https://oftc.irclog.whitequark.org/linux-sunxi
apritzel has quit [Ping timeout: 480 seconds]
apritzel has joined #linux-sunxi
Xalius has quit [Quit: Leaving]
apritzel has quit []
apritzel has joined #linux-sunxi
Mangy_Dog has quit [Ping timeout: 480 seconds]
ftg has quit [Read error: Connection reset by peer]
samueldr has joined #linux-sunxi
lurchi_ has joined #linux-sunxi
lurchi__ has quit [Ping timeout: 480 seconds]
ssvb has quit [Quit: Leaving]
cmeerw has joined #linux-sunxi
cmeerw has quit [Ping timeout: 480 seconds]
pentabarf has joined #linux-sunxi
prefixcactus has joined #linux-sunxi
pentabarf has quit [Ping timeout: 480 seconds]
kayterina[m] has joined #linux-sunxi
<libv> what do you guys think, should i spend a few minutes throwing people, who are here already, out of the freenode channel?
<swiftgeek> maybe borrow a bot from coreboot
ats has joined #linux-sunxi
ats_ has quit [Ping timeout: 480 seconds]
<rellla> libv: is it possible to send them a personal message, too? i guess most of the people have this channel active in their bouncer. as i had...
jelly is now known as Guest290
* rellla is thinking about doing that with #lima and #cedrus, too...
<rellla> and afaik you will get this personal popup message once you open your irc client. i would go one step further and throw all people, because these ones may have missed the move to oftc.
<rellla> and then they are probably wondering, why no conversation is happening anymore on the channel - if they don't backread or don't read the topic. they can't speak anyway...
Xalius has joined #linux-sunxi
Guest290 is now known as jelly
warpme_ has joined #linux-sunxi
<libv> i would just do it all manually i guess
<libv> but i'll probably do it on monday, while childsitting
<karlp> why does it matter?
<libv> the fewer people are there, the fewer incentive people have to stay there i guess
<libv> but you're probably right
<apritzel> if people can't speak anyway, they will sooner or later drop out anyway, so why do we need to do something anyway?
<apritzel> not everyone uses IRC on a daily or even weekly basis
<libv> right, i guess i just want to get this over and done with asap, and move on
<karlp> just leave yourself, out of sight, out of mind :)
tnovotny has joined #linux-sunxi
<warpme_> jernej: re: jout q about g31 on 2 h616 boards i have: yes. it is not working. panfrost loads, seesm to report ok caps but dmeg every 60sec is full of:
<libv> apritzel: thanks :)
<warpme_> maybe issue is with wrong interrupt config in DT (so panfrost not receiving int when gpu finishes job)?
<apritzel> warpme_: your interrupts looked alright, though
<warpme_> arh (was hoping it was only me....)
<apritzel> warpme_: at least when compared to the manual
<apritzel> you can try that other interrupt
<warpme_> jernej: did you get working g31 on h616 at all?
<apritzel> warpme_: 94 instead of 97
<warpme_> apritzel: so something like this:
<warpme_> <------><------><------><------> <GIC_SPI 94 IRQ_TYPE_LEVEL_HIGH>;
<warpme_> <------><------><------>interrupts = <GIC_SPI 95 IRQ_TYPE_LEVEL_HIGH>,
<warpme_> <------><------><------><------> <GIC_SPI 96 IRQ_TYPE_LEVEL_HIGH>,
Mangy_Dog has joined #linux-sunxi
<apritzel> warpme_: but the PX30 seems to use the same IRQs as your first version (so not using "event")
<apritzel> so that's probably not it
<warpme_> apritzel: with 94 instead 97 the issue is the same. int are like this (this is with 94 instead 97):
gsz has joined #linux-sunxi
<apritzel> warpme_: yeah, as told above, was worth a shot anyway
<apritzel> warpme_: do you see 0 interrupts when using 97 as well?
tnovotny_ has joined #linux-sunxi
tnovotny has quit [Remote host closed the connection]
<warpme_> apritzel: nope. with 97 it look like this:
<warpme_> so it look like ints are ok
<warpme_> timeout is exactly every 60sec. a bit long period as for gpu jobs....
<warpme_> does everybody knows new irc server with panfrost channel?
<apritzel> warpme_: the logs say it moved to OFTC
<jernej> warpme_: I didn't work with H616 GPU yet
<jernej> just check dt node in bsp and compare it with yours
<jernej> they should contain same information
<warpme_> jernej: as source to compare i was using tvbox android decompiled dt.
<warpme_> may you point me to bsp for h616?
<jernej> I'm on a phone now, so no
<warpme_> ah ok :-)
<jernej> check xunlong github
ssvb has joined #linux-sunxi
cmeerw has joined #linux-sunxi
tnovotny_ has quit []
prefixcactus has quit [Ping timeout: 480 seconds]
tuxd3v_ has joined #linux-sunxi
vagrantc has joined #linux-sunxi
tuxd3v_ has quit []
<libv> tuxd3v_: i set it to only allow people who are registered can talk as we had some spammers already
<libv> heh
<jernej> warpme_: it looks fine to me...
<jernej> you don't need to specify interrupt parent, since gic is already default
<warpme_> jernej: thx. panforst bbrezillon says: "assuming both your amlogic and allwinner SoC embed the same GPU (G31), and one works while the other doesn't, I'd bet on a kernel side issue (PM or cache related)"
<jernej> yeah, I concur
<jernej> warpme_: are you 100% you properly powered it?
<jernej> *sure
<warpme_> as i see interrupts counters are growing with time it tells me that gpu is somehow alive and constantly doing some jobs.... this will suggest rather cache coherency hypothessis....
<warpme_> re: PM: well. i "guess" by PM names. in OpiZero2 i see dcdcc is named "vdd-gpu-sys" - so i routed this to mali-supply....
pentabarf has joined #linux-sunxi
<warpme_> issue for me it that panfrost dmesg verbosity is very difficult to get conclusions what is wrong....
<jernej> yeah, dcdcc is correct
<warpme_> re: "don't need to specify interrupt parent" - oh for sure there are many places asking for clanup/optimization. but i think this might be 2nd pass - when we will get it working. so far h616 starts to see pretty good at starter phase: usb, hdmi, eth, cec, vdec, ir are basically working. gl loads but not works. no audio.
pentabarf has quit []
<jernej> warpme_: there is another special bit for GPU, this time for secure memory
<jernej> make sure that 0x3007054 bit 0 is 0
<warpme_> for sure many subsystems are need polishing but imho having as starter an working media playback may encourage new ppl who will maybe help polish those subsystems...
<jernej> at same place that you added hack for power
<warpme_> oh qll. let me try
tuxd3v has joined #linux-sunxi
<tuxd3v> hello all
<jernej> hello :)
<tuxd3v> :)
<tuxd3v> apritzel, what's the news?
Daanct12 has joined #linux-sunxi
<warpme_> jernej: so i tried with adding in uboot shell mw.l 0x3007054 0 . not helps. still job timeouts...
<jernej> ok, I'll compare vanilla bifrost driver from ARM web page and AW variant from BSP
<warpme_> many thx. i might be unavaliable next 2..3h - you know - familly commitments :-)
<jernej> well, don't expect anything quickly from me either :)
Danct12 has quit [Ping timeout: 480 seconds]
<warpme_> jernej: btw: event with timeouts - after waiting long enough - i see apearing single gui elements on screen....
<warpme_> event->even with job timetouts
choozy has joined #linux-sunxi
pentabarf has joined #linux-sunxi
choozy has quit [Ping timeout: 480 seconds]
<jernej> warpme_: I don't see anything special: http://ix.io/3oa8
ftg has joined #linux-sunxi
<apritzel> tuxd3v: I bisected it down
<apritzel> first tried releases, then within the 5.6 merge window
<apritzel> the culprit is this one: dc56ecb81a0a
<apritzel> tuxd3v: that introduced CONFIG_SERIAL_8250_16550A_VARIANTS
<apritzel> if I enable this, it works for me
<apritzel> judging by the commit message, that patch allows to avoids some delay(!) in the serial driver probe
<apritzel> it's not the fix, we probably need to add some delay in the hci_bcm driver
<apritzel> tuxd3v: it would be good if you could confirm that it fixes it for you as well
<jernej> warpme_: there is slight possibility that gpu1 clock needs to be enabled too
<jernej> you can enable it by setting bit 31 0x3001674 (using devmem2 for example)
BorgCuba has joined #linux-sunxi
<tuxd3v> apritzel, compiling :)
Daaanct12 has joined #linux-sunxi
Daaanct12 has quit []
Danct12 has joined #linux-sunxi
<tuxd3v> apritzel, that commit is from: Fri Jan 10 18:25:13 2020 -0800
<tuxd3v> it could be indeed that delay that prevent the system from loading firmware, as it timeouts later..
Daanct12 has quit [Ping timeout: 480 seconds]
<jernej> apritzel: it would be interesting to check which flags are set on AW and try with them directly
<jernej> to avoid delay
<apritzel> jernej: you saw this BT issue as well, didn't you?
<jernej> I'm not 100% it is the same issue
<jernej> since I had SDIO problems
<jernej> I concentrated more on wifi than bt
<apritzel> I see. the commit above is for the 8250 driver, so I am still trying to wrap my head around this
<apritzel> it seems like megi and wens had their hand in the BT driver as well, and I see some commits tuning delays
<megi> hmm, I have a problem since around 5.6 with BT firmware loading on A83T tablet, too
<megi> I worked it around by reducing baud rate
<apritzel> megi: 5.6-rc1 is exactly when this was introduced
<megi> I'll retry with that patch reverted
<apritzel> megi: no need for that, just enable CONFIG_SERIAL_8250_16550A_VARIANTS
<megi> ok
<apritzel> (that's what the patch does: introducing this symbol, and skipping some delay when it's not defined)
<apritzel> megi: it was pretty reliable for me, as it survived the bisect over 14 steps
<megi> hmm, yeah, I don't have this enabled, good
<apritzel> megi: maybe it is related to the RTS line?
<apritzel> IIUC it affects the delays during UART probe, I'd guess that those serdev probes happen only after the UART is up
<megi> might be, my device also uses rts/cts
<megi> there's a switch to higher baudrate in that BT driver
<megi> it happens after the switch
<megi> or during it
<megi> commenting out that line disables the switch
<megi> I've seen occasional/rather rare BT initialization failures on A64/pinephone too
<apritzel> Does the UART come up in some default baudrate, and the BT driver increases it?
<megi> BT driver first configures the BT chip via 115200 and tells it to up the baudrate, and then switches the baud rate on the host side too
<apritzel> megi: but that's just a parent change, right? From 24MHz to "PLL_PERIPH"? No actual PLL frequency change involved?
<megi> I don't know what happens on the clk side
<tuxd3v> apritzel, out of the first 5 times, I got 5 times bluetooth working :)
<tuxd3v> I continue to reboot :)
<megi> me too
<megi> one time test, so far
<tuxd3v> apritzel, you unlocked the Holy grail!! :)
<apritzel> megi: that's on that tablet?
<megi> yes
pentabarf has quit []
<tuxd3v> this is my bluetooth def config section: https://paste.debian.net/hidden/877d18c7/
<tuxd3v> my device uses broadcom related stuff( ampak 6212 )
<megi> mine is ampak 6210
<tuxd3v> humm even the very annoying messages of timeout's disappeared :)
<apritzel> tuxd3v: I think the timeout was the actual problem: the driver couldn't communicate with the chip. Any other error reported (firmware loading, etc) was a subsequent error
<tuxd3v> apritzel, agree
<jernej> apritzel: do you think some quirk is actually needed or it's just delay that helps?
akanouras has joined #linux-sunxi
<apritzel> jernej: I don't know. It's a bit puzzling, as the added delay is in the UART driver.
<apritzel> I'd guess that delays only matter once the BT driver comes into play, which happens after the UART driver
Xalius has quit [Quit: Leaving]
<tuxd3v> some tests done on bluetooth: https://paste.debian.net/hidden/7a212336/
<tuxd3v> I made some 20 reboots working every single time
<tuxd3v> now I am preparing to test, shutdown and cold boot, then I will swithc to boot from sdcard :)
<tuxd3v> but yeah its working .. nice discovery :)
jkm has quit [Ping timeout: 480 seconds]
<tuxd3v> apritzel, it just works!!, sdcard, emmc you name it ;)
choozy has joined #linux-sunxi
gsz has quit []
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #linux-sunxi
cmeerw has quit [Ping timeout: 480 seconds]
choozy has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
BorgCuba has quit [Quit: Leaving]
pentabarf has joined #linux-sunxi