<Xogium>
peterm6881: boot does hang at that line, not reset
<Xogium>
it resets after reached target path units
<peterm6881>
yeah you're right, you can always trust me to fuck it up
<peterm6881>
thats how i am
<peterm6881>
ironically , I copied and pasted that from an activaely rebooting system
<peterm6881>
actively
<peterm6881>
and STILL managed to get the wrong message
<Xogium>
so for now I'm going to rename the tag and readd the dontuse, since its broken as hell
<Xogium>
if that seems fine with you that is
<Xogium>
to be fair I think they will not care about our problem, because one: we don't user their software, two: we don't use their vendor kernel nor u-boot
<Xogium>
*use their
<Xogium>
vendors expect you to use their stuff, not mainline
<peterm6881>
I emailed a correction, but weidly Gmail didnt copy you and Steve in, so I had to forward you the correction
<peterm6881>
weirdly, obviously
<peterm6881>
im pretty certain sipeed will encounter the same problem
<peterm6881>
besides, they are bound to know their stuff is outdated junk in comparison
<peterm6881>
but I take your point
<Xogium>
oh they know its outdated, but since they never touched mainline with a 10 foot pole
<peterm6881>
im confident they will be able to replicate that problem, and may already be aware of it, and may even have a solution
<peterm6881>
but we'll see. certainly it cant do any harm
peterm6881 has quit [Remote host closed the connection]
<Xogium>
I've set up a getty on ttyGS0 in the usual manner (I'm running up-to-date Jessie). Sweet jesus that's old
<Xogium>
+1 Saved me hours! Use serial-getty@.service not getty@.service
<Xogium>
TTYVTDisallocate comes from systemd, but the problem is that the getty@.service unit is meant for virtual consoles. For serial consoles the serial-getty@service should be used.
<Xogium>
but we already do this
<Xogium>
besides for them, its not a *boot* issue its a *reboot* shutdown problem
<Xogium>
the tty comes up okay, but when they shutdown or do a soft reboot the board refuses to power off
<Xogium>
and all that because they used the wrong getty service from the start, apparently
<peterm6881>
can we compare the getty service configuration files from the OPi0 with the Lichee ones?
<peterm6881>
there MUST be a difference
<Xogium>
how could there be ? We use the very same systemd release in both cases, and we definitely did not modify the serial-getty
<Xogium>
I've just checked the speedsaver repo for orange pi, there is no modified serial-getty service
<peterm6881>
fuckity fuck
<peterm6881>
:)
<peterm6881>
whats the path to those files on the device?
<peterm6881>
fuck it, im having a look
<Xogium>
should be someplace in /usr/lib
<Xogium>
/usr/lib/systemd/system
<Xogium>
probably
<peterm6881>
yeah theres a serial-getty@.service
<peterm6881>
thats the file that does the magic right?
<Xogium>
yep
<Xogium>
gest instantiated for every serial console
<peterm6881>
wa?
<Xogium>
that's what the @ means
<Xogium>
you get that service running for every serial console available
<Xogium>
serial-getty@ttyS0, @ttyGS0 in our case
<Xogium>
instead of makin multiple services for each getty instance you want to run for every device, you invoke the same thing with a different tty name
<Xogium>
*making
<peterm6881>
I have to say, Qterminal is shit
<peterm6881>
cant fucking scroll, cant select text that isnt in the active window
<Xogium>
well that's serial terminal
<Xogium>
I kind of get to cheat with my screen reader because it can copy / paste everything it says
<Xogium>
this sucks so damn much. I wish there was jtag
<peterm6881>
its literally impossible for me to read the serial-getty@.service file on the lichee
<peterm6881>
if i paste the contents of it here from the OPi0, would yoiu be able to compare the 2?
<peterm6881>
its 46 lines
<peterm6881>
Qterminal fucks literally every single thing up when i access the same file on the Lichee, so i cant compare
<Xogium>
how come you were able to access on the orange pi but not the lichee ?
<Xogium>
but no matter, yes, they are practically identical, as I did say before ;)
<Xogium>
I just checked on my end
<Xogium>
the more this goes, the more it feels like a driver bug that is so bad that you get violently ejected out of userspace all the way back to u-boot. Sure is looking more likely
<peterm6881>
the text gets all fucked up on the lichee
<peterm6881>
in the Qterminal window
<Xogium>
and on orange pi it doesn't ?
<peterm6881>
even using uart0
<peterm6881>
no
<Xogium>
lol
<Xogium>
what's this joke, honestly...
<peterm6881>
yeah its toally unusable
<peterm6881>
so i cant commpare the two files
<peterm6881>
compare
<Xogium>
yeah no worries I've done that
<peterm6881>
totally ffs
<peterm6881>
when you say practically identical, are there any differences?
<Xogium>
did it get this way because I enabled gadget ? Or was it always unusable ?
<Xogium>
nop I mean they are literally the exact same file
<peterm6881>
i dont think it was ever a problem before, but then i didnt use the bridge much
<Xogium>
fair enough
<peterm6881>
i think this rendering issue came in with gadget
<peterm6881>
what are you thinking?
<peterm6881>
but like i said, i didnt use it much
<Xogium>
well, I think its getting more likely to be hardware going bonkers, or software aka kernel driver problem that causes everything to freeze and get reset back to u-boot
<Xogium>
I've never seen a driver issue so violent that linux gets reset as if you had pressed a reset button, but everything's possible, certainly
<Xogium>
but I wish I had an osciloscope to be able to measure the voltage and make sure there's no major drops or spikes when that happens
<peterm6881>
theres a file called usb-gadget.target in the same directory
<Xogium>
yeah its a target
<peterm6881>
are they the same?
<Xogium>
its not an unit, its more or less used to group other units together
<peterm6881>
i think we've gone past the point where we need to vaporise Gadget Serial
<Xogium>
like getty target is used to regroup all the units for gettys together and have them launch at a precise moment in the boot sequence
<peterm6881>
lets kill it unless Sipeed respond
<Xogium>
well to be honest, we've gone past the point where we need to favorise gadet, at all. For all we know, it'll act crazily with every other gadgets
<peterm6881>
i agree I think the driver for V3s simply doesnt work, and nobnody noticed because nobody tried it
<peterm6881>
nobody
<Xogium>
that's a giant showstopper for makers
<peterm6881>
well thats their problem
<peterm6881>
somebody will fix it
<peterm6881>
no way im looking for another soc
<Xogium>
given that board has been around since 2017... I'm kind of starting to dout that
<peterm6881>
we have to focus on our core use case
<Xogium>
and seeing as I would need gadget for my myna use, I think that makes the v3/v3s definitely a brick wall
<peterm6881>
yes but Cesar Wu and icenowy only worked on it for 6 months, and literally 3 people bought it
<Xogium>
yeah, like 99% of allwinner SoCs
<peterm6881>
so lets agree to never speak about Gadget Serial again, and kill it in its cradle of filth
<Xogium>
the arm boards are a joke
<peterm6881>
to be fair, Orange Pi Zero keeps on trucking
<Xogium>
yeah, but they took a higher end SoC
<peterm6881>
we havent had nearly so much problems, H2+ and H3 are way out in front
<peterm6881>
we just cant fucking use them, and they run too hot
<peterm6881>
revert Gadget and lets move on, fuck it
<peterm6881>
agreed?
<Xogium>
fair enough with me
<peterm6881>
what do you need Gadget for, just curious
<peterm6881>
it'll be interesting to see if any of those 3 guys from Sipeed reply
<Xogium>
ah mass storage to extract data from the board without removing the micro sd card, also for update
<Xogium>
device update over usb
<peterm6881>
yeah it would have been nice, but we've done all we can
<peterm6881>
we need to kill it for now
<peterm6881>
perhaps we could pay someone to write a driver for it
<peterm6881>
dunno
<peterm6881>
certainly at the minute whatever has been implemented is a bag of shite
<peterm6881>
and nobody noticed because nobody is working with V3s atm
<Xogium>
I mean, it could be systemd. But there's like 1% chance of it being systemd
<peterm6881>
i appreciate its too much work to rule out systemd
<peterm6881>
I dont wanna take you out of your comfort zone into unknown territory
<peterm6881>
i agree i think its a driver issue, something broke in V3s
<peterm6881>
because OPi0 is perfect, never had any problems talking to it over Gadget
<peterm6881>
so either Allwinner fucked something up in the hardware implementation, OR, if its a different kernel driver for the V3s, its broke
<Xogium>
let me check the driver real quick
<peterm6881>
either way, we're screwed, for now
<peterm6881>
yeah that would be useful
<peterm6881>
work your magic ;)
<peterm6881>
I can browse at will through OPi0, never fails to boot, all done through one micro usb cable
<Xogium>
well wouldn't you know it: they use the very same freaking driver, configured exactly the same way
<peterm6881>
good work Xogium
<Xogium>
smells more and more like a hardware fuckup then
<peterm6881>
yes I dont think theres any getting away from that now
<Xogium>
either on the lichee side or on the SoC itself
<peterm6881>
thats a good question, I'd need to check the schematic
<peterm6881>
i think its all in the SoC
<peterm6881>
i dont think there is anything in between
<Xogium>
prolly not
<peterm6881>
yup straight off the soc to the micro usb
<peterm6881>
also, you were right, there is only one USB on the soc.
<peterm6881>
they just split it between the micro usb socket and the pin header
<Xogium>
ouch
<Xogium>
that's mean
<peterm6881>
;)
<peterm6881>
are you tired?
<peterm6881>
I appreciate all your brain power today
<Xogium>
not really tired yet
<peterm6881>
its been useful exploring this, if ultimately fruitless
<peterm6881>
does seem like a hardware issue with the soc
<peterm6881>
nil we can do about that
<Xogium>
I can't confirm this at 100000% but yeah... I see honestly no other explanation for the truly strange behaviors
<peterm6881>
id be amazed if Sipeed respond
<peterm6881>
I wonder what Allwinner fucked up
<peterm6881>
you would think they know what they're doing by now, although to be fair I dont think i would personally get to far fabricating a SoC ;)
<peterm6881>
too
<Xogium>
neither would I lol
<peterm6881>
:)
<Xogium>
such complex things those are
<peterm6881>
sadly, we have to kill it. Nice try Xogium
<peterm6881>
ill eat my hat if Sipeed reply
<Xogium>
one of these days, x86 will become like arm. No more sodimm ram for laptops, no mor dimm for desktop, nor more big as hell gpu, no tiny square shaped cpu
<Xogium>
just a whole bunch of chips on a motherboard
<peterm6881>
maybe the current chips fixed it
<Xogium>
x86 will be full of SoCs
<Xogium>
it'll have to do this if it wants to keep on living. We're reaching out the power limits of x86 in the current state it is in. On the die of every chip for x86, there's more space for thermal management than anything else
<peterm6881>
or, no new chips will be made and there will be a zombie apocalypse of former electronics manufacturing employees
<peterm6881>
because the lead time became one millenium
<peterm6881>
wait, is x86 still a thing?
<peterm6881>
wow
<Xogium>
the rtx4080 gpu nvidia is working on can peak at 450w load. With tiny spikes the double of that. You need a freaking 1000w psu to run your pc and avoid brownouts
<Xogium>
I'm talking x86 as is x86_64, not i686 and similar
<Xogium>
to be clear we'r reaching out the power limit of the current form factor of that architecture
<Xogium>
*we're
<Xogium>
takes all of that to manage truly amazing graphics and performances, whereas in arm, the m2 from apple is leaving anyone in the dust. Its performance is comparable to the most badass nvidia gpu or very nearly there, while the whole machine draws a quarter of what the gpu alone can draw
<peterm6881>
are you becoming an mlg?
<Xogium>
mlg ?
<peterm6881>
major league gamer
<peterm6881>
how do you know so much about this stuff? This is the kinf of language Ross talks :)
<peterm6881>
kind
<Xogium>
well.. I've learned a ton of stuff since I started with embeded and with getting interested in getting a custom built pc
<peterm6881>
i havent heard of the M2 from Apple, why wasnt I told about this..
<Xogium>
one day, custom built pc won't be possible anymore. Best build one while you still can
<peterm6881>
you're an enthusiast
<Xogium>
even apple is realising that arm is the way to go forward
<peterm6881>
I should get the linux credit card guy to email Allwinner
<Xogium>
why they built the m2
<peterm6881>
he used the V3s, remember him? Allwinner reached out to HIM
<peterm6881>
no wait
<peterm6881>
thats wrong
<Xogium>
actually no, he used the tiny SoC in the lichee nano and in the mangopi
<peterm6881>
he used that chip thats on the Mango Pi,
<peterm6881>
yeah, my head is officially fried ;)
<Xogium>
I completely blanked out on the name of that SoC lol
<Xogium>
besides we already contacted him once, and he never replied
<peterm6881>
fc100s and fc200s
<peterm6881>
fuck what a trip this has been
<Xogium>
yeah that's them
<peterm6881>
its funny, you never think a simple request for Gadget Serial would lead to 2 days head scratching and an impasse
<peterm6881>
to discover the chip is cooked
<Xogium>
want me to be honest ? I think the whole of the el cheapo allwinner SoC are bullshit. Higher end like h3 are more respectabl
<Xogium>
*respectable
<peterm6881>
im becoming way more sympathetic to that view
<peterm6881>
its strange, you would think in this space the maximum traction and support would come from the most affordable devices. Not so
<Xogium>
not only are they bullshit from a mainline linux/u-boot or what have you perspective because as usually noone ever bothers mainlining but community folks and they quicly seem to get abandoned when they do get in mainline, suffice to look at how many time icenowy started porting something and then ended up dropping it. But they have crappy hardware design on top, it appears
<peterm6881>
its possible Allwinner fixed it, theres no way to find out
<Xogium>
yes, it could be
<peterm6881>
we only have Lichees to test on
<Xogium>
we have no idea if there were later SoC revisions that fixed that, let alone which SoC revision the mangopi is using
<Xogium>
err lichee
<peterm6881>
nice wrok finsing the driver is the same, and the serial gadget service config file is the same
<peterm6881>
work
<peterm6881>
finding ffs
<peterm6881>
yeah fuck knows Xogium
<peterm6881>
ok heres the policy
<peterm6881>
we proceed as follows
<Xogium>
granted I was pretty sure they were similar, if not identical. Allwinner tends to use the same pinout and way of making things should they prove to work out well
<Xogium>
that's why the usb on the SoCs for h3 and v3s are the same, so is the uart
<Xogium>
and i2c, if I'm not mistaken
<peterm6881>
switch off Gadget Serial, wait for Sipeed to respond, and revisit when we get our first prototypes, agreed?
<Xogium>
fair enough with me
<peterm6881>
I think we need to crack on down this road, for lack of alternatives
<Xogium>
the tag for 19 will stay there for historical purpose, but has dontuse as suffix, so people are aware its currently a bad tag
<Xogium>
Revert "board/speedsaver: disable v4l subsystem and enable usb gadget serial."
<Xogium>
This is so broken right now that it isn't workth the hassle. Suspecting a SoC defect or a driver problem, the latter being unlikely given this same driver works on orange pi zero without a problem.
<peterm6881>
yuck
<peterm6881>
I messageed Allwinner Tech via their web portal but im guaranteed no response
<peterm6881>
Hi,
<peterm6881>
Were you aware of this issue early in the production cycle of the V3s and has it now been resolved?
<peterm6881>
Regards,
<peterm6881>
We want to use ttyGS0 Gadget Serial on the V3s, but when we enable it in kernel we get a bootloop until we start listening on the port. This appears to be a hardware fault, because the same kernel, systemd and Gadget Serial kernel driver for the H2+ / H3 works fine.
<peterm6881>
Peter McKinley
<Speedsaver>
peterm6881: Error: "Ltd" is not a valid command.
<peterm6881>
Speedsaver Ltd
<peterm6881>
meh
<peterm6881>
you should get some rest
<Xogium>
I'm too hot again
<peterm6881>
Clear with periodic clouds
<peterm6881>
°C°F
<peterm6881>
13
<peterm6881>
Humidity: 88%
<peterm6881>
Precipitation: 3%
<peterm6881>
Wind: 4 mph
<peterm6881>
currently 13 degrees Celcius in Fivemiletown, just sayin'
<peterm6881>
you're always welcome here
<peterm6881>
i emailed Sochip for a response from Allwinner