Mr__Anderson has quit [Remote host closed the connection]
Cubietruck_noob has joined #linux-sunxi
paulk-collins has quit [Quit: Leaving]
atsampson has quit [Ping timeout: 245 seconds]
atsampson has joined #linux-sunxi
wigyori has quit [Ping timeout: 260 seconds]
<willmore>
Yay, my Opi zero is up and running. Well done, everybody!
<Cubietruck_noob>
cheers! :)
<willmore>
Up next is the PC 2, but that'll be a while longer it seems. Can't wait to play with the SPI NOR. My little 150mil chips haven't showed up yet so I can't modify my zero. :(
<apritzel>
willmore: but it's worth it ;-)
<apritzel>
willmore: how big are the chips you ordered?
<willmore>
apritzel, 16Mb. Seems I picked the right size.
<apritzel>
willmore: I soldered a 128Mbit lately and it worked like a charm
<willmore>
I do have some 128Mb that I mounted on adapters, so I can use them as well. They're in the larger SOIC package, though--can't put them on the zero.
<apritzel>
willmore: well, with 128Mbit you can boot a kernel + initrd from it as well
<willmore>
Yep. That sounds like a lot of fun.
<apritzel>
willmore: huh? The OPi Zero has a SoIC footprint
<apritzel>
or do you mean DIP?
<willmore>
nope, there's the wide SOIC and the 150 mil narrow ones. The ones I have are larger than that.
<apritzel>
ah
<willmore>
Sorry, I'm saying the sizes wrong. The little ones like the 0 wants are the 205 mil SOIC.
<willmore>
The ones I have are the 300 mil SOIC.
<willmore>
It also have 16 pins. NC's...NC's everywhere....
<willmore>
apritzel, were your 128Mb in the little 205mil package?
<apritzel>
willmore: yes, just checked the datasheet
<apritzel>
but apparently they come only in this size anyway for SOIC8
<apritzel>
Winbond W25Q128FV
<willmore>
What's the issue with the Opi zero not using the lower 1.1V out of the regulator for slower speeds?
<willmore>
apritzel, yep, that's the chip. I just got the wrong package when I ordered mine.
<apritzel>
willmore: which maker and model are yours?
<willmore>
Not sure. I'll have to find where I put them. It was before we had guests over and I had to 'clean'.
<willmore>
They may have been winbond.
wigyori has joined #linux-sunxi
<apritzel>
what is the problem with the voltage you are talking about?
<willmore>
I saw a note on the beta support for the zero that the dual voltage regulator for the CPU wasn't currently being used. Maybe that's inaccurate?
<apritzel>
willmore: do you mean the Compatibility section in the Wiki?
<apritzel>
since I have just hacked some I2C code to drive the SY8106 from firmware let me check that ...
<willmore>
apritzel, yes.
<apritzel>
the missing UART pins silk screen is really annoying ...
<willmore>
apritzel, from a quick look at the datasheet, it really looks like the Zero and the One (and other SY8106 using boards) could implement proper multi-voltage DVFS.
<willmore>
The state transitions might be a little slow due to the need to program the regulator with slow 400KHz I2C.
<apritzel>
willmore: do you expect to switch that often that it should matter?
<apritzel>
willmore: also you can adjust the slope
<willmore>
apritzel, I was thinking of interactive changes. But, 1ms is plenty fast and that's 400 clock cycles.
<willmore>
In the PC world, they are working harder and harder on reducing that time. 1ms would be horribly slow for them. It's just a mindset thing. I'll come around.
<apritzel>
mmh, but the OPi Zero doesn't have an I2C adjustable regulator? (according to the schematics)
orly_owl_ has quit [Quit: leaving]
<apritzel>
the CPUX regulator is a SY8113B
<apritzel>
without I2C
mzki has quit [Ping timeout: 256 seconds]
<apritzel>
also S_TWI_{SCK,SDA} are not connected
<apritzel>
there is just PL6, which switches between 1.1V and 1.3V
<willmore>
Ahh, okay, I'm think of something else, then.
<apritzel>
this is also as the Wiki says ;-)
<apritzel>
I checked the regulator on the OPi PC 2, though, it works there
<willmore>
So, only the zero and the one use the 1.1/1.3V split setup and the rest use a more flexable regulator? Or is that oversimplifying it?
<willmore>
Thinking only about Orange Pi boards.
<apritzel>
I have no real insight into all those variants: 32bit, you know ...
<jelle>
I think the PC uses it too willmore
<apritzel>
but the OPi PC 2 uses a simple I2C regulator, unlike the A64 boards, which have a more sophisticated AXP regulator
<jelle>
nice variarty of things :p
<willmore>
apritzel, right, they use a more tablet focused controller with battery control and such.
<willmore>
As much as I like ethernet, I have to think the zero would have been better off without it. They could have laid down the USB connector--or even had two of them side by side.
<willmore>
Oh, well, it is what it is.
bugzc has joined #linux-sunxi
<apritzel>
what's the use of a board without Ethernet? ;-)
<Cubietruck_noob>
RF nesh sensor network, like with tiny Arduino! ;)
<Cubietruck_noob>
*mesh
Da_Coynul has joined #linux-sunxi
jmcneill has joined #linux-sunxi
<jmcneill>
Hi folks. Has anybody noticed any issues with SD idma timeouts on H3 boards?
<jmcneill>
Timeouts or transfer errors
apritzel has quit [Ping timeout: 268 seconds]
<jmcneill>
I've got a batch of cards that are very good at triggering the issue on FreeBSD w/ NanoPi NEO, and another card (different vendor and size) that can trigger it too but takes a bit more effort
<KotCzarny>
shouldn't they just go on one Q8 page with variants paragraph?
<KotCzarny>
or at least feature common q8 template
jernej has joined #linux-sunxi
<KotCzarny>
tkaiser: regarding 4.9 numbers update, as you've said iops is more important, did it change?
Ultrasauce has joined #linux-sunxi
Ultrasauce_ has joined #linux-sunxi
Ultrasauce_ has quit [Remote host closed the connection]
iamfrankenstein has joined #linux-sunxi
perr has joined #linux-sunxi
perr has joined #linux-sunxi
perr has quit [Changing host]
scream has joined #linux-sunxi
lemonzest has joined #linux-sunxi
wens has quit [Ping timeout: 245 seconds]
<jonkerj>
hmm, trying to compile u-boot 2016.11 for h3 (opi+). Default config suggests to select sy8106a-based board (which makes sense), but cannot find a combination of i2c config options that actually compile
wens has joined #linux-sunxi
<jonkerj>
can anyone gist/pastebin a config for sy8106a-based h3 (ie, opi-pc, opi-plus) for me?
<KotCzarny>
hmm, afair i just used the default for h3
<jonkerj>
that's what I thought I did too
<KotCzarny>
maybe 2016.11 just broke something?
<KotCzarny>
retry with older one?
<jonkerj>
the whole point of this excercise is to check if .11 fixes a u-boot shell bug I encounter on .09 :-)
<KotCzarny>
copy .09 config ?
<jonkerj>
that's what I did
<jonkerj>
well, I am going to fiddle some more
jernej has quit [Ping timeout: 260 seconds]
<miasma>
2016.11 defaults worked for me too
<miasma>
jonkerj: what's your gcc version
paulk-collins has joined #linux-sunxi
<MoeIcenowy>
Prepare to purchase a Banana Pi M2 Ultra
bugzc has quit [Ping timeout: 260 seconds]
Putti has joined #linux-sunxi
<miasma>
hm, spi is listed as upcoming feature in 4.10 kernel for h3, but the status matrix shows '?'. http://linux-sunxi.org/Linux_mainlining_effort#Status_Matrix I guess it's not updated yet? or is it still uncertain? I also saw a spi implementation in armbian forums which seems to come from a different author
<KotCzarny>
if one prefers cube form factor over flat brick..
<KotCzarny>
they could've integrated heatsink too.
<tkaiser>
willmore: And the voltage regulation problem only affected 'official' OS images from Xunlong, it seems they still have no clue how to write settings properly to fex file to get the voltage switched (values are interpreted octal and not decimal)
<tkaiser>
KotCzarny: As soon as Xunlong solders SPI NOR flash to the Zero and puts a useful bootloader on it, maybe they could provide another such 'HAT' with a good USB-to-SATA bridge and a mSATA slot on it.
scream has quit [Remote host closed the connection]
scream has joined #linux-sunxi
<miasma>
oh boy :)
silviop has joined #linux-sunxi
<silviop>
Hi, u-boot recognize only 512Mb on my A33 1Gb tablet (and kernel too) , what does i need to change in .config or dts ?
<miasma>
tkaiser: if only there was a way to share the disk with three usb-sata bridges :) would increase the bandwidth nicely :)
<KotCzarny>
miasma: arent you thinking around the problem from the wrong direction?
<miasma>
:)
<KotCzarny>
solution is: sata3/usb3 capable board
<tkaiser>
KotCzarny: Regarding IOPS: not possible to produce reasonable numbers since depends largely on device tested and also filesystem. So I would have to test with an old kernel and my setup now again and that's not worth the efforts especially given that no one is interested in these numbers (most people still have no idea that random IO on those SBC is more important then that they're after)
<miasma>
KotCzarny: even sata 1 would do
<tkaiser>
miasma: Depends on the SoC in question and the use case.
<KotCzarny>
tkaiser: you should create 'tkaiser's standard test setup', hehe, to have comparable numbers
<tkaiser>
miasma: A20/R40 for example are slower when using SATA than USB/UAS when it's about writing.
<miasma>
hm okay
<miasma>
tkaiser: there were some benchmarks in armbian forums. i noticed that many of them were using some synthetic, non-portable benchmarks. wouldn't some real world tests with e.g. vorbis/h264 media work too?
LargePrime has quit [Ping timeout: 268 seconds]
<tkaiser>
miasma: what do you mean? IMO you need to test out the storage individually but when it's about specific use cases (NAS for example) then this also. Since different things happen. Especially with good old A20 you can ran in situations where you are bottlenecked by CPU in a NAS setup you won't notice when you either test network or storage individually.
<miasma>
sorry, this wasn't nas related, just a general question
<tkaiser>
miasma: You were talking about 'non-portable'?
<ssvb>
miasma: spi is the same in a64/h3 as in the older allwinner chips, it's only a matter of enabling it
<ssvb>
miasma: and then we have the generic spi driver quality problems, which apply to every soc
<miasma>
tkaiser: the forums posts mentioned some benchmarks that provide some numbers that aren't comparable with results from other versions or when used with other compiler settings
<miasma>
ssvb: ok
<miasma>
tkaiser: e.g. sysbench
<tkaiser>
miasma: That was sysbench. This thing can only be used to be a lightweight alternative to cpuburn ;) But not to measure performance.
<miasma>
something like the phoronix test suite would be better, but they have tons of silly, useless benchmarks
<miasma>
like compile time benchmarks with different compiler versions, which are not comparable at all
<tkaiser>
miasma: Moronix test suite is even worse for exactly that reason.
<miasma>
but they have some individual tests that might be ok. e.g. c-ray
<miasma>
c-ray is a bit silly too, but it can measure the quality of the compiler's codegen and the SoC's execution pipeline
<miasma>
crypto and media benchmarks are also useful for many users
<ssvb>
miasma: c-ray is a very small piece of code, and its performance heavily depends on whether the compiler does or does not inline one particular function
<ssvb>
miasma: that's why there are wild differences in c-ray results between different compilers and compiler versions
<miasma>
sure
<jonkerj>
miasma/KotCzarny: you are right, clean / deconfig compiles
<jonkerj>
think I may have a breaking option enabled, my bad
<KotCzarny>
jonkerj: i remember that uboot menuconfig is fragile, so its more uboot problem
LargePrime has joined #linux-sunxi
<KotCzarny>
even enabling/disabling same option could lead to non-compiling config
<ssvb>
miasma: basically, phoronix prefers "broken" benchmarks, which tend to provide unstable performance results due to some minor variances :-)
<miasma>
ssvb: but assuming the same gcc/llvm version and same settings, it can show some differences in the execution speed. i'm sure that there are better raytrace examples, but raytracing might demonstrate the execution performance of branching general purpose code quite well in some cases
<ssvb>
miasma: that's how you can get bars of notably different length on charts, and also post articles about major "regressions" and major "improvements" when the performance flips up or down
<miasma>
yea, the benchmark articles are mostly stupid
<miasma>
even worse are the articles about some raid configurations with usb sticks. 5 sticks in raid5. 10 tested filesystem, but the tests don't include fat or exfat for some reason
<KotCzarny>
maybe we should start sunxi-news? ;)
apritzel has joined #linux-sunxi
bonbons has joined #linux-sunxi
netlynx has joined #linux-sunxi
<miasma>
i just mentioned this cause I find it pretty annoying that there are no good benchmarks available online for comparing arm SoC performance. maybe I haven't found them yet. another annoying thing is, people use the legacy kernels for allwinner devices and the energy readings are all silly
<miasma>
e.g. orange pi pc was using 3,3W when idle on one page. my opi pcs only need 800-1000 mW
<scelestic>
when using btrfs is lzo recommended for compression or should i stick with the default? (emmc on orangepi plus)
<ssvb>
miasma: how do you manage to get 3.3W on an idle system with any kernel?
<ssvb>
the legacy kernel is not that bad either
<tkaiser>
ssvb: Use a crappy PSU and measure between wall and PSU
<miasma>
ssvb: dunno, i might find that link again, but i don't have it here now. i think i found it from opi forums
<KotCzarny>
miasma: low energy is for devices that: 1/ are always-on, 2/ do something sporadically
<ssvb>
tkaiser: yeah, that's probably how they measure it
<tkaiser>
scelestic: I would use lzo for daily use.
<tkaiser>
ssvb: We had a user in Armbian forum complaining about high consumption of an A20 device. +6W, turned out to be just a contact problem since after unplugging everything consumption with PSU included was back below 2W
<tkaiser>
miasma: But as usual with such passive benchmarking stuff you have to be careful. There are situations where specific fs performan totally differently.
<scelestic>
miasma: thanks!
<tkaiser>
miasma: btrfs with enabled compression has drawbacks when you deal with large files. And in general you get worse performance when you constantly write files that are smaller than btrfs' default blocksize (4k). And a way higher write amplification
<miasma>
sure. i just think that on low power devices the cpu load is a significant factor
<miasma>
but btrfs doesn't support lz4 now, i guess
<scelestic>
lz4 would have been nice yeah
<Wizzup>
MoeIcenowy: I see... I thought it was NAND somehow...
<miasma>
you have more options with squashfs, but some formats like lzma/xz are way too slow for arm devices atm
<scelestic>
i'm still using my sheevaplug until i have my opi fully running, i think anything is an improvement at this point
<miasma>
scelestic: there's even more potential with these codecs, but sadly they're not open source http://www.radgametools.com/oodle.htm. might take years before we have up to date transparent compression with free file systems
Mr__Anderson has joined #linux-sunxi
<scelestic>
mh, i guess i should have looked up if mainline uboot supported btrfs, it seems it doesnt?
<KotCzarny>
make separate boot partition, problem solved
<scelestic>
yeah guess i'll do that then
leviathanch has joined #linux-sunxi
<silviop>
I try to add mem=1024M to kernel but it no boot.
<silviop>
There is someone with 1Gb A33 Tablet?
<silviop>
and with mem=512M too
fdcx has quit [Ping timeout: 265 seconds]
fdcx has joined #linux-sunxi
fdcx_ has joined #linux-sunxi
<scelestic>
odd, i can create partitions on emmc but i cannot use them, fdisk -l doenst show it however 'print' in fdisk does show them, mkfs refuses to format however
fdcx has quit [Remote host closed the connection]
fdcx_ has quit [Remote host closed the connection]
mzki has joined #linux-sunxi
fdcx has joined #linux-sunxi
fdcx_ has joined #linux-sunxi
fdcx__ has joined #linux-sunxi
<tkaiser>
scelestic: Always worked for me, installed yesterday 4.9 with btrfs on OPI Plus 2E directly to eMMC
<scelestic>
tkaiser: i assume you had to create a ext4 boot partition?
<tkaiser>
scelestic: Yep, in Armbian we use 64 MB in the meantime for this (since /boot lives there)
<scelestic>
odd
<scelestic>
well whatever it was it's now fixed
BenG83 has joined #linux-sunxi
silviop has quit [Ping timeout: 252 seconds]
my123_ is now known as my123
my123 is now known as my123_
<KotCzarny>
scelestic: gpt vs mbr ?
<scelestic>
KotCzarny: not sure to be honest, i just got the warning it couldnt re-read the partition table and even after a clean boot, not it just worked without a hitch
<KotCzarny>
might be that some fs was mouted at the time
<Seppoz>
so the internal temp sensor of the A20 is completeley crap
<Seppoz>
for anyone who ever tires to use it dont
<Seppoz>
at 70°C it shows like 35
<Seppoz>
at 40 it shown like 25
apritzel has quit [Ping timeout: 268 seconds]
<Wizzup>
MoeIcenowy: what is fSD? eMMC?
<scelestic>
KotCzarny: yeah i thought so as well but it did the same after a reboot, well it's fixed now
<scelestic>
tkaiser: i noticed a few times it scaled back to about 1.2ghz in cpu freq, i guess it was throttling?
<KotCzarny>
i think its for legacy kernel
<freemangordon>
KotCzarny: any idea which driver shall I use?
jstein_ is now known as jstein
<tkaiser>
scelestic: Yes, but I'm surprised since cpufreq should go down a little bit more. But maybe I remember wrong and those values are ok. The large Xunlong boards show pretty good heat dissipation
<scelestic>
tkaiser: cool, the room is fairly cool anyway so maybe that helps as well, ambient temperature is about 17C
<jmcneill>
on the topic of H3, has anybody seen SD card error issues on it?
<freemangordon>
yes, and don;t understand what I am supposed to do :(
<freemangordon>
"First get a working display driver. "
<jmcneill>
I've got some cards here that are otherwise reliable on other SoCs, that have fairly repeatable errors on H3 boards (both NanoPi NEO and OrangePi Plus 2E)
<tkaiser>
scelestic: True, that helps also. My tests were all done with at least 23¡C ambient temp.
<ssvb>
freemangordon: what are you trying to use the mali driver for?
<ssvb>
freemangordon: X11 or framebuffer?
<freemangordon>
ssvb: X11
<ssvb>
ok
<ssvb>
also with X11 you have a choice between ump and dmabuf
<tkaiser>
jmcneill: Not that I know of, only the 'usual' problems when cards are crap.
<tkaiser>
jmcneill: Different brands or all the same?
<freemangordon>
ssvb: but there was a note that 3d accell does not work without ump
<freemangordon>
and I need 3d accell
<KotCzarny>
(and mainline)
<freemangordon>
:nod:
<freemangordon>
i can;t find any clear howto :(
<ssvb>
freemangordon: these mali instructions were written maybe 3 years ago
<ssvb>
we just need to update them and that's all
<freemangordon>
ssvb: could be, however in the meantime I have no idea how to build the driver :)
<ssvb>
you need either ump or dmabuf
<ssvb>
ump is an old thing, dmabuf is new
<freemangordon>
ah, I see. You say that I will have 3d even without ump?
<ssvb>
yes
<jmcneill>
tkaiser: I've got one batch of the same brand that fail almot the same way, and one from a different brand that fails the same way (but it's a bit harder to trigger)
<freemangordon>
ssvb: thanks, going to try
<tkaiser>
jmcneill: Which brand exactly?
<jmcneill>
Lexar for the easy to fail ones, SanDisk for the less easy to fail one
<ssvb>
freemangordon: do you have a usable display driver already?
<ssvb>
because mali is just a memory to memory accelerator
<jmcneill>
OTOH I have a different, slower Lexar card that doesn't fail at all
<freemangordon>
ssvb: I guess so, I have xfce running
<freemangordon>
i tried to build xf86-whateveritwas-turbo driver, but it said there is no /dev/mali
jstein_ has joined #linux-sunxi
jstein is now known as Guest61332
<ssvb>
if you have no /dev/mali, then your mali kernel driver has problems
<ssvb>
or you don't have it at all :-)
<freemangordon>
ssvb: I have no malo.ko at all
jstein____ has joined #linux-sunxi
<freemangordon>
I don;t have it, thus I am trying to build it :)
<KotCzarny>
fmg: using fbturbo even without mali speeds things a lot
<tkaiser>
jmcneill: Hmm... never used Lexar and only limited SanDisk experiences (they all work flawlessly) since I prefer Samsung EVO due to superiour random IO performance in the meantime.
<freemangordon>
KotCzarny: but how to build it, I have no idea what parameters to give to make
<freemangordon>
thus I tried in auto mode
Guest61332 has quit [Read error: Connection reset by peer]
<ssvb>
freemangordon: I would guess that somebody already has the mali kernel driver adapted to allwinner devices
<KotCzarny>
hmm, maybe you are missing some build dependency, it was some time since i last built it
<ssvb>
freemangordon: you can try to search for some kernel branches
<ssvb>
freemangordon: a somewhat tricky thing is that the mali kernel driver from the arm website still needs some glue layer to take care of things such as the clocks setup
<jmcneill>
tkaiser: Yeah I normally stick to SanDisk (although I have a failing one here) or Samsung. Lexar were on sale and I needed a handful at the time.
jstein_ has quit [Ping timeout: 260 seconds]
<ssvb>
freemangordon: but nowadays you can probably describe this stuff in the dts
<jmcneill>
tkaiser: trying to figure out if it's a bug in the freebsd driver or a known issue with the soc
<jmcneill>
tkaiser: I found and fixed a bunch of bugs yesterday but none of them seem to be the cause of the issues I'm seeing :/
jstein has joined #linux-sunxi
<freemangordon>
hmm, doing just make seems to run
<ssvb>
freemangordon: still I guess that some people have already done this, so it's best if you take somebody's else work rather than doing this adaptation yourself from scratch
jstein____ has quit [Ping timeout: 246 seconds]
<freemangordon>
ssvb: yes, MoeIcenowy said she has it running
<tkaiser>
jmcneill: Did you a test for counterfeit cards? F3 or H2testw already used?
<MoeIcenowy>
wens: I seem to got a A33 tablet with DSI panel...
KB3VGW has joined #linux-sunxi
<MoeIcenowy>
it seems that this is not supported, even in U-Boot?
<KB3VGW>
wigyori, you around
iamfrankenstein1 has joined #linux-sunxi
iamfrankenstein has quit [Ping timeout: 250 seconds]
iamfrankenstein1 is now known as iamfrankenstein
<zoobab>
@tkaiser is there any mainline version for the orangepi zero?
<tkaiser>
zoobab: sure, MoeIcenowy and dgp have repos with (almost) full support
BuddyZhang1 has quit [Ping timeout: 260 seconds]
<zoobab>
when I plug it on my laptop in OTG, it shows up as:
<zoobab>
$ lsusb
<zoobab>
Bus 002 Device 009: ID 1f3a:efe8 Onda (unverified) V972 tablet in flashing mode
<MoeIcenowy>
zoobab: this is the FEL mode
<zoobab>
does Armbian 3.4 gives a serial console in gadget mode like the CHIP?
<tkaiser>
zoobab: We only enabled this on the 2 boards without Ethernet at all: OPi Lite and NanoPi Air. Since with legacy kernel sometime unloading of USB gadget drivers fails
<tkaiser>
zoobab: But it's just adding g_serial to /etc/modules and activate another getty
<zoobab>
:-)
<tkaiser>
zoobab: But accessing through Ethernet will work also of course (sshd)
silviop has joined #linux-sunxi
<willmore>
tkaiser, WRT the voltage regulator issues on the 'official' Opi zero images: Thanks, that clears up my confusion.
<tkaiser>
willmore: This more primitive voltage regulator is used on all H3 boards except those larger Oranges that use SY8106. So also all NanoPi and even Banana Pi M2+. But on the latter they somehow forgot to enable switching so the Banana runs constantly at 1.3V and overheats more easily
<KotCzarny>
forgot?
<KotCzarny>
or got lazy to save a fraction of a cent or something
<MoeIcenowy>
tkaiser: just for curious, do you have all the boards supported by Armbian?
<tkaiser>
KotCzarny: Nope, it's the same voltage regulator. Lion Wang told us back then they would have chosen another since 'logistic' but that's not the case. Maybe they f*cked up just pin mapping. No one knows, no one cares.
<tkaiser>
MoeIcenowy: Sure, even those that not exist (yet) ;)
cnxsoft has joined #linux-sunxi
<tkaiser>
MoeIcenowy: http://www.armbian.com/download/ -- only challenge is Wi-Fi/BT since everything else is the same anyway (more or less).
<tkaiser>
MoeIcenowy: Based on yet available fex file FriendlyARM's new NanoPi M1 Plus uses exactly the same settings like Orange Pi Plus 2E and the only difference is also the primitive voltage regulation and one led has another color. And no idea about size of DRAM/eMMC.
jernej has quit [Ping timeout: 250 seconds]
<Seppoz>
is there a way to intercept the console output via uart0 from userspace?
<Seppoz>
or redirect console
<KotCzarny>
console=... in kernel cmdline
<Seppoz>
i want to change it during runtime
cnxsoft has quit [Quit: cnxsoft]
<KotCzarny>
maybe use netconsole ?
<KotCzarny>
you can always grab it on the same machine too
<KotCzarny>
might be related too, but depends on init system you are using
<Seppoz>
first one seems not like what i want
<zoobab>
I don't think you can change it after boot
<Seppoz>
im via ssh on the machine and id like to see the console
<Seppoz>
ok
<KotCzarny>
you can have multiple consoles
<silviop>
My A33 1Gb tablet report only 512Mb in u-boot and kernel , does i need to change sometingh in arch/arm/mach-sunxi/dram_sun8i_a33.c ?
<KotCzarny>
and with netconsole you can justrun socat and connect to it locally
<Seppoz>
ok let me look into that then
<willmore>
tkaiser, what's a 'larger Orange'? They're all pretty small, IMHO. Well less than the Pine64--which I would call a 'large' SBC. ;) But I actually know what you mean. KotCzarny's link to the comparison chart helped.
<KotCzarny>
willmore:feelfree to update the table if you find incosistences or new devices (tedious task but wiki isnt setup to autocreate such tables)
<willmore>
KotCzarny, If I do, I will.
<freemangordon>
MoeIcenowy: well, still no /dev/dri here :(
<freemangordon>
silviop: my 1GB tablet reports 512 as well, but it is the same in /proc/meminfo under Android :)
<KotCzarny>
maybe there are really 512MB ? ;)
<KotCzarny>
*they
<freemangordon>
yeah ;)
<MoeIcenowy>
freemangordon: will you mind to give me your dmesg?
<Guest97474>
I'm trying to get 8 audio channels out of the HDMI. I have checked many SBCs and non of them shows 8 audio channels discovered by ALSA. The Linux SW I'm using is working with ALSA and JACK.
mzki has joined #linux-sunxi
<MoeIcenowy>
freemangordon: no, sun4i-drm can kick up simplefb
<MoeIcenowy>
but I cannot still know why your sun4i-drm is not started
<freemangordon>
MoeIcenowy: do I need some out-of-tree module?
<MoeIcenowy>
yes for gpu, but no for drm
<freemangordon>
so, with your a33 branch drm should work?
<MoeIcenowy>
yes
<freemangordon>
hmm
<MoeIcenowy>
could you try to build sun4i-drm as modules?
<freemangordon>
ok
whaf has quit [Remote host closed the connection]
whaf has joined #linux-sunxi
<silviop>
http://linux-sunxi.org/ET_Q8_V2.0 see the photo two big 512Mb chips!! on android meminfo report 900Mb (is normal to report less than full dram)
<freemangordon>
silviop: well, then maybe you should tweak the board dts
scream has quit [Remote host closed the connection]
Pepes has joined #linux-sunxi
Pepes is now known as Pepe
<silviop>
in sun8i-A33 i found only
<silviop>
memory {
<silviop>
reg = <0x40000000 0x80000000>;
<silviop>
};
<silviop>
the region is 2Gb size
<silviop>
i found detection routine in arch/arm/mach-sunxi/dram_sun8i_a33.c
<silviop>
where there is a built in
<silviop>
struct dram_para para = {
<silviop>
.cs1 = 0,
<silviop>
.bank = 1,
<silviop>
.rank = 1,
<silviop>
.rows = 15,
<silviop>
.bus_width = 16,
<silviop>
.page_size = 2048,
<silviop>
};
<freemangordon>
MoeIcenowy: now I have "[drm] Initialized" in dmesg
<KotCzarny>
it wasthere too before
<freemangordon>
but still no /dev/drm/card0
<freemangordon>
in the meanwhile sub4i modules are loaded
<KotCzarny>
but it lookslike its just genericstart of the code, shoyuldnt it report what was enabled etc?
<Wizzup>
freemangordon: it should be /dev/dri not /dev/drm
<Wizzup>
may be a typo but fyi
<freemangordon>
yeah, a typo
<freemangordon>
the only thing with d* in /dev/ is disk :)
<Wizzup>
:)
Guest97474 has quit [Ping timeout: 260 seconds]
<freemangordon>
silviop: maybe u-boot adjusts the size of RAM
<MoeIcenowy>
yes u-boot adjusts it
<KotCzarny>
MoeIcenowy: can you show your dmesg for comparison?
<freemangordon>
yeah, might help
<MoeIcenowy>
temporarily cannot...
<MoeIcenowy>
I'm hacking on my another A33 tablet
<MoeIcenowy>
which is still even stuck on U-Boot
<jelle>
I should investigate how well my two A33 tablets work
jernej has joined #linux-sunxi
<MoeIcenowy>
usually A33 tablets can work well...
<MoeIcenowy>
but this time I met a difficult tablet :-(
<freemangordon>
MoeIcenowy: does pipelines need to be explicitly enabled?
<MoeIcenowy>
freemangordon: see my boot.cmd
<jelle>
MoeIcenowy: sure they worked some time ago :)
<jelle>
MoeIcenowy: just checking which parts are missing + should reallly update the wiki
<willmore>
Has anyone used the wifi on the Orange Pi zero with the legacy kernel? The latency seems very high. Is it sleeping agressively or something?
<freemangordon>
MoeIcenowy: there is nothing about the pipelines, just for the display engine
<tkaiser>
willmore: dpg thinks it might be an interrupt problem. Please see Armbian H2/H3 forum and there the 'OPi Zero went to market' thread (last page)
<tkaiser>
willmore: And also 'iwconfig wlan0 power off' might be an idea
<freemangordon>
ok, I will enable DEBUG in the modules and retry, to see what happens
<freemangordon>
MoeIcenowy: USB kbd does not work with your config
<freemangordon>
I guess I need to tweak it
<MoeIcenowy>
freemangordon: have you copied the modules :-)
<jmcneill>
tkaiser: f3 checks out on the Lexar card
<freemangordon>
MoeIcenowy: sure, but there is noone to load them :)
fl_0 has quit [Quit: STRG + Q]
<KotCzarny>
/etc/rc.local?
<MoeIcenowy>
freemangordon: have you depmod'ed?
<freemangordon>
yes
<MoeIcenowy>
oh strange
<freemangordon>
but nothing in /etc/rc.local to load them
<freemangordon>
I made gadget drivers build-int and ow it works
<KotCzarny>
i meant, load them via that file
<tkaiser>
jmcneill: What exactly?
<jmcneill>
Data OK: 14.92 GB (31293184 sectors)
<jmcneill>
Data LOST: 0.00 Byte (0 sectors)
<tkaiser>
jmcneill: And performance numbers ok?
<jmcneill>
hard to tell i'm using a crappy usb card reader
<jmcneill>
Average reading speed: 17.69 MB/s
<jmcneill>
about what i would expect from it
<jmcneill>
(doesn't support UHS-I)
<tkaiser>
jmcneill: Ah, ok. Well I had a card a while back that showed no problems in crappy SD card reader but failed always when trying to be written from my MacBook (assumption: too fast --> then problems occured)
<jmcneill>
I've got 5 cards (Lexar and SanDisk) that show the problem across 3 boards (OPi+ 2E and NanoPi NEO)
<jmcneill>
Is there an armbian image for nanopi neo? I'd like to see if I can reproduce the problem there
<freemangordon>
MoeIcenowy: hmm, now it works
<MoeIcenowy>
drm?
<freemangordon>
yes
<MoeIcenowy>
is the display correct on the board?
<MoeIcenowy>
s/board/panel
<freemangordon>
no idea, will start lxde to check
KB3VGW has quit [Ping timeout: 265 seconds]
<freemangordon>
MoeIcenowy: well, don;t know if it is correct, but lxde works
<MoeIcenowy>
oh it seems that no one have implemented MIPI DSI support for A33?
Gerwin_J has quit [Quit: Gerwin_J]
Keziolio has quit [Ping timeout: 256 seconds]
<freemangordon>
MoeIcenowy: ok, do you have time and will now to help me with enabling 3d?
<MoeIcenowy>
freemangordon: try to build github.com/mripard/sunxi-mali
<jmcneill>
tkaiser: seems to be stuck in a reboot loop
<jmcneill>
tkaiser: I login, start to create an account, watchdog fires
<tkaiser>
jmcneill: There's one potential problem with the 'legacy' images: They want to reboot once (automatic rootfs resize)
<jmcneill>
it's happened more than once..
<jmcneill>
seems to have settled at least
<tkaiser>
jmcneill: IIRC I read this today in the forum that we might have a race condition here. At least when the first unprivileged user account can not be created a reboot is issued (we should change that)
<jmcneill>
ah that was it then
<jmcneill>
the resize reboot happened while i was creating the first
<jmcneill>
tried to create it again and it already existed
<MoeIcenowy>
however, it seems that modern gtk3 programs runs better on armsoc...
<MoeIcenowy>
ssvb: should I use any -O or -mcpu options?
<ssvb>
there should be nothing that runs faster on armsoc, of course except for the 3d stuff (which just works as opposed to not being accelerated at all)
<ssvb>
MoeIcenowy: for tinymembench just use the default build options, all the relevant stuff is implemented in assembly anyway
<MoeIcenowy>
ssvb: will the usage of them affect the score?
<MoeIcenowy>
if it will, I will rerun the bench
<ssvb>
only the C implementations
<ssvb>
yes, you can try it both with your options and using the defaults just to see if there is any difference
<ssvb>
BTW, what is the DRAM clock speed on your tablet?
Pepe has joined #linux-sunxi
<MoeIcenowy>
ssvb: failsafe, 432MHz
<ssvb>
I see
<MoeIcenowy>
in q8_a33_tablet_800x480_defconfig
<MoeIcenowy>
oh stock script.bin uses 408...
gzamboni has quit [Read error: No route to host]
terra854 has quit [Quit: Connection closed for inactivity]
<freemangordon>
YAY, hildon-desktop works with 3d acceleration :)
<freemangordon>
download it, for some reason FF wants quicktime plugin to play it
<MoeIcenowy>
exciting!
<freemangordon>
it is still rough around the edges, but I guess it will be fixed
pg12 has quit [Ping timeout: 258 seconds]
pg12 has joined #linux-sunxi
<jmcneill>
tkaiser: good news, no issues with the card in armbian.. now i get go to back to bug hunting :/
vagrantc has joined #linux-sunxi
vagrantc has quit [Changing host]
vagrantc has joined #linux-sunxi
<tkaiser>
jmcneill: Good luck!
<freemangordon>
MoeIcenowy: any hint on how to make TS act like a mouse?
<KotCzarny>
fmg: you da man!
<freemangordon>
:)
<KotCzarny>
now ssvb can't say mali is useless ;)
<MoeIcenowy>
freemangordon: how do you mean "act like a mouse"?
<MoeIcenowy>
jernej: I successfully made simplefb run with your h3_hdmi branch.
<jmcneill>
tkaiser: thanks for your help!
<ssvb>
KotCzarny: it's not useless, but rather has some niche uses
<freemangordon>
MoeIcenowy: well, there is something weird, if I have real mouse attached, I can click task switcher etc. but with TS I can only close a program, the other touches doesn't get registered by hildon-desktop
<freemangordon>
I bet it is some bug in h-d, but still
<MoeIcenowy>
I don't know...
<freemangordon>
ok
<ssvb>
MoeIcenowy: do you have a tinymembench log to share?
<MoeIcenowy>
a program can always diff touches with clicks with XInput protocol
<freemangordon>
ssvb: I'd rather not, I just followed what MoeIcenowy said, and it seems she knows the things
<tkaiser>
jmcneill: BTW: Did you hear about the Pinebook? Should be rather easy to support given the state of FreeBSD's A64 support now. But HDMI/audio is missing, right?
<MoeIcenowy>
freemangordon, ssvb: who really knows things is mripard ;-)
<ssvb>
freemangordon: well, you have solved your problem for now, but it's give and take
<ssvb>
freemangordon: the next person will have much less problems when getting mali/armsoc up and running if you update the wiki
<freemangordon>
ssvb: I know what you mean
<KotCzarny>
hmm, obligatory, run glxgears? ;)
<ssvb>
do you mean es2gears?
<MoeIcenowy>
KotCzarny: in the world of Mali, glxgears is useless, you need es2gears_x11
<KotCzarny>
in any case if you didnt get it, i was joking
<KotCzarny>
:)
<freemangordon>
ssvb: But I'd rather do it when I am about to prepare sdcard image, so to be sure I am writing the correct stuff in the wiki
<MoeIcenowy>
I can run a real glxgears with Mali ;-)
<freemangordon>
me too, me too
<freemangordon>
both (es2_gears and glxgears) give ~60fps
<KotCzarny>
vsync ?
<freemangordon>
I guess disabled
<freemangordon>
as there is a bit of tearing
<MoeIcenowy>
https://github.com/ptitSeb/gl4es this is a GL 1.5 implementation on GLES 1.1, which is not only rich enough to run glxgears but also can run some games ;-)
<freemangordon>
when scrolling between desktops in h-d
<KotCzarny>
glshim fork?
<MoeIcenowy>
yes, but has more function
<freemangordon>
glshim is still supported?
<buZz>
its still used
<KotCzarny>
looks like someone took the matter in own hands and just forked
<freemangordon>
anyway, me gtg for a while, bbl
<MoeIcenowy>
glshim seems to be now even unable to run a glxgears :-(
<KotCzarny>
there is so much fragmented work done, but somehow it gets pieced together
whaf has quit [Ping timeout: 246 seconds]
<MoeIcenowy>
freemangordon: you also ensured a compatible for 1024x600 Q8 panels :-)
<MoeIcenowy>
at least for your Q8's panel ;-)
silviop has quit [Ping timeout: 252 seconds]
Da_Coynul has joined #linux-sunxi
<jmcneill>
tkaiser: yeah apart from hdmi/audio, a64 works well. those things are on my list.. trying to sort out h3 first
<MoeIcenowy>
jmcneill: freebsd?
whaf has joined #linux-sunxi
tkaiser has left #linux-sunxi [#linux-sunxi]
tkaiser has joined #linux-sunxi
<MoeIcenowy>
jernej: how are you H3 HDMI U-Boot patch cleaning progesss?
<MoeIcenowy>
s/sss/ss
ericxdu_ has quit [Ping timeout: 268 seconds]
ericxdu has joined #linux-sunxi
Da_Coynul has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<jernej>
MoeIcenowy: I think I did most of the cleaning, except defining HDMI and DE registers and registers bits
<MoeIcenowy>
jernej: do you have a WIP branch for the cleaning?
<MoeIcenowy>
I tested this branch on both H3 and A33 (sun8iw7,w5)
<jelle>
MoeIcenowy: are you working on h3 hdmi fb support for u-boot?
jstein has quit [Remote host closed the connection]
<MoeIcenowy>
jelle: jernej is working on it
<jelle>
oh nice
<MoeIcenowy>
i'm only trying to help him to clean it
<MoeIcenowy>
as I have sun8i device with legacy de1 ;-)
Mr__Anderson has quit [Ping timeout: 256 seconds]
Mr__Anderson has joined #linux-sunxi
<jelle>
I only got as far as getting HPD to work iirc
Mr__Anderson has quit [Client Quit]
Mr__Anderson has joined #linux-sunxi
scream has joined #linux-sunxi
The_Loko has quit [Quit: Leaving]
whaf has quit [Remote host closed the connection]
tkaiser has joined #linux-sunxi
whaf has joined #linux-sunxi
reinforce has quit [Quit: Leaving.]
<MoeIcenowy>
jernej: trying to run H3 code on A64 (surely) fails
Da_Coynul has joined #linux-sunxi
yann-kaelig has quit [Quit: Leaving]
Da_Coynul has quit [Client Quit]
whaf has quit [Quit: Leaving.]
<jmcneill>
tkaiser: do you recall the purpose of the gpio set lines in boot.cmd for nanopi neo?
lemonzest has quit [Quit: Leaving]
whaf has joined #linux-sunxi
<tkaiser>
jmcneill: Most probably not needed. I'll check.
<jmcneill>
thanks
<jmcneill>
(running out of ideas here, arg)
<tkaiser>
jmcneill: I think that's the result that we share one config file for all sun8i devices currently. PG11 is GPIO to turn on the USB-to-SATA bridge on Orange Pi Plus and Plus 2 IIRC (to have rootfs on a disk connected there).
<tkaiser>
jmcneill: And PL10 is green led. The latter isn't needed any more (jernej told me maybe 5 times already)
<tkaiser>
So on the NEO both useless anyway
<jernej>
MoeIcenowy: I have few minutes now. For A64, you have to change at least address of TCON registers
<jernej>
tkaiser: Vanilla U-Boot (unchanged) doesn't turn on any LEDs. I made a patch which turns on LEDs based on DT
<jernej>
but I'm not sure if there is interest to merge such patch into mainline U-Boot
<tkaiser>
jernej: Armbian uses your patch: h3-enable-power-led.patch so setting PL10 to on in boot script is useless (but doesn't cause any harm)
<tkaiser>
jernej: Nope, I was wrong, we use an earlier patch from you not using DT but hardcoded PL10
<jernej>
I thought so :)
<Wizzup>
MoeIcenowy: what did you mean about my tablet using fSD instead of NAND?
<tkaiser>
Maybe that patch was from Mikhail... don't know. Do you include your DT variant in OpenELEC?
<tkaiser>
jmcneill: Hmm... might work now with initrd support but in the past there was AFAIK no way to put rootfs behind GL830 (which is a feature though IMO ;) )
<jmcneill>
ahh
<jmcneill>
well wait, u-boot loads the dts doesn't it?
<jmcneill>
oh you are loading dts from sata
<tkaiser>
Nope, but I've to admit that I'm really clueless regarding u-boot
<jmcneill>
ditto
<jmcneill>
i'm giving up on h3 mmc today
<jmcneill>
on the bright side i've fixed a bunch of issues in the driver
<jmcneill>
(just not the issue i'm trying to solve)
bugzc has joined #linux-sunxi
<Wizzup>
MoeIcenowy: I thought I saw NAND on the device
Mr__Anderson has quit [Quit: Leaving.]
bonbons has quit [Quit: Leaving]
paulk-collins has quit [Quit: Leaving]
IgorPec has quit [Ping timeout: 256 seconds]
Andy-D has joined #linux-sunxi
Andy-D has quit [Remote host closed the connection]
sunxi_fan1 has joined #linux-sunxi
Ntemis has joined #linux-sunxi
Andy-D has joined #linux-sunxi
scream has quit [Remote host closed the connection]