<arokux2>
wens: hm.. how can u-boot patch make kernel see the phy?
<arokux2>
wens: it was the problem in u-boot, phy was visible, but gmac won't work due to dma
<steev>
arokux2: it's not really a concern, just saying keep in mind that while that patch is good (seriously, i like it) - it can be overridden, especially if someone is using a kernel that a distro created rather than compiled their own
<arokux2>
steev: alright
<arokux2>
mnemoc: please apply on stage: [PATCH] defconfigs: Let kernel name contains the hash of the tree
arokux2 has quit [Remote host closed the connection]
<wens>
arokux: not sure, either u-boot correctly inits gmac, then kernel just works or something
n01 has joined #linux-sunxi
fredy has quit [Excess Flood]
<arokux>
wens: weirdness
fredy has joined #linux-sunxi
<arokux>
wens: are you still passing clk_ignore_unused? you can try to remove this to see what happens, but this is just random advice
<arokux>
wens: but does gmac work completely in u-boot or only phy? you'd like to test ping/tftp to really know.
<wens>
arokux: it loads kernel using tftp and boots :)
<arokux>
wens: nice :)
Quarx has joined #linux-sunxi
<arokux>
wens: the next stopper for jemk was that data actually comes from phy but gmac won't react on it. I don't know how jemk detected this.
n01 has quit [Ping timeout: 272 seconds]
<arokux>
( wens: the last was dma problem )
<wens>
arokux: it seems data is not being sent out, don't see any DHCP packets from my board with tcpdump
tzafrir has joined #linux-sunxi
HeHoPMaJIeH has joined #linux-sunxi
cubear has joined #linux-sunxi
iamfrankenstein has joined #linux-sunxi
iamfrankenstein has quit [Excess Flood]
maz_ has joined #linux-sunxi
rellla has joined #linux-sunxi
wolfy has joined #linux-sunxi
maz_ has quit [Ping timeout: 244 seconds]
<cubear>
Is cubieboard 2 powerful enough for OpenCV video processing with a decent framerate? I.e. face recognition?
<cubear>
I read on google some people were having framerate problems
<Montjoie>
arokux, you can delete the page
<oliv3r>
hno: yeah but i had forgotten I ddid that, so was like 'how's that even possible' and i rm -r build/*u-boot and still got error ...
eebrah_ has quit [Read error: Connection reset by peer]
panda84kde has joined #linux-sunxi
rellla2 has joined #linux-sunxi
<oliv3r>
Turl: arokux for YEARS i used the 'detailed' list, especially in my win98 days. I even switched to detailed view on my first years of linux. Once I did my first Ubuntu install I noticed the default was icon mode, and at first I switched it back, I got to use it a few times and actually quite like it. I hardly need to sort stuff, I switch to detailed sorted mode if I have to find stuff, but I now love icon mode
<oliv3r>
wens: awesome work; if gmac works for both modes in u-boot, we're 90% there. just gotta talk ot the deisngware u-boot maintaner about our suggested changes and merge them
<libv>
oliv3r: good point, will add that later as well
bamvor has quit [Quit: Leaving]
bamvor has joined #linux-sunxi
n01 has quit [Ping timeout: 260 seconds]
n01 has joined #linux-sunxi
jemk has joined #linux-sunxi
_massi_ has joined #linux-sunxi
jemk_ has joined #linux-sunxi
jemk has quit [Write error: Broken pipe]
jemk_ is now known as jemk
wolfy has quit [Remote host closed the connection]
wolfy has joined #linux-sunxi
hramrach has joined #linux-sunxi
Black_Horseman has quit [Remote host closed the connection]
vicenteH has quit [Ping timeout: 246 seconds]
<gzamboni>
i cant understand, i'm trying to use libspi, i do activate the spi0 in my fex for my cubieboard but i dont have the SPI0-CLK in the output when i open the device with a c app. do i have to set a clock setting somewhere else ?
<Montjoie>
yes, see tegra_aes_get_random() in drivers/crypto/tegra-aes.c
<Montjoie>
it seems simple
<wens>
arokux, jemk: enables gmac on cb2 for u-boot.
<arokux>
can anybody please try to register at www.cubieforums.com? i'm not getting their activation e-mail, and admin says my provider is rejecting the mails. but I have tried two mail address already.
HeHoPMaJIeH has quit [Ping timeout: 268 seconds]
FR^2 has joined #linux-sunxi
<arokux>
wens: nice. sorry I cannot test it, ct here.
<oliv3r>
mnemoc: i'm not sure it can hurt, can it? i mean it's the version string only; people paste their dmesg, we know immediatly which kernel they use. advanced users that mangle it; will know this and probably say so
<arokux>
libv: any reports on that?
<oliv3r>
even allwinner said they burned their hands on powerVR and don't want to use it anymore then they have to (they have to since they payed money for it, and need to sell some volume because of ROI)
<arokux>
oliv3r: really?
<arokux>
oliv3r: and how did they "burn hands" on it? just interesting, they get blobs for it, what else do they need?
<binaryferret>
Has anyone had any luck with getting an ektf2k driver for 3.4 linux-sunxi working?
<mnemoc>
arokux: documentation is one thing, sure. but I mean something broader
<arokux>
mnemoc: you want a *hug* from Eva? :)
<mnemoc>
arokux: touching details about official linux support, community support, and mali vs pvr
<mnemoc>
arokux: basically yes
prasannapete has quit [Quit: prasannapete]
<arokux>
mnemoc: you know, money is talking there. unless they get lots of Linux customers they won't do anything, I think
<arokux>
mnemoc: but do they get those customers now?
<mnemoc>
android is linux, tizen is linux, xmbc is linux
<mnemoc>
servers are linux
<arokux>
mnemoc: who should really care about us are all the companies which produce SoMs and boards
<mnemoc>
and those are AW customers
<arokux>
mnemoc: but I get the feeling that they are happy there is free (as in beer) code for them at sunxi.org.
<mnemoc>
that's why the letter should be directed to AW and not to SoM/board manufacturers
<arokux>
mnemoc: they actually get docs from allwinner, but they do nothing to share it with us. the reason is, they regard this as there winning bit against the others. like they'll be able to release some feature earlier.
<mnemoc>
as they can increase their marketshare in that world just by realizing that android is only ONE frontend of Linux
<arokux>
ah, you want to try to open their eyes broader, I see.
<arokux>
consider Cubietech, they were holding the driver for the GMAC back, till ct was released.
<arokux>
there are other examples I know of.
<mnemoc>
but after that they released it
<arokux>
yes, not bad, true.
<mnemoc>
I don't mean exposing everything before the product is done. I mean doing their secrets aware of this broader market share they have
<mnemoc>
and then, once the product is ready, release everything they can to hit the market strongerly
<arokux>
mnemoc: the secret was SoC doc, not their awesome feature
<mnemoc>
they did voluntarely share the latest user manuals, they weren't leaked
<mnemoc>
a10's was leaked
<arokux>
mnemoc: there is no doc on EMAC
<arokux>
GMAC*
<mnemoc>
even if those documents are poor they did show intention
<mnemoc>
arokux: because they simply don't own the right
<mnemoc>
EMAC/GMAC is not their IP
<arokux>
hmm... I see. but the question is: do they try to ask to publish it?
<mnemoc>
but they can design their future products considering the importance of been able to open a bit more
<mnemoc>
arokux: we can't know as that is private between allwinner and davicom or dw
<oliv3r>
using IP that is allready supported makes things a lot easier, even for them
<arokux>
JohnDoe_71Rus: perform the command manually and show it then.
<arokux>
JohnDoe_71Rus: who knows what is inside of your uEnv.txt etc.
<JohnDoe_71Rus>
but work for me u-boot-hwpack
<JohnDoe_71Rus>
the only difference is in the u-boot
<wens>
arokux: i have my cb2 booted with usb ethernet now, but not sure what to look for about stmmac problems
<wens>
and the system is wierdly sluggish :(
<arokux>
wens: I hope it is not usb...
<arokux>
wens: jemk might know what you could investigate further.
<wens>
arokux: it was fine the first five minutes. it could be stmmac trying to resend stuff or something
<arokux>
I see...
<arokux>
wens: if it is a module, you can rmmod it to see if the sluggischness is caused by stmmac.
<wens>
arokux: i have my led set to flash according to cpu usage, and it's not flashing a lot
<wens>
arokux: good idea. i'll rebuild with stmmac as module.
<arokux>
JohnDoe_71Rus: the problem could be you image size. if the kernel is so fat u-boot can override something in memory. make sure u-boot will not relocate anything
<arokux>
JohnDoe_71Rus: yes, different commits. and find the particular commit when bug appears, but be sure to reproduce the issue first.
<mnemoc>
512MB just like Tsvetan's lime
<mnemoc>
and also 1GHz cortex-a8
<mnemoc>
so it's a more reasonable competitor
<arokux>
mnemoc: but no SATA
<mnemoc>
the pi is a marketing BS
<Tsvetan>
and no native HDMI
<mnemoc>
go lime go
<Tsvetan>
BBB use converter
<arokux>
Tsvetan: is there some thoughts LIME with 2xethernet? :)
<arokux>
Tsvetan: and wlan :)
<arokux>
ok.. it won't be LIME anymore..
<Tsvetan>
arokux A10 have no 2 MACs
<mnemoc>
Tsvetan: that why I said, feel free to add the lime to our frontpage and provide a comparisson table with the "mainstream" competitors
<binaryferret>
I have an A13 Generic whitelabel tablet. I have the Fedora-19 r1 image running on an SD card. Works, able to log in. The issue I'm having is that when I install the 8188eu with modprobe, it's still not detecting any wifi interfaces.
<mnemoc>
binaryferret: you might need to modify the script.bin and pre-enable the corresponding usbc
<binaryferret>
I did a test with no SD card in, and let it boot to android. Used adb shell and then did rmmod 8188eu, and this brough the wireless down on android. This is basically what made me think that this was the driver I required.
<Tsvetan>
mnemoc I will do this during the weekend
<binaryferret>
mnemoc: Thanks mnemoc, I'm still new to all this, and I have got my own build of the kernel running with the same issue, so I thought I'd go back to the image. I have used the script.bin from the actual tab, and not the standard script.bin that came with it.
<arokux>
Tsvetan: have you asked for any support by AW for GMAC, you had problems with it initially.
<mnemoc>
binaryferret: the script.bin describes your board. it's critical to use the right one
<Tsvetan>
arokux this was long time ago, yes we asked and we didnt received
<Tsvetan>
not because they didnt want to help
<Tsvetan>
but we just cant speak chinese and they cant english
<arokux>
gosh
<binaryferret>
I will have a look at how to pre-enable the corresponding usbc, I just assumed that if using the script.bin that was from the tablet itself, then it would work.
<arokux>
Tsvetan: there is software that can help you both!
<mnemoc>
Tsvetan: after eva added me to linked in some days ago I realized she has a degree in english literature, not in marketing or businesses
<mnemoc>
Tsvetan: so she is there because her english :p
<arokux>
binaryferret: just do: usb_used = 1 in [usbc2] and [usbc1] (just to be sure) I never remember which one is used by wlan
<mnemoc>
usbc2 normally
<mnemoc>
but it's about the initial power state, not onlt the _used
<mnemoc>
for "power saving" allwinners turns on/off the usbc when loading/unloading the wifi module
<binaryferret>
In the .bin that I converted to fex from my tab I have usbc0 and usbc1 defined. Both have the key/value of usb_used = 1
<hramrach>
hmm, lime is interesting. cheaper than a10s olinuxino, more expensive than a13 olinuxino, has sata and hdmi but sucky miniature pins
<binaryferret>
Ok cheers. It's suddently hit me, that I bet a document exists somewhere discussing all these various settings.
<mnemoc>
notmart: and usb_wifi_usbc_num = 1 tells you which usbc is connected to the wifi
<hramrach>
I guess many people don't use the pins so won't mind but I read about these pins on the a20 som bending easily so even with just shields it sucks
<binaryferret>
OOoo Thanks mnemoc and arokux
<hramrach>
even worse than what cubieboard has :S
<binaryferret>
Dinner time, then I'll make changes. Thanks again.
<oliv3r>
Tsvetan: lime brings out i2s? VGA/Composite boards etc?
<cubear>
what's the pin spacing on that lime thing? 2.0mm?
<mnemoc>
it seems to expose every pin
<mnemoc>
0.05"
<mnemoc>
so 1.27mm
<cubear>
buttz
<mnemoc>
tiny
<oliv3r>
looks like it's mini-IDE
jemk has quit [Read error: Operation timed out]
<JohnDoe_71Rus>
micro
<cubear>
I'll need to buy new connectors
<hramrach>
2.5" ide is 2mm, right?
<mnemoc>
cubear: Tsvetan said they will sell contectors and a "cape" with 0.1" and lvds
<oliv3r>
hramrach: 1mm?
<Tsvetan>
these connectors are cheap
<Tsvetan>
1.27 mm step
<oliv3r>
US size, bleh
<cubear>
well I was talking about pcboard-to-wire style connectors like the crimp sockets
<Tsvetan>
Imperial ;)
<mnemoc>
Tsvetan: not when distributors charge absurd shipping prices and large minimal order
<oliv3r>
Lime 2.0; gbit PHY, A20 only, second memory chip underneath! or instead of FLASH, more ram and only boot from SD ;)
<cubear>
and you need a special tool for those crimp pins
<hramrach>
2mm are not expensive either. just hard to get
<oliv3r>
cubear: i just use pliers and solder if I have to
<cubear>
I found 2mm ones for cubieboard on Farnell
<oliv3r>
ebay has a lot of connectors etc
<oliv3r>
but trying to find them is hard
<oliv3r>
what's the 2 JST connectoo called cb3 and olimex use for battery connectors?
<hramrach>
yes, it's just that *every* store that sels something like connectors has the 2.54 ones
<oliv3r>
Tsvetan: i think it could be technicalyl possible to detect various memory controler parameters, but we have zero docs, so probably not in my lifetime :)
<Tsvetan>
oliv3r simple write and read back can solve this simply
<Tsvetan>
write 32 bit bus value read and see if your 32 bit are same
<Tsvetan>
if only 16 bits are same then bus is 16 bit
<Tsvetan>
then write and read pattern to different memory addresses to see where RAM start and ends
<Tsvetan>
it should be few lines of code
<Tsvetan>
which to config these two memory parameters
<oliv3r>
we talked about this on the ML a while ago
<Tsvetan>
the speed and timing let stay in FEX
<oliv3r>
u-boot doesn't use fex :p
<Tsvetan>
Android image do it so its possible
<oliv3r>
android does it?!
<oliv3r>
noway
<Tsvetan>
yes
<oliv3r>
noway!
<oliv3r>
show me
<Tsvetan>
we just programmed Android image for OLinuXino (with 32bit bus) to LIME and it runs
<oliv3r>
anyway, we talked about it and concluded not yet, there's so many things that we don't know
<Tsvetan>
without changes
<Tsvetan>
I can log the console messages
<oliv3r>
i think yalk about io-width vs bus-width
<Tsvetan>
and post
<oliv3r>
you talk*
<oliv3r>
one was guessed to be for 'drive strenght'
<oliv3r>
e.g. if you have 2-4 chips you need more 'power' to drive them, and we expected that to be the same
<oliv3r>
same = set by the io/bus width parameters
rz2k has quit [Ping timeout: 246 seconds]
<oliv3r>
i think hno said, the CPU _always_ communicates with the ram at 32 bit, the difference is
<oliv3r>
16 bit = 2 chips, 32 bit is 1 chip
<arokux>
Tsvetan: Lime is $39, A10 or A20?
<Tsvetan>
arokux: A31
<Tsvetan>
hahahaha
<Tsvetan>
there is picture with A10
<arokux>
Tsvetan: yes, ok, I've checked your blog :p here you were saying that both are ready, and saying the price, so I was puzzled.
<Tsvetan>
yesterday when I blogged we had just A10
<mripard>
mnemoc: heard something about what?
<arokux>
Tsvetan: cann't amazon sell your boards with free shipping? :(
<Tsvetan>
today we assembled A20 too and surprisingly it works with 512MB too
<mripard>
hramrach: doesn't the GMAC have its own DMA controller?
<arokux>
Tsvetan: no price for A20-LIME so far?
<Tsvetan>
arokux no
<binaryferret>
Fex
<binaryferret>
Ignore that.
<Tsvetan>
as I didnt know if it will work at all, but seems it works
<hramrach>
mripard: does it? it depends on HAS_DMA in kconfig but not sure if just selecting that would work
<arokux>
Tsvetan: I see. what about amazon?
<Tsvetan>
we have to make some benchmark now to see if it worth to have DualCore A7 on 512MB and how it compare to Single Core A8
<Tsvetan>
arokux only the cheese in the mouse trap is free
<hramrach>
I the mainlining effort page suggests that ethernet uses the main dma controller but that might apply to emac only
<arokux>
Tsvetan: true. :) I was just curios.
<naobsd>
I really have interest to A20-LIME
iamfrankenstein has quit [Quit: Nettalk6 - www.ntalk.de]
<arokux>
Tsvetan: hm.. I've just realized amazon will be only local.
<naobsd>
Tsvetan: please check power consumption too
<naobsd>
under cpu load
<Tsvetan>
naobsd yes, lot of things to check
HeHoPMaJIeH has quit [Remote host closed the connection]
<arokux>
mripard: we were curious if allwinner was releasing/communicating some doc/info to recently. so mnemoc asked if they maybe communicated with you off-the list, so to say. it is interesting if they start to (pure) Linux as a target.
<arokux>
mripard: start to see* ...
<wingrime>
oliv3r: should we have speical ML for maline?
<wingrime>
*err mainline
iamfrankenstein has joined #linux-sunxi
<arokux>
wingrime: it was discussed. mripard is reluctant to have a special list, he insists on sending everything to upstream only without even CCing linux-sunxi. (not that I like it)
<arokux>
wingrime: are you going to do some mainline work? :)
<binaryferret>
ls
<wingrime>
arokux: I have IR driver
<wingrime>
arokux: but still not able send it to ML
<wingrime>
err
<wingrime>
Mainline
<arokux>
wingrime: because of clocks?
<wingrime>
yes
<wingrime>
arokux: I think I will try bring sata, but , same
<wingrime>
Tsvetan: LIME realy 30$ ???
<wingrime>
Tsvetan: thats awesome
<arokux>
wingrime: 39$
<arokux>
wingrime: LIME-A10
<wingrime>
arokux: mass production can beat rPI
* arokux
dreams of LIME mass produced
<Tsvetan>
wingrime order me 100K LIMEs and I will do them for you at $25 ;)
<wingrime>
Tsvetan: are you still soldeded componets by hand
<arokux>
I hate those RPi ppl
<arokux>
they think they have invented everything :((
<wigyori>
hmm
<wigyori>
this lime stuff looks nice
<Tsvetan>
wingrime you greatly overestimate my soldering skills :)))))
<Tsvetan>
I do not solder manually all boards we sell :))))
<wingrime>
Tsvetan: currently, A10 most opensource on market
<wingrime>
Tsvetan: no idea what I will do with 100k ))))))
<arokux>
Tsvetan: I (as programmer) wonder. You had an A10-board already, why producing LIME take so much time, isn't it copy-paste and throwing out features?
<Tsvetan>
wingrime but you will have best price :)))
<wolfy>
wingrime: you preserve one , purchased at $25 and throw away the others, as you do not need them
<arokux>
wingrime: give them as gits to russian children на новый год!
<Tsvetan>
arokux as hardware engineer I also do wonder why it takes you so much time to move to Kernel 3.12 isnt is just copy and paste from kernel 3.0?
<arokux>
(for new year)
<wingrime>
arokux: I have no such amounts of money
<arokux>
Tsvetan: :) I can explain you that since I understand this. but can you explain me the problems of hw devs? :p
<wingrime>
arokux: testing, fixing,etc
<Tsvetan>
I guess same is here ;)
<arokux>
wingrime: do you need to pay customs if ordering at Olimex?
<oliv3r>
wingrime: not sure if splitting the ML is a good idea; libv thinks its a bad idea to split users from devs
<oliv3r>
wingrime: i think ian got sata to work! but i haven't tested it all; a little busy still
<wingrime>
oliv3r: no, I thinkied not about stilt ml's
<wingrime>
oliv3r: simply move mainline messages to other
* arokux
will happily test SATA, but no SATA HDD around
<oliv3r>
wingrime: thunderbird message filter if cc or to = linux@vger -> mainline folder
vicenteH has quit [Read error: Operation timed out]
<wingrime>
oliv3r: I use gmane
<wingrime>
oliv3r: but without any [MAINLINE] thats hard to figure out what is it
<wingrime>
oliv3r: OR maybe we should ask all peolple use [3.4] tag for all non ml
rwmjones has joined #linux-sunxi
* arokux
wishes wingrime has stepped up as he begged mripard to CC linux-sunxi
<wens>
I can test SATA on mainline next week if needed. I have HDDs and SSDs lying around
<mripard>
hramrach: yeah, from what I looked at, gmac has its own dma controller
<arokux>
mnemoc: how should I know? e-mail bamvor he always replies timely.
<mripard>
while emac uses the "global" one
<mnemoc>
arokux: I thouyght you were into that too
<arokux>
mripard: why then it wouldn't work in u-boot without dcaches disabled or it is not related?
* arokux
is still a noob
<mripard>
arokux: yes, allwinner and I are in contact
<mripard>
not a very regular one, but we exchange a few mails
<arokux>
mripard: some nice news?
<mripard>
arokux: and I'm not against cc'ing linux-sunxi, I'm not convinced to aggressively tell anyone that they *must* send it to linux-sunxi.
<wigyori>
arokux: for the gmac stuff, you're talking about what Zalan put together, right?
<mripard>
arokux: not really yet :(
<arokux>
wigyori: 1) gmac in u-boot by jemk, 2) gmac in sunxi-3.4 taken from cubieboard tree (by cubietech). Zalan patches allow gmac to talk to mii too, something wens is also working on.
<mripard>
and it won't work with dcaches precisely because it uses dma, and does something in the cpu's back :)
<arokux>
mripard: I see :)
<wigyori>
arokux: gotcha - for the kernel side stuff, do you know if they're hooked up already? i'm in daily contact with Zalan, so just in case you want to pass something on
<arokux>
wigyori: which kernel?
<arokux>
wigyori: there are too kernel we care of.
<wigyori>
Zalan's focusing on mainline
<arokux>
wigyori: his patches are against -3.4 though.
<arokux>
wigyori: gmac is not yet working in mainline - only partially.
<wigyori>
hrm
<arokux>
wigyori: so only in u-boot for cb2/ct and -3.4 ct and maybe cb2
<wingrime>
mripard: we defenetly need place for sunxi only, mainline patches, so CC is a good thing
* arokux
happy wingrime steps up
<arokux>
:)
<arokux>
~ 15k e-mail in my mailbox...
<wingrime>
arokux: BT and wifi still In way for CT
<mnemoc>
after one week you forgot you wanted to read the article :p
<arokux>
mnemoc: :D it is the same with me!!
<Tsvetan>
A20 is faster and need less power :)
<mnemoc>
Tsvetan: but how much more expensive would it make the lime?
<Tsvetan>
mnemoc you mean with A20?
<mnemoc>
yes
* arokux
is also curious
<Tsvetan>
I don't know I have to convince AW to give me good price
<mnemoc>
but you are already buying A20's
<Tsvetan>
yes, but I can't make low cost LIME with the prices I pay now for A20
<mnemoc>
ok
<arokux>
if allwinner agrees it means they believe in pure Linux
<mnemoc>
an A20 lime would allow you to run xen, linux/android + RTOS
<arokux>
+ gigabit LAN
<arokux>
err, Eth*
<Tsvetan>
and A20 require 30% less power and runs x1.7 times faster than A10 :-)
<mnemoc>
just thinking in the applications of such small board as replacement of MCU + fancy interface
<wigyori>
does the a20 sata support sata port multipliers?
<mnemoc>
no
<Tsvetan>
we just did some measurements LIME with A10 need 1W to run with A20 it needs 0.7W
<mnemoc>
wingrime: you need a multiplier that "fakes" a single drive
<mnemoc>
Tsvetan: nice^2
<wigyori>
mnemoc: okay - i was thinking along the lines of replacing f.e. hp microserver-grade boxes with a nas/microserver based on the lime, or a similar board
<arokux>
wigyori: no, no sata port multipl
<WarheadsSE>
Thoughts: would fbturbo work on non-sunxi?
<WarheadsSE>
ala: XE303 ?
<mnemoc>
wigyori: in this price range, and with A20's GMAC you can make a farm of one-drive-nodes
<mnemoc>
WarheadsSE: iirc it even works awesome on the Rpi
<WarheadsSE>
orly
<mnemoc>
that's why it was renamed from sunxifb to fbturbo
<WarheadsSE>
I need to tpeak to these people about putting shit in the AUR that is ARM specific.
bsdfox\ has quit [Ping timeout: 264 seconds]
<WarheadsSE>
because they REALLY need to stop doing that
<binaryferret>
Still having issues getting wifi drivers working on A13 tab. I'm using a script.bin from the tablet, with the usb_host_init_state = 1 as recommend earlier. I'm also using modprobe 8188eu with options rtw_power_mgnt=0 rtw_enusbss=0. However alas no device.
<arokux>
WarheadsSE: you do not have something to offer them in return.
<arokux>
WarheadsSE: why is it so bad? :(
<WarheadsSE>
arokux: to some degree, you are correct
<Turl>
WarheadsSE: anything I should be aware of? I do use yaourt :p
<binaryferret>
Turl: When doing ifconfig it now displays wlan0
<binaryferret>
Turl: What did that do? replaced the wrong id with the one that actual exists?
<WarheadsSE>
Turl: don't, its bad, dangerous, and full of stupid assumptions
<binaryferret>
well I say 'wrong id' the one that came with the image.
<arokux>
WarheadsSE: I see arch devs keep saying it all the time.. "do not use yaourt", but should I build it by hand? there are potentially myriad of deps!! we are not in ancient distro times anymore.
<Turl>
I'm going to add some extra remarks and send it to the list later today
<hramrach>
and what is required for u-boot to patch in the mac?
<oliv3r>
bootenv is probably clenest
<hramrach>
does it patch it automagically to ethernet0 or do you need to do something?
<oliv3r>
arokux: is there some universal network USB driver in u-boot?
atsampson has quit [*.net *.split]
diego71 has quit [*.net *.split]
mouchon has quit [*.net *.split]
setkeh has quit [*.net *.split]
dwilkins has quit [*.net *.split]
ojn has quit [*.net *.split]
<hramrach>
there is not universal USB network but there are some usbnet drivers
<Turl>
hramrach: automagical, as long as you set the mac on uboot
<Turl>
(ethaddr property iirc)
dwilkins has joined #linux-sunxi
atsampson has joined #linux-sunxi
diego71 has joined #linux-sunxi
mouchon has joined #linux-sunxi
setkeh has joined #linux-sunxi
ojn has joined #linux-sunxi
Quarx has quit [Read error: Connection reset by peer]
tzafrir has joined #linux-sunxi
<hramrach>
yes, seems to work
<hramrach>
thanks
<hramrach>
how do you set module parameters?
<hramrach>
module.parameter=value ?
<hramrach>
actually non-module
<Turl>
parameter=value?
<hramrach>
on commandline
<Turl>
hramrach: I just sent the patch to the lists, you're cc'ed
<Turl>
hramrach: you mean on modprobe?
<hramrach>
I don't modprobe when the driver is built-in
<Turl>
ah
<Turl>
module.parameter=stuff on kernel cmdline I think
<hramrach>
so the doc is bogus :/
<hramrach>
it says mudule=parameter:value,..
<Turl>
it may be that way too
<Turl>
but I've seen i915.stuff=blah on cmdline
<Turl>
(although that might've been a module)
<hramrach>
it should be the same module or not
<Turl>
maybe it supports both syntaxes
<hramrach>
and the module=parameter:value,.. is probably old way
<Turl>
docs on Documentation/ tend to get a bit obsolete
<hramrach>
I guess
<Turl>
SubmittingPatches still recommends using diff
<Turl>
hramrach: what hardware did you test the patch on, btw?
<hramrach>
which patch?
<hramrach>
the mac address?
<Turl>
yep
<hramrach>
I patched it in manually on cb2
<hramrach>
but I do not get the damn debug prints
<hramrach>
there is pr_Debug in the driver which never gets printed
<Turl>
you need to add a special #define to enable those I think
<Turl>
hramrach: try #define DEBUG on the top of the file, before any header
bennhuang has quit [Quit: Leaving]
<Turl>
hramrach: or change it to a pr_err or something
<hramrach>
it's a config option
<hramrach>
CONFIG_DYNAMIC_DEBUG
<Turl>
yeah but then you have to echo something on sysfs to enable them
n01 has quit [Read error: Connection reset by peer]
n01 has joined #linux-sunxi
<hramrach>
or pass an argument
* hramrach
building debug kernel
paulk-collins has joined #linux-sunxi
<hramrach>
ok, now it works
<hramrach>
but ethernet still does not
iamfrankenstein has joined #linux-sunxi
<Turl>
what issue are you having?
Fusing has joined #linux-sunxi
Fusing has quit [Read error: Connection reset by peer]
<hramrach>
I compiled the designware driver into Linux and it fails
<hramrach>
I will try on CT. Should be easier with full-featured phy
<hramrach>
and if it fails the same way at least I can see it's not because of that
tzafrir has quit [Ping timeout: 264 seconds]
<hramrach>
is there the patch for using gmac in u-boot on cb2 somewhere?
<hramrach>
cannot find it
<mnemoc>
hramrach: on the ML
<mnemoc>
posted by jens kuske yesterday
<hramrach>
that's for CT but it's supposed to work with cb2 too
<mnemoc>
i thought the cb2 was using emac anyway
<hramrach>
it can use either it seems
<mnemoc>
but switching to gmac should improve the performance due to the separated dma
iamfrankenstein has quit [Read error: Connection reset by peer]
<pfdm>
mnemoc: does that means you can do gigabit connection with cb2 ?
pirea has joined #linux-sunxi
<pirea>
bakc
<pirea>
back
<mnemoc>
pfdm: no. the phy only allows 10/100
<pfdm>
mnemoc: Ok. anyway I barely do 2Mo/s using scp on my cb2. If I could reach 10 i'd be happy !
<mnemoc>
try switchimg to perfor,ance governor
<oliv3r>
anybdoy tried sunxi KVM yet?
<Turl>
and use a lighter crypto on ssh
<hramrach>
and get the crypto accel working
<Turl>
With arcfour perf is pretty decent
<hramrach>
but we have aes accel ;-)
<oliv3r>
is it confirmed that the designware mac has it's own DMA controller?
Super-noob has quit []
<hramrach>
since the dw driver actually works it's quite likely
shineworld has quit [Quit: Leaving]
<pfdm>
I use a lighter cypher, and the performance optimization of the wiki .
<Turl>
pfdm: do you use a armhf distro?
<pfdm>
yes, archlinux powered !
Black_Horseman has joined #linux-sunxi
<oliv3r>
can someone explain to me how it's possible taht some IP has their own dma controller? i mean, the soc has one, how do they 'talk'
<Turl>
pfdm: ah, cb2
<Turl>
my tests are on A10 (cb1 like)
<Turl>
pfdm: if you feel like experimenting, try two simultaneous transfers via two connections
<pfdm>
Turl : yes cb2, but the A20 should be more powerfull ?
<pfdm>
Turl ? what do you mean two connections ? two ssh session ?
<Turl>
pfdm: it's dual core, but each A7 core may independently perform more slowly than the old A8 core
notmart has quit [Quit: notmart terminated!]
<Turl>
pfdm: yes, two independent ssh sessions
<pfdm>
I wonder if it's not a cache issue with disk io or the sd -reader two slow
<pfdm>
I can try from 2 different laptop at the same time
<Turl>
oh, you are transfering from/to sd card?
<Turl>
that may be the limit then
<pfdm>
Turl: to sdcard, I burned my sata hdd trying to power it (feeling stupid)
<Turl>
ouch
<oliv3r>
burned how did you burn it?
<Turl>
pfdm: you can make a tmpfs and try to upload files there
<pfdm>
I know the sdcard can do pretty well !
<pfdm>
oliv3r: The disk was resetting all the time when doing intensive io, so i decide not to power it through cb but using an external power, i used an external case board and use the pin for the fan...and dead !
<pfdm>
Turl : Yes I can try tmpfs too see
<oliv3r>
pin from the fan?!?!
<Turl>
oliv3r: molex?
<oliv3r>
but those are usualyl 12V
<Turl>
yeah, and disk is 5v right?
<oliv3r>
exactly
<Turl>
so it burns :)
<oliv3r>
lol literally
<oliv3r>
i blew upa pc once
<pfdm>
oliv3r: yes the +- two pins connector, I thought it was a much smaller fan but i was wrong...
<pfdm>
but if some cb team hear me, it would be nice to have a sata cable with separate data and power, specially if the board can't deliver enough for the disk
<ssvb>
pfdm: cortex-a8 runs circles around cortex-a7 for the crypto stuff
<pfdm>
ssvb: alright, let's hope hw crypto will help then
<ssvb>
pfdm: "Rescheduling for dual-issue pipeline resulted in 22% improvement on Cortex A8 core"
<ssvb>
pfdm: and then "Profiler-assisted and platform-specific optimization resulted in 16% improvement on Cortex A8 core"
<pfdm>
ssvb: nothing is done for A7 now, or there is nothing to be done ?
<ssvb>
pfdm: this is just a good example with nice comments, but the other assembly optimized functions from openssl have similar performance properties
<oliv3r>
pfdm: it is seperate :p
<oliv3r>
but you should have a beefy PSU
<oliv3r>
needing two psu's is silly
<ssvb>
pfdm: the difference is that A8 is a true superscalar dual-issue processor capable of sustained execution of ~2 instructions per cycle with proper tuning
<ssvb>
pfdm: but A7 has only very limited dual-issue support
AreaScout has quit []
<ssvb>
pfdm: on compiler generated code A8 is only a little bit faster per mhz, but with hand optimized assembly A8 leaps ahead
<pfdm>
ssvb: wow, means when gcc will do a better job, A10 will beat A20
<pfdm>
oliv3r : what do you mean needing to psu ? one for the board one for the hdd ?
<ssvb>
pfdm: I don't think that this is going to happen any time soon :) Itanium was designed to rely on clever compilers, but this did not work well
<pfdm>
ssvb: yeah remember my old IA64 ...never used at full capacity :(