<Turl>
drachensun: what did you type? your error seems to have an extra .a
<drachensun>
yup, trying again
<drachensun>
hey alright
<drachensun>
thanks for the help
<Turl>
drachensun: can you try this and let me know if it works still? http://sprunge.us/HLJW
<Turl>
I always use -j2 so it's probably a concurrency problem I just haven't hit :)
<drachensun>
that would actually make sense
<drachensun>
whats the default, if your dont set a value for j?
<Turl>
1 if you use make
<Turl>
if you use mka, or brunch and friends, it's $(number of cores)
<drachensun>
so that should be -j2 for me too
<drachensun>
I am building for jellybean
<Turl>
yeah same
<drachensun>
ok, I saw IceCreamSandwich/ in that patch and wanted to make sure
<drachensun>
well this is in my virtual ubunut 10.04 time machine
<drachensun>
so it only has one core assigned I think
<drachensun>
I applied the patch and started building for my target again, I'll see if it comes back up
* Turl
fired a build too
<Marex>
Turl: is there any howto for building android images for mk802 ?
<Turl>
Marex: nobody with an mk802 cared to make a device tree for it yet, so not really
<Marex>
Turl: android images ... I mean the blob that you write into the device's memory and it boots android magically
<drachensun>
well one comes before the other
<Marex>
drachensun: why is that so, I'd expect some ancient kernel would work with android 4.x because they contain the same ABI (or how does cyanogenmod do it with the kernels ?)
<drachensun>
ok, so you mean repacking a stock android image with the latest kernel?
<Marex>
jesus
<Marex>
drachensun: the other way around -- building a new userland and using an old kernel with it
<drachensun>
lol gotcha
<Turl>
Marex: userland is device-dependent, you still need a device tree
<Marex>
Turl: are we talking about the same DT ?
naobsd has joined #arm-netbook
juandie has quit [Quit: Leaving.]
<Marex>
(android is a really sad story, indeed :( )
<lundman>
interesting, scatterlist seems to have some maximum ->length value
ibrah has joined #arm-netbook
sv has quit [Remote host closed the connection]
askit has joined #arm-netbook
sv has joined #arm-netbook
bsdfox_ has quit [Read error: Connection reset by peer]
bsdfox_ has joined #arm-netbook
bsdfox_ has quit [Changing host]
bsdfox_ has joined #arm-netbook
Quarx has joined #arm-netbook
cat_x301 has quit [Ping timeout: 256 seconds]
ZaEarl has quit [Ping timeout: 260 seconds]
stefanro1 has quit [Quit: Leaving.]
WarheadsSE has quit [Ping timeout: 256 seconds]
stefanro has joined #arm-netbook
mSquare has quit [Quit: Leaving.]
ZaEarl has joined #arm-netbook
rellla has joined #arm-netbook
rellla2 has joined #arm-netbook
rellla has quit [Ping timeout: 252 seconds]
<hno>
crap, SPL somehow broken again.
rellla2 is now known as rellla
* stefanro
is not responsible this time ;)
<hno>
No, it looks like it's me this time. Forgot to test a cold boot in the register #define cleanup.
<stefanro>
hno: btw: seems you cleanup for submission is going along quite good - is there anything that i might do to help out?
<stefanro>
hno: even though time is a bit limited right now
<hno>
stefanro, I sent a mail to the list with a little todo.
<hno>
on that is close to your heart might be updating wemac driver to current mii bus registration api?
* stefanro
has to admit that he has never used this api before
<stefanro>
hno: i'm not sure if this is required for upstreaming
<hno>
No idea. You know those requirement better. All I know is that the code for miiphy_register do say "This API is now deprecated. Please use mdio_alloc and mdio_register, instead." but I see now that README.drivers.eth says miiphy_register.
<hno>
so seems the transition is not really started yet.
ibrah has quit [Ping timeout: 268 seconds]
cat_x301 has joined #arm-netbook
<stefanro>
hno: yes, i would drop this api move for now
ppord has joined #arm-netbook
<slapin_>
hno: ping
<hno>
slapin_, pong
<slapin_>
hno: good morning.
<hno>
morning
<slapin_>
hno: what I don't understand about this NAND stuff - it seems that read/write commands are handled differently in the controller than other commands.
<hno>
slapin_, In what way?
<slapin_>
hno: their output data are not in SRAM, at least not in same SRAM as the rest of commands
<hno>
slapin_, odd. There should be no other SRAM, and it need to buffer the data to do ECC.
<slapin_>
hno: at least successful read produce no valid information
<slapin_>
hno: I start to think that DMA trick is unavoidable, which is too bad :(
<slapin_>
hno: also, original code executes read command without really caring for its output, like it will be read automagically
<slapin_>
hno: see for example _read_single_page_spare
<slapin_>
hno: or I miss something obvious here....
<slapin_>
on NAND_CMD_READ0 I tried executing 2 command sequence (00 C C C R R 30 -> D D D) which was expected to write something to SRAM, but it didn't
<slapin_>
hno: 4-command sequence (00 ... 05 .... e0 .... 30 -> ...) is the same
Holo_ has quit [Remote host closed the connection]
Thomas42 has quit [Ping timeout: 260 seconds]
slapin_ has quit [Ping timeout: 246 seconds]
slapin has joined #arm-netbook
steev has quit [Excess Flood]
arokux has quit [Remote host closed the connection]
jeremb has joined #arm-netbook
WarheadsSE has quit [Ping timeout: 240 seconds]
ppord has quit [Quit: Leaving.]
WarheadsSE has joined #arm-netbook
jeremb has quit [Remote host closed the connection]
Thomas42 has joined #arm-netbook
Holo_ has joined #arm-netbook
jeremb has joined #arm-netbook
<hno>
slapin, ok, context switching to NAND. Where were we?
popolon has joined #arm-netbook
Hexxeh has joined #arm-netbook
<hno>
slapin, I do not think the controller have any hardcoded commands. Is there any difference is control bits set?
steev has joined #arm-netbook
<hno>
slapin, in my taces I never seen byte count higher than 0x400., even when the DMA transfer is obviously much larger.
MindBeat has quit [Quit: Leaving.]
jeremb has quit [Remote host closed the connection]
ppord has joined #arm-netbook
<hno>
Argh.. it's back at returning 00 00 00 00 00 00 00 again. Thought I had that fixed.
<focus_well>
How does one do a Lubuntu from scratch? I'm trying to make my own A10 ARM board and then put Linux on it. I found the MK802 and the Lubuntu on that is really excellent! The gist of my current understanding is that you install tool chain to compile kernel for the CPU in already working Lubuntu such as the MK802, and then after that its a blank.
<focus_well>
I have installed the tool chain and it compiled Gambas (a GUI program like VB) from sources so I know compiler bits at least work.
<hno>
focus_well, you should not need to rebuild the kernel. All A10 boards uses the same kernel.
alcides has joined #arm-netbook
alcides has joined #arm-netbook
alcides has quit [Changing host]
vanner has left #arm-netbook [#arm-netbook]
<focus_well>
ok, thats good, so I copy the already built kernel, but then next what do I do say to get a bash prompt on newly designed board with nothing on it?
<hno>
focus_well, you need a userspace root with a bash prompt.
<focus_well>
hno: is that something I compile and build, or is that something that is available pre-built?
<hno>
the lubuntu one you have is a candidate.
<hno>
focus_well, I would recommend playing a bit with qemu to learn the boot procedure, kernel arguments etc.
<slapin>
hno: NAND NAND NAND, might you play a bit with code on pastebin?
<focus_well>
hno: hmmm.. I used qemu a long time ago, is there a qemu image for the A10?
<hno>
focus_well, your question is mostly architecture neutral. You can use x86 while learning to speed things up.
<focus_well>
hno: I have 2 MK802s to play with and a couple of 8GB uSD cards. So real hardware is not a problem.
<hno>
focus_well, it's much quicker doing this with a VM. But you can do it using real hardware ans swapping cards if you like.
<hno>
I would wait with the hardware until you have a basic understanding of how booting Linux works.
<focus_well>
hno: real harware I prefer. So basically you tell get the linux, put onto uSD, and find a way to make that boot into bash. Other pray hard until your prayers are answered?
<focus_well>
Otherwise pray hard?
<hno>
There is no praying needed. Only doing.
<focus_well>
Good :-)
<focus_well>
Thats progress!
<hno>
But you need to have a understanding of what is bootloader, kernel, initramfs, rootfs, userspace.
<focus_well>
hno: I have good understanding linux
<focus_well>
but insufficient detailed knowledge of the bits that connect everything together
<cheng>
the better way is have a good understanding of each steps (at least it make senses to you) before going hands one
<cheng>
hands on.
<hno>
focus_well, so you are familiar with kernel command line, rood fs specifications, how to make a rootfs, how to build a kernel?
<hno>
and how to make an initramfs?
<focus_well>
its no to all, I only know about them, the boot loader I know a bit more about because I make SDCard read/write hardware and software
<hno>
then spend an hour on learning these things. It helps a lot. Kernel compile is not the important.
<focus_well>
very detailed knowledge about sectors and registers for controlling SDcard
<hno>
that is not needed. software already take care of the register details.
<focus_well>
so I guess the next step is getting the boot sector to boot a linux
<focus_well>
or is the bootloader that I have to inform as the boot loader already knows how to read sdcard
<hno>
Both CPU, bootloader and kernel knows how to read sdcard.
<hno>
I assume you have UART console?
<focus_well>
I got PC, and USB to serial, so I assume I can wire some pins to A10 to get UART console? (Sorry to ask I am out of my depth here with A10 hardware for now.)
<hno>
Ok, back to learning desk then. Don't dive into this on real hardware without console access.
<focus_well>
hno: my plan was I got MK802, so do everything on uSD and then get my own A10 to boot from uSD. The A10 will go on a SO-DIMM 200 pin card and it will have on it the uSD socket
<focus_well>
MK802 I can buy for $30 - so buy 10 if need be! :-)
<focus_well>
I did not want console based programming
<hno>
focus_well, sure. And it only have UART on small test pads so it will take you some time to get the console running on each.
<hno>
plus you need to be careful so you actually get an A10 based one. MK802 exists in many flavors.
<hno>
Having console access is critical to understanding these things.
<focus_well>
I know console is good, but I saw an opportunity when I note A10 can boot directly from SD. if I have a working MK802, then perfect everything by swapping code between different uSD cards
<hno>
focus_well, without console you are blindfolded. Everything must be close to 100% right before you see any output on HDMI.
<focus_well>
hno: I take in your advise. I like the blindfold. What I was going to do is run Linux and get a small program to flash LED. That way I know it booted and doing things. I am aiming for 5" LCD instead of HDMI - because the board won't have HDMI.
<hno>
LCD is even more complex to get right than HDMI.
<focus_well>
Once I got LCD to show a bash prompt, nothing to hold me back
<focus_well>
I got LCD I make work with other ARM - non Linux software
kaspter has quit [Ping timeout: 256 seconds]
<focus_well>
So it is a matter of configuring LCD driver for Linux with the LCD I know works, and getting Linux to boot to bash
<focus_well>
hno: if things get really difficult I get console working
<hno>
OlinuXIno A13 probably a better dev platform for you than MK802. And maybe A13 is even the better choice of CPU if you only want LCD.
<hno>
max 512MB memory, no HDMI. no LAN, no SATA.
<focus_well>
The A13 RAM is limiting - 512MB
<focus_well>
Need 1GB, LAN, Sata optional, but prefer I have it
<hno>
Cubieboard is it then.
<focus_well>
Yes but I hold back from ordering cubie because I don't know when they ship
<hno>
Note: there is only one day left of preorder.
<focus_well>
I will place order for 2 then - but do you know when they might ship?
<hno>
Hopefully before christmas.
<hno>
but it's tight.
<focus_well>
thats why i resist - may be i risk it
<focus_well>
hno: thanks I go study more
cheng has quit [Quit: Leaving]
<slapin>
is it theoretically possible *some random tablet* or mk802 to upgrade to SATA or LAN?
<rm>
there are dongles for USB to SATA, USB to Ethernet
<rm>
or do you mean native A10's
juandie has joined #arm-netbook
<hno>
slapin, only if you desolder the A10 and rework the PCB to connect the pins.
<rm>
afaik at least Ethernet requires quite a bit of circuitry (PHY, transceiver, etc) in addition to what's in A10
<rm>
not sure about SATA
<hno>
SATA PHY is internal. But it's 3GHz traces.
<hno>
Only connector, a resistor, some capacitors and power connected to the SATA module power pins of the A10.
<hno>
Ah, right, a 2.5V power source as well to connect there.
<slapin>
hno: any SATA-enabled schematics?
<hno>
slapin, yes. both cubieboard and olinuxino A10.
* slapin
is prowd of owning THE ONLY battery-backed mk802
<slapin>
hno: please help me identify mk802 clones
<slapin>
hno: mine new 512MB board does have axp209
<slapin>
hno: but layout is different from what is on so called mk802ii
<hno>
slapin, get yourself some cubieboards.
<slapin>
hno: I don't understand this indiegogo thingie
<hno>
what is unclear about it?
<slapin>
hno: I was not able to figure out at time how to buy 3 boards, and ordered one.
juandie has quit [Quit: Leaving.]
arokux has joined #arm-netbook
<slapin>
hno: I know it will not be enough for me, but as I see time is over for second thoughts.
<slapin>
so I am still on sticks and tablets.
<slapin>
how much does OlinuXIno cost?
<slapin>
hno: does cubieboard have battery pins, by the way?
<slapin>
and also, how hard is to cheat A10's SATA to connect eSATA raid box?
<slapin>
and where can I buy a10 kits (a10 + axp209)?
<hno>
slapin, no battery pins in this generation.
<hno>
you can buy A10 kits from olimex
<slapin>
hno: too bad, I'd like to play with a battery on cubieboard :(
<hno>
A10 SATA connects fine to SATA RAID boxes with built-in RAID. It does not work with port multipliers however.
<slapin>
hno: about NAND - can you please make your opinion on reading stuff?
<slapin>
hno: my eSATA "raid" box is really multiplier.
<slapin>
hno: is it hardware or software issue?
<hno>
I don't have much idea on what is going on with the NAND read commands. But don't see that the controller should make much difference.
<slapin>
hno: what requirements are to work with multipliers? I use dreamplug, where multipliers work, and don't see there any special support.
<slapin>
hno: do you have an oscilloscope or logic analyser?
<slapin>
hno: I'd like to see what happens on NAND's pins...
<slapin>
this is all too weird
<slapin>
so if NAND controller is considered not that intelligent, then why there is nothing in SRAM after read command?
<hno>
slapin, I do have an OLS. But SUMP Java client is not playing well with my kernel.
<slapin>
hno: no other clients?
<hno>
there is a python client, but it is very basic and do not support compression so it's hard to capture high frequency signals.
<hno>
trace buffer spase of OLS is very limited.
<slapin>
hno: :(
<slapin>
actually only <20 bytes needed
tzafrir has joined #arm-netbook
<hno>
I will try connecting up the OlinuXIno A13 with SUMP later. I can make it run with the help of GDB.
<hno>
Or I think I remember how. Was a couple of months ago.
* slapin
read ML... how do they connect SATA to STM32F I wonder...
<hno>
What?
<hno>
Don't remember any discussion about STM32F SATA. Ony SDCARD.
<slapin>
hno: EOMA board...
<hno>
Yes
<hno>
EOMA have SATA.
sspiff has quit [Ping timeout: 245 seconds]
<hno>
EOMA CPU module -> IO Board -> SATA device. Not touching the microcontroller.
<oliv3r>
so the 'missing' clock gates can be figured out by trying and seeing what does or does no longer work?
<mnemoc>
oliv3r: semantically it makes more sense the cubieboard2 to have the double of extensions, and the cubieboard4 to have the quad core soc
<oliv3r>
slapin: so basically its a transistor that allows the clock signal to pass or not
<mnemoc>
to give some meaning to the number
<slapin>
oliv3r: might be so
<oliv3r>
mnemoc: cubebieboard++
<oliv3r>
slapin: in 'laymans terms' i ment :)
<oliv3r>
i just ask, because there's no clock gating defined for things like AC97, SPDIF, keyboard etc
<slapin>
oliv3r: it is usually several components, but yes
<oliv3r>
and some spots where they could very well be hiding in the gating register
<slapin>
hmmmmm
<oliv3r>
hmm, the doc is quite specifc, anyway, what triggerd me, is there are timer registers for said components, but no gating entries in the gateing registers, BUT there are holes in the gating registers
<oliv3r>
(page 47 and 48)
<oliv3r>
so by disabling/enabling certain gates, the ac97/spdif etc gates should be 'findable'
<slapin>
oliv3r: or not
<slapin>
oliv3r: depends if there are any gates
<oliv3r>
slapin: :(
<oliv3r>
well there's some 'reserved holes' e.g. 'we don't want you to see this' in the docs and some aren't even in the src
<oliv3r>
well not in ccmu_regs
<slapin>
oliv3r: they omitted lots of peripherals in the doc, no wonder
<slapin>
which of these meles is a2000g?
<oliv3r>
slapin: yeah most of it is in the source, but some isn't in the source either, so they are just 'hiding' it?
<oliv3r>
slapin: there is only an a1000g i belive, and hipboi said it's not visible in the front yet
<oliv3r>
probably has to replicate across the frontend mirrors first
<oliv3r>
i am guessing
<hno>
slapin, need a picture of the board to be able to help locating serial.
<mnemoc>
slapin: aliexpress moderates new products, the a2000g is not just published on the catalog
<stefanro>
hno: might be that i can find some time for the sr32 cleanup - which branch is the one i should use?
<slapin>
mnemoc: any direct links?
<mnemoc>
slapin: only after it passes moderation
<slapin>
:(
<slapin>
I will spend all my money by then :(
<oliv3r>
i wanna buy now!
<hno>
oliv3r, ther is schematics, full board layout CAD drawings and component supply available for nexus 10?
<mnemoc>
"it for your own safety" (tm)
<oliv3r>
hno: of course not, they where talking software! the statement up above claimed 'full' meaning software too
<oliv3r>
safety? pff! safety is for wussies!
<mnemoc>
oliv3r: if you ignore the gpu and vpu, software is 100% free/open
<mnemoc>
and those two are on the road of been REed
* slapin
got a cruel plan building A13 "stick" using components from balcony, olimex, and building board at iteadstudio.com
<hno>
oliv3r, what missing AHB gates?
<mnemoc>
slapin: the a13 is awful for an stick
<slapin>
this is far cheaper than building S3C24xx boxes
<slapin>
mnemoc: I solder large BGA packages very badly
<mnemoc>
S3C24xx is so 4ya
cheng has joined #arm-netbook
<hno>
oliv3r, see clock diagram in the datasheet (short version with pinout)
<hno>
stefanro, always sunxi-current if you want a sane life with git. sunxi-patchqueue also works but is continously rebased.
<slapin>
mnemoc: A13 is TQFP, and I can solder RAM chips
<stefanro>
hno: okay, thats what i thought
* hno
is not entirely comfortable with TQFP of the spacing & size of A13, but could probably do it.
<oliv3r>
hno: ac97 clock gates
<oliv3r>
hno even the short datasheet has things like 'NC-CLK' under the nand-clk which I think was sata clk iirc (see linux-sunxi.org/A10/CCM top section)
<oliv3r>
or the IIS-clk combined with a nother 'NC' clock
hipboi has quit [Remote host closed the connection]
<oliv3r>
slapin: good plan, I've solderd a GDDR ram module too, solderd the pads, put chip on, used hot air blowdryer to 'melt' it in place :)
<oliv3r>
hno: and finally, KEYPAD-CLK shows up in the diagram, but is 'missing' from CCM_AHBCLK_GATE[01] in source
<L84Supper>
mnemoc: so what I think you are saying is that is other than all the closed stuff it's all open
<hno>
oliv3r, KEYPAD & AC97 is on APB
<mnemoc>
L84Supper: only that all closed stuff in the a13-olinuxino is optional
<mnemoc>
L84Supper: and that eventually those will be open too
<oliv3r>
hno: ah! (Wasn't that far yet :p)
<oliv3r>
mnemoc: eventually :( Faster! Faster! :) sorry, i'll get back to work
<hno>
oliv3r, there is several different logic busses depending on the speed of the pheriperal.
<pucko>
gah... I lost the URL for the Mele Debian armhf image... anyone remembers?
<hno>
oliv3r, most peripherals also have a function clock.
<hno>
both clocks are needed for operation. The bus clock controls the CPU facing side (registers etc) and the function clock the operation of the device.
<oliv3r>
ah, dual facing I didn't know
<hno>
the bus clock is only gated. function clocks have divisors
<L84Supper>
is somebody working on reversing the video decoder/accel in the a10/13? is that part of Lima?
<hno>
and some pheripherals have additional clocks derived by th
<hno>
their own.
<oliv3r>
so you enable/disable the clock (gate) on the CPU facing side; and you setup devisors for the perp. facing side, check. I may write a little abbout that on the wiki
<hno>
i.e. UART have a baud rate clock divisor dividing the UART function clock to a sensible baud rate.
<oliv3r>
L84Supper: ian ... i forgot, he posted on the ML and on g+
<L84Supper>
so allwinner is shy when it comes to sharing any of this source
<hno>
shu of sharing any source currently :(
<L84Supper>
nothing like building trust :)
<L84Supper>
maybe Luke just heard them wrong last year :)
avernos has quit [Remote host closed the connection]
<hno>
L84Supper, no, lkcl knows where they stand.
<L84Supper>
"sure we will share all the source with you in exchange for community support..... except for big chunks of it"
mSquare has quit [Quit: Leaving.]
<hno>
for the A10 the situation is reasonable. The SDK have full kernel & u-boot sources.
<oliv3r>
i think once linux-sunxi work becomes more known and more used, it may trigger allwinner into seeing that what is happening here is a good thing (for them)
<L84Supper>
oliv3r: maybe, I haven't met any of them or checked into their backgrounds. I wonder what their track record was like with their management before.
<L84Supper>
is it the same team as Boxchip?
<hno>
oliv3r, maybe. But it's too late for any serious effect. Allwinner already had their boom with A10 sales. Now focus is on next generation.
<hno>
L84Supper, should be.
<hno>
with some adjustments from being several companies merged together.
<L84Supper>
hno: I agree, they should be giving all the source away for the A10 now that it's old. What's left to control? .... unless you have issues with control
<oliv3r>
hno: that may be true, but A40 may be very close related so could get a jump start
<oliv3r>
sad but true, linux-sunxi will just follow after the facts and nothing exciting will happen in the industry
<mnemoc>
L84Supper: lichee2/2.6.36 was released once totally obsoleted by lichee3/3.0.8
<hno>
L84Supper, they already have given away as much source as they can realistically do.
<oliv3r>
libcedarx!
<mnemoc>
Turl: did you decide to take vaapi or REing cedarx as univ. project?
<L84Supper>
hno: since GPU and video decoder source is under NDA from ARM?
<hno>
cedarx is still their main IP.
<hno>
The MALI GPU is under NDA from ARM.
<L84Supper>
is the video decoder/accel (mpeg, h.264 etc decoder) part of mali?
<mnemoc>
L84Supper: that's cedarx, not mali
<L84Supper>
where do they get the cedarx block from?
<hno>
cedarx video & audio codec is their is considered their main IP block, what sets the company apart from others.
<L84Supper>
did they design it themselves?
<hno>
full history is unknown, but it's been part of their whole series of CPUs, noticeably improved with each generation.
<L84Supper>
oh ok, and they won't share source for those blocks, got it now
<slapin>
I don't know how this all is related to closed drivers
<slapin>
I understand they don't want to share their IP blocks, but drivers are just controlling interface...
<hno>
slapin, acceleration hardware and software using it goes in tandem.
<L84Supper>
so if someone were to reverse the cedarx it would give some of what they consider 'secrets' away to how to build the multimedia IP blocks
<slapin>
I don't think I will understand anything re IP blocks building from looking at driver source...
ppord has quit [Ping timeout: 260 seconds]
<L84Supper>
now I'm starting to understand the pathology
<slapin>
I don't even know what NAND controller thinks now.
<hno>
slapin, IP concept also covers software.
<L84Supper>
part of the reasons why register specs are not shared is that it provides evidence of how the hardware works and what patents might be violated
<oliv3r>
mnemoc: vaapi? what happened to openmax!
<focus_well>
hardly true about IP secrets - everyone knows how to build multimedia blocks - all they loose are sales if their chips are not selling
<L84Supper>
it could easily be covered by just moving some registers around but that's too much work for some ASIC teams
<slapin>
L84Supper: if you cover video you 100% know that hundreds of patents are violated, just knowing it is video decoder which is not paid to MPEG LA
<libv>
L84Supper: ?
<libv>
L84Supper: there's millions to be earned by patent cases
<libv>
L84Supper: it's trivial enough to RE one of these blobs, and point out to a judge if it is violating some patents, especially given the amount of money, and this REing time available
<focus_well>
The best way around mpeg is to create your own alternative video formats and support one for sure and edge out the closed ones
<libv>
focus_well: and now for the real world...
<focus_well>
that means a bit of VHDL to add into the cores and then some software to use it
<slapin>
focus_well: many patents are not algorithm or format related, they cover whole process of video decoding and encoding
<L84Supper>
yeah, they aren't going to support a proprietary standard
<libv>
again, the availability of code does not help with IP issues
<focus_well>
libv: we all make money eventually doing our own open source projects - so all it needs is one or more persons to invest in this in organized way to phase out locked systems
<libv>
focus_well: go at it. now.
<focus_well>
libv: I sure will - just starting out on freeing up the KiCAD designs for modular SO-DIMM PCBs and after that who knows
<focus_well>
the money is mine but designs are of no use to me after they are built so I give them away
<slapin>
any available design on 72-pin SIMM module for KiCAD, and is DRAM chips are available? my AXPpci board is collecting dust without memory :(
<focus_well>
not that I know of - I'm just starting out - I don't know where it take me
<oliv3r>
there, lots of stuff up on the wiki. so much my fingers hurt now! More tomorrow maybe :)
<mnemoc>
he is designing SoMs, not DRAMs
<L84Supper>
the a10 was successful because of its features at a low price....
<libv>
oliv3r: :)
<focus_well>
mnemoc: interchangeable SoMs so if one manufacturer falls over then switch to another one
<L84Supper>
the main cost of developing an ARM SOC is the ARM cpu + GPU license + the other IP blocks you buy (and the masks)
<focus_well>
that way those who can't open source their drivers will sooner fade
<oliv3r>
well until the a40 comes out, the a10 will remain successful. if the a40 has the same pinout as the a10; it will be an instant success imo
<focus_well>
L84Supper: there is new CPU + Linux all running on open VHDL code at opencores.org - no need for arm
<oliv3r>
problem with opencores, you need to be able to manufacture said CPU's
<mnemoc>
focus_well: are you designing also some boards where to put these SoMs?
<L84Supper>
focus_well: I don't think people are interested in supporting those cores
<oliv3r>
and that requires quite a lot of money (and quite some expertise). BUT I agree, in the far future, aboslutly
<oliv3r>
L84Supper: they will be, once you can actually buy hardware with those cores
<slapin>
I think crosdfinding some SoC would be interesting experience
<slapin>
hno: the ones near USB?
<focus_well>
that what I think, one of us make it big, don't forget your roots, and fund an open CPU with open VHDL and Linux
<slapin>
I mean wifi usb
<focus_well>
if you make money, you can't take it with you, so spread it around on open sourced projects :-)
<hno>
You almost always first to a CPU in an FPGA, then when tested & verified you resynthesize it for silicon.
<L84Supper>
focus_well: I wish there were enough people interested in purchasing such devices
<slapin>
focus_well: you can spend all money on vodka and girls, unlimited amounts of money.
<focus_well>
i'd be bored i tell ya!
<L84Supper>
what most people seem to want is either a status symbol or cheapest that works even if it's closed
<hno>
slapin, yes those two. You need to find GND elsewhere, preferably not the same spot you wire the wireless module..
<slapin>
L84Supper: enough people buy mk802, and its usefulness is questionable, so there will be enough people for open CPU, of properly adverticed.
<L84Supper>
slapin: have any references I could show to a bank, investors, venture cap. etc etc?
<slapin>
L84Supper: you can say in your ad, that successful people use only open SoCs/CPUs and it will become status symbol.
<L84Supper>
I just don't see the numbers adding up
<focus_well>
whats missing is a linux + LCD module that runs gambas (a VB like program). That can go into any machine. I got MK802 with Gambas working.
<L84Supper>
heh, I'm all behind you trying that
<focus_well>
So all I need to do is put it into SoM and make available drivers for 3", 4", 5", 7", 10" LCDs
<slapin>
L84Supper: if you don't see a numbers, make fake ones.
<L84Supper>
if the RaspberryPi has to fab their own device they would't be breaking even yet
<L84Supper>
has/had
<slapin>
focus_well: don't forget of quite popular 1.5" and 2", 2.2"
<slapin>
L84Supper: RaspberryPi has good sponsors, no?
<focus_well>
The MK802 image is here (not for public release) http://www.gplsquared.com/mk802/mk802.html with gambas3 working and a sample program that boots up. The sample program can even talk!
<hno>
L84Supper, Infrant used an open core in their CPU which was used in ReadyNAS. Later bought by Netgear. They switched to an off the shelf ARM cpu & X86 models nowdays however, mostly because there now is suitable and not too expensive ARM CPUs for their market.
<traeak>
hehe
* slapin
feels bad about talking programs/computers
<focus_well>
to make it talk - just type into the box below the talk button and press the button - it comes out the HDMI speakers
<traeak>
i wish the fpgas were cheaper
<focus_well>
The program is serial port communications - so if you send it RS232 data, it logs it to a file. If you send the data with <s> text </s>, it will parse it and talk that bit
vanner has joined #arm-netbook
<focus_well>
good for RS232 terminal that talks!
<focus_well>
That is kind of application I think people will buy
* slapin
is sick of kid's talking toys :(
<L84Supper>
organize the community, show a vendor that they will make $$ by making open hardware, if its real they will do it
* slapin
thinks machines should not talk - that's human's privilege
<focus_well>
traeak: fpgas will never be cheap - because inside is a lot of waste and redundancy
<L84Supper>
get 1M devs to put up $10 ea
<traeak>
focus_well: point taken, it's a technology thing
<traeak>
well things are changing
<traeak>
10 years ago the options were pretty bleak
<focus_well>
L84Supper: I intend to release the SoM KiCAD, and that will allow anyone to put their chip to a SoM and get it talking to LCD - instant winner
mikey_w has quit [Remote host closed the connection]
<focus_well>
With gambas running on it, they can use like VB and design their applications very quickly and get it on to the SoM and make money
<hno>
focus_well, FPGAs are very cheap compared to not having FPGAs
<focus_well>
Of course, if other chip makers want in, they need to their SoMs
<focus_well>
and Linux
<focus_well>
and LCD :-)
<slapin>
my old OR project FPGA board costs $10
<focus_well>
and if their stuff is closed, their Lunux and SoMs will not work or be cheap
<slapin>
focus_well: and how to hide trade secrets? :)
<hno>
slapin, does not seem to be the NAND clock, unless I am setting it completely wrong.
<slapin>
hno: can you do register dump on original u-boot?
<hno>
not @480MHz. Need to reflash first.
<hno>
@408MHz both works.
* slapin
looks for some automatic mmc card inserter/remover/writer
<slapin>
I can't do this stuff at work, and at home I have too little time :(
<Turl>
lundman: why is rm so slow on zfs? :/
<hno>
slapin, there is loady command in u-boot.
<rm>
actually I prefer btrfs
<rm>
maybe that's why :p
<Turl>
rm :P
<rm>
and for the record I think porting of ZFS to Android (by lundman ) while an impressive feat
<rm>
using ZFS on Android to me seems compeletely insane
<rm>
like using rocket fuel in your lawnmower
<slapin>
hno: it is cool, but I need clean boots to test u-boot
<jinzo>
Damn, even the Allwinners datasheets are inconsistent.
<slapin>
hno: jtag might help, probably, but I still don't understand how it is generally attached to Allwinner devices...
<hno>
slapin, seems we are missing something in the initialization. Works if I first let Allwinner nand driver access the NAND and then ours. Still @480MHz.
<slapin>
hno: well, probably before NAND initialization then
<slapin>
please dump registers to compare...
<hno>
slapin, it's still loaded by SPL.
<hno>
slapin, I will. But need to put the kids to bed first.
<slapin>
hno: ok, I'm going home, afk for 1-2 hours.
NIN11 has joined #arm-netbook
vinifm has joined #arm-netbook
NIN11 has quit [Ping timeout: 245 seconds]
<zoobab>
hi
<zoobab>
just getting my Chromebook now
ZaEarl has joined #arm-netbook
slash_random has quit [Ping timeout: 240 seconds]
freakazoid0223 has quit [Quit: Leaving]
<sv>
awesome zoobab
<sv>
the ARM ones with 2gb of ram?
<mnemoc>
the exynos 5 chromebook is the only relevant one for this channel :)
Mehhh has quit [Ping timeout: 245 seconds]
pawel5870 has joined #arm-netbook
<drachensun>
jinzo I'm around now
<Turl>
drachensun: hey
<Turl>
drachensun: did you get android running yet? :)
<drachensun>
turl: It built for the zatab, still hanging on my new tree, I have to figure out some settings and play with the defconfig for the kernel
<drachensun>
turl: but yeah, its looking good
pawel5870 has quit [Read error: Operation timed out]
<drachensun>
turl: is there any guide or anything for how to boot cm from an SD?
techn has quit [Ping timeout: 252 seconds]
<drachensun>
oh never mind, everything in the zip is pretty clear
Mehhh has joined #arm-netbook
Mehhh_ has joined #arm-netbook
<Workboot>
Is journalling really appropriate for flash storage?
<specing>
no
Mehhh has quit [Ping timeout: 245 seconds]
pawel5870 has joined #arm-netbook
slash_random has joined #arm-netbook
Mehhh_ has quit [Ping timeout: 245 seconds]
Sternennebel has joined #arm-netbook
<jelly-home>
but a fully logging file system is like journaling forever!
techn has joined #arm-netbook
<hno>
jelly-home, there is very big differences between journaling and log-structured.
merbzt has joined #arm-netbook
<hno>
slapin, it's still a mystery. But it is not any clock setting. Leaning more towards a timing issue in reset (0xff).
rellla has joined #arm-netbook
ibrah has joined #arm-netbook
Sternennebel has quit [Read error: Connection reset by peer]
bsdfox has joined #arm-netbook
ibrah has quit [Quit: ChatZilla 0.9.89 [Firefox 16.0.2/20121024073032]]
bsdfox_ has quit [Ping timeout: 252 seconds]
Almamuetya12 has quit [Ping timeout: 245 seconds]
vinifm has quit [Ping timeout: 246 seconds]
<mnemoc>
hno: the kernel nfd driver has a #define called USE_SYS_CLOCK (or similar), if enabled we don't see the nand .... it came enabled
<mnemoc>
USE_SYS_CLK
tinti has quit [Quit: Leaving]
<Turl>
fuu, directfb website down :(
<slapin>
hno: timing issue in reset command? but how?
<slapin>
hno: so partial reads are to be implemeted and my ONFI spec is not so clear about this...
<slapin>
hno: it reads starting 512 bytes of some page
<lundman>
never thought of rm as slow, but not tried it on linux
<slapin>
hno: when I choose different pages it reads different things
<lundman>
NAME SIZE ALLOC FREE CAP HEALTH ALTROOT
<lundman>
zpool1 81.2T 252K 81.2T 0% ONLINE -
<lundman>
woo
<lundman>
shame its for work
<slapin>
hno: when same pages, it reads same things
<slapin>
lundman: that's what you pay for zfs nifty features
<rm>
Turl, what
<rm>
who cares desktop or whatever
<rm>
prefix "eatmydata" to your rm command
<libv>
drachensun: how dare olimex label this thing an open platform
<slapin>
lundman: when zfs will be in legally-feasible state to include in Linux kernel?
<traeak>
not very usefulwhen empty
<rm>
it NOPs fsync, which rm might do a bit too often; in any case this seemed to help on btrfs
<libv>
in some sense, yes, but at least they picked the SoC which currently comes closest
<drachensun>
my thoughts exactly, sure its not perfect
<libv>
but really, /., pffff
<libv>
what are these guys smoking
<traeak>
lundman: what kind of stuff is going on the storage?
<libv>
and now on to real work :)
<drachensun>
but when I looked at all the other SOC's this has the best community support and most available source
<traeak>
drachensun: that's a pretty sad state of affairs
<libv>
drachensun: yeah, and then notice all the RPi fanboys on /.
<libv>
they still do not want to own up to the fact that the videocore runs the whole show
<drachensun>
traek: No argument :)
<slapin>
TI SoCs are a bit better supported, but these are a bit harder to get...
<hno>
well, 2K in total, but the upper half seems to be mostly used for collecting spare data.
<traeak>
i dumped my rpi off to some sucker, got almost what i paid for it.
<traeak>
s/to/on/
<ibot>
traeak meant: i dumped my rpi off on some sucker, got almost what i paid for it.
<drachensun>
yup, similar state for the video drivers, its much slower and you cant get it in tablets or netbooks as far as I know
<drachensun>
without hacking your own device together
<hno>
slapin, Allwinner uses a randomizer in all written data I think.
<Turl>
pretty much the only way to get an OMAP is with pandaboard/beagleboard
<Turl>
(in board format that is, you could buy a phone too)
<drachensun>
slapin: The sad thing is TI wont even sell you there current gen stuff
<slapin>
hno: is randomizer predictable?
<drachensun>
and sitara processors are VERY expensive in comparison to the chinese ones
<hno>
It's seeded in the ECC register.
<hno>
slapin, with a static seed per page.
<drachensun>
they are pretty capable though, but one processor is almost as much as a MK802 from the cheapest sources
<hno>
slapin, disabled only when checking for erased or bad pages I think.
<slapin>
hno: can you play with it? I think we can make command to disable or enable compatibility for testing. I want to concentrate on making it work properly (partial reads).
<hno>
slapin, use of the randomizer is optional, but increases NAND lifetime and reliability.
<hno>
have you pushed your read code?
<slapin>
I will push now
<hno>
Still fighting with rhe initialization issue. But quite clueless on what it really is about.
<slapin>
hno: pushed to work branch
<slapin>
hno: it was rebased to your mtd branch
<slapin>
hno: why so little NAND controllers out there implement randomizers?
<slapin>
hno: and what mechanics for randomizer to prolong NAND lifetime if seeds are fixed?
<hno>
It reduces the chances that large areas have the same cell charge.
gzamboni has quit [Ping timeout: 265 seconds]
<hno>
if I remember right.
<slapin>
hno: where can I read about it?
rellla has quit [Quit: rellla]
gzamboni has joined #arm-netbook
<lundman>
phew
<slapin>
thinking is hard :(
<lundman>
I conquer :)
cat_x301 has quit [Quit: Leaving.]
<hno>
slapin, merged your changes. You may want to update.
<slapin>
hno: ?
<slapin>
hno: interesting, is it possible to make u-boot NAND stuff to output 0x05 command and read page in parts... can't see example of that anywhere...
<hno>
slapin, no idea. Further Allwinner code do not read in parts, they do a large DMA transfer, but with nand controller byte count set to 1K.
<hno>
some magic of some kind there.
<hno>
slapin, I have some other changes in my repo.
<hno>
and github is down..
<hno>
no it's not. my git is confused.
gzamboni has quit [Ping timeout: 252 seconds]
<Turl>
[ 3.170000] [DISP] alloc_pages fail!
<Turl>
<1>Unable to handle kernel NULL pointer dereference at virtual address 00000000
<Turl>
libv: any idea? ^ :P
<hno>
Turl, the second is poor error checking.
<ZaEarl>
that reminds me of the fun that was MacOS 8. They stored the system interrupt tablet at 0, without any memory protection. Any write to a null pointer would overwrite the system interrupt tablet. Boom.
<Turl>
CONFIG_FB_SUNXI=y
<Turl>
# CONFIG_FB_SUNXI_RESERVED_MEM is not set
<Turl>
CONFIG_FB_SUNXI_LCD=y
<Turl>
CONFIG_FB_SUNXI_HDMI=y
<Turl>
maybe I need to enable the reserved mem for it?
<hno>
ZaEarl, there is several systems doign that stupidiy with their interrupt table.
<slapin>
hno: does u-boot's mtd struct contains buffer of page size?
<slapin>
also somebody needs to set proper page size in controller
<hno>
slapin, I seem to get the same non-random data in every page. lots of 0x30, lots of 0xff, lots of 0x00, lots of 0x30.
<focus_it>
Just noticed - cubieboard has raised $86,000 - $36,000 more than the $50,000 they were seeking. May be with some of the profit cubie should think about start making own SoC
<focus_it>
:-)
<libv>
Turl: sounds like a plan
<libv>
but poor checking indeed, file an issue
<focus_it>
Its about $30,000 to make a SoC, but if you don't have the infrastructure its about $250,000 to $300,000 to make it happen