Turl changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask and wait! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
<hramrach> Turl: do you have CONFIG_ARM_PSCI=y ?
<Turl> yeah
<Turl> it brings up the core
<Turl> but it's all weird
<Turl> apps hang, doesn't reboot, etc
<hramrach> did not try that new smp yet
<Turl> hramrach: ok, it works with hansg's tree
<Turl> must be something I'm missing
<hramrach> well, hansg's tree has support for it
morfoh has quit [Quit: leaving]
<Turl> hramrach: what's the config option for gmac?
<Turl> nvm
<hramrach> it's called stmmac
<hramrach> and the commit for smp is probably f712e27e95121dff08af9659e657bfba29ee4174
<hramrach> CONFIG_NET_VENDOR_STMICRO=y
<hramrach> CONFIG_STMMAC_ETH=y
<hramrach> CONFIG_STMMAC_PLATFORM=y
NightShade has quit [Ping timeout: 272 seconds]
<Turl> I'm using sunxi-next instead of -devel, derp
<Turl> no wonder gmac didn't work :)
<Turl> [ ID] Interval Transfer Bandwidth
<Turl> [ 3] 0.0-10.0 sec 864 MBytes 725 Mbits/sec
<Turl> not bad
tomboy64 has quit [Ping timeout: 272 seconds]
morfoh has joined #linux-sunxi
<wens> Turl: working now?
<Turl> wens: yeah, it was a bad patch cable
<Turl> clearly not GbE capable :)
<wens> :)
<Turl> wens: I pasted some iperf output ^ :)
<wens> nice! hadn't the time to test GbE throughput yet
<morfoh> Turl: got a new one from santa? :)
boycottg00gle has quit [Remote host closed the connection]
<Turl> wens: note that's only one way traffic; if you do simultaneous both ways it's half and half
<Turl> morfoh: not really, I had it sitting on my desk for quite a while now :p
<Turl> but my card reader was incomplete, missing its cable, so I couldn't get uboot
<Turl> I found a compatible cable today and got it all set up :)
<morfoh> Turl: :)
<wens> Turl: any chance you could get cpu usage during the test?
<Turl> wens: if you give me half an hour, sure
<wens> take your time
<morfoh> Turl: *cough* soonish(tm)!?!? *cough*
* morfoh hides :)
<morfoh> anyhow... time to go sleep I guess :)
<morfoh> have fun Turl & wens :)
<Turl> morfoh: did I test that thing? :p I forgot now :S
<Turl> morfoh: g'nite
<Turl> wens: is it using DMA or PIO btw?
<wens> Turl: it has a dedicated DMA controller. AFAIK it does a DMA burst for each packet
<wens> so you get 1 IRQ for each packet. don't know if it supports IRQ coalescing
<Turl> wens: I see some wemac errors btw
<Turl> failed to set mac on uboot
<Turl> linux says stmmac_open: failed PTP initialisation
<wens> not sure about the u-boot error, didn't dig into it
<wens> failed PTP init is ok, it just means our gmac doesn't support it
<wens> PTP is a time sync protocol like NTP
<spv> Hmm, it looks like the nand driver from Allwinner has a lot of secret-sauce for interacting with certain parts
rz2k has quit []
egbert has quit [Ping timeout: 246 seconds]
<Turl> wens: iperf seems to peg the cpu to 100%
<Turl> so it's probably cpu bound
<wens> try larger packets?
<wens> btw, can anyone test jumbo frames?
egbert has joined #linux-sunxi
<Turl> how would one test that?
tomboy64 has joined #linux-sunxi
<wens> not sure... set MTU to 9000 and run ping/iperf?
<wens> ping with large packet size of course
tomboy64 has quit [Ping timeout: 272 seconds]
<Turl> wens: eth0: invalid MTU, max MTU is: 3838
<wens> :(
<spv> turl: which gmac driver are you using?
<Turl> spv: stmmac on mainline
<Turl> spv: does the other one fare better?
<wens> should be the same
<wens> both share the same code:
<wens> max_mtu = SKB_MAX_HEAD(NET_SKB_PAD + NET_IP_ALIGN);
<spv> I haven't tried anything but stmmac
<spv> was wondering if you were talking about some "new" one.
<Turl> wens: maybe my router is the bottleneck
<spv> if you're going through the switch part, I doubt it.
<Turl> yeah, it's switched
<spv> those have been capable of wire-speed for a _long_ time.
<spv> How much throughput have you guys been getting?
<spv> I haven't run a benchmark myself ...
<Turl> spv: see above, I pasted an iperf one way
<spv> oh wow nice
<spv> 725 is not bad
<Turl> ~600-700 one way, ~300 something when going both ways
<wens> i dont have another GbE capable linux system to run iperf :(
<Turl> wens: no desktop or laptop made in the last ~5y?
rings_IIV has joined #linux-sunxi
<wens> no spare one that i can reuse
<wens> i keep my boards at work
<Turl> ah
<spv> TIL mixing-and-matching kernels with OSes can be time-consuming
mrnuke has quit [Remote host closed the connection]
mrnuke has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
tomboy64 has joined #linux-sunxi
\\Mr_C\\ has quit [Quit: .]
bsdfox has quit [Ping timeout: 240 seconds]
bsdfox has joined #linux-sunxi
<spv> Have any of you guys gotten the error "stmmac_open: DMA engine initialization failed" upon boot, in relation to the gmac?
bsdfox has quit [Ping timeout: 264 seconds]
bsdfox has joined #linux-sunxi
tomboy64 has quit [Ping timeout: 264 seconds]
Gerwin_J has quit [Quit: Gerwin_J]
tomboy64 has joined #linux-sunxi
tomboy65 has joined #linux-sunxi
tomboy64 has quit [Ping timeout: 245 seconds]
bsdfox has quit [Ping timeout: 245 seconds]
bsdfox has joined #linux-sunxi
bsdfox has quit [Ping timeout: 272 seconds]
bsdfox has joined #linux-sunxi
tomboy65 has quit [Ping timeout: 265 seconds]
bsdfox has quit [Ping timeout: 252 seconds]
bsdfox has joined #linux-sunxi
maz_ has joined #linux-sunxi
pwhalen has joined #linux-sunxi
pwhalen has quit [Changing host]
pwhalen has joined #linux-sunxi
bsdfox has quit [Read error: Operation timed out]
bsdfox has joined #linux-sunxi
maz_ has quit [Ping timeout: 240 seconds]
bsdfox has quit [Ping timeout: 240 seconds]
bsdfox has joined #linux-sunxi
Quarx has joined #linux-sunxi
<spv> wens: did you ever run into issues with the GMAC where the DMA would refuse to reset?
<wens> spv: no
<wens> does power cycling help?
<spv> negative
<spv> Do you think it would make a difference if the emac were brought up in U-boot?
<spv> (because it is right now)
<spv> i.e. would the DMA resetting depend on the pin mux state?
<wens> it shouldn't
<spv> makes sense ...
<spv> hmm
<wens> emac shouldn't be an issue, as they are seperate cores
<wens> pin mux is external, so ...
<spv> I was wondering if resetting the DMA, in the core, required that it communicate with a PHY or something ...
<spv> like resetting the DMA necessarily reset the PHY
pwhalen has quit [Ping timeout: 265 seconds]
<spv> and the PHY, presumably, would depend on the pin mux state?
<spv> since it is external
<wens> never thought of that
<wens> but if the DT is correct, the kernel should set the pinmux for you, before driver _probe
<spv> wens: what board do you have? Cubieboard2?
<wens> both CubieBoard2 and CubieTruck
<spv> I only have the CubieTruck atm
<wens> gmac support is already merged into u-boot-sunxi
<wens> though i don't recall running into problems using emac in u-boot and gmac in linux on cb2
<spv> I so do hope that there isn't a U-boot dependence on this.
<spv> I've worked on platforms that have U-boot dependence, and it's a pain
<spv> There was one that refused to initialize a nand controller unless the nand controller registers had already been correctly programmed by a loader or something
<spv> (in case you're curious, it's fsl_elbc_nand.c)
<wens> btw, gmac in hans' sunxi-devel tree is the old patch series, which requires u-boot to setup the clock registers
<spv> Well, I'm using sunxi-devel so that might explain it
<spv> where can I find the new patch series?
<spv> sunxi-next?
<wens> i suggest you update u-boot for now
<wens> i'm still cleaning and rebasing the new series
pwhalen has joined #linux-sunxi
pwhalen has joined #linux-sunxi
pwhalen has quit [Changing host]
ssvb has quit [Remote host closed the connection]
<spv> I grabbed your repo only to realize it doesn't have MMC support, which breaks the way I boot
panda84kde has joined #linux-sunxi
<spv> and yes you were right that updating u-boot would resolve the gmac problems
shineworld has joined #linux-sunxi
<wens> I plan to post the new series in a few days
<wens> after that hans can pull it into sunxi-devel
_massi_ has joined #linux-sunxi
notmart has joined #linux-sunxi
<spv> I just ran some iperfs and got similar results to you.
<spv> no, wait, it was Turl, nvm
bsdfox has quit [Ping timeout: 272 seconds]
ZetaNeta has quit [Read error: Operation timed out]
n01 has joined #linux-sunxi
bsdfox has joined #linux-sunxi
<spv> anyone here use kexec to do design iteration?
<wens> can one load dtb with kexec?
<spv> I was just looking into that ... I know that on powerpc one does
<spv> but the man page for what is in the arch repository suggests no ...
NightShade has joined #linux-sunxi
vector80 has joined #linux-sunxi
<wens> i2c0 node missing in cubietruck dt...
<spv> Does anything hang off of it?
<spv> I tend to not end up using those expansions ....
ZetaNeta has joined #linux-sunxi
ZetaNeta has quit [Changing host]
ZetaNeta has joined #linux-sunxi
<spv> wens: dtb is supported
<spv> in kexec
<vector80> oliv3r: here?
<vector80> Now it seems I have be able to boot with TF card + eMMC at the same time on debian...
<vector80> Here is the log: http://fpaste.org/64938/96904138/
<vector80> Now, I want to write this interra3_emmc.bin to my eMMC, but for this, I should be able to mount it first, right?
juanfont has joined #linux-sunxi
popolon has joined #linux-sunxi
popolon has quit [Changing host]
popolon has joined #linux-sunxi
juanfont has quit [Ping timeout: 260 seconds]
<spv> if it's just a binary blob you may not need to mount it
<spv> depends on what it represents
<spv> mount implies that there is a file system
<vector80> spv: thanks.
<vector80> Actually, in my eMMC,
<vector80> there is AW's LinuxBSP image
<vector80> As you see from the log, they use 7 partitions
<vector80> I feel I must do fdisk or something
<vector80> because partition table for debian is different
<vector80> So, How can I copy this debian image to my eMMC ...
<vector80> I am thinking this...
cajg_ is now known as cajg
<spv> if the debian image includes the partition table then you just overwrite all the things
<vector80> oliv3r know how to do, but I feel he is in vacation now..
<spv> what does the debian image look like?
<spv> is it a 2GB flat file?
<spv> or something like that?
<vector80> It is from olimex A20 wiki page
<vector80> it is made for SD card booting
<vector80> I booted my new board with TF card, and TF card has their image
<vector80> it basically has two partitions I think
<vector80> But my eMMC has 7
<spv> Let me check the A20 wiki ...
tomboy65 has joined #linux-sunxi
<spv> Are you saying you have one of these:
<spv> ?
<vector80> yeah
<vector80> that one
<spv> there are two ways to copy it over
<spv> through filesystems or through raw reads / writes
ZetaNeta has quit [Ping timeout: 252 seconds]
<spv> However, I suspect that there's more to the debian card than just the two partitions
<spv> so I'd probably vote for raw reads / writes
<spv> unless you want to manually play around with bootloaders (which is what I do)
<vector80> wait, let me understand...
hramrach has quit [Ping timeout: 240 seconds]
<kz1> spv: do you know where the hook to enable / disable "reboot bootloader" resides?
<oliv3r> vector80: i'm not gone for vacation :p
<spv> kz1: I'm not certain what you mean by "reboot bootloader"
<spv> would you please elaborate?
<kz1> ex, adb reboot bootloader -> boot fastboot image
<n01> wens: I still have a (huge) problem. It seems like the ISR for NMI in irqchip kind of disable the GIC since when the ISR of regmap is called, all the interrupts are disabled and I get nothing from i2c so everything just halt, no console, no time, nothing
<spv> I do not have enough android experience to say, no.
<kz1> I am trying to enable it so that I don't have to use a UART
<spv> the closest I have is that android would likely be calling the machine_restart command
<spv> and the relevant hooks for that are in the kernel
<spv> in ...
<spv> linux/arch/arm/mach-sunxi/sunxi.c
<spv> assuming you're using the sunxi-devel branch
<spv> from linux-sunxi
<spv> rather, those are the endpoints for the restart, the hook should be whether or not one is allowed to call init 0
<spv> which to me is android specific ...
<spv> or rather, seems so
<spv> anyway, I've said enough about something I don't 100% understand, so I'll leave it at that.
<kz1> in my case the reboot command is called but u-boot does not recognise the "bootloader" flag
<oliv3r> vector80: no need to mount, we ar3e at step 1, test emmc3.bin to see if it loads properly, if you get a u-boot prompt with that, we can continue onwards
<kz1> however "reboot recovery" works
notmart has quit [Quit: notmart terminated!]
<spv> I was unaware that it was possible for u-boot to know to not automatically start the kernel ...
<spv> I would probably end up modifying the boot script I put in /boot/ to implement this ..
<oliv3r> kz1: we do not have fastboot in u-boot yet
<oliv3r> kz1: and as far as i know, adb reboot bootloader should bring up the u-boot console, afaik I don't see why that won't work, as it is simply a reboot command
<oliv3r> i think adb updates the env partition and tells u-boot to not continue booting
<oliv3r> but mind you, if you are booting from nand, and using boot0, boot1, boot.axf and then u-boot; you have to ask allwinner for support as that is their stuff
uwe_ has quit [Ping timeout: 272 seconds]
atiti has joined #linux-sunxi
<kz1> oliv3r: thanks, I'm trying to figure out wtf is going on… but it's doing my head in
<oliv3r> kz1: what is exactly the problem?
<kz1> zero documentation...
ZetaNeta has joined #linux-sunxi
<oliv3r> kz1: chances that you need adb reboot bootloader are minimal
<oliv3r> kz1: you can just reboot and interact with u-boot and tell it to stop
<oliv3r> it should respond to any keypress
<oliv3r> IF you are using allwinner u-boot; they disabled this
<spv> Would that not require UART?
<oliv3r> but you can simply extract the u-boot env partition, flip a few bits and write it back
<kz1> my goal is to do a batch firmware update on multiple devices
<oliv3r> spv: yes, but without uart, what'st he point in rebooting to the bootloader
<oliv3r> the only way to interact witht he bootloader is via u-boot
<spv> I think what kz1 wants to do is add some u-boot scripts to automate the firmware load process
shineworld has quit [Quit: Leaving]
<oliv3r> kz1: make a master SD card, bioot from that, and update partitions that way
<kz1> oliv3r:
<oliv3r> kz1: with an SD card, you can boot a linux kernel, with nand support and do anything you'd expect from within linux
<kz1> IIUC the sdcard has to be removed after the firmware update
<spv> if you want it to boot from nand, yes
<oliv3r> boot from sd card, automatically do the update from scripts, remove sd card
<oliv3r> if you want to update via usb
<oliv3r> you can in theory
<oliv3r> but then you have to force your board into FEL mode
<kz1> so I have this awesome situation where I cannot boot from and sdcard, cannot get access to fastboot via console and cannot use phoenixsuit
<oliv3r> you currently cannot do fastboot or load u-boot and update via usb
<oliv3r> u-boot has zero usb support
<oliv3r> even if you use arokux's patches and add usb support
<spv> kz1: which part of the system are you trying to upgrade?
<oliv3r> which would be host mode, not client mode, our u-boot has no nand driver
<oliv3r> kz1: if your really adventerous, you could try; boot android normally, modify the boot and env partitions that forces an tftp boot, and boot/update via tftp
<kz1> :-) I also don't have tftp support
<kz1> If I break it down I have only usb support
<spv> FEL mode it is?
<kz1> specifically for batch updates
<spv> which is anti-FEL
<oliv3r> and u-boot doesn't do USB
<oliv3r> so you only can do anything with FEL
<wens> n01: that should be expected. GIC ISR calls NMI ISR
<oliv3r> kz1: and in order to use FEL mode, you have to forcefully enable fel
<kz1> Maybe I should ask a completely different question?
<kz1> Can anyone think of the most efficient method for enabling batch firmware updates over usb on linux?
<wens> n01: you may need to split the I2C stuff into ISR bottom half (workqueue, etc)
<spv> kz1: you mean you have like 100 devices connected over USB to a single computer, and want to upgrade them all?
<n01> wens: i2c irq is already setup with a threaded irq
<kz1> n01: yep
ZetaNeta has quit [Ping timeout: 264 seconds]
<oliv3r> kz1: i wonder how much data you can overwrite on the active partition, i bet you run into something and get a crash at somepoint
<oliv3r> kz1: i don't think you can
<oliv3r> kz1: not right now; we need some patches to u-boot to enable that
<spv> it might be possible to kill everything except some bare essentials, transfer those essentials into a tmpfs chroot jail, and then modify the nand?
<oliv3r> kz1: the onyl way I see that happening, is inserting an SD card that forces the device into FEL mode; and then do everything via FEL
<spv> naturally worried about init ... but that mostly resides in memory
<oliv3r> but if you have to go through that trouble, you might aswell use a master SD card
<n01> I have: GIC ISR -> NMI ISR -> handle_level_irq ISR, then _not_ in interrupt context REGMAP-IRQ ISR that tries to access to i2c but no interrupt is received than -> deadlock on wait_*
<spv> looks like kz1 may not have access to such?
<oliv3r> no sd slot?
<oliv3r> no fel button; not possible :p
<oliv3r> what you want can not be done at this time
<oliv3r> if someone where to pickup the MTD stuff and complete that, then yes, fastboot should be possilbe
<wens> n01: I think threaded irq still has a small part with interrupts disabled?
<spv> I was actually going to work on the MTD stuff ...
<oliv3r> spv: slapin was going to spend his winter hollidays too on it
<oliv3r> sad thing is, many commercial people want it, none want to sponsor a dev for it :)
<wens> when I learned basic kernel programming, there wasn't threaded irqs though
<kz1> I can contribute time and/or money to the MTD code
<n01> wens: need to check this but the i2c driver is well tested (i2c-mv64xxx) :(
enrico_ has joined #linux-sunxi
<spv> wens: you are correct, but it's only the part that disables the interrupt and schedules a softirq
ZetaNeta has joined #linux-sunxi
<n01> wens: now request_irq() is alias for request_threaded_irq()
<oliv3r> kz1: well that's a grand step forward
<wens> i see
<oliv3r> kz1: probably best to send a message to our mailling list informing that you are willing to sponsor a dev (or two) to work on mtd support for u-boot + kernel)
<wens> no wonder there are so many kernel threads
<kz1> ok
netlynx has joined #linux-sunxi
ZetaNeta has quit [Ping timeout: 272 seconds]
uwe_ has joined #linux-sunxi
<wens> n01: do you plan on doing all the functions in AXP209?
<n01> wens: at least I want to have irqchip for NMI and MFD
<n01> now I'm using a simple input driver for power button to test the irq chain
<wens> I'm more interested in readouts and maybe power off
ZetaNeta has joined #linux-sunxi
<wens> is there any way to use magic-sysrq over serial console?
<oliv3r> wens: if you enable it in your kernel; yes
<wens> oliv3r: but how do you send the sequence?
<wens> ah, found the sequence for screen
<spv> OMG IT'S SO USEFUL
<spv> especially when you have terrible bugs that break the system
<spv> break + g gives you prompt iirc
<spv> err, for kgdb
kivutar has joined #linux-sunxi
<wens> you can get a stack trace, among other things
<spv> also the set of all things that are on the scheduler's queue
<wens> n01: how about getting a stack trace :)
<n01> wens: no interrupts -> no serial console
<wens> :(
<spv> magic sysrq is NMI
<spv> iirc
<n01> let's try :)
<spv> now that I think about it ... I have no idea how that would be implemented ...
<spv> maybe it's not NMI ...
wolfy has joined #linux-sunxi
<n01> nop, nothing with sysrq ... back to the old printk for debugging ;)
<spv> oliv3r: does the u-boot-sunxi repo include the BSP code from AW?
<spv> n01: apologies :(
pwhalen has quit [Ping timeout: 245 seconds]
VargaD_ has joined #linux-sunxi
<wens> wonder if arokux is coming back?
VargaD has quit [Ping timeout: 252 seconds]
VargaD_ is now known as VargaD
<rm> under some different silly nickname! you never know!
<wens> what?
<libv> rm: what was silly about te nickname "arokux"?
<rm> I am just making fun of it due to the fact he was so insistent about not revealing his real name
<rm> not even in patches etc
<rm> if not for that, then nothing I guess
<libv> i think he contributed quite a bit, so it is sad to see that he left
<n01> I don't think he left forever
<libv> n01: since he is in uni, he could be busy with exams
<n01> I think he was also busy with protesting or something (ukraine iirc)
<rm> orz
<libv> oh, was that arokux? wasn't that slapin?
<n01> hummm, I could be wrong
<JohnDoe_71Rus> busy for 15 days...
<wens> anyway, doing some research on usb otg and phys
<oliv3r> spv: what do you mean?
<oliv3r> libv: i think he's some sub teacher; he was rather old 30+ i think
<spv> so, when we received the Cubieboards / Cubietrucks
<spv> there was software installed on the nand already
<spv> particularly, Android
<spv> with U-boot
<oliv3r> spv: yeah that's lichee kernel with lichee u-boot
<oliv3r> spv: that u-boot doesn't do memory init etc
<spv> eww
<spv> what does?
<oliv3r> boot0/boot1
<oliv3r> they use u-boot to load the kernel
torindel has joined #linux-sunxi
kivutar has quit [Read error: Operation timed out]
kz1_ has joined #linux-sunxi
<vector80> oliv3r:
Nyuutwo_ has joined #linux-sunxi
<vector80> oliv3r: I am sorry, I was messing up with SATA now I am back
<vector80> now, where should I write this emmc.bin ?
diego has joined #linux-sunxi
diego is now known as Guest88545
torindel_ has joined #linux-sunxi
popolon_ has joined #linux-sunxi
andoma_ has joined #linux-sunxi
utente_ has joined #linux-sunxi
speakman_ has joined #linux-sunxi
speakman_ has quit [Changing host]
speakman_ has joined #linux-sunxi
torindel has quit [Ping timeout: 272 seconds]
popolon has quit [Ping timeout: 272 seconds]
panda84kde has quit [Ping timeout: 272 seconds]
slopez_ has quit [Ping timeout: 272 seconds]
Nyuutwo has quit [Ping timeout: 272 seconds]
utente has quit [Ping timeout: 272 seconds]
kz1 has quit [Ping timeout: 272 seconds]
speakman has quit [Ping timeout: 272 seconds]
andoma has quit [Ping timeout: 272 seconds]
slopez has joined #linux-sunxi
kz1_ is now known as kz1
AreaScout has joined #linux-sunxi
<oliv3r> vector80: dd if=emmc.bin of=/dev/mmcblk0
<oliv3r> seek=8 bs=1024
<vector80> ok wait
<oliv3r> standard stuff
quink_ has joined #linux-sunxi
quink_ has quit [Quit: Leaving]
torindel_ is now known as torindel
<spv> Are we certain the emmc.bin doesn't include 8KiB of padding at the front?
cajg has quit [Ping timeout: 240 seconds]
<vector80> yes thats my question also
cajg has joined #linux-sunxi
hero100 has joined #linux-sunxi
<spv> where was emmc.bin from? did you just DD from a mmcblk0 device?
<spv> you mentioned you had that Debian SD card
<vector80> no oliv3r has made it
<spv> oh kk
<spv> then he'd know
<spv> he / she
<vector80> ok, I wrote it toeMMC
<vector80> now reboot ?
<spv> probably should run sync
<vector80> sync is ok too
<vector80> take out SD card and try to boot?
<vector80> I will do it
<spv> I vote yes, even though I only have a little context
<vector80> well, it seems it is working
<vector80> of course not booting
<spv> but do you get to U-boot?
<vector80> yes
<spv> that's a plus
<vector80> I get uboot console
<vector80> what next ?
<oliv3r> vector80: whatever you want; it means that the patch works
<oliv3r> vector80: meaning you can now boot from emmc with u-boot
<oliv3r> you can now make partitions on emmc
<oliv3r> copy debian or what not
<oliv3r> anything goes
<oliv3r> i'll commit your changes
<oliv3r> also we probably want to update your fex
<vector80> Alright, thank you very much again
<oliv3r> i'll push this board.cfg etc to u-boot repository now
<vector80> ok, but I think I must give you my log first
<vector80> pls wait, need to reboot
vector80 has quit [Remote host closed the connection]
<oliv3r> windows and its reboots
<spv> wait, booting from eMMC didn't work?
<oliv3r> spv: i thin it did; he got u-boot console
<oliv3r> which means SPL loads u-boot from emmc fine
<spv> yes it does
<oliv3r> hence, u-boot is working normally
<spv> what was broken?
<oliv3r> nothing broken
<oliv3r> new board that needed supporting
<oliv3r> and vector can't do it himself
<spv> Allwinner made a new board?
<oliv3r> but since we have zero boards with support for sdc2 booting (emmc)
<spv> oh, I see
<spv> sdc2
<oliv3r> spv: na his company made it itnernally
<spv> vector80: how much volume did your company have to come up with for Allwinner to take you seriously enough to sell A10s or A20s?
andoma_ is now known as andoma
<oliv3r> he's not here right now :)
<spv> oh
<spv> whoops
<oliv3r> but i think witts designed the base of the board
<oliv3r> and they went ontop of that and added their own propriatery stuff
<spv> At least working with A20's has provided me with some extra stuff to do over winter holidays
<spv> granted it's essentially wasted work.
<Wizzup> why wasted?
<oliv3r> spv: what is wasted wsork
<oliv3r> i broke my olimex a20 :(
<spv> well, I feel like the A10's and A20's won't be around forever
<spv> the IP that is used in them will be thrown out for something sensible
<oliv3r> spv: a10 IP is used in A20
<oliv3r> a20 ip is used in a23 and possibly beyond
<libv> spv: this is true for _all_ hw
<wens> oliv3r: how?
<spv> And a lot of devices are going to use allwinner?
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<oliv3r> mind you; allwinner is a for profit company, they won't change IP just to make it better
<oliv3r> only if its really really needed
<spv> same goes for all
<oliv3r> spv: the A10 was once the most widely used SoC
<oliv3r> only the quad RK took their leading spot
<spv> that I did not know.
<spv> quad RK?
<oliv3r> the A23 looks like it can compete nicely with budget stuff
<wens> Rockchip RK3188
<wens> quad A9 i think
ZetaNeta has quit [Ping timeout: 272 seconds]
<oliv3r> the A23+ will be very similar to a20/a31 and hopefully be very popular again :)
<mnemoc> the rk3188 relies in a hell of a lot of companion chips to do anything
<mnemoc> not even hdmi is native
<spv> A lot of this work could be accelerated by having adequate documentation of the hardware
<spv> why is it so hard to get that?
<oliv3r> because they 'buy' IP and sign nda's
<oliv3r> some docs are available, a lot really if you thinka bout it
<oliv3r> but some core docs are missing (nand controller, memory controller)
<Wizzup> spv: regardless, the allwinner stuff is nice for years to come
<spv> well, that's comforting
<mnemoc> 1) not a priority in allwinner's business model, 2) they don't own all the IPs involved, 3) patent trolls scare the crap out of them
<oliv3r> spv: also, the sunxi socs are atm THE most open soc's around
<spv> (no sarcasm intended, even though it sounded like it)
<oliv3r> spv: name one soc that has an opensource bootloader
<oliv3r> spv: none taken
<Wizzup> isn't u-boot open?
<spv> hmm, don't the TI Sitara processors have open source bootloaders?
<Wizzup> or is that a higher level
<spv> for that matter, all freescale and TI SoCs have full documentation minus graphics hardware
<oliv3r> Wizzup: for a10 etc it is'
<oliv3r> Wizzup: but any other soc that hs their sources out?
<Wizzup> cubox solidrun maybe
<oliv3r> spv: and then, which soc's can be had for 7 USD or less :)
<oliv3r> so a10 willr emain a very interesting target for a while
<spv> All of them if you get enough volume
<oliv3r> and it uses mali :)
vector80 has joined #linux-sunxi
<oliv3r> spv: noway
<vector80> Sorry, I am back
<spv> maybe not the high-end ones
<oliv3r> spv: how much will an exynox set you back :)
<mnemoc> spv: go an make a REAL survey regarding what's available for the different TI chips
<oliv3r> spv: anyway, A10 has been around for 2 years now? and still used today :)
<oliv3r> (olimex lime)
<spv> exynos is also a good deal faster than the Ax
<oliv3r> spv: that it is :)
<mnemoc> try to buy an exynos5
<Wizzup> exynos odroid u3 is 60 dollar
<oliv3r> spv: but cheap vendors don't care about really
<oliv3r> vector80: yes, that's perect; that means u-boot can boot from emmc3
<oliv3r> vector80: now boot your debian sd card, fdisk the partitions, copy all the files I guess to test etc
<vector80> clear
<oliv3r> vector80: it's all really easy from here on out :)
<vector80> Now, I must insert SD card again
<vector80> and change bootargs again
<oliv3r> vector80: in thoery, you can make 2 partitions, same size as what you have on the SD card now; copy all files over (cp -ar)
<vector80> oliv3r: I have confusion about this point
<oliv3r> vector80: how so
<vector80> that debian SD card image is suitable for a 4Gbyte SD card
<vector80> I wrote it to an 8Gbyte SDcard
<oliv3r> doesn't matter
<vector80> when I run df, it shows 4Gbyte only
<oliv3r> fdisk /dev/mmcblk0 (or 1 whatever is your mmc)
<oliv3r> vector80: of coruse
<vector80> and my eMMC is 8Gbyte
<oliv3r> a full 'dd' image, includes the partition table
<vector80> How can I make correct partitioning for my eMMC ?
<oliv3r> so if you dd something, you do a 1:1 copy of everythign
<vector80> yes but I want whole 8Gbyte
<oliv3r> that's why it's usually a bad idea to work with images
<oliv3r> fdisk /dev/mmcblk0
<oliv3r> make your partitions, format them, mount
<vector80> ok wait let me boot debian again
<mnemoc> Wizzup: the odroid u3 is an awesome step, and a reaction to the inexpensive chinese-soc based boards. but feature wise it's far behind the olinuxinos, cubietruck or the radxa rock
<oliv3r> vector80: copy files, it's all really bog standard linux stuff you have to deal with
<Wizzup> agreed.
<Wizzup> btw how is olinuxinos comparwed to cubie
<Wizzup> I may want to get lots of either for work
<Wizzup> they seems... the same,mostly
<Wizzup> seem*
<oliv3r> vector80: fdisk will write a partition table to the start of the disk, your u-boot comes at 8kb into the emmc (which is after the partitiont able) the first partition should start at sector 63 or 2048 (better) and shouldn't influence your bootloader
<oliv3r> Wizzup: olimex is opensource hardware :)
<oliv3r> Wizzup: i want a lime!
<vector80> oliv3r: Ok I will start it from 2048
<Wizzup> oliv3r: sounds good
<Wizzup> I'd need to talk to siome of the olimex dduudes soon then
<oliv3r> vector80: default fdisk will do that
<vector80> oliv3r: now, my problem is, how can I modify loadkernel script to let it get uImage from my SDcard
<vector80> curently, when I do mmcinfo, it only shows my eMMC, it doesn't detect my Sdcard
<vector80> when I do printenv, it shows device=mmc
<vector80> There is no setting / selection for mmc0 or mmc2
<vector80> You mean, boot from Sdcard directly?
<vector80> Let me try it
<vector80> ah yess
<vector80> got your point now
<vector80> so sorry for being so stupid
<vector80> booting debian now..
<vector80> boot complete
<vector80> try fdisk
<vector80> ok, deleted all partitions
<vector80> and then I pressed "n" to create a new partition
<vector80> first sector is 2048
<vector80> what must be the last sector?
<spv> up to you
<spv> the rest of the emmc?
<spv> single-partition system?
<vector80> As I know, for this SDcard, there are two partitions, one is 16Mbyte which holds uboot and kernel and script.bin
<vector80> and the second partition is all remaining
<vector80> Should I do the same?
<spv> up to you
<spv> some people like putting /boot early in a device because they're afraid their boot-loader can't handle addressing beyond some limit
<spv> (actually true for older x86 systems)
<oliv3r> vector80: but 63 was used since the 80's so should work fine :)
<oliv3r> gotta go shopping no :)
<vector80> Here is my partitioning
<vector80> does it look fine ?
<oliv3r> vector80: you can ONLY boot from mmc0 OR from mmc2, that's why you have two u-boots now. one for mmc, one for emmc :)
<vector80> How can I select which one to build during compile sunxi-uboot ?
<oliv3r> vector80: with our u-boot, it is completly up to you; u-boot can read ext4 just fine
<oliv3r> vector80: allwinner u-boot can only read fat32 or rather, allwinner boot0/boot1 cn only read fat32
n01 has quit [Read error: Connection reset by peer]
<oliv3r> hence your first partition needs to be fat32 so it can load u-boot from it
<oliv3r> vector80: when you checkout u-boot, you ill find a file boards.cfg, if you open that ith your text editor
<oliv3r> you see CONFIG_SOMETHING_MMC=2
<oliv3r> change that to =0 (or leave out the entire string, as it defaults to 0) you build a u-boot suitable for mmc
<vector80> very very clear
<vector80> Ok then, did you see my partitioning table above?
<oliv3r> you could copy the line, one with mmc=2, one ith mmc=0 and name the targets differently, but i don't think you'll swap that often
<oliv3r> compile once, keep the file marked somewhere sane for your SD card sounds logical
<oliv3r> vector80: partition table is really up to you and your usecase
<oliv3r> depending on ho the system is used etc
Black_Horseman has quit [Quit: Zwi se logou mou!!!]
<vector80> clear
<vector80> now, can I format these partitions ?
<vector80> such as mkfs.vfat /dev/mmcblk0p1 ?
<oliv3r> vector80: personally, i would have 1 boot partition to hold kernel, script.bin etc, 1 ro only partition for the OS, 1 rw partition to store settings. and in the OS have an aufs like partition with something from tmpfs, so that most write disk access is done to the memory version
<vector80> mkfs.ext4 /dev/mmcblk0p2
<oliv3r> i ouldn't use vfat, it's not needed
<oliv3r> mkfs.ext4 /dev/mmblk0p1 yes
<oliv3r> but look into f2fs
<oliv3r> that's optimized for flash devices like usb, emmc, sd etc
<oliv3r> it's faster and more reliable
<oliv3r> look it up :)
<oliv3r> gtg no for real
<vector80> oliv3r: you are much more experienced than me for such things, pls tell me your best ideas to setup this partitioning
<vector80> f2fs
<vector80> I have never heard about it
<vector80> does debian have its formatting tool ?
wolfy has quit [Ping timeout: 252 seconds]
<vector80> waww this f2fs is doing cleaning in the background ?
<mnemoc> all modern filesystems do continious maintainance, not only of fsck time
<vector80> so, ext4 also can do it, right?
<vector80> but the point is, how can I find its formatting tool for debian..
<spv> now that I think about it, these devices boot a lot faster than what I'm used to.
<vector80> f2fs boot faster ?
<spv> like, 6 seconds to login prompt
<spv> no, just A20's in general
<spv> I'm used to around 90 seconds.
<vector80> hmm, actually, I see login prompt in 6 seconds in android, but not for this debian image, may be it need optimisations...
<vector80> Debian seems to have f2fs-tools package, but apt-get install f2fs-tools fails... May be I need to add to my /etc/apt/sourcees.lists ?
<spv> so how many writes does it take to fubar a microSD card?
<spv> sudo apt-get update?
<vector80> yes, I needed to add deb http://ftp.de.debian.org/debian jessie main to my sources.list file
<vector80> and then apt-get update
<vector80> and now f2fs-tools package is installed succesfully
<vector80> Just for testing, I will format both partitions as ext4 for beginning
<spv> finally figured out why kexec wasn't working
<spv> *phew*
<vector80> Does it look fine now: http://fpaste.org/64966/09245138/
<vector80> ?
n01 has joined #linux-sunxi
<spv> How much do we know about the A7's that are in the AW parts? Are they standard Cortex-A7s, or were they modified?
<hero100> ext4 is not suitable for flash
<spv> use ubifs
<vector80> well I don't have Flash, I have eMMC
<spv> yeah emmc does its own wear leveling and such
<vector80> I am doing a 1:1 copy of all files in SD to my eMMC
<vector80> Lets see what will happen..
<vector80> Did really nobody used an eMMC before ?
<Turl> spv: I use tftp to iterate, just reboot and let uboot load fresh stuff
<spv> yeah kexec is not supported on a20 yet
<spv> tftpboot is likely the right way to go
<spv> alright I'm out for the night ttyl
popolon_ has quit [Quit: #freenode]
popolon has joined #linux-sunxi
popolon has quit [Read error: Connection reset by peer]
popolon has joined #linux-sunxi
nieuwbie has joined #linux-sunxi
<buZz> ben je bi?
<nieuwbie> Hey. What image should I coppy on my sdcard to get something like init ram on my tablet
<nieuwbie> I want to fsck one of corrupted nand partitons.
<buZz> any would do
<buZz> 'like init ram' <- you mean, not mounting the nand
<buZz> afaik no SD image for sunxi automounts nand
<nieuwbie> buZz: Well i dont know the right terminology here. I just want to be able to fsck my nand cause my tablet got stuck into bootloop and by analyzing what uart showed me I suspect its something with nande
<buZz> yeah, so any SD image should do
Guest88545 is now known as panda84kde
<nieuwbie> buZz: any from sunxi-tools/bin or is there another location where I can find other ones?
<buZz> eh, there are SD images in sunxi-tools?
<buZz> i highly doubt that :D
<buZz> this is a good list; http://linux-sunxi.org/Bootable_OS_images
<nieuwbie> buZz: well I found fel-sdboot there
wolfy has joined #linux-sunxi
<buZz> thats a tool, yes, not a images
<buZz> -s
nove has joined #linux-sunxi
<nieuwbie> buZz: thank you.
<buZz> yw :)
<hero100> what does "sunxi" mean?
<gzamboni> a japonese food ? :P im kiding
<gzamboni> its how the Allwinner ICs plataform is called
<gzamboni> sunXi as it can be sun4i, sun5i ...
<gzamboni> check the linux-sunxi.org wiki
<hero100> yeah I know that, just wonder does it has some special meaning
<Wizzup> buZz: well, all gentoo stage 3 work
<buZz> what, no stage 4?
<buZz> tsss
<mnemoc> afaik allwinner people pronounces it separated, as sun-X-i, sun-4-i, ...
nieuwbie has quit [Remote host closed the connection]
nieuwbie has joined #linux-sunxi
paulk-collins has joined #linux-sunxi
<vector80> oliv3r, mnemoc, spv: Thank you very much for your helps. I copied whole rootfs from SDcard to my eMMC, and than, removed SD card and boot again
<vector80> Debian succesfully booted, and running now
<vector80> Unfortunately, I feel great speed issue
<gzamboni> i thought those eMMC were fast
<wolfy> theoretically they are faster than sd
<mnemoc> but a good SD card is far more expensive than a normal emmc of the same capacity...
<mnemoc> btw, the 30c3 talk about REing SD card gives good hints on how livesuit is probably probing the dram on flash
vector80 has quit [Remote host closed the connection]
vector80 has joined #linux-sunxi
hero100 has quit [Quit: Leaving]
<wens> did anyone extract the new kernel from A23 SDK yet?
<wens> I seriously don't want to download the whole thing
<mnemoc> wens: I added a file listing to the same dir in dl.
<mnemoc> but haven't had time to look into it yet
wolfy has quit [Ping timeout: 252 seconds]
rz2k has joined #linux-sunxi
nieuwbie has quit [Quit: ERC Version 5.3 (IRC client for Emacs)]
slapin has quit [*.net *.split]
binaryferret has quit [*.net *.split]
utente_ has quit [Ping timeout: 264 seconds]
n01 has quit [Ping timeout: 240 seconds]
n01 has joined #linux-sunxi
slapin has joined #linux-sunxi
binaryferret has joined #linux-sunxi
<oliv3r> vector80: i'm not much more experienced :p i'm not working for an embedded company :D
<vector80> :)
<vector80> Thas ok also, it was a great achievement for me though
<vector80> Now I am testing was is working, what is not working..
<vector80> For example, none of the GPIO's are working :(
<vector80> Ethernet is working
<vector80> HDMI is working
<vector80> USBs are not working
<vector80> As you know, my fex file was ok for all these in the android
<vector80> I am wondering how to debug each of these..
<vector80> in /sys/class/gpio, export interface is not working
<oliv3r> buZz: nee sorry
<buZz> hopeloos :)
<vector80> oliv3r: Can I git checkout uboot-sunxi now?
fredy has quit [Excess Flood]
<oliv3r> vector80: sure
<oliv3r> vector80: if you compile yourself, you should get an identical emmc.bin :)
<oliv3r> vector80: idaelly, get the sunxi-bsp
<oliv3r> github.com/linux-sunxi/sunxi-bsp
<vector80> and then, compile whole BSP ?
fredy has joined #linux-sunxi
<oliv3r> ./configure; make u-boot
<oliv3r> make linux
<oliv3r> i use those 2
<vector80> in ./configure, I should select a board, right?
<vector80> such as "interra-3"
<vector80> ?
ZetaNeta has joined #linux-sunxi
iamfrankenstein has joined #linux-sunxi
<ZetaNeta> hi
<ZetaNeta> Anyone know a HD camera i can attach to cubie
<ZetaNeta> Interface doesnt matter as long as i can find a ~1m extension cable
<ZetaNeta> All usb webcams arent meant to be used in portable recording, and ofcourse HD on webcam doesnt mean HD
<ZetaNeta> Nyuutwo_, Also, i found a 65ah car bat
<oliv3r> vector80: yes, the bsp should normally pull in sunxi-boards, u-boot and the kernel at the least as submodules
<oliv3r> vector80: http://linux-sunxi.org/BSP
<oliv3r> vector80: ignore the 'hwpack' section
<oliv3r> vector80: actually; that page is crap
<n01> sooo plans for new year's eve?
<oliv3r> mnemoc: how do you get the submodules again on the sunxi-bsp? the wiki is missing those instructions
<oliv3r> git clone sunxi-bsp; then what
<mnemoc> the makefile clones them on-demand
<oliv3r> but if you run configure first, it needs sunxi-boards
<oliv3r> so git clone sunxi-bsp; ./configure board fails?
<mnemoc> ./configure should do what's needed to get the list of boards
<oliv3r> okay, good then
<oliv3r> vector80: ./configure interra-3
<oliv3r> vector80: make (or only make u-boot && make linux)
<mnemoc> it's supposed to be "magic".... but I hate git submodules anyway... they are very painful to maintain
HdkR has quit [Ping timeout: 240 seconds]
<vector80> oliv3r: what is the difference of using BSP, instead of getting uboot and kernel and compile seperately ?
<oliv3r> easier
<mnemoc> it's only automatizing what you normaly do
<vector80> interra-3 invalid board name
<mnemoc> "building your own is not that hard!" (tm) .... against the "download this dd-able image"
<mnemoc> vector80: you need an interra-3.fex in sunxi-boards ;-)
<vector80> Ok, can I give you now?
<oliv3r> i submitted that
<mnemoc> commit it locally, then submit
<vector80> you mean upload via SCP, and how to commit ?
<mnemoc> submodules are bound to a given commit, and so it doesn't get automatically updated :\
<oliv3r> i committed it allready
<vector80> Ok let me upload
<mnemoc> vector80: cd sunxi-boards; git pull --rebase
<vector80> When I do that, I get error, I am trying to understand what it is
<vector80> Let me upload my working fex file to SCP server now
<mnemoc> try: cd sunxi-boards; git rebase origin/master
<vector80> ok wait
<vector80> thank you very much
<vector80> it is cloning now
<vector80> let me see if it compile fine or not
<vector80> I will try compare binaries
<mnemoc> cloning? it should have been cloned already... only changed the HEAD to the current one
<vector80> make: *** No rule to make target `interra-3_config'. Stop.
<mnemoc> that means u-boot doesn't know about your board yet
<vector80> but as I know, oliv3r already added my board to boards.cfg ?
<vector80> 1) git clone git://github.com/linux-sunxi/sunxi-bsp.git
<mnemoc> as I said, the sunxi-bsp uses submodules, submodules are bound to a given commit, not a branch. so new commits are not immediatelly available to the BSP
<vector80> Ok, so, oliv3r's commit is not available for me now, right? Then what do I must do?
<mnemoc> cd u-boot-sunxi; git pull --rebase
<vector80> ok let me do it again
<mnemoc> and if it fails, show me the error
<vector80> I should be stulluk user?
<mnemoc> git rebase origin/sunxi
<mnemoc> vector80: no, unrelated
<vector80> ok , then?
<mnemoc> 16:47:50 < mnemoc> git rebase origin/sunxi
<vector80> yes I did that
<vector80> then, try configure again?
<mnemoc> no, make u-boot
<vector80> ok wait
<mnemoc> but cd .. first
paulk-collins has quit [Ping timeout: 265 seconds]
<oliv3r> commit log i messed up :)
popolon has quit [Ping timeout: 264 seconds]
<vector80> ok, seems it will compile but failed with missing compiler... arm--linux-gnueabihf... My kubuntu is 12.04.3 LTS, and it can't find it via apt-get update
<vector80> Let me try to find how to add it to sources.list
<oliv3r> IF you have an arm hf compiler installed
<oliv3r> check the first line of Makefile
<oliv3r> or
popolon has joined #linux-sunxi
popolon has joined #linux-sunxi
popolon has quit [Changing host]
slapin has quit [Ping timeout: 272 seconds]
<oliv3r> make u-boot CROSS_COMPILE=arm-pc-linux-arm-hf or the like
<vector80> What happens if I don't use a hard float toolchain ?
slapin has joined #linux-sunxi
<vector80> ok found an HF toolchain, let me try that one
<oliv3r> kernel and u-boot do not use floats' so no problem
<oliv3r> but to compilee your own ap you may want it :-)
HdkR has joined #linux-sunxi
<vector80> ok compiled u-boot succesfully
<vector80> let me compare it with yours
<vector80> When I say: "make linux" it started to git clone...
<vector80> I fell it will gonna take a lonnggg time
<oliv3r> sure, it will clone linux-sunxi
<oliv3r> e.g. the kernel
<oliv3r> when it's done, maybe cd linux-sunxi; git pull; cd ..
<oliv3r> :)
<mnemoc> git pull doesn't work out of the box because submodules are branch-less
<vector80> clear
<vector80> my compiled uboot is 226.400 bytes
<vector80> however emmc.bin is 258.672bytes
<vector80> this compiled one is for SD card, right?
<vector80> for eMMC, I have to change config..
<oliv3r> vice versa
<oliv3r> but check boards.cfg to be sure
<oliv3r> and it all depends on the compiler what is created
<oliv3r> and my file is u-boot-with-spl.bin
<vector80> oh really
<vector80> I checked u-boot.bin
<vector80> my file is u-boot-sunxi-with-spl.bin
<vector80> that one is more clear to your size, 259.132bytes
<vector80> I should use this bin file?
<vector80> oh my god
<vector80> in boards.cfg, you have selected GMAC
<vector80> but my board have EMAC
<vector80> Can we fix this ?
<mnemoc> make the fix, commit it, git format-patch and git send-email
<Turl> vector80: you can use both on A20
<Turl> make sure you choose the right phy type though
<vector80> yes, I was surprised about that on uboot console
<vector80> When I use Olinuxino uboot, it was displayed EMAC
<vector80> but on your uboot, it was displayed MII
<vector80> ok I make the fixes on the boards.cfg, now, I want to commit and go back to home , it will gonnna take a lonng time, it is very bad rain in istanbul :(
<vector80> Terible traffic jam :(
<vector80> How to commit ?
<oliv3r> vector80: you use GMAC, but you don't have RGMII, so you are usiong MII, e.g. 100 mbit
<vector80> What is FAST_MBUS ?
<mnemoc> [Rumour]
<vector80> Ok, I get cold today, I must go home and take some rest... Thank you soo much for your helps, I hope to continue tomorrow
<Turl> mnemoc: it's like your dream laptop if you ignore that bit ;)
<mnemoc> :p
<mnemoc> charbax has been teasing with that for a while too
<Turl> vector80: I wish it were cold over here :p suffering 36C now
<HdkR> Google/Samsung has been working on ChromeOS support with Exynos 5420 for months now
<HdkR> They've got two boards with the 5420 on it
<HdkR> It should have come out quite a while ago :|
<mnemoc> and they missed the xmas season...
<mnemoc> maybe to keep HP in the game
<HdkR> HP can get punched in the kidney
<mnemoc> Google takes their partnerships seriously
<mnemoc> they will not punch any partner in the kidney
<HdkR> Nobody in their right mind should have a laptop charge from micro-usb
<mnemoc> *g*
<Turl> HdkR: why not? I think that bit is genious :)
<Turl> I mean, you charge tablets already...
<HdkR> A good thing about the Samsung Chromebook is that the chromebook charges in like 2 hours
<mnemoc> tablets don't need as much power as a laptop :p
<Turl> mnemoc: arm/bay trail laptops are basically a glorified tablet
<Turl> a tablet with a laptop :p
<Turl> with a keyboard* fail
<mnemoc> with a huge display to light
<Turl> it's just 1-3 inches bigger than a tablet
<mnemoc> compare the surface of a 7" display with the surface of a 13.3" display
<mnemoc> it's almost 4 times
<Turl> there's 10" tablets
<Turl> ...
<mnemoc> and the tablets go to sleep much quicker than laptops
<mnemoc> Turl: still more than double of the surface
<Turl> mnemoc: the chromebook was 11 anyway
<Turl> 10->11 is like nothing :)
<HdkR> 20" tablet
<HdkR> HP is a bunch of weirdos
<mnemoc> same HP who told Wozniak making personal computers was stupid
<wens> Turl: any suggestions on how to do usb clock properly? arokux's patch splits them between clk and reset controllers
<mnemoc> I prefer VESA mountable computers to plug behind any normal display
<HdkR> One day I'll have enough desk space for a monitor
<mnemoc> Turl: does u-boot-sunxi/sunxi support CT's PHY ?
<HdkR> haha
<wens> mnemoc: it does # does u-boot-sunxi/sunxi support CT's PHY ?
<mnemoc> damn dnsmasq then..... # dhcp BOOTP broadcast 1 BOOTP broadcast 2 ....
<mnemoc> and nothing in dnsmasq logs about it :<
<Turl> wens: the reset stuff should be registered as reset controller and the clocks as clocks
<Turl> wens: are you going to tackle usb next?
<wens> Turl: but they are in the same register?
<wens> looking at usb otg ATM
<Turl> mnemoc: once you merge the gmac patches, it should. But it wasn't merged last I checked
<mnemoc> ow
<wens> oliv3r pushed them yesterday
<Turl> then it should be good to go
<mnemoc> yesterday... I think my u-boot is from xmas...
<Turl> mnemoc: do you see mii as network?
<Turl> then a rebuild is in order
<mnemoc> MII devices: 'emac'
<Turl> old uboot then
<Turl> get fresh one, then it should say mii0 instead of emac
<mnemoc> old is a very relative term :p
<Turl> outdated*
<Turl> better? :p
<Turl> wens: yeah, they're on the same register
<Turl> wens: that's why the same driver should register both things, so there's no race
<mnemoc> who needs to build things :p
<Turl> J o.O
<mnemoc> for .tar.xz
<Turl> I never use those
<Turl> my tar autodetects them or sth :)
<mnemoc> $ sudo dd if=u-boot-sunxi-cubietruck-20131230T134217-c8bcaed/u-boot-sunxi-with-spl.bin of=/dev/sdb bs=1024 seek=8 conv=sync
<mnemoc> yes!
<mnemoc> DHCP client bound to address 10.42.0.200
<mnemoc> thank you :)
<Turl> :)
<mnemoc> have I mentioned I love this community?
<Turl> mnemoc: we hadn't noticed :p
<wens> crap... usb is a mess... need to do a control module and phys
<mnemoc> Turl: =)
<Turl> wens: notice that usb otg is a different module than the host one
<Turl> wens: mainline supports it via musb
<wens> that's right, but musb only supports the core
<Turl> there was some work to make it work with that on linux-sunxi
<Turl> one of the two modes was working
<wens> mode detection all that stuff is a seperate control module shared by all three controllers
<wens> that is used is access the usb phys
<mnemoc> Turl: just tried | tar -xv and failed :( mine requires -J
<Turl> mnemoc: drop the -
<mnemoc> tar: Archive is compressed. Use -J option
<mnemoc> tar: Error is not recoverable: exiting now
<mnemoc> :p
<mnemoc> smart enough to detect but not to do "the right thing"
<Turl> mnemoc: just tried it, fails the same :p
<Turl> mnemoc: try wget + tar xvf file.tar.xz
<mnemoc> so you were just BSing me
<Turl> mnemoc: haha
<Turl> mnemoc: but bsdtar does the right thing
<Turl> curl...|bsdtar xvf -
jemk has joined #linux-sunxi
<ZetaNeta> Hi, i need a HD camera for a cubie
<ZetaNeta> What you can say?
<ZetaNeta> I mean a "Normal Camera"
<ZetaNeta> Not a webcam
<Turl> and what's the difference?
<Turl> :p
<ZetaNeta> because HD webcams are 1) Not meant for "portable"
<ZetaNeta> 2) HD webcam doesnt related to HD at all
<ZetaNeta> Well, it have quite a big resolution
<Turl> those listed there have at least 720p (HD)
<Turl> some even do fullHD (1080p)
<ZetaNeta> But while webcams are being optimized into lowlight places, and webchats
<ZetaNeta> I need something what cant be damaged by sun XD
<Turl> buy some wifi cctv camera then
<ZetaNeta> and have atleast somekind of being shockproof
<ZetaNeta> Turl, You get how much those cost? How much power they eat?
<ZetaNeta> No... i need something what i can attach to my cubie by a wire XD
<ZetaNeta> i2c for example
<Turl> dunno :) but they're made to watch places outside
<Turl> so they often are rain, storm and whatever-proof
<mnemoc> ZetaNeta: any supported by the kernel you want to use
<ZetaNeta> Turl, This cam is something what i can attach by a long wire to cubie, put in my "specially" prepared pocket, and record my trips for ~12 hours
<Turl> I don't think you'd get any decent video via i2c
<mnemoc> wens: can serverip be passed by the dhcp server?
<ZetaNeta> Any interface. But better if not wireless
<Turl> mnemoc: dhcp can pass the server ip I think, yeah
<Turl> I did it on dnsmasq once I think
<Turl> ZetaNeta: buy a gopro and use that
<ZetaNeta> Turl, Not a go
<ZetaNeta> I wanted to get one... but too much expensive for alot of feature i dont need
<ZetaNeta> And it need like a bagfull of batteries and 64gb sd cards to go for ~12 hours
<ZetaNeta> + it cant play mp3
<ZetaNeta> + You cant control it with a ps3 joystick
<ZetaNeta> I am going to build a cubieboard camera
<Turl> mnemoc: try something like dhcp-boot=pxelinux,server.name,192.168.1.100
<ZetaNeta> But i need to find some HD camera :P
<ZetaNeta> And something to do the microphone
<nove> ZetaNeta, my choice, would be to use a normal handheld camera, with optical zoom, and the availability to transmit the video feed by usb or wifi
<mnemoc> oliv3r: is this expected in 3.10? CPU1: failed to boot: -38
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<Turl> mnemoc: very likely, yes
<mnemoc> good :)
<Turl> mnemoc: you need PSCI or PCSI, I never get the name right
<wens> i get this on my cb2 on mainline, too. but not ct
<Turl> on your kernel
<Turl> and an uboot supporting it
<Turl> which currently requires some extra branches from people
<ZetaNeta> nove, Erm... well.... My setup already cost +200 eur... eheh
<nove> ZetaNeta, to have image quality that was what i would use
<mnemoc> time to start woriking on my heretic hybrid kernel
<ZetaNeta> and i expected that getting a hd camera and a mic would be much cheaper than getting a Cubietruck, ssd and a 65ah car battery....
jemk has quit [Remote host closed the connection]
<nove> ZetaNeta, and how do you plan to record save the video?
jemk has joined #linux-sunxi
paulk-collins has joined #linux-sunxi
<ZetaNeta> nove, ?
<nove> ZetaNeta, from the camera output to a file in the ssd, how ?
<ZetaNeta> nove, Again, i didnt get it.
<ZetaNeta> You mean what will be betwen the ssd and camera, to do the converting and etc?
<ZetaNeta> A cubietruck ofcourse!
<nove> ZetaNeta, you want HD, but Cubietruck can't encode this by software?
<ZetaNeta> why?
<nove> ZetaNeta, you need to use harware encoding, or to have a camera to does that for you
<ZetaNeta> nove, Afaik everything what is not proprietary, can be done by software
<ZetaNeta> and isnt the A1whatevertheygotincubietruck powerful enough?
<nove> ZetaNeta, to encoder h264 1080p video in software in real-time, i think is not
<mnemoc> Turl: sent size: 4 option: 54 server-identifier 10.42.0.254 ..... didn't end up in serverip var :|
<ZetaNeta> And what hw encoding does cubie provide?
<nove> ZetaNeta, you have hardware encoding by the means of the cedar hardware block
<Turl> mnemoc: are you using "tftpboot" ?
<ZetaNeta> crap.. can both cedar and cpu be loaded with encoding?
<mnemoc> Turl: -v
<ZetaNeta> in "parallel"?
<mnemoc> Turl: dhcp and tftp are working, from dnsmasq
<ZetaNeta> Because thats the main thing cubie is for in that project
<nove> ZetaNeta, there is only possible to use it now, by the mean of a closed source binary, that is not in a state not very usable
<ZetaNeta> I may even leave the mp3 functionality, if it prevents the recording
<Turl> mnemoc: run "tftpboot" on uboot shell
<mnemoc> *** ERROR: `serverip' not set
<ZetaNeta> nove, "So crappy..."
<Turl> mnemoc: "tftp"?
<Turl> I never remember the right command
<mnemoc> Turl: same
<Turl> mnemoc: bootp?
<ZetaNeta> well, i guess i need to find a "normal camera", which will stream it to cubie. Right?
<mnemoc> DHCP client bound to address 10.42.0.200
<mnemoc> Turl: I do get dnsip var, but not serverip :|
<mnemoc> also ipaddr is set.... but I want serverip :p
<ZetaNeta> but... will cubie be capable of realtime "receive-store" scheme
<nove> ZetaNeta, the good news, is that the cedar hardware encoder is in the process of starting to be reverse-engineering
<Turl> mnemoc: dhcp?
<ZetaNeta> nove, Will it be done by this spring?
<mnemoc> Turl: dhcp is working, and sets ipaddr, dnsip, gatewayip and netmask
<ZetaNeta> XD
<mnemoc> Turl: but not serverip
<Turl> mnemoc: what did you add on dnsmasq?
<ZetaNeta> Because i actually need this project just for ~2 weeks... after that it will be unlikely to be touched. except the cubietruck, it got too much power to lose.
<mnemoc> Turl: it seems I need to pass "option 66" for serverip
<mnemoc> server-identifier is 54
<Turl> mnemoc: is 10.42.0.254 the ip of dnsmasq?
<mnemoc> yes
<mnemoc> the primary ip of eth0
<Turl> mnemoc: then you can just use dhcp-boot=filename
<mnemoc> (my uplink is wlan0)
<nove> ZetaNeta, well it would be go if done by then, but is more that simples reverse-engineering, is also need to have it integrated in the right places ffmpeg/gstreamer,, this will take more time
<ZetaNeta> /zeta is mad
<mnemoc> Turl: how will u-boot/dhcp know where to load the uImage ?
<mnemoc> Turl: also, what's the usual address for the initramfs?
Wizzup has quit [Ping timeout: 272 seconds]
<Turl> mnemoc: I never use the automagic way, dunno
<nove> ZetaNeta, how about dashboard cameras?
<ZetaNeta> nove, Can they stream it to cubie?
<ZetaNeta> and again, will cubie be able to "receive and save it" fast enough?
<nove> ZetaNeta, no idea, maybe there are some that can
<ZetaNeta> i hardly imagine a dash cam which can
<mnemoc> Turl: these knowledge in the wiki is very disperse... but that page doesn't talk about automatic serverip from dhcp or about where to load the initramfs :p
<Turl> mnemoc: I never use a separate ramdisk
<Turl> but you should be able to put it anywhere in ram
<Turl> may need to do the ffffff trick so uboot doesn't relocate it if stuff explodes
<mnemoc> what's the diff between `env set var val` and `setenv var val` ?
<Turl> mnemoc: one requires you to type one char less
<mnemoc> sent size: 12 option: 66 tftp-server 10.42.0.254
<mnemoc> and still serverip not set :\
<Turl> mnemoc: can't you just hardcode it?
<mnemoc> sure I can
<Turl> env set serverip 10.42.0.254; saveenv :) and then you can do sth productive :p
<mnemoc> just wanted to set up the env "correctly"
<Turl> mnemoc: I hardcode a mac and ip on my env :p I don't even use dhcp
<mnemoc> setenv serverip ${gatewayip} is better than hardcoded
<mnemoc> but still not "correct"
paulk-collins has quit [Remote host closed the connection]
<keebler> If anyone mines dogecoin, I'll pay a 25K dogecoin bounty to anyone who and save me the time and compile a .deb binary for the allwinner A10 and A20.
<keebler> and/can^
<keebler> arg
<keebler> for dogecoin-qt
<keebler> That's a whole $6!
<keebler> :)
enrico_ has quit [Quit: Bye]
<eagles0513875> hey guys
<eagles0513875> can anyone give me the revision for when mmc was pushed to the repos please so i can pass that onto buildroot as I am having issues atm booting using buildroot as i get an mmc error
<eagles0513875> oliv3r: hey
<Turl> keebler: what deb? debian already supports the arch, so for most of them you can find them on the official repos
eagles0513875 has quit [Changing host]
eagles0513875 has joined #linux-sunxi
eagles0513875 has joined #linux-sunxi
<keebler> A .deb for the dogecoin-qt?
eagles0513875 is now known as eagles051387
<Turl> ah
<Turl> keebler: is it even supported on linux? last I checked it only had mac and windows support
<keebler> Well, it's just a variation of the litecoin client, and it has a -git for archlinux. https://aur.archlinux.org/packages/dogecoin-qt-git/https://aur.archlinux.org/packages/dogecoin-qt-git/
<keebler> oops
paulk-collins has joined #linux-sunxi
ssvb has joined #linux-sunxi
<keebler> nvm
<keebler> just need to compile it for sunxi
<Turl> no source packages there, just i386 bins
<keebler> yeah. I only meant it as an example of something compiled for debian already.
nove has quit [Quit: nove]
<keebler> seems he pulled the source from github.
panda84kde has quit [Ping timeout: 260 seconds]
Quarx has quit [Read error: Connection reset by peer]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
_massi_ has quit [Remote host closed the connection]
n01 has quit [Ping timeout: 272 seconds]
sholom has joined #linux-sunxi
sholom has quit [Remote host closed the connection]
FR^2 has joined #linux-sunxi
jemk has quit [Read error: Operation timed out]
<mnemoc> does a dtb-based kernel consider bootarguments given by u-boot, or only those within the .dtb?
<mripard> only those in /chosen/bootargs
<mnemoc> thanks
<mripard> the address of the DTB and the address of the bootargs are both stored in r2
<mripard> so they are mutually exclusive
<mnemoc> i was fearing that :p
<mnemoc> btw, what's the usual address to load the initramfs?
<mnemoc> or everyone embeds it in the kernel?
jemk has joined #linux-sunxi
<mripard> I use 0x43000000
<mripard> but there's no real requirements iirc, or "usual" address
<mnemoc> so the same as traditionally used for script.bin
<mripard> yeah, but that would work just the same at 44 or 49, or whatever.
<mnemoc> recently I saw a commit removing /chosen/bootargs from the sunxi dts files... what's the rationale of that if it's the only way of passing bootargs like the initial speed of ttyS0 and root= ?
<mripard> it's set by u-boot
<mripard> u-boot will patch the DTB just before booting the kernel
<mnemoc> i see
<mripard> so the one provided by default always gets overridden
<mnemoc> but our u-boot doesn't do dtb yet... or it does?
<mnemoc> unless I add console=ttyS0,115200 to chosen/bootargs in the .dts my kernel gets silent after [ 0.372093] 1c28000.serial: ttyS0 at MMIO 0x1c28000 (irq = 33) is a U6_16550A
ZetaNeta has quit [Ping timeout: 252 seconds]
<mripard> it does for all but the A31
<mripard> for like a year or so
<mnemoc> .oO
<mnemoc> then why is mine not hotpatching chosen/bootargs .... uhm
<mripard> you have bootargs filled in u-boot ?
<mnemoc> so run setargs; bootm should do...
<mnemoc> bootargs=console=ttyS0,115200 root=/dev/mmcblk0p1 loglevel=8 panic=10
<mnemoc> weee!
ssvb has quit [Ping timeout: 252 seconds]
<mnemoc> I was missing bootpath to make the magic work
ZetaNeta has joined #linux-sunxi
ssvb has joined #linux-sunxi
<ZetaNeta> hi
<ZetaNeta> anyone heard my question, or my irc timed out before anyone could?
<mnemoc> you were out for 15 minutes until 15 minutes ago
<oliv3r> wens: btw, usb-otg was a mentor graphics IP; it should be somewhat supported
tomboy64 has joined #linux-sunxi
tomboy65 has quit [Ping timeout: 265 seconds]
<ZetaNeta> oliv3r, I choose to go with a Sony AS15
<ZetaNeta> And use cubie to "receive and store"
<ZetaNeta> And i got 2 choices
<ZetaNeta> Wifi, in which il be a brave hero breaking through probably proprietary protocols
<ZetaNeta> and microsd
<ZetaNeta> i can either have to take it out every few hours, insert it into cubie, press a button on a ps3 joystick, and wait for few minutes till the mp3 starts playing michael jackson
<ZetaNeta> either, somehow find something what can emulate a hispeed class microsd card, to transparently write to cubies ssd
<ZetaNeta> like you know those "MicroSD breakout" cards which can give you serial over microsd.
<ZetaNeta> But here, i need to provide a "Virtual MicroSD" to the camera
<ZetaNeta> know such a device?
<mnemoc> afaik it's not hard to fake SD cards programatically
<mnemoc> Turl: congratulations mr. sunxi clock official maintainer!
techn_ has quit [Ping timeout: 272 seconds]
techn_ has joined #linux-sunxi
techn__ has joined #linux-sunxi
techn_ has quit [Ping timeout: 265 seconds]
vector80 has quit [Quit: Konversation terminated!]
vector80 has joined #linux-sunxi
kivutar has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
<vector80> hi there... How can I test & compare my SD card and eMMC read/write speeds ?
rz2k has quit [Read error: Connection reset by peer]
<mnemoc> if you have android you can try comparing the results of 0xbenchmark when running from both
<mnemoc> that will give you a nice overall comparison
<vector80> mnemoc: thank you... I made such test: I boot debian from one Class10 SD card, and then tried to copy a 800+MByte file over SCP.. Speed was around 2.9Mbyte / sec
<vector80> And then, I boot from eMMC, and copied same file over SCP, speed was exactly same, 2.9Mbyte/sec
<vector80> But when I check the datasheet of eMMC, it mentiones around 11Mbyte/sec for write speed
<vector80> Is it possible that this can be caused by driver ?
ganbold_ has quit [Ping timeout: 245 seconds]
<mnemoc> be sure to test with cpufreq in performance
<vector80> what is that mean?
<vector80> It doesn't do this automatically?
<mnemoc> set the cpufreq governor to performance
<mnemoc> it's probably using ondemand
<vector80> Via console command ?
<mnemoc> and working at very low freq
<mnemoc> see the Cpufreq page on the wiki for details
<vector80> Are these minimum speeds?
<mnemoc> 120MHz sounds pretty low, but I bet it can go lower
<vector80> Should I install cpupower package?
<mnemoc> just set the governor to performance
<vector80> in linux-sunxi, I see this page: http://linux-sunxi.org/Cpufreq
<mnemoc> only set both governors to performance
<mnemoc> you don't need to tune ondemand for benchmarking
<vector80> ok, it seems it is 912Mhz now, but write speed is still 2.9Mbps
<vector80> Let me try other PC
<vector80> Tried a different PC, result is same
<vector80> speed never exceed 3.0Mbps
<vector80> 3 days ago, we have checked schematics and A20 datasheet with oliv3r, and A20 doesn't have 8line interface for any SD
<vector80> Actually our eMMC is connected to A20 via 8line, but 4 pins seems reserved
<vector80> Does 4bit operation reduce the speed dramatically ?
<eagles051387> hey does anyone know the revision in git that introduces support for mmc?
<andhe> http://pastebin.com/Y2m2qweg <-- while building sunxi-devel + wens wifi.... is there some Kconfig select/depends missing?
<Turl> mnemoc :)
iamfrankenstein has quit [Read error: Connection reset by peer]
<mnemoc> Turl: what are you smiling at? :p
jemk has quit [Remote host closed the connection]
<mnemoc> andhe: looking at http://dl.linux-sunxi.org/nightly/linux-sunxi/linux-sunxi-devel/ it seems sunxi_defconfig builds just fine
<andhe> mnemoc: maybe it already has CONFIG_RESET_CONTROLLER enabled?
<mnemoc> the nightlies reset to defconfig every time
<Turl> mnemoc :p
<mnemoc> o_O
Gerwin_J has joined #linux-sunxi
<Turl> O.o
* mnemoc scratches his head
<oliv3r> atleast it's not your balls :p
<Turl> lol
<oliv3r> vector80: you really have to ask allwinner about the 4 lines, but taht's 'mmc+' and i don't think a10 or a20 does that at all
<mnemoc> :)
<vector80> oliv3r: thank you
<mnemoc> oliv3r: btw, tagged sunxi-v3.10.25-r0
<mnemoc> oliv3r: if you can merge the rest of sunxi-next...
<vector80> I will try to add a tmpfs to my fstab, lets see if the performance will be different or not
<oliv3r> vector80: but we dont' have that infomration; only allwinner does :)
<oliv3r> tmpfs is an FS in 'ram' :p
<oliv3r> vector80: format to f2fs and see if that's better
<vector80> ok, I will try f2fs
<vector80> but before that
<vector80> I am really confused now
<mnemoc> tmpfs will tell you the max your dram/cpu can give
<vector80> I added this line to my fstab: none /tmp tmpfs defaults 0 0
<vector80> and rebooted
<oliv3r> tmpfs is a fs in ram
<vector80> after reboot, when I run df -k, I get this: none 463400 0 463400 0% /tmp
<vector80> But my DRAM is 1Gbyte ???
<vector80> Where is the rest ?
<mnemoc> bad u-boot config?
<vector80> When I do "free", I get 926.804
<mnemoc> look at the top of dmesg
<vector80> wait, let me give you
<mnemoc> also see what u-boot tells when greeting
<oliv3r> also, doesn't tmpfs only do half the ram by default
<oliv3r> would be silly to do all of it :)
<oliv3r> you could run out of ram :)
<mnemoc> Memory: 448MB 512MB = 960MB total
Gerwin_J has quit [Quit: Gerwin_J]
<oliv3r> strange way of writing it though; never seen it like that before
<mnemoc> normally people doesn't have chunks removed from the middle of their ram :p
<oliv3r> yeah what's causing that
<mnemoc> mali
<oliv3r> ahh
<oliv3r> never noticed that
<oliv3r> but booted mainline mostly
<mnemoc> [ 0.000000] Machine: sun7i
<mnemoc> [ 0.000000] Memory cut off:
<mnemoc> [ 0.000000] MALI : 0x5c000000 - 0x5fffffff ( 64 MB)
<mnemoc> U-Boot SPL 2014.01-rc1-09147-g5cf4ad2-dirty (Dec 24 2013 - 15:40:05)
<mnemoc> Board: Interra-3
<mnemoc> DRAM: 1024 MiB
<mnemoc> vector80: ---^ as I said, wrong u-boot configuration
<oliv3r> how is it wrong? he has 960 MB total, sounds about right does it not? wasn't it a 1GB board?
<andhe> mripard: ping
<vector80> Thank you very much,but what is wrong, I also couldn't understand?
<mnemoc> so you have 1GB or 2GB?
<vector80> 1Gb
<vector80> 1Gbyte
<mnemoc> :\
<mnemoc> so everything is right
<vector80> Where is the remaining RAM ?
<vector80> I mean, where is 448 and where is 512 ?
<mnemoc> tmpfs will now allocate more than half unless you tell it so
<buZz> not*
<vector80> how to tell it allocate more than half ?
<buZz> df -h
<vector80> wait
<buZz> you will see size of tmpfs mounts
<vector80> yes, what is that devtmpfs ?
<buZz> 453M
<buZz> thats a bit big :P
<oliv3r> vector80: tmpfs is a filesystem completly in RAM
<oliv3r> vector80: so when you store a file in tmpfs (say /tmp) then when you power off, its gone
<oliv3r> vector80: it is 1/2 your ram by default, but it doesn't USE all of that
<vector80> oliv3r: I know it, since my eMMC is too slow than my expectation, I thought I can use tmpfs for transferring big files over network.
<oliv3r> if you save 1 MiB in your tmpfs, you loose 1 MiB of your RAM (until you remove that file)
<oliv3r> vector80: how do you get files into tmpfs? loading it FROM emmc :p
<oliv3r> vector80: you should compare speed between emmc and mmc
<oliv3r> it SHOULD be the same really
<oliv3r> or close enough
<vector80> Yes, they are both 2.9Mbps
<oliv3r> since afterall; emmc and mmc IS the same, it's just the package is different (card vs chip)
<vector80> This is really disappointing
<oliv3r> ask allwinner though about the 8 bit thing, maybe the controller secretly does support it :)
<vector80> When I use eMMC, I thought I could get the speeds written on the datasheet
<vector80> Have a look at the speeds written here
<vector80> I will ask to allwinner about it
<vector80> Currently, it seems NAND is faster than eMMC ?
iamfrankenstein has joined #linux-sunxi
<oliv3r> i get about 5mb i think from nand
<vector80> yes, thats too strange
<vector80> I tried to copy a file to /tmp
<vector80> and I get same speed, around 3.0Mbps
<oliv3r> that's probably best speed test
<oliv3r> let me read the datasheet
<oliv3r> But i'm guessing this is a limitation of 4bit
<vector80> I feel something wrong with my board
<oliv3r> who designed the 8 bit emmc interconnect? wits or you guys?
<vector80> I just wrote it to my technical specification, and they connected to A20
<oliv3r> vector80: atleast the emmc does TRIM
<oliv3r> that's a big plus
<oliv3r> not sure if our mmc driver supports trim though :)
<vector80> oliv3r: I currently don't have a NAND based A20 board here, can you try FTP to your board, and transfer a file, and let me know your speed?
<oliv3r> i have nothing connected here right now :)
<oliv3r> mnemoc: is biooting his atm though :)
<mnemoc> eh?
<mnemoc> I don't have nand in 3.10
<vector80> mnemoc: Are you using MMC ?
<mnemoc> booting from mmc, yes... but my kernel doesn't have a driver for it yet
<vector80> cubie ?
<mnemoc> cubietruck
<oliv3r> vector80: emmc MAY need some driver additions? they talk about iNAND etc
<vector80> oliv3r: I am not thinking about eMMC now, I am thinking, why even from DRAM, transfer speed doesn't exceed 3Mbps ?
<oliv3r> tmpfs -> lan?
<vector80> yes
<vector80> I am trying to say that
<oliv3r> performance governor?
pirea has joined #linux-sunxi
<vector80> around 912Mhz
<vector80> after reboot, should I change again?
<oliv3r> sustained read, 70 MB/s
<oliv3r> maybe it's a typo and should be Mb/s :)
<oliv3r> megabit
<vector80> WAWWW
<vector80> oh sorry
<vector80> So, I feel something wrong with my board
<oliv3r> yeah might be
<oliv3r> but ask witts/allwinner
<oliv3r> since witts designed your board
<vector80> I think you didn't work with them before :)
<vector80> Anyway :)
<vector80> When I increasedto performance, I get about 5Mbps
<vector80> 5Mbyte/sec
<vector80> but speed is still 912Mhz, how to increase to 1.2Ghz ?
<oliv3r> you mean over clock it?: )
<morfoh> mnemoc: you tested my last initrd yet? ;)
<vector80> yes
<vector80> actually, 1.2Ghz should be within the specs,
<vector80> not overclock ?
<vector80> am I wrong ?
<oliv3r> no, overclock :)
<oliv3r> you MAY get it to run at 1.1 GHz if you have a really good sample
<mnemoc> morfoh: Failed to execute /init
<morfoh> :/
<vector80> oliv3r: what I understand from you is that, A20 does never work at 1.2Ghz ?
<vector80> Its max speed is about 1Ghz ?
andhe has quit [Read error: Connection reset by peer]
eagles051387 is now known as eagles0513875
pirea has quit [Read error: Connection reset by peer]
ah has joined #linux-sunxi
ah is now known as andhe
Gerwin_J has joined #linux-sunxi
iamfrankenstein has quit [Ping timeout: 240 seconds]
AreaScout has quit []
popolon has quit [Quit: Quitte]
iamfrankenstein has joined #linux-sunxi
bsdfox has quit [Ping timeout: 264 seconds]
FDCX has joined #linux-sunxi
netlynx has quit [Quit: Leaving]
<oliv3r> vector80: a10 could run at 1102 MHz
<oliv3r> vector80: A20 couldn't hit that, it's slightly slower
<mnemoc> Unable to handle kernel paging request at virtual address f1c00024 .... narf
<oliv3r> mnemoc: what are you hacking on!
<mnemoc> trying to read the SC on 3.10
<oliv3r> SmartCard driver?
<oliv3r> yeah you may be missing a lot of patches
<mnemoc> nah, system control
<oliv3r> ohh
<oliv3r> what's that?
<oliv3r> mind you, 3.10 isn't heavily tested!
vector80 has quit [Remote host closed the connection]
<mnemoc> where the chip_id lives
<oliv3r> Ohhh, sramc
<oliv3r> yeah
<oliv3r> soc_detect stuff :)
<mnemoc> yup
<mnemoc> part of the quest of letting a (mostly) script.bin powered 3.10 generate a .dts
<oliv3r> mnemoc: you should talk to mripard about that too
<oliv3r> something babelfish