<hramrach>
If there were really two of them where would the other one get the link? There is only one physical port
<mripard>
wens: and what if you use gmac all the way?
<mripard>
maybe it's just the phy that needs to be reset
<hramrach>
how do you use gmac?
<wens>
mripard: can't transmit packets
<hramrach>
it works allright with 3.4 booting from SD with no net in u-boot and never got it working in mainline
<wens>
mripard: dma still needs to be worked out
<mripard>
wens: I thought you said it was working in u-boot
<wens>
mripard: it works in u-boot, not in kernel
<hramrach>
it is working in u-boot
<hramrach>
I download hte kernel over gmac
apo has quit [Ping timeout: 245 seconds]
<mripard>
so. if you are using both in u-boot and in the kernel, it doesn't work either?
apo has joined #linux-sunxi
<hramrach>
no
<wens>
same here
<mripard>
wens: you told me gmac was working in the kernel 2 minutes ago...
<wens>
on cb2 stmmac only sees one phy though
<wens>
mripard: did I?
paulk-collins has joined #linux-sunxi
<hramrach>
not clear enough
<hramrach>
wens | mripard: if i use emac in u-boot, then gmac in kernel, or vice versa, the phy in the kernel doesn't work
<hramrach>
seems to imply that it does work if the drivers are same in both
<mripard>
ok, so, gmac doesn't work at all
<mripard>
you're using the st-micro one, right?
apo has quit [Read error: Connection reset by peer]
<wens>
sorry for the consufion
<wens>
mripard: that's right
apo has joined #linux-sunxi
<hramrach>
mripard: see the link
<wens>
so is hramrach
<mripard>
have you tried the various different compatible it has?
<hramrach>
just picked some random driver that seemed to match what is in u-boot
<hramrach>
no. how do I pick which to use?
<hramrach>
tried like 2 of them
<mripard>
I have no idea, you're the one that looked into it :)
<hramrach>
also the driver has way more parameters than what is set in u-boot
<mripard>
I just swa that it has several of them
<hramrach>
like DT sets IO size for start
<mripard>
with a version of the IP
<mripard>
that enables a few stuff
<wens>
mripard: compatibles fall into 2 groups, none of them work
<hramrach>
or disables, yes
<mripard>
maybe it's just that you don't have the right set of features for the IP
<hramrach>
yes, there are quite a few
<hramrach>
but wth the u-boot driver uses
<mripard>
you can't compare a u-boot driver with a kernel ones
<wens>
according to the driver, gmac should be an older version of the IP that doesn't have a feature set register
<mripard>
u-boot drivers are usually minimal
<hramrach>
you can. it has to set up the features somehow
<hramrach>
yes, that's what I get. not sure if it's selected or detected
<mripard>
anyway
<hramrach>
but using the version for Exynos which is recent I get no features detected
<mripard>
it looks like it has two clocks in the A20
<mripard>
and you enable only one of it in your dt
<wens>
mripard: the second clock is for selecting MII/RGMII
<wens>
mripard: and it's more of a register flag than a clock
maz_ has joined #linux-sunxi
<mripard>
it sets up the parent clock it uses, the dividor, delay and has an invertor
<mripard>
it looks a lot like a clock to me.
<mripard>
and anyway
<mripard>
your partent clock isn't right
<mripard>
it should be 49
<mripard>
17 is nthte EMAC
jemk has joined #linux-sunxi
<hramrach>
it's set to 17 in u-boot as well afaict
<hramrach>
that's where I copied it from
<wens>
hramrach: it's 17 in the second clock register
<wens>
hramrach: so you need to add 32
maz_ has quit [Ping timeout: 248 seconds]
<mripard>
so the clock will be gated at the end of the kernel boot, and you won't get anything out of the gmac anymore
maz_ has joined #linux-sunxi
<hramrach>
hmm, I get link with clock 49
<hramrach>
on phy1
<wens>
hramrach: are you using cb2 or ct?
<hramrach>
CT
<hramrach>
eth0: PHY ID 001cc915 at 0 IRQ 0 (stmmac-0:00)
<hramrach>
eth0: PHY ID 001cc915 at 1 IRQ 0 (stmmac-0:01) active
<hramrach>
libphy: stmmac-0:01 - Link is Up - 100/Full
<wens>
hramrach: you may need to set GMAC clock, though u-boot might already have done it
maz_ has quit [Read error: Connection reset by peer]
<hramrach>
I get all 32 PHYs with 0s on cb2 so there is something wrong eth0: PHY ID 00000000 at 0 IRQ 0 (stmmac-0:00) active
<hramrach>
but I use emac for downloading the kernel there
<wens>
hramrach: if you use gmac for downloading the kernel, you get 1 phy, but no network activity after boot
<hramrach>
so there is something that is not enabled in the kernel that u-boot does for you, definitely
<wens>
hramrach: i'm going to dig through cubietech 3.4 kernel
<hramrach>
hmm, when I omit the line which selects phy1 I get same result on phy0
<hramrach>
so looks like a bug that shows the same phy twice
maz_ has joined #linux-sunxi
Black_Horseman has quit [Quit: Zwi se logou mou!!!]
<mnemoc>
Turl: has anyone tried to port android 4.4 yet?
specing has quit [Ping timeout: 260 seconds]
<shineworld>
4.4 in A10/A20 should be very interesting
<mnemoc>
their new vm will "revive" all those 512MB sunxi devices out there
<mnemoc>
but the freaking blobs...
<mnemoc>
unless someone parts the RE mali and cedarx code to run on Android ;-)
<mnemoc>
ports*
<mnemoc>
a fully free android 4.4 for sunxi...
<mnemoc>
</dreaming>
<shineworld>
so main problem is : hard to have binary parts updated to 4.4 for A10/A20 ?
<mnemoc>
that's even the reason why galaxy nexus is not getting 4.4
<mnemoc>
f* blobs
<mnemoc>
but we do have reasonably usable RE code
<shineworld>
I thought that RE code is only for linux...
<mnemoc>
the high level API, yes
<shineworld>
however this is very out of my capabilities unfortunately...
<mnemoc>
making OSS bindings for android would give this platform a new level of openness
<shineworld>
After a year in android matter my skill with it aren't as I would like
<shineworld>
yeah, actually I fond that allwinner powered toys are one of most OSS toys that I've found at moment
<mnemoc>
hitting the android world with a similar level of openness would bring a lot of new devs
<shineworld>
I'm using android devices to do HMI (human machine interfaces) for industrial plants and I like results
<mnemoc>
you don't need fancy video decoding or 3d acceleration on an HMI
<shineworld>
unfortunately I'm moved to different platform than allwinner because I wasn't able to get working Android Accessory with any linux-sunxi kernel
<shineworld>
I've got always kernel problems with USB switching mode
enrico_ has joined #linux-sunxi
<shineworld>
so I'm using some rockchip (sigh) hardware at moment
<mnemoc>
we have usb for mainline, they don't ;-)
<shineworld>
when I run a simple UartAccessoryTest in my Linux uartaccessory emulator with Allwinner the kernel get crazy and USB enter in a infinite loop o
<shineworld>
I will happy if some USB gurus run the UartAccessoryTest on any android allwinner powered board and connect it with usb to a Linux PC running uartaccessory emulator...
<shineworld>
you will catch a loop in USB sw-udc.c
<hramrach>
the problem with android is there are no android devs, no docs, nothing
<shineworld>
I've tried to debug USB matter but the bad thing is that accessory mode close adb every time so any debug is very HARD to do
<shineworld>
my knowledge of linux is little ... too much little to afford a debug and understand what is wrong in code
<shineworld>
so I'm moved to rockchip (for the moment)
<shineworld>
obviously any my google powered terminal works fine :)
<shineworld>
but some customer require low-cost tablets to use how HMI interfaces
<hramrach>
rk is good for that
<hramrach>
seems lowest cost tablets atm are rk29xx
<shineworld>
BUT ... there is a BUT ... they don't share sources
<shineworld>
and so my capabilities to "mod" android os are very limited
<shineworld>
with previous Allwinner tablets I've modified a lot android sources to be more "industrial need friendly"
<shineworld>
Unfortunately to connect my external motion control board with Allwinner the only solution is to use IOIO board which use a different approach (an ADB client instead of AOA device)
Soru has quit [Remote host closed the connection]
wingrime has joined #linux-sunxi
<shineworld>
but ADB client is very slow than direct USB AOA approach
<JohnDoe_71Rus>
shineworld: hi
<shineworld>
hi JohnDoe_71Rus
<shineworld>
cya
shineworld has quit [Quit: Leaving]
Sonicadvance1 has quit [Ping timeout: 260 seconds]
<plaes>
unfortunately I cannot test the uart this weekend, as I'm on the road :S
<libv>
it should just work :)
_massi_ has quit [Quit: Leaving]
<plaes>
yeah, though I stupidly forgot to pack the usb->uart dongle :S
codingrobot has quit [Ping timeout: 264 seconds]
codingrobot has joined #linux-sunxi
tomee^ has joined #linux-sunxi
Upgreydd has joined #linux-sunxi
jemk has quit [Ping timeout: 252 seconds]
<Upgreydd>
hi guys i have a little problem, i've doing this http://linux-sunxi.org/Building_on_Ubuntu and installed gcc-arm-linux-gnueabi and gcc-arm-linux-gnueabihf... when i'm compiling i've get following error: error: required section '.rel.plt' not found in the linker script
<Upgreydd>
should i add some things to PATH compiler?
<Upgreydd>
variable*
<Dreadlish>
what are you compiling?
<Upgreydd>
Dreadlish: make hwpack-install SD_CARD=/dev/sdX ROOTFS=your_rootfs.tar.gz
<Upgreydd>
i'm doing this ;]
<Upgreydd>
with mine SD_CARD and rootfs ;]
<Upgreydd>
make hwpack-install SD_CARD=/dev/sdb ROOTFS=/home/upgreydd/sunxi/linaro-raring-server-20131021-529.tar.gz
<Dreadlish>
eh
<Dreadlish>
so you didn't set the toolchain?
<Upgreydd>
Dreadlish: but i should use gcc-arm-linux-gnueabi or gcc-arm-linux-gnueabihf ??
<Dreadlish>
if your hardware support hardfloat, it's better to use hf
<Upgreydd>
Dreadlish: ok i see, and what path i should add? i've tried to find it using "find / -name *gnueabihf*" but there's a lot of patches can you give me advice where i can find gcc-arm-linux-gnueabihf/bin ?
<Dreadlish>
eh
<Dreadlish>
/usr/arm-linux-gnueabihf/
<Dreadlish>
somewhere there
<Upgreydd>
thank you very much ;)
<Dreadlish>
if it isn't there should be similar directory there
<Upgreydd>
i was looking in /usr/lib
<Dreadlish>
bins should be linked in /usr/bin
<Dreadlish>
so you hadn't to add anything
<Upgreydd>
anything to $PATH?
<Dreadlish>
yup
<Upgreydd>
sooo why i'm getting this error?
<Upgreydd>
.rel.plt
<plaes>
funny, the touchscreen controller has another arm chip on it.. :S
<Dreadlish>
Upgreydd: which version of gcc do you have?
eebrah has joined #linux-sunxi
<Dreadlish>
(the arm one)
<Upgreydd>
Dreadlish: gcc version 4.8.1 (Ubuntu/Linaro 4.8.1-10ubuntu8
<Dreadlish>
4.8.1 and kernel isn't good choice.
<Dreadlish>
try to get something like 4.7
<Upgreydd>
Dreadlish: this is output of: gcc -v
<Upgreydd>
ok i'll
<Dreadlish>
4.7.2 or so
<Dreadlish>
linaro hosts it on their site
<Dreadlish>
afaik
<Upgreydd>
i've installed 4.7.3-7
<Upgreydd>
Dreadlish: thank you, looks you've solved mine problem ;]
eebrah has quit [Ping timeout: 264 seconds]
eebrah has joined #linux-sunxi
eebrah is now known as Guest77334
Taduro has quit [Read error: Connection reset by peer]
Guest77334 is now known as eebrah_
mouchon has quit [Ping timeout: 252 seconds]
Taduro has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Black_Horseman has quit [Changing host]
<oliv3r>
hno or anybody else, if i'm not mistaken, io-width was the per chip width, whereas bus-width was the widt of all chips combined. so 4 8-bit chips give 32-bit bus width, and 2 16-bit chips give also 32 bit bus width.
<oliv3r>
so single chip boards need to have an io-width of 32bits or is the bus-width 16 bits there?
<oliv3r>
in any case, cubietruck and cb2 have the same bus/io which is not right as cb2 has 2 and ct has 4 memory chps
Upgreydd has quit [Quit: Page closed]
cubear has quit [Quit: Leaving]
mouchon has joined #linux-sunxi
<torbenh3>
arokux: ping
<torbenh3>
wens: got a git repo with the gmac u-boot ?
<torbenh3>
ah.... found arokux repo
<torbenh3>
mmmm... doesnt look as sweet as the patches posted to the ML
<mnemoc>
shame on arokux
<mnemoc>
hno: can you comment on the gmac patches for u-boot?
<bizd>
I did put cubian on my cubieboard 1 and compiled the sunxi git kernel becuase I needed some missing kernel features, everything worked out fine but now my xbmc doesn't seem to do hw accel with HD Video anymore, mali is loaded though. Is there anything I need to do in addition after having
<bizd>
compiled the kernel on my wn ? Also what I see as a difference in the working an not working xbmc log is tahat it doesnt seem to be able to open cedar saying "ERROR: A10: open failed. (-4)"
Black_Horseman has quit [Quit: Αποχώρησε]
<tomboy64>
sounds to me like you need to relink the mali driver
<tomboy64>
(but i'm not an expert on that matter)
<bizd>
tomboy64: How can I do that ? Can you please explain ?
<tomboy64>
bizd: from where you got the sunxi-sources there should be a directory that offers the mali drivers
<tomboy64>
which in turn should provide instructions on what to do with them
<bizd>
tomboy64: yes the mali drivers are in the kernel as I understood and I did install all kernel modules after compiling the kernel and mali is also loaded
<tomboy64>
then i can't help you. i presumed you were still using the module your machine was shipped with against a newly compiled kernel
apo has quit [Ping timeout: 248 seconds]
<bizd>
The difference to the cubie original kernel on the image is the "+" infront of the version number 3.4.67+ is this probably some other kernel source cubie creator used ?
<bizd>
tomboy64: ah ok no :)
<bizd>
s/cubie/cubian
eebrah_ has quit [Ping timeout: 245 seconds]
eebrah_ has joined #linux-sunxi
apo_ has quit [Remote host closed the connection]
<tomboy64>
bizd: possible. or maybe he just used some additional features that are not enabled per default in the sunxi-kernel
<tomboy64>
you might want to ask in #cubian if there's a dev around
rz2k has quit []
FR^2 has joined #linux-sunxi
<bizd>
tomboy64: ok ;) they forwarded me to cubieboard
<mnemoc>
the + is added when you aren't sitting on a tag
<plaes>
does anybody know how to spot g-sensor on the board?
<tomee^>
btw. does sunxi-mali kernel support >r3p0 (both units of Mali400MP?)?