<brain_>
damn you all and your mainlines! i'm stuck on 3.0 until I get a new tablet or force the company to hand over custom ft5x module.
<libv>
brain_: put $nick to work and get yourself the custom module
<brain_>
Lol true, after looking at another users dmesg of the module loading Im gonna try tackle that once this CM9 gets done building so I know the code is complete.
<brain_>
another smaller issue is my nand layout...were do you define if you need nanda-i or nanda-j, etc. I have been reusing the same nand.ko code since its the only one I can find that works with my tablets a-j layout.
<brain_>
cubieboard kernel i think also has the correct nand code but it didnt work when i tried porting it into a later kernel
<libv>
oh, and stop chasing upstream.
<libv>
at least when you chase your tail, you know where you are when you stop.
dwilkins has joined #linux-sunxi
<brain_>
that makes sense.
FDCX_ has quit [Ping timeout: 245 seconds]
<hno>
brain_, partition table is on the nand. But there is two major Allwinner nand driver versions (incompatible).
popolon has quit [Quit: Quitte]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
diego71 has quit [Ping timeout: 256 seconds]
diego71 has joined #linux-sunxi
akaizen_ has quit [Remote host closed the connection]
FDCX_ has joined #linux-sunxi
ykchavan has joined #linux-sunxi
FDCX_ has quit [Ping timeout: 264 seconds]
<steev>
specing: is he 10? i assumed he was young... or something isn't right with him
ykchavan has quit [Ping timeout: 245 seconds]
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
akaizen has joined #linux-sunxi
akaizen has quit [Remote host closed the connection]
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
Soru has quit [Remote host closed the connection]
ykchavan has joined #linux-sunxi
fldc has quit [Ping timeout: 245 seconds]
x_x|eebrah is now known as o_O|eebrah
akaizen has joined #linux-sunxi
Xaros has quit [Ping timeout: 246 seconds]
akaizen has quit [Remote host closed the connection]
akaizen has joined #linux-sunxi
akaizen has quit [Remote host closed the connection]
<brain_>
Success! CM9 is running smoothly! Wifi works, Touch works, storage works, sound works! I had to do a few more ugly hacks to the source but so far so good. gotta fix a few minor bugs after i get some sleep. Thanks everyone.
<brain_>
Also using prebuilt kernel and my modules.
Brod_ has joined #linux-sunxi
techn_ has joined #linux-sunxi
iamfrankenstein has quit [*.net *.split]
Seppoz has quit [*.net *.split]
techn__ has quit [*.net *.split]
Brodomir has quit [*.net *.split]
hramrach has quit [*.net *.split]
Seppoz has joined #linux-sunxi
soul has joined #linux-sunxi
soul is now known as Guest82821
atsampso1 has joined #linux-sunxi
Guest82821 has quit [Read error: Connection reset by peer]
MadSpark- has joined #linux-sunxi
speakman_ has joined #linux-sunxi
mturquet1e has joined #linux-sunxi
buZz__ has joined #linux-sunxi
speakman_ has quit [Changing host]
speakman_ has joined #linux-sunxi
Turl_ has joined #linux-sunxi
Turl has quit [Ping timeout: 264 seconds]
mturquette has quit [Ping timeout: 264 seconds]
atsampson has quit [Ping timeout: 264 seconds]
hno has quit [Ping timeout: 264 seconds]
buZz has quit [Ping timeout: 264 seconds]
speakman has quit [Ping timeout: 264 seconds]
rm has quit [Ping timeout: 264 seconds]
MadSpark has quit [Ping timeout: 264 seconds]
jeremb has quit [Ping timeout: 264 seconds]
hno has joined #linux-sunxi
hno has joined #linux-sunxi
hno has quit [Changing host]
jeremb_ has joined #linux-sunxi
jeremb_ is now known as jeremb
brain__ has joined #linux-sunxi
iamfrankenstein has joined #linux-sunxi
iamfrankenstein has quit [Excess Flood]
tzafrir_laptop has joined #linux-sunxi
rm has joined #linux-sunxi
MadSpark has joined #linux-sunxi
tgaz has quit [Ping timeout: 245 seconds]
brain_ has quit [Ping timeout: 245 seconds]
eagles0513875 has quit [Ping timeout: 245 seconds]
tzafrir has quit [Ping timeout: 245 seconds]
woprr_ has joined #linux-sunxi
merbanan has quit [Ping timeout: 245 seconds]
MadSpark- has quit [Ping timeout: 245 seconds]
HeHoPMaJIeH has quit [Ping timeout: 245 seconds]
ykchavan has quit [Read error: Connection reset by peer]
woprr has quit [Ping timeout: 245 seconds]
hipboi has joined #linux-sunxi
stulluk has joined #linux-sunxi
<stulluk>
Hi, I have finally be able to setup my A20 board for 1920x1080. My sys_config.fex file part for LCD0 setting is like this: http://pastie.org/8332190
<stulluk>
However, with this setting, my mouse pointer is moving slowly, and screen is flickering, and display is slow
<stulluk>
I have modified lcd_dclk_frq and set to 156, but it become worse now...
<stulluk>
Make it more slower
<stulluk>
Is there a guide or tutorial, how to set all these parameters correctly ?
akaizen has joined #linux-sunxi
rellla has joined #linux-sunxi
\\Mr_C\\ has quit []
rellla2 has joined #linux-sunxi
n01 has joined #linux-sunxi
rellla has quit [Ping timeout: 264 seconds]
akaizen is now known as 36DABGD3Q
akaizen has joined #linux-sunxi
tgaz has joined #linux-sunxi
36DABGD3Q has quit [Read error: Connection reset by peer]
jeremb has joined #linux-sunxi
jeremb has quit [Changing host]
eagles0513875 has joined #linux-sunxi
o_O|eebrah is now known as O_O|eebrah
HeHoPMaJIeH has joined #linux-sunxi
HeHoPMaJIeH has joined #linux-sunxi
HeHoPMaJIeH has quit [Changing host]
panda84kde has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
hramrach has joined #linux-sunxi
ykchavan has joined #linux-sunxi
ykchavan has quit [Client Quit]
cajg_ has joined #linux-sunxi
notmart has joined #linux-sunxi
notmart has joined #linux-sunxi
cajg has quit [Ping timeout: 245 seconds]
arokux2 has joined #linux-sunxi
FR^2 has joined #linux-sunxi
stulluk has quit [Remote host closed the connection]
focus has quit [Remote host closed the connection]
gnufan has joined #linux-sunxi
focus has joined #linux-sunxi
fldc has joined #linux-sunxi
cajg_ is now known as cajg
kaspter has joined #linux-sunxi
gnufan has quit [Ping timeout: 240 seconds]
atiti has joined #linux-sunxi
FDCX_ has joined #linux-sunxi
gnufan has joined #linux-sunxi
hramrach has quit [Ping timeout: 240 seconds]
rz2k has joined #linux-sunxi
soldoKyn has joined #linux-sunxi
hramrach has joined #linux-sunxi
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<soldoKyn>
hi to all! I noticed that in sunxi-3.4 branch of linux-sunxi git repository there isn't the sun7i-a20-olinuxino-micro.dts file, while there is in official torvalds git repo...
<arokux>
focus, you may want to add a link on our wiki to you image. it will get more testing :)
<oliv3r>
a23 might actually be sun8i :)
<arokux>
oliv3r, I'll delete A23 since nobody has it anyway
<focus>
arokux: done! thank you for suggestion :)
<arokux>
thanks focus I'm really glad you work closer with sunxi.org now :)
<focus>
Next target is to get Linaro installed on Sata and chroot to boot from sata and install a whole bunch of other packages to beef up the distro
<focus>
any idea where i should set the chroot command so that linut boots from uSD and then switches to sata?
<focus>
automatically
<oliv3r>
arokux: i wowuldn't be supprised if a23 will become popular
<arokux>
focus, are you sure it is called chroot? you just need to supply boot parameter to the kernel
<oliv3r>
a23 is basically dualcore a13
<arokux>
oliv3r, we'll add it then.
<arokux>
focus, root=/dev/device
<arokux>
focus, i.e. root=/dev/sda1
<oliv3r>
focus: yeah what arokux said, load the kernel from SD (+initramfs if you wish) but assume from the kernel that you are booting from sata as nomral
<oliv3r>
the kernel won't know nor care it was booted over mmc
<oliv3r>
u-boot actually just loads it into ram, and then loads the kernel; at that point the kernel knows nothing so boots whatever you tell it too
<arokux>
focus, you should make sure the kernel can talk to SATA and EXT4 (or whatever the filesystem you'll have on it) so SATA and FS support should either be in intiramfs or built-in
<focus>
ah - so the /dev/sda is already already known to kernel?
<focus>
oop arokux you answer
<focus>
sata needs to be compiled into linux and not loaded as module somewhere I read
<arokux>
focus, root=/dev/sda1 <--- pay attention to the "1" because you actually need to supply a partition, not a disk device file
<arokux>
focus, making it built-in is the easiest option, but maybe not the most flexible one.
<oliv3r>
focus: if you load sata as module, it needs to be in your initramfs
<focus>
ok, so just set root=/dev/sda1 in setenv where its normally set and it should boot?
<arokux>
focus, yes, instead of root=/dev/mmcblk0p2
<arokux>
focus, if you've used it.
<focus>
I try this evening some combinations and report back :)
<focus>
thanks guys!
<arokux>
focus, but then rootfs should be on /dev/sda1
<focus>
yes - I format and copy the full distro into there - i got 250GB ssd to play with for this
<arokux>
focus, and do not forget to generate boot.scr
<TheViking>
One bad sectory on my whole hdd and windows acts like the heavens are crashing down.
buZz__ is now known as buZz
<oliv3r>
Tsvetan: because EAGLE isn't opensource :p
<oliv3r>
Tsvetan: that's why I ask! in the hopes the webpages are wrong and you did use KiCAD or gEDA :)
<Tsvetan>
oliv3r so what?
<oliv3r>
TheViking: wrong channel :p
<oliv3r>
Tsvetan: it breaks my heart :(
<Tsvetan>
IIRC there is no requirement to make OpenSource boards only with Open Source CADs :))))
<oliv3r>
Tsvetan: that is true
<Tsvetan>
Eagle is used for most of the OpenSource hobby project if you look at Sparkfun, Adafruit, SeedStudio they all use Eagle and this is not coincedence
<TheViking>
IIt's a conspiracy
<Tsvetan>
Eagle is decent low cost reliable tool which is mature and present for decades
<Tsvetan>
and it allow you to extend its capabilities with your own scripting language
<oliv3r>
Tsvetan: dangerousprototypes; i'm well aware
<oliv3r>
Tsvetan: but most of those hobbiests are windows users :p
<Tsvetan>
you have access to all databases and you can make your own commands which incredible increase the productivity
<oliv3r>
Tsvetan: aye, i've done some dabbeling with eagle YEARS ago
<Tsvetan>
Eagle runs on Linux and MacOs too
iamfrankenstein has joined #linux-sunxi
<oliv3r>
yeah but gEDA for example does not; so a lot of hobbiests don't use it
<oliv3r>
i played with gEDA and it was nice, but not yet 'ready for prime time'
<Tsvetan>
I try both gEDA and KiCAd and they are lighting years behind Eagle
<oliv3r>
really that bad?
<oliv3r>
that's sad
<Tsvetan>
yes, I guess crytical mass is missing there
<Tsvetan>
just because one project is open source do not make it better or easier to work with
<Tsvetan>
it have to attract crytical mass of users and developers to move forward
<Tsvetan>
and Eagle with these ULP scripts gives you unlimited flexibility once you learn it
<Tsvetan>
which doesnt exist even in the $$$$ CAD products
<Tsvetan>
its like CAD which have Linux shell language
<oliv3r>
heheh
<oliv3r>
yeah I understand where you are comming from of course
<tgaz>
Tsvetan: what kind of scripts are common to make?
<Tsvetan>
tgaz you have access to the netlist, components, every single object in the PCB and you have C like language which allow you to manipulate all these objects
<tgaz>
Tsvetan: but is there one kind that is more common than others? (placing groups, rule checks, parameterized components)
\\Mr_C\\ has quit []
<tgaz>
or is it just all-over? :)
<Tsvetan>
tgaz I do not understand your question
<Tsvetan>
go to Eagle site, there is lot of docs and tutorials
<tgaz>
i'm just curious what problems people solve with the scripts
<Tsvetan>
tgaz add new commands which do not exist
<focus>
Tsvetan: KiCAD not light years behind, just a different way of doing things, everything runs off text files and so its incredibly powerful when you know how to work it http://www.gplsquared.com/SoM2/SoM2.html
<Tsvetan>
focus ok
<oliv3r>
but what Tsvetan says is true, Eagle has a huge momentum in the electronic hobbiest market
<focus>
only thing it don't have is a good built in autorouter (yet)
<oliv3r>
and while, of course, it would be much cooler if it where made using opensource tools; we have to overlook that bit sometimes :)
<Tsvetan>
focus :-)
<oliv3r>
otherwise, we couldn't even drink water anymore
<oliv3r>
since water purification plants probably run windows in their scada systems
<Tsvetan>
oliv3r I do not like talibanish thinking :) nor I have problems to use Windows or other OS if necessary
<focus>
hate windoosh - its the king kong wrecker of projects
<focus>
can't build anything before its broken, infected, or no sources to fix things
<Tsvetan>
yes, there are many things which could be done easier under linux, but if I have to sit on computer with Windows I will not act like prima donna :)
<focus>
I would sing like a soprano with their balls cut off
<oliv3r>
Tsvetan: lol :p Tsvetan dressed up as a prima donna :)
<Tsvetan>
:))))
<mnemoc>
o_O
<oliv3r>
at work im forced to work like a prima donna :(
<mnemoc>
...
<Tsvetan>
its like to say I cant drive any other car than BMW :))
<oliv3r>
i don't own a car :)
<Tsvetan>
ok, you may like BMW and to miss some features when you sit on Opel, but you can move from point.A to point.B with any car
<oliv3r>
I would never drive a Ford though
\\Mr_C\\ has joined #linux-sunxi
<Tsvetan>
Ive driven Moskvitch once upon a time, so nothing can scare me
<oliv3r>
hahaha
<oliv3r>
well fords use windows for their onboard OS
<oliv3r>
*sshiver*
<Tsvetan>
oliv3r you must be kidding me
<Tsvetan>
probably for the entertaining
<Tsvetan>
but not driven your engine and brakes I hope
<oliv3r>
Tsvetan: you never know; but ford made a deal, maybe just like nokia
<oliv3r>
:)
<oliv3r>
Lexus actually does use linux for their onboard 'infotainment'
stulluk has quit [Ping timeout: 242 seconds]
<Tsvetan>
what I read few days ago is that MS bought Nokia mostly for the patents they have, now MS racket Android manufacturers $3.6 bm per year for patents, with Nokia patents they want to make this up to $8 bn
<Tsvetan>
Nokia didnt licensee these patents they had them just to protect themself from other company claims
<oliv3r>
yeah
<oliv3r>
microsof tis a pattent troll/shill
<oliv3r>
they bought nokia for very cheap
<oliv3r>
they send in their elop drone; he ruined the company, now he goes back and they bought it cheap
<oliv3r>
shame
<Tsvetan>
yes the definition of patent troll is company which do not manufacture but just licensee the patents :)
<oliv3r>
and those android patents aren't even valid; but manufactures are 'affraid of the big microsoft' so they just pay
hipboi has quit [Quit: Leaving]
stulluk has joined #linux-sunxi
Black_Horseman has quit [Remote host closed the connection]
<ssvb>
techn_: ping
ganbold_ has joined #linux-sunxi
O_O|eebrah is now known as o_O|eebrah
TheViking has quit [Ping timeout: 250 seconds]
FDCX_ has quit [Remote host closed the connection]
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
popolon has joined #linux-sunxi
<ssvb>
Turl_: do you know if Android can easily move to r3p2-01rel2 version of mali?
massi_ has joined #linux-sunxi
FDCX_ has joined #linux-sunxi
deasy has joined #linux-sunxi
hipboi has joined #linux-sunxi
<oliv3r>
n01: you where talking about hte resitive touchscreen driver yeah? the we recently had a driver found for the temp sensor that even works on a10; so cpu temp monitoring should be possible with that sensor
<n01>
oliv3r: it wasn't me sorry
FDCX_ has quit [Ping timeout: 248 seconds]
<oliv3r>
n01: your right, vinifvr
nove has joined #linux-sunxi
iamfrankenstein has quit [Ping timeout: 245 seconds]
Seppoz has quit [Remote host closed the connection]
<nove>
in a A13 tablet, usb-boot boots old kernel 3.0 fine, but with new compiled 3.4 gives no signs of live, beyond that panic=xxx affects the reboot time
<nove>
i don't have serial console, any ideas of how to tackle this
deasy has quit [Remote host closed the connection]
arokux2 has quit [Remote host closed the connection]
deasy has joined #linux-sunxi
<soldoKyn>
arokux: I noticed that sun7i_defconfig in sunxi-3.4 branch missed CONFIG_SUNXI_EMAC ... is that right? I had to enable that flag to be able to work with ethernet with my new hand-compiled kernel...
<arokux>
soldoKyn, no, it's not, please submit a patch with CONFIG_SUNXI_EMAC=m
iamfrankenstein has joined #linux-sunxi
hansg has joined #linux-sunxi
hansg has quit [Client Quit]
<deasy>
plop
<deasy>
someone know where find the hwpack for a20/cubieboard2
<buZz>
~ i'm deasy like sunday moooorningggg
* deasy
touch buZz
<deasy>
are you ok?
<deasy>
:p
<buZz>
;)
<mnemoc>
the sunxi-bsp doesn't even know about a20 yet :|
<buZz>
deasy: i never used a hwpack for a20 btw
stulluk has quit [Ping timeout: 245 seconds]
stulluk has joined #linux-sunxi
<stulluk>
Hi, does anybody know correct sys_config.fex setting for 1920x1080 on A20 ?
<stulluk>
I mean hv, ht, hbp...etc settings?
<hno>
nove, a little hard to say anything without serial console I am afraid.
<hno>
nove, but likely something silly obvious..
<hno>
are you using an initramfs for the rootfs?
<nove>
hno, no initramfs, has hoping to get lcd to display
<atsampso1>
can you pass something like "console=tty0 disp.screen0.output_type=1" as kernel args? (dunno if that makes sense on the A13...)
<nove>
also tried usb gadget serial, and ether but no signs
<nove>
atsampso1, console=tty0 is there, will try the rest
dlan has quit [Ping timeout: 260 seconds]
<popolon>
nove, do you use direct HDMI or a a HDMI->DVI converter ?
dlan has joined #linux-sunxi
<popolon>
oh, on a tablet, not a cubie
<nove>
popolon, is a a13 tablet, doing usb-boot, kernel 3.0 works, but 3.4
rz2k has quit []
gnufan has quit [Ping timeout: 264 seconds]
* tgaz
wonders if using makefiles for dep-based init would be a novel approach...
pitillo has quit [Ping timeout: 240 seconds]
ganbold__ has joined #linux-sunxi
<hno>
nove, root on nand or sd?
<hno>
and is root specified correctly? And corresponding driver built-in?
<nove>
hno, sd
ganbold_ has quit [Ping timeout: 245 seconds]
<hno>
nove, have you specified rootwait?
<nove>
hno, 3.0 displays console in lcd, even with root=invalid
<nove>
yes, rootwait
<hno>
nove, is the lcd & hdmi & fbcon drivers built-in in your kernel .config?
<specing>
tgaz: why reinvent that? you already have dep-bases init systems...
<tgaz>
specing: for the fun of going where no man has gone before ;)
<specing>
rather spend your time helping libv RE those malis
pitillo has joined #linux-sunxi
<Turl_>
ssvb: assuming you get matching android blobs and the kernel gets updated, it shouldn't be too hard
<ssvb>
Turl_: the question is whether we have these blobs already or not
<nove>
hno, yes builtin from sun5i_defconfig, also set framebuffer console support, and compiled in fonts
<tgaz>
specing: i'd love to. let me get the blinking LED tutorial written first, and i'll get back to you. i have a cb2 that i'd love to have some foss graphics on.
<ssvb>
Turl_: I just want to know if we can move the kernel to r3p2-01rel2 mali or still need to keep both r3p0 and r3p2-01rel2 there
<hno>
tgaz, not sure no man has gone there before... but there is good reasons why it is not commonplace to use make for init..
<Turl_>
ssvb: I haven't seen those blobs
<Turl_>
ssvb: but then I haven't looked for them either :)
<Turl_>
ssvb: btw why rel2 and not rel3?
<ssvb>
Turl_: ask Benn :)
<ssvb>
Turl_: that's the version of the new userspace mali blobs
<Turl_>
:)
<rellla2>
ssvb: is there any WIP branch existing anywhere for the new mali already?
<wingrime>
ssvb: blob use call back to opensourced part
<wingrime>
ssvb: It can explain why patck get this cache error
<nove>
it works
<nove>
wingrime, i also got slow decoding at 3.0, by starting/stoping(ctrl-C) mplayer, dmesg has disp failed (something layer scaler)
<wingrime>
nove: on a13>
<wingrime>
?
<nove>
wingrime, yes, i only have a13, i will right now try 3.4
<wingrime>
nove: whats different with before?
<nove>
wingrime, with 3.0? no difference, i only haven't tested multiples start/stop mplayer
soldoKyn has quit [Quit: Konversation terminated!]
<arokux>
nove, what tablet do you own?
djahma has joined #linux-sunxi
<techn_>
ssvb
<nove>
arokux, some random one
<arokux>
nove, you do not want to tell or what?
<nove>
wingrime, here 3.4, is playing ok (like 3.0)
<wingrime>
... no idea what wrong with me ...
<ssvb>
wingrime: yes, libvecore is a small low level library, and then there is an open source higher level wrapper for cedar
o_O|eebrah has quit [Ping timeout: 260 seconds]
ZetaNeta has joined #linux-sunxi
<ssvb>
wingrime: does cedar decode videos correctly on a20 with cached mapping enabled in your tests?
<wingrime>
ssvb: cache issue must be fixable there
Paullo612 has joined #linux-sunxi
eebrah has joined #linux-sunxi
<djahma>
hi there. I'm trying to compile a kernel for my A10 polaroid tablet that would make use of wifi and touchscreen out of the boot. I've managed to get fedora working with their inet97 board specs and kernel but no touchscreen nor any wifi or bluetooth working. Since linux-sunxi 3.4 doesn't compile and ends with an error, I tried using the kernel.org 3.4 and 3.10 kernels. They compiled, but my tablet refuses to boot them and I can't find out why
eebrah is now known as Guest9303
<ssvb>
wingrime: supposedly android should have a correct higher level wrapper code for a20
<nove>
arokux, doesn't have printed that looks like a model number, and i also don't care
<buZz>
djahma: kernel.org kernels dont support our platform
<buZz>
or they do, but you wont get any proper drivers yet
<buZz>
no hdmi, no mali, etc
Guest9303 is now known as eebrah_
eebrah_ is now known as ibrah
<ssvb>
wingrime: the cpu cache in cortex-a7 is a bit more aggressive than in cortex-a8, so it's more likely to expose bugs
<nove>
wingrime, that was with fddev, i will try now with sunxifb
<arokux>
djahma, your best bet is to be specific about the sunxi-3.4 build failures and get that fixed by us
<arokux>
djahma, and be sure to use the latest one from git.
<ssvb>
techn_: I wanted to ask about the android & mali kernel module version
<ssvb>
techn_: do you know which mali version is used by the latest android images?
<arokux>
buZz, very basic mainline support started from 3.10 or so..
<arokux>
djahma, ^
<djahma>
sunxi 3.4 kernel compilation attempts with the gnueabihf toolchain I found in synaptic in my ubuntu 12.04 end in an error like so: drivers/built-in.o: In function `rcu_read_unlock_sched_notrace':
<djahma>
/home/djahma/Documents/arm/linux-sunxi/linux-sunxi-sunxi-3.4/include/linux/rcupdate.h:843: undefined reference to `__tracepoint_mmc_blk_rw_end'
<Paullo612>
wingrime: Hello, do you know anythig about those h264 engine regs that marked as ??? at VE guide?
<techn_>
ssvb: I have seen atleast r3p2 version string somewhere.. but I can try to find if there is newer versions somewhere
<wingrime>
ssvb: or this mustbe changed for cortex a7?
<ssvb>
wingrime: it looks correct if it this code is actually executed when needed
<wingrime>
Paullo612: there is much registers that blob not touch
<wingrime>
Paullo612: even general block 0-0x100 have many untouched regs
<techn_>
ssvb: actually I saw a13 firmware with r3p2 mali
<ssvb>
wingrime: I mean that the missed cache flush ioctl can be harmless on a8, but trigger bugs on a7 because the cache allocation policy is different (write-allocate vs. read-allocate)
<ssvb>
techn_: thanks, maybe that's it (r3p2-01rel2)
<wingrime>
ssvb: we need firstly see how libvdpau works on a20
<Paullo612>
ssvb: about the cache -- i tried to use ioctl to flush cache with no luck in my VP8 implementation (on A10). I also tried __clear_cache() gcc function with no luck ether. __clear_cache() does baisicaly the same as ioctl
<Paullo612>
wingrime: I tryed to fill them all with 0xffffffff today, and read them back. By that value we can determine register type (RAM, Ve RAM, data, etc). For example, all Ve RAM data regs looks like 0xfffffc00.
<djahma>
arokux cheers, i'll follow these steps more thoroughly. I remember having an issue with kernel.org where compiling a uImage required providing a LOAD_ADDRESS. Will I need to specify that again? and how does uboot boots the uImage? does it just look for a uImage and start it like magic?
<arokux>
djahma, I've replied to you
<wingrime>
ssvb: cache can be droped form userspace ?
<arokux>
djahma, in an e-maill I mean
<djahma>
arokux. I see that. That's kind of you, I'll follow your guidance
<arokux>
djahma, read that wiki page, everything is described there. LOAD_ADDRESS is for mainline kernel. you need mainline kernel if you want to help to add support of the sunxi to the mainline. otherwise if you just want a feature full kernel you go for sunxi-3.4
<ssvb>
Paullo612: not really, __clear_cache is doing different things, it exists to support self modifying code and JIT
jemk has joined #linux-sunxi
<wingrime>
Paullo612: we have no code , no blob that use that regs
<wingrime>
Paullo612: Hopefully blob have all mpeg reg names with it
<wingrime>
Paullo612: but even with names , some regs not used
<wingrime>
hipboi: cedarx are allwinner's IP ?
<Paullo612>
wingrime: Seems like there is another trigger reg in h264 engine.
<nove>
wingrime, is playing better than 3.0, big bunny 1080p, with audio, no frames dropped
<wingrime>
nove: thats cool see more man with working libvdpau
<wingrime>
nove:I also have same with a10
<wingrime>
Paullo612: are you getting irq
<wingrime>
?
massi_ has quit [Quit: Sto andando via]
<wingrime>
Paullo612: and witch reg numer?
<Paullo612>
ssvb: Ok. But cache flush using ioctl does not work ether. Only uncommenting vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot); in kernel module helps.
djahma has quit [Ping timeout: 245 seconds]
<ZetaNeta>
Whats the cheapest allwinner netbook?
<ssvb>
Paullo612: can you check if using writecombine instead of noncached also works?
<wingrime>
Paullo612: when you do flush?
<ssvb>
yeah, that's a good question
<ZetaNeta>
Because i like thinked of investing into 3d printer, and "print cubieboard powered laptops"
<buZz>
yeah, build a reprap!
<Paullo612>
wingrime: I'm not shure. Needs some testing. But when i try to fill with 0xffffffff 0x02bc-0x02cc everything just hangs (like when triggering some functions using MACC_H264_TRIG with trash in address regs)
<wingrime>
Paullo612: you just get IRQ
<wingrime>
Paullo612: mostly sure
FR^2 has quit [Quit: Connection reset by peer]
<Paullo612>
ssvb: And when should i flush?
<wingrime>
Paullo612: can you try write 0xffffffff to mpeg's MACC_MPEG_CTR_MB?
<ssvb>
Paullo612: after writing the encoded bitstream data to memory and before banging the cedar registers to start decoding
<Paullo612>
ssvb: it's obuious. Ok. I'll try to move flush function around the code and see what happend.
<steev>
is there a way to buy a cubieboard2 without it being shipped via USPS? USPS is absolutely turrible
<ssvb>
steev: do you want to order another cubieboard2 and check which of them gets delivered first? ;)
<steev>
ssvb: yes
<steev>
considering the cubieboard2 has been IN san antonio since the 13th
<steev>
and still hasn't been delivered, and i can't actually CALL that post office to find out wtf is going on
<steev>
and apparently this phone number that i do have, you call them and they *email* that post office
<Paullo612>
wingrime: Should i enable mpeg engine first?
<wingrime>
Paullo612: yes of couse
fredy has quit [Excess Flood]
<wingrime>
steev: thats nothing comparea russian post service
arokux2 has joined #linux-sunxi
<ssvb>
Paullo612: in any case, the __clear_cache performs "clean D line to the point of unification", which means that the dirty L1 data cache lines are only pushed to L2 cache (which is the point of unification between I and D) but not necessarily reach memory
<wingrime>
steev: 2.5 mouth to get cb2
<ssvb>
Paullo612: still would be very interesting to know if writecombine helps
fredy has joined #linux-sunxi
<ssvb>
Paullo612: the ioctl from cedar should clean the cache to the point of coherency, at least in theory
<steev>
...
<steev>
so i just got hung up on, but before i was, i was told that "since it's registered mail, it's under lock and key and they won't redeliver it"
<Paullo612>
ssvb: Should i use writecombine with or without ioctl?
<steev>
why they didn't tell me this yesterday when i called to get it redelivered, i have no idea
<ssvb>
Paullo612: writecombine should be just a faster alternative to noncached (no ioctls necessary)
<ssvb>
Paullo612: but it's surely interesting why ioctl does not help
<Paullo612>
ssvb: Ok, i'll try it. Need some time to rebuild the kernel.
rz2k has joined #linux-sunxi
<ssvb>
techn_, wingrime, nove: what is the clock frequency of mali in your a13 devices?
<techn_>
ssvb: 160Mhz
<techn_>
~
<nove>
ssvb, how to get it?
<ssvb>
techn_: hmm, does PLL4 run at 320MHz?
<ssvb>
nove: "modprobe mali" and then watch dmesg log
<techn_>
ssvb: "mali clock set completed, clock is 180000000 Hz"
<wingrime>
Paullo612: can you try some regs from mpeg engine
<wingrime>
Paullo612: it have some regs are closely to h264
<ssvb>
nove: and cedar must be also running at a rather low clock frequency
<wingrime>
Paullo612: so, try same registers like h264 but with 0x1xx base
<ssvb>
techn_: could you possibly experiment a bit with PLL4 clock frequency and the Mali divisor to find how fast it can run for real?
<techn_>
ssvb: sure.. just tell me where to tweak :p
<wingrime>
ssvb: jemk tested cedarx on 24 Mhz , it decode video but slow
<Paullo612>
wingrime: Ok. I'll play with it later.
<ssvb>
techn_: try to increase the PLL4 clock frequency little by little and run something like glmark2-es2 to confirm the performance increase and that it actually works
<techn_>
ssvb: I have android rootfs set
<techn_>
so I can try some android benchmarks :p
<wingrime>
ssvb: thats not real stability test....
<Paullo612>
wingrime: What city are you from?
<ssvb>
techn_: PLL4 is also used by Cedar, so it's better not to load Cedar module while you are experimenting with Mali
<techn_>
ssvb: a10 and a20 are using pll5?
<wingrime>
Paullo612: yekateriburg
<ssvb>
wingrime: it will break when Mali does not work at all (actually start freezing a bit, with some spam posted to dmesg log)
<Paullo612>
:( So far from here. I'm from St. Petersburg.
<ssvb>
wingrime: but surely if glmark2-es2 works without problems, it does not mean that mali is really stable
<wingrime>
Paullo612: yeax
<ssvb>
techn_: pll5 is used for dram
<ssvb>
wingrime: the A13 manual specifies 381MHz for Mali, I think this is a little bit faster than 180MHz ;)
<libv>
180MHz?
<libv>
yikes
<libv>
can it even fill 800x400 lcds fast enough with that?
<techn_>
ssvb: what is parent for pll4 ? could that be also wrong?
focus_it has joined #linux-sunxi
<ssvb>
techn_: pll4 is a parent for mali and cedar, otherwise it runs by itself (from 24-MHz crystal), you can check the a13 user manual
djahma has joined #linux-sunxi
<ssvb>
libv: I don't have a13 myself, but this misconfiguration looks like a bug that nobody bothered to fix
<ssvb>
libv: or it might be that a13 really can't run mali at a high clock speed because of some hw issue
<ssvb>
libv: do you know any gpu stability stress test?
<libv>
you should be able to run many different instances of glmark2-es2 tests together
<libv>
that should keep things pretty busy
mcbrick has joined #linux-sunxi
<techn_>
furmark :)
<ZetaNeta>
Is there a cheap alternatives to olinuxino/cubie?
<oliv3r>
ZetaNeta: oinuxino not cheap enough?
<ZetaNeta>
Nope
<techn_>
cheaper version of cheap
<oliv3r>
ZetaNeta: one of those android 'sticks'
<ZetaNeta>
I mean a dev board. And better if it have cheap LCDs for it
<jelly-home>
ZetaNeta: which features are you willing to give up?
<ZetaNeta>
jelly-home, Gime some time......
<ZetaNeta>
Erm.......
<techn_>
ZetaNeta: buy cheap a13 tablet
<oliv3r>
i don't think you'll find a cheaper dev board thena olimex
<jelly-home>
"none, I just want it CHEAP but with all the open source and gpio exposed and a nice box!"
<ZetaNeta>
All the "special" developer stuff, like gpio. I am unlikely to use that.....
<ZetaNeta>
I like am investing into a 3d printer. And now i wanna print cheap ARM netbooks, and get profit
atsampso1 is now known as atsampson
* ZetaNeta
hopes that chineese will not just "eat him" in this business
<ZetaNeta>
Lan, LCD, USB. (Keyboard and mouse.... if not TS screen)
<nove>
ZetaNeta, convert tablets in netbooks
<ZetaNeta>
Thats what i need
<ZetaNeta>
nove, People would just then buy "keyboards for tablets"
<ZetaNeta>
than my thing
<wingrime>
ZetaNeta: marsboard?
<ZetaNeta>
wingrime, Googling.....
<ZetaNeta>
Hey, looks nice
<ZetaNeta>
only 50 bucks........
<ZetaNeta>
I am aiming to have atleast something RPI "price"
<steev>
woo, the manager is going to deliver it today
<wingrime>
ZetaNeta: rpi in my city local stores costs 80$
<nove>
ZetaNeta, i mean get the tablet insides, and put them in a netbook case
<ZetaNeta>
I mean something like 20 bucks, exept shipping
<ZetaNeta>
nove, Thinked about this
<wingrime>
ZetaNeta: tablet+ case with usbkeyboard
<ZetaNeta>
But tablets also cost alot.... And lack LAN and USB........
<ZetaNeta>
Through the last thing could be fixed
<wingrime>
ZetaNeta: tablet have usualy OTG and wifi
<ZetaNeta>
But USB-to-LAN and USB hubs will make it more expencive
<ZetaNeta>
wingrime, But LAN would be very good
<wingrime>
ZetaNeta transformer for lan is expensive
<ZetaNeta>
wingrime, Exactly what i sad
<ZetaNeta>
That USB to Ethernet is quite expencive, for a laptop targeting ~40 bucks
<wingrime>
ZetaNeta: RPi looks best for you
<ZetaNeta>
RPi is a old, slow, proprietary, piece of........
<ZetaNeta>
Through, i can take a look
<wingrime>
ZetaNeta: a13 tablets chap enought
<ZetaNeta>
"Bring me the cheapest one!"
<wingrime>
ZetaNeta: but I can also say a10 a20 slow
<wingrime>
ZetaNeta: I want something with cortex a15!
<wingrime>
ZetaNeta: and normal ram controller
<ZetaNeta>
wingrime, In my opinion, Allwinner is always faster than RPI
<ZetaNeta>
Through, i maybe wrong.....
<wingrime>
ZetaNeta: thats I think faster than Rpi
<wingrime>
ZetaNeta: but when you comparea aw's with tegra 3 ...
<ZetaNeta>
wingrime, We are comparing DevBoards? Right? Not the dirty smartphones
<wingrime>
ZetaNeta: want fast --- odroid
<wingrime>
ZetaNeta: hw are same with smarphones
<oliv3r>
ZetaNeta: develop your own 'pi' board; the A10/A20 is 'only' 7 USD in volume
<atsampson>
bonus marks if you can make it have the same footprint as a RPi, saving you having to design your own case ;)
<ZetaNeta>
wingrime, True, but i thinked that all those "Super HW" like a big graphic card is only a smartphone thingy
<ZetaNeta>
And anyway, its out of the budget to make a Tegra 3 netbook
<jelly-home>
tho, good luck keeping the price with eth phy down to $20 with anything under... what, 10k units?
<ZetaNeta>
oliv3r, Genius!
<ZetaNeta>
oliv3r, Where can i get em, and how to make the board!
<jelly-home>
ZetaNeta: ask lkcl how much %^@$# time and resources it takes to develop a product from scratch, with just a SoC in sight
<buZz>
i wish A10/A20 was used in cellphones
<oliv3r>
ZetaNeta: gEDA or KiCAD
<buZz>
would make picking a new one a lot easier
<oliv3r>
you can buy low number of chips from olimex
<ZetaNeta>
oliv3r, Will try
<atsampson>
I'd probably be more interested in an ATX-size board with, say, 24 A20 chips on, though...
<wingrime>
ZetaNeta: also check eoma-68 price for ref
<oliv3r>
ZetaNeta: but olimex is about to release the A13/A23 CoM (computer on a module)
<ssvb>
wingrime: yeah, I will give this series a try right now
<wingrime>
techn_: I not understand point of you patchset
<oliv3r>
After appearing on the list for the first time in 2012, Microsoft notably dropped off the list entirely this year.
<oliv3r>
*g*
<oliv3r>
big supprise there
<oliv3r>
they didn't imrpove linux really; only added their hypver-V crap
<oliv3r>
and now that it's in; they don't have to do anything really anymore
<wingrime>
techn_: negative duty ?
<oliv3r>
which is fine, but it makes it sound like the worked really hard before
rellla has joined #linux-sunxi
<wingrime>
oliv3r: can you review thats very small patch [PATCH] sunxi:emac Make EMAC builds by default as module
<wingrime>
oliv3r: no any comments around my patches
<oliv3r>
wingrime: the patch is fine; it's a political desision that's not upto me
<oliv3r>
personally, all the allwinner intergrated core IP should be available as Y or M imo
<wingrime>
oliv3r: ... thats only default chose
<oliv3r>
if one really doesn't want it; just rm ip.ko or just rebuild kernel without it
<oliv3r>
wingrime: i agree
<oliv3r>
i'll mail it
<steev>
what all works/doesn't work on mainline?
<mnemoc>
wingrime: ack it on the ML please
<oliv3r>
wingrime: there :)
<oliv3r>
steev: not too much; but emac for one :)
<wingrime>
mnemoc: already
<steev>
oliv3r: haha
<mnemoc>
steev: you can get a prompt booting an initramfs ;-)
<steev>
mnemoc: indeed
<wingrime>
who want ACK emacs deafult build as module?
<steev>
i've got it all put together, but unfortunately, i've gotta do some work for a couple more hours first
<wingrime>
*emac
<steev>
wingrime: i would but uh, i normally only ever give Tested-By
<oliv3r>
wingrime: allready merged :p
<arokux2>
mnemoc, hm.. why not as default in Konfig?
<mnemoc>
the defconfig, not the default m
<mnemoc>
arokux2: waiting for comments
<mnemoc>
and a patch :p
<arokux2>
mnemoc, I thought one liners are easier to edit then to apply a patch
<mnemoc>
arokux2: not from a phone :p
<arokux2>
mnemoc, :D
<mnemoc>
also, not sure if depends on are desired
<mnemoc>
and also oliv3r's comment about the phy
<mnemoc>
so it deserves a real discussion/patch
<arokux2>
oliv3r, but RTL PHY is a board thing, not SoC's?
<oliv3r>
yep, that's a seperate chip
<wingrime>
arokux2: you can comment my patch
<oliv3r>
you can connect a random amount of PHY's to the SoC's mac
<oliv3r>
wingrime: your patch is allready merged
<wingrime>
oliv3r: where
<oliv3r>
wingrime: ML
<mnemoc>
which one?
<arokux2>
oliv3r, so what does it have to do with emac?
<oliv3r>
wingrime: linux-sunxi
<oliv3r>
arokux2: a network connection requires a few bits and pieces
<mnemoc>
which patch, wingrime has a ton of pending commits
<wingrime>
[PATCH] sunxi:emac Make EMAC builds by default as module
<oliv3r>
mnemoc: true :p
<oliv3r>
wingrime: that one is pushed
<oliv3r>
arokux2: first, you obviously have the CPU, IRQ's and busses; so lets ignore that for a minute
<mnemoc>
oliv3r: no no... only the sun7i_defconfig part
<mnemoc>
not the default in Kconfig
<arokux2>
oliv3r, wait. i've got it. but different boards use different PHYs?
<oliv3r>
arokux2: you then have the MAC layer, which connects to the AHB bus in this case, it basically IS the networking interface, but you can't just connect networking wires to a SoC :)
<mnemoc>
they all use the same, from AW's reference design
<oliv3r>
arokux2: a PHY connects to a MAC and the actual ethernet transormers
<mnemoc>
and cheapest available I supose
<oliv3r>
arokux2: the PHY does things like autonegotiation, transimission etc etc
<wingrime>
I just want comments , community's answer , ACK or NACK, none of this worser
<arokux2>
oliv3r, let me ask: is PHY on the SoC or is it external to it?
<oliv3r>
arokux2: the PHY and MAC are connected via a bus called MII (or R(educed)MII or GMII (for gigabit)
<oliv3r>
arokux2: it depends, intel cards have everything in 1 chip
<arokux2>
oliv3r, ... sunxi here :)
<oliv3r>
arokux2: some socs have it all onboard (i don't know of any), A10 has the MAC and PHY seperated
<oliv3r>
arokux2: SoC has MAC (Ethernet MAC)
<oliv3r>
A20 has GMAC (Gigabit ethernet MAC)
<oliv3r>
but they both need to talk to a PHY to connect to the outside world :)
<wingrime>
arokux2: thats fex will decide witch one we have
<wingrime>
arokux2: we have two separated modules or all build in it does not matter
<oliv3r>
dts in mainline determines it
<mnemoc>
built-in imposes too much, m is good
<oliv3r>
on x86, the code decides which one i suppose
<oliv3r>
for emac, M is good, as most devices don't use it
<mnemoc>
and hope to be able to create the right platform_devices by scanning script.bin soon
<arokux2>
if PHY is external to a SoC (and potentially can be different in each board), why should we add PHY module to a SoC's defconfig?
<oliv3r>
arokux2: we don't
<mnemoc>
arokux2: one important detail is that our defconfig aren't aimed at mainline, but to real life users
<wingrime>
mnemoc: thats not enought , and more , it outdated
<wingrime>
mnemoc: better parse to dts
<oliv3r>
we add emac as M, since the mac is actually INSIDE the soc, so it's always potentially there
<arokux2>
oliv3r, right.
<mnemoc>
wingrime: when our dts model settles we can teach fexc to output a dts
<arokux2>
well, if for all realworld scenarios there is the same PHY why not build it as a module then? (with dependency on EMAC)
<atsampson>
I would have thought it makes sense to have everything that's Allwinner-specific built as modules, so breakages get caught quickly...
<mnemoc>
arokux2: fine with me
<oliv3r>
arokux2: but the PHY, could be one of thousands, or not at all
<mnemoc>
but 99% chances of been the same
<arokux2>
oliv3r, 1. there is always the same PHY so far
<mnemoc>
so fair default
<mnemoc>
as m
<arokux2>
oliv3r, 2. there is not a big deal.
<arokux2>
oliv3r, PHY isn't something big
<hno>
mnemoc, odd.. seems error is gone now. Something about aiksmet.com or something like that. Didn't save it unfortunatley.
<oliv3r>
arokux2: MANY devices do not have a PHY at all though ;)
<mnemoc>
hno: o_O .... I didn't know we were using a CDN....
<hno>
it was an inline error page in the rsponse page when saving the edit.
<wingrime>
arokux2: phy external becose IC is cheap enouth and tablets not need it, and phy IP const mouney and ASIC place, then chip smaller then cheaper
MadSpark has quit [Remote host closed the connection]
<arokux2>
wingrime, got it.
<arokux2>
oliv3r, I see. but sane default is sane default. it is not a minimum config, right?
<mnemoc>
does patrick wood have a github? chinese driver renaming patches aren't ML friendly :(
<hno>
mnemoc, a wordpress CDM module wrongly enabled which should not be there?
<hno>
CDN.
<hno>
err, mediawiki even..
<mnemoc>
Turl_: -----^ ?
<Turl_>
mnemoc: pong, what's up?
<ssvb>
mnemoc: the Chinese comments are mangled in his g2d files move patch
<mnemoc>
Turl_: do we have something using aiksmet.com in the wiki?
<wingrime>
mnemoc: thats is not realy big for poll
<mnemoc>
ssvb: that's why I hoped I could do a pull from somewhere
<Turl_>
mnemoc: I don't know aiksmet, we're not using akamai afaik (CDN), but we do have akismet the antismap solution
<wingrime>
mnemoc: he have git but I don't know nick
<mnemoc>
Turl_: so akismet's antispam is failing on hno
<mnemoc>
wingrime: I found two patrick woods on github, but no linux-sunxi fork :<
<Turl_>
I was considering turning it off as spammers seem to have moved on to some other side
<Turl_>
site*
<Turl_>
hno: what error are you getting?
<wingrime>
mnemoc: I found him
<hno>
it didn't block anything, just spewed a screenful of mediawiki error / traceback.
<oliv3r>
as they quickly got superseeded by EOMA68-A20
<arokux2>
oliv3r, it is not mimimum defconfig and it is not an opportunistic (default) config
<arokux2>
oliv3r, so the CPU in my head is confused.
<Turl>
hno: I find it odd that lkcl_ would push for closed HW
<hno>
Turl, the plan is to get it opened up, but not quite there yet.
<oliv3r>
arokux2: many voices count :p
<oliv3r>
arokux2: but the defconfig is for most users, and for the hwpack
<oliv3r>
so all functinality for the SoC should have drivers/support; period
<techn_>
wingrime: that pwm patch set is just base for next disp patch
<mnemoc>
(and for testing commits before pushing)
<oliv3r>
if you don't have all soc functionality is kinda silly
<arokux2>
oliv3r, so TS should be built by defconfig?
<oliv3r>
M
<wingrime>
techn_: I will wait
<techn_>
what?
<wingrime>
techn_: your disp patch
<techn_>
I'll send it after pwm stuff is aced
<techn_>
acked
<techn_>
not aced ;p
<hno>
arokux2, yes, TS drivers should all be included in defconfig.
<oliv3r>
well the TS question goes two way
<wingrime>
techn_: can't do this unill I get what poit of this
<oliv3r>
the sunxi-ts (the resistive one) should absolutly be M; as its part of the SoC
<oliv3r>
the other ones should be too; since a lot of sunxi hardware is tablets
<oliv3r>
wingrime: what emac patch; yours I acked
<hno>
Yes.
MadSpark has quit [Ping timeout: 264 seconds]
<ssvb>
techn_: sun4i has PLL4 set to 960MHz, mali and cedar are just using different divisors (mali uses divisor 3 for 320MHz, and cedar is initially using divisor 6 for 160MHz but allows to change it at runtime via ioctl)
<hno>
and none should be Y as there is nothing that depends on them for booting.
<ssvb>
techn_: sun5i can possibly do the same
* mnemoc
hopes A23 is sun7i and not sun8i
<wingrime>
oliv3r: [PATCH] sunxi:emac Make EMAC builds by default as module
<wingrime>
oliv3r: not see anything
<oliv3r>
wingrime: it's acked AND pushed by mnemoc
<hno>
I would argue that EMAC should be built-in.
<mnemoc>
oliv3r: different commit
<oliv3r>
I think atleast M; Y would make some sense too
<oliv3r>
mnemoc: oh
ynezz has quit [Ping timeout: 264 seconds]
<oliv3r>
mnemoc: now i'm confused
<wingrime>
oliv3r: no menmoc pushed only defconfig patch and not mine
<mnemoc>
oliv3r: I pushed a change to sun7i_defconfig
<techn_>
wingrime: it's pwm driver and should be handled as is.. compare it to other pwm drivers
<oliv3r>
i think i deleted the mail then :p
<hno>
we really should make a new sun7i_defconfig based on sun4i_defconfig. the current one is too different.
<mnemoc>
i'm fine with emac as default y too.... but might bring troubles with gmac and the phy
<oliv3r>
i would do M
<wingrime>
techn_: they use negative cycle value?
<wingrime>
mnemoc: we have fex for it
<oliv3r>
ideally i wanna use the GMAC driver for both 10/100 PHY's and 1000 PHY's and leave the EMAC pure for sun4i
<techn_>
wingrime: negative cycle value is workaround in current driver
<hno>
oliv3r, mnemoc, EMAC should be Y in defconfig to allow for network booting without requiring the use of an initramfs.
<mnemoc>
hno: but the phy? and gmac?
<oliv3r>
hno: i agree with mnemoc here kind of
<oliv3r>
but
<Turl>
mnemoc: or maybe I'm messing it up with A20 and A31 order
<oliv3r>
as long as sun7i and sun4i need to be seperate kernels, GMAC=Y on sun7i; EMAC=
<oliv3r>
Y on sun4i
<hno>
emac and gmac should coexists in the kernel. And yes, relevant PHY drivers also need Y obviously.
<wingrime>
hno: we neeed change many options for net boot
<mnemoc>
+1
<mnemoc>
good point, nfs y?
<arokux2>
i'm up to it. nfs as y +1
<hno>
yes, and some more.
<ssvb>
mnemoc: I'm primarily using nfs, +1
<wingrime>
mnemoc: better make netboot_config_a10
<Turl>
nfs client right?
<hno>
wingrime, why?
<arokux2>
I think we first need a definition for _defconfig, then it's easy.
<wingrime>
hno: tablets
<wingrime>
mnemoc: my point stll have enabled thinngs only in soc
<hno>
wingrime, the normal _defconfig should suit all uses. There can then be optimized ones for specific uses such as tablets where only a fraction of the drivers are needed.
MadSpark has joined #linux-sunxi
<arokux2>
if you are building emac as Y, then _defconfig is like sane default config which is *not* the minimal one
<oliv3r>
but normal users don't care if there's a driver in the kernel built in
<oliv3r>
also, fex disables unused stuff anyway
<mnemoc>
for each soc we have headless server, linux desktop and android use cases :( all VERY different
<hno>
yes, but same defconfig can apply (as long as you accep Android crap in server/desktop)
<hno>
server/desktop is all runtime configuration, do not need different compiletime config.
<wingrime>
oliv3r: but bug kernel ins't normal too
<mnemoc>
hno: GPU/VPU have strong memory penalty
<hno>
mnemoc, yes, but can be disabled at kernel cmdline.
<mnemoc>
ok
<oliv3r>
wingrime: i don't think it'll be HUGE
<wingrime>
mnemoc: we will count how much people for y vs m
<wingrime>
mnemoc: but we can have m in kconfig and y in defconfigs
<mnemoc>
ture
<mnemoc>
true
<hno>
yes, kconfig should default to minimal.
<mnemoc>
damn... still 3099 unread sunxi mails and 2021 unread arm-netbook mails :'(
<hno>
would not even set most things as m in kconfig, just leave them off.
<mnemoc>
hno: default off means fatter defconfig
<hno>
yes.
<wingrime>
mnemoc: all stuff must correctly setups if user have no config and run menuconfig with arch=ARM , and just set soc to sunxi, all must work without any config
<hno>
but many of the modules are not sunxi specific.
<mnemoc>
for testing purposes all sunxi-related stuff needs to be either m or y in sunxi defconfigs
<hno>
wingrime, sure, all config should be possible to set via menuconfig, and default kconfig should boot.
<hno>
mnemoc, not only for testing, but also for providing a sane default that suits most users.
<wingrime>
mnemoc: so, lets define what must be in kconfig defaults and default configs for each soc
<mnemoc>
most users need far more than sunxi-related stuff.... but where to draw the line is the problem
<mnemoc>
rm: wanna join the discussion?
<oliv3r>
default config should be minimal yes; but also the most usable for the most common scenario's
<oliv3r>
that's NFS for dev boards, TS for tablets
<mnemoc>
but we can't really have defconfigs for each soc for dev boards and for tablets separated
<wingrime>
mnemoc: my point now, have soc stuff in y/m default in kconfig, all other things must be in defconfigs
<oliv3r>
it's a sane default; everybody can always tweak
<oliv3r>
fedora kernels aren't all tiny and minimal
<oliv3r>
distro*
<hno>
no generic distro kernel is tiny and minimal... you only see those built for a single device.
<arokux2>
hno, +1
<hno>
mnemoc, we have a choice.between fat _defconfig, or separate config for hwpacks & build testing.
<arokux2>
exactly.
<mnemoc>
hno: I prefer to reduce complexity, so minimizing the number of parallel configs
<arokux2>
that is why first of all we need _definition_ of a defconfig.
<mnemoc>
i.e. fat _defconfig, as hramrach has submited a couple of times
<Turl>
you could reuse rm's or hansg's defconfigs
<mnemoc>
blindly dropping a defconfig on top of another kills history and reasoning
<arokux2>
mnemoc, +1 as Allwinner's devices aren't that tiny, so extra kernel image bits won't do any harm. (the size of the kernel image is the only disadvantage with the fat config that I see)
<wingrime>
i think def config can be fat, but basic soc stuff must just work form kconfig
<mnemoc>
defconfigs count as code too.... things need to be reviewable
<mnemoc>
wingrime: agree on Kconfig y/m and fat _defconfig
<mnemoc>
wingrime: but disagree in dropping an unreviewable patch on top
<wingrime>
mnemoc: not get in
<Turl>
mnemoc: I meant for buildtesting and hwpacking
<arokux2>
mnemoc, +1 for not dropping a random config on top of ours.
<hno>
most Kconfig stuff is quite reasonable, except for USB-OTG which is a mess..
<arokux2>
if we decide for a fat defconfig we'll have one quick by our own
<hno>
the sun7i_defconfig is "a random config".
<wingrime>
mnemoc: are you about patrick patch?
<oliv3r>
how is sunxi not for 1 specific device?
<arokux2>
hno, submit patches which clean it. :)
<hno>
$ wc -l sun?i_defconfig
<hno>
323 sun4i_defconfig
<hno>
1218 sun7i_defconfig
<hno>
250 sun5i_defconfig
<wingrime>
hno: we have more configs
<hno>
arokux2, I am trying to find time to clean up the defconfigs and fix kconfig stuff, but likely not in the next weeks.
<Turl>
hno: probably not 'savedefconfig'
<wingrime>
mnemoc: >but disagree in dropping an unreviewable patch on top
<hno>
Turl, it is a savedefconfig. Just very different from the other.
<wingrime>
mnemoc: not understand what are about
<arokux2>
hno, no problem, good that we know *what* should be done then bit by bit we'll do it.
<wingrime>
oliv3r: facepalm
<wingrime>
oliv3r: you acked other patch
<hno>
oliv3r, what do you mean by "how is sunxi not for 1 specific device?"
<wingrime>
hno: dynamic selection
<wingrime>
hno: sun5i-sun5i possible and not so difficult
<wingrime>
err
<wingrime>
sun5i-sun4i
<ssvb>
mnemoc: the Patrick's patches are missing Signed-off-by in stage/sunxi-3.4 :-/
<hno>
wingrime, I don't follow.
<wingrime>
hno: we can merge sun5i-sun4i for make it run single kernel
<hno>
wingrime, yes, but not likely to happen in sunxi-3.4 I think. But is a hard requirement for mainline.
<wingrime>
hno: If I send many patches happend
<hno>
wingrime, and is a quite different discussion from _defconfig
<wingrime>
hno: thats better than _defconfigs
<mnemoc>
ssvb: :(
<wingrime>
mnemoc: how about official night-builds?
<oliv3r>
hno: well fedora kernel is only for x86
<arokux2>
wingrime, first _defconfigs then you send your patches and reduce the code bloat even more.
<mnemoc>
wingrime: the bsp needs some love to support the current u-boot and a20
<oliv3r>
this kernel would only be for sunxi :)
<hno>
wingrime, you are welcome to unify the sunxi-3.4 kernel if you think it is worth the effort. Personally I would prefer to spend that energy on improving mainline.
<hno>
oliv3r, no, Fedora have kernels for many arches, including ARM.
<wingrime>
hno: ....
<arokux2>
ok, with everything discussed here, could we vote on what a defconfig should be?
<mnemoc>
unifying 3.4 helps mainlining too
* arokux2
is voting for a fat _defconfig
* hno
is voting for a fat _defconfig, maybe supplemented with a minimal one for just the minimal soc stuff without having to use menuconfig.
* wingrime
fat _defconfig (reasonable big) and soc stuff in default m/y (depending on boot requiment) in Kconfig
<oliv3r>
yeah but _defconfig for us, is onyl for sunxi anyway
<arokux2>
wingrime, !!! not "and" those things are not connected.
<arokux2>
oliv3r, it is called sunXi_defconfig, so yes it is only for sunxi.
<wingrime>
arokux2: worable configuration must be buildable without defconfigs only with kconfig
<wingrime>
*workable
<oliv3r>
so it should be a minimal config, but all sunxi hardware drivers, and all most common drivers included
<hno>
wingrime, yes ofcourse.
<arokux2>
wingrime, agree, but this is a *separate* issue.
<oliv3r>
it should have nfs boot support, input/output support that's most common
<wingrime>
arokux2: actualy linked ))
<arokux2>
wingrime, but separate :p
<hno>
wingrime, I don't see them linked even. Only separate. But fixing kconfig means smaller _defconfig.
<wingrime>
hno: all wifi / ts drivers must be in _defconfigs I think
<steev>
does the graphics stuff in 3.4 support CEA modes? i have a monitor here (SOYO) that is connected via hdmi->dvi, and unfortunately it's edid is kinda messed up, but if you force CEA modes, it works fine
<wingrime>
hno: End user must have more chances get workable system after firstbuild
<oliv3r>
wingrime: and end user will most likly rely ond efconfig
<steev>
defconfig is for the minimal to get the board itself running, not for adding in every wifi/ts driver
<wingrime>
oliv3r: not all wifi drivers in defconfig
<oliv3r>
it should have all the most common wifi/ts drivers
<oliv3r>
in the case of sunxi
<hno>
yes.
<oliv3r>
all drivers that have been fexified
<steev>
no
<oliv3r>
yes
<steev>
yes
<wingrime>
oliv3r: for example I have tablet that wifi not in config
<steev>
sorry, i thought you meant *all* wifi
<oliv3r>
wingrime: it should be
<oliv3r>
steev: all fexified wifi, all USB wifi i would even think too
<oliv3r>
actually, all usb devices should be M anyway
<steev>
no
<arokux2>
guys.. wait.
<oliv3r>
so a user can hotplug stuff
<steev>
only what would be on the board
<wingrime>
oliv3r: also, new drivers for ts in a20 sdk
<oliv3r>
since that's usb
<hno>
wingrime, we are not discussion what is missing, we are discussing what should be in _defconfig.
<steev>
that's up to distributions/distributors
<arokux2>
have we decided for a fat sunXi_defconfig?
<wingrime>
hno: thats example
<oliv3r>
arokux2: medium :)
<oliv3r>
for example, there's no point in having scsi adapters, PCI devices etc etc (even as modules) in :p
<oliv3r>
but USB devices, most/all; since a user could plug in random device
<wingrime>
arokux2: resonable fat for all usercases , mostly relay on modules
<arokux2>
oliv3r, why? what you actually lose if you have a nice beefy config where everything sane is supported?
<ZetaNeta>
so.....
<ZetaNeta>
how do i port linux to my TAB750?
<oliv3r>
arokux2: huge kernel, making it slow :p
<oliv3r>
ZetaNeta: did you read my email on the list?
<ZetaNeta>
oh
<hno>
arokux, have you seen an sunxi device with a PCI SCSI adapter?
<oliv3r>
ZetaNeta: follow those breadcrumbs and you will succeed
<arokux2>
oliv3r, PCI? what PCI?? PCI isn't even near the hardware we are dealing with
<oliv3r>
arokux2: but you wanted a fat kernel
<oliv3r>
arokux2: i can hook up a pci bus somehow via ductape to a few GPIO's
<oliv3r>
:p
<hno>
can we stop arguing about the same goal?
<arokux2>
oliv3r, fat = boot from NFS root, everything for Ethernet
<oliv3r>
arokux2: i'd concider that medium :D
<oliv3r>
anyway, bed time :)
<oliv3r>
nn
<arokux2>
oliv3r, :)
<arokux2>
ok
jemk has quit [Remote host closed the connection]
<arokux2>
oliv3r, fat is allyes_config
<arokux2>
oliv3r, for your definitions
<wingrime>
bedtime
wingrime has quit [Quit: Lost terminal]
<hno>
Ok. I think we are all mostly in agreement, except maybe steev?
<steev>
if it's hardware that will be *on board* it should be in the defconfig
<steev>
random usb devices, no
<hno>
so no USB storage in _defconfig?
<arokux2>
USB storage should be in, everybody can use it.
<hno>
I would draw the line on board or commonly used with sunxi devices.
<arokux2>
we should go for commonly used, I think
<arokux2>
stuff will be in the modules anyway... kernel image size won't get bigger
<hno>
yes.
<arokux2>
only build time will get affected
<arokux2>
but for hacking only first build will take time.
<hno>
or use kconfig defaults.
<hno>
and only enable what you need for your hacking..
<arokux2>
Turl, around?
<arokux2>
steev, so you agree with us?
<Turl>
arokux2: sup
<steev>
arokux2: yeah, i still don't think we should just add in any usb wifi devices (and i say that while owning just about ever chipset that has a wifi device that's in the kernel)
<popolon>
do you have information about Allwinner A40 ?
<popolon>
Cortex-A15 based ?
<popolon>
hop this is a mali GPU (400 or T6xx or other...)
focus_it has quit [Quit: Leaving]
panda84kde has quit [Ping timeout: 264 seconds]
<hramrach>
problem with not all USB wifi is how do you pick which you include in defconfig
<hramrach>
only adapters shipped with some boards? and which boards? and when a new board ships with different adapter you add it? and all old kernel builds will lack the driver