hno changed the topic of #linux-sunxi to: /Allwinner/sunxi development discussion - Don't ask to ask. Just ask and wait! - See | | Logs at
<leviathanch_> Turl: how do you pass this image to the kernel?
<leviathanch_> since u-boot refuses it
<leviathanch_> and I can't figure out the correct entry point
<leviathanch_> somehow
* leviathanch_ is using mkimage
* leviathanch_ googles
<leviathanch_> Turl: found it
<leviathanch_> :-)
<Turl> leviathanch_: I use CONFIG_INITRAMFS_SOURCE and embed it into the kernel
<leviathanch_> ok
naobsd has quit [Quit: Page closed]
robb83 has joined #linux-sunxi
robb83 has quit [Client Quit]
hipboi has joined #linux-sunxi
egbert has joined #linux-sunxi
egbert_ has quit [Ping timeout: 264 seconds]
<Turl> leviathanch_: any luck?
<Turl> leviathanch_: I just finished implementing automagic reparenting
<Turl> mturquette: are you around by any chance?
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
<leviathanch_> Turl: awesome!
<leviathanch_> and no
<leviathanch_> I had not any luck booting the initrd
<Turl> leviathanch_: it barfs about machine id or other seemingly random stuff?
<leviathanch_> worse
<leviathanch_> wait a moment
<leviathanch_> undefined instruction
<leviathanch_> pc : [<40008008>] lr : [<7ff93e28>]
<leviathanch_> sp : 7fe72a40 ip : 00000007 fp : 40328028
<leviathanch_> r7 : 7ffbc858 r6 : 40008000 r5 : 7ffbc858 r4 : 00000000
<leviathanch_> r10: 7fe7494c r9 : ffe0010e r8 : 7fe72f48
<leviathanch_> r3 : 00002504 r2 : 40ffa000 r1 : 000010bb r0 : ea000006
<leviathanch_> Flags: nZCv IRQs off FIQs off Mode SVC_32
<leviathanch_> Resetting CPU ...
<Turl> can you paste the full log from uboot? (use a pastebin)
<leviathanch_> ok
<Turl> XIP Kernel Image ... OK
<Turl> wingrime had that issue too
<Turl> it's not XIP
<Turl> leviathanch_: try loading the kernel somewhere else
<Turl> leviathanch_: I load it @0x40000000
<leviathanch_> nope
<leviathanch_> still the same issue
<Turl> still says XIP? really? :/
<leviathanch_> yes
<leviathanch_> ... -.-
<Turl> it should relocate it to 40008000 as per load address
<leviathanch_> I'm using bootm
<leviathanch_> that's correct
<leviathanch_> right?
<leviathanch_> ^_^"
<Turl> yeah, bootm 0xkernel 0xramdisk 0xdt
<Turl> leviathanch_: paste your uboot env
[7] has quit [Disconnected by services]
<leviathanch_> hmm
TheSeven has joined #linux-sunxi
<leviathanch_> obviously it doesn't like it when I'm loading the initrd
<leviathanch_> maybe I'm overwriting parts of the kernel in RAM?
<leviathanch_> done something wrong with the memory usage calculus=
<leviathanch_> ?
<leviathanch_> or maybe
<leviathanch_> hmm
<leviathanch_> wait a sec
<Turl> you can load the kernel last if you fear memory being overwritten
<leviathanch_> yes
<Turl> although your addresses are 16M apart
<Turl> I'd drop that mem=128M btw
<leviathanch_> ok
<Turl> leviathanch_: try replacing the address on *just* the ext4 load to 0x40007FC0
<Turl> that should let you boot
<Turl> we should ask wingrime then what was his final solution (if any) to this problem
<leviathanch_> you mean 0x40007FC0 as loadaddress for uImage
<leviathanch_> ?
<Turl> leviathanch_: yes
<Turl> keep the one on bootm the same though
<Turl> if that doesn't work, try with 0x40007FC0 on both places (bootm & ext4load)
<leviathanch_> then it stops
<Turl> I can imagine it complaining about the machine id
<leviathanch_> ?!
<Turl> load dt to 0x50000000, initrd to 0x51000000
<leviathanch_> ok
<Turl> and set fdt_high 0xffffffff
<Turl> and initrd_high 0xffffffff
<leviathanch_> uhm
<leviathanch_> all ones?!
<Turl> yep, 32 bit -1U :)
<Turl> that's the special value that means 'do not relocate'
<leviathanch_> ok
<leviathanch_> you mean like
<leviathanch_> setenv initrd_high 0xffffffff
<leviathanch_> ?
<Turl> leviathanch_: yes
<leviathanch_> uhm
<leviathanch_> it's not doing anything...
<leviathanch_> it's just hanging after
<leviathanch_> "Starting kernel..."
<Turl> leviathanch_: try with 40007fc0 on bootm too then
<leviathanch_> as you can see
<leviathanch_> my load address
<leviathanch_> is as well the address I give to bootm
<leviathanch_> except of course I'd be using reserved variables right now
<leviathanch_> of which I didn't know...
<Turl> leviathanch_: did you try ?
<leviathanch_> yes
<leviathanch_> and it doesn't work
Legitsu has joined #linux-sunxi
<Legitsu> hello is there a kernel dev in the house
<leviathanch_> multiple
<leviathanch_> why?
<Legitsu> I seem to be stuck with a project
<leviathanch_> which one?
<Legitsu> I got a one of these Rk3188 sticks a MK908 V3 and for the life of me I can't get any kernel I build
<Legitsu> save for one hack job of a source tree to work
<Legitsu> so what I need is a way to debug the kernel with KGDB
<Turl> leviathanch_: enable EARLY_PRINTK, DEBUG_LL, choose the sunxi uart on the submenu, and pass "earlyprintk" on kernel cmdline and play again then :)
<Legitsu> over usb or IP
<Legitsu> with a stick that wont boot
<Legitsu> ideas ?
<Turl> leviathanch_: get a serial console on it
<Legitsu> I wish I could but I am old and my eyes and hands
<Legitsu> wont do that kind of soldering anymore
<Legitsu> I saw omegamoon did a console mod
<Legitsu> he makes it look huge I can't even see the traces todo the work
<Legitsu> its odd because if you take this same non-working kernel and throw it on a V1 Mk908 it runs like a topp
<Turl> get someone else to do it for you? :p
<Legitsu> I am at a complete loss as to how to attack the problem
<Legitsu> iv tried litterally every Option in the defconfig
<Legitsu> i think I may have accually started to wear out the NAND
<Legitsu> from flashing kernels
<Turl> well, if you had a serial console, DEBUG_LL and EARLYPRINTK would get most of the job done from typical crashes
<Legitsu> I did take it to a local shop and showed them the pictures from omegamoon
<Legitsu> they said impossible
<Legitsu> bah if i was 10 years younger
<Turl> leviathanch_: btw, there's #linux-rockchip
<Turl> maybe they know
<leviathanch_> uhm
<leviathanch_> first I need to try sleeping once again
<Legitsu> whats sleep
<leviathanch_> it's already morning around here
<leviathanch_> again
<leviathanch_> >_>
<Legitsu> is that what the button on keyboard is for ?
<Turl> :)
<Legitsu> Iv never accually pressed it
<Turl> leviathanch_: I should too, past midnight
<Turl> Legitsu: PCs usually crash when you touch that
* Legitsu gets the caffine injections
* Legitsu presses it
<Legitsu> cool
<Legitsu> I can turn my computer on and off with that button
<leviathanch_> Turl: I'll try early printk tomorrow
<leviathanch_> gn8
<leviathanch_> :-)
<Legitsu> will early printk work fast enough
<Legitsu> because it litterally gets to the boot splash and reboots
<Legitsu> I would't even know where to add one
* Legitsu sighs
<Turl> Legitsu: it'll print on the serial console (that's the early part)
<Legitsu> that I dont have
<Turl> night
* Legitsu picks the stick up and throws it out the nearest window
<Legitsu> anyone want a slightly abused MK908
<Legitsu> only got run over with the truck twice
<Legitsu> I wonder if I could mail my stick to somebody and let them mess with it lol I don't even wanna look at the pos at this point
<Legitsu> ill even pay shipping just to get if off my desk so it can no longer taunt me
robb83 has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
ykchavan has joined #linux-sunxi
Legitsu has left #linux-sunxi ["Once you know what it is you want to be true, instinct is a very useful device for enabling you to know that it is"]
Ehsand has joined #linux-sunxi
Black_Horseman has quit [Ping timeout: 240 seconds]
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
Black_Horseman has quit [Changing host]
Black_Horseman has quit [Remote host closed the connection]
robb83 has quit [Quit: - A hand crafted IRC client]
atiti has joined #linux-sunxi
eebrah_ has joined #linux-sunxi
<ykchavan> Is there any link/webpage where there is mapping for known tablet/board with corresponding script.fex name?
<ykchavan> e.g. A10-MID (512mb) matches with hyundai_a7hd.fex
eebrah_ has quit [Ping timeout: 246 seconds]
<Tsvetan> allwinner sent me encoding and decoding demo
Ehsand has quit [Remote host closed the connection]
<oliv3r> Tsvetan: awesome
<oliv3r> wingrime: ^
atiti has quit [Ping timeout: 246 seconds]
Ehsand has joined #linux-sunxi
<Tsvetan> now I want to stream video with A20-SOM-EVB 2Mpix camera :)
shineworld has joined #linux-sunxi
<oliv3r> :)
<oliv3r> lets ses how complete their demo's are
<drachensun> hows your Mandarin?
<drachensun> looks like the comments are english
<oliv3r> drachensun: excellent why?\
<oliv3r> :p
<oliv3r> looks like enoding is h264 only
<oliv3r> but it's a start
arokux has joined #linux-sunxi
<drachensun> I was just joking because the docs they attached to that demon are in Mandarin (
<drachensun> can you actually read it?
<hno> oliv3r, yes, video capture is h264 only. At least according to specifications.
rellla has joined #linux-sunxi
<hno> wonder if the new demo differs anything from the earlier A10 demos.
<oliv3r> hmm good point
<oliv3r> drachensun: nah; i wish i could, but yeah saw that too :)
<oliv3r> hno: well i'm curious which codecs cedarX can encode
ykchavan has quit [Read error: No route to host]
<drachensun> <--- feels silly
<drachensun> I got excited for a sec
<drachensun> I just read they released the 3.2 SDK for A31
<drachensun> but I can't find a link anywhere, all the info seems to be in Chinese and google translate is having a hell of a time with it today, I can't seem to make any headway
<hno> 3.2?
<drachensun> they actually released it while back I think but news takes longer to reach the english news feed I have noticed
<drachensun> I got that number from somewhere else
<hno> oliv3r, specifications I have seen only says: HD Video Encoding (H.264 High Profile). Support encoding in H.264 format, 1080p @ 60, 720p @ 100 fps
n01 has joined #linux-sunxi
<hno> and jpeg.
<hno> Hm.. A20 spefifications is lower. 1080p@30
boycottg00gle has joined #linux-sunxi
boycottg00gle has quit [Remote host closed the connection]
boycottg00gle has joined #linux-sunxi
<jukivili> arokux: ok.. so.. musb config for host-only
<arokux> jukivili, yes :)
<jukivili> which defconfig you use as base?
<arokux> jukivili, well, first that, then I'll test everything else, I thought it is a good start..
<arokux> jukivili, hm... do no know actually ((( was long time ago
<Amitabha> who was interested on the eoma laptop?
<shineworld> A31... 5s speedy boot ... IS A DREAM ... I want it on A20 :)
<drachensun> shineworld: it sips power too, when its in that shutdown
<drachensun> I like the A31, I need to bring some of you guys around to it :)
<shineworld> drachensun, so isn't a true boot but a sort of hybernation ?
<mnemoc> why the *censored* are days so short?????!!!!
<drachensun> shineworld: I think so, its got an extra core for low power that brings it up quickly
<drachensun> at least I think that is how they are doing it
<drachensun> an ar100 or something? its not an Arm
<mnemoc> Amitabha: I would love an eoma laptop, but with "retina"+ display, real keyboard and 6h+ battery
<mnemoc> 11.6-13"
<jukivili> arokux: sun4i_defconfig + musb host-mode .. use stage/sunxi-3.4
<Amitabha> I saw a few emails, there are quite a few designs ideas around, right?
<Amitabha> mnemoc: is that display sourced already?
<oliv3r> hno: that's odd, especialyl since a20 is overclocked cedar mostly. and that example doesn't even do a JPEG encoding demo, which is a pitty; because i was hoping to get that
<oliv3r> Tsvetan: ^
<mnemoc> Amitabha: see
<mnemoc> Amitabha: see CHIMEI INNOLUX N116HSE-EA1
<mnemoc> and N133HSE-EA1
<drachensun> mnemoc: I'm supposed to have a sample coming or something like that
<mnemoc> but those are 1080p only
<drachensun> with an A31 though
<drachensun> of something like that I meant
<oliv3r> mnemoc: 1080p is 95% 'retina'
<drachensun> the tablet and keyboard are separate pieces but it docks
<mnemoc> not sure if eoma can drive more than that
massi has joined #linux-sunxi
arokux has quit [Ping timeout: 256 seconds]
<Amitabha> oh, from the mailing list i understood the project was stopped?
<oliv3r> mnemoc: 1080p is gonna be hard for the a10/a20
<oliv3r> mnemoc: i memory bandwith is a huge issue
<mnemoc> LP129QE1-SPA1 is 2560×1700 is 12.9"
<mnemoc> d/is/in/2
<mnemoc> err
<mnemoc> s/is/in/2
<mnemoc> like the chromebook pixel
<drachensun> can any Allwinner processor drive a display with that resolution?
BluesBoy has joined #linux-sunxi
<mnemoc> a31 maybe
<oliv3r> but only because of the dual channel memory controller
<drachensun> Display
<drachensun> Dual-Channel LVDS 1920x1080@60fps
<drachensun> 4 lanes MIPI DSI 1920x1200@60fps
<drachensun> Dual-channel RGB/CPU LCD 2048x1536@60fps
<drachensun> Integrated HDMI1.4 1080p@60fps
<oliv3r> lets pray that A40 will be better
<Amitabha> A40 already?
<Amitabha> lol
<mnemoc> A40 should be just an A20 but quad a7
<Amitabha> have not read much recently..
<drachensun> dont blink, you'll miss a product life cycle
<Amitabha> indeed!
<oliv3r> and i'm hoping it will be similiar to a20 and a31
<oliv3r> e.g. a20 with dual channel memory controller etc
<mnemoc> yes, damn memory bw
<oliv3r> well a lot of performance issues we have now is due to the 400~ MHz limitation of the single channel memory controller we have now
<Tsvetan> I hope A40 will be pin to pin with A20 :)
<Tsvetan> so no need for new re-design
<Tsvetan> yesterday hynix ddr3 memory factory burned in fire and price of ddr3 jumped 15% up
<Amitabha> :O
<mnemoc> hope you had stock...
<Tsvetan> I placed order few days ago, but didnt got confirmation
atiti has joined #linux-sunxi
<mnemoc> :(
<drachensun> ah hell
<drachensun> thats going to hit the whole market
<oliv3r> Tsvetan: yeah but hynix allready said t hat it shouldn't affect memory pricing
<oliv3r> Tsvetan: those reltailers are dirty hounds
<oliv3r> Tsvetan: thanks for the encoding demo; but it's missing jpeg encoding unfortunatly :(
<arokux1> jukivili, thanks. anything else that could be tested?
wingrime has joined #linux-sunxi
<Tsvetan> oliv3r memory price doubled for the last 8 month, so they are high enough, but this factory alone was with 10% of world sales capacity and droping the supply with 10% cant be insignificant
<hno> oliv3r, I have seen jpeg encoding in earlier demos.
<oliv3r> Tsvetan: well samsung or seagate also promissed harddrive prices shouldn't be up by more then 5% as the flooding had little impact, yet prices doubled and took more then a year and are still 25% more then what they where
<oliv3r> hno: oh really? interesting
<oliv3r> wingrime: ^
<drachensun> tsvetan: so 15% more, what was the actual cost of 1Gb?
<drachensun> I want to estimate the overall cost difference
<drachensun> on a tablet with 2Gb
<arokux1> Tsvetan, in the sdk 2.0 the kernel tree contains .git, however everything inside points to nonexisting directories, do you happen to have them?
<oliv3r> arokux1: nah, they do this on purpouse/by accident
<oliv3r> 1) they don't want you to have it
<Tsvetan> drachensun - the sales stop yesterday, the offers jump by 15%
<oliv3r> 2) i think they have an android tree and hence, the repo reference
<Tsvetan> nobody want to sell stock until see where the prices are going to be headed
<hno> arokux1, all SDK releases look like that
<drachensun> ok, but what was the cost before?
atiti has quit [Read error: Operation timed out]
<Tsvetan> drachensun I guess this depend from where you buy ;)
<arokux1> maybe you just were to shy to ask :p
<drachensun> I'm seeing sodimm prices of $7-20
<jukivili> arokux1: I think host-mode is enough for start.. is that cable that you got for musb device-mode capable?
<drachensun> I assume the raw chips are cheaper
vicenteH has quit [Ping timeout: 245 seconds]
<arokux1> jukivili, yes, actually is was used for device mode at the beginning. now I need to figure out what to do with two loose cables now, so that I can connect a usb device.
<arokux1> jukivili, but ok, let's start in host mode. so with that config you have given I compile and boot the kernel and plug a usb stick and it should work?
<jukivili> arokux1: yes.. hopefully it should show up
ykchavan has joined #linux-sunxi
<JohnDoe_71Rus> is it posible use hdmi and cvbs at the same time? cubie2 A20
<JohnDoe_71Rus> android
atiti has joined #linux-sunxi
ykchavan has quit [Ping timeout: 245 seconds]
<arokux1> mnemoc, your sanitize scripts are too cryptic for me (
<mnemoc> find -txpe -f | xargs -r .....
_enrico_ has joined #linux-sunxi
popolon has joined #linux-sunxi
<arokux1> mnemoc, the encoding of the code drop is iso-8859-1, but you seem to be fishing for GB* ones
Black_Horseman has joined #linux-sunxi
Black_Horseman has quit [Changing host]
Black_Horseman has joined #linux-sunxi
<oliv3r> UTF!
<mnemoc> arokux1: the script tries different encodings, you can try iso-8859-1 on the list too
vicenteH has joined #linux-sunxi
<mnemoc> the purpose is to standarize as utf8
<mnemoc> previous SDKs and leaks didn't have any iso-8859-1 bit
<mnemoc> only assorted variants of chinese
<arokux1> mnemoc, ok, maybe it's is better you do sanitization yourself... for now I've found a good point in android-3.4 to commit on top. should I upload AW's tree to github?
notmart has joined #linux-sunxi
notmart has joined #linux-sunxi
<mnemoc> i can't atm
<arokux1> ok, then
Ehsand has quit [Quit: Leaving.]
ykchavan has joined #linux-sunxi
\\Mr_C\\ has quit []
wingrime has quit [Ping timeout: 264 seconds]
iamfrankenstein has joined #linux-sunxi
Soru_ has quit [Read error: Connection reset by peer]
Soru has joined #linux-sunxi
<arokux1> jukivili, do you know if usb host is somehow special about dma aligning?
deasy has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
shineworld has quit [Quit: Sto andando via]
hramrach has joined #linux-sunxi
BluesBoy has quit [Quit: HydraIRC -> <- *I* use it, so it must be good!]
hipboi has quit [Quit: Leaving]
Black_Horseman has quit [Ping timeout: 264 seconds]
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
<arokux1> the recommended tree for users is 3.4 now?
eebrah_ has joined #linux-sunxi
eebrah_ has quit [Remote host closed the connection]
atiti has quit [Ping timeout: 240 seconds]
JohnDoe_71Rus has quit [Read error: Connection reset by peer]
hramrach has quit [Remote host closed the connection]
hramrach has joined #linux-sunxi
eebrah_ has joined #linux-sunxi
vicenteH` has joined #linux-sunxi
vicenteH has quit [Remote host closed the connection]
atiti has joined #linux-sunxi
eebrah_ has quit [Remote host closed the connection]
vicenteH` is now known as VicenteH
JohnDoe_71Rus has joined #linux-sunxi
panda84kde has joined #linux-sunxi
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<jukivili> arokux1: well, it's not.. but there is workaround for buggy usb drivers that pass wrong kind of buffers to usb-hosts which can result memory corruption (of driver's internal structures)
<arokux1> jukivili, ah, was waiting for you.
<arokux1> jukivili, the code from the sdk-2.0 doesn't have dma fixes
<jukivili> arokux1: at first I thought it was problem with dma-alignment with ehci/ohci.. but in the end it was not
<arokux1> jukivili, I was reading your thread a bit
<jukivili> one such (still) buggy driver is the upstream rtl8192cu
<arokux1> jukivili, but then the usb drivers should be fixed and we should drop the fixes?
hipboi has joined #linux-sunxi
<arokux1> jukivili, as far as I understood if ARCH_DMA_MINALIGN is defined, kmalloc will return correct alignment, or?
<jukivili> arokux1: I tried that.. getting 'compile-only tested' patches for large number random usb-driver in upstream is.. challenging
<arokux1> jukivili, ok, but the other SoC hci-driver seem to ignore this issue? only tegra is doing something like this
<jukivili> jukivili: yes, but networking drivers pass unaligned buffers.. it's allowed since that unaligned head/tail of buffer is not used by networking usb-drivers while urb is in-flight
fredy has quit [Excess Flood]
<arokux1> jukivili, how about dropping the fixes and waiting for a user of the broken driver. if he shows up, we can fix the broken driver and ask the user to test it?
fredy has joined #linux-sunxi
<jukivili> arokux1: I'd say drop those fixes for mainline version and keep for sunxi-3.4
<arokux1> jukivili, so that we do not need to fix anything in sunxi-3.4 usb drivers?
<jukivili> arokux1: right
<arokux1> jukivili, but then we won't have a chance to fish the users with hardware :)
<arokux1> jukivili, can you show what exactly is wrong?
Black_Horseman has quit [Ping timeout: 240 seconds]
<jukivili> arokux1: multiple _usb_read_sync can be active from different threads at same time.. and function takes buffer from preallocated buffer that has element size less than cacheline
<arokux1> jukivili, I see
<arokux1> jukivili, but shouldn't usb framework take care of it?
<jukivili> arokux1: then usb-host dma-maps those buffers that are on same cacheline.. at least sunxi and rasbperrybi does not like this too much
Tartarus_ has quit [Quit: ZNC -]
Tartarus has joined #linux-sunxi
<jukivili> arokux1: since function is 'sync', it's probably safe to fix _usb_read_sync to use mutex_lock/mutex_unlock there with single buffer
<arokux1> jukivili, are you sure those are the usb-drivers that are buggy? I mean it would be much nicer if usb framework takes care of it?
<jukivili> arokux1: maybe there would be performance issues if usb-core would be change to handle non-dma'able buffers (alloc new tmpbuf, copy, etc). Most of the drivers do the right thing so to fix small set of buggy drivers and cause performance regression for other makes little sense.
<arokux1> jukivili, ok, but in your earlier posts you said host controller drivers should take care of it?
<arokux1> jukivili, I'm confused a bit
<jukivili> jukivili: I was wrong then
vinifr has joined #linux-sunxi
<arokux1> jukivili, ah, ok. do you know a usb driver that respects dma alignment?
<arokux1> jukivili, yes, I've seen it too.
<arokux1> jukivili, (better to read here: )
<arokux1> jukivili, here I see usb host controller drivers should take care of this problem:
<jukivili> So, at first I thought the problems with rtl8192cu were because sunxi ehci not handling unaligned dma (and fixing unaligned dma with sunxi ehci fixed the issues with rtl8192cu).
<jukivili> because if you handle dma-unaligned buffers by copying to/from temporary buffer, cachelines don't overlap anymore
<arokux1> jukivili, that is clear. but I'm still confused about who is really buggy...
BluesBoy has joined #linux-sunxi
<jukivili> jukivili: rtl8192cu for using non-dma'able buffers
<arokux1> jukivili, but some guy Ming Lei says that "most of in-tree host controller can handle non-aligned buffer"
<arokux1> jukivili, and "If transfer buffers share one same cache line, it should be bug in driver, not fault of host controller."
<jukivili> arokux1: yes.. and sunxi ehci/ohci can handle non-aligned buffers too
<jukivili> arokux1: rtl8192cu is buggy, because transfer buffers share same cacheline
<arokux1> jukivili, sunxi ehci/ohci can handle them because there are "dma fixes", but I've seen something like this only in tegra
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier.]
BluesBoy has quit [Quit: Leaving]
<arokux1> jukivili, so I cannot understand why the claim "most in-tree host drivers handle not aligned buffers correctly"
<jukivili> arokux1: Problem with rtl8192cu was that there were multiple places that were wrong, and fixing one place didn't yield full success.. driver kept crashing machine. So I started looking at sunxi ehci and found working solution that was copying unaligned-dma fix from terga to sunxi.
<jukivili> arokux1: only one in-tree host driver (terga) does not handle non-aligned buffers correctly
<jukivili> well.. there are few more if you dig deeper but not the majority
<arokux1> jukivili, hm... I thought "dma fixes" was there to handle the buffers correctly...
<jukivili> arokux1: dma fixes in sunxi ehci/ohci?
<arokux1> jukivili, ok. let me ask: why for are {tegra,sunxi}_ehci_map_urb_for_dma functions? I've thought they are for handling non-aligned buffers correctly?
panda84kde has quit [Quit: Konversation terminated!]
<jukivili> arokux1: The sunxi_*_map_urb_for_dma , that was before I knew better. Now they are still there to prevent buggy drivers from crashing kernel.
<jukivili> argh! .. hit <enter> too early
vinifr has quit [Quit: Saindo]
<jukivili> arokux1: sunxi_*_map_urb_for_dma was to fix sunxi-ehci to not crash with rtl8192. At first I thought that sunxi-ehci could not handle unaligned buffers. But I was wrong, rtl8192cu is buggy.
<arokux1> jukivili, ok. what makes sunxi-echi able to handle unaligned buffers?
<jukivili> I don't know about terga_*_map_urb_for_dma, maybe their hardware cannot handle unaligned-dma (or they fixed same bug as I did)
<jukivili> arokux1: there is no reason why not.. the sunxi_*_map_urb_for_dma even had off-by-one error at beginning (temporary buffers misaligned by one).. none has reported problems with that
<jukivili> problem with rtl8192cu is not about unaligned buffers, it's problem with overlapping buffers
<arokux1> jukivili, so that is not the code in sunxi-ehci but somewhere else that makes sure it works fine with unaligned buffers?
<jukivili> overlapping in sense 'two buffers in same cacheline'
ykchavan has quit [Ping timeout: 256 seconds]
<jukivili> arokux1: well.. kind of
<jukivili> arokux1: 'somewhere else' means dma-map all cachelines that hold part of buffer
atiti has quit [Ping timeout: 256 seconds]
<arokux1> jukivili, this was confusing. as all the conversations implied there is some extra work to be done in *each* host controller driver so that it can deal with non-aligned buffers. but now it is clear it is done at some central place. am I correct?
VicenteH has quit [Ping timeout: 264 seconds]
<jukivili> arokux1: well.. there is not extra code to handle non-aligned buffers. usb-core just calls dma_map_single/etc,
<jukivili> no extra central code, but if host-driver cannot work with unaligned buffer that is mapped with dma_map_single/etc, then it needs hacks like in terga driver
naobsd has joined #linux-sunxi
<arokux1> jukivili, ok. so my question now: if the sunxi_*_map_urb_for_dma tries to fix what is already working fine, why do we need it?
<arokux1> jukivili, as the purpose of sunxi_*_map_urb_for_dma is to make sure everything is fine if non-aligned buffer is passed. but it works already?
<jukivili> arokux1: it's preventing random crashes with buggy usb-drivers
<jukivili> arokux1: but yes, it could be removed and buggy drivers be fixed
<arokux1> jukivili, so sunxi_*_map_urb_for_dma is guarding against overlapping buffers in one cacheline really?
<jukivili> yes
<arokux1> jukivili, if so, it catches much more than that and causes copying overhead if the buffers do not overlap in one cacheline, but are non-aligned
<jukivili> yes.. true.. right.. better to get rid of it and just fix rtl8192cu :)
boycottg00gle has quit [Remote host closed the connection]
<arokux1> jukivili, ok...... now I clearly understand everything. thanks.
<arokux1> jukivili, is rtl8192cu fixed in mainline or is it still allocates buffers that are in one cacheline? I think you said it is still buggy, right?
<jukivili> yes.. I don't have hw (anymore) to test fix
deasy has joined #linux-sunxi
<arokux1> jukivili, I see. well, let us remove the sunxi_*_map_urb_for_dma and wait for the user with rtl8192cu? :)
<arokux1> jukivili, btw, which wifi module does Mele A1000 use?
<jukivili> arokux1: wiki does not mention about wifi in Mele
<arokux1> jukivili, it seems it has RTL8192cu
<jukivili> does lsusb show it?
<arokux1> jukivili, I'll have access to the device in the evening
<arokux1> jukivili, but from what I see here:,8618
<arokux1> jukivili, Mele has RTL8192cu
<arokux1> jukivili, and I have Mele
<arokux1> :)
<arokux1> jukivili, so if you know how to fix RTL8192cu, you can send a patch and I'll test it. Are you interested?
<jukivili> already working on it :)
<arokux1> jukivili, you are quick.
vinifr has joined #linux-sunxi
<arokux1> jukivili, btw, do you know how to trigger the bug, if sunxi_*_map_urb_for_dma are not there?
<jukivili> just use rtlwifi/rtl8192cu driver and try to connect to wpa protected AP with wpa_applicant
<jukivili> actually.. even scanning might enough to freeze
<arokux1> ok.
<arokux1> jukivili, here 3.4.4_4749 is the newest one, release in 2012/11/12. since then they are supporting mainline?
<jukivili> arokux1: not that one
<jukivili> arokux1: the upstream rtl8192cu .. driver/net/wireless/rtlwifi/
<arokux1> jukivili, so they are different? I thought they pushed their driver upstream.
<arokux1> jukivili, but yes, theirs code is completely different
<arokux1> jukivili, so upstream linux driver is not supported by realtek?
<jukivili> it's kind of supported, there's one developer working with realtek if I ahve understood correctly
<arokux1> jukivili, ok. you seem to have push lots of patches against it :)
<arokux1> pushed*
<arokux1> jukivili, cool!
<arokux1> jukivili, I assume the patch is against sunxi-3.4?
<arokux1> jukivili, once it works, I'll send patches that drop "dma fixes"
<jukivili> arokux1: yes
<mnemoc> is there anything critical missing in stage-3.4 ?
rellla2 has joined #linux-sunxi
<mnemoc> want to merge that junk, but i don't have time to think :(
rellla has quit [Ping timeout: 245 seconds]
<mnemoc> (or read the 3k pending mails)
<oliv3r> make a dir, move all mails from the merge date to that dir; then start reading from there
<oliv3r> everytime you have a few minutes, scan the old mails
<mnemoc> with two jobs and no permanent residence location that's pretty much impossible
VicenteH has joined #linux-sunxi
<leviathanch_> Turl: resolved _one_ issue
<leviathanch_> Turl: the load address must not be 0 but not the actual load address either
<leviathanch_> e.g. 0x42 will do
<leviathanch_> -.-
<leviathanch_> in mkimage
<arokux1> leviathanch_, what are you doing?
<leviathanch_> Turl: I'm trying to boot the linux-next kernel
<leviathanch_> ;-)
<leviathanch_> with an initrd
<leviathanch_> but the cpio image I embedd doesn't get used at all
<arokux1> leviathanch_, have you seen this?
atiti has joined #linux-sunxi
rellla2 has quit [Quit: Nettalk6 -]
vinifr has quit [Quit: Saindo]
<jukivili> what works with mainline? mmc? ethernet?
<oliv3r> ethernet
<oliv3r> mnemoc: :(
<jukivili> so getting even host-mode only musb support would be useful for having storage device
<arokux1> jukivili, yes.. very useful :), but usb host controller is even more useful since wifi module is behind one of them
rellla has joined #linux-sunxi
<mnemoc> oliv3r: but it seems i'll finally get a permanent residence location by the begining of october, and with real internet connection
<mnemoc> a year since my crap^Wdisinvolvement began :/
<oliv3r> a year allready?! wow
eebrah_ has joined #linux-sunxi
<hipboi> oliv3r, yes, a year now
<hipboi> this time of last year, cubieboard was just announced
jdeisenberg has joined #linux-sunxi
<jdeisenberg> I bought a Mele M5 set top box. Is there an existing version of Linux that will work on it?
<jdeisenberg> It's an Allwinner A20 ARM Cortex A7 1GB RAM, 8GB ROM.
<hipboi> jdeisenberg, you can try the cubieboard image
<hipboi> jdeisenberg, and replace the script.bin
<hipboi> jdeisenberg, sorry, cubieboard2
<jdeisenberg> Thank you. I tried the Mele A1000/2000, and it didn't work at all. ( I was using the script)
<arokux1> jdeisenberg, Mele A1000/2000 have A10 inside, you have A20 as cb2.
<jdeisenberg> That would explain a lot. :)
<jdeisenberg> I'm currently using a hackberry as the server for my intro to Linux course. It should be able to handle approx. 30 people using a shell simultaneously.
<jdeisenberg> Decided to see if I could use the Mele M5 also.
<arokux1> wow, there was a lot of politics...,10087,10087
<arokux1> or still is.. I somehow missed it
<hipboi> jdeisenberg, in that case, you need a new board
<hipboi> jdeisenberg, i think mele m5 can hardly handle that 30 people using a shell simultaneously.
<arokux1> hipboi, any updates for cubietruck, linux-sunxi community deserves to be informed first :)
tinti has quit [Read error: No route to host]
<hipboi> arokux1, only 50 samples will be made first
tinti has joined #linux-sunxi
<jdeisenberg> hipboi: The hackberry seems to be holding up OK. Time will tell.
<arokux1> hipboi, all of them already sold?
<hipboi> arokux1, no
<hipboi> arokux1, most will be sent to distributors as samples
<hipboi> i have about 5 can be handled freely
<arokux1> hipboi, will we have a chance to buy one of those? or should we wait for the second batch?
<hipboi> and 2 are reserved for the writer of the book about cubieboard
<hipboi> so there are 3 left
<hipboi> arokux1, so i think yes
<jdeisenberg> I'm also looking forward to the cubietruck (and the cubieboard is wonderful as well)
<arokux1> hipboi, so I'd like to have one :) but I'd like you to address the questions in my e-mail, have you got it?
<hipboi> jdeisenberg, i have some thing for you
<hipboi> jdeisenberg,
<hipboi> arokux1, no
wingrime has joined #linux-sunxi
<arokux1> hipboi, is it your e-mail: ? check for a message with subject "Preorder for Cubietruck"
<hipboi> arokux1, yes, the address is right
<jdeisenberg> hipboi: Looks interesting; thanks.
<hipboi> arokux1, not received, even in the spam
<jdeisenberg> I'm currently running Fedora on the hackberry, and have used GIMP and LibreOffice on it. It works pretty well; I haven't run into any major speed issues.
<hipboi> arokux1, cubietruck will use the same kernel as cubieboard
<hipboi> arokux1, case will be available later, not now
<arokux1> hipboi, what about the rest?
<wingrime> hipboi: aw have any plans about cortex a15 ?
<hipboi> wingrime, cortex-a15 will be beginning of next year
<hipboi> following the steps of allwinner
\\Mr_C\\ has quit []
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
rellla has quit [Quit: Nettalk6 -]
eebrah_ has quit [Remote host closed the connection]
<Tsvetan> is cortex-15 really good move for allwinner
<Tsvetan> rockchip seems are ahead again with cortex-a12 in development
<Tsvetan> to be released at same time
\\Mr_C\\ has joined #linux-sunxi
ykchavan has joined #linux-sunxi
<Tsvetan> wingrime something interesting
rz2k has joined #linux-sunxi
<wingrime> hipboi: very nice
<wingrime> Tsvetan: nice
<hipboi> wingrime, thanks
<wingrime> Tsvetan: we have strange codec AVS jizun
<wingrime> Tsvetan: its looks like some china-specific
<wingrime> oliv3r: ping
nove has joined #linux-sunxi
wingrime_ has joined #linux-sunxi
<arokux1> hipboi, what about usb hub, wifi and pcb layout of the cubietruck?
<hipboi> pcb layout is not available for sure
hramrach has quit [Remote host closed the connection]
<hipboi> wifi is from bcm
<arokux1> hipboi, is wifi supported by mainline?
ykchavan has quit [Read error: No route to host]
hramrach has joined #linux-sunxi
<Turl> leviathanch_: can you paste a log, including uboot output?
massi has quit [Quit: Sto andando via]
mnemoc has quit [Ping timeout: 264 seconds]
<wingrime> ssvb: ping
mnemoc has joined #linux-sunxi
jdeisenberg has left #linux-sunxi [#linux-sunxi]
<wingrime> hipboi: also I hope aw will sometimes fix, that two (possible) hw bugs in disp
n01 has quit [Ping timeout: 245 seconds]
rz2k has quit []
hipboi has quit [Quit: Leaving]
hramrach has quit [Remote host closed the connection]
hramrach has joined #linux-sunxi
deasy has joined #linux-sunxi
ZetaNeta has joined #linux-sunxi
ZetaNeta has joined #linux-sunxi
ZetaNeta has quit [Changing host]
<oliv3r> wingrime: pong
\\Mr_C\\ has quit []
<mturquette> Turl: i am now
<Turl> mturquette: :)
arokux has joined #linux-sunxi
<wingrime> oliv3r: hipboy sayed cortex a15 at year end
<Turl> mturquette: would you be willing to take something like this for composite?
<oliv3r> wingrime: hopefully big.LITTLE
_enrico_ has quit [Quit: Bye]
<arokux> why are you guys so happy to see new hardware coming, there is very little code for the existing chips :((
<oliv3r> arokux: 99% change it'll be compatible with what we have
<oliv3r> and if it's without powerVR its good
<oliv3r> allwinner isn't gonna respin everything and start from scratch;t hey'll build uppon what they have sof ar
<wingrime> arokux: corex a7 in a20
<oliv3r> and a7 == a15
<arokux> == but?
<wingrime> arokux: much faster
<arokux> ok, then :)
<mturquette> Turl: seems reasonable. i guess you really don't want to use a custom clock type?
<Turl> mturquette: I am using a custom clock type
<Turl> mturquette: but with composite I don't need to reimplement muxing and gating :)
<Turl> mturquette: is the diff for automagic reparenting
<Turl> I'd say it's pretty generic code, except for the fact that I cannot call clk_round_rate on itself because that calls determine_rate again
<arokux> hm.. something strange, why to mkdir to rm -rf it?
<Turl> arokux: weird way of 'mkdir lib' ?
<arokux> Turl, yes, very weird
<arokux> considering cp -r is enough
robb83 has joined #linux-sunxi
<WarheadsSE> arokux: also, A15 has the virtualization capability
<Turl> WarheadsSE: so does A7
<WarheadsSE> mm, didn't think the a7 had that.. just cause I hadn't bothered to look :P
<WarheadsSE> both have vfpv4 & neon, so
<WarheadsSE> makes sense
<arokux> anybody with sunxi-3.4 on mele?
<Turl> WarheadsSE: A7 is ISA-compatible with A15, so it should
<WarheadsSE> agreed
atsampson has quit [Ping timeout: 248 seconds]
<oliv3r> WarheadsSE: they are identical; a15 is just faster, but less power efficient
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<Turl> mripard: ping
<arokux> Turl, what you have running on your mele?
<Turl> mripard: I thought Greg was going to take sid for 3.12, maybe an ack + asking him to carry the dt would be best instead of queuing the dt bit for 3.13
<Turl> arokux: 3.4+debian
<arokux> Turl, our 3.4? it freezes for me (
tinti has quit [Ping timeout: 260 seconds]
<Turl> yes, although I haven't updated it in a while
<oliv3r> Turl: yeah greg said he'd take the sid driver for 3.12-rc1
<Turl> arokux: built on aug 28, uname doesn't list the exact hash unfortunately
<Turl> but it should be stage/sunxi-3.4 around that date
<Turl> oliv3r: yeah and maxime queued the dt for 3.13
<Turl> oliv3r: :p
<arokux> hm, I do not think something bad was done to the code
<arokux> Turl, could you post your kernel log plz?
<Turl> anything specific you're looking for?
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
<arokux> Turl, (mmcblk0p2): error: couldn't mount because of unsupported optional features (240)
<Turl> arokux: that's because you're probably used ext4 and it's not enabled in kernel
<Turl> you've*
<arokux> Turl, but later it works
<oliv3r> later it gets loaded byt a module?
<Turl> probably the mount code trying with all FS's :)
<arokux> but that is not the biggest problem
<arokux> the problem is the serial shell freezes
<Turl> arokux: it's trying ext3, then 2, then 4 and it works there
<oliv3r> maybe you formated using ACL or something and your kernel driver doesn't have it?
<oliv3r> like feature 240 and feature 244 ;)
tinti has joined #linux-sunxi
<oliv3r> disable ext3 and ext2 drivers and enable ext4 driver to handle both ext3 and ext2
<arokux> Turl, which config do you use?
<oliv3r> +2 points for turl for spotting it
<Turl> arokux: one of my own, all the things I don't need off :)
<arokux> Turl, wireless is off?
<Turl> yeah, I don't use wifi
atsampson has joined #linux-sunxi
<arokux> :D
<arokux> but that's not important
<arokux> Turl, is 8192cu is built as a module at least?
<Turl> $ zgrep 8192 /proc/config.gz
<Turl> CONFIG_RTL8192CU=m
<Turl> # CONFIG_RTL8192CU_SW is not set
<Turl> but it's not loaded
nove has quit [Quit: nove]
<arokux> Turl, I guess CONFIG_RTL8192CU_SW should be set
<arokux> but if not difficult try modprobe 8192cu
<arokux> Turl, I get this
<Turl> arokux: what's the problem?
<arokux> <3>hub 3-0:1.0: unable to enumerate USB device on port 1
<Turl> probably because it has no power
<Turl> did you edit the fex to power on the wifi port always?
<arokux> Turl, no
<arokux> i.e. I haven't touched it
<arokux> Turl, the power is set to all ports, right?
<arokux> not sure if these settings are correct
<arokux> i.e. sdio_wifi_used = 0
<hno> arete74_, that fex have power control for all three USB controllers.
<hno> sorry, mean arokux ^
<arokux> hno, have you had wifi running with ours sunxi-3.4?
<hno> haven't tried in a while. And when I last tried main focus was trying to get the mainline driver running, don'lt like the Realtek vendor driver very much.
<arokux> maybe my cleanups has broken usb, checking master sunxi-3.4
alcides has joined #linux-sunxi
<arokux> hno, is "sdio_wifi_used = 0" ok?
ykchavan has joined #linux-sunxi
<arokux> oh, it is indeed the case, something is broken in stage/sunxi-3.4
<Turl> arokux: sdio is unrelated, mele's wifi is USB afaik
<arokux> most probably one of my usb cleanups has broken smth
<arokux> Turl, thanks, I have no idea what sdio is..
\\Mr_C\\ has joined #linux-sunxi
<Turl> arokux: SD I/O
<hno> arokux, sdio is using the SD/MMC bus, same as sdcards.
deasy has joined #linux-sunxi
<wingrime> ssvb: ping
wingrime_ has quit [Ping timeout: 264 seconds]
alcides has quit [Remote host closed the connection]
eebrah_ has joined #linux-sunxi
geecko has joined #linux-sunxi
<arokux> hno, for some reason wlan adapter is identified as RTL8188CUS is it correct?
eebrah_ has quit [Remote host closed the connection]
<arokux> shit is it is correct
<arokux> how come people here,8618,10993,quote=1 believed it has RTL8192cu
<arokux> jukivili, ^
<jukivili> yes, it's identified as rtl8188cus, but rtl8192cu handles it
<arokux> jukivili, so 8192cu needs to be loaded?
<jukivili> arokux: default config has MAC80211 disable, you need to enable it before rtlwifi/rtl8192cu config option becomes visible
<jukivili> 8192cu is vendor driver
<jukivili> not mainline
<arokux> jukivili, I'm booting sunxi-3.4 now
<arokux> jukivili, hm... with default config I can see wlan0
<arokux> but nothing works with it
wingrime has quit [Ping timeout: 260 seconds]
<jukivili> default config enables vendor driver 'CONFIG_RTL8192CU_SW'
<arokux> jukivili, that is right. what should I enable then?
<arokux> there is: <M> cfg80211 - wireless configuration API
<jukivili> disable CONFIG_RTL8192CU_SW, enable CONFIG_MAC80211, CONFIG_RTLWIFI and CONFIG_RTL8192CU
<hno> I have never got the mainline rtl8192cu driver working well.
<arokux> Generic IEEE 802.11 Networking Stack (mac80211)
<jukivili> arokux: that's the one
<arokux> is there a possibility to search through a list in menuconfig?!
wingrime has joined #linux-sunxi
<jukivili> arokux: it's easiest to edit .config and then run 'make oldconfig'
<oliv3r> arokux: press the / key
<oliv3r> :)
<oliv3r> that should open a search box
<oliv3r> jukivili: it's not recommended to hand edit .config ;0
<jukivili> :O
<arokux> yes.. "/" is not what I want
<wingrime> ssvb: ping
<oliv3r> arokux: you wanted to 'search'
<oliv3r> otherwise press the highlihgted letter
<oliv3r> tha'ts why it's a highlight, so you can quickly switch
ykchavan has quit [Quit: Leaving]
<arokux> to search but in list item not in "everywhere"
<oliv3r> what are you searching for
<arokux> oliv3r, :) for a feature to search in list items :P
<arokux> oliv3r, suppose usb stuff is somewhere, I do not know highlited letter, so I want to search for this item
<oliv3r> but yeah, a ctrl-f would be nioce
<arokux> ok, there will be a path to it, but at the end node there is a big list
<Turl> newer kconfigs have a (N) next to the options
<Turl> you just press it and jump there :)
<oliv3r> i know what you mean now and face the same issue :)
<arokux> rtlwifi: Firmware rtlwifi/rtl8192cufw.bin not available
<arokux> :(((((( I solve one problem for another to appear!
<oliv3r> but that one is esay
<wingrime> oliv3r: can we claim that a10 boot fully opensource without BROM sources?
BJfreeman has quit [Ping timeout: 268 seconds]
<oliv3r> wingrime: for a while
<oliv3r> but not fully usable imo
<oliv3r> it's all testing stuff (lima cedar)
<wingrime> oliv3r: but BROM
<Turl> wingrime: BROM is not upgradable
<arokux> oliv3r, there no such bin in the kernel tree.... where is it then?
<oliv3r> and we might get it eventually as it's GPL violating
<oliv3r> lkcl_: ^
<wingrime> Turl: problem, that we can't claim that a10 boot fully opensourced...
<Turl> wingrime: FSF doesn't consider that as software
<wingrime> Turl: but it is
<arokux> jukivili, thanks! where should it be put?
<oliv3r> wingrime: and no soc is; i don't even think x86 is
<jukivili> /lib/firmware/rtlwifi/
iamfrankenstein has quit [Ping timeout: 264 seconds]
notmart has quit [Quit: notmart terminated!]
<wingrime> oliv3r: x86 worse here, rootkit in hw very posible nowdays
<Turl> wingrime: it is ROM, can't be replaced
<oliv3r> well coreboot makes it partially ok
<Turl> wingrime: only way to replace is replace HW
<oliv3r> but microcode in CPU is like BROM
<Turl> so BROM == hw for fsf
<wingrime> Turl: but it can be gpl'ed
<arokux> wingrime, why "problem, that we can't claim that a10 boot fully opensourced..."
<arokux> ?
<Turl> wingrime: by that criteria not fully opensource because you don't have ARM SoC design documents
<wingrime> Turl: ok, will wait opensparc/openrisc
<oliv3r> A10 is 100% GPL according to the FSF
<Turl> wingrime: RMS use MIPS PC
<Turl> I don't think free MIPS cores available?
<oliv3r> :p
<arokux> why do you guys care?
<oliv3r> but once (if ever) we have BROM code; well we kinda have a 100% free cpu according to oliver spec :)
<arokux> the most important is that our work will remain gpl
<arokux> and not stolen
<Turl> arokux: I dunno, I don't care for BROM as long as it calls my code :p
<wingrime> Turl: also there some ARM clones
<Turl> wingrime: does Lemote use them?
<oliv3r> well one nice thing about BROM code is that we can see if the 24k is 'fixable'
<wingrime> Turl: no
<Turl> wingrime: then? :)
<wingrime> Turl: if you buy some big fpga you can run them
<wingrime> Turl: also if you have money , you can do ASIC
<oliv3r> but BROM not being open is acceptable
<Turl> I don't have ASIC-level money :'(
<oliv3r> it's exactly the same as an intel or Amd cpu
<oliv3r> they also start running code from the bios at address XXX
<oliv3r> same with BROM :)
<wingrime> Turl: ~200k$
<Turl> oliv3r: intel or amd bios is on nvram :P
<oliv3r> Turl: start saving
<oliv3r> Turl: nono, i'm talking about BROM
<oliv3r> not bios
<Turl> well, they probably have a brom too that runs bios then
<Turl> wingrime: btw I added nodes for A20 mod0 clocks and other stuff
<Turl> wingrime: I will finish cleaning branch and push this weekend/next week
<oliv3r> and they have microcode in the CPU
<oliv3r> which is like the bro
<oliv3r> so there exists no 100% fully free CPU
<oliv3r> besides, after that, you start going after VHDL designs :)
<oliv3r> in 50 years from now, that may all be the case
<oliv3r> once it's proven how superior open source * is
<wingrime> Turl: whats with pll and sata clocks?
<arokux> jukivili, i'm one step further.
<Turl> wingrime: yes, pll6_sata
<Turl> haven't tried yet but should work
<Turl> docs look same as a10
<wingrime> Turl: very nice
<wingrime> Turl: also mripard have any success with mmc?
<oliv3r> wingrime: i think someone else picked that up now
<wingrime> oliv3r: sdio are same
<oliv3r> wingrime: leviathanch_ i think was starting on it
<oliv3r> speakign of lemote
<oliv3r> wingrime: nice one :)
<oliv3r> geek porn :)
<oliv3r> wingrime: when are you guys doing a jpeg encoder RE? :)
<wingrime> oliv3r: no, realy, we need something for trace
<oliv3r> wingrime: hno said we used to have a demo encoder
<wingrime> oliv3r: thats russain sparc
<wingrime> hno: ping
<Turl> wingrime: leviathanch_ picked it up, but still not working afaik
<wingrime> Turl: also, there some @problem@ with sata, I can't explain regs and bits it use
<Turl> wingrime: not in doc? (do we have doc?)
<arokux> jukivili, ok, so I have now called wpa_supplicant but haven't yet assigned IP to the wlan0
<wingrime> Turl: we have no docs for it
<jukivili> arokux: are you testing with or without dma-hack in sunxi-ehci
<wingrime> Turl: no docs for sata
<oliv3r> wingrime: i think it's from one of the first encoding demo's (or maybe it was only decdoing demo with jepg encoder added)
<wingrime> Turl: but its synaptic's ip
<arokux> jukivili, with sunxi_hcd_*_urb_for_dma
<Turl> wingrime: synaptics? or synopsys?
<wingrime> Turl: second
<arokux> jukivili, note, this is sunxi-3.4, not stage.
<Turl> wingrime: same as UART
<wingrime> Turl: but phy looks aw's
<arokux> jukivili, I do not know how to assign IP in ubuntu, dhclient wlan0 just hangs
<wingrime> Turl you have to configure some regs, for access to IP
<jukivili> arokux: well.. to get working rtl8192cu you'd need to try compat-wireless..
<arokux> jukivili, why?
<jukivili> arokux: a lot of patches and fixes have gone in to mainline since 3.4 for rtl8192cu
<arokux> jukivili, shit.
<jukivili> arokux: for testing dma-hack and that rtl8192cu patch sunxi-3.4 is enough
<arokux> jukivili, ok, I just want to test dam-hack for now.
<jukivili> arokux: because, if you now take off the dma-hack and try rtl8192cu, you should be able to crash your mele quite quickly
<arokux> jukivili, so I've issued:
<arokux> wpa_supplicant -B -D nl80211 -i interface -c /etc/wpa_supplicant/foobar.conf
<arokux> interface=wlan0
<arokux> now there should be something meaningful from: iw dev wlan0 link
<arokux> but I get "Not connected."
<arokux> jukivili, so you say it is ok?
<jukivili> does scanning work.. iw dev wlan0 scan
<arokux> jukivili, yes, scanning works.
<jukivili> if I remember right, without dma-hack scanning caused crash.
<arokux> jukivili, ok, I'm removing hack and testing.
<arokux> jukivili, btw, my cleanups has broken smth :(
<jukivili> :/
<hno> wingrime yes?
<hno> better to do as topic says....
<wingrime> hno: jpeg encoder
<wingrime> hno: are you tested demo?
<wingrime> hno: from aw
<oliv3r> hno: you mentioned the original demo with jpeg encoder demo; is it the one from the repo?
<arokux> jukivili, scanning works without dma hacks
<jukivili> arokux: how about wpa_supplicant
<arokux> jukivili, works too.
<arokux> jukivili, next I need to set IP
<hno> wingrime, haven't tested.
<arokux> jukivili, but i'm not ubuntu guy, and dhclient wlan0 hangs
<oliv3r> you puretty much wpa_supplicant for WPA2 stuff
<jukivili> arokux: ok.. lets remove the dma-hack. I'll try revive my mk802 and test with it later.
<arokux> jukivili, oh! wait! there are backtraces in kernel log!
<jukivili> :)
<arokux> I thought the kernel should crash, but it didn't
Olaff has quit [Quit: Cassos]
<arokux> jukivili, is that a "correct" crash?
<jukivili> it's not, just some bug in cfg80211
<arokux> jukivili, :(
<jukivili> or in rtl8192cu, but unrelated to what we are trying to catch
<hno> arokux, can't find the full demo today, but I do have libjpgenc.a and headers with a JpegEnc() call.
<hno> wingrime^
<arokux> hno, :)
<oliv3r> wingrime: so should be relativly easy to get working
<hno> the JpegEnc() call looks trivial. A10_enc_dec/include/enc_type.h:extern int JpegEnc(void * pBufOut, int * bufSize, int addrY, int addrC, int width, int height, int pixelFmt);
<wingrime> hno: nice
<hno> but... not entirely obvious.
<wingrime> hno: but who need jpeg encoding
<oliv3r> thumbnail generation, camera etc
<wingrime> hno: we have jpeg decoder but there no any real users
<wingrime> hno: also png decoder
<arokux> jukivili, let me know you catch the crash
<oliv3r> with libjpeg using it; there will be a lot of real users
<hno> well, I assume it would be meaningful in a photo slideshow app, to not burn too much battery on decoding largeish images.
<wingrime> oliv3r: jpeg lib work line - by - line
<hno> the decoder that is. I don't know who needs encoding.
<wingrime> oliv3r: you can try write some replacement
<oliv3r> wingrime: i doubt it can't work
<oliv3r> wingrime: for the very simple reason, that libcedarX uses libjpeg for jpeg enc/dec itself :)
<arokux> if stuff was set to be build as modules and I disable support for modules with the stuff be built into the image?
<oliv3r> they took libjpeg, and hacked in cedarX and made a big ugly blob for it:)
<wingrime> oliv3r: thats here for mjpeg, you can make surviliance camera with web interface on a10
<arokux> I should have bought i7
<wingrime> arokux: I still have core2d on mainpc
<arokux> kernel compilation takes time
<oliv3r> phenom II here :)
<oliv3r> x6 and happy for it
<oliv3r> only pentium-M 2.1 Ghz is intel of all pc's I have
<arokux> hardware is cheap in comparison to you waiting time
<oliv3r> waiting for what?
<oliv3r> i compile a kernel in less then a minute
<oliv3r> but i build on my server, which is only x4 Phenom and 1ghz less; but still 3 minutes
<arokux> oliv3r, so you build a specific module? that works of course
<oliv3r> wingrime: anyway, a fully featured libjpeg should be easy and really cool to be able to offload both decoding (browsing a gallery etc) and encoding (creating camera snapshots, creating lots of thumbnails etc) and if it can be used for mpeg2 encoding :p that's still good; though with h264 encoder that's a little less intersting
<oliv3r> arokux: no
<oliv3r> arokux: rm -r build; make
<oliv3r> make -j 10 to be fair
<arokux> oliv3r, where are the 10 cores comming from?
<oliv3r> number of cores *2 +1 :p
<oliv3r> but i find 13 abit over kill :)
<oliv3r> a make job is never 100% cpu for long in a kernel build
<oliv3r> firefox/libreoffice has some big modules that take almost a minute to compile yeah; then it's too much
<oliv3r> but on my x4 i do make -j 8 (the bsp does it for me)
<oliv3r> and when working on kernel stuff, only the first make takes that long
<oliv3r> all other makes are insane fast, since stuff i haven't touched, doesn't have to get recompiled anyway
<arokux> -j5 will put 100% on each of my 4 cores
<oliv3r> you have intel right?
<oliv3r> in that case, yeah -j5 probably will; as it's 2 cores with 2 HT fake cores, so your doing 2 * 2 + 1 anyhow :)
<oliv3r> if you have 4 actual cores, and then 2 HT ontop, make -j 10 sounds better :)
<arokux> model name: Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz
<oliv3r> yeah that's a SMT one
<oliv3r> so 2 real cores
<arokux> well i7 has HT, i5 not :((
<arokux> no, it has 4 cores
<jukivili> i5-3570K has 4 cores, no smt
<arokux> smt?
<oliv3r> Simultanious Multi Threading
<oliv3r> or sumat; Hyperthreading :)
<arokux> ok, HT in intel case
Soru has quit [Read error: No route to host]
<oliv3r> SMT is 'generic'
<arokux> ok
<oliv3r> though i don't think anybody else does it
<arokux> so you have 4 cores + SMT?
<oliv3r> think differently
<oliv3r> each core, gets one SMT unit
<oliv3r> so you see 2 cores
<oliv3r> they act as if they where 2 cores
<oliv3r> but basic stuff is shared between both cores (like ALU)
<oliv3r> so 2 cores, 1 ALU (very short sighted)
<oliv3r> which you can imagine isn't 2x the performance
<oliv3r> so 4 cores with SMT, your PC will see '8' cores
<oliv3r> but you don't have 8 ALU's, only 4
<oliv3r> but in the 3570k's case, your fine, that one ias only 4 core real cores, no SMT at all
<leviathanch_> Turl: are you using arch/arm/configs/multi_v7_defconfig in order to build you kernel?
Soru has joined #linux-sunxi
<oliv3r> leviathanch_: we both use that one pretty much; i think he also strips unused stuff
<arokux> oliv3r, yes, but how you are able to run -j10 :P
<oliv3r> leviathanch_: i remove all the other soc's i don't use, freescale etc
<oliv3r> arokux: well there are some performance gains with SMT, so on 2 cores 2 SMT (4 total) i'd run -j5
<oliv3r> arokux: with 6 cores (real cores that I have), 10 isn't even x2 :)
<arokux> hm.. intel cpu with 6 cores will be expensive
<jukivili> also if you are compiling on spinning disk some of the compiler processes are blocked by I/O, so running extra can help overall compiling throughput
<oliv3r> yeah
<oliv3r> i have spinning rust :)
<Turl> leviathanch_: pretty much yep
<oliv3r> 2 raptor velocistar in raid10
<oliv3r> so it's not horribly slow
<oliv3r> on my server i build on a 8 disk raid5
<arokux> no, SSD here
<arokux> leviathanch_, have you seen the wiki page I've shown to you?
<oliv3r> i wouldn't compile on a SSD :)
<oliv3r> i'd maybe do a tmpfs dir or ccache on tmpfs to build fast
<oliv3r> but trashing my ssd with lots of writes; i dunno :)
<jukivili> let it burn.. you'll buy bigger one anyway in few years
<arokux> exactly, you lifetime is much more expensive
<leviathanch_> arokux: yes, didn't help
<arokux> leviathanch_, from which branch is your kernel?
<arokux> ok
<wingrime> oliv3r: a20 still have zeroed serials?
<arokux> leviathanch_, which board?
<leviathanch_> cubieboard2
<oliv3r> wingrime: the ones i have; yeah
<leviathanch_> arokux: it has to be an A1X issue if I'm not doing anything wrong
<leviathanch_> since it's refusing to load the shell script within the initrd
<wingrime> oliv3r: even after new manual?
<wingrime> oliv3r: they not moved/masked it?
robb83 has quit [Quit: - A hand crafted IRC client]
<oliv3r> manual showed us that sid is MUCH bigger on a20; but all zero
<oliv3r> wingrime: see my commit :)
ssvb has quit [Ping timeout: 264 seconds]
<wingrime> oliv3r: cortex a15 cool
<wingrime> oliv3r: realy nice news form hipboy
<oliv3r> wingrime: but we guessed that 6 months ago i think
<oliv3r> it's only to be expected
<oliv3r> the bigger question is, mali vs powervr
<oliv3r> dual channel ram vs single channel
ssvb has joined #linux-sunxi
<oliv3r> upgraded defe and befe
<wingrime> oliv3r: I prefer modern 2d accel
<wingrime> oliv3r: mali/powervr does not matter, I want openGL
<wingrime> oliv3r: not ES
<wingrime> oliv3r: also question for mali
<wingrime> oliv3r: can lima in future emulate mali
<wingrime> *opengl
<wingrime> libv: it possilbe build opengl stack ontop mali hw?
<wingrime> libv: also can mali400 do iDCT ?
<wingrime> libv: or IQ, or colorconversion
<wingrime> oliv3r: according mali150 it can do iDCT and DCT and color space transform....
<wingrime> oliv3r: if you looks to docs
<arokux> jukivili, still around?
<oliv3r> wingrime: at 'what cost' however, but still kinda cool to have, offload ogg encoding/decodign :D
<arokux> this patch breaks things, the code indeed does something...
<arokux> jukivili, ^
<jukivili> arokux: hm.. maybe it's something that have to be run atleast once, by sunxi-ehci/ohci or by sw_usb_sunxi/sunxi-musb
<jukivili> initializing all usb-phys
<arokux> jukivili, who knows....
<jukivili> but now I have to sleep
<arokux> jukivili, good night
geecko has quit [Ping timeout: 260 seconds]
vinifr has joined #linux-sunxi
<arokux> mnemoc, please remove 5443e361 from stage/sunxi-3.4, it is broken.
<arokux> how do you guys quickly reboot your boards? I disconnect/connect power cable, but wonder if it can damage the board...
<Turl> arokux: `reboot`
<arokux> Turl, but if smth should be copied to SD card?
<Turl> arokux: `reboot` does a clean system reboot
<Turl> if something needs syncing it'll sync, etc
<arokux> Turl, say you want to write a new kernel to sd and you want to do it every minute
<Turl> arokux: I don't copy kernels to the sd
<arokux> Turl, tftp?
<Turl> yup
<Turl> can build & test many kernels/sec ;)
<Turl> /minute*
<leviathanch_> Turl: well
<leviathanch_> it appears to be crashing in the background
<leviathanch_> when I build the kernel from your branch
<leviathanch_> sunxi-clk
<leviathanch_> ...
<leviathanch_> [ 1.139107] VFP support v0.3: implementor 41 architecture 2 part 30 variant 7 rev 4
<leviathanch_> [ 1.151075] Registering SWP/SWPB emulation handler
<leviathanch_> [ 1.146776] ThumbEE CPU extension supported.
<leviathanch_> then byebye
<leviathanch_> that's why I wasn't able to get anything from sd card or initrd
<leviathanch_> it mayb have been working alread
<leviathanch_> *already
<leviathanch_> but since it crashed later on without an error message
<leviathanch_> I didn't get it
<leviathanch_> -.-
<arokux> leviathanch_, what is your config?
<leviathanch_> arokux: are you working on sunxi-clk as well?
<leviathanch_> maybe you could post me your defconfig
<leviathanch_> given your build is booting properly
<arokux> leviathanch_, i booted sunxi-next from mripards branch with multiv7 config
vinifr has quit [Remote host closed the connection]
<leviathanch_> arokux: ok
uwe_ has quit [Ping timeout: 245 seconds]
<leviathanch_> Turl: ok
<leviathanch_> it's really an issue within your branch
<leviathanch_> it just bootet fine into the initrd on sunxi-next
<arokux> leviathanch_, mripards works?
<leviathanch_> ...
<leviathanch_> arokux: yes
<arokux> leviathanch_, I've use a few clock commits from Turl's branch
<arokux> used*
<leviathanch_> hmm
<arokux> leviathanch_, take a look here:
<arokux> leviathanch_, you wont to work on clocks?
<arokux> want*
<leviathanch_> no
<leviathanch_> I'm working on mmc
<arokux> leviathanch_, it does not need DMAengine?
<leviathanch_> it has it's own
<leviathanch_> *its
<leviathanch_> you control it with the IDMA registers of the SD/MMC controller
<arokux> leviathanch_, ok. if you want clock stuff from Turl's branch, you could cherry pick needed commits as i did
<Turl> leviathanch_: what branch are you working on?
<Turl> s/branch/hardware/
<leviathanch_> Turl: sunxi-clk
<leviathanch_> ah
<leviathanch_> cubieboard2
<Turl> maybe my branch is missing A20 patches
<Turl> I'm testing as I work on A10S hardware, which is sun5i
<leviathanch_> Turl: should I just switch over to sunxi-next and cherrypick your patches as arokux suggested?
<Turl> that would be the easiest route I suppose
<leviathanch_> ok
<Turl> leviathanch_: I found a couple of bugs that I fixed, implemented automagic reparenting and added dt nodes for A10S/A13/A20
<Turl> I need to do a bit more testing and review, but I should be pushing an updated branch one of these days
Black_Horseman has joined #linux-sunxi
<leviathanch_> Turl: ok
popolon has quit [Quit: Quitte]
<arokux> now I have a multimeter
<leviathanch_> arokux: wow, I've got an oscilloscope
<leviathanch_> two multimeters
<leviathanch_> soldering iron and edging equipment
<leviathanch_> ;-)
<leviathanch_> ah, and four fishing rods
<leviathanch_> xD
<arokux> which multimeters?
<arokux> leviathanch_, ^
<Turl> leviathanch_: I have neither of those :(
arokux has quit [Remote host closed the connection]
wingrime has quit [Ping timeout: 256 seconds]
naobsd has quit [Quit: Page closed]
<leviathanch_> Turl: well
<leviathanch_> helps with debugging
<leviathanch_> btw
<leviathanch_> it's one of your patches
<leviathanch_> which prevents the cubieboard from booting
<leviathanch_> after cherry picking them
<leviathanch_> also next stopped working
<leviathanch_> when I do clk_prepare_enable
<leviathanch_> wth happens?!
<leviathanch_> basically the mod0 should just care about it's own business as well as the ahb hate
<leviathanch_> *gate
<leviathanch_> but somehow it's killing the clocking