ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
moofish has joined #linux-sunxi
<moofish> *Hypothetically* if someone happened to have laid their hands on lichee for a1x,a20 and a31 - Who would be best to talk to about making use of it?
vinifm has joined #linux-sunxi
<Turl> moofish: you mean kernel source?
<Turl> moofish: mnemoc ^^^
<moofish> Turl: ;)
<moofish> Turl: Check PM
<vinifm> Turl, ping
<Turl> vinifm: pong
<vinifm> ?
<Turl> vinifm: there is some patches on ML for USB, can you try with those?
<vinifm> you think I should delete my post and post to https://groups.google.com/forum/#!topic/linux-sunxi/lnfpFszuUH8
<vinifm> ?
moofish has left #linux-sunxi [#linux-sunxi]
<vinifm> the kernel is dying in sunxi_hcd_free_temp_setup() or sunxi_hcd_free_temp_buffer()
<vinifm> from sunxi_hcd_unmap_urb_for_dma(): sw_hci_sunxi.c
ZaEarl has joined #linux-sunxi
<vinifm> i have to go, bye
vinifm has quit [Quit: Ex-Chat]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
rz2k has joined #linux-sunxi
Skaag has quit [Read error: Connection reset by peer]
Skaag has joined #linux-sunxi
jochensp has quit [Ping timeout: 248 seconds]
ZaEarl has quit [Ping timeout: 244 seconds]
hansg has joined #linux-sunxi
jochensp has joined #linux-sunxi
rz2k has quit []
<Turl> mnemoc: ping
<mnemoc> Turl:
<Turl> mnemoc: may I /msg?
<mnemoc> anytime
<gzamboni> mnemoc did you start using kicad ?
<mnemoc> gzamboni: haven't had time yet
<gzamboni> im finishing routing :)
<gzamboni> will send gerber files to make the pcb next week after revision
hansg has quit [Remote host closed the connection]
<mnemoc> gzamboni: what pcb?
<gzamboni> the extension board for cubieboard
<gzamboni> i think i will try out to make an A13 with smsc lan8512
<gzamboni> the tfpg package so i can solder it by hand
<gzamboni> A13, sdram, nand, mmc and the mac+phy from smsc with gpios of course
<mnemoc> nice
<gzamboni> without usb , hdmi, audio or any other connector, all them will be in the gpios
<gzamboni> maybe i will try out to set up power over ethernet in the circuit
ganbold_ has joined #linux-sunxi
<mnemoc> gzamboni: so you'll have a 2mm socket on one side, and male 0.01" pins on the other?
<mnemoc> doh. just noticed the png is the cubieboard itself and not an equaly sized extension board
vinifm has joined #linux-sunxi
rellla has joined #linux-sunxi
<gzamboni> :p this is just for people to know the positioning of the headers. DWG is proprietary and i coudnt open in linux here
<gzamboni> hipboi posted only the positioning in the dwg
<mnemoc> there is dxf iirc
<mnemoc> but yes, a png is good
<gzamboni> its not complete, i will have to add some stuffs like the drills
<mnemoc> hno: /q ?
<vinifm> mnemoc: as soon as I get in the business I'll test the USB patch
<mnemoc> vinifm: great!
<mnemoc> vinifm: general feedback about stage/sunxi-3.0 is greatly appreciated too
<vinifm> i am using stage/sunxi-3.0
<mnemoc> :)
<vinifm> because sunxi-3.0 does not work for a13-olinuxino-micro
<mnemoc> explain
<vinifm> actually I do not know why :)
<vinifm> memory maybe, 256MB
<xenoxaos> i think the mali is hardcoded for memory location in the upper half of 512 iirc
<mnemoc> yes, but for a sub-512 board you can disable mali, cedarx and the rest of the reserves
<vinifm> ok
gzamboni has quit [Read error: Connection reset by peer]
<vinifm> so, now #arm-netbook is for discussion about what?
<mnemoc> arm-netbooks, eoma, random sexy new arm-based devices, etc etc
<vinifm> hum :)
<libv> vinifm: anything that isn't sunxi :)
gzamboni has joined #linux-sunxi
hramrach has joined #linux-sunxi
vinifm has quit [Quit: Ex-Chat]
hansg has joined #linux-sunxi
paulk-desktop has joined #linux-sunxi
ZaEarl has joined #linux-sunxi
hansg has quit [Quit: Leaving]
ganbold_ has quit [Remote host closed the connection]
ganbold__ has joined #linux-sunxi
paulk-desktop has quit [Quit: Ex-Chat]
ZaEarl has quit [Ping timeout: 244 seconds]
vinifm has joined #linux-sunxi
vinifm has quit [Client Quit]
vinifm has joined #linux-sunxi
<vinifm> mnemoc, I did poorly on the delete posts from https://mail.google.com/mail/u/0/?shva=1#inbox/13d2398ae09d087b?
<mnemoc> it's a mailing list. you can't delete/edit posts
<mnemoc> stuff is already sent to everyone's mail
<vinifm> ok
<vinifm> Does anyone know why all printk is shown twice when using stage/sunxi3.0 on a13-olinuxino-micro?
<hramrach> why can't you diff a moved file in git?
<vinifm> sorry, i did not understand
<hramrach> vinifm: that's unrelated
<hramrach> the double printk is probably due to early console not going away when the regular console is enabled
<hramrach> note that early messages only apper once
<Turl> hramrach: you can, use the -M flag to find renames
<hramrach> Turl: I have the file *everywhere* but whatever file I diff git always diffs with /dev/null
<Turl> did you change the file much?
<hramrach> and, I already did git mv but I want to swap the mv with some changes to the file
<Turl> if it's too different it'll consider it a different file
<Turl> you can use -C to find copies too
<hramrach> that's completely fine
<hramrach> so long as it diffs the damn file with whatever I have on disk
<Turl> there's also --find-copies-harder
<hramrach> it's right there. in *both* the place from where the file was moved from and in the destination
<hramrach> but git uses /dev/null, not the on-disk file
<hramrach> well, you check out the latest commit, move the file someplace else, roll back before the move, move the modified file back, get diff :s
<hramrach> but without such dances git is totally unable to give you diff of a file that has moved. Because you told it you moved it. so whtever is on the disk cannot possible be the moved file, right?
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
ganbold__ has quit [Remote host closed the connection]
ganbold__ has joined #linux-sunxi
* Turl still doesn't understand what you tried to do hramrach
<Turl> you had a file named A and renamed/moved to B and git is diffing it with /dev/null?
<hramrach> yes, exactly that
<Turl> git diff -M doesn't work?
<hramrach> well, maybe
<hramrach> but it *was* aware that the file has moved
<hramrach> otherwise it would pick the on-disk file and not /dev/null
<hramrach> oh, and I was making diff with not checkoed out branch, too
<Turl> -M works, I just tried with a test repo
rz2k has joined #linux-sunxi
<hramrach> good to know :)
<mnemoc> problem is git doesn't really track file. it tracks their content
<hramrach> if it was so it would use the on-disk content
<mnemoc> and so finding copies or moves is expensive
<hramrach> I did have the damn file
<mnemoc> git doesn't care
<hramrach> but since it was a moved file it would not diff it
<hramrach> obviously it does care
<hramrach> I have some general idea how git works
<hramrach> just some details of git oeration are really asinine
<mnemoc> that article explains how git internals work
<hramrach> like when you tell it that a file moved it will *never* difff that file with something on-disk anymore
<mnemoc> and why things work like they do
<hramrach> I am sure this detail is not fundamental to git operation
<Turl> hramrach: if you use -M it does
<hramrach> which is totally obvious that you need to detect moved files when you diff a file that is not moved
<hramrach> in whatever sense you would want it the file is not moved
<hramrach> except it's marked as moved in git
<hramrach> so it will never diff it unless you include move detection
<hramrach> sooooo awesome
<hramrach> there you go content tracking and not file tracking
<mnemoc> Turl: the a13 code in the last drop includes hno's watchdog driver
<hramrach> hehe
<mnemoc> at least they didn't mangle the header :)
<Turl> mnemoc: hah really
<Turl> mnemoc: did they fix the bug it has?
<mnemoc> Turl: no, identical to what we have in sunxi-3.0
* mnemoc hates wireless headsets that don't let you wake them up :|
ssvb has quit [Quit: Leaving]
Skaag has quit [Ping timeout: 256 seconds]
ssvb has joined #linux-sunxi
Skaag has joined #linux-sunxi
<vinifm> anyone ever seen this error: http://pastebin.com/7fdHPAbm
<vinifm> bug*
<hramrach> you mean it is always disconnecting and reconnecting?
<hramrach> not sure I saw exact smae thing
<hramrach> but wpa_supplicant is definitely a piece of software that breaks in various ways
<hramrach> try a different version if avalilable in your distro
<vinifm> it is deauthenticating itself
<hramrach> indeed
<hramrach> thre is a possibility that reason 3 has some actual meaning
<hramrach> and different configuration whould not trigger it perhaps
<hramrach> but the damn thing ahs printf so it could tell you what reason 3 is
<vinifm> could be driver?
<hramrach> if you really want to know find a wpa_supplicant mailing list or whatever
<hramrach> or use gdb ;-)
<hramrach> sadly the time when you do not need gdb to connect to a wireless network has not yet come :\
<hramrach> does nobody ever use wireless /o\
<hramrach> vinifm: I have a wifi card that used to work flawlessly sometime mid between squeeze and wheezy
<hramrach> after upgrade to wheezy it stopped working
<hramrach> in Fedora it locks up
<vinifm> :0
<vinifm> i am using debian squeeze
<hramrach> the problem is the working version was somewhere in testing in-between, at least for my card
<mnemoc> hramrach: after this change to sunxi-tools, anything pending regarding nand partitions?
<hramrach> yes, the kernel patchset on ml
Skaag has quit [Ping timeout: 244 seconds]
<mnemoc> drachensun's ?
<hramrach> I just sent one
<mnemoc> 1m
<mnemoc> oh
<hramrach> well, his stuff seems low-level from the topic so hopefully not conflicting
* mnemoc nods
<hramrach> the patchset I sent only removes part of the nand driver that duplicates partition-generic.c in block
<mnemoc> drachensun: btw, I have the source-drop related commits on hold until I finishing putting up the sanitized lichee3 tracking branches. need to have a clear idea of what is merged and what's not.
<mnemoc> drachensun: shouldn't pass from this weekend
<hramrach> btw I suspect the sunxi partitioning code was bogus to begin with because it gives weird minors to the partitions
<hramrach> but also by using partition-generic we lose the sinxi ioctls to make partitions ro/wo/no-access
<hramrach> I am not sure they are available for generic partitions and I have no idea if not implementingthem would break something on android
<hramrach> sure I would not mind having nanda accessible again but the ioctl failing might get noticed if they have error checking
<drachensun> drachensun: Ok
<drachensun> mnemoc: ok
<drachensun> oops
Skaag has joined #linux-sunxi
<Turl> hramrach: basically what we need to make sure of is
<Turl> hramrach: 1) does it still expose nanda, b, etc as before by default?
<Turl> hramrach: 2) can you mount -o ro /dev/nandX /somewhere?
<Turl> and I think that's it
<hramrach> Turl: 1) is done
vinifm has quit [Quit: Saindo]
<hramrach> 2) is done by the generic partition code, just tested
<hramrach> the remaining problem is you can't build nand as module now
<hramrach> which was introduced by the first patch which did generic partitions
<hramrach> it links sumbols from nanda
<hramrach> nand*
<Turl> well, most people make it builtin
<Turl> but that's no excuse for breaking the module support
<hramrach> well, I did not break it
<hramrach> but removed the code that was used before ;-)
<mnemoc> Turl: do you know if there is a kernel common 3.3 beside https://android.googlesource.com/kernel/common.git's android-3.3 ?
<hramrach> Turl: you have some idea how the other partition drivers work?
<hramrach> the nand one relies on nand driver symbols to get nand size
<Turl> mnemoc: korg's -stable series?
<Turl> hramrach: no idea
<mnemoc> Turl: but that isn't android-ized
<Turl> mnemoc: well, there's qcom, ti, nvidia trees too
<Turl> but it wouldn't make any sense to use them :)
<mnemoc> Turl: thing is the drop has files like Documentation/devicetree/bindings/arm/arch_timer.txt which shouldn't exist in a 3.3 tree
ZaEarl has joined #linux-sunxi
<Turl> mnemoc: maybe they cherrypicked stuff for A7
<jelly-home> Turl: free shipping to where? Not .hr apparently
<jochensp> mnemoc: is there a new source drop?
<Turl> jelly-home: I get free shipping using china post on that one
<jelly-home> I don't get an option for CNP at all
<Turl> change shipcountry
<mnemoc> jochensp: will publish the stuff soon. need to sanitize a couple of things more
<jochensp> mnemoc: some new audio drivers for the A10 in there as well?
<mnemoc> haven't had time to review the new a10 code yet, but there is quite a bunch of code to review/merge, so I need to push decent sanitized branches so people can review their favourite soc/subsystem and merge what's valuable
<jochensp> mnemoc: awesome, will have a look when it's published
<Turl> mnemoc: try merging this > git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git timers-v3.4-rc4
<Turl> they probably cherrypicked it
<jelly-home> I've done some testing of SATA on a cubieboard, and it seems it uses quite a bit of cpu. In addition, it had issues talking to a consumer SSD on 3.0Gbps speeds when I tried to dd the whole disk. http://paste.debian.net/239224/ The transfer got stable after reneg. to 1.5Gbps and pulled a decent 110MB/s. sata on A10 does DMA, right?
<mnemoc> Turl: will push it based on android-3.3 + v3.3 for now
<Turl> jelly-home: what A10 hardware revision are you using?
<Turl> it's printed on dmesg early on boot
<jelly-home> this? [ 0.000000] chip-id: A10 (AW1623 revision C)
<Turl> yeah
<Turl> jelly-home: what cpu governor are you using btw?
<jelly-home> interrupt count / MB looks about the same as ahci on an intel H55 chipset, around 8 per 1MB read = 900/s, so I _guess_ it does DMA not PIO
<jelly-home> Turl: ondemand with 400-1000MHz, it goes up fairly early on and stays at 1GHz
<Turl> there should be DMA support on the hardware
<Turl> jelly-home: if you set it to performance and run the same test, can you spot what's eating the cpu time?
<jelly-home> from top: dd_rescue 65% kswapd0 10-18% (even after I disabled all swap)
<Turl> iirc kswapd also handles freeing of pages and compaction
<jelly-home> from vmstat, 0-1% user, 92% system, 7-8% iowait
<jelly-home> so... cache usage and/or memory bandwidth is not that well on (this) A10?
<Turl> anything interesting on /proc/<pid>/stack ?
<jelly-home> of dd_rescue? Lemme see.
<Turl> yes
<jelly-home> huh, no stack in /proc/.../ on this Linux version 3.4.24-a10-aufs+ (max@lynx) (gcc version 4.7.2 (Ubuntu/Linaro 4.7.2-1ubuntu1) ) #1 PREEMPT Sun Jan 27 00:28:21 CET 2013
<jelly-home> (it's not my kernel, but the berryboot-a10's one, seemed based on stage/sunxi-3.4)
<Turl> google says you may need to enable CONFIG_STACKTRACE to use it
<Turl> (which sounds sensible, but I haven't checked)
<jelly-home> could be, # CONFIG_STACKTRACE is not set, and it's present on my pc workstation where there is /proc/*/stack
Skaag has quit [Ping timeout: 245 seconds]
* jelly-home can only go so far without having to build a kernel
<Turl> :)
Skaag has joined #linux-sunxi
<jelly-home> amusing thing, if I let some mplayer play music (off a nfs share unrelated to sata) at the same time, it takes 10% cpu but it's subtracted off _kswapd0_ not the dd userspace
<jelly-home> perhaps I should not trust these numbers
<mnemoc> Turl: there is something else missing between the v3.3 and the timers patch. but we need to compose a sane base.