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
tinti has joined #linux-sunxi
egbert has quit [Disconnected by services]
egbert_ has joined #linux-sunxi
zub has quit [Ping timeout: 240 seconds]
zub has joined #linux-sunxi
rav0_ is now known as rav0
eebrah has joined #linux-sunxi
rellla has joined #linux-sunxi
Undertasker has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
shineworld has joined #linux-sunxi
hansg has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
n01|work|afk is now known as n01
eebrah has quit [Ping timeout: 264 seconds]
eebrah has joined #linux-sunxi
Undertasker has quit [Ping timeout: 245 seconds]
slapin has quit [Quit: Lost terminal]
Undertasker has joined #linux-sunxi
focus has quit [Remote host closed the connection]
focus has joined #linux-sunxi
ZaEarl has quit [Ping timeout: 245 seconds]
Undertasker has quit [Ping timeout: 248 seconds]
wingrime has joined #linux-sunxi
<wingrime> mnemoc: you should disable registration on wiki
<wingrime> quickly as possible
Undertasker has joined #linux-sunxi
<mnemoc> wingrime: ouch
<shineworld> after we can again contribute ?
<mnemoc> Turl had setup captcha and questions to avoid that...
<hramrach> they cracked the captcha I guess ;-)
<wingrime> mnemoc: add header like "go to #linux-sunxi for manual reg or simular"
<hramrach> which sucks but is better than dealing with bazillion spammer accounts
<wingrime> and you shuld remove users without contrib
<wingrime> one good sql req to DB
<mnemoc> registration disabled
<wingrime> mnemc: I saw e-book readers with sun3i in local store
<wingrime> E200 and F15
<mnemoc> nice
<mnemoc> got one? :)
<wingrime> price
<mnemoc> :(
<wingrime> sun5i have same price
<wingrime> but it totaly crap HW
<wingrime> TFT ebook readers
<wingrime> If I saw e-paper , I maybe think about
<wingrime> mnemoc: You can consider that sun3i are still in market
<mnemoc> i know. that's why I'm sad no one cares for it
<wingrime> mnemoc: but a10 less-than before - now 2-cores form AM logic and RK more common
<rellla> <wingrime>and you shuld remove users without contrib <- +1
<mnemoc> as it's for hacking it doesn't really matter if it's a usable e-reader or a crappy tft... it will be taken apart anyway
<mnemoc> rellla: if you point me to an extension I can install to do that I will
<wingrime> mnemoc: pricing are realy insane
<mnemoc> that part is indeed sad
<mnemoc> short before tom quit allwinner he had promised some sun3i devkits
<mnemoc> from the piles in their office
<wingrime> that will be great If I get one
<mnemoc> s/piles/stacks/
<mnemoc> unfortunatelly he left AW...
<wingrime> mnemoc: I can bye some cheap sunxi form refubrished store
<mnemoc> .oO(good... the word "pile" in english exists and means what I expected)o
<mnemoc> sun3i?
<wingrime> mnemoc: but It little far form me
<hramrach> what is allwinner e200 cpu?
<wingrime> mnemoc: may tablets "repaired"
<wingrime> *many
<mnemoc> hramrach: I think all allwinner arm9 are sun3i
<wingrime> uCOS - I saw commented part in drivers
<mnemoc> their melis thing is a fork of a commercial RTOS iirc
<mnemoc> need to go. back in 30m
<hramrach> it does not say arm9
<hramrach> it says e200
<mnemoc> f1e200
<mnemoc> first column
<wingrime> hramrach: thank good that is not MIPS
<hramrach> huh, melis OS
<hramrach> with 64mb ram not much use but sould suffice for reader I guess
<wingrime> I don't relay think I need this even for hacking
<hramrach> if it could boot from SD you could put real system on it
<hramrach> and it's an off-the-mill reader to start with
<hramrach> but maybe hard to find by now
<hramrach> hmmm, the e-ink one is actually pretty new
<wingrime> I saw "tft" I don't remember e-link
<wingrime> maybe
<hramrach> they have multiple models
<hramrach> most are tft
<hramrach> the e-ink one has no wifi
<hramrach> probably to save batteries
<hramrach> no datasheet for f1e200
<hramrach> not even the datasheet for managers :s
Undertasker has quit [Ping timeout: 260 seconds]
Undertasker has joined #linux-sunxi
<rellla> don't know if this one can handle batch deleting. easiest way would be to directly edit db-users-table imo and delete all users without edits
bhoj2 is now known as bhoj
tinti has quit [Quit: Leaving]
<Turl> mnemoc: ouch, more spammers
<Turl> we need to rotate the questions
vicenteH has joined #linux-sunxi
calris has quit [Read error: Connection reset by peer]
zub has quit [Ping timeout: 246 seconds]
<Turl> mnemoc: done, I've rotated the questions now
<oliv3r> stupid spammers should get a live; it's highly frustrating
<oliv3r> bah, now the recent page is messed up again; can't see what's new :)
zub has joined #linux-sunxi
hipboi|cubie has joined #linux-sunxi
<Turl> yeah, I started banhammering the obvious spammy accounts now
<oliv3r> make registration require confirmation; they can ask for it on irc? though it'll be slow
<Turl> feel free to kick the rest out
<oliv3r> do we have recaptcha by google? the read books thing?
<Turl> oliv3r: that'd let many people out (ask on irc)
<Turl> we have manual question captcha, try to register an account and see :P
<oliv3r> Turl: i have an account! :p
<rellla> Turl: wouldn't it be better to delete all users without edits instead of banning them one by one?
<oliv3r> well recaptcha works reasonable that +sunxi related questions
<oliv3r> rellla: i agree; it's a lot of 'crap' filling the db
<oliv3r> can you sort users based on edits?
<oliv3r> though a simple SQL query should sort that
<Turl> rellla: can you do that? I'm a mediawiki noob
<oliv3r> so how can I see any recent wiki changes :(
<rellla> Turl: me too ;)
<oliv3r> oh, increase the length :)
<oliv3r> i know it's user 'areta74' or something
<rellla> wanted to do a "SELECT * FROM user" on http://linux-sunxi.org/Special:Maintenance/sql - no output. hm.
<oliv3r> permissions?
<rellla> most likely not enough ;)
hipboi|cubie has quit [Ping timeout: 256 seconds]
<mnemoc> rellla: I see three groups (bot, administrator, bureaucrat), you belong to the last 2
<Turl> I used show tables and it didn't work either
<rellla> i think this sql-script doesn't have any function at all as it is blanked in mediawiki-wiki
<Turl> iirc it's running with postgres though, might not be the right command
<oliv3r> Turl: i use postgres for years; what's the SQL query you are trying to send?
<RaYmAn> Turl: \d
<Turl> oliv3r: 'show tables' :P
<RaYmAn> but it might need direct access to 'psql' command?
<rellla> maybe dangerous... for temporary use only
<mnemoc> rellla: Asksql installed
<mnemoc> hope no bureaucrat decides to abuse...
<hramrach> you can remove once this is finished
<oliv3r> Turl: \d :p
<oliv3r> but tehre's also a nother way ;)
<Turl> :)
<rellla> something goes wrong, as i only get returned a blank page...
<mnemoc> rellla: slap Turl harder :)
<Turl> mnemoc: why? :(
<mnemoc> you are root, so you are to blame :p
<Turl> sudo psql says there's no root role
<Turl> I gave up trying to run sql a big ago :P
<Turl> bit*
<oliv3r> Turl: psql -U postfix <databasename>
<oliv3r> that should work as root, opening any database
<oliv3r> Turl: otherwise, psql -U <webuser that can write to the database> <databasename>
<oliv3r> psql -U <$user> is what happens if you ommit -U :)
<oliv3r> and $user would be root :p
<oliv3r> and i doubt there's a 'root' user in the database :)
<mnemoc> Turl: `sudo su - sunxi-mw` should turn you into the same user that runs the php of the wiki
<mnemoc> rellla: what is the super query I need to run?
<mnemoc> to kill all zombies
<Turl> oliv3r: mysql always has a root user :p
<rellla> SELECT * FROM user WHERE useredit_count=NULL
<rellla> ?
<rellla> just to proof to get the right ones;
<Turl> but make a full db backup before running any destructive operation :)
Undertasker has quit [Ping timeout: 264 seconds]
<rellla> yeah, backup!
<oliv3r> backup twice
<oliv3r> just in case your backup needs a backup!
<oliv3r> Turl: the postgresql super user is called 'postgres'
<oliv3r> btw, postfix is obivously wrong :p i should have written postgres :)
<Turl> :) I was wondering what did postfix had to do with all this
<oliv3r> yeah; i have postfix and postgres on the same box, and often make this mistake ;)
<oliv3r> muscle memory I guess :)
vicenteH has quit [Read error: Operation timed out]
Undertasker has joined #linux-sunxi
<mnemoc> http://sprunge.us/Fcie <--- wiki users with no edits
<rellla> i think completely deleting users from database is a little bit tricky, as there are references to user.user_id in so many other tables... should this rows all be deleted?
<hramrach> preferably the rows that show up in reent changes should be ;-)
<mnemoc> http://sprunge.us/PjiA <--- these don't have the email address confirmed
<hramrach> does the editcount go to 0 when the page user has edited is deleted?
<hramrach> seems like some old users are in the list, too
<mnemoc> http://sprunge.us/cBMN <--- 0 edits, no email authenticated, ordered by registration time
<rellla> i'd prefer to take the last ones.
* mnemoc wonders if mw people properly uses delete cascade and friends...
<hramrach> It does not seem the earlier ones are really much different
<mnemoc> btw, the table is called mediawiki.mwuser, not 'user'
<mnemoc> that's probably why the queries on the web interface didn't work
<hramrach> tons of users registered in 08 and 09 as well
<hramrach> I don't know why registering users when not using them but I guess that's just a bot fail
<Turl> hramrach: maybe they're using them, but we require a captcha when posting pages too iirc
<hramrach> yes, that's why it might be a botfail
<hramrach> crack captcha for registration but not for posting
<oliv3r> do we allow users to do anything if their account isn't e-mail authenticated? I'm supprised it doesn't get automatically cleaned after a certain amount of time
<oliv3r> actually, i'm supprised there is no 'auto cleanup' function in mediawiki's user management system
<oliv3r> e.g. set a flag on the user account; user gets deleted
<mnemoc> Turl: those of the last paste haven't verified their mail address
<mnemoc> so they can't edit anyway
<mnemoc> oliv3r: mw's user management is crap
<mnemoc> personally I only allow user creation from a url (with unique and temporal token) sent by mail
<mnemoc> so no ghost accounts are possible
<mnemoc> but mw people seems to be very fond to anon edits
<oliv3r> well, initially; wikipedia was a 'edit ya'll' kinda thing
<oliv3r> but (l)users quickly abused that
ZaEarl has joined #linux-sunxi
hipboi has joined #linux-sunxi
<mnemoc> hipboi: hi, is it possible to get some A20 based cubieboard prototypes to start working on sun7i support?
<hipboi> mnemoc, i will try to get some
<hipboi> it's possible
<hipboi> they have released the first sdk for a20
rz2k has joined #linux-sunxi
<mnemoc> we don't really need the sdk
<mnemoc> only a device where to test
<hipboi> i mean they release the sdk to customer which means the software is stable now
<hipboi> at least no bug of the chip
<mnemoc> it would be awesome if we could get the newer commits. our -dev branches end around march 2nd
<Turl> hipboi: nice :)
<oliv3r> software is stable, or hardware is stable?
<oliv3r> eitherway, awesome that A20 is becoming reality :)
* mnemoc wants an A40 :p
<oliv3r> I was typing that, but didn't wanna sound overly excited :p
<oliv3r> little.BIG and mali
<oliv3r> and with the rest of the core being identical to A10, so our work isn't in vain :)
<arete74> qrt579
bhoj has left #linux-sunxi ["Leaving"]
hansg has quit [Quit: Leaving]
<mnemoc> hipboi: anything there we don't have in our wiki?
<slapin_nb> it'd be good to have design guide for A13 with sample DDR3 layout...
* slapin_nb dreams too much...
<hipboi> mnemoc, it's allwinner official support wiki
<mnemoc> yes, i've seen it before.... but couldn't find any new content
<hipboi> a31/a20 stuff is new i think
<mnemoc> gzamboni: can you help importing content?
<gzamboni> sure
<gzamboni> im importing the csi table i mentionned
tinti has joined #linux-sunxi
<mnemoc> please try to keep things categorized and interlinked
<gzamboni> i created a new item in "Generic Hardware Hacking of Allwinner SoCs:"
<gzamboni> as the CSI is present also in A13, A10s and A20
<mnemoc> [[CSI]] is ok. just don't forget to categorize and redirect [[A10/CSI]] and friends, and leave sections to describe the controller or the CSI-related diffs across SoCs
<gzamboni> ok
art103 has joined #linux-sunxi
hipboi has quit [Quit: Leaving]
Undertasker has quit [Ping timeout: 256 seconds]
Slash has joined #linux-sunxi
Slash has quit [Client Quit]
slash_m_a_x has joined #linux-sunxi
Undertasker has joined #linux-sunxi
slash_m_a_x has quit [Read error: Connection reset by peer]
<jinzo> what's the call on A10s? flop?
<mnemoc> sad they removed sata from the A10s
vicenteH has joined #linux-sunxi
<jinzo> and I wonder how price-competative it can actually be to A10 for example
<jinzo> now that they're churning out A31 and A20
<jinzo> the latter albeit slowly
<mnemoc> the key of the A10s is size
<mnemoc> very important on hdmi dongles
<jinzo> but most of the hdmi dongles manufacturers (as most china manufacturers really) only see the "cores" - and I don't know how can A10s compete with cheap RK dual cores.
<jinzo> but, I guess the time will tell.
<jinzo> The other day I had a dream about a compleatly open SoC with support in mainline kernel and open source GPU drivers
rellla has quit [Ping timeout: 255 seconds]
n01 is now known as n01|away
<hramrach> thechnically rk3066 is in quite good standing - dual core cpu, 4 core gpu, some drivers exist, support under development, 2GB ram device is sold
<hramrach> but the rk developers use forum as the primary communication medium so getting any info ..
<hramrach> meh
<jinzo> there's some driver/open source kernel? Didn't know that - the last I checked I couldn't find any (but that was single core days)
<hramrach> yes, there is some
<hramrach> can't really tell how good
<hramrach> how the hell do I tell what SoC models the platform driver supports?
<mnemoc> it's done the other way around
<mnemoc> the mach- or the .dts registers platform devices
<hramrach> you boot it on the SoC and iether it works or fails
<mnemoc> and so the corresponding platform driver gets asked to probe it
<hramrach> I have no .dts
<mnemoc> "modern" way is to include that info in the .dts file. old way is hardcoded in the mach- code
<hramrach> it's 3.4 kernel
<mnemoc> allwinner's way is, script.bin
<hramrach> so the script.bin tells what platform you have
torqu3e has quit [Quit: torqu3e]
<mnemoc> but allwinner's way works backward. instead of making the platform load the driver, the driver checks (on load) if the script.bin has it enabled
<mnemoc> the plat/script.h API was written to allow us to make a generic platform-driver registrator reading the script.bin just once
<mnemoc> using script.bin you can "distinguish" between a sun5i with hdmi and one without
<mnemoc> but it's still limited to a single mach
<hramrach> either way you can't tell what the driver supports by looking at it
<hramrach> it's defined in script.bin
<wingrime> hramrach: we can detect soc
<hramrach> and the script.bin exists on the devices if at all
<wingrime> more generaly , you can simply check HW registers
<hramrach> I was thinking about getting the prestigio reader but since it does not come with linux it has no script.bin
<hramrach> so I guess it's a lost case
<hramrach> will work as reader presumably but no Linux support
<wingrime> hramrach: possible have
<mnemoc> "prestigio reader" is?
<mnemoc> it has microsd, and usb (assuming OTG)
<hramrach> I would really like to get my hands on a device with e-ink display to try a stab at e-ink usable UI and page flipping
<mnemoc> so using fel and a uart/jtag breakout you should get a pretty decent headstart
<hramrach> assuming it's supported platform at all
<hramrach> and I can manage to swithc on fel and/or jtag
<mnemoc> we have sun3i code from the initial 2.6.36 release
<mnemoc> it lacks of pm, but should be similar to the a10's
<mnemoc> and using the jtag breakout you can snif the rest
<hramrach> but sunxi does not ship this SoC with linux at all, officially
<mnemoc> it seems they thought about doing so when they started the A10
<mnemoc> and then found it... not interesting
<wingrime> hramrach: we have more important problem with a10s
<hramrach> there is Olinuxino for that
<mnemoc> the sun3i code initially imported into 2.6.36 (android 2.3) is very incomplete
<hramrach> very nice board
<mnemoc> hramrach: are they selling it yet?
<hramrach> yes, I see it in stores
<wingrime> I don't think we need support sun3i at all, It have many variants
<mnemoc> nice
<hramrach> the olinuxino or the reader?
<hramrach> afaik both are readily available
<mnemoc> having one variant working is enough to attract a lot of russians who have already worked with F13 readers
<mnemoc> these F1e200 are pretty new
<hramrach> well, the thing is I am not the person to write support for a platform, even with partial sources available
<wingrime> AW not see linux on e200
<hramrach> it takes too much time
<hramrach> I don't feel that much bored
<wingrime> hramrach: e200 have not wifi simply becose some one must write wifi stack for eCOS
<hramrach> heh
<wingrime> hramrach: wifi stack is not easy at all
<hramrach> I know. It's half-broken on every system.
<hramrach> maybe just quarter-broken on some
<wingrime> and more generaly what do you can do with wifi on reader without browser
<hramrach> somebody would have to write a browser or e-store app for eCOS
<hramrach> and it drains batteries
<wingrime> all about economical effectivity
<mnemoc> or fix basic sun3i/f1e200 linux support ;-)
<wingrime> mnemoc: eCOS are RTOS
<hramrach> eCOS is much lighter so probably better for simple reader
<hramrach> rt is maybe good idea for something with media player
slash_m_a_x has joined #linux-sunxi
<hramrach> lokk how difficult it is to get audio on sunxi linux
<wingrime> hramrach: thay simply use old code base
<mnemoc> sun3i for openinkpot.org :)
<wingrime> mnemoc: we have freeRTOS
<mnemoc> true
<hramrach> heh, they did not want to pay to those mpeg-la
<wingrime> RTOS are much faster but can't do much
<hramrach> no mp3 support on the reader but you can play wav, flac, ogg, aac
<hramrach> they can do anything but many applications aren't designed for realtime
<wingrime> hramrach: no mp3 in public specs
<hramrach> plus with 64M ram you can't really expect miracles
<wingrime> hramrach: I have chep ADSL modem with linux have 32 MB RAM
<wingrime> and 4 ROM SPI flash
<hramrach> mnemoc: they have other e-ink reader with Linux based on some low-end rockchip chip
<wingrime> very old linux kernel
<mnemoc> cedarx isn't really an advantage when aiming at a router...
<mnemoc> or g2d
<wingrime> mnemoc: not all sun4i have cedarx
<wingrime> *sun3i
<hramrach> this one has limited cedarx/different media accel - no encoding. and no graphics it seems
<wingrime> you realy need cedar on elink?
<wingrime> *e-ink
<hramrach> no.
<hramrach> but the chip has it
<hramrach> the e200
<hramrach> there is simply no way to play video on e-ink
n01 has joined #linux-sunxi
<wingrime> why AW don't make cortex a8 with SGX?
slash_m_a_x has quit [Ping timeout: 256 seconds]
<wingrime> How much HW revisons a10 have?
<hramrach> why sgx?
<wingrime> opengl
<hramrach> blob
<wingrime> sgx better for hdmi pc
<wingrime> hramrach: do you know non-blob video for mobile?
<hramrach> actually if lima is completed mali is going to be better, most likely
<oliv3r> wingrime: i think 3 that we know of (a, b and c) and of course the internal FPGA version
torqu3e has joined #linux-sunxi
<wingrime> hramrach: I still not see HW accel linux desktops
<oliv3r> oh arete74 is here! i like your howto; deffinatly going to try it
<hramrach> we have g3d for that
<hramrach> g2d
<wingrime> aw drop g2d in a13 a10s
<hramrach> sucks
<wingrime> and not sure maybe in some others
<hramrach> are you sure they are dropped?
<oliv3r> i MAY pickup a a10s dongle (don't need sata port on a hdmi dongle anyway) to experiment if its cheap enough; but may wait (enough to play with) for a possible A20* dongle
<hramrach> mali is well capable of some blending
<wingrime> android don't need g2d
<hramrach> so is the main CPU when it comes to that
<hramrach> but g2d is sillicone specifically designed for that
<hramrach> and I have presumably 3d accelerated gnome desktop
<hramrach> but since doublebuffering does not work dues to some disagreement between mali blob and X server it's not as accelerated as it could be
<hramrach> you can have software blended xfce/lxde or whatever using XRender
<hramrach> it's faster for small effects but sucks for window resizing
<hramrach> android does not allow window resizing so does not have this problem ;-)
<oliv3r> also isn't adreno in mainline now? 'freedreno'
<wingrime> adreno must be simular with radeon
<hramrach> why?
<oliv3r> i think they swapped the g2d unit for the IET engine (flikker filter)
<oliv3r> wingrime: adreno comes FROM radeon, taht's for sure, hence the name :)
<hramrach> and it's spelled similar fwiw
<wingrime> hramrach: I still have questions form your dma test
<wingrime> 1) git log
<wingrime> 2) /dev/nand ??
<jukivil1> mnemoc: I think my email to sunxi-list got stuck to moderation since I've changed email address
<oliv3r> when they start dumping those A10s sticks i'll get one, 35 USD from DX is cheap; that's about 25Euro's
<oliv3r> drop 10 USD and i'll get it
<mnemoc> jukivil1: whitelisted
<jukivil1> thanks
jukivil1 is now known as jukivili
<hramrach> the fact ardeno is some old ATI IP does not mean it's similar to any of the AMD graphics
<hramrach> it may be but you would really have to look at the actual arch specs
<hramrach> even the desktop cards went through several generations which are quite different at times
paulk-desktop has joined #linux-sunxi
<mnemoc> gosh... so many pending commits :(
<wingrime> mnemoc: add header about registration tom linux-sunxi mainpage
<mnemoc> i reenabled registration.... now that the questions were changed and the app updated
<hramrach> wingrime: /dev/nand is the whole nand
<hramrach> I could use nandb but nanda is too small
<mnemoc> the whole *logical* nand
<wingrime> hramrach: you results are opposite....
<hramrach> I have no result whatsoever
<hramrach> same speed with or without patch
<wingrime> I need some one to test "responce" with android
<wingrime> with next patches
<oliv3r> hramrach: check out the freedreno source. It was so easy to write, because it was SO similar to the radeon. Ask libv or check phoronix
<oliv3r> mnemoc: many commits is good, it means we are working hard :)
<oliv3r> what mnemoc means, boot0/1 are stored before the logical nand (inaccessable to the normal OS)
<hramrach> wingrime: both devices I use for testing are quite dodgy, unfortunately
<wingrime> I still don't know are dma patches workable.....
<hramrach> but all are
<hramrach> wemac has some issues
slash_m_a_x has joined #linux-sunxi
<hramrach> the nand performance varies wildly with uptime
<hramrach> nobody else tested usb
<wingrime> hramrach: dma code are realy strange
<wingrime> hramrach: It need rewite
<hramrach> but the device I use is an old 5.4k rpm disk so it might be too slow
<wingrime> hramrach: and more one It call callback in irq context
<wingrime> hramrach: mmc code too use irq context too much
<wingrime> hramrach: wemac: send next packet It dma callback (irq contextx)
<wingrime> *in
<wingrime> That realy crap
<hramrach> you use workques for that in sane drivers, presumably
art103 has quit [Quit: Page closed]
<hramrach> but working in irq context did not break anything immediately so AW did it that way :s
<wingrime> problem is..... you lower performance when do in workqeues but
<hramrach> seriously
<hramrach> the nand performance is so lame that you can't possibly lower it much by in-kernel handling
<hramrach> and workques should really help a single core system overall
<wingrime> better io-wait make UI more responce and user feel quick interface
<hramrach> keep the cpu free for iqr handling
<hramrach> so that you can read flash *and* transfer network data
<wingrime> small irq are requrement for RTOS
<hramrach> and working sound drivers ..
<wingrime> and it means UI smoothness
<oliv3r> the nand is EXTREMLY slow; 5-15 MB/s max
<hramrach> I measured 3.1 MB/s consistently on a long running system
<hramrach> rebooted and it went to 13-14MB/s
<oliv3r> yeah that's really crap
<oliv3r> i've never seen 13 MB/s :p
<wingrime> hramrach: nand driver have own cache
<oliv3r> wingrime: you mean the sram .. (c3 top of my head?)
<wingrime> see driver
<hramrach> I did 1G reads to overflow it
<wingrime> no, I mean sunxi_nand driver have internal cache
<hramrach> it can't possibly cache that
<wingrime> it can possibe explain why you have 13 Mb/s at boot
<hramrach> possibly
<hramrach> but how does it cache 1G data?
<wingrime> you don't need cache some thing you read to null
<hramrach> meybe it uses some strange eviction algorithm that keep the earlier data
<wingrime> anyone used lxde ?
<oliv3r> briefly
<hramrach> same
<oliv3r> i htink linaro uses lxde
<wingrime> xfce
<Dreadlish> linaro uses lxde
<Dreadlish> alip image
<oliv3r> hmm, fedora 18 uses xfce; linaro alip lxde
<wingrime> mnemoc: do you have plans about next versions beyond 3.4?
<hramrach> mainline
<oliv3r> i would hope so
<wingrime> mainline is a long term
<oliv3r> though an intermdiate version may be required. 3.9 + our patches (dps, usb)
<wingrime> but be can go smooth if we begin use new kernel with deivers mained
<wingrime> *mainlined
<hramrach> last I heard it's 3.4 and mainline
<wingrime> such speed we see mainline at 4.0
<oliv3r> that's why I say, we may need an intermdiate version
<wingrime> I think we can use new kernel and "mainlined driver"
<wingrime> becose we can't test "good" driver replacments
<hramrach> there are no 'good' driver replacements so far
<wingrime> exists
<wingrime> uart
<hramrach> but with the nand mtd driver and wemac you can get bootable and somewhat usable system
<hramrach> uart is good too
<wingrime> good wemac need dma
<wingrime> but without dma it still possible
<hramrach> yes, dma is kind of missing
<n01> mripard_: which is the difference between dt-for-3.10 and core-for-3.10 branches?
<hramrach> also the bridge between mtd and vfat
<wingrime> hramrach: I realy don't thnk is dma difficult
<wingrime> hramrach: i prefer ext2 for boot
<wingrime> and uboot can it
<hramrach> is it even possible to use the current nand content with mtd?
<mripard_> n01: the dt branch is only for changes on the *.dts
<hramrach> it's not like ext2 is different from vfat when it comes to mtd compatibility
<mripard_> core is for the core changes, on the code itself
<n01> ok so for the wdt driver I use the core-for-3.10?
<mripard_> both actually
<wingrime> hramrach: have ext4 mtd compat? last I failed using etx4 on mtd with 2.6.35
<n01> ok
<hramrach> shouldn't that be 'export this mtd as plain block device'?
<wingrime> mripard: can we replace our uart with new driver?
<wingrime> as kconfig choose
<hramrach> so that the brdige code does wear levelling and bad block amnagement and the filisystem is oblivious as ever
<wingrime> ext4 have badblock
<mripard_> wingrime: there's no "new" driver, we're using the designware uart driver
<mripard_> and I guess it's already there in 3.4
<wingrime> mripard: thay used synopsis IP for it?
<mripard_> that's what they told me yes
<hramrach> nothing designware enabled
<wingrime> mripard: mtd can be mainlined?
<mripard_> which one? the one posted on the mailing list?
<hramrach> it's in early stage of development
<wingrime> mripard: some benchmarks?
<hramrach> but technically it's an acceptable driver, in general
<n01> mripard_: what about drivers-for-3.10? :)
<mripard_> n01: hmmm
<mripard_> n01: ignore this one :)
<n01> :D ok
<hramrach> I have only CONFIG_SERIAL_8250
<wingrime> hramrach: dma also can be used with uart
<hramrach> it can be used with anything
<hramrach> but does the 8250 driver support that?
<hramrach> does not seem to have an option for dma
<wingrime> hramrach: dma support DRQ form UART
<hramrach> oh, CONFIG_SERIAL_8250_SUNXI
<mripard_> wingrime: it would need a bit more effort for the mtd, and rework it a bit to use the standard kernel API
<mripard_> but it seems reasonnable
<mripard_> the only thing we miss to make it work is still the DMA
<wingrime> mripard: dma actualy simpler than mtd
<wingrime> mripard: It can be added easy
<mripard_> I'm not saying it's hard
<mripard_> I'm saying it's not there yet
<wingrime> 1) config channel 2) ask for send 3) wait end/or send
<wingrime> but problem is,,,,kernel have not realy dma framework
<wingrime> Or I have not found this one in source
<wingrime> every soc use own-way dma
<mripard_> there's dmaengine for that
<mripard_> which is precisely that :)
<wingrime> give me example
<wingrime> I have not found anything
<hramrach> Documentation/crypto/async-tx-api.txt:4
<wingrime> crypto
<hramrach> include/linux/dmaengine.h
<hramrach> include/sound/dmaengine_pcm.h
<hramrach> serial driver with dma drivers/tty/serial/amba-pl011.c
<wingrime> show me some implemen on arm
<hramrach> amba is arm
<hramrach> include/linux/amba/pl08x.h
<hramrach> drivers/dma/amba-pl08x.c
rav0_ has joined #linux-sunxi
<wingrime> thanks
<hramrach> that's as far as git grep gets me ;-)
<oliv3r> include/linux/dmaengine.h
<oliv3r> sounds exactly what is needed
rav0 has quit [Ping timeout: 256 seconds]
wingrime has quit [Ping timeout: 245 seconds]
slash_m_a_x has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]
<vicenteH> hello
<vicenteH> I have an A13 based F900 mainboard
<vicenteH> I have been searching for a serial port in it, but have not found it.
<vicenteH> Any clues about it, to get booting console?
<slapin_nb> look at A13's datasheet, there is no much choice
<slapin_nb> then use your multimeter with appropriate pins
<slapin_nb> and testpoints
<vicenteH> thanks slapin_nb
rav0__ has joined #linux-sunxi
rav0_ has quit [Ping timeout: 260 seconds]
<hramrach> oh yeah
<hramrach> a13 has pins
<shineworld> for all android users which would to put android building out to SD instead of run pack-cmd and put all in nand, I've made a simple (raw) script that automatize all.
Undertasker has quit [Quit: Bye]
rellla has joined #linux-sunxi
<vicenteH> have any of you extracted kernel image from boot.img partition and converted it into an uboot kernel image?
<vicenteH> and booted successfully, of course.
<vicenteH> with the SD card.
<shineworld> yes
<shineworld> I've do
<shineworld> just some mins ago
<vicenteH> I tried but it did not boot
<shineworld> there is a wrong document in linux-sunxi about Booting android in SD
<shineworld> try to use:
<vicenteH> is a line like that enough:
<vicenteH> mkimage -A arm -O linux -T kernel -C none -a 0x40008000 -e 0x40008000 -d nandc_boot-zImage uImage
<vicenteH> ?
<shineworld> mkimage -A ARM -C none -T kernel -O linux -a 40008000 -e 40008000 -d $BUILD_PATH/obj/KERNEL_OBJ/arch/arm/boot/zImage mnt/bootloader/uImage
<shineworld> yes
<shineworld> change onlgy my path because went by a script
<shineworld> so
<shineworld> mkimage -A ARM -C none -T kernel -O linux -a 40008000 -e 40008000 -d zImage uImage
<vicenteH> thanks shineworld
<shineworld> uhm similar mine works
ZaEarl has quit [Ping timeout: 245 seconds]
<shineworld> I start by another point, not from livesuit image but directly by output of compilation
<vicenteH> I made a backup from my android nand partitions,
<vicenteH> then installed a bad CWM recovery.
<shineworld> I don't know yet how to install CMW recovery
<vicenteH> and it messed up everything, not being able to access anymore to nandf.
<vicenteH> so now I need a kernel to be able to detect nand right.
<shineworld> ah ok
<shineworld> however boot from SD = 43 S, boot from SD (class 4) = 37S so...
<shineworld> 37 S
<shineworld> nand of cubie isn't a fast device
<vicenteH> I tried too to create a copy of the old android in SD card, but it failed in initialization.
<vicenteH> May be the kernel from a Debian image I used is not suitable.
<shineworld> what is your hardware
<vicenteH> allwinner A13 f900 mainboard
<vicenteH> a tablet device.
<shineworld> ah ok ... mine is a dev board without any limit about hack :)
<vicenteH> great for you :)
<shineworld> I don't know if tablets powered by allwinner are same
<vicenteH> well, it uses same processor, etc.
<shineworld> I'm using yet 3.0.56+ because I've got a very slow performances on LAN using last from git
<shineworld> I'm just understanding the things and what goes fine and what don't goes fine
<vicenteH> sadly the kernels I have tried do not detect right the nand in my device.
ZaEarl has joined #linux-sunxi
<shineworld> root@android:/proc # cat version
<shineworld> Linux version 3.0.52 (shine@shine) (gcc version 4.4.3 (GCC) ) #4 PREEMPT Sun Apr 7 17:24:30 CEST 2013
<shineworld> yes I'm using that...
<shineworld> tomorrow I will try again the last
<vicenteH> I gonna convert an old image now into uImage and try again.
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
<shineworld> uImage is for u-boot however (I guess) :)
paulk-desktop has quit [Quit: Ex-Chat]
<vicenteH> shineworld: in the end I unbricked my device
<vicenteH> thanks for your help :)
<shineworld> you are welcome
theborger has joined #linux-sunxi
jukivil1 has joined #linux-sunxi
<n01> uhm mripard_ merging both the branches I'm not able anymore to reach the login. I'm stuck at "turn off boot console earlycon0" ... is this a known issue or it is just me?
hno` has joined #linux-sunxi
xnxs has joined #linux-sunxi
y0gurt has quit [*.net *.split]
hno has quit [*.net *.split]
WarheadsSE has quit [*.net *.split]
xenoxaos has quit [*.net *.split]
jukivili has quit [*.net *.split]
xnxs is now known as xenoxaos
WarheadsSE has joined #linux-sunxi
shineworld has left #linux-sunxi ["Leaving"]
rz2k has quit []
tinti has quit [Read error: Connection reset by peer]
n01 has quit [Ping timeout: 248 seconds]
tinti has joined #linux-sunxi
ZaEarl has quit [Read error: Connection reset by peer]
ZaEarl has joined #linux-sunxi
torqu3e has quit [Quit: torqu3e]