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
bfree_ has quit [Ping timeout: 252 seconds]
bfree_ has joined #linux-sunxi
bfree_ is now known as bfree
techn_ has quit [Read error: Connection reset by peer]
rz2k has quit []
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
Skaag has joined #linux-sunxi
rz2k has joined #linux-sunxi
user has quit [Ping timeout: 272 seconds]
rellla has joined #linux-sunxi
ZaEarl has quit [Ping timeout: 244 seconds]
rz2k has quit []
<gzamboni> nice work techn
hramrach has quit [Ping timeout: 276 seconds]
hramrach has joined #linux-sunxi
andoma has joined #linux-sunxi
<arete74> Assert::IsFalse(dPerc> dPercMax,csMess);
<arete74> a20 aviable from march
<theOzzieRat> nice :)
rellla2 has joined #linux-sunxi
<mnemoc> hramrach: does the kernel part of the BLKRRPART thing on the stage/ branches work for you?
<hramrach> mnemoc: it fails when the partition is mounted
<mnemoc> ouch
<hramrach> I changed it to the same size only
<mnemoc> and, does a BLKRRPART-patched nand-part work with a not-BLKRRPART-patched kernel?
<hramrach> yes, it just fails
<hramrach> but the BLKRRPART part in kernel does not work
<hramrach> I resized nand2 but sfdisk reports same size
<hramrach> oh, it's just broken wrt the last partition
<hramrach> it occupies the space until the end of flash regardless of the specified size
<hramrach> so I guess you can say it works
<hramrach> mnemoc: I did not test with non-BLKRRPART kernel but
<hramrach> 1) when nand1 is not present (because you have nanda) BLKRRPART is not attempted
<hramrach> 2) when it fails the failure is just reported and does not affect nand-part operation
<mnemoc> :(
<mnemoc> should I just remove it from stage for now?
<hramrach> the BLKRRPART patch?
<hramrach> it appears to work correctly mostly
<hramrach> the only weird part is that if you specify small last partition it is extended to the end of the nand device
jochensp has quit [*.net *.split]
<hramrach> presumably if you specified too large one it would be truncated
<mnemoc> i can't test myself atm, but i'm not really confortable of breaking nand-part for non-BLKRRPART-capable kernels
<mnemoc> s/of/with/
<hramrach> it will not break
<hramrach> it will at most report that BLKRRPART failed
<mnemoc> ah, ok
<mnemoc> i was fearing something for fatal
<mnemoc> more*
<mnemoc> damn. can't type :|
<hramrach> and it will not build on non-linux systems without adding some #ifdefs if that's a concern
jochensp has joined #linux-sunxi
<mnemoc> e.g. android?
<hramrach> e.g. FreeBSD
<hramrach> I have no idea about android
<mnemoc> there is functional freebsd for sunxi already???!
<hramrach> nand-part is partitioning tool
<mnemoc> yes, but runs natively
<hramrach> yuo can technically run it anywhere
ganbold_ has joined #linux-sunxi
<Turl> is the default to behave as previously? (eg nanda?)
<hramrach> it writes a block device you specify
<mnemoc> i mean. it would be a concern if there is any non-linux system running already. for future OSs, it's a future concern :)
<hramrach> it does not even care what the kernel does nanda vs nand1 vs nothing
<mnemoc> Turl: can you give a try to the stage branches?
<mnemoc> i'm not even sure if they boot :|
<mnemoc> but would like to see what can be merged back, and what not
<mnemoc> it's growing too fat already
<Turl> hramrach: I mean, if I boot with that patch on an old system, does it expose nanda or nand1?
<Turl> exposing nand1 is changing behavior and it'd break a lot of stuff
<mnemoc> nanda vs nand1 is not nand-part related, it's kernel related. but yes, I agree it will harm
<hramrach> mnemoc: you could technically run it on amd64 and set up the nand partitioning on a block device (or file if nand-part supports)
<hramrach> usefulness of that is disputable because you typically don't get access to sunxi nand on other platforms
<mnemoc> :)
<hramrach> Turl: the nand-part patch only uses the presence of /dev/nand1 as heuristic to try and do BLKRRPART on the device it is partitioning. It does not ever access anything but the device it partitions - the whole raw nand
<Turl> mnemoc: weren't we discussing the kernel nand patch on staging? o.O
<rm> ^^ L1 cache is shared between all cores??
<hramrach> I am not sure
<mnemoc> Turl: almost. I reacted to a pull request for sunxi-tools :)
<hramrach> wrt that patch on staging it will break android unless the nanda aliases are present
<mnemoc> rm: uh
<hramrach> but they aren't
<mnemoc> :<
<hramrach> but reather than removing the patch it would be useful to add aliases
<mnemoc> +1
<hramrach> actually it will break a lot of Linux setups too
<hramrach> if I was booting off nand I would not be able to boot after patching the kernel with that
<Turl> yeah that was my fear :S
<hramrach> but I don't know any good example of driver with block device aliases where to steal the code from
<Turl> on a related topic, 'super chip' from allwinner on Q3 :p
<mnemoc> it has been rumored they are working on an A15
<mnemoc> personally I would be happy with a humble A40 (quad a7 variant of the A10)
<Turl> what would be an A15? dual core A30 with sgx?
<mnemoc> A31 and A31s are... meh
<mnemoc> I mean cortex-a15
<Turl> ah
<Turl> damn chinese marketing :)
<mnemoc> yeah
<Turl> A10 will still be in market btw
<libv> too bad the a40 will only have a mali-400mp2
<Turl> and "several new chips" on Q2
<Turl> maybe A40
<libv> it will be vastly preferred over the sgx crap though
<mnemoc> +1
<Turl> quad A7 + Mali400MP4 would be awesome
<mnemoc> hopefully not sharing L1 :p
<Turl> can that be shared? o.O
<mnemoc> see rm's comment above
<Turl> where does it say it's shared?
<Turl> another AXP, AXP221s
<rm> the picture implies
<mnemoc> there is only one L1 in the diagram
<rm> but I don't know really
<Turl> it's a picture, it has no technical value
<mnemoc> true
<rm> that's why I asked you :)
<rm> if it's not shared, then it's 64KB per core
<rm> which is kind of low, considering the A10 has 256KB of L1
<rm> (the other possibility, 256 KB of L1 cache PER CORE is way too good to be true)
<Turl> exynos 5 dual has 32KB (Instruction)/32KB (Data) L1 Cache and 1MB L2 Cache
<rm> bingo
<rm> 32K+32K per core
<rm> the picture is still not completely correct
<rm> but at least it gives us those numbers
<rm> and: my mistake @ the A10 has 256KB of L1
<rm> it has 32+32+256(L2)
<Turl> so not that bad then
<Turl> on par with the top market players
vinifm has joined #linux-sunxi
vinifm has quit [Quit: Ex-Chat]
merbanan has joined #linux-sunxi
hipboi has joined #linux-sunxi
ZaEarl has joined #linux-sunxi
<hramrach> where is CONFIG_ANDROID defined?
<mnemoc> if you want an android kernel start from an android defconfig
<mnemoc> crane or nuclear or elite
<hramrach> I want CONFIG_ANDROID to select a symol
<Turl> hramrach: staging/android/Kconfig likely
<mnemoc> / on make menuconfig will tell you the dependency tree
<Turl> mnemoc: on 3.something they added a 'go to' button too :D
<mnemoc> nice
<hramrach> thanks
<hramrach> but it probably makes more sense to add to defconfig
<mnemoc> the idea (at least mine) was to have for each SoC one android, one linux desktop and one not-cedarx/mali (aka server)
<mnemoc> we have the first 2
<hramrach> mnemoc: the driver does have support for the old device names
<hramrach> they are just configured by CONFIG_ANDROID
<hramrach> which is kind of useless
<mnemoc> so we need a new CONFIG_
<mnemoc> and default it to y for a while
<hramrach> yes, just added it
<hramrach> but did not try the kernel with it so far
<mnemoc> what ver are you testing? 3.0 or 3.4?
<hramrach> 3.4
<hramrach> so I enabled the compat code and it looks like the device number for the compat device s is different :s
<Turl> mripard: hi, I just rebased the sunxi-clk branch with a bit better patch split and cleaner support for clk on 8250_dw
<mripard> ok
drachensun has joined #linux-sunxi
<drachensun> wow, thanks libv
<drachensun> I was thinking the channel had gotten awfully quiet
<mnemoc> welcome :)
<Turl> drachensun: need to read /topic more often I guess :p
<drachensun> heh yup
<Turl> mripard: I'll probably email it tomorrow if I don't find any further issues
<libv> it is rather amazing to what extent #arm-netbook had become the linux-sunxi development channel
<Turl> yeah haha
<libv> even now most of the talk in there is sunxi :)
<drachensun> I forgot to set the signed off by on those patches, should I resend them with that?
oliv3r has joined #linux-sunxi
<oliv3r> we have a new channel? :) no longer hijacking arm-netbook?
hramrach has quit [Ping timeout: 276 seconds]
ganbold_ has quit [Remote host closed the connection]
ZaEarl has quit [Ping timeout: 244 seconds]
hipboi has quit [Quit: Leaving]
paulk-desktop has joined #linux-sunxi
rellla2 has quit [Quit: Nettalk6 - www.ntalk.de]
techn_ has joined #linux-sunxi
merbanan has quit [Ping timeout: 276 seconds]
merbanan has joined #linux-sunxi
rz2k has joined #linux-sunxi
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
jelly-home has joined #linux-sunxi
* mnemoc wonders if the eva in charbax's video is *the* eva, or another eva
merbzt has joined #linux-sunxi
<rz2k> whats the status of the gpio driver in our kernel and does it work with a13?
<Turl> mnemoc: I think it's her, yes
<Turl> she says she's part of the marketing department at allwinner
<mnemoc> rz2k: the gpiolib driver still in the stage branches, need feedback on the 3.0 side. eint support for a13 still no included because we haven't added code to distinguish between sun5i variants
<mnemoc> Turl: she actually sounds like she understands the things
<mnemoc> Turl: and pretty good english. no wonder she controls the company
<Turl> mnemoc: well, nobody will buy you chips if you don't know them well :)
<mnemoc> most chinese sellers don't have any idea what they are selling
<mnemoc> and not long ago the public info about the A10 didn't include sata
<Turl> yeah, but sellers don't buy stuff from allwinner
<Turl> sellers just.. sell
<mnemoc> :)
<mnemoc> fair point
<Turl> mnemoc: if you go to a local PC store and ask them, I bet they don't know half of the specs of the PCs on sale
<drachensun> oh man, yeah hearing one of the kids try to convince me why I need the $200 hdmi cable is priceless
<mnemoc> *g*
<drachensun> the sad thing is, its not just commission seeking, they really believe all the bs they spew since they have no conception of what the word digital really means
merbzt has quit [Ping timeout: 245 seconds]
ZaEarl has joined #linux-sunxi
<rz2k> mnemoc: and what about old 'ugly' driver?
<mnemoc> still there
<mnemoc> both use the same script.bin section
<mnemoc> but the 'ugly' driver is very limited
<mnemoc> if you can test stage/sunxi-3.0 I can graduate the stuff :)
merbzt has joined #linux-sunxi
merbzt has quit [Ping timeout: 264 seconds]
<rz2k> mnemoc: the only thing I need is to generate arb. digital signal on one of any pins on a13
<rz2k> cant really decide what to use :/
<theOzzieRat> i have been testing the new gpiolib on A10 cubieboard, including interrupts. It appears to work well for me. I am also using Lorenzo Nava's [PATCH] Sunxi spi board IRQ setting with GPIO management which also works for me.
<mnemoc> rz2k: the ugly driver works fine for outputs
<mnemoc> theOzzieRat: if a not-merged patch works good for you, please reply it telling so
<theOzzieRat> ok, will do
<theOzzieRat> mnemoc, done. I am only new at this stuff so feel free to provide guidance.
<theOzzieRat> I have also found a bug in spi_sunxi.c, but was waiting for that patch to get merged before submitting.
paulk-desktop has quit [Quit: Ex-Chat]
ZaEarl has quit [Quit: Ex-Chat]
rz2k has quit []
<gzamboni> theOzzieRat is it possible to use gpio interrupts in the userspace ?
<gzamboni> i supose maybe using system calls, do you have any code exemple ?
<gzamboni> found a little example using the edge from the sys filesystem gpio control: https://www.ridgerun.com/developer/wiki/index.php/Gpio-int-test.c
<gzamboni> i havent tested yet the new gpio driver
<gzamboni> but the ugly one was working also with inputs here
<theOzzieRat> gzamboni, i don't believe so
<theOzzieRat> im going to use the interrupt to unblock a blocked read
<theOzzieRat> basically I have copied spidev.c, made it use dynamic device numbers and starting work on using interrupts