hno changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See | | Logs at
leowt has quit [Quit: leowt]
BJfreeman has quit [Quit: had a good time]
ganbold_ has quit [Ping timeout: 268 seconds]
<naobsd> hno: I modified Cubieboard2_FEL line which has only "sun7i:CUBIEBOARD2,SPL_FEL" in boards.cfg, and add some undefs in include/configs/sunxi-common.h
<naobsd> hno: but spl/u-boot-spl.bin is still 11KB
<naobsd> "make cubieboard_fel_config" (for A10 board) also makes 11KB spl.bin
<naobsd> all *.bin in sunxi-tools/bin/ are less than 8K
<naobsd> ssvb: all 7zip benchmarks are less than 500mA, very interesting :) no peripherals except uSD card, right?
<naobsd> hmm, probably spl for fel is not ready yet in this u-boot...
<naobsd> "0x2000-0x5cff Free for program use" hmm?
<naobsd> I couldn't write 0x4000- on cb2
<naobsd> anyway I think all I need is dram/clock/power to load real u-boot
<naobsd> it should be smaller than 8k...
ganbold_ has joined #linux-sunxi
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest58692
_BJFreeman is now known as BJfreeman
Guest58692 has quit [Ping timeout: 252 seconds]
ganbold_ has quit [Ping timeout: 246 seconds]
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest50614
_BJFreeman is now known as BJfreeman
Guest50614 has quit [Ping timeout: 252 seconds]
\\Mr_C\\ has quit []
\\Mr_C\\ has joined #linux-sunxi
rz2k has joined #linux-sunxi
Tsvetan has quit [Ping timeout: 268 seconds]
BJfreeman has quit [Quit: had a good time]
ganbold_ has joined #linux-sunxi
Quarx has joined #linux-sunxi
foubarre has quit [Remote host closed the connection]
ganbold_ has quit [Ping timeout: 268 seconds]
Quarx has quit [Read error: Connection reset by peer]
ganbold_ has joined #linux-sunxi
wingrime has joined #linux-sunxi
sachin_ has joined #linux-sunxi
sachin_ is now known as tas_Devil
paulk-desktop has joined #linux-sunxi
<ssvb> naobsd: yes, just uSD card and ethernet
tas_Devil has quit [Ping timeout: 252 seconds]
<ssvb> also disabling anything framebuffer related in the kernel reduces idle power consumption from ~250mA to something like ~215mA
<naobsd> sounds good
tas_Devil has joined #linux-sunxi
<naobsd> do you have same measurement with cb1?
<ssvb> I don't have the original CubieBoard, only Mele A2000 with A10
rellla has joined #linux-sunxi
<naobsd> I see
<naobsd> btw I was surprised 1080p fb consumes memory bus :(
<ssvb> yes, the framebuffer needs to be scanned out 60 times per second, that's not a laughing matter (even theoretically that's ~500MB/s memory bandwidth reduction, but in reality it causes more disruption)
<ssvb> are you testing 1080p with A20?
<naobsd> 1920*1080*4*60=500MB/s
<naobsd> nothing is wrong ;)
<naobsd> no, currently no video out
<naobsd> I didn't use so much, only few testing mainly about booting
jukivili has quit [Ping timeout: 256 seconds]
<hramrach_> we have like 16bit 480MHz ddr2 so it's theoretically like almost 2GB/s
<hramrach_> but I you never get 100% of bus bandwidth.
rellla has quit [Quit: Nettalk6 -]
rellla has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
sanka has joined #linux-sunxi
fluo75 has quit [Ping timeout: 250 seconds]
jukivili has joined #linux-sunxi
<hramrach_> what do I need to make a bootable card?
<hramrach_> this is the second unbootable image I have
rellla has quit [Quit: Nettalk6 -]
<ssvb> hramrach_: cubieboard2?
<naobsd> same as a10
<hramrach_> and 1
<hramrach_> will try to clear the first 1mb
<hramrach_> maybe it does not like junk around the bootloader
<ssvb> there is only script.bin pitfall in the case of cubieboard2
<hramrach_> yes, they wipe it
<hramrach_> but does not help :s
<ssvb> try harder :D
<hramrach_> zero with a hammer ;-)
<hramrach_> actually, this image was zero to start with
<hramrach_> so this is not it
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest86552
_BJFreeman is now known as BJfreeman
Guest86552 has quit [Ping timeout: 252 seconds]
rellla has joined #linux-sunxi
egbert has joined #linux-sunxi
<hramrach_> yes, did what they say in that tutorial and it does not boot
<hramrach_> just boots from flash regardless of card inserted
<hramrach_> the card is readable - when I put it in the slot and the mmc is not fubared I can mount the linux rootfs there
tas_Devil has quit [Ping timeout: 252 seconds]
BJfreeman has quit [Quit: had a good time]
servili007 has quit [Quit: Leaving]
n01 has joined #linux-sunxi
leowt has joined #linux-sunxi
<naobsd> hmm
<naobsd> hramrach_: do you have another SD card?
<hramrach_> yes, I have several
<naobsd> and tried already?
<hramrach_> did not try every card yet
<hramrach_> I have this 16G A-data UHS-I card which worked with the old cubie image so I guess I can try that
<hramrach_> hmm, looks like the UHS-I card really had something to it
<hramrach_> the 15G Patriot card is really slow
<hramrach_> does not boot off the UHS-I card either
<hramrach_> so probably not the card
CaCtus491 has quit [Ping timeout: 245 seconds]
CaCtus491 has joined #linux-sunxi
<hramrach_> heh, it failed
<hramrach_> but now it boots. was wrong spl
<hramrach_> why does it complain about wrong machine??
<hramrach_> I mihgt have aggressive memory timings in script.bin bat that should not affect CPU registers
<hramrach_> you have u-boot-spl.bin and sunxi-spl.bin and only the latter works
<hramrach_> not sure what the other is supposed to be
<hramrach_> Error: unrecognized/unsupported machine ID (r1 = 0x000010bb).
tzafrir has quit [Ping timeout: 240 seconds]
<hramrach_> so the bootloader is bogus :/
<hramrach_> it passes wrong system type
<hramrach_> or both the kernel and the original bootloader are bogus and bug-compatible
<hramrach_> ***
<oliv3r> hramrach_: script.bin dram_para does not do anything to either uboot-spl nor to boot0
<hramrach_> the machine id is hardcoded somewhere in u-boot
<oliv3r> oh that's right, 3.0 and 3.4 use machine ID
<oliv3r> 3.8+ don't use Machine ID anymore
<hramrach_> linux wants 0x00000f35 but u-boot has 10BB
<oliv3r> hramrach_: it's on the ML somewhere
<oliv3r> tom replied to someone else having that exact same error
<hramrach_> hmm, just redefining the machine id
<naobsd> setenv machid 0xf35
<naobsd> saveenv
<naobsd> is necessary with hno's u-boot
<naobsd> or modify kernel source
<hramrach_> I changed it in the u-boot header
<naobsd> is there any docs which mention u-boot-spl.bin?
<hramrach_> did not try to find any
<naobsd> sunxi-spl.bin is always used for a10 and same for a20
<naobsd> (as far as I know)
<hramrach_> yes, wrote the wrong spl to the card
<naobsd> anyway, problem is gone now, everything should be ok :)
<hramrach_> it's probably for pxe
<hramrach_> well, not exactly pxe. tftp
<hramrach_> not sure how you get the spl anywhere
<naobsd> well, sorry, I couldn't understand "not sure how you get the spl anywhere"
<hno> naobsd, u-boot-spl.bin is the actual SPL binary. sunxi.spl.bin is the same wrapped in a sunxi boot header added by mksunxiboot.
<hno> and u-boot-spl is the ELF version, useful if you jtag debug or the like...
<oliv3r> crap
<oliv3r> mbus can either be 'memory bus' or Media bus (for v4l and csi and cameras n shit
<hramrach_> I received the uSD JTAG thingy with cubieboard
<hramrach_> but nothing to connect to the other end :(
<oliv3r> hramrach_: it's not just uSD -> JTAG it's also uSD -> UART
<oliv3r> you can use a bus blaster/pirate for the JTAG end
<hramrach_> well, we are past the point when I would need to, anyway
<oliv3r> and a regular USB uart (that you did get) with the UARt port
<naobsd> well, I know such a thing...
<hramrach_> yes, that one works fine
<naobsd> script.bin need to be midified to use uSD -> UART, right?
<oliv3r> think so
<naobsd> then uSD -> UART works only when executing kernel, right?
<naobsd> hno: I want to try fel/usbboot, but currently u-boot-spl.bin for fel is larger than 8KB
<naobsd> I guess fel binaries under sunxi-tools/bin/ were built from old u-boot
<naobsd> (lichee u-boot?)
<hno> naobsd, have you built the fel version for your board? see boards.cfg.
<hno> fel and noraml spl are different u-boot boards.
<naobsd> I did "make cubieboard_fel_config" and "make cubieboard2_fel_config", both are about 11KB
<naobsd> a lot of functions are compiled/linked
<hno> naobsd, 11KB is fine.
<oliv3r> hno: have you got any idea what MBUS is?
<oliv3r> is it 'memory bus' or media bus
<hno> oliv3r, no idea.
<naobsd> fel write 0x2000 file will hang if file is larger than 8KB
<oliv3r> a13 manual mentions it in the DEFE section
<naobsd> fel write 0x4000/fel clear 0x4000 are also hang
<oliv3r> hno: well mbus is being enabled in spl; so it's interesting at the least for something
<naobsd> on my cubieboard2
<oliv3r> just looking over stuff trying to figure out why spl doesn't work on sun4i
<naobsd> could anyone succeed fel/usbboot on cubieboard2?
<hramrach_> just got a SD image working
<naobsd> can anyone try "fel clear 0x4000 1" with your board?
<n01> mripard_: ping
<hno> naobsd, -rwxrwxr-x. 1 henrik henrik 11632 23 jun 14.57 build/Cubieboard2_FEL/spl/u-boot-spl.bin
<hno> works fine on my Cubieboard2.
<naobsd> hmmmmmmmm
<naobsd> I should try different machine :(
<mripard_> n01: pong
<naobsd> hno: thank. I understand it's my env issue
<hno> ?
<hno> what env?
<naobsd> hno: I understand it works fine on your environment. then my environment should have problem
<n01> mripard_: no reply for my ping for wdt patch. any idea?
<naobsd> something in my environment
<Turl> n01: how much did you wait?
<mripard_> n01: I'll ping the guy
<naobsd> all I need was someone's confirmation
<n01> last ping was 7 days ago
<n01> mripard_: thank you
<hno> naobsd, how are you entering fel mode?
<hno> and do "fel ver" work?
<naobsd> connect USB miniB while pushing fel button
<oliv3r> mripard_: have read gregk's commend about the userspace race? He suggests to use the regular api, but that won't give us a Binary sysfs entry, will it?
<hno> naobsd, "fel clear 0x4000 1" also works fine for me.
<naobsd> fel ver works fine: AWUSBFEX soc=00165100(unknown) 00000001 ver=0001 44 08 scratchpad=00007e00 00000000 00000000
<naobsd> hmmmm
<Turl> oliv3r: did you have a look at the patch I linked the other day?
<mripard_> oliv3r: no, I didn't, you told me Turl answered your question (and I admit I totally forgot to take a look at it for my curiosity)
<oliv3r> Turl: yeah, there they just change create_sysfs_file to the standard stuff
<oliv3r> but i don't think you get a BINARY entry
<oliv3r> so i'm kinda hopeing greg overlooked that
<Turl> oliv3r: it created a class and other stuff from what I recall
<Turl> I don't have my mail client in here to look it up agai
<Turl> n
<oliv3r> Turl: yeah it did a lot of things, but greg said 'use the standard stuff and it will be smaller
<oliv3r> adding a class instead of the sysfs_create won't make it smaller; the sysfs stuff only adds a very few lines; so not sure if htat is the answer
<naobsd> I can do "./fel write 0x2000 bin/fel-boot-cubieboard.bin; ./fel exec 0x2000", I can see "DRAM:" in serial console (then it hangs, probably expected result)
<oliv3r> naobsd: what a20 u-boot branch are you using?
<oliv3r> sounds like you have an a10 u-boot
<naobsd> but "fel write 0x2000 11kb-bin" never complete
<naobsd> oliv3r: hno's wip/a20. but it's not u-boot source issue. fel is not working properly
<naobsd> I thought 11kb is issue, but hno confirmed it's ok
<oliv3r> well the 'hangs after the DRAM:' bit triggered me :) the line above clearly says sun7i?
<naobsd> oliv3r: I said I pushed bin for cubieboard1 into cubieboard2
<naobsd> I never said hang after DRAM: is problem
<oliv3r> ok :)
<naobsd> bin for cubieboard1 is smaller than 8K so my insane fel binary can handle it
<naobsd> hmm
<naobsd> lets try dd if=u-boot-spl.bin of=xxx bs=8192 count=1 ;)
<naobsd> all I need is init code, right? ;)
<oliv3r> heh, i do bs=1024 count=8 :D
<oliv3r> spl is all you need to get DRAM: xxx MiB to work yeah
<naobsd> aaaaaaaaaaarrrrrrrrrrrrrrhhhhhh
<naobsd> I found where is the problem: my brain
<hramrach_> get a replacement?
<naobsd> really sorry
<naobsd> I need to get better brain at first
<hramrach_> sadly last time I looked they did not sell any brain upgrades :s
<naobsd> I noticed u-boot-spl.bin (copied from build machine) is not 11KB when trying to run dd...
<naobsd> ah
<naobsd> shell history function must be evil, my brain is not wrong!!! ;)
<naobsd> anyway
<naobsd> fel/usbboot works fine ;)
<naobsd> sorry for noise!
<hramrach_> cool
<hramrach_> how slow is it to download whole kernel over usb?
<naobsd> real0m12.872s
<hramrach_> for whole kernel?
<naobsd> 45088+246+242800+11632+3402368=3702134 bytes transfered
<oliv3r> hno: well i can tell you that sun4i works up until i broke it with my last local patch ;)
edeloget has joined #linux-sunxi
<naobsd> 3402368 is uImage
<hramrach_> nice
<naobsd> apt-get update my-brain
<naobsd> "E: The update command takes no arguments"
<naobsd> so cool...
<hramrach_> you should use upgrade ;-)
<hramrach_> hehe, I crashed gparted
<hramrach_> resized my flash image to fill whole card and BOOM
<naobsd> u-boot-spl.bin is added too
<naobsd> (sorry, I don't have enough time to write something in wiki)
<hno> naobsd, FEL SPL for cubieboard & cubiebord2 is almost the same size.
<hno> -rwxrwxr-x. 1 henrik henrik 11632 23 jun 14.57 build/Cubieboard2_FEL/spl/u-boot-spl.bin
<hno> -rwxrwxr-x. 1 henrik henrik 11704 23 jun 15.36 build/Cubieboard_FEL/spl/u-boot-spl.bin
<naobsd> yes
<naobsd> please don't mind, my brain was too bad
<naobsd> really sorry
tzafrir has joined #linux-sunxi
<n01> what's happening to github? I get a lot of 404
<hramrach_> use git://
<hramrach_> http is unrealiable :/
<hramrach_> 404 might also mean they have actually taken the stuff down
<hramrach_> but I get random 403 and 5XX errors too
<hno> naobsd, there likely is some KB that can be shaved off, the early fel boot binaries wer about 7KB in size. But as 11KB well within allowed size I have not cared to try to minimize the FEL SPL further.
<hno> instead focused to minimize the source code differences.
<naobsd> yes, 11KB is fine, I copied 22KB (probably not for fel) spl bin. I wonder why I believed there is 8KB limit :(
<hno> 22KB is the full MMC SPL.
<naobsd> yup
<naobsd> really sorry
<naobsd> I always copied mmc one instead of fel one compiled on different directory :(
<naobsd> anyway
<naobsd> I think SUNXI_EMAC word can be removed from _FEL in boards.cfg
<hno> 0x2000-0x5d00 is available + some at other places.
<naobsd> yes, I read your allwinner-info too
<hno> The EMAC driver is only built in full u-boot, not SPL.
<hno> Ideally I'd like to get rid of the _FEL boards entirely and always have both SPL versions built, but it's the simplest way of having the FEL SPL built without chaning too much in the u-boot SPL makefiles.
edeloget has quit [Ping timeout: 250 seconds]
<naobsd> hmm, emac.o is build, but not linked, right?
<hno> no, it's not even built in the spl buildö.
<hno> spl only uses the objects in the spl/ directory.
<hno> and main u-boot the normal ones..
<naobsd> just "make" after "make cubieboard2_fel_config" does some unnecessary thing? there is ./drivers/net/sunxi_emac.o
<naobsd> (I did "make mrproper" at 1st)
<naobsd> surely no emac.o under spl/drivers/
<hno> make with no target builds both spl and full u-boot.
<hno> if you only want to build the spl them "make spl/u-boot-spl.bin"
<naobsd> I just do it
<naobsd> only tools/ and spl/ have .o
<naobsd> I understand right way now...
<naobsd> anyway, SUNXI_EMAC word is not necessary in _FEL line
<hno> It is neccesary. Otherwise full u-boot in a FEL boot can not be used for netbooting.
<naobsd> ah
<hno> there is not really any reason why the full u-boot should differ between a FEL build and a normal build.
<naobsd> I used u-boot.bin built by non-fel target, but
<hno> well... maybe the RAM boot thing should only be in the FEL version, but that's it.
leowt_ has joined #linux-sunxi
leowt_ has quit [Client Quit]
<naobsd> if I don't need mmc spl, I can use u-boot.bin built by fel target... I understand know :(
<hno> yes
<hno> and if I do the change I am thinking about on the line above then you will need the FEL u-boot.bin to do a full boot via FEL.
<naobsd> probably I like single full u-boot.bin ;)
leowt has quit [Ping timeout: 276 seconds]
<naobsd> single target per board, it builds full u-boot.bin and 2 spl.bin
<naobsd> if I can modify Makefiles easily ;)
<naobsd> anyway, sorry for taking your time for my bad brain
<hno> I don't know... not entirely comfortable with normal u-boot looking for a uimage signature at a fixed ram address and assuming it's a boot.scr if one if found.
<Turl> hno: reminds me of script.bin
<hno> Turl, yes.. kind of. But at least it fails gracefully if no ramboot boot.scr is found.
<hno> falling back on normal boot sequence.
tas_Devil has joined #linux-sunxi
<bfree> thankfully I didn't run into any extra sillyness with an external initrd (or nbdroot) on the cubie2 ;) the "setenv machid 0xf35" appeared here just in time to save me saying "me too" on that "problem"
<hno> bfree, official assignment is sun7i MACH_SUN7I SUN7I 4283
<hno> but allwinner selected a random number even if one was already assigned to them..
<bfree> I got/guessed that from what you had said earlier :-/ was just too lazy to recompile anything considering I see both the 3.3 and wip/a20 as just transient bootstrappers ;)
<naobsd> it's time to make "community kernel" which just has proper machid ;)
<hno> yes
<naobsd> I just added "setenv ..." on my web site ;)
<hno> how hard can it be to add A20 to 3.4 kernel...
<hramrach_> seems either the mbr format changed on the nand on sun7i or the nandlib hides the mbr
<hramrach_> because nand_part does not like what it sees on /dev/nand
<hramrach_> the nand is 3989504 blocks on a10 and 3866624 on a20 so there is something missing
<hno> hramrach_, NAND size is not an absolute thing. Differs slightly from device to device even with the exact same NAND chip.
<hno> what do /dev/nand content look like?
<hno> Hmm.. Alejandro appears to have got the branch point slightly wrong when importing A20 sources.
<hramrach_> hno: how much of the content do you want?
<hno> hramrach_, first 64 KB should be quite sufficient I think.
n01 has quit [Remote host closed the connection]
<hramrach_> hmm, the 3.3 a20 linux does not have lots of stuff so just building the sunxi nand driver for it is not going to work
<hno> hramrach_, what do you mean?
<hno> hramrach_, that do look like a sunxi MBR to me. But haven't tried nandpart on it.
<hramrach_> drivers/block/sunxi_nand/nfd/nand_blk.c:1496:5: error: implicit declaration of function ‘script_parser_fetch’ [-Werror=implicit-function-declaration]
<hramrach_> drivers/block/sunxi_nand/nfd/nand_blk.c:1062:4: error: implicit declaration of function ‘gpio_request_ex’ [-Werror=implicit-function-declaration]
<hramrach_> if that looks like mbr then it changed format
<hno> seems so.
<hramrach_> because nand_part does not like it
<hno> hramrach_, I do have A20 NAND driver sources. One moment.
<hno> (GPL version)
<hramrach_> apt on a20 is slooooooooooooooooooooooooooooooooow
<bfree> hramrach_: try in on a raspberry pi :-p
<bfree> s/in/apt/
<hramrach_> does not look good. it does not even show same numbers:
<hramrach_> check partition table copy 0: BAD! 18225e3c, 6bc2c502
<hramrach_> check partition table copy 1: BAD! 0, 67bbaf86
<hramrach_> check partition table copy 2: BAD! 0, 67bbaf86
<hramrach_> check partition table copy 3: BAD! 0, 67bbaf86
<hramrach_> is the GPL version available somewhere?
Soru has quit [Read error: Connection reset by peer]
<hno> Imported and pushing A20 tree to my github.repo as imported/lichee-3.3-a20 when push have completed.
<hno> will merge in the cubieboard2 changes later.
<hramrach_> ok
<hno> It will take a while... only 1Mbit uplink from here.
Soru has joined #linux-sunxi
<hramrach_> oh yeah, consumer link
<hno> ADSL copper link.
<hramrach_> who would ever upload anyting
<hno> uploads are only used by torrent pirates.
<hramrach_> well, the sig is softw311 vs softw411
<hramrach_> and the 411 partition table is much larger
<hramrach_> as in has way more empty space
* hramrach_ pirating Ubuntu CDs
_BJFreeman has joined #linux-sunxi
<hno> hramrach_, something while you wait for the push to complete:
_BJFreeman is now known as BJfreeman
<Turl> hramrach_: damn pirate! you will leave canonical employees homeless if you continue your pirating activities
<Turl> :)
Soru has quit [Read error: Connection reset by peer]
<jelly-home> hno: are those comments supposed to be chinese? The encoding looks broken
<hramrach_> what else
<hramrach_> there are files with CHinese comments in multiple encodings in single flie in the AW trees :/
<jelly-home> assuming Big5 it seems to make some sense
<jelly-home> unsigned int version; // 唳掛陓洘ㄛ 0x00000100
<jelly-home> unsigned int copy; //煦杅
<jelly-home> unsigned int index; //菴撓跺MBR掘爺
<jelly-home> unsigned char magic[8]; //"softw411"
Soru has joined #linux-sunxi
* jelly-home has no idea whether that's valid
<hramrach_> when it looks ok it's valid ;-)
<hno> jelly-home, I haven't recoded them. AW seems to be using Guobiao encoding most of the time.
<hno> never seen Big5 encoding in files from AW.
<hramrach_> yes, browser likes it as GBK
<hramrach_> and looks ok in that too
<jelly-home> hno: thanks, //保留数据,匹配分区信息128字节 (/ / Retain data, 128 bytes matching partition information)
<hno> hramrach_, push completed.
<hramrach_> cool
<hramrach_> looks like the partition structure changed from 64byte to 128
<hramrach_> so for nand_part to handle that it will need an overhaul
<hramrach_> just updating the defines from the header you posted seems to work
<hramrach_> but ideally there should be format detection and stuff
<hramrach_> drachensun: patch for nand-part to decipher new partition format
<hramrach_> seems to work for me but ymmw
<hramrach_> also note that version and magic fiealds are unused in nand-part because if they were used the new format would be rejected
<hramrach_> there are new fields keydata and stamp. not sure nand-part not touching them is good enough
ganbold__ has joined #linux-sunxi
ganbold_ has quit [Ping timeout: 256 seconds]
<drachensun> Ok, so this will need your other patch to make the whole nand as a device too right?
<hramrach_> yes, without the device you get nothing to partition
<drachensun> ok, thanks, I'll try it soon
<drachensun> the old bin2fex, fex2bin doesn't work on the A31 files either
<hramrach_> if you want to write a partition arrange for the new fields to get zeroed. current code does not seem to do that at first glance
<drachensun> its a big pain, I was all set to try and boot this up on a tablet today
<hramrach_> did not try the fex2bin tbh
<hramrach_> just copied the known good bin from android
<hramrach_> is ther a working fex2bin?
<drachensun> yeah, they provide a binary 'script' tool
<drachensun> its called during the packing process
<drachensun> but it doesn't seem to do the reverse
<hramrach_> hmm
<drachensun> I guess I could pack it by hand
<hramrach_> I guess I don't have that
<drachensun> I'm not sure if it really needs the fex anywhere
<hramrach_> thanks for help ppl
<hramrach_> I have a somewhat bootable SD card and looks like I can acces the nand so can start breaking stuff ;-)
_BJFreeman has joined #linux-sunxi
cajg has quit [Ping timeout: 264 seconds]
BJfreeman is now known as Guest19215
_BJFreeman is now known as BJfreeman
Guest19215 has quit [Ping timeout: 252 seconds]
cajg has joined #linux-sunxi
tzafrir has quit [Ping timeout: 276 seconds]
sanka has quit [Quit: Leaving]
<ssvb> drachensun: could you please run on A31 hardware?
<ssvb> I'm particularly interested in the latency test to compare it with
<ssvb> just to confirm that the test can see 1M of L2 cache on A31
tzafrir has joined #linux-sunxi
<hramrach_> ssvb: do the results vary a lot?
<hramrach_> I have a bit different numbers but it's 796 vs 806 or something like that
<hno> drachensun, have they changed script.bin format, or do bin2fex reject some value?
<hramrach_> probably default cpu governor, too
<ssvb> hramrach_: the results may depend on dram clock and other things
<drachensun> it was a value
<drachensun> ssvb: I'm installing git now to try it
<drachensun> fexc-bin: script.bin: size: 35168 (66 sections)
<drachensun> E: fexc-bin: script.bin: 3g_para.bb_vbat: unknown GPIO port type 12
<ssvb> drachensun: thanks!
<hno> drachensun, do you have the original fex?
<hramrach_> running 480MHz presumably
<hramrach_> how do I tell?
<drachensun> hno: not for this device, I've got that mele tablet here and its got a uart on it, so I can debug with it but I've only got the rip of nanda
<drachensun> mele has not been able to provide a firmware image so far
<drachensun> they gave me one right away for the new mele a31 tv box
<drachensun> but not the tablet
<ssvb> hramrach_: lookup the u-boot sources for the physical address of dram controller registers, then use devmem2
<hramrach_> heh
<ssvb> hramrach_: or check a10-meminfo tool (and fix it if necessary)
<hramrach_> do we not have meminfo for a20 yet?
<hno> a10meminfo tells the DRAM frequency used on A10/A13/A20.
<hno> they are all pretty much the same...
<ssvb> hno: great, so the dram controller hw registers addresses are the same in A20
<hno> mostly.
<hramrach_> ssvb: I am running at the default cpu freqencyy, did not overclock
<hramrach_> 480 MHz
<ssvb> hramrach_: the current preliminary tinymembench result in the wiki is from running the original linaro firmware on nand, I still need to figure out why it has better memset performance
<ssvb> hramrach_: something must be configured in a bit different way
<bfree> hramrach_: just seen your saying "HDMI display currently does not work" ... I had a VT on hdmi using 3.3 and an existing debian rootfs I'd been using on cubie1. installed xserver-xorg-video-fbdev and confirmed X also works
<hramrach_> yes, the difference in memset is quite big
<hramrach_> bfree: I get no output on screen. it's probably broken fixed-requency driver
<hramrach_> frequency
<hramrach_> and your screen happens to decode whatefver it is configured to send
<drachensun> ssvb: roughly how long does it take to run?
<hramrach_> a few nimutes maybe
<drachensun> hmmm I probably should have shutdown X before I ran it
<ssvb> drachensun: maybe up to 10 minutes, and yes, it is better to run it on undisturbed system
<bfree> hramrach_: hmm might be, seems it's running 1280x720 (though it's a 1920x1080 screen)
<hramrach_> bfree: it runs 1280x720@50 and my screen is 1280x1024@60
<hramrach_> the dimensions might get overlooked but it really insists on 60Hz
<drachensun> ssvb: So how does it compare?
<bfree> hramrach_: yep just tried playing with the kernel args, no change :-/ anyway I'd still think most hdmi devices should support 720p50 so not as bad as "don't work" (I'd guess your screen is dvi). a regression still of course
wingrime has quit [Ping timeout: 256 seconds]
<ssvb> drachensun: thanks, memory bandwidth available to the CPU is similar to A20, the L2 cache is clearly 1MB
ganbold__ has quit [Remote host closed the connection]
<ssvb> drachensun: are we supposed to expect better results from dual-channel, etc?
<hramrach_> do we have dual-channel? is it used in the board on which that runs?
<hramrach_> thechnically a31 *can* have wider memory bus but that does not mean it is used
<hramrach_> lots of copper to put in the pcb
<ssvb> hramrach_: or it just has lower clock frequency to compensate wider bus :) iirc, oliv3r mentioned 240MHz for dram on A31 earlier, this might explain worse latency for random accesses in a 64MB memory block
<hramrach_> yesh, you have wider bus but so lame PCB that you have to underclock it a lot to work. quite possible
<drachensun> yup my EVB is 240Mhz for the dram
<drachensun> I'm not sure they are all like that though
<drachensun> let me check the mele config
<drachensun> onda972 is 312
<drachensun> so is the mele
<hramrach_> could give better results
<drachensun> The tablets wont boot up yet, well they boot but I have no access
<drachensun> yeah I might try the mele box soon but I'm going to try and fix bin2fex first
<ssvb> drachensun: may I add your pastebin info to wiki?
<drachensun> sure
<ssvb> thanks
<hramrach_> yeah, I get higher memset rate with the factory image too
<drachensun> oliv3r: I figured out whats wrong with the A31 meminfo
<hramrach_> maybe different cache settings?
<drachensun> its trying to divide by 0
tzafrir has quit [Read error: Operation timed out]
<hramrach_> with 432 MHz ram, no less
<drachensun> oliv3r:
<naobsd> ssvb: is this work with a20 3.3 kernel?
<hramrach_> other than frequency hte ram parameters seem same
<hramrach_> naobsd: it will od something
<hramrach_> but g2d is broken and mali needs new binary blobs because kernel side is now r3p2
<ssvb> naobsd: yes, but it will not use g2d, and there is no mali
<naobsd> hmm
<ssvb> hramrach_: currently it just fails the disp version check and refuses to take it into use
<hramrach_> it does not work anyway
<naobsd> is userland mali binary board specific?
<hramrach_> I tried to compile it into the kernel and it complains about something
<naobsd> s/board/SoC/
<hramrach_> naobsd: it's not board specific afaik
<drachensun> hno: I wasn't using the latest fexc tools it seems. I pulled the latest and got all setup to start test-build-test and my first test worked with the latest source.
<hramrach_> naobsd: they used ericsson blobs on odriod, obviously
<hramrach_> you can too if you don't ming the license or find ones licensed for AW
<hramrach_> well, if you are on a31 there will never be any mali
<naobsd> I thought about this
<hramrach_> yes, those
<hramrach_> were licensed for ericsson evb originally and then they put in the correct license for odroid, presumably
<hramrach_> does not allow using the drivers on another board
tzafrir has joined #linux-sunxi
<drachensun> yeah, no mali
<drachensun> I have to figure out this libhybis stuff now
<naobsd> I see
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest1449
_BJFreeman is now known as BJfreeman
Guest1449 has quit [Ping timeout: 252 seconds]
tas_Devil has quit [Ping timeout: 252 seconds]
jemk has joined #linux-sunxi
^hawk^ has joined #linux-sunxi
[hawk] has quit [Ping timeout: 268 seconds]
^hawk^ has quit [Read error: Connection reset by peer]
vincenzo has joined #linux-sunxi
n01 has joined #linux-sunxi
<hno> naobsd, A10 and A20 have different MALI. There is no guarantee one userland build will work on the other. Hopefully, but...
rellla has quit [Remote host closed the connection]
<rz2k> hno: userspace libs are one for everything
<rz2k> I used our libs on mp4 system
<hno> rz2k, all I know is that the MALI configuration is part of the build process when the libs are built.
cajg has quit [Ping timeout: 246 seconds]
Soru__ has joined #linux-sunxi
Soru has quit [Ping timeout: 268 seconds]
cajg has joined #linux-sunxi
oliv3r has quit [Ping timeout: 252 seconds]
rz2k has quit []
oliv3r has joined #linux-sunxi
<oliv3r> drachensun: did you have to recompile?
<oliv3r> drachensun: clock calculation doesn't go through; i take it you took out the mul and the div
<oliv3r> but thanks, that's very usefull indeed
Soru has joined #linux-sunxi
Soru__ has quit [Ping timeout: 268 seconds]
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest10277
_BJFreeman is now known as BJfreeman
Guest10277 has quit [Ping timeout: 252 seconds]
<hno> oliv3r, I haven't had time to look at your patches yet.
<oliv3r> drachensun: the div by 0 is also interesting, means they set up the clock much differently, but i think we heard that before, i think on a31 register setting 0, is value 1
<hno> just get a register dump...
<hno> then work out what it means.
<oliv3r> hno: the big one i'm redoing with a proper split and after i've done sun4i check; was initially gonna do submit patch first; then check for sun4i; but almost done finding why sun4i won't boot
<oliv3r> hno: i tried to rewite a10meminfo for a31
<oliv3r> well like an initial version first
<oliv3r> see how badly it goes
<oliv3r> going pretty bad
<oliv3r> memory controller looks so differnt, it makes you want to cry
ganbold has quit [Ping timeout: 268 seconds]
jinzo has quit [Quit: Stahp, stahp, WHAT ARE YOU DOING?!]
jemk has quit [Quit: bye]
n01 has quit [Ping timeout: 248 seconds]
<rm> hm, trying to set up an MK802II, now it hangs under load
<rm> somehow didn't notice that when I used it earlier
<rm> I wonder which AXP it has, and if my kernel is for a wrong one
<rm> should I try a kernel without AXP support, will that fix the CPU to maximum voltage?
paulk-desktop has quit [Quit: Ex-Chat]
<hno> oliv3r, good. I'll wait for next version then :)
<hno> It is expected the A31 memory controller is different.
Soru has quit [Read error: Connection reset by peer]
soul has joined #linux-sunxi
soul has quit [Read error: Connection reset by peer]
soul has joined #linux-sunxi
soul has quit [Read error: Connection reset by peer]
derethor has joined #linux-sunxi
Soru has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest13525
_BJFreeman is now known as BJfreeman
Guest13525 has quit [Ping timeout: 252 seconds]
tzafrir has quit [Ping timeout: 256 seconds]
vincenzo has quit [Remote host closed the connection]
vincenzo has joined #linux-sunxi
vicenteH has quit [Ping timeout: 246 seconds]