<n01_>
mripard: thank you for replying on LAKML about the WDT driver. Life is a bit complicated for me at the moment
<mripard>
n01_: that's ok, don't worry :)
<mripard>
if you have any comments and not enough time to make the needed changes, feel free to tell me, I'll do them if needed.
<mripard>
(keeping your author name, of course.)
shineworld has joined #linux-sunxi
<shineworld>
sorry for silly question, I was far for some weeks, but where I can find u-boot git for A20 ? Looking in main u-boot branch of linux-sunxi/u-boot I can look only board.cfg for cubieboard v1. Thank you for reply
<rellla>
is stage/sunxi-3.4 useable or broken atm?
<n01_>
mripard: thank you, I hope things will get better soon
<mripard>
n01_: I hope so as well :)
gzamboni_ has quit [Read error: Connection reset by peer]
<hramrach__>
but I don't use hwpacks myself so I have no idea if it actually works
<H6r6tiqu6>
si it just a kernel ?
Yaku-noob has joined #linux-sunxi
<H6r6tiqu6>
*is
<hramrach__>
kernel, u-boot, modules
vicenteH has joined #linux-sunxi
<hramrach__>
should be as normal hwpack unless I missed something
<H6r6tiqu6>
okay
rellla has joined #linux-sunxi
* hramrach__
puts making image with standard machine id on todo list
<buZz>
0xf73?
<hramrach__>
the other I guess
<H6r6tiqu6>
hramrach__: what are the differences between the hwpack and the sdcard image ?
<hramrach__>
you can use the hwpack with another image such as cubian
<hramrach__>
the image is just iamge you write to a card
<H6r6tiqu6>
okay...
<H6r6tiqu6>
I'm not sure to undestrand the purpose of the hwpack x)
<hramrach__>
if you want a feature of cubian such as the morse-code IP address it might work is you update it with the hwpack
rz2k has joined #linux-sunxi
<hramrach__>
the full image is quite reliable but has no features
<hramrach__>
you have to add them on your own
<H6r6tiqu6>
meeeeeh...
<H6r6tiqu6>
I just want a system to boot on x)
<H6r6tiqu6>
apt-get is my friend, then
<hramrach__>
and vim and make
<H6r6tiqu6>
yep(
<H6r6tiqu6>
*'
<shineworld>
hramrach__, in boards.cfg (u-boot) there are two cubieboard2 configs : Cubieboard2 and Cubieboard2_FEL. What is the main difference ? I've choose first, is right ?
<hramrach__>
FEL is for booting over USB
<shineworld>
ok right the first so...
<shineworld>
thank you
<wigyori>
that's similar for the a13 olinuxino board, right?
Soru has quit [Read error: Operation timed out]
soul has joined #linux-sunxi
hansg has joined #linux-sunxi
Yaku-noob_ has joined #linux-sunxi
Yaku-noob has quit [Ping timeout: 246 seconds]
Yaku-noob_ is now known as Yaku-noob
dodo has joined #linux-sunxi
<dodo>
Does work LAN with Cubieboard2 Android 4.2.2 ? I don't find connections to lan with netcfg
<ssvb>
hansg, mnemoc: after a20 is merged into 3.4, can we drop 3.0 support?
<dodo>
eth0 says: eth0 DOWN 0.0.0.0/0 0x00001002 22:a0:a0:03:c2:88
<hansg>
ssvb, that is up to mnemoc
<ssvb>
hansg, mnemoc: I'm just a bit tired of the memory reservation mess, and 3.4 with CMA backported is a nice and clean solution
paulk-desktop has quit [Quit: Ex-Chat]
<oliv3r>
not until we can verify all android users can use 3.4 reliably
<oliv3r>
i have a uSD breakout board and once i have a few minutes to set it up to get output, i'll try 3.4
<oliv3r>
but i'll have to port my touchscreen driver over somehow :(
<oliv3r>
android users are still an important usergroup unfortunatly
<ssvb>
hansg: or actually, what do you really want to be done about it?
<ssvb>
oliv3r: the android users seem to be holding us back, with no real benefit
<oliv3r>
well not the android users themselves
<oliv3r>
but we have too little android 3.0 users that actually use our kernel reporting whether 3.4 works reliably or not
<oliv3r>
but as I said, gotta get that uSD thing working to output stuff
<oliv3r>
then compile 3.4
<oliv3r>
add touchscreen support and use it )
<oliv3r>
i want to do some mainline work tonight, then try to do uSD stuffs
PiyushVerma has joined #linux-sunxi
<oliv3r>
wingrime: are yo running 3.4 on your tablet?
<ssvb>
that's a sad situation, and it is really holding us back (for example, I got no feedback about the dynamic allocation vs. boot time reservation for mali memory in android)
<ssvb>
even though I tried to ping some people
vrga has joined #linux-sunxi
<oliv3r>
our biggest problem right now, second to the patches coming in ;) is people testing and running stuff
<oliv3r>
actually reporting stuff
vrga has left #linux-sunxi [#linux-sunxi]
ykchavan has joined #linux-sunxi
<wingrime>
oliv3r: now 3.0
<ssvb>
why 3.0?
<wingrime>
oliv3r: but 3.4 have some problem related suspend
<oliv3r>
you don't run android?
<wingrime>
oliv3r: if on 3,4 xorg are running resome crashes
<wingrime>
oliv3r: stock on nand
<dodo>
well now working, was hard to enable network in android settings because are by default in chinese
<oliv3r>
wingrime: we (me too) should start upgrading stock kernel to 3.4
<wingrime>
oliv3r: reason?
<oliv3r>
testing 3.4 with android, as a replacement for 3.0
<oliv3r>
so that we can retire our 3.0 tree
<oliv3r>
and focus on 3.4
<wingrime>
oliv3r: I know
<wingrime>
oliv3r: but look like problem related framebuffer or disp
<oliv3r>
wingrime: oh, ssvb can fix that ;)
<ssvb>
oliv3r: I can't, at least not unless I can reproduce it myself :(
<ssvb>
oliv3r: and I have been relatively less motivated lately, but that's another story
<oliv3r>
i can motivate you
<ssvb>
:)
<oliv3r>
if you make it happen; there will be boobies
<wingrime>
oliv3r: more strange I have same bug using second tabelet with fedora
<wingrime>
oliv3r: so it defenetly reproduceably
<oliv3r>
can you dd your sd card to an image and upload it somewhere?
<wingrime>
oliv3r: it happends without ssvb driver
<oliv3r>
that way, ssvb can download it and verify it
<wingrime>
oliv3r; in second case this is hans fedora with hans kernel without sunxifb
<wingrime>
oliv3r: just enable PM in config and try suspend
<wingrime>
oliv3r: with XFCE button on top
<hansg>
ssvb, about: http://thread.gmane.org/gmane.comp.hardware.netbook.arm.sunxi/628 as mentioned in the thread, I think the Kconfig options should only set some defaults (so as to be able to build a kernel for android which does not need special cmdline options to make android work) , and that everything should be fully dynamic, controlled from the kernel cmdline
<hansg>
So:
<hansg>
1) There will still be an option to reserve memory for the mali driver, controlled from the kernel cmdline, the location may be dynamic
<hansg>
2) If there is no reserved memory the code will fallback to dynamic memory
<hansg>
3) The default amount to reserve (when no kernel cmdline option is given) is set from Kconfig
<ssvb>
hansg: the boot time mali memory reservation is only supposedly needed for android, and even this is not certain
<ssvb>
hansg: android needs a separate kernel config anyway
<hansg>
ssvb, hmm, so you're suggesting to make the reservation / no reservation a Kconfig option only, and I guess then also with a fixed size?
<ssvb>
hansg: I don't want to change android behavior (as I'm neigher using nor testing it)
<hansg>
ssvb, same for me (neither using nor testing it, nor do I want to break it).
Black_Horseman has quit [Remote host closed the connection]
<ssvb>
hansg: but for linux, the existence of weird cmdline kernel options is confusing, and the people will surely try to misuse them
<hansg>
ssvb, so lets have (keep) the Kconfig option for Android, and maybe document it can / should be turned off for non android builds. And then for non android builds (so the kconfig option off) simply switch to all dynamic memory ?
<hansg>
ssvb, I would be happy to merge a patch doing something like that in my 3.4-a20-wip tree, provided that it has been tested with the mali driver under regular Linux. I know the mali driver is of limited use under regular linux but I don't want to break it
<ssvb>
hansg: somebody will try to ensure that mali works on a20 with 3.4 as soon as 3.4 can boot on it and is "stable"
<ssvb>
hansg: probably mali can work even now with some minor tweaks for the register addresses and the other hw configuration (A20 is MP2 and A10 was MP1)
<hansg>
ssvb, I was not talking about testing on the a20, just that I would like to see a tested (on the A10) patch-set to merge in my 3.4-wip-a20 tree. which is where most recent sunxi development is currently happening
<hansg>
ssvb, also I have the a20 booting and stable, as long as you don't try to use hdmi audio :)
<ssvb>
hansg: so basically your git branch should be now used for development instead of mnemoc's sunxi-3.4 and stage/sunix-3.4?
<ssvb>
hansg: I was kinda waiting for it to be merged to the linux-sunxi github repository
<mnemoc>
hansg: hi, from your huge sent of branches... is there a one for me to blindly pull and solve a month of offline-ness in one pass?
<mnemoc>
s/sent/set/
<wingrime>
mnemoc: you finaly return))
<mnemoc>
i'm melting, but i'm here :)
<wingrime>
mnemoc: current branch need merge manypatches
<wingrime>
mnemoc: I glad you return, maybe I will thik about rewrite axp crap
<mnemoc>
wingrime: can you do a pull branch for me?
<wingrime>
mnemoc: currently - no
<mnemoc>
:(
<wingrime>
mnemoc: but I have still unmerget tuchscreen driver, some dma patchs
<wingrime>
mnemoc: simply not sure about 100% workable
<wingrime>
ssvb: I not have any ethernet enabled device yet
<hansg>
Where I'm slowly but steadily bringing the a20 up. which is going well
<mnemoc>
working from an awfully hot shared dorm using the phone as modem. i can pull in whatever you consider proper
<hansg>
Since you were offline I've also started merging various other patches there.
<mnemoc>
but can't read the ton of unread mails i have after a month offilne
<ssvb>
mnemoc: before talking about the new patches, what is the status of the current stage branch?
<mnemoc>
no idea
<wingrime>
mnemoc: why not give hansg right to aprove ?
<mnemoc>
there is plently people with commit access
<ssvb>
mnemoc: is the current stage stable? hramrach__ seemed to have some regressions with it
soul has quit [Remote host closed the connection]
<hansg>
I can probably reshuffle patches to move the a20 stuff to the end of the history and then do a branch for you with everything I've reviewed and merged
<mnemoc>
your call
<mnemoc>
ssvb: not nice to know there are regressions in stage/sunxi-3.4 :(
<wingrime>
hansg: I recent hear that otg fix are make regession with android?
<hansg>
wingrime, my tree also has the fix to the fix, fixing the regression
<wingrime>
hansg:nice
<wingrime>
mnemoc: what city?
<mnemoc>
berlin
<hansg>
(or so I've deducted from the patches I've merged)
<hansg>
mnemoc, yeah the weather is too hot in Delft too
<mnemoc>
wingrime: i'm used to the non-stop flow of cold air from the atlantic
<hramrach__>
I cannot play video at all. creating output always fails with vlc
<ssvb>
hramrach__: with the stage branch?
<wingrime>
mnemoc: I still wait your soc detect and we can try begin make unifited kernel
<hramrach__>
yes, with the stage branch built sometime around start of the week or so
<ssvb>
hramrach__: have you tried to bisect it?
<mnemoc>
will pull that in in 10m. let me go to the toilet for water. brb
<hansg>
hramrach__, it would be good to bisect it
<hramrach__>
did not try yet
<hramrach__>
was offline most of the week
<wingrime>
mnemoc: here 26 C today
<ssvb>
hansg: maybe a call for testing in the mailing list would be a good idea? for the current sunxi-3.4 vs. stage/sunxi-3.4 vs. your new branch?
<hansg>
ssvb, a call for testing never hurts
<derethor>
which device do you need for testing?
<derethor>
cubieboard2?
soul has joined #linux-sunxi
jlj has quit [Read error: Operation timed out]
<ssvb>
hansg: and first graduating stage/sunxi-3.4 to sunxi-3.4 before piling a lot of new stuff may be also reasonable
<Turl>
hramrach__: what's your wiki username?
<mnemoc>
ssvb: that was my idea, until the surprise with hramrach's regression :(
<Turl>
hramrach__: hopefully I found the right user in the wiki, you should be able to skip the spam filter fully now
jlj has joined #linux-sunxi
<Turl>
just don't become a spambot overnight please :)
<Turl>
mripard: Tsvetan ping :)
<Tsvetan>
pong
<Turl>
Tsvetan: btw, are you going to drop it from all boards?
<Tsvetan>
Turl: yes
<Tsvetan>
no use of this EEPROM
<Turl>
I thought it was a cool feature
<mripard>
Turl: pong?
<Turl>
mripard: you know the A10S board currently has an EEPROM right?
<Tsvetan>
Turl: what you can do with this EEPROM
<mnemoc>
for products it's good to have where to store a serial number, or mac address, or ....
<mripard>
well, yeah, you sent a patch enabling it :)
<Turl>
mripard: well, Tsvetan is going to drop it on newer revisions
<Turl>
mripard: how should we handle it?
<Tsvetan>
but all this you can store in script.bin or file anyway
<Turl>
Tsvetan: store mac address, keys
<Turl>
Tsvetan: yes but script.bin is not used on mainline
<mnemoc>
Tsvetan: but makes updating more complex
<Turl>
and file is "per sdcard"
<Tsvetan>
or NAND flash
<Turl>
if I put the same sdcard on another device it keeps the mac address
<mnemoc>
Tsvetan: if it's stored on eeprom and images are identical
<mripard>
Turl: it's I2C, so I guess we can always load it
<mnemoc>
keeping uniqueness of each device on hw
<mripard>
and see if we can talk to it :)
<Tsvetan>
well we didnt do any new assembly without EEPROM, if you find it useful we can keep it
<Tsvetan>
I just though its useless
<Turl>
I think hno liked the EEPROM too
<Tsvetan>
as MAC etc can be stored in file too
<Tsvetan>
and you can set the MAC in user space when read it from this file
<mnemoc>
re-flash, game over
<wingrime>
Tsvetan: I thick it needed on a20 only becose a10 have SID for it
<mripard>
but I agree with mnemoc and Turl, maybe it's because I'm a software guy, but I find it much more convenient to have everything hardware-related stored in the hardware :)
<hansg>
My tree has a patch to use the random parts of the sid to generate a unique mac
<mnemoc>
also in the case of products, you can keep a signed crypto key on the eeprom to safely identify the unit
<mnemoc>
again reflashing-safe
<hansg>
But the a20 on the cubie2 has an all empty sid (all 0), so I need to add a patch to detect that and fall back to a random mac if none is set through another means
<mripard>
and you don't have to worry wether you boot from the NAND or the SD Card, etc.
<hansg>
I've got to go, see you all later.
hansg has quit [Quit: Leaving]
<wingrime>
hansg: random mac is good idea only for first boot
<Tsvetan>
this EEPROM is cheap so no problem to keep it, I just though that nobody is using it until today I saw this patch from Turl
<Tsvetan>
:)
<mripard>
Tsvetan: and I started a discussion on the kernel mailning list on friday to support such usage of an EEPROM easily possible in the kernel actually :)
<mripard>
exactly by thinking about the AT24 in the olinuxinos :)
jlj has quit [Ping timeout: 256 seconds]
<wingrime>
Tsvetan: if sid are 0 than we can test eerom, after if eerom crc32 fails than generate new mac and store to eeprom
<Tsvetan>
ok, then we will keep it
<Turl>
:)
<wingrime>
Turl: we can boot using SPI ? can we i2c?
<wingrime>
i remeber we can SPI
jlj has joined #linux-sunxi
<Turl>
wingrime: I think SPI NOR is supported as boot method
<Turl>
but I don't think you can boot from i2c
<Tsvetan>
even if you can it will be slow
<Tsvetan>
40K/sec
<derethor>
i bought a cubieboards2 right now
<derethor>
i hope we can test the kernel here too
<derethor>
my olinuxino seems to be broken :/
<Turl>
Tsvetan: btw, did you write anything to the EEPROM before shipping?
<wingrime>
derethor: It nearly impossible...
<Turl>
Tsvetan: mine has something on it but does not look like a mac address
<Tsvetan>
here in Bulgaria the prices of the raw materials through EU distributors are x3 times higher than if you buy from China and imprt
<Tsvetan>
just the import here is very easy
<rz2k>
wingrime: also try to attend to "ExpoElectronica" and ".newElectronics" exhibition next year in Moscow. I've made many new contacts for my company there, iirc, I've had even cheaper pcb prototyping, but I cant remember the names.
<wingrime>
rz2k: I am in yekaterinburg
<wingrime>
rz2k: but it not change anything
<rz2k>
pselectro has office there, iirc.
<wingrime>
Tsvetan: are you not heared about russian mail logistics....
<rz2k>
also check if your board is cheaper if done in quantity, for exmaple 10 prototypes can be cheaper than one in some cases
<rm>
wingrime, hola from Tyumen o/
<wingrime>
rz2k: and more one , I need find some 'not common' components
<Tsvetan>
wingrime posts here are same, in post here work only ppl who cant find other decent job, low salaries, no motivation
dodo has quit [Quit: Page closed]
shineworld has quit [Remote host closed the connection]
* rz2k
hopes nothing horrible will happen to UPS/DHL soon
<Tsvetan>
rz2k :)
<wingrime>
Tsvetan: mltiply it with russia size
<rz2k>
this is only way to not wait till next year
<derethor>
what difference, in spain, people fight hard to become a postmail worker
<derethor>
it is a secure job for life
<wingrime>
rz2k: UPS/DHL have problems with
<wingrime>
customs
<wingrime>
?
<wingrime>
,
<wingrime>
Tsvetan: are you traced DDR3 by your self ?
Yaku-noob has quit [Ping timeout: 246 seconds]
<wingrime>
rm: cool
<Tsvetan>
wingrime: not really, I have 12 developers here and If I have to route the PCBs myself something will be very wrong :)))
zeRez has joined #linux-sunxi
<rz2k>
wingrime: no, if you will not go more than 200 euro per month by them
<rz2k>
its their limit
<rz2k>
if you go more - your package will be stopped and you will need to fill out documents needed for customs clearance (why you bought it, what is and etc.)
<wingrime>
rz2k: +400$ recently?
<rz2k>
wingrime: I checked with calculator on ps-electro. without silkscreen 10 4-layer boards with gold immersion will be 4400rub. this is good price, i think. 440 ($20) per board...
<rz2k>
no idea about new rules, month before my pal had a package with $520 price stopped :(
<wingrime>
Tsvetan: where I can find hi-freq SMD transformrer ?
<Tsvetan>
have no idea, we do not use such
<wingrime>
Tsvetan: why, inductors are good prower filter, it strange that you not use it...
<Tsvetan>
transformers and inductors are differnt things ;)
<wingrime>
Tsvetan: same vendors
<Tsvetan>
not for sure
<Turl>
mnemoc: :)
<wingrime>
Tsvetan: so it more one problem, I know whan I can bye ADC and FPGA but I have no idea when I can get transformer for ADC frontend
<mnemoc>
and finally a new stage/sunxi-3.4
<Tsvetan>
if this is for prototypes you have no much choices: Digikey, Mouser, Farnell
<wingrime>
Tsvetan: in case 100pics ? (If It will be usable)
<mripard>
and yes, because the address is parsed by the i2c core
dwilkins has joined #linux-sunxi
<hramrach__>
Turl: thanks for enabling the spamming
* hramrach__
spams wiki
* Turl
hits hramrach__ with spam cans
<hramrach__>
ouch :<
pirea_ has quit [Quit: Leaving]
<Turl>
:)
dwilkins has quit [Excess Flood]
<hramrach__>
it still asks for captcha but does not break because of spam filter
<Turl>
hmm, we need an A10S-OLinuXino page on the wiki
<Turl>
hramrach__: yeah captcha is not bypassed by this I think
<hramrach__>
still if the filter blocks the thing I just added what is supposed to pass it?
<Turl>
hramrach__: maybe androtab.info is engaged with spammy activities
<Turl>
I don't think the github links would trigger it
<hramrach__>
like somebody added links to their site to spam for credibility?
dwilkins has joined #linux-sunxi
<Turl>
yeah I dunno, just a guess
<Turl>
it doesn't look like they would, looking at their site
<hramrach__>
also it looks like the filter just fails and does not trigger
<Turl>
which filter?
<hramrach__>
the spam filter
<Turl>
which one, akismet?
<Turl>
if akismet rejects it you're out of luck, indeed
<hramrach__>
yes
<hramrach__>
it said blacklist error
<Turl>
but so far you're the first th be hit by a false positive
<Turl>
(at least that I'm aware)
<hramrach__>
does blacklist error mean the filter failed or that it rejected the diff?
<oliv3r>
still backreading; but as hansg said, itśs a stage branch; i say put it there and let people test it, if its all good; great, if not, well I guess we have to fix/bisect it
<Turl>
hramrach__: I would have to see the full error
<oliv3r>
sharing the burden isn bad either :)
<Turl>
hramrach__: "akismet.com blacklist error" means it's been clasified spam
<hramrach__>
The following text is what triggered our spam filter: http://akismet.com blacklist error
<hramrach__>
then it works wrong
<hramrach__>
there is supposed to be some un-spam feature
<hramrach__>
good luck finding it in the wiki interface
<hramrach__>
also if the comments don't pass teh filter it does not learn they should pass :/
dwilkins has joined #linux-sunxi
<Turl>
hramrach__: yeah the extension is half implemented, and half crippled due to how akismet works
<mnemoc>
:(
<Turl>
hramrach__: you submit the text, and it returns ham or spam
<Turl>
it doesn't tell you what exactly triggered it though
<hramrach__>
secret sauce
<hramrach__>
still you need the spam/un-spam feature for it to work effectively
<mnemoc>
submitting current content for review can also be a good idea
<Turl>
hramrach__: yeah, but mediawiki doesn't have such support either
<mnemoc>
in can the "spam" part is in the current content and not in the change
<mnemoc>
s/can/case/
<Turl>
if it turns out to bring more trouble than solutions to the table we can disable it
<Turl>
for known good contributors you can add them to the people group and it'll let them bypass that as well as edit user pages (which spammers seem to use solely for their purposes)
<mnemoc>
what about changing the error to "we have doubts if you are a bot or not, please join us on irc to sort out the problem"
wingrime has quit [Ping timeout: 246 seconds]
<mnemoc>
and then mark the user as People
<jelly-home>
that's what the bots will do come the Skynet Revolution time
<Turl>
mnemoc: sounds good to me
<Turl>
mnemoc: do you remember which page had the irc instructions?
<mnemoc>
nope.... [[IRC]] ?
<Turl>
that should work
<mnemoc>
and less scary than "akismet.com blacklist error"
Yaku-noob has quit [Quit: ChatZilla 0.9.90 [Firefox 22.0/20130618035212]]
<mnemoc>
the tool doesn't smell very reliable anyway
<Turl>
it works pretty well for WP comments in my experience
<mnemoc>
so let's blame the plugin
<Turl>
We have doubts if you are a spambot or not, please <a href="/IRC">join us on IRC</a> to sort this issue out.
<Turl>
I changed the message to that
<hramrach__>
the thing is that WP comments and l-s wiki content are way different
<hramrach__>
adding just 4 links is 99% spam on WP but okish on technical wiki
<hno>
hi mnemoc!
<mnemoc>
hi hno :)
<hramrach__>
hi :)
<Turl>
hramrach__: yeah you have a point
<Turl>
it does have a content type field set to wiki though
shineworld has joined #linux-sunxi
soul has quit [Read error: Connection reset by peer]
Black_Horseman has quit [Remote host closed the connection]
soul has joined #linux-sunxi
<oliv3r>
SID is very unreliable, iso far, only AW progrms it, and sometimes its all zeroś's. the eeprom is great for easily modifiable stuff, as said above, mac being the primary one. and also as said above, change the sdcard from device to device, identical mac, clone card, identical mac, so per device mac is good via SID/EEPROM
<hramrach__>
mnemoc: I tried older kernel and I still cannot play video
<oliv3r>
also, a10, a12, a13, a10s and a20 all have sid, only a31 i haven foun it yet, but don have hardware
<mnemoc>
hramrach__: can you try different sunxi tags? to narrow it a bit
<hramrach__>
I am not sure it's the kernel or some other stuff I accidentally changed
<mnemoc>
then test the oldest kerel you remember worked
<hramrach__>
I generally use some random merge of multiple branches :/
<hramrach__>
but I have old kernel on ssd maybe
<oliv3r>
wingrime: have you not seen the RTL-SDR movement, very cheap SDR hardware
<oliv3r>
mripard: do you have more background info on that SMP bit thing?
<mripard>
oliv3r: well, without that bit set, the CPU won't enable the caches
<mripard>
I still have to test that, but it seems to be the reason why I have these hangs in the kernel boot on the A31
<mripard>
and the A20 is concerned as well, since it's a Cortex-A7
<hramrach__>
mnemoc: last kernel I used was based on c7d23318c
<hramrach__>
will try to copy it back
<hramrach__>
mnemoc: I had a typo in the vlc invocation
<hramrach__>
and the vlc error mesages are soooo helpful
<mnemoc>
hramrach__: so no kernel regression? :)
<hramrach__>
did not try the latest kernel yet but looks like it was that typo
<hramrach__>
am I the only one who playes video with latest kernel? :s
<hramrach__>
anyone is seeing reboots as not exactly reliable?
<hramrach__>
seems sometimes the kernel gets stuck powering off
<hramrach__>
maybe it was broken revision. did not see that before
<hramrach__>
but booting from nand gets stuck like every other time from the very first time I installed Linux there
<mnemoc>
what happened to the plan of making ffmpeg or libva bindings? instead of per-player bs
<hramrach__>
nobody wrote them
<ssvb>
mnemoc: I still have plans for vdpau
<mnemoc>
any pseudo-standard api is better than per-player
<ssvb>
just some other things need to be done first
<mnemoc>
assult aw's hq?
<ssvb>
not really :)
<hramrach__>
rellla: the 1080i sample plays reasonably well in vlc
<hramrach__>
there is no noticable tearing or anything
<hramrach__>
it's no quick changing scene, though
<rellla>
hramrach__: so its time to find out, whats wrong with xbmc..
<rellla>
yeah ~20s
<rellla>
haven't found some time to test it yet...
<rellla>
due to spam cleaning ;)
<hramrach__>
yesh, really bad in xbmc
<Turl>
rellla: :p
<hramrach__>
ok, the latest kernel works too
heffer_ is now known as heffer
heffer has quit [Changing host]
heffer has joined #linux-sunxi
<hramrach__>
e75be343e97 with some random patches on top
<hramrach__>
where do you put links to hwpacks?
tinti has quit [Ping timeout: 260 seconds]
ykchavan has quit [Read error: No route to host]
shineworld has quit [Remote host closed the connection]
ykchavan has joined #linux-sunxi
ykchavan has left #linux-sunxi ["Leaving"]
<oliv3r>
hramrach__: in kernel we use u32 anyway :)
tinti has joined #linux-sunxi
notmart has quit [Quit: notmart terminated!]
<oliv3r>
mripard: if hno is busy, iĺl look at it tomorrow
<oliv3r>
i need to start a todo list :S
<oliv3r>
mripard: so appearantly boot01 doesn't set it either, if your having toruble now
<mripard>
yeah, but I looked at their A31 kernel and don't see it either
* hramrach__
spams wiki some more
tinti has quit [Quit: Leaving]
<oliv3r>
mripard: so why does their kernel run right?
<oliv3r>
btw, tom said, a31 is conciderd 'a flop' by aw managment :)
<oliv3r>
maybe they forgot the bit :)
<mripard>
the info comes from two ARM engineers
<mripard>
so they shouldn't have "forgot" it
<oliv3r>
pirea_ ping
<mripard>
but rather patched whatever was coming from the A7 from ARM :)
<oliv3r>
mripard: from the ML? oh i belive that is true
<oliv3r>
maybe AW forgot it
<mripard>
actually, someone only pointed it to me
<mripard>
a colleague of mine was speaking with one of the ARM engineers, saying that I had weird hangs at boot with the A31
<mripard>
and they just said that it was probably this issue
<mripard>
but the kernel we all have on the A20/A31 seems to be working fine though
<mripard>
without these hangs
<oliv3r>
well the ML post said the kernel assumes bootloader does it
<mripard>
so either we don't have the right boot0/1 code, or the right u-boot one, or the kernel one
<mripard>
yeah, it's a design decision from mainline :)
<oliv3r>
we dont' officially have boot0 for a31
<oliv3r>
and AW doesn follow ml :p
<mripard>
the CPU doesn't care at all who sets this bit
<hramrach__>
everyone set it ! ;-)
<mripard>
yeah, but I'd expect them to at least listen to ARM :)
<oliv3r>
but if you haven't found it in boot/kernel
<hramrach__>
I mean you could set it in boot1 *and* u-boot
<oliv3r>
yet, they dn have issues
<mripard>
I haven't been able to run the boot0/boot1/u-boot code they provide.
<mripard>
so for all I know, we know nothing :)
<mripard>
except for what ARM tells us
<hramrach__>
should it not run without cache as well?
<oliv3r>
mripard: you haven booted anything yet? :p
<oliv3r>
mripard: btw, read gregś's mail? :)
<oliv3r>
he made me chase a ghost for a week or two
<mripard>
yeah, I did
<mripard>
see why I was telling you that no-one's a god, even gkh :)
<mripard>
and to ask him questions :)
<mripard>
oliv3r: and I do have a booting kernel
<hramrach__>
no no no, everyone's a god ;-)
<mripard>
only with binaries provided by allwinner.
<hramrach__>
you did not sprinkle your binaries with enough Chinese magic sauce
<mripard>
:)
<oliv3r>
so you boot stock boot/u-boot with your on kernel and performance is poop, but their kernel wroks fine and they don't set the bit
<mripard>
no, actually, that's what is odd
<oliv3r>
have you checked if the bit is said
<mripard>
once booted, and even during the boot, it's not *that* slow
<mripard>
in console, you don't even notice it
<mripard>
but if you look at the timestamps, you have two times where there's a pause of about 6s
<mripard>
at decompression and just after the dentry/inode cache
<oliv3r>
i noticed that
<oliv3r>
sounds resonable
<hno>
mripard, I am not quite up to speed on A20 and A31 matters, but is there anything specific you want me to look into on A20? (don't have an A31 yet)
<oliv3r>
so have you logged in and dumped the register that sets the bit?
<mripard>
hno: I haven't started on the A20 yet :)
<mripard>
but I plan to start working on it quite soon
<oliv3r>
hno: appearantly thereś as SMP cache enable bit for cotrex a7
<oliv3r>
the linux kernel expects the bootloader to set it
<mripard>
especially if these hangs are because of a bit not set
<mripard>
we will notice the same thing with the A20 I guess
<mripard>
but yes, like oliv3r says
<hno>
any reference?
<oliv3r>
and with u-boot-spl for a31, itĺl be hard to test the bit
<mripard>
the Cortex-A7 will enable its caches only if the SMP bit is set, and mainline developpers said that it was up to the bootloader to set it
<hno>
u-boot do set the bit on a number of different cpus or boards. But nothing generic.
BJfreeman has quit [Quit: had a good time]
<hno>
boot0 do set it.
<hno>
mmu_set_smp
<hno>
Hmm.. but call is commented...
<hno>
mripard, can you please dump BROM from your A31? It's easily done using fel.
<hno>
fel write 0xffff0000 0x8000 A31-BROM.bin
<hno>
Err.. I don't rmember the syntax of my own tool.
<hno>
fel read 0xffff0000 0x8000 A31-BROM.bin
<mripard>
hno: I will
<hno>
A20 FEL seems to have code setting it, but can't find any in BROM...
<oliv3r>
hno what branch do i need for nand + uSD uart boot
<hno>
which nand?
<oliv3r>
i want to debug my stock tablet soon
<oliv3r>
can we read the smp bit from uboot console?
<hno>
if using AW bootloader then it's configurable.
<hno>
don't know, but it's easily read with ARM instructions.
<oliv3r>
yeah, but what branch?
<hno>
oliv3r, you need to be more specific.
<oliv3r>
ah ok so we can check it at the very least
<oliv3r>
hno i want to replace my stock nand bootloader with one that does libnand but also uSD
<hno>
you don't need to replace u-boot for that.
<hno>
it's set in script.bin.
<oliv3r>
ah ok some said i did
<oliv3r>
ok, good ill experiment
<hno>
remember to disable the MMC...
<oliv3r>
boot0 doesnt output via it thoguh, does ir
<oliv3r>
since boot0 is to small to parse script.bin
hramrach__ has quit [Ping timeout: 240 seconds]
<hno>
It does if you flash with an image configured for it. And on some of my devices there is a little key combo magic to trigger debug mode where uart0 is routed to PF0..
MadSpark- has quit [Ping timeout: 264 seconds]
MadSpark has joined #linux-sunxi
<oliv3r>
ah, nice
<hno>
but key combo is boot1, not boot0.
<oliv3r>
well the kernel wont boot so boot1 uboot will be enough
tkoskine has quit [Ping timeout: 264 seconds]
hark has quit [Ping timeout: 264 seconds]
<oliv3r>
didnt know they patched script.bin into uboot
tkoskine has joined #linux-sunxi
hark has joined #linux-sunxi
hramrach__ has joined #linux-sunxi
<hno>
mripard, do you have a working boot0 binary?
<mripard>
hno: yes, both boot0 and boot1
<mripard>
and those I can send them to you right row
<hno>
Ok.
<hno>
arm-linux-gnu-objdump -D -m arm -b binary ...
<hno>
and look for somethig like
<hno>
168: ee110f30 mrc 15, 0, r0, cr1, cr0, {1}
<hno>
16c: e3800040 orr r0, r0, #64 ; 0x40
<hno>
170: ee010f30 mcr 15, 0, r0, cr1, cr0, {1}
<mripard>
just pushed it to the build server.
<hno>
Ok. Will look at them later. Need to get some sleep now.
<derethor>
good! my first custom buildroot image boots!!!
<derethor>
lots of fun
dragonn has quit [Ping timeout: 276 seconds]
<hno>
mripard, oliv3r http://ur1.ca/ekonw should read the register. Dump it into memory with the help of mm.l 0x51000000, then go 0x51000000 to execute, and md.l 0x50000000 to dump out the register value.