<hno>
you should be able to add plugins to the provided nginx
<mnemoc>
recompiling :p
<mnemoc>
hno: do you have a suggested tutorial to get that thing done with squid?
<mnemoc>
not fond to have build-your-own things on servers
<mnemoc>
was hoping to find a reverse prodxy dedicated to throttling and DoS handling
egbert_ is now known as egbert
<hno>
In squid delay pools it's done using a large pool size and slowish refill rate.
* mnemoc
fears the big cephalopods
<hno>
There is some commercial ones which is have that focus, but those only make sense in multi-gigabit use cases.
<hno>
To handle a DDoS requires you to have a significant amount of bandwidth available to swallow the attack.
<hno>
seems our CONFIG_ dependencies is not quite in shape.
<Yaku>
that new usb implementation is parallel to otg or replace otg ?
<mnemoc>
new usb implementation?
<Yaku>
that new protocol you´ve been talking about before
<Yaku>
well i was interested in using my arduino more together with android in the future too
<hno>
Yaku, it's not using OTG.
<hno>
expects that the pheriperial that you connect to the Android phone/whatever acts as an USB HOST and queries a hardcoded USB vendor+device ID on the Android.
<hno>
the perpherial must also power the USB bus.
<hno>
so the Android with all it's smart CPU power is reduced to act as an USB device and might even charge from the pheriperial.
<gzamboni>
mnemoc i can set up a mirror if you want. i have a 100mbps server with unlimited traffic
<mnemoc>
gzamboni: a mirror of the wiki doesn't make much sense :p
<gzamboni>
i thought the bandwitdth came from the dl.linux-sunxi.org
<mnemoc>
I don't see much activity in the access logs of dl.l-s.o nor dl.cb.o... but not measuring bw
<mnemoc>
please mail me your ssh public key so you can get rsync
<gzamboni>
k
<mnemoc>
so much stuff to do and so little time :(
<hno>
mnemoc & gzamboni, doesn't it make more sense with a cache than a mirror?
<mnemoc>
dl. only has stitic files
<mnemoc>
static*
<mnemoc>
the wiki/php part already has caching, but that consumes the same bw
<hno>
Hmm.. no insmod/modprobe in my Debian image.
<mnemoc>
o_O
<gzamboni>
we can do a dns round robin an i can cache the wiki
* mnemoc
wonders if mediawiki will like that
<mnemoc>
gzamboni: let's try caching dl. first
<mnemoc>
where is your server?
<mnemoc>
.fr?
<gzamboni>
yes
<gzamboni>
i have other services on it, but i dont use a lot of the bandwidth
<gzamboni>
i just sent you the pub key
<mnemoc>
preferred username? gzamboni?
<gzamboni>
:) yes
<mnemoc>
wow, dss. years without seeing one of those
<mnemoc>
14GB between 9:00-10:00...
<rm>
<gzamboni> we can do a dns round robin an i can cache the wiki
<rm>
so it's a matter of luck which copy you will edit!
<rm>
sounds pretty exciting
<mnemoc>
POST pass-throw :)
<gzamboni>
yeah, edit doesnt use an specific hostname
<gzamboni>
wont work
<mnemoc>
dl first, it's simpler :p
<gzamboni>
ok
<mnemoc>
gzamboni: try to ssh to linux-sunxi.org
<mnemoc>
will remove dl. from https now
<gzamboni>
one sec
<mnemoc>
but hno is probably right about doing a caching proxy instead of a mirror for dl.
tzafrir has quit [Ping timeout: 245 seconds]
<gzamboni>
i sent you the worng key, i will send you a new one, i'm sorry
<mnemoc>
gzamboni: paste it on /q
<mnemoc>
or mail, both will work for me
<gzamboni>
just mailed it
<gzamboni>
the other one was an old one i dont use anymore
<gzamboni>
still dss :)
<mnemoc>
try
<mnemoc>
/srv/http/sunxi/linux-sunxi.org/dl and /srv/http/tom/cubieboard.org/dl
<gzamboni>
not working :(
<hno>
caching requires no preparations outside DNS and avoids downloading crap which no one ever visits.
<hno>
well, a little configuration of the cache obviously, but other than that.
<mnemoc>
+1
<gzamboni>
i never did it
<hno>
gzamboni, want me to set up a Squid for you?
<rm>
I'd say the first step should be to try understanding what's with the unusually high traffic
<mnemoc>
hno: if you have time it would be appreciated if you can setup the squid delay pool on linux-sunxi.org as well :p
<gzamboni>
hno: i would like to learn it :)
<mnemoc>
just a pass through but throtling people downloading more than 500m in an hour or so
<mnemoc>
damn. 1h of battery left :(
<rm>
for example recently we had a website randomly suspended by web hoster due to high load
<gzamboni>
mnemoc, it keeps asking for the password
<rm>
that was a great hint to check the access log
<rm>
turns out someone was bruteforcing a password to a wordpress blog
<rm>
many attempts per second; and this was what caused the high load
<mnemoc>
there is no password. let me inject my key and see
<mnemoc>
gzamboni: worked fine... so it's not the account
<gzamboni>
humm, one sec
<mnemoc>
and i doubt ubuntu's openssh doesn't support dsa by default
<mnemoc>
at least I see a dsa key in /etc/ssh
<mnemoc>
will need to move to another place to charge the laptop... but no irc from there :( ... the damn squid proxy only allows http/https out :(
<mnemoc>
gzamboni: still have 20m, so please hurry :p
<mnemoc>
normally have 6h of battery but last night forgot to plug it
<gzamboni>
6h , what laptop do you have ?
<mnemoc>
thinkpad x120e
<mnemoc>
with an e-350 (and ssd) inside
<gzamboni>
bought it without the windows license ?
<mnemoc>
nope, had to pay the ms tax
<hno>
mnemoc, freenode have web chat.
<mnemoc>
*g* ... good point
<hno>
plus I can relay an SSH for you on port 443 if you like
<mnemoc>
will squid let it pass?
<mnemoc>
well... I connect using the web chat thing once there (~30m) and we can test :)
<mnemoc>
bbl
<hno>
Squid will pass it if wrapped up in an CONNECT.
<hno>
Yes! USB ethernet dongle works on the A13, so can run apt-get natively to add missing packages..
<gzamboni>
im reading about squid, never used and it seems to be a nice solution
<gzamboni>
can it do imap caching ?
<hno>
no, squid is an HTTP proxy.
<rm>
yaay dongles rule
<hno>
now I just need to figure out what packages are missing.. no ping, no automatic loading of modules, no most things...
<rm>
hno, is it possible to add HTTPS to that example config? even if it will be without caching, just a "CONNECT"-like redirection
<hno>
rm, sure, but it needs to act as an https server then. The use of https on each side of the proxy is indenendent.
<rm>
well it could just redirect the port 443 connection to the actual server
<hno>
well, it is a proxy not a tunnel/relay.
<hno>
most people doing this also want to cache, and in many cases not even using https on the actual server.
<rm>
you may ask what's the point and why not use iptables-based redirection, and the point would be to have these connections also recorded in the same Squid access log
<rm>
ok
<hno>
rm, it would only be able to record source IP + start time of session + total amount of traffic.
<gzamboni>
i supose the squid-cache works only for GET requests, right ? how it will deal with post and GET querystrings when they are used to update info in the database by php for instance ?
<hno>
gzamboni, if the web application is well behaved then it will work just fine.
<hno>
for PHP you usually have to think a bit to at all allow Squid to cache. Default in PHP is "nothing can be cached ever by anyone"
<mnemoc_webchat>
hno: any easy way to test if I can do something other than real https over this http proxy?
<hno>
mnemoc, I have an SSH on home.hno.se:443
<gzamboni>
brb lunch is ready :)
<hno>
mnemoc_webchat, if the proxy do not require authentication then "telnet ip.of.proxy 3128; CONNECT home.hno.se:443 HTTP/1.0<enter><enter>"
<gzamboni>
mnemoc_webchat, if you can create an hostname A record to 37.59.53.179 like dl2.linux-sunxi.org
<gzamboni>
for the moment i will just rsync it and set it up on apache
<mnemoc_webchat>
HTTP/1.0 200 Connection established SSH-2.0-OpenSSH_5.8
<mnemoc_webchat>
I suppose that's a good sign :)
<hno>
it is.
<RaYmAn>
sslh is pretty good for those cases too if you only have one ip. It lets you share e.g. port 443 between https, ssh and some others: http://www.rutschle.net/tech/sslh.shtml
<rellla>
whats the best email address to get in contact with allwinner? service@...?
<RaYmAn>
it's also nice cause if people go to the ip (e.g. after seing in logs), they'll get a valid website :P
<hno>
rellla, not sure there is a best one..
<rellla>
hno: now that i know, how eva looks like, maybe she would be the one to write at...
* rellla
wants to make them know of the linux cedarx-lib bug.
<hno>
Yes, Eva is the primary contact. Official channel always goes via her.
<hno>
RaYmAn, thanks, looks handy. Wonder what logics they fingerprint SSH clients by.. silence?
<hno>
yep, silence.
tzafrir has joined #linux-sunxi
<RaYmAn>
:)
<rellla>
thanks hno, and her email is?
<hno>
Don't remember. Have given up on trying to email her.
<rellla>
thats the problem of blacklisting :P
eebrah has joined #linux-sunxi
<rm>
thanks hno, set up one of my VPSes in Germany as a caching reverse proxy in front of my home web server (not used for much though) :)
<rm>
actually I use Squid there as my 'forward' proxy
<rm>
so it was only a matter of adding some more config lines
<hno>
rm, good. Feel free to ask if needed.
<hno>
annoying.. rtl8192cu module seems to almos twork. It finds the card, configures it, and even associates with the access point, but seems to not actually transmit any packets.
<mnemoc_webchat>
hno: /q
<rm>
not sure what's up with rtl8192cu. I think for me it did transmit packets, even
<rm>
but locked up soon
<rm>
and/or failed to re-associate
<lunra>
Sorry if this sounds obvious, but have you guys tried turning powersave off? http://linux-sunxi.org/Wifi
<rm>
that's for 8192cu
<rm>
it works fine with powersave off
<rm>
but what we were talking about is rtl8192cu, a different driver
<hno>
8192cu works for me, but lacks many features I need.
<hno>
an effect of not really being an Linux wifi driver.
<lunra>
rtl8192cu and 8192cu are not the same thing? Or am I misreading?
<Dreadlish>
probably yes
<hno>
It's two different drivers for the same chipsets.
<hno>
rtl8192cu is the mainline driver. 8192cu is from Realtek.
<lunra>
Ah
<eebrah>
hno: the mainline driver has more features than the vendors driver?
Yaku has quit []
mnemoc_webchat has quit [Ping timeout: 245 seconds]
Turl|web has joined #linux-sunxi
Yaku has joined #linux-sunxi
<Turl|web>
hi mripard_
<hno>
eebrah, yes, as is always the case.
<hno>
gah.. the version numbers of the modified hostapd that realtek ship do not exists.
<mripard_>
Turl|web: hi
<Turl|web>
mripard_: your clock patches look good
<Turl|web>
mripard_: are they 3.10 or 3.11 material?
<mripard_>
Turl|web: I don't know, it will depend on what mike thinks
<mripard_>
It would be best to have it in 3.10, but 3.10 works well on the A13 without it, so...
<Turl|web>
yeah, it's a subset of the gates
<mripard_>
yes
vakkov has joined #linux-sunxi
cajg has joined #linux-sunxi
Turl has joined #linux-sunxi
rz2k has joined #linux-sunxi
Turl|web has quit [Quit: Page closed]
<gzamboni>
Is it possible to compile the linux kernel modules directly to current used filesystem lib dir ? i was thinking in compiling modules directly from the arm a10 to its own debian filesystem /lib
<gzamboni>
so no need to crosscompile, remove sdcard put at the development pc overwrite, remove, reput it to the cubieboard to try the modifications out
<gzamboni>
or maybe crosscompile it and overwrite the current modules from the lib dir using ssh, so no need to always remove/put the sdcard
<Dreadlish>
ech
<Dreadlish>
just get gcc for arm
<Dreadlish>
get kernel headers
<Dreadlish>
and compile it
<Dreadlish>
it isn't any problem...
<Dreadlish>
it can take years, but should work...
<gzamboni>
mnemoc rsync form the linuxsunxi dl: sent 84098 bytes received 31241031217 bytes 8667253.52 bytes/sec
Yaku has quit []
<gzamboni>
i will crosscompile and overwrite it using ssh, it will be faster, i just wanted to know if i can touch the modules files on the lib dir when the system kernel is runing
<shineworld>
so, what about open accessory ? Is a doom for me at this point...
tinti has quit [Ping timeout: 256 seconds]
<shineworld>
Seller don't would to stop my order so I've purchased 100 tablets which couldn't be used .... sigh
<shineworld>
I don't see so much differences in android sources between AOSP and openbox in usb related code
n01 has joined #linux-sunxi
<shineworld>
my guilty... rule : try a piece compatibility before to purchase it
<shineworld>
;|
vakkov has quit [Ping timeout: 268 seconds]
n01 has quit [Quit: leaving]
Yaku has joined #linux-sunxi
vakkov has joined #linux-sunxi
n01 has joined #linux-sunxi
tinti has joined #linux-sunxi
Yaku has quit []
tinti has quit [Remote host closed the connection]
<mnemoc>
gzamboni: cubieboard's dl is much fatter
vakkov has quit [Quit: Leaving]
Yaku has joined #linux-sunxi
<mnemoc>
gzamboni: added dl-fr1.D instead of a dl2.D
<mnemoc>
gzamboni: will turn the current into dl-de1.D and set dl.D into two CNAMEs
<mnemoc>
gzamboni: but try to see the alternative of setting up a caching proxy
<mnemoc>
to avoid syncronization issues
<hno>
mnemoc, you can't have two CNAMEs.. only A
<mnemoc>
:(
<mnemoc>
damn
<hno>
CNAME means "this entry is really named..."
<hno>
can't be combined with anything else, including CNAME. Must be alone.
<mnemoc>
so dl needs the set of A and AAAAs directly
<hno>
yes.
<mnemoc>
in which case I wonder if it makes sense to create the extra dl-foo.
<hno>
it's not hard to automate keeping them in sync.
<mnemoc>
maybe if someone explicitly wants one server and not a random (rr)
<hno>
yes.
<hno>
if using bind then put the address records in a separate file and include in both labels.
<mnemoc>
using a free dns hosting (pointhq.com), not my own... so I'm limited to the annoyances of a web interface
<mnemoc>
but we won't be adding mirrors daily anyway
<hno>
indeed. Only gets slighly annoying when changing one of the mirrors..
<gzamboni>
mnemoc, im trying out squid3 here
<mnemoc>
cool
hozer has quit [Ping timeout: 268 seconds]
<gzamboni>
if i pass 5tb/month they will reduce it to 10mbps, but i dont think it will happend
<mnemoc>
same case for me
<mnemoc>
but I'm getting worried with these 10+GB/h notifications
<gzamboni>
yes, lets load balance it
<mnemoc>
i we *really* become popular enough to justify >10TB/M I have no problem in paying the $ for extra TBs at 100Mbps, but not if it's because of abusers
<hno>
mnemoc, so need to dig a bt in the logs to see what is going on.
<mnemoc>
yes
<hno>
hmm.. wonder if this massive update of rtlwifi will work. Merged all that could be merged reasonably clean from 3.9
<mnemoc>
but first I need to use the little spare time to move cubieboard.org to the new server (to return the old and save the $$), and then I'll install squid in front to handle the abusers
<hno>
it did work.
<mnemoc>
into 3.4?
<hno>
yes. But I only merged commits touching rtlwifi so likely broke a number of other drivers.
<mnemoc>
*g*
<mnemoc>
isn't there maintained rtlwifi tree out there somewhere?
<hno>
mainline
<mnemoc>
:(
<mnemoc>
hope for a consistent but fixed 3.4 tree from rtlwifi people
<mnemoc>
hoped*
<hno>
can try to bisect which of the changes made a difference.
<hno>
there is only 54 commits.
<mnemoc>
:(
* mnemoc
needs to finish forward porting the rtl8188eu driver from stage/3.0 to stage/3.4 ...
<hno>
who is using RTL8188EU?
<mnemoc>
too many people :(
<hno>
I guess it's the successor to RTL8188CUS. Pretty much the same specifications.
<mnemoc>
the rtl drivers are a mess
tinti has joined #linux-sunxi
<hno>
Any drivers which originates out-of-tree is per definition a mess. Mess to maintain, generally broken, and years behind.
<mnemoc>
some of the rtl monster drivers even have filenames in chinese :\
<hno>
mainline rtlwifi do not even mention 8188EU
zub has quit [Ping timeout: 256 seconds]
<mnemoc>
14GB 15:00-16:00... meh
<mnemoc>
we had under 200M/h until two days ago
zub has joined #linux-sunxi
<mnemoc>
but $work$ first :|
<gzamboni>
dl-fr1.linux-sunxi.org:81 i did put at port 81 for testing it seems like its redirecting
<gzamboni>
is it normal ?
<gzamboni>
what is strange is that i configured to proxy dl.linux-sunxi.org and not www.linux-sunxi.org
<gzamboni>
i cleared the cache in firefox and its fine now
<mnemoc>
always test with curl. browsers are evil
<gzamboni>
;)
<gzamboni>
i have to set some resources limits and put it to port 80 now
<gzamboni>
thanks for the help
<Turl>
download mirror? nice :)
<shineworld>
At lest 04-30 13:54:23 I/20080411: android.app.cts.SystemFeaturesTest#testUsbAccessory PASS
<shineworld>
*least
torqu3e has quit [Ping timeout: 268 seconds]
gzamboni has quit [Read error: Connection reset by peer]
gzamboni has joined #linux-sunxi
* hno
tests with telnet
<hno>
shineworld, what made it work?
<shineworld>
ah just executed CTS test suite but I've already checked that first part of connection with device (FTDI FT311D) works because I recover Manufacturer, model, description, version, etc...
_BJFreeman has joined #linux-sunxi
<shineworld>
unfortunately after that nothing work and connection is not established
hozer has joined #linux-sunxi
<shineworld>
I'm not an expert in that so I'm trying to understand
<hozer>
is there a recommended/clean way to have multiple kernels to boot?
<shineworld>
- checking the AOSP code vs openbox code of
<shineworld>
- and after trying to understand if kernel code and usb drivers are ok
<shineworld>
but last is very hard for me
<shineworld>
unfortunately in AOA doc is only write that AOA is supported by all devices with 3.1+ (honeycomb)
BJFreeman has quit [Ping timeout: 255 seconds]
<shineworld>
so I've purchased a lot of tablets A13 powered, with 4.0.4 and kernel 3.0.8+ thinking that I can connect the FTDI USB to UART converter to it
<shineworld>
but doesn't work....
<shineworld>
same thing for cubieboard
<shineworld>
probably is a kernel problem or missing
<shineworld>
sometime when I attach the USB to FTDI FT311D I lost also adb connection (redirected to TCP with adb tcpip + adb connect x.x.x.x
<shineworld>
something goes very bad
<hozer>
interesting
<hozer>
I seem to have data corruption on my SDcard
<hno>
hozer, did get massive data corruption when I tried using discard mount option on one of my cards.
<hozer>
it's not so much massive as if I do a lot of IO random single bits seem to get flipped
<hno>
shineworld, is FT311D an AOA device?
<shineworld>
of course
<hozer>
I suppose since I want the filesystem on nand anyway and not SD ..
* hozer
debootstraps
<shineworld>
already connected with success to Asus Nexus 7 and Samsung Galaxy Advance S with back-port library (because sgas is 2.3.6)
<hozer>
oh, and what is the best filesystem to use on NAND? Is ext4 going to have flash wearout issues with journaling?
<xenoxaos>
well, ext4 is used on cell phones
<xenoxaos>
and iirc the nand driver for sunxi does some wear leveling
<xenoxaos>
i could be wrong though
<hno>
yes the Allwinner block emualtion layer does wear leveling to some extent.
<hozer>
so, why didn't they just use /dev/mtd ?
<hozer>
NIH syndrome?
<hno>
shineworld, There seems to be support in our 3.4 kernel. No idea if it works or even compiles (haven't tried)
<shineworld>
could 3.4 be used with android openbox ?
<shineworld>
just yes/no then I will try
<hno>
the kernel part is in the android usb gadget driver, with the open accessory part in f_accessory.c
<hno>
I can't answer. Not following Android developments. Others here know Android far better.
<hno>
Turl maybe?
<Turl>
what's up?
* Turl
reads
<Turl>
what do you guys want to know?
<hno>
mnemoc, pushed the rtlwifi changes to my github repo.
<hno>
sunxi-3.4-rtlwifi branch
<shineworld>
I've to use Android Open Accessory Mode to connect AllWinner tablets (and cubieboard based prototype) to FDTI FT311D chip which is an USB to UART/IO/PWM chip
mdfe has joined #linux-sunxi
<Turl>
shineworld: iirc the usb device needs to act as host and implement some protocol for android
<hno>
shineworld, you got it the wrong way around. You connect the cubieboard to the FTDI chip...
<hno>
Turl, yes. The question is about the status of our kernels for Android usage. The kernel driver part is in the Google Android gadget driver.
<shineworld>
with some minimal modifications
<Turl>
do we have f_accesory? is it enabled?
<hno>
Only looked in stage/sunxi-3.4 and it's there, unconditionally built if the gadget driver is built (same driver also used by adb)
rz2k has quit []
vicenteH has quit [Ping timeout: 240 seconds]
<Turl>
then it should work I guess
<shineworld>
I've only followed what wrote in that page : http://www.ftdichip.com/Android.htm - With Nexus 7, SGSA and SG S2 (CyanogenMod 10.1) worked fine at first try
<Turl>
what problem do you have with sunxi?
<shineworld>
ok so I will download and build kernel from github.com/linux-sunxi for 3.4
<Turl>
it should work on 3.0 too
<shineworld>
work only first part and often crash adb (I'm using adb tcpip 5555 - adb connect to device ip)
<shineworld>
I can pastebin dmesg and logcat
<Turl>
paste them please
<shineworld>
just 5 min to power up the system
<shineworld>
and I will add also N7 logs to compare
user_2 has joined #linux-sunxi
<techn_>
shineworld: so wip/usb didn't help :(
<techn_>
?
tzafrir has quit [Read error: Operation timed out]
<shineworld>
I've got some problems to implement it
<shineworld>
So I'm back to original openbox sources + very limited fixes (only big bugs)
<shineworld>
I've passed 6 hours waiting the build phase ....
<shineworld>
however I've already downloaded the entire wip kernel
<shineworld>
so I can try just after catch the log for you
<shineworld>
techn_, At first, when I connect the FT311D EV board to cubieboard the dialog "Open UART Loopback when this USB accessory is connected?" appear. So first part of AOA support seem to works fine.
<shineworld>
techn_, after a little pop up window show "Accessory connected" and "Manufacturer, ID etc info"
<shineworld>
That is what I've said "OK, works, check the purchase of tablets"
<shineworld>
But when I try to put a data (is a loopback UART program which send entered string and get any UART data from dev board) none work and EV Board don't light the connect led.
<shineworld>
looking at logcat and dmesg at this point there are errors
<techn_>
:(
<shineworld>
a sec for capture
<techn_>
but it's better than earlier?
<shineworld>
no same result
<shineworld>
now catch the log, send to dropbox and after I try to use wip kernel (if compile right)
<shineworld>
move that to android is enough fast because I'm running on uSD
<shineworld>
what take me crazy is that after (not ever) adbd seem don't work or my device result often 'offline' at adb devices
<shineworld>
like now....
<shineworld>
every time I unconnect usb from FT311D ev board to connect to computer android os THINK to have the same Accessory Board and reopen the signed APP and show manufacturer info match...
<shineworld>
Is crazy
<shineworld>
reboot
<shineworld>
I need to switch adb to tcpip because after dev board connection/disconnection cubieboard OTG USB don't connect to ADBD
<techn_>
or is it becouse this line: E/ProductConfig( 913): cubieboard not have framework config class
<shineworld>
this is only a useless message because is used MELE configuration
<shineworld>
in ProductConfig
<shineworld>
I've to add cubieboardSpec but before I need to contact matson
<shineworld>
I've tried in cubieforum but nothing at moment
<shineworld>
I've other openbox issues solved to submit to github (that is closed at moment)
<shineworld>
I've to choice : fork the git or have possibility to submit patch
<shineworld>
I don't like to fork
<shineworld>
is a waste of energies and code
<techn_>
but for me dmesg log looks fine.. expect that "binder: 304:304 transaction failed 29189, size 92-0" row
<techn_>
following errors cause this print: err_alloc_t_failed:err_bad_call_stack:err_empty_call_stack:err_dead_binder:err_invalid_target_handle:err_no_context_mgr_node:
<paulk-desktop>
The wiki seems unable to send mails
<paulk-desktop>
Mailer returned: Unknown error in PHP's mail() function.
<paulk-desktop>
that's to confirm the email address to edit pages
<techn_>
you could add more debug for binder, set "debug_mask" module parater to 0xffff
<shineworld>
how to set (sorry I'm a noob in kernel matter)
<shineworld>
at least in linux kernel because I write firmware by 25 years
<techn_>
I'm not sure if this works but add "binder.debug_mask=65535" to kernel command line. Like in here http://linux-sunxi.org/Display#HDMI
<shineworld>
I/InputQueue-JNI( 313): Sending finished signal for input channel '41212d10 AtchDlg:com.android.launcher/com.android.launcher2.Launcher (client)' since it is being unregistered while an input message is still in progress.
<shineworld>
I/InputQueue-JNI( 313): Ignoring finish signal on channel that is no longer registered.