IRC log for #htc-linux on 20130204

00:04.52*** part/#htc-linux Cotulla (~myfakemai@109.205.253.11)
03:14.53*** join/#htc-linux Bry8Star (~Bry8Star@gateway/tor-sasl/bry8star)
03:45.43*** join/#htc-linux rajkosto (~rajkosto@wan.rajkonet.info)
04:10.58*** join/#htc-linux bartman (~bart@2607:f2c0:f00e:700::dd)
04:58.36*** join/#htc-linux d3tul3 (~detule@unaffiliated/d3tul3)
05:16.13*** join/#htc-linux eR^zeRa` (~zzeratul@ip-84-42-202-42.net.upcbroadband.cz)
06:00.16*** join/#htc-linux DuperMan (Duper@85-250-109-40.bb.netvision.net.il)
06:16.40*** join/#htc-linux Rajko (~rajkosto@wan.rajkonet.info)
06:28.55*** join/#htc-linux rajkosto (~rajkosto@wan.rajkonet.info)
06:50.50*** join/#htc-linux kiozen (~kiozen@p578a42db.dip0.t-ipconnect.de)
07:23.10*** join/#htc-linux DuperMan (Duper@93-172-160-252.bb.netvision.net.il)
08:20.55*** join/#htc-linux lamikr (lamikr@nat/nokia/x-mwlghdedojnejdud)
09:08.29*** join/#htc-linux arif-ali (~arif-ali@81.144.163.60)
09:38.26*** join/#htc-linux Bry8Star (~Bry8Star@gateway/tor-sasl/bry8star)
10:36.35*** join/#htc-linux noobhands_ (d03032a2@gateway/web/freenode/ip.208.48.50.162)
10:43.16*** join/#htc-linux gauner1986 (~Adium@87.253.171.220)
10:43.49fakkersup gauner1986
10:45.35gauner1986hey fakker
10:50.01fakkerhows tricks?
10:51.04gauner1986currently boring myself at work
10:51.56fakkerSNAP
10:52.23styblagauner1986, you could be my twin.
10:56.26gauner1986hehe.. but today is an exception actually. ;)
10:56.50fakkermine isn't
10:57.07gauner1986hope to buy a htc m7 soon
10:57.18fakkerhmm
10:57.30fakkeryou still got the iphone?
10:57.42gauner1986yeah - company phone
10:57.50fakkeri still got mine - but i left that company
10:57.51fakkerhaha
11:08.41styblagauner1986: lucky you. I'm quite constantly bored at work.
11:08.59gauner1986styble: what's your job?
11:09.17styblagauner1986: quality assurance/sustaining.
11:09.31gauner1986software?
11:09.37styblayes
11:09.58gauner1986ah ok.. i'm sitting right next to my qa-team. ;)
11:10.08gauner1986but i'm on the developer-side luckily
11:11.40styblaI don't think that's the problem in general way.
11:13.35styblaanyway ...
11:14.18*** join/#htc-linux Rajko (~rajkosto@cable-94-189-245-214.dynamic.sbb.rs)
11:15.02*** join/#htc-linux bitrot (~rajkosto@wan.rajkonet.info)
11:53.01*** join/#htc-linux d3tul3 (~detule@pool-96-234-132-77.bltmmd.east.verizon.net)
12:41.34*** join/#htc-linux conantroutman (~conantrou@cpc5-pert5-2-0-cust768.16-4.cable.virginmedia.com)
13:03.03noobhandshttp://www.gizmodo.co.uk/2013/02/this-is-what-happens-when-your-baker-doesnt-understand-english/
13:16.33*** join/#htc-linux d3tul3 (~detule@unaffiliated/d3tul3)
14:05.00*** join/#htc-linux arif-ali (~arif-ali@81.144.163.60)
14:10.07*** join/#htc-linux arif-ali (~arif-ali@81.144.163.60)
14:13.42*** join/#htc-linux arif-ali (~arif-ali@81.144.163.60)
14:31.55*** join/#htc-linux arif-ali (~arif-ali@81.144.163.60)
15:51.23*** join/#htc-linux kiozen (~kiozen@ppp-93-104-93-124.dynamic.mnet-online.de)
17:04.11*** join/#htc-linux rudeblunt (4fa80aec@gateway/web/freenode/ip.79.168.10.236)
17:04.27rudeblunthey cockpoo?
17:04.43rudeblunt~seen Cotulla
17:04.54aptcotulla <~myfakemai@109.205.253.11> was last seen on IRC in channel #htc-linux, 1d 1h 7m 59s ago, saying: 'he is not here'.
17:04.55rudeblunt~seen Cottula
17:04.56apti haven't seen 'cottula', rudeblunt
17:05.10rudeblunt~seen Marc
17:05.10aptmarc is currently on #htc-linux (7d 1h 50m 16s). Has said a total of 119 messages. Is idling for 1d 1h 24m 50s, last said: 'I did a quick search and couldn't find it'.
17:05.21rudebluntso
17:05.35fakkerhey cockpoo
17:05.46rudebluntI've managed to boot the current 3.06 leo kernel
17:06.06Marcok
17:06.13Marcwith loading the stupid initrd?
17:06.17rudebluntnot by creating a boot.img
17:06.26rudebluntyep
17:06.40rudeblunti think I've confirmed cockpoo's theory
17:07.05rudebluntif you put initrd inside kernel
17:07.14rudebluntit actually boots
17:07.42rudebluntcomplains about not finding some stuff
17:07.54Marcwhich is normal
17:07.56jonpryuncompressed kernels require less space between kernel and initrd
17:08.34rudebluntand stays alive ... (but i think all processes spawned by init crash)
17:08.46Marchm, we are currently only using gzip
17:08.54rudebluntnot even initlogo.rle displays
17:09.23Marcthat might actually be an mdp fail
17:09.41rudebluntcare to see my dmesg?
17:09.50Marcsure
17:09.58Marcit'd be nice if you'd upload your compiled kernel, too
17:10.24MarcI was actually trying to compile cLK with a 16M offset yesterday
17:10.35Marcgot kind of stuck while creating the nbh ...
17:10.41Marcand didn't want to brick my HD2 ^^
17:12.08rudeblunthttp://pastebin.com/0MA3BBtf
17:13.31rudebluntactually all I did was uncompressing my initrd and copy it over /usr/initramfs_data.cpio
17:14.12Marcok
17:14.25Marcdid you enable blk_dev_ram?
17:14.49Marcalso: the first part of init is complaining about you having the 2.6.32 modules ^^
17:14.56rudebluntyes
17:14.59rudebluntI know
17:15.09rudebluntbut that should't be an issue
17:15.21rudeblunt(or so I think)
17:16.05Marcok, qmi_open is also failing
17:16.10Marcwho would've thought? ^^
17:18.20*** join/#htc-linux the-leviathan (~quassel@2001:470:26:484:6ef0:49ff:fee6:8dca)
17:18.49rudebluntanyway... still kinda busy
17:19.19rudebluntbut I tink now you have sort of a way to continue Marc ...
17:19.24rudeblunt*think
17:19.43rudeblunteven whitout MAGLDR2
17:19.44rudeblunt:)
17:25.03rudebluntlast but not least ... i was trying to boot a NativeSD rom
17:25.40*** join/#htc-linux a-st (~a-st@unaffiliated/a-st)
17:25.54rudebluntand kernel config is default
17:30.49*** join/#htc-linux paulk-desktop (~paulk@lib33-1-82-233-88-171.fbx.proxad.net)
17:34.51*** join/#htc-linux gauner1986 (~Adium@hmbg-4d06f88b.pool.mediaWays.net)
17:42.26*** join/#htc-linux gassed (faxed@c-76-30-161-228.hsd1.tx.comcast.net)
17:53.15*** join/#htc-linux the-leviathan (~quassel@c-82-192-226-27.customer.ggaweb.ch)
18:17.57*** join/#htc-linux Cotulla (~myfakemai@109.205.253.11)
18:18.10Cotullahello
18:20.25Marchi
18:20.29Marcand nice pics on twitter
18:20.35Cotullayea
18:20.35noobhandsfu
18:23.27Cotulladid u try with boot.img?
18:24.04Marc<rudeblunt> if you put initrd inside kernel
18:24.12Cotullawhat?
18:24.19Marc<rudeblunt> I've managed to boot the current 3.06 leo kernel
18:24.31Cotullagood
18:24.34Cotullalike I tell
18:25.11CotullaI never was cockpoo :P
18:25.15Cotullait was some another guy
18:36.16rudebluntsorry cockpoo
18:36.18rudeblunt:)
18:37.20rudebluntwell, now there is another issue
18:38.45rudeblunti tried to flash a boot image with a bigger initrd offset
18:39.19rudebluntbut MAGLDR could never find the kernel
18:39.20Cotulla?
18:39.34rudebluntremember?
18:39.50rudebluntyou said : create a boot.img and flash it?
18:40.16rudeblunti used abootimg
18:40.33rudebluntand created it
18:41.16rudebluntbut MAGLDR always thought NAND device was yaffs2 and failed with "Kernel not found" or something...
18:41.52rudebluntso I gave up and ended placing initramfs inside kernel
18:44.33Cotullahm
18:44.37Cotullabut it supports two formats
18:44.41CotullaI tested that
18:57.52rudebluntanyway... everything will be fixed in MAGLDR2 ... riiiiiiiiiiiiight?
18:58.02CotullaI increased offset to 16
18:58.05Cotullamaybe should set to 32
18:58.10Cotullait works for me
18:58.39rudebluntit must be that
18:58.40Cotullabut new region system makes not clear which config it should take by loading kernel as file from SD card
18:58.50Marcjust fyi: offset is set to 16M in cLK's main repository (not in the HD2 version though O.o)
18:59.10Cotullabut it's not important?
18:59.19Cotullaboot.img specify initrd separately
18:59.37Marcit's important if you just flash it to boot
18:59.51Marcor use the fastboot boot
18:59.52Cotullayes but ever fastboot is working with such format
19:02.01rudebluntmaybe I was doing someting wrong ... I created a boot.img and flashed it to boot partition with flash_image ...
19:04.16Cotullau need attributes for partition
19:04.28Cotullathere were 4 types
19:04.39CotullaYAFFS2 Kernel/Recovery and RAW
19:05.54rudebluntgot it ... so my bad
19:06.24CotullaMAGLDR2  doesn't include YAFFS2 support
19:06.27Cotullaonly RAW images
19:07.20rudebluntk
19:11.40jonpryCotulla, how is qdsp port?
19:15.44Cotullalooking to stuffs jonpry
19:15.53Cotullalooking for possible replacements
19:15.55Cotullafor commands
19:16.02Cotulla*HVM commands
19:17.21*** part/#htc-linux Cotulla (~myfakemai@109.205.253.11)
19:17.40*** join/#htc-linux Cotulla (~myfakemai@109.205.253.11)
19:17.53Cotullau jonpry?
19:19.49Cotullaalso fundamental problems like divide 64 TLBs to 6 fixed
19:19.53Cotullaor make it dynamic
19:20.01Cotullatracked by tlbmiss
19:20.25Cotullalike 10 TLB per thread
19:20.32Cotullaand 4 system ones
19:22.18CotullaI think I will test SW MMU in qMAGLDR before kernel
19:30.39jonpryi don't think its't that complicated
19:31.06jonprylike there is a kernel page table
19:31.19jonpryand then per process page table
19:32.13jonpryand when a core schedules a task. we set just keep track of which user pagetable is assigned to every core
19:32.26jonpryso on tlb miss
19:32.51jonpryif it's kernel region. just use kernel page tables and create a global entry in the FILO
19:33.10jonpryif not we look up what process is schedules, and create ASID specific entry
19:33.56Cotullaright
19:34.03Cotullabut u need debug that
19:34.33jonpryeven that isn't so bad
19:34.46jonprycause like the kernel will boot all the way to init with no user page table support
19:35.12jonpryso we can have ramconsole and everything working first
19:35.25Cotullaright too
19:39.33jonpryhow does ring0 protection work?
19:39.58jonpryi don't see anything in the docs that suggest how to protect memory
19:40.33Cotullawhat kind protection?
19:43.58Cotullau mean
19:44.01Cotullakernel / user mode I guess
19:44.02Cotulla?
19:45.19jonpryyeah
19:46.02Cotullahm I don't see inside TLB
19:46.05Cotullasuch entries
19:46.31Cotullamaybe
19:46.35Cotullaalso via different ASID?
19:47.22Cotullaor some special mode
19:47.25Cotullahm
19:47.38CotullaI guess it should exist
19:48.38jonprysyscall is done with a trap0
19:49.32CotullaLinux one
19:49.34Cotulla,
19:49.36Cotulla?
19:49.58jonpryuser to kernel
19:50.02Cotullayes
20:06.00jonprythere is PTE_U
20:07.14Cotulla?
20:07.55jonprypage table flag
20:08.02jonprypage accessible to user mode
20:08.30jonprythe trick is dropping into usermode
20:08.45jonpryGSM.UM looks like it has something to do with it
20:09.07jonpryer. GSR
20:09.18Cotullaoh in HVM
20:09.51Cotullawhat is GSR?
20:10.04jonpryguest status register
20:10.12Cotullahm
20:10.17Cotullasound like it fake
20:10.43jonprymust have some hardware support
20:10.47jonpryor TLB would hit
20:11.01Cotullawell it can use ASIDs as well
20:11.13Cotullaor clear TLB at switch
20:11.36jonprybut then kernel can't use global entries
20:11.56Cotullaright
20:13.40jonpryand have to flush tlb on context switch exit
20:14.16CotullaI am loking
20:27.01*** join/#htc-linux BabelO (~fcr@unaffiliated/babelo)
20:29.37jonprydoing it with ASID is going to be nasty
20:29.54jonprythe whole syscall will need to be redone
20:30.22Cotullayea
20:30.26jonprycause like trap0 will then cause a TLB miss
20:31.08jonprysomehow the miss will have to cause the syscall to happen
20:31.46jonprybut i guess that is probably the easiest thing
20:32.30Cotullalooks like it has supervisor and user mode
20:33.42jonprybut we don' have any idea how it is programmed
20:43.41CotullaI will find it I think
21:02.26*** join/#htc-linux gauner1986 (~Adium@hmbg-4d06f88b.pool.mediaWays.net)
21:14.02*** join/#htc-linux zeusk_ (dfbe5524@gateway/web/freenode/ip.223.190.85.36)
21:14.04zeusk_hi
21:14.32Cotullahey
21:15.11zeusk_it gets to init now ?
21:15.17Cotullawho
21:15.25zeusk_3.0, leo
21:15.28zeusk_linux
21:15.31Cotullalong time already
21:15.48zeusk_hm
21:23.15zeusk_jonpry, you got nexus 4 ?
21:26.39Cotullahe is busy
21:26.51Cotullahe is trying to start engine in his boat :)
21:27.31zeusk_lol
21:36.51Cotullalooks like they have a special tlb bit
21:37.03Cotullawhich is kinda super visor access flag
21:40.09zeusk_it has same modes as arm ? (svc, sys, usr etc..)
21:40.18Cotullano
21:40.49zeusk_then ?
21:45.20Cotullasvc and normal seems
21:46.24zeusk_hmm
21:51.35Cotullahis brain blows up
22:01.56jonpryi got engine going good
22:02.57jonpryhttp://www.youtube.com/watch?v=Wl2ZayrBlL4
22:04.14jonpryit's better now. turns out that was not running on all cylinders
22:05.22detulewhat the hell
22:06.26jonpry?
22:07.07detulehow much better?
22:07.22detuledamnit should've been an engineer
22:07.26jonprywell it can move the boat
22:07.33Cotullacool
22:07.36jonpryand doesn't smoke
22:07.46jonpryor leave diesel fuel all over the water
22:07.54Cotullahow many cylinders?
22:07.58jonpry4
22:08.28jonpryit's like 30kw and 400 kg
22:08.38Cotullahow many h.p?
22:08.42jonpry40
22:08.53Cotullagood
22:09.04Cotulla;)
22:09.29jonpryi might take it on the maiden voyage this week
22:09.34Cotullaintersting if trap0 and trap1 goes to same supervisor mode or not
22:09.46jonpryi think so
22:09.59jonpryand hvm drops back to quest for trap0
22:10.08jonpryer guest
22:10.16Cotullahm yes inside real handler
22:11.59Cotullaokay
22:12.11Cotullaso I think it half clear
22:12.25jonpryi'm going to try and write the TLB in pseudo code later
22:12.47jonpryi think it needs  7 ASID
22:12.53Cotullado u have extinguisher in the boat?
22:13.04jonpryand 2 permanent entries per core
22:13.15jonpryit's called the ocean
22:13.20Cotulla:O
22:13.20Cotulla:D
22:13.36Cotullawell
22:13.50Cotullawhy 7 ASIDs?
22:14.11jonprycontext switch just switches to ASID 0
22:14.36Cotullahm
22:14.41Cotullaand kernel in global
22:14.46Cotullasupervisor
22:14.47Cotullaone
22:14.52jonpryyeah
22:15.17jonpryeach thread needs a PTE_X PTE_R for the tlb handler
22:15.31jonpryand a PTE_W that the handler can store the context in
22:15.32Cotullayes
22:15.43Cotullaa fixed entry
22:15.50jonpry2
22:15.59Cotullamaybe cover kernel by one 16M entry
22:16.21jonpryno highpage?
22:16.36jonprydma, or ioremap?
22:16.41Cotullahm
22:16.46Cotullaor we can just copy tlbmiss code
22:16.49Cotullato special location
22:17.00jonpryi think so
22:17.08Cotullait can be 64K
22:17.10CotullaXRW
22:17.11Cotullatogether
22:17.17jonpryno!
22:17.17*** join/#htc-linux Mavy-bnc (~mavfree@46.249.37.173)
22:17.25jonprycan't write kernel from userland
22:17.38Cotullabut it should be switched to supervisor mode
22:17.48jonpryyeah but we can't use that
22:17.52Cotullaat tlbmiss entry
22:17.56jonpryASID won't switch on tlbmiss
22:17.57CotullaI guess?
22:18.14jonpryat least i don't think so
22:18.33Cotullawell but TLB entry with Global should be always valid
22:18.36Cotullafor any ASIDs?
22:18.42jonpryhrm. can user programs just change ASID?
22:18.56Cotulladon't think so
22:18.56jonpryi think global is not an option
22:19.01jonprythat's why 7 ASID
22:19.02CotullaI think SSR is not accessable from user mode
22:19.22jonpryhow exactly do we set something as user mode?
22:19.25Cotullaalso it looks like TLB with supervisor bit can access anything without rights
22:19.33jonpryif that can be done, then everything is different
22:19.36CotullaI should try that in MAGLDR
22:19.45Cotullahm it's good question
22:19.53CotullatrapX  user -> svc
22:20.02Cotullareturn  svc -> user
22:20.11Cotullabut the initial state . . .
22:20.44jonpryi couldn't find anything in the kernel source
22:21.29jonpryjust uses vmrte
22:21.46Cotulla__switch_to
22:21.47Cotullamaybe?
22:25.05jonprydon't think so
22:26.08jonpryi was looking at the GAS sources
22:26.23jonpryfiguring maybe they compile the HVM with it
22:26.37Cotullait supports registers
22:26.42Cotullayes
22:26.50jonpryno mention of GSR
22:26.57Cotullawhere u found GSR?
22:27.05CotullaI see that name the first time
22:27.10jonpry80-nb419-3
22:27.15Cotullaoh
22:27.19Cotullait's accessed via function
22:27.25CotullaI think it's fake
22:27.28Cotullajust variable
22:27.46Cotullathere is remark that it can be accessed directly in the feature fversions
22:27.54jonprythere must be some way to set it
22:28.04jonpryusermode that is
22:28.05Cotullait can be SW solution
22:28.26jonprysw solution to make tlb miss?
22:28.48Cotullamaybe it's possible to change SSR
22:28.53Cotullasvc -> user
22:28.59jonprywhat is SSR?
22:29.25CotullaSupervisor Status Register
22:29.31CotullaI think
22:29.39jonprywhere did you here about this?
22:30.02CotullaI read it somewhere I think
22:30.16jonprygas generates code for it?
22:30.35Cotullayes
22:30.45CotullaI am used it in qMAGLDR to get thread number for example
22:30.46Cotulla<PROTECTED>
22:30.46Cotullaget_hwthread_num:
22:30.46Cotulla<PROTECTED>
22:30.46Cotulla<PROTECTED>
22:30.46Cotulla<PROTECTED>
22:32.10jonpry<PROTECTED>
22:32.10jonpry<PROTECTED>
22:32.10jonpry<PROTECTED>
22:32.10jonpry<PROTECTED>
22:32.33Cotullathat ones I also used
22:32.41jonpryssr is register 6 on v4
22:32.41Cotullau setup TLBIDX
22:32.53Cotullaand then specify 64bit value
22:35.55CotullaTLB format is differnet between V2 and V4 AFAIK
22:36.47jonpryREG_ELR?
22:37.14jonpryhttp://pastebin.com/ECbw4eb1
22:37.26CotullaELR is Exception return address
22:37.52Cotullalike in tlbmiss handler I can get code address
22:37.55Cotullawith problem
22:38.09jonpryin v4 gsr is real
22:38.11Cotullaoh shit stid and htid
22:38.22Cotullasuper visor tid and hyper visor tid?
22:38.35jonprygsr is g1
22:46.17Cotullahm bad
22:50.28*** join/#htc-linux rajkosto (~rajkosto@wan.rajkonet.info)
22:56.19*** join/#htc-linux jonpry (~jon@2602:306:c417:8aa0:15a4:8791:5d38:cca3)
23:04.16jonprymaybe it's a property of the hypervisor
23:04.32jonpryand gsr writes work fine on v2?
23:06.17Cotullahm don't think so
23:06.21Cotullav2 is rather odl
23:06.22Cotullaold
23:06.53jonprysomehow the hypervisor is setting this usermode thing
23:07.53jonpryint reg_htid;   /* v2/v3 id was store as part of reg_ssr */
23:08.04jonpryint reg_sgp0;   /* User mode same as reg_resgp on v2/3 */
23:08.14jonpryint reg_stid;   /* Re-named TID register */
23:11.45Cotullayes
23:11.46Cotullaso
23:11.51Cotullamy code takes tid
23:11.54Cotullafrom ssr
23:11.56Cotullaon V2
23:17.28Cotullaso u need to do a lot of things via different way
23:17.30Cotullaon V4
23:17.32Cotullalooks like
23:21.22jonpryugh
23:21.32jonpryat least usermode is possible on v4
23:44.23*** join/#htc-linux arrrghhh (~arrrghhh@unaffiliated/arrrghhh)
23:50.03Cotullaon V2 too
23:50.09Cotullathere is super bit seems
23:56.23*** join/#htc-linux arrrghhh_ (~arrrghhh@unaffiliated/arrrghhh)
23:57.03jonpryyou cleared it?
23:57.46Cotullanot yet
23:57.51Cotullawill try it
23:58.00Cotullabut in qMAGLDR it's clear
23:58.47Cotullamaybe create git project
23:59.10Cotullaor too early

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.