**** BEGIN LOGGING AT Fri Jun 04 02:59:57 2010 Jun 04 05:04:18 shr-u connects to wifi easy Jun 04 05:17:07 slaxxer: with iliwi? Jun 04 05:17:21 yup Jun 04 05:23:26 it rocks... Jun 04 05:23:35 super fast Jun 04 05:42:40 qtmoko is very fast Jun 04 11:04:02 JaMa|Wrk: hi, http://www.bsdmn.com/openmoko/kernel/touchscreen_ignoreunexpectedintr_34.patch Jun 04 11:05:45 JaMa|Wrk: proper fix for touchscreen issue Jun 04 11:05:53 gena2x: great! will try today Jun 04 11:06:01 gena2x: will you attach it to that bug? Jun 04 11:07:37 JaMa|Wrk: i'll prepare&test similar fix for .29 and publish both Jun 04 11:07:53 JaMa|Wrk: today i hope Jun 04 11:08:02 thanks a lot Jun 04 11:08:30 it's my first proper patch (i hope) to kernel Jun 04 11:08:54 i'll be happy if someone will use it Jun 04 11:09:26 i mean wifi patch is more workaround than fix Jun 04 11:09:58 so this is like a small archievent :) Jun 04 11:12:21 :) if the same issue is also on other devices, then after some testing in linux-openmoko, you should send it to upstream Jun 04 11:14:36 i thinks issue is possible on other devices, as where seems race as a minimum. 'seems' because it's unspecified in docs how hardware works in certain situation. Jun 04 11:15:12 :) Jun 04 12:10:07 JaMa|Wrk: found how to make patch to look a bit better, so i finally be a bit different. Jun 04 12:11:03 JaMa|Wrk: (move global var to global struct)... Jun 04 12:16:26 gena2x: hi, and dont use C++ style comments :) Jun 04 12:17:14 gena2x: /* IIRC this is only allowed comment style in kernel ;-) */ Jun 04 12:19:31 radekp: hi, and thanks. i've already fixed one such case but missed other... damn c++ custom :) Jun 04 16:14:08 cando2 Jun 04 19:00:16 JaMa|Wrk: final patch posted along with description Jun 04 19:00:59 radekp: ^^ Jun 04 19:01:18 for both kernel variants Jun 04 19:02:08 spend much time with .29 case something were changed in debian and kernels with my previous config were unable to find sd card Jun 04 19:02:38 no -O2 kernel basically works here Jun 04 19:02:43 s/no/now/ Jun 04 19:02:44 gena2x meant: now -O2 kernel basically works here Jun 04 19:03:50 tomorrow i'll depart for several days from internet so if something need related to patch, try it today. Jun 04 20:14:39 hows the debian install.sh working today? Jun 04 20:38:32 gena2x: your kernel config...are needed drivers compiled in? Jun 04 20:38:54 ie: usb networking Jun 04 21:04:55 undrwater: what do you mean? Jun 04 21:05:18 undrwater: i mean which kernel are you speaking about? Jun 04 21:06:25 undrwater: if about kernel from optimization effort, it has usb networking as a module Jun 04 21:06:46 <[Rui]> gena2x: no WS for me, so far, with your kernel Jun 04 21:07:06 <[Rui]> hi all Jun 04 21:07:27 [Rui]: i told it is as almost stable as ordinary Jun 04 21:07:48 <[Rui]> gena2x: yeah, but I'm just reporting the progress so far Jun 04 21:08:40 [Rui]: oh, no need now, i need that in January :) just enjoy faster system :) Jun 04 21:08:49 <[Rui]> gena2x: lol Jun 04 21:09:10 [Rui]: and don't forget to report bugs :) Jun 04 21:12:09 undrwater: but in general it's easy to check by zcat /proc/config.gz|grep CDCETHER Jun 04 21:12:14 i think this one Jun 04 21:13:37 [Rui]: but is shr now feels much better, like in .32? ;) Jun 04 21:13:46 <[Rui]> gena2x: not yet like in .32 Jun 04 21:14:06 <[Rui]> gena2x: I notice it quite a lot in elmdentica, my identi.ca/twitter client Jun 04 21:14:14 [Rui]: i noticed some huge speedup of sd card subsystem Jun 04 21:14:27 <[Rui]> gena2x: as I'm not yet using ecore_con_url, I am using libcurl in a blocking way Jun 04 21:14:35 <[Rui]> and fetching the pages is MUCH, MUCH faster Jun 04 21:14:47 <[Rui]> also a lot noticeable under gprs Jun 04 21:15:48 i gone crazy than i first noticed that is reason of several years of speking about slowness of neo :) Jun 04 21:16:46 <[Rui]> gena2x: when I tried .32 and .34 I noticed the phone wasn't dragging anylonger Jun 04 21:16:54 <[Rui]> I mean, it does a little but... Jun 04 21:17:01 <[Rui]> it's quite bearable Jun 04 21:17:25 <[Rui]> while 2.6.29 is nigh-unbearable :) Jun 04 21:19:04 in fact speed of any non-debug kernel is close to expected from 400mhz cpu Jun 04 21:19:38 but sd speed might be much better. Jun 04 21:20:45 nand speed can be much better (now we are calculating checksum in software for each read/write, while our hw can do that for us) Jun 04 21:21:36 someone also IMO should review memcpy speed. we have x10 times slower than N800 Jun 04 21:22:17 so would be nice to he how it should be thoretically and is current speed is expected Jun 04 21:23:03 That's quite involved. From memory, the memory bus on the FR is lots slower Jun 04 21:24:36 i think need to just check specs for our 2442b. Jun 04 21:25:15 i'm not expert in ram but i always like to learn, and best learn is real, live problem Jun 04 21:25:34 yes. Jun 04 21:26:12 that's why like my freerunner more and more :) Jun 04 21:33:58 i think than om needs stable kernel no matter will it be .29, .32. or .34 or which compiler it built, and wonder much why nobody applied patch 29.rc3 to 29.final Jun 04 21:37:53 32/34 have better ubifs support, what can't be ignored. also btrfs more gadget devices.. Jun 04 21:38:27 Because the number of devs on the project is small, and they are working on stuff they know. Jun 04 21:39:23 right about ubifs. but about nand speed i think first thing should be hardware ecc, this should give big speedup in both read and write. Jun 04 21:39:38 yeah. Jun 04 21:39:42 ubifs is good. Jun 04 21:39:52 There are some other areas which would be perhaps hepful. Jun 04 21:40:12 For example - as I understand it - swap is not in fact sane for swap. Jun 04 21:40:16 oh, i thought too few than considering rc3->final Jun 04 21:40:32 i am not right. it's better to use .28 for stable kernel Jun 04 21:41:30 In that it does not completely optimise - or even partially - for linear swap. Jun 04 21:41:43 I note also: Jun 04 21:43:04 Linux Nokia-N900-02-8 2.6.28-omap1 #1 PREEMPT Thu Apr 15 09:47:09 EEST 2010 armv7l unknown Jun 04 21:43:34 yeah, nothing wrong with using stable kernel Jun 04 21:44:03 i didn't understood about swap in fact Jun 04 21:44:23 To recap. Jun 04 21:44:32 Flash has eraseblocks of around 140K. Jun 04 21:44:45 nand? or sd? Jun 04 21:45:20 ok, consider sd. but we can't directly access it Jun 04 21:45:24 If you do small block writes, that can often involve reading a 140K block, erasing it, and writing back 140K with 4K of your new data Jun 04 21:45:28 they are both the same Jun 04 21:45:39 they are very different Jun 04 21:45:42 it's just that SD has a controller that pretends it's a block device. Jun 04 21:45:51 They use the same underlying chips. Jun 04 21:46:04 yes, but still different for us Jun 04 21:46:17 The SD controller is not particularly smart. Small block write performance - generally - blows - in both cases. Jun 04 21:46:42 As in blows to the extent of hitting a few tens of kbytes/s write speed for small random blocks Jun 04 21:46:46 As can occur with swap Jun 04 21:47:23 i to read from nand you have to read whole block too? Jun 04 21:47:32 s/i/i think/ Jun 04 21:47:32 gena2x meant: i think to read from nand you have to read whole block too? Jun 04 21:47:47 No. Jun 04 21:47:57 ok, anyway i got what you wanted to say. Jun 04 21:48:08 http://www.mail-archive.com/linux-kernel@vger.kernel.org/msg170028.html is somewhat relevant Jun 04 21:49:06 oh, interesting read Jun 04 21:58:20 reading on nand can be performed only in pages too i thing Jun 04 21:58:26 s/thing/think/ Jun 04 21:58:26 gena2x meant: reading on nand can be performed only in pages too i think Jun 04 21:59:20 onyway i understood you concerns about swap Jun 04 21:59:21 the eraseblocks are of similar size. Jun 04 21:59:36 From memory you can read blocks at I think 512 byte boundaries Jun 04 21:59:42 it's beena while since I read it Jun 04 21:59:42 i am not using swap :) Jun 04 22:00:29 on neo, and only using it because of tmpfs on other systems Jun 04 22:00:41 if i don't want tmpfs i disable swap Jun 04 22:00:48 ah Jun 04 22:01:34 no need in fact, as something will go in swap while main memory will be used for file buffering Jun 04 22:02:51 and i hate swap because of possibility to go too deep in it because of some bug. Jun 04 22:03:05 then someone starts allocating more and more. Jun 04 22:03:39 this way of system death is hate so much that i am ready to disable swap. Jun 04 22:03:42 yes, that can be an issue. Jun 04 22:03:50 When it works well, it's good. Jun 04 22:04:18 The right way is clearly to write the applications that are normally resident so that the inactive parts are not there. Jun 04 22:04:41 But... Jun 04 22:05:10 i think right way for applications is not bother. much better to keep code easy. Jun 04 22:05:39 and if you sure you need huge ram - just allocate and map some file Jun 04 22:05:53 :) Jun 04 22:06:38 something like this... Jun 04 22:08:55 oh, have to go to sleep, bye Jun 04 22:09:02 <[Rui]> gena2x: good night! Jun 04 22:57:52 hey guys just wondering would it be hard to convert a free runner from Europe setup to north amircan setup ? Jun 04 23:00:29 Very hard. Jun 04 23:00:38 How competant are you at dense SMT rework Jun 04 23:01:23 my brain would tell me removing the GSM modem and putting a US version in would do the trick, but i suppect it build in Jun 04 23:01:45 still wondering why the freerunner is the only phone where you need a different model in europe then the US.... Jun 04 23:03:02 No, it's not. Jun 04 23:03:05 Many models do that Jun 04 23:03:09 there is no GSM modem Jun 04 23:03:16 it's integrated onto the motherboard Jun 04 23:03:33 Only quadband GSM works 'worldwide' (sort-of) Jun 04 23:03:38 triband are all US or EU Jun 04 23:04:13 You need to - as I understnad it - swap several filter components that may not be easily available, and reprogram the modem firmware in a non-trivial manner. Jun 04 23:04:54 still don't get why a US iphone for example works in europe but a free runner gives so much problems apperantly.. Jun 04 23:05:38 or for that mather my any HTC phone i ever touched, they all work in the US. and there europian build .. Jun 04 23:07:28 so i really don't see why a freerunner could not be build like that :\ but i must be missing some piece of information somewhere Jun 04 23:07:40 At the time it was designed, there wasn't an easily available quadband chipset. Jun 04 23:08:07 Mobile phone chipsets are not easily available. Jun 04 23:08:22 Most makers are not interested in talking about small orders. Jun 04 23:08:27 (~10K) Jun 04 23:09:59 SpeedEvil: so i guess its a design limit ( for me its rather problematic to say the least ) Jun 04 23:10:18 yes. Jun 04 23:10:42 Sorry. Jun 04 23:13:28 hhm i guess there are 2 roads i can travel 1 get my hands on the design for a better gsm modem and build it into a freerunner ( most likely i'm way out of my league on that 1 ), or get 2 phones :\ both seems less the perfect or alternatively i may be able to find a US OR europe phone provider that works on the 1900 Mhz band that should in theory get me out of the woods ... Jun 04 23:13:36 so thats 3 roads Jun 04 23:14:38 where are you wanting it to work in europe? Jun 04 23:15:01 You may be able to find a provider with spectrum only in a band that works with it. Jun 04 23:24:00 SpeedEvil: i wish it was easyer like software easy or connect a few wires or replace a single chip easy but i guess i'm not that lucky Jun 05 00:32:08 Deathzor: It possibly is replace a couple of components. Jun 05 00:32:39 Deathzor: The problem is that the components involved are likely 1mm*2mm, and are surrounded closely by lots of 1*0.5mm components Jun 05 00:39:58 SpeedEvil: i read into that sure it can be done but your most likely not gonna do it ;) Jun 05 00:40:44 Basically **** ENDING LOGGING AT Sat Jun 05 02:59:57 2010