**** BEGIN LOGGING AT Thu Jan 23 02:59:59 2014 Jan 23 07:37:55 Are all physical keys on the N900 ones the software sees? Jan 23 07:38:09 so running `xev` should show a keycode for each one? Jan 23 07:38:51 (as opposed to what happens on a lot of laptop keyboards, where the `fn` button modifies what keys are sent and is not itself seen) Jan 23 07:43:50 that depends, either side Jan 23 07:44:05 either side? Jan 23 07:44:14 yeah, on laptop as well Jan 23 07:44:45 Well, pretty much all laptops have a special fn key. Jan 23 07:44:54 on N900 all kbd keys and the volume +/- keys are genuine "keys", with keycode and all Jan 23 07:44:58 a lot of them also use it as speciial modifier the OS doesn't see. Jan 23 07:45:00 Okay Jan 23 07:45:20 the power and cam are buttons Jan 23 07:45:28 (eg, as a numpad that overlays iopjklbnm or something. Jan 23 07:45:30 ) Jan 23 07:45:53 well, that's highly depending on laptop hardware Jan 23 07:46:59 Yeah, as I said, pretty much all of them have the `fn` key .. probably most of them have at least a few such uses as modifiers for normal keys, which aren't seen by the software. Jan 23 07:47:01 but yes, many laptops process this sort of stuff in the keyboard MCU Jan 23 07:47:20 my Aspire One uses `fn` for home/end/pgup/pgdown Jan 23 07:48:05 and it's not up to the OS to figure out that fn+up = home .. the keyboard sends the normal keycode for home over its PS/2 interface. Jan 23 07:48:41 er, fn+pgup = home Jan 23 07:50:12 where it could, if it weren't for the screen brightness controls, which are independent of the OS, have just been another key that OSes treated as modifiers when configured properly. Jan 23 07:50:39 (and other similar things, like enable/disable touchpad, switch display output, ..) Jan 23 09:35:38 is this still around somewhere? http://talk.maemo.org/showthread.php?t=66713 Jan 23 09:36:10 (n900 recovery system to be loaded into ram via flasher) Jan 23 10:19:25 hi Jan 23 10:19:44 for noia 770 the datasheets: http://www.s-manuals.com/phone/nokia_n770 Jan 23 10:20:19 and much more on http://www.s-manuals.com/phone/nokia Jan 23 10:20:52 for nokia 770 the datasheets: http://www.s-manuals.com/phone/nokia_n770 Jan 23 10:21:33 so anyone is free to open it and to watch how is did it, or to hack it Jan 23 10:21:47 happy hacking Jan 23 10:21:53 :-) Jan 23 10:24:19 is there a way tro charge the battery of nokia 770 with the usb ? Jan 23 10:50:19 sshing in my n900 with key auth is not working Jan 23 10:52:31 ssh -vvv user@nokikana gives http://pastie.org/private/wrqiwlrl3ofqkabhlcp3w Jan 23 10:53:39 How did you try setting it up? Jan 23 10:55:01 im moving my stuff over to "new" device, previous stopped charging Jan 23 10:55:14 new device is freshly flashed and working Jan 23 10:55:45 installed openssh and copied /home/user/.ssh over from the old one Jan 23 10:56:13 pw authenticated login works Jan 23 10:56:40 and you can't ssh out from the new phone? Jan 23 10:56:47 (the issue isn't sshing into it?) Jan 23 10:56:47 in Jan 23 10:56:57 and i can ssh both in and out Jan 23 10:57:04 in will require you set it up again on the machine you want to connect from. Jan 23 10:57:09 only inbound key auth fails Jan 23 10:57:25 unless you copied the keys from sshd itself Jan 23 10:57:26 which will be in /etc somewhere. Jan 23 10:57:37 /etc/ssh Jan 23 10:57:57 I suspect it'll work if you just copy /etc/ssh from the old device to the new one and restart sshd. Jan 23 10:58:49 uh oh, why would i need the host key of old n900? Jan 23 10:59:06 or do you mean copying over the conf? Jan 23 11:00:19 Hmm .. dunno, maybe it shouldn't. Jan 23 11:00:37 does n900 openssh server log somewhere? Jan 23 11:00:46 it might provide some clue Jan 23 11:05:05 don't really know what was up, but setting password for user made key auth work :D Jan 23 11:05:41 if i upload a shell script in my e-mail, and i try to send it, whe trhe nokia will try to download it, it will be opened as a file text that a shell script, how to solve this problem ? Jan 23 11:09:03 what would you want it to do? (i.e. what's the problem) Jan 23 11:09:12 natsukao: you expect the email client to _execute_ the script when you open it? Jan 23 11:09:40 i upload the shell script in the email Jan 23 11:09:52 natsukao: you expect the email client to _execute_ the script when you open it? Jan 23 11:10:09 I dont think it'd be good behavior :P Jan 23 11:10:13 then i downloaded with nokia 770, but i don't see the shell script but the source code Jan 23 11:10:18 ashley: ask microsoft ;) Jan 23 11:10:27 r00t|d430, :D Jan 23 11:10:45 natsukao: with a shell script, there is no difference between the script or it's source... Jan 23 11:10:49 natsukao: shell script _is_ a text file Jan 23 11:11:03 natsukao: do you want to run it on your 770? Jan 23 11:11:19 how to start a shell script on shell ? Jan 23 11:11:25 yes, correct Jan 23 11:11:33 sh /path/to/file.sh Jan 23 11:11:46 (or chmod +x and without 'sh') Jan 23 11:11:47 thanks Jan 23 11:17:40 correct, now works Jan 23 11:17:45 thans Jan 23 11:18:42 when i uploaded the shell script the proprieties were changed from a simple file text in *.html Jan 23 11:19:56 and so chmod +x could not works, then i found this little error, i removed *.html and then i did chmod +x namefile ad pushed the key enter Jan 23 11:20:12 then sh wargames Jan 23 11:22:24 the filename (or extension) is irrelevant, you can execute it even when it's named .html Jan 23 11:22:36 as long as the content wasn't actually html-formatted Jan 23 11:22:53 it is only a shell script http://pastebin.com/3A5h2QbJ Jan 23 11:23:16 trying to execute an html file as a script usually creates lots of junk files, due to all the redirection symbols ;) Jan 23 11:24:53 but i nver saved as *.html it was changed from mail.riseup.net such as from gmail Jan 23 11:25:08 who cares, really? Jan 23 11:25:26 i got a little problem with nokia 770, i an't insert MMC card Jan 23 11:26:01 so for me was necessary to upload from the laptop and to download from the nokia 770 Jan 23 11:26:04 you mean, it's not detected when you insert it? Jan 23 11:26:18 nokia 770, can't works without mmc card Jan 23 11:26:23 why not juse use scp or obex or something? Jan 23 11:26:50 no, it is not that the problem, the fact is not that does not close Jan 23 11:27:19 and so it means, to left it 1 month to the nokia, only for this stupid thing, Jan 23 11:30:03 why is this not on maemo.org? REAL recovery mode for n900, without requiring a bootable rootfs... http://archive.is/lP093 Jan 23 13:26:30 hm. Jan 23 13:26:40 I reflashed my n900 Jan 23 13:26:51 updated, installed cssu-testing Jan 23 13:27:00 installed u-boot (first time) Jan 23 13:27:10 (u-boot-flasher package) Jan 23 13:27:25 and kernel-cssu-bootimg Jan 23 13:27:45 and when I boot my n900 and select cssu kernel (or default one) Jan 23 13:28:29 it loads the kernel, etc etc, then it says "starting kernel", and after few-fewteen (fewteen xD) seconds it turns off Jan 23 13:28:49 i can reflash only kernel from flasher-3.5, and it boots normally Jan 23 13:28:58 what did I do wrong? Jan 23 13:29:14 is updating time from gsm network supposed to work with n900? Jan 23 13:29:18 pr1.3 Jan 23 13:30:00 as far as i remember my previous n900 had sync working Jan 23 13:30:04 TTilus: never worked for me with any carrier Jan 23 13:30:14 or any phone for that matter Jan 23 13:30:34 so ntpd it is then... Jan 23 13:30:45 It works for me I think Jan 23 13:30:57 Poland, Orange Jan 23 13:31:14 (but with latest cssu-t) Jan 23 13:32:06 ok, maybe ntpd only after upgrading to cssu, which im about to do Jan 23 13:32:40 ashley: what you did wrong: you are using a system that provides no diagnostic messages during boot... Jan 23 13:55:57 r00t|d430, what can I do? is there any flag or something for "debug mode"? Jan 23 13:56:52 ashley: i don't know... Jan 23 13:57:02 ashley: serial console, good luck ;) Jan 23 14:08:28 :/ Jan 23 14:09:01 TTilus, was working for me with orange and free, not with SFR Jan 23 14:09:06 I guess u-boot is not for me :f Jan 23 14:09:14 ... depends on carrier ... some didn't respect gsm protocol Jan 23 14:37:12 Is there any alternative to u-boot? Jan 23 14:52:25 KhertanAtwork: ok, good to know Jan 23 15:57:00 freemangordon: Hello Jan 23 15:57:26 freemangordon: How SR will work if one enables VDD1 but not VDD2? Jan 23 16:29:22 hxka: VDD2 doesn't get controlled by SR :) Jan 23 16:29:35 VDD2 is L3 interconnect IIRC Jan 23 16:29:49 voltage that is Jan 23 16:31:19 So it's useless then? Jan 23 16:32:37 yay Jan 23 16:32:57 hxka: hmm? why is that? Jan 23 16:32:58 I downloaded pali's kernel pack with pr1.3 kernel with framebuffer Jan 23 16:33:08 if you enable it, the VDD2 lotage is controlled Jan 23 16:33:17 *voltage Jan 23 16:33:31 and I can see why it doesn't work with u-boot Jan 23 16:33:31 hxka: I am not sure I get you right Jan 23 16:33:32 freemangordon: I meant it's useless to enable only VDD1 Jan 23 16:34:00 no, why? VDD1 is MPU/DSP voltage Jan 23 16:34:15 but ofc it is better to have both enabled Jan 23 16:34:31 Oh, thanks Jan 23 16:35:16 Another question: is there any benefit in lowering the voltages in k-p-s profile if both VDD1 and VDD2 are enabled anyway? Jan 23 16:41:19 Hm Jan 23 16:41:50 Also, I reported earlier that my N900 is unstable on any frequency above 600 MHz. That turned out to be not true: it's only unstable if VDD2 is enabled. Jan 23 16:42:33 I think that I "hm" too much xD Jan 23 16:44:33 Maxdamantus: the back cover does have a magnet to control mounting and unmounting /media/mmc1 Jan 23 16:49:04 hxka: I believe you shouldn't adjust voltages if using SR, this will make the device unstable. Either use SR or "legacy mode" like ideal, starving etc... not both. Jan 23 16:50:27 hxka: you should not touch the voltages when using SR Jan 23 16:50:49 freemangordon: Thanks, that's what I thought Jan 23 16:51:26 hxka: BTW I did not fix SR VDD2 calibrations the unlike those for VDD1, maybe this is the reason Jan 23 16:51:59 freemangordon: Can I help somehow with this? Jan 23 16:52:10 hxka: yep, just a minute to check something Jan 23 16:52:16 Sure Jan 23 16:52:56 could you gimme the output of "cat /sys/power/efuse_vdd2" Jan 23 16:52:56 ACTION gives freemangordon Jan 23 16:53:06 Gatta_Negra: hi! Jan 23 16:53:14 how's your hal sensor? Jan 23 16:53:58 http://ix.io/a4Q <-- this is what I get when I boot a kernel from u-boot Jan 23 16:54:12 which kernel is this? Jan 23 16:54:15 after a few seconds device turn's off Jan 23 16:54:21 stock with framebuffer Jan 23 16:54:26 from pali's pack Jan 23 16:54:43 ashley: do you pass atags from u-boot to the kernel? Jan 23 16:54:44 http://atrey.karlin.mff.cuni.cz/~pali/u-boot-images_2012.04-1.tar.gz Jan 23 16:54:46 from there Jan 23 16:55:08 hxka: did you see ^^^ about /sys/... ? Jan 23 16:55:12 mh Jan 23 16:55:14 freemangordon: http://paste.kde.org/p2760daa5 Jan 23 16:55:18 thanks Jan 23 16:55:18 pali's not online so idk Jan 23 16:55:27 freemangordon, I don't know, I use bootmenu.d/*.item from this pack Jan 23 16:55:36 ask freemangordon, he's usually our surrogate Pali Jan 23 16:56:04 there is "ITEM_REUSEATAGS="1"" in there Jan 23 16:56:11 hmm, weird Jan 23 16:56:38 oops Jan 23 16:56:52 I accidendaly parted Jan 23 16:57:15 Ashley`: : wait, you should use u-boot from repos afaik Jan 23 16:57:25 u-boot is from repos Jan 23 16:57:44 I just got kernel with framebuffer consolle from that thread, to see what's going on Jan 23 16:57:55 freemangordon, no progress Jan 23 16:58:01 Gatta_Negra: :( Jan 23 16:58:06 why is that? Jan 23 16:58:18 i am having construction works and repair works at the apartment so i did not have any time to deal with it Jan 23 16:58:25 oh, I see Jan 23 16:58:32 i am even moving back to htc until better times Jan 23 16:58:46 oh, htc is better? Jan 23 16:58:53 no, just faster Jan 23 16:58:57 heh Jan 23 16:58:58 and i can use it for work Jan 23 16:59:04 as otherwise i have to carry two phones Jan 23 16:59:30 but i love my n900 and i will keep it for future experiments Jan 23 16:59:39 kerio: could you give the output from "cat /sys/power/efuse_vdd2" Jan 23 16:59:39 it can work without sim card also Jan 23 17:00:02 freemangordon: k one sec Jan 23 17:00:04 Gatta_Negra: what is that HTC does but n900 doesn't? out of curiosity Jan 23 17:00:46 freemangordon: 009aceef 009a94b3 00aac695 Jan 23 17:02:23 hxka: kerio: yeah, the same shit like with VDD1 calibrations :( Jan 23 17:02:31 wut Jan 23 17:02:48 the third calibration value is one and the same on every device Jan 23 17:03:05 ok whatevs Jan 23 17:03:07 00aac695 Jan 23 17:03:36 well, I guess I should make another patch Jan 23 17:06:11 freemangordon, any idea? (with that u-boot) :/ Jan 23 17:07:25 hmm, no. so you can't boot any kernel with u-boot installed? Jan 23 17:09:03 yup Jan 23 17:09:25 kernel-power kernel-cssu, nokia kernel Jan 23 17:10:06 it hangs up at "starting kernel..." and shut's down after a few seconds Jan 23 17:10:43 (with that kernel with framebuffer it shows kernel panic, and shuts down after few seconds) Jan 23 17:13:54 Ashley`: the only thing I can think of is that your rootfs is screwed somehow. Jan 23 17:14:48 Ashley`: you you somehow managed to install some wrong version of u-boot Jan 23 17:14:58 *or* you Jan 23 17:16:17 I've flashed that rootfs today (with emmc ofc) Jan 23 17:16:42 and it boots and works fine without u-boot Jan 23 17:17:06 I've tried uboot from extras and extras-devel (aren't they the same version?) Jan 23 17:17:10 brb Jan 23 17:17:43 Ashley`: what HW revision is your device? Jan 23 17:18:44 2101 Jan 23 17:19:51 same as mine Jan 23 17:19:55 hi, dos11 :) Jan 23 17:20:41 Ashley`: I'm afraid you'll have to wait for Pali to appear Jan 23 17:21:18 oh, okay. Thanks :) Jan 23 17:22:35 freemangordon, what's his nickname here? (so my xchat can notify me when he come) Jan 23 17:22:51 its Pali, but he is MIA Jan 23 17:23:34 MIA? Jan 23 17:23:42 What does that mean? :s Jan 23 17:24:00 Missing In Action Jan 23 17:24:05 Oh. Jan 23 17:29:02 yeah, his laptop is in repairing for the last ~2 months :( . at least I hope this is still the reason Jan 23 17:30:52 (( uh oh, why would i need the host key of old n900?)) either that, or you edit (or just delete) ~/.ssh/known_hosts on your PC you try to connect to N900 from Jan 23 17:32:02 you also might have to reset known_hosts on N900 Jan 23 17:33:07 aaah, you failed with password entry on N900. Yes that's another reason Jan 23 17:42:17 freemangordon, I've found something Jan 23 17:42:21 freemangordon, http://mg.pov.lt/maemo-ssu-irclog/%23maemo-ssu.2012-07-19.log.html#t2012-07-19T15:08:15 Jan 23 17:44:17 Ashley`: install correct modules Jan 23 17:47:25 DocScrutinizer05: he has a different problem Jan 23 17:47:33 the modules are in place Jan 23 17:47:36 she, but... yeah, maybe Jan 23 17:47:53 she? ok :) Jan 23 17:48:21 aah, nope. He Jan 23 17:48:47 DocScrutinizer05: the same problem with kernel failing to mount rootfs after u-boot, -ENODEV Jan 23 17:49:17 or maybe she. Dank, is Ashley` same as ashley? Jan 23 17:49:23 dnag even Jan 23 17:49:36 she, and nope :P Jan 23 17:49:49 I don't even know who ashley is :P Jan 23 17:49:57 joe phenix Jan 23 17:50:06 jack Jan 23 17:50:10 whatever Jan 23 17:50:11 (I still don't know xD) Jan 23 17:50:30 our wiki maintainer Jan 23 17:52:46 heh some days ago got a spare n900 and, well, gf came back from a trip with a dislocated usb connector on hers, nice timing Jan 23 17:52:58 aynway, yes it seems I missed some details in your problem report. It's a nasty thing in uBoot that sometimes causes uBoot to fail epically. Dunno if it's related to uBoot version or whatever Jan 23 17:52:59 bah, just got it tweaked for my work use :p Jan 23 17:53:13 will borrow it to her anyway until I get the connector fixed Jan 23 17:53:31 (if possible) Jan 23 17:53:33 prolly an incompatibility between uBoot fake memory config reported to kernel, and actual hardware in device Jan 23 17:54:12 joga: Jan 23 17:54:15 ~usbfix Jan 23 17:54:16 extra, extra, read all about it, usbfix is http://talk.maemo.org/showthread.php?t=75920 - and **NEVER** use epoxy (unless you want to seal your device for underwater) Jan 23 17:54:54 it's always possible, just sometimes not exactly easy Jan 23 17:58:20 yeah, need to do some research and open the thing at least first Jan 23 17:58:40 no rush so I'll just set up the spare for her first Jan 23 18:05:06 hm... Jan 23 18:05:15 is there kexec for maemo? Jan 23 18:05:50 I can't find "Copy from SIM card" in contacts again :F Jan 23 18:06:25 it's in phone mode, showing reception, pin given and in contacts there's just sync from other device, import from new account and import contacts (file/folder) Jan 23 18:06:45 though I did enter phone mode only after it was booted Jan 23 18:07:02 Ashley`: Yeah, in kernel-power Jan 23 18:07:25 And there is kexec-tools package in extras-testing and devel Jan 23 18:07:31 Didn't test it though Jan 23 18:08:02 oh, I see. Thanks. Jan 23 18:12:35 joga: try pkill osso-addressbook Jan 23 18:12:45 Is there any alternative to u-boot? :P Jan 23 18:12:55 hxka, ah, probably would've helped, reboot did at least Jan 23 18:14:26 Ashley`: not really Jan 23 18:14:48 :c Jan 23 18:15:05 Ashley`: OTOH yiu actually don't need uBoot except for nitdroid and that's .... not needed either Jan 23 18:15:23 i want to boot archlinux :P Jan 23 18:15:29 ooh Jan 23 18:15:57 (I have it in chroot... had (reflash), but watchdoge doesn't like pacman) Jan 23 18:16:16 your problem AFAIK is with your RAM banks hw config Jan 23 18:16:34 is it fixable somehow? Jan 23 18:16:49 well, probably in uBoot it actually is fixable Jan 23 18:18:03 or just maybe you actually got a bad block in your kernel partition Jan 23 18:19:40 I don't know how NOLO/stock-kernel is handling bad blocks, but somehow they are supposed to do Jan 23 18:20:37 alas all this is pretty obscure, thanks to NOLO being a blob Jan 23 18:20:41 can I check somehow if I've got bad blocks? Jan 23 18:21:53 err yep, there is some way to check. Seems your link you posted refers to a IRC chat that relates to it Jan 23 18:22:43 another possible approach is to cat /dev/mtd and compare to the content supposed to be there Jan 23 18:25:13 mtd_utils seem to have some useful functions Jan 23 18:27:06 mtdinfo /dev/mtd3 Jan 23 18:27:13 doesn't show bad blocks Jan 23 18:29:14 Eraseblock size: 131072 bytes, 128.0 KiB Jan 23 18:29:15 Amount of eraseblocks: 16 (2097152 bytes, 2.0 MiB) Jan 23 18:29:56 I'll just give up, meh. Jan 23 18:30:01 one bad block might already suffice to reduce available space below the threshold where uBoot(+kernel) fits. even when bad blocks are handled correctly Jan 23 18:31:17 Ashley`: could you please do a `cat /dev/mtd3 >kerneldump.bin` and provide that file somewhere for analysis? Jan 23 18:31:21 okay Jan 23 18:32:04 ooh wait, you can't get that done except in rescueOS, since when you flash uBoot it doesn't boot any normal kernel anymore, right? Jan 23 18:32:31 i can boot it from flasher-3.5 Jan 23 18:32:50 well, a dump of stock kernel NAND partition will help too. It still should indicate if there are bad blocks Jan 23 18:33:07 I'm on kp52 now Jan 23 18:33:09 Ashley`: right, you can load kernel to RAM Jan 23 18:33:09 Apparently there is nandtest to check if blocks are bad Jan 23 18:33:21 hxka: NOOOOOO! Jan 23 18:33:42 STAY AWAY FROM NANDTEST! Jan 23 18:34:08 DocScrutinizer05, http://ashley.re/kernel.bin Jan 23 18:34:12 it's a sure way to semi-brick your N900 Jan 23 18:34:21 Ashley`: thanks Jan 23 18:34:27 it is 2.6.28.10-power52 Jan 23 18:34:42 Well, OBVIOUSLY you'll have to reflash kernel after that Jan 23 18:34:42 i think Jan 23 18:35:45 hxka: yeah, or reflash CAL after accidentally erasing the config partition Jan 23 18:36:36 nandtest doesn't know about partitions afaik Jan 23 18:37:02 Oh wait Jan 23 18:37:09 -k, --keep Restore existing contents after test Jan 23 18:37:18 good luck! Jan 23 18:37:35 DocScrutinizer05, can I disable watchdog somehow? Jan 23 18:37:41 yes Jan 23 18:37:44 flasher Jan 23 18:38:08 --no-foobarwd Jan 23 18:38:13 or sth like that Jan 23 18:38:28 there are 3(?) different wds Jan 23 18:39:20 DocScrutinizer05, but it's not persistent, yep? Jan 23 18:39:24 flasher --set-rd-flags=no-omap-wd,no-ext-wd,no-lifeguard-reset Jan 23 18:39:40 also --enable-rd-mode Jan 23 18:40:00 Because watchdog is rebooting my phone when I try to pacman -S whatever Jan 23 18:40:13 Ashley`: it's persistent until you run flasher with --disable-rd-mode Jan 23 18:41:02 Ashley`: it is persistent Jan 23 18:41:06 but, Is it good to run phone in that r&d mode all the time? Jan 23 18:41:37 no Jan 23 18:41:45 Well, it drains your battery because of constant kb backlight Jan 23 18:41:56 it eats more battery, and watchdogs are for there a reason Jan 23 18:42:29 I dont like random reboots :f Jan 23 18:42:49 kb backlight is not constant but signalling SoC clock etc. And it can get disabled via a sysfs node Jan 23 18:43:12 but there is other stuff that's enabled in R&D mode, like uart3 (serial console) Jan 23 18:43:43 I just want to install packages in arch chroot without reboots Jan 23 18:44:14 hmm, that's a problem with swap hogging CPU Jan 23 18:44:25 aiui Jan 23 18:46:41 a ulimit for pacman might already help Jan 23 18:47:00 Not exactly hogging cpu Jan 23 18:47:39 Just that the cpu has nothing to do. Jan 23 18:47:40 As code isn't coming in from storage fast enough Jan 23 18:47:41 OMAP seems to have slow I/O. Jan 23 18:47:50 Not particulary Jan 23 18:48:05 when it swaps, whole system works slowly and getting unresponsive. Jan 23 18:48:27 Because the storage is slow Jan 23 18:48:31 well, that's a property of... *every* system Jan 23 18:48:45 [22:47:34] a ulimit for pacman might already help <- yep, it would just crash or malfunction due to lack of RAM :) Jan 23 18:49:03 then pacman is defect Jan 23 18:49:19 swap suxx on every system but on n900 it can be so bad it would cause watchdog reboot :D Jan 23 18:49:55 ashley; dunno what kind of setup you have, but, if you're running on MicroSD, get a fast one, I recommend Samsung Pro, Sandisk Extreme (available in two speed grades, the slower rated one is faster), or Sandisk Ultra android edition Jan 23 18:50:00 no, on N900 you *have* a wd that reboots, while other usual PCs don't Jan 23 18:50:16 ShadowJK, I don't have microsd :P Jan 23 18:50:22 Don't use ext2, it's uber slow on flash. ext4 is preferrable Jan 23 18:50:31 [22:49:57] then pacman is defect <- no, that's how memory allocation failures are looking :) Jan 23 18:50:40 echo 8192 > /sys/block/mmcblk0/queue/nr_requests Jan 23 18:50:48 t3st3r: I'm no noob Jan 23 18:50:56 Speeds up IO quite alot IME Jan 23 18:51:28 when pacman fails with malloc when no 512MB+ of memory available, then pacman *is* broken in my book Jan 23 18:51:32 DocScrutinizer05> in fact PC can have watchdog too. But it both more rare to have watchdog on PC and it's harder to make it swapping SO badly. Jan 23 18:51:34 t3st3r; I have swap on microsd, I've never experienced watchdog reboot due to swapstorm :) Jan 23 18:51:40 can someone help me with dbus-scripts. I'm trying to make someplayer skip to next track with zoom keys. The latest cssu update broke it Jan 23 18:51:47 And I've sometimes been up to gig of swap in use :s Jan 23 18:51:52 t3st3r: thanks, never heard of that¡ Jan 23 18:51:58 [22:51:15] Don't use ext2, it's uber slow on flash. ext4 is preferrable <- are there recent kernels for n900 to run it safely? Jan 23 18:52:42 DocScrutinizer05> IIRC N900 is really paranoid about watchdogs: 2 watchdogs + some software monitor. Jan 23 18:52:54 so? Jan 23 18:53:04 t3st3r; also PC even with slow laptop harddrive has 100+ IOPS performance, the N900 emmc is <10, cheap SD cards around 1. SSDs 10,000 or more. Jan 23 18:53:34 t3st3r; kernel-power has backported ext4 Jan 23 18:53:48 b1101: what's zoom keys? Jan 23 18:53:56 ShadowJK> as I told I got impression OMAP I/O is "quite slow". Though idea with fast SD could work I guess. Jan 23 18:54:06 the volume key Jan 23 18:54:25 but still, i/o with SD or eMMC tend to cause CPU load. Jan 23 18:54:32 t3st3r; my Asus tablet with nvidia tegra has "slower I/O" than N900 :) Jan 23 18:54:52 a particularly funny idea to abuse volume keys for something other than volume control during mp3 playback Jan 23 18:54:55 but well Jan 23 18:55:01 DocScrutinizer05, http://talk.maemo.org/showthread.php?t=70405&page=52 the last post. I have no idea what to do with those things Jan 23 18:55:43 ShadowJK> oh, really? :D Jan 23 18:56:18 t3st3r; the big issue is that writing anything takes seconds in the worst case, and there's loads of little writes going on all the time... the worst case scenario becomes more likely the more stuff is going on simultaneously, which also creates greater memory pressure, so stuff is constantly waiting for some piece of code to get read in, but first needs to wait for the write to complete Jan 23 18:57:02 t3st3r; yes, that tablet has a slower emmc than N900.. Jan 23 18:58:10 Quad core, but the cores mostly idle, and nothing happens in browser, browser having first written url history then checking disk cache blocked until adding the url to url history completes Jan 23 18:59:34 ShadowJK> I guess in worst case it's infamous bug 12309 could kick in. Jan 23 18:59:36 04Bug https://bugs.maemo.org/12309 menu icon not graphical Jan 23 18:59:48 Roughly speaking, a 8 megabyte write takes just as long as a 4 kilobyte write, updating a database file involves something like minimum 4 such 4k writes, which works out to the equivalent of writing a 32 megabyte file... Jan 23 18:59:51 bot, you lame :P Jan 23 19:00:26 flash is block device... Jan 23 19:00:38 It pretends to be a block device Jan 23 19:01:34 It is actually a block device. Minimum write unit is page (several kb). But sometimes it takes erase of large block (hundreds Kb or more). Jan 23 19:01:39 ~wiki erasepage Jan 23 19:01:41 I couldn't find a matching article in wikipedia, look for yerselves: http://en.wikipedia.org/wiki/Special:Search?search=erasepage&go=Go Jan 23 19:02:06 t3st3r; the emmc in my asus tablet has 8M erase block Jan 23 19:02:11 Some advanced controllers with more than 1 IC attached can operate on "erase groups", these are even more. Jan 23 19:02:37 4M, 6M, 8M and 12M "effective" erase blocks are thee norm now Jan 23 19:02:46 Maybe more than 1 flash IC attached and controller haves erase group width 8Mb? Jan 23 19:02:56 possibly Jan 23 19:03:51 From my point of view it's not much of a difference whether it's 2x4M grouped together or 1x8M, the net result is the same, worst case the emmc does 8M read-modify-write for every request Jan 23 19:03:57 I don't see how all this factoid-tossing is helping Ashley` Jan 23 19:04:54 nr_requests bigger and swap on uSD should sort out ashley Jan 23 19:04:55 DocScrutinizer05> to be honest I don't see how to help with it at all. Maybe fast SDHC with swap on it could improve it a bit? Jan 23 19:04:57 b1101: I don't know Jan 23 19:05:55 also maybe thumb build would help? More RAM for programs after all. Jan 23 19:05:57 t3st3r: thanks! Jan 23 19:06:19 Larger queue size helped me, I saw an increase in average request size as reported by iostat, which means less read-modify-write type write amplification on the emmc, which means higher net throughput :D Jan 23 19:06:58 ShadowJK> is this tunable system-wide? Jan 23 19:07:14 It's per block device Jan 23 19:07:46 ahh then it looks like nice thing to try to select better blocking for flash devices. Jan 23 19:08:43 Ashley`: after t3st3r stopped discussing basics with ShadowJK, I suggest you ask ShadowJK what he suggests to improve the watchdog timeout problem Jan 23 19:08:46 With modern kernel we'd use the flash aware I/O scheduler, and the flash friendly filesystem (f2fs) :) Jan 23 19:09:07 Don't use ext2, it's uber slow on flash. ext4 is preferrable Jan 23 19:09:17 echo 8192 > /sys/block/mmcblk0/queue/nr_requests Jan 23 19:09:59 [23:09:40] With modern kernel we'd use the flash aware I/O scheduler, and the flash friendly filesystem (f2fs) :) <- I wonder if it possible on n900 :D Jan 23 19:10:03 thanks, i'll try (i need to install arch again - i was reflashing my device today) Jan 23 19:11:07 as for ext2/3 you can try to hint it to use stride supposedly equal to erase block of card. Jan 23 19:11:22 *sigh* Jan 23 19:11:24 and it could be "ext3 minus journal". Jan 23 19:11:25 Before I changed that tunable, large highspeed file transfers were basically impossible for me Jan 23 19:11:45 ShadowJK> what kernel you use for ext4? Jan 23 19:11:55 :-S Jan 23 19:12:03 in 2.6.28 it was experimental/unsafe oO Jan 23 19:12:10 t3st3r; ext wouldn't allow it, but I wanted to try run with gigantic journal, on the order of 32g journal on 64g card Jan 23 19:12:27 t3st3r; power52, it has backported ext4 fixes Jan 23 19:12:50 oh, really? So it haves less scary ext4? Jan 23 19:13:04 RLY Jan 23 19:13:22 yes Jan 23 19:13:45 I think Estel runs ext4 /home Jan 23 19:14:05 me too lazy to edit startup scripts and only have it on uSD Jan 23 19:14:21 (and a bunch of speed critical stuff from /home on uSD and symlinked) Jan 23 19:14:55 because originally there was a lot of bugs fixed, up to total volume blow-up. Jan 23 19:15:01 ya Jan 23 19:15:31 that's why we got backports of a working version Jan 23 19:15:35 and I only dared to use Ext4 at about 2.6.38 or so... Jan 23 19:15:56 It's not a fair comparison, but I've blown up the ext3 on $home about 5 times, and ext4 on uSD 0 times Jan 23 19:16:20 I havent blown up any EXTs because I'm careful :) Jan 23 19:16:31 * DocScrutinizer05 heads out Jan 23 19:18:32 btw, I think the misconception that I/O is busywait comes from the myriad of cpu monitors that treat all not-idle cpu time as busy, and shows widget claiming cpu is fully utilized :) Jan 23 19:21:16 Despite my optimized setup, the typic web browsing session for me is 20-30% cpu wa, meaning that the cpu was idling despite processes wanting to run, as i/o the process depends on is in progress Jan 23 19:22:41 Or in other words, 30% of websurfing system load consists on waiting for the emmc and uSD to do their job :/ Jan 23 19:25:51 ShadowJK: I wasn't to suggest cpu was busy-waiting. Hogged meant something different than busy in my staement Jan 23 19:28:46 the real problem I think is [swapd] not doing preemption Jan 23 19:29:21 DocScrutinizer05, what you're going to do with that kernel.bin from me? Jan 23 19:29:47 maybe I find time to compare it to a "sane" image Jan 23 19:30:23 ah... one thing... Jan 23 19:30:48 You wanted me to do that dump with u-boot flashed into nand? Jan 23 19:31:47 It's kinda funny how much my finetuning of my N800 has changed over the years, as I've continously upgraded the SD cards in it Jan 23 19:32:12 I wonder if it's already past the point where the SD cards have more CPU power than the N800 itself :) Jan 23 19:32:29 DocScrutinizer05; what do you mean by preemption? Jan 23 19:32:55 I mean kernel can't run [wd] Jan 23 19:34:10 Well it would run it if it was still in ram Jan 23 19:34:50 yeah, not sure if a 'normal' (even kernel) process is supposed to block kernel from multitasking Jan 23 19:35:22 waiting for the emmc and uSD to do their job :/ <- means it could theoretically be 30% faster at same horsepower :) Jan 23 19:35:34 then otoh sw-reset (wd) could as well be the dsme one, or whoknowswhat Jan 23 19:35:57 this is backupmenu's watchdog software: Jan 23 19:36:05 ( while true ; do echo x ; sleep 1 ; done ) > $dev 2>/dev/null < /dev/null & Jan 23 19:36:42 I wonder what process does the equivalent in fremantle... mce? Jan 23 19:36:44 yeah, that's how all watchdogs work Jan 23 19:36:51 dsme Jan 23 19:37:21 and another one is embedded in kernel I guess Jan 23 19:37:34 disclaimer: AFAIK Jan 23 19:38:00 Ashley`: don't worry Jan 23 19:38:27 So if dsme isn't in ram when its timer expires, and there's heavy I/O going on, it probably doesn't make it back in time Jan 23 19:38:47 :nod: Jan 23 19:38:56 though dsme should be ram-locked Jan 23 19:39:25 (dunno how that's done but I know it is possible and used in maemo a lot) Jan 23 19:40:08 DocScrutinizer05, http://talk.maemo.org/showthread.php?t=68125 Jan 23 19:40:16 DocScrutinizer05, there is something about ramlocking Jan 23 19:40:23 might be intentional, when a relatively small thing like dsme doesn't respond quickly enough, the UI responds even slower, so intentionally let it reboot? Jan 23 19:45:58 319M swap in use on N800 :) Jan 23 20:16:47 ShadowJK> do you have N800? Jan 23 20:16:58 that old tablet from Nokia? Jan 23 20:17:25 I have a N800, N810, 4 broken N900, 2 working N900 Jan 23 20:18:26 echo 8192 > /sys/block/mmcblk0/queue/nr_requests Jan 23 20:18:42 oh Jan 23 20:18:44 can't create /sys/block/.../nr_requests: Read only filesystem Jan 23 20:18:54 missing an iosched in that path somewhere Jan 23 20:19:21 /sys/block/mmcblk0/queue/nr_requests Jan 23 20:19:31 hm Jan 23 20:19:36 looks identical :P Jan 23 20:19:55 Look around in /sys/block/mmcblk0/ for it? Jan 23 20:20:05 there is io_shed Jan 23 20:20:13 iosched* Jan 23 20:20:26 but there isn't nr_requests Jan 23 20:20:43 (also, I can't for example change sheduler) Jan 23 20:20:46 what kernel is this Jan 23 20:20:51 (i could some time ago...) Jan 23 20:20:53 kp52 Jan 23 20:21:18 same here Jan 23 20:21:39 You dont want anything except cfq anyway Jan 23 20:22:39 (i was just testing something with scheduler) Jan 23 20:22:49 (anyway, it was in /sys, and it wasn't ro...) Jan 23 20:22:57 brb... Jan 23 20:23:14 I have in /sys/block/mmcblk*/queue/ : hw_sector_size max_hw_sectors_kb nomerges read_ahead_kb scheduler iosched max_sectors_kb nr_requests rq_affinity Jan 23 20:23:25 is mmcblk1 present in /sys and more populated? Jan 23 20:24:08 oh, wait Jan 23 20:24:17 there isnt mmcblk0 Jan 23 20:24:24 there is mmcblk0p* Jan 23 20:24:27 brb Jan 23 20:26:11 uh Jan 23 20:26:13 nope Jan 23 20:26:15 sorry Jan 23 20:26:20 mmcblk0p is different Jan 23 20:26:26 i forgot about -d switch Jan 23 20:26:40 and i did ls /sys/block/mmc* Jan 23 20:27:00 instead ls -d /sys/block/mmc* Jan 23 20:27:02 :P Jan 23 20:28:07 37 Jan 23 20:28:17 *sigh* Jan 23 20:43:14 Ashley`: That's why tab completion exists Jan 23 21:04:05 Ashley`: I checked your kernel.bin, it's ok Jan 23 21:05:17 oh, thanks Jan 23 21:05:53 But it means that I still dont know what causes the problem :f Jan 23 21:28:01 probably your memory hw architecture then Jan 23 21:28:25 my initial suspition Jan 23 21:28:51 suspicion* Jan 23 21:29:51 could somebody check a x86-linux "bug" I seem to encounter? Jan 23 21:31:46 bug? Jan 23 21:33:19 in X11 make your kbd numlock LED shine. (I.E. set numlock). Then switch to console (ctrl-alt-F1). switch back to X11 (usually alt-F7). watch numlock LED. press SHIFT-key-> LED off. press capslock ->numlock+capslock LED on. (press capslock again to unlock capslock) Jan 23 21:34:45 can you reproduce this or is it just me? Jan 23 21:35:54 i suppose it's just you . ...Going to console the numlock led remains in the same condition Jan 23 21:35:55 I can reproduce it Jan 23 21:36:35 xes: yes Jan 23 21:38:56 hxka: thanks Jan 23 21:39:19 I guess it's a bug in kbd LED control driver Jan 23 21:40:20 at least finally I managed to find how to reproduce it. Before it drove me nuts at times Jan 23 21:40:41 DocScrutinizer05: sorry, can't reproduce (kde 4.11.5, kernel 3.12.8) Jan 23 21:42:25 kde 4.8.5 (4.8.5) "release 2" Jan 23 21:42:57 kernel 3.4.63-2.44-desktop Jan 23 21:43:01 ..but i remember that the kde version of fedora 17 had some strange numlock issue (kde 4.9.?) Jan 23 21:43:46 4.8.5 seems close enough to 4.9 Jan 23 21:45:46 could be related: https://bugs.kde.org/show_bug.cgi?id=183308 Jan 23 21:45:48 04Bug 183308: was not found. Jan 23 21:47:03 https://bugs.kde.org/show_bug.cgi?id=183308#c31 Jan 23 21:47:04 04Bug 183308: was not found. Jan 23 22:00:56 rather exactly my issue than just related Jan 23 22:08:35 1798 user 93124 S /usr/bin/hildon-status-menu Jan 23 22:08:38 hmmm. Jan 23 22:11:54 hmm? Jan 23 22:12:03 i dont remember it growing that big Jan 23 22:12:12 err Jan 23 22:13:24 user 1766 0.0 0.2 3936 536 ? Ss 2013 0:00 /usr/bin/hildon-status-menu Jan 23 22:13:25 user 1768 0.2 3.1 98240 7620 ? Ssl 2013 248:12 /usr/bin/hildon-status-menu Jan 23 22:13:46 ah ok. was just remembering wrong then as i updated my ancient cssu Jan 23 22:16:47 the numlock should change as you switch between the console and Xorg, if it's in a different state in both. Jan 23 22:17:33 each console can have the locks in different states. Jan 23 22:19:16 * Maxdamantus remembers writing programs that read and wrote to devices in /dev/input/* to make the lights make funny patterns as you typed. Jan 23 22:23:10 ah, nvm, didn't understand the problem at first. Jan 23 22:43:09 DocScrutinizer05: did you know http://archive.is/lP093 (http://wiki.meego.com/ARM/N900/Using_Rescue_Initrd), an n900 rescue system to be loaded into ram via flasher? Jan 23 22:43:39 hmm? Jan 23 22:43:50 you can boot it even if your rootfs is shot Jan 23 22:43:58 ~rescueos Jan 23 22:43:58 somebody said rescueos was http://n900.quitesimple.org/rescueOS/ Jan 23 22:44:39 "This project is _NOT_ the meego rescue initrd." Jan 23 22:44:39 k Jan 23 22:45:18 http://n900.quitesimple.org/rescueOS/documentation.txt Jan 23 22:45:42 that's what i copied from, yes Jan 23 22:47:17 anyway, thanks... should have found that earlier Jan 23 22:48:47 wow, wiki.meego.com still exists? thought they nukes meego.com completely Jan 23 22:48:54 nuked* Jan 23 22:49:24 that url is a copy of that one page Jan 23 22:49:31 aaah Jan 23 22:51:12 sorry, I'm a tad grumpy Jan 23 22:51:33 thanks for letting us know, anyway Jan 23 22:52:51 i was just wondering why i did not run into that (or rescueos) earlier... seems a bit underdocumented Jan 23 22:53:19 hmm yeah, rescueOS for sure could use some more PR Jan 23 22:53:37 nin101 is a shy sissi ;-P Jan 23 22:54:21 ~factinfo rescueos Jan 23 22:54:21 rescueos -- created by DocScrutinizer05 <~HaleBopp@openmoko/engineers/joerg> at Sun Sep 2 16:57:34 2012 (508 days); last modified at Thu Aug 29 14:08:40 2013 by kerio!~kerio@Maemo/community/contributor/kerio; it has been requested 49 times, last by DocScrutinizer05, 10m 23s ago. Jan 23 22:54:26 hmm Jan 23 22:54:52 the meego reascue image has the cute advatange of not having a keyboard map, and having usb networking but no single network client or server Jan 23 22:55:07 lol Jan 23 22:55:26 yeah Jan 23 22:55:29 ;-P Jan 23 22:55:35 extremely useful Jan 23 22:55:50 the only special characters you can type are <>., cursor keys and enter Jan 23 22:56:03 :nod: Jan 23 22:56:14 absolutely sufficient, no? X-P Jan 23 22:56:21 yeah :D Jan 23 22:56:22 sshd Jan 23 22:56:29 ooh, i guess you have Jan 23 22:56:30 kerio, as i said, no clients or servers Jan 23 22:56:34 no tab either Jan 23 22:56:39 ...oh Jan 23 22:56:58 and busybox has a - in it's name, so you can't call applets that are not symlinked (like telnetd?) Jan 23 22:57:21 well, i guess rescueos is a bit better, will try Jan 23 22:57:38 ctrl-i ? Jan 23 22:57:56 rescueos is awesome, I heard Jan 23 22:57:59 good one Jan 23 22:59:40 on maemo.cloud-7 there's an older version that still uses a kernel compatible with the stock maemo kernel Jan 23 23:00:05 nin101 thought it's leete to go for new kernel Jan 23 23:00:08 iirc Jan 23 23:00:31 thanks, will try... Jan 23 23:00:33 later Jan 23 23:07:21 http://maemo.cloud-7.de/NIN101_N900_rescue_os/nin101.uni.cx/n900-new/206.253.166.96/N900/rescueOS/CHANGELOG Jan 23 23:16:34 ShadowJK: nice -20 swapoff /dev/mmcblk1p2 && #<-- does that really work? (the nice -20) Jan 23 23:16:54 i like how rescueos contains 'fdformat' Jan 23 23:17:04 lol Jan 23 23:17:34 messybox-paua, eh? Jan 23 23:18:21 well, at least all applets are linked, unlike on the meego initrd Jan 23 23:23:12 ShadowJK: actually my mind boggles form those 5 scriptlines in http://enivax.net/jk/n900/reswap.sh.txt Jan 23 23:23:58 swapon /dev/mmcblk0p3 >/dev/null 2>&1 || echo cow Jan 23 23:24:00 nice -20 swapoff /dev/mmcblk1p2 && Jan 23 23:24:01 iostat -md mmcblk0p2 | grep mmcblk0p2 | awk '{print $6}' >/root/reswap.stat Jan 23 23:24:03 /home/user/swapon -p 10 /dev/mmcblk1p2 && Jan 23 23:24:04 nice -20 swapoff /dev/mmcblk0p3 && Jan 23 23:24:06 swapon /dev/mmcblk0p3 Jan 23 23:24:19 ------ last two lines are especially mindboggling Jan 23 23:25:41 without the last line it might make sense Jan 23 23:33:30 ShadowJK: aiui you rely on nice -20 blocking the parent script from execution until the background process finished? **** ENDING LOGGING AT Fri Jan 24 02:59:59 2014