**** BEGIN LOGGING AT Sat Jul 09 02:59:58 2016 Jul 09 05:41:01 vajb: poser :P Jul 09 07:23:06 DocScrutinizer05: cheers. yeah didnt see any question there either :) Jul 09 10:56:50 Vajb: i'd love to have that one, lol Jul 09 10:57:23 pretty easy, speedswap batteries Jul 09 10:58:38 i've done that a couple of times, yes Jul 09 10:59:39 LMD means nothing, not even when CI=0 Jul 09 11:00:06 it's not very convenient tbh .. esp. the need for wall charger. most times when you need that extra power, it's because you do not have access to 'charging' Jul 09 11:00:28 aah that sort of howswap, yeah Jul 09 11:00:37 hotswap even Jul 09 11:02:17 re "access to charging": there are pretty small (~1.5 .. 2 BL-5J volume sizes) USB power banks Jul 09 11:02:44 even with photovoltaic cell Jul 09 11:03:55 :) Jul 09 11:07:58 https://www.amazon.com/Dual-USB-Waterproof-Solar-Battery-Charger/dp/B01EZ91C76/ref=sr_1_11?ie=UTF8&qid=1468062437&sr=8-11&keywords=usb+power+bank+solar Jul 09 11:08:49 https://www.amazon.com/dp/B01H3CFFF2?psc=1 Jul 09 11:09:25 the latter actually looks like a nifty smart critter Jul 09 11:12:05 1200mAh tho Jul 09 11:12:22 well i mean, the first one claims 5Ah Jul 09 11:12:26 for 10 monies Jul 09 11:12:31 there's no way that's true Jul 09 11:15:18 https://www.amazon.com/Mini-Portable-Retractable-Motorola-Blackberry/dp/B01EJ69VF0/ref=sr_1_28?s=wireless&ie=UTF8&qid=1468062883&sr=1-28&keywords=usb+mini+power+bank Jul 09 11:38:48 burn in stress test for N900 CPU: rsync a full eMMC via ssh Jul 09 11:39:19 temperature 52 C Jul 09 11:40:57 how do you tell the temperature? Jul 09 11:41:36 http://wstaw.org/m/2016/07/09/plasma-desktopEA2277.png Jul 09 11:41:49 temp? simple: bq27200.sh Jul 09 11:42:09 chip temp of bat gauge (_not_ battery temp!) Jul 09 11:50:43 i hope you're using a recent enough sshd Jul 09 11:50:54 (and ssh, but that's easier) Jul 09 11:55:20 chacha20 is measurably faster than other ciphers Jul 09 11:58:52 sshd on N900 is as recent as it gets, I guess Jul 09 11:59:09 IOW ancient Jul 09 12:00:16 faster would be just inserting microsd, rsync to it and rsync again to target Jul 09 12:00:44 that's not the purpose of the setup though Jul 09 12:00:48 order of magnitude at least faster Jul 09 12:02:12 my cron on PC doesn't control a robot Jul 09 12:02:38 my server without AES-NI does 195MB/s of aes128gcm, 178MB/s of aes128cbc and 580MB/s of chacha20 Jul 09 12:03:00 almost 3x speedup just by updating openssh Jul 09 12:03:24 kerio: that's very useful info, but I dunno what to make of it in context of rsync to N900 Jul 09 12:03:33 rsync goes over ssh Jul 09 12:04:02 unless you set up rsyncd Jul 09 12:04:05 so, you say there's a openssh version for N900 I did miss to notice? Jul 09 12:04:34 i don't think it got packaged by anyone actually Jul 09 12:04:40 see? Jul 09 12:05:00 i just asked fmg to compile it for me Jul 09 12:05:04 I dunno what to make of it in context of rsync to N900 Jul 09 12:05:08 and i untarred it in /usr/local Jul 09 12:05:25 ok whatever forget i said anything Jul 09 12:07:58 transfer rate is pretty constant at a pathetic 500kB/s Jul 09 12:08:01 (Question from Oksana) Installed update from kernel power 51 to kernel power 53, phone does not boot, gives "Can not flash kernel, required files not found #00000025.62017712". Alas I have non-working USB port, so can not look at files and fix them by hand... Jul 09 12:08:38 ouch Jul 09 12:09:18 o.o Jul 09 12:10:57 did you have multiboot installed? Jul 09 12:14:18 I guess when device refuses to boot (how the heck you get the diagnostic output then?) then your only chance is USB, by whatever means. If jack is broken, there's still the 2 testpads under battery Jul 09 12:16:09 (From Oksana) I am trying to fix USB port (googling videos), the phone is in a boot loop, it shows this Guru Meditation message and shuts down and then boots again. Jul 09 12:16:21 however since the diagnostic output has to have a source, the device actually boots, just maybe not entering normal interactive GUI state. bootmenu/backupmenu shell is your best bet Jul 09 12:18:36 iirc kernel installation hooks in a script to do the actual kernel flashing during next boot. BM should kick in before that, so you could remove that kernel flashing hooked-in script Jul 09 12:19:37 hi svetlana! Jul 09 12:19:58 best wishes and good luck to oksana Jul 09 12:24:01 http://paste.opensuse.org/49384982 Jul 09 12:24:29 # Scheduled updates are to be collected from here: Jul 09 12:24:30 STAGING_DIR=/var/lib/softupd/staging Jul 09 12:26:19 sbin/preinit: http://paste.opensuse.org/39905617 Jul 09 12:27:40 I'm not 100% sure this is the right location to look at Jul 09 12:27:52 but odds are it is Jul 09 12:31:42 ~usbfix Jul 09 12:31:42 from memory, usbfix is http://talk.maemo.org/showthread.php?t=75920 - and **NEVER** use epoxy (unless you want to seal your device for underwater), or https://www.youtube.com/watch?v=fYz5nIHH0iY#t=1866, you will basically need two irons: a small good one (or better hot-air reflow) and a 60+ Watt Jul 09 12:33:59 desparate measure: solder the two data wires (green, yellow?) of a USB cable to the pair of USB testpoints under battery, solder GND (black) to ground testpad nearby (do NOT use battery minus which isn't a GND pin!) Jul 09 12:35:04 http://wiki.maemo.org/N900_Hardware_Hacking#Debug_ports Jul 09 12:35:23 J5300/01 Jul 09 12:35:26 data Jul 09 12:35:55 J2000 Jul 09 12:35:57 pin 1 :: GND (MIDDLE upper point, right to J2062 !) Jul 09 12:36:50 you probably also want to solder red wire of cable to pin 3 :: USB VBUS 5V Jul 09 12:38:12 unless charging via USB jack still works, which means the VBUS pin still has contact and so you can provide 5V via the normal way Jul 09 12:39:17 test: when device power up when connecting it to a charger, then VBUS still works Jul 09 12:42:32 How to enter the backup menu? Multiboot is installed Jul 09 12:50:23 gryllida: just google your error code Jul 09 12:50:31 there is tmo thread that could help you Jul 09 12:50:51 and most likely multiboot is the culprit of failed update Jul 09 12:51:07 ~maemo-multiboot Jul 09 12:51:08 hmm... maemo-multiboot is deprecated, and a horrible hack. PROBLEMS WITH NITDROID/MULTIBOOT? reflash rootfs&kernel aka COMBINED Jul 09 12:53:13 this factoid been defined after literally dozens of failed efforts to recover from a broken multiboot (multiboot is NOT uBoot!) Jul 09 12:54:03 multiboot basically reflashes kernel on *every* boot, which is.. well what the factoid says: a *horrible* hack Jul 09 12:54:44 no wonder its broken, it similar to the /system/etc/install-recovery.sh annoyance on Android. Jul 09 12:55:07 aka no surprise Jul 09 12:55:12 * Juest waves Jul 09 12:57:03 and obviously when you flash a new kernel, next multiboot will fail epically since it has no idea that kernel (and matching modules!!) changed, and so multiboot will flash something inappropriate or will throw random crappy errors Jul 09 13:01:26 it is a recoverable brick at least? Jul 09 13:01:32 * Juest wonders Jul 09 13:02:03 >>Plzzz tel cz my n900 is screeveedd << OHMY!! Jul 09 13:02:20 o.o Jul 09 13:02:20 https://talk.maemo.org/showthread.php?t=72879&page=2 Jul 09 13:03:43 Juest: multiboot? -->reflash Jul 09 13:03:54 juest, with working usb port you cannot brick n900 Jul 09 13:04:04 the only way is to nuke nolo Jul 09 13:04:10 no Jul 09 13:04:13 but that requires some skill Jul 09 13:04:15 not even that Jul 09 13:04:29 Well, apparently wikiwide didnt armor the usb port Jul 09 13:04:42 nolo? wat Jul 09 13:04:47 nevermind Jul 09 13:04:51 ~nolo Jul 09 13:04:55 o.O Jul 09 13:05:10 NOkia bootLOader Jul 09 13:05:18 Right Jul 09 13:05:26 ~coldflashing Jul 09 13:05:31 weird its not in infobot, would expected it Jul 09 13:05:35 ~coldflas Jul 09 13:05:37 ~coldflash Jul 09 13:05:37 [coldflash] http://mg.pov.lt/maemo-irclog/%23maemo.2010-10-31.log.html#t2010-10-31T23:34:01 (jacekowski's instructions), or http://wiki.maemo.org/Updating_the_firmware/Cold_Flashing Jul 09 13:05:49 Thank you. Jul 09 13:10:08 yw Jul 09 13:11:26 ~nolo is NOkia bootLOader, Nokia's version of uboot-alike first (actually 2nd) stage bootloader for NIT like N900 Jul 09 13:11:26 DocScrutinizer05: okay Jul 09 13:12:00 ~nit Jul 09 13:12:00 somebody said nit was Nokia Internet Tablet Jul 09 13:12:10 hehe Jul 09 13:12:30 nice addition btw Jul 09 13:13:43 ~nolo is also somebody said it was a nasty character in Norwegian languange Jul 09 13:13:43 okay, DocScrutinizer05 Jul 09 13:14:08 ~nolo Jul 09 13:14:08 rumour has it, nolo is NOkia bootLOader, Nokia's version of uboot-alike first (actually 2nd) stage bootloader for NIT like N900, or somebody said it was a nasty character in Norwegian languange Jul 09 13:14:10 * Juest wonders is there a way to properly execute maemo shortcuts on xepyhr since shift+control is used to grab and release mouse/keyboard control Jul 09 13:14:34 reconfigure xephyr shortcuts Jul 09 13:14:51 either bypassing xephyr and using X directly or yeah that, i dont know where the config is located Jul 09 13:15:02 nfc Jul 09 13:15:33 its sort of "fragile" the launcher only seems to work properly when using the shortcut on the desktop, it didnt work with a terminal directly Jul 09 13:15:54 launcher? Jul 09 13:16:01 .desktop Jul 09 13:16:03 with a terminal? Jul 09 13:16:17 .desktop's commandline Jul 09 13:16:31 .desktop are intricate Jul 09 13:16:32 sdk is a ubuntu lucid vm.... Jul 09 13:17:01 often a run-standalone helps Jul 09 13:17:15 what you mean? hmm Jul 09 13:17:25 run-standalone Jul 09 13:17:35 * Juest adventures into /etc about xephyr Jul 09 13:17:50 run-standalone.sh Jul 09 13:17:57 path? Jul 09 13:18:01 err Jul 09 13:18:14 /usr/bin/run-standalone.sh Jul 09 13:18:22 ah ok Jul 09 13:22:10 anyway you can't run e.g /usr/share/applications/hildon/camera-ui.desktop as a cmdline binary or script Jul 09 13:23:39 you got some dbus related info in those .desktop: Jul 09 13:23:43 X-Osso-Service=com.nokia.cameraui Jul 09 13:23:44 X-Osso-Type=application/x-executable Jul 09 13:24:09 and some supposedly cmdline path: Exec=/usr/bin/camera-ui Jul 09 13:24:45 ... i've tried running /usr/bin/start_xephyr.sh and it didnt work properly from a terminal Jul 09 13:24:46 often the latter is sinmply a symlink to maemo-launcher Jul 09 13:25:28 starting xephyr is tricky Jul 09 13:25:32 s/starting // Jul 09 13:25:33 DocScrutinizer05 meant: xephyr is tricky Jul 09 13:26:24 where are you running it ? Jul 09 13:26:54 Apparently only it works properly by using the desktop launcher Jul 09 13:27:01 solely on a terminal doesnt work Jul 09 13:27:18 And i am still trying to find Xephyr's config files Jul 09 13:29:35 Okay, it works from a oot terminal Jul 09 13:29:45 s/oot/root Jul 09 13:29:57 no screen :( Jul 09 13:30:05 ah, not allowed Jul 09 13:30:54 Weird, now i simply do not have a display :/ Jul 09 13:32:38 Great... now i dont have a screen Jul 09 13:34:49 I am unable to run it from a standard terminal Jul 09 13:35:07 root terminal works but the sdk is not designed to be ran as root Jul 09 13:35:21 now maemo is there but in some sort of limbo Jul 09 13:35:34 it broke for the session Jul 09 13:37:10 feel free to join at juesto.pw:6000 pw: maemosdk Jul 09 13:37:22 (vnc) Jul 09 13:38:08 it also has few glitches.... Jul 09 13:38:15 like changing themes breaks ui Jul 09 13:39:22 sorry, last time I used maemo-sdk + xephyr must be like 6 years ago, maybe longer Jul 09 13:40:44 damnit, it makes me feel tired Gryllida Jul 09 13:41:59 i cant find anything about tweaking xephyr Jul 09 13:43:58 is xephyr an an emulator ? Jul 09 13:44:29 man xephyr its not a emulator but a Nested X Jul 09 13:44:38 in this case its some sort of emulator Gryllida Jul 09 13:44:48 fremantle sdk emulator Jul 09 13:44:52 one does not maemo until they are soldering Jul 09 13:45:02 o_O? Jul 09 13:45:18 meep pm Gryllida Jul 09 13:48:34 our solder is too big, and no flux, this tiny thing. feynman's radio life was more fun, at least the items were large enough for the soldering iron to stick its end nto them freely Jul 09 13:48:57 :( Jul 09 13:51:15 Lol, i cant delete in maemo configuration editor Jul 09 13:52:42 scrollwheel doesnt even work Jul 09 13:52:45 neither hold :/ Jul 09 13:53:08 at least not in that app Jul 09 13:53:58 and theres no camera on the sdk Jul 09 13:54:47 Hello? Jul 09 13:56:50 well idk Jul 09 13:56:57 apparently its the app itself Jul 09 13:59:00 wtf, upgrade ability is broken somewhat Jul 09 14:03:52 lol Jul 09 14:03:56 now anything fials to install Jul 09 14:05:37 Juest: you're talking random? Jul 09 14:06:00 perhaps, i am still upset with this half-working thing Jul 09 14:10:31 great, i managed to freeze fremantle sdk Jul 09 14:10:33 ffs Jul 09 15:14:32 do we have a report of someone restoring bluetooth functionality after clearing CAL? Jul 09 15:14:59 Gryllida: don't even think of soldering SMT unless you got a proper noclean flux Jul 09 15:15:24 kerio: not afaik, but why whouldn't it work? Jul 09 15:16:00 istr that guy who did nandtest on everything didn't have a working bluetooth afterwards Jul 09 15:16:07 the bluetooth MAC is stored in CAL Jul 09 15:16:08 you're not supposed to clear CAL, restore it to arbitrary proper content Jul 09 15:16:30 what if critical failure happens Jul 09 15:17:37 ^^^ Jul 09 15:17:53 we're soldering to the testing pads Jul 09 15:18:24 hold on weren't there some mtdblk devices Jul 09 15:18:35 that behaved like shitty block devices Jul 09 15:18:43 hm? Jul 09 15:19:01 hold on what's log Jul 09 15:19:19 http://maemo.cloud-7.de/dd_if%3dmtd1ro.bin Jul 09 15:19:36 log = mtdoops Jul 09 15:20:04 oh right Jul 09 15:20:07 how do i read that again? Jul 09 15:20:19 DocScrutinizer05: are you sure that dd is enough? Jul 09 15:20:34 like, have you tried a restore? Jul 09 15:20:54 nanddump has a bunch of options regarding OOB data and bad blocks ._. Jul 09 15:21:36 sp-oops-extract Jul 09 15:22:36 http://maemo.cloud-7.de/mtd_debug_read_mtd1_0_0x60000.bin Jul 09 15:23:32 why isn't cal-tool able to dump and restore CAL from a higher level, though? Jul 09 15:24:03 no idea Jul 09 15:25:13 maybe because cal-tool isn't meant for that, there's another more generic cal thing made by iirc pali Jul 09 15:25:56 because there is no backup/restore tool for cal Jul 09 15:27:05 https://dev.openwrt.org/browser/packages/utils/calvaria Jul 09 15:30:46 http://maemo.cloud-7.de/_hidden/cncmods-net/www.cncmods.net/files/calvaria Jul 09 15:35:29 http://paste.opensuse.org/85827438 Jul 09 15:39:22 http://paste.opensuse.org/95893210 Jul 09 15:44:12 seems calvaria lacks write capabilities, however that should be trivial Jul 09 15:44:57 and anyway you could check if a raw data write to NAND CAL partition (nandwrite et al) would result in a CAL content that looks OK Jul 09 15:45:51 you even could edit the image file before to your liking, and check that file with calvaria to look OK before you flash it Jul 09 15:46:23 edit manually that is Jul 09 15:47:18 I dunno if openwrt has improved the calvaria tool Jul 09 15:52:13 noteworthy that niokia's phenix service tool knows to write e.g. ALS calib data to CAL, so I guess that must be a generic way Nokia uses to calibrate their phones - phenix is universal for all Nokia devices Jul 09 15:53:14 Phoenix Jul 09 15:53:33 http://www.allmobiletools.net/2014/12/nokia-phoenix-service-software-201415.html Jul 09 15:57:18 note that this particular site looks like scam Jul 09 15:57:41 I explicitly don't recommend to download anything there Jul 09 15:59:32 anyway here's more info http://nokiarevolution.com/how-to-flash-your-nokia-device-via-phoenix-usb-dead-flashing/ Jul 09 16:07:17 kerio: nope, I actually didn't try to restore yet. Anyway what could possibly go wrong that's worse than an already borked CAL when you try to restore it? Jul 09 16:07:36 the point is that we should figure out a way to not have a borked CAL Jul 09 16:07:50 but i don't think a block-by-block copy is the answer Jul 09 16:08:14 is CAL used somewhere else? Jul 09 16:08:25 because i don't understand how you could come up with something like that Jul 09 16:08:28 I *guess* when you write the dump to CAL and dd dump the CAL content after that and compare to original image, and they don't differ, I'd dare to say you won Jul 09 16:08:30 versus just a file on jffs2 Jul 09 16:08:59 again, it's legacy Jul 09 16:09:13 prolly used on all Nokia phones Jul 09 16:09:13 yeah but is it used on some dumber nokia phone? Jul 09 16:09:15 k Jul 09 16:11:07 we already discused and verified that a simple bindmount or the like should suffice to use a image file instead of a mtd partition for cal* in maemo Jul 09 16:11:37 except for flasher of course Jul 09 16:11:49 since that's not 'in maemo' Jul 09 16:12:05 (From Oksana) With such huge soldering iron and no flux, it is hopeless to do anything even with such large test pads.. Jul 09 16:12:24 oh, you can change the lock code and it'll get "appended" to your bindmounted file? Jul 09 16:12:32 and when you remove the bindmount the lock code is still the old one? Jul 09 16:12:37 neato Jul 09 16:12:46 yes, please don't ruin device by trying to solder on it with cheap tools and lack of decent flux Jul 09 16:12:50 it never will work Jul 09 16:13:35 kerio: yes Jul 09 16:14:07 DocScrutinizer05 would probably like louis rossmann Jul 09 16:14:12 Oksana says she can only hope the camera is still okay. (I say we would maybe try to get access to proper tools at uni, or at a local store; the lack of flux immediately disappointed me in the beginning) Jul 09 16:14:30 I don't know that guy Jul 09 16:14:58 makes videos on youtube where he repairs mostly macbook motherboards Jul 09 16:15:03 yells at apple Jul 09 16:15:07 get *decent* flux, NO acid flux! Jul 09 16:15:08 yells at stupid repair shops Jul 09 16:15:39 cheers Jul 09 16:15:59 ~usbfix Jul 09 16:15:59 i guess usbfix is http://talk.maemo.org/showthread.php?t=75920 - and **NEVER** use epoxy (unless you want to seal your device for underwater), or https://www.youtube.com/watch?v=fYz5nIHH0iY#t=1866, you will basically need two irons: a small good one (or better hot-air reflow) and a 60+ Watt Jul 09 16:16:04 see the video Jul 09 16:16:32 became famous after yelling at another (very popular) youtuber for making a video where a video card was "fixed" by putting it in an oven Jul 09 16:21:32 Gryllida: https://youtu.be/b9FC9fAlfQE?t=939 Jul 09 16:28:26 good video, shows how it works and the picture is not blurry, and has useful info. thank you Jul 09 16:33:53 https://youtu.be/b9FC9fAlfQE?t=1560 Jul 09 16:34:02 Gryllida: yw Jul 09 16:34:15 no learning without competent info source Jul 09 16:36:41 (From Oksana) Curious, what this problem is caused by, a lot of people must be using kernel power 53, and many of them have probably upgraded from kernel power 51. So what exactly was the problem with upgrding to kernel power 53? Jul 09 16:38:34 you have multiboot installed Jul 09 16:40:25 https://youtu.be/b9FC9fAlfQE?t=1850 Jul 09 16:40:46 Gryllida: what KotCzarny said# Jul 09 16:40:54 ~maemo-multiboot Jul 09 16:40:54 i guess maemo-multiboot is deprecated, and a horrible hack. PROBLEMS WITH NITDROID/MULTIBOOT? reflash rootfs&kernel aka COMBINED Jul 09 16:41:42 [2016-07-09 Sat 14:53:13] this factoid been defined after literally dozens of failed efforts to recover from a broken multiboot (multiboot is NOT uBoot!) Jul 09 16:41:43 [2016-07-09 Sat 14:54:03] multiboot basically reflashes kernel on *every* boot, which is.. well what the factoid says: a *horrible* hack Jul 09 16:43:25 ok Jul 09 16:50:10 https://youtu.be/b9FC9fAlfQE?t=2211 Jul 09 16:50:31 take care! you NEED a decent regulated soldering iron Jul 09 16:52:09 soldering has its times Jul 09 16:52:15 just not today Jul 09 16:52:42 tried to made bridging over two pairs of pads Jul 09 16:53:06 problem was that area was like 1mm x 2mm at max Jul 09 16:53:26 needless to say i failed miserably Jul 09 16:53:46 should have sticked to graphite instead... Jul 09 16:54:52 and what i was doing? Tried to fix Nokia 5130 "no sound from headset" problem Jul 09 16:56:11 Gryllida: https://youtu.be/J5Sb21qbpEQ?t=91 Jul 09 16:57:20 thank you most kindly; I was expecting this to be easy, but the soldering material isn't sticky enough, and I should definitely know better Jul 09 16:57:59 I will watch all this when I'm a bit more awake, with Oksana, and plan from there Jul 09 17:00:18 Gryllida: good plan :-) you're welcome Jul 09 17:01:02 honestly no expert can do a decent job with cheesy tools. You need a certain minimum that's listed in last URL Jul 09 17:01:45 regulated soldering iron, FLUX!!!, solder wick. Then you're already almost there Jul 09 17:02:23 hot air comes in very handy when you also got solder paste etc Jul 09 17:04:02 and decent solder wire of course Jul 09 17:05:07 I think in another few years we will be able to open a phone fix company Jul 09 17:05:13 https://youtu.be/J5Sb21qbpEQ?t=686 Jul 09 17:05:25 already have a lot of screwdrivers, internet to buy replacement parts, now soldering iron and related tools, ... what else? :) Jul 09 17:06:12 a pretty logo and a pretty site + some time for flying a cheaper beta program to gain experience Jul 09 17:06:14 or something Jul 09 17:06:34 (or some course... whatever these guys usually begin from) Jul 09 17:09:16 DocScrutinizer05: Phoenix boots N900 into TEST mode and uses softupd daemon and it just update cal keys, but do not support dump/restore Jul 09 17:10:21 I didn't think it supports dump/restore Jul 09 17:10:48 dump/restore is a basically alien concept to CAL Jul 09 17:26:31 however I wonder how Phoenix does e.g. ALS CAL calibration via softupd, when maemo has no tool to write CAL records Jul 09 17:27:03 must be raw NAND partition access then, no? Jul 09 17:28:04 unless flasher has some secrete function to update CAL records Jul 09 17:29:03 or Phoenix uploads a executable to RAM and runs it Jul 09 17:31:01 flasher indeed also updates CAL records, it notes the software version there Jul 09 17:31:30 so I wonder how flasher and phoenix do this magic Jul 09 17:32:21 BME also seems to write to CAL Jul 09 17:32:55 RE the softupd? Jul 09 17:37:32 what for? Jul 09 17:38:05 it's way easier to augment calvaria to learn to write records Jul 09 17:38:56 the only unclear (unspecified) behavior in CAL is when it's getting full Jul 09 17:39:55 until then new records are simply appended and supersede any prior records for same key Jul 09 17:40:19 iirc Jul 09 17:41:30 iirc that's also how bad NAND blocks get handled. Simply read back the record you just wrote, and when that fails simply write it once more Jul 09 17:43:17 note that writing to erased NAND is simple. Just erasing NAND is not as simple Jul 09 17:45:40 you only can erase complete erase pages (set all bits to e.g. zero) and then you can set each individual bit to one at arbotrary times Jul 09 17:47:06 that's btw what TRIM is all about Jul 09 17:47:27 Pali: does powerkernel know TRIM/discrad? Jul 09 17:48:01 I noticed all our flash storage isn't mounted with -o discard Jul 09 17:48:37 this would make for a *huge* improvement, also and particularly for swap Jul 09 17:50:01 IIRC ext3 does not support discard Jul 09 17:50:25 it does Jul 09 17:50:48 https://wiki.ubuntuusers.de/SSD/TRIM/ Jul 09 17:51:32 ext2, ext3 :: Kernel 2.6.38 Jul 09 17:52:09 but that is batched mode Jul 09 17:52:11 not "discard" Jul 09 17:52:13 aah yes Jul 09 17:52:21 just noticed that Jul 09 17:52:47 fstrim is using that batched mode Jul 09 17:53:05 I didn't find fstrim for maemo Jul 09 17:53:06 but anyway, everything needs new kernel Jul 09 17:53:13 nothing for 2.6.28 Jul 09 17:53:19 yep Jul 09 17:53:33 that's why I asked about powerkernel Jul 09 17:53:42 nothing Jul 09 17:54:11 only battery modules were backported and all is same as upstream kernel Jul 09 17:55:15 worthwhile to ponder how to get TRIM support Jul 09 17:57:14 for swap it's important to make swap page size == flash erase page size Jul 09 17:57:18 and align Jul 09 18:15:57 for the interested audience: http://www.thessdreview.com/daily-news/latest-buzz/garbage-collection-and-trim-in-ssds-explained-an-ssd-primer/ Jul 09 18:18:30 cute, ty Jul 09 18:19:04 DocScrutinizer05: also checkout fstrim Jul 09 18:19:12 fstrim is used on a mounted filesystem to discard (or "trim") blocks Jul 09 18:19:13 which are not in use by the filesystem. This is useful for solid-state Jul 09 18:19:13 drives (SSDs) and thinly-provisioned storage. Jul 09 18:23:19 I know :-9 Jul 09 18:23:24 :-) even Jul 09 18:24:14 the key statement: >>When the OS deletes the file, it simply marks the space used for that data as free in its logical data table. With HDDs, the OS does not need to tell the storage device anything about the deletion because it would simply write something new into that same physical location in the future. In the case of an SSD, it only becomes aware that the data is deleted (or invalid) when the OS tries to write to that location Jul 09 18:24:15 again<< Jul 09 18:26:29 and >>The advantage of the TRIM command is that it enables the SSD’s GC to skip the invalid data rather than moving it, thus saving time not rewriting the invalid data. This results in a reduction of the number of erase cycles on the flash memory and enables higher performance during writes<< Jul 09 18:38:24 https://www.youtube.com/watch?v=x6lqYU4j7no Jul 09 18:43:55 * DocScrutinizer05 wonders how hard it could be to implement fstrim-alike brute force TRIM to a range of eMMC (read: swap) without any kernel support Jul 09 18:45:08 reswap only makes true sense with TRIM of the deactivated and depleted swap partition Jul 09 18:46:23 this whole TRIM issue also suggests an explanation why devices slow down when you have much data on eMMC Jul 09 18:46:37 the eMMC lacks overprovisioning then Jul 09 18:46:59 which significantly burdens the GC Jul 09 18:52:01 hmmm, re implementing batched TRIM, it would need a way to talk directly to the storage bus, which is probably only allowed to the kernel driver Jul 09 18:52:46 so backporting the according IOCTL should suffice for that Jul 09 18:56:47 >> * This program uses FITRIM ioctl to discard parts or the whole filesystem * online (mounted). You can specify range (start and length) to be * discarded, or simply discard whole filesystem.<< https://github.com/karelzak/util-linux/blob/master/sys-utils/fstrim.c Jul 09 19:11:17 hmmm http://lxr.free-electrons.com/source/drivers/mmc/core/core.c#L2052 dunno if I wound the right runway to land Jul 09 19:12:51 anyway I guess backporting the low level mmc driver can't be too hard since at least the API shouldn't have changed since lernel 2.6.xx Jul 09 19:13:00 kernel even Jul 09 19:15:24 http://lxr.free-electrons.com/source/drivers/mmc/core/core.c#L2198 Jul 09 19:57:40 Pali: whenever you boot your mainline linux kernel, can you see if mounting a partition with discard actually works as intended? Jul 09 19:58:10 do not know, I have never used discard Jul 09 19:59:12 just mount a partition with -o discard Jul 09 19:59:17 idk how to test if it worked tho Jul 09 19:59:33 I'm not going to test any discard on any machine Jul 09 19:59:37 :c Jul 09 19:59:55 it is buggy and unsupported on lot of places Jul 09 20:04:33 ok Jul 09 21:28:52 moin ;) Jul 10 00:59:51 whats the easiest way to check i was able to move swap to sd successfully other than pulling the back cover and see if it messes up? Jul 10 01:04:11 iotop Jul 10 01:05:05 cat /proc/swaps Jul 10 01:05:47 Linkandzelda: ^^^ Jul 10 01:05:52 hmm Jul 10 01:06:14 its showing /dev/mmcblk1p1 with priority -1 and /dev/mmcblk0p3 with priority 0 Jul 10 01:06:45 that looks pretty good Jul 10 01:07:12 seems like they are both in use, though 80mb used on the emmc and 3.6mb used on microsd Jul 10 01:08:49 swapoff /dev/mmcblk0p3 Jul 10 01:09:45 that did it, guess i should run that on startup Jul 10 01:09:47 alas the standard messybox swapon doesn't know priorities Jul 10 01:12:38 otherwise swapon -p5 /dev/mmcblk1p1 Jul 10 01:12:57 i added the swapoff to a startup script, rebooting now to see the result Jul 10 01:14:59 for w/e reason that didnt run as expected Jul 10 01:18:12 ok it works with a sleep, thanks DocScrutinizer05 Jul 10 01:21:19 * DocScrutinizer05 hmmms at gconftool -g /system/osso/af/mmc-swap Jul 10 01:26:57 Linkandzelda: less +'/swaps=`blkid' /etc/event.d/rcS-late Jul 10 01:27:30 less /etc/fstab : Jul 10 01:27:40 # This file was generated by script /etc/event.d/rcS-late Jul 10 01:27:41