| 00:00.42 | *** join/#edev flavioribeiro (~flavio@li20-198.members.linode.com) |
| 00:01.33 | *** mode/#edev [+o Mirell] by ChanServ |
| 00:01.42 | *** mode/#edev [+q flyback!*@*] by Mirell |
| 00:02.55 | Mirell | prpplague: Give me some debugging tips on proving that the timer on a system is horked. |
| 00:04.16 | prpplague | Mirell: is it the primary timer that is used for jiffies? |
| 00:04.24 | prpplague | Mirell: or a secondary timer? |
| 00:06.17 | SpeedEvil | Oh! /me realises someone here may know. |
| 00:06.40 | SpeedEvil | Anyone aware of any projects to break down power consumption more? |
| 00:07.01 | SpeedEvil | For example - per-process times at various CPU frequencies? |
| 00:07.14 | SpeedEvil | Or interrupt timings. |
| 00:08.34 | SpeedEvil | For example - over 10 seconds there are 45 hw interrupts. If there are 20 in 1ms - they are pretty cheap, and can be ignored compared to the other 20 that are spaced out by .5s and stop the system goind to deepest selep |
| 00:10.52 | prpplague | dave says don't make me laugh - http://imagebin.ca/view/R4Hh6F.html |
| 00:14.40 | *** part/#edev flavioribeiro (~flavio@li20-198.members.linode.com) |
| 00:14.44 | *** join/#edev flavioribeiro (~flavio@li20-198.members.linode.com) |
| 00:20.55 | prpplague | wonders if Mirell ran away |
| 00:33.23 | Mirell | prpplague: Primary. |
| 00:33.31 | Mirell | Sorry, was reading documentation. |
| 00:33.50 | prpplague | Mirell: np, is it booting the kernel? |
| 00:34.05 | Mirell | Basically, this system is periodically just losing time according to the logs, because I see NTPD go crazy and adjust 90 seconds every three minutes for an hour. |
| 00:34.18 | prpplague | Mirell: ahh |
| 00:34.23 | Mirell | And I want to figure out some sort of test to prove/disprove a faulty timer. |
| 00:34.40 | Mirell | I'm pretty sure it's hardware, but I don't know how to pinpoint that. |
| 00:35.01 | prpplague | Mirell: you can all the jiffies calculation function at random intervals and report the value to the logs |
| 00:36.02 | *** join/#edev calculus (~calculus@gentoo/user/calculus) |
| 00:36.24 | SpeedEvil | I have a sort of prototype thing that polls /proc/*/stat every 5s |
| 00:36.40 | SpeedEvil | that would tell you if it's a process doing it |
| 00:37.48 | Mirell | Hmm. |
| 00:38.55 | prpplague | Mirell: the jiffies calc routine should come out almost identical every time it is called |
| 00:39.34 | prpplague | Mirell: you can do a basic test by rebooting the device about 20 times and checking the jiffies value |
| 00:39.52 | prpplague | Mirell: if it varies more that +- 20 points, you have an issue |
| 00:40.43 | prpplague | Mirell: some systems like the omap3 series use a 32khz clock to determine the operating freq, and if the 32khz clock is unstable can calculate a bogus value |
| 00:41.38 | prpplague | Mirell: can you tell me which SoC it is? |
| 00:41.49 | SpeedEvil | Also |
| 00:42.00 | Mirell | prpplague: It's actually not embedded, just an Intel Core 2 Duo. |
| 00:42.06 | SpeedEvil | Is it 90s+-13s, or is it 93.844 seconds |
| 00:42.09 | SpeedEvil | (for example) |
| 00:42.10 | prpplague | Mirell: oh |
| 00:42.46 | Mirell | SpeedEvil: adjusting local clock by 69.638914s |
| 00:42.48 | Mirell | For example. |
| 00:43.26 | prpplague | Mirell: yea for that i would definely recalc the jiffes |
| 00:43.44 | prpplague | Mirell: if it isn't dead on you have some sort of clock issue |
| 00:44.54 | Mirell | prpplague: Have an example? |
| 00:46.28 | prpplague | Mirell: can you pastebin a bootlog/dmesg? |
| 01:02.53 | *** join/#edev bgamari_ (~ben@physnat56.physics.umass.edu) |
| 01:03.35 | prpplague | bgamari_: hey bud |
| 01:03.47 | prpplague | bgamari_: how goes your beagle hacking? |
| 01:04.01 | bgamari_ | Things have been a bit slow recently |
| 01:04.18 | *** join/#edev SpeedEvil (1000@tor/regular/SpeedEvil) |
| 01:04.30 | bgamari_ | Need to produce some data so the hardware has been on the back burner for a while |
| 01:04.54 | bgamari_ | I think I finally have the beagle-daq design finalized now, however |
| 01:05.01 | bgamari_ | I've been meaning to order the PCBs |
| 01:05.23 | bgamari_ | Most of the silicon is on order at Newark |
| 01:07.34 | bgamari_ | I've been trying to keep the noise down |
| 01:26.17 | bgamari_ | Is the openembedded's gcc recipe supposed to produce a full gcc toolchain for the target? |
| 01:47.24 | Mirell | Oh. |
| 01:47.25 | Mirell | Huh. |
| 01:47.27 | Mirell | Duh. |
| 01:47.39 | Mirell | cat /proc/cpuinfo | grep bogo && cat /proc/cpuinfo | grep "MHz" |
| 01:47.39 | Mirell | bogomips : 4388.94 |
| 01:47.39 | Mirell | bogomips : 4486.02 |
| 01:47.39 | Mirell | cpu MHz : 800.000 |
| 01:47.46 | Mirell | Deeeerrr. Right. Screwed up. |
| 02:06.51 | solar | Mirell: try http://dev.gentoo.org/~vapier/vpenis |
| 02:10.52 | *** join/#edev miromiro (~miromiro@unaffiliated/miromiro) |
| 02:10.52 | *** join/#edev bsdfox (~bsdfox@unaffiliated/bsdfox) |
| 02:10.52 | *** join/#edev Ian_Daniher (~it@nat-pool-128-107.olin.edu) |
| 02:12.31 | Mirell | Heh. |
| 02:12.43 | Mirell | But the main point was, the timer between the two cores is whacked. |
| 02:15.22 | solar | oh heh. yeah |
| 02:23.21 | *** join/#edev djerome (~djerome@ip24-251-158-189.ph.ph.cox.net) |
| 02:23.37 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 02:49.25 | *** join/#edev roxfan (dunno@84.184-66-87.adsl-dyn.isp.belgacom.be) |
| 04:13.55 | *** join/#edev mbuf (~user@118.102.130.6) |
| 05:43.29 | *** join/#edev pytey (~pytey@howler.blatero.com) |
| 05:57.25 | *** part/#edev djerome (~djerome@ip24-251-158-189.ph.ph.cox.net) |
| 06:59.30 | *** join/#edev linac (~lin@121.11.255.19) |
| 07:05.07 | ohsix | http://hackaday.com/2010/06/22/ti-makes-a-big-bid-for-the-hobby-market/ |
| 10:01.50 | *** join/#edev kergoth_ (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 10:14.21 | *** join/#edev davidc__ (~davidc__@S01060002b360aacd.vc.shawcable.net) |
| 10:17.30 | *** join/#edev pirho (debian-tor@gateway/tor-sasl/pirho) |
| 11:23.05 | *** join/#edev btraynor (~btraynor@gromit.mixdown.ca) |
| 11:25.41 | *** part/#edev btraynor (~btraynor@gromit.mixdown.ca) |
| 12:04.35 | *** join/#edev gustavoz (~gustavoz@host247.201-252-51.telecom.net.ar) |
| 12:27.29 | *** join/#edev calculus (~calculus@gentoo/user/calculus) |
| 13:05.30 | *** join/#edev devilsadvocate (~devilsadv@202.3.77.208) |
| 13:44.26 | *** join/#edev sjhill (~sjhill@home.bethel-hill.org) |
| 13:44.26 | *** mode/#edev [+o sjhill] by ChanServ |
| 13:50.10 | *** join/#edev mbuf (~mbuf@124.123.137.162) |
| 13:52.46 | Mirell | The thing I find curious so far, is that on a default Linux install these days, /dev/rtc is rarely activated. |
| 14:04.56 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 14:09.24 | kergoth | morning |
| 14:11.34 | *** join/#edev bgamari_ (~ben@physnat56.physics.umass.edu) |
| 14:28.40 | *** join/#edev linac (~lin@new1-26-153-a8.bta.net.cn) |
| 14:36.51 | *** join/#edev sjhill (~sjhill@home.bethel-hill.org) |
| 14:36.51 | *** mode/#edev [+o sjhill] by ChanServ |
| 14:39.11 | *** join/#edev sjhill (~sjhill@home.bethel-hill.org) |
| 14:39.11 | *** mode/#edev [+o sjhill] by ChanServ |
| 14:41.37 | *** join/#edev pirho (debian-tor@gateway/tor-sasl/pirho) |
| 14:42.51 | *** join/#edev sjhill (~sjhill@home.bethel-hill.org) |
| 14:42.51 | *** mode/#edev [+o sjhill] by ChanServ |
| 14:46.40 | *** join/#edev prpplague (~danders@nat/ti/x-mtrlqqbtaupwcscs) |
| 14:58.58 | *** join/#edev CosmicPenguin (~nobody@207-114-132-30.static.twtelecom.net) |
| 14:59.00 | *** mode/#edev [+o CosmicPenguin] by ChanServ |
| 15:00.22 | prpplague | CosmicPenguin: hey bud |
| 15:02.28 | Mirell | prpplague: So I figured it out. |
| 15:02.37 | prpplague | Mirell: do tell |
| 15:02.41 | Mirell | prpplague: First Core: bogomips : 4388.94 |
| 15:02.50 | Mirell | prpplague: Second Core: bogomips : 4486.02 |
| 15:03.06 | prpplague | ahh multicore |
| 15:03.20 | CosmicPenguin | I didn't do it |
| 15:03.30 | prpplague | CosmicPenguin: we have video |
| 15:04.07 | CosmicPenguin | That could be anybody's ass |
| 15:04.37 | prpplague | hehe |
| 15:05.01 | Mirell | I really don't think most people have a birthmark that looks like Diana Ross |
| 15:12.11 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 15:13.24 | prpplague | Mirell: i thought it looked more like Mahmoud Ahmadinejad |
| 15:47.58 | chouimat | morning |
| 15:52.52 | *** join/#edev devilsadvocate (~devilsadv@202.3.77.231) |
| 15:56.06 | CosmicPenguin | Oh, wow - Italy just got eliminated |
| 15:56.48 | chouimat | cool |
| 15:58.14 | *** join/#edev landley (~landley@m365636d0.tmodns.net) |
| 15:58.14 | *** mode/#edev [+o landley] by ChanServ |
| 16:04.17 | SpeedEvil | On no! |
| 16:04.24 | SpeedEvil | Where will I get my olives from! |
| 16:04.34 | CosmicPenguin | Uh - Greece? |
| 16:09.42 | *** join/#edev wmat (~wmat@CPE001b1168d6ea-CM00111adf5f4a.cpe.net.cable.rogers.com) |
| 16:51.46 | *** join/#edev mbuf (~mbuf@124.123.137.162) |
| 16:52.09 | *** part/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 16:52.29 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 16:53.32 | *** join/#edev gustavoz (~gustavoz@host247.201-252-51.telecom.net.ar) |
| 16:58.31 | prpplague | Mirell: so what did you do to fix the issue? |
| 17:05.11 | penguinece | Using a GCC ARM cross compiler, my own linker script (gnu ld). My device has ROM at 0x08000000 and RAM at 0x20000000, a difference of about 402MB. When I objcopy an executable from ELF to binary, I get a huge (~403MB) file as a result. Nothing is actually written to RAM upon programming (data is copied there on executable startup). Is there a way to let the symbols exist without creating 402MB of zeros in my binary? |
| 17:05.32 | penguinece | I'm thinking its a gnu ld option that I'm missing, or not using correctly |
| 17:05.48 | penguinece | but I'm already a bit scared of my linker script, so... |
| 17:06.33 | evilmoo | penguince: "-j text"? |
| 17:06.57 | penguinece | evilmoo, that flag with what command? |
| 17:07.07 | penguinece | objcopy? |
| 17:07.08 | evilmoo | objcopy |
| 17:07.52 | evilmoo | er "-j .text" |
| 17:07.55 | penguinece | I still need the symbols to point to those addresses, though |
| 17:08.19 | penguinece | hmm |
| 17:08.23 | penguinece | I will try it though, thanks |
| 17:08.27 | evilmoo | I don't understand your point |
| 17:08.53 | penguinece | the startup script that copies the data into RAM (and zero fills bss) uses those symbols to write to RAM |
| 17:09.07 | penguinece | that is, to know where to write in RAM |
| 17:09.45 | SpeedEvil | If you want an unformatted binary - without any relocation information like elf has - it cannot have segments or other portions of the file offset by 400 meg - without having 400 meg of zeros in. |
| 17:09.54 | SpeedEvil | (modulo sparse filesystems) |
| 17:10.16 | evilmoo | penguince: is the end result here a ROM, or an ELF file? |
| 17:12.13 | penguinece | evilmoo, the end result is a binary that is being packed into a USB DFU file |
| 17:12.48 | evilmoo | maybe you should put your linker script on pastebin so we can see it |
| 17:12.52 | penguinece | SpeedEvil, If ignoring the RAM section doesn't work, I'm thinking I might just end up doing 1 ELF into 2 separate BINs |
| 17:14.22 | penguinece | http://pastebin.com/Kmj6CxkM |
| 17:15.56 | evilmoo | so out of all of this, you want just flash, or option_bytes_rom + flash? |
| 17:17.03 | penguinece | just flash. but there is code in flash that references symbols with addresses in RAM |
| 17:17.19 | evilmoo | if the code is compiled, and your loader initializes RAM for you, then symbols are irrelevant |
| 17:17.59 | penguinece | okay, I'll give it a try -j'ing just the flash sections |
| 17:18.03 | penguinece | thansk |
| 17:20.56 | evilmoo | i think what you want is |
| 17:21.20 | evilmoo | objcopy -O binary -j text-section input.elf.file output.flash.file |
| 17:21.55 | penguinece | I did: |
| 17:21.57 | penguinece | arm-eabi-objcopy -O binary -j .isr_vector -j .text -j .init -j .fini -j .rodata arp.eabi arp.bin |
| 17:22.18 | penguinece | those are all the sections with VMA and LMA at 0800xxxx |
| 17:32.47 | evilmoo | success? failure? fiery ruin? |
| 17:36.53 | penguinece | arm-eabi-objcopy -O elf32-littlearm -j .isr_vector -j .text -j .init -j .fini -j .rodata -j .data -j .init_array -j .fini_array -j .jcr -j .bss arp.eabi arp.elf <---- produces working executable (flashed via jtag) |
| 17:37.25 | penguinece | arm-eabi-objcopy -O binary -j .isr_vector -j .text -j .init -j .fini -j .rodata -j .data -j .init_array -j .fini_array -j .jcr -j .bss arp.eabi arp.bin <---- produces stat size 25864 byte binary |
| 17:37.29 | penguinece | so, I think success |
| 17:37.30 | penguinece | I hope. |
| 17:38.03 | penguinece | -j'd all sections with LMA in 0x0800xxxx, though some of them had VMA 0x2000xxxx |
| 17:38.13 | penguinece | so... I think I did what I wanted |
| 17:38.16 | evilmoo | you can also use -R to remove sections |
| 17:38.33 | evilmoo | which might make your statement shorter |
| 17:38.34 | penguinece | haha, that would probably have been way more efficient |
| 17:38.36 | penguinece | yeah |
| 17:39.09 | penguinece | thanks, I hope. now to continue writing the DFU wrapper |
| 17:39.20 | evilmoo | :) |
| 17:47.24 | landley | What's the difference between elf2flt and objcopy -o binary? |
| 17:47.44 | evilmoo | dunno, never heard of elf2flt |
| 18:17.12 | *** join/#edev roxfan (dunno@91.180.243.205) |
| 18:25.18 | landley | It makes execute in place binflat files. |
| 18:27.40 | evilmoo | interesting |
| 18:27.53 | evilmoo | guess i need to do more embedded work :) |
| 18:32.20 | Mirell | prpplague: It was hardware. |
| 18:35.05 | Mirell | landley: So, here's an odd question. How exactly would you test an HPET? |
| 18:35.29 | *** join/#edev flo_lap (~fuchs@Maemo/community/contributor/florian) |
| 18:35.38 | landley | Test it in what way? |
| 18:35.47 | landley | Accuracy? |
| 18:35.56 | landley | Granularity? |
| 18:36.04 | penguinece | flavor. |
| 18:36.06 | landley | Wakeup latency? |
| 18:37.13 | landley | evilmoo: it lets linux mmap flash and run out of it without copying pages to dram. |
| 18:38.31 | Mirell | landley: Accuracy |
| 18:39.11 | landley | You'd need to check it against an external clock. |
| 18:40.10 | landley | Basically you reinvwnt ntpd, or check two clocks in the same machine against each other. |
| 18:40.11 | landley | reinvent |
| 18:40.34 | Mirell | Ah, right, get two instances of a /dev/rtc.. |
| 18:40.44 | landley | yup |
| 18:41.37 | landley | I note that half the reason for GPS is to accurately set your clock. |
| 18:41.59 | landley | To the nanosecond. |
| 18:43.28 | Mirell | Yah, that always amused me about GPS. |
| 18:43.49 | Mirell | All right, to keep time, we have to launch satellites. Wait... |
| 18:44.34 | penguinece | in some places, you could always be a rebel and use LORAN |
| 18:46.00 | Mirell | Ah yes. |
| 18:46.14 | Mirell | What is the radio frequency based atomic clock one.. |
| 18:47.17 | penguinece | WWV |
| 18:47.34 | penguinece | though they're no LORAN |
| 18:47.53 | penguinece | or GPS |
| 19:22.37 | sjhill | Mirell: GPS was designed for global, all-weather time synchronization and positioning |
| 19:23.14 | sjhill | for military i.e. battlefield, so your point is probably moot |
| 19:24.00 | sjhill | prpplague: never saw the url for the repository, no hurry, i dropped off IRC for a bit |
| 19:26.24 | prpplague | sjhill: hmm, i thought i pasted it, but let me check |
| 19:27.54 | prpplague | sjhill: git://git.omapzoom.org/kernel/omap.git |
| 19:32.49 | sjhill | thx |
| 19:35.12 | prpplague | np |
| 19:38.35 | chouimat | wants to useless conference call to be over soon |
| 19:38.42 | evilmoo | hang up |
| 19:38.50 | chouimat | can't ... |
| 19:39.30 | evilmoo | :( |
| 19:43.44 | *** join/#edev calculus (~calculus@gentoo/user/calculus) |
| 19:51.37 | landley | Mirell: gps works by timing signal propogation delays for signals traveling at the speed of light. |
| 19:51.54 | landley | This requires accurate clocks. |
| 19:54.44 | landley | The sattelites just go "ping" at fixed intervals, you receiver figures out how late each ping is and converts that to a speed of light distance, then solves for a point where you could be that far away from each sattelite. |
| 19:54.59 | landley | The one in deep space isn't it. |
| 20:45.59 | SpeedEvil | you only get a solution in space for three or fewer sats IIRC |
| 20:46.14 | SpeedEvil | four |
| 21:06.17 | *** join/#edev cablewhore69 (~dev@c-71-206-95-254.hsd1.fl.comcast.net) |
| 21:07.53 | cablewhore69 | Can I add my own target arch? I need an armv6b instead of armv6l. is there more to it than changing the one config file's 'endian' field and building from source? |
| 21:12.40 | cablewhore69 | {taps on mic} "tap tap tap.... is this thing on?" |
| 21:29.25 | Sphere_ | hello |
| 21:29.34 | Sphere_ | your mic is working |
| 21:31.01 | roxfan | add your own arch to what? |
| 21:37.38 | *** join/#edev prpplague (~Dave@ppp-70-249-145-66.dsl.rcsntx.swbell.net) |
| 21:38.02 | *** join/#edev SpeedEvil (~user@tor/regular/SpeedEvil) |
| 22:08.45 | *** join/#edev landley (~landley@m465636d0.tmodns.net) |
| 22:08.45 | *** mode/#edev [+o landley] by ChanServ |
| 22:13.28 | *** join/#edev davidc__ (~davidc__@S01060002b360aacd.vc.shawcable.net) |
| 22:15.28 | *** join/#edev davidc__ (~davidc__@S01060002b360aacd.vc.shawcable.net) |
| 22:39.42 | *** join/#edev SpeedEvil (~user@tor/regular/SpeedEvil) |
| 22:59.11 | *** join/#edev calculus (~calculus@gentoo/user/calculus) |
| 23:07.00 | *** join/#edev penguinece (~penguinec@magritte.eece.maine.edu) |
| 23:47.18 | *** join/#edev flyback- (~butyl@c-98-219-129-239.hsd1.pa.comcast.net) |
| 23:47.22 | flyback- | woot |
| 23:47.31 | flyback- | I love my dual opteron esxi box |
| 23:47.46 | flyback- | soon as its tested I can start building cross compiler vm's :O |
| 23:52.48 | *** join/#edev kergoth_ (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 23:55.13 | *** mode/#edev [+q flyback-!*@*] by Mirell |
| 23:56.01 | Mirell | Hmm.. |
| 23:57.44 | *** join/#edev SpeedEvil (~user@tor/regular/SpeedEvil) |