**** BEGIN LOGGING AT Wed Jul 27 02:59:58 2016 Jul 27 11:29:54 hi Jul 27 11:30:22 i'm a cs student Jul 27 11:30:37 i'm currently learning a lot about os design/development Jul 27 11:30:46 and would try to build my own (simple) os Jul 27 11:31:05 i'm thinking of doing it on a sbc because of fixed hardware Jul 27 11:31:14 bon giorno Jul 27 11:31:24 have you read tannenbaum's book? Jul 27 11:31:27 yup Jul 27 11:31:29 good Jul 27 11:31:37 nice Jul 27 11:32:02 is the bbb a good sbc to build an os from scatch? Jul 27 11:32:09 now get the OS kit and start building that Jul 27 11:32:34 https://www.cs.utah.edu/flux/oskit/ Jul 27 11:33:25 dead link? Jul 27 11:33:34 nope, it works here Jul 27 11:33:51 however Jul 27 11:34:12 would you think the bbb is a good scb for this purpose Jul 27 11:35:32 well, depends what you want to do Jul 27 11:35:47 you will have to write quite a few drivers yourself, as the hardware is rather complex Jul 27 11:36:01 mh Jul 27 11:36:07 something like an cortex-m3 based system would be simpler, but less powerful of course Jul 27 11:36:28 and cortex-m might be too slow for many OS concepts described in Tanenbaum Jul 27 11:37:05 lol Jul 27 11:37:34 are you referring to usb controllers etc Jul 27 11:37:34 you know, when tannenbaums book was written that the clock speed of the average cpu was in the order of 1-2MHz and an addition took 2-10 clock cycles? Jul 27 11:37:54 for the first edition, sure Jul 27 11:37:58 minix 3 Jul 27 11:38:00 is rather new Jul 27 11:38:05 an cortex-m3 is about 1000 times faster than that Jul 27 11:38:19 well, minix3 is supposed to be run on a pc Jul 27 11:38:26 that's a different beast entirely Jul 27 11:38:31 koth what drivers were you referring to Jul 27 11:38:41 drivers for everything Jul 27 11:38:46 you are writing your own os Jul 27 11:38:50 traf: start with uart Jul 27 11:38:52 http://wiki.osdev.org/Main_Page has a lot of nice ideas too Jul 27 11:38:52 you will need to write each driver yourself Jul 27 11:38:56 each and everyone Jul 27 11:39:51 traf: also note that a lot of stuff isnt written in tannenbaums book Jul 27 11:40:02 traf: stuff that you will run into when actually writing an OS Jul 27 11:40:11 mh Jul 27 11:40:22 traf: also, the very idea of an high performance microkernel is fundamentally flawed Jul 27 11:41:13 (a high performance OS kernel will never be an microkernel, and a microkernel will never be high performance) Jul 27 11:41:22 thinkfat: what's uart about (can't look for it right now) Jul 27 11:41:47 KotH: assuming we are talking about the hardware we know these days. Jul 27 11:41:59 LetoThe2nd: it's independent of hardware Jul 27 11:42:24 KotH: i don't think so. Jul 27 11:42:34 LetoThe2nd: the problem lies in the fact that the process scheduler needs detailed insight into what the other subsystems (memroy manager, communication manager, etc) do to perform well Jul 27 11:42:48 KotH: well i'm not committed to a microkernel architecture Jul 27 11:42:56 would be interesting to implement one nontheless Jul 27 11:42:59 traf: btw: have a look at l4 Jul 27 11:43:12 traf: they have good literature on real life kernel and os development Jul 27 11:43:31 KotH: original minix was written for 8088 PC with 512KB RAM, but if you start looking into OS design _today_, why limit yourself by choice of hardware? Jul 27 11:43:53 traf: if you knwo what you are doing, you can implement an l4 derivate within a couple of days/weeks (plus writing the drivers) Jul 27 11:44:04 KotH: in a (however theoretical) world where a single-cycle-context-switch-scheduler-in-hardware could do that, its not fundamentally flawed. Jul 27 11:44:22 LetoThe2nd: yes, but which process do you switch to? Jul 27 11:44:26 the point is that its mostly orthogonal on current hardware architectures, agreed there. Jul 27 11:44:45 LetoThe2nd: the decision will be always sub-optimal if the scheduler doesnt have the information it needs to Jul 27 11:45:03 LetoThe2nd: and i'm not talking about a 1-2% hit, but something in the order of 20-50% Jul 27 11:45:29 KotH: and i am not talking about a real-world, general case solution. Jul 27 11:45:38 KotH: sorry maybe i'm a bit dump but "14 kernel" and "14 operating system" yield nothing on google Jul 27 11:45:40 LetoThe2nd: ah.. ok Jul 27 11:45:59 what should i look for Jul 27 11:46:07 KotH: i think it is *theoretically* possible to generate a high-performance microkernel system for a given, known usecase. Jul 27 11:46:22 LetoThe2nd: but then you do what all theoreticians do: limit yourself to one solvable problem that has very limited real impact inorder to have have a solvable problem Jul 27 11:46:45 traf: 1) read Jul 27 11:46:48 traf: 2) read Jul 27 11:46:52 traf: 3) ... Jul 27 11:46:54 traf: 4) profit Jul 27 11:46:58 KotH: you're right, but still it is the one example which suffices to kill any mathematical proof. Jul 27 11:47:12 LetoThe2nd: *g* Jul 27 11:47:37 yep i appreciate the tip but i didn't understand what to look for Jul 27 11:47:38 14 what Jul 27 11:47:47 traf: l4, not 14 Jul 27 11:47:55 lol, thanks Jul 27 11:47:58 KotH: i'll give you 'generally speaking, microkernel and high-performance are orthogonal' :) Jul 27 11:48:18 KotH: and one ping. exactly one ping :-) Jul 27 11:48:58 koth: what about gnu hurd, isn't that a microkernel Jul 27 11:49:03 LetoThe2nd: *ping* Jul 27 11:49:18 traf: hurd is a complete OS, not just a kernel Jul 27 11:49:32 microkernel based Jul 27 11:49:35 traf: and they started of with X, switched to mach, switched to l4... Jul 27 11:49:41 * LetoThe2nd once met a os dev who wrote a mirkokernel. Jul 27 11:49:59 LetoThe2nd: hasnt everyone done that once? Jul 27 11:50:22 KotH: re-read, re-think-re-interpret :-) Jul 27 11:51:10 lol Jul 27 12:41:49 hi, I would like to modify /boot/uEnv.txt. This file is just here to load capes on the bbb's start, a I right ? My bbb is 4.6.2-bone3 kernel, should I go to 3.8 example or 4.1 I'm missed Jul 27 12:44:14 + I want to configure the pins, can I comment this line cmdline=coherent_pool=1M quiet cape_universal=enable Jul 27 14:18:55 Hi People Jul 27 14:49:03 Hello Jul 27 14:58:50 KotH: ahh hurd is a kernel, with the gnu packages it could become a OS someday. but hurd is just a kernel. Jul 27 15:08:26 zykotick9: uhmm.. that's why they are switching from one ukernel to another every couple of years? Jul 27 15:09:43 KotH: timeline for Hurd https://xkcd.com/1508/ ;) Jul 27 15:09:56 :) Jul 27 17:18:41 Is there a reason why boot/uEnv.txt wont load my overlay, but i can load it manually and from /etc/default/capemgr? Jul 27 18:17:40 laurittr: assuming your overlay resides at /lib/firmware thats probably because the uEnv.txt is executed by the bootloader and at that time, the fs is not actually mounted yet, the overlay is not actually there. Jul 27 20:37:42 * ds2 shakes fist at DT again Jul 27 20:40:40 * zmatt at the posix api in its entirety Jul 27 20:41:34 why does it seem like every at first sight sane mechanism turns out to contain some violently obnoxious behaviour that apparently noone will fix due to backwards compat Jul 27 20:45:27 What's the recommended way to disable the PRU on the bbg? Jul 27 20:45:41 why would it be enabled? Jul 27 20:46:05 im trying to prevent the kernel from trying to load the fw on boot because its causing boot time slowdown Jul 27 20:46:55 decap chip, and laser it out Jul 27 20:47:40 Well i guess i should have just said: Anyone have a solution for fixing boot time error "Direct firmware load for am335x-pru0-fw failed with error -2" Jul 27 20:52:56 hammer_gaidin: dike it out of DT Jul 27 20:53:21 or remove the relevant kernel module(s), if it's compiled as modules Jul 27 20:54:01 or rebuild the kernel with custom config... you can save a LOT of boot time with that Jul 27 20:57:49 hammer_gaidin: In that instance it would be jettison remoteproc most likely. Jul 27 21:01:15 what can I delete to make some space on emmc. bbb memory is full since I switched to bone kernel. It seems some version kernel take more space than other. Jul 27 21:01:18 or put the board in a warp bubble Jul 27 21:04:57 Ragnorok: only specifically for pru, you still need it for the cortex-m3 if you want to use the more advanced power management states Jul 27 21:06:35 dirkk: it's hard to telepathically examine your bbb, but there's usually plenty of junk if you go hunting for it Jul 27 21:06:45 you've installed another kernel, have you also ever removed old ones? :P Jul 27 21:07:03 expand the filesys!? :P Jul 27 21:13:12 zmatt: BBB doesn't use a M3 does it? I know the original question was for a BBG, but some point I'll need some power management and I don't have remoteproc on the BBB. Jul 27 21:13:29 BBB and BBG is same thing Jul 27 21:13:42 the am335x has an integrated cortex-m3 in the wakeup power domain Jul 27 21:14:06 Urkle. Won't remoteproc hose uio-pruss? Jul 27 21:14:06 e.g. to manage suspend/resume Jul 27 21:14:53 nothing to do with each other, "remoteproc" is just for loading code onto a remote processor Jul 27 21:15:38 its use for the m3 (to load power management firmware) and its use for pruss (competitor to uio_pruss) have nothing to do with each other really Jul 27 21:15:48 Right. But doesn't it try to load the PRUs "for" me when it's loaded? Or is that bypassable? Jul 27 21:16:20 it's a different driver Jul 27 21:16:27 Cool. Thanks! Jul 27 21:16:27 it just uses the same framework Jul 27 21:16:53 zmatt: I did apt-get clean. How to know if there is still kernel that I don't use ? Jul 27 21:16:56 the m3 is CONFIG_WKUP_M3_RPROC Jul 27 21:17:11 I gotto go catch a train, I'll be back later Jul 27 21:17:18 l8r Jul 27 21:17:43 (and apt-get clean does not remove any packages whatsoever) **** ENDING LOGGING AT Thu Jul 28 02:59:58 2016