**** BEGIN LOGGING AT Sat Aug 01 02:59:58 2015 Aug 01 03:59:51 Hi, i'm new to beaglebone black Aug 01 04:01:21 I have just brought a kit(BBB). But when I plugin the usb and install the drivers, the board will automatically disconnect the usb-ethernet channel. Aug 01 04:01:31 anyone has any idea? Aug 01 04:03:03 I could connect to the BBB by usb-ethernet at first. But after a few second, it is disconnected. Aug 01 10:00:43 09:43:37<@jme> Update: We finally got the last db imported. There were a few issues with its backup, but I'm now replaying the binlogs. Aug 01 10:00:46 09:43:51<@jme> They're going much, much faster. Aug 01 10:00:57 elinux should be coming back soon then Aug 01 10:08:50 \o/ Aug 01 10:10:49 importing backups, replaying binlogs..... sounds like something unhealthy happened to the server? Aug 01 10:13:39 PEBKAC, AFAIU it was an operator mistake, compounded by some backup issues Aug 01 10:14:09 operator fumbles can indeed be unhealthy to a server Aug 01 10:30:19 #whois Aug 01 10:35:53 Billgle_: schnapp der en kaffi.. du bruchsch en Aug 01 10:40:44 KotH: caffe? Aug 01 10:44:56 mrpackethead2: nai.. kaffi Aug 01 10:45:06 mrpackethead2: nöd das grusige zügs won er trinked! Aug 01 10:45:52 :> Aug 01 10:51:15 咖啡？ Aug 01 10:51:41 Kāfēi? Aug 01 10:52:13 kawhe? Aug 01 12:17:40 funny, the combo of the am335x lcd controller and the on-board hdmi framer can in theory handle quite a few funky formats Aug 01 12:21:33 like yuv422 with 12-bit per component (36-bit color) at 720p/1080i or close to it Aug 01 12:29:13 or 8-bit per component, which would only occupy two of the three sets of pins { d1-d4, d7-d10, d12-d15 } (free choice which two) + pixel clock (and optionally hsync/vsync/oe depending on one's desire to mess with embedded sync and such) Aug 01 12:29:27 freeing up the remaining pins for other uses Aug 01 12:29:53 not sure if you can convince the linux framebuffer to produce it though, heh Aug 01 12:32:24 hdmi audio also occupies unnecessarily many pins... McASP can generate S/PDIF and the HDMI framer can accept it, freeing up the afsx and aclkx pins Aug 01 12:48:21 hi Aug 01 12:48:56 i want to add my companmy website backlink in this page (http://beagleboard.org/resources/) Aug 01 12:49:02 so how to do that? Aug 01 12:50:50 are you there? Aug 01 14:03:20 oh hey, this register map uses a mix of big- and little-endian... that's... nice Aug 01 17:03:21 using this flasher image (https://rcn-ee.com/rootfs/2015-07-08/flasher/BBB-eMMC-flasher-ubuntu-14.04.2-console-armhf-2015-07-08-2gb.img.xz) how can I enable networking over USB? Aug 01 18:53:56 is the eMMC flasher image different than the regular sd image? Aug 01 18:54:45 trying to flash eMMC and can't tell if it's working Aug 01 19:05:02 tlab: I think the difference is just in a file or two Aug 01 19:05:45 tlab: flashing should not take longer than 30min. if you have a connection to the debug UART you might see progress (don't know if current flasher sends output there) Aug 01 19:06:08 everything I find points to elinux.org, but it is down.. looks like there might be a sh file on the regular image Aug 01 19:06:10 20min actually is already generous. I'd expect 10-15 Aug 01 19:06:18 yes, there is in /opt Aug 01 19:06:39 yeah I see nothing special on the uart Aug 01 19:07:39 it does boot off the sd card tho, when it's inserted! Aug 01 19:07:41 I think the led pattern should change if it's flashing and when it's done Aug 01 19:08:12 yeah I've tried to flash with the image I find online, but just seems to do boot off sd card Aug 01 19:21:09 http://rcn-ee.net/rootfs/bb.org/release/2015-03-01/console/ Aug 01 19:21:20 flasher image, yay 8) Aug 01 19:34:21 yeah lots of information displayed when it's really flashing the eMMC, not just when it's a regular boot and *I think it's updating* Aug 01 20:07:44 I'm surprised there is no offical flasher image, just sd card image Aug 02 00:38:11 tlab: current images are in https://rcn-ee.com/rootfs/bb.org/testing/ Aug 02 00:38:16 (note: .com rather than .net ) Aug 02 00:45:47 matt: yeah I found that after some searching by accident. I only noticed it on someones web link Aug 02 00:46:06 the .com vs .net situation is really confusing Aug 02 00:46:22 the two dirs looks the same, except .net contains old images Aug 02 00:48:38 (the latest testing image is rarely more than a week old) Aug 02 01:30:33 I woulda personally hoped rcn wouldn't ADD to the existing mess and confusion of bbb images/documentation .. think we should lean on him a bit to fix that .. and risk the link breakage. Aug 02 01:32:07 dont' lean too hard Aug 02 01:32:11 he might run away. Aug 02 01:32:32 I doubt it ..hes pretty hardcore lol Aug 02 01:33:27 like the rest .. he wouldn't still be here if he wasn't lol Aug 02 01:34:50 And to be actively contributing .. I gotta take off my hat :p Aug 02 01:39:11 I noticed the latest dirs also have console images for, not just the bbx15, but also the omap5-uevm Aug 02 01:41:02 now that's hardcore :P Aug 02 01:41:27 pretty sure he's slightly insane lol Aug 02 01:41:35 nah Aug 02 01:41:56 or addicted .. or something lol Aug 02 01:42:49 I do wonder though how the people who wrote the C++ STL are coping mentally Aug 02 01:45:13 do they cope? are they mental?! lol Aug 02 01:48:46 I mean, sometimes with C++ I think "ahh wait I can make a class for this that hides the gory details behind pretty syntax sugar") and then my heart darkens as I realize what would be necessary to make it happen, that after writing that header file I'll be slightly less human than before, a small part will have died inside Aug 02 01:49:55 could be worse .. could be a .net addict .. :p Aug 02 01:50:19 then it's all fancy syntax sugar ;d Aug 02 01:51:26 c# is mind-numbing due to rigidity and repetitive boilerplate Aug 02 01:51:51 afaik it isn't depraved the same way c++ is Aug 02 01:52:00 its a fairly good n00b language tbh Aug 02 01:54:15 exactly... originally I just used c++ as c with some nice syntax sugar (struct inheritance is very handy) and metaprogramming capabilities... especially for embedded systems, anything that can be moved from runtime to compile-time is a win Aug 02 01:55:37 but it's a trap, before you know it you find yourself thinking with maniacal laughter "AHAHA I CAN SOLVE THIS WITH VARIADIC TEMPLATES!" Aug 02 01:58:57 yeah. Don't. lol Aug 02 02:02:50 I already did... I'm not sure I can be saved anymore Aug 02 02:05:10 ah well Aug 02 02:05:45 besides, SFINAE is in a way more horrid than variadic templates Aug 02 02:06:03 oook lol Aug 02 02:06:28 (either, or especially the combination, is also especially good at turning any typo inside an error message the size of the epic poem) Aug 02 02:06:34 *into Aug 02 02:07:49 I really wish there was a *good* low-level programming language Aug 02 02:09:36 I still need to post a plea on the clang and/or gcc mailing lists for getting at least some attributes to explain some of the restrictions of memory-mapped I/O to the optimizer Aug 02 02:09:51 assembler works :P Aug 02 02:09:59 since right now, all you have is 'volatile', which is an extremely blunt instrument Aug 02 02:10:08 not really Aug 02 02:10:28 its a tool .. learn to use it wisely :P lol Aug 02 02:10:53 call me weird, but I want readable code Aug 02 02:11:17 learn python?! rofl .. *duck* Aug 02 02:12:26 take for example an Ethernet DMA descriptor, which is rougly: struct Desc { Desc *next; void *buf; u16 buflen; u16 offset; u16 pktlen; u16 flags; }; Aug 02 02:13:31 with a minor problem: the SRAM in the ethernet subsystem meant specifically for the DMA descriptors doesn't support 8- or 16-bit writes. any such write will silently corrupt the rest of the word Aug 02 02:14:26 aren't words 16 bits? Aug 02 02:14:45 thurgood: back in the '80s yes Aug 02 02:15:08 ah.. that's about when I learned assembly Aug 02 02:15:13 on ARM, POWER, and any other modern architecture I know of a word is 32-bit Aug 02 02:15:30 (note: x86 isn't a modern architecture, it's just one being kept alive for way too long) Aug 02 02:17:58 x64 is :p Aug 02 02:18:18 but word length is defined as dependant on the architecture iirc Aug 02 02:18:34 veremit: so to fix this you have to replace the four u16 fields by two u32 fields (have fun naming them), and make all writes to them read-modify-update with the update a volatile write Aug 02 02:18:58 you fixing some driver , zmatt ? or targetting something obscure :P Aug 02 02:20:09 the ethernet subsystem I'm describing is TI's... afaik same problem in all versions thereof Aug 02 02:20:37 anyhow, but with 'volatile' you lose all possibility for optimization Aug 02 02:20:57 unless you do your own :p Aug 02 02:21:38 I thought ethernet Worked (tm) ... Aug 02 02:22:00 its not as bad as the Pi .. *Cough* lol Aug 02 02:22:03 right, at the sacrifice of maintainability.... it's idiotic to have to sabotage the optimizer and then subsequently have to do your own Aug 02 02:22:19 I ran into this long before I even had a BBB Aug 02 02:23:09 it continues to make a nice example because it's so frustrating: it's perfectly normal RAM, just with the restriction that only 32-bit writes or bigger are allowed Aug 02 02:23:22 even though the descriptor structure has 16-bit fields Aug 02 02:23:46 and to add insult to injury: the actual memory-mapped registers of the ethernet subsystem *do* support writes of any size Aug 02 02:24:24 turning writes into read-modify-update is something you really want to avoid Aug 02 02:24:39 considering a write is typically 1 cycle, and a read typically 150 cycles Aug 02 02:25:03 (when targeting a peripheral) Aug 02 02:27:15 and the linux cpsw driver *works* yes... it's pretty awful and inefficient code, and will crash if you're honest about the fact that you only use one of the two ethernet interfaces (hence the error message in the kernel log about not being able to find PHY for slave 1 or something) Aug 02 02:28:42 though rcn-ee found the patch for that in ti's git repo, so maybe that one is fixed in the latest kernel, haven't checked yet Aug 02 02:30:51 eh?! lol Aug 02 02:31:24 most code is ahack .. if you're totally honext about it lol Aug 02 02:31:35 yeah, if you say you have only one interface, it'll allocate some array with only one element rather than two Aug 02 02:31:47 and then the irq handlers just blindly index at 0 and 1 anyway Aug 02 02:32:15 it is, but some more than other Aug 02 02:32:46 take .. erm... musb for instance lol Aug 02 02:33:51 (I ran into fun issues like that of the ethernet descriptor RAM thanks to having a year or two of baremetal programming experience on a close ancestor of the AM335x, the DM814x) Aug 02 02:34:32 if you don't mind, no I'm not going to take musb thank you Aug 02 02:34:35 ;P Aug 02 02:34:47 what's yuor background zmatt .. you seem to have a deep interest in some .. Aug 02 02:34:50 erm Aug 02 02:35:00 odd things lol Aug 02 02:35:11 why odd? :) Aug 02 02:35:54 yeah I was gonna rephrase and hit 'enter' not 'backspace' .. gotta love that one :D Aug 02 02:36:12 I started out baremetal mostly to be able to carefully test our hardware before moving to booting linux Aug 02 02:36:41 ported a Forth I'd written some time before that to be able to interactively poke around and test things Aug 02 02:36:42 what 'hardware' is this? Aug 02 02:38:45 a board we designed that's now an expensive mistake collecting dust on a shelf after it became apparent that post-omapgeddon TI had no intention of supporting the SoC we had used (or its heavily customized linux kernel and software stack) and full baremetal development would take too much time Aug 02 02:40:09 ah yea .. you don't tend to get 'support' from SoC vendors .. not like the effort the community generally generates Aug 02 02:40:23 well it seemed well-maintained Aug 02 02:41:04 problem is .. you neve rknow when they're gonna turn life support off lol Aug 02 02:41:04 and then *poof* it no longer was, git commits came to a halt, disclaimer showed up on product and wiki pages Aug 02 02:41:11 xactly Aug 02 02:41:31 in fact the whole C6-Integra concept disappeared Aug 02 02:41:40 the 811x has only been released as an automotive part Aug 02 02:42:18 someone somewhere in management said 'no. Aug 02 02:42:27 I would have loved to have a board with the 811x instead of the am335x Aug 02 02:42:41 automotive you're always going to have a market .. until someone invents the teleport. Aug 02 02:43:07 likewise SoC for phones and media boxes Aug 02 02:43:21 tablets, etc Aug 02 02:44:20 811x would have been kickass though... basically DM814x minus some of the heavier video shit but with all the real-time goodies (pwmss, pruss) from am335x merged in Aug 02 02:45:49 sadly, these days,you really gotta have ya commercial head on .. not the engineer one .. :/ Aug 02 02:46:11 "wouldn't it be nice if ... " don't really sway it :/ Aug 02 02:46:17 I know Aug 02 02:46:29 if it don't bring in the $$ . they don't wanna know Aug 02 02:46:44 it's just sad to see such a nice processor series being treated in such a way Aug 02 02:47:23 the 81xx processors also clearly had received more careful attention using the design phase Aug 02 02:47:23 lol as if it wre a person .. cute lol Aug 02 02:47:44 unlike the mess that is the am335x Aug 02 02:48:38 (it's still a cool SoC, don't get me wrong) Aug 02 02:48:38 I suspect the am335x was the best cut & paste engineering around Aug 02 02:49:01 uhuh Aug 02 02:49:48 revision 1.0 actually had as erratum that its BOOTP requests claimed it was a DM814x Aug 02 02:50:09 hahaha Aug 02 02:51:27 that's also why the L4LS interconnect is full of holes (even if you account for the target agents being listed as "reserved") Aug 02 02:51:49 that's where the peripherals moved to L4WKUP were located on the DM814x Aug 02 02:51:56 (DM814x had no L4WKUP yet) **** ENDING LOGGING AT Sun Aug 02 02:59:58 2015