**** BEGIN LOGGING AT Tue Oct 06 02:59:58 2015 Oct 06 10:01:09 Hello, I asked here yesterday about a problem with flashing eMMC with the newest BBB images and the problem still isn't solved. Oct 06 10:04:08 Long story short - I cannot run the 192... webpage and when I try to run BBB standalone with screen pluged in, all that appears is is a penguin logo in top left corner. I tried with many flashers versions from 2015, nothing works. Does anybody have some solutions? Oct 06 10:05:02 Btw, right before flashing the new image, the old one worked perfectly (from around 2-3 years ago) Oct 06 10:12:45 I think the official latest is from march Oct 06 10:13:11 do you have a usb-UART adaptor to plug into the debug port? Oct 06 10:15:35 unfortunatelly not Oct 06 10:16:00 you should get one, they are cheap Oct 06 10:16:02 i tried the one from march and the one from 2015-07-28, neither works Oct 06 10:16:18 define "neither works" Oct 06 10:16:34 like I said above, the symptoms are the same for both Oct 06 10:16:53 "I cannot run the 192... webpage and when I try to run BBB standalone with screen pluged in, all that appears is is a penguin logo in top left corner. " Oct 06 10:16:57 well. then start at square 1 and do the #exactsteps dance Oct 06 10:17:25 tried that already with many combinations Oct 06 10:17:27 I'm trying to return to the last working for me version from 2013 Oct 06 10:18:20 well the insrtuctions are pretty straight-forward, there's not much place for mistakes Oct 06 10:18:34 there are a few Oct 06 10:18:50 Maybe not included in the instructions.. Oct 06 10:19:08 is your BBB rev A, B or C? Oct 06 10:20:04 that's the full reference: Beaglebone Black (A5) (CC-9990001457) Oct 06 10:20:15 ok, so 2G eMMC Oct 06 10:20:17 bought in 07.2013 Oct 06 10:20:33 do you want the gui or console only version? Oct 06 10:21:33 daaamn Oct 06 10:21:53 so I was flashing the 4gigs version on 2 gigs memory? Oct 06 10:22:30 well the console version would be enough if I was still able to open it through clould9 - is that possible? Oct 06 10:22:48 I don't think that comes with any fancyness Oct 06 10:22:59 http://elinux.org/Beagleboard:BeagleBoneBlack_Debian#BBB_.28All_Revs.29_eMMC_Flashers Oct 06 10:23:04 choose one of those two Oct 06 10:26:38 ok I'll try that Oct 06 10:27:23 also upon first boot after flashing, give it some time. it tends to rearrange things on first boot and doesn't show up via usb Oct 06 10:27:52 ok thank you very much, it will probabaly solve the problem Oct 06 10:28:03 well, for sure it will :) Oct 06 10:29:01 I was checking the wrong reference number all the time, that seems to be unchanged over different revisions, that's why I saw it was 4gigs in the specifications Oct 06 10:29:58 :) Oct 06 10:49:34 ok tbr , just confirmed - works well. tahnks again and have a nice day :) Oct 06 10:50:12 nie ma za co. miłej zabawy z bigielkiem :> Oct 06 10:51:21 :D Oct 06 12:44:56 hi Oct 06 12:45:44 i need help, if i compiling a kernel, how to select/change a module? Oct 06 12:46:33 for ex. if I need to select to compiling new chip sound type: tlv https://github.com/beagleboard/linux/tree/4.1/sound/soc/codecs Oct 06 12:47:49 how to force a compiling of tlv320aic32x4.c tlv320aic32x4.h to create a file object tlv320aic32x4.o? Oct 06 12:48:45 I try to compiling, but after I see only tlv320aic3x.o Oct 06 12:51:08 somebody can help me? Oct 06 12:52:37 hmm, the kernel config option hasn't been given a prompt Oct 06 12:53:36 zmatt: hi, menuconfig? Oct 06 12:54:05 without a prompt it won't appear there, and in fact can only be selected by another kernel option Oct 06 12:55:00 you can try editing the Kconfig to give it a prompt to allow you to enable it in menuconfig, but there's presumably some reason it hasn't been given a prompt Oct 06 12:55:12 zmatt: okay, were I select a new object? Oct 06 12:55:58 zmatt:general Kconfig, in main path of kernel source code? Oct 06 12:56:11 sound/soc/codecs/Kconfig Oct 06 12:57:30 or select SND_AM43XX_SOC_EPOS_EVM Oct 06 12:57:48 that is a prompt option, and it selects SND_SOC_TLV320AIC31XX Oct 06 12:58:32 wait, that's 31x Oct 06 12:58:41 zmatt: if I add this line in Kconfig: + select SND_SOC_TLV320AIC3204 if I2C Oct 06 12:58:44 is okay? Oct 06 12:58:52 no Oct 06 12:59:21 zmatt: my codec is compatible with: TLV320AIC32x4 Oct 06 13:01:32 below config SND_SOC_TLV320AIC32X4 Oct 06 13:02:18 add some description after "tristate" Oct 06 13:02:54 then go into menuconfig, the option should appear (with the description you gave it) Oct 06 13:03:07 zmatt: okay tristate "TI TLV320AIC26 Codec support" if SND_SOC_OF_SIMPLE depends on SPI Oct 06 13:03:45 just a description Oct 06 13:04:04 like tristate "Texas Instruments TLV320AIC32x4 CODECs" Oct 06 13:04:32 +config SND_SOC_TLV320AIC3204 + tristate + depends on I2C + config SND_SOC_TLV320AIC3X tristate Oct 06 13:04:59 ehh, huh Oct 06 13:05:16 just add the description after the "tristate" that's already there Oct 06 13:05:24 line 641 Oct 06 13:05:38 zmatt: after I add this line in Kconfig, just run menuconfig? Oct 06 13:06:11 yes, it will appear as new option Oct 06 13:06:12 zmatt: line 641, okay Oct 06 13:06:41 https://github.com/beagleboard/linux/blob/4.1/sound/soc/codecs/Kconfig#L641 Oct 06 13:06:42 :P Oct 06 13:07:13 zmatt: sorry, is just inclused? Oct 06 13:07:44 adding the description is necessary to have the option appear in menuconfig so you can select it Oct 06 13:08:29 anyway, I'm afk Oct 06 13:10:11 zmatt: okay Oct 06 13:17:55 zmatt:you are gentleman, for help i can contact you in private mail? Oct 06 14:02:13 fr1: just ask her Oct 06 14:02:18 there is no need for secrecy Oct 06 14:02:27 unless money is involved Oct 06 14:08:58 @av500: yes non secret, absolutely no money Oct 06 14:09:12 then stay here and discuss here Oct 06 14:09:16 then just ask here Oct 06 14:11:25 @av500, tbr: clear Oct 06 14:12:32 @av500, tbr: zmatt, helped me a lot, that's all Oct 06 14:24:17 But the point of public fora like this one is that a) not everyone is an expert in all things and b) discussions generate histroy that others with the same question can find. Oct 06 14:32:54 and just because I happened to be helpful doesn't mean I want to be your personal helpdesk (no, you can not contact me in private email) Oct 06 14:33:24 ^^^ Oct 06 14:34:13 fr1: There is an exceedingly good missive on "How to Ask QUestions the Smart Way", written by Eric S. Raymond. It is hosted on catb.org, and also linked from http://beagleboard.org/chat. Oct 06 14:34:20 fr1: I suggest reading through it. Oct 06 14:54:46 * tbr has that on an alias in his irc client Oct 06 15:14:30 zmatt: okay, sorry Oct 06 15:16:33 agmlego:... Oct 06 15:18:14 ===FOR ALL== I don't have problem to ask help in public chat Oct 06 15:19:08 Thank anyone help me Oct 06 15:55:28 Hi, i have a Beaglebone Rev A6 running on Angstrom. The operating system on my PC is Ubuntu 14.04. I can access my beaglebone via ssh. How can i mount my beaglebone root path at an folder on my Pc? Oct 06 15:57:51 I recommend you update your BBB to the current Debian Oct 06 15:58:27 http://beagleboard.org/latest-images Oct 06 15:58:43 It is an normal Beaglebone not an BBB. Oct 06 15:59:19 same applies .. Oct 06 15:59:32 all the 'Beagle' family share a common image now Oct 06 16:00:19 Alaif: then its a Beagleboard Oct 06 16:00:24 not a beaglebone Oct 06 16:01:03 it is this one here: http://beagleboard.org/bone Oct 06 16:01:44 ie. a 'white' :P av500 Oct 06 16:02:10 Thanks for the hint with debian. Is Debian missing something that Angstrom supports? Oct 06 16:04:07 ok, so a BBW Oct 06 16:04:29 Alaif: nobody supports angstrom any more on BB/BBB/BBW Oct 06 16:04:42 ok Oct 06 16:05:06 Alaif, rmmod g_multi and then re modprobe it with file= Oct 06 16:05:22 wait, angstrom doesn't have modprobe, use insmod... Oct 06 16:05:32 and then wreak havoc by having it mounted twice Oct 06 16:06:29 its not got emmc .. just write another (u)SD Card :P with debian Oct 06 16:07:19 but tbh .. since it's not got emmc .. just plug the uSD into your pc :P much easier! Oct 06 16:09:40 Ok, is Debian lacking anything that Amstrong has? I am asking because i am trying to familiarize with a project at university that i have been asked to continue from another student. Oct 06 16:09:52 debian is much more user-friendly Oct 06 16:10:00 veremit: but that's the easy way out.... I suggest setting up a BOOTP/TFTP/NFS configuration to allow booting a linux system and then load the gadget driver Oct 06 16:10:09 shup zmatt :P lol Oct 06 16:12:13 (the BOOTP and TFTP'ing something part is easy, but you'd need to build an SPL that can fetch the full u-boot.img over TFTP...) Oct 06 16:12:54 Alaif, it's been 2013 since the last version of angstrom shipped, if we lacked something, it's been added (and then some) Oct 06 16:13:55 of course an application can still have dependencies on something that's been obsoleted Oct 06 16:14:16 like the armel abi.. ;) Oct 06 16:14:42 Ok, thanks for all the information. I will try if it works whit debian. Oct 06 16:14:56 rcn-ee: angstrom was still armel ?? Oct 06 16:15:02 Bye Oct 06 16:15:30 zmatt, it was "softfp".. but built for armv7-a... so to run Angstrom binaries on debian, you need the "armel" linker... Oct 06 16:16:41 zmatt, omap4 doesn't like gpt. ;) this partuuid is a fun rabit hole... Oct 06 16:21:40 does omap4 require a DOS partition? if not, why would it care at all about the partition scheme used? Oct 06 16:21:48 *DOS boot partition Oct 06 16:22:30 (even if it does => hybrid partition format) Oct 06 16:22:51 omap4 was the first (well that other part after the dm37xx) to support raw mmc boot.. so just trying to get my omap4 boot with gpt partitions.. Oct 06 16:23:46 raw mmc boot means ROM doesn't care, so that's a matter of making sure u-boot is compiled with GPT support Oct 06 16:25:04 ahh.. and it wasn't enabled for the target till rc4... and here's i'm still using rc3... Oct 06 16:26:09 lol :p Oct 06 16:26:20 "that other part after the dm37xx" .. doesn't sound very specific Oct 06 16:26:40 omap4430 es1 is apparently older (or at least, has a lower jtag id) than the 37xx Oct 06 16:26:53 then again, es1 seems to be "let's speak of this no more" Oct 06 16:27:35 (I've never found any datasheet, TRM, or errata for it) Oct 06 16:28:18 zmatt, it's that single core 1.5Ghz part.. the first one with a bootrom that supported raw boot, i never had one of those kits.. dm81xx? Oct 06 16:29:51 Netra (dm816x) is about contemperaneous with omap4 I think, or possibly slightly older... does it support raw eMMC boot? *checks* Oct 06 16:30:30 yep Oct 06 16:32:14 makes sense since Netra, the grandfather of the whole dm81xx / am335x / dra6xx happy family is omap4-generation and has a lot in common with it Oct 06 16:34:23 the later chips inherit most of their architecture from Netra (except for its FAPLL which was apparently "yeah, nice idea... *toss into trash*") Oct 06 16:34:46 except for am5/dra7 which directly descend from omap lineage of course Oct 06 16:34:53 and then scraped part of the bootloader in am5/dra7... Oct 06 16:35:00 am5 isn't a descendent Oct 06 16:35:06 of netra Oct 06 16:35:21 more keystone.. Oct 06 16:35:29 at least they fixed the ROMBL bug of O4 in AM33xx Oct 06 16:35:30 no, omap5, very obvious omap5 Oct 06 16:36:00 the one where it would be more anal about partition table than windows/linux Oct 06 16:36:03 it's omap4 -> omap5 -> vayu and netra -> centaurus -> subarctic -> aegis Oct 06 16:36:10 with plenty of cross-pollination of course Oct 06 16:36:49 and stuff borrowed from keystone, or even their c2000 microcontrollers (pwmss) Oct 06 16:37:14 tbr part of that was a broken x-loader... and it was flashed into nand.. ;) Oct 06 16:37:37 rcn-ee: panda didn't have nand Oct 06 16:37:50 netra & children don't have x-loader anyway Oct 06 16:38:21 tbr the original beagleboard did.. x-loader was hardcoded for 63.. ;) Oct 06 16:38:33 at least on GP, dunno HS versions Oct 06 16:38:59 rcn-ee: the 63 bug was a whole different can of worms and source of cargo cult Oct 06 16:39:08 (I've seen the X-LOADER string in ROM, but that doesn't say much since ROM seems to be full of disused crap) Oct 06 16:39:32 yeap tbr it was fun to deal with.. Oct 06 16:40:46 rcn-ee: the 81xx are really nice chips btw, they're just being treated like shit by TI Oct 06 16:41:05 * tbr has been seeing X15 early samples popping up and wondered: a) which bugs will that have b) how do I get my hands on one. The BBB hooked up to @IoToilets wants out of there and demands a more powerful successor as IoT gateway Oct 06 16:41:49 tbr ping jkrinder ;) Oct 06 16:42:12 I do wonder what chip problem(s) have been delaying public release of the am5 for... almost a year now? Oct 06 16:42:21 must be an expensive oops Oct 06 16:42:24 would also undoubtedly help as we'll end up fielding questions here Oct 06 16:42:37 rcn-ee: mhm, haven't seen him here lately Oct 06 16:42:51 I thought they were being held up by TI .. Oct 06 16:43:08 veremit: hence I'm saying "public release of am5", not "public release of x15" Oct 06 16:43:22 tbr, his wife just had a baby yesterday, so he's not going to be around for a few days.. Oct 06 16:43:29 zmatt.. I'm talkin about x15 .. per tbr Oct 06 16:43:34 ah, that's quite understandable Oct 06 16:43:50 veremit: ... which houses the am5, hence is also held up Oct 06 16:44:01 veremit: well, the borad still needs to get ready for launch. so some early boards are circulating Oct 06 16:44:24 sweet, bootrom with gpt works fine on this ancient omap4 now.... now to hack up initramfs to pass root=PARTUUID=<> to the kernel.. Oct 06 16:44:32 tbr... thought they were Very early boards .. and if so, nearly a year in circulation I thought ... Oct 06 16:44:47 original release was set for last februari Oct 06 16:44:55 zmatt .. right Oct 06 16:44:59 rcn-ee: nice Oct 06 16:45:29 interesting, and google has an am572x-evm page indexed at TI but clicking on it yields a 404 Oct 06 16:45:58 lol Oct 06 16:46:35 cpu itself also still doesn't have a product page, not even a preliminary one Oct 06 16:46:54 yeah that doesn't suprise me .. they're really not ready with it Oct 06 16:47:00 zmatt, this one works: http://www.ti.com/lit/ug/spruhz6/spruhz6.pdf (AM572x rev 1.1) Oct 06 16:47:09 the page google indexed really was a test I guess lol Oct 06 16:47:12 the TRM is still there yes Oct 06 16:47:38 probably gauging interest Oct 06 16:47:43 still the same version that's been out there for a long while Oct 06 16:48:02 what's the last date on the changelog ? :p Oct 06 16:48:18 veremit, October 2014... Oct 06 16:48:20 I bet Q1 2015 or before .. Oct 06 16:48:26 ScORE! Oct 06 16:48:52 veremit: it was out there before the februari-release was announced, obviously Oct 06 16:48:53 :P Oct 06 16:49:00 i'm pretty sure this x15 on my desk has 1.2 silcon, and i know production x15's should have rev 2.0... Oct 06 16:49:02 as I might expect lol Oct 06 16:49:23 there was a "reset" problem on 1.1/1.2 henced it got deleayed.. Oct 06 16:49:25 rcn-ee: yeah, so r1.x is probably also going the way of "let's not speak of this anymore" Oct 06 16:49:53 which is not entirely surprising considering the complexity of the chip Oct 06 16:51:02 it's crazy how much of a beast this chip is, when you realize even more has been efused-disabled from the dra7xx line. ;) Oct 06 16:51:16 yup, no EVEs for us Oct 06 16:51:21 :/ that's no fun .. Oct 06 16:51:50 rcn-ee: I actually noticed that our dm814x evm has a centaurus part with all features enabled except customer eFUSE Oct 06 16:51:54 i know, we will have to do image recognition via the dual c6000's instead of eve.. ;) Oct 06 16:52:06 but all automotive shit is enabled, and even TPPSS Oct 06 16:52:17 yeah Oct 06 16:52:56 i think part of the problem with eve, the person who designed it left ti, so no one knows how to support it... Oct 06 16:53:02 lol Oct 06 16:53:06 nice ... Oct 06 16:53:47 EVE was actually given birth on a Centaurus-derivative, dispite claims by TI that it is a new feature of the TDA2 Oct 06 16:53:56 that almost happed to the pruss on am33xx, but he fought to keep it enabled... Oct 06 16:53:57 Centaurus Eve, aka TDA1MEV Oct 06 16:54:10 \o/ Oct 06 16:54:16 PRU is really amazing Oct 06 16:54:36 it's one of the best features.. ;) and we get 4 on the x15.. Oct 06 16:55:05 hmm I need to get another beagle .. Oct 06 16:55:06 and two Benelli instances Oct 06 16:55:38 and IVA-HD... (closed source blobs for that probably?) Oct 06 16:56:10 yeah iva-hd is closed.. but there's been 5 blob updates over the last 4 months... Oct 06 16:56:15 and dual-core SGX, and a 2d bitblit Oct 06 16:56:30 iva-hd would also be quite hard to figure out without public docs Oct 06 16:56:49 I mean, I could probably get the arm9 cores up, but those hw accels are a different story Oct 06 16:57:21 (if you look at the already highly cryptic hw accels of iva2.2 that are documented) Oct 06 16:59:29 one of the Benellis was freely usable right, with the other one dedicated to looking after the closed-source video stuff? Oct 06 17:00:16 or whatever they were called nowadays... IPU or something Oct 06 17:02:11 rcn-ee: I assumed the lack of EVEs enabled on the am5 would be due to feature-binning... i.e. the dies with working EVEs get a TDA2 label slapped on Oct 06 17:02:37 if they're working but deliberately disabled that would be really sad Oct 06 17:03:25 as for support, just do a "community" release of whatever docs are there and we'll take it from there :P Oct 06 17:03:31 zmatt, i'm pretty sure they are working, jkrinder's been fighting to keep them enabled.. i'm pretty sure it's just the support issue.. Oct 06 17:04:59 argh Oct 06 17:06:12 there's probably as much collective expertise in the community as there is at TI .. lol .. -facepalm- Oct 06 17:06:18 like, same with the dm81xx, a ton of stuff isn't publicly documented yet also not really supported by TI... if you're not going to support it, then at least hand over a garbage bag containing whatever you do have on it to allow people to figure out it Oct 06 17:07:06 and as far as resource goes .. community has it every time .. unless someone 's payin the bill for the engineer's time .. ! Oct 06 17:22:22 hello Oct 06 17:42:10 <[Butch]> Does anyone have any suggestions for those of us who upgraded to Mac OS X 10.11 (El Capitan) only to discover that the USB gadget drivers don't work? When the BBB comes up, it mounts properly on the desktop, but the USB gadget interface doesn't come up, so I can't ping/ssh into my BBB any more. :-( Oct 06 17:42:18 <[Butch]> And yes, I could use the serial port, but I'm hoping for something a bit more useable. Oct 06 17:44:52 [Butch], http://joshuawise.com/horndis Oct 06 17:45:28 <[Butch]> So, no. :-( Oct 06 17:45:40 <[Butch]> At least not yet. Oct 06 17:46:00 [Butch], "no" and "we" aren't working on it... contact "jwise"... Oct 06 17:46:22 his source is located here: https://github.com/jwise/HoRNDIS/ Oct 06 17:53:07 thats a bit shit lol Oct 06 17:53:27 but then I have my own opinions about usb-gadget Oct 06 17:54:01 <[Butch]> rcn-ee: Thanks for the info. Oct 06 17:54:52 Sorry [Butch], all these locked os' just make things painful (windows with it's signing key/ mac with it's ktext's..) Oct 06 18:18:20 [Butch]: my first intuition would be: recompile kernel with EEM instead of RNDIS Oct 06 18:19:00 rcn-ee: the mac driver is open source though Oct 06 18:19:18 www.opensource.apple.com/source/AppleUSBCDCDriver/ Oct 06 18:20:25 (that's how I knew it supports EEM) Oct 06 18:20:49 zmatt, that's now 2 reasons to switch to eem... it just works in windows/mac.. ;) Oct 06 18:21:14 EEM also works on windows? I thought you said it didn't Oct 06 18:22:07 if windows does EEM I'd suggest ditching RNDIS this second Oct 06 18:24:01 wiki... (RNDIS) is a Microsoft proprietary protocol Oct 06 18:24:07 so why do we need a signed driver? Oct 06 18:24:39 VID/PID Oct 06 18:24:42 because windows needs "signed drivers" for any device, even if it's an .inf file telling it to use its own fucking class driver Oct 06 18:24:50 same with acm Oct 06 18:24:50 it isnt so much a driver as a pointer Oct 06 18:25:09 yeah it's a ini file of a few lines Oct 06 18:25:27 you can subvert all that by faking the VID/PID Oct 06 18:25:36 I was about to suggest that Oct 06 18:26:14 somebody seriously needs to hit Microsoft with a clue-by-four about the concept of USB class drivers Oct 06 18:26:36 but godaddy want's the $10K for a signing cert! ;) Oct 06 18:27:17 solution is simple. don't call it USB. return random numbers on the VID/PID side Oct 06 18:27:23 push that out to numerous products. Oct 06 18:27:46 then submit a patch for Linux to recogonize it by other means. Oct 06 18:27:58 linux will already recognize it Oct 06 18:28:00 as well mac Oct 06 18:28:15 not after you randomize the VID/PID Oct 06 18:28:27 and if we use eem, mac's networks is now fixed... Oct 06 18:28:30 for recognized usb classes, vid/pid is only used to choose whether any workarounds/hacks are used Oct 06 18:28:38 but we need to fix the winblows... Oct 06 18:28:41 that's a good point Oct 06 18:28:55 that's the whole point of the concept of usb classes Oct 06 18:29:05 the other way is to compile a special version of putty that uses libusb Oct 06 18:29:13 so a class driver can support *all* devices for that class Oct 06 18:29:24 except for the 'quirks' list Oct 06 18:29:58 buggy hardware? *acts shocked* ;P Oct 06 18:30:59 there are MSC devices that break on Linux Oct 06 18:31:14 msdn claims "If a device that belongs to a supported device class is connected to a system, Windows automatically loads the class driver, and the device functions with no additional driver required." Oct 06 18:31:26 ah Oct 06 18:31:38 wait... Oct 06 18:31:39 that's since windows 8 apparently Oct 06 18:31:51 RNDIS can't be a class device Oct 06 18:31:58 wiki... (RNDIS) is a Microsoft proprietary protocol Oct 06 18:32:01 no it isn't Oct 06 18:32:04 that by definition.... Oct 06 18:32:25 but I'd still expect Microsoft's own driver to recognize it for similar reasons Oct 06 18:32:51 rcn-ee: are your images officially USB certified? Oct 06 18:33:05 ds2, hell no... Oct 06 18:33:31 then you can have it use any numbers that happen to be located in a field that USB uses for VID/PID Oct 06 18:33:54 if those numbers happens to match usage by a device recognized by their driver.... that's just a coincidence Oct 06 18:33:57 laughs.. true.. i can use any vid/pid... as they can't take "my" vid, as i don't have one. ;) Oct 06 18:34:14 afterall you don't have a USB device... you just have a serial protocol Oct 06 18:34:26 windows 10 apparently finally recognizes usb serial devices without custom .inf file Oct 06 18:34:47 so microsoft is slowly gaining clue Oct 06 18:35:07 call it a LSB - Liberated Serial Bus Oct 06 18:35:09 :D Oct 06 18:35:20 libreSB Oct 06 18:35:40 no mention about any usb net driver other than MBIM though Oct 06 18:36:51 rcn-ee: what's the root pw verdict? Oct 06 18:37:53 waiting for jkrinder's answer, as he's the sign-off for beagleboard.org's images... but i'm nuking the "root/root" for normal elinux images. ;) Oct 06 18:38:35 usually has a crappy root pw, but ssh only accepts public-key authentication Oct 06 18:38:39 /me Oct 06 18:38:45 on BBBs that is Oct 06 18:39:26 ^ that's part of the change, disable root over ssh (with password authencation (which is either or "root")) Oct 06 18:39:31 shutting off network connectivity is more reasonable Oct 06 18:39:58 ds2: except for that being the primary connectivity I use Oct 06 18:40:15 rcn-ee: if it is network reachable, people can use nodejs to screw the box over anyways Oct 06 18:40:20 no Oct 06 18:40:26 keep it local network accessible Oct 06 18:40:26 for a while I actually compiled my kernels without usb support :P Oct 06 18:40:29 i.e. no default route Oct 06 18:41:04 ds2, with bonescript too... there's a few lots of leaks all over the image... but linux.xdds use ssh... Oct 06 18:41:31 but cutting off non local traffic covers those too Oct 06 18:41:40 oh wait we're talking about the newbie-introductory images Oct 06 18:41:47 ds2 bonescript allows you to remote "flash" a board via: http://beagleboard.org/updater (just stick in the ip) Oct 06 18:42:18 rcn-ee: it should really use a form of public-key authentication then Oct 06 18:42:23 IIRC, that should work just fine Oct 06 18:42:47 it just needs to download, not listening Oct 06 18:42:54 that updated should be javascript tricks Oct 06 18:43:01 zmatt, i though about reading a *.pub into the "usb flash" drive.. How do we explain that to userr? Oct 06 18:43:01 doh 1145...need to go Oct 06 18:43:32 rcn-ee: what do you mean? Oct 06 18:44:07 zmatt, so with an out of box image, how do we tell the user to upload a *.pub key to enable ssh... (maybe thru the web interface?) Oct 06 18:44:36 rcn-ee: the "it should really use a form of public-key authentication" was about remote flashing Oct 06 18:44:51 zmatt, ah that! yeah... ;) Oct 06 18:45:55 and a web interface thing for setting up the ssh public key sounds like a good idea, and not too hard Oct 06 18:46:27 tell them to ssh-keygen -t ed25519 or if that fails -t ecdsa Oct 06 18:46:28 :P Oct 06 18:46:49 paste the public key into browser field, apply Oct 06 18:47:19 on windows can putty generate a key now days? or is still a pain? Oct 06 18:47:47 putty can generate keys, but they're not in openssh format and you have to tell it to convert them, last I checked Oct 06 18:47:52 ahh puttygen... Oct 06 18:47:52 so, more confusing steps Oct 06 18:48:50 yeah, we want the confusing down to a minimal. ;) Oct 06 18:49:04 I can certainly understand that :) Oct 06 18:49:10 is there no way to make a terminal appear *in* a browser window? Oct 06 18:49:19 surely someone made a terminal emulator in javascript Oct 06 18:49:29 how would it connect to things? Oct 06 18:49:31 cloud9 ide has one built in (port 4000) Oct 06 18:49:36 websocket Oct 06 18:49:45 websockets need a web server backing them Oct 06 18:49:48 they aren't raw sockets Oct 06 18:49:49 it already has Oct 06 18:49:53 ah Oct 06 18:49:57 cloud9ide: (and it's also logged in as root) Oct 06 18:49:59 I popped in in the middle of the conversation, sorry Oct 06 18:50:49 I believe I've seen websocket powered terminal emulators... Oct 06 18:50:58 but I might be thinking of websocket powered vnc into a terminal :) Oct 06 18:51:13 rcn-ee: problem solved, doesn't require any client-side effort... though if we expect users to be *that* newbie I'd also support requiring the usb connection for meddling with the BBB and by default allow only outgoing ethernet connections Oct 06 18:51:23 endrift, https://github.com/aluzzardi/wssh yeap websocket.. Oct 06 18:51:37 neat Oct 06 18:51:50 then again, most people have a LAN that's adequately secure Oct 06 18:52:48 zmatt, yeah takes more work for pass thru, so we can be lax on security, where as can be plugged directly, so root needs to be disable there.. Oct 06 18:52:49 alternatively, a post-install wizard asking it whether their LAN is trustworthy (e.g. home) or not (e.g. university network) Oct 06 18:56:43 how does this sound as default: public key only, root disabled, password authentication allowed, root disabled Oct 06 18:56:55 haha, crazy thought, but could in principle help them by generating a keypair for them and deliver it in a putty-compatible format (perhaps along with putty) on a read-only usb mass storage gadget Oct 06 18:57:38 We "regenerate" ssh-keys on first bootup with the "flasher"... so we could coyp the *.pub key to the usb flash... Oct 06 18:57:39 rcn-ee: I was pondering whether you can configure ssh's policy based on IP address / interface Oct 06 18:57:53 zmatt found: https://unix.stackexchange.com/questions/128444/disable-ssh-password-authentication-on-specific-interface Oct 06 18:57:59 check! Oct 06 18:58:05 192.168.7.x is always used.. for usb0... Oct 06 18:58:35 option 2 on that site is prefect for us. ;) Oct 06 18:59:45 though tbh, inventing ways to things like this as newbie-proof as possible isn't really my idea of fun... I'm glad I'm not you guys ;-) Oct 06 18:59:56 though I understand the motivation Oct 06 19:01:15 Is it possible to create a storage partition on a blank sd card when booting from the eMMC on the BBB? Oct 06 19:02:04 Jackix, sure, pass "file=/dev/mmcblk0" thru g_multi... Oct 06 19:02:28 obviously: "file=/dev/mmcblk0p1" ;) Oct 06 19:03:10 so, a while back I worked on a Windows machine often enough to also try to make things more comfortable, I setup something that creates a nice GNUish environment in a properly working terminal emu, and can ssh-keygen and ssh like any normal *nix... I'll check again what that setup was, probably cygwin though I'm not sure anymore since it looked noticably uncygwinnish Oct 06 19:04:22 it's quite possible to preconfigure a small but adequate environment like that and serve it to them via g_multi when the system is running Oct 06 19:04:42 allowing instructions from there on to be the same on windows, mac, and linux Oct 06 19:05:00 (for e.g. generating ssh keys) Oct 06 19:05:01 Wow.. that seem to be way out of my technical skill level. But ill google around and see what i can find on it.. Oct 06 19:05:21 rcn-ee: don't you mean mmcblk1p1 ? Oct 06 19:05:26 Jackix, i'm assuming you want it to pop up over usb right? Oct 06 19:05:49 zmatt, depends on the condition of the cd line at kernel "boot".. Oct 06 19:06:05 rcn-ee: argh, can't we fix this with an udev rule? Oct 06 19:06:10 actually, that's not even a question Oct 06 19:06:14 we can fix this with an udev rule Oct 06 19:06:21 zmatt, partuuid!!! ;) Oct 06 19:06:32 no, make mmc1 always mmcblk1 Oct 06 19:07:03 the dt-mmc maintainer said no..... Oct 06 19:07:04 Well i want to run it over vnc and use it as a stand alone computer, but for the time beeing im connecting over usb. (if that was what you ment :) ) Oct 06 19:07:05 as a side effect, you wouldn't need partuuid since u-boot's numbering would always agree with the device name Oct 06 19:07:32 hmm Oct 06 19:07:42 is that a trick only network interfaces can do? Oct 06 19:08:04 it would be seriously pleasant if mmc0 were always mmcblk0 and mmc1 always mmcblk1 Oct 06 19:09:40 the OF-way of doing this would be using /aliases but few kernel drivers bother to check there if they have a designated number assigned already Oct 06 19:10:10 (e.g. the generic spi framework does it, but the mcspi driver doesn't use the generic spi framework nor does it do it itself) Oct 06 19:11:02 okay found the last aliase mmc patch: http://permalink.gmane.org/gmane.linux.kernel.mmc/21519 (and disussion) Oct 06 19:11:11 rcn-ee: What I would like to have is the OS running from the eMMC and have all my programs/files ect. to be located on my sd card. That way the OS runs fast via the eMMC and my files/programs can be stored and run from the sd card. Oct 06 19:11:36 ugg wrong link disussion: http://comments.gmane.org/gmane.linux.kernel.mmc/21519 Oct 06 19:12:08 Jackix, the eMMC (even with 8bit) isn't that fast... you can get faster 4-bit microSD cards... Oct 06 19:12:32 rcn-ee: depends very much also on whether you have the Micron or Kingston Oct 06 19:12:37 the latter is much, MUCH, faster Oct 06 19:12:42 and you have no idea which you'll get Oct 06 19:13:51 rcn-ee: the rejection sounds like bullshit to me Oct 06 19:14:05 there's no reason to not reserve mmcblk0-2 for the on-chip controllers Oct 06 19:14:15 zmatt, i agree... Oct 06 19:14:18 Oh, well i have an old no ancient SD card and thats why i thought it was a better idea, and besides i wanted to try it. Mostly for the knowledge that comes with it :) Oct 06 19:14:22 wonder if it's time to bring it up again... Oct 06 19:14:23 zmatt i patched my mmc driver to assign ids on port rather then on appraerance .. once n for all Oct 06 19:14:35 rob_w: exactly Oct 06 19:14:59 cool... the omap4 bootrom can do something the imx6's can't... gpt partition breaks imx6... ;) Oct 06 19:15:05 rcn-ee: also, to answer the original poster's doubt: /aliases *is* afaik the correct place for it Oct 06 19:15:45 ok ok, no reason for doubt, *grabs the OF spec* Oct 06 19:17:41 laughs!! zmatt is brought up again 4 days ago: https://lkml.org/lkml/2015/10/2/522 Oct 06 19:18:22 ah, and look how simple it is... I remembered correctly there's already a convenience function for it Oct 06 19:19:43 the only problem I see is that the ids are not reserved in this way, if it's done at card detect time Oct 06 19:20:09 e.g. if you add an external mmc controller it could claim mmcblk0 Oct 06 19:20:50 that's where steffen's patch is better.. mmc0 => mmcblk0 Oct 06 19:21:38 haven't looked at it, does it prevent other devices from claiming the IDs? Oct 06 19:22:58 they just need to be specified in the dtb... Oct 06 19:23:00 also, mmc_alloc_host sounds like something that ought not be dependent on card-detect Oct 06 19:23:13 that's the same with the new patch Oct 06 19:23:28 you define /aliases/mmc0 = &mmc0; Oct 06 19:23:29 etc Oct 06 19:23:52 (i still need to patch dtc to accept that syntax as substitute for /aliases { mmc0 = &mmc0; }; ) Oct 06 19:26:45 ok I'm getting more and more confused here Oct 06 19:27:29 rcn-ee: are the kernel modules in ti-sgx-es8-modules-3.14.54-ti-xenomai-r77 doing anything without the utilities like pvrsrvctl from TI's package? Oct 06 19:27:40 rcn-ee: and can wayland work with those modules? Oct 06 19:27:56 juodumas, those are just the sgx modules built-against the kernel in the repo... Oct 06 19:28:32 run: sgx_create_package.sh from this repo branch: https://github.com/RobertCNelson/ti-linux-kernel-dev/tree/ti-linux-3.14.y Oct 06 19:28:32 I don't see how allocation of the mmc_host can be delayed by card detect, given that card detect config is parsed in mmc_of_parse which already need an mmc_host to have been allocated Oct 06 19:28:54 rcn-ee: you still haven't packaged the sgx stuff? ;) Oct 06 19:29:36 not the old broken stuff.. ti's going to push egl/wayland stuff for "am3/am4/am5" to the git repo. ;) Oct 06 19:29:49 (v4.1.x+ only) Oct 06 19:29:54 rcn-ee: when? :) Oct 06 19:29:58 that / doesn't make sense... am3 = lcdc, am4/5 isn't Oct 06 19:30:00 when it works.. Oct 06 19:30:07 zmatt, dmabuff! Oct 06 19:30:44 I'll take your word for it that that is a valid retort to my objection... I don't really know enough about it Oct 06 19:30:52 rcn-ee: so wayland won't work with 3d accel even if I install the utils with sgx_create_package.sh? Oct 06 19:31:14 zmatt: http://git.ti.com/gitweb/?p=ti-linux-kernel/ti-linux-kernel.git;a=commit;h=d84004699f61a3bbf112d59c0e106f1c47762e8c Oct 06 19:31:37 juodumas, last time i tried it, it just hardcrashed... Oct 06 19:31:49 thanks. will try on stretch Oct 06 19:32:18 juodumas, my "weston" hacks are here: https://github.com/RobertCNelson/boot-scripts/tree/master/tools/wm fbdev worked ... Oct 06 19:32:19 be sure to use FRONT and not FLIP Oct 06 19:32:44 any suggestion on how to get notified when/if ti's going to push the egl/wayland stuff? Oct 06 19:33:16 or wait, did FLIP work? I know the fliptest thing utility crashed badly, and on E2E people are also having serious problem with FLIP Oct 06 19:34:33 juodumas, i was told the 2nd release of "ti2015.0x" ti2015.01 just came out 2 weeks ago.. maybe a quarter from now... Oct 06 19:36:34 juodumas, setup a cron job to mirror this repo: http://git.ti.com/gitweb/?p=graphics/omap5-sgx-ddk-um-linux.git;a=summary and just watch the commits' ;) then you'll know.. Oct 06 19:39:22 rcn-ee: ok, so the latest patch doesn't actually fix mmcblk numbering, steffen's patch does... the new version is a just a simplified version of steffen's second patch Oct 06 19:39:40 I agree with the simplification, though I'm not sure what that numbering is used for if not for mmcblk Oct 06 19:39:54 neither addresses my concern though Oct 06 19:40:24 thanks again. will you then make lxqt run on wayland by default on the official debian image? :) Oct 06 19:41:01 shouldn't it then be called lwqt ? Oct 06 19:42:59 looking back at lxde's history, i think it will take another rewrite before it can be renamed... Oct 06 19:43:03 juodumas, that's what i wanted to do with qt5/lxqt... it looks like stretch qt5/kde might have wayland just work... Oct 06 19:43:45 rcn-ee, do you know of sgx support with xorg is beeing worked on for am437x ? Oct 06 19:44:04 latest sdks exclude xorg for some reason .. i just hope they dont drop the support at all Oct 06 19:44:13 it might just work with stretch in 6 months, it's not there yet Oct 06 19:44:36 rob_w, xorg is done with sgx, they ripped everything... Oct 06 19:44:47 or sgx is done with xorg. ;) Oct 06 19:45:02 so its dropped ? or is it finished ? Oct 06 19:45:09 rcn-ee: ahh, starting to become clearer, mmc_host != mmcblk, hence mmc_hosts are currently being numbered correctly already (simply because they are probed in the right order) but that doesn't help getting the mmcblk numbering right Oct 06 19:46:17 not sure when, if ever, I'd run into the mmc_host numbering then (well, in /sys/class/mmc_host obviously, but I mean practically) but it's good to have both right Oct 06 19:47:12 my objection then doesn't apply to mmc_host since those are allocated early and remain fixed, but my concern remains valid for mmcblk: the ids are not being reserved Oct 06 19:48:26 easiest workaround would be scanning for all aliases present, and use the highest + 1 as min val for all mmcblk devices which do not have an id obtained from of_alias_get_id Oct 06 19:49:15 the patch as-is would suffice for fixing boot though Oct 06 19:49:33 (probably) Oct 06 19:52:10 and look there, a convenience function exists for that too: of_alias_get_highest_id Oct 06 19:53:02 used e.g. by i2c core in exactly the same way Oct 06 19:53:10 __i2c_first_dynamic_bus_num = retval + 1; Oct 06 19:56:48 rcn-ee: looks like all the ingredients for fixing this properly are there Oct 06 19:56:58 and precedent is there also Oct 06 19:57:51 and then you can stop fiddling with uuids ;) Oct 06 19:58:24 lol Oct 06 20:07:27 rcn-ee: the new patch is actually more correct than Steffen's version for dealing with mmc_host .. (and more concise) Oct 06 20:08:03 zmatt, should we just run with ben's and call it good? ;) Oct 06 20:08:41 ben's is just mmc_host though, i.e. steffen's "2/2" Oct 06 20:08:50 it still needs an equivalent of steffen's "1/2" Oct 06 20:10:03 but steffen uses the of id as min for idr_alloc while ben tries to get the of id (exactly, not "at least") and falls back to dynamic alloc on failure Oct 06 20:10:28 the distinction is important when later dynamic allocation is patched up to use a minimum value analogous to __i2c_first_dynamic_bus_num Oct 06 20:12:18 which I do think is important, but for now you can get boot fixed without it... assuming nobody also has an external mmc card reader attached which responds faster to probe than the internal one Oct 06 20:16:51 Wayland clearly works with `weston-launch -- --backend=fbdev-backend.so --use-gl` (resizing and rotating windows is smooth) Oct 06 20:18:09 But running a QtQuick example with QT_QPA_PLATFORM=wayland exits with "EGL not available", "failed to create EGLContext" etc. :( Oct 06 20:49:36 sooooo Oct 06 20:49:42 what's with the missing install-me.sh scripts on rcn-ee ? Oct 06 20:50:35 nyt, got rid of them... ;) all kernel *.deb's are now in the repo... Oct 06 20:50:56 sudo apt-get update ; sudo apt-get install linux-image-<*> ; sudo reboot Oct 06 20:51:03 ah Oct 06 20:51:29 was trying to update an old bbb for the wife Oct 06 20:52:08 nyt, so [cd /opt/scripts/tools/ ; git pull ; sudo ./update_kernel.sh] so update your kernel just fine (3.8.x) Oct 06 20:52:46 already on it ;) saw a ml post Oct 06 20:52:48 ty Oct 06 20:53:01 had to reset then pull Oct 06 20:53:15 sweet Oct 06 20:55:40 rcn-ee: how usable or unusable are your 4.x branches? Oct 06 20:56:26 ds2, it's about 98% there... Oct 06 20:56:42 compared to v3.8.x Oct 06 20:56:52 rcn-ee: what would you recommend as an experimental setup for rpmsg/remoteproc on the PRUs? Oct 06 20:57:02 that's the 2%... Oct 06 20:57:05 :P Oct 06 20:57:24 hehe Oct 06 20:57:30 so... in v4.1.x-ti, "uio_pruss" is broken... ti's forcing down remoteproc_pruss.. Oct 06 20:57:50 what exactly is either of them doing that you can't do with plain uio ? Oct 06 20:57:57 I can fix that Oct 06 20:58:04 in the other v4.1/v4.2/v4.3 (bone) we have patches for good old uoi_pruss... Oct 06 20:58:35 fine. 3.8 it is Oct 06 20:58:48 I still prefer 3.2 ;) Oct 06 20:59:11 community wants uio_pruss, ti wants remoteproc_pruss... ;) Oct 06 20:59:28 so we will keep patching it for uio_pruss to work. ;) Oct 06 21:00:00 remoteproc has its niceness Oct 06 21:00:01 this sounds very fixable... I mean, hell, fuck both and use /dev/mem ;P or uio_pdrv_genirq if you need a pruss -> a8 irq Oct 06 21:00:16 community also wants to run a 200Kps ADC through userspace Oct 06 21:00:21 but isnt remoteproc the kernel way to do things ? Oct 06 21:00:30 remoteproc is the "kernel" way.. Oct 06 21:00:36 why only 200Ksps ? I've done 1Msps in userspace Oct 06 21:00:36 way forward... Oct 06 21:00:47 the counter arguement is the kernel way is dead ;) Oct 06 21:00:51 die DT die! Oct 06 21:00:58 lol Oct 06 21:01:07 rob_w, but remoteproc_pruss is not 100% compatible with uio_pruss... ;) Oct 06 21:01:15 users just load binaries... Oct 06 21:01:20 didnt suprise me rcn-ee Oct 06 21:01:36 rcn-ee: i'd think objcopy can bridge that Oct 06 21:01:41 though usually I just configure the adc with EDMA set up to write to a circular buffer, so userspace can read the latest measurements from the buffer whenever it wants to Oct 06 21:02:10 zmatt: EDMA to userspace? heheh Oct 06 21:02:20 to a piece of ocmc I reserve using DT Oct 06 21:02:32 and yes I setup EDMA from userspace Oct 06 21:02:33 :) Oct 06 21:02:42 silly hacks Oct 06 21:02:51 just do it in the kernel or go use a cortex-m :P Oct 06 21:02:59 the kernel driver already reserves the necessary resources thanks to DT declaration Oct 06 21:03:14 so I can actually do this safely from userspace without any risk of running foul with the kernel Oct 06 21:03:28 (EDMA is designed to serve multiple lords anyway) Oct 06 21:04:35 ds2: I'd love to use the cortex-m, but annoyingly it doesn't have L3 access, only L4WKUP, which limits its utility Oct 06 21:05:13 zmatt: I mean a different chip like the STM32's Oct 06 21:05:37 although it occurred to me it may be possible to setup an EDMA job, triggerable by the m3 in some way, which copies an EDMA QDMA job from the m3's memory to itself Oct 06 21:05:46 what's the point of having all the hw protections when all you do is bypass it Oct 06 21:06:08 sounds like a worm? Oct 06 21:06:29 aforementioned proposal is not to be taken too seriously ;) Oct 06 21:06:52 but it would allow the m3 to perform arbitrary L3 transactions Oct 06 21:11:17 interacting with a peripheral from userspace is a lot safer than doing it from the kernel exactly thanks to memory protections Oct 06 21:12:13 any mistake in kernelspace is Bad™, while in userspace it will typically lead to not much worse than a confused peripheral which can be reset if needed Oct 06 21:12:56 (granted, the EDMA setup is a bit more delicate, but very easy and EDMA isn't memory-mapped except for that one-time setup) Oct 06 21:15:20 (I mean, this is my EDMA setup code -> http://pastebin.com/E3ZrZ1pf ) Oct 06 23:44:43 quick q, hooked up to a 1920x1080 capable display, what res will the bbb pick? Oct 06 23:51:05 nyt: depends on EDID data Oct 06 23:51:40 nyt: see here -> http://elinux.org/Beagleboard:BeagleBoneBlack_HDMI Oct 06 23:51:57 ours picked 1280x1024, with the screen then cheerfully scaling that up to 1080 lines, with hideous results obviously Oct 06 23:52:15 juse seeing some odd behavior, hooked up to a samsung 60" tv, with display format in 16:9 Oct 06 23:52:20 its like there's overscan Oct 06 23:52:44 Corners: +0+0 -0+0 -0-0 +0-0 Oct 06 23:52:44 -geometry 1920x1080+0+0 Oct 06 23:52:52 Screen 0: minimum 1920 x 1080, current 1920 x 1080, maximum 1920 x 1080 Oct 06 23:52:57 but everything says 1920x1080 Oct 06 23:53:09 then you have 1920x1080 @ low refresh rate Oct 06 23:53:20 tv thinks you're a video source, and you get overscanned Oct 06 23:54:04 the tv may have a setting to fix that Oct 06 23:54:10 didnt see one bug will dig more :/ Oct 06 23:54:23 it's not really a bug Oct 06 23:54:41 overscanning is normal for video sources Oct 06 23:55:49 my TV has a "game mode" setting or something like that to fix it Oct 06 23:56:04 there shouldnt be any overscan on a digital source Oct 06 23:56:06 overscan is archaic Oct 06 23:56:17 samsung tv is in game mode, found a fix though Oct 06 23:56:18 http://www.storageforum.net/forum/showthread.php/8905-How-to-Disable-overscan-on-a-Samsung-quot-SMART-quot-tv-solved Oct 06 23:56:33 In the end it turns out the solution was exactly what all the forums said, you need to plug the PC in to the HDMI port labeled DVI and then change the name of that port to PC/DVI Oct 06 23:56:39 ah yes Oct 06 23:57:00 sorry, it's been a while since I configured my tv for it (also samsung) but you're right Oct 06 23:58:17 it's not just for the overscan though, also disables other stuff to "polish up" video sources but is incredibly annoying when using it as a monitor Oct 06 23:58:33 (especially since the processing adds noticable delay) Oct 07 00:06:23 nyt: I'm curious what fps you got then Oct 07 00:06:31 can't be very high at that res Oct 07 00:06:35 i believe 24fps Oct 07 00:06:43 according to wiki Oct 07 00:06:47 and feeling the mouse stutter Oct 07 00:06:51 im not at the location that it's installed now Oct 07 00:06:53 just vpn'd into the bbb Oct 07 00:06:55 ah Oct 07 00:07:03 let me see if i can pull that info Oct 07 00:09:51 xrandr not showing :/ Oct 07 00:09:54 if you want all the gory details, I wrote a little util that pulls the raw config from the lcdc -> http://gerbil.xs4all.nl/lcd-util.tgz Oct 07 00:10:20 1920x1080 0.0* Oct 07 00:10:34 also lets you reconfigure the clock polarity (press ctrl-C instead of enter to _not_ change the setting) Oct 07 00:10:34 0.0 lol Oct 07 00:10:52 well, elinux wiki says max it can do is 24hz at 1920x1080 Oct 07 00:10:52 so Oct 07 00:10:54 in my experience the reset default is actually wrong Oct 07 00:11:01 mouse wasn't smooth as 60hz, im going to run with that its at 24hz Oct 07 00:12:48 (pressing enter cycles through four options for clock polarity, so you can check which produces cleanest results... typically there's also a one-pixel shift resulting from one or two of the choices) Oct 07 00:14:08 the shift was my original motivation to write the util... I had a screen where the rightmost column of pixels actually ended up as the first column of the next line, and the last pixel ended up at the topright of the next frame Oct 07 00:14:25 topleft Oct 07 00:14:26 sorry Oct 07 00:20:50 I force a 1440x1080R@50 mode myself on a 1080p monitor at work Oct 07 00:21:49 i.e. reduced width, but the crispness of native resolution and adequate fps Oct 07 00:22:56 discovered that mode fails with -bone kernels though, requires a -ti kernel (due to issue in the clock divider calculation) **** ENDING LOGGING AT Wed Oct 07 02:59:58 2015