**** BEGIN LOGGING AT Sat Sep 19 02:59:59 2015 Sep 19 08:48:25 Hello, after "opkg update" and "opkg upgrade", I can't communicate with it through USB Sep 19 08:48:32 is there any tipp? Sep 19 08:52:18 Tipp? Sep 19 08:53:32 Oh my god. All are idled! Sep 19 08:53:37 unbelieveable Sep 19 08:53:49 But thanks for this plattform Sep 19 09:04:12 gfast: try again later this day Sep 19 09:56:33 hey guys, anyone knows details about the AM5728 IPU1/2? is there an ISS like in the OMAP5432? Sep 19 12:29:39 Hello Sep 19 12:33:58 hi, is posting a URL i could use some help with allowed here? Sep 19 12:35:32 i guess its allowed? the EFIDroid(bootloader for android/linux) developer may implement uboot support for any SBC(Beagle,Cubieboard,RaspberryPi,ODROID and others) with uboot and possibly others if he recieves enough requests so you can switch between android or any other device supported OS without having to modify boot.ini every boot - Sep 19 12:35:33 https://github.com/efidroid/projectmanagement/issues/3 just add a +1 and your device name, URL, and if possible URL to the uboot source for your device; if its one thats allready on the list just add +1 and model Sep 19 18:24:00 getting error: No protocol specified. cannot display :0.0; Need help Sep 19 18:25:45 I tried both the angstrom dist and the Debian dist. Same problem. Sep 19 21:57:24 hi Sep 19 21:58:18 why VDD_5V doe not work? Sep 19 23:25:15 (probably because you're powering the BBB via USB? why people doe not stick around for a while after asking question...) Sep 19 23:52:54 zmatt .. why do people insist on thinking the best way of accessing the beagle is via the otg connector .. oh we have a usb client cable supplied with it .. Sep 20 00:05:43 it's not an otg connector Sep 20 00:05:54 otg is micro-AB, the BBB has mini-B Sep 20 00:11:11 we actually also commonly power BBBs via usb, since we have tons of usb cables and insufficient 5V DC adapters with the right plug Sep 20 00:11:58 (or actually, I *know* we should have a heap of those somewhere in our building, I just haven't been able to find the box they're in) Sep 20 00:17:13 lilCodie: I don't understand the proposal... Sep 20 00:19:14 lilCodie: I'd never heard of EFIDroid before, but I guess it's an EFI(-like) implementation for ARM, and apparently used to boot Android ? Sep 20 00:23:06 its basically like grub - so you can switch between adroid/ubuntu/other distro at boot without having to modify the uboot setup Sep 20 00:24:08 well u-boot supports that too actually Sep 20 00:24:44 though not graphically Sep 20 00:24:56 yeah - plus youd have to do something like http://omappedia.org/wiki/Multiboot_using_u-boot Sep 20 00:25:58 but then just replace u-boot... I'd cheer you on for that, if the result sucks less than u-boot at least Sep 20 00:29:33 if you only target reasonably modern ARM systems (rpi 1 is on the borderline for that) then that will also simplify things considerably compared to u-boot's desire to support almost any device in the world, most of them ancient and crufty Sep 20 00:29:38 well - if something is made that is more modular so that UEFI or UBOOT (or whatever else your SBC/PC uses) can boot it would enable it to do things uboot cant do like USB based ethernet booting - and still work on a wide range of devices - since some SBC companys use custom uboot code(like ODROID/Hardkernel) Sep 20 00:29:57 true Sep 20 00:30:47 note btw that on the AM335x, booting via ethernet or usb (rndis) is actually supported by ROM already Sep 20 00:31:09 that's usb as device though, not sure what you meant with usb-based ethernet booting Sep 20 00:31:27 I guess you might mean booting over the rpi's horrible ethernet Sep 20 00:31:41 ^ or over the ODROID XU4s gigabit ethernet Sep 20 00:32:06 I pity you already for having to deal with musb ;-) Sep 20 00:33:23 xD thats why we are trying to see how many people might actually use EFIDroid w/ uboot Sep 20 00:33:37 hence the +1s and uboot source ect Sep 20 00:34:10 ethernet on the am335x is a piece of cake to use though (*don't* look at the linux kernel driver or the u-boot driver, they're horrible implementations) Sep 20 00:35:41 how is the am335x havent had the joy of using one yet Sep 20 00:36:19 or most of the BB products, doing that soon though - friend has a DIY hydroponics setup im helping him with, going to do ph and other sensors on the BBB Sep 20 00:37:14 not as nice as the related dm814x (that's the first TI processor I worked with and done a lot of baremetal coding on), but it's still quite nice Sep 20 00:37:30 usb sucks... but that's more of a general statement than one about the am335x Sep 20 00:37:45 xD Sep 20 00:38:02 the main disappoint was what happened to PRCM (Power, Reset, and Clock Management) Sep 20 00:39:02 the dm814x uses simplified omap4 PRCM... not any of the fancy hardware-managed stuff the OMAPs have, only manual software control, but simple and clean Sep 20 00:39:06 well-organised register blocks Sep 20 00:40:00 the am335x uses the same PRCM, but organised by throwing bits into registers while blindfolded and drunk Sep 20 00:41:21 it has an interesting arrangement for sleep modes though, haven't played with that yet Sep 20 00:41:53 (the am335x has a separate cortex-m3 for that) Sep 20 00:42:44 the big benefit of the TI SoCs is documentation... even though it's still not as complete as I'd prefer Sep 20 00:43:09 hmm interesting Sep 20 00:43:22 I mean, I used to complain about TI docs (still do from time to time) Sep 20 00:43:45 but then I looked at the sad little "manual" that's published for that broadcom thing on the rpi Sep 20 00:44:06 and the community-maintained errata on a wiki because broadcom doesn't publish any Sep 20 00:44:58 whereas TI has the link to the errata prominently near the top of the product page, right under the link to the datasheet... this is something I really praise them for Sep 20 00:45:18 yeah >.< they make you pay for licensing fees for the full dox if i remeber right Sep 20 00:45:26 TI by comparison is heaven Sep 20 00:46:02 yup, for the AM335x you can just download the TRM, don't even need free registration or anything Sep 20 00:46:07 and that's ~5000 pages Sep 20 00:47:16 some people just love the devil they know(broadcom/Rpi or other things) i know ive been guilty of that sometimes Sep 20 00:47:56 ask em to learn something new n theyll run xD Sep 20 00:48:02 I suspect plenty of rpi users don't even know the ARM processor isn't the boot cpu :P Sep 20 00:50:06 (or why the part code "bcm2708" shows up in various places even though it's a bcm2835 (rpi1) / bcm2836 (rpi2) soc) Sep 20 00:52:26 (also, if you see "bcm2709" anywhere, go forth and fix that to bcm2708, which may require fixing something else that assumes bcm2708 implies anything about the ARM subsystem present -- it doesn't) Sep 20 00:55:26 (bcm2708 == videocore 4 + peripherals) Sep 20 00:57:10 a preference for what you're familiar with isn't too strange though Sep 20 00:57:33 given that any chip and any platform will have its... eccentricies Sep 20 01:00:20 yeah - at the moment i use the Samsung's Exynos 5422 - its a HMP chip (basically 4 big + 4 LITTLE cores) Cortex™-A15 2Ghz and Cortex™-A7 Sep 20 01:00:52 don't get too spoiled with that much CPU power XD Sep 20 01:01:45 xD ikr Sep 20 01:01:54 64gb EMMC 2gb ram Sep 20 01:02:05 and usb 2 + 3 Sep 20 01:02:10 so bit spoiled Sep 20 01:02:24 on the other hand, there are more metrics by which you can measure a chip Sep 20 01:03:20 I wonder what its fastest reaction time is (say if you make a general-purpose output follow a general-purpose input) Sep 20 01:04:01 and compare that to the AM335x PRUSS :) Sep 20 01:04:38 havent tried :P n right now its running ubuntu - kinda bloated Sep 20 01:05:09 its nice though, it can run run my web ide, desktop enviroment and anything i want and fits in the palm of my hand and has a screen on it for my debug outputs if i want to run it headless - or hdmi or vnc or ssh if i need to mess with it Sep 20 01:06:06 PRUSS contains two cores of a custom 32-bit risc architecture, running at 200 MHz even though they're _non-pipelined_ Sep 20 01:06:37 nice - didnt know it had that Sep 20 01:06:38 so a 32-bit compare-and-branch takes 1 cycle, regardless of whether the branch is taken or not Sep 20 01:06:59 and it has gpio wired directly onto cpu registers Sep 20 01:10:13 all of the GPIOs? Sep 20 01:12:19 not all processor GPIOs are PRU GPIOs... in fact those are different pinmux settings (since PRU GPI/GPO obviously doesn't go through the main GPIO controllers) Sep 20 01:14:32 each PRU core has 16 GPOs and 17 GPIs but iirc you can't use all of them, at least on a BBB Sep 20 01:14:37 I'd need to check the pinmux lists Sep 20 01:14:51 hmm interesting Sep 20 01:16:06 of course the PRU cores can also access normal GPIOs, but at nowhere near the same performance of course... and since you'd go via the main interconnect you also lose predictable timing Sep 20 01:16:42 the reason the PRU cores are non-pipelined and have such direct IOs is because they were designed to simulate all kinds of bus protocols in software Sep 20 01:20:02 so SPI I2C ect? Sep 20 01:20:22 more like obscure industrial shit like profibus, modbus, etc Sep 20 01:20:45 I mean, they can do SPI too obviously Sep 20 01:20:59 (in fact, at higher data rates than the actual SPI peripherals) Sep 20 01:21:33 xD whats the fastest bus it can emulate in software Sep 20 01:23:04 well they're also used for EtherCAT Sep 20 01:23:57 in which case they're connected to MII Sep 20 01:24:43 so they can apparently deal with real-world data at at least 25 MHz Sep 20 01:26:29 but you can also configure the GPI and/or GPO to serial mode, in which case I think they go to 100 MHz Sep 20 01:27:40 (or possibly the full 200 MHz, but you can't output that clock without violating the datasheet) Sep 20 01:28:18 you'd have to ask someone with experience with PRUSS what the limits are you can push them to Sep 20 01:28:23 pfft rules are made to be broken :P Sep 20 01:30:32 oh it's quite possible it works but just violates the timing values stated in the datasheet Sep 20 01:32:01 do you have to sign an NDA at all with TI to get datasheets? Sep 20 01:32:48 http://www.ti.com/product/am3358 Sep 20 01:33:09 datasheet, errata, TRM links at the top of the page Sep 20 01:33:10 :P Sep 20 01:33:26 of course it's not like that with all their SoCs Sep 20 01:34:05 some are... slightly less open... http://www.ti.com/product/dra626 Sep 20 01:34:50 and public TRMs don't include all info Sep 20 01:37:02 hows it going? Sep 20 01:37:38 the crypto accelerators almost never documented for example Sep 20 01:38:02 even though some of them have linux drivers :P Sep 20 01:38:26 the debug subsystem is often also poorly documented, if at all Sep 20 01:38:48 xD Sep 20 01:38:57 the AM335x TRM is also unusually uninformative about the L3 and L4 interconnects Sep 20 01:39:13 you can find quite good docs on those in the DM814x TRM though Sep 20 01:41:26 it only lacks documentation on the L3 firewalls... you can find documentation about those e.g. in the omap4 TRM, though if you ever want to configure those I suggest poking me about them since in DM81xx and later processors there are some details that aren't in the omap4 trm ;) Sep 20 01:43:35 the best public docs on the debug subsystem can be found in the DM816x TRM, and a scrap of info in the AM335x TRM Sep 20 01:44:07 (the DM816x, DM814x, and AM335x all have the same DebugSS copy/pasted) Sep 20 01:44:25 ill have to look at those sometime, so what are you up to tonight? Sep 20 01:45:22 finishing my exploration of exception handling on ARM EABI (+ the parts implemented by everyone yet documented nowhere) Sep 20 01:45:44 which was triggered by wanting to throw C++ exceptions from inside signal handlers (which works now) Sep 20 01:47:16 but I've come to the conclusion that Paul Brooks was wrong... which surprises me since it looked like he's one of the few people on earth who actually know how this shit works Sep 20 01:47:33 cool - im just watching the new espisode of doctor who and mucking about with the xu4 Sep 20 01:48:29 (in that a custom "personality function" isn't necessary for the signal restorer function) Sep 20 01:48:41 how's the docs for that exynos soc? Sep 20 01:48:54 no clue, havent signed the NDA yet Sep 20 01:49:04 ok, so no public docs Sep 20 01:49:15 atleast for the chip the xu4 uses Sep 20 01:49:22 others do i think Sep 20 01:50:13 too bad, the specs and price look interesting, but no documentation is a big turn-off (no, I won't sign an NDA) Sep 20 01:51:34 well i just emailed them about the NDA - maybe i can get you a semi redacted version :P Sep 20 01:51:55 nah, I don't have time to play with more hardware anyway Sep 20 01:54:39 I've already cursed the guy who made me aware of this thing's existence -> http://inversepath.com/usbarmory ... since now I want to play with one Sep 20 01:56:51 ikr Sep 20 01:57:51 it's rare to encounter a SoC where the OTPs are still available... and documented Sep 20 01:58:05 I guess the iMX53 is old enough that freescale doesn't care anymore Sep 20 01:58:47 <3 that things device emulation Sep 20 01:58:50 imx53 is pretty old .. but freescale stuff is better documented and supported imho Sep 20 01:59:51 veremit: still, for the iMX53 they really just dumped internal docs onto the internet Sep 20 02:00:10 like " can be overriden by a metal layer 6 change" Sep 20 02:01:15 like, okay, lemme get my dremel and reaaaaly fine-tipped soldering iron ;-) Sep 20 02:02:04 hot air rework .. tis the way :D Sep 20 02:02:20 but I gotta go do some pcb routing in a few ... Sep 20 02:03:54 for the prototype of the rad1o (the CCC badge this year) they actually had to do a patch involving drilling through the pcb under an angle to access a BGA ball Sep 20 02:04:01 -respect- Sep 20 02:04:36 hi, still no one out there with information about the extention headers of the x15? Sep 20 02:06:04 eww. lol nice :D Sep 20 02:06:04 * ds2 makes a note to shove as many critical traces under unused balls as possible Sep 20 02:06:28 uavcamm: I think once that info is released it'll circulate fast enough that'll be hard to miss if you're tracking info about the x15 at all Sep 20 02:08:25 lilCodie: so, just curious... how bloated is this efidroid thing? :) Sep 20 02:09:22 since intuitively I'd suspect EFI isn't particularly lean Sep 20 02:10:22 sure. butt it would be nice to do some work before the whole circut is released... but i think i have to wait :) Sep 20 02:11:21 it started out being named as grub4android - and i just started trying to help him get support and maybe help him code it - assuming by the name hes using grub as a reference - but who knows Sep 20 02:11:56 trying to get my xu4 dev enviroment back running at the moment (wiped it a bit ago) Sep 20 02:12:14 well grub and efi are rather unrelated codebases... I suspect it's more likely he dumped one in favor of the other Sep 20 02:14:53 uavcamm: if you want to design an expansion board already in advance of release, it's quite possible such info may be available under NDA Sep 20 02:15:21 ill have to look at the code - heres the github https://github.com/efidroid Sep 20 02:18:23 looks like http://www.tianocore.org/edk2/ Sep 20 02:18:57 yup, that's what I was looking at also Sep 20 02:20:34 https://github.com/tianocore/edk2/tree/master/BeagleBoardPkg < interesting Sep 20 02:21:05 like I said, I don't see why you'd want to be chainloaded by u-boot Sep 20 02:21:30 the real work is in the hardware drivers you need anyway Sep 20 02:21:58 compared to that, platform initialization is almost nothing, at least on TI processors Sep 20 02:22:10 basically just the memory controller Sep 20 02:22:53 wondering how easy it would be to do those since the uboot code for the xu4 is well...bloated at the least hardware wise atleast https://github.com/hardkernel/u-boot/tree/odroid-v2010.12/drivers Sep 20 02:22:55 and optionally some PLLs / clock mux, although ROM already setups up sensible defaults so that's not really needed Sep 20 02:24:14 'sensible'? he he he he he he Sep 20 02:25:03 ds2: core/peripheral PLLs are at their final value unless you want to go OPP50, MPU PLL is at a safe value that's fast enough for a bootloader Sep 20 02:26:12 so, what am I missing? Sep 20 02:27:10 ddr and display PLLs still need to be setup with platform-specific values (in fact display PLL setup should be deferred until you got EDID data) Sep 20 02:28:03 lilCodie: I've set up the LCD controller from baremetal code, that was quite easy Sep 20 02:29:00 memory controller is also not much more than writing the correct values to config registers Sep 20 02:29:20 functional vs sensible Sep 20 02:29:45 ds2: what would you consider sensible then? Sep 20 02:30:49 rom only does things with stuff it uses Sep 20 02:31:08 IIRC - the clocks to the RGMII stuff is not sensible on the am33x Sep 20 02:31:38 ethernet boot works, so apparently they are Sep 20 02:31:40 had to chase down a mysterious bug where code to init stuff was lost from u-boot Sep 20 02:31:46 not when gige is involved Sep 20 02:32:03 ok that's possible, I don't have an am335x with rgmii Sep 20 02:32:41 the bug was things worked with TI uboot but not upstream Sep 20 02:32:47 kernel didn't touch that clock Sep 20 02:33:19 that doesn't sound like ROM was involved at all Sep 20 02:34:42 in any case, doesn't have much to do with PLL setup Sep 20 02:35:26 since the core PLL isn't a low-jitter PLL and not usable for rmii (known erratum) I can't imagine it's usable for rgmii either? Sep 20 02:37:17 how did you get gige on a beagle, ds2?1 :p Sep 20 02:37:37 I'm guessing it's am335x hardware other than a bbb Sep 20 02:37:43 ;) zmatt Sep 20 02:41:05 ds2: hmm, though I don't see any alternative... odd Sep 20 02:42:52 I guess rmii could easily be more jitter-sensitive than rgmii due to a single clock being used for tx and rx Sep 20 02:43:12 rgmii just includes the clock with data along each direction Sep 20 02:46:56 ds2: unless you can be more specific I'm afraid I'm going to dismiss your complaint as invalid... as far as I can tell the PLL setup done by ROM is correct Sep 20 02:47:16 i think he went afk :P Sep 20 02:49:54 * zmatt shrugs Sep 20 02:51:54 veremit: HW mods Sep 20 02:52:20 lilCodie: there's good reason though for the BBB bootloader to _not_ include any driver for graphics output... plenty of people use those pins for other purposes Sep 20 02:52:27 zmatt: the fix was porting 2 lines of constants to shove into a PLL Sep 20 02:52:42 i wish I had my notes handy on what the exact change was. this was stuff from last year Sep 20 02:53:09 ds2: then I'd be very curious which values Sep 20 02:53:16 zmatt: the symptoms were if you probe the RGMII clock line, it turns out to be something other then the right value (50MHz?) Sep 20 02:53:30 zmatt: the values were from cut and pasting the equiv. code on the TI U-boot Sep 20 02:53:41 sounds tracable Sep 20 02:53:54 iirc rgmii clock frequency actually depends on link speed Sep 20 02:54:04 all I am claims is the ROM does not init it to the right value Sep 20 02:54:05 but not 100% sure Sep 20 02:54:12 s/right/sensible/ Sep 20 02:54:26 IIRC, yes but it is different from the MII clock Sep 20 02:54:32 since RGMII is always DDR Sep 20 02:54:38 well tbh my first suspicion would be that it does but u-boot then fucks up the settings Sep 20 02:54:44 that sounds very u-boot to me Sep 20 02:54:57 the stock (broken) u-boot didn't do anything Sep 20 02:55:06 i had thought it was the case Sep 20 02:55:39 zmatt: for now we were looking at doing the boot process via UART or GPIOs(Dipswitch maybe) so it could be completely headless Sep 20 02:55:59 there aren't that many upstream u-boot users on gigE HW Sep 20 02:55:59 lilCodie: uart is fine Sep 20 02:56:50 the real solution is the kernel should init things properly rather then relying on the precursors to do it Sep 20 02:57:09 and thanks to the @#%@#$$#@%$#@#$ changes, the old u-boot couldn't boot the later builds Sep 20 02:57:20 ds2: yes and no, the kernel should not reinitialize the ethernet switch subsystem at all if it's properly initialized Sep 20 02:57:40 zmatt: this is the clock stuff Sep 20 02:57:52 not the ethernet subsystem Sep 20 02:58:08 yes, so that part should be safe Sep 20 02:58:23 the cpsw driver should tossed in the garbage and rewritten from scratch though Sep 20 02:58:34 but Sep 20 02:58:43 on a tangent.... Sep 20 02:58:47 I'm going to grab jtag and check the PLL values at boot :P Sep 20 02:59:31 do you know how well does the am335x ethernet stuff work ina pure switch mode? where the cpsw is just a 3 port switch Sep 20 02:59:39 Cortex-A8 + 2 physical external port Sep 20 02:59:49 it works perfectly on the dm814x **** ENDING LOGGING AT Sun Sep 20 02:59:58 2015