**** BEGIN LOGGING AT Fri Mar 10 03:00:04 2017 Mar 10 03:00:26 are they at the same temperature? Mar 10 03:06:13 not sure Mar 10 03:06:56 maybe I should indeed hunt for one of the on-board temperature sensors and log that info too Mar 10 03:13:48 I wonder if the DLLs and such run on core supply or the ddr3 supply Mar 10 05:50:47 uhhhhhh.... waitwhat? Mar 10 05:56:12 ah, pebkac Mar 10 08:48:38 hi Mar 10 08:48:56 Is anybody here? This is the first time i am using IRC Mar 10 08:49:18 I have some questions about beaglebonegreenwireless Mar 10 08:49:35 I have some questions about beaglebone green wireless Mar 10 08:50:41 1. how do we connect to the serial port? I know we can connect through ssh, but i am considering to reflash the image with the new image that i will build with yocto. So, I am curious to know how can i use the serial port. Mar 10 08:50:59 there seems to be only one micro usb port available and that is being used for power Mar 10 10:19:37 mantralasan: presumably the BBGW still has the normal serial console header? Mar 10 10:21:54 http://mysticengineering.com/debug.buddy/images/beaglebone-black-serial.jpg Mar 10 12:31:15 can we install ubuntu on bb ? Mar 10 12:52:57 LK: yeah, see http://elinux.org/BeagleBoardUbuntu Mar 10 14:42:16 zmatt: do you also think this am335x suspend scheme is a bit odd? Mar 10 14:43:34 zmatt: I've re-read the sources time and again but it's pretty strange that the M3 races with the A8 to deep sleep. Mar 10 14:59:25 yes it's absolutely not how I understood the docs Mar 10 15:01:55 how it should work IMHO is that you set up a command in the IPC registers, swith MPU modulemode to 0, then call WFI Mar 10 15:02:14 what happens if the m3 tries to disable the l3 interconnect while the mpu isn't sleeping? or if the mpu wakes itself up while the l3 is disabled? Mar 10 15:02:29 exactly Mar 10 15:02:34 I'm not getting it Mar 10 15:02:46 it was meant as a suggestion of the next thing to try :) Mar 10 15:03:30 it is quite doable to fix the race conditions in this though Mar 10 15:04:07 but the best way depends on how exactly things behave Mar 10 15:04:47 zmatt: well what I'm seeing is that the M3 seems to disable the control module and so I cannot even write to m3_tx_ev_eoi Mar 10 15:04:59 I get a data abort immediately Mar 10 15:05:07 yeah that one I saw coming Mar 10 15:05:08 hello. Where can i find a c or c++ working sample about using the serial ports of a BBB running the current image? Mar 10 15:05:23 everything i find seems not working / obsolete.... Mar 10 15:07:29 Parduz: once you've made sure the uart is enabled and pinmux is setup (which is system setup, not something you do in your application), using the serial port is no different using serial ports in general on linux Mar 10 15:08:31 enabling and pinmux setup can be done 1. using cape-universal and the config-pin utility, or 2. via a device tree overlay, or 3. by customizing the main device tree Mar 10 15:08:34 whichever you prefer Mar 10 15:08:52 yep, UART is working if i use a terminal Mar 10 15:09:35 my problem is that i'm a linux noob, but a win programmer from 20 years. Mar 10 15:09:49 so i'm trying to grasp what to do looking for samples. Mar 10 15:10:12 but i think i'm missing some fundamental point, as nothing works Mar 10 15:11:28 this is what i was trying: Mar 10 15:11:29 https://press3.mcs.anl.gov/forest/hardware/beaglebone-black-software/examples-and-tutorials/ Mar 10 15:12:48 and this is what i've done in the uEnv.txt to enable the serial: Mar 10 15:12:50 cape_enable=bone_capemgr.enable_partno=BB-UART2 Mar 10 15:12:58 which seems working Mar 10 15:23:57 zmatt: what I'm also not getting is the interrupt from A8 to the M3 on A8 executing wfi Mar 10 15:24:31 zmatt: however I know that MPU is off, I can peek the PM_PER registers through the M3 Mar 10 15:27:16 thinkfat: you mean when MPU has modulemode=off but an interrupt still pending? Mar 10 15:28:21 Parduz: ok, if it's working in the terminal then note that you're not looking for beaglebone-specific info anymore, just "using a serial port on linux" Mar 10 15:33:41 zmatt: no, no interrupt pending Mar 10 15:34:00 ok.... Mar 10 15:34:06 zmatt: if there's an interrupt pending the wfi immediately returns and there is transition to OFF Mar 10 15:34:08 thnks Mar 10 15:34:50 zmatt: well strictly speaking it's not an interrupt from the A8 to the M3 but a PRCM interrupt to the M3 Mar 10 15:35:13 zmatt: but it's not happening, I have a breakpoint on the m3 in the exception handler and it's not firing Mar 10 15:35:38 thinkfat: of course it's not, the a8 never enters wfi state since it's falling through the wfi Mar 10 15:35:51 what do you mean by "and there is transition to OFF" though? Mar 10 15:38:04 zmatt: the MPU PM goes OFF Mar 10 15:38:42 huh, how? Mar 10 15:39:00 PM shouldn't allow that if MPU is active Mar 10 15:41:05 well it's not. Mar 10 15:41:22 it's what I'm trying to tell you. the MPU is off, but the interrupt to the M3 has not happened. Mar 10 15:42:33 that... is not good Mar 10 15:43:23 it's not a big problem either though Mar 10 15:45:18 it is very weird however Mar 10 15:45:41 I must be missing some setup step Mar 10 15:45:43 since WFI when an interrupt is pending should execute as a nop Mar 10 15:45:56 where did I say there's an interrupt pending? Mar 10 15:46:09 oh then I misunderstood Mar 10 15:47:09 ahhh, now I see where I misparsed you Mar 10 15:48:06 I parsed "16:34 < thinkfat> zmatt: if there's an interrupt pending the wfi immediately returns and [then] there is transition to OFF" Mar 10 15:49:01 overlooking what you said right above is... "no, no interrupt pending", .. "if there's an interrupt pending the wfi immediately return[, whereas] there is transition to OFF [hence clearly no interrupt is pending]" Mar 10 15:49:20 sorry, now I'm following :) Mar 10 15:50:04 ok, so this is WFI while modulemode=off and requested power domain state is off? Mar 10 15:50:19 or, what were the conditions exactly? Mar 10 15:50:28 what you said Mar 10 15:50:35 and the clock domain? Mar 10 15:50:52 PM_MPU_PWRSTST is 0 Mar 10 15:51:09 that should cover the clock domain, I guess Mar 10 15:51:22 well not its requested state Mar 10 15:52:41 my understanding is that CM_MPU_CLKSTCTRL.CLKTRCTRL should be SW_SLEEP to allow the power domain to enter a low power state Mar 10 15:54:18 of course just because that's how it's *supposed* to work very obviously doesn't mean that it is how it actually works ;) Mar 10 15:56:09 iirc that's how it worked when I tested prcm on the dm814x Mar 10 15:59:11 zmatt: that's what the M3 firmware does Mar 10 16:01:51 then I'm still unclear about the scenario and/or observations, since you said the m3 isn't getting the irq that would normally trigger it to do such things Mar 10 16:03:10 zmatt: well, there are two irqs actually. you can write to the mbox and cause an IRQ, that's what the suspend code in linux does, and also the PRCM(?) should interrupt the M3 once the A8 enters WFI state with modulemode=off Mar 10 16:03:38 zmatt: that latter is what I'm not seeing Mar 10 16:07:42 does the a8 actually disable modulemode in this case, or does it just send the message and then enter wfi and let the m3 do the rest? Mar 10 16:07:42 ah Mar 10 16:09:39 ah? Mar 10 16:14:28 forgot something: PRM_IRQENABLE_M3 Mar 10 16:14:42 hehe, that was the other thing I thought of but hadn't asked yet Mar 10 16:21:14 ok, it seems I can get the A8 to sleep and wakeup :-) Mar 10 16:21:49 and then it's killed by trying to output something on UART0, which is idled :-) Mar 10 16:24:26 yes of course, it should either reenable those itself or leave it to the m3 but wait for an ok from there Mar 10 16:40:05 ok, gotta run, weekend starts Mar 10 18:50:05 Hello, "VDD_CORE" subsystem refer to. I mean in http://processors.wiki.ti.com/index.php/AM572x_General_Purpose_EVM_HW_User_Guide#APM_Sense_Resistors, If I want to monitor power of the CPU I monitor for VDD_MPU. But what about VDD_CORE? Mar 10 20:11:53 Catzilla: see chapter 3.4 PRCM Subsystem Integration of the TRM for a detailed listing of what's on which voltage rail Mar 10 20:12:20 though a brief summary would be, "almost all core logic, except those that have dedicated core supplies" Mar 10 20:12:29 gm Mar 10 20:13:03 zmatt: found some detail I overlooked before, and I change I made in the firmware that I'd rather not have done :) Mar 10 20:13:36 mhh? :) Mar 10 20:13:49 I misunderstood the suspend handling completely Mar 10 20:14:04 it's actually a two-stage process Mar 10 20:14:20 there's a prepare step and an execute step Mar 10 20:14:45 the prepare step is done through a mailbox interrupt and just set the stage Mar 10 20:15:09 the actual execute step is done when the A8 enters wfi and wakes up the M3 Mar 10 20:15:44 you mean the a8 waits for acknowledge of the first step from the m3? Mar 10 20:15:45 that way it's actually race free and none of the ill effects i have observed should be happening Mar 10 20:16:02 yes, it waits for an interrupt from the m3, then it enters wfi Mar 10 20:16:20 that sounds very inefficient an unnecessary? Mar 10 20:17:25 indeed, because the first step does nothing but enable the prcm interrupt in the M3 Mar 10 20:17:59 then it goes back to sleep again and when the prcm interrupt wakes it up again, it executes the sleep program Mar 10 20:19:18 I wonder how much time all this times... TI bothered to hook the STM up to the L4WK for logging purposes, but I don't think they ever bothered implementing use thereof did they? Mar 10 20:25:34 zmatt: i have one other problem left to solve: when the wakeup event arrives, the m3 wakes up allright, but then it catches a hard-fault exception trying to re-lock the PLLs Mar 10 20:26:57 why / under what condition? Mar 10 20:27:27 that sounds odd actually, prcm should always be accessible Mar 10 20:28:49 also, hardfault? but why? Mar 10 20:29:49 that suggests it got a fault on top of a fault, but its fault handlers are deadloops Mar 10 20:30:31 hmm, which m3 revision is it btw? Mar 10 20:30:38 2p0, I think Mar 10 20:31:26 zmatt: the m3 tries to access something in the control module, and i think that is idled Mar 10 20:34:05 who did that, the a8 or itself? if it did it itself then it's just an idiot, if it's the a8 then they need to have a talk about who's responsible to turning off the lights where. and still, why hardfault? it has a busfault handler that should be taken Mar 10 20:36:39 racing to turn off modules is not a problem if it's modules that neither of them need to access at that point anymore, or if the module will be kept enabled by prcm anyway for some reason until the last user is done Mar 10 20:36:40 zmatt: well, that is the call trace I get with jtag Mar 10 20:36:59 zmatt: wakeup -> pll-relock -> hardfault Mar 10 20:38:33 what does the hardfault status register say? Mar 10 20:39:32 (0xE000ED2C) Mar 10 20:43:00 I'll have a look when I have the board connected again Mar 10 20:43:28 dumping all of 0xE000ED24 - 0xE000ED38 would be useful for fault decoding purposes Mar 10 20:44:25 what did you do to trigger this bug? Mar 10 20:46:26 zmatt: I tried waking from DS1 Mar 10 20:46:39 zmatt: not sure someone ever verified that Mar 10 20:48:21 hmm, it would be nice if openocd had built-in decoding of faults when trapping a fault (vector catch, lockup, or just noticing entry into a fault) Mar 10 20:48:34 sounds strange to me if noone ever verified that Mar 10 20:48:39 in all this time Mar 10 20:50:26 zmatt: well, no idea. Linux uses DS0 and the PDK doesn't seem to support DS1 either Mar 10 20:50:37 hmmm Mar 10 20:50:59 hmmmm Mar 10 20:51:16 strange, nothing in my software seems to ever touch control_sysconfig Mar 10 20:51:55 ds1 is apparently implemented in starterware according to http://processors.wiki.ti.com/index.php/AM335x_Power_Consumption_Summary Mar 10 20:57:11 ds0 vs ds1 really shouldn't matter from the m3's point of view, all modules it can access are in pd_wkup, not pd_core Mar 10 20:58:46 zmatt: for me it's an intermediate step, something to verify the infrastructure without going through the pain of shutting down and restoring PER :) Mar 10 20:59:42 I understand, but ds0 has been validated in linux right? Mar 10 21:00:57 I can imagine it might be easier in some ways since you know exactly in what state you'll find the peripherals that are in pd_per Mar 10 21:01:20 no that's not true actually, it may have been cancelled at the last moment Mar 10 21:01:44 (that's an interesting test case to deliberately trigger though) Mar 10 21:02:15 ds0 entry falling through the wfi Mar 10 21:03:54 btw do you have any experience with the "Software supervised transition completed" irq from prcm? I wonder when exactly that triggers Mar 10 21:06:30 zmatt: no, and the documentation is very "lean" about these aspects Mar 10 21:06:45 that part I noticed Mar 10 21:07:22 I'm also trying to remember if the slave port of a module is accessible if modulemode=off but idle transition is prevented by sysconfig.idlemode=no_idle Mar 10 21:31:24 which is the best package I can get? Mar 10 21:56:57 which is the best package I can buy? Mar 10 21:57:07 depends on your needs Mar 10 21:57:12 what's the best color of the rainbow? Mar 10 21:58:40 alright I see..well I wanna use it for my third year project at university for advance robotics Mar 10 22:00:04 any ideas? Mar 10 22:01:36 i think there's a variant of the beaglebone black specifically targeted at robotics Mar 10 22:02:25 http://beagleboard.org/boards Mar 10 22:02:31 that lists some of the differences between boards Mar 11 01:23:42 bwlergh, gcc still produces abominable code when you use bitfields :( **** ENDING LOGGING AT Sat Mar 11 03:00:01 2017