**** BEGIN LOGGING AT Wed Nov 06 03:00:06 2019 Nov 06 12:49:30 m Nov 06 13:53:23 rah: with both caps (C24 and C30) removed, no pull-up: 0 failures in 13891 power cycles Nov 06 14:55:00 rcn-ee[m]: in my device tree, I have vdd_ddr as vin for the vtt_fixed regulator, but you have vdd_3v3. do you know which is right? Nov 06 14:55:49 look like software wouldn't even touch it. Nov 06 15:00:10 jkridner: for the ai? it this u-boot vs the kernel, i thought they were in sync? Nov 06 15:00:42 u-boot. I might have grabbed the wrong one. I'll show you my patches in a minute. Nov 06 15:01:36 I'll lean towards what is in https://github.com/beagleboard/BeagleBoard-DeviceTrees/blob/v4.14.x-ti/src/arm/am5729-beagleboneai.dts Nov 06 15:01:55 ah, yes, the uboot dts is back from a few months ago. Nov 06 15:02:02 I pulled the patches out of https://github.com/beagleboard/beaglebone-ai Nov 06 15:02:07 it was just something that worked at the time.. Nov 06 15:02:23 if it is out-of-date, we should probably nuke it and refer to what is not out-of-date. Nov 06 15:03:13 i agree, i'll reync with your u-boot patches.. (including the pinmux change) Nov 06 15:04:56 jkridner: vtt will derive from ddr Nov 06 15:05:44 oh actually it uses both Nov 06 15:06:29 and at some point, our dts labels should match the am5/dra7 that are in u-boot/mainline, espically when they are the same function.. some of the nodes were nameed by embest kinda randomly, but they worked.. Nov 06 15:09:43 jkridner: vdd_ddr is the supply for vtt, but vdd_3v3 is used to enable it... for some reason Nov 06 15:11:00 https://github.com/jadonk/u-boot/pull/new/rcn-ai-rebase-2019.10 Nov 06 15:11:04 building and testing now. Nov 06 15:11:33 zmatt: yeah, odd. I'll trust they got the sequencing right. Nov 06 15:12:17 crap... missed a file in the merge. Nov 06 15:13:55 well the thing is this vtt regulator *needs* some supply that's > 2.3V to make its built-in undervoltage lockout happy. this is a separate input from the actual supply input, for which the ddr supply voltage is used, which is 1.5V (ddr3) or 1.35V (ddr3l) Nov 06 15:14:00 history re-written. ;) Nov 06 15:17:21 I have trouble believing the am572x would need it (and the ram chip certainly won't), but there's no harm is delaying VTT until the 3.3V is up either (it's only needed when the ram is in use) Nov 06 15:25:42 rcn-ee[m]: k, finally building. patch set is a total mess right now. I'll clean it up once it works again. I still need to figure out a magic delta between your code and mainline w/ my patches. Nov 06 15:26:43 grrr.... still broken. :-( Nov 06 15:27:43 hmmm.... didn't get the cape eeprom printouts either. Nov 06 15:34:58 rcn-ee[m]: well, at least with your default environment it boots, so I must have been doing something wrong in the boot process. Nov 06 15:36:21 issue #1: MMC: no pinctrl state for default mode ... omap_hsmmc_init_setup: timedout waiting for cc2! Nov 06 15:36:21 mmc@480d1000 - probe failed: -110 Nov 06 15:37:18 issue #2: "BeagleBone: cape eeprom: i2c_probe:" never prints. Nov 06 15:59:01 I have just flashed the newest debian distribution 9.9 on BeagleBoneBlack Wireless and when I open Cloud9 IDE it gives me an error of: http://192.168.7.2/ Nov 06 15:59:02 Failed to write to 'state.settings'. ENOSPC: no space left on device, write . Nov 06 16:01:36 Linux421: yes, the current lxqt image on beagleboard.org is too big for eMMC, grab this version for now.. https://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Debian_Stretch_LXQt_Snapshot Nov 06 16:02:06 or preferably use the iot image (if you don't actually need a desktop environment running on your BBB) Nov 06 16:02:27 Linux421: ti-opencl/ti-tidl for the x15/bbai made the image too big.. Nov 06 16:02:30 those have now been removed in teh default lxqt images Nov 06 16:02:43 they shouldn't be on any image for the am335x Nov 06 16:03:18 so should I be grabbing this one: bone-eMMC-flasher-debian-9.11-lxqt-armhf-2019-11-02-4gb.img.xz? Nov 06 16:03:38 one of jason's goals to make unified image... "TPL" u-boot support. Nov 06 16:04:00 Linux421: yes, that one will fit... on the 4gb emmc, with a full lxqt desktop. Nov 06 16:04:35 presumably *barely* fit :P Nov 06 16:05:04 rcn-ee[m]: while I can sort of understand the wish I'm also kinda leaning towards "ew" Nov 06 19:06:34 rats... my rebase wasn't quite right: https://github.com/jadonk/u-boot/commits/ai-rebase-squash2 Nov 06 19:52:56 rcn-ee[m]: hopefully the 4th time is the charm: https://github.com/jadonk/u-boot/commits/ai-rebase-squash4 Nov 06 22:41:50 So another device tree question for the Beaglebone AI. I want to setup a SPI port that has P9.17-CS, P9.18-SDI, P9.21-SDO, P9.22-SCLK using the device tree. Nov 06 22:43:16 However, I believe that the default for SDI is that it idles high. So I want to make P9.18 a GPIO input with a pulldown that after I start up my ADC I can then set it to SDI. Nov 06 22:44:47 *Because I'm talking to an ADC that if P9.18-SDI is high during power up it won't work. So I have to power up the ADC, wait a second, then set P9.18 to SDI. Unless there is a way to make the SDI line just idle low when it isn't being used. Nov 06 22:46:03 I know how to setup the spi bus pins in the device tree, but switching P9.18 from GPIO with pulldown to SDI at runtime is less clear to me. I believe that it can be done using the pinmux-helper, but the details aren't clear to me. Nov 06 23:54:54 jkridner[m]: I don't think the util in that repo is the model conversion util. The only binary there is eve_test_dl_algo.out; the stuff floating around references a tidl_import.out tool Nov 06 23:57:47 regardless of the name, it wants some config file Nov 06 23:58:08 going to scrub the logs for hints but if I am on a wild goose chase.... Nov 07 00:02:59 looks like those binaries were deleted in Dec 2018 **** ENDING LOGGING AT Thu Nov 07 02:59:57 2019