Daniel J. Posted February 8, 2017 Posted February 8, 2017 Like whaaaaaa... Okay, so long and frustrating story short, I have been having inexplicable stability issues on my Banana Pro. I have blamed the SD card, the power supply, the micro USB cable, but none of those leads seemed to actually lead anywhere. I connected up a Fluke multimeter to the SATA power output, set to peak min mode, to see if I get any undervolting... Nope, no problem there. I ran stress tests on my sd card on my computer - it was PERFECT. I was completely stumped... Until, after getting an Anker charging cable, and having a really good run with the new Armbian release - the system had been working for an hour without glitches, apt-get worked, I have upgraded and installed mc, ran "stress" for 20 minutes with no issues... I decided to try if Wifi will work... (I suspected it might not, as I was (still am) getting errors like these in dmesg: brcmf_cfg80211_reg_notifier: not a ISO3166 code (0x30 0x30) brcmf_p2p_create_p2pdev: timeout occurred brcmf_cfg80211_add_iface: add iface p2p-dev-wlan0 type 10 failed: err=-5So following the "getting started" guide, I kicked off nmtui-connect, typed in my password for my 2.4 wifi ap, and then all hell broke loose. Everything fell apart right there and then, with error messages like these: mmc0: card stuck in programming state! mmc_erase: group start error -5, status 0x0 ext4fs: This should never happen etc.Well... that's when I looked at my board, neatly inside a Chinese enclosure, with a Chinese RP-SMA pigtail connected up to the WiFi, the cable actually touching the board... So I'm like "wait a minute right there, this can't be no coincidence"... So I removed the pigtail, re-flashed the memory card, rebooted, and for a while things were working. I upgraded the system, installed mc, tried connecting to the WiFi (didn't work without antenna), tried scanning for APs (detected the closer ones successfully)... I was pretty happy, posted here about how WiFi interference was probably the real culprit, left the battery to charge, with the system running "stress" constantly, checking in now and then on AC voltages in /sys/power (between 4.8V and 5.1V), and CPU temp (below 61C)... and then, after two hours, I decided to install Privoxy.. and right enough, during apt-get, the MC card errors again. Stuck in programming state, i/o error, error -5 sending status command, the whole deal. Again. Really I don't get it. EDIT 2: I booted up again, this time pulled off the battery to make sure that charging doesn't accidentally undervolt anything... Voltage safely above 5V according to /sys/power... Logged in again, sudo apt-get install privoxy again, mmc error again, remounted read-only, that's it. It's an EVO32 - the SD card people keep recommending as THE card to use with SBCs... According to testing it's 100% good...
Daniel J. Posted February 8, 2017 Author Posted February 8, 2017 Some experimentation on, privoxy installed, so while for a while I was suspecting some kind of "bad sectors" on the SD card after it failed twice on the exact same operation, it seems that probably isn't the case at all. I put the pigtail back, screwed on an antenna, booted in running from battery, and successfully connected to wifi... doing an apt-get update through wlan without issues. I don't know about power. I have no idea what counts as "undervolted" for this board. AFAIK the Allwinner 20 operates at 3.3V, and SD cards require 2.7V at a minimum... I seriously doubt even 4.5V would be a real undervolt condition, I mean the whole thing also runs off of a 3.7V Li-Ion battery... Spikes are a possibility of course, but I had the Fluke on it set to Peak Detect - that's analog peak detection, anything that causes things to malfunction in the board should be picked up... and the USB power input side never went below 5V...
Recommended Posts