Curious NAND corruption (?) incident


Binky

Death's Steed
Staff member
Joined
May 28, 2003
Messages
6,997
Location
16A (TO)
A few days ago, I picked up my (1GHz, running recent standard firmware, clocked at 1.05GHz) pandora up off the table where it had been charging. The table gets very little direct sunlight, but the day was reasonably warm and humid.

I noticed that the whole unit was really quite warm - warmer than you'd expect when idle - perhaps as warm as if I had CPU, screen backlight and wlan all at high/full power

When I last left it, wlan was enabled but idle with power-saving on. screen was set to dim (and closed!) and I had a few pnds open (but not doing much)

I opened the lid - black screen, no response to keyboard/nubs

First thought: power management failure: unplug charger (official pandora one btw), hard reset, remove battery.

Inspect battery: Battery looks fine (no bulges, etc)

Reinsert battery: boots, gets stuck towards end of boot, complaining of missing files, and giving UBIFS errors

Login at cli terminal: (using alt+arrows) screen fills with pairs of hex digits (I don't get to see anything before these, because it scrolls off the screen)

Does eventually give me a command prompt, where I copy what little is in /home onto an SD card for safekeeping.

Soft reboot, login - same errors

Reflash from SD card - seems to work fine

While this isn't a huge problem, given that it seems to be a rare occurence, and thanks to the isolation of system and user files, no data was lost, I'm curious.

What, in the name of His Noodliness, actually happened?

Why was my pandora so hot, and what happened to my internal memory? I can see that overheating might corrupt memory, but there doesn't seem to have been any long-term hardware failure. It runs quite happily now with temperatures reported as 25-40°C

Any ideas?

Has this happened before to anyone?
 
Last edited by a moderator:
Maybe your NAND was full or near-full? I've had a similar corruption while having lots of -dbg packages installed that filled the NAND, and doing lots of reboots to test kernels. It seems ubifs doesn't really like being full or near-full.
 
Last edited by a moderator:
Back
Top