Nondeterministic boot errors

From one day to the next, my TP-Link TL-WR841N stopped working. I connected to its serial port and found that is in boot loop. Most of the times it sais

## Booting image at 9f020000 ...
   Uncompressing Kernel Image ... ERROR: LzmaDecode.c, 543

Decoding error = 1
LZMA ERROR 1 - must RESET board to recover�

But sometimes it also

## Booting image at 9f020000 ...
   Uncompressing Kernel Image ... OK

Starting kernel ...

If it gets past this point, another error occurs, for example:

[    2.976951] Starting init: /bin/sh exists but couldn't execute it (error -5)

But a different error every time. One time it got very far, but then just stopped and did nothing until I power cycled it.
So my router turned into a random error machine. I was thinking, could be something wrong with the flash-chip? Has anyone here seen something like this before?

If reflashing OpenWrt doesn't help, you are likely to experience hardware issues, probably flash related but eventually also RAM faults (which can manifest as decompression and/ or read-write errors as well). A faulty power supply could also be a cause, due to fluctuating voltages, insufficient filtering of the aging capacitors or over-current.

Given the hardware in question, further debugging (beyond just flashing OpenWrt again, once) or trying to fix it would not be economically sensible, as you can get better -fully functional- devices for less money on the second hand markets.

3 Likes

I replaced capacitors and power supply and reflashed OpenWrt. Error persists. Unfortunately, I have no working oscilloscope, so I can't check if the voltage ripples. I will try replacing the RAM now.

Unlikely to be power ripples, especially with an external wall-wart power supply. Suggest just switch out the wall wart.

If repairing for the fun of it and you have the capability for fine SMT work, then the far likelier candidate is flash. You're sort of working from least-likely to most-likely right now. Assess odds of:

  • 95% chance flash
  • 4% chance RAM
  • 1% chance power/other (ie: cracked solder joint, etc)