Re: Anyone else seen this.

Theo <[email protected]> writes:
> In uk.d-i-y The Natural Philosopher <[email protected]> wrote:
>> Recently my router (Draytek 2726) has been spontaneously rebooting.
>> 
>> Examination of the logs show a kernel panic caused by 'attempt to 
>> execute an illegal instruction'.
>> 
>> My assumption, possibly erroneous, is that the actual firmware code is 
>> corrupted
>> The router is rated only to +45°C but that is the recommended limit for 
>> flash memory, and the room temperature has been up towards 40°C. which 
>> doesn't leave a great deal of margin for router internal heat production.
>> My query is whether anyone else has seen issues with flash memory in the 
>> recent hear wave.
>> 
>> Surely phones get hotter than this?
>> 
>> Genuinely interested in other's experiences.

No computer misbehavior here.

> You can get bit flips from cosmic rays.  That's a reason servers have ECC
> memory:
>
> https://en.wikipedia.org/wiki/Soft_error

If the behavior is consistent across reboot that suggests a problem
somewhere more persistent than RAM. Flash corruption would be my first
guess.

> Not sure to what extent flash is affected, but many routers nowadays copy
> their OS into RAM before execution: 'execute in place' is still possible,
> but sometimes they store a compressed image in flash to reduce flash cost
> and decompress into RAM on boot.

Page caching can make RAM corruption very persistent (albeit not across
a reboot). A while back chiark’s /bin/cat developed a single-bit error
in the cached copy of the code section, which was used (without
reloading) every time it is executed.

> It could also be a RAM chip on the way out.

-- 
https://www.greenend.org.uk/rjk/