pcengines / apu2-documentation

Documentation and scripts for building and adjusting PC Engines APU2 firmware
https://pcengines.github.io/apu2-documentation/
208 stars 45 forks source link

Constant reports of "Hardware Error" #329

Open srcshelton opened 6 months ago

srcshelton commented 6 months ago

PCEngines firmware version

v4.19.0.1

APU variant

apu2e

OS and OS Version

Linux 5.15.32, 6.0.7, 6.1.4

Affected component(s), peripheral(s) or functionality

The errors are reported as "corrected"… but there are a lot of them!

Brief summary

Very frequently (several times a day, at least) I'm seeing the following being logged via syslog():

[ T1030] [Hardware Error]: Corrected error, no action required.
[ T1030] [Hardware Error]: CPU:0 (16:30:1) MC0_STATUS[-|CE|-|AddrV|-|-|-]: 0x9400000000010015
[ T1030] [Hardware Error]: Error Addr: 0x00008ae14b79c000
[ T1030] [Hardware Error]: MC0 Error: L1 TLB multimatch.
[ T1030] [Hardware Error]: cache level: L1, tx: DATA

How reproducible

This has happened across a range of Linux kernels from early 5.x through to at least Linux 6.1 - and I'll test with more recent kernels shortly.

How to reproduce

Steps to reproduce the behavior:

  1. Boot Linux
  2. Watch the system logs
  3. Even if fully correctable, there's something amiss here...

Expected behavior

No L1 TLB multimatch errors.

Actual behavior

Consistent errors over time.

arjacy commented 6 months ago

ECC RAM detecting a memory error?

srcshelton commented 6 months ago

Potentially, but I assume the TLB multimatch is something like this:

https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1575932.html

… which looks to have been fixed years ago in the general case, but still happening here?

(Plus, I'm booting with nopti and still getting this)