FW 5.16.21 creates crash logs every now and then

[mergen3107]

I recently got my JB'ed Kindle Scrive updated to 5.16.21 (because I just forgot to block OTA).

During usage, it creates crash logs files in documents folder and well as /mnt/us/ ("root" when connected via USB).

In documents they are like:
cmd_10725_037-juno_16021_barolo_bellatrix3-428978_crash_Sep_25_21.52.22_2024.sdr
cmd_10725_037-juno_16021_barolo_bellatrix3-428978_crash_Sep_25_21.52.22_2024.txt
cmd_15905_037-juno_16021_barolo_bellatrix3-428978_crash_Sep_26_23.12.36_2024.tgz

or similar, but for powerd.

In /mnt/us/ they have similar names, but archives are about 30 MB:
cmd_15905_037-juno_16021_barolo_bellatrix3-428978.core
Indexer_Dump_1727406757193.txt
powerd_3586_037-juno_16021_barolo_bellatrix3-428978.core

I don't know might be causing it.

Does anybody experience the same?
@Marek

Maybe Kindle is complaining because I have logging, OTA, and minerva blocked
But this wasn't the case on 5.16.9 or 5.16.3.

[scribe_user]

I also had similar issues on PW5, Scribe on 5.16.10 and above, with logs disabled. Reverted both of them to 5.16.8

[mergen3107]

Thanks!

I think there is some kind of a watchdog somewhere.

This also happens when I install stuff with MRPI or restart Kindle while it is plugged to PC. Although, these were present even on earliest Kindle Scribe FWs, like 5.16.2

[mergen3107]

I disabled /app/bin/KPPMainAppWatchdog , that appeared in 5.16.7. Nothing bad happened, but logs are also still here, so I don’t know

[mergen3107]

Long story short:

1) To stop some logs, it is enough to add a new empty file called DISABLE_CORE_DUMP in the /mnt/us/ (visible root via USB). Heavy core dumps won't show up anymore. Only ones that show up are Indexer Dumps (a few tens of KBs) every now and then. Context: I also have blocked all logging, and indexer dumps says something like "this is a diff between files on device and cc.db" - I am pretty sure this is because there is a mismatch between cloud (which never sees any logs from my Kindle Scribe) and local files

2) To hard stop EVERYTHING, block /usr/bin/dump-stack . This will prevent ALL logs. Side effect: usbnet session is very laggy in between actions. Most probably because Kindle still tries to run dump-stack, but encounters wrong exit statuses.

In the end, I rolled back to 5.16.2.0.1 and plan to stay here for a long time.
hf versions seem to be much more chatty in terms of dumps and stuff.


P.S. Also, I just figured I have been missing out on this little gem of a GitHub page
and it works only on non-hf kindle due to browser support.
https://github.com/KindleModding/KWebBrew

[Frogm4n]

Thanks for figuring this out! I've had the big coredump problem on my Oasis 2 when running KOReader (no framework) to have enough memory to read page scan-only PDFs.

[vvvad]

Originally Posted by mergen3107

Long story short:

2) To hard stop EVERYTHING, block /usr/bin/dump-stack . This will prevent ALL logs. Side effect: usbnet session is not very laggy in between actions. Most probably because Kindle still tries to run dump-stack, but encounters wrong exit statuses.

Do you mean now very laggy rather than not very laggy? If so you could try replacing dump-stack with just a exit 0 or replace with /bin/true depending on whether its a script or binary

[mergen3107]

Thanks for inputs!

Yeah, my bad. Probably was going between “not very responsive” and “very laggy” and mixed them both.

That’s cool! I’ll try that once I have my Scribe replaced…