[computer] adventures in *nix

[kaberett]

My Toshiba laptop (purchased by the DSA) finally had the keyboard finally give up the ghost entirely (this being the proximate cause of the trigger finger, which is doing very nicely with copious overnight splinting), having already had substantial problems with various other bits of its hardware in addition to overall slowing down a lot, and it was just... time for a new laptop.

So I went and bought a new-to-me refurbed Dell laptop (Latitude 7280) and it's generally doing rather well. Except that it does this fascinating thing, after a few but inconsistent hours of use, of just... losing ext4 and everything on it, and also producing the error message [drm:gen8_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun.

Which shows up a lot with not much concrete by way of obvious fixes, so herewith a list of things I've experimented with so far:

• checking the SSD health, all of which is fine
  
• swapping from "RAID On" to "AHCI" in BIOS (because some people suggested this might help; changed the precise nature of the symptoms in terms of things staying usable slightly longer and also getting an exciting blue-green screen gradient, but didn't actually fix the problem)
  
• looking for any sort of auto power-down on the SSD that Debian might be handling inelegantly (and failing to find one)
  
• confirming I'm really definitely on a later kernel version than the one that seems to be associated with this bug being introduced
  
• confirming that I have the various xorg bits of video-intel that might reasonably be expected, and that they are (as these things go) up to date
  
• actually updating my BIOS

... so now I get to wait a day or so and see if the BIOS update magically fixes things, and if not I probably need to start faffing with my x11 config in slightly greater earnest.

Comments

[vass]

...yikes.

*peers at your shark with great interest, from a safe distance*

[kaberett]

Downside: updating the BIOS did not, in fact, solve the problem.

Upside: the behaviour I'm seeing is now consistent with all the other reports of this bug, where previously it was not! Which gives me greater confidence that fucking around with graphics drivers will achieve anything at all...

[kaberett]

The Debian wiki links out to the Fedora wiki and fora (because of course it does), leading to:

The xorg.conf.new belives that my monitor is some sort of a hypercube with most of its surfaces placed in the dimensions mere mortals cannot see...

[vass]

*______* IT'S ALL FULL OF STARS

That is a Marvel movie of a shark.

[kaberett]

Setting the kernel parameter to 5500 as described in this askubuntu thread made the errors new and exciting! This time, the first thing to conk out was the CLOCK in the PANEL. And sudo commands failed with an I/O error rather than a command-not-found!

I am calling this progress, I've reduced the number to 200 as suggested, and I'm going to see what happens in the next 24 hours or so. (If I still get an excitingly different and vaguely-more-functional error message tomorrow, I'll reduce it to 0 as yet other sources have suggested!

Meanwhile, in an attempt to muddy the waters, I just did a dist-upgrade, so frankly WHO THE FUCK KNOWS.

[ewx]

Is there any kernel output associated with ext4 going wrong? Leaving dmesg --follow running may be the best bet for that, if kernel.log is beyond reach when things go wrong.

[kaberett]

kernel.log is indeed beyond reach, but thank you for the dmesg suggestion; it is duly sitting and thinking about what it's done, and I'll report back if my various other hacking around the underbrush hasn't fixed the issue...

[silveradept]

That's not any fun at all, but regarding comments above, yay for it starting to be consistent as a problem?

[kaberett]

WHO KNOWS IT'S ALL VERY EXCITING. (I don't particularly need to be doing any work so I'm not losing anything, which makes this an entertaining adventure rather than an utter misery.)

(PS dunno if you've spotted but Magisk Manager now ALSO has a "hide" option in settings -- if you do that then delete all your Magisk-related .zip files the Pokemon should Go again, something something something apparently Niantic are now querying for rooting-related software and interpreting the resulting error message as an indicator of presence or absence and blocking accordingly, if you haven't already fixed that...)

[silveradept]

I have been needing to use the hide option for a good while now, but I was able to anticipate the latest hilarity by fortuitously reading the right thread on xda, and so things are still going fine, and that makes Celebi research on my radar now.