Machine check exception

[pulkomandy]

Happened while compiling webkit_gtk on 32bit

[waddlesplash]

Is this the same machine as your recent tickets #19025 and #18983? Perhaps something's beginning to go wrong with it?

[pulkomandy]

These other two tickets are on my main 64 bit machine, which I bought new in 2022, I sure hope nothing is wrong with it at this point.

This one is on my "32 bit" machine (I mean I have a 32 bit Haiku install on it), which I also use a lot under Windows and is perfectly stable there.

Random crashes are not super uncommon for me (maybe once a month or so), but sometimes I don't take time to make a bugreport unless it's reproductible. I've been more careful about logging everything these last few weeks because we're due for a release soon.

This was also the case on my previous laptop, which is now running as a Linux server and also having no stability problems in that use.

So, stability isn't as great as it should, if for example we want to get TuneTracker systems to use Haiku again. For me these crashes are OK, I just reboot the machine and move on with my day. For them it means the radio station stops broadsating.

[pulkomandy]

I had another machine check exceptio non my 64bit machine tonight while compiling webkit. Unfortunately I couldn't take a picture of it. Apparently there was another KDL triggered on top of it, that covered it with a longer backtrace.

After that my keyboard was not usable anylore in KDL.

[pakyr]

I am not sure if it is the same issue (please advise if I should open a new ticket) but my Dell XPS 9560 which used to run Haiku flawlessly now KDLs with a machine check exception on startup. I will attach an image.

[pakyr]

KDL on latest nightly

[waddlesplash]

This one looks more like #19078, but the two may be related.

Does beta5 TC0 have the same problem?

[waddlesplash]

mmlr reports he also has this problem on recent nightlies:

turns out the machine check comes from 8cb8c3d75679dfb6852dfd150f1bd5bc883cce8e

So, it's the change to ACPI to use writeback instead of uncached memory that's triggering this, somehow.

[pakyr]

No, it doesn't happen on beta5_tc0 (at least when booted from live USB). Also, the most recent state on my machine before the current nightly was hrev58067, which works fine when booted into.

[pulkomandy]

I had more of these today, this time on my 64bit laptop. The machine was idle with just an ssh session open to another machine.

[pulkomandy]

It happens when I close the laptopelid, probably related to recent acpi changes

[waddlesplash]

This should be fixed in hrev58140.

[pakyr]

Can confirm I am no longer experiencing this as of hrev58146

[waddlesplash]

Excellent!