PANIC: vm_page_fault

[bbjimmy]

hrev58503

using any browser, after about 1.5 minutes.

[waddlesplash]

That hrev is from 2023. Are you sure that's the right number?

[bbjimmy]

Sorry, that is what was shown in the syslog buffer. I'm running hrev58503.

Why does the kdl text not show the current hrev, and why does the syslog buffer show the wrong hrev?

[waddlesplash]

The backtrace is already shown in the first image, that's why I removed the second.

Are you really sure that the syslog buffer is wrong?

[waddlesplash]

There's a ticket in #19299 about including the hrev in the panic message.

A reliable way to tell what kernel you are really running is (before the KDL occurs) to check the package it's from: catattr SYS:PACKAGE /system/kernel....

[bbjimmy]

syslog cache ( syslog kdl command)

[waddlesplash]

That's the revision of the bootloader. I guess you have an EFI loader installed from back then.

[waddlesplash]

Is there anything else interesting at the very end of the syslog cache? I/O errors perhaps?

[bbjimmy]

end of syslog cache

[waddlesplash]

Indeed, a read error, but a very odd one.

This looks like #18717 then: crashes in block_cache after I/O errors, so it's probably not a regression, but something's "shaken it loose" in your case. Perhaps the recent performance improvements uncovered a race? Or a hardware problem even?

[waddlesplash]

Reading over the code, I found a problem which may affect this ticket but not #18717. Fixed in hrev58506, please see if anything's different after that.

Does booting into an earlier hrev also have these odd SATA errors or ones like them, and just no KDLs? Or at least that odd SATA "Probe" right before the errors start?

[bbjimmy]

hrev58506 seems to have resolved the kdl.

[waddlesplash]

Excellent, thanks for testing!

I would keep an eye on those SATA errors, though. If this is an SSD you might try running "fstrim".