RE: Kernel pagefault in BETA1 and BETA2.

From: Mark Millard <marklmi_at_yahoo.com>
Date: Mon, 12 May 2025 16:53:28 UTC
Philip Homburg <pch-fbsd-3-stable_at_u-1.phicoh.com> wrote on
Date: Mon, 12 May 2025 16:25:53 UTC :

> I have an AMD Ryzen 5 system that runs 14.2 just fine but upgrading to
> 14.3-BETA1 and 14.3-BETA2 results in a kernel pagefault after boot.
> The system continues to boot just fine.
> 
> Fatal trap 12: page fault while in kernel mode
> cpuid = 2; apic id = 02
> fault virtual address = 0xdcf7be70
> fault code = supervisor read data, page not present
> instruction pointer = 0x20:0xdcd1f68f
> stack pointer = 0x28:0xfffffe01139e6c50
> frame pointer = 0x28:0x1
> code segment = base 0x0, limit 0xfffff, type 0x1b
> = DPL 0, pres 1, long 1, def32 0, gran 1
> processor eflags = interrupt enabled, resume, IOPL = 0
> current process = 0 (thread taskq)
> rdi: fffffe01139e6d18 rsi: fffff80001cb4c60 rdx: 0000000000000000
> rcx: 00000000dcd20108 r8: 00000000dcd20320 r9: 0000000000000004
> rax: 00000000dcf7bd30 rbx: fffffe01139e6de8 rbp: 0000000000000001
> r10: 0000000000000000 r11: 000000000000000b r12: fffff80001cc1258
> r13: 0000000000000000 r14: fffffe01139e6e20 r15: fffffe01139e6dc0
> trap number = 12
> EFI RT page fault

That "EFI RT page fault" is the important context when
it shows up.

https://lists.freebsd.org/archives/freebsd-current/2025-January/006862.html

reports about EFI runtime page fault notices:

QUOTE (of Konstantin Belousov)
The reporting of the faults during the calls into EFI RT was added recently.
Before that, such faults were silently ignored. Now, the report is printed
and then the fault is ignored.

There is no actionable items for users; a developer might be interested.
END QUOTE

Hopefully, release notes or some such will indicate the new
behavior and its status, if for no other reason than to
cut down on the volume of support questions.

Getting a notable volume of such notices at some stage after
starting a boot sequence would likely justify sending in a
report about it with notes about the kind of context that
got the issue.

===
Mark Millard
marklmi at yahoo.com