Call acpi_pxm_set_proximity_info() slightly earlier.

[markj]

Summary

This function is responsible for setting pc_domain in each pcpu
structure. Call it from the main function that starts APs, rather than
a separate SYSINIT. This makes it easier to close the window where
UMA's per-CPU slab allocator may be called while pc_domain is
uninitialized. In particular, the allocator uses pc_domain to allocate
domain-local pages, so allocations before this point end up using domain
0 for everything.

[kib]

sys/dev/acpica/acpi_pxm.c

655 ↗

(On Diff #71534)

What is the point of this unmap ? We still do not return stealed memory back.

[markj]

sys/dev/acpica/acpi_pxm.c

655 ↗

(On Diff #71534)

I guess it is just defensive, to help ensure that the cpus table is not used once the kernel has bootstrapped.

[markj]

sys/x86/acpica/srat.c

64 ↗

(On Diff #71534)

I meant to delete this whole function.

[markj]

Remote srat_set_cpus(). acpi_pxm_set_cpu_locality() is called during
cpu_mp_start() now.

[kib]

sys/dev/acpica/acpi_pxm.c

655 ↗

(On Diff #71534)

I think it would be nice to return the memory.

[markj]

sys/dev/acpica/acpi_pxm.c

655 ↗

(On Diff #71534)

To do that we would need to create a new vm_phys_seg[] entry for the memory, which is going to be a single 4KB page on most systems. I guess that is fine since x86 is VM_PHYSSEG_DENSE, so vm_phys_paddr_to_vm_page() does not need to iterate over the entries. I will add an XXX comment for now.