Re-factor the various m68k platforms' identifycpu() functions, and bring
some consistency to how the CPU/FPU/MMU/cache information is displayed.
Each platform now provides a machine_set_model() function that is called
from cpu_startup_common(). This function is respondible for calling
cpu_setmodel() with platform-specific model description. Then, common code
is responsible for printing the model information, along with the CPU
and cache info, but separately (and this is no longer included in the
hw.model sysctl node; at least it's now consistent!)
Some examples:
Qemu 10.1.2 Virt platform
MC68040+MMU+FPU, 4k+4k on-chip I/D caches
(hw.model == "Qemu 10.1.2 Virt platform")
Motorola MVME-147
MC68030 CPU+MMU @ 25MHz, MC68882 FPU
[43 lines not shown]
pmapmi: Define and use pmap_md_kernel_vaddr_{p,to_paddr} in pmap_extract.
Allow an architecture to provide a function that extracts the PA from the
VA that the kernel text/data/bss are mapped at. This allows pmap_extract
to work for kernel mappings that aren't PAGE_SIZE sized.
micphy: Fill out mii_mpd_model in the sc struct.
The rest of the code expects this to be filled in and the driver
seems stable with this change (at least on my machine).
"ship it!" jmcneill@
micphy: Apply errata workaround for KSZ9031
This fixes ethernet on an olimex a64 board, and should be pulled up
to -11, -10, -9...
Thanks jmcneill for pointing out the problem.
PR port-arm/60193: olimex a64 - no carrier with emac0 ethernet
