[Etherboot-developers] Etherboot thoughts....

[Eric W B. <ebi...@ln...>]

The freebsd pxeemu code is broken in 5.0.6.

In pxeemu_entry there is the snippet of code:

		/* Switch Stacks */
		__asm__("xorl %%eax, %%eax\n"
			"movw initsp, %%ax\n"
			"addl %0, %%eax\n"
			"movl %%eax, %%esp\n"
			: : "i" (RELOC));

Which references a local variable in start32.S which
because it was local was actually renamed: initial_regs + 12....

So for those people wanting PXE the freebsd code is a place to
start but it could really use some touch ups to even comppile ;(

Beyond that I've got my work to date in good pretty good shape to
start checking it in (on the development branch, of course).
Tomorrow is a holiday here so I will only get to it if I start
going stir crazy.

The topics look like

virt_to_bus, virt_to_phys, and ioremap for portability and
relocatibility.

Code to build a compressesed LinuxBIOS romimage.

Unplugged cable detection for the eepro100.

Renamed xstart xstart16, and added xstart32.  So we have a common
function between osloader.c and the external programs.  This allows
us to tear down virtual addresses when we are relocated.

Improved debugging code for detecting LinuxBIOS bugs.

The LinuxBIOS stdint compile fix.

Replacement of _real_call instead _prot_to_real && _real_to_prot,
in the assembly code, to allow for relocation.

Reversion of hacks in the pci code.  Deleting membase and not fixing
all of the drivers that use it is unacceptible.  I have fixed the
fa311.c driver to call pcbios_read_config_dword(... PCI_RESOURCE_1 ..)
for itself.

Relocate.c to actually relocate LinuxBIOS.

Fixes to the rtc timer code to:
A) Print the cpu Mhz so buggy implementations of timer2 may be
   detected.
B) Use a smaller value that actually works with the AMD766
   southbridge.
C) Derive TICKS_PER_MS from a more precise definition of
   CLOCK_TICK_RATE

I've got to go or I would check it in now...

Eric