Re: [Etherboot-developers] 5.1.3 make changes?

[<ebi...@ln...>]

ke...@us... (Ken Yap) writes:

> >There seems to be a change in the 5.1.3 Makefile.  Previously in 5.0.x 
> >one could do:
> >
> >$ make bin32/mx98715.lzdsk
> >cat bin/przloader.bin bin32/tulip.huf > bin32/mx98715.lzrom
> >./makerom.pl  -p 0x10d9,0x0531 -i'mx98715.lzrom 5.0.8 Etherboot (GPL)' 
> >bin32/mx98715.lzrom
> >PCI header at 0x1c and PnP header at 0x34
> >cat bin/boot1a.bin bin32/mx98715.lzrom > bin32/mx98715.lzdsk.x
> >dd if=bin32/mx98715.lzdsk.x of=bin32/mx98715.lzdsk bs=36k conv=sync 2> 
> >/dev/null
> >rm -f bin32/mx98715.lzdsk.x
> >
> >But now in 5.1.3 you get:
> >
> >$ make bin32/mx98715.lzdsk
> >make: *** No rule to make target `bin32/mx98715.lzdsk'.  Stop.
> 
> Ok, this was puzzling, but the reason is not that the .lzdsk rule has
> gone away, but that it had not been updated to match the .dsk rule.
> It seems that the rule to fix should look like:
> 
> %.lzdsk:        %.lzrom $(DISKLOADER)
>         cat $(DISKLOADER) $< > $@.x
>         dd if=$@.x of=$@ bs=36k conv=sync 2> /dev/null 
>         $(RM) $@.x
> 
> if it's to match .dsk. I assume that $(START16) is not needed any more.

Yes.  It is correct to remove $(START16) I accidentally added
it in that case.  The .rom image has $(START16) included earlier, and
including it twice is just silly.  It appeared non-trivial and low
priority to remove the assumptions of the disk loader that it was
loading a ROM image, which is why I made that mistake.  That does mean
disk images need to be below 32K or so though. 
 
> The whole area needs a good cleanup, but I don't know which prefix files
> can jump directly into the .img. Eric, could we have a crash course on
> how to jump into the new .img/.lzimgs so that I can have a go at fixing
> the prefixes for 5.1.4? I assume $(DISKLOADER) works?

Crash course directly using the .img file.
A) Just jump to the first byte of the .img. (Only a stack and return address are expected)
B) $(START16) is a intermediate prefix that does the 16bit->32bit transition,
   and is needed, when you have a 16bit prefix.
C) We might want to consolidate compressors and just use nrv2b (works for any size image
   and decompresses in 32bit mode). The old lzhuf compressor is still present and all of
   lzdisk etc still refer to it.
D) Except for lilo I think everything works.  I just missed something there...
E) The goal of having stackable prefixes like $(DISKLOADER) $(START16) ....
   is primarily so that the same prefix can be use weather or not compression is present
   or not, simplify maintenance by decoupling those variables.
G) Yep I left a few rough edges in 5.1.x my apologies.
H) I was very careful in one or two cases (.rom and .dsk to be certain no relocation would be
   in effect, so drivers that were not yet updated would still work)

Eric