> Date: Thu, 30 May 2019 18:09:01 +0200 > From: "Peter J. Philipp" <p...@centroid.eu> > > On Sat, May 25, 2019 at 07:58:36PM +0200, Karel Gardas wrote: > > Could you be also so kind and share information about how do you build > > OpenBSD for PPC64 and how do you boot whatever you get from the build. I'm > > especially curious if you build bsd.rd and somehow boot that and how exactly > > do you invoke Qemu and also what Qemu version exactly do you use for that. > > > > Thanks, > > Karel > > Hi Karel, > > Last weeks answer wasn't to my own satisfaction, so I'll try again. Last > year I spent considerable time and effort to make the bsd kernel compile for > aim64 arch. aim64 arch is like macppc arch, and powerpc64 is like powerpc, > only that some variables are different and the asm has been changed. ie. > ppc_mtmsr() asm macro has been changed to ppc_mtmsrd() (doubleword to > indicate that 64-bit flag can be set on the MSR register, it's the highest > bit also called PSL_SF found in /sys/arch/powerpc64/include/psl.h, this I > also had changed). Everything was changed around ie. a long in powerpc would > be a uint32_t in powerpc64 unless the variable size can go higher to 64 bit, > in > that case it would be uint64_t). > > So I get to build a 64-bit kernel whoowee, it's like a little lightbulb that > goes poof when I apply current to it. Basically the kernel starts in > locore0.S, > and it branches off into the ofwreal.S asm file, here it stopped last year at > entry into a C routine. I had given up there. Closer examination with qemu > though this year showed that it can go further into C routines, I'm not sure > if it was the fact that I compiled with gcc (I tried it out, basically and it > worked). I was experimenting a bit with the temporary 64-bit bridge which is > a bit one can set to have 32-bit functionality on opcodes like mfsrin. This > for a moment got me so far as pmap.c but crashed/panic'ed on data that was > supposed to be gotten from openfirmware. The fact that the stdout console > was not initialized showing me the panic string was another indicator that > these were awry. > > Last year I also hacked on the bootloader ofwboot to boot 64-bit ELF images. > This works for the most part and it starts in 32-bit mode and hands off the > execution to the kernel in 32-bit mode, which then does the following opcodes > to set 64-bit mode: > > mfmsr %r0 > lis %r21, 1 > insrdi %r0, %r21, 1, 0 /* set PSL_SF */ > mtmsrd %r0 > isync > nop > > This I learned from reading the FreeBSD asm and it's not changed much except > perhaps the register #'s. Ultimately this is called first, and I don't think > I'll see an alignment issue either as 64-bit opcodes are 4 bytes just like > 32-bit opcodes. > > OK if I didn't already say it I'm focusing on bsd kernel not bsd.rd, when the > day comes when bsd kernel boots but panics on no root filesystem is when I > am going to work on bsd.rd. That day is still a long ways off I think as I > gotta have locore0.S, open firmware routines, and locore.S right. locore.S > still nees a lot of work and I'm not even there yet because I'm approaching > the bootstrapping as it executes, and once I'M past the openfirmware I know > it will panic or do funky things once the cpu and memory bootstrapping is > done. > > Does this better explain where I am at? There is still a lot to do, and > I'm not really that proud of what I have accomplished thus far, but I started > not knowing any asm, not knowing much in general about kernel work, sure I > poked in ther kernel before but mostly in the networking areas, this is > totally > new territory. And it's hard. I still can't turn away/walk away, somehow I > keep thinking I can make the lightbulb glow. I'm approaching this different > though than last year. Last year I wanted to get something done by christmas. > This year there is no due date. I'm curious to see how far I can take it.
Ome thing to keep in mind with hte G5 is that it is a bit strange and event when running in 32-bit mode it takes traps in 64-bit mode. And for some reason the bits in the MSR register aren't where you expect them to be. This is what the nop_32_Ns/Ne labels are about in the powerpc locore..S Those bits are there to fix this up on G5 and are overwritten with NOP instructions on older 32-bit CPUs.
