On 27/01/2017 15:06, Markus Armbruster wrote: >> Is rom_add_file_fixed supposed to mark the memory region read-only? Or >> should it be created read-only before calling rom_add_file_fixed? > I'm not familiar with the memory API, but the fact that you're getting > RW suggests that you need to mark read-only yourself. Suggest to > examine similar code elsewhere for additional clues. > >>> The monkey-patching of dinfo->is_default is to silence the "Orphaned >>> drive without device" warning. Naughty. >>> >>> This use of -drive if=pflash is highly unorthodox. To suggest better >>> ways, we need to understand the role of "the virtual ramdisk". Can you >>> explain? How does it work on a physical system? >> AFAIK except for simulators it's only used on FPGA boards. It's what >> the comment says: a kind of initrd, but it resides outside of the >> partition RAM. So, it is a RAM region pre-filled with data in the way >> not visible to the guest. >> >> I would have used an -initrd option instead of -pflash, but it >> required a -kernel option, and there is no sensible way to use the >> -kernel with Solaris. >> If QEMU had something like a nvram-disk - it would have worked here the best. > > Have you considered pressing -bios into service? -option-rom? -object > memory-backend-file? Paolo, any recommendations?
Yeah, I think -bios is closest. By the way, where do I get all the .rom files for the Niagara board? "make check" complains about them. Paolo