Stuart Henderson <s...@spacehopper.org> wrote: > On 2018/01/09 15:13, Mike Larkin wrote: > > On Tue, Jan 09, 2018 at 10:39:30PM +0000, Stuart Henderson wrote: > > > There was a new release of SeaBIOS recently which includes its own > > > serial console redirection code. This updates SeaBIOS, enables > > > CONFIG_SERCON and removes SGABIOS and the option rom-related > > > code/config. > > > > > > Working here in a quick test with OpenBSD guest (also gets rid of the > > > doubled characters in the boot loader seen before with SGABIOS) but I'm > > > not making heavy use of vmm myself. > > > > > > After building, to test: > > > > > > fw_update -p /path/to/packages/amd64/ -D unsigned > > > > > > To revert to the old one in case of problems, pkg_delete vmm-firmware > > > and re-run fw_update with no command line. > > > > cool, seems ok to me. if we do a quick sanity test with a linux guest > > bootloader (eg can you see grub), then I think this can become the default. > > > > thanks sthen > > Ah damn, sorry, my testing methodology was bogus. It needs more than just > the config option, also it needs a fw_cfg variable to actually enable it, > which I had forgotten to workaround. > > That's easy enough and works for HD boots (I've now got Alpine setup to > test) but something is wrong with CD boots (both with CONFIG_SERCON and > with SGABIOS). > > No urgency (afaik) with the update, and I definitely don't want to cause > any work for you Mike :) but paging ccardenas if you are interested in > taking a look? > > It hangs after "pci dev 00:04.0 using legacy (0.9.5) virtio mode" if > cdrom is used. > > (I mostly looked at updating SeaBIOS because I thought it would be nice > to remove the Apache-license SGABIOS code). > > Updated diff with the config part I missed before at > https://junkpile.org/seabios-1.11.0-diff,2 >
The latest SeaBIOS requires additional opcodes to be implemented in our vioscsi driver (at least report_luns). I won't know which other ones need implementing until I handle report_luns. The default behavior in the vmd vioscsi driver is to not respond to unhandled opcodes which gives the behavior of the VM "hanging" as it's waiting on a response. I'll see what I can do in the next few days. +--+ Carlos