On Wed, 19 Aug 2015 19:31:30 +0200 Gerd Hoffmann <kra...@redhat.com> wrote:
> On Mi, 2015-08-19 at 09:57 +0200, Marc Marí wrote: > > On Wed, 19 Aug 2015 00:27:54 +0200 > > Gerd Hoffmann <kra...@redhat.com> wrote: > > > > > On Di, 2015-07-21 at 18:06 +0200, Marc Marí wrote: > > > > It's probably not good to override all the ROM options. Better > > > > ways can be discussed. > > > > > > source code for the option rom doing linux kernel (and multiboot > > > kernel) boot is in pc-bios/optionrom/ (qemu tree). Patching that > > > one to use fw_cfg dma if available should get simliar speedups, > > > but without compatiblity issues. > > > > > > > The problem with patching the current ROMs, is that they are > > written in assembly. So they would need a complete rewrite in C. > > Writing a DMA guest in assembly I think it is a nightmare (but it > > can be done, of course). > > Yes, doesn't look that easy. Especially when probing for the dma > fwcfg at rumtime. We could have just two versions of the roms and > let qemu pick the correct one depending in whenever fwcfg dma is > enabled or not. > > Doing this in seabios looks appealing too as seabios needs the code > for fw_cfg support anyway and it is already written in C. It needs > some more care through so things like bootorder and pxeboot continue > to work. Maybe seabios could match the rom name (the one which > appears in the boot menu) to figure whenever it should kick a direct > kernel boot. Have you looked at the other email thread?: http://www.seabios.org/pipermail/seabios/2015-August/009603.html There's some work missing, but a basic sketch of this ROM problem is here: http://www.seabios.org/pipermail/seabios/2015-August/009606.html Thanks Marc _______________________________________________ SeaBIOS mailing list SeaBIOS@seabios.org http://www.seabios.org/mailman/listinfo/seabios