We already know that you not one of the faint of heart. Nate
On Sat, Jan 10, 2009 at 11:49 PM, Gabe Black <[email protected]> wrote: > Intel provides an implementation of at least part of EFI that we can use > which has various licenses associated with it. The open source parts are > here https://www.tianocore.org/. I have a copy of the Beyond BIOS book > listed in the upper right from work which I've started working through. > I don't know enough about EFI at the moment or about what that provides > to know how much we'd need to add. EFI actually comes closer to an > operating system than the thin flimsy firmware interface the traditional > BIOS provides, so it would likely be a lot of work to cook one up > totally from scratch. Before anyone jumps to the conclusion that we > should therefore use a traditional BIOS, keep in mind that those are > almost always written entirely in assembly and in 16 bit real mode with > severe space limitations. They work in such a way that there are > basically no higher language compilers that target them, so the tools > you have to work with are an assembler and maybe gdb. The interfaces are > also VERY crufty and also complex and poorly documented. Something like > linuxbios helps to get partway there, but just figuring out linuxbios > itself is purported to not be for the faint of heart. > > Gabe > > nathan binkert wrote: >> My only question would be, is there an existing bios from qemu or >> linuxbios or something like that which we can adapt and will be >> easier/faster? I think it's acceptable to use GPL code for that sort >> of thing since it will be totally separate from M5. >> >> Nate >> >> On Sat, Jan 10, 2009 at 6:22 PM, Steve Reinhardt <[email protected]> wrote: >> >>> I think the only reason it's in a separate repository is because it's >>> derived from some Compaq/HP code and probably has a different license. >>> If we were to implement an EFI BIOS from scratch (presumably just a >>> fake one that follows the API) then I think it could live in the main >>> repo. >>> >>> Steve >>> >>> On Sat, Jan 10, 2009 at 4:44 PM, Ali Saidi <[email protected]> wrote: >>> >>>> Yea the alpha-system repository. Although it could be called something >>>> else. At one point we talked about having a system directory in m5 >>>> that contained alpha/sparc/... code. That is probably a good bet. >>>> >>>> Ali >>>> >>>> On Jan 10, 2009, at 8:48 PM, Gabe Black wrote: >>>> >>>> >>>>> We actually have an implementation for Alpha's equivalent, right? The >>>>> console binary? Do we have that in a separate repository? Would we do >>>>> that for our BIOS of whichever flavor? >>>>> >>>>> Gabe >>>>> >>>>> nathan binkert wrote: >>>>> >>>>>> My guess is that people want to run Linux, OpenSolaris, and maybe >>>>>> someday windows. I think that OpenSolaris can probably do EFI. >>>>>> Also, >>>>>> the bios emulation is a software thing that you can probably load >>>>>> as a >>>>>> loader and runs on top of EFI, so maybe we can use that if it >>>>>> matters. >>>>>> >>>>>> I guess this is a vote in favor of EFI if it is indeed easier to >>>>>> implement. >>>>>> >>>>>> Nate >>>>>> >>>>>> >>>>>> On Fri, Jan 9, 2009 at 1:20 PM, <[email protected]> wrote: >>>>>> >>>>>> >>>>>>> According to Wikipedia, the following started supporting EFI at >>>>>>> some point: >>>>>>> >>>>>>> Linux with elilo >>>>>>> HP-UX on IA-64 >>>>>>> OpenVMS >>>>>>> OSX >>>>>>> Windows 2000 on Itanium >>>>>>> Windows Server 2003 for IA-64 >>>>>>> Windows XP 64-bit Edition >>>>>>> Windows blah blah. >>>>>>> >>>>>>> EFI is the only supported mechanism to boot OSX, I believe, and >>>>>>> BIOS is the only >>>>>>> supported mechanism to boot basically anything not in the list >>>>>>> above. The list >>>>>>> fortunately has most important OSes, but that would include DOS >>>>>>> for example. >>>>>>> >>>>>>> Basically, OSX has switched completely because they control the >>>>>>> whole hardware >>>>>>> stack. I think everyone else is switching too, but because people >>>>>>> apparently >>>>>>> still want to run windows 95 on their 16 way super servers they're >>>>>>> taking >>>>>>> longer. >>>>>>> >>>>>>> Gabe >>>>>>> >>>>>>> Quoting nathan binkert <[email protected]>: >>>>>>> >>>>>>> >>>>>>> >>>>>>>> I'd say whichever is easier. What guests are in each camp? >>>>>>>> >>>>>>>> Nate >>>>>>>> >>>>>>>> On Fri, Jan 9, 2009 at 10:50 AM, <[email protected]> wrote: >>>>>>>> >>>>>>>> >>>>>>>>> I don't plan on doing any work on this in the near future, but >>>>>>>>> what are >>>>>>>>> >>>>>>>>> >>>>>>>> people's >>>>>>>> >>>>>>>> >>>>>>>>> opinions about implementing an EFI BIOS for m5 rather than a >>>>>>>>> traditional >>>>>>>>> >>>>>>>>> >>>>>>>> BIOS? I >>>>>>>> >>>>>>>> >>>>>>>>> think EFI would be easier to implement and work with and easier >>>>>>>>> to get >>>>>>>>> documentation and support tools for, but a traditional BIOS would >>>>>>>>> >>>>>>>>> >>>>>>>> potentially >>>>>>>> >>>>>>>> >>>>>>>>> be compatible with more guests. EFI supports a compatibility >>>>>>>>> layer, but at >>>>>>>>> >>>>>>>>> >>>>>>>> that >>>>>>>> >>>>>>>> >>>>>>>>> point we'd be approximating implementing both, I think. That may >>>>>>>>> be ok >>>>>>>>> >>>>>>>>> >>>>>>>> because I >>>>>>>> >>>>>>>> >>>>>>>>> think most BIOS services are ignored by most modern guests, with >>>>>>>>> the fairly >>>>>>>>> minor exception of a boot loader. >>>>>>>>> >>>>>>>>> Gabe >>>>>>>>> >>>>>>>>> _______________________________________________ >>>>>>>>> m5-dev mailing list >>>>>>>>> [email protected] >>>>>>>>> http://m5sim.org/mailman/listinfo/m5-dev >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> m5-dev mailing list >>>>>>>> [email protected] >>>>>>>> http://m5sim.org/mailman/listinfo/m5-dev >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>> _______________________________________________ >>>>>> m5-dev mailing list >>>>>> [email protected] >>>>>> http://m5sim.org/mailman/listinfo/m5-dev >>>>>> >>>>>> >>>>> _______________________________________________ >>>>> m5-dev mailing list >>>>> [email protected] >>>>> http://m5sim.org/mailman/listinfo/m5-dev >>>>> >>>>> >>>> _______________________________________________ >>>> m5-dev mailing list >>>> [email protected] >>>> http://m5sim.org/mailman/listinfo/m5-dev >>>> >>>> >>> _______________________________________________ >>> m5-dev mailing list >>> [email protected] >>> http://m5sim.org/mailman/listinfo/m5-dev >>> >>> >>> >> _______________________________________________ >> m5-dev mailing list >> [email protected] >> http://m5sim.org/mailman/listinfo/m5-dev >> > > _______________________________________________ > m5-dev mailing list > [email protected] > http://m5sim.org/mailman/listinfo/m5-dev > > _______________________________________________ m5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/m5-dev
