Date: Sun, 1 Nov 2015 16:54:41 +0100 From: Emmanuel Vadot <manu@bidouilliste.com> To: freebsd-hackers@freebsd.org Subject: Re: EFI Variables Message-ID: <20151101165441.ee31e416fa004637879561c7@bidouilliste.com> In-Reply-To: <201510280727.19357.ganael.laplanche@corp.ovh.com> References: <6ce779725aab266bc85e92f0ee2186b6@megadrive.org> <201510280727.19357.ganael.laplanche@corp.ovh.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 28 Oct 2015 07:27:19 +0100 Ganael Laplanche <ganael.laplanche@corp.ovh.com> wrote: > On Tuesday, October 27, 2015 07:24:23 PM Emmanuel Vadot wrote: > > Hi Emmanuel, > > > I'm currently hacking around the loader.efi > > Great :) > > > I've also added the list and get command to the not working "nvram" > > command. > > I had myself posted a PR to fix that command as well as add a verbose switch > and the ability to specify a variable name, see : > > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=202614 > > > For the "set" subcommand I think that the best way to handle it is : > > "nvram set myvar data" -> This will set the variable myvar to data with > > the freebsd guid (if there is any) > > > > and > > > > "nvram set myvar guid data" -> This will force the guid to <guid> > > It can be useful to set variables containing *strings*, but will hardly handle > binary stuff :/ > > I am not sure whether it should be the loader's job to set variables... I can > think of changing the boot order, but it may be difficult to get it right by > hand and would probably require an upper-level tool, such as efibootmgr on > Linux. > > > I'll look tomorrow how to access efivars once the kernel is booted so > > we can set some from some userland tool (especially the boot related > > one). > > Yes, this is interesting as the current kernel (amd64) does not provide access > to EFI variables at all. > > 10.x/ia64 provided access to EFI variables through libefi(3) and io(4). It > should be possible to import that code to other archs too, but you'll have to > save the entry point to the Runtime Services Tables and maybe set a Virtual > Address Map too (not sure about that point). > > Best regards, > The entry point for the runtime service isn't a problem, I set it as a metadata in loader.efi (this is done the same way for the graphic output protocol and efifb). I don't know if there is a way to call a function from it's physical addresses (doing it directly kernel panic), if not we need to set a Virtual Address Map and Convert all the addresses. -- Emmanuel Vadot <manu@bidouilliste.com>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20151101165441.ee31e416fa004637879561c7>