The GUID has already been declared in EmbeddedPkg/Include/Guid/Fdt.h. I have just pushed some helper functions to load FDT from Firmware Volume (svn rev15905) or File System (svn rev15908). These functions could be used by VirtFdtDxe.c.
> -----Original Message----- > From: Ard Biesheuvel [mailto:ard.biesheu...@linaro.org] > Sent: 26 August 2014 11:12 > To: Laszlo Ersek > Cc: Olivier Martin; edk2-devel@lists.sourceforge.net; Peter Maydell; > Christoffer Dall; Andrew Jones; Ilias Biris; Leif Lindholm > Subject: Re: [PATCH 01/10] Add minimal support for passing a device > tree image > > On 26 August 2014 11:39, Laszlo Ersek <ler...@redhat.com> wrote: > > On 08/25/14 21:19, Ard Biesheuvel wrote: > >> This adds the possibility to include a DTB blob into the firmware > image, and > >> have it installed as a configuration under the correct GUID at UEFI > init time. > >> > >> Signed-off-by: Ard Biesheuvel <ard.biesheu...@linaro.org> > >> --- > [...] > > > > I'm trying to familiarize myself with the structure of this series. > > > > Before I get to the real meat (or what I consider to be the "real > > meat"), I'd like to ask why we need this patch. I searched the other > > patches for "FdtTableDxe", and I got no hits. > > > > This patch does two things: > - declare a GUID for a device tree configuration table > - implement a driver that takes an embedded FV containing a device > tree blob and install it as a configuration table under the declared > GUID > > The latter part is not used in this series, but the GUID /is/ used by > VirtFdtDxe.c to install the device tree as a config table. This is > what allows the kernel to boot straight from UEFI without having to > pass -dtb arguments and have the EFI stub perform the load. Granted, > this is somewhat of a bonus, and the EFI stub does support dtb=, but > we disable it for secure boot because loading unauthenticated data and > install it right into the heart of the kernel is considered a security > hazard. > > So i could split it up and move the part we need under ArmPkg, and > drop the other half, I suppose. > I know Olivier is also working on device tree integration, so perhaps > he could chime in as well? > > @Olivier:? > > > As far as I understand, this functionality is useful for the linaro > > tree, and some of the physical hardware platforms. In a virtual > machine > > though, the DTB will always come from QEMU, so for aarch64 VM > support, > > this patch doesn't appear to be necessary. > > > > The point is not where the device tree came from, but where it goes to > ... > > > Can you drop it from this series? (Along with 02/10, using the FDF > > change you described before.) I'd like to reduce the things I must > > understand to a minimum. > > > > Additionally, adding anything at all to MdePkg and MdeModulePkg is > > pretty hard in upstream edk2, so if we don't have to, let's not try. > :) > > If you drop the first two patches, then your series remains in the > > confines of ArmPkg and ArmPlatformPkg. I can't stress enough how much > > easier that makes it for your series to get applied. > > > > Agreed. > > -- > Ard. ------------------------------------------------------------------------------ Slashdot TV. Video for Nerds. Stuff that matters. http://tv.slashdot.org/ _______________________________________________ edk2-devel mailing list edk2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/edk2-devel
