[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Obtaining the UUID of the system for a PXE boot
From: |
Andrey Borzenkov |
Subject: |
Re: Obtaining the UUID of the system for a PXE boot |
Date: |
Tue, 23 Jul 2013 20:45:03 +0400 |
В Tue, 23 Jul 2013 18:24:55 +0200
Holger Goetz <address@hidden> пишет:
> Hi,
>
> looks i could need some help here.
>
> I've moved the smbios module from commands/i386 to commands and added
> some casts to make the compiler happy. As I need a x86_64-efi grub.
> What i see now is that the "_SM_" signature is not found: "Failed to
> locate entry point structure" some added grub_dprintf's show all zeros
> in the memory area.
> Actually when i use the "dump" command to display the memory in the area
> that gets swiped (0xf0000..0x10000) it (also) only displays "00"s.
> Is there some memory mapping required in x86_64 mode to access the
> memory area properly? I did cross check w/ the Linux kernel efi code and
> syslinux all basically do the same searching.
>
You need to fetch SMBIOS table address from EFI system table. See as
example
https://git.ipxe.org/ipxe.git/blob/HEAD:/src/interface/efi/efi_smbios.c
You can use grub_machine_acpi_get_rsdpv2() as example implementation.
Just add SMBIOS GUID and any sanity checks as required.
> Thanks,
> Holger
>
>
>
> On 19.07.2013 18:31, David Michael wrote:
> > Hi,
> >
> > On Fri, Jul 19, 2013 at 10:24 AM, Holger Goetz <address@hidden> wrote:
> >> Yes it's towards the right direction. But it is 32bit only if i understand
> >> correctly, and it basically is a memory access to fixed/hardcoded MEMORY
> >> address (0x80000001). to pick the veondor id and machine info.
> > It's true that I've mostly been using the module on 32-bit virtual
> > machines, so it hasn't really been tested elsewhere. However, I'm not
> > sure I understand what you mean by the hard-coded memory address. The
> > first function grub_smbios_locate_eps searches for the SMBIOS entry
> > point structure as described in the spec. The table entries are then
> > read at the table address found in the EPS, not a hard-coded location.
> >
> >> I have only 64bit - UEFI here - therefore the approach w/ first searching
> >> the SMBIOS infoblock in memory is probably required. And then properly walk
> >> through the info-tables/blocks to get to the UUID entry. It doesn'T need to
> >> be a fixed info to be retrieved from the SMBIOS memory - maybe a generic
> >> function to query/search a specific entry and return that to be assigned to
> >> a variable would be more flexible.
> > The module's command-line interface does use a (dumb) query/search
> > method. You can specify the desired entry's type and/or handle and
> > the data to retrieve from it. For example the following command
> > prints the machine name (i.e. the string at offset 5 in an entry with
> > type 1).
> >
> > smbios -t 1 -s 5
> >
> > I think you may have found the first patch I sent in that old thread,
> > which was for different functionality. The SMBIOS module can be
> > downloaded from the list archive[1].
> >
> > Unfortunately it doesn't have any convenient functions to output a
> > usable UUID. It shouldn't take much to add one: the variable "entry"
> > in the "main" function is a pointer to the matched entry, so entry[8]
> > through entry[23] is the UUID in a call "smbios -t 1 ...". I've
> > verified these bytes correspond to dmidecode output on my physical
> > hardware with the following.
> >
> > for i in 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
> > do smbios -t 1 -b $i
> > done
> >
> > If the module isn't salvageable on UEFI, maybe I can send out an
> > updated version whenever I upgrade to such a system.
> >
> > Thanks.
> >
> > David
> >
> > [1] https://lists.gnu.org/archive/html/grub-devel/2013-04/binx8am8MvVSh.bin
> >
> > _______________________________________________
> > Grub-devel mailing list
> > address@hidden
> > https://lists.gnu.org/mailman/listinfo/grub-devel
> >
> >
>