[bug #61101] UEFI PXE boot over VLAN is failing with 'could not resolve hardware address' msg

[Scott Serr]

Follow-up Comment #4, bug #61101 (project grub):

I had some discussion with TJ_ on IRC/#grub.  TJ was very helpful, here are
some relevant links and findings we discussed:

This thread is related:
https://lists.gnu.org/archive/html/help-grub/2017-01/msg00002.html "UEFI-PXE
boot to grub2 with BIOS configured VLAN tagging?"

VLAN Trunk: The system trying to PXE boot is on a trunk port, not access port.
 The post to the help-grub has a better use case than I do, "...think
hypervisor host with just one interface which is a VLAN trunk."

Found 2 links to informative UEFI documentation:
-
https://edk2-docs.gitbook.io/edk-ii-uefi-driver-writer-s-guide/25_network_driver_design_guidelines/253_simple_network_protocol_implementations
-
https://edk2-docs.gitbook.io/edk-ii-uefi-driver-writer-s-guide/25_network_driver_design_guidelines

iPXE vision explaining poor UEFI features:
- https://ipxe.org/efi/vision

>From TJ: there's a very interesting comment in the source of
grub-core/net/drivers/efi/efinet.c::grub_efi_net_config_real() around line 353
about how EFI deals with PXE.

A workaround that TJ proposed (that is possibly better than using my old grub
0.97 for these specific boots): UEFI pxe -> DHCP -> iPXE -> grub2.  "Host UEFI
PXE DHCP/TFTP fetches the iPXE UEFI executable as the PXE payload. iPXE can
then be configured to do the same again (DHCP/TFTP) to load GRUB"

Possibly a hint: 
My main test system is a Dell with IDRAC on a shared nic. The IDRAC is also
configured to use the same same VLAN.  Odd thing, the IDRAC is very unstable
during these attempted PXE boots, the IDRAC itself hangs/reboots seemingly
every time the system tries to PXE boot.  The IDRAC is stable via network when
not doing these PXE boot tests.  It is also stable when accessed from USB
(instead of network) while doing these PXE boot tests.

Potential debug options:
- build all needed modules into the grub image for better diagnostic info
- (hack) adding the option to set a vlan tag number from a variable would be
useful as a test

I will attempt to:
- build modules in base image and post (in text) all the grub variable (using
a text console instead of graphical)
- get a workaround by bouncing through iPXE
- setup a qemu environment where I can test and share configuration with
others


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?61101>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/