IPV4 UEFI grub2 PXE boot, network unavailable immediately after grub starts

IPV4 UEFI grub2 PXE boot, network unavailable immediately after grub starts

[alexander.burmashev]

Hello!
Per Daniel's suggestion reaching grub-devel, since people might be 
already working on this.
For reference: https://savannah.gnu.org/bugs/?63245


TLDR: during UEFI PXE boot + IPv4 grub drops to cmdline, unable to get 
grub config, network is not available. We have a huge fleet of machines, 
and with latest 2.06 we are regularly hitting this bug during PXE boot. 
https://savannah.gnu.org/bugs/?63245
I am wondering if someone else is hitting this issue as well and have 
some ideas/WIP to share.

It is indeed random, it could be 1 boot out of 1000, or 5 boots in a 
row. Nature of the bug makes it hard to debug.

There are multiple machines with identical hardware provisioned at the 
same time, and it is absolutely random which will be hit by this bug. If 
you reboot the affected box, most likely it will come back just fine and 
with no issue.

i was not able to catch any abnormal traffic before it happens via 
tcpdump, nothing suspicious in tftp log, though i might have not been 
trying/checking hard enough. According to grub variables everything is 
setup correctly, so there is no issue with dhcp traffic ?

The exact loop grub2 is entering is permacycling over 
https://git.savannah.gnu.org/cgit/grub.git/tree/grub-core/net/drivers/efi/efinet.c#n68

i was able to reproduce the issue both with RHEL 8 ( rhel-8.7.0 ) grub2 
patchset and with latest upstream grub2.

bug filer for https://savannah.gnu.org/bugs/?63245 mentioned in his 
thread with intel that he was hitting that with one the latest canonical 
grubs.