Regression on RPi with commit 4cbb293 (efi_loader: consider no-map property of reserved memory)

[Peter Robinson]

Hey Heinrich,

Not sure why this hasn't come up before as the issue with testing goes
back to 2021.01-rc1/merge window, or if it's something like a newer
firmware (I currently have one from 2021-02-25) but I'm seeing issues
with the Raspberry Pi bringing up the secondary CPUs in Linux [1].

Ine of our users worked out it was a u-boot issue and with bisecting I
got it down to down to the "efi_loader: consider no-map property of
reserved memory" commit 4cbb293. I don't know too much about SMP
bringing online CPUs on the RPi raspberry pi but I wonder if that
commit is clobbering some piece of memory needed for that (I vaguely
remember some mention of spin tables).

I see it on both the original RPi3 and the 3B+, what are your thoughts?

Peter

[    0.008498] Remapping and enabling EFI services.
[    0.009304] smp: Bringing up secondary CPUs ...
[    5.042966] CPU1: failed to come online
[    5.042994] CPU1: failed in unknown state : 0x0
[   10.085989] CPU2: failed to come online
[   10.086017] CPU2: failed in unknown state : 0x0
[   15.128975] CPU3: failed to come online
[   15.129003] CPU3: failed in unknown state : 0x0
[   15.129114] smp: Brought up 1 node, 1 CPU
[   15.129138] SMP: Total of 1 processors activated.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1921924