fix 'uninitialized value $ucode_vars{"AVAIL"}' from AMD ucode

[mmitch]

This should fix the warning uninitialized value $ucode_vars{"AVAIL"} when no microcode is available for an AMD CPU (issue #191).
I have fixed this by adopting the logic for Intel CPUs.

This means that the case we know the microcode inside the CPU but we don't have any firmware files to see if it is current or not now maps to NRM_UNKNOWN (like the Intel code did in that case) instead of NRM_CURRENT (which was exactly the source of the bug: if we are current we need to know an AVAIL version, but we didn't).

Unfortunately, I have no system to really test this on: all my current CPUs (Intel as well as AMD) have no microcode at all available for them.

[liske]

Thanks!

[anarcat]

It seems like this introduced a regression that all our AMD systems are now marked "UNKNOWN" here: #249

@mmitch could you review #285 to see if it makes sense? thanks!

update: i'm actually rooting for #288 now, see #249 (comment) for more context.

[mmitch]

I have Debian 12 Bookworm with a Ryzen 5700G and no firmware updates available.
I've tried both #285 and #288 - on my systems, there is no difference, both work exactly the same:

• default needrestart 3.6-4 from Debian gives me Use of uninitialized value $processor in concatenation (.) or string at /usr/share/perl5/NeedRestart/uCode.pm line 61. when I don't have the cpuid kernel module loaded AND I use the needrestart -v. Apart from that, I get the message Failed to check for processor microcode upgrades.
• both Fix AMD Microcode Check #285 and fix AMD ucode checking in non-debug mode #288 fix the uninitialized value error. They also both report Failed to check for processor microcode upgrades as expected

I have read the discussion in #249 regarding UNKNOWN and the Nagios output mode and would be in favor of the "UNKNOWN should not map to a Nagios error" argumentation given.

My home-grown monitoring system uses needrestart -b (batch mode) and as long as UNKNOWN still results in return code 0 from the needrestart call (like it currently is with vanilla, #285 and #288), I'm happy.

Same goes for the apt integration in Debian: As long as the return code is unaffected, needrestart can print the warning about UNKNOWN without any side effects.