Support FDT setup for CPU caches with high number of sets

[ckatsak]

Changes

The first commit improves the logging/reporting of errors during FDT creation (e.g., failures while parsing CPU cache information from sysfs).

The second commit changes the type of field number_of_sets in struct vmm::arch::aarch64::cache_info::CacheEntry from u16 to u32, thus allowing support for hosts with CPU caches with a number of sets higher than u16::MAX (e.g., NVIDIA GH200 Grace Hopper).

Reason

When we attempted to boot a microVM on a GH200 host following the Getting Started guide (with a minor modification, to download the aarch64 kernel binary), building the microVM failed, and the following was logged:

 . . .
2024-10-22T16:11:02.966071476 [anonymous-instance:fc_api:INFO:src/firecracker/src/api_server/parsed_request.rs:69] The API server received a Put request on "/actions" with body "{\n        \"action_type\": \"InstanceStart\"\n    }".
2024-10-22T16:11:02.966135189 [anonymous-instance:main:DEBUG:src/vmm/src/builder.rs:399] event_start: build microvm for boot
2024-10-22T16:11:02.972842006 [anonymous-instance:main:DEBUG:src/vmm/src/devices/acpi/vmgenid.rs:61] vmgenid: building VMGenID device. Address: 0x801ffff0. IRQ: 36
2024-10-22T16:11:02.972877750 [anonymous-instance:main:DEBUG:src/vmm/src/devices/acpi/vmgenid.rs:69] vmgenid: writing new generation ID to guest: 0xc9a01e17708cc1e2bc0518f9d2cc685e
2024-10-22T16:11:02.973158455 [anonymous-instance:main:INFO:src/vmm/src/lib.rs:833] Vmm is stopping.
2024-10-22T16:11:02.986500410 [anonymous-instance:fc_api:ERROR:src/firecracker/src/api_server/parsed_request.rs:190] Received Error. Status code: 400 Bad Request. Message: Start microvm error: System configuration error: Failed to create a Flattened Device Tree for this aarch64 microVM.
2024-10-22T16:11:02.986543450 [anonymous-instance:fc_api:DEBUG:src/firecracker/src/api_server/mod.rs:114] Total previous API call duration: 20487 us.

while this was the response of the API server:

{"fault_message":"Start microvm error: System configuration error: Failed to create a Flattened Device Tree for this aarch64 microVM."}

The first commit improves the reporting of failures while setting up FDT, by expanding the doc comments of the vmm::arch::aarch64::ConfigurationError::SetupFDT variant (thus also its implementation of std::fmt::Display, thanks to crates thiserror and displaydoc) to include the underlying vmm::arch::aarch64::fdt::FdtError.
As a result, the improved logs end up being as follows:

 . . .
2024-10-22T16:16:59.912859809 [anonymous-instance:fc_api:INFO:src/firecracker/src/api_server/parsed_request.rs:69] The API server received a Put request on "/actions" with body "{\n        \"action_type\": \"InstanceStart\"\n    }".
2024-10-22T16:16:59.912930657 [anonymous-instance:main:DEBUG:src/vmm/src/builder.rs:399] event_start: build microvm for boot
2024-10-22T16:16:59.919450370 [anonymous-instance:main:DEBUG:src/vmm/src/devices/acpi/vmgenid.rs:61] vmgenid: building VMGenID device. Address: 0x801ffff0. IRQ: 36
2024-10-22T16:16:59.919484802 [anonymous-instance:main:DEBUG:src/vmm/src/devices/acpi/vmgenid.rs:69] vmgenid: writing new generation ID to guest: 0x9934a6187f502a62579e99b35bed8b8f
2024-10-22T16:16:59.919760995 [anonymous-instance:main:INFO:src/vmm/src/lib.rs:833] Vmm is stopping.
2024-10-22T16:16:59.930398232 [anonymous-instance:fc_api:ERROR:src/firecracker/src/api_server/parsed_request.rs:190] Received Error. Status code: 400 Bad Request. Message: Start microvm error: System configuration error: Failed to create a Flattened Device Tree for this aarch64 microVM: Read cache info error: Invalid cache configuration found for number_of_sets: number too large to fit in target type
2024-10-22T16:16:59.930448408 [anonymous-instance:fc_api:DEBUG:src/firecracker/src/api_server/mod.rs:114] Total previous API call duration: 17606 us.

API server's HTTP response becomes more descriptive as well:

{"fault_message":"Start microvm error: System configuration error: Failed to create a Flattened Device Tree for this aarch64 microVM: Read cache info error: Invalid cache configuration found for number_of_sets: number too large to fit in target type"}

The second commit fixes the actual bug for the GH200 host (unveiled by and shown in the improved logs above). In this case, host CPU's number of sets for the unified L3 cache is higher than u16::MAX:

$ bat --style header /sys/devices/system/cpu/cpu0/cache/index3/{level,size,number_of_sets}
File: /sys/devices/system/cpu/cpu0/cache/index3/level
3

File: /sys/devices/system/cpu/cpu0/cache/index3/size
116736K

File: /sys/devices/system/cpu/cpu0/cache/index3/number_of_sets
155648

Changing the type of the number_of_sets field in struct vmm::arch::aarch64::cache_info::CacheEntry from u16 to u32 suffices to fit a high number of CPU cache sets, thus allowing the microVM to be built and booted without any problem.

License Acceptance

By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache 2.0 license. For more information on following Developer
Certificate of Origin and signing off your commits, please check
CONTRIBUTING.md.

PR Checklist

• If a specific issue led to this PR, this PR closes the issue.
• The description of changes is clear and encompassing.
• Any required documentation changes (code and docs) are included in this
  PR.
• API changes follow the Runbook for Firecracker API changes.
• User-facing changes are mentioned in CHANGELOG.md.
• All added/changed functionality is tested.
• New TODOs link to an issue.
• Commits meet
  contribution quality standards.

---

• This functionality cannot be added in rust-vmm.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.07%. Comparing base (5ef02a8) to head (aa50499).
Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4869   +/-   ##
=======================================
  Coverage   84.07%   84.07%           
=======================================
  Files         251      251           
  Lines       28052    28052           
=======================================
  Hits        23586    23586           
  Misses       4466     4466           

Flag	Coverage Δ
5.10-c5n.metal	84.71% <ø> (ø)
5.10-m5n.metal	84.69% <ø> (ø)
5.10-m6a.metal	84.00% <ø> (ø)
5.10-m6g.metal	80.70% <100.00%> (ø)
5.10-m6i.metal	84.69% <ø> (-0.01%)	⬇️
5.10-m7g.metal	80.70% <100.00%> (ø)
6.1-c5n.metal	84.71% <ø> (ø)
6.1-m5n.metal	84.69% <ø> (-0.01%)	⬇️
6.1-m6a.metal	84.00% <ø> (ø)
6.1-m6g.metal	80.69% <100.00%> (-0.01%)	⬇️
6.1-m6i.metal	84.69% <ø> (ø)
6.1-m7g.metal	80.70% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bchalios]

Hey guys,

thanks a lot for your contribution. It generally LGTM. Could you fix the style issue? It's not in your changes, but your changes did make redundant a u32::from call (since the argument is now a u32) and clippy complains about it.

[bchalios]

Thanks a lot for the contribution