[RFC] fix incorrect GSI numbering causing device limit on aarch64

[Swchexit]

I’m opening this as an RFC because I would appreciate feedback on my proposed solution. Please see my detailed analysis and reasoning on the issue page #4207 . Additionally, I’m encountering timeout problems running the test suite with my current setup and would appreciate assistance verifying this patch.

Changes

Introduce two new constants (GSI_BASE and GSI_MAX) in the layout.rs files for both aarch64 and x86_64. These constants correctly adjust the GSI numbering for aarch64 VMs, aligning Firecracker’s interrupt assignments with Linux’s expectations and ensuring that the interrupts provided in the device tree and registered via irqfd are correctly numbered.

Reason

To resolve issue #4207 , where incorrect GSI numbering caused by using an offset of 32 (IRQ_BASE) resulted in an artificial limitation of 64 devices on aarch64 microVMs. This change removes that limitation, enabling aarch64 microVMs to correctly use up to 95 devices.

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

• I have read and understand CONTRIBUTING.md.
• I have run tools/devtool checkstyle to verify that the PR passes the
  automated style checks.
• I have described what is done in these changes, why they are needed, and
  how they are solving the problem in a clear and encompassing way.
• I have updated any relevant documentation (both in code and in the docs)
  in the PR.
• I have mentioned all user-facing changes in CHANGELOG.md.
• If a specific issue led to this PR, this PR closes the issue.
• When making API changes, I have followed the
  Runbook for Firecracker API changes.
• I have tested all new and changed functionalities in unit tests and/or
  integration tests.
• I have linked an issue to every new TODO.

---

• This functionality cannot be added in rust-vmm.

[codecov]

Codecov Report

Attention: Patch coverage is 88.88889% with 1 line in your changes missing coverage. Please review.

Project coverage is 82.91%. Comparing base (5ce383e) to head (b46706b).
Report is 7 commits behind head on main.

Files with missing lines	Patch %	Lines
src/vmm/src/device_manager/mmio.rs	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5283      +/-   ##
==========================================
+ Coverage   82.86%   82.91%   +0.05%     
==========================================
  Files         250      250              
  Lines       26897    26897              
==========================================
+ Hits        22288    22302      +14     
+ Misses       4609     4595      -14     

Flag	Coverage Δ
5.10-c5n.metal	83.35% <100.00%> (ø)
5.10-m5n.metal	83.35% <100.00%> (+<0.01%)	⬆️
5.10-m6a.metal	82.56% <100.00%> (-0.01%)	⬇️
5.10-m6g.metal	79.17% <83.33%> (ø)
5.10-m6i.metal	83.34% <100.00%> (-0.01%)	⬇️
5.10-m7a.metal-48xl	82.55% <100.00%> (?)
5.10-m7g.metal	79.17% <83.33%> (ø)
5.10-m7i.metal-24xl	83.31% <100.00%> (?)
5.10-m7i.metal-48xl	83.30% <100.00%> (?)
5.10-m8g.metal-24xl	79.17% <83.33%> (?)
5.10-m8g.metal-48xl	79.17% <83.33%> (?)
6.1-c5n.metal	83.39% <100.00%> (-0.01%)	⬇️
6.1-m5n.metal	83.40% <100.00%> (ø)
6.1-m6a.metal	82.61% <100.00%> (-0.01%)	⬇️
6.1-m6g.metal	79.16% <83.33%> (ø)
6.1-m6i.metal	83.39% <100.00%> (+<0.01%)	⬆️
6.1-m7a.metal-48xl	82.60% <100.00%> (?)
6.1-m7g.metal	79.17% <83.33%> (ø)
6.1-m7i.metal-24xl	83.40% <100.00%> (?)
6.1-m7i.metal-48xl	83.41% <100.00%> (?)
6.1-m8g.metal-24xl	79.16% <83.33%> (?)
6.1-m8g.metal-48xl	79.17% <83.33%> (?)

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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Manciukic]

aarch64 complains about:

error: useless conversion to the same type: `u32`
   --> src/vmm/src/arch/aarch64/fdt.rs:365:13
    |
365 |             dev_info.irq.unwrap().into(),
    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: consider removing `.into()`: `dev_info.irq.unwrap()`
    |
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#useless_conversion
    = note: `-D clippy::useless-conversion` implied by `-D warnings`
    = help: to override `-D warnings` add `#[allow(clippy::useless_conversion)]`
error: useless conversion to the same type: `u32`
   --> src/vmm/src/arch/aarch64/fdt.rs:386:13
    |
386 |             dev_info.irq.unwrap().into(),
    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: consider removing `.into()`: `dev_info.irq.unwrap()`
    |
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#useless_conversion

[Manciukic]

We also need to update a few entries in the documentation:

• add a changelog entry describing the fix
• remove the 64 device limitation here:

  firecracker/docs/device-api.md

  Lines 128 to 132 in 931a233

  	## Known device limitations
  	If more than 64 devices are configured for a VM in total on aarch64, only first
  	64 of them are functional
  	([related issue](https://github.com/firecracker-microvm/firecracker/issues/4207)).

PS: kudos for finding the problem and fixing the issue! Thanks!

[Swchexit]

Hi, I've fixed the clippy error, updated the CHANGELOG, and removed the description of the 64-device limit from the docs. I will try to run the extended test as soon as I fix my settings.

I will also clean up commits if there are no further changes to be made.

[roypat]

Alright, I think the last failures we can fix by just disabling the memory monitor in the "too many devices" test (pass monitor_memory=False to microvm_factory.build()).