virtio: avoid spurious config change interrupt on DRIVER_OK

[Bhargav (bharqav)]

Summary

Fixes Issue #4049 - "virtio-pci: spurious configuration-change interrupt (ISR bit 1) raised at DRIVER_OK"

The VirtIO transport currently updates config_generation when the
device reaches DRIVER_OK.

The DRIVER_OK transition does not represent a change to the
device-specific configuration space. However, updating the generation
during this transition also causes a configuration-change interrupt to
be signaled during device initialization.

This change removes that spurious update and adds regression coverage
for MMIO and PCI transports.

Changes

• Remove update_config_generation() from the DRIVER_OK completion path.
• Preserve the existing FEATURES_OK config-generation update.
• Update initialization expectations to reflect the correct generation.
• Verify that DRIVER_OK does not generate a configuration-change
  interrupt for MMIO.
• Verify that DRIVER_OK does not generate a configuration-change
  interrupt for PCI.
• Remove existing test expectations that depended on the spurious
  initialization interrupt.

Testing

• cargo fmt --check -p virtio — passed
• git diff --check — passed
• cargo test -p virtio — unable to execute locally because the
  Windows SDK required by the MSVC toolchain is unavailable
  (kernel32.lib).

CI validation is therefore required for the full test suite.

[github-actions]

⚠️ Unsafe Code Detected

This PR modifies files containing unsafe Rust code. Extra scrutiny is required during review.

For more on why we check whole files, instead of just diffs, check out the Rustonomicon

[Bhargav (bharqav)]

@microsoft-github-policy-service agree

[github-actions]

1 Petri tests failed (0 unstable)

[Steven Malis (smalis-msft)]

Thanks for your submission, someone will take a look soon! Don't worry about any SNP test failures, they're having some issues right now.

[github-actions]

2 Petri tests failed (0 unstable)