arm64: fix tpidr maybe null

[hujun260]

Note: Please adhere to Contributing Guidelines.

Summary

arm64: fix tpidr maybe null

Before the MPU initialization, the up_update_task(this_cpu()) function is called at a time when hardware cache coherency is not yet enabled. In certain critical scenarios, Core 1 reads a zero value for tcb from the global variable g_assignedtask and stores this zero value into the tpidr register. This results in subsequent interrupt handlers reading a zero tcb, causing an exception.

Impact

arm64

Testing

qemu-armv8a:nsh_smp

[nuttxpr]

[Experimental Bot, please feedback here]

Yes, this PR likely meets the basic NuttX requirements but is missing crucial information.

Here's why it's borderline and what's missing:

• Summary: While it explains the why and what, it lacks a clear explanation of how the fix works. What code was changed to prevent the zero value from being stored? This is the most important part of the summary.
• Impact: Stating "arm64" is insufficient. It needs to specify:
  • Is new feature added? Is existing feature changed? (This is a bug fix, so existing feature changed)
  • Impact on user: Likely NO, but explicitly state it.
  • Impact on build: Likely NO, but explicitly state it.
  • Impact on hardware: YES, arm64. Be specific which arm64 chips/boards are affected.
  • Impact on documentation: Likely NO, but explicitly state it unless documentation was updated to reflect the issue/fix.
  • Impact on security: YES, potentially. Explain the security implications of a null tpidr. A crash due to a null pointer exception is a security issue if it can be triggered by user input or external events.
  • Impact on compatibility: Likely NO, but explicitly state it.
• Testing: "qemu-armv8a:nsh_smp" is insufficient. Provide:
  • Build Host: Your OS, CPU architecture, compiler, and version.
  • MORE detailed Target: Specify the QEMU version and the exact configuration used.
  • Crucially missing: Logs. There are no "before" and "after" logs demonstrating the fix. This is a critical requirement. Show the error occurring before the fix and the successful operation after the fix. Ideally, the logs should demonstrate the tpidr value being correctly populated.

To make this PR acceptable, you need to add the missing details outlined above. Specifically, clarify how the fix works, expand on the impact section answering all the questions, and provide detailed build host information and actual before/after logs demonstrating the fix.

[xiaoxiang781216]

do we need change other arch too

[hujun260]

Yes, the current community code implements percpu storage for this_task on arm64, armv7a, armv7r, and armv8r archs, all using this approach.