[Bug 294027] safexcel: driver attaches but registers no algorithms with OCF without warning

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=294027

            Bug ID: 294027
           Summary: safexcel: driver attaches but registers no algorithms
                    with OCF without warning
           Product: Base System
           Version: Unspecified
          Hardware: arm64
                OS: Any
            Status: New
          Severity: Affects Some People
          Priority: ---
         Component: arm
          Assignee: [email protected]
          Reporter: [email protected]

I would like to report an issue related to crypto driver behavior and
observability on ARM-based systems.

On systems using safexcel(4), the driver may successfully attach during boot
but fail to register any algorithms with the Open Crypto Framework (OCF). In
this state:

The driver appears present (e.g., via kldstat)

No hardware crypto acceleration is actually used

The system silently falls back to software crypto

No warning or indication is emitted in dmesg or system logs

This makes the issue difficult to detect and diagnose.

In testing, this appears to occur in configurations where early consumers of
crypto (such as geli(8) encrypted swap) are initialized during boot. It is
possible that initialization order or resource contention prevents safexcel(4)
from registering its algorithms, though this has not been definitively
confirmed.

Regardless of the specific cause, the lack of visibility when a crypto driver
attaches but does not register any algorithms can lead to confusion and
prolonged troubleshooting.

I would like to propose that the system emit a warning in cases where:

A crypto driver attaches successfully

But registers zero algorithms with OCF

For example:

WARNING: safexcel0 attached but did not register any crypto algorithms.
Hardware acceleration may be unavailable.

This would improve observability and help identify hardware crypto issues
across platforms.

Thank you for your time and consideration. I am happy to provide additional
details or assist with testing if needed.

-- 
You are receiving this mail because:
You are the assignee for the bug.