Re: [PATCH v20 14/14] KVM: selftests: arm64: Add GCS EXLOCK exception emulation test

[Mark Brown <broonie@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 1314 bytes --]

On Thu, Sep 10, 2026 at 06:20:30PM +0100, Leonardo Bras wrote:

> The part of re-running test_udf_exception() with HFGITR_EL2 set, IIUC, it
> that part of fine-grained traps. It seems correct, although I am not seeing
> any tests of the EC syndrome being 0x1A to make sure it's in use.

Now that I look again this isn't actually effective anyway, we would
need to run the guest code at emulated EL1 which for something that's
basically just a double check.  I'll just delete the extra runs.

> There are also a lot of expect_exlock being used as an int, even if it's a 
> bool. I understand the idea of summing it as we expect it to be 1 on true 
> or 0 on false, but is it really ok?
> Maybe we could change it to int, so we just set it when we enter 
> test_sysreg() or test_udf_exception() like:

> -	expect_exlock = enable_exlock;
> +	expect_exlock = enable_exlock ? 1 : 0;

The conversions between int and and boolean values are part of the C
spec and widely used within the kernel, we don't need to open code it.
A zero in a boolean context is false, any other integer is true.
Similarly false in an integer context is 0, and true is 1.

> If I got the workings right, and both above suggestions (0x1A and 
> bool->int) are either fixed or considered not an issue, then:

Thanks.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]