Fix ARM32 floating point registers unwinding in PAL

[davidwrighton]

While building issue #56522 with PR #56919 it was discovered that float registers are not properly being unwound on Linux Arm32 platforms. See PR #56919 for details on how to repro.

We should fix this before we ship 6.0 and if the fix is simple enough we may consider backport to 5.0.

[dotnet-issue-labeler]

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[VincentBu]

Failed again in: runtime-coreclr jitstress 20210810.2

Failed test:

CoreCLR Linux arm Checked jitstress2 @ (Ubuntu.1804.Arm32.Open)Ubuntu.1804.Armarch.Open@mcr.microsoft.com/dotnet-buildtools/prereqs:ubuntu-18.04-helix-arm32v7-bfcd90a-20200121150440

- JIT/IL_Conformance/Convert/TestConvertFromIntegral/TestConvertFromIntegral.sh

Error message:

Return code:      1
Raw output file:      /root/helix/work/workitem/uploads/Reports/JIT.IL_Conformance/Convert/TestConvertFromIntegral/TestConvertFromIntegral.output.txt
Raw output:
BEGIN EXECUTION
/root/helix/work/correlation/corerun TestConvertFromIntegral.dll ''
No exception in DynamicConvertFromSystem.SingleToSystem.UInt640Opconv.ovf.u8
No exception in DynamicConvertFromSystem.SingleToSystem.UInt640Opconv.ovf.u8.un
The number of failed tests: 2
Expected: 100
Actual: 101
END EXECUTION - FAILED
Test Harness Exitcode is : 1
To run the test:

set CORE_ROOT=/root/helix/work/correlation
/root/helix/work/workitem/JIT/IL_Conformance/Convert/TestConvertFromIntegral/TestConvertFromIntegral.sh
Expected: True
Actual:   False


Stack trace
   at JIT_IL_Conformance._Convert_TestConvertFromIntegral_TestConvertFromIntegral_._Convert_TestConvertFromIntegral_TestConvertFromIntegral_sh() in JIT.IL_Conformance.XUnitWrapper.dll:token 0x6000004+0x295

[davidwrighton]

@janvorli had some ideas on how to fix this, but wasn't able to put them together before he went on vacation. As he has dramatically more experience than others on the team in this area, I'd like to wait for him to address this problem.

[mangod9]

Is this a regression?

[davidwrighton]

No, this isn't a regression.