Re: [kvm-unit-tests PATCH v2 2/2] s390x: pgm interrupt handler and a way to test them

[David Hildenbrand <david@redhat.com>]

On 31.05.2017 14:51, Paolo Bonzini wrote:
> 
> 
> On 31/05/2017 14:39, David Hildenbrand wrote:
>> +	expect_pgm_int();
>> +	*((unsigned int*)-1) = 1;
>> +	check_pgm_int_code(PGM_INT_CODE_ADDRESSING);
> 
> QEMU gets
> 
> ABORT: selftest: Unexpected program interrupt: 5 at 0x12b4c, ilen 4
> 
> It seems that QEMU doesn't point the PSW at the next instruction:
> 
> 	IN: main
> 	0x0000000000012b44:  lghi       %r1,-1
> 	0x0000000000012b48:  lhi        %r3,1
> 	0x0000000000012b4c:  st %r3,0(%r1)
> 	0x0000000000012b50:  lghi       %r2,5
> 	0x0000000000012b54:  brasl      %r14,0x128e8
> 	0x0000000000012b4c:  st %r3,0(%r1)
> 	0x0000000000012b50:  lghi       %r2,5
> 	0x0000000000012b54:  brasl      %r14,0x128e8     # this is expect_pgm_int
> 
> 	Trace 0x7f5ff6c7f520 [0: 0000000000012b44] main
> 	Trace 0x7f5ff6c7bc20 [0: 000000000001004e] 
> 	Trace 0x7f5ff6c7c960 [0: 0000000000012918] handle_pgm_int
> 	Trace 0x7f5ff6c7d280 [0: 00000000000100a4] 
> 
> and now it gets another program interrupt:

Exactly, the emulator doesn't properly forward the PSW.

> 
> 	IN: main
> 	0x0000000000012b4c:  st %r3,0(%r1)
> 	0x0000000000012b50:  lghi       %r2,5
> 	0x0000000000012b54:  brasl      %r14,0x128e8
> 
> 	Trace 0x7f5ff6c7f600 [0: 0000000000012b4c] main
> 	Trace 0x7f5ff6c7bc20 [0: 000000000001004e] 
> 	Trace 0x7f5ff6c7c960 [0: 0000000000012918] handle_pgm_int
> 
> Thanks,
> 
> Paolo
> 

Yes, known problem as described in the cover letter. Happens with
current TCG and old KVM. Patch for TCG is on the list.

Thanks for testing!

-- 

Thanks,

David