Re: Odd PSW Address does not trigger a specification exception.

[Thomas Huth]

On 30/01/2023 18.31, Harold Grovesteen wrote:
> While doing some testing, I encountered the situation where an odd PSW
> address is introduced into the active PSW by means of a BC instruction.
> The odd address is used by Qemu for successful instruction fetching.

Do you have, by any chance, a small test program that allows to reproduce 
the wrong behavior?

> I am using qemu-7.2.0-rc4. I have noticed in this list some patches on
> or about 9 Jan 2023 related to displacement addresses and the
> introduction of "gen_psw_addr_disp". The version of qemu I am using
> does not contain that function, so I can not validate whether it
> addresses the odd instruction address I encountered in my testing.

I assume that's this patch here:

 https://lists.gnu.org/archive/html/qemu-devel/2023-01/msg01591.html

?

... it hasn't been merged yet, I think, but it also doesn't look like there 
is any code in there to deal with odd addresses, as far as I can see at a 
first quick glance.

 Thomas

> z/Architecture Principles of Operation manual -13 on page 5-13 states:
>
> "A specification exception due to an odd branch address and access
> exceptions due to fetching of the instruction at the branch location
> are not recognized as part of the branch operation but instead are
> recognized as exceptions associated with the execution of the
> instruction at the branch location."
>
> The specification exception was not recognized at all. Rather the
> instruction coded at the odd address, LGFI, was fetched and executed.
>
> It was not until the X'00', also at an odd address, immediately
> following the LGFI was itself fetched for execution that the program
> ceased operation with an operation exception program interrupt.
>
> I am hoping that this is sufficient information to determine if this is
> a known reaction by Qemu to this condition and whether the new patches
> will address this.  If neither is the case, FYI.
>
> If you require details, please contact me here or off list.
>
> Thanks,
> Harold Grovesteen