Re: [risu PATCH 1/4] s390x: Add basic s390x support to the C code

[Thomas Huth <thuth@redhat.com>]

On 04/09/2023 16.19, Ilya Leoshkevich wrote:
> On Mon, 2023-09-04 at 16:00 +0200, Thomas Huth wrote:
>> With these changes, it is now possible to compile the "risu" binary
>> for s390x hosts.
>>
>> Signed-off-by: Thomas Huth <thuth@redhat.com>
>> ---
>>   risu_reginfo_s390x.c | 142
>> +++++++++++++++++++++++++++++++++++++++++++
>>   risu_reginfo_s390x.h |  23 +++++++
>>   risu_s390x.c         |  48 +++++++++++++++
>>   test_s390x.S         |  32 ++++++++++
>>   4 files changed, 245 insertions(+)
>>   create mode 100644 risu_reginfo_s390x.c
>>   create mode 100644 risu_reginfo_s390x.h
>>   create mode 100644 risu_s390x.c
>>   create mode 100644 test_s390x.S
> 
> Looks really interesting! I was doing similar qemu-system-s390x testing
> with a bunch of ad-hoc scripts, and there are quite a few unresolved
> problems still, especially in the error handling area.
> 
> Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>
> 
> I have one small question (see below).
> 
> [...]
> 
>> +/* reginfo_is_eq: compare the reginfo structs, returns nonzero if
>> equal */
>> +int reginfo_is_eq(struct reginfo *m, struct reginfo *a)
>> +{
>> +    int i;
>> +
>> +    if (m->psw_mask != a->psw_mask || m->psw_addr != a->psw_addr) {
>> +        return 0;
>> +    }
>> +
>> +    /* Skip return address register and stack register for
>> comparison */
> 
> Is this because of ASLR? In this case, would it be possible to build a
> non-PIE binary and switch to a private stack at the beginning? This
> could be useful for the other architectures as well.

Ah, no, it's not due to ASLR ... I just experienced some crashes during 
development and saw that other targets like m68k skip the stack pointer 
here, too, so I did it the same way on s390x.
(But I finally discovered that the crashes were due to other reasons that I 
then fixed)

Looking through the code, it seems like there is already support for an 
alternate signal handler stack:

 
https://git.linaro.org/people/peter.maydell/risu.git/commit/?id=ad82a069e8d6a21842bbb265

... maybe it's working on s390x, too ... I'll give it a try.

  Thomas