Re: [PATCH qemu] s390x/css: fix PMCW invalid mask

[Halil Pasic <pasic@linux.ibm.com>]

On Fri, 17 Dec 2021 18:13:47 +0100
Pierre Morel <pmorel@linux.ibm.com> wrote:

> >> Previously, we required bits 5, 6 and 7 to be zero (0x07 == 0b111). But,
> >> as per the principles of operation, bit 5 is ignored in MSCH and bits 0,
> >> 1, 6 and 7 need to be zero.  
> > 
> > On a second thought, don't we have to make sure then that bit 5 is
> > ignored?
> > 
> > static void copy_pmcw_from_guest(PMCW *dest, const PMCW *src)
> > {
> >      int i;
> > 
> >      dest->intparm = be32_to_cpu(src->intparm);
> >      dest->flags = be16_to_cpu(src->flags);
> >      dest->devno = be16_to_cpu(src->devno);
> > 
> > Here we seem to grab flags as a whole, but actually we would have to
> > mask of bit 5.  
> 
> Why?
> If this bit is ignored by the machine shouldn't we just ignore it?
> Forcing it to 0 or to 1 is purely arbitrary no?

We do the masking later on:
IOInstEnding css_do_msch(SubchDev *sch, const SCHIB *orig_schib)
{
[..]
    /* Only update the program-modifiable fields. */
    schib->pmcw.intparm = schib_copy.pmcw.intparm;
    oldflags = schib->pmcw.flags;
    schib->pmcw.flags &= ~(PMCW_FLAGS_MASK_ISC | PMCW_FLAGS_MASK_ENA |
                  PMCW_FLAGS_MASK_LM | PMCW_FLAGS_MASK_MME |
                  PMCW_FLAGS_MASK_MP);
    schib->pmcw.flags |= schib_copy.pmcw.flags &
            (PMCW_FLAGS_MASK_ISC | PMCW_FLAGS_MASK_ENA |
             PMCW_FLAGS_MASK_LM | PMCW_FLAGS_MASK_MME |
             PMCW_FLAGS_MASK_MP);
[..]

I just didn't read far enough. We do that for a while now.

The PoP says that the machine shall ignore other fields
of the PMCW when an MSCH is performed. I.e. we should not update
"our" pmcw.flags bit 5 from 0 to 1 even if 1 was supplied, and
thus STSCH should keep storing the bit 5 as 0 even if there was
a MSCH with bit 5 set.

Regards,
Halil