On Wed, 29 Jun 2022 22:47:19 -0500
Scott Cheloha <[email protected]> wrote:
> To be perfectly clear, you are concerned about this scenario:
>
> > > + if (ci->ci_dec_deferred && newcpl < IPL_CLOCK) {
> > > + ppc_mtdec(0);
>
> /* DEC interrupt fires *here*. */
> /* We jump to decrint() and then call decr_intr(). */
>
> > > + ppc_mtdec(UINT32_MAX); /* raise DEC exception */
> > > + }
>
> I think it's possible for the DEC exception to occur in that spot.
> However, external/DEC *interrupts* are explicitly disabled, so I don't
> think that we will jump to decrint() until the next time we do
>
> ppc_intr_enable(1);
I missed the ppc_intr_disable(), which disables PSL_EE, in
macintr_splx and openpic_splx. You are correct, it can't call
decr_intr until ppc_intr_enable(1).
ppc_dflt_splx also looks good, because we don't enable PSL_EE until
we switch to macintr_splx or openpic_splx.
> > Would this be better?
> >
> > ppc_mtdec(1 >> UINT32_MAX);
> > ppc_mtdec(UINT32_MAX);
>
> I assume you meant to type
>
> ppc_mtdec(UINT32_MAX >> 1);
Yes, I meant UINT32_MAX >> 1, but you have persuaded me that the
existing ppc_mtdec(0) is correct, and no change is necessary. I
will continue running your diff with ppc_mtdec(0).
