Re: [RFC PATCH] accel/tcg: clear all TBs from a page when it is written to

[Richard Henderson]

On 8/14/24 16:09, Nicholas Piggin wrote:
> > > @@ -1107,6 +1107,9 @@ tb_invalidate_phys_page_range__locked(struct 
> > > page_collection *pages,
> > >        TranslationBlock *current_tb = retaddr ? tcg_tb_lookup(retaddr) : NULL;
> > >    #endif /* TARGET_HAS_PRECISE_SMC */
> > >    
> > > +    start &= TARGET_PAGE_MASK;
> > > +    last |= ~TARGET_PAGE_MASK;
> > > +
> > >        /* Range may not cross a page. */
> > >        tcg_debug_assert(((start ^ last) & TARGET_PAGE_MASK) == 0);
> >
> > This would definitely break SMC.
>
> They can't invalidate the instruction currently being executed?

They can.  But that's where adjusting start/last would break things, because we've lost 
track of exactly what's being modified, which changes the result of "did we modify the 
current instruction".


r~