Re: [PATCH 1/1] target/riscv: Clear vstart_qe_zero flag

[Richard Henderson]

On 2/17/24 00:53, Daniel Henrique Barboza wrote:
> This patch is replacing mark_vs_dirty() with finalize(), that does call 
> mark_vs_dirty() and
> set start_eq_zero = true, but it's missing the start_eq_zero update for store 
> functions
> because of these ifs.
>
> We could just remove these ifs and finalize() all the time. To keep the 
> existing logic
> (i.e. not set vs_dirty for writes) I would do, in this same patch:
>
>
> >       if (!is_store) {
> >           mark_vs_dirty(s);
> >       }
> > +     s->start_eq_zero = true;
>
>
> This would make these load/stores functions different from the rest, without a 
> finalize()
> call, but given that they're already difference sine vs_dirty() is conditional 
> I guess
> it's fine.
>
>
> What do you think?

I think it's required to have stores set dirty unconditionally, before the 
operation.

Consider a store that traps on the 2nd element, leaving vstart = 2, and exiting to the 
main loop via exception.  The exception enters the kernel page fault handler.  The kernel 
may need to fault in the page for the process, and in the meantime task switch.

If vs dirty is not already set, the kernel won't know to save vector state on 
task switch.


r~