Re: 10.0 crashes on leaving X

On Wed, 24 Apr 2024 at 21:02, matthew green <[email protected]> wrote:
>
> interesting.  the panic only happens on the sun4m becuase the
> sun4c version of this function continues instead of panicking.
>
> ---
> void
> pmap_protect4_4c(struct pmap *pm, vaddr_t sva, vaddr_t eva, vm_prot_t prot)
> ...
>         for (va = sva; va < eva;) {
> ...
>                 if (rp->rg_nsegmap == 0) {
>                         va = nva;
>                         continue;
> --- vs ---
> pmap_page_protect4m(struct vm_page *pg, vm_prot_t prot)
> ...
>         while (pv != NULL) {
> ...
>                 rp = &pm->pm_regmap[vr];
>                 if (rp->rg_nsegmap == 0)
>                         panic("pmap_remove_all: empty vreg");
> ---
>
> i wonder if there is some race condition happening.  don't
> worry, we believe you're seeing this :)

Worth adding a log line to sun4c, for people to give a spin in -current? :)

David