[PATCH v5 2/4] armv8: mmu: teach the pagetable dumper to show explicit FAULT maps

[Ilias Apalodimas]

On Tue, 26 May 2026 at 17:10, Casey Connolly <casey.connolly at linaro.org> wrote:
>
>
>
> On 25/05/2026 13:35, Ilias Apalodimas wrote:
> > On Mon, 18 May 2026 at 12:00, Casey Connolly <casey.connolly at linaro.org> wrote:
> >>
> >> When a region is explicitly unmapped (like with
> >> mmu_change_region_attr(.... PTE_TYPE_FAULT)) the address translation
> >> still remains but won't be used since the region is marked invalid.
> >> Print these regions when we dump the pagetable to help with debugging.
> >>
> >> Signed-off-by: Casey Connolly <casey.connolly at linaro.org>
> >> ---
> >>  arch/arm/cpu/armv8/cache_v8.c | 8 ++++++--
> >>  1 file changed, 6 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/arch/arm/cpu/armv8/cache_v8.c b/arch/arm/cpu/armv8/cache_v8.c
> >> index 97a1c39ed268..2800ba57a464 100644
> >> --- a/arch/arm/cpu/armv8/cache_v8.c
> >> +++ b/arch/arm/cpu/armv8/cache_v8.c
> >> @@ -533,9 +533,9 @@ static void __pagetable_walk(u64 addr, u64 tcr, int level, pte_walker_cb_t cb, v
> >>
> >>                 if (exit)
> >>                         return;
> >>
> >> -               if (pte_type(&pte) == PTE_TYPE_FAULT)
> >> +               if (!pte)
> >>                         continue;
> >>
> >>                 attrs = pte & ALL_ATTRS;
> >>                 /* If we're currently inside a block or set of pages */
> >> @@ -572,9 +572,9 @@ static void __pagetable_walk(u64 addr, u64 tcr, int level, pte_walker_cb_t cb, v
> >>
> >>                         /* Go down a level */
> >>                         __pagetable_walk(_addr, tcr, level + 1, cb, priv);
> >>                         state[level] = WALKER_STATE_START;
> >> -               } else if (pte_type(&pte) == PTE_TYPE_BLOCK || pte_type(&pte) == PTE_TYPE_PAGE) {
> >> +               } else {
> >>                         /* We foud a block or page, start walking */
> >>                         entry_start = pte;
> >>                         state[level] = WALKER_STATE_REGION;
> >>                 }
> >> @@ -707,8 +707,12 @@ static bool pagetable_print_entry(u64 start_attrs, u64 end, int va_bits, int lev
> >>  {
> >>         u64 _addr = start_attrs & GENMASK_ULL(va_bits, PAGE_SHIFT);
> >>         int indent = va_bits < 39 ? level - 1 : level;
> >>
> >> +       /* Ignore empty blocks */
> >> +       if (!start_attrs)
> >> +               return false;
> >
> > I am not sure I am following this one. This one get the value of
> > TTBR_ELx. What are we trying to prevent? Perhaps we should move the
> > check to walk_pagetable()?
>
> Not sure what you mean about TTBR_ELx here. I think you're right that it

start_attrs is only defined in dump_pagetable(). That's only used with tbbr.

> should be moved, I think I added it because I hit a case where there was
> a table with no blocks in it (which obviously should never happen if the
> mmu code is working properly).

Exactly

>
> I can just drop it then if we see empty FAULT mappings from address 0
> with a size of 1 granule we know there's a bug in the MMU code :D

I tried without it on QEMU and didn't get any splats.

Cheers
/Ilias
>
>
> >
> > Cheers
> > /Ilias
> >> +
> >>         printf("%*s", indent * 2, "");
> >>         if (PTE_IS_TABLE(start_attrs, level))
> >>                 printf("[%#016llx]%19s", _addr, "");
> >>         else
> >>
> >> --
> >> 2.53.0
> >>
>
> --
> // Casey (she/her)
>