On Mon, May 13, 2024 at 09:33:57PM -0700, Kees Cook wrote:
> In function 'page_fixed_fake_head',
> inlined from '_compound_head' at ../include/linux/page-flags.h:251:24,
> inlined from '__dump_page' at ../mm/debug.c:123:11:
> ../include/asm-generic/rwonce.h:44:26: warning: array subscript 9 is outside array bounds of 'struct page[1]' [-Warray-bounds=]
>
> (Not noted in this warning is that the code passes through page_folio()
> _Generic macro.)
>
> It doesn't like that it can see that "precise" is exactly one page, so
> looking at page[1] later is going to freak out. I suspect this may be
> "impossible" at run-time, but I'm not 100% sure. Regardless, the compiler
> can't tell.
Actually, I'm not sure that I can tell that it's impossible.
I think we just need to open-code page_folio() here so that we don't
get into the fixed fake head palaver. Something like this, although
it's only compile-tested.
diff --git a/mm/debug.c b/mm/debug.c
index e3ff3ac19fa1..47ba8b0a4872 100644
--- a/mm/debug.c
+++ b/mm/debug.c
@@ -110,19 +110,22 @@ static void __dump_page(const struct page *page)
{
struct folio *foliop, folio;
struct page precise;
+ unsigned long head;
unsigned long pfn = page_to_pfn(page);
unsigned long idx, nr_pages = 1;
int loops = 5;
again:
memcpy(&precise, page, sizeof(*page));
- foliop = page_folio(&precise);
- if (foliop == (struct folio *)&precise) {
+ head = precise.compound_head;
+ if ((head & 1) == 0) {
+ foliop = (struct folio *)&precise;
idx = 0;
if (!folio_test_large(foliop))
goto dump;
foliop = (struct folio *)page;
} else {
+ foliop = (struct folio *)(head - 1);
idx = folio_page_idx(foliop, page);
}
