On Tue, Jun 17, 2025 at 11:49:20AM +0200, David Hildenbrand wrote:

On 16.06.25 13:58, Alistair Popple wrote:

quoted

Nothing uses PFN_DEV anymore so no need to create devmap pXd's when
mapping a PFN. Instead special mappings will be created which ensures
vm_normal_page_pXd() will not return pages which don't have an
associated page. This could change behaviour slightly on architectures
where pXd_devmap() does not imply pXd_special() as the normal page
checks would have fallen through to checking VM_PFNMAP/MIXEDMAP instead,
which in theory at least could have returned a page.

However vm_normal_page_pXd() should never have been returning pages for
pXd_devmap() entries anyway, so anything relying on that would have been
a bug.

Signed-off-by: Alistair Popple <apopple@nvidia.com>

---

Changes since v1:

  - New for v2
---
  mm/huge_memory.c | 12 ++----------
  1 file changed, 2 insertions(+), 10 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index b096240..6514e25 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c

@@ -1415,11 +1415,7 @@ static int insert_pmd(struct vm_area_struct *vma, unsigned long addr,
  		add_mm_counter(mm, mm_counter_file(fop.folio), HPAGE_PMD_NR);
  	} else {
  		entry = pmd_mkhuge(pfn_t_pmd(fop.pfn, prot));
-
-		if (pfn_t_devmap(fop.pfn))
-			entry = pmd_mkdevmap(entry);
-		else
-			entry = pmd_mkspecial(entry);
+		entry = pmd_mkspecial(entry);
  	}
  	if (write) {
  		entry = pmd_mkyoung(pmd_mkdirty(entry));

@@ -1565,11 +1561,7 @@ static void insert_pud(struct vm_area_struct *vma, unsigned long addr,
  		add_mm_counter(mm, mm_counter_file(fop.folio), HPAGE_PUD_NR);
  	} else {
  		entry = pud_mkhuge(pfn_t_pud(fop.pfn, prot));
-
-		if (pfn_t_devmap(fop.pfn))
-			entry = pud_mkdevmap(entry);
-		else
-			entry = pud_mkspecial(entry);
+		entry = pud_mkspecial(entry);
  	}
  	if (write) {
  		entry = pud_mkyoung(pud_mkdirty(entry));

Why not squash this patch into #3, and remove the pmd_special() check from
vm_normal_page_pmd() in the same go? Seems wrong to handle the PMD/PUD case
separately.

Yeah, that was mostly because "someone" (and thankyou btw, it was somewhat my
mess) changed all this while I was working on it :-) I wanted to make the rebase
fixups obvious but will squash them for v3.

But now I am confused why some pte_devmap() checks are removed in patch #3,
while others are removed in #7.

Why not split it up into (a) stop setting p*_devmap() and (b) remove
p*_devmap().

Logically makes more sense to me ... :)

Heh. You're right. For various reasons this patch series has gone through a
couple of reorderings, mainly to get rid of unused stuff early in the series but
that didn't work out due to that RISC-V bug. I needed a break from silly rebase
build errors so this was a good checkpoint.

But I've reworked things for v3 to get the ordering a bit more sensible.

-- 
Cheers,

David / dhildenb