riscv: Fix set_huge_pte_at() for NAPOT mapping

As stated by the privileged specification, we must clear a NAPOT
mapping and emit a sfence.vma before setting a new translation.

Fixes: 82a1a1f ("riscv: mm: support Svnapot in hugetlb page")
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20240117195741.1926459-2-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

Author: Alexandre Ghiti

arch/riscv/mm/hugetlbpage.c

@@ -177,13 +177,36 @@ pte_t arch_make_huge_pte(pte_t entry, unsigned int shift, vm_flags_t flags)
 	return entry;
 }
 
+static void clear_flush(struct mm_struct *mm,
+			unsigned long addr,
+			pte_t *ptep,
+			unsigned long pgsize,
+			unsigned long ncontig)
+{
+	struct vm_area_struct vma = TLB_FLUSH_VMA(mm, 0);
+	unsigned long i, saddr = addr;
+
+	for (i = 0; i < ncontig; i++, addr += pgsize, ptep++)
+		ptep_get_and_clear(mm, addr, ptep);
+
+	flush_tlb_range(&vma, saddr, addr);
+}
+
+/*
+ * When dealing with NAPOT mappings, the privileged specification indicates that
+ * "if an update needs to be made, the OS generally should first mark all of the
+ * PTEs invalid, then issue SFENCE.VMA instruction(s) covering all 4 KiB regions
+ * within the range, [...] then update the PTE(s), as described in Section
+ * 4.2.1.". That's the equivalent of the Break-Before-Make approach used by
+ * arm64.
+ */
 void set_huge_pte_at(struct mm_struct *mm,
 		     unsigned long addr,
 		     pte_t *ptep,
 		     pte_t pte,
 		     unsigned long sz)
 {
-	unsigned long hugepage_shift;
+	unsigned long hugepage_shift, pgsize;
 	int i, pte_num;
 
 	if (sz >= PGDIR_SIZE)
@@ -198,7 +221,22 @@ void set_huge_pte_at(struct mm_struct *mm,
 		hugepage_shift = PAGE_SHIFT;
 
 	pte_num = sz >> hugepage_shift;
-	for (i = 0; i < pte_num; i++, ptep++, addr += (1 << hugepage_shift))
+	pgsize = 1 << hugepage_shift;
+
+	if (!pte_present(pte)) {
+		for (i = 0; i < pte_num; i++, ptep++, addr += pgsize)
+			set_ptes(mm, addr, ptep, pte, 1);
+		return;
+	}
+
+	if (!pte_napot(pte)) {
+		set_ptes(mm, addr, ptep, pte, 1);
+		return;
+	}
+
+	clear_flush(mm, addr, ptep, pgsize, pte_num);
+
+	for (i = 0; i < pte_num; i++, ptep++, addr += pgsize)
 		set_pte_at(mm, addr, ptep, pte);
 }