[POWERPC] Add ability to 4K kernel to hash in 64K pages

This adds the ability for a kernel compiled with 4K page size
to have special slices containing 64K pages and hash the right type
of hash PTEs.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>

4 files changed, 46 insertions, 16 deletions

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 5226f701634e..ecd459dd1baf 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -546,9 +546,15 @@ config NODES_SPAN_OTHER_NODES
         def_bool y
         depends on NEED_MULTIPLE_NODES
 
+config PPC_HAS_HASH_64K
+        bool
+        depends on PPC64
+        default n
+
 config PPC_64K_PAGES
         bool "64k page size"
         depends on PPC64
+        select PPC_HAS_HASH_64K
         help
           This option changes the kernel logical page size to 64k. On machines
           without processor support for 64k pages, the kernel will simulate
diff --git a/arch/powerpc/mm/hash_low_64.S b/arch/powerpc/mm/hash_low_64.S
index e64ce3eec36e..4762ff7c14df 100644
--- a/arch/powerpc/mm/hash_low_64.S
+++ b/arch/powerpc/mm/hash_low_64.S
@@ -615,6 +615,9 @@ htab_pte_insert_failure:
         li      r3,-1
         b       htab_bail
 
+#endif /* CONFIG_PPC_64K_PAGES */
+
+#ifdef CONFIG_PPC_HAS_HASH_64K
 
 /*****************************************************************************
  *                                                                           *
@@ -870,7 +873,7 @@ ht64_pte_insert_failure:
         b       ht64_bail
 
 
-#endif /* CONFIG_PPC_64K_PAGES */
+#endif /* CONFIG_PPC_HAS_HASH_64K */
 
 
 /*****************************************************************************
diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
index 5610ffb14211..028ba4ed03d2 100644
--- a/arch/powerpc/mm/hash_utils_64.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -420,7 +420,7 @@ static void __init htab_finish_init(void)
         extern unsigned int *htab_call_hpte_remove;
         extern unsigned int *htab_call_hpte_updatepp;
 
-#ifdef CONFIG_PPC_64K_PAGES
+#ifdef CONFIG_PPC_HAS_HASH_64K
         extern unsigned int *ht64_call_hpte_insert1;
         extern unsigned int *ht64_call_hpte_insert2;
         extern unsigned int *ht64_call_hpte_remove;
@@ -648,7 +648,11 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
                         return 1;
                 }
                 vsid = get_vsid(mm->context.id, ea);
+#ifdef CONFIG_PPC_MM_SLICES
+                psize = get_slice_psize(mm, ea);
+#else
                 psize = mm->context.user_psize;
+#endif
                 break;
         case VMALLOC_REGION_ID:
                 mm = &init_mm;
@@ -678,13 +682,21 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 
 #ifdef CONFIG_HUGETLB_PAGE
         /* Handle hugepage regions */
-        if (HPAGE_SHIFT &&
-            unlikely(get_slice_psize(mm, ea) == mmu_huge_psize)) {
+        if (HPAGE_SHIFT && psize == mmu_huge_psize) {
                 DBG_LOW(" -> huge page !\n");
                 return hash_huge_page(mm, access, ea, vsid, local, trap);
         }
 #endif /* CONFIG_HUGETLB_PAGE */
 
+#ifndef CONFIG_PPC_64K_PAGES
+        /* If we use 4K pages and our psize is not 4K, then we are hitting
+         * a special driver mapping, we need to align the address before
+         * we fetch the PTE
+         */
+        if (psize != MMU_PAGE_4K)
+                ea &= ~((1ul << mmu_psize_defs[psize].shift) - 1);
+#endif /* CONFIG_PPC_64K_PAGES */
+
         /* Get PTE and page size from page tables */
         ptep = find_linux_pte(pgdir, ea);
         if (ptep == NULL || !pte_present(*ptep)) {
@@ -707,9 +719,7 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
         }
 
         /* Do actual hashing */
-#ifndef CONFIG_PPC_64K_PAGES
-        rc = __hash_page_4K(ea, access, vsid, ptep, trap, local);
-#else
+#ifdef CONFIG_PPC_64K_PAGES
         /* If _PAGE_4K_PFN is set, make sure this is a 4k segment */
         if (pte_val(*ptep) & _PAGE_4K_PFN) {
                 demote_segment_4k(mm, ea);
@@ -751,12 +761,14 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
                         mmu_psize_defs[mmu_vmalloc_psize].sllp;
                 slb_flush_and_rebolt();
         }
+#endif /* CONFIG_PPC_64K_PAGES */
 
+#ifdef CONFIG_PPC_HAS_HASH_64K
         if (psize == MMU_PAGE_64K)
                 rc = __hash_page_64K(ea, access, vsid, ptep, trap, local);
         else
+#endif /* CONFIG_PPC_HAS_HASH_64K */
                 rc = __hash_page_4K(ea, access, vsid, ptep, trap, local);
-#endif /* CONFIG_PPC_64K_PAGES */
 
 #ifndef CONFIG_PPC_64K_PAGES
         DBG_LOW(" o-pte: %016lx\n", pte_val(*ptep));
@@ -812,19 +824,22 @@ void hash_preload(struct mm_struct *mm, unsigned long ea,
         /* Get VSID */
         vsid = get_vsid(mm->context.id, ea);
 
-        /* Hash it in */
+        /* Hash doesn't like irqs */
         local_irq_save(flags);
+
+        /* Is that local to this CPU ? */
         mask = cpumask_of_cpu(smp_processor_id());
         if (cpus_equal(mm->cpu_vm_mask, mask))
                 local = 1;
-#ifndef CONFIG_PPC_64K_PAGES
-        __hash_page_4K(ea, access, vsid, ptep, trap, local);
-#else
+
+        /* Hash it in */
+#ifdef CONFIG_PPC_HAS_HASH_64K
         if (mm->context.user_psize == MMU_PAGE_64K)
                 __hash_page_64K(ea, access, vsid, ptep, trap, local);
         else
-                __hash_page_4K(ea, access, vsid, ptep, trap, local);
 #endif /* CONFIG_PPC_64K_PAGES */
+                __hash_page_4K(ea, access, vsid, ptep, trap, local);
+
         local_irq_restore(flags);
 }
 
diff --git a/arch/powerpc/mm/tlb_64.c b/arch/powerpc/mm/tlb_64.c
index fd8d08c325eb..2bfc4d7e1aa2 100644
--- a/arch/powerpc/mm/tlb_64.c
+++ b/arch/powerpc/mm/tlb_64.c
@@ -143,16 +143,22 @@ void hpte_need_flush(struct mm_struct *mm, unsigned long addr,
          */
         addr &= PAGE_MASK;
 
-        /* Get page size (maybe move back to caller) */
+        /* Get page size (maybe move back to caller).
+         *
+         * NOTE: when using special 64K mappings in 4K environment like
+         * for SPEs, we obtain the page size from the slice, which thus
+         * must still exist (and thus the VMA not reused) at the time
+         * of this call
+         */
         if (huge) {
 #ifdef CONFIG_HUGETLB_PAGE
                 psize = mmu_huge_psize;
 #else
                 BUG();
-                psize = pte_pagesize_index(pte); /* shutup gcc */
+                psize = pte_pagesize_index(mm, addr, pte); /* shutup gcc */
 #endif
         } else
-                psize = pte_pagesize_index(pte);
+                psize = pte_pagesize_index(mm, addr, pte);
 
         /* Build full vaddr */
         if (!is_kernel_addr(addr)) {