s390/mm: cleanup page table definitions
Improve the encoding of the different pte types and the naming of the
page, segment table and region table bits. Due to the different pte
encoding the hugetlbfs primitives need to be adapted as well. To improve
compatability with common code make the huge ptes use the encoding of
normal ptes. The conversion between the pte and pmd encoding for a huge
pte is done with set_huge_pte_at and huge_ptep_get.
Overall the code is now easier to understand.
Reviewed-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
diff --git a/arch/s390/mm/dump_pagetables.c b/arch/s390/mm/dump_pagetables.c
index 3ad65b0..46d517c 100644
--- a/arch/s390/mm/dump_pagetables.c
+++ b/arch/s390/mm/dump_pagetables.c
@@ -53,7 +53,7 @@
seq_printf(m, "I\n");
return;
}
- seq_printf(m, "%s", pr & _PAGE_RO ? "RO " : "RW ");
+ seq_printf(m, "%s", pr & _PAGE_PROTECT ? "RO " : "RW ");
seq_printf(m, "%s", pr & _PAGE_CO ? "CO " : " ");
seq_putc(m, '\n');
}
@@ -105,12 +105,12 @@
}
/*
- * The actual page table walker functions. In order to keep the implementation
- * of print_prot() short, we only check and pass _PAGE_INVALID and _PAGE_RO
- * flags to note_page() if a region, segment or page table entry is invalid or
- * read-only.
- * After all it's just a hint that the current level being walked contains an
- * invalid or read-only entry.
+ * The actual page table walker functions. In order to keep the
+ * implementation of print_prot() short, we only check and pass
+ * _PAGE_INVALID and _PAGE_PROTECT flags to note_page() if a region,
+ * segment or page table entry is invalid or read-only.
+ * After all it's just a hint that the current level being walked
+ * contains an invalid or read-only entry.
*/
static void walk_pte_level(struct seq_file *m, struct pg_state *st,
pmd_t *pmd, unsigned long addr)
@@ -122,14 +122,14 @@
for (i = 0; i < PTRS_PER_PTE && addr < max_addr; i++) {
st->current_address = addr;
pte = pte_offset_kernel(pmd, addr);
- prot = pte_val(*pte) & (_PAGE_RO | _PAGE_INVALID);
+ prot = pte_val(*pte) & (_PAGE_PROTECT | _PAGE_INVALID);
note_page(m, st, prot, 4);
addr += PAGE_SIZE;
}
}
#ifdef CONFIG_64BIT
-#define _PMD_PROT_MASK (_SEGMENT_ENTRY_RO | _SEGMENT_ENTRY_CO)
+#define _PMD_PROT_MASK (_SEGMENT_ENTRY_PROTECT | _SEGMENT_ENTRY_CO)
#else
#define _PMD_PROT_MASK 0
#endif
diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c
index 1f5315d1..5d758db 100644
--- a/arch/s390/mm/gup.c
+++ b/arch/s390/mm/gup.c
@@ -24,7 +24,7 @@
pte_t *ptep, pte;
struct page *page;
- mask = (write ? _PAGE_RO : 0) | _PAGE_INVALID | _PAGE_SPECIAL;
+ mask = (write ? _PAGE_PROTECT : 0) | _PAGE_INVALID | _PAGE_SPECIAL;
ptep = ((pte_t *) pmd_deref(pmd)) + pte_index(addr);
do {
@@ -55,8 +55,8 @@
struct page *head, *page, *tail;
int refs;
- result = write ? 0 : _SEGMENT_ENTRY_RO;
- mask = result | _SEGMENT_ENTRY_INV;
+ result = write ? 0 : _SEGMENT_ENTRY_PROTECT;
+ mask = result | _SEGMENT_ENTRY_INVALID;
if ((pmd_val(pmd) & mask) != result)
return 0;
VM_BUG_ON(!pfn_valid(pmd_val(pmd) >> PAGE_SHIFT));
diff --git a/arch/s390/mm/hugetlbpage.c b/arch/s390/mm/hugetlbpage.c
index 121089d..b0bd0ae 100644
--- a/arch/s390/mm/hugetlbpage.c
+++ b/arch/s390/mm/hugetlbpage.c
@@ -8,21 +8,107 @@
#include <linux/mm.h>
#include <linux/hugetlb.h>
+static inline pmd_t __pte_to_pmd(pte_t pte)
+{
+ int none, prot;
+ pmd_t pmd;
+
+ /*
+ * Convert encoding pte bits pmd bits
+ * .IR.....wdtp ..R...I.....
+ * empty .10.....0000 -> ..0...1.....
+ * prot-none, clean .11.....0001 -> ..1...1.....
+ * prot-none, dirty .10.....0101 -> ..1...1.....
+ * read-only, clean .01.....0001 -> ..1...0.....
+ * read-only, dirty .01.....0101 -> ..1...0.....
+ * read-write, clean .01.....1001 -> ..0...0.....
+ * read-write, dirty .00.....1101 -> ..0...0.....
+ * Huge ptes are dirty by definition, a clean pte is made dirty
+ * by the conversion.
+ */
+ if (pte_present(pte)) {
+ pmd_val(pmd) = pte_val(pte) & PAGE_MASK;
+ if (pte_val(pte) & _PAGE_INVALID)
+ pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
+ none = (pte_val(pte) & _PAGE_PRESENT) &&
+ (pte_val(pte) & _PAGE_INVALID);
+ prot = (pte_val(pte) & _PAGE_PROTECT);
+ if (prot || none)
+ pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
+ } else
+ pmd_val(pmd) = _SEGMENT_ENTRY_INVALID;
+ return pmd;
+}
+
+static inline pte_t __pmd_to_pte(pmd_t pmd)
+{
+ pte_t pte;
+
+ /*
+ * Convert encoding pmd bits pte bits
+ * ..R...I..... .IR.....wdtp
+ * empty ..0...1..... -> .10.....0000
+ * prot-none, young ..1...1..... -> .10.....0101
+ * read-only, young ..1...0..... -> .01.....0101
+ * read-write, young ..0...0..... -> .00.....1101
+ * Huge ptes are dirty by definition
+ */
+ if (pmd_present(pmd)) {
+ pte_val(pte) = _PAGE_PRESENT | _PAGE_LARGE | _PAGE_DIRTY |
+ (pmd_val(pmd) & PAGE_MASK);
+ if (pmd_val(pmd) & _SEGMENT_ENTRY_INVALID)
+ pte_val(pte) |= _PAGE_INVALID;
+ else {
+ if (pmd_val(pmd) & _SEGMENT_ENTRY_PROTECT)
+ pte_val(pte) |= _PAGE_PROTECT;
+ else
+ pte_val(pte) |= _PAGE_WRITE;
+ }
+ } else
+ pte_val(pte) = _PAGE_INVALID;
+ return pte;
+}
void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
- pte_t *pteptr, pte_t pteval)
+ pte_t *ptep, pte_t pte)
{
- pmd_t *pmdp = (pmd_t *) pteptr;
- unsigned long mask;
+ pmd_t pmd;
+ pmd = __pte_to_pmd(pte);
if (!MACHINE_HAS_HPAGE) {
- pteptr = (pte_t *) pte_page(pteval)[1].index;
- mask = pte_val(pteval) &
- (_SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO);
- pte_val(pteval) = (_SEGMENT_ENTRY + __pa(pteptr)) | mask;
- }
+ pmd_val(pmd) &= ~_SEGMENT_ENTRY_ORIGIN;
+ pmd_val(pmd) |= pte_page(pte)[1].index;
+ } else
+ pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO;
+ *(pmd_t *) ptep = pmd;
+}
- pmd_val(*pmdp) = pte_val(pteval);
+pte_t huge_ptep_get(pte_t *ptep)
+{
+ unsigned long origin;
+ pmd_t pmd;
+
+ pmd = *(pmd_t *) ptep;
+ if (!MACHINE_HAS_HPAGE && pmd_present(pmd)) {
+ origin = pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN;
+ pmd_val(pmd) &= ~_SEGMENT_ENTRY_ORIGIN;
+ pmd_val(pmd) |= *(unsigned long *) origin;
+ }
+ return __pmd_to_pte(pmd);
+}
+
+pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep)
+{
+ pmd_t *pmdp = (pmd_t *) ptep;
+ pte_t pte = huge_ptep_get(ptep);
+
+ if (MACHINE_HAS_IDTE)
+ __pmd_idte(addr, pmdp);
+ else
+ __pmd_csp(pmdp);
+ pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
+ return pte;
}
int arch_prepare_hugepage(struct page *page)
@@ -58,7 +144,7 @@
ptep = (pte_t *) page[1].index;
if (!ptep)
return;
- clear_table((unsigned long *) ptep, _PAGE_TYPE_EMPTY,
+ clear_table((unsigned long *) ptep, _PAGE_INVALID,
PTRS_PER_PTE * sizeof(pte_t));
page_table_free(&init_mm, (unsigned long *) ptep);
page[1].index = 0;
diff --git a/arch/s390/mm/pageattr.c b/arch/s390/mm/pageattr.c
index 80adfbf..9903974 100644
--- a/arch/s390/mm/pageattr.c
+++ b/arch/s390/mm/pageattr.c
@@ -118,7 +118,7 @@
pte = pte_offset_kernel(pmd, address);
if (!enable) {
__ptep_ipte(address, pte);
- pte_val(*pte) = _PAGE_TYPE_EMPTY;
+ pte_val(*pte) = _PAGE_INVALID;
continue;
}
pte_val(*pte) = __pa(address);
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index a8154a1..b9d35d6 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -161,7 +161,7 @@
struct gmap_rmap *rmap;
struct page *page;
- if (*table & _SEGMENT_ENTRY_INV)
+ if (*table & _SEGMENT_ENTRY_INVALID)
return 0;
page = pfn_to_page(*table >> PAGE_SHIFT);
mp = (struct gmap_pgtable *) page->index;
@@ -172,7 +172,7 @@
kfree(rmap);
break;
}
- *table = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | mp->vmaddr;
+ *table = mp->vmaddr | _SEGMENT_ENTRY_INVALID | _SEGMENT_ENTRY_PROTECT;
return 1;
}
@@ -258,7 +258,7 @@
return -ENOMEM;
new = (unsigned long *) page_to_phys(page);
crst_table_init(new, init);
- if (*table & _REGION_ENTRY_INV) {
+ if (*table & _REGION_ENTRY_INVALID) {
list_add(&page->lru, &gmap->crst_list);
*table = (unsigned long) new | _REGION_ENTRY_LENGTH |
(*table & _REGION_ENTRY_TYPE_MASK);
@@ -292,22 +292,22 @@
for (off = 0; off < len; off += PMD_SIZE) {
/* Walk the guest addr space page table */
table = gmap->table + (((to + off) >> 53) & 0x7ff);
- if (*table & _REGION_ENTRY_INV)
+ if (*table & _REGION_ENTRY_INVALID)
goto out;
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + (((to + off) >> 42) & 0x7ff);
- if (*table & _REGION_ENTRY_INV)
+ if (*table & _REGION_ENTRY_INVALID)
goto out;
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + (((to + off) >> 31) & 0x7ff);
- if (*table & _REGION_ENTRY_INV)
+ if (*table & _REGION_ENTRY_INVALID)
goto out;
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + (((to + off) >> 20) & 0x7ff);
/* Clear segment table entry in guest address space. */
flush |= gmap_unlink_segment(gmap, table);
- *table = _SEGMENT_ENTRY_INV;
+ *table = _SEGMENT_ENTRY_INVALID;
}
out:
spin_unlock(&gmap->mm->page_table_lock);
@@ -345,17 +345,17 @@
for (off = 0; off < len; off += PMD_SIZE) {
/* Walk the gmap address space page table */
table = gmap->table + (((to + off) >> 53) & 0x7ff);
- if ((*table & _REGION_ENTRY_INV) &&
+ if ((*table & _REGION_ENTRY_INVALID) &&
gmap_alloc_table(gmap, table, _REGION2_ENTRY_EMPTY))
goto out_unmap;
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + (((to + off) >> 42) & 0x7ff);
- if ((*table & _REGION_ENTRY_INV) &&
+ if ((*table & _REGION_ENTRY_INVALID) &&
gmap_alloc_table(gmap, table, _REGION3_ENTRY_EMPTY))
goto out_unmap;
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + (((to + off) >> 31) & 0x7ff);
- if ((*table & _REGION_ENTRY_INV) &&
+ if ((*table & _REGION_ENTRY_INVALID) &&
gmap_alloc_table(gmap, table, _SEGMENT_ENTRY_EMPTY))
goto out_unmap;
table = (unsigned long *) (*table & _REGION_ENTRY_ORIGIN);
@@ -363,7 +363,8 @@
/* Store 'from' address in an invalid segment table entry. */
flush |= gmap_unlink_segment(gmap, table);
- *table = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | (from + off);
+ *table = (from + off) | (_SEGMENT_ENTRY_INVALID |
+ _SEGMENT_ENTRY_PROTECT);
}
spin_unlock(&gmap->mm->page_table_lock);
up_read(&gmap->mm->mmap_sem);
@@ -384,15 +385,15 @@
unsigned long *table;
table = gmap->table + ((address >> 53) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV))
+ if (unlikely(*table & _REGION_ENTRY_INVALID))
return ERR_PTR(-EFAULT);
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 42) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV))
+ if (unlikely(*table & _REGION_ENTRY_INVALID))
return ERR_PTR(-EFAULT);
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 31) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV))
+ if (unlikely(*table & _REGION_ENTRY_INVALID))
return ERR_PTR(-EFAULT);
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 20) & 0x7ff);
@@ -422,11 +423,11 @@
return PTR_ERR(segment_ptr);
/* Convert the gmap address to an mm address. */
segment = *segment_ptr;
- if (!(segment & _SEGMENT_ENTRY_INV)) {
+ if (!(segment & _SEGMENT_ENTRY_INVALID)) {
page = pfn_to_page(segment >> PAGE_SHIFT);
mp = (struct gmap_pgtable *) page->index;
return mp->vmaddr | (address & ~PMD_MASK);
- } else if (segment & _SEGMENT_ENTRY_RO) {
+ } else if (segment & _SEGMENT_ENTRY_PROTECT) {
vmaddr = segment & _SEGMENT_ENTRY_ORIGIN;
return vmaddr | (address & ~PMD_MASK);
}
@@ -517,8 +518,8 @@
page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
mp = (struct gmap_pgtable *) page->index;
list_for_each_entry_safe(rmap, next, &mp->mapper, list) {
- *rmap->entry =
- _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | mp->vmaddr;
+ *rmap->entry = mp->vmaddr | (_SEGMENT_ENTRY_INVALID |
+ _SEGMENT_ENTRY_PROTECT);
list_del(&rmap->list);
kfree(rmap);
flush = 1;
@@ -545,13 +546,13 @@
/* Convert the gmap address to an mm address. */
while (1) {
segment = *segment_ptr;
- if (!(segment & _SEGMENT_ENTRY_INV)) {
+ if (!(segment & _SEGMENT_ENTRY_INVALID)) {
/* Page table is present */
page = pfn_to_page(segment >> PAGE_SHIFT);
mp = (struct gmap_pgtable *) page->index;
return mp->vmaddr | (address & ~PMD_MASK);
}
- if (!(segment & _SEGMENT_ENTRY_RO))
+ if (!(segment & _SEGMENT_ENTRY_PROTECT))
/* Nothing mapped in the gmap address space. */
break;
rc = gmap_connect_pgtable(address, segment, segment_ptr, gmap);
@@ -586,25 +587,25 @@
while (address < to) {
/* Walk the gmap address space page table */
table = gmap->table + ((address >> 53) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV)) {
+ if (unlikely(*table & _REGION_ENTRY_INVALID)) {
address = (address + PMD_SIZE) & PMD_MASK;
continue;
}
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 42) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV)) {
+ if (unlikely(*table & _REGION_ENTRY_INVALID)) {
address = (address + PMD_SIZE) & PMD_MASK;
continue;
}
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 31) & 0x7ff);
- if (unlikely(*table & _REGION_ENTRY_INV)) {
+ if (unlikely(*table & _REGION_ENTRY_INVALID)) {
address = (address + PMD_SIZE) & PMD_MASK;
continue;
}
table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
table = table + ((address >> 20) & 0x7ff);
- if (unlikely(*table & _SEGMENT_ENTRY_INV)) {
+ if (unlikely(*table & _SEGMENT_ENTRY_INVALID)) {
address = (address + PMD_SIZE) & PMD_MASK;
continue;
}
@@ -687,7 +688,7 @@
continue;
/* Set notification bit in the pgste of the pte */
entry = *ptep;
- if ((pte_val(entry) & (_PAGE_INVALID | _PAGE_RO)) == 0) {
+ if ((pte_val(entry) & (_PAGE_INVALID | _PAGE_PROTECT)) == 0) {
pgste = pgste_get_lock(ptep);
pgste_val(pgste) |= PGSTE_IN_BIT;
pgste_set_unlock(ptep, pgste);
@@ -752,7 +753,7 @@
page->index = (unsigned long) mp;
atomic_set(&page->_mapcount, 3);
table = (unsigned long *) page_to_phys(page);
- clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/2);
+ clear_table(table, _PAGE_INVALID, PAGE_SIZE/2);
clear_table(table + PTRS_PER_PTE, 0, PAGE_SIZE/2);
return table;
}
@@ -878,7 +879,7 @@
pgtable_page_ctor(page);
atomic_set(&page->_mapcount, 1);
table = (unsigned long *) page_to_phys(page);
- clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+ clear_table(table, _PAGE_INVALID, PAGE_SIZE);
spin_lock_bh(&mm->context.list_lock);
list_add(&page->lru, &mm->context.pgtable_list);
} else {
@@ -1198,9 +1199,9 @@
list_del(lh);
}
ptep = (pte_t *) pgtable;
- pte_val(*ptep) = _PAGE_TYPE_EMPTY;
+ pte_val(*ptep) = _PAGE_INVALID;
ptep++;
- pte_val(*ptep) = _PAGE_TYPE_EMPTY;
+ pte_val(*ptep) = _PAGE_INVALID;
return pgtable;
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c
index 8b268fc..e1299d4 100644
--- a/arch/s390/mm/vmem.c
+++ b/arch/s390/mm/vmem.c
@@ -69,7 +69,7 @@
pte = alloc_bootmem(PTRS_PER_PTE * sizeof(pte_t));
if (!pte)
return NULL;
- clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY,
+ clear_table((unsigned long *) pte, _PAGE_INVALID,
PTRS_PER_PTE * sizeof(pte_t));
return pte;
}
@@ -101,7 +101,7 @@
!(address & ~PUD_MASK) && (address + PUD_SIZE <= end)) {
pud_val(*pu_dir) = __pa(address) |
_REGION_ENTRY_TYPE_R3 | _REGION3_ENTRY_LARGE |
- (ro ? _REGION_ENTRY_RO : 0);
+ (ro ? _REGION_ENTRY_PROTECT : 0);
address += PUD_SIZE;
continue;
}
@@ -118,7 +118,7 @@
!(address & ~PMD_MASK) && (address + PMD_SIZE <= end)) {
pmd_val(*pm_dir) = __pa(address) |
_SEGMENT_ENTRY | _SEGMENT_ENTRY_LARGE |
- (ro ? _SEGMENT_ENTRY_RO : 0);
+ (ro ? _SEGMENT_ENTRY_PROTECT : 0);
address += PMD_SIZE;
continue;
}
@@ -131,7 +131,8 @@
}
pt_dir = pte_offset_kernel(pm_dir, address);
- pte_val(*pt_dir) = __pa(address) | (ro ? _PAGE_RO : 0);
+ pte_val(*pt_dir) = __pa(address) |
+ pgprot_val(ro ? PAGE_KERNEL_RO : PAGE_KERNEL);
address += PAGE_SIZE;
}
ret = 0;
@@ -154,7 +155,7 @@
pte_t *pt_dir;
pte_t pte;
- pte_val(pte) = _PAGE_TYPE_EMPTY;
+ pte_val(pte) = _PAGE_INVALID;
while (address < end) {
pg_dir = pgd_offset_k(address);
if (pgd_none(*pg_dir)) {
@@ -255,7 +256,8 @@
new_page =__pa(vmem_alloc_pages(0));
if (!new_page)
goto out;
- pte_val(*pt_dir) = __pa(new_page);
+ pte_val(*pt_dir) =
+ __pa(new_page) | pgprot_val(PAGE_KERNEL);
}
address += PAGE_SIZE;
}