Skip to content

Commit

Permalink
s390/mm: add support for 2GB hugepages
Browse files Browse the repository at this point in the history 
This adds support for 2GB hugetlbfs pages on s390.

Reviewed-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
  • Loading branch information
@gerald-schaefer
gerald-schaefer authored and Martin Schwidefsky committed Jul 6, 2016
1 parent 46210c4 commit d08de8e
Show file tree
Hide file tree
Showing 8 changed files with 233 additions and 64 deletions.
5 changes: 4 additions & 1 deletion arch/s390/include/asm/hugetlb.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,10 @@ static inline int prepare_hugepage_range(struct file *file,
static inline void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
pte_t *ptep)
{
pte_val(*ptep) = _SEGMENT_ENTRY_EMPTY;
if ((pte_val(*ptep) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
pte_val(*ptep) = _REGION3_ENTRY_EMPTY;
else
pte_val(*ptep) = _SEGMENT_ENTRY_EMPTY;
}

static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
Expand Down
1 change: 1 addition & 0 deletions arch/s390/include/asm/page.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
#define HPAGE_SIZE (1UL << HPAGE_SHIFT)
#define HPAGE_MASK (~(HPAGE_SIZE - 1))
#define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT)
#define HUGE_MAX_HSTATE 2

#define ARCH_HAS_SETCLEAR_HUGE_PTE
#define ARCH_HAS_HUGE_PTE_TYPE
Expand Down
71 changes: 49 additions & 22 deletions arch/s390/include/asm/pgtable.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -306,6 +306,9 @@ static inline int is_module_addr(void *addr)
#define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
#endif

#define _REGION_ENTRY_BITS 0xfffffffffffff227UL
#define _REGION_ENTRY_BITS_LARGE 0xffffffff8000fe27UL

/* Bits in the segment table entry */
#define _SEGMENT_ENTRY_BITS 0xfffffffffffffe33UL
#define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff0ff33UL
Expand Down Expand Up @@ -573,7 +576,7 @@ static inline int pud_none(pud_t pud)
{
if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
return 0;
return (pud_val(pud) & _REGION_ENTRY_INVALID) != 0UL;
return pud_val(pud) == _REGION3_ENTRY_EMPTY;
}

static inline int pud_large(pud_t pud)
Expand All @@ -593,17 +596,25 @@ static inline unsigned long pud_pfn(pud_t pud)
return (pud_val(pud) & origin_mask) >> PAGE_SHIFT;
}

static inline int pmd_large(pmd_t pmd)
{
return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
}

static inline int pmd_bad(pmd_t pmd)
{
if (pmd_large(pmd))
return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
}

static inline int pud_bad(pud_t pud)
{
/*
* With dynamic page table levels the pud can be a region table
* entry or a segment table entry. Check for the bit that are
* invalid for either table entry.
*/
unsigned long mask =
~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
return (pud_val(pud) & mask) != 0;
if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
return pmd_bad(__pmd(pud_val(pud)));
if (pud_large(pud))
return (pud_val(pud) & ~_REGION_ENTRY_BITS_LARGE) != 0;
return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
}

static inline int pmd_present(pmd_t pmd)
Expand All @@ -616,11 +627,6 @@ static inline int pmd_none(pmd_t pmd)
return pmd_val(pmd) == _SEGMENT_ENTRY_INVALID;
}

static inline int pmd_large(pmd_t pmd)
{
return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
}

static inline unsigned long pmd_pfn(pmd_t pmd)
{
unsigned long origin_mask;
Expand All @@ -631,13 +637,6 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
}

static inline int pmd_bad(pmd_t pmd)
{
if (pmd_large(pmd))
return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
}

#define __HAVE_ARCH_PMD_WRITE
static inline int pmd_write(pmd_t pmd)
{
Expand Down Expand Up @@ -1081,6 +1080,7 @@ static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
#define pte_page(x) pfn_to_page(pte_pfn(x))

#define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
#define pud_page(pud) pfn_to_page(pud_pfn(pud))

/* Find an entry in the lowest level page table.. */
#define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
Expand Down Expand Up @@ -1238,6 +1238,19 @@ static inline void __pmdp_idte(unsigned long address, pmd_t *pmdp)
: "cc" );
}

static inline void __pudp_idte(unsigned long address, pud_t *pudp)
{
unsigned long r3o;

r3o = (unsigned long) pudp - pud_index(address) * sizeof(pud_t);
r3o |= _ASCE_TYPE_REGION3;
asm volatile(
" .insn rrf,0xb98e0000,%2,%3,0,0"
: "=m" (*pudp)
: "m" (*pudp), "a" (r3o), "a" ((address & PUD_MASK))
: "cc");
}

static inline void __pmdp_idte_local(unsigned long address, pmd_t *pmdp)
{
unsigned long sto;
Expand All @@ -1250,8 +1263,22 @@ static inline void __pmdp_idte_local(unsigned long address, pmd_t *pmdp)
: "cc" );
}

static inline void __pudp_idte_local(unsigned long address, pud_t *pudp)
{
unsigned long r3o;

r3o = (unsigned long) pudp - pud_index(address) * sizeof(pud_t);
r3o |= _ASCE_TYPE_REGION3;
asm volatile(
" .insn rrf,0xb98e0000,%2,%3,0,1"
: "=m" (*pudp)
: "m" (*pudp), "a" (r3o), "a" ((address & PUD_MASK))
: "cc");
}

pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);

#ifdef CONFIG_TRANSPARENT_HUGEPAGE

Expand Down
3 changes: 3 additions & 0 deletions arch/s390/mm/gmap.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -430,6 +430,9 @@ int __gmap_link(struct gmap *gmap, unsigned long gaddr, unsigned long vmaddr)
VM_BUG_ON(pgd_none(*pgd));
pud = pud_offset(pgd, vmaddr);
VM_BUG_ON(pud_none(*pud));
/* large puds cannot yet be handled */
if (pud_large(*pud))
return -EFAULT;
pmd = pmd_offset(pud, vmaddr);
VM_BUG_ON(pmd_none(*pmd));
/* large pmds cannot yet be handled */
Expand Down
45 changes: 44 additions & 1 deletion arch/s390/mm/gup.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -128,6 +128,44 @@ static inline int gup_pmd_range(pud_t *pudp, pud_t pud, unsigned long addr,
return 1;
}

static int gup_huge_pud(pud_t *pudp, pud_t pud, unsigned long addr,
unsigned long end, int write, struct page **pages, int *nr)
{
struct page *head, *page;
unsigned long mask;
int refs;

mask = (write ? _REGION_ENTRY_PROTECT : 0) | _REGION_ENTRY_INVALID;
if ((pud_val(pud) & mask) != 0)
return 0;
VM_BUG_ON(!pfn_valid(pud_pfn(pud)));

refs = 0;
head = pud_page(pud);
page = head + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
do {
VM_BUG_ON_PAGE(compound_head(page) != head, page);
pages[*nr] = page;
(*nr)++;
page++;
refs++;
} while (addr += PAGE_SIZE, addr != end);

if (!page_cache_add_speculative(head, refs)) {
*nr -= refs;
return 0;
}

if (unlikely(pud_val(pud) != pud_val(*pudp))) {
*nr -= refs;
while (refs--)
put_page(head);
return 0;
}

return 1;
}

static inline int gup_pud_range(pgd_t *pgdp, pgd_t pgd, unsigned long addr,
unsigned long end, int write, struct page **pages, int *nr)
{
Expand All @@ -144,7 +182,12 @@ static inline int gup_pud_range(pgd_t *pgdp, pgd_t pgd, unsigned long addr,
next = pud_addr_end(addr, end);
if (pud_none(pud))
return 0;
if (!gup_pmd_range(pudp, pud, addr, next, write, pages, nr))
if (unlikely(pud_large(pud))) {
if (!gup_huge_pud(pudp, pud, addr, next, write, pages,
nr))
return 0;
} else if (!gup_pmd_range(pudp, pud, addr, next, write, pages,
nr))
return 0;
} while (pudp++, addr = next, addr != end);

Expand Down
Loading

0 comments on commit d08de8e

Please sign in to comment.