blob: 3bcf269f8f55470097ac56680685321bf13e62ba [file] [log] [blame]
/* SPDX-License-Identifier: GPL-2.0 */
#define H_PTE_INDEX_SIZE 8
#define H_PMD_INDEX_SIZE 10
#define H_PUD_INDEX_SIZE 7
#define H_PGD_INDEX_SIZE 8
* 64k aligned address free up few of the lower bits of RPN for us
* We steal that here. For more deatils look at pte_pfn/pfn_pte()
#define H_PAGE_COMBO _RPAGE_RPN0 /* this is a combo 4k page */
#define H_PAGE_4K_PFN _RPAGE_RPN1 /* PFN is for a single 4k page */
#define H_PAGE_BUSY _RPAGE_RPN44 /* software: PTE & hash are busy */
#define H_PAGE_HASHPTE _RPAGE_RPN43 /* PTE has associated HPTE */
* We need to differentiate between explicit huge page and THP huge
* page, since THP huge page also need to track real subpage details
/* PTE flags to conserve for HPTE identification */
* we support 16 fragments per PTE page of 64K size.
#define H_PTE_FRAG_NR 16
* We use a 2K PTE page fragment and another 2K for storing
* real_pte_t hash index
#ifndef __ASSEMBLY__
#include <asm/errno.h>
* With 64K pages on hash table, we have a special PTE format that
* uses a second "half" of the page table to encode sub-page information
* in order to deal with 64K made of 4K HW pages. Thus we override the
* generic accessors and iterators here
#define __real_pte __real_pte
static inline real_pte_t __real_pte(pte_t pte, pte_t *ptep, int offset)
real_pte_t rpte;
unsigned long *hidxp;
rpte.pte = pte;
* Ensure that we do not read the hidx before we read the PTE. Because
* the writer side is expected to finish writing the hidx first followed
* by the PTE, by using smp_wmb(). pte_set_hash_slot() ensures that.
hidxp = (unsigned long *)(ptep + offset);
rpte.hidx = *hidxp;
return rpte;
* shift the hidx representation by one-modulo-0xf; i.e hidx 0 is respresented
* as 1, 1 as 2,... , and 0xf as 0. This convention lets us represent a
* invalid hidx 0xf with a 0x0 bit value. PTEs are anyway zero'd when
* allocated. We dont have to zero them gain; thus save on the initialization.
#define HIDX_UNSHIFT_BY_ONE(x) ((x + 0xfUL) & 0xfUL) /* shift backward by one */
#define HIDX_SHIFT_BY_ONE(x) ((x + 0x1UL) & 0xfUL) /* shift forward by one */
#define HIDX_BITS(x, index) (x << (index << 2))
#define BITS_TO_HIDX(x, index) ((x >> (index << 2)) & 0xfUL)
static inline unsigned long __rpte_to_hidx(real_pte_t rpte, unsigned long index)
return HIDX_UNSHIFT_BY_ONE(BITS_TO_HIDX(rpte.hidx, index));
* Commit the hidx and return PTE bits that needs to be modified. The caller is
* expected to modify the PTE bits accordingly and commit the PTE to memory.
static inline unsigned long pte_set_hidx(pte_t *ptep, real_pte_t rpte,
unsigned int subpg_index,
unsigned long hidx, int offset)
unsigned long *hidxp = (unsigned long *)(ptep + offset);
rpte.hidx &= ~HIDX_BITS(0xfUL, subpg_index);
*hidxp = rpte.hidx | HIDX_BITS(HIDX_SHIFT_BY_ONE(hidx), subpg_index);
* Anyone reading PTE must ensure hidx bits are read after reading the
* PTE by using the read-side barrier smp_rmb(). __real_pte() can be
* used for that.
/* No PTE bits to be modified, return 0x0UL */
return 0x0UL;
#define __rpte_to_pte(r) ((r).pte)
extern bool __rpte_sub_valid(real_pte_t rpte, unsigned long index);
* Trick: we set __end to va + 64k, which happens works for
* a 16M page as well as we want only one iteration
#define pte_iterate_hashed_subpages(rpte, psize, vpn, index, shift) \
do { \
unsigned long __end = vpn + (1UL << (PAGE_SHIFT - VPN_SHIFT)); \
unsigned __split = (psize == MMU_PAGE_4K || \
psize == MMU_PAGE_64K_AP); \
shift = mmu_psize_defs[psize].shift; \
for (index = 0; vpn < __end; index++, \
vpn += (1L << (shift - VPN_SHIFT))) { \
if (!__split || __rpte_sub_valid(rpte, index)) \
do {
#define pte_iterate_hashed_end() } while(0); } } while(0)
#define pte_pagesize_index(mm, addr, pte) \
extern int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
unsigned long pfn, unsigned long size, pgprot_t);
static inline int hash__remap_4k_pfn(struct vm_area_struct *vma, unsigned long addr,
unsigned long pfn, pgprot_t prot)
if (pfn > (PTE_RPN_MASK >> PAGE_SHIFT)) {
WARN(1, "remap_4k_pfn called with wrong pfn value\n");
return -EINVAL;
return remap_pfn_range(vma, addr, pfn, PAGE_SIZE,
__pgprot(pgprot_val(prot) | H_PAGE_4K_PFN));
#define H_PMD_TABLE_SIZE ((sizeof(pmd_t) << PMD_INDEX_SIZE) + \
(sizeof(unsigned long) << PMD_INDEX_SIZE))
#define H_PMD_TABLE_SIZE (sizeof(pmd_t) << PMD_INDEX_SIZE)
#define H_PUD_TABLE_SIZE ((sizeof(pud_t) << PUD_INDEX_SIZE) + \
(sizeof(unsigned long) << PUD_INDEX_SIZE))
#define H_PUD_TABLE_SIZE (sizeof(pud_t) << PUD_INDEX_SIZE)
#define H_PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
static inline char *get_hpte_slot_array(pmd_t *pmdp)
* The hpte hindex is stored in the pgtable whose address is in the
* second half of the PMD
* Order this load with the test for pmd_trans_huge in the caller
return *(char **)(pmdp + PTRS_PER_PMD);
* The linux hugepage PMD now include the pmd entries followed by the address
* to the stashed pgtable_t. The stashed pgtable_t contains the hpte bits.
* [ 000 | 1 bit secondary | 3 bit hidx | 1 bit valid]. We use one byte per
* each HPTE entry. With 16MB hugepage and 64K HPTE we need 256 entries and
* with 4K HPTE we need 4096 entries. Both will fit in a 4K pgtable_t.
* The top three bits are intentionally left as zero. This memory location
* are also used as normal page PTE pointers. So if we have any pointers
* left around while we collapse a hugepage, we need to make sure
* _PAGE_PRESENT bit of that is zero when we look at them
static inline unsigned int hpte_valid(unsigned char *hpte_slot_array, int index)
return hpte_slot_array[index] & 0x1;
static inline unsigned int hpte_hash_index(unsigned char *hpte_slot_array,
int index)
return hpte_slot_array[index] >> 1;
static inline void mark_hpte_slot_valid(unsigned char *hpte_slot_array,
unsigned int index, unsigned int hidx)
hpte_slot_array[index] = (hidx << 1) | 0x1;
* For core kernel code by design pmd_trans_huge is never run on any hugetlbfs
* page. The hugetlbfs page table walking and mangling paths are totally
* separated form the core VM paths and they're differentiated by
* VM_HUGETLB being set on vm_flags well before any pmd_trans_huge could run.
* pmd_trans_huge() is defined as false at build time if
* CONFIG_TRANSPARENT_HUGEPAGE=n to optimize away code blocks at build
* time in such case.
* For ppc64 we need to differntiate from explicit hugepages from THP, because
* for THP we also track the subpage details at the pmd level. We don't do
* that for explicit huge pages.
static inline int hash__pmd_trans_huge(pmd_t pmd)
return !!((pmd_val(pmd) & (_PAGE_PTE | H_PAGE_THP_HUGE)) ==
static inline int hash__pmd_same(pmd_t pmd_a, pmd_t pmd_b)
return (((pmd_raw(pmd_a) ^ pmd_raw(pmd_b)) & ~cpu_to_be64(_PAGE_HPTEFLAGS)) == 0);
static inline pmd_t hash__pmd_mkhuge(pmd_t pmd)
return __pmd(pmd_val(pmd) | (_PAGE_PTE | H_PAGE_THP_HUGE));
extern unsigned long hash__pmd_hugepage_update(struct mm_struct *mm,
unsigned long addr, pmd_t *pmdp,
unsigned long clr, unsigned long set);
extern pmd_t hash__pmdp_collapse_flush(struct vm_area_struct *vma,
unsigned long address, pmd_t *pmdp);
extern void hash__pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
pgtable_t pgtable);
extern pgtable_t hash__pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
extern pmd_t hash__pmdp_huge_get_and_clear(struct mm_struct *mm,
unsigned long addr, pmd_t *pmdp);
extern int hash__has_transparent_hugepage(void);
#endif /* __ASSEMBLY__ */
#endif /* _ASM_POWERPC_BOOK3S_64_HASH_64K_H */