2013-04-10 19:48:00 +07:00
|
|
|
/*
|
|
|
|
* arch/arm64/mm/hugetlbpage.c
|
|
|
|
*
|
|
|
|
* Copyright (C) 2013 Linaro Ltd.
|
|
|
|
*
|
|
|
|
* Based on arch/x86/mm/hugetlbpage.c.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License version 2 as
|
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
* GNU General Public License for more details.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/hugetlb.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/err.h>
|
|
|
|
#include <linux/sysctl.h>
|
|
|
|
#include <asm/mman.h>
|
|
|
|
#include <asm/tlb.h>
|
|
|
|
#include <asm/tlbflush.h>
|
|
|
|
#include <asm/pgalloc.h>
|
|
|
|
|
|
|
|
int pmd_huge(pmd_t pmd)
|
|
|
|
{
|
2015-07-01 19:08:31 +07:00
|
|
|
return pmd_val(pmd) && !(pmd_val(pmd) & PMD_TABLE_BIT);
|
2013-04-10 19:48:00 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int pud_huge(pud_t pud)
|
|
|
|
{
|
2014-05-15 21:19:22 +07:00
|
|
|
#ifndef __PAGETABLE_PMD_FOLDED
|
2015-07-01 19:08:31 +07:00
|
|
|
return pud_val(pud) && !(pud_val(pud) & PUD_TABLE_BIT);
|
2014-05-15 21:19:22 +07:00
|
|
|
#else
|
|
|
|
return 0;
|
|
|
|
#endif
|
2013-04-10 19:48:00 +07:00
|
|
|
}
|
|
|
|
|
2015-12-18 02:31:26 +07:00
|
|
|
static int find_num_contig(struct mm_struct *mm, unsigned long addr,
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
pte_t *ptep, size_t *pgsize)
|
2015-12-18 02:31:26 +07:00
|
|
|
{
|
|
|
|
pgd_t *pgd = pgd_offset(mm, addr);
|
|
|
|
pud_t *pud;
|
|
|
|
pmd_t *pmd;
|
|
|
|
|
|
|
|
*pgsize = PAGE_SIZE;
|
|
|
|
pud = pud_offset(pgd, addr);
|
|
|
|
pmd = pmd_offset(pud, addr);
|
|
|
|
if ((pte_t *)pmd == ptep) {
|
|
|
|
*pgsize = PMD_SIZE;
|
|
|
|
return CONT_PMDS;
|
|
|
|
}
|
|
|
|
return CONT_PTES;
|
|
|
|
}
|
|
|
|
|
|
|
|
void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
|
|
|
|
pte_t *ptep, pte_t pte)
|
|
|
|
{
|
|
|
|
size_t pgsize;
|
|
|
|
int i;
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
int ncontig;
|
2015-12-18 02:31:26 +07:00
|
|
|
unsigned long pfn;
|
|
|
|
pgprot_t hugeprot;
|
|
|
|
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
if (!pte_cont(pte)) {
|
2015-12-18 02:31:26 +07:00
|
|
|
set_pte_at(mm, addr, ptep, pte);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
ncontig = find_num_contig(mm, addr, ptep, &pgsize);
|
2015-12-18 02:31:26 +07:00
|
|
|
pfn = pte_pfn(pte);
|
|
|
|
hugeprot = __pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^ pte_val(pte));
|
|
|
|
for (i = 0; i < ncontig; i++) {
|
|
|
|
pr_debug("%s: set pte %p to 0x%llx\n", __func__, ptep,
|
|
|
|
pte_val(pfn_pte(pfn, hugeprot)));
|
|
|
|
set_pte_at(mm, addr, ptep, pfn_pte(pfn, hugeprot));
|
|
|
|
ptep++;
|
|
|
|
pfn += pgsize >> PAGE_SHIFT;
|
|
|
|
addr += pgsize;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
pte_t *huge_pte_alloc(struct mm_struct *mm,
|
|
|
|
unsigned long addr, unsigned long sz)
|
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
|
|
|
pud_t *pud;
|
|
|
|
pte_t *pte = NULL;
|
|
|
|
|
|
|
|
pr_debug("%s: addr:0x%lx sz:0x%lx\n", __func__, addr, sz);
|
|
|
|
pgd = pgd_offset(mm, addr);
|
|
|
|
pud = pud_alloc(mm, pgd, addr);
|
|
|
|
if (!pud)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (sz == PUD_SIZE) {
|
|
|
|
pte = (pte_t *)pud;
|
|
|
|
} else if (sz == (PAGE_SIZE * CONT_PTES)) {
|
|
|
|
pmd_t *pmd = pmd_alloc(mm, pud, addr);
|
|
|
|
|
|
|
|
WARN_ON(addr & (sz - 1));
|
|
|
|
/*
|
|
|
|
* Note that if this code were ever ported to the
|
|
|
|
* 32-bit arm platform then it will cause trouble in
|
|
|
|
* the case where CONFIG_HIGHPTE is set, since there
|
|
|
|
* will be no pte_unmap() to correspond with this
|
|
|
|
* pte_alloc_map().
|
|
|
|
*/
|
2016-03-18 04:19:11 +07:00
|
|
|
pte = pte_alloc_map(mm, pmd, addr);
|
2015-12-18 02:31:26 +07:00
|
|
|
} else if (sz == PMD_SIZE) {
|
|
|
|
if (IS_ENABLED(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) &&
|
|
|
|
pud_none(*pud))
|
|
|
|
pte = huge_pmd_share(mm, addr, pud);
|
|
|
|
else
|
|
|
|
pte = (pte_t *)pmd_alloc(mm, pud, addr);
|
|
|
|
} else if (sz == (PMD_SIZE * CONT_PMDS)) {
|
|
|
|
pmd_t *pmd;
|
|
|
|
|
|
|
|
pmd = pmd_alloc(mm, pud, addr);
|
|
|
|
WARN_ON(addr & (sz - 1));
|
|
|
|
return (pte_t *)pmd;
|
|
|
|
}
|
|
|
|
|
|
|
|
pr_debug("%s: addr:0x%lx sz:0x%lx ret pte=%p/0x%llx\n", __func__, addr,
|
|
|
|
sz, pte, pte_val(*pte));
|
|
|
|
return pte;
|
|
|
|
}
|
|
|
|
|
|
|
|
pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
|
|
|
|
{
|
|
|
|
pgd_t *pgd;
|
|
|
|
pud_t *pud;
|
2017-06-09 00:25:26 +07:00
|
|
|
pmd_t *pmd;
|
2015-12-18 02:31:26 +07:00
|
|
|
|
|
|
|
pgd = pgd_offset(mm, addr);
|
|
|
|
pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
|
|
|
|
if (!pgd_present(*pgd))
|
|
|
|
return NULL;
|
2017-06-09 00:25:26 +07:00
|
|
|
|
2015-12-18 02:31:26 +07:00
|
|
|
pud = pud_offset(pgd, addr);
|
2017-06-09 00:25:26 +07:00
|
|
|
if (pud_none(*pud))
|
2015-12-18 02:31:26 +07:00
|
|
|
return NULL;
|
2017-06-09 00:25:26 +07:00
|
|
|
/* swap or huge page */
|
|
|
|
if (!pud_present(*pud) || pud_huge(*pud))
|
2015-12-18 02:31:26 +07:00
|
|
|
return (pte_t *)pud;
|
2017-06-09 00:25:26 +07:00
|
|
|
/* table; check the next level */
|
|
|
|
|
2015-12-18 02:31:26 +07:00
|
|
|
pmd = pmd_offset(pud, addr);
|
2017-06-09 00:25:26 +07:00
|
|
|
if (pmd_none(*pmd))
|
2015-12-18 02:31:26 +07:00
|
|
|
return NULL;
|
2017-06-09 00:25:26 +07:00
|
|
|
if (!pmd_present(*pmd) || pmd_huge(*pmd))
|
2015-12-18 02:31:26 +07:00
|
|
|
return (pte_t *)pmd;
|
2017-06-09 00:25:26 +07:00
|
|
|
|
2015-12-18 02:31:26 +07:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
|
|
|
|
struct page *page, int writable)
|
|
|
|
{
|
|
|
|
size_t pagesize = huge_page_size(hstate_vma(vma));
|
|
|
|
|
|
|
|
if (pagesize == CONT_PTE_SIZE) {
|
|
|
|
entry = pte_mkcont(entry);
|
|
|
|
} else if (pagesize == CONT_PMD_SIZE) {
|
|
|
|
entry = pmd_pte(pmd_mkcont(pte_pmd(entry)));
|
|
|
|
} else if (pagesize != PUD_SIZE && pagesize != PMD_SIZE) {
|
|
|
|
pr_warn("%s: unrecognized huge page size 0x%lx\n",
|
|
|
|
__func__, pagesize);
|
|
|
|
}
|
|
|
|
return entry;
|
|
|
|
}
|
|
|
|
|
|
|
|
pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
|
|
|
|
unsigned long addr, pte_t *ptep)
|
|
|
|
{
|
|
|
|
pte_t pte;
|
|
|
|
|
|
|
|
if (pte_cont(*ptep)) {
|
|
|
|
int ncontig, i;
|
|
|
|
size_t pgsize;
|
|
|
|
bool is_dirty = false;
|
|
|
|
|
2017-07-07 05:39:33 +07:00
|
|
|
ncontig = find_num_contig(mm, addr, ptep, &pgsize);
|
2015-12-18 02:31:26 +07:00
|
|
|
/* save the 1st pte to return */
|
2017-07-07 05:39:33 +07:00
|
|
|
pte = ptep_get_and_clear(mm, addr, ptep);
|
2016-11-08 12:44:39 +07:00
|
|
|
for (i = 1, addr += pgsize; i < ncontig; ++i, addr += pgsize) {
|
2015-12-18 02:31:26 +07:00
|
|
|
/*
|
|
|
|
* If HW_AFDBM is enabled, then the HW could
|
|
|
|
* turn on the dirty bit for any of the page
|
|
|
|
* in the set, so check them all.
|
|
|
|
*/
|
2017-07-07 05:39:33 +07:00
|
|
|
++ptep;
|
|
|
|
if (pte_dirty(ptep_get_and_clear(mm, addr, ptep)))
|
2015-12-18 02:31:26 +07:00
|
|
|
is_dirty = true;
|
|
|
|
}
|
|
|
|
if (is_dirty)
|
|
|
|
return pte_mkdirty(pte);
|
|
|
|
else
|
|
|
|
return pte;
|
|
|
|
} else {
|
|
|
|
return ptep_get_and_clear(mm, addr, ptep);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
int huge_ptep_set_access_flags(struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, pte_t *ptep,
|
|
|
|
pte_t pte, int dirty)
|
|
|
|
{
|
|
|
|
if (pte_cont(pte)) {
|
|
|
|
int ncontig, i, changed = 0;
|
|
|
|
size_t pgsize = 0;
|
|
|
|
unsigned long pfn = pte_pfn(pte);
|
|
|
|
/* Select all bits except the pfn */
|
|
|
|
pgprot_t hugeprot =
|
|
|
|
__pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^
|
|
|
|
pte_val(pte));
|
|
|
|
|
2017-07-07 05:39:33 +07:00
|
|
|
pfn = pte_pfn(pte);
|
|
|
|
ncontig = find_num_contig(vma->vm_mm, addr, ptep,
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
&pgsize);
|
2017-07-07 05:39:33 +07:00
|
|
|
for (i = 0; i < ncontig; ++i, ++ptep, addr += pgsize) {
|
|
|
|
changed |= ptep_set_access_flags(vma, addr, ptep,
|
2015-12-18 02:31:26 +07:00
|
|
|
pfn_pte(pfn,
|
|
|
|
hugeprot),
|
|
|
|
dirty);
|
|
|
|
pfn += pgsize >> PAGE_SHIFT;
|
|
|
|
}
|
|
|
|
return changed;
|
|
|
|
} else {
|
|
|
|
return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void huge_ptep_set_wrprotect(struct mm_struct *mm,
|
|
|
|
unsigned long addr, pte_t *ptep)
|
|
|
|
{
|
|
|
|
if (pte_cont(*ptep)) {
|
|
|
|
int ncontig, i;
|
|
|
|
size_t pgsize = 0;
|
|
|
|
|
2017-07-07 05:39:33 +07:00
|
|
|
ncontig = find_num_contig(mm, addr, ptep, &pgsize);
|
|
|
|
for (i = 0; i < ncontig; ++i, ++ptep, addr += pgsize)
|
|
|
|
ptep_set_wrprotect(mm, addr, ptep);
|
2015-12-18 02:31:26 +07:00
|
|
|
} else {
|
|
|
|
ptep_set_wrprotect(mm, addr, ptep);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void huge_ptep_clear_flush(struct vm_area_struct *vma,
|
|
|
|
unsigned long addr, pte_t *ptep)
|
|
|
|
{
|
|
|
|
if (pte_cont(*ptep)) {
|
|
|
|
int ncontig, i;
|
|
|
|
size_t pgsize = 0;
|
|
|
|
|
2017-07-07 05:39:33 +07:00
|
|
|
ncontig = find_num_contig(vma->vm_mm, addr, ptep,
|
arm64: hugetlb: refactor find_num_contig()
Patch series "Support for contiguous pte hugepages", v4.
This patchset updates the hugetlb code to fix issues arising from
contiguous pte hugepages (such as on arm64). Compared to v3, This
version addresses a build failure on arm64 by including two cleanup
patches. Other than the arm64 cleanups, the rest are generic code
changes. The remaining arm64 support based on these patches will be
posted separately. The patches are based on v4.12-rc2. Previous
related postings can be found at [0], [1], [2], and [3].
The patches fall into three categories -
* Patch 1-2 - arm64 cleanups required to greatly simplify changing
huge_pte_offset() prototype in Patch 5.
Catalin, Will - are you happy for these patches to go via mm?
* Patches 3-4 address issues with gup
* Patches 5-8 relate to passing a size argument to hugepage helpers to
disambiguate the size of the referred page. These changes are
required to enable arch code to properly handle swap entries for
contiguous pte hugepages.
The changes to huge_pte_offset() (patch 5) touch multiple
architectures but I've managed to minimise these changes for the
other affected functions - huge_pte_clear() and set_huge_pte_at().
These patches gate the enabling of contiguous hugepages support on arm64
which has been requested for systems using !4k page granule.
The ARM64 architecture supports two flavours of hugepages -
* Block mappings at the pud/pmd level
These are regular hugepages where a pmd or a pud page table entry
points to a block of memory. Depending on the PAGE_SIZE in use the
following size of block mappings are supported -
PMD PUD
--- ---
4K: 2M 1G
16K: 32M
64K: 512M
For certain applications/usecases such as HPC and large enterprise
workloads, folks are using 64k page size but the minimum hugepage size
of 512MB isn't very practical.
To overcome this ...
* Using the Contiguous bit
The architecture provides a contiguous bit in the translation table
entry which acts as a hint to the mmu to indicate that it is one of a
contiguous set of entries that can be cached in a single TLB entry.
We use the contiguous bit in Linux to increase the mapping size at the
pmd and pte (last) level.
The number of supported contiguous entries varies by page size and
level of the page table.
Using the contiguous bit allows additional hugepage sizes -
CONT PTE PMD CONT PMD PUD
-------- --- -------- ---
4K: 64K 2M 32M 1G
16K: 2M 32M 1G
64K: 2M 512M 16G
Of these, 64K with 4K and 2M with 64K pages have been explicitly
requested by a few different users.
Entries with the contiguous bit set are required to be modified all
together - which makes things like memory poisoning and migration
impossible to do correctly without knowing the size of hugepage being
dealt with - the reason for adding size parameter to a few of the
hugepage helpers in this series.
This patch (of 8):
As we regularly check for contiguous pte's in the huge accessors, remove
this extra check from find_num_contig.
[punit.agrawal@arm.com: resolve rebase conflicts due to patch re-ordering]
Link: http://lkml.kernel.org/r/20170524115409.31309-2-punit.agrawal@arm.com
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Cc: David Woods <dwoods@mellanox.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-07-07 05:39:29 +07:00
|
|
|
&pgsize);
|
2017-07-07 05:39:33 +07:00
|
|
|
for (i = 0; i < ncontig; ++i, ++ptep, addr += pgsize)
|
|
|
|
ptep_clear_flush(vma, addr, ptep);
|
2015-12-18 02:31:26 +07:00
|
|
|
} else {
|
|
|
|
ptep_clear_flush(vma, addr, ptep);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-04-10 19:48:00 +07:00
|
|
|
static __init int setup_hugepagesz(char *opt)
|
|
|
|
{
|
|
|
|
unsigned long ps = memparse(opt, &opt);
|
2015-12-18 02:31:26 +07:00
|
|
|
|
2013-04-10 19:48:00 +07:00
|
|
|
if (ps == PMD_SIZE) {
|
|
|
|
hugetlb_add_hstate(PMD_SHIFT - PAGE_SHIFT);
|
|
|
|
} else if (ps == PUD_SIZE) {
|
|
|
|
hugetlb_add_hstate(PUD_SHIFT - PAGE_SHIFT);
|
|
|
|
} else {
|
2016-05-20 07:11:08 +07:00
|
|
|
hugetlb_bad_size();
|
2015-12-18 02:31:26 +07:00
|
|
|
pr_err("hugepagesz: Unsupported page size %lu K\n", ps >> 10);
|
2013-04-10 19:48:00 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("hugepagesz=", setup_hugepagesz);
|