Commit 0db501f7 authored by Ebru Akagunduz's avatar Ebru Akagunduz Committed by Linus Torvalds

mm, thp: convert from optimistic swapin collapsing to conservative

To detect whether khugepaged swapin is worthwhile, this patch checks the
amount of young pages.  There should be at least half of HPAGE_PMD_NR to
swapin.

Link: http://lkml.kernel.org/r/1468109451-1615-1-git-send-email-ebru.akagunduz@gmail.comSigned-off-by: default avatarEbru Akagunduz <ebru.akagunduz@gmail.com>
Suggested-by: default avatarMinchan Kim <minchan@kernel.org>
Acked-by: default avatarRik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Boaz Harrosh <boaz@plexistor.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent 47f863ea
...@@ -13,7 +13,7 @@ ...@@ -13,7 +13,7 @@
EM( SCAN_EXCEED_NONE_PTE, "exceed_none_pte") \ EM( SCAN_EXCEED_NONE_PTE, "exceed_none_pte") \
EM( SCAN_PTE_NON_PRESENT, "pte_non_present") \ EM( SCAN_PTE_NON_PRESENT, "pte_non_present") \
EM( SCAN_PAGE_RO, "no_writable_page") \ EM( SCAN_PAGE_RO, "no_writable_page") \
EM( SCAN_NO_REFERENCED_PAGE, "no_referenced_page") \ EM( SCAN_LACK_REFERENCED_PAGE, "lack_referenced_page") \
EM( SCAN_PAGE_NULL, "page_null") \ EM( SCAN_PAGE_NULL, "page_null") \
EM( SCAN_SCAN_ABORT, "scan_aborted") \ EM( SCAN_SCAN_ABORT, "scan_aborted") \
EM( SCAN_PAGE_COUNT, "not_suitable_page_count") \ EM( SCAN_PAGE_COUNT, "not_suitable_page_count") \
...@@ -47,7 +47,7 @@ SCAN_STATUS ...@@ -47,7 +47,7 @@ SCAN_STATUS
TRACE_EVENT(mm_khugepaged_scan_pmd, TRACE_EVENT(mm_khugepaged_scan_pmd,
TP_PROTO(struct mm_struct *mm, struct page *page, bool writable, TP_PROTO(struct mm_struct *mm, struct page *page, bool writable,
bool referenced, int none_or_zero, int status, int unmapped), int referenced, int none_or_zero, int status, int unmapped),
TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped), TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped),
...@@ -55,7 +55,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd, ...@@ -55,7 +55,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd,
__field(struct mm_struct *, mm) __field(struct mm_struct *, mm)
__field(unsigned long, pfn) __field(unsigned long, pfn)
__field(bool, writable) __field(bool, writable)
__field(bool, referenced) __field(int, referenced)
__field(int, none_or_zero) __field(int, none_or_zero)
__field(int, status) __field(int, status)
__field(int, unmapped) __field(int, unmapped)
...@@ -108,14 +108,14 @@ TRACE_EVENT(mm_collapse_huge_page, ...@@ -108,14 +108,14 @@ TRACE_EVENT(mm_collapse_huge_page,
TRACE_EVENT(mm_collapse_huge_page_isolate, TRACE_EVENT(mm_collapse_huge_page_isolate,
TP_PROTO(struct page *page, int none_or_zero, TP_PROTO(struct page *page, int none_or_zero,
bool referenced, bool writable, int status), int referenced, bool writable, int status),
TP_ARGS(page, none_or_zero, referenced, writable, status), TP_ARGS(page, none_or_zero, referenced, writable, status),
TP_STRUCT__entry( TP_STRUCT__entry(
__field(unsigned long, pfn) __field(unsigned long, pfn)
__field(int, none_or_zero) __field(int, none_or_zero)
__field(bool, referenced) __field(int, referenced)
__field(bool, writable) __field(bool, writable)
__field(int, status) __field(int, status)
), ),
...@@ -138,25 +138,28 @@ TRACE_EVENT(mm_collapse_huge_page_isolate, ...@@ -138,25 +138,28 @@ TRACE_EVENT(mm_collapse_huge_page_isolate,
TRACE_EVENT(mm_collapse_huge_page_swapin, TRACE_EVENT(mm_collapse_huge_page_swapin,
TP_PROTO(struct mm_struct *mm, int swapped_in, int ret), TP_PROTO(struct mm_struct *mm, int swapped_in, int referenced, int ret),
TP_ARGS(mm, swapped_in, ret), TP_ARGS(mm, swapped_in, referenced, ret),
TP_STRUCT__entry( TP_STRUCT__entry(
__field(struct mm_struct *, mm) __field(struct mm_struct *, mm)
__field(int, swapped_in) __field(int, swapped_in)
__field(int, referenced)
__field(int, ret) __field(int, ret)
), ),
TP_fast_assign( TP_fast_assign(
__entry->mm = mm; __entry->mm = mm;
__entry->swapped_in = swapped_in; __entry->swapped_in = swapped_in;
__entry->referenced = referenced;
__entry->ret = ret; __entry->ret = ret;
), ),
TP_printk("mm=%p, swapped_in=%d, ret=%d", TP_printk("mm=%p, swapped_in=%d, referenced=%d, ret=%d",
__entry->mm, __entry->mm,
__entry->swapped_in, __entry->swapped_in,
__entry->referenced,
__entry->ret) __entry->ret)
); );
......
...@@ -27,7 +27,7 @@ enum scan_result { ...@@ -27,7 +27,7 @@ enum scan_result {
SCAN_EXCEED_NONE_PTE, SCAN_EXCEED_NONE_PTE,
SCAN_PTE_NON_PRESENT, SCAN_PTE_NON_PRESENT,
SCAN_PAGE_RO, SCAN_PAGE_RO,
SCAN_NO_REFERENCED_PAGE, SCAN_LACK_REFERENCED_PAGE,
SCAN_PAGE_NULL, SCAN_PAGE_NULL,
SCAN_SCAN_ABORT, SCAN_SCAN_ABORT,
SCAN_PAGE_COUNT, SCAN_PAGE_COUNT,
...@@ -500,8 +500,8 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma, ...@@ -500,8 +500,8 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
{ {
struct page *page = NULL; struct page *page = NULL;
pte_t *_pte; pte_t *_pte;
int none_or_zero = 0, result = 0; int none_or_zero = 0, result = 0, referenced = 0;
bool referenced = false, writable = false; bool writable = false;
for (_pte = pte; _pte < pte+HPAGE_PMD_NR; for (_pte = pte; _pte < pte+HPAGE_PMD_NR;
_pte++, address += PAGE_SIZE) { _pte++, address += PAGE_SIZE) {
...@@ -580,11 +580,11 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma, ...@@ -580,11 +580,11 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
VM_BUG_ON_PAGE(!PageLocked(page), page); VM_BUG_ON_PAGE(!PageLocked(page), page);
VM_BUG_ON_PAGE(PageLRU(page), page); VM_BUG_ON_PAGE(PageLRU(page), page);
/* If there is no mapped pte young don't collapse the page */ /* There should be enough young pte to collapse the page */
if (pte_young(pteval) || if (pte_young(pteval) ||
page_is_young(page) || PageReferenced(page) || page_is_young(page) || PageReferenced(page) ||
mmu_notifier_test_young(vma->vm_mm, address)) mmu_notifier_test_young(vma->vm_mm, address))
referenced = true; referenced++;
} }
if (likely(writable)) { if (likely(writable)) {
if (likely(referenced)) { if (likely(referenced)) {
...@@ -869,7 +869,8 @@ static int hugepage_vma_revalidate(struct mm_struct *mm, unsigned long address) ...@@ -869,7 +869,8 @@ static int hugepage_vma_revalidate(struct mm_struct *mm, unsigned long address)
static bool __collapse_huge_page_swapin(struct mm_struct *mm, static bool __collapse_huge_page_swapin(struct mm_struct *mm,
struct vm_area_struct *vma, struct vm_area_struct *vma,
unsigned long address, pmd_t *pmd) unsigned long address, pmd_t *pmd,
int referenced)
{ {
pte_t pteval; pte_t pteval;
int swapped_in = 0, ret = 0; int swapped_in = 0, ret = 0;
...@@ -887,12 +888,19 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm, ...@@ -887,12 +888,19 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm,
if (!is_swap_pte(pteval)) if (!is_swap_pte(pteval))
continue; continue;
swapped_in++; swapped_in++;
/* we only decide to swapin, if there is enough young ptes */
if (referenced < HPAGE_PMD_NR/2) {
trace_mm_collapse_huge_page_swapin(mm, swapped_in, referenced, 0);
return false;
}
ret = do_swap_page(&fe, pteval); ret = do_swap_page(&fe, pteval);
/* do_swap_page returns VM_FAULT_RETRY with released mmap_sem */ /* do_swap_page returns VM_FAULT_RETRY with released mmap_sem */
if (ret & VM_FAULT_RETRY) { if (ret & VM_FAULT_RETRY) {
down_read(&mm->mmap_sem); down_read(&mm->mmap_sem);
if (hugepage_vma_revalidate(mm, address)) { if (hugepage_vma_revalidate(mm, address)) {
/* vma is no longer available, don't continue to swapin */ /* vma is no longer available, don't continue to swapin */
trace_mm_collapse_huge_page_swapin(mm, swapped_in, referenced, 0);
return false; return false;
} }
/* check if the pmd is still valid */ /* check if the pmd is still valid */
...@@ -900,7 +908,7 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm, ...@@ -900,7 +908,7 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm,
return false; return false;
} }
if (ret & VM_FAULT_ERROR) { if (ret & VM_FAULT_ERROR) {
trace_mm_collapse_huge_page_swapin(mm, swapped_in, 0); trace_mm_collapse_huge_page_swapin(mm, swapped_in, referenced, 0);
return false; return false;
} }
/* pte is unmapped now, we need to map it */ /* pte is unmapped now, we need to map it */
...@@ -908,7 +916,7 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm, ...@@ -908,7 +916,7 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm,
} }
fe.pte--; fe.pte--;
pte_unmap(fe.pte); pte_unmap(fe.pte);
trace_mm_collapse_huge_page_swapin(mm, swapped_in, 1); trace_mm_collapse_huge_page_swapin(mm, swapped_in, referenced, 1);
return true; return true;
} }
...@@ -916,7 +924,7 @@ static void collapse_huge_page(struct mm_struct *mm, ...@@ -916,7 +924,7 @@ static void collapse_huge_page(struct mm_struct *mm,
unsigned long address, unsigned long address,
struct page **hpage, struct page **hpage,
struct vm_area_struct *vma, struct vm_area_struct *vma,
int node) int node, int referenced)
{ {
pmd_t *pmd, _pmd; pmd_t *pmd, _pmd;
pte_t *pte; pte_t *pte;
...@@ -973,7 +981,7 @@ static void collapse_huge_page(struct mm_struct *mm, ...@@ -973,7 +981,7 @@ static void collapse_huge_page(struct mm_struct *mm,
* If it fails, we release mmap_sem and jump out_nolock. * If it fails, we release mmap_sem and jump out_nolock.
* Continuing to collapse causes inconsistency. * Continuing to collapse causes inconsistency.
*/ */
if (!__collapse_huge_page_swapin(mm, vma, address, pmd)) { if (!__collapse_huge_page_swapin(mm, vma, address, pmd, referenced)) {
mem_cgroup_cancel_charge(new_page, memcg, true); mem_cgroup_cancel_charge(new_page, memcg, true);
up_read(&mm->mmap_sem); up_read(&mm->mmap_sem);
goto out_nolock; goto out_nolock;
...@@ -1084,12 +1092,12 @@ static int khugepaged_scan_pmd(struct mm_struct *mm, ...@@ -1084,12 +1092,12 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
{ {
pmd_t *pmd; pmd_t *pmd;
pte_t *pte, *_pte; pte_t *pte, *_pte;
int ret = 0, none_or_zero = 0, result = 0; int ret = 0, none_or_zero = 0, result = 0, referenced = 0;
struct page *page = NULL; struct page *page = NULL;
unsigned long _address; unsigned long _address;
spinlock_t *ptl; spinlock_t *ptl;
int node = NUMA_NO_NODE, unmapped = 0; int node = NUMA_NO_NODE, unmapped = 0;
bool writable = false, referenced = false; bool writable = false;
VM_BUG_ON(address & ~HPAGE_PMD_MASK); VM_BUG_ON(address & ~HPAGE_PMD_MASK);
...@@ -1177,14 +1185,14 @@ static int khugepaged_scan_pmd(struct mm_struct *mm, ...@@ -1177,14 +1185,14 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
if (pte_young(pteval) || if (pte_young(pteval) ||
page_is_young(page) || PageReferenced(page) || page_is_young(page) || PageReferenced(page) ||
mmu_notifier_test_young(vma->vm_mm, address)) mmu_notifier_test_young(vma->vm_mm, address))
referenced = true; referenced++;
} }
if (writable) { if (writable) {
if (referenced) { if (referenced) {
result = SCAN_SUCCEED; result = SCAN_SUCCEED;
ret = 1; ret = 1;
} else { } else {
result = SCAN_NO_REFERENCED_PAGE; result = SCAN_LACK_REFERENCED_PAGE;
} }
} else { } else {
result = SCAN_PAGE_RO; result = SCAN_PAGE_RO;
...@@ -1194,7 +1202,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm, ...@@ -1194,7 +1202,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
if (ret) { if (ret) {
node = khugepaged_find_target_node(); node = khugepaged_find_target_node();
/* collapse_huge_page will return with the mmap_sem released */ /* collapse_huge_page will return with the mmap_sem released */
collapse_huge_page(mm, address, hpage, vma, node); collapse_huge_page(mm, address, hpage, vma, node, referenced);
} }
out: out:
trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced, trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment