NR_FILE_MAPPED accounting in mm/rmap.c (for /proc/meminfo "Mapped" and
/proc/vmstat "nr_mapped" and the memcg's memory.stat "mapped_file") is
slightly flawed for file or shmem huge pages.
It is well thought out, and looks convincing, but there's a racy case
when the careful counting in page_remove_file_rmap() (without page lock)
gets discarded. So that in a workload like two "make -j20" kernel builds
under memory pressure, with cc1 on hugepage text, "Mapped" can easily
grow by a spurious 5MB or more on each iteration, ending up implausibly
bigger than most other numbers in /proc/meminfo. And, hypothetically,
might grow to the point of seriously interfering in mm/vmscan.c's
heuristics, which do take NR_FILE_MAPPED into some consideration.
Fixed by moving the __mod_lruvec_page_state() down to where it will not
be missed before return (and I've grown a bit tired of that oft-repeated
but-not-everywhere comment on the __ness: it gets lost in the move here).
Does page_add_file_rmap() need the same change? I suspect not, because
page lock is held in all relevant cases, and its skipping case looks safe;
but it's much easier to be sure, if we do make the same change.
Fixes: dd78fedde4b9 ("rmap: support file thp")
Signed-off-by: Hugh Dickins <[email protected]>
---
If this were thought serious enough to backport (I don't feel strongly,
but it is something I keep in my own trees), it needs a little more care
near "out", because the mm/munlock series has removed some action there.
mm/rmap.c | 31 ++++++++++++++-----------------
1 file changed, 14 insertions(+), 17 deletions(-)
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1238,14 +1238,14 @@ void page_add_new_anon_rmap(struct page *page,
void page_add_file_rmap(struct page *page,
struct vm_area_struct *vma, bool compound)
{
- int i, nr = 1;
+ int i, nr = 0;
VM_BUG_ON_PAGE(compound && !PageTransHuge(page), page);
lock_page_memcg(page);
if (compound && PageTransHuge(page)) {
int nr_pages = thp_nr_pages(page);
- for (i = 0, nr = 0; i < nr_pages; i++) {
+ for (i = 0; i < nr_pages; i++) {
if (atomic_inc_and_test(&page[i]._mapcount))
nr++;
}
@@ -1262,11 +1262,12 @@ void page_add_file_rmap(struct page *page,
VM_WARN_ON_ONCE(!PageLocked(page));
SetPageDoubleMap(compound_head(page));
}
- if (!atomic_inc_and_test(&page->_mapcount))
- goto out;
+ if (atomic_inc_and_test(&page->_mapcount))
+ nr++;
}
- __mod_lruvec_page_state(page, NR_FILE_MAPPED, nr);
out:
+ if (nr)
+ __mod_lruvec_page_state(page, NR_FILE_MAPPED, nr);
unlock_page_memcg(page);
mlock_vma_page(page, vma, compound);
@@ -1274,7 +1275,7 @@ void page_add_file_rmap(struct page *page,
static void page_remove_file_rmap(struct page *page, bool compound)
{
- int i, nr = 1;
+ int i, nr = 0;
VM_BUG_ON_PAGE(compound && !PageHead(page), page);
@@ -1289,12 +1290,12 @@ static void page_remove_file_rmap(struct page *page, bool compound)
if (compound && PageTransHuge(page)) {
int nr_pages = thp_nr_pages(page);
- for (i = 0, nr = 0; i < nr_pages; i++) {
+ for (i = 0; i < nr_pages; i++) {
if (atomic_add_negative(-1, &page[i]._mapcount))
nr++;
}
if (!atomic_add_negative(-1, compound_mapcount_ptr(page)))
- return;
+ goto out;
if (PageSwapBacked(page))
__mod_lruvec_page_state(page, NR_SHMEM_PMDMAPPED,
-nr_pages);
@@ -1302,16 +1303,12 @@ static void page_remove_file_rmap(struct page *page, bool compound)
__mod_lruvec_page_state(page, NR_FILE_PMDMAPPED,
-nr_pages);
} else {
- if (!atomic_add_negative(-1, &page->_mapcount))
- return;
+ if (atomic_add_negative(-1, &page->_mapcount))
+ nr++;
}
-
- /*
- * We use the irq-unsafe __{inc|mod}_lruvec_page_state because
- * these counters are not modified in interrupt context, and
- * pte lock(a spinlock) is held, which implies preemption disabled.
- */
- __mod_lruvec_page_state(page, NR_FILE_MAPPED, -nr);
+out:
+ if (nr)
+ __mod_lruvec_page_state(page, NR_FILE_MAPPED, -nr);
}
static void page_remove_anon_compound_rmap(struct page *page)
On Wed, Mar 2, 2022 at 5:46 PM Hugh Dickins <[email protected]> wrote:
>
> NR_FILE_MAPPED accounting in mm/rmap.c (for /proc/meminfo "Mapped" and
> /proc/vmstat "nr_mapped" and the memcg's memory.stat "mapped_file") is
> slightly flawed for file or shmem huge pages.
>
> It is well thought out, and looks convincing, but there's a racy case
> when the careful counting in page_remove_file_rmap() (without page lock)
> gets discarded. So that in a workload like two "make -j20" kernel builds
> under memory pressure, with cc1 on hugepage text, "Mapped" can easily
> grow by a spurious 5MB or more on each iteration, ending up implausibly
> bigger than most other numbers in /proc/meminfo. And, hypothetically,
> might grow to the point of seriously interfering in mm/vmscan.c's
> heuristics, which do take NR_FILE_MAPPED into some consideration.
>
> Fixed by moving the __mod_lruvec_page_state() down to where it will not
> be missed before return (and I've grown a bit tired of that oft-repeated
> but-not-everywhere comment on the __ness: it gets lost in the move here).
>
> Does page_add_file_rmap() need the same change? I suspect not, because
> page lock is held in all relevant cases, and its skipping case looks safe;
> but it's much easier to be sure, if we do make the same change.
>
> Fixes: dd78fedde4b9 ("rmap: support file thp")
> Signed-off-by: Hugh Dickins <[email protected]>
Reviewed-by: Yang Shi <[email protected]>
> ---
> If this were thought serious enough to backport (I don't feel strongly,
> but it is something I keep in my own trees), it needs a little more care
> near "out", because the mm/munlock series has removed some action there.
>
> mm/rmap.c | 31 ++++++++++++++-----------------
> 1 file changed, 14 insertions(+), 17 deletions(-)
>
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -1238,14 +1238,14 @@ void page_add_new_anon_rmap(struct page *page,
> void page_add_file_rmap(struct page *page,
> struct vm_area_struct *vma, bool compound)
> {
> - int i, nr = 1;
> + int i, nr = 0;
>
> VM_BUG_ON_PAGE(compound && !PageTransHuge(page), page);
> lock_page_memcg(page);
> if (compound && PageTransHuge(page)) {
> int nr_pages = thp_nr_pages(page);
>
> - for (i = 0, nr = 0; i < nr_pages; i++) {
> + for (i = 0; i < nr_pages; i++) {
> if (atomic_inc_and_test(&page[i]._mapcount))
> nr++;
> }
> @@ -1262,11 +1262,12 @@ void page_add_file_rmap(struct page *page,
> VM_WARN_ON_ONCE(!PageLocked(page));
> SetPageDoubleMap(compound_head(page));
> }
> - if (!atomic_inc_and_test(&page->_mapcount))
> - goto out;
> + if (atomic_inc_and_test(&page->_mapcount))
> + nr++;
> }
> - __mod_lruvec_page_state(page, NR_FILE_MAPPED, nr);
> out:
> + if (nr)
> + __mod_lruvec_page_state(page, NR_FILE_MAPPED, nr);
> unlock_page_memcg(page);
>
> mlock_vma_page(page, vma, compound);
> @@ -1274,7 +1275,7 @@ void page_add_file_rmap(struct page *page,
>
> static void page_remove_file_rmap(struct page *page, bool compound)
> {
> - int i, nr = 1;
> + int i, nr = 0;
>
> VM_BUG_ON_PAGE(compound && !PageHead(page), page);
>
> @@ -1289,12 +1290,12 @@ static void page_remove_file_rmap(struct page *page, bool compound)
> if (compound && PageTransHuge(page)) {
> int nr_pages = thp_nr_pages(page);
>
> - for (i = 0, nr = 0; i < nr_pages; i++) {
> + for (i = 0; i < nr_pages; i++) {
> if (atomic_add_negative(-1, &page[i]._mapcount))
> nr++;
> }
> if (!atomic_add_negative(-1, compound_mapcount_ptr(page)))
> - return;
> + goto out;
> if (PageSwapBacked(page))
> __mod_lruvec_page_state(page, NR_SHMEM_PMDMAPPED,
> -nr_pages);
> @@ -1302,16 +1303,12 @@ static void page_remove_file_rmap(struct page *page, bool compound)
> __mod_lruvec_page_state(page, NR_FILE_PMDMAPPED,
> -nr_pages);
> } else {
> - if (!atomic_add_negative(-1, &page->_mapcount))
> - return;
> + if (atomic_add_negative(-1, &page->_mapcount))
> + nr++;
> }
> -
> - /*
> - * We use the irq-unsafe __{inc|mod}_lruvec_page_state because
> - * these counters are not modified in interrupt context, and
> - * pte lock(a spinlock) is held, which implies preemption disabled.
> - */
> - __mod_lruvec_page_state(page, NR_FILE_MAPPED, -nr);
> +out:
> + if (nr)
> + __mod_lruvec_page_state(page, NR_FILE_MAPPED, -nr);
> }
>
> static void page_remove_anon_compound_rmap(struct page *page)