[PATCH 3/4] mm: swapoff: take notice of completion sooner

From: Hugh Dickins
Date: Mon Apr 08 2019 - 15:59:38 EST


The old try_to_unuse() implementation was driven by find_next_to_unuse(),
which terminated as soon as all the swap had been freed. Add inuse_pages
checks now (alongside signal_pending()) to stop scanning mms and swap_map
once finished. The same ought to be done in shmem_unuse() too, but never
was before, and needs a different interface: so leave it as is for now.

Fixes: b56a2d8af914 ("mm: rid swapoff of quadratic complexity")
Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
---

mm/swapfile.c | 19 +++++++++++--------
1 file changed, 11 insertions(+), 8 deletions(-)

--- 5.1-rc4/mm/swapfile.c 2019-04-07 19:15:01.269054187 -0700
+++ linux/mm/swapfile.c 2019-04-07 19:17:13.291957539 -0700
@@ -2051,11 +2051,9 @@ retry:

spin_lock(&mmlist_lock);
p = &init_mm.mmlist;
- while ((p = p->next) != &init_mm.mmlist) {
- if (signal_pending(current)) {
- retval = -EINTR;
- break;
- }
+ while (si->inuse_pages &&
+ !signal_pending(current) &&
+ (p = p->next) != &init_mm.mmlist) {

mm = list_entry(p, struct mm_struct, mmlist);
if (!mmget_not_zero(mm))
@@ -2082,7 +2080,9 @@ retry:
mmput(prev_mm);

i = 0;
- while ((i = find_next_to_unuse(si, i, frontswap)) != 0) {
+ while (si->inuse_pages &&
+ !signal_pending(current) &&
+ (i = find_next_to_unuse(si, i, frontswap)) != 0) {

entry = swp_entry(type, i);
page = find_get_page(swap_address_space(entry), i);
@@ -2123,8 +2123,11 @@ retry:
* separate lists, and wait for those lists to be emptied; but it's
* easier and more robust (though cpu-intensive) just to keep retrying.
*/
- if (si->inuse_pages)
- goto retry;
+ if (si->inuse_pages) {
+ if (!signal_pending(current))
+ goto retry;
+ retval = -EINTR;
+ }
out:
return (retval == FRONTSWAP_PAGES_UNUSED) ? 0 : retval;
}