z3fold: fix reclaim lock-ups

commit 6098d7e136692f9c6e23ae362c62ec822343e4d5 upstream.

Do not try to optimize in-page object layout while the page is under
reclaim.  This fixes lock-ups on reclaim and improves reclaim
performance at the same time.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20180430125800.444cae9706489f412ad12621@gmail.com
Signed-off-by: Vitaly Wool <vitaly.vul@sony.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: <Oleksiy.Avramchenko@sony.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/mm/z3fold.c b/mm/z3fold.c
index 39e1912..ddfb20c 100644 (file)
--- a/mm/z3fold.c
+++ b/mm/z3fold.c
@@ -144,7 +144,8 @@ enum z3fold_page_flags {
        PAGE_HEADLESS = 0,
        MIDDLE_CHUNK_MAPPED,
        NEEDS_COMPACTING,
-       PAGE_STALE
+       PAGE_STALE,
+       UNDER_RECLAIM
 };
 
 /*****************
@@ -173,6 +174,7 @@ static struct z3fold_header *init_z3fold_page(struct page *page,
        clear_bit(MIDDLE_CHUNK_MAPPED, &page->private);
        clear_bit(NEEDS_COMPACTING, &page->private);
        clear_bit(PAGE_STALE, &page->private);
+       clear_bit(UNDER_RECLAIM, &page->private);
 
        spin_lock_init(&zhdr->page_lock);
        kref_init(&zhdr->refcount);
@@ -748,6 +750,10 @@ static void z3fold_free(struct z3fold_pool *pool, unsigned long handle)
                atomic64_dec(&pool->pages_nr);
                return;
        }
+       if (test_bit(UNDER_RECLAIM, &page->private)) {
+               z3fold_page_unlock(zhdr);
+               return;
+       }
        if (test_and_set_bit(NEEDS_COMPACTING, &page->private)) {
                z3fold_page_unlock(zhdr);
                return;
@@ -832,6 +838,8 @@ static int z3fold_reclaim_page(struct z3fold_pool *pool, unsigned int retries)
                        kref_get(&zhdr->refcount);
                        list_del_init(&zhdr->buddy);
                        zhdr->cpu = -1;
+                       set_bit(UNDER_RECLAIM, &page->private);
+                       break;
                }
 
                list_del_init(&page->lru);
@@ -879,25 +887,35 @@ static int z3fold_reclaim_page(struct z3fold_pool *pool, unsigned int retries)
                                goto next;
                }
 next:
-               spin_lock(&pool->lock);
                if (test_bit(PAGE_HEADLESS, &page->private)) {
                        if (ret == 0) {
-                               spin_unlock(&pool->lock);
                                free_z3fold_page(page);
                                return 0;
                        }
-               } else if (kref_put(&zhdr->refcount, release_z3fold_page)) {
-                       atomic64_dec(&pool->pages_nr);
+                       spin_lock(&pool->lock);
+                       list_add(&page->lru, &pool->lru);
+                       spin_unlock(&pool->lock);
+               } else {
+                       z3fold_page_lock(zhdr);
+                       clear_bit(UNDER_RECLAIM, &page->private);
+                       if (kref_put(&zhdr->refcount,
+                                       release_z3fold_page_locked)) {
+                               atomic64_dec(&pool->pages_nr);
+                               return 0;
+                       }
+                       /*
+                        * if we are here, the page is still not completely
+                        * free. Take the global pool lock then to be able
+                        * to add it back to the lru list
+                        */
+                       spin_lock(&pool->lock);
+                       list_add(&page->lru, &pool->lru);
                        spin_unlock(&pool->lock);
-                       return 0;
+                       z3fold_page_unlock(zhdr);
                }
 
-               /*
-                * Add to the beginning of LRU.
-                * Pool lock has to be kept here to ensure the page has
-                * not already been released
-                */
-               list_add(&page->lru, &pool->lru);
+               /* We started off locked to we need to lock the pool back */
+               spin_lock(&pool->lock);
        }
        spin_unlock(&pool->lock);
        return -EAGAIN;