mm: hwpoison: refactor refcount check handling

Memory failure will report failure if the page still has extra pinned
refcount other than from hwpoison after the handler is done.  Actually
the check is not necessary for all handlers, so move the check into
specific handlers.  This would make the following keeping shmem page in
page cache patch easier.

There may be expected extra pin for some cases, for example, when the
page is dirty and in swapcache.

Link: https://lkml.kernel.org/r/20211020210755.23964-5-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Suggested-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 64 insertions, 29 deletions

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 54a5d28ea601..bbd433e96790 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -807,12 +807,44 @@ static int truncate_error_page(struct page *p, unsigned long pfn,
 	return ret;
 }
 
+struct page_state {
+	unsigned long mask;
+	unsigned long res;
+	enum mf_action_page_type type;
+
+	/* Callback ->action() has to unlock the relevant page inside it. */
+	int (*action)(struct page_state *ps, struct page *p);
+};
+
+/*
+ * Return true if page is still referenced by others, otherwise return
+ * false.
+ *
+ * The extra_pins is true when one extra refcount is expected.
+ */
+static bool has_extra_refcount(struct page_state *ps, struct page *p,
+			       bool extra_pins)
+{
+	int count = page_count(p) - 1;
+
+	if (extra_pins)
+		count -= 1;
+
+	if (count > 0) {
+		pr_err("Memory failure: %#lx: %s still referenced by %d users\n",
+		       page_to_pfn(p), action_page_types[ps->type], count);
+		return true;
+	}
+
+	return false;
+}
+
 /*
  * Error hit kernel page.
  * Do nothing, try to be lucky and not touch this instead. For a few cases we
  * could be more sophisticated.
  */
-static int me_kernel(struct page *p, unsigned long pfn)
+static int me_kernel(struct page_state *ps, struct page *p)
 {
 	unlock_page(p);
 	return MF_IGNORED;
@@ -821,9 +853,9 @@ static int me_kernel(struct page *p, unsigned long pfn)
 /*
  * Page in unknown state. Do nothing.
  */
-static int me_unknown(struct page *p, unsigned long pfn)
+static int me_unknown(struct page_state *ps, struct page *p)
 {
-	pr_err("Memory failure: %#lx: Unknown page state\n", pfn);
+	pr_err("Memory failure: %#lx: Unknown page state\n", page_to_pfn(p));
 	unlock_page(p);
 	return MF_FAILED;
 }
@@ -831,7 +863,7 @@ static int me_unknown(struct page *p, unsigned long pfn)
 /*
  * Clean (or cleaned) page cache page.
  */
-static int me_pagecache_clean(struct page *p, unsigned long pfn)
+static int me_pagecache_clean(struct page_state *ps, struct page *p)
 {
 	int ret;
 	struct address_space *mapping;
@@ -868,9 +900,13 @@ static int me_pagecache_clean(struct page *p, unsigned long pfn)
 	 *
 	 * Open: to take i_rwsem or not for this? Right now we don't.
 	 */
-	ret = truncate_error_page(p, pfn, mapping);
+	ret = truncate_error_page(p, page_to_pfn(p), mapping);
 out:
 	unlock_page(p);
+
+	if (has_extra_refcount(ps, p, false))
+		ret = MF_FAILED;
+
 	return ret;
 }
 
@@ -879,7 +915,7 @@ out:
  * Issues: when the error hit a hole page the error is not properly
  * propagated.
  */
-static int me_pagecache_dirty(struct page *p, unsigned long pfn)
+static int me_pagecache_dirty(struct page_state *ps, struct page *p)
 {
 	struct address_space *mapping = page_mapping(p);
 
@@ -923,7 +959,7 @@ static int me_pagecache_dirty(struct page *p, unsigned long pfn)
 		mapping_set_error(mapping, -EIO);
 	}
 
-	return me_pagecache_clean(p, pfn);
+	return me_pagecache_clean(ps, p);
 }
 
 /*
@@ -945,9 +981,10 @@ static int me_pagecache_dirty(struct page *p, unsigned long pfn)
  * Clean swap cache pages can be directly isolated. A later page fault will
  * bring in the known good data from disk.
  */
-static int me_swapcache_dirty(struct page *p, unsigned long pfn)
+static int me_swapcache_dirty(struct page_state *ps, struct page *p)
 {
 	int ret;
+	bool extra_pins = false;
 
 	ClearPageDirty(p);
 	/* Trigger EIO in shmem: */
@@ -955,10 +992,17 @@ static int me_swapcache_dirty(struct page *p, unsigned long pfn)
 
 	ret = delete_from_lru_cache(p) ? MF_FAILED : MF_DELAYED;
 	unlock_page(p);
+
+	if (ret == MF_DELAYED)
+		extra_pins = true;
+
+	if (has_extra_refcount(ps, p, extra_pins))
+		ret = MF_FAILED;
+
 	return ret;
 }
 
-static int me_swapcache_clean(struct page *p, unsigned long pfn)
+static int me_swapcache_clean(struct page_state *ps, struct page *p)
 {
 	int ret;
 
@@ -966,6 +1010,10 @@ static int me_swapcache_clean(struct page *p, unsigned long pfn)
 
 	ret = delete_from_lru_cache(p) ? MF_FAILED : MF_RECOVERED;
 	unlock_page(p);
+
+	if (has_extra_refcount(ps, p, false))
+		ret = MF_FAILED;
+
 	return ret;
 }
 
@@ -975,7 +1023,7 @@ static int me_swapcache_clean(struct page *p, unsigned long pfn)
  * - Error on hugepage is contained in hugepage unit (not in raw page unit.)
  *   To narrow down kill region to one page, we need to break up pmd.
  */
-static int me_huge_page(struct page *p, unsigned long pfn)
+static int me_huge_page(struct page_state *ps, struct page *p)
 {
 	int res;
 	struct page *hpage = compound_head(p);
@@ -986,7 +1034,7 @@ static int me_huge_page(struct page *p, unsigned long pfn)
 
 	mapping = page_mapping(hpage);
 	if (mapping) {
-		res = truncate_error_page(hpage, pfn, mapping);
+		res = truncate_error_page(hpage, page_to_pfn(p), mapping);
 		unlock_page(hpage);
 	} else {
 		res = MF_FAILED;
@@ -1004,6 +1052,9 @@ static int me_huge_page(struct page *p, unsigned long pfn)
 		}
 	}
 
+	if (has_extra_refcount(ps, p, false))
+		res = MF_FAILED;
+
 	return res;
 }
 
@@ -1029,14 +1080,7 @@ static int me_huge_page(struct page *p, unsigned long pfn)
 #define slab		(1UL << PG_slab)
 #define reserved	(1UL << PG_reserved)
 
-static struct page_state {
-	unsigned long mask;
-	unsigned long res;
-	enum mf_action_page_type type;
-
-	/* Callback ->action() has to unlock the relevant page inside it. */
-	int (*action)(struct page *p, unsigned long pfn);
-} error_states[] = {
+static struct page_state error_states[] = {
 	{ reserved,	reserved,	MF_MSG_KERNEL,	me_kernel },
 	/*
 	 * free pages are specially detected outside this table:
@@ -1096,19 +1140,10 @@ static int page_action(struct page_state *ps, struct page *p,
 			unsigned long pfn)
 {
 	int result;
-	int count;
 
 	/* page p should be unlocked after returning from ps->action().  */
-	result = ps->action(p, pfn);
+	result = ps->action(ps, p);
 
-	count = page_count(p) - 1;
-	if (ps->action == me_swapcache_dirty && result == MF_DELAYED)
-		count--;
-	if (count > 0) {
-		pr_err("Memory failure: %#lx: %s still referenced by %d users\n",
-		       pfn, action_page_types[ps->type], count);
-		result = MF_FAILED;
-	}
 	action_result(pfn, ps->type, result);
 
 	/* Could do more checks here if page looks ok */