mm: fix page reference leak in soft_offline_page()
authorDan Williams <dan.j.williams@intel.com>
Sun, 24 Jan 2021 05:01:52 +0000 (21:01 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 24 Jan 2021 18:34:52 +0000 (10:34 -0800)
The conversion to move pfn_to_online_page() internal to
soft_offline_page() missed that the get_user_pages() reference taken by
the madvise() path needs to be dropped when pfn_to_online_page() fails.

Note the direct sysfs-path to soft_offline_page() does not perform a
get_user_pages() lookup.

When soft_offline_page() is handed a pfn_valid() && !pfn_to_online_page()
pfn the kernel hangs at dax-device shutdown due to a leaked reference.

Link: https://lkml.kernel.org/r/161058501210.1840162.8108917599181157327.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: feec24a6139d ("mm, soft-offline: convert parameter to pfn")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memory-failure.c

index 04d9f154a130d28f38d26dfb0a897656d747cd71..e9481632fcd1b5e83fff6c87f4ad7354bb0b40bf 100644 (file)
@@ -1885,6 +1885,12 @@ static int soft_offline_free_page(struct page *page)
        return rc;
 }
 
+static void put_ref_page(struct page *page)
+{
+       if (page)
+               put_page(page);
+}
+
 /**
  * soft_offline_page - Soft offline a page.
  * @pfn: pfn to soft-offline
@@ -1910,20 +1916,26 @@ static int soft_offline_free_page(struct page *page)
 int soft_offline_page(unsigned long pfn, int flags)
 {
        int ret;
-       struct page *page;
        bool try_again = true;
+       struct page *page, *ref_page = NULL;
+
+       WARN_ON_ONCE(!pfn_valid(pfn) && (flags & MF_COUNT_INCREASED));
 
        if (!pfn_valid(pfn))
                return -ENXIO;
+       if (flags & MF_COUNT_INCREASED)
+               ref_page = pfn_to_page(pfn);
+
        /* Only online pages can be soft-offlined (esp., not ZONE_DEVICE). */
        page = pfn_to_online_page(pfn);
-       if (!page)
+       if (!page) {
+               put_ref_page(ref_page);
                return -EIO;
+       }
 
        if (PageHWPoison(page)) {
                pr_info("%s: %#lx page already poisoned\n", __func__, pfn);
-               if (flags & MF_COUNT_INCREASED)
-                       put_page(page);
+               put_ref_page(ref_page);
                return 0;
        }