Re: [PATCH v2 1/8] mm: dump_page: print head page's refcount, for compound pages

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 1/29/20 10:23 PM, John Hubbard wrote:
On 1/29/20 2:59 PM, Matthew Wilcox wrote:
...
I have a hunk in my current tree which looks like this:

@@ -77,6 +77,11 @@ void __dump_page(struct page *page, const char *reason)
                 pr_warn("page:%px refcount:%d mapcount:%d mapping:%px index:%#lx
\n",
                         page, page_ref_count(page), mapcount,
                         page->mapping, page_to_pgoff(page));
+       if (PageTail(page)) {
+               struct page *head = compound_head(page);
+               pr_warn("head:%px mapping:%px index:%#lx\n",
+                       head, head->mapping, page_to_pgoff(head));
+       }
         if (PageKsm(page))
                 pr_warn("ksm flags: %#lx(%pGp)\n", page->flags, &page->flags);
         else if (PageAnon(page))

I wonder if we can combine these two patches in some more useful way?

I also think we probably want a sanity check that 'head' and 'page'
are within a sane range of each other (ie head < page and head +
MAX_ORDER_NR_PAGES > page) to protect against a struct page that contains
complete garbage.


OK, here's a go at combining those. I like the observation, implicit in your
diffs, that PageTail rather than PageCompound is the key differentiator in
deciding what to print. How's this look:

diff --git a/mm/debug.c b/mm/debug.c
index a90da5337c14..944652843e7b 100644
--- a/mm/debug.c
+++ b/mm/debug.c
@@ -75,12 +75,31 @@ void __dump_page(struct page *page, const char *reason)
       */
      mapcount = PageSlab(page) ? 0 : page_mapcount(page);

-    if (PageCompound(page))
-        pr_warn("page:%px refcount:%d mapcount:%d mapping:%px "
-            "index:%#lx compound_mapcount: %d\n",
-            page, page_ref_count(page), mapcount,
-            page->mapping, page_to_pgoff(page),
-            compound_mapcount(page));
+    if (PageTail(page)) {
+        struct page *head = compound_head(page);
+
+        if ((page < head) || (page >= head + MAX_ORDER_NR_PAGES)) {
+            /*
+             * Page is hopelessly corrupted, so limit any reporting
+             * to information about the page itself. Do not attempt
+             * to look at the head page.
+             */
+            pr_warn("page:%px refcount:%d mapcount:%d mapping:%px "
+                "index:%#lx (corrupted tail page case)\n",
+                page, page_ref_count(page), mapcount,
+                page->mapping, page_to_pgoff(page));
+        } else {
+            pr_warn("page:%px compound refcount:%d mapcount:%d "
+                "mapping:%px index:%#lx compound_mapcount:%d\n",
+                page, page_ref_count(head),
+                mapcount, head->mapping, page_to_pgoff(head),
+                compound_mapcount(page));
+
+            if (page_ref_count(page) != 0)
+                pr_warn("page:%px PROBLEM: non-zero refcount (==%d) on "
+                    "this tail page\n", page, page_ref_count(page));

...ahem, I sorta botched the above statement, because that should
be outside (just below) the "else" statement--it can be done whether or
not the page fails the safety/bounds check. :)

thanks,
--
John Hubbard
NVIDIA




[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux