On Thu, Oct 22, 2015 at 10:21:36AM +0900, Minchan Kim wrote:
> Hello Hugh,
> 
> On Wed, Oct 21, 2015 at 05:59:59PM -0700, Hugh Dickins wrote:
> > On Thu, 22 Oct 2015, Minchan Kim wrote:
> > > 
> > > I added the code to check it and queued it again but I had another oops
> > > in this time but symptom is related to anon_vma, too.
> > > (kernel is based on recent mmotm + unconditional mkdirty for bug fix)
> > > It seems page_get_anon_vma returns NULL since the page was not page_mapped
> > > at that time but second check of page_mapped right before try_to_unmap seems
> > > to be true.
> > > 
> > > Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
> > > Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
> > > page:ffffea0001cfbfc0 count:3 mapcount:1 mapping:ffff88007f1b5f51 index:0x600000aff
> > > flags: 0x4000000000048019(locked|uptodate|dirty|swapcache|swapbacked)
> > > page dumped because: VM_BUG_ON_PAGE(PageAnon(page) && !PageKsm(page) && !anon_vma)
> > 
> > That's interesting, that's one I added in my page migration series.
> > Let me think on it, but it could well relate to the one you got before.
> 
> I will roll back to mm/madv_free-v4.3-rc5-mmotm-2015-10-15-15-20
> instead of next-20151021 to remove noise from your migration cleanup
> series and will test it again.
> If it is fixed, I will test again with your migration patchset, then.

I tested mmotm-2015-10-15-15-20 with test program I attach for a long time.
Therefore, there is no patchset from Hugh's migration patch in there.
And I added below debug code with request from Kirill to all test kernels.

diff --git a/mm/rmap.c b/mm/rmap.c
index ddfb9be72366..1c23b70b1f57 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -513,6 +513,13 @@ struct anon_vma *page_lock_anon_vma_read(struct page *page)
 
        anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
        root_anon_vma = READ_ONCE(anon_vma->root);
+
+       if (root_anon_vma == NULL) {
+               printk("anon_vma %p refcount %d\n", anon_vma,
+                       atomic_read(&anon_vma->refcount));
+               VM_BUG_ON_PAGE(1, page);
+       }
+
        if (down_read_trylock(&root_anon_vma->rwsem)) {
                /*
                 * If the page is still mapped, then this anon_vma is still


1. mmotm-2015-10-15-15-20 + kirill's pte_mkdirty

1st trial:
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
BUG: Bad rss-counter state mm:ffff88007f1ed780 idx:1 val:488
BUG: Bad rss-counter state mm:ffff88007f1ed780 idx:2 val:24

2nd trial:

Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
BUG: Bad rss-counter state mm:ffff8800a5cca680 idx:1 val:512
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS

2. mmotm-2015-10-15-15-20-no-madvise_free, IOW it means git head for
54bad5da4834 arm64: add pmd_[dirty|mkclean] for THP.

1st trial:
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
BUG: Bad rss-counter state mm:ffff88007f4c2d80 idx:1 val:511
BUG: Bad rss-counter state mm:ffff88007f4c2d80 idx:2 val:1

2nd trial:
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
Adding 4191228k swap on /dev/vda5.  Priority:-1 extents:1 across:4191228k FS
anon_vma ffff880000089aa0 refcount 0
page:ffffea0001a2ea40 count:3 mapcount:1 mapping:ffff880000089aa1 index:0x6000047a9

I tested it with KVM which guest system has 12 core and 3G memory.
In mmotm-2015-10-15-15-20-no-madvise_free, I tweaked test program does
madvise_dontneed intead of madvise_free via below patch

For the testing,

        gcc -o oops oops.c
        ./memcg_test.sh

I will be off from now on so please understand late response
but I hope my test program will reproduce it in your machine.

diff --git a/oops.c b/oops.c
index e50330a..c8298f8 100644
--- a/oops.c
+++ b/oops.c
@@ -8,7 +8,7 @@
 #include <errno.h>
 #include <signal.h>
 
-#define MADV_FREE 5
+#define MADV_FREE 4
 
 int pid;