Re: [bisected commit 0fc9d10] NFS-server corruption with 3.4

[Konstantin Khlebnikov]

Proper fix in attachment.

Konstantin Khlebnikov wrote:
> OGAWA Hirofumi wrote:
> > Konstantin Khlebnikov<khlebnikov@xxxxxxxxxx>   writes:
> >
> > > Hmm, very interesting!
> > > Please try this patch, it must fix the problem and print some numbers to debug.
> >
> > I think the bug is in radix_tree_for_each_contig().
> >
> > radix_tree_next_slot() returns NULL if the slot was NULL (i.e. there is
> > hole). But, slot == NULL is not meaning to stop iterate here. Actually,
> > if slot is NULL, it gets next chunk.
> >
> > Bang.
>
> Yeah, you are right, I already found this too.
> Currently I think how to fix this more accurately...
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

radix-tree: fix contiguous iterator

From: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxx>

This patch fixes bug in macro radix_tree_for_each_contig().
If radix_tree_next_slot() sees NULL in next slot it returns NULL, but following
radix_tree_next_chunk() switches iterating into next chunk. As result iterating
becomes non-contiguous and breaks vfs "splice" and all its users.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxx>
Reported-and-bisected-by: Hans de Bruin <jmdebruin@xxxxxxxxx>
Reported-and-bisected-by: Ondrej Zary <linux@xxxxxxxxxxxxxxxxxxxx>
Reported-and-bisected-by: Toralf Förster <toralf.foerster@xxxxxx>
---
 include/linux/radix-tree.h |    5 ++++-
 lib/radix-tree.c           |    3 +++
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 0d04cd6..ffc444c 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -368,8 +368,11 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 			iter->index++;
 			if (likely(*slot))
 				return slot;
-			if (flags & RADIX_TREE_ITER_CONTIG)
+			if (flags & RADIX_TREE_ITER_CONTIG) {
+				/* forbid switching to the next chunk */
+				iter->next_index = 0;
 				break;
+			}
 		}
 	}
 	return NULL;
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 86516f5..3ac50dc 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -673,6 +673,9 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 	 * during iterating; it can be zero only at the beginning.
 	 * And we cannot overflow iter->next_index in a single step,
 	 * because RADIX_TREE_MAP_SHIFT < BITS_PER_LONG.
+	 *
+	 * This condition also used by radix_tree_next_slot() to stop
+	 * contiguous iterating, and forbid swithing to the next chunk.
 	 */
 	index = iter->next_index;
 	if (!index && iter->index)