Patch "libceph: don't allow bidirectional swap of pg-upmap-items" has been added to the 4.13-stable tree

This is a note to let you know that I've just added the patch titled

libceph: don't allow bidirectional swap of pg-upmap-items

to the 4.13-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
libceph-don-t-allow-bidirectional-swap-of-pg-upmap-items.patch
and it can be found in the queue-4.13 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

From: Ilya Dryomov <[email protected]>
Date: Mon, 18 Sep 2017 12:21:37 +0200
Subject: libceph: don't allow bidirectional swap of pg-upmap-items

From: Ilya Dryomov <[email protected]>

commit 29a0cfbf91ba997591535a4f7246835ce8328141 upstream.

This reverts most of commit f53b7665c8ce ("libceph: upmap semantic changes").

We need to prevent duplicates in the final result. For example, we can currently take

[1,2,3] and apply [(1,2)] and get [2,2,3]

or

[1,2,3] and apply [(3,2)] and get [1,2,2]

The rest of the system is not prepared to handle duplicates in the result set like this.

The reverted piece was intended to allow

[1,2,3] and [(1,2),(2,1)] to get [2,1,3]

to reorder primaries. First, this bidirectional swap is hard to implement in a way that also prevents dups. For example, [1,2,3] and [(1,4),(2,3),(3,4)] would give [4,3,4] but would we just drop the last step we'd have [4,3,3] which is also invalid, etc. Simpler to just not handle bidirectional swaps. In practice, they are not needed: if you just want to choose a different primary then use primary_affinity, or pg_upmap (not pg_upmap_items).

---
 net/ceph/osdmap.c |   33 ++++++++++++++++++++++++---------
 1 file changed, 24 insertions(+), 9 deletions(-)

@@ -2445,19 +2445,34 @@ static void apply_upmap(struct ceph_osdm
 
 	pg = lookup_pg_mapping(&osdmap->pg_upmap_items, pgid);
 	if (pg) {
-		for (i = 0; i < raw->size; i++) {
-			for (j = 0; j < pg->pg_upmap_items.len; j++) {
-				int from = pg->pg_upmap_items.from_to[j][0];
-				int to = pg->pg_upmap_items.from_to[j][1];
+		/*
+		 * Note: this approach does not allow a bidirectional swap,
+		 * e.g., [[1,2],[2,1]] applied to [0,1,2] -> [0,2,1].
+		 */
+		for (i = 0; i < pg->pg_upmap_items.len; i++) {
+			int from = pg->pg_upmap_items.from_to[i][0];
+			int to = pg->pg_upmap_items.from_to[i][1];
+			int pos = -1;
+			bool exists = false;
 
-				if (from == raw->osds[i]) {
-					if (!(to != CRUSH_ITEM_NONE &&
-					      to < osdmap->max_osd &&
-					      osdmap->osd_weight[to] == 0))
-						raw->osds[i] = to;
+			/* make sure replacement doesn't already appear */
+			for (j = 0; j < raw->size; j++) {
+				int osd = raw->osds[j];
+
+				if (osd == to) {
+					exists = true;
 					break;
 				}
+				/* ignore mapping if target is marked out */
+				if (osd == from && pos < 0 &&
+				    !(to != CRUSH_ITEM_NONE &&
+				      to < osdmap->max_osd &&
+				      osdmap->osd_weight[to] == 0)) {
+					pos = j;
+				}
 			}
+			if (!exists && pos >= 0)
+				raw->osds[pos] = to;
 		}
 	}
 }

Patches currently in stable-queue which might be from [email protected] are

queue-4.13/libceph-don-t-allow-bidirectional-swap-of-pg-upmap-items.patch