Fixed mkdir when inserting into a non-end block

This was an oversight on my part when adding strict ordering to
directories. Unfortunately now we can't take advantage of the atomic
creation of tail+dir entries. Now we need to first create the tail, then
create the actually directory entry. If we lose power, the orphan is
cleaned up like orphans created during remove.

Note that we still take advantage of the atomic tail+dir entries if we
are an end block. This is actually because this corner case is
complicated to _not_ do atomically, needing to update the directory we
just committed to.

.travis.yml

@@ -126,7 +126,7 @@ jobs:
         - mkdir mount/littlefs
         - cp -r $(git ls-tree --name-only HEAD) mount/littlefs
         - cd mount/littlefs
-        - ls
+        - ls -flh
         - make -B test_dirs test_files QUIET=1
 
       # Automatically update releases

lfs.c

@@ -1746,29 +1746,58 @@ int lfs_mkdir(lfs_t *lfs, const char *path) {
 
     // build up new directory
     lfs_alloc_ack(lfs);
-
     lfs_mdir_t dir;
     err = lfs_dir_alloc(lfs, &dir);
     if (err) {
         return err;
     }
 
-    dir.tail[0] = cwd.tail[0];
-    dir.tail[1] = cwd.tail[1];
+    // find end of list
+    lfs_mdir_t pred = cwd;
+    while (pred.split) {
+        err = lfs_dir_fetch(lfs, &pred, pred.tail);
+        if (err) {
+            return err;
+        }
+    }
+
+    // setup dir
+    dir.tail[0] = pred.tail[0];
+    dir.tail[1] = pred.tail[1];
     err = lfs_dir_commit(lfs, &dir, NULL);
     if (err) {
         return err;
     }
 
-    // get next slot and commit
-    cwd.tail[0] = dir.pair[0];
-    cwd.tail[1] = dir.pair[1];
+    // current block end of list?
+    if (!cwd.split) {
+        // update atomically
+        cwd.tail[0] = dir.pair[0];
+        cwd.tail[1] = dir.pair[1];
+    } else {
+        // update tails, this creates a desync
+        pred.tail[0] = dir.pair[0];
+        pred.tail[1] = dir.pair[1];
+        lfs_global_orphans(lfs, +1);
+        err = lfs_dir_commit(lfs, &pred,
+                LFS_MKATTR(LFS_TYPE_SOFTTAIL, 0x1ff,
+                    pred.tail, sizeof(pred.tail),
+                NULL));
+        if (err) {
+            return err;
+        }
+        lfs_global_orphans(lfs, -1);
+    }
+
+    // now insert into our parent block
     lfs_pair_tole32(dir.pair);
     err = lfs_dir_commit(lfs, &cwd,
-            LFS_MKATTR(LFS_TYPE_SOFTTAIL, 0x1ff, cwd.tail, sizeof(cwd.tail),
             LFS_MKATTR(LFS_TYPE_DIRSTRUCT, id, dir.pair, sizeof(dir.pair),
             LFS_MKATTR(LFS_TYPE_DIR, id, path, nlen,
-            NULL))));
+            (!cwd.split)
+                ? LFS_MKATTR(LFS_TYPE_SOFTTAIL, 0x1ff,
+                    cwd.tail, sizeof(cwd.tail), NULL)
+                : NULL)));
     lfs_pair_fromle32(dir.pair);
     if (err) {
         return err;

tests/test_corrupt.sh

@@ -89,7 +89,7 @@ do
     rm -rf blocks
     mkdir blocks
     lfs_mktree
-    chmod a-w blocks/$b
+    chmod a-w blocks/$b || true
     lfs_mktree
     lfs_chktree
 done