Prevented unnecessary superblock rewrites if old version in superblock chain

Because multiple, out-of-date superblocks can exist in our superblock
chain, we need to be careful to make sure newer superblock entries
override older superblock entries.

If we see an older on-disk minor version in the superblock chain, we
were correctly overriding the on-disk minor version, but we were also
leaving the "needs superblock" bit set in our consistency state.

This isn't a hard-error, but would lead to a superblock rewrite every
mount. The rewrite would make no progress, as the out-of-date version is
effectively immutable at this point, and just waste prog cycles.

This should fix that by clearing the "needs superblock" bit if we see a
newer on-disk minor version.

SPEC.md

@@ -441,9 +441,10 @@ Superblock fields:
 
 7. **Attr max (32-bits)** - Maximum size of file attributes in bytes.
 
-The superblock must always be the first entry (id 0) in a metadata pair as well
-as be the first entry written to the block. This means that the superblock
-entry can be read from a device using offsets alone.
+The superblock must always be the first entry (id 0) in the metadata pair, and
+the name tag must always be the first tag in the metadata pair. This makes it
+so that the magic string "littlefs" will always reside at offset=8 in a valid
+littlefs superblock.
 
 ---
 #### `0x2xx` LFS_TYPE_STRUCT

lfs.c

@@ -2191,7 +2191,8 @@ static int lfs_dir_splittingcompact(lfs_t *lfs, lfs_mdir_t *dir,
                 // we can do, we'll error later if we've become frozen
                 LFS_WARN("Unable to expand superblock");
             } else {
-                end = begin;
+                // duplicate the superblock entry into the new superblock
+                end = 1;
             }
         }
     }
@@ -2358,7 +2359,9 @@ fixmlist:;
 
             while (d->id >= d->m.count && d->m.split) {
                 // we split and id is on tail now
-                d->id -= d->m.count;
+                if (lfs_pair_cmp(d->m.tail, lfs->root) != 0) {
+                    d->id -= d->m.count;
+                }
                 int err = lfs_dir_fetch(lfs, &d->m, d->m.tail);
                 if (err) {
                     return err;
@@ -4466,6 +4469,7 @@ static int lfs_mount_(lfs_t *lfs, const struct lfs_config *cfg) {
             // found older minor version? set an in-device only bit in the
             // gstate so we know we need to rewrite the superblock before
             // the first write
+            bool needssuperblock = false;
             if (minor_version < lfs_fs_disk_version_minor(lfs)) {
                 LFS_DEBUG("Found older minor version "
                         "v%"PRIu16".%"PRIu16" < v%"PRIu16".%"PRIu16,
@@ -4473,10 +4477,11 @@ static int lfs_mount_(lfs_t *lfs, const struct lfs_config *cfg) {
                         minor_version,
                         lfs_fs_disk_version_major(lfs),
                         lfs_fs_disk_version_minor(lfs));
-                // note this bit is reserved on disk, so fetching more gstate
-                // will not interfere here
-                lfs_fs_prepsuperblock(lfs, true);
+                needssuperblock = true;
             }
+            // note this bit is reserved on disk, so fetching more gstate
+            // will not interfere here
+            lfs_fs_prepsuperblock(lfs, needssuperblock);
 
             // check superblock configuration
             if (superblock.name_max) {

tests/test_superblocks.toml

@@ -14,6 +14,24 @@ code = '''
     lfs_unmount(&lfs) => 0;
 '''
 
+# make sure the magic string "littlefs" is always at offset=8
+[cases.test_superblocks_magic]
+code = '''
+    lfs_t lfs;
+    lfs_format(&lfs, cfg) => 0;
+
+    // check our magic string
+    //
+    // note if we lose power we may not have the magic string in both blocks!
+    // but we don't lose power in this test so we can assert the magic string
+    // is present in both
+    uint8_t magic[lfs_max(16, READ_SIZE)];
+    cfg->read(cfg, 0, 0, magic, lfs_max(16, READ_SIZE)) => 0;
+    assert(memcmp(&magic[8], "littlefs", 8) == 0);
+    cfg->read(cfg, 1, 0, magic, lfs_max(16, READ_SIZE)) => 0;
+    assert(memcmp(&magic[8], "littlefs", 8) == 0);
+'''
+
 # mount/unmount from interpretting a previous superblock block_count
 [cases.test_superblocks_mount_unknown_block_count]
 code = '''
@@ -28,7 +46,6 @@ code = '''
     lfs_unmount(&lfs) => 0;
 '''
 
-
 # reentrant format
 [cases.test_superblocks_reentrant_format]
 reentrant = true
@@ -135,6 +152,39 @@ code = '''
     lfs_unmount(&lfs) => 0;
 '''
 
+# make sure the magic string "littlefs" is always at offset=8
+[cases.test_superblocks_magic_expand]
+defines.BLOCK_CYCLES = [32, 33, 1]
+defines.N = [10, 100, 1000]
+code = '''
+    lfs_t lfs;
+    lfs_format(&lfs, cfg) => 0;
+    lfs_mount(&lfs, cfg) => 0;
+    for (int i = 0; i < N; i++) {
+        lfs_file_t file;
+        lfs_file_open(&lfs, &file, "dummy",
+                LFS_O_WRONLY | LFS_O_CREAT | LFS_O_EXCL) => 0;
+        lfs_file_close(&lfs, &file) => 0;
+        struct lfs_info info;
+        lfs_stat(&lfs, "dummy", &info) => 0;
+        assert(strcmp(info.name, "dummy") == 0);
+        assert(info.type == LFS_TYPE_REG);
+        lfs_remove(&lfs, "dummy") => 0;
+    }
+    lfs_unmount(&lfs) => 0;
+
+    // check our magic string
+    //
+    // note if we lose power we may not have the magic string in both blocks!
+    // but we don't lose power in this test so we can assert the magic string
+    // is present in both
+    uint8_t magic[lfs_max(16, READ_SIZE)];
+    cfg->read(cfg, 0, 0, magic, lfs_max(16, READ_SIZE)) => 0;
+    assert(memcmp(&magic[8], "littlefs", 8) == 0);
+    cfg->read(cfg, 1, 0, magic, lfs_max(16, READ_SIZE)) => 0;
+    assert(memcmp(&magic[8], "littlefs", 8) == 0);
+'''
+
 # expanding superblock with power cycle
 [cases.test_superblocks_expand_power_cycle]
 defines.BLOCK_CYCLES = [32, 33, 1]
@@ -221,6 +271,7 @@ code = '''
     lfs_unmount(&lfs) => 0;
 '''
 
+
 # mount with unknown block_count
 [cases.test_superblocks_unknown_blocks]
 code = '''