Tweaked wording of littlefs2-rust link in README.md

Add "chamelon" to the related projects section.

.github/workflows/test.yml

@@ -473,6 +473,42 @@ jobs:
           path: status
           retention-days: 1
 
+  # run compatibility tests using the current master as the previous version
+  test-compat:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v2
+        if: ${{github.event_name == 'pull_request'}}
+      # checkout the current pr target into lfsp
+      - uses: actions/checkout@v2
+        if: ${{github.event_name == 'pull_request'}}
+        with:
+          ref: ${{github.event.pull_request.base.ref}}
+          path: lfsp
+      - name: install
+        if: ${{github.event_name == 'pull_request'}}
+        run: |
+          # need a few things
+          sudo apt-get update -qq
+          sudo apt-get install -qq gcc python3 python3-pip
+          pip3 install toml
+          gcc --version
+          python3 --version
+      # adjust prefix of lfsp
+      - name: changeprefix
+        if: ${{github.event_name == 'pull_request'}}
+        run: |
+          ./scripts/changeprefix.py lfs lfsp lfsp/*.h lfsp/*.c
+      - name: test-compat
+        if: ${{github.event_name == 'pull_request'}}
+        run: |
+          TESTS=tests/test_compat.toml \
+            SRC="$(find . lfsp -name '*.c' -maxdepth 1 \
+                -and -not -name '*.t.*' \
+                -and -not -name '*.b.*')" \
+            CFLAGS="-DLFSP=lfsp/lfsp.h" \
+            make test
+
   # self-host with littlefs-fuse for a fuzz-like test
   fuse:
     runs-on: ubuntu-22.04

Makefile

@@ -1,15 +1,5 @@
-ifdef BUILDDIR
-# bit of a hack, but we want to make sure BUILDDIR directory structure
-# is correct before any commands
-$(if $(findstring n,$(MAKEFLAGS)),, $(shell mkdir -p \
-	$(BUILDDIR)/ \
-	$(BUILDDIR)/bd \
-	$(BUILDDIR)/runners \
-	$(BUILDDIR)/tests \
-	$(BUILDDIR)/benches))
-endif
+# overrideable build dir, default is in-place
 BUILDDIR ?= .
-
 # overridable target/src/tools/flags/etc
 ifneq ($(wildcard test.c main.c),)
 TARGET ?= $(BUILDDIR)/lfs
@@ -163,6 +153,18 @@ TESTFLAGS  += --perf-path="$(PERF)"
 BENCHFLAGS += --perf-path="$(PERF)"
 endif
 
+# this is a bit of a hack, but we want to make sure the BUILDDIR
+# directory structure is correct before we run any commands
+ifneq ($(BUILDDIR),.)
+$(if $(findstring n,$(MAKEFLAGS)),, $(shell mkdir -p \
+	$(addprefix $(BUILDDIR)/,$(dir \
+		$(SRC) \
+		$(TESTS) \
+		$(TEST_SRC) \
+		$(BENCHES) \
+		$(BENCH_SRC)))))
+endif
+
 
 # commands
 
@@ -514,6 +516,9 @@ $(BUILDDIR)/runners/bench_runner: $(BENCH_OBJ)
 $(BUILDDIR)/%.o $(BUILDDIR)/%.ci: %.c
 	$(CC) -c -MMD $(CFLAGS) $< -o $(BUILDDIR)/$*.o
 
+$(BUILDDIR)/%.o $(BUILDDIR)/%.ci: $(BUILDDIR)/%.c
+	$(CC) -c -MMD $(CFLAGS) $< -o $(BUILDDIR)/$*.o
+
 $(BUILDDIR)/%.s: %.c
 	$(CC) -S $(CFLAGS) $< -o $@
 

README.md

@@ -226,6 +226,13 @@ License Identifiers that are here available: http://spdx.org/licenses/
   to create images of the filesystem on your PC. Check if littlefs will fit
   your needs, create images for a later download to the target memory or
   inspect the content of a binary image of the target memory.
+  
+- [littlefs2-rust] - A Rust wrapper for littlefs. This project allows you
+  to use littlefs in a Rust-friendly API, reaping the benefits of Rust's memory
+  safety and other guarantees.
+
+- [littlefs-disk-img-viewer] - A memory-efficient web application for viewing
+  littlefs disk images in your web browser.
 
 - [mklfs] - A command line tool built by the [Lua RTOS] guys for making
   littlefs images from a host PC. Supports Windows, Mac OS, and Linux.
@@ -243,8 +250,12 @@ License Identifiers that are here available: http://spdx.org/licenses/
   MCUs. It offers static wear-leveling and power-resilience with only a fixed
   _O(|address|)_ pointer structure stored on each block and in RAM.
 
+- [chamelon] - A pure-OCaml implementation of (most of) littlefs, designed for
+  use with the MirageOS library operating system project. It is interoperable
+  with the reference implementation, with some caveats.
 
 [BSD-3-Clause]: https://spdx.org/licenses/BSD-3-Clause.html
+[littlefs-disk-img-viewer]: https://github.com/tniessen/littlefs-disk-img-viewer
 [littlefs-fuse]: https://github.com/geky/littlefs-fuse
 [FUSE]: https://github.com/libfuse/libfuse
 [littlefs-js]: https://github.com/geky/littlefs-js
@@ -256,3 +267,5 @@ License Identifiers that are here available: http://spdx.org/licenses/
 [SPIFFS]: https://github.com/pellepl/spiffs
 [Dhara]: https://github.com/dlbeer/dhara
 [littlefs-python]: https://pypi.org/project/littlefs-python/
+[littlefs2-rust]: https://crates.io/crates/littlefs2
+[chamelon]: https://github.com/yomimono/chamelon

SPEC.md

@@ -1,10 +1,10 @@
 ## littlefs technical specification
 
-This is the technical specification of the little filesystem. This document
-covers the technical details of how the littlefs is stored on disk for
-introspection and tooling. This document assumes you are familiar with the
-design of the littlefs, for more info on how littlefs works check
-out [DESIGN.md](DESIGN.md).
+This is the technical specification of the little filesystem with on-disk
+version lfs2.1. This document covers the technical details of how the littlefs
+is stored on disk for introspection and tooling. This document assumes you are
+familiar with the design of the littlefs, for more info on how littlefs works
+check out [DESIGN.md](DESIGN.md).
 
 ```
    | | |     .---._____
@@ -133,12 +133,6 @@ tags XORed together, starting with `0xffffffff`.
 '-------------------'                    '-------------------'
 ```
 
-One last thing to note before we get into the details around tag encoding. Each
-tag contains a valid bit used to indicate if the tag and containing commit is
-valid. This valid bit is the first bit found in the tag and the commit and can
-be used to tell if we've attempted to write to the remaining space in the
-block.
-
 Here's a more complete example of metadata block containing 4 entries:
 
 ```
@@ -191,6 +185,53 @@ Here's a more complete example of metadata block containing 4 entries:
                                              '---- most recent D
 ```
 
+Two things to note before we get into the details around tag encoding:
+
+1. Each tag contains a valid bit used to indicate if the tag and containing
+   commit is valid. After XORing, this bit should always be zero.
+
+   At the end of each commit, the valid bit of the previous tag is XORed
+   with the lowest bit in the type field of the CRC tag. This allows
+   the CRC tag to force the next commit to fail the valid bit test if it
+   has not yet been written to.
+
+2. The valid bit alone is not enough info to know if the next commit has been
+   erased. We don't know the order bits will be programmed in a program block,
+   so it's possible that the next commit had an attempted program that left the
+   valid bit unchanged.
+
+   To ensure we only ever program erased bytes, each commit can contain an
+   optional forward-CRC (FCRC). An FCRC contains a checksum of some amount of
+   bytes in the next commit at the time it was erased.
+
+   ```
+   .-------------------. \      \
+   |  revision count   | |      |
+   |-------------------| |      |
+   |     metadata      | |      |
+   |                   | +---.  +-- current commit
+   |                   | |   |  |
+   |-------------------| |   |  |
+   |       FCRC       ---|-. |  |
+   |-------------------| / | |  |
+   |        CRC       -----|-'  /
+   |-------------------|   |
+   |      padding      |   |        padding (does't need CRC)
+   |                   |   |
+   |-------------------| \ |    \
+   |      erased?      | +-'    |
+   |         |         | |      +-- next commit
+   |         v         | /      |
+   |                   |        /
+   |                   |
+   '-------------------'
+   ```
+
+   If the FCRC is missing or the checksum does not match, we must assume a
+   commit was attempted but failed due to power-loss.
+
+   Note that end-of-block commits do not need an FCRC.
+
 ## Metadata tags
 
 So in littlefs, 32-bit tags describe every type of metadata. And this means
@@ -785,3 +826,41 @@ CRC fields:
    are made about the contents.
 
 ---
+#### `0x5ff` LFS_TYPE_FCRC
+
+Added in lfs2.1, the optional FCRC tag contains a checksum of some amount of
+bytes in the next commit at the time it was erased. This allows us to ensure
+that we only ever program erased bytes, even if a previous commit failed due
+to power-loss.
+
+When programming a commit, the FCRC size must be at least as large as the
+program block size. However, the program block is not saved on disk, and can
+change between mounts, so the FCRC size on disk may be different than the
+current program block size.
+
+If the FCRC is missing or the checksum does not match, we must assume a
+commit was attempted but failed due to power-loss.
+
+Layout of the FCRC tag:
+
+```
+        tag                          data
+[--      32      --][--      32      --|--      32      --]
+[1|- 11 -| 10 | 10 ][--      32      --|--      32      --]
+ ^    ^     ^    ^            ^- fcrc size       ^- fcrc
+ |    |     |    '- size (8)
+ |    |     '------ id (0x3ff)
+ |    '------------ type (0x5ff)
+ '----------------- valid bit
+```
+
+FCRC fields:
+
+1. **FCRC size (32-bits)** - Number of bytes after this commit's CRC tag's
+   padding to include in the FCRC.
+
+2. **FCRC (32-bits)** - CRC of the bytes after this commit's CRC tag's padding
+   when erased. Like the CRC tag, this uses a CRC-32 with a polynomial of
+   `0x04c11db7` initialized with `0xffffffff`.
+
+---

lfs.c

@@ -300,14 +300,12 @@ static inline int lfs_pair_cmp(
              paira[0] == pairb[1] || paira[1] == pairb[0]);
 }
 
-#ifndef LFS_READONLY
-static inline bool lfs_pair_sync(
+static inline bool lfs_pair_issync(
         const lfs_block_t paira[2],
         const lfs_block_t pairb[2]) {
     return (paira[0] == pairb[0] && paira[1] == pairb[1]) ||
            (paira[0] == pairb[1] && paira[1] == pairb[0]);
 }
-#endif
 
 static inline void lfs_pair_fromle32(lfs_block_t pair[2]) {
     pair[0] = lfs_fromle32(pair[0]);
@@ -411,12 +409,16 @@ static inline bool lfs_gstate_hasorphans(const lfs_gstate_t *a) {
 }
 
 static inline uint8_t lfs_gstate_getorphans(const lfs_gstate_t *a) {
-    return lfs_tag_size(a->tag);
+    return lfs_tag_size(a->tag) & 0x1ff;
 }
 
 static inline bool lfs_gstate_hasmove(const lfs_gstate_t *a) {
     return lfs_tag_type1(a->tag);
 }
+
+static inline bool lfs_gstate_needssuperblock(const lfs_gstate_t *a) {
+    return lfs_tag_size(a->tag) >> 9;
+}
 #endif
 
 static inline bool lfs_gstate_hasmovehere(const lfs_gstate_t *a,
@@ -533,6 +535,7 @@ static int lfs_file_outline(lfs_t *lfs, lfs_file_t *file);
 static int lfs_file_flush(lfs_t *lfs, lfs_file_t *file);
 
 static int lfs_fs_deorphan(lfs_t *lfs, bool powerloss);
+static void lfs_fs_prepsuperblock(lfs_t *lfs, bool needssuperblock);
 static int lfs_fs_preporphans(lfs_t *lfs, int8_t orphans);
 static void lfs_fs_prepmove(lfs_t *lfs,
         uint16_t id, const lfs_block_t pair[2]);
@@ -4215,14 +4218,23 @@ static int lfs_rawmount(lfs_t *lfs, const struct lfs_config *cfg) {
 
     // scan directory blocks for superblock and any global updates
     lfs_mdir_t dir = {.tail = {0, 1}};
-    lfs_block_t cycle = 0;
+    lfs_block_t tortoise[2] = {LFS_BLOCK_NULL, LFS_BLOCK_NULL};
+    lfs_size_t tortoise_i = 1;
+    lfs_size_t tortoise_period = 1;
     while (!lfs_pair_isnull(dir.tail)) {
-        if (cycle >= lfs->cfg->block_count/2) {
-            // loop detected
+        // detect cycles with Brent's algorithm
+        if (lfs_pair_issync(dir.tail, tortoise)) {
+            LFS_WARN("Cycle detected in tail list");
             err = LFS_ERR_CORRUPT;
             goto cleanup;
         }
-        cycle += 1;
+        if (tortoise_i == tortoise_period) {
+            tortoise[0] = dir.tail[0];
+            tortoise[1] = dir.tail[1];
+            tortoise_i = 0;
+            tortoise_period *= 2;
+        }
+        tortoise_i += 1;
 
         // fetch next block in tail list
         lfs_stag_t tag = lfs_dir_fetchmatch(lfs, &dir, dir.tail,
@@ -4258,12 +4270,29 @@ static int lfs_rawmount(lfs_t *lfs, const struct lfs_config *cfg) {
             uint16_t minor_version = (0xffff & (superblock.version >>  0));
             if ((major_version != LFS_DISK_VERSION_MAJOR ||
                  minor_version > LFS_DISK_VERSION_MINOR)) {
-                LFS_ERROR("Invalid version v%"PRIu16".%"PRIu16,
-                        major_version, minor_version);
+                LFS_ERROR("Invalid version "
+                        "v%"PRIu16".%"PRIu16" != v%"PRIu16".%"PRIu16,
+                        major_version, minor_version,
+                        LFS_DISK_VERSION_MAJOR, LFS_DISK_VERSION_MINOR);
                 err = LFS_ERR_INVAL;
                 goto cleanup;
             }
 
+            // 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
+            if (minor_version < LFS_DISK_VERSION_MINOR) {
+                LFS_DEBUG("Found older minor version "
+                        "v%"PRIu16".%"PRIu16" < v%"PRIu16".%"PRIu16,
+                        major_version, minor_version,
+                        LFS_DISK_VERSION_MAJOR, LFS_DISK_VERSION_MINOR);
+            #ifndef LFS_READONLY
+                // note this bit is reserved on disk, so fetching more gstate
+                // will not interfere here
+                lfs_fs_prepsuperblock(lfs, true);
+            #endif
+            }
+
             // check superblock configuration
             if (superblock.name_max) {
                 if (superblock.name_max > lfs->name_max) {
@@ -4373,13 +4402,22 @@ int lfs_fs_rawtraverse(lfs_t *lfs,
     }
 #endif
 
-    lfs_block_t cycle = 0;
+    lfs_block_t tortoise[2] = {LFS_BLOCK_NULL, LFS_BLOCK_NULL};
+    lfs_size_t tortoise_i = 1;
+    lfs_size_t tortoise_period = 1;
     while (!lfs_pair_isnull(dir.tail)) {
-        if (cycle >= lfs->cfg->block_count/2) {
-            // loop detected
+        // detect cycles with Brent's algorithm
+        if (lfs_pair_issync(dir.tail, tortoise)) {
+            LFS_WARN("Cycle detected in tail list");
             return LFS_ERR_CORRUPT;
         }
-        cycle += 1;
+        if (tortoise_i == tortoise_period) {
+            tortoise[0] = dir.tail[0];
+            tortoise[1] = dir.tail[1];
+            tortoise_i = 0;
+            tortoise_period *= 2;
+        }
+        tortoise_i += 1;
 
         for (int i = 0; i < 2; i++) {
             int err = cb(data, dir.tail[i]);
@@ -4458,13 +4496,22 @@ static int lfs_fs_pred(lfs_t *lfs,
     // iterate over all directory directory entries
     pdir->tail[0] = 0;
     pdir->tail[1] = 1;
-    lfs_block_t cycle = 0;
+    lfs_block_t tortoise[2] = {LFS_BLOCK_NULL, LFS_BLOCK_NULL};
+    lfs_size_t tortoise_i = 1;
+    lfs_size_t tortoise_period = 1;
     while (!lfs_pair_isnull(pdir->tail)) {
-        if (cycle >= lfs->cfg->block_count/2) {
-            // loop detected
+        // detect cycles with Brent's algorithm
+        if (lfs_pair_issync(pdir->tail, tortoise)) {
+            LFS_WARN("Cycle detected in tail list");
             return LFS_ERR_CORRUPT;
         }
-        cycle += 1;
+        if (tortoise_i == tortoise_period) {
+            tortoise[0] = pdir->tail[0];
+            tortoise[1] = pdir->tail[1];
+            tortoise_i = 0;
+            tortoise_period *= 2;
+        }
+        tortoise_i += 1;
 
         if (lfs_pair_cmp(pdir->tail, pair) == 0) {
             return 0;
@@ -4514,13 +4561,22 @@ static lfs_stag_t lfs_fs_parent(lfs_t *lfs, const lfs_block_t pair[2],
     // use fetchmatch with callback to find pairs
     parent->tail[0] = 0;
     parent->tail[1] = 1;
-    lfs_block_t cycle = 0;
+    lfs_block_t tortoise[2] = {LFS_BLOCK_NULL, LFS_BLOCK_NULL};
+    lfs_size_t tortoise_i = 1;
+    lfs_size_t tortoise_period = 1;
     while (!lfs_pair_isnull(parent->tail)) {
-        if (cycle >= lfs->cfg->block_count/2) {
-            // loop detected
+        // detect cycles with Brent's algorithm
+        if (lfs_pair_issync(parent->tail, tortoise)) {
+            LFS_WARN("Cycle detected in tail list");
             return LFS_ERR_CORRUPT;
         }
-        cycle += 1;
+        if (tortoise_i == tortoise_period) {
+            tortoise[0] = parent->tail[0];
+            tortoise[1] = parent->tail[1];
+            tortoise_i = 0;
+            tortoise_period *= 2;
+        }
+        tortoise_i += 1;
 
         lfs_stag_t tag = lfs_dir_fetchmatch(lfs, parent, parent->tail,
                 LFS_MKTAG(0x7ff, 0, 0x3ff),
@@ -4537,10 +4593,17 @@ static lfs_stag_t lfs_fs_parent(lfs_t *lfs, const lfs_block_t pair[2],
 }
 #endif
 
+#ifndef LFS_READONLY
+static void lfs_fs_prepsuperblock(lfs_t *lfs, bool needssuperblock) {
+    lfs->gstate.tag = (lfs->gstate.tag & ~LFS_MKTAG(0, 0, 0x200))
+            | (uint32_t)needssuperblock << 9;
+}
+#endif
+
 #ifndef LFS_READONLY
 static int lfs_fs_preporphans(lfs_t *lfs, int8_t orphans) {
     LFS_ASSERT(lfs_tag_size(lfs->gstate.tag) > 0x000 || orphans >= 0);
-    LFS_ASSERT(lfs_tag_size(lfs->gstate.tag) < 0x3ff || orphans <= 0);
+    LFS_ASSERT(lfs_tag_size(lfs->gstate.tag) < 0x1ff || orphans <= 0);
     lfs->gstate.tag += orphans;
     lfs->gstate.tag = ((lfs->gstate.tag & ~LFS_MKTAG(0x800, 0, 0)) |
             ((uint32_t)lfs_gstate_hasorphans(&lfs->gstate) << 31));
@@ -4559,6 +4622,45 @@ static void lfs_fs_prepmove(lfs_t *lfs,
 }
 #endif
 
+#ifndef LFS_READONLY
+static int lfs_fs_desuperblock(lfs_t *lfs) {
+    if (!lfs_gstate_needssuperblock(&lfs->gstate)) {
+        return 0;
+    }
+
+    LFS_DEBUG("Rewriting superblock {0x%"PRIx32", 0x%"PRIx32"}",
+            lfs->root[0],
+            lfs->root[1]);
+
+    lfs_mdir_t root;
+    int err = lfs_dir_fetch(lfs, &root, lfs->root);
+    if (err) {
+        return err;
+    }
+
+    // write a new superblock
+    lfs_superblock_t superblock = {
+        .version     = LFS_DISK_VERSION,
+        .block_size  = lfs->cfg->block_size,
+        .block_count = lfs->cfg->block_count,
+        .name_max    = lfs->name_max,
+        .file_max    = lfs->file_max,
+        .attr_max    = lfs->attr_max,
+    };
+
+    lfs_superblock_tole32(&superblock);
+    err = lfs_dir_commit(lfs, &root, LFS_MKATTRS(
+            {LFS_MKTAG(LFS_TYPE_INLINESTRUCT, 0, sizeof(superblock)),
+                &superblock}));
+    if (err) {
+        return err;
+    }
+
+    lfs_fs_prepsuperblock(lfs, false);
+    return 0;
+}
+#endif
+
 #ifndef LFS_READONLY
 static int lfs_fs_demove(lfs_t *lfs) {
     if (!lfs_gstate_hasmove(&lfs->gdisk)) {
@@ -4643,7 +4745,7 @@ static int lfs_fs_deorphan(lfs_t *lfs, bool powerloss) {
                     }
                     lfs_pair_fromle32(pair);
 
-                    if (!lfs_pair_sync(pair, pdir.tail)) {
+                    if (!lfs_pair_issync(pair, pdir.tail)) {
                         // we have desynced
                         LFS_DEBUG("Fixing half-orphan "
                                 "{0x%"PRIx32", 0x%"PRIx32"} "
@@ -4736,7 +4838,12 @@ static int lfs_fs_deorphan(lfs_t *lfs, bool powerloss) {
 
 #ifndef LFS_READONLY
 static int lfs_fs_forceconsistency(lfs_t *lfs) {
-    int err = lfs_fs_demove(lfs);
+    int err = lfs_fs_desuperblock(lfs);
+    if (err) {
+        return err;
+    }
+
+    err = lfs_fs_demove(lfs);
     if (err) {
         return err;
     }

lfs.h

@@ -21,14 +21,14 @@ extern "C"
 // Software library version
 // Major (top-nibble), incremented on backwards incompatible changes
 // Minor (bottom-nibble), incremented on feature additions
-#define LFS_VERSION 0x00020005
+#define LFS_VERSION 0x00020006
 #define LFS_VERSION_MAJOR (0xffff & (LFS_VERSION >> 16))
 #define LFS_VERSION_MINOR (0xffff & (LFS_VERSION >>  0))
 
 // Version of On-disk data structures
 // Major (top-nibble), incremented on backwards incompatible changes
 // Minor (bottom-nibble), incremented on feature additions
-#define LFS_DISK_VERSION 0x00020000
+#define LFS_DISK_VERSION 0x00020001
 #define LFS_DISK_VERSION_MAJOR (0xffff & (LFS_DISK_VERSION >> 16))
 #define LFS_DISK_VERSION_MINOR (0xffff & (LFS_DISK_VERSION >>  0))
 

scripts/changeprefix.py

@@ -107,7 +107,10 @@ def main(from_prefix, to_prefix, paths=[], *,
         elif no_renames:
             to_path = from_path
         else:
-            to_path, _ = changeprefix(from_prefix, to_prefix, from_path)
+            to_path = os.path.join(
+                os.path.dirname(from_path),
+                changeprefix(from_prefix, to_prefix,
+                    os.path.basename(from_path))[0])
 
         # rename contents
         changefile(from_prefix, to_prefix, from_path, to_path,