Fixed ENOSPC issues with zero-granularity blocks

Result of testing on zero-granularity blocks, where the prog size and
read size equals the block size. This represents SD cards and other
traditional forms of block storage where we don't really get a benefit
from the metadata logging.

Unfortunately, since updates in both are tested by the same script,
we can't really use simple bash commands. Added a more complex
script to simulate corruption. Fortunately this should be more robust
than the previous solutions.

The main fixes were around corner cases where the commit logic fell
apart when it didn't have room to complete commits, but these were
fixable in the current design.

lfs.c

@@ -886,6 +886,11 @@ split:
         // drop caches and create tail
         lfs->pcache.block = 0xffffffff;
 
+        if (ack == -1) {
+            // If we can't fit in this block, we won't fit in next block
+            return LFS_ERR_NOSPC;
+        }
+
         lfs_mdir_t tail;
         int err = lfs_dir_alloc(lfs, &tail, dir->split, dir->tail);
         if (err) {
@@ -1971,7 +1976,7 @@ int lfs_file_close(lfs_t *lfs, lfs_file_t *file) {
 }
 
 static int lfs_file_relocate(lfs_t *lfs, lfs_file_t *file) {
-relocate:;
+    while (true) {
         // just relocate what exists into new block
         lfs_block_t nblock;
         int err = lfs_alloc(lfs, &nblock);
@@ -2014,6 +2019,10 @@ relocate:;
 
         file->block = nblock;
         return 0;
+
+relocate:
+        continue;
+    }
 }
 
 static int lfs_file_flush(lfs_t *lfs, lfs_file_t *file) {
@@ -2067,6 +2076,7 @@ static int lfs_file_flush(lfs_t *lfs, lfs_file_t *file) {
                 }
 
                 break;
+
 relocate:
                 LFS_DEBUG("Bad block at %d", file->block);
                 err = lfs_file_relocate(lfs, file);
@@ -2091,6 +2101,7 @@ relocate:
 }
 
 int lfs_file_sync(lfs_t *lfs, lfs_file_t *file) {
+    while (true) {
         int err = lfs_file_flush(lfs, file);
         if (err) {
             return err;
@@ -2109,30 +2120,39 @@ int lfs_file_sync(lfs_t *lfs, lfs_file_t *file) {
             }
 
             // either update the references or inline the whole file
-        if (!(file->flags & LFS_F_INLINE)) {
             int err = lfs_dir_commit(lfs, &cwd,
-                    LFS_MKATTR(LFS_TYPE_CTZSTRUCT, file->id,
-                        &file->ctz.head, sizeof(file->ctz),
                     LFS_MKATTR(LFS_FROM_ATTRS, file->id, file->cfg->attrs, 0,
-                    NULL)));
-            if (err) {
-                return err;
-            }
-        } else {
-            int err = lfs_dir_commit(lfs, &cwd,
+                    (file->flags & LFS_F_INLINE) ?
                         LFS_MKATTR(LFS_TYPE_INLINESTRUCT, file->id,
-                            file->cache.buffer, file->ctz.size,
-                    LFS_MKATTR(LFS_FROM_ATTRS, file->id, file->cfg->attrs, 0,
-                    NULL)));
+                            file->cache.buffer, file->ctz.size, NULL) :
+                        LFS_MKATTR(LFS_TYPE_CTZSTRUCT, file->id,
+                            &file->ctz.head, sizeof(file->ctz), NULL)));
             if (err) {
-                return err;
+                if (err == LFS_ERR_NOSPC && (file->flags & LFS_F_INLINE)) {
+                    goto relocate;
                 }
+                return err;
             }
 
             file->flags &= ~LFS_F_DIRTY;
         }
 
         return 0;
+
+relocate:
+        // inline file doesn't fit anymore
+        file->block = 0xfffffffe;
+        file->off = file->pos;
+
+        lfs_alloc_ack(lfs);
+        err = lfs_file_relocate(lfs, file);
+        if (err) {
+            return err;
+        }
+
+        file->flags &= ~LFS_F_INLINE;
+        file->flags |= LFS_F_WRITING;
+    }
 }
 
 lfs_ssize_t lfs_file_read(lfs_t *lfs, lfs_file_t *file,
@@ -3304,6 +3324,7 @@ static int32_t lfs_parent(lfs_t *lfs, const lfs_block_t pair[2],
 // TODO rename to lfs_dir_relocate?
 static int lfs_relocate(lfs_t *lfs,
         const lfs_block_t oldpair[2], const lfs_block_t newpair[2]) {
+    // TODO name lfs_dir_relocate?
     // find parent
     lfs_mdir_t parent;
     int32_t tag = lfs_parent(lfs, oldpair, &parent);

tests/corrupt.py

@@ -0,0 +1,39 @@
+#!/usr/bin/env python
+
+import struct
+import sys
+import os
+
+def main(*paths):
+    # find most recent block
+    file = None
+    rev = None
+    for path in paths:
+        try:
+            nfile = open(path, 'r+b')
+            nrev, = struct.unpack('<I', nfile.read(4))
+
+            assert rev != nrev
+            if not file or ((rev - nrev) & 0x80000000):
+                file = nfile
+                rev = nrev
+        except IOError:
+            pass
+
+    # go to last commit
+    tag = 0
+    while True:
+        try:
+            ntag, = struct.unpack('<I', file.read(4))
+        except struct.error:
+            break
+
+        tag ^= ntag
+        file.seek(tag & 0xfff, os.SEEK_CUR)
+
+    # lob off last 3 bytes
+    file.seek(-((tag & 0xfff) + 3), os.SEEK_CUR)
+    file.truncate()
+
+if __name__ == "__main__":
+    main(*sys.argv[1:])

tests/test_move.sh

@@ -59,7 +59,7 @@ tests/test.py << TEST
     lfs_rename(&lfs, "b/hello", "c/hello") => 0;
     lfs_unmount(&lfs) => 0;
 TEST
-truncate -s-7 blocks/6
+tests/corrupt.py blocks/{6,7}
 tests/test.py << TEST
     lfs_mount(&lfs, &cfg) => 0;
     lfs_dir_open(&lfs, &dir[0], "b") => 0;
@@ -86,8 +86,8 @@ tests/test.py << TEST
     lfs_rename(&lfs, "c/hello", "d/hello") => 0;
     lfs_unmount(&lfs) => 0;
 TEST
-truncate -s-7 blocks/8
-truncate -s-7 blocks/a
+tests/corrupt.py blocks/{8,9}
+tests/corrupt.py blocks/{a,b}
 tests/test.py << TEST
     lfs_mount(&lfs, &cfg) => 0;
     lfs_dir_open(&lfs, &dir[0], "c") => 0;
@@ -166,7 +166,7 @@ tests/test.py << TEST
     lfs_rename(&lfs, "b/hi", "c/hi") => 0;
     lfs_unmount(&lfs) => 0;
 TEST
-truncate -s-7 blocks/7
+tests/corrupt.py blocks/{6,7}
 tests/test.py << TEST
     lfs_mount(&lfs, &cfg) => 0;
     lfs_dir_open(&lfs, &dir[0], "b") => 0;
@@ -193,8 +193,8 @@ tests/test.py << TEST
     lfs_rename(&lfs, "c/hi", "d/hi") => 0;
     lfs_unmount(&lfs) => 0;
 TEST
-truncate -s-7 blocks/9
-truncate -s-7 blocks/b
+tests/corrupt.py blocks/{8,9}
+tests/corrupt.py blocks/{a,b}
 tests/test.py << TEST
     lfs_mount(&lfs, &cfg) => 0;
     lfs_dir_open(&lfs, &dir[0], "c") => 0;

tests/test_orphan.sh

@@ -17,7 +17,7 @@ tests/test.py << TEST
 TEST
 # corrupt most recent commit, this should be the update to the previous
 # linked-list entry and should orphan the child
-truncate -s-14 blocks/8
+tests/corrupt.py blocks/{8,9}
 tests/test.py  << TEST
     lfs_mount(&lfs, &cfg) => 0;