Fixed broken wear-leveling when block_cycles = 2n-1

This was an interesting issue found during a GitHub discussion with
rmollway and thrasher8390.

Blocks in the metadata-pair are relocated every "block_cycles", or, more
mathy, when rev % block_cycles == 0 as long as rev += 1 every block write.

But there's a problem, rev isn't += 1 every block write. There are two
blocks in a metadata-pair, so looking at it from each blocks
perspective, rev += 2 every block write.

This leads to a sort of aliasing issue, where, if block_cycles is
divisible by 2, one block in the metadata-pair is always relocated, and
the other block is _never_ relocated. Causing a complete failure of
block-level wear-leveling.

Fortunately, because of a previous workaround to avoid block_cycles = 1
(since this will cause the relocation algorithm to never terminate), the
actual math is rev % (block_cycles+1) == 0. This means the bug only
shows its head in the much less likely case where block_cycles is a
multiple of 2 plus 1, or, in more mathy terms, block_cycles = 2n+1 for
some n.

To workaround this we can bitwise or our block_cycles with 1 to force it
to never be a multiple of 2n.

(Maybe we should do this during initialization? But then block_cycles
would need to be mutable.)

---

There's a few unrelated changes mixed into this commit that shouldn't be
there since I added this as part of a branch of bug fixes I'm putting
together rather hastily, so unfortunately this is not easily cherry-pickable.

lfs.c

@@ -1486,7 +1486,7 @@ static int lfs_dir_compact(lfs_t *lfs,
     // increment revision count
     uint32_t nrev = dir->rev + 1;
     if (lfs->cfg->block_cycles > 0 &&
-            (nrev % (lfs->cfg->block_cycles+1) == 0)) {
+            (nrev % ((lfs->cfg->block_cycles+1)|1) == 0)) {
         if (lfs_pair_cmp(dir->pair, (const lfs_block_t[2]){0, 1}) == 0) {
             // oh no! we're writing too much to the superblock,
             // should we expand?
@@ -1524,6 +1524,7 @@ static int lfs_dir_compact(lfs_t *lfs,
         } else {
             // we're writing too much, time to relocate
             tired = true;
+            //nrev += 1;//lfs->cfg->block_cycles & 1; // TODO do this here?
             goto relocate;
         }
     }
@@ -1631,17 +1632,33 @@ static int lfs_dir_compact(lfs_t *lfs,
 
             // TODO huh?
             LFS_ASSERT(commit.off % lfs->cfg->prog_size == 0);
+            // update gstate
+            lfs->gdelta = (lfs_gstate_t){0};
+            if (!relocated) {
+                lfs->gdisk = lfs->gstate;
+            }
+
+    // TODO here??
+    if (relocated) {
+        // update references if we relocated
+        LFS_DEBUG("Relocating %"PRIx32" %"PRIx32" -> %"PRIx32" %"PRIx32,
+                oldpair[0], oldpair[1], dir->pair[0], dir->pair[1]);
+        int err = lfs_fs_relocate(lfs, oldpair, dir->pair);
+        if (err) {
+            // TODO make better
+            //dir->pair[1] = oldpair[1]; //
+            return err;
+        }
+//        LFS_DEBUG("Relocated %"PRIx32" %"PRIx32" -> %"PRIx32" %"PRIx32,
+//                oldpair[0], oldpair[1], dir->pair[0], dir->pair[1]);
+    }
+
             // successful compaction, swap dir pair to indicate most recent
             lfs_pair_swap(dir->pair);
             dir->rev = nrev;
             dir->count = end - begin;
             dir->off = commit.off;
             dir->etag = commit.ptag;
-            // update gstate
-            lfs->gdelta = (lfs_gstate_t){0};
-            if (!relocated) {
-                lfs->gdisk = lfs->gstate;
-            }
         }
         break;
 
@@ -1669,20 +1686,6 @@ relocate:
         continue;
     }
 
-    // TODO here??
-    if (relocated) {
-        // update references if we relocated
-        LFS_DEBUG("Relocating %"PRIx32" %"PRIx32" -> %"PRIx32" %"PRIx32,
-                oldpair[0], oldpair[1], dir->pair[0], dir->pair[1]);
-        int err = lfs_fs_relocate(lfs, oldpair, dir->pair);
-        if (err) {
-            // TODO make better
-            //dir->pair[1] = oldpair[1]; //
-            return err;
-        }
-        LFS_DEBUG("Relocated %"PRIx32" %"PRIx32" -> %"PRIx32" %"PRIx32,
-                oldpair[0], oldpair[1], dir->pair[0], dir->pair[1]);
-    }
 
     return 0;
 }
@@ -1709,6 +1712,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
 
     // calculate changes to the directory
     bool hasdelete = false;
+    lfs_mdir_t olddir = *dir; // TODO is this a good idea?
     for (int i = 0; i < attrcount; i++) {
         if (lfs_tag_type3(attrs[i].tag) == LFS_TYPE_CREATE) {
             dir->count += 1;
@@ -1729,6 +1733,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
         lfs_mdir_t pdir;
         int err = lfs_fs_pred(lfs, dir->pair, &pdir);
         if (err && err != LFS_ERR_NOENT) {
+            *dir = olddir;
             return err;
         }
 
@@ -1761,6 +1766,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
             if (err == LFS_ERR_NOSPC || err == LFS_ERR_CORRUPT) {
                 goto compact;
             }
+            *dir = olddir;
             return err;
         }
 
@@ -1773,6 +1779,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
         if (!lfs_gstate_iszero(&delta)) {
             err = lfs_dir_getgstate(lfs, dir, &delta);
             if (err) {
+                *dir = olddir;
                 return err;
             }
 
@@ -1784,6 +1791,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
                 if (err == LFS_ERR_NOSPC || err == LFS_ERR_CORRUPT) {
                     goto compact;
                 }
+                *dir = olddir;
                 return err;
             }
         }
@@ -1794,6 +1802,7 @@ static int lfs_dir_commit(lfs_t *lfs, lfs_mdir_t *dir,
             if (err == LFS_ERR_NOSPC || err == LFS_ERR_CORRUPT) {
                 goto compact;
             }
+            *dir = olddir;
             return err;
         }
 
@@ -1812,6 +1821,7 @@ compact:
         int err = lfs_dir_compact(lfs, dir, attrs, attrcount,
                 dir, 0, dir->count);
         if (err) {
+            *dir = olddir;
             return err;
         }
     }
@@ -3845,7 +3855,7 @@ int lfs_fs_traverse(lfs_t *lfs,
                     LFS_TRACE("lfs_fs_traverse -> %d", err);
                     return err;
                 }
-            } else if (lfs_gstate_hasorphans(&lfs->gstate) &&
+            } else if (/*lfs_gstate_hasorphans(&lfs->gstate) && TODO maybe report size-dirs/2 ? */
                     lfs_tag_type3(tag) == LFS_TYPE_DIRSTRUCT) {
                 // TODO HMMMMMM HMMMMMMMMMMMMMMMMMMM
                 for (int i = 0; i < 2; i++) {

tests/test_exhaustion.toml

@@ -333,3 +333,113 @@ exhausted:
     // check we increased the lifetime by 2x with ~10% error
     LFS_ASSERT(run_cycles[1]*110/100 > 2*run_cycles[0]);
 '''
+
+[[case]] # test that we wear blocks roughly evenly
+define.LFS_ERASE_CYCLES = 0xffffffff
+define.LFS_BLOCK_COUNT = 256 # small bd so test runs faster
+define.LFS_BLOCK_CYCLES = [5, 4, 3, 2, 1]
+#define.LFS_BLOCK_CYCLES = [4, 2]
+define.CYCLES = 100
+define.FILES = 10
+code = '''
+    lfs_format(&lfs, &cfg) => 0;
+    lfs_mount(&lfs, &cfg) => 0;
+    lfs_mkdir(&lfs, "roadrunner") => 0;
+    lfs_unmount(&lfs) => 0;
+
+    uint32_t cycle = 0;
+    while (cycle < CYCLES) {
+        lfs_mount(&lfs, &cfg) => 0;
+        for (uint32_t i = 0; i < FILES; i++) {
+            // chose name, roughly random seed, and random 2^n size
+            sprintf(path, "roadrunner/test%d", i);
+            srand(cycle * i);
+            size = 1 << 4; //((rand() % 10)+2);
+
+            lfs_file_open(&lfs, &file, path,
+                    LFS_O_WRONLY | LFS_O_CREAT | LFS_O_TRUNC) => 0;
+
+            for (lfs_size_t j = 0; j < size; j++) {
+                char c = 'a' + (rand() % 26);
+                lfs_ssize_t res = lfs_file_write(&lfs, &file, &c, 1);
+                assert(res == 1 || res == LFS_ERR_NOSPC);
+                if (res == LFS_ERR_NOSPC) {
+                    goto exhausted;
+                }
+            }
+
+            err = lfs_file_close(&lfs, &file);
+            assert(err == 0 || err == LFS_ERR_NOSPC);
+            if (err == LFS_ERR_NOSPC) {
+                goto exhausted;
+            }
+        }
+
+        for (uint32_t i = 0; i < FILES; i++) {
+            // check for errors
+            sprintf(path, "roadrunner/test%d", i);
+            srand(cycle * i);
+            size = 1 << 4; //((rand() % 10)+2);
+
+            lfs_file_open(&lfs, &file, path, LFS_O_RDONLY) => 0;
+            for (lfs_size_t j = 0; j < size; j++) {
+                char c = 'a' + (rand() % 26);
+                char r;
+                lfs_file_read(&lfs, &file, &r, 1) => 1;
+                assert(r == c);
+            }
+
+            lfs_file_close(&lfs, &file) => 0;
+        }
+        lfs_unmount(&lfs) => 0;
+
+        cycle += 1;
+    }
+
+exhausted:
+    // should still be readable
+    lfs_mount(&lfs, &cfg) => 0;
+    for (uint32_t i = 0; i < FILES; i++) {
+        // check for errors
+        sprintf(path, "roadrunner/test%d", i);
+        lfs_stat(&lfs, path, &info) => 0;
+    }
+    lfs_unmount(&lfs) => 0;
+
+    LFS_WARN("completed %d cycles", cycle);
+
+    // check the wear on our block device
+    lfs_testbd_wear_t minwear = -1;
+    lfs_testbd_wear_t totalwear = 0;
+    lfs_testbd_wear_t maxwear = 0;
+    // skip 0 and 1 as superblock movement is intentionally avoided
+    for (lfs_block_t b = 2; b < LFS_BLOCK_COUNT; b++) {
+        lfs_testbd_wear_t wear = lfs_testbd_getwear(&cfg, b);
+        printf("%08x: wear %d\n", b, wear);
+        assert(wear >= 0);
+        if (wear < minwear) {
+            minwear = wear;
+        }
+        if (wear > maxwear) {
+            maxwear = wear;
+        }
+        totalwear += wear;
+    }
+    lfs_testbd_wear_t avgwear = totalwear / LFS_BLOCK_COUNT;
+    LFS_WARN("max wear: %d cycles", maxwear);
+    LFS_WARN("avg wear: %d cycles", totalwear / LFS_BLOCK_COUNT);
+    LFS_WARN("min wear: %d cycles", minwear);
+
+    // find standard deviation^2
+    lfs_testbd_wear_t dev2 = 0;
+    for (lfs_block_t b = 2; b < LFS_BLOCK_COUNT; b++) {
+        lfs_testbd_wear_t wear = lfs_testbd_getwear(&cfg, b);
+        assert(wear >= 0);
+        lfs_testbd_swear_t diff = wear - avgwear;
+        dev2 += diff*diff;
+    }
+    dev2 /= totalwear;
+    LFS_WARN("std dev^2: %d", dev2);
+    assert(dev2 < 8);
+'''
+