Github/littlefs
1
Fork 0
Code Actions Activity

Added generalize fuzz testing for B-trees, fixed single-child bug

Browse Source 
A single child is just another condition to watch out for during B-tree
merge, since a single-child obviously can't have a sibling.

This is a good safety to have, but I was surprised this can happen. But
it turns out to be quite easy since our rbyds defer the B-tree
operations until compaction. A merge down to a single child won't
propagate the merge until the parent compacts.
This commit is contained in:
Christopher Haster
2023-03-13 13:21:32 -05:00
parent c55cc4d010
commit dd5af724fd
2 changed files with 285 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lfs.c
+6 -2
View File
@@ -2232,6 +2232,7 @@ static int lfsr_rbyd_append(lfs_t *lfs, lfsr_rbyd_t *rbyd,
other_id_ = id_;
} else if (tag == LFSR_TAG_SHRINK) {
LFS_ASSERT(id < rbyd->weight);
LFS_ASSERT(lfsr_data_len(data) <= rbyd->weight);
// noop?
if (lfsr_data_len(data) == 0) {
return 0;
@@ -3320,8 +3321,6 @@ static lfs_ssize_t lfsr_btree_get(lfs_t *lfs,
buffer, size);
}
// TODO drop the root during merges
// TODO inline?
static int lfsr_btree_commit(lfs_t *lfs,
lfsr_btree_t *btree, lfs_size_t id,
lfsr_rbyd_t *rbyd, const struct lfsr_attr *attrs) {
@@ -3821,6 +3820,11 @@ static int lfsr_btree_commit(lfs_t *lfs,
goto abort;
}
// only child? can't merge
if (pweight == parent.weight) {
goto abort;
}
// last child? try the left sibling
lfs_ssize_t sid;
lfs_ssize_t sdelta;
tests/test_btree.toml
+279 -8
View File
@@ -452,7 +452,6 @@ code = '''
memmove(&sim[id+1], &sim[id], sim_size-id);
sim[id] = alphas[i % 26];
sim_size += 1;
printf("%c\n", alphas[i % 26]);
}
// check that btree matches sim
@@ -2068,10 +2067,282 @@ code = '''
}
'''
#
## [cases.test_btree_split]
## [cases.test_btree_split_fuzz]
#
## [cases.test_btree_general_fuzz]
#
## TODO also test copy-on-write!
# TODO [cases.test_btree_split]
# TODO [cases.test_btree_split_fuzz]
# Some more general fuzz testing
[cases.test_btree_general_fuzz]
defines.N = [1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024]
defines.ITER = 100
in = 'lfs.c'
code = '''
const char *alphas = "abcdefghijklmnopqrstuvwxyz";
// iterate through severals seeds that we can reproduce easily
for (uint32_t seed = 1; seed < ITER+1; seed++) {
// create lfs here since we need to reset each iteration, we're
// space constrained and we can't expect gc to work at this point
lfs_t lfs;
lfs_init(&lfs, cfg) => 0;
// create free lookahead
memset(lfs.free.buffer, 0, lfs.cfg->lookahead_size);
lfs.free.off = 0;
lfs.free.size = lfs_min(8*lfs.cfg->lookahead_size,
lfs.cfg->block_count);
lfs.free.i = 0;
lfs_alloc_ack(&lfs);
// create a btree
lfsr_btree_t btree = LFSR_BTREE_NULL;
// set up a simulation to compare against
//
// fun fact this is slower than our actual tree! unfun fact this is
// starting to be a problem...
char *sim = malloc(N);
lfs_size_t sim_size = 0;
memset(sim, 0, N);
uint32_t prng = seed;
for (lfs_size_t i = 0; i < (N-REMAINING); i++) {
// choose a pseudo-random op
uint8_t op = TEST_PRNG(&prng) % 3;
// choose a pseudo-random id
lfs_size_t id = TEST_PRNG(&prng) % (sim_size+1);
if (op == 0 || id == sim_size) {
// push to btree
lfsr_btree_push(&lfs, &btree, id,
LFSR_TAG_INLINED, 1,
&alphas[i % 26], 1) => 0;
// push to sim
memmove(&sim[id+1], &sim[id], sim_size-id);
sim[id] = alphas[i % 26];
sim_size += 1;
} else if (op == 1) {
// update btree
lfsr_btree_update(&lfs, &btree, id,
LFSR_TAG_INLINED, 1,
&alphas[i % 26], 1) => 0;
// update sim
sim[id] = alphas[i % 26];
} else {
// pop from btree
lfsr_btree_pop(&lfs, &btree, id) => 0;
// pop from sim
memmove(&sim[id], &sim[id+1], sim_size-(id+1));
sim_size -= 1;
}
}
// check that btree matches sim
printf("expd: [");
bool first = true;
for (lfs_size_t i = 0; i < sim_size; i++) {
if (!first) {
printf(", ");
}
first = false;
printf("%c", sim[i]);
}
printf("]\n");
printf("btree: 0x%x.%x 0x%x w%d\n",
btree.u.trunk.block,
btree.u.trunk.limit,
btree.tag,
btree.weight);
assert(btree.weight == sim_size);
uint8_t buffer[4];
lfsr_tag_t tag_;
lfs_size_t id_;
lfs_size_t weight_;
for (lfs_size_t i = 0; i < sim_size; i++) {
lfsr_btree_get(&lfs, &btree, i,
&tag_, &id_, &weight_,
buffer, 4) => 1;
assert(tag_ == LFSR_TAG_INLINED);
assert(id_ == i);
assert(weight_ == 1);
assert(memcmp(buffer, &sim[i], 1) == 0);
}
// and no extra elements
lfsr_btree_get(&lfs, &btree, sim_size,
&tag_, &id_, &weight_,
buffer, 4) => LFS_ERR_NOENT;
// clean up sim
free(sim);
}
'''
[cases.test_btree_general_sparse_fuzz]
defines.N = [1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024]
defines.W = 5
defines.ITER = 100
in = 'lfs.c'
code = '''
const char *alphas = "abcdefghijklmnopqrstuvwxyz";
// iterate through severals seeds that we can reproduce easily
for (uint32_t seed = 1; seed < ITER+1; seed++) {
// create lfs here since we need to reset each iteration, we're
// space constrained and we can't expect gc to work at this point
lfs_t lfs;
lfs_init(&lfs, cfg) => 0;
// create free lookahead
memset(lfs.free.buffer, 0, lfs.cfg->lookahead_size);
lfs.free.off = 0;
lfs.free.size = lfs_min(8*lfs.cfg->lookahead_size,
lfs.cfg->block_count);
lfs.free.i = 0;
lfs_alloc_ack(&lfs);
// create a btree
lfsr_btree_t btree = LFSR_BTREE_NULL;
// set up a simulation to compare against
//
// fun fact this is slower than our actual tree! unfun fact this is
// starting to be a problem...
char *sim = malloc(N);
lfs_size_t *sim_weights = malloc(N*sizeof(lfs_size_t));
lfs_size_t sim_size = 0;
memset(sim, 0, N);
memset(sim_weights, 0, N*sizeof(lfs_size_t));
uint32_t prng = seed;
for (lfs_size_t i = 0; i < (N-REMAINING); i++) {
// choose a pseudo-random op
uint8_t op = TEST_PRNG(&prng) % 3;
// choose a pseudo-random id
lfs_size_t id = TEST_PRNG(&prng) % (sim_size+1);
// choose a pseudo-random weight
lfs_size_t weight = 1 + (TEST_PRNG(&prng) % W);
// calculate actual id in btree space
lfs_size_t weighted_id = 0;
for (lfs_size_t j = 0; j < id; j++) {
weighted_id += sim_weights[j];
}
if (op == 0 || id == sim_size) {
// push to btree
lfsr_btree_push(&lfs, &btree, weighted_id,
LFSR_TAG_INLINED, weight,
&alphas[i % 26], 1) => 0;
// push to sim
memmove(&sim[id+1], &sim[id], sim_size-id);
memmove(&sim_weights[id+1], &sim_weights[id],
(sim_size-id)*sizeof(lfs_size_t));
sim[id] = alphas[i % 26];
sim_weights[id] = weight;
sim_size += 1;
} else if (op == 1) {
// update btree
lfsr_btree_update(&lfs, &btree,
weighted_id+sim_weights[id]-1, LFSR_TAG_INLINED, weight,
&alphas[i % 26], 1) => 0;
// update sim
sim[id] = alphas[i % 26];
sim_weights[id] = weight;
} else {
// remove from btree
lfsr_btree_pop(&lfs, &btree,
weighted_id+sim_weights[id]-1) => 0;
// remove from sim
memmove(&sim[id], &sim[id+1], sim_size-(id+1));
memmove(&sim_weights[id], &sim_weights[id+1],
(sim_size-(id+1))*sizeof(lfs_size_t));
sim_size -= 1;
}
}
// check that btree matches sim
printf("expd: [");
bool first = true;
for (lfs_size_t i = 0; i < sim_size; i++) {
if (!first) {
printf(", ");
}
first = false;
printf("%c", sim[i]);
}
printf("]\n");
printf("btree: 0x%x.%x 0x%x w%d\n",
btree.u.trunk.block,
btree.u.trunk.limit,
btree.tag,
btree.weight);
lfs_size_t total_weight = 0;
for (lfs_size_t j = 0; j < sim_size; j++) {
total_weight += sim_weights[j];
}
assert(btree.weight == total_weight);
uint8_t buffer[4];
lfsr_tag_t tag_;
lfs_size_t id_;
lfs_size_t weight_;
for (lfs_size_t i = 0; i < sim_size; i++) {
// calculate actual id in btree space
lfs_size_t weighted_id = 0;
for (lfs_size_t j = 0; j < i; j++) {
weighted_id += sim_weights[j];
}
lfsr_btree_get(&lfs, &btree, weighted_id+sim_weights[i]-1,
&tag_, &id_, &weight_,
buffer, 4) => 1;
assert(tag_ == LFSR_TAG_INLINED);
assert(id_ == weighted_id+sim_weights[i]-1);
assert(weight_ == sim_weights[i]);
assert(memcmp(buffer, &sim[i], 1) == 0);
}
// and no extra elements
lfsr_btree_get(&lfs, &btree, total_weight,
&tag_, &id_, &weight_,
buffer, 4) => LFS_ERR_NOENT;
// also test that we can traverse the tree without prior knowledge
id_ = -1;
for (lfs_size_t i = 0; i < sim_size; i++) {
// calculate actual id in btree space
lfs_size_t weighted_id = 0;
for (lfs_size_t j = 0; j < i; j++) {
weighted_id += sim_weights[j];
}
lfsr_btree_get(&lfs, &btree, id_+1,
&tag_, &id_, &weight_,
buffer, 4) => 1;
assert(tag_ == LFSR_TAG_INLINED);
assert(id_ == weighted_id+sim_weights[i]-1);
assert(weight_ == sim_weights[i]);
assert(memcmp(buffer, &sim[i], 1) == 0);
}
lfsr_btree_get(&lfs, &btree, id_+1,
&tag_, &id_, &weight_,
buffer, 4) => LFS_ERR_NOENT;
// clean up sim
free(sim);
}
'''