Skip to contentSkip to conversation

reftable: avoid reading and writing empty keys

PATCH Superseded Han-Wen Nienhuys via GitGitGadget 3 participants 33 messages Jan 12, 2022 – Feb 23, 2022 Documentation

lore
Share & tools
lore

Han-Wen Nienhuys via GitGitGadget

author
#0 lore raw

cover reftable: avoid reading and writing empty keys

this series makes sure that the object record does not have to consider empty keys (and therefore, a NULL memcpy destination)

while we're at it add some more tests, and fix a naming mistake.

Han-Wen Nienhuys (7):
Documentation: object_id_len goes up to 31
reftable: reject 0 object_id_len
reftable: add a test that verifies that writing empty keys fails
reftable: avoid writing empty keys at the block layer
reftable: ensure that obj_id_len is >= 2 on writing
reftable: add test for length of disambiguating prefix
reftable: rename writer_stats to reftable_writer_stats

 Documentation/technical/reftable.txt |   2 +-
 reftable/block.c                     |  27 ++++---
 reftable/block_test.c                |   5 ++
 reftable/reader.c                    |   5 ++
 reftable/readwrite_test.c            | 103 ++++++++++++++++++++++++++-
 reftable/reftable-writer.h           |   2 +-
 reftable/writer.c                    |   9 +--
 7 files changed, 135 insertions(+), 18 deletions(-)

base-commit: 90d242d36e248acfae0033274b524bfa55a947fd
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1185%2Fhanwen%2Fobj-id-len-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1185/hanwen/obj-id-len-v1
Pull-Request: https://github.com/git/git/pull/1185

Signature · 1 line
gitgitgadget
author
#1 lore raw

PATCH 1/7 Documentation: object_id_len goes up to 31

From: Han-Wen Nienhuys <[email protected]>

The value is stored in a 5-bit field, so we can't support more without a format version upgrade.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 Documentation/technical/reftable.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/technical/reftable.txt +1 -1
index d7c3b645cfb..6a67cc4174f 100644
@@ -443,7 +443,7 @@ Obj block format
 Object blocks are optional. Writers may choose to omit object blocks,
 especially if readers will not use the object name to ref mapping.
 
-Object blocks use unique, abbreviated 2-32 object name keys, mapping to
+Object blocks use unique, abbreviated 2-31 byte object name keys, mapping to
 ref blocks containing references pointing to that object directly, or as
 the peeled value of an annotated tag. Like ref blocks, object blocks use
 the file's standard block size. The abbreviation length is available in
Signature · 1 line
gitgitgadget
author
#2 lore raw

PATCH 2/7 reftable: reject 0 object_id_len

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1, but we forbid 0, so we can we can be sure that we never read a 0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/reader.c | 5 +++++
 1 file changed, 5 insertions(+)
reftable/reader.c +5 -0
index 006709a645a..21843f5e935 100644
@@ -155,6 +155,11 @@ static int parse_footer(struct reftable_reader *r, uint8_t *footer,
 	r->log_offsets.is_present = (first_block_typ == BLOCK_TYPE_LOG ||
 				     r->log_offsets.offset > 0);
 	r->obj_offsets.is_present = r->obj_offsets.offset > 0;
+	if (r->obj_offsets.is_present && !r->object_id_len) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
+
 	err = 0;
 done:
 	return err;
Signature · 1 line
gitgitgadget
author
#3 lore raw

PATCH 3/7 reftable: add a test that verifies that writing empty keys fails

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

Empty keys can only be written as ref records with empty names. The log record has a logical timestamp in the key, so the key is never empty.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
reftable/readwrite_test.c +24 -0
index 70c7aedba2c..a315c8992e8 100644
@@ -602,6 +602,29 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_empty_key(void)
+{
+	struct reftable_write_options opts = { 0 };
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	struct reftable_ref_record ref = {
+		.refname = "",
+		.update_index = 1,
+		.value_type = REFTABLE_REF_DELETION,
+	};
+	int err;
+
+	reftable_writer_set_limits(w, 1, 1);
+	err = reftable_writer_add_ref(w, &ref);
+	EXPECT(err == REFTABLE_API_ERROR);
+
+	err = reftable_writer_close(w);
+	EXPECT(err == REFTABLE_EMPTY_TABLE_ERROR);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_key_order(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -681,6 +704,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_table_read_write_seek_index);
 	RUN_TEST(test_table_refs_for_no_index);
 	RUN_TEST(test_table_refs_for_obj_index);
+	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
 	return 0;
Signature · 1 line
gitgitgadget
author
#4 lore raw

PATCH 4/7 reftable: avoid writing empty keys at the block layer

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The public interface (reftable_writer) already ensures that keys are written in strictly increasing order, and an empty key by definition fails this check.

However, by also enforcing this at the block layer, it is easier to verify that records (which are written into blocks) never have to consider the possibility of empty keys.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/block.c      | 27 +++++++++++++++++----------
 reftable/block_test.c |  5 +++++
 reftable/writer.c     |  3 +--
 3 files changed, 23 insertions(+), 12 deletions(-)
reftable/block.c +17 -10
index 855e3f5c947..8725eaaf64f 100644
@@ -88,8 +88,9 @@ uint8_t block_writer_type(struct block_writer *bw)
 	return bw->buf[bw->header_off];
 }
 
-/* adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
-   success */
+/* Adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
+   success. Returns REFTABLE_API_ERROR if attempting to write a record with
+   empty key. */
 int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 {
 	struct strbuf empty = STRBUF_INIT;

@@ -105,8 +106,14 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 	int is_restart = 0;
 	struct strbuf key = STRBUF_INIT;
 	int n = 0;
+	int err = -1;
 
 	reftable_record_key(rec, &key);
+	if (!key.len) {
+		err = REFTABLE_API_ERROR;
+		goto done;
+	}
+
 	n = reftable_encode_key(&is_restart, out, last, key,
 				reftable_record_val_type(rec));
 	if (n < 0)

@@ -118,16 +125,11 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 		goto done;
 	string_view_consume(&out, n);
 
-	if (block_writer_register_restart(w, start.len - out.len, is_restart,
-					  &key) < 0)
-		goto done;
-
-	strbuf_release(&key);
-	return 0;
-
+	err = block_writer_register_restart(w, start.len - out.len, is_restart,
+					    &key);
 done:
 	strbuf_release(&key);
-	return -1;
+	return err;
 }
 
 int block_writer_finish(struct block_writer *w)

@@ -324,6 +326,9 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
 	if (n < 0)
 		return -1;
 
+	if (!key.len)
+		return REFTABLE_FORMAT_ERROR;
+
 	string_view_consume(&in, n);
 	n = reftable_record_decode(rec, key, extra, in, it->br->hash_size);
 	if (n < 0)

@@ -350,6 +355,8 @@ int block_reader_first_key(struct block_reader *br, struct strbuf *key)
 	int n = reftable_decode_key(key, &extra, empty, in);
 	if (n < 0)
 		return n;
+	if (!key->len)
+		return -1;
 
 	return 0;
 }
reftable/block_test.c +5 -0
index 4b3ea262dcb..5112ddbf468 100644
@@ -42,6 +42,11 @@ static void test_block_read_write(void)
 			  header_off, hash_size(GIT_SHA1_FORMAT_ID));
 	reftable_record_from_ref(&rec, &ref);
 
+	ref.refname = "";
+	ref.value_type = REFTABLE_REF_DELETION;
+	n = block_writer_add(&bw, &rec);
+	EXPECT(n == REFTABLE_API_ERROR);
+
 	for (i = 0; i < N; i++) {
 		char name[100];
 		uint8_t hash[GIT_SHA1_RAWSZ];
reftable/writer.c +1 -2
index 35c8649c9b7..e3c042b9d84 100644
@@ -238,14 +238,13 @@ static int writer_add_record(struct reftable_writer *w,
 
 	writer_reinit_block_writer(w, reftable_record_type(rec));
 	err = block_writer_add(w->block_writer, rec);
-	if (err < 0) {
+	if (err == -1) {
 		/* we are writing into memory, so an error can only mean it
 		 * doesn't fit. */
 		err = REFTABLE_ENTRY_TOO_BIG_ERROR;
 		goto done;
 	}
 
-	err = 0;
 done:
 	strbuf_release(&key);
 	return err;
Signature · 1 line
gitgitgadget
author
#8 lore raw

PATCH 5/7 reftable: ensure that obj_id_len is >= 2 on writing

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

When writing the same hash many times, we might decide to use a length-1 object ID prefix for the ObjectID => ref table, which is out of spec.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 37 +++++++++++++++++++++++++++++++++++++
 reftable/writer.c         |  4 +++-
 2 files changed, 40 insertions(+), 1 deletion(-)
reftable/readwrite_test.c +37 -0
index a315c8992e8..b4371b75724 100644
@@ -602,6 +602,42 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_min_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 2);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -707,5 +743,6 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
reftable/writer.c +3 -1
index e3c042b9d84..f94af531351 100644
@@ -508,7 +508,9 @@ static void object_record_free(void *void_arg, void *key)
 static int writer_dump_object_index(struct reftable_writer *w)
 {
 	struct write_record_arg closure = { .w = w };
-	struct common_prefix_arg common = { NULL };
+	struct common_prefix_arg common = {
+		.max = 1,		/* obj_id_len should be >= 2. */
+	};
 	if (w->obj_index_tree) {
 		infix_walk(w->obj_index_tree, &update_common, &common);
 	}
Signature · 1 line
gitgitgadget
author
#9 lore raw

PATCH 6/7 reftable: add test for length of disambiguating prefix

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The ID => ref map is trimming object IDs to a disambiguating prefix. Check that we are computing their length correctly.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
reftable/readwrite_test.c +38 -0
index b4371b75724..5f45cee39d6 100644
@@ -638,6 +638,43 @@ static void test_write_object_id_min_length(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		ref.value.val1[15] = i;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 16);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -743,6 +780,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_length);
 	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
Signature · 1 line
gitgitgadget
author
#10 lore raw

PATCH 7/7 reftable: rename writer_stats to reftable_writer_stats

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

This function is part of the reftable API, so it should use the reftable_ prefix

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c  | 8 ++++----
 reftable/reftable-writer.h | 2 +-
 reftable/writer.c          | 2 +-
 3 files changed, 6 insertions(+), 6 deletions(-)
reftable/readwrite_test.c +4 -4
index 5f45cee39d6..b1ff46c18a9 100644
@@ -100,7 +100,7 @@ static void write_table(char ***names, struct strbuf *buf, int N,
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	for (i = 0; i < stats->ref_stats.blocks; i++) {
 		int off = i * opts.block_size;
 		if (off == 0) {

@@ -239,7 +239,7 @@ static void test_log_write_read(void)
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	EXPECT(stats->log_stats.blocks > 0);
 	reftable_writer_free(w);
 	w = NULL;

@@ -633,7 +633,7 @@ static void test_write_object_id_min_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 2);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 2);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }

@@ -670,7 +670,7 @@ static void test_write_object_id_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 16);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 16);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }
reftable/reftable-writer.h +1 -1
index a560dc17255..db8de197f6c 100644
@@ -143,7 +143,7 @@ int reftable_writer_close(struct reftable_writer *w);
 
    This struct becomes invalid when the writer is freed.
  */
-const struct reftable_stats *writer_stats(struct reftable_writer *w);
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w);
 
 /* reftable_writer_free deallocates memory for the writer */
 void reftable_writer_free(struct reftable_writer *w);
reftable/writer.c +1 -1
index f94af531351..495784e6294 100644
@@ -688,7 +688,7 @@ static int writer_flush_block(struct reftable_writer *w)
 	return writer_flush_nonempty_block(w);
 }
 
-const struct reftable_stats *writer_stats(struct reftable_writer *w)
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w)
 {
 	return &w->stats;
 }
Signature · 1 line
gitgitgadget
#5 lore raw

Re: [PATCH 4/7] reftable: avoid writing empty keys at the block layer

In reply to Han-Wen Nienhuys via GitGitGadget

reftable/block_test.c static void test_block_read_write(void) · line 49

Quoted 13 lines · Han-Wen Nienhuys via GitGitGadget
reftable/block_test.c +5 -0
index 4b3ea262dcb..5112ddbf468 100644
@@ -42,6 +42,11 @@ static void test_block_read_write(void)
 			  header_off, hash_size(GIT_SHA1_FORMAT_ID));
 	reftable_record_from_ref(&rec, &ref);
 
+	ref.refname = "";
+	ref.value_type = REFTABLE_REF_DELETION;
+	n = block_writer_add(&bw, &rec);
+	EXPECT(n == REFTABLE_API_ERROR);
+

The preimage of this hunk has been invalidated by your 9c498398 (reftable: make reftable_record a tagged union, 2021-12-22).

I see that the hn/reftable-coverity-fixes topic, which the commit is a part of, has been expecting a reroll since last year---are you plannning to rebuild that series after landing this series first?

#6 lore raw

Re: [PATCH 4/7] reftable: avoid writing empty keys at the block layer

Quoted 11 lines · Junio C Hamano

reftable/block_test.c static void test_block_read_write(void) · line 49

Quoted 5 lines
reftable/block_test.c +5 -0
+     ref.refname = "";
+     ref.value_type = REFTABLE_REF_DELETION;
+     n = block_writer_add(&bw, &rec);
+     EXPECT(n == REFTABLE_API_ERROR);
+

The preimage of this hunk has been invalidated by your 9c498398 (reftable: make reftable_record a tagged union, 2021-12-22).

I see that the hn/reftable-coverity-fixes topic, which the commit is a part of, has been expecting a reroll since last year---are you

I sent a reroll on Jan 12. As that has had more scrutiny, it's best if that lands first. This series is less urgent.

Quoted 1 line

plannning to rebuild that series after landing this series first?

Signature · 11 lines
Han-Wen Nienhuys - Google Munich
I work 80%. Don't expect answers from me on Fridays.
--

Google Germany GmbH, Erika-Mann-Strasse 33, 80636 Munich

Registergericht und -nummer: Hamburg, HRB 86891

Sitz der Gesellschaft: Hamburg

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
#7 lore raw

Re: [PATCH 4/7] reftable: avoid writing empty keys at the block layer

Quoted 5 lines · Han-Wen Nienhuys
Quoted 2 lines

I see that the hn/reftable-coverity-fixes topic, which the commit is a part of, has been expecting a reroll since last year---are you

I sent a reroll on Jan 12. As that has had more scrutiny, it's best if that lands first. This series is less urgent.

Hmph, it probably was missed during pre-release freeze.

https://lore.kernel.org/git/e16bf0c5212ae85daa0d6aa2c78d551824b542bd.1640199396.git.gitgitgadget@gmail.com/

not showing the updated ones did not help, either but stuff outside the upcoming release are not urgent right now, so it is OK.

Thanks.

author
#11 lore raw

v2 cover reftable: avoid reading and writing empty keys

In reply to Han-Wen Nienhuys via GitGitGadget

this series makes sure that the object record does not have to consider empty keys (and therefore, a NULL memcpy destination)

while we're at it add some more tests, and fix a naming mistake.

Han-Wen Nienhuys (7):
Documentation: object_id_len goes up to 31
reftable: reject 0 object_id_len
reftable: add a test that verifies that writing empty keys fails
reftable: avoid writing empty keys at the block layer
reftable: ensure that obj_id_len is >= 2 on writing
reftable: add test for length of disambiguating prefix
reftable: rename writer_stats to reftable_writer_stats

 Documentation/technical/reftable.txt |   2 +-
 reftable/block.c                     |  27 ++++---
 reftable/block_test.c                |   5 ++
 reftable/reader.c                    |   5 ++
 reftable/readwrite_test.c            | 105 ++++++++++++++++++++++++++-
 reftable/reftable-writer.h           |   2 +-
 reftable/writer.c                    |   9 ++-
 7 files changed, 136 insertions(+), 19 deletions(-)

base-commit: 45fe28c951c3e70666ee4ef8379772851a8e4d32
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1185%2Fhanwen%2Fobj-id-len-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1185/hanwen/obj-id-len-v2
Pull-Request: https://github.com/git/git/pull/1185

Range-diff vs v1:

 1:  2d2e1177ff7 = 1:  80d29e8f269 Documentation: object_id_len goes up to 31
 2:  747c9e9a4c8 = 2:  4c1a19fc4ae reftable: reject 0 object_id_len
 3:  4eefedb0d07 = 3:  600b115f8b1 reftable: add a test that verifies that writing empty keys fails
 4:  e4c1cc58265 ! 4:  ba036ee8543 reftable: avoid writing empty keys at the block layer
     @@ reftable/block.c: int block_reader_first_key(struct block_reader *br, struct str
      
       ## reftable/block_test.c ##
      @@ reftable/block_test.c: static void test_block_read_write(void)
     + 	block_writer_init(&bw, BLOCK_TYPE_REF, block.data, block_size,
       			  header_off, hash_size(GIT_SHA1_FORMAT_ID));
     - 	reftable_record_from_ref(&rec, &ref);
       
     -+	ref.refname = "";
     -+	ref.value_type = REFTABLE_REF_DELETION;
     ++	rec.u.ref.refname = "";
     ++	rec.u.ref.value_type = REFTABLE_REF_DELETION;
      +	n = block_writer_add(&bw, &rec);
      +	EXPECT(n == REFTABLE_API_ERROR);
      +
 5:  3a72aba447c = 5:  2bd3d44ba57 reftable: ensure that obj_id_len is >= 2 on writing
 6:  a5dfa048884 = 6:  82d36ee0e0d reftable: add test for length of disambiguating prefix
 7:  37aa7744c84 ! 7:  c6ffdb3471c reftable: rename writer_stats to reftable_writer_stats
     @@ reftable/readwrite_test.c: static void test_log_write_read(void)
       	n = reftable_writer_close(w);
       	EXPECT(n == 0);
       
     +-	stats = writer_stats(w);
     ++	stats = reftable_writer_stats(w);
     + 	EXPECT(stats->log_stats.blocks > 0);
     + 	reftable_writer_free(w);
     + 	w = NULL;
     +@@ reftable/readwrite_test.c: static void test_log_zlib_corruption(void)
     + 	n = reftable_writer_close(w);
     + 	EXPECT(n == 0);
     + 
      -	stats = writer_stats(w);
      +	stats = reftable_writer_stats(w);
       	EXPECT(stats->log_stats.blocks > 0);
Signature · 1 line
gitgitgadget
author
#12 lore raw

PATCH v2 1/7 Documentation: object_id_len goes up to 31

From: Han-Wen Nienhuys <[email protected]>

The value is stored in a 5-bit field, so we can't support more without a format version upgrade.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 Documentation/technical/reftable.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/technical/reftable.txt +1 -1
index d7c3b645cfb..6a67cc4174f 100644
@@ -443,7 +443,7 @@ Obj block format
 Object blocks are optional. Writers may choose to omit object blocks,
 especially if readers will not use the object name to ref mapping.
 
-Object blocks use unique, abbreviated 2-32 object name keys, mapping to
+Object blocks use unique, abbreviated 2-31 byte object name keys, mapping to
 ref blocks containing references pointing to that object directly, or as
 the peeled value of an annotated tag. Like ref blocks, object blocks use
 the file's standard block size. The abbreviation length is available in
Signature · 1 line
gitgitgadget
author
#13 lore raw

PATCH v2 2/7 reftable: reject 0 object_id_len

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1, but we forbid 0, so we can we can be sure that we never read a 0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/reader.c | 5 +++++
 1 file changed, 5 insertions(+)
reftable/reader.c +5 -0
index 00906e7a2de..54b4025105c 100644
@@ -155,6 +155,11 @@ static int parse_footer(struct reftable_reader *r, uint8_t *footer,
 	r->log_offsets.is_present = (first_block_typ == BLOCK_TYPE_LOG ||
 				     r->log_offsets.offset > 0);
 	r->obj_offsets.is_present = r->obj_offsets.offset > 0;
+	if (r->obj_offsets.is_present && !r->object_id_len) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
+
 	err = 0;
 done:
 	return err;
Signature · 1 line
gitgitgadget
author
#15 lore raw

PATCH v2 3/7 reftable: add a test that verifies that writing empty keys fails

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

Empty keys can only be written as ref records with empty names. The log record has a logical timestamp in the key, so the key is never empty.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
reftable/readwrite_test.c +24 -0
index 605ba0f9fd4..fd5922e55f6 100644
@@ -667,6 +667,29 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_empty_key(void)
+{
+	struct reftable_write_options opts = { 0 };
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	struct reftable_ref_record ref = {
+		.refname = "",
+		.update_index = 1,
+		.value_type = REFTABLE_REF_DELETION,
+	};
+	int err;
+
+	reftable_writer_set_limits(w, 1, 1);
+	err = reftable_writer_add_ref(w, &ref);
+	EXPECT(err == REFTABLE_API_ERROR);
+
+	err = reftable_writer_close(w);
+	EXPECT(err == REFTABLE_EMPTY_TABLE_ERROR);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_key_order(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -746,6 +769,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_table_read_write_seek_index);
 	RUN_TEST(test_table_refs_for_no_index);
 	RUN_TEST(test_table_refs_for_obj_index);
+	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
 	return 0;
Signature · 1 line
gitgitgadget
author
#16 lore raw

PATCH v2 4/7 reftable: avoid writing empty keys at the block layer

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The public interface (reftable_writer) already ensures that keys are written in strictly increasing order, and an empty key by definition fails this check.

However, by also enforcing this at the block layer, it is easier to verify that records (which are written into blocks) never have to consider the possibility of empty keys.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/block.c      | 27 +++++++++++++++++----------
 reftable/block_test.c |  5 +++++
 reftable/writer.c     |  3 +--
 3 files changed, 23 insertions(+), 12 deletions(-)
reftable/block.c +17 -10
index 2170748c5e9..4a095afe1e2 100644
@@ -88,8 +88,9 @@ uint8_t block_writer_type(struct block_writer *bw)
 	return bw->buf[bw->header_off];
 }
 
-/* adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
-   success */
+/* Adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
+   success. Returns REFTABLE_API_ERROR if attempting to write a record with
+   empty key. */
 int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 {
 	struct strbuf empty = STRBUF_INIT;

@@ -105,8 +106,14 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 	int is_restart = 0;
 	struct strbuf key = STRBUF_INIT;
 	int n = 0;
+	int err = -1;
 
 	reftable_record_key(rec, &key);
+	if (!key.len) {
+		err = REFTABLE_API_ERROR;
+		goto done;
+	}
+
 	n = reftable_encode_key(&is_restart, out, last, key,
 				reftable_record_val_type(rec));
 	if (n < 0)

@@ -118,16 +125,11 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 		goto done;
 	string_view_consume(&out, n);
 
-	if (block_writer_register_restart(w, start.len - out.len, is_restart,
-					  &key) < 0)
-		goto done;
-
-	strbuf_release(&key);
-	return 0;
-
+	err = block_writer_register_restart(w, start.len - out.len, is_restart,
+					    &key);
 done:
 	strbuf_release(&key);
-	return -1;
+	return err;
 }
 
 int block_writer_finish(struct block_writer *w)

@@ -332,6 +334,9 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
 	if (n < 0)
 		return -1;
 
+	if (!key.len)
+		return REFTABLE_FORMAT_ERROR;
+
 	string_view_consume(&in, n);
 	n = reftable_record_decode(rec, key, extra, in, it->br->hash_size);
 	if (n < 0)

@@ -358,6 +363,8 @@ int block_reader_first_key(struct block_reader *br, struct strbuf *key)
 	int n = reftable_decode_key(key, &extra, empty, in);
 	if (n < 0)
 		return n;
+	if (!key->len)
+		return -1;
 
 	return 0;
 }
reftable/block_test.c +5 -0
index fa2ee092ec0..cb88af4a563 100644
@@ -42,6 +42,11 @@ static void test_block_read_write(void)
 	block_writer_init(&bw, BLOCK_TYPE_REF, block.data, block_size,
 			  header_off, hash_size(GIT_SHA1_FORMAT_ID));
 
+	rec.u.ref.refname = "";
+	rec.u.ref.value_type = REFTABLE_REF_DELETION;
+	n = block_writer_add(&bw, &rec);
+	EXPECT(n == REFTABLE_API_ERROR);
+
 	for (i = 0; i < N; i++) {
 		char name[100];
 		uint8_t hash[GIT_SHA1_RAWSZ];
reftable/writer.c +1 -2
index 944c2329ab5..d54215a50dc 100644
@@ -240,14 +240,13 @@ static int writer_add_record(struct reftable_writer *w,
 
 	writer_reinit_block_writer(w, reftable_record_type(rec));
 	err = block_writer_add(w->block_writer, rec);
-	if (err < 0) {
+	if (err == -1) {
 		/* we are writing into memory, so an error can only mean it
 		 * doesn't fit. */
 		err = REFTABLE_ENTRY_TOO_BIG_ERROR;
 		goto done;
 	}
 
-	err = 0;
 done:
 	strbuf_release(&key);
 	return err;
Signature · 1 line
gitgitgadget
author
#19 lore raw

PATCH v2 5/7 reftable: ensure that obj_id_len is >= 2 on writing

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

When writing the same hash many times, we might decide to use a length-1 object ID prefix for the ObjectID => ref table, which is out of spec.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 37 +++++++++++++++++++++++++++++++++++++
 reftable/writer.c         |  4 +++-
 2 files changed, 40 insertions(+), 1 deletion(-)
reftable/readwrite_test.c +37 -0
index fd5922e55f6..35142eb070e 100644
@@ -667,6 +667,42 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_min_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 2);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -772,5 +808,6 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
reftable/writer.c +3 -1
index d54215a50dc..5e4e6e93416 100644
@@ -515,7 +515,9 @@ static void object_record_free(void *void_arg, void *key)
 static int writer_dump_object_index(struct reftable_writer *w)
 {
 	struct write_record_arg closure = { .w = w };
-	struct common_prefix_arg common = { NULL };
+	struct common_prefix_arg common = {
+		.max = 1,		/* obj_id_len should be >= 2. */
+	};
 	if (w->obj_index_tree) {
 		infix_walk(w->obj_index_tree, &update_common, &common);
 	}
Signature · 1 line
gitgitgadget
author
#21 lore raw

PATCH v2 6/7 reftable: add test for length of disambiguating prefix

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The ID => ref map is trimming object IDs to a disambiguating prefix. Check that we are computing their length correctly.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
reftable/readwrite_test.c +38 -0
index 35142eb070e..a1b835785a3 100644
@@ -703,6 +703,43 @@ static void test_write_object_id_min_length(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		ref.value.val1[15] = i;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 16);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -808,6 +845,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_length);
 	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
Signature · 1 line
gitgitgadget
author
#22 lore raw

PATCH v2 7/7 reftable: rename writer_stats to reftable_writer_stats

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

This function is part of the reftable API, so it should use the reftable_ prefix

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c  | 10 +++++-----
 reftable/reftable-writer.h |  2 +-
 reftable/writer.c          |  2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)
reftable/readwrite_test.c +5 -5
index a1b835785a3..469ab79a5ad 100644
@@ -100,7 +100,7 @@ static void write_table(char ***names, struct strbuf *buf, int N,
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	for (i = 0; i < stats->ref_stats.blocks; i++) {
 		int off = i * opts.block_size;
 		if (off == 0) {

@@ -239,7 +239,7 @@ static void test_log_write_read(void)
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	EXPECT(stats->log_stats.blocks > 0);
 	reftable_writer_free(w);
 	w = NULL;

@@ -330,7 +330,7 @@ static void test_log_zlib_corruption(void)
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	EXPECT(stats->log_stats.blocks > 0);
 	reftable_writer_free(w);
 	w = NULL;

@@ -698,7 +698,7 @@ static void test_write_object_id_min_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 2);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 2);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }

@@ -735,7 +735,7 @@ static void test_write_object_id_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 16);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 16);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }
reftable/reftable-writer.h +1 -1
index a560dc17255..db8de197f6c 100644
@@ -143,7 +143,7 @@ int reftable_writer_close(struct reftable_writer *w);
 
    This struct becomes invalid when the writer is freed.
  */
-const struct reftable_stats *writer_stats(struct reftable_writer *w);
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w);
 
 /* reftable_writer_free deallocates memory for the writer */
 void reftable_writer_free(struct reftable_writer *w);
reftable/writer.c +1 -1
index 5e4e6e93416..6d979e245ff 100644
@@ -695,7 +695,7 @@ static int writer_flush_block(struct reftable_writer *w)
 	return writer_flush_nonempty_block(w);
 }
 
-const struct reftable_stats *writer_stats(struct reftable_writer *w)
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w)
 {
 	return &w->stats;
 }
Signature · 1 line
gitgitgadget
#17 lore raw

Re: [PATCH v2 4/7] reftable: avoid writing empty keys at the block layer

In reply to Han-Wen Nienhuys via GitGitGadget

reftable/block.c int block_writer_add(struct block_writer *w, struct reftable_record *rec) · line 115

Quoted 11 lines · Han-Wen Nienhuys via GitGitGadget
reftable/block.c +5 -0
@@ -105,8 +106,14 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 	int is_restart = 0;
 	struct strbuf key = STRBUF_INIT;
 	int n = 0;
+	int err = -1;
 
 	reftable_record_key(rec, &key);
+	if (!key.len) {
+		err = REFTABLE_API_ERROR;
+		goto done;
+	}

OK; we get an API_ERROR when trying to write a bad one. And ...

reftable/block.c int block_iter_next(struct block_iter *it, struct reftable_record *rec) · line 338

Quoted 6 lines
reftable/block.c +2 -0
@@ -332,6 +334,9 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
 	if (n < 0)
 		return -1;
 
+	if (!key.len)
+		return REFTABLE_FORMAT_ERROR;

... we get a FORMAT_ERROR when the data we try to read is bad (i.e. not our fault). OK.

reftable/block.c int block_reader_first_key(struct block_reader *br, struct strbuf *key) · line 367

Quoted 6 lines
reftable/block.c +2 -0
@@ -358,6 +363,8 @@ int block_reader_first_key(struct block_reader *br, struct strbuf *key)
 	int n = reftable_decode_key(key, &extra, empty, in);
 	if (n < 0)
 		return n;
+	if (!key->len)
+		return -1;

It is curious that this gets a different error out of the same sequence, i.e. decode-key did not return an error but the length of the key happens to be 0, not FORMAT_ERROR.

Quoted 7 lines
reftable/writer.c +0 -0
index 944c2329ab5..d54215a50dc 100644
@@ -240,14 +240,13 @@ static int writer_add_record(struct reftable_writer *w,
 
 	writer_reinit_block_writer(w, reftable_record_type(rec));

reftable/writer.c static int writer_add_record(struct reftable_writer *w, · line 250

Quoted 10 lines
reftable/writer.c +1 -2
 	err = block_writer_add(w->block_writer, rec);
-	if (err < 0) {
+	if (err == -1) {
 		/* we are writing into memory, so an error can only mean it
 		 * doesn't fit. */
 		err = REFTABLE_ENTRY_TOO_BIG_ERROR;
 		goto done;
 	}
 
-	err = 0;

Is this "doesn't fit" related to "we catch 0-length keys", or an unrelated fix was included in this step by "rebase -i" mistake?

#20 lore raw

Re: [PATCH v2 5/7] reftable: ensure that obj_id_len is >= 2 on writing

In reply to Han-Wen Nienhuys via GitGitGadget

reftable/writer.c static void object_record_free(void *void_arg, void *key) · line 520

Quoted 12 lines · Han-Wen Nienhuys via GitGitGadget
reftable/writer.c +3 -1
index d54215a50dc..5e4e6e93416 100644
@@ -515,7 +515,9 @@ static void object_record_free(void *void_arg, void *key)
 static int writer_dump_object_index(struct reftable_writer *w)
 {
 	struct write_record_arg closure = { .w = w };
-	struct common_prefix_arg common = { NULL };
+	struct common_prefix_arg common = {
+		.max = 1,		/* obj_id_len should be >= 2. */
+	};

It feels somewhat strange that we have to set .max to set the floor for the minimum length, but given the way update_common() uses and maintains the member, it is more like "max size we have seen so far", and by pretending that we have already seen common prefix of length 1, we'd force ourselves that we need at least 2 to differentiate.

Quoted 3 lines
reftable/writer.c +0 -0
 	if (w->obj_index_tree) {
 		infix_walk(w->obj_index_tree, &update_common, &common);
 	}
#23 lore raw
In reply to Han-Wen Nienhuys via GitGitGadget
Quoted 4 lines · Han-Wen Nienhuys via GitGitGadget

this series makes sure that the object record does not have to consider empty keys (and therefore, a NULL memcpy destination)

while we're at it add some more tests, and fix a naming mistake.

Looking good. Will queue.

Thanks.

#14 lore raw

Re: [PATCH v2 2/7] reftable: reject 0 object_id_len

In reply to Han-Wen Nienhuys via GitGitGadget
Quoted 4 lines · Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1, but we forbid 0, so we can we can be sure that we never read a

s/we can we can/we can/;

Quoted 23 lines

0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/reader.c | 5 +++++
 1 file changed, 5 insertions(+)
reftable/reader.c +5 -0
index 00906e7a2de..54b4025105c 100644
@@ -155,6 +155,11 @@ static int parse_footer(struct reftable_reader *r, uint8_t *footer,
 	r->log_offsets.is_present = (first_block_typ == BLOCK_TYPE_LOG ||
 				     r->log_offsets.offset > 0);
 	r->obj_offsets.is_present = r->obj_offsets.offset > 0;
+	if (r->obj_offsets.is_present && !r->object_id_len) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
+
 	err = 0;
 done:
 	return err;
#18 lore raw

Re: [PATCH v2 4/7] reftable: avoid writing empty keys at the block layer

In reply to Junio C Hamano
Quoted 10 lines · Junio C Hamano

reftable/block.c int block_reader_first_key(struct block_reader *br, struct strbuf *key) · line 367

Quoted 6 lines
reftable/block.c +2 -0
@@ -358,6 +363,8 @@ int block_reader_first_key(struct block_reader *br, struct strbuf *key)
      int n = reftable_decode_key(key, &extra, empty, in);
      if (n < 0)
              return n;
+     if (!key->len)
+             return -1;

It is curious that this gets a different error out of the same sequence, i.e. decode-key did not return an error but the length of the key happens to be 0, not FORMAT_ERROR.

fixed.

Quoted 14 lines
Quoted 14 lines
reftable/writer.c +1 -1
@@ -240,14 +240,13 @@ static int writer_add_record(struct reftable_writer *w,
 
      writer_reinit_block_writer(w, reftable_record_type(rec));
      err = block_writer_add(w->block_writer, rec);
-     if (err < 0) {
+     if (err == -1) {
              /* we are writing into memory, so an error can only mean it
               * doesn't fit. */
              err = REFTABLE_ENTRY_TOO_BIG_ERROR;
              goto done;
      }
Quoted 4 lines

reftable/writer.c static int writer_add_record(struct reftable_writer *w, · line 250

Quoted 1 line
reftable/writer.c +0 -1
-     err = 0;

Is this "doesn't fit" related to "we catch 0-length keys", or an unrelated fix was included in this step by "rebase -i" mistake?

We don't want to reinterpret API_ERROR (from block_writer_add) as ENTRY_TOO_BIG_ERROR, so we have to tweak the condition here.

Signature · 11 lines
Han-Wen Nienhuys - Google Munich
I work 80%. Don't expect answers from me on Fridays.
--

Google Germany GmbH, Erika-Mann-Strasse 33, 80636 Munich

Registergericht und -nummer: Hamburg, HRB 86891

Sitz der Gesellschaft: Hamburg

Geschäftsführer: Paul Manicle, Liana Sebastian
author
#24 lore raw

v3 cover reftable: avoid reading and writing empty keys

In reply to Han-Wen Nienhuys via GitGitGadget

this series makes sure that the object record does not have to consider empty keys (and therefore, a NULL memcpy destination)

while we're at it add some more tests, and fix a naming mistake.

Han-Wen Nienhuys (7):
Documentation: object_id_len goes up to 31
reftable: reject 0 object_id_len
reftable: add a test that verifies that writing empty keys fails
reftable: avoid writing empty keys at the block layer
reftable: ensure that obj_id_len is >= 2 on writing
reftable: add test for length of disambiguating prefix
reftable: rename writer_stats to reftable_writer_stats

 Documentation/technical/reftable.txt |   2 +-
 reftable/block.c                     |  27 ++++---
 reftable/block_test.c                |   5 ++
 reftable/reader.c                    |   5 ++
 reftable/readwrite_test.c            | 105 ++++++++++++++++++++++++++-
 reftable/reftable-writer.h           |   2 +-
 reftable/writer.c                    |   9 ++-
 7 files changed, 136 insertions(+), 19 deletions(-)

base-commit: 45fe28c951c3e70666ee4ef8379772851a8e4d32
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1185%2Fhanwen%2Fobj-id-len-v3
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1185/hanwen/obj-id-len-v3
Pull-Request: https://github.com/git/git/pull/1185

Range-diff vs v2:

 1:  80d29e8f269 = 1:  80d29e8f269 Documentation: object_id_len goes up to 31
 2:  4c1a19fc4ae ! 2:  68e7bc32ff8 reftable: reject 0 object_id_len

@@ Commit message
reftable: reject 0 object_id_len

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1,
- but we forbid 0, so we can we can be sure that we never read a
- 0-length key.
+ but we forbid 0, so we can be sure that we never read a 0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
 3:  600b115f8b1 = 3:  8b5aebdb07a reftable: add a test that verifies that writing empty keys fails
 4:  ba036ee8543 ! 4:  a9372cacd1b reftable: avoid writing empty keys at the block layer
     @@ reftable/block.c: int block_reader_first_key(struct block_reader *br, struct str
       	if (n < 0)
       		return n;
      +	if (!key->len)
     -+		return -1;
     ++		return REFTABLE_FORMAT_ERROR;
       
       	return 0;
       }
 5:  2bd3d44ba57 = 5:  0b8a42399dd reftable: ensure that obj_id_len is >= 2 on writing
 6:  82d36ee0e0d = 6:  bdccd969475 reftable: add test for length of disambiguating prefix
 7:  c6ffdb3471c = 7:  72499a14e38 reftable: rename writer_stats to reftable_writer_stats
Signature · 1 line
gitgitgadget
author
#26 lore raw

PATCH v3 2/7 reftable: reject 0 object_id_len

From: Han-Wen Nienhuys <[email protected]>

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1, but we forbid 0, so we can be sure that we never read a 0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/reader.c | 5 +++++
 1 file changed, 5 insertions(+)
reftable/reader.c +5 -0
index 00906e7a2de..54b4025105c 100644
@@ -155,6 +155,11 @@ static int parse_footer(struct reftable_reader *r, uint8_t *footer,
 	r->log_offsets.is_present = (first_block_typ == BLOCK_TYPE_LOG ||
 				     r->log_offsets.offset > 0);
 	r->obj_offsets.is_present = r->obj_offsets.offset > 0;
+	if (r->obj_offsets.is_present && !r->object_id_len) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
+
 	err = 0;
 done:
 	return err;
Signature · 1 line
gitgitgadget
author
#25 lore raw

PATCH v3 1/7 Documentation: object_id_len goes up to 31

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The value is stored in a 5-bit field, so we can't support more without a format version upgrade.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 Documentation/technical/reftable.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/technical/reftable.txt +1 -1
index d7c3b645cfb..6a67cc4174f 100644
@@ -443,7 +443,7 @@ Obj block format
 Object blocks are optional. Writers may choose to omit object blocks,
 especially if readers will not use the object name to ref mapping.
 
-Object blocks use unique, abbreviated 2-32 object name keys, mapping to
+Object blocks use unique, abbreviated 2-31 byte object name keys, mapping to
 ref blocks containing references pointing to that object directly, or as
 the peeled value of an annotated tag. Like ref blocks, object blocks use
 the file's standard block size. The abbreviation length is available in
Signature · 1 line
gitgitgadget
author
#28 lore raw

PATCH v3 4/7 reftable: avoid writing empty keys at the block layer

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The public interface (reftable_writer) already ensures that keys are written in strictly increasing order, and an empty key by definition fails this check.

However, by also enforcing this at the block layer, it is easier to verify that records (which are written into blocks) never have to consider the possibility of empty keys.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/block.c      | 27 +++++++++++++++++----------
 reftable/block_test.c |  5 +++++
 reftable/writer.c     |  3 +--
 3 files changed, 23 insertions(+), 12 deletions(-)
reftable/block.c +17 -10
index 2170748c5e9..34d4d073692 100644
@@ -88,8 +88,9 @@ uint8_t block_writer_type(struct block_writer *bw)
 	return bw->buf[bw->header_off];
 }
 
-/* adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
-   success */
+/* Adds the reftable_record to the block. Returns -1 if it does not fit, 0 on
+   success. Returns REFTABLE_API_ERROR if attempting to write a record with
+   empty key. */
 int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 {
 	struct strbuf empty = STRBUF_INIT;

@@ -105,8 +106,14 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 	int is_restart = 0;
 	struct strbuf key = STRBUF_INIT;
 	int n = 0;
+	int err = -1;
 
 	reftable_record_key(rec, &key);
+	if (!key.len) {
+		err = REFTABLE_API_ERROR;
+		goto done;
+	}
+
 	n = reftable_encode_key(&is_restart, out, last, key,
 				reftable_record_val_type(rec));
 	if (n < 0)

@@ -118,16 +125,11 @@ int block_writer_add(struct block_writer *w, struct reftable_record *rec)
 		goto done;
 	string_view_consume(&out, n);
 
-	if (block_writer_register_restart(w, start.len - out.len, is_restart,
-					  &key) < 0)
-		goto done;
-
-	strbuf_release(&key);
-	return 0;
-
+	err = block_writer_register_restart(w, start.len - out.len, is_restart,
+					    &key);
 done:
 	strbuf_release(&key);
-	return -1;
+	return err;
 }
 
 int block_writer_finish(struct block_writer *w)

@@ -332,6 +334,9 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
 	if (n < 0)
 		return -1;
 
+	if (!key.len)
+		return REFTABLE_FORMAT_ERROR;
+
 	string_view_consume(&in, n);
 	n = reftable_record_decode(rec, key, extra, in, it->br->hash_size);
 	if (n < 0)

@@ -358,6 +363,8 @@ int block_reader_first_key(struct block_reader *br, struct strbuf *key)
 	int n = reftable_decode_key(key, &extra, empty, in);
 	if (n < 0)
 		return n;
+	if (!key->len)
+		return REFTABLE_FORMAT_ERROR;
 
 	return 0;
 }
reftable/block_test.c +5 -0
index fa2ee092ec0..cb88af4a563 100644
@@ -42,6 +42,11 @@ static void test_block_read_write(void)
 	block_writer_init(&bw, BLOCK_TYPE_REF, block.data, block_size,
 			  header_off, hash_size(GIT_SHA1_FORMAT_ID));
 
+	rec.u.ref.refname = "";
+	rec.u.ref.value_type = REFTABLE_REF_DELETION;
+	n = block_writer_add(&bw, &rec);
+	EXPECT(n == REFTABLE_API_ERROR);
+
 	for (i = 0; i < N; i++) {
 		char name[100];
 		uint8_t hash[GIT_SHA1_RAWSZ];
reftable/writer.c +1 -2
index 944c2329ab5..d54215a50dc 100644
@@ -240,14 +240,13 @@ static int writer_add_record(struct reftable_writer *w,
 
 	writer_reinit_block_writer(w, reftable_record_type(rec));
 	err = block_writer_add(w->block_writer, rec);
-	if (err < 0) {
+	if (err == -1) {
 		/* we are writing into memory, so an error can only mean it
 		 * doesn't fit. */
 		err = REFTABLE_ENTRY_TOO_BIG_ERROR;
 		goto done;
 	}
 
-	err = 0;
 done:
 	strbuf_release(&key);
 	return err;
Signature · 1 line
gitgitgadget
author
#27 lore raw

PATCH v3 3/7 reftable: add a test that verifies that writing empty keys fails

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

Empty keys can only be written as ref records with empty names. The log record has a logical timestamp in the key, so the key is never empty.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
reftable/readwrite_test.c +24 -0
index 605ba0f9fd4..fd5922e55f6 100644
@@ -667,6 +667,29 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_empty_key(void)
+{
+	struct reftable_write_options opts = { 0 };
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	struct reftable_ref_record ref = {
+		.refname = "",
+		.update_index = 1,
+		.value_type = REFTABLE_REF_DELETION,
+	};
+	int err;
+
+	reftable_writer_set_limits(w, 1, 1);
+	err = reftable_writer_add_ref(w, &ref);
+	EXPECT(err == REFTABLE_API_ERROR);
+
+	err = reftable_writer_close(w);
+	EXPECT(err == REFTABLE_EMPTY_TABLE_ERROR);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_key_order(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -746,6 +769,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_table_read_write_seek_index);
 	RUN_TEST(test_table_refs_for_no_index);
 	RUN_TEST(test_table_refs_for_obj_index);
+	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
 	return 0;
Signature · 1 line
gitgitgadget
author
#29 lore raw

PATCH v3 5/7 reftable: ensure that obj_id_len is >= 2 on writing

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

When writing the same hash many times, we might decide to use a length-1 object ID prefix for the ObjectID => ref table, which is out of spec.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 37 +++++++++++++++++++++++++++++++++++++
 reftable/writer.c         |  4 +++-
 2 files changed, 40 insertions(+), 1 deletion(-)
reftable/readwrite_test.c +37 -0
index fd5922e55f6..35142eb070e 100644
@@ -667,6 +667,42 @@ static void test_write_empty_table(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_min_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 2);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -772,5 +808,6 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
reftable/writer.c +3 -1
index d54215a50dc..5e4e6e93416 100644
@@ -515,7 +515,9 @@ static void object_record_free(void *void_arg, void *key)
 static int writer_dump_object_index(struct reftable_writer *w)
 {
 	struct write_record_arg closure = { .w = w };
-	struct common_prefix_arg common = { NULL };
+	struct common_prefix_arg common = {
+		.max = 1,		/* obj_id_len should be >= 2. */
+	};
 	if (w->obj_index_tree) {
 		infix_walk(w->obj_index_tree, &update_common, &common);
 	}
Signature · 1 line
gitgitgadget
author
#30 lore raw

PATCH v3 6/7 reftable: add test for length of disambiguating prefix

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

The ID => ref map is trimming object IDs to a disambiguating prefix. Check that we are computing their length correctly.

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
reftable/readwrite_test.c +38 -0
index 35142eb070e..a1b835785a3 100644
@@ -703,6 +703,43 @@ static void test_write_object_id_min_length(void)
 	strbuf_release(&buf);
 }
 
+static void test_write_object_id_length(void)
+{
+	struct reftable_write_options opts = {
+		.block_size = 75,
+	};
+	struct strbuf buf = STRBUF_INIT;
+	struct reftable_writer *w =
+		reftable_new_writer(&strbuf_add_void, &buf, &opts);
+	uint8_t hash[GIT_SHA1_RAWSZ] = {42};
+	struct reftable_ref_record ref = {
+		.update_index = 1,
+		.value_type = REFTABLE_REF_VAL1,
+		.value.val1 = hash,
+	};
+	int err;
+	int i;
+
+	reftable_writer_set_limits(w, 1, 1);
+
+	/* Write the same hash in many refs. If there is only 1 hash, the
+	 * disambiguating prefix is length 0 */
+	for (i = 0; i < 256; i++) {
+		char name[256];
+		snprintf(name, sizeof(name), "ref%05d", i);
+		ref.refname = name;
+		ref.value.val1[15] = i;
+		err = reftable_writer_add_ref(w, &ref);
+		EXPECT_ERR(err);
+	}
+
+	err = reftable_writer_close(w);
+	EXPECT_ERR(err);
+	EXPECT(writer_stats(w)->object_id_len == 16);
+	reftable_writer_free(w);
+	strbuf_release(&buf);
+}
+
 static void test_write_empty_key(void)
 {
 	struct reftable_write_options opts = { 0 };

@@ -808,6 +845,7 @@ int readwrite_test_main(int argc, const char *argv[])
 	RUN_TEST(test_write_empty_key);
 	RUN_TEST(test_write_empty_table);
 	RUN_TEST(test_log_overflow);
+	RUN_TEST(test_write_object_id_length);
 	RUN_TEST(test_write_object_id_min_length);
 	return 0;
 }
Signature · 1 line
gitgitgadget
author
#31 lore raw

PATCH v3 7/7 reftable: rename writer_stats to reftable_writer_stats

In reply to Han-Wen Nienhuys via GitGitGadget

From: Han-Wen Nienhuys <[email protected]>

This function is part of the reftable API, so it should use the reftable_ prefix

Signed-off-by Han-Wen Nienhuys <[email protected]>
---
 reftable/readwrite_test.c  | 10 +++++-----
 reftable/reftable-writer.h |  2 +-
 reftable/writer.c          |  2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)
reftable/readwrite_test.c +5 -5
index a1b835785a3..469ab79a5ad 100644
@@ -100,7 +100,7 @@ static void write_table(char ***names, struct strbuf *buf, int N,
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	for (i = 0; i < stats->ref_stats.blocks; i++) {
 		int off = i * opts.block_size;
 		if (off == 0) {

@@ -239,7 +239,7 @@ static void test_log_write_read(void)
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	EXPECT(stats->log_stats.blocks > 0);
 	reftable_writer_free(w);
 	w = NULL;

@@ -330,7 +330,7 @@ static void test_log_zlib_corruption(void)
 	n = reftable_writer_close(w);
 	EXPECT(n == 0);
 
-	stats = writer_stats(w);
+	stats = reftable_writer_stats(w);
 	EXPECT(stats->log_stats.blocks > 0);
 	reftable_writer_free(w);
 	w = NULL;

@@ -698,7 +698,7 @@ static void test_write_object_id_min_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 2);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 2);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }

@@ -735,7 +735,7 @@ static void test_write_object_id_length(void)
 
 	err = reftable_writer_close(w);
 	EXPECT_ERR(err);
-	EXPECT(writer_stats(w)->object_id_len == 16);
+	EXPECT(reftable_writer_stats(w)->object_id_len == 16);
 	reftable_writer_free(w);
 	strbuf_release(&buf);
 }
reftable/reftable-writer.h +1 -1
index a560dc17255..db8de197f6c 100644
@@ -143,7 +143,7 @@ int reftable_writer_close(struct reftable_writer *w);
 
    This struct becomes invalid when the writer is freed.
  */
-const struct reftable_stats *writer_stats(struct reftable_writer *w);
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w);
 
 /* reftable_writer_free deallocates memory for the writer */
 void reftable_writer_free(struct reftable_writer *w);
reftable/writer.c +1 -1
index 5e4e6e93416..6d979e245ff 100644
@@ -695,7 +695,7 @@ static int writer_flush_block(struct reftable_writer *w)
 	return writer_flush_nonempty_block(w);
 }
 
-const struct reftable_stats *writer_stats(struct reftable_writer *w)
+const struct reftable_stats *reftable_writer_stats(struct reftable_writer *w)
 {
 	return &w->stats;
 }
Signature · 1 line
gitgitgadget
#32 lore raw
In reply to Han-Wen Nienhuys via GitGitGadget
Quoted 4 lines · Han-Wen Nienhuys via GitGitGadget

this series makes sure that the object record does not have to consider empty keys (and therefore, a NULL memcpy destination)

while we're at it add some more tests, and fix a naming mistake.

Looking good. Let's mark it for 'next' and below soonish.

Thanks.

Quoted 53 lines

Han-Wen Nienhuys (7):
Documentation: object_id_len goes up to 31
reftable: reject 0 object_id_len
reftable: add a test that verifies that writing empty keys fails
reftable: avoid writing empty keys at the block layer
reftable: ensure that obj_id_len is >= 2 on writing
reftable: add test for length of disambiguating prefix
reftable: rename writer_stats to reftable_writer_stats

 Documentation/technical/reftable.txt |   2 +-
 reftable/block.c                     |  27 ++++---
 reftable/block_test.c                |   5 ++
 reftable/reader.c                    |   5 ++
 reftable/readwrite_test.c            | 105 ++++++++++++++++++++++++++-
 reftable/reftable-writer.h           |   2 +-
 reftable/writer.c                    |   9 ++-
 7 files changed, 136 insertions(+), 19 deletions(-)

base-commit: 45fe28c951c3e70666ee4ef8379772851a8e4d32
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1185%2Fhanwen%2Fobj-id-len-v3
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1185/hanwen/obj-id-len-v3
Pull-Request: https://github.com/git/git/pull/1185

Range-diff vs v2:

 1:  80d29e8f269 = 1:  80d29e8f269 Documentation: object_id_len goes up to 31
 2:  4c1a19fc4ae ! 2:  68e7bc32ff8 reftable: reject 0 object_id_len

@@ Commit message
reftable: reject 0 object_id_len

The spec says 2 <= object_id_len <= 31. We are lenient and allow 1,
- but we forbid 0, so we can we can be sure that we never read a
- 0-length key.
+ but we forbid 0, so we can be sure that we never read a 0-length key.

Signed-off-by Han-Wen Nienhuys <[email protected]>
 3:  600b115f8b1 = 3:  8b5aebdb07a reftable: add a test that verifies that writing empty keys fails
 4:  ba036ee8543 ! 4:  a9372cacd1b reftable: avoid writing empty keys at the block layer
     @@ reftable/block.c: int block_reader_first_key(struct block_reader *br, struct str
       	if (n < 0)
       		return n;
      +	if (!key->len)
     -+		return -1;
     ++		return REFTABLE_FORMAT_ERROR;
       
       	return 0;
       }
 5:  2bd3d44ba57 = 5:  0b8a42399dd reftable: ensure that obj_id_len is >= 2 on writing
 6:  82d36ee0e0d = 6:  bdccd969475 reftable: add test for length of disambiguating prefix
 7:  c6ffdb3471c = 7:  72499a14e38 reftable: rename writer_stats to reftable_writer_stats