git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v2 3/3] reftable: prevent 'update_index' changes after adding records

From
Karthik Nayak <karthik.188@gmail.com>
Date
Jan 21, 2025, 03:34 UTC
Message-ID
<20250121-461-corrupted-reftable-followup-v2-3-37e26c7a79b4@gmail.com>
In-Reply-To
<20250121-461-corrupted-reftable-followup-v2-0-37e26c7a79b4@gmail.com>

The function `reftable_writer_set_limits()` allows updating the 'min_update_index' and 'max_update_index' of a reftable writer. These values are written to both the writer's header and footer.

Since the header is written during the first block write, any subsequent changes to the update index would create a mismatch between the header and footer values. The footer would contain the newer values while the header retained the original ones.

To fix this bug, prevent callers from updating these values after any record is written. To do this, modify the function to return an error whenever the limits are modified after any record adds. Check for record adds within `reftable_writer_set_limits()` by checking the `last_key` variable, which is set whenever a new record is added.

Modify all callers of the function to anticipate a return type and handle it accordingly.

Helped-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Karthik Nayak <karthik.188@gmail.com>
---
 refs/reftable-backend.c         | 20 +++++++++++++++-----
 reftable/reftable-error.h       |  1 +
 reftable/reftable-writer.h      | 24 ++++++++++++++----------
 reftable/stack.c                |  6 ++++--
 reftable/writer.c               | 13 +++++++++++--
 t/unit-tests/t-reftable-stack.c |  8 +++++---
 6 files changed, 50 insertions(+), 22 deletions(-)
diff --git a/refs/reftable-backend.c b/refs/reftable-backend.c
index 6814c87bc618229ac8a70b904be3f850371ad876..9cfb0cb26721a9425c3b4a374f7b41e192037315 100644
--- a/refs/reftable-backend.c
+++ b/refs/reftable-backend.c
@@ -1443,7 +1443,9 @@ static int write_transaction_table(struct reftable_writer *writer, void *cb_data
 	 * multiple entries. Each entry will contain a different update_index,
 	 * so set the limits accordingly.
 	 */
-	reftable_writer_set_limits(writer, ts, ts + arg->max_index);
+	ret = reftable_writer_set_limits(writer, ts, ts + arg->max_index);
+	if (ret < 0)
+		goto done;
 
 	for (i = 0; i < arg->updates_nr; i++) {
 		struct reftable_transaction_update *tx_update = &arg->updates[i];
@@ -1766,7 +1768,9 @@ static int write_copy_table(struct reftable_writer *writer, void *cb_data)
 	deletion_ts = creation_ts = reftable_stack_next_update_index(arg->be->stack);
 	if (arg->delete_old)
 		creation_ts++;
-	reftable_writer_set_limits(writer, deletion_ts, creation_ts);
+	ret = reftable_writer_set_limits(writer, deletion_ts, creation_ts);
+	if (ret < 0)
+		goto done;
 
 	/*
 	 * Add the new reference. If this is a rename then we also delete the
@@ -2298,7 +2302,9 @@ static int write_reflog_existence_table(struct reftable_writer *writer,
 	if (ret <= 0)
 		goto done;
 
-	reftable_writer_set_limits(writer, ts, ts);
+	ret = reftable_writer_set_limits(writer, ts, ts);
+	if (ret < 0)
+		goto done;
 
 	/*
 	 * The existence entry has both old and new object ID set to the
@@ -2357,7 +2363,9 @@ static int write_reflog_delete_table(struct reftable_writer *writer, void *cb_da
 	uint64_t ts = reftable_stack_next_update_index(arg->stack);
 	int ret;
 
-	reftable_writer_set_limits(writer, ts, ts);
+	ret = reftable_writer_set_limits(writer, ts, ts);
+	if (ret < 0)
+		goto out;
 
 	ret = reftable_stack_init_log_iterator(arg->stack, &it);
 	if (ret < 0)
@@ -2434,7 +2442,9 @@ static int write_reflog_expiry_table(struct reftable_writer *writer, void *cb_da
 		if (arg->records[i].value_type == REFTABLE_LOG_UPDATE)
 			live_records++;
 
-	reftable_writer_set_limits(writer, ts, ts);
+	ret = reftable_writer_set_limits(writer, ts, ts);
+	if (ret < 0)
+		return ret;
 
 	if (!is_null_oid(&arg->update_oid)) {
 		struct reftable_ref_record ref = {0};
diff --git a/reftable/reftable-error.h b/reftable/reftable-error.h
index f4048265629fe456207b88620658193f770a84f0..a7e33d964d0cfe5546f588d26c0fcb66ab326828 100644
--- a/reftable/reftable-error.h
+++ b/reftable/reftable-error.h
@@ -30,6 +30,7 @@ enum reftable_error {
 
 	/* Misuse of the API:
 	 *  - on writing a record with NULL refname.
+	 *  - on writing a record before setting the writer limits.
 	 *  - on writing a reftable_ref_record outside the table limits
 	 *  - on writing a ref or log record before the stack's
 	 * next_update_inde*x
diff --git a/reftable/reftable-writer.h b/reftable/reftable-writer.h
index 5f9afa620bb00de66c311765fb0ae8c6f56401ae..1ea014d389cc47f173279e3234a82f3fcbc807a0 100644
--- a/reftable/reftable-writer.h
+++ b/reftable/reftable-writer.h
@@ -124,17 +124,21 @@ int reftable_writer_new(struct reftable_writer **out,
 			int (*flush_func)(void *),
 			void *writer_arg, const struct reftable_write_options *opts);
 
-/* Set the range of update indices for the records we will add. When writing a
-   table into a stack, the min should be at least
-   reftable_stack_next_update_index(), or REFTABLE_API_ERROR is returned.
-
-   For transactional updates to a stack, typically min==max, and the
-   update_index can be obtained by inspeciting the stack. When converting an
-   existing ref database into a single reftable, this would be a range of
-   update-index timestamps.
+/*
+ * Set the range of update indices for the records we will add. When writing a
+ * table into a stack, the min should be at least
+ * reftable_stack_next_update_index(), or REFTABLE_API_ERROR is returned.
+ *
+ * For transactional updates to a stack, typically min==max, and the
+ * update_index can be obtained by inspeciting the stack. When converting an
+ * existing ref database into a single reftable, this would be a range of
+ * update-index timestamps.
+ *
+ * The function should be called before adding any records to the writer. If not
+ * it will fail with REFTABLE_API_ERROR.
  */
-void reftable_writer_set_limits(struct reftable_writer *w, uint64_t min,
-				uint64_t max);
+int reftable_writer_set_limits(struct reftable_writer *w, uint64_t min,
+			       uint64_t max);
 
 /*
   Add a reftable_ref_record. The record should have names that come after
diff --git a/reftable/stack.c b/reftable/stack.c
index 531660a49f0948c33041831ee0d740feacb22b2f..9649dbbb04c51e106ee752f14481bbad381cb348 100644
--- a/reftable/stack.c
+++ b/reftable/stack.c
@@ -1058,8 +1058,10 @@ static int stack_write_compact(struct reftable_stack *st,
 
 	for (size_t i = first; i <= last; i++)
 		st->stats.bytes += st->readers[i]->size;
-	reftable_writer_set_limits(wr, st->readers[first]->min_update_index,
-				   st->readers[last]->max_update_index);
+	err = reftable_writer_set_limits(wr, st->readers[first]->min_update_index,
+					 st->readers[last]->max_update_index);
+	if (err < 0)
+		goto done;
 
 	err = reftable_merged_table_new(&mt, st->readers + first, subtabs_len,
 					st->opts.hash_id);
diff --git a/reftable/writer.c b/reftable/writer.c
index 740c98038eaf883258bef4988f78977ac7e4a75a..03acbdbcce75fd51820c5fb016bd94f0f7f4914a 100644
--- a/reftable/writer.c
+++ b/reftable/writer.c
@@ -179,11 +179,20 @@ int reftable_writer_new(struct reftable_writer **out,
 	return 0;
 }
 
-void reftable_writer_set_limits(struct reftable_writer *w, uint64_t min,
-				uint64_t max)
+int reftable_writer_set_limits(struct reftable_writer *w, uint64_t min,
+			       uint64_t max)
 {
+	/*
+	 * The limits should be set before any records are added to the writer.
+	 * Check if any records were added by checking if `last_key` was set.
+	 */
+	if (w->last_key.len)
+		return REFTABLE_API_ERROR;
+
 	w->min_update_index = min;
 	w->max_update_index = max;
+
+	return 0;
 }
 
 static void writer_release(struct reftable_writer *w)
diff --git a/t/unit-tests/t-reftable-stack.c b/t/unit-tests/t-reftable-stack.c
index aeec195b2b1014445d71c5db39a9795017fd8ff2..b23edf18a7d75b0c2292490ad06d4dfaaa571e79 100644
--- a/t/unit-tests/t-reftable-stack.c
+++ b/t/unit-tests/t-reftable-stack.c
@@ -103,7 +103,8 @@ static void t_read_file(void)
 static int write_test_ref(struct reftable_writer *wr, void *arg)
 {
 	struct reftable_ref_record *ref = arg;
-	reftable_writer_set_limits(wr, ref->update_index, ref->update_index);
+	check(!reftable_writer_set_limits(wr, ref->update_index,
+					  ref->update_index));
 	return reftable_writer_add_ref(wr, ref);
 }
 
@@ -143,7 +144,8 @@ static int write_test_log(struct reftable_writer *wr, void *arg)
 {
 	struct write_log_arg *wla = arg;
 
-	reftable_writer_set_limits(wr, wla->update_index, wla->update_index);
+	check(!reftable_writer_set_limits(wr, wla->update_index,
+					  wla->update_index));
 	return reftable_writer_add_log(wr, wla->log);
 }
 
@@ -961,7 +963,7 @@ static void t_reflog_expire(void)
 
 static int write_nothing(struct reftable_writer *wr, void *arg UNUSED)
 {
-	reftable_writer_set_limits(wr, 1, 1);
+	check(!reftable_writer_set_limits(wr, 1, 1));
 	return 0;
 }
 

-- 
2.47.0
Previous: Karthik NayakNext: Patrick Steinhardt
Message 13 of 30 in “refs: small followups to the migration corruption fix”
  1. 0/3 refs: small followups to the migration corruption fixKarthik Nayak, Jan 17, 2025
  2. 1/3 refs: mark `ref_transaction_update_reflog()` as staticKarthik Nayak, Jan 17, 2025
  3. Patrick SteinhardtJan 17, 2025
  4. Karthik NayakJan 20, 2025
  5. 2/3 refs: use 'uint64_t' for 'ref_update.index'Karthik Nayak, Jan 17, 2025
  6. 3/3 reftable: prevent 'update_index' changes after header writeKarthik Nayak, Jan 17, 2025
  7. Patrick SteinhardtJan 17, 2025
  8. Karthik NayakJan 20, 2025
  9. Karthik NayakJan 20, 2025
  10. 0/3 refs: small followups to the migration corruption fixKarthik Nayak, Jan 21, 2025
  11. 1/3 refs: mark `ref_transaction_update_reflog()` as staticKarthik Nayak, Jan 21, 2025
  12. 2/3 refs: use 'uint64_t' for 'ref_update.index'Karthik Nayak, Jan 21, 2025
  13. 3/3 reftable: prevent 'update_index' changes after adding recordsKarthik Nayak, Jan 21, 2025
  14. Patrick SteinhardtJan 21, 2025
  15. Karthik NayakJan 21, 2025
  16. 0/3 refs: small followups to the migration corruption fixKarthik Nayak, Jan 22, 2025
  17. 1/3 refs: mark `ref_transaction_update_reflog()` as staticKarthik Nayak, Jan 22, 2025
  18. 2/3 refs: use 'uint64_t' for 'ref_update.index'Karthik Nayak, Jan 22, 2025
  19. 3/3 reftable: prevent 'update_index' changes after adding recordsKarthik Nayak, Jan 22, 2025
  20. Patrick SteinhardtJan 22, 2025
  21. Junio C HamanoJan 22, 2025
  22. Junio C HamanoJan 22, 2025
  23. undefined behavior in unit tests, was Re: [PATCH v3 3/3] reftable: prevent 'update_index' changes after adding recordsJeff King, Feb 1, 2025
  24. Phillip WoodFeb 1, 2025
  25. Patrick SteinhardtFeb 3, 2025
  26. Junio C HamanoFeb 3, 2025
  27. Jeff KingFeb 3, 2025
  28. Patrick SteinhardtFeb 3, 2025
  29. Karthik NayakFeb 3, 2025
  30. Jeff KingFeb 3, 2025

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.