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

[PATCH 13/20] commit-graph: detect out-of-bounds extra-edges pointers

From
Jeff King <peff@peff.net>
Date
Oct 9, 2023, 21:05 UTC
Message-ID
<20231009210538.GM3282181@coredump.intra.peff.net>
In-Reply-To
<20231009205544.GA3281950@coredump.intra.peff.net>

If an entry in a commit-graph file has more than 2 parents, the fixed-size parent fields instead point to an offset within an "extra edges" chunk. We blindly follow these, assuming that the chunk is present and sufficiently large; this can lead to an out-of-bounds read for a corrupt or malicious file.

We can fix this by recording the size of the chunk and adding a bounds-check in fill_commit_in_graph(). There are a few tricky bits:

  1. We'll switch from working with a pointer to an offset. This makes
     some corner cases just fall out naturally:
      a. If we did not find an EDGE chunk at all, our size will
         correctly be zero (so everything is "out of bounds").
      b. Comparing "size / 4" lets us make sure we have at least 4 bytes
         to read, and we never compute a pointer more than one element
         past the end of the array (computing a larger pointer is
         probably OK in practice, but is technically undefined
         behavior).
      c. The current code casts to "uint32_t *". Replacing it with an
         offset avoids any comparison between different types of pointer
         (since the chunk is stored as "unsigned char *").
  2. This is the first case in which fill_commit_in_graph() may return
     anything but success. We need to make sure to roll back the
     "parsed" flag (and any parents we might have added before running
     out of buffer) so that the caller can cleanly fall back to
     loading the commit object itself.
     It's a little non-trivial to do this, and we might benefit from
     factoring it out. But we can wait on that until we actually see a
     second case where we return an error.

As a bonus, this lets us drop the st_mult() call. Since we've already done a bounds check, we know there won't be any integer overflow (it would imply our buffer is larger than a size_t can hold).

The included test does not actually segfault before this patch (though you could construct a case where it does). Instead, it reads garbage from the next chunk which results in it complaining about a bogus parent id. This is sufficient for our needs, though (we care that the fallback succeeds, and that stderr mentions the out-of-bounds read).

Signed-off-by: Jeff King <peff@peff.net>
---
 commit-graph.c          | 20 ++++++++++++++------
 commit-graph.h          |  1 +
 t/t5318-commit-graph.sh |  8 ++++++++
 3 files changed, 23 insertions(+), 6 deletions(-)
diff --git a/commit-graph.c b/commit-graph.c
index 9b80bbd75b..e4860841fc 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -433,7 +433,8 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	read_chunk(cf, GRAPH_CHUNKID_OIDFANOUT, graph_read_oid_fanout, graph);
 	read_chunk(cf, GRAPH_CHUNKID_OIDLOOKUP, graph_read_oid_lookup, graph);
 	read_chunk(cf, GRAPH_CHUNKID_DATA, graph_read_commit_data, graph);
-	pair_chunk_unsafe(cf, GRAPH_CHUNKID_EXTRAEDGES, &graph->chunk_extra_edges);
+	pair_chunk(cf, GRAPH_CHUNKID_EXTRAEDGES, &graph->chunk_extra_edges,
+		   &graph->chunk_extra_edges_size);
 	pair_chunk_unsafe(cf, GRAPH_CHUNKID_BASE, &graph->chunk_base_graphs);
 
 	if (s->commit_graph_generation_version >= 2) {
@@ -899,7 +900,7 @@ static int fill_commit_in_graph(struct repository *r,
 				struct commit_graph *g, uint32_t pos)
 {
 	uint32_t edge_value;
-	uint32_t *parent_data_ptr;
+	uint32_t parent_data_pos;
 	struct commit_list **pptr;
 	const unsigned char *commit_data;
 	uint32_t lex_index;
@@ -931,14 +932,21 @@ static int fill_commit_in_graph(struct repository *r,
 		return 1;
 	}
 
-	parent_data_ptr = (uint32_t*)(g->chunk_extra_edges +
-			  st_mult(4, edge_value & GRAPH_EDGE_LAST_MASK));
+	parent_data_pos = edge_value & GRAPH_EDGE_LAST_MASK;
 	do {
-		edge_value = get_be32(parent_data_ptr);
+		if (g->chunk_extra_edges_size / sizeof(uint32_t) <= parent_data_pos) {
+			error("commit-graph extra-edges pointer out of bounds");
+			free_commit_list(item->parents);
+			item->parents = NULL;
+			item->object.parsed = 0;
+			return 0;
+		}
+		edge_value = get_be32(g->chunk_extra_edges +
+				      sizeof(uint32_t) * parent_data_pos);
 		pptr = insert_parent_or_die(r, g,
 					    edge_value & GRAPH_EDGE_LAST_MASK,
 					    pptr);
-		parent_data_ptr++;
+		parent_data_pos++;
 	} while (!(edge_value & GRAPH_LAST_EDGE));
 
 	return 1;
diff --git a/commit-graph.h b/commit-graph.h
index 20ada7e891..1f8a9de4fb 100644
--- a/commit-graph.h
+++ b/commit-graph.h
@@ -95,6 +95,7 @@ struct commit_graph {
 	const unsigned char *chunk_generation_data;
 	const unsigned char *chunk_generation_data_overflow;
 	const unsigned char *chunk_extra_edges;
+	size_t chunk_extra_edges_size;
 	const unsigned char *chunk_base_graphs;
 	const unsigned char *chunk_bloom_indexes;
 	const unsigned char *chunk_bloom_data;
diff --git a/t/t5318-commit-graph.sh b/t/t5318-commit-graph.sh
index 492460157d..05bafcfe5f 100755
--- a/t/t5318-commit-graph.sh
+++ b/t/t5318-commit-graph.sh
@@ -879,4 +879,12 @@ test_expect_success 'reader notices too-small commit data chunk' '
 	test_cmp expect.err err
 '
 
+test_expect_success 'reader notices out-of-bounds extra edge' '
+	check_corrupt_chunk EDGE clear &&
+	cat >expect.err <<-\EOF &&
+	error: commit-graph extra-edges pointer out of bounds
+	EOF
+	test_cmp expect.err err
+'
+
 test_done
-- 
2.42.0.884.g35e1fe1a6a
Previous: Jeff KingNext: Taylor Blau
Message 34 of 67 in “bounds-checks for chunk-based files”
  1. 0/20 bounds-checks for chunk-based filesJeff King, Oct 9, 2023
  2. 01/20 chunk-format: note that pair_chunk() is unsafeJeff King, Oct 9, 2023
  3. Taylor BlauOct 10, 2023
  4. Jeff KingOct 11, 2023
  5. 02/20 t: add library for munging chunk-format filesJeff King, Oct 9, 2023
  6. Taylor BlauOct 10, 2023
  7. 03/20 midx: stop ignoring malformed oid fanout chunkJeff King, Oct 9, 2023
  8. Taylor BlauOct 10, 2023
  9. Jeff KingOct 11, 2023
  10. 04/20 commit-graph: check size of oid fanout chunkJeff King, Oct 9, 2023
  11. Taylor BlauOct 11, 2023
  12. Taylor BlauOct 11, 2023
  13. Jeff KingOct 11, 2023
  14. 05/20 midx: check size of oid lookup chunkJeff King, Oct 9, 2023
  15. 06/20 commit-graph: check consistency of fanout tableJeff King, Oct 9, 2023
  16. Taylor BlauOct 11, 2023
  17. 07/20 midx: check size of pack names chunkJeff King, Oct 9, 2023
  18. Taylor BlauOct 11, 2023
  19. Jeff KingOct 11, 2023
  20. 08/20 midx: enforce chunk alignment on readingJeff King, Oct 9, 2023
  21. Taylor BlauOct 11, 2023
  22. Taylor BlauOct 11, 2023
  23. Jeff KingOct 11, 2023
  24. 09/20 midx: check size of object offset chunkJeff King, Oct 9, 2023
  25. Taylor BlauOct 11, 2023
  26. 10/20 midx: bounds-check large offset chunkJeff King, Oct 9, 2023
  27. Taylor BlauOct 11, 2023
  28. Jeff KingOct 11, 2023
  29. 11/20 midx: check size of revindex chunkJeff King, Oct 9, 2023
  30. Taylor BlauOct 11, 2023
  31. 12/20 commit-graph: check size of commit data chunkJeff King, Oct 9, 2023
  32. Taylor BlauOct 11, 2023
  33. Jeff KingOct 11, 2023
  34. 13/20 commit-graph: detect out-of-bounds extra-edges pointersJeff King, Oct 9, 2023
  35. Taylor BlauOct 11, 2023
  36. 14/20 commit-graph: bounds-check base graphs chunkJeff King, Oct 9, 2023
  37. Taylor BlauOct 11, 2023
  38. 15/20 commit-graph: check size of generations chunkJeff King, Oct 9, 2023
  39. 16/20 commit-graph: bounds-check generation overflow chunkJeff King, Oct 9, 2023
  40. 17/20 commit-graph: check bounds when accessing BDAT chunkJeff King, Oct 9, 2023
  41. Taylor BlauOct 11, 2023
  42. Jeff KingOct 11, 2023
  43. 18/20 commit-graph: check bounds when accessing BIDX chunkJeff King, Oct 9, 2023
  44. Taylor BlauOct 11, 2023
  45. 19/20 commit-graph: detect out-of-order BIDX offsetsJeff King, Oct 9, 2023
  46. Taylor BlauOct 11, 2023
  47. 20/20 chunk-format: drop pair_chunk_unsafe()Jeff King, Oct 9, 2023
  48. Taylor BlauOct 11, 2023
  49. Jeff KingOct 11, 2023
  50. 0/8 chunk-format: introduce `pair_chunk_expect()` convenience APITaylor Blau, Oct 13, 2023
  51. 1/8 chunk-format: introduce `pair_chunk_expect()` helperTaylor Blau, Oct 13, 2023
  52. 2/8 commit-graph: read `OIDF` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  53. 3/8 commit-graph: read `CDAT` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  54. 4/8 commit-graph: read `GDAT` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  55. 5/8 commit-graph: read `BIDX` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  56. Taylor BlauOct 13, 2023
  57. Junio C HamanoOct 14, 2023
  58. Jeff KingOct 20, 2023
  59. 6/8 midx: read `OIDF` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  60. Junio C HamanoOct 13, 2023
  61. 7/8 midx: read `OIDL` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  62. 8/8 midx: read `OOFF` chunk with `pair_chunk_expect()`Taylor Blau, Oct 13, 2023
  63. Jeff KingOct 20, 2023
  64. 21/20 t5319: make corrupted large-offset test more robustJeff King, Oct 14, 2023
  65. Junio C HamanoOct 14, 2023
  66. Jeff KingOct 15, 2023
  67. Junio C HamanoOct 15, 2023

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.