commit-graph: fix memory leak when not writing graph

Commit Message

Patrick Steinhardt Dec. 18, 2023, 10:02 a.m. UTC

When `write_commit_graph()` bails out writing a split commit-graph early
then it may happen that we have already gathered the set of existing
commit-graph file names without yet determining the new merged set of
files. This can result in a memory leak though because we only clear the
preimage of files when we have collected the postimage.

Fix this issue by dropping the condition altogether so that we always
try to free both preimage and postimage filenames. As the context
structure is zero-initialized this simplification is safe to do.

Signed-off-by: Patrick Steinhardt <[e-mail]>
---
 commit-graph.c | 19 ++++++++-----------
 1 file changed, 8 insertions(+), 11 deletions(-)


base-commit: 1a87c842ece327d03d08096395969aca5e0a6996

Comments

Taylor Blau Jan. 5, 2024, 7:11 p.m. UTC | #1

On Mon, Dec 18, 2023 at 11:02:28AM +0100, Patrick Steinhardt wrote:
> When `write_commit_graph()` bails out writing a split commit-graph early
> then it may happen that we have already gathered the set of existing
> commit-graph file names without yet determining the new merged set of
> files. This can result in a memory leak though because we only clear the
> preimage of files when we have collected the postimage.
>
> Fix this issue by dropping the condition altogether so that we always
> try to free both preimage and postimage filenames. As the context
> structure is zero-initialized this simplification is safe to do.

Looks obviously good to me, thanks for finding and fixing.

Thanks,
Taylor

Patrick Steinhardt Jan. 15, 2024, 7:08 a.m. UTC | #2

On Fri, Jan 05, 2024 at 02:11:22PM -0500, Taylor Blau wrote:
> On Mon, Dec 18, 2023 at 11:02:28AM +0100, Patrick Steinhardt wrote:
> > When `write_commit_graph()` bails out writing a split commit-graph early
> > then it may happen that we have already gathered the set of existing
> > commit-graph file names without yet determining the new merged set of
> > files. This can result in a memory leak though because we only clear the
> > preimage of files when we have collected the postimage.
> >
> > Fix this issue by dropping the condition altogether so that we always
> > try to free both preimage and postimage filenames. As the context
> > structure is zero-initialized this simplification is safe to do.
> 
> Looks obviously good to me, thanks for finding and fixing.

Cc'ing Junio so that this fix doesn't fall off the radar. I thought I
saw the topic in "seen" once, but either I misremember or it got dropped
from there.

Patrick

Patch

diff --git a/commit-graph.c b/commit-graph.c
index acac9bf6e1..afe0177ab2 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -2617,19 +2617,16 @@  int write_commit_graph(struct object_directory *odb,
 	oid_array_clear(&ctx->oids);
 	clear_topo_level_slab(&topo_levels);
 
-	if (ctx->commit_graph_filenames_after) {
-		for (i = 0; i < ctx->num_commit_graphs_after; i++) {
-			free(ctx->commit_graph_filenames_after[i]);
-			free(ctx->commit_graph_hash_after[i]);
-		}
-
-		for (i = 0; i < ctx->num_commit_graphs_before; i++)
-			free(ctx->commit_graph_filenames_before[i]);
+	for (i = 0; i < ctx->num_commit_graphs_before; i++)
+		free(ctx->commit_graph_filenames_before[i]);
+	free(ctx->commit_graph_filenames_before);
 
-		free(ctx->commit_graph_filenames_after);
-		free(ctx->commit_graph_filenames_before);
-		free(ctx->commit_graph_hash_after);
+	for (i = 0; i < ctx->num_commit_graphs_after; i++) {
+		free(ctx->commit_graph_filenames_after[i]);
+		free(ctx->commit_graph_hash_after[i]);
 	}
+	free(ctx->commit_graph_filenames_after);
+	free(ctx->commit_graph_hash_after);
 
 	free(ctx);