Re: [PATCH] builtin/blame: destroy initialized commit_info only

From: Jeff King <peff@peff.net>
Date: Feb 9, 2015, 23:37 UTC
Message-ID: <20150209233707.GA20462@peff.net>
In-Reply-To: <20150209232435.GB24814@peff.net>

On Mon, Feb 09, 2015 at 06:24:35PM -0500, Jeff King wrote:

Show 9 quoted lines

> Clang's address sanitizer has compiler support, so it does get to see
> this memory and could put a canary value in for each loop iteration. But
> it doesn't. Instead, you're supposed to use the "memory sanitizer" to
> catch uninitialized memory.
> 
> I tried that, but got overwhelmed with false positives. Like valgrind,
> it has problems accepting that memory written by zlib is actually
> initialized. But in theory, if we went to the work to annotate some
> false positives, it should be able to find this problem.

I got rid of the false positives here, through a combination of compiling with NO_OPENSSL (since it otherwise doesn't know that git_SHA1_Final is initializing hashes), and this patch which lets it assume that the output of zlib (at least for these cases) is always initialized:

diff --git a/sha1_file.c b/sha1_file.c
index 30995e6..28c8f84 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -1682,6 +1682,7 @@ unsigned long get_size_from_delta(struct packed_git *p,
 	git_zstream stream;
 	int st;
 
+	memset(delta_head, 0, 20);
 	memset(&stream, 0, sizeof(stream));
 	stream.next_out = delta_head;
 	stream.avail_out = sizeof(delta_head);
@@ -1973,6 +1974,7 @@ static void *unpack_compressed_entry(struct packed_git *p,
 	buffer = xmallocz_gently(size);
 	if (!buffer)
 		return NULL;
+	memset(buffer, 0, size);
 	memset(&stream, 0, sizeof(stream));
 	stream.next_out = buffer;
 	stream.avail_out = size + 1;


Sadly, though, the test case in question runs to completion. It does not
seem to detect our use of uninitialized memory. :(

-Peff