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

Re: [PATCH 3/6] Make sure fsck_commit_buffer() does not run out of the buffer

From
Junio C Hamano <gitster@pobox.com>
Date
Aug 28, 2014, 20:59 UTC
Message-ID
<xmqqr4008h4p.fsf@gitster.dls.corp.google.com>
In-Reply-To
<alpine.DEB.1.00.1408281646450.990@s15462909.onlinehome-server.info>
Johannes Schindelin <johannes.schindelin@gmx.de> writes:
Show 12 quoted lines
> So far, we assumed that the buffer is NUL terminated, but this is not
> a safe assumption, now that we opened the fsck_object() API to pass a
> buffer directly.
>
> So let's make sure that there is at least an empty line in the buffer.
> That way, our checks would fail if the empty line was encountered
> prematurely, and consequently we can get away with the current string
> comparisons even with non-NUL-terminated buffers are passed to
> fsck_object().
>
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> ---
Heh, I probably should have read this one before commenting on 2/6.

It makes me feel somewhat uneasy to insist that there must be a blank line in the commit object, even though from the very first implementation of "commit-tree" I think we have always had a blank there at the end of the header, even when you feed nothing as the message to it.

I think the new check is OK, but the message should be s/empty line/end of header/ or something. It is not like we require an empty line in the log message proper.

Show 44 quoted lines
>  fsck.c | 23 +++++++++++++++++++++++
>  1 file changed, 23 insertions(+)
>
> diff --git a/fsck.c b/fsck.c
> index dd77628..db6aaa4 100644
> --- a/fsck.c
> +++ b/fsck.c
> @@ -237,6 +237,26 @@ static int fsck_tree(struct tree *item, int strict, fsck_error error_func)
>  	return retval;
>  }
>  
> +static int must_have_empty_line(const void *data, unsigned long size,
> +	struct object *obj, fsck_error error_func)
> +{
> +	const char *buffer = (const char *)data;
> +	int i;
> +
> +	for (i = 0; i < size; i++) {
> +		switch (buffer[i]) {
> +		case '\0':
> +			return error_func(obj, FSCK_ERROR,
> +				"invalid message: NUL at offset %d", i);
> +		case '\n':
> +			if (i + 1 < size && buffer[i + 1] == '\n')
> +				return 0;
> +		}
> +	}
> +
> +	return error_func(obj, FSCK_ERROR, "invalid buffer: missing empty line");
> +}
> +
>  static int fsck_ident(const char **ident, struct object *obj, fsck_error error_func)
>  {
>  	char *end;
> @@ -284,6 +304,9 @@ static int fsck_commit_buffer(struct commit *commit, const char *buffer,
>  	unsigned parent_count, parent_line_count = 0;
>  	int err;
>  
> +	if (must_have_empty_line(buffer, size, &commit->object, error_func))
> +		return -1;
> +
>  	if (!skip_prefix(buffer, "tree ", &buffer))
>  		return error_func(&commit->object, FSCK_ERROR, "invalid format - expected 'tree' line");
>  	if (get_sha1_hex(buffer, tree_sha1) || buffer[40] != '\n')
Previous: Johannes SchindelinNext: Jeff King
Message 9 of 68 in “Improve tag checking in fsck and with transfer.fsckobjects”
  1. 0/6 Improve tag checking in fsck and with transfer.fsckobjectsJohannes Schindelin, Aug 28, 2014
  2. 1/6 Refactor type_from_string() to avoid die()ing in case of errorsJohannes Schindelin, Aug 28, 2014
  3. Junio C HamanoAug 28, 2014
  4. 2/6 Accept object data in the fsck_object() functionJohannes Schindelin, Aug 28, 2014
  5. Junio C HamanoAug 28, 2014
  6. Jeff KingAug 29, 2014
  7. Jeff KingAug 29, 2014
  8. 3/6 Make sure fsck_commit_buffer() does not run out of the bufferJohannes Schindelin, Aug 28, 2014
  9. Junio C HamanoAug 28, 2014
  10. Jeff KingAug 29, 2014
  11. 4/6 fsck: check tag objects' headersJohannes Schindelin, Aug 28, 2014
  12. Junio C HamanoAug 28, 2014
  13. Junio C HamanoAug 28, 2014
  14. Jeff KingAug 29, 2014
  15. Junio C HamanoAug 31, 2014
  16. Jeff KingSep 3, 2014
  17. Junio C HamanoSep 3, 2014
  18. Jeff KingSep 4, 2014
  19. Jeff KingAug 29, 2014
  20. Junio C HamanoSep 2, 2014
  21. Jeff KingSep 3, 2014
  22. 5/6 Add regression tests for stricter tag fsck'ingJohannes Schindelin, Aug 28, 2014
  23. 6/6 Make sure that index-pack --strict fails upon invalid tag objectsJohannes Schindelin, Aug 28, 2014
  24. 0/6 Improve tag checking in fsck and with transfer.fsckobjectsJohannes Schindelin, Sep 10, 2014
  25. Johannes SchindelinSep 10, 2014
  26. Junio C HamanoSep 10, 2014
  27. Junio C HamanoSep 10, 2014
  28. Johannes SchindelinSep 11, 2014
  29. 0/6 Improve tag checking in fsck and with transfer.fsckobjectsJohannes Schindelin, Sep 11, 2014
  30. 1/6 Refactor type_from_string() to avoid die()ing in case of errorsJohannes Schindelin, Sep 11, 2014
  31. 2/6 Accept object data in the fsck_object() functionJohannes Schindelin, Sep 11, 2014
  32. 3/6 Make sure fsck_commit_buffer() does not run out of the bufferJohannes Schindelin, Sep 11, 2014
  33. 4/6 fsck: check tag objects' headersJohannes Schindelin, Sep 11, 2014
  34. 5/6 Add regression tests for stricter tag fsck'ingJohannes Schindelin, Sep 11, 2014
  35. 6/6 Make sure that index-pack --strict checks tag objectsJohannes Schindelin, Sep 11, 2014
  36. Junio C HamanoSep 11, 2014
  37. Junio C HamanoSep 11, 2014
  38. 0/3 hash-object --literallyJunio C Hamano, Sep 11, 2014
  39. 1/3 hash-object: reduce file-scope staticsJunio C Hamano, Sep 11, 2014
  40. 2/3 hash-object: pass 'write_object' as a flagJunio C Hamano, Sep 11, 2014
  41. 3/3 hash-object: add --literally optionJunio C Hamano, Sep 11, 2014
  42. Johannes SchindelinSep 12, 2014
  43. 0/6 Improve tag checking in fsck and with transfer.fsckobjectsJohannes Schindelin, Sep 12, 2014
  44. 1/6 Refactor type_from_string() to avoid die()ing in case of errorsJohannes Schindelin, Sep 12, 2014
  45. 2/6 Accept object data in the fsck_object() functionJohannes Schindelin, Sep 12, 2014
  46. 3/6 Make sure fsck_commit_buffer() does not run out of the bufferJohannes Schindelin, Sep 12, 2014
  47. 4/6 fsck: check tag objects' headersJohannes Schindelin, Sep 12, 2014
  48. 5/6 Add regression tests for stricter tag fsck'ingJohannes Schindelin, Sep 12, 2014
  49. 6/6 Make sure that index-pack --strict checks tag objectsJohannes Schindelin, Sep 12, 2014
  50. Junio C HamanoSep 12, 2014
  51. Johannes SchindelinSep 13, 2014
  52. 1/6 Refactor type_from_string() to avoid die()ing in case of errorsJohannes Schindelin, Sep 10, 2014
  53. 2/6 Accept object data in the fsck_object() functionJohannes Schindelin, Sep 10, 2014
  54. 3/6 Make sure fsck_commit_buffer() does not run out of the bufferJohannes Schindelin, Sep 10, 2014
  55. Junio C HamanoSep 10, 2014
  56. Johannes SchindelinSep 11, 2014
  57. Junio C HamanoSep 11, 2014
  58. Eric SunshineSep 10, 2014
  59. 4/6 fsck: check tag objects' headersJohannes Schindelin, Sep 10, 2014
  60. Junio C HamanoSep 10, 2014
  61. 5/6 Add regression tests for stricter tag fsck'ingJohannes Schindelin, Sep 10, 2014
  62. Junio C HamanoSep 10, 2014
  63. Johannes SchindelinSep 11, 2014
  64. 6/6 Make sure that index-pack --strict fails upon invalid tag objectsJohannes Schindelin, Sep 10, 2014
  65. Junio C HamanoSep 10, 2014
  66. Johannes SchindelinSep 11, 2014
  67. Junio C HamanoSep 11, 2014
  68. Johannes SchindelinSep 11, 2014

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.