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

Re: [PATCH v2 04/12] builtin/for-each-repo: remove unnecessary argv copy to plug leak

From
Junio C Hamano <gitster@pobox.com>
Date
Jul 26, 2021, 20:02 UTC
Message-ID
<xmqqmtq8j12y.fsf@gitster.g>
In-Reply-To
<20210725130830.5145-5-andrzej@ahunt.org>
andrzej@ahunt.org writes:
Show 7 quoted lines
> From: Andrzej Hunt <ajrhunt@google.com>
>
> cmd_for_each_repo() copies argv into args (a strvec), which is later
> passed into run_command_on_repo(), which in turn copies that strvec onto
> the end of child.args. The initial copy is unnecessary (we never modify
> args). We therefore choose to just pass argv directly into
> run_command_on_repo(), which lets us avoid the copy and fixes the leak.
Makes sense.  There is no reason to make these copies.
Show 91 quoted lines
>
> LSAN output from t0068:
>
> Direct leak of 192 byte(s) in 1 object(s) allocated from:
>     #0 0x7f63bd4ab8b0 in realloc (/usr/lib64/libasan.so.4+0xdc8b0)
>     #1 0x98d7e6 in xrealloc wrapper.c:126
>     #2 0x916914 in strvec_push_nodup strvec.c:19
>     #3 0x916a6e in strvec_push strvec.c:26
>     #4 0x4be4eb in cmd_for_each_repo builtin/for-each-repo.c:49
>     #5 0x410dcd in run_builtin git.c:475
>     #6 0x410dcd in handle_builtin git.c:729
>     #7 0x414087 in run_argv git.c:818
>     #8 0x414087 in cmd_main git.c:949
>     #9 0x40e9ec in main common-main.c:52
>     #10 0x7f63bc9fa349 in __libc_start_main (/lib64/libc.so.6+0x24349)
>
> Indirect leak of 22 byte(s) in 2 object(s) allocated from:
>     #0 0x7f63bd445e30 in __interceptor_strdup (/usr/lib64/libasan.so.4+0x76e30)
>     #1 0x98d698 in xstrdup wrapper.c:29
>     #2 0x916a63 in strvec_push strvec.c:26
>     #3 0x4be4eb in cmd_for_each_repo builtin/for-each-repo.c:49
>     #4 0x410dcd in run_builtin git.c:475
>     #5 0x410dcd in handle_builtin git.c:729
>     #6 0x414087 in run_argv git.c:818
>     #7 0x414087 in cmd_main git.c:949
>     #8 0x40e9ec in main common-main.c:52
>     #9 0x7f63bc9fa349 in __libc_start_main (/lib64/libc.so.6+0x24349)
>
> See also discussion about the original implementation below - this code
> appears to have evolved from a callback explaining the double-strvec-copy
> pattern, but there's no strong reason to keep that now:
>   https://lore.kernel.org/git/68bbeca5-314b-08ee-ef36-040e3f3814e9@gmail.com/
>
> Signed-off-by: Andrzej Hunt <andrzej@ahunt.org>
> ---
>  builtin/for-each-repo.c | 14 ++++----------
>  1 file changed, 4 insertions(+), 10 deletions(-)
>
> diff --git a/builtin/for-each-repo.c b/builtin/for-each-repo.c
> index 52be64a437..fd86e5a861 100644
> --- a/builtin/for-each-repo.c
> +++ b/builtin/for-each-repo.c
> @@ -10,18 +10,16 @@ static const char * const for_each_repo_usage[] = {
>  	NULL
>  };
>  
> -static int run_command_on_repo(const char *path,
> -			       void *cbdata)
> +static int run_command_on_repo(const char *path, int argc, const char ** argv)
>  {
>  	int i;
>  	struct child_process child = CHILD_PROCESS_INIT;
> -	struct strvec *args = (struct strvec *)cbdata;
>  
>  	child.git_cmd = 1;
>  	strvec_pushl(&child.args, "-C", path, NULL);
>  
> -	for (i = 0; i < args->nr; i++)
> -		strvec_push(&child.args, args->v[i]);
> +	for (i = 0; i < argc; i++)
> +		strvec_push(&child.args, argv[i]);
>  
>  	return run_command(&child);
>  }
> @@ -31,7 +29,6 @@ int cmd_for_each_repo(int argc, const char **argv, const char *prefix)
>  	static const char *config_key = NULL;
>  	int i, result = 0;
>  	const struct string_list *values;
> -	struct strvec args = STRVEC_INIT;
>  
>  	const struct option options[] = {
>  		OPT_STRING(0, "config", &config_key, N_("config"),
> @@ -45,9 +42,6 @@ int cmd_for_each_repo(int argc, const char **argv, const char *prefix)
>  	if (!config_key)
>  		die(_("missing --config=<config>"));
>  
> -	for (i = 0; i < argc; i++)
> -		strvec_push(&args, argv[i]);
> -
>  	values = repo_config_get_value_multi(the_repository,
>  					     config_key);
>  
> @@ -59,7 +53,7 @@ int cmd_for_each_repo(int argc, const char **argv, const char *prefix)
>  		return 0;
>  
>  	for (i = 0; !result && i < values->nr; i++)
> -		result = run_command_on_repo(values->items[i].string, &args);
> +		result = run_command_on_repo(values->items[i].string, argc, argv);
>  
>  	return result;
>  }
Previous: andrzej@ahunt.orgNext: andrzej@ahunt.org
Message 38 of 51 in “Fix all leaks in tests t0002-t0099: Part 2”
  1. 00/12 Fix all leaks in tests t0002-t0099: Part 2andrzej@ahunt.org, Jun 20, 2021
  2. 01/12 fmt-merge-msg: free newly allocated temporary strings when doneandrzej@ahunt.org, Jun 20, 2021
  3. Elijah NewrenJun 21, 2021
  4. 02/12 environment: move strbuf into block to plug leakandrzej@ahunt.org, Jun 20, 2021
  5. Elijah NewrenJun 21, 2021
  6. René ScharfeJun 26, 2021
  7. 03/12 builtin/submodule--helper: release unused strbuf to avoid leakandrzej@ahunt.org, Jun 20, 2021
  8. 04/12 builtin/for-each-repo: remove unnecessary argv copy to plug leakandrzej@ahunt.org, Jun 20, 2021
  9. Elijah NewrenJun 21, 2021
  10. 05/12 diffcore-rename: move old_dir/new_dir definition to plug leakandrzej@ahunt.org, Jun 20, 2021
  11. Elijah NewrenJun 21, 2021
  12. 06/12 ref-filter: also free head for ATOM_HEAD to avoid leakandrzej@ahunt.org, Jun 20, 2021
  13. Elijah NewrenJun 21, 2021
  14. 07/12 read-cache: call diff_setup_done to avoid leakandrzej@ahunt.org, Jun 20, 2021
  15. Elijah NewrenJun 21, 2021
  16. 08/12 convert: release strbuf to avoid leakandrzej@ahunt.org, Jun 20, 2021
  17. Elijah NewrenJun 21, 2021
  18. 09/12 builtin/mv: free or UNLEAK multiple pointers at end of cmd_mvandrzej@ahunt.org, Jun 20, 2021
  19. 10/12 builtin/merge: free found_ref when doneandrzej@ahunt.org, Jun 20, 2021
  20. Elijah NewrenJun 21, 2021
  21. 11/12 builtin/rebase: fix options.strategy memory lifecycleandrzej@ahunt.org, Jun 20, 2021
  22. Phillip WoodJun 20, 2021
  23. Elijah NewrenJun 21, 2021
  24. Phillip WoodJun 22, 2021
  25. Andrzej HuntJul 25, 2021
  26. Phillip WoodJul 27, 2021
  27. 12/12 reset: clear_unpack_trees_porcelain to plug leakandrzej@ahunt.org, Jun 20, 2021
  28. Elijah NewrenJun 21, 2021
  29. Elijah NewrenJun 21, 2021
  30. Andrzej HuntJul 25, 2021
  31. Christian CouderJul 26, 2021
  32. 00/12 Fix all leaks in tests t0002-t0099: Part 2andrzej@ahunt.org, Jul 25, 2021
  33. 01/12 fmt-merge-msg: free newly allocated temporary strings when doneandrzej@ahunt.org, Jul 25, 2021
  34. Junio C HamanoJul 26, 2021
  35. 02/12 environment: move strbuf into block to plug leakandrzej@ahunt.org, Jul 25, 2021
  36. 03/12 builtin/submodule--helper: release unused strbuf to avoid leakandrzej@ahunt.org, Jul 25, 2021
  37. 04/12 builtin/for-each-repo: remove unnecessary argv copy to plug leakandrzej@ahunt.org, Jul 25, 2021
  38. Junio C HamanoJul 26, 2021
  39. 06/12 ref-filter: also free head for ATOM_HEAD to avoid leakandrzej@ahunt.org, Jul 25, 2021
  40. Junio C HamanoJul 26, 2021
  41. 07/12 read-cache: call diff_setup_done to avoid leakandrzej@ahunt.org, Jul 25, 2021
  42. Junio C HamanoJul 26, 2021
  43. 05/12 diffcore-rename: move old_dir/new_dir definition to plug leakandrzej@ahunt.org, Jul 25, 2021
  44. Junio C HamanoJul 26, 2021
  45. 09/12 builtin/mv: free or UNLEAK multiple pointers at end of cmd_mvandrzej@ahunt.org, Jul 25, 2021
  46. 08/12 convert: release strbuf to avoid leakandrzej@ahunt.org, Jul 25, 2021
  47. Junio C HamanoJul 26, 2021
  48. 10/12 builtin/merge: free found_ref when doneandrzej@ahunt.org, Jul 25, 2021
  49. 12/12 reset: clear_unpack_trees_porcelain to plug leakandrzej@ahunt.org, Jul 25, 2021
  50. 11/12 builtin/rebase: fix options.strategy memory lifecycleandrzej@ahunt.org, Jul 25, 2021
  51. Junio C HamanoJul 26, 2021

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.