Re: [PATCH] fix null pointer dereference

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 21, 2021, 17:25 UTC
Message-ID: <xmqqft0oberu.fsf@gitster.g>
In-Reply-To: <pull.983.git.git.1616323936790.gitgitgadget@gmail.com>

"Kleber Tarcísio via GitGitGadget"  <gitgitgadget@gmail.com> writes:

> From: =?UTF-8?q?Kleber=20Tarc=C3=ADsio?= <klebertarcisio@yahoo.com.br>
> Subject: Re: [PATCH] fix null pointer dereference

Thanks, but see Documentation/SubmittingPatches for general guidelines. Our commit title begins with "<area>:" so that when this change is buried among hundreds of other commits in "git shortlog --no-merges", the readers can tell what it is about.

	Subject: [PATCH] submodule-helper: avoid unchecked malloc()

perhaps.

> The malloc function can return null when the memory allocation fails. This commit adds a condition to handle these cases properly. https://cwe.mitre.org/data/definitions/476.html

Overlong line. Also when you can say something without forcing readers to refer to external material, do so (and if you do not think you can, try harder ;-).

In this case, I think you do not need to say anything more than

	submodule-helper.c::submodule_summary_callback() calls
	malloc() and uses the returned value without checking for
	NULLness.  Use xmalloc() instaed.

Show 10 quoted lines

> diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c
> index 9d505a6329c8..92349d715a78 100644
> --- a/builtin/submodule--helper.c
> +++ b/builtin/submodule--helper.c
> @@ -1215,6 +1215,8 @@ static void submodule_summary_callback(struct diff_queue_struct *q,
>  		if (!S_ISGITLINK(p->one->mode) && !S_ISGITLINK(p->two->mode))
>  			continue;
>  		temp = (struct module_cb*)malloc(sizeof(struct module_cb));
> +		if (!temp) 
> +			die(_("out of memory"));

And this should be just

- 		temp = (struct module_cb*)malloc(sizeof(struct module_cb));
+ 		temp = (struct module_cb*)xmalloc(sizeof(struct module_cb));

without any "check and die" on its own.

Note that if this were a new code that adds a call to xmalloc(), competent reviewers would say it should be spelled more like so:

 		temp = xmalloc(sizeof(*temp));

to lose unnecessary cast and to prepare for future evolution of the code (e.g. the type of 'temp' may change from 'struct module_cb' to somethng else). But for this "malloc is wrong, use xmalloc instead" fix, we do not mix such a code improvement in the same patch.

Thanks.