Re: [BUG] 3ce1217d6cd5 ima patch causes s390 to crash on boot

From: Heiko Carstens
Date: Fri Nov 22 2013 - 11:36:30 EST


On Fri, Nov 22, 2013 at 05:23:49PM +0100, Roberto Sassu wrote:
> Another problem that I found is that strsep()
> modifies the source buffer by replacing the separator
> character with '\0'. In particular, this function
> modifies static data initialized at the beginning
> of the ima_template.c file. Maybe, this is causing
> the kernel panic. I already sent a patch to fix this
> problem (attached to the email) even if it is not
> supposed to land on the 3.13 kernel. Let me know
> if this fixes the issue. Otherwise, I will check
> the code more in depth.

With your patch below applied the kernel boots again.
So it should go into 3.13 (or a different fix).

Thanks!

> From 2d3aa1c0328c44ecc3af7de162791c8cddfb6dfd Mon Sep 17 00:00:00 2001
> From: Roberto Sassu <roberto.sassu@xxxxxxxxx>
> Date: Wed, 6 Nov 2013 13:51:35 +0100
> Subject: [RFC][PATCH 2/4] ima: make a copy of template_fmt in
> template_desc_init_fields()
>
> This patch makes a copy of the 'template_fmt' function argument so that
> the latter will not be modified by strsep(), which does the splitting by
> replacing the given separator with '\0'.
>
> Signed-off-by: Roberto Sassu <roberto.sassu@xxxxxxxxx>
> ---
> security/integrity/ima/ima_template.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/security/integrity/ima/ima_template.c b/security/integrity/ima/ima_template.c
> index 7bcff5c..bb33576 100644
> --- a/security/integrity/ima/ima_template.c
> +++ b/security/integrity/ima/ima_template.c
> @@ -113,13 +113,19 @@ static int template_desc_init_fields(char *template_fmt,
> struct ima_template_field ***fields,
> int *num_fields)
> {
> - char *c, *template_fmt_ptr = template_fmt;
> + char *c, *template_fmt_ptr, *template_fmt_copy = NULL;
> int template_num_fields = template_fmt_size(template_fmt);
> int i, result = 0;
>
> if (template_num_fields > IMA_TEMPLATE_NUM_FIELDS_MAX)
> return -EINVAL;
>
> + /* copying is needed as strsep() modifies the original buffer */
> + template_fmt_copy = kstrdup(template_fmt, GFP_KERNEL);
> + if (template_fmt_copy == NULL)
> + return -ENOMEM;
> +
> + template_fmt_ptr = template_fmt_copy;
> *fields = kzalloc(template_num_fields * sizeof(*fields), GFP_KERNEL);
> if (*fields == NULL) {
> result = -ENOMEM;
> @@ -139,6 +145,7 @@ static int template_desc_init_fields(char *template_fmt,
> *num_fields = i;
> return 0;
> out:
> + kfree(template_fmt_copy);
> kfree(*fields);
> *fields = NULL;
> return result;
> --
> 1.8.1.4
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/