[OE-core] [PATCH] busybox.inc: Add sanity check to test if the suid binary provides sh
Andrej Valek
andrej.valek at siemens.com
Fri Nov 10 07:17:08 UTC 2017
I don't think so, that adding bbfatal into this recipe is a good
solution. Someone could build the busybox with /bin/sh linking into
.suid eg. for more secure. We can leave live with this one as before.
Andrej
On 11/09/2017 02:25 PM, Nathan Rossi wrote:
> Add a sanity check during the do_compile task to fail if the suid
> busybox provides /bin/sh. This is considered as a hard fail since not
> only is providing sh as suid problematic for security reasons but also
> because the sh configured for suid is less functional than the nosuid
> configured sh and breaks a number of required features (e.g. 64-bit
> test).
>
> Signed-off-by: Nathan Rossi <nathan at nathanrossi.com>
> Cc: Ross Burton <ross.burton at intel.com>
> ---
> meta/recipes-core/busybox/busybox.inc | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/meta/recipes-core/busybox/busybox.inc b/meta/recipes-core/busybox/busybox.inc
> index 4012f921c6..157aea3968 100644
> --- a/meta/recipes-core/busybox/busybox.inc
> +++ b/meta/recipes-core/busybox/busybox.inc
> @@ -183,6 +183,12 @@ do_compile() {
> oe_runmake busybox.links
> mv busybox.links busybox.links.$s
> done
> +
> + # hard fail if sh is being linked to the suid busybox (detects bug 10346)
> + if grep -q -x "/bin/sh" busybox.links.suid; then
> + bbfatal "busybox suid binary incorrectly provides /bin/sh"
> + fi
> +
> # copy .config.orig back to .config, because the install process may check this file
> cp .config.orig .config
> # cleanup
>
More information about the Openembedded-core
mailing list