[PATCH] samples/landlock: Document best-effort approach for LANDLOCK_ACCESS_FS_REFER
Mickaël Salaün
mic at digikod.net
Thu Nov 3 14:21:32 UTC 2022
Thanks Günther. Here are small changes:
On 30/10/2022 07:11, Günther Noack wrote:
> Add a comment to clarify how to handle best-effort backwards
> compatibility for LANDLOCK_ACCESS_FS_REFER.
>
> The "refer" access is special because these operations are always
> forbidden in ABI 1, unlike most other operations, which are permitted
> when using Landlock ABI levels where they are not supported yet.
>
> Signed-off-by: Günther Noack <gnoack3000 at gmail.com>
> ---
> samples/landlock/sandboxer.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/samples/landlock/sandboxer.c b/samples/landlock/sandboxer.c
> index fd4237c64fb2..901acb383124 100644
> --- a/samples/landlock/sandboxer.c
> +++ b/samples/landlock/sandboxer.c
> @@ -234,7 +234,21 @@ int main(const int argc, char *const argv[], char *const *const envp)
> /* Best-effort security. */
> switch (abi) {
> case 1:
> - /* Removes LANDLOCK_ACCESS_FS_REFER for ABI < 2 */
> + /*
> + * Removes LANDLOCK_ACCESS_FS_REFER for ABI < 2
> + *
> + * Note: The "refer" operations (file renaming and linking
> + * across different directories) are always forbidden when using
> + * Landlock with ABI 1.
> + *
> + * If only ABI 1 is available, the sample tool knowingly forbids
s/the sample tool/this sandboxer/
> + * refer operations.
> + *
> + * If a program *needs* to do refer operations after enabling
> + * Landlock, it can not use Landlock at ABI level 1. To be
> + * compatible across different kernels, such programs should
> + * fall back to not using Landlock instead.
To be compatible with different kernel versions, such programs should
then fall back to not restrict themselves at all if the running kernel
only supports ABI 1.
> + */
> ruleset_attr.handled_access_fs &= ~LANDLOCK_ACCESS_FS_REFER;
> __attribute__((fallthrough));
> case 2:
>
> base-commit: 4bc90a766ea5af69c12ca1ea00b7fc5fe1d68831
More information about the Linux-security-module-archive
mailing list