Re: [RFC PATCH 1/2] kfence: Allow re-enabling KFENCE after system startup

From: Marco Elver
Date: Fri Mar 04 2022 - 13:14:28 EST


On Thu, 3 Mar 2022 at 04:15, Tianchen Ding <dtcccc@xxxxxxxxxxxxxxxxx> wrote:
>
> If once KFENCE is disabled by:
> echo 0 > /sys/module/kfence/parameters/sample_interval
> KFENCE could never be re-enabled until next rebooting.
>
> Allow re-enabling it by writing a positive num to sample_interval.
>
> Signed-off-by: Tianchen Ding <dtcccc@xxxxxxxxxxxxxxxxx>

The only problem I see with this is if KFENCE was disabled because of
a KFENCE_WARN_ON(). See below.

> ---
> mm/kfence/core.c | 16 ++++++++++++++--
> 1 file changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/mm/kfence/core.c b/mm/kfence/core.c
> index 13128fa13062..19eb123c0bba 100644
> --- a/mm/kfence/core.c
> +++ b/mm/kfence/core.c
> @@ -55,6 +55,7 @@ EXPORT_SYMBOL_GPL(kfence_sample_interval); /* Export for test modules. */
> #endif
> #define MODULE_PARAM_PREFIX "kfence."
>
> +static int kfence_enable_late(void);
> static int param_set_sample_interval(const char *val, const struct kernel_param *kp)
> {
> unsigned long num;
> @@ -65,10 +66,11 @@ static int param_set_sample_interval(const char *val, const struct kernel_param
>
> if (!num) /* Using 0 to indicate KFENCE is disabled. */
> WRITE_ONCE(kfence_enabled, false);
> - else if (!READ_ONCE(kfence_enabled) && system_state != SYSTEM_BOOTING)
> - return -EINVAL; /* Cannot (re-)enable KFENCE on-the-fly. */
>
> *((unsigned long *)kp->arg) = num;
> +
> + if (num && !READ_ONCE(kfence_enabled) && system_state != SYSTEM_BOOTING)

Should probably have an 'old_sample_interval = *((unsigned long
*)kp->arg)' somewhere before, and add a '&& !old_sample_interval',
because if old_sample_interval!=0 then KFENCE was disabled due to a
KFENCE_WARN_ON(). Also in this case, it should return -EINVAL. So you
want a flow like this:

old_sample_interval = ...;
...
if (num && !READ_ONCE(kfence_enabled) && system_state != SYSTEM_BOOTING)
return old_sample_interval ? -EINVAL : kfence_enable_late();
...

Thanks,
-- Marco