It seems like in the process of refactoring pwm_config() to utilize the
newly-introduced pwm_apply_state() API, some args/bounds checking was
dropped.
In particular, I noted that we are now allowing invalid period
selections, e.g.:
  # echo 1 > /sys/class/pwm/pwmchip0/export
  # cat /sys/class/pwm/pwmchip0/pwm1/period
  100
  # echo 101 > /sys/class/pwm/pwmchip0/pwm1/duty_cycle
  [... driver may or may not reject the value, or trigger some logic bug ...]
It's better to see:
  # echo 1 > /sys/class/pwm/pwmchip0/export
  # cat /sys/class/pwm/pwmchip0/pwm1/period
  100
  # echo 101 > /sys/class/pwm/pwmchip0/pwm1/duty_cycle
  -bash: echo: write error: Invalid argument
This patch reintroduces some bounds checks in both pwm_config() (for its
signed parameters; we don't want to convert negative values into large
unsigned values) and in pwm_apply_state() (which fix the above described
behavior, as well as other potential API misuses).
Fixes: 5ec803edcb70 ("pwm: Add core infrastructure to allow atomic updates")
Signed-off-by: Brian Norris <briannorris@chromium.org>
Acked-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
 {
        int err;
 
-       if (!pwm)
+       if (!pwm || !state || !state->period ||
+           state->duty_cycle > state->period)
                return -EINVAL;
 
        if (!memcmp(state, &pwm->state, sizeof(*state)))
 
        if (!pwm)
                return -EINVAL;
 
+       if (duty_ns < 0 || period_ns < 0)
+               return -EINVAL;
+
        pwm_get_state(pwm, &state);
        if (state.duty_cycle == duty_ns && state.period == period_ns)
                return 0;