On 04/10/2019 11:52:06-0700, Yizhuo wrote:
> Inside function snvs_rtc_enable(), variable "lpcr" could be
> uninitialized if regmap_read() returns -EINVAL. However,"lpcr"
> is used later in the if statement, which is potentially unsafe.
>
This is not true, regmap_read either lock up the cpu or not fail in this
driver.
> Similar cases happened in function snvs_rtc_irq_handler() with
> variable "lpsr" and function snvs_rtc_read_alarm() with variables
> "lptar", "lpsr". The patch for those two are not easy since
> -EINVAL is not an acceptable return value for these functions.
>
> Signed-off-by: Yizhuo <yzhai003@xxxxxxx>
> ---
> drivers/rtc/rtc-snvs.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/rtc/rtc-snvs.c b/drivers/rtc/rtc-snvs.c
> index 757f4daa7181..dadcc3e193b2 100644
> --- a/drivers/rtc/rtc-snvs.c
> +++ b/drivers/rtc/rtc-snvs.c
> @@ -124,12 +124,16 @@ static int snvs_rtc_enable(struct snvs_rtc_data *data, bool enable)
> {
> int timeout = 1000;
> u32 lpcr;
> + int ret;
>
> regmap_update_bits(data->regmap, data->offset + SNVS_LPCR, SNVS_LPCR_SRTC_ENV,
> enable ? SNVS_LPCR_SRTC_ENV : 0);
>
> while (--timeout) {
> - regmap_read(data->regmap, data->offset + SNVS_LPCR, &lpcr);
> + ret = regmap_read(data->regmap,
> + data->offset + SNVS_LPCR, &lpcr);
> + if (ret)
> + return ret;
>
> if (enable) {
> if (lpcr & SNVS_LPCR_SRTC_ENV)
> --
> 2.17.1
>
--
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
