If the feature query fails during this driver's probe, we memset the
features bits to 0, but then we continue to overwrite that with whatever
junk we read back when the feature probing failed. In the case of a SPI
device, it may be complete junk, causing the caller of this function to
think certain features are present when they aren't. Don't copy over the
bad message on failure so we can fail more gracefully.
Fixes: e4244ebddae2 ("platform/chrome: Introduce a new function to check EC
features.")
Cc: Benson Leung <[email protected]>
Cc: Enric Balletbo i Serra <[email protected]>
Cc: Gwendal Grignou <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Signed-off-by: Stephen Boyd <[email protected]>
---
drivers/mfd/cros_ec_dev.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/mfd/cros_ec_dev.c b/drivers/mfd/cros_ec_dev.c
index 1d6dc5c7a19d..d37c79426c87 100644
--- a/drivers/mfd/cros_ec_dev.c
+++ b/drivers/mfd/cros_ec_dev.c
@@ -113,10 +113,10 @@ static int cros_ec_check_features(struct cros_ec_dev *ec,
int feature)
dev_warn(ec->dev, "cannot get EC features: %d/%d\n",
ret, msg->result);
memset(ec->features, 0, sizeof(ec->features));
+ } else {
+ memcpy(ec->features, msg->data, sizeof(ec->features));
}
- memcpy(ec->features, msg->data, sizeof(ec->features));
-
dev_dbg(ec->dev, "EC features %08x %08x\n",
ec->features[0], ec->features[1]);
--
Sent by a computer through tubes
