Hi Maoyi, On Wed, Nov 18, 2026 at 06:15:55PM +0800, Maoyi Xie wrote: > wilco_ec_get_property() copies rs.length bytes from rs.data into > prop_msg->data without checking rs.length, which comes from the EC. > Both buffers hold WILCO_EC_PROPERTY_MAX_SIZE (4) bytes, so a larger > length reads past rs on the stack and writes past the caller's message. > > Return -EBADMSG for such a length, as the file does for other bad > responses. > > Found with a CodeQL-based static analysis tool that we are developing. > I used an LLM to help write this patch and a userspace harness. The > harness runs the unmodified properties.c against a model of the EC > under AddressSanitizer. It reports the copy for lengths 5 to 255, and > nothing with this patch. Not tested on hardware, only compile-tested > with W=1. > > Fixes: 0c0b7ea23aed ("platform/chrome: wilco_ec: Add property helper library") > Cc: stable@vger.kernel.org > Signed-off-by: Maoyi Xie A Reviewed-by: Benson Leung Thanks, Benson > --- > drivers/platform/chrome/wilco_ec/properties.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/platform/chrome/wilco_ec/properties.c b/drivers/platform/chrome/wilco_ec/properties.c > index 9951c8db04da065..4ca7e874804c28e 100644 > --- a/drivers/platform/chrome/wilco_ec/properties.c > +++ b/drivers/platform/chrome/wilco_ec/properties.c > @@ -70,6 +70,8 @@ int wilco_ec_get_property(struct wilco_ec_device *ec, > ret = send_property_msg(ec, &rq, &rs); > if (ret < 0) > return ret; > + if (rs.length > WILCO_EC_PROPERTY_MAX_SIZE) > + return -EBADMSG; > > prop_msg->length = rs.length; > memcpy(prop_msg->data, rs.data, rs.length); > > base-commit: 7e1392a42ea288d70a266f6553fe06f57f8647ea > -- > 2.34.1 >