Skip to contentSkip to conversation

signedness bug in lbs_auth_to_authtype()

DISCUSSION Dan Carpenter 1 participant 1 message Mar 28, 2012

lore
Share & tools
lore
#0 lore raw

There is a signedness bug in lbs_auth_to_authtype() and it messes up the error handling.

drivers/net/wireless/libertas/cfg.c
106 static u8 lbs_auth_to_authtype(enum nl80211_auth_type auth_type) 107 {
108 int ret = -ENOTSUPP;
109

We return -ENOTSUPP but it's casted to u8 making it 244.

   351          auth->header.type = cpu_to_le16(TLV_TYPE_AUTH_TYPE);
   352          auth->header.len = cpu_to_le16(sizeof(*auth)-sizeof(auth->header));
   353          auth->auth = cpu_to_le16(lbs_auth_to_authtype(auth_type));
                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

We don't check for errors here.

   354          return sizeof(*auth);
   355  }
  1061          /* convert auth_type */
  1062          ret = lbs_auth_to_authtype(sme->auth_type);
  1063          if (ret < 0)
                    ^^^^^^^

This doesn't work because 244 is larger than zero.

1064 goto done;

I would just change the u8 to an int, except that I felt that maybe we should be checking for errors on the first time this function is called. I wasn't sure how to handle that error.

regards,
dan carpenter