Hi Jukka On Fri, Jun 1, 2012 at 9:02 AM, Jukka Rissanen <jukka.rissa...@linux.intel.com> wrote: > The problem was seen with valgrind. We tried to get hostname > without domain part and if server returns an error, we could > end up in situation where domain_len == 0 even if append_domain > is set to TRUE. So check that if domain_len is 0, then do not > try to memcpy. > --- > src/dnsproxy.c | 21 +++++++++++++++------ > 1 files changed, 15 insertions(+), 6 deletions(-) > > diff --git a/src/dnsproxy.c b/src/dnsproxy.c > index efbc195..f56c30c 100644 > --- a/src/dnsproxy.c > +++ b/src/dnsproxy.c > @@ -1554,14 +1554,23 @@ static int forward_dns_reply(unsigned char *reply, > int reply_len, int protocol, > domain_len = strlen((const char *)ptr + host_len + 1); > > /* > - * remove the domain name and replace it by the end > - * of reply. > + * Remove the domain name and replace it by the end > + * of reply. Check if the domain is really there > + * before trying to copy the data. The domain_len can > + * be 0 because if the original query did not contain > + * a domain name, then we are sending two packets, > + * first without the domain name and the second packet > + * with domain name. The append_domain is set to true > + * even if we sent the first packet without domain > + * name. In this case we end up in this branch.
The description in the log message and this one are different. Are you sure about the one in the log message? If it receives an error, this part of the code would not be reached. As per this comment, couldn't we just not set the append_domain flag if it was not appended? Lucas De Marchi _______________________________________________ connman mailing list connman@connman.net http://lists.connman.net/listinfo/connman
