On Thu, 2013-10-17 at 14:38 +0200, Peter Meerwald wrote:
> From: Peter Meerwald <[email protected]>
>
> fix a crash because cache is NULL; this happens often, not always on the
> first name to resolve
>
> initialization of the cache happens all over the place
>
> cache_update() already has a create_cache() in a special-case path, but not
> in the main path
>
> observed with 1.19; may have to do provisioning config
As git doesn't do line wrapping please keep the commit message to less
than 80 chars per line. There is a four space indentation prepended, so
normally one should wrap at column 72 or so.
> connmand[4347]: src/network.c:connman_network_unref_debug() 0xc97b0 name
> Wired ref 3 by src/network.c:462:check_dhcpv6()
> connmand[4347]: src/ntp.c:send_timeout() send timeout (retries 1)
> connmand[4347]: src/ntp.c:send_timeout() send timeout (retries 2)
> connmand[4347]: src/ntp.c:send_timeout() send timeout (retries 3)
> connmand[4347]: src/ntp.c:__connman_ntp_stop()
> connmand[4347]: src/dnsproxy.c:udp_listener_event() Received 27 bytes (id
> 0x76ea)
> connmand[4347]: src/dnsproxy.c:parse_request() id 0x76ea qr 0 opcode 0
> qdcount 1 arcount 0
> connmand[4347]: src/dnsproxy.c:parse_request() query google.at.
> connmand[4347]: src/dnsproxy.c:resolv() server 192.168.233.10 enabled 1
> connmand[4347]: src/dnsproxy.c:forward_dns_reply() Received 189 bytes (id
> 0x2d11)
> connmand[4347]: src/dnsproxy.c:forward_dns_reply() req 0xc81c0 dstid 0x2d11
> altid 0xf6ca rcode 0
> connmand[4347]: src/dnsproxy.c:cache_update() offset 0 hdr 0xbe828b88 msg
> 0xbe828b88 rcode 0
> connmand[4347]: src/dnsproxy.c:parse_response() qr 1 qdcount 1
> connmand[4347]: Aborting (signal 11) [/usr/sbin/connmand]
Good catch! I wonder why this has hit only you and only now.
> ---
> src/dnsproxy.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/src/dnsproxy.c b/src/dnsproxy.c
> index 65806ca..97b07aa 100644
> --- a/src/dnsproxy.c
> +++ b/src/dnsproxy.c
> @@ -1468,7 +1468,11 @@ static int cache_update(struct server_data *srv,
> unsigned char *msg,
> * This is needed so that we can cache both A and AAAA
> * records for the same name.
> */
> - entry = g_hash_table_lookup(cache, question);
> + if (!cache) {
> + create_cache();
> + entry = NULL;
> + } else
> + entry = g_hash_table_lookup(cache, question);
> if (!entry) {
> entry = g_try_new(struct cache_entry, 1);
> if (!entry)
This code does the job for sure. Since the same code also exists for the
(err == -ENOMSG || err == -ENOBUFS)... path, I'd suggest to move the
cache creation from its current location to right before 'rsplen = ...'.
With that all possible future cases will be covered.
Cheers,
Patrik
_______________________________________________
connman mailing list
[email protected]
https://lists.connman.net/mailman/listinfo/connman
