Hi Alexandru, On 07/02/2014 10:18 AM, Alexandru Costache wrote: > Hi, > > Further testing revealed that d099d36c8654260524c423eff4edcad0171e24aa > <http://git.kernel.org/cgit/network/connman/connman.git/commit/?id=d099d36c8654260524c423eff4edcad0171e24aa> > does not fix all crashing conditions when connecting to a network without > dhcp services. ipv4ll_available_cb() and dhcp_retry_cb() may both fail when > either dhcp->network is NULL or has just been freed. Removing the timeout > and the source of the callback when dhcp is invalidated has fixed the rest > of the failures.
I see. > > I have added this fix as an attachment because I'm not able to use > git-send-email for the moment. Please let me know if there's something I > should change in this patch or if I need to attach it somewhere else. Okay. Note you can use your gmail account to send this patch as well via 'git send-email'. There are example configuration how to get this working. I use a msmtp setup to get this working. > > Thanks, > Alexandru > > > 0001-dhcp-Fix-further-crashes.patch > > > From 2c0c84abc2fa198f4b88e7b0e14f78e9c26b6f03 Mon Sep 17 00:00:00 2001 > From: Alexandru Costache <[email protected]> > Date: Tue, 1 Jul 2014 10:42:44 -0400 > Subject: [PATCH] dhcp: Fix further crashes when connected to network without > dhcp > > There are several crashing conditions in dhcp_retry_cb() and in > ipv4ll_available_cb(), when service is NULL or dhcp->network is > either NULL or has been freed. Fixed all by removing timeout and > ipv4ll callback when invalidating dhcp. > --- > src/dhcp.c | 14 ++++++++++++++ > 1 file changed, 14 insertions(+) > > diff --git a/src/dhcp.c b/src/dhcp.c > index c717f84..d20cc3c 100644 > --- a/src/dhcp.c > +++ b/src/dhcp.c > @@ -70,6 +70,8 @@ static void dhcp_free(struct connman_dhcp *dhcp) > g_free(dhcp); > } > > +static void ipv4ll_stop_client(struct connman_dhcp *dhcp); > + > /** > * dhcp_invalidate: Invalidate an existing DHCP lease > * @dhcp: pointer to the DHCP lease to invalidate. > @@ -131,6 +133,12 @@ static void dhcp_invalidate(struct connman_dhcp *dhcp, > bool callback) > __connman_ipconfig_set_gateway(ipconfig, NULL); > __connman_ipconfig_set_prefixlen(ipconfig, 0); > > + if (dhcp->timeout > 0) > + g_source_remove(dhcp->timeout); I think you should also do a 'dhcp->timeout = 0' to be on the safe side. > + > + if (ipv4ll_running) > + ipv4ll_stop_client(dhcp); > + > if (dhcp->callback && callback) > dhcp->callback(dhcp->network, false, NULL); > } > @@ -221,6 +229,9 @@ static gboolean dhcp_retry_cb(gpointer user_data) > > dhcp->timeout = 0; > > + if (!dhcp->network) > + return FALSE; > + That is not necessary since connman_service_lookup_from_network() has the exact test from above. > service = connman_service_lookup_from_network(dhcp->network); > if (!service) > return FALSE; > @@ -471,6 +482,9 @@ static void ipv4ll_available_cb(GDHCPClient > *ipv4ll_client, gpointer user_data) > > DBG("IPV4LL available"); > > + if (!dhcp->network) > + return; > + Same here. > service = connman_service_lookup_from_network(dhcp->network); > if (!service) > return; Thanks, Daniel _______________________________________________ connman mailing list [email protected] https://lists.connman.net/mailman/listinfo/connman
