On Fri, Aug 22, 2014 at 04:36:37AM +0000, Serge Hallyn wrote:
> Quoting Tycho Andersen (tycho.ander...@canonical.com):
>
> Thanks, Tycho. Just one remaining request below. Other than that,
>
> Acked-by: Serge E. Hallyn <serge.hal...@ubuntu.com>
>
> > +static bool lxcapi_checkpoint(struct lxc_container *c, char *directory,
> > bool stop, bool verbose)
> > +{
> > + int netnr, status;
> > + struct lxc_list *it;
> > + bool error = false;
> > + pid_t pid;
> > +
> > + if (!criu_ok(c))
> > + return false;
> > +
> > + if (mkdir(directory, 0700) < 0 && errno != EEXIST)
> > + return false;
> > +
> > + netnr = 0;
> > + lxc_list_for_each(it, &c->lxc_conf->network) {
> > + char *veth = NULL, *bridge = NULL, veth_path[PATH_MAX],
> > eth[128];
> > + struct lxc_netdev *n = it->elem;
> > + int pret;
> > +
> > + pret = snprintf(veth_path, PATH_MAX,
> > "lxc.network.%d.veth.pair", netnr);
> > + if (pret < 0 || pret >= PATH_MAX) {
> > + error = true;
> > + goto out;
> > + }
> > +
> > + veth = lxcapi_get_running_config_item(c, veth_path);
> > + if (!veth) {
> > + /* criu_ok() checks that all interfaces are
> > + * LXC_NET{VETH,NONE}, and VETHs should have this
> > + * config */
> > + assert(n->type == LXC_NET_NONE);
> > + break;
> > + }
> > +
> > + pret = snprintf(veth_path, PATH_MAX, "lxc.network.%d.link",
> > netnr);
> > + if (pret < 0 || pret >= PATH_MAX) {
> > + error = true;
> > + goto out;
> > + }
> > +
> > + bridge = lxcapi_get_running_config_item(c, veth_path);
> > + if (!bridge) {
> > + error = true;
> > + goto out;
> > + }
> > +
> > + pret = snprintf(veth_path, PATH_MAX, "%s/veth%d", directory,
> > netnr);
> > + if (pret < 0 || pret >= PATH_MAX || print_to_file(veth_path,
> > veth) < 0) {
> > + error = true;
> > + goto out;
> > + }
> > +
> > + pret = snprintf(veth_path, PATH_MAX, "%s/bridge%d", directory,
> > netnr);
> > + if (pret < 0 || pret >= PATH_MAX || print_to_file(veth_path,
> > bridge) < 0) {
> > + error = true;
> > + goto out;
> > + }
> > +
> > + if (n->name) {
> > + assert(strlen(n->name) <= 127);
> > + strncpy(eth, n->name, 128);
>
> This again is an API function so I don't want asserts there if we can help
> it. I was thinking
>
> strncpy(eth, n->name, 127);
>
> Is there a good reason to do it this way?
Well, I guess we could also test if it is 127 or less, I was just
thinking that if it is over 127, you're going to get a non-working
container, so best to bail vs. continuing on knowing things will
break. You're right though, assert is probably the wrong thing here.
Maybe just a test and then return false?
Tycho
> > + } else
> > + sprintf(eth, "eth%d", netnr);
> > +
> > + pret = snprintf(veth_path, PATH_MAX, "%s/eth%d", directory,
> > netnr);
> > + if (pret < 0 || pret >= PATH_MAX || print_to_file(veth_path,
> > eth) < 0)
> > + error = true;
> > +
> > +out:
> > + free(veth);
> > + free(bridge);
> > + if (error)
> > + return false;
> > + }
>
> _______________________________________________
> lxc-devel mailing list
> lxc-devel@lists.linuxcontainers.org
> http://lists.linuxcontainers.org/listinfo/lxc-devel
_______________________________________________
lxc-devel mailing list
lxc-devel@lists.linuxcontainers.org
http://lists.linuxcontainers.org/listinfo/lxc-devel
