diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c --- a/sys/kern/kern_jail.c +++ b/sys/kern/kern_jail.c @@ -3440,6 +3440,7 @@ #endif case AF_LOCAL: case AF_ROUTE: + case AF_NETLINK: break; default: if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) diff --git a/sys/netlink/netlink_ctl.h b/sys/netlink/netlink_ctl.h --- a/sys/netlink/netlink_ctl.h +++ b/sys/netlink/netlink_ctl.h @@ -81,6 +81,7 @@ bool nl_has_listeners(int netlink_family, uint32_t groups_mask); bool nlp_has_priv(struct nlpcb *nlp, int priv); struct ucred *nlp_get_cred(struct nlpcb *nlp); +bool nlp_unconstrained_vnet(const struct nlpcb *nlp); /* netlink_generic.c */ struct genl_cmd { diff --git a/sys/netlink/netlink_domain.c b/sys/netlink/netlink_domain.c --- a/sys/netlink/netlink_domain.c +++ b/sys/netlink/netlink_domain.c @@ -36,6 +36,7 @@ #include #include #include +#include #include #include #include @@ -111,6 +112,10 @@ MPASS(group_id <= NLP_MAX_GROUPS); --group_id; + /* TODO: add family handler callback */ + if (!nlp_unconstrained_vnet(nlp)) + return; + nlp->nl_groups[group_id / 64] |= (uint64_t)1 << (group_id % 64); } @@ -212,6 +217,12 @@ return (priv_check_cred(nlp->nl_cred, priv) == 0); } +bool +nlp_unconstrained_vnet(const struct nlpcb *nlp) +{ + return (nlp->nl_unconstrained_vnet); +} + struct ucred * nlp_get_cred(struct nlpcb *nlp) { @@ -308,6 +319,7 @@ nlp->nl_process_id = curproc->p_pid; nlp->nl_linux = is_linux; nlp->nl_active = true; + nlp->nl_unconstrained_vnet = !jailed_without_vnet(so->so_cred); NLP_LOCK_INIT(nlp); refcount_init(&nlp->nl_refcount, 1); nl_init_io(nlp); diff --git a/sys/netlink/netlink_generic.c b/sys/netlink/netlink_generic.c --- a/sys/netlink/netlink_generic.c +++ b/sys/netlink/netlink_generic.c @@ -31,6 +31,7 @@ #include #include #include +#include #include #include #include diff --git a/sys/netlink/netlink_route.c b/sys/netlink/netlink_route.c --- a/sys/netlink/netlink_route.c +++ b/sys/netlink/netlink_route.c @@ -93,6 +93,11 @@ } else if (cmd->priv != 0) NLP_LOG(LOG_DEBUG3, nlp, "priv %d check passed for msg %s", cmd->priv, cmd->name); + if (!nlp_unconstrained_vnet(nlp) && (cmd->flags & RTNL_F_ALLOW_NONVNET_JAIL) == 0) { + NLP_LOG(LOG_DEBUG2, nlp, "jail check failed for msg %s", cmd->name); + return (EPERM); + } + bool need_epoch = !(cmd->flags & RTNL_F_NOEPOCH); if (need_epoch) diff --git a/sys/netlink/netlink_var.h b/sys/netlink/netlink_var.h --- a/sys/netlink/netlink_var.h +++ b/sys/netlink/netlink_var.h @@ -61,6 +61,7 @@ bool nl_task_pending; bool nl_tx_blocked; /* No new requests accepted */ bool nl_linux; /* true if running under compat */ + bool nl_unconstrained_vnet; /* true if running under VNET jail (or without jail) */ struct nl_io_queue rx_queue; struct nl_io_queue tx_queue; struct taskqueue *nl_taskqueue; diff --git a/sys/netlink/route/iface.c b/sys/netlink/route/iface.c --- a/sys/netlink/route/iface.c +++ b/sys/netlink/route/iface.c @@ -32,6 +32,7 @@ #include #include #include +#include #include #include #include @@ -61,6 +62,7 @@ struct nl_writer *nw; struct nlmsghdr hdr; struct nlpcb *so; + struct ucred *cred; uint32_t fibnum; int family; int error; @@ -833,6 +835,8 @@ continue; if (ifa->ifa_addr->sa_family == AF_LINK) continue; + if (prison_if(wa->cred, ifa->ifa_addr) != 0) + continue; wa->count++; if (!dump_iface_addr(wa->nw, ifp, ifa, &wa->hdr)) return (ENOMEM); @@ -856,6 +860,7 @@ struct netlink_walkargs wa = { .so = nlp, .nw = npt->nw, + .cred = nlp_get_cred(nlp), .family = attrs.ifa_family, .hdr.nlmsg_pid = hdr->nlmsg_pid, .hdr.nlmsg_seq = hdr->nlmsg_seq, @@ -977,7 +982,7 @@ .cmd = NL_RTM_GETLINK, .name = "RTM_GETLINK", .cb = &rtnl_handle_getlink, - .flags = RTNL_F_NOEPOCH, + .flags = RTNL_F_NOEPOCH | RTNL_F_ALLOW_NONVNET_JAIL, }, { .cmd = NL_RTM_DELLINK, @@ -997,6 +1002,7 @@ .cmd = NL_RTM_GETADDR, .name = "RTM_GETADDR", .cb = &rtnl_handle_getaddr, + .flags = RTNL_F_ALLOW_NONVNET_JAIL, }, { .cmd = NL_RTM_NEWADDR, diff --git a/sys/netlink/route/route_var.h b/sys/netlink/route/route_var.h --- a/sys/netlink/route/route_var.h +++ b/sys/netlink/route/route_var.h @@ -48,7 +48,8 @@ int flags; }; -#define RTNL_F_NOEPOCH 0x01 +#define RTNL_F_NOEPOCH 0x01 /* Do not enter epoch when handling command */ +#define RTNL_F_ALLOW_NONVNET_JAIL 0x02 /* Allow command execution inside non-VNET jail */ bool rtnl_register_messages(const struct rtnl_cmd_handler *handlers, int count); diff --git a/sys/netlink/route/rt.c b/sys/netlink/route/rt.c --- a/sys/netlink/route/rt.c +++ b/sys/netlink/route/rt.c @@ -513,6 +513,8 @@ wa->count++; if (wa->error != 0) return (0); + if (!rt_is_exportable(rt, nlp_get_cred(wa->nlp))) + return (0); wa->dumped++; rt_get_rnd(rt, &wa->rnd); @@ -606,6 +608,9 @@ RIB_RUNLOCK(rnh); + if (!rt_is_exportable(rt, nlp_get_cred(nlp))) + return (ESRCH); + IF_DEBUG_LEVEL(LOG_DEBUG2) { char rtbuf[NHOP_PRINT_BUFSIZE] __unused, nhbuf[NHOP_PRINT_BUFSIZE] __unused; FIB_LOG(LOG_DEBUG2, fibnum, family, "getroute completed: got %s for %s", @@ -1026,6 +1031,7 @@ .cmd = NL_RTM_GETROUTE, .name = "RTM_GETROUTE", .cb = &rtnl_handle_getroute, + .flags = RTNL_F_ALLOW_NONVNET_JAIL, }, { .cmd = NL_RTM_DELROUTE,