diff options
| -rwxr-xr-x | src/services/vyos-netlinkd | 95 |
1 files changed, 84 insertions, 11 deletions
diff --git a/src/services/vyos-netlinkd b/src/services/vyos-netlinkd index fda946050..9627f9b28 100755 --- a/src/services/vyos-netlinkd +++ b/src/services/vyos-netlinkd @@ -15,6 +15,7 @@ # along with this program. If not, see <http://www.gnu.org/licenses/>. import importlib.util +import os import re import sys import syslog @@ -33,6 +34,7 @@ from vyos.configquery import op_mode_config_dict from vyos.ifconfig import Section from vyos.utils.boot import boot_configuration_complete from vyos.utils.commit import commit_in_progress2 +from vyos.utils.file import read_file from vyos.utils.dict import dict_search from vyos.utils.process import cmdl from vyos.utils.process import is_systemd_service_active @@ -54,15 +56,31 @@ IFACE_RE = re.compile(r"^(?:eth|br|bond|wlan|pppoe|sstpc|wwan)") _dynamic_qos_interfaces: set[str] = set() # Per-interface previous operstate, used to suppress DHCP restarts on -# UP-to-UP re-notifications (e.g. post-migration gratuitous-ARP events). +# UP-to-UP re-notifications (e.g. post-migration gratuitous-ARP events), and +# to detect edges missed while netlink events were drained during a commit. _iface_prev_operstate: dict[str, str] = {} +# True while the main loop is draining netlink events because a config +# commit holds the lock. When it clears we reconcile DHCP clients against +# live operstate so a transition that only happened inside the drain window +# is not lost (T9086). +_in_commit_skip = False + def match_iface(ifname: str) -> bool: """ Helper function returning true if interface name is a match for further processing (e.g. restart of DHCP(v6) client) """ return IFACE_RE.match(ifname) is not None +def _read_operstate(ifname: str) -> Optional[str]: + """Read live kernel operstate from sysfs (uppercase), or None if unavailable. + + Values match IFLA_OPERSTATE names used by RTM_NEWLINK (UP, DOWN, ...). + See Documentation/ABI/testing/sysfs-class-net. + """ + fname = f'/sys/class/net/{ifname}/operstate' + return read_file(fname, defaultonfailure='').strip().upper() or None + def _is_dynamic_qos_iface(ifname: str) -> bool: """ Helper function returning true if interface requires QoS re-apply after first address assignment. @@ -162,7 +180,45 @@ def _handle_dhcp_events(operstate: Optional[str], ifname: str) -> None: return None +def _reconcile_dhcp_operstate() -> None: + """After a commit drain window, sync DHCP clients to live operstate (T9086). + + Events discarded while commit_in_progress2() was true may have included a + real UP/DOWN edge. Compare each interface's live sysfs operstate against + the last one we acted on and run the normal DHCP handler only where they + disagree. + + Interfaces never seen before are seeded into the tracker without acting - + conf_mode owns DHCP bring-up across commits for those. + """ + try: + ifnames = os.listdir('/sys/class/net') + except OSError as e: + syslog.syslog(syslog.LOG_ERR, f'Failed to list interfaces for reconcile: {e}') + return + + for ifname in ifnames: + if not match_iface(ifname): + continue + + operstate = _read_operstate(ifname) + if operstate not in ['UP', 'DOWN']: + continue + + prev = _iface_prev_operstate.get(ifname) + if prev is None: + _iface_prev_operstate[ifname] = operstate + continue + if prev == operstate: + continue + + syslog.syslog(syslog.LOG_DEBUG, + f'Reconcile {ifname}: prev={prev} current={operstate}') + _handle_dhcp_events(operstate, ifname) + def main(): + global _in_commit_skip + syslog.openlog(ident="vyos-netlinkd", logoption=syslog.LOG_PID, facility=syslog.LOG_DAEMON) @@ -200,19 +256,36 @@ def main(): continue try: - # Wait for up to 1 second for a netlink message + # Wait for up to 1 second for a netlink message. The timeout also + # lets us notice a commit ending even with no further netlink + # traffic, so we can reconcile (T9086). rlist, _, _ = select.select([fd], [], [], 1.0) - if not rlist: - # timeout - retry - continue - # Check if a config commit is in progress before processing any - # messages. This avoids blocking per-message and reduces unnecessary - # calls to commit_in_progress2() if commit_in_progress2(): - syslog.syslog(syslog.LOG_DEBUG, - 'Config commit in progress, skipping netlink events') - sleep(1) + # Drain the socket instead of leaving messages queued, so a + # stale event (e.g. a DOWN from a disable that was already + # re-enabled) can't be processed as fresh once the commit + # ends (T9086). + if not _in_commit_skip: + syslog.syslog(syslog.LOG_DEBUG, + 'Config commit in progress, draining netlink events without acting') + _in_commit_skip = True + if rlist: + try: + ipr.get() + except NetlinkError as e: + syslog.syslog(syslog.LOG_ERR, + f'Netlink error while draining during commit: {e}') + continue + + if _in_commit_skip: + _in_commit_skip = False + syslog.syslog(syslog.LOG_INFO, + 'Config commit finished, reconciling DHCP client state with operstate') + _reconcile_dhcp_operstate() + + if not rlist: + # timeout - retry continue # Receive and process any messages |