pid1: set SYSTEMD_NSS_DYNAMIC_BYPASS=1 env var for dbus-daemon
There's currently a deadlock between PID 1 and dbus-daemon: in some
cases dbus-daemon will do NSS lookups (which are blocking) at the same
time PID 1 synchronously blocks on some call to dbus-daemon. Let's break
that by setting SYSTEMD_NSS_DYNAMIC_BYPASS=1 env var for dbus-daemon,
which will disable synchronously blocking varlink calls from nss-systemd
to PID 1.
In the long run we should fix this differently: remove all synchronous
calls to dbus-daemon from PID 1. This is not trivial however: so far we
had the rule that synchronous calls from PID 1 to the dbus broker are OK
as long as they only go to interfaces implemented by the broke itself
rather than services reachable through it. Given that the relationship
between PID 1 and dbus is kinda special anyway, this was considered
acceptable for the sake of simplicity, since we quite often need
metadata about bus peers from the broker, and the asynchronous logic
would substantially complicate even the simplest method handlers.
This mostly reworks the existing code that sets SYSTEMD_NSS_BYPASS_BUS=
(which is a similar hack to deal with deadlocks between nss-systemd and
dbus-daemon itself) to set SYSTEMD_NSS_DYNAMIC_BYPASS=1 instead. No code
was checking SYSTEMD_NSS_BYPASS_BUS= anymore anyway, and it used to
solve a similar problem, hence it's an obvious piece of code to rework
like this.
Issue originally tracked down by Lukas Märdian. This patch is inspired
and closely based on his patch:
https://github.com/systemd/systemd/pull/22038
Fixes: #15316
Co-authored-by: Lukas Märdian <slyon@ubuntu.com>
This commit is contained in:
co-authored by
Lukas Märdian
parent
cec16155e3
commit
de90700f36
+5
-5
@@ -1865,11 +1865,11 @@ static int build_environment(
|
||||
our_env[n_env++] = x;
|
||||
}
|
||||
|
||||
/* If this is D-Bus, tell the nss-systemd module, since it relies on being able to use D-Bus look up dynamic
|
||||
* users via PID 1, possibly dead-locking the dbus daemon. This way it will not use D-Bus to resolve names, but
|
||||
* check the database directly. */
|
||||
if (p->flags & EXEC_NSS_BYPASS_BUS) {
|
||||
x = strdup("SYSTEMD_NSS_BYPASS_BUS=1");
|
||||
/* If this is D-Bus, tell the nss-systemd module, since it relies on being able to use blocking
|
||||
* Varlink calls back to us for look up dynamic users in PID 1. Break the deadlock between D-Bus and
|
||||
* PID 1 by disabling use of PID1' NSS interface for looking up dynamic users. */
|
||||
if (p->flags & EXEC_NSS_DYNAMIC_BYPASS) {
|
||||
x = strdup("SYSTEMD_NSS_DYNAMIC_BYPASS=1");
|
||||
if (!x)
|
||||
return -ENOMEM;
|
||||
our_env[n_env++] = x;
|
||||
|
||||
+1
-1
@@ -376,7 +376,7 @@ typedef enum ExecFlags {
|
||||
EXEC_APPLY_TTY_STDIN = 1 << 2,
|
||||
EXEC_PASS_LOG_UNIT = 1 << 3, /* Whether to pass the unit name to the service's journal stream connection */
|
||||
EXEC_CHOWN_DIRECTORIES = 1 << 4, /* chown() the runtime/state/cache/log directories to the user we run as, under all conditions */
|
||||
EXEC_NSS_BYPASS_BUS = 1 << 5, /* Set the SYSTEMD_NSS_BYPASS_BUS environment variable, to disable nss-systemd for dbus */
|
||||
EXEC_NSS_DYNAMIC_BYPASS = 1 << 5, /* Set the SYSTEMD_NSS_DYNAMIC_BYPASS environment variable, to disable nss-systemd blocking on PID 1, for use by dbus-daemon */
|
||||
EXEC_CGROUP_DELEGATE = 1 << 6,
|
||||
EXEC_IS_CONTROL = 1 << 7,
|
||||
EXEC_CONTROL_CGROUP = 1 << 8, /* Place the process not in the indicated cgroup but in a subcgroup '/.control', but only EXEC_CGROUP_DELEGATE and EXEC_IS_CONTROL is set, too */
|
||||
|
||||
+1
-1
@@ -1681,7 +1681,7 @@ static int service_spawn(
|
||||
return -ENOMEM;
|
||||
|
||||
/* System D-Bus needs nss-systemd disabled, so that we don't deadlock */
|
||||
SET_FLAG(exec_params.flags, EXEC_NSS_BYPASS_BUS,
|
||||
SET_FLAG(exec_params.flags, EXEC_NSS_DYNAMIC_BYPASS,
|
||||
MANAGER_IS_SYSTEM(UNIT(s)->manager) && unit_has_name(UNIT(s), SPECIAL_DBUS_SERVICE));
|
||||
|
||||
strv_free_and_replace(exec_params.environment, final_env);
|
||||
|
||||
Reference in New Issue
Block a user