From: Yu Watanabe <watanabe.yu+github@gmail.com>
Date: Tue, 25 Oct 2022 16:18:05 +0000 (+0900)
Subject: core/device: verify device syspath on switching root
X-Git-Tag: v253-rc1~429^2
X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=b6c86ae28149c4abb2f0bd6acab13153382da9e7;p=thirdparty%2Fsystemd.git

core/device: verify device syspath on switching root

Otherwise, if a device is removed while switching root, then the
corresponding .device unit will never go to inactive state.

This replaces the code dropped by cf1ac0cfe44997747b0f857a1d0b67cea1298272.

Fixes #25106.
---

diff --git a/src/core/device.c b/src/core/device.c
index 7e354b2b4a7..6e07f2745b4 100644
--- a/src/core/device.c
+++ b/src/core/device.c
@@ -305,6 +305,19 @@ static int device_coldplug(Unit *u) {
                 found &= ~DEVICE_FOUND_UDEV;
                 if (state == DEVICE_PLUGGED)
                         state = DEVICE_TENTATIVE;
+
+                /* Also check the validity of the device syspath. Without this check, if the device was
+                 * removed while switching root, it would never go to inactive state, as both Device.found
+                 * and Device.enumerated_found do not have the DEVICE_FOUND_UDEV flag, so device_catchup() in
+                 * device_update_found_one() does nothing in most cases. See issue #25106. Note that the
+                 * syspath field is only serialized when systemd is sufficiently new and the device has been
+                 * already processed by udevd. */
+                if (d->deserialized_sysfs) {
+                        _cleanup_(sd_device_unrefp) sd_device *dev = NULL;
+
+                        if (sd_device_new_from_syspath(&dev, d->deserialized_sysfs) < 0)
+                                state = DEVICE_DEAD;
+                }
         }
 
         if (d->found == found && d->state == state)