linux-kernel - [PATCH v2] PCI/PM: Prevent runtime suspend before devices are fully initialized

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20251017122123.v2.1.I60a53c170a8596661883bd2b4ef475155c7aa72b@changeid>
Date: Fri, 17 Oct 2025 12:21:23 -0700
From: Brian Norris <briannorris@...omium.org>
To: Bjorn Helgaas <bhelgaas@...gle.com>
Cc: "Rafael J . Wysocki" <rafael@...nel.org>,
	linux-kernel@...r.kernel.org,
	Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>,
	linux-pm@...r.kernel.org,
	Lukas Wunner <lukas@...ner.de>,
	linux-pci@...r.kernel.org,
	Brian Norris <briannorris@...omium.org>
Subject: [PATCH v2] PCI/PM: Prevent runtime suspend before devices are fully initialized

Today, it's possible for a PCI device to be created and
runtime-suspended before it is fully initialized. When that happens, the
device will remain in D0, but the suspend process may save an
intermediate version of that device's state -- for example, without
appropriate BAR configuration. When the device later resumes, we'll
restore invalid PCI state and the device may not function.

Prevent runtime suspend for PCI devices by deferring pm_runtime_enable()
until we've fully initialized the device.

More details on how exactly this may occur:

1. PCI device is created by pci_scan_slot() or similar
2. As part of pci_scan_slot(), pci_pm_init() enables runtime PM; the
   device starts "active" and we initially prevent (pm_runtime_forbid())
   suspend -- but see [*] footnote
3. Underlying 'struct device' is added to the system (device_add());
   runtime PM can now be configured by user space
4. PCI device receives BAR configuration
   (pci_assign_unassigned_bus_resources(), etc.)
5. PCI device is added to the system in pci_bus_add_device()

The device may potentially suspend between #3 and #4.

[*] By default, pm_runtime_forbid() prevents suspending a device; but by
design, this can be overridden by user space policy via

  echo auto > /sys/bus/pci/devices/.../power/control

Thus, the above #3/#4 sequence is racy with user space (udev or
similar).

Notably, many PCI devices are enumerated at subsys_initcall time and so
will not race with user space. However, there are several scenarios
where PCI devices are created later on, such as with hotplug or when
drivers (pwrctrl or controller drivers) are built as modules.

Signed-off-by: Brian Norris <briannorris@...omium.org>
Cc: <stable@...r.kernel.org>
---

Changes in v2:
 * Update CC list
 * Rework problem description
 * Update solution: defer pm_runtime_enable(), instead of trying to
   get()/put()

 drivers/pci/bus.c | 3 +++
 drivers/pci/pci.c | 1 -
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
index f26aec6ff588..fc66b6cb3a54 100644
--- a/drivers/pci/bus.c
+++ b/drivers/pci/bus.c
@@ -14,6 +14,7 @@
 #include <linux/of.h>
 #include <linux/of_platform.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 #include <linux/proc_fs.h>
 #include <linux/slab.h>

@@ -375,6 +376,8 @@ void pci_bus_add_device(struct pci_dev *dev)
 		put_device(&pdev->dev);
 	}

+	pm_runtime_enable(&dev->dev);
+
 	if (!dn || of_device_is_available(dn))
 		pci_dev_allow_binding(dev);

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index b14dd064006c..f792164fa297 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3226,7 +3226,6 @@ void pci_pm_init(struct pci_dev *dev)
 	pci_pm_power_up_and_verify_state(dev);
 	pm_runtime_forbid(&dev->dev);
 	pm_runtime_set_active(&dev->dev);
-	pm_runtime_enable(&dev->dev);
 }

 static unsigned long pci_ea_flags(struct pci_dev *dev, u8 prop)
-- 
2.51.0.858.gf9c4a03a3a-goog