PCI/AER: Work around use-after-free in pcie_do_fatal_recovery()
authorThomas Tai <thomas.tai@oracle.com>
Thu, 26 Jul 2018 17:13:04 +0000 (12:13 -0500)
committerBjorn Helgaas <helgaas@kernel.org>
Thu, 26 Jul 2018 17:13:04 +0000 (12:13 -0500)
When an fatal error is received by a non-bridge device, the device is
removed, and pci_stop_and_remove_bus_device() deallocates the device
structure.  The freed device structure is used by subsequent code to send
uevents and print messages.

Hold a reference on the device until we're finished using it.  This is not
an ideal fix because pcie_do_fatal_recovery() should not use the device at
all after removing it, but that's too big a project for right now.

Fixes: 7e9084b36740 ("PCI/AER: Handle ERR_FATAL with removal and re-enumeration of devices")
Signed-off-by: Thomas Tai <thomas.tai@oracle.com>
[bhelgaas: changelog, reduce get/put coverage]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
drivers/pci/pcie/err.c

index f7ce0cb0b0b70a48902010c3cea0fb1707e5ead9..f02e334beb457da586857736fc8fc45530847fe3 100644 (file)
@@ -295,6 +295,7 @@ void pcie_do_fatal_recovery(struct pci_dev *dev, u32 service)
 
        parent = udev->subordinate;
        pci_lock_rescan_remove();
+       pci_dev_get(dev);
        list_for_each_entry_safe_reverse(pdev, temp, &parent->devices,
                                         bus_list) {
                pci_dev_get(pdev);
@@ -328,6 +329,7 @@ void pcie_do_fatal_recovery(struct pci_dev *dev, u32 service)
                pci_info(dev, "Device recovery from fatal error failed\n");
        }
 
+       pci_dev_put(dev);
        pci_unlock_rescan_remove();
 }