PCI/keystone: Fix race in installing chained IRQ handler
authorThomas Gleixner <tglx@linutronix.de>
Sun, 21 Jun 2015 18:16:09 +0000 (20:16 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Thu, 25 Jun 2015 09:57:01 +0000 (11:57 +0200)
Fix a race where a pending interrupt could be received and the handler
called before the handler's data has been setup, by converting to
irq_set_chained_handler_and_data().

Search and conversion was done with coccinelle:

@@
expression E1, E2, E3;
@@
(
-if (irq_set_chained_handler(E1, E3) != 0)
-   BUG();
|
-irq_set_chained_handler(E1, E3);
)
-irq_set_handler_data(E1, E2);
+irq_set_chained_handler_and_data(E1, E3, E2);

@@
expression E1, E2, E3;
@@
(
-if (irq_set_chained_handler(E1, E3) != 0)
-   BUG();
...
|
-irq_set_chained_handler(E1, E3);
...
)
-irq_set_handler_data(E1, E2);
+irq_set_chained_handler_and_data(E1, E3, E2);

Reported-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Murali Karicheri <m-karicheri2@ti.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: linux-pci@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
drivers/pci/host/pci-keystone.c

index 75333b0..47cf0c1 100644 (file)
@@ -223,10 +223,9 @@ static void ks_pcie_setup_interrupts(struct keystone_pcie *ks_pcie)
        /* MSI IRQ */
        if (IS_ENABLED(CONFIG_PCI_MSI)) {
                for (i = 0; i < ks_pcie->num_msi_host_irqs; i++) {
-                       irq_set_chained_handler(ks_pcie->msi_host_irqs[i],
-                                               ks_pcie_msi_irq_handler);
-                       irq_set_handler_data(ks_pcie->msi_host_irqs[i],
-                                            ks_pcie);
+                       irq_set_chained_handler_and_data(ks_pcie->msi_host_irqs[i],
+                                                        ks_pcie_msi_irq_handler,
+                                                        ks_pcie);
                }
        }
 }