]> www.infradead.org Git - users/dwmw2/linux.git/commitdiff
x86/hotplug: Silence APIC only after all interrupts are migrated
authorAshok Raj <ashok.raj@intel.com>
Thu, 27 Aug 2020 04:12:10 +0000 (21:12 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 3 Sep 2020 09:27:06 +0000 (11:27 +0200)
commit 52d6b926aabc47643cd910c85edb262b7f44c168 upstream.

There is a race when taking a CPU offline. Current code looks like this:

native_cpu_disable()
{
...
apic_soft_disable();
/*
 * Any existing set bits for pending interrupt to
 * this CPU are preserved and will be sent via IPI
 * to another CPU by fixup_irqs().
 */
cpu_disable_common();
{
....
/*
 * Race window happens here. Once local APIC has been
 * disabled any new interrupts from the device to
 * the old CPU are lost
 */
fixup_irqs(); // Too late to capture anything in IRR.
...
}
}

The fix is to disable the APIC *after* cpu_disable_common().

Testing was done with a USB NIC that provided a source of frequent
interrupts. A script migrated interrupts to a specific CPU and
then took that CPU offline.

Fixes: 60dcaad5736f ("x86/hotplug: Silence APIC and NMI when CPU is dead")
Reported-by: Evan Green <evgreen@chromium.org>
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Tested-by: Evan Green <evgreen@chromium.org>
Reviewed-by: Evan Green <evgreen@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/lkml/875zdarr4h.fsf@nanos.tec.linutronix.de/
Link: https://lore.kernel.org/r/1598501530-45821-1-git-send-email-ashok.raj@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kernel/smpboot.c

index 9674321ce3a3b6aa4cd93f78ae7e3cb2bd287f10..8367bd7a9a810cc303bd3e7644a8accb3e5add22 100644 (file)
@@ -1599,14 +1599,28 @@ int native_cpu_disable(void)
        if (ret)
                return ret;
 
-       /*
-        * Disable the local APIC. Otherwise IPI broadcasts will reach
-        * it. It still responds normally to INIT, NMI, SMI, and SIPI
-        * messages.
-        */
-       apic_soft_disable();
        cpu_disable_common();
 
+        /*
+         * Disable the local APIC. Otherwise IPI broadcasts will reach
+         * it. It still responds normally to INIT, NMI, SMI, and SIPI
+         * messages.
+         *
+         * Disabling the APIC must happen after cpu_disable_common()
+         * which invokes fixup_irqs().
+         *
+         * Disabling the APIC preserves already set bits in IRR, but
+         * an interrupt arriving after disabling the local APIC does not
+         * set the corresponding IRR bit.
+         *
+         * fixup_irqs() scans IRR for set bits so it can raise a not
+         * yet handled interrupt on the new destination CPU via an IPI
+         * but obviously it can't do so for IRR bits which are not set.
+         * IOW, interrupts arriving after disabling the local APIC will
+         * be lost.
+         */
+       apic_soft_disable();
+
        return 0;
 }