[PATCH 3.13.y-ckt 09/30] x86/irq: Call chip->irq_set_affinity in proper context
From: Kamal Mostafa
Date: Wed Feb 10 2016 - 16:47:45 EST
3.13.11-ckt35 -stable review patch. If anyone has any objections, please let me know.
---8<------------------------------------------------------------
From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
commit e23b257c293ce4bcc8cabb2aa3097b6ed8a8261a upstream.
setup_ioapic_dest() calls irqchip->irq_set_affinity() completely
unprotected. That's wrong in several aspects:
- it opens a race window where irq_set_affinity() can be interrupted and the
irq chip left in unconsistent state.
- it triggers a lockdep splat when we fix the vector race for 4.3+ because
vector lock is taken with interrupts enabled.
The proper calling convention is irq descriptor lock held and interrupts
disabled.
Reported-and-tested-by: Borislav Petkov <bp@xxxxxxxxx>
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Jiang Liu <jiang.liu@xxxxxxxxxxxxxxx>
Cc: Jeremiah Mahler <jmmahler@xxxxxxxxx>
Cc: andy.shevchenko@xxxxxxxxx
Cc: Guenter Roeck <linux@xxxxxxxxxxxx>
Cc: Joe Lawrence <joe.lawrence@xxxxxxxxxxx>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1601140919420.3575@nanos
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
[ kamal: backport to 3.19-stable: context ]
Signed-off-by: Kamal Mostafa <kamal@xxxxxxxxxxxxx>
---
arch/x86/kernel/apic/io_apic.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index e63a5bd..94a3320 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -3610,6 +3610,7 @@ void __init setup_ioapic_dest(void)
{
int pin, ioapic, irq, irq_entry;
const struct cpumask *mask;
+ struct irq_desc *desc;
struct irq_data *idata;
if (skip_ioapic_setup == 1)
@@ -3625,7 +3626,9 @@ void __init setup_ioapic_dest(void)
if ((ioapic > 0) && (irq > 16))
continue;
- idata = irq_get_irq_data(irq);
+ desc = irq_to_desc(irq);
+ raw_spin_lock_irq(&desc->lock);
+ idata = irq_desc_get_irq_data(desc);
/*
* Honour affinities which have been set in early boot
@@ -3636,6 +3639,7 @@ void __init setup_ioapic_dest(void)
mask = apic->target_cpus();
x86_io_apic_ops.set_affinity(idata, mask, false);
+ raw_spin_unlock_irq(&desc->lock);
}
}
--
1.9.1