kick_handler() doesn't have an irq_enter/exit pair, but it's used for
handling SMP IPIs which require work to be done in softirqs, which are
invoked from irq_exit() when the hard irq nest count reaches 0.
The scheduler_ipi() callback in the IPI handler calls irq_enter/exit
itself, but this is inside kick_handler()'s spin lock critical section,
so if an invoked softirq issues an IPI the kick_handler() will be
re-entered on the same CPU and will deadlock.
This is easily fixed by adding the missing irq_enter/exit to
kick_handler() so that the hard irq nest count doesn't reach 0 until
after the spin lock has been released.
Ideally the spin lock protected handler list will also be replaced by a
lockless RCU protected list since it is certainly mostly read. That can
be done in a later change though.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
* pass it as an argument.
*/
#include <linux/export.h>
+#include <linux/hardirq.h>
+#include <linux/irq.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/types.h>
TBIRES
kick_handler(TBIRES State, int SigNum, int Triggers, int Inst, PTBI pTBI)
{
+ struct pt_regs *old_regs;
struct kick_irq_handler *kh;
struct list_head *lh;
int handled = 0;
trace_hardirqs_off();
+ old_regs = set_irq_regs((struct pt_regs *)State.Sig.pCtx);
+ irq_enter();
+
/*
* There is no need to disable interrupts here because we
* can't nest KICK interrupts in a KICK interrupt handler.
WARN_ON(!handled);
+ irq_exit();
+ set_irq_regs(old_regs);
+
return tail_end(ret);
}