xtime_lock vs update_process_times

Commit d3d74453c34f8fd87674a8cf5b8a327c68f22e99 ("hrtimer: fixup the
HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback") broke several archs, and since
only Russell bothered to merge the fix, and Greg to ACK his arch, I'm
sending this for merger.

I have confirmation that the Alpha bit results in a booting kernel.
That leaves: blackfin, frv, sh and sparc untested.

The deadlock in question was found by Russell:

  IRQ handle
    -> timer_tick() - xtime seqlock held for write
      -> update_process_times()
        -> run_local_timers()
          -> hrtimer_run_queues()
            -> hrtimer_get_softirq_time() - tries to get a read lock

Now, Thomas assures me the fix is trivial, only do_timer() needs to be
done under the xtime_lock, and update_process_times() can savely be
removed from under it.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Greg Ungerer <gerg@uclinux.org>
CC: Richard Henderson <rth@twiddle.net>
CC: Bryan Wu <bryan.wu@analog.com>
CC: David Howells <dhowells@redhat.com>
CC: Paul Mundt <lethal@linux-sh.org>
CC: William Irwin <wli@holomorphy.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 files changed, 28 insertions, 33 deletions

diff --git a/arch/alpha/kernel/time.c b/arch/alpha/kernel/time.c
index 1dd50d07693..75480cab089 100644
--- a/arch/alpha/kernel/time.c
+++ b/arch/alpha/kernel/time.c
@@ -119,13 +119,8 @@ irqreturn_t timer_interrupt(int irq, void *dev)
 	state.partial_tick = delta & ((1UL << FIX_SHIFT) - 1); 
 	nticks = delta >> FIX_SHIFT;
 
-	while (nticks > 0) {
-		do_timer(1);
-#ifndef CONFIG_SMP
-		update_process_times(user_mode(get_irq_regs()));
-#endif
-		nticks--;
-	}
+	if (nticks)
+		do_timer(nticks);
 
 	/*
 	 * If we have an externally synchronized Linux clock, then update
@@ -141,6 +136,12 @@ irqreturn_t timer_interrupt(int irq, void *dev)
 	}
 
 	write_sequnlock(&xtime_lock);
+
+#ifndef CONFIG_SMP
+	while (nticks--)
+		update_process_times(user_mode(get_irq_regs()));
+#endif
+
 	return IRQ_HANDLED;
 }
 
diff --git a/arch/blackfin/kernel/time.c b/arch/blackfin/kernel/time.c
index 5bd64e341df..9bdc8f99183 100644
--- a/arch/blackfin/kernel/time.c
+++ b/arch/blackfin/kernel/time.c
@@ -137,9 +137,6 @@ irqreturn_t timer_interrupt(int irq, void *dummy)
 
 	do_timer(1);
 
-#ifndef CONFIG_SMP
-	update_process_times(user_mode(get_irq_regs()));
-#endif
 	profile_tick(CPU_PROFILING);
 
 	/*
@@ -161,6 +158,11 @@ irqreturn_t timer_interrupt(int irq, void *dummy)
 			last_rtc_update = xtime.tv_sec - 600;
 	}
 	write_sequnlock(&xtime_lock);
+
+#ifndef CONFIG_SMP
+	update_process_times(user_mode(get_irq_regs()));
+#endif
+
 	return IRQ_HANDLED;
 }
 
diff --git a/arch/frv/kernel/time.c b/arch/frv/kernel/time.c
index 925fb0199a0..69f6a4ef5d6 100644
--- a/arch/frv/kernel/time.c
+++ b/arch/frv/kernel/time.c
@@ -63,6 +63,7 @@ static irqreturn_t timer_interrupt(int irq, void *dummy)
 	/* last time the cmos clock got updated */
 	static long last_rtc_update = 0;
 
+	profile_tick(CPU_PROFILING);
 	/*
 	 * Here we are in the timer irq handler. We just have irqs locally
 	 * disabled but we don't know if the timer_bh is running on the other
@@ -73,8 +74,6 @@ static irqreturn_t timer_interrupt(int irq, void *dummy)
 	write_seqlock(&xtime_lock);
 
 	do_timer(1);
-	update_process_times(user_mode(get_irq_regs()));
-	profile_tick(CPU_PROFILING);
 
 	/*
 	 * If we have an externally synchronized Linux clock, then update
@@ -99,6 +98,9 @@ static irqreturn_t timer_interrupt(int irq, void *dummy)
 #endif /* CONFIG_HEARTBEAT */
 
 	write_sequnlock(&xtime_lock);
+
+	update_process_times(user_mode(get_irq_regs()));
+
 	return IRQ_HANDLED;
 }
 
diff --git a/arch/m68knommu/kernel/time.c b/arch/m68knommu/kernel/time.c
index 89cdbcaeb45..0ccfb2ad638 100644
--- a/arch/m68knommu/kernel/time.c
+++ b/arch/m68knommu/kernel/time.c
@@ -42,14 +42,12 @@ irqreturn_t arch_timer_interrupt(int irq, void *dummy)
 	/* last time the cmos clock got updated */
 	static long last_rtc_update=0;
 
+	if (current->pid)
+		profile_tick(CPU_PROFILING);
+
 	write_seqlock(&xtime_lock);
 
 	do_timer(1);
-#ifndef CONFIG_SMP
-	update_process_times(user_mode(get_irq_regs()));
-#endif
-	if (current->pid)
-		profile_tick(CPU_PROFILING);
 
 	/*
 	 * If we have an externally synchronized Linux clock, then update
@@ -67,6 +65,10 @@ irqreturn_t arch_timer_interrupt(int irq, void *dummy)
 	}
 
 	write_sequnlock(&xtime_lock);
+
+#ifndef CONFIG_SMP
+	update_process_times(user_mode(get_irq_regs()));
+#endif
 	return(IRQ_HANDLED);
 }
 
diff --git a/arch/sh/kernel/timers/timer-cmt.c b/arch/sh/kernel/timers/timer-cmt.c
index 499e07beebe..71312324b5d 100644
--- a/arch/sh/kernel/timers/timer-cmt.c
+++ b/arch/sh/kernel/timers/timer-cmt.c
@@ -100,16 +100,7 @@ static irqreturn_t cmt_timer_interrupt(int irq, void *dev_id)
 	timer_status &= ~0x80;
 	ctrl_outw(timer_status, CMT_CMCSR_0);
 
-	/*
-	 * Here we are in the timer irq handler. We just have irqs locally
-	 * disabled but we don't know if the timer_bh is running on the other
-	 * CPU. We need to avoid to SMP race with it. NOTE: we don' t need
-	 * the irq version of write_lock because as just said we have irq
-	 * locally disabled. -arca
-	 */
-	write_seqlock(&xtime_lock);
 	handle_timer_tick();
-	write_sequnlock(&xtime_lock);
 
 	return IRQ_HANDLED;
 }
diff --git a/arch/sh/kernel/timers/timer-mtu2.c b/arch/sh/kernel/timers/timer-mtu2.c
index b7499a2a918..463cd08f951 100644
--- a/arch/sh/kernel/timers/timer-mtu2.c
+++ b/arch/sh/kernel/timers/timer-mtu2.c
@@ -100,9 +100,7 @@ static irqreturn_t mtu2_timer_interrupt(int irq, void *dev_id)
 	ctrl_outb(timer_status, MTU2_TSR_1);
 
 	/* Do timer tick */
-	write_seqlock(&xtime_lock);
 	handle_timer_tick();
-	write_sequnlock(&xtime_lock);
 
 	return IRQ_HANDLED;
 }
diff --git a/arch/sparc/kernel/pcic.c b/arch/sparc/kernel/pcic.c
index 4cd5d7818dc..a6a6f982337 100644
--- a/arch/sparc/kernel/pcic.c
+++ b/arch/sparc/kernel/pcic.c
@@ -713,10 +713,10 @@ static irqreturn_t pcic_timer_handler (int irq, void *h)
 	write_seqlock(&xtime_lock);	/* Dummy, to show that we remember */
 	pcic_clear_clock_irq();
 	do_timer(1);
+	write_sequnlock(&xtime_lock);
 #ifndef CONFIG_SMP
 	update_process_times(user_mode(get_irq_regs()));
 #endif
-	write_sequnlock(&xtime_lock);
 	return IRQ_HANDLED;
 }
 
diff --git a/arch/sparc/kernel/time.c b/arch/sparc/kernel/time.c
index 00b393c3a4a..cfaf22c05bc 100644
--- a/arch/sparc/kernel/time.c
+++ b/arch/sparc/kernel/time.c
@@ -128,10 +128,6 @@ irqreturn_t timer_interrupt(int irq, void *dev_id)
 	clear_clock_irq();
 
 	do_timer(1);
-#ifndef CONFIG_SMP
-	update_process_times(user_mode(get_irq_regs()));
-#endif
-
 
 	/* Determine when to update the Mostek clock. */
 	if (ntp_synced() &&
@@ -145,6 +141,9 @@ irqreturn_t timer_interrupt(int irq, void *dev_id)
 	}
 	write_sequnlock(&xtime_lock);
 
+#ifndef CONFIG_SMP
+	update_process_times(user_mode(get_irq_regs()));
+#endif
 	return IRQ_HANDLED;
 }