[PATCH 37/37] s390: Prevent from cputime leaks

[Frederic Weisbecker <fweisbec@gmail.com>]

The s390 clock has a higher granularity than nanoseconds. 1 nanosec
equals 4.096 in s390 cputime_t. Therefore we leak a remainder while
flushing the cputime through cputime_to_nsecs().

For more precision, make sure we keep that remainder on cputime
accumulators for later accounting.

Reported-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Wanpeng Li <wanpeng.li@hotmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
---
 arch/s390/include/asm/cputime.h |  4 ++++
 arch/s390/kernel/vtime.c        | 33 +++++++++++++++++++++------------
 2 files changed, 25 insertions(+), 12 deletions(-)

diff --git a/arch/s390/include/asm/cputime.h b/arch/s390/include/asm/cputime.h
index d1c407d..86b5e3f 100644
--- a/arch/s390/include/asm/cputime.h
+++ b/arch/s390/include/asm/cputime.h
@@ -33,6 +33,10 @@ static inline unsigned int cputime_to_usecs(const cputime_t cputime)
 	return (__force unsigned long long) cputime >> 12;
 }
 
+static inline cputime_t nsecs_to_cputime(u64 cputime)
+{
+	return (__force cputime_t) cputime * CPUTIME_PER_USEC * NSEC_PER_USEC;
+}
 
 u64 arch_cpu_idle_time(int cpu);
 
diff --git a/arch/s390/kernel/vtime.c b/arch/s390/kernel/vtime.c
index b4a3e9e..922f959 100644
--- a/arch/s390/kernel/vtime.c
+++ b/arch/s390/kernel/vtime.c
@@ -94,8 +94,14 @@ static inline u64 update_tsk_timer(unsigned long *tsk_vtime, u64 new)
 {
 	u64 delta;
 
-	delta = new - *tsk_vtime;
-	*tsk_vtime = new;
+	/*
+	 * Since nsecs is less granular than cputime_t in s390,
+	 * the conversion to nsecs is rounded. Make sure we don't
+	 * lose the remainder.
+	 */
+	delta = cputime_to_nsecs(new - *tsk_vtime);
+	*tsk_vtime += nsecs_to_cputime(delta);
+
 	return delta;
 }
 
@@ -124,7 +130,8 @@ static void account_system_index_scaled(struct task_struct *p,
  */
 static int do_account_vtime(struct task_struct *tsk)
 {
-	u64 timer, clock, user, guest, system, hardirq, softirq, steal;
+	u64 timer, clock, delta;
+	u64 user, guest, system, hardirq, softirq, steal;
 
 	timer = S390_lowcore.last_update_timer;
 	clock = S390_lowcore.last_update_clock;
@@ -161,18 +168,19 @@ static int do_account_vtime(struct task_struct *tsk)
 				   READ_ONCE(S390_lowcore.hardirq_timer));
 	softirq = update_tsk_timer(&tsk->thread.softirq_timer,
 				   READ_ONCE(S390_lowcore.softirq_timer));
-	S390_lowcore.steal_timer +=
-		clock - user - guest - system - hardirq - softirq;
+
+	delta = nsecs_to_cputime(user + guest + system + hardirq + softirq);
+	S390_lowcore.steal_timer += clock - delta;
 
 	/* Push account value */
 	if (user) {
-		account_user_time(tsk, cputime_to_nsecs(user));
-		tsk->utimescaled += cputime_to_nsecs(scale_vtime(user));
+		account_user_time(tsk, user);
+		tsk->utimescaled += scale_vtime(user);
 	}
 
 	if (guest) {
-		account_guest_time(tsk, cputime_to_nsecs(guest));
-		tsk->utimescaled += cputime_to_nsecs(scale_vtime(guest));
+		account_guest_time(tsk, guest);
+		tsk->utimescaled += scale_vtime(guest);
 	}
 
 	if (system)
@@ -187,11 +195,12 @@ static int do_account_vtime(struct task_struct *tsk)
 
 	steal = S390_lowcore.steal_timer;
 	if ((s64) steal > 0) {
-		S390_lowcore.steal_timer = 0;
-		account_steal_time(cputime_to_nsecs(steal));
+		u64 nsecs = cputime_to_nsecs(steal);
+		S390_lowcore.steal_timer -= nsecs_to_cputime(nsecs);
+		account_steal_time(nsecs);
 	}
 
-	return virt_timer_forward(user + guest + system + hardirq + softirq);
+	return virt_timer_forward(delta);
 }
 
 void vtime_task_switch(struct task_struct *prev)
-- 
2.7.4