sched/cputime: Fix steal_account_process_tick() to always return jiffies
authorChris Friesen <cbf123@mail.usask.ca>
Sun, 6 Mar 2016 05:18:48 +0000 (23:18 -0600)
committerIngo Molnar <mingo@kernel.org>
Tue, 8 Mar 2016 11:24:56 +0000 (12:24 +0100)
The callers of steal_account_process_tick() expect it to return
whether a jiffy should be considered stolen or not.

Currently the return value of steal_account_process_tick() is in
units of cputime, which vary between either jiffies or nsecs
depending on CONFIG_VIRT_CPU_ACCOUNTING_GEN.

If cputime has nsecs granularity and there is a tiny amount of
stolen time (a few nsecs, say) then we will consider the entire
tick stolen and will not account the tick on user/system/idle,
causing /proc/stats to show invalid data.

The fix is to change steal_account_process_tick() to accumulate
the stolen time and only account it once it's worth a jiffy.

(Thanks to Frederic Weisbecker for suggestions to fix a bug in my
first version of the patch.)

Signed-off-by: Chris Friesen <chris.friesen@windriver.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/56DBBDB8.40305@mail.usask.ca
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/cputime.c

index 01d9898bc9a222d22291fb7b69b187981b74a1ae..75f98c5498d55d38b0a36bba2719907742166fe1 100644 (file)
@@ -262,21 +262,21 @@ static __always_inline bool steal_account_process_tick(void)
 #ifdef CONFIG_PARAVIRT
        if (static_key_false(&paravirt_steal_enabled)) {
                u64 steal;
-               cputime_t steal_ct;
+               unsigned long steal_jiffies;
 
                steal = paravirt_steal_clock(smp_processor_id());
                steal -= this_rq()->prev_steal_time;
 
                /*
-                * cputime_t may be less precise than nsecs (eg: if it's
-                * based on jiffies). Lets cast the result to cputime
+                * steal is in nsecs but our caller is expecting steal
+                * time in jiffies. Lets cast the result to jiffies
                 * granularity and account the rest on the next rounds.
                 */
-               steal_ct = nsecs_to_cputime(steal);
-               this_rq()->prev_steal_time += cputime_to_nsecs(steal_ct);
+               steal_jiffies = nsecs_to_jiffies(steal);
+               this_rq()->prev_steal_time += jiffies_to_nsecs(steal_jiffies);
 
-               account_steal_time(steal_ct);
-               return steal_ct;
+               account_steal_time(jiffies_to_cputime(steal_jiffies));
+               return steal_jiffies;
        }
 #endif
        return false;