Commit 6441402b1f173fa38e561d3cee7c01c32e5281ad
1 parent
72d31053f6
Exists in
master
and in
20 other branches
clockevents: prevent cpu online to interfere with nohz
Impact: rare hang which can be triggered on CPU online.
tick_do_timer_cpu keeps track of the CPU which updates jiffies
via do_timer. The value -1 is used to signal, that currently no
CPU is doing this. There are two cases, where the variable can
have this state:
boot:
necessary for systems where the boot cpu id can be != 0
nohz long idle sleep:
When the CPU which did the jiffies update last goes into
a long idle sleep it drops the update jiffies duty so
another CPU which is not idle can pick it up and keep
jiffies going.
Using the same value for both situations is wrong, as the CPU online
code can see the -1 state when the timer of the newly onlined CPU is
setup. The setup for a newly onlined CPU goes through periodic mode
and can pick up the do_timer duty without being aware of the nohz /
highres mode of the already running system.
Use two separate states and make them constants to avoid magic
numbers confusion.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Showing 3 changed files with 12 additions and 7 deletions Side-by-side Diff
kernel/time/tick-common.c
| ... | ... | @@ -33,7 +33,7 @@ |
| 33 | 33 | */ |
| 34 | 34 | ktime_t tick_next_period; |
| 35 | 35 | ktime_t tick_period; |
| 36 | -int tick_do_timer_cpu __read_mostly = -1; | |
| 36 | +int tick_do_timer_cpu __read_mostly = TICK_DO_TIMER_BOOT; | |
| 37 | 37 | DEFINE_SPINLOCK(tick_device_lock); |
| 38 | 38 | |
| 39 | 39 | /* |
| ... | ... | @@ -148,7 +148,7 @@ |
| 148 | 148 | * If no cpu took the do_timer update, assign it to |
| 149 | 149 | * this cpu: |
| 150 | 150 | */ |
| 151 | - if (tick_do_timer_cpu == -1) { | |
| 151 | + if (tick_do_timer_cpu == TICK_DO_TIMER_BOOT) { | |
| 152 | 152 | tick_do_timer_cpu = cpu; |
| 153 | 153 | tick_next_period = ktime_get(); |
| 154 | 154 | tick_period = ktime_set(0, NSEC_PER_SEC / HZ); |
| ... | ... | @@ -300,7 +300,8 @@ |
| 300 | 300 | if (*cpup == tick_do_timer_cpu) { |
| 301 | 301 | int cpu = first_cpu(cpu_online_map); |
| 302 | 302 | |
| 303 | - tick_do_timer_cpu = (cpu != NR_CPUS) ? cpu : -1; | |
| 303 | + tick_do_timer_cpu = (cpu != NR_CPUS) ? cpu : | |
| 304 | + TICK_DO_TIMER_NONE; | |
| 304 | 305 | } |
| 305 | 306 | spin_unlock_irqrestore(&tick_device_lock, flags); |
| 306 | 307 | } |
kernel/time/tick-internal.h
kernel/time/tick-sched.c
| ... | ... | @@ -221,7 +221,7 @@ |
| 221 | 221 | */ |
| 222 | 222 | if (unlikely(!cpu_online(cpu))) { |
| 223 | 223 | if (cpu == tick_do_timer_cpu) |
| 224 | - tick_do_timer_cpu = -1; | |
| 224 | + tick_do_timer_cpu = TICK_DO_TIMER_NONE; | |
| 225 | 225 | } |
| 226 | 226 | |
| 227 | 227 | if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE)) |
| ... | ... | @@ -303,7 +303,7 @@ |
| 303 | 303 | * invoked. |
| 304 | 304 | */ |
| 305 | 305 | if (cpu == tick_do_timer_cpu) |
| 306 | - tick_do_timer_cpu = -1; | |
| 306 | + tick_do_timer_cpu = TICK_DO_TIMER_NONE; | |
| 307 | 307 | |
| 308 | 308 | ts->idle_sleeps++; |
| 309 | 309 | |
| ... | ... | @@ -468,7 +468,7 @@ |
| 468 | 468 | * this duty, then the jiffies update is still serialized by |
| 469 | 469 | * xtime_lock. |
| 470 | 470 | */ |
| 471 | - if (unlikely(tick_do_timer_cpu == -1)) | |
| 471 | + if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE)) | |
| 472 | 472 | tick_do_timer_cpu = cpu; |
| 473 | 473 | |
| 474 | 474 | /* Check, if the jiffies need an update */ |
| ... | ... | @@ -570,7 +570,7 @@ |
| 570 | 570 | * this duty, then the jiffies update is still serialized by |
| 571 | 571 | * xtime_lock. |
| 572 | 572 | */ |
| 573 | - if (unlikely(tick_do_timer_cpu == -1)) | |
| 573 | + if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE)) | |
| 574 | 574 | tick_do_timer_cpu = cpu; |
| 575 | 575 | #endif |
| 576 | 576 |