Kcpustat is not correctly supported on nohz_full CPUs. The tick doesn't fire and the cputime therefore doesn't move forward. The issue has shown up after the vanishing of the remaining 1Hz which has made the stall visible. We are solving that with checking the task running on a CPU through RCU and reading its vtime delta that we add to the raw kcpustat values. We make sure that we fetch a coherent raw-kcpustat/vtime-delta couple sequence while checking that the CPU referred by the target vtime is the correct one, under the locked vtime seqcount. Only CPUTIME_SYSTEM is handled here as a start because it's the trivial case. User and guest time will require more preparation work to correctly handle niceness.
Showing the changes to include/linux (1 of 2 files). Show all
Unified · Split
| @@ -78,6 +78,17 @@ static inline unsigned int kstat_cpu_irqs_sum(unsigned int cpu) · 77 unmodified lines | ||
| 78 | 78 | return kstat_cpu(cpu).irqs_sum; |
| 79 | 79 | } |
| 80 | 80 | |
| 81 | +#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN | |
| 82 | +extern u64 kcpustat_field(struct kernel_cpustat *kcpustat, | |
| 83 | + enum cpu_usage_stat usage, int cpu); | |
| 84 | +#else | |
| 85 | +static inline u64 kcpustat_field(struct kernel_cpustat *kcpustat, | |
| 86 | + enum cpu_usage_stat usage, int cpu) | |
| 87 | +{ | |
| 88 | + return kcpustat->cpustat[usage]; | |
| 89 | +} | |
| 90 | +#endif | |
| 91 | + | |
| 81 | 92 | extern void account_user_time(struct task_struct *, u64); |
| 82 | 93 | extern void account_guest_time(struct task_struct *, u64); |
| 83 | 94 | extern void account_system_time(struct task_struct *, int, u64); |