进程时间片

colinchhe

1. 进程在调用内核函数运行时，这时运行的内核函数有没有优先级这个概念的？如果有的话是不是代表了该进程的优先级？
   2.当进程调用内核函数时，这时进程的时间片有没有消耗，还是说运行在用户态才会消耗时间片？

T-Bagwell

1. 进程在调用内核函数运行时，这时运行的内核函数有没有优先级这个概念的？如果有的话是不是代表了该进程的 ...
colinchhe 发表于 2010-10-21 19:50

  2. 昨天和朋友讨论过了，有，有个patch，可以让他没有，细节http://git.kernel.org/?p=linux/kernel/git/x86/linux-2.6-tip.git;a=commitdiff;h=b52bfee445d315549d41eacf2fa7c156e7d153d5

论坛ID

我看书上说，现在的调度器不用时间片了

colinchhe

我用的是2.6.31版本，不用时间片，那么用什么了？

colinchhe

2. 昨天和朋友讨论过了，有，有个patch，可以让他没有，细节;a=commitdiff;h=b52bfee445d315549d41 ...
T-Bagwell 发表于 2010-10-21 19:52

   
  问题1，是不是代表当前的调用这个内核函数的进程优先级？

问题2，可以通过补丁形式让进程在调用内核函数时不消耗时间片？

T-Bagwell

问题1，是不是代表当前的调用这个内核函数的进程优先级？

问题2，可以通过补丁形式让进 ...
colinchhe 发表于 2010-10-22 14:19

1. 
   
2. does
   
3. the fine granularity accounting of user, system, hardirq, softirq times.
   
4. Adding that option on archs like x86 will be challenging however, given the
   
5. state of TSC reliability on various platforms and also the overhead it will
   
6. add in syscall entry exit.
   
7. 
   
8. Instead, add a lighter variant that only does finer accounting of
   
9. hardirq and softirq times, providing precise irq times (instead of timer tick
   
10. based samples). This accounting is added with a new config option
    
11. CONFIG_IRQ_TIME_ACCOUNTING so that there won't be any overhead for users not
    
12. interested in paying the perf penalty.
    
13. 
    
14. This accounting is based on sched_clock, with the code being generic.
    
15. So, other archs may find it useful as well.
    
16. 
    
17. This patch just adds the core logic and does not enable this logic yet
    
18. 
    
19. 
    
20. 
    
21. 
    
22. 
    
23. 
    
24. 
    
25. +static DEFINE_PER_CPU(u64, cpu_hardirq_time);
    
26. +static DEFINE_PER_CPU(u64, cpu_softirq_time);
    
27. +
    
28. +static DEFINE_PER_CPU(u64, irq_start_time);
    
29. +static int sched_clock_irqtime;
    
30. +
    
31. +void enable_sched_clock_irqtime(void)
    
32. +{
    
33. +       sched_clock_irqtime = 1;
    
34. +}
    
35. +
    
36. +void disable_sched_clock_irqtime(void)
    
37. +{
    
38. +       sched_clock_irqtime = 0;
    
39. +}
    
40. +
    
41. +void account_system_vtime(struct task_struct *curr)
    
42. +{
    
43. +       unsigned long flags;
    
44. +       int cpu;
    
45. +       u64 now, delta;
    
46. +
    
47. +       if (!sched_clock_irqtime)
    
48. +               return;
    
49. +
    
50. +       local_irq_save(flags);
    
51. +
    
52. +       now = sched_clock();
    
53. +       cpu = smp_processor_id();
    
54. +       delta = now - per_cpu(irq_start_time, cpu);
    
55. +       per_cpu(irq_start_time, cpu) = now;
    
56. +       /*
    
57. +        * We do not account for softirq time from ksoftirqd here.
    
58. +        * We want to continue accounting softirq time to ksoftirqd thread
    
59. +        * in that case, so as not to confuse scheduler with a special task
    
60. +        * that do not consume any time, but still wants to run.
    
61. +        */
    
62. +       if (hardirq_count())
    
63. +               per_cpu(cpu_hardirq_time, cpu) += delta;
    
64. +       else if (in_serving_softirq() && !(curr->flags & PF_KSOFTIRQD))
    
65. +               per_cpu(cpu_softirq_time, cpu) += delta;
    
66. +
    
67. +       local_irq_restore(flags);
    
68. +}
    
69. +
    
70. +#endif
    
71. +
    

复制代码

就是说，如果sched_clock_irqtime=0,就return了，就不now = sched_clock();了