学习9-runqlat有一些问题求教

[niebowen666]

请问在handle_switch函数里对hist进行更新后是eBPF程序是如何把数据传输给用户空间命令行显示的呢？而且我也没有看到把hist进行存储，以供下一次更新。
有大佬知道怎么回事吗？

[Officeyutong]

• handle_switch updated data stored in map hist
• userspace program can access the same map as kernel program does, so it can also access hist
• hist doesn't need to be stored by your program, its handled by kernel, and only a fd is provided to your program so you can operate on it

[niebowen666]

Thanks for your answer!
I have run the program by bash
The command is ./ecli package.json without any parameters.
I view the code, the information about nums of process among different durations is stored in the maps hists
Because I ran the program without any parameters, the hists has only one elem(hist) whose key is -1.
All of information to user space is stored in slots in the hist struct.

There is no user space code. So I am wondering how the bash print the result like this:

Tracing run queue latency... Hit Ctrl-C to end.
     usecs               : count     distribution
         0 -> 1          : 233      |***********                             |
         2 -> 3          : 742      |************************************    |
         4 -> 7          : 203      |**********                              |
         8 -> 15         : 173      |********                                |
        16 -> 31         : 24       |*                                       |
        32 -> 63         : 0        |                                        |
        64 -> 127        : 30       |*                                       |
       128 -> 255        : 6        |                                        |
       256 -> 511        : 3        |                                        |
       512 -> 1023       : 5        |                                        |
      1024 -> 2047       : 27       |*                                       |
      2048 -> 4095       : 30       |*                                       |
      4096 -> 8191       : 20       |                                        |
      8192 -> 16383      : 29       |*                                       |
     16384 -> 32767      : 809      |****************************************|
     32768 -> 65535      : 64       |***                                     |

I konw the workflow about the kernel space, but I haven't seen any code about output in user space.
Looking forward to your answer!
@Officeyutong

[Officeyutong]

Thanks for your answer! I have run the program by bash The command is ./ecli package.json without any parameters. I view the code, the information about nums of process among different durations is stored in the maps hists Because I ran the program without any parameters, the hists has only one elem(hist) whose key is -1. All of information to user space is stored in slots in the hist struct.

There is no user space code. So I am wondering how the bash print the result like this:

Tracing run queue latency... Hit Ctrl-C to end.
     usecs               : count     distribution
         0 -> 1          : 233      |***********                             |
         2 -> 3          : 742      |************************************    |
         4 -> 7          : 203      |**********                              |
         8 -> 15         : 173      |********                                |
        16 -> 31         : 24       |*                                       |
        32 -> 63         : 0        |                                        |
        64 -> 127        : 30       |*                                       |
       128 -> 255        : 6        |                                        |
       256 -> 511        : 3        |                                        |
       512 -> 1023       : 5        |                                        |
      1024 -> 2047       : 27       |*                                       |
      2048 -> 4095       : 30       |*                                       |
      4096 -> 8191       : 20       |                                        |
      8192 -> 16383      : 29       |*                                       |
     16384 -> 32767      : 809      |****************************************|
     32768 -> 65535      : 64       |***                                     |

I konw the workflow about the kernel space, but I haven't seen any code about output in user space. Looking forward to your answer! @Officeyutong

eunomia-bpf automatically detects map types and decides how to show that. Related logic is in https://github.com/eunomia-bpf/eunomia-bpf/blob/291304bf415817d2f5b46a3655d9f9594b8672a6/bpf-loader-rs/bpf-loader-lib/src/export_event/event_handlers/sample_map.rs#L118

For runqlat, there is a special mark in its header file (See

bpf-developer-tutorial/src/9-runqlat/runqlat.bpf.c

Line 38 in 5e62255

/// @sample {"interval": 1000, "type" : "log2_hist"}

), eunomia-bpf will parse this header to retrive these special marked maps, and use the corresponding way to show the data in it