[OLK-6.6] Backport sched_ext to OLK-6.6 · Pull Request !19088 · openEuler/kernel

基于6.6内核现有的bpf_struct_ops和注册机制，Backport目前upstream 6.12已经合入的sched_ext相关patches以及部分必要依赖到OLK-6.6内核分支上，用于实现在openEuler 6.6内核上支持sched_ext的主要功能实现用户态编程调度器。
upstream patches: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=88264981f2082248e892a706b2c5004650faac54
在aarch64和x86_64架构上所有测试用例和单元测试都测试通过。
单元测试结果：
```
[root@kvm1 sched_ext]# ./runner 
===== START =====
TEST: create_dsq
DESCRIPTION: Create and destroy a dsq in a loop
OUTPUT:
ok 1 create_dsq # 
=====  END  =====
===== START =====
TEST: enq_last_no_enq_fails
DESCRIPTION: Verify we fail to load a scheduler if we specify the SCX_OPS_ENQ_LAST flag without defining ops.enqueue()
OUTPUT:
ok 2 enq_last_no_enq_fails # 
=====  END  =====
===== START =====
TEST: enq_select_cpu_fails
DESCRIPTION: Verify we fail to call scx_bpf_select_cpu_dfl() from ops.enqueue()
OUTPUT:
ok 3 enq_select_cpu_fails # 
=====  END  =====
===== START =====
TEST: ddsp_bogus_dsq_fail
DESCRIPTION: Verify we gracefully fail, and fall back to using a built-in DSQ, if we do a direct dispatch to an invalid DSQ in ops.select_cpu()
OUTPUT:
ok 4 ddsp_bogus_dsq_fail # 
=====  END  =====
===== START =====
TEST: ddsp_vtimelocal_fail
DESCRIPTION: Verify we gracefully fail, and fall back to using a built-in DSQ, if we do a direct vtime dispatch to a built-in DSQ from DSQ in ops.select_cpu()
OUTPUT:
ok 5 ddsp_vtimelocal_fail # 
=====  END  =====
===== START =====
TEST: dsp_local_on
DESCRIPTION: Verify we can directly dispatch tasks to a local DSQs from osp.dispatch()
OUTPUT:
ok 6 dsp_local_on # 
=====  END  =====
===== START =====
TEST: exit
DESCRIPTION: Verify we can cleanly exit a scheduler in multiple places
OUTPUT:
ok 7 exit # 
=====  END  =====
===== START =====
TEST: hotplug
DESCRIPTION: Verify hotplug behavior
OUTPUT:
EXIT: unregistered from BPF (hotplug event detected (1 going online))
EXIT: unregistered from BPF (hotplug event detected (1 going offline))
EXIT: unregistered from the main kernel (cpu 1 going online, exiting scheduler)
EXIT: unregistered from the main kernel (cpu 1 going offline, exiting scheduler)
EXIT: unregistered from the main kernel (expected hotplug seq 35 did not match actual 36)
ok 8 hotplug # 
=====  END  =====
===== START =====
TEST: init_enable_count
DESCRIPTION: Verify we do the correct amount of counting of init, enable, etc callbacks.
OUTPUT:
ok 9 init_enable_count # 
=====  END  =====
===== START =====
TEST: maximal
DESCRIPTION: Verify we can load a scheduler with every callback defined
OUTPUT:
ok 10 maximal # 
=====  END  =====
===== START =====
TEST: maybe_null
DESCRIPTION: Verify if PTR_MAYBE_NULL work for .dispatch
OUTPUT:
libbpf: prog 'maybe_null_fail_dispatch': BPF program load failed: Permission denied
libbpf: prog 'maybe_null_fail_dispatch': -- BEGIN PROG LOAD LOG --
reg type unsupported for arg#0 function maybe_null_fail_dispatch#5
0: R1=ctx(off=0,imm=0) R10=fp0
; void BPF_STRUCT_OPS(maybe_null_fail_dispatch, s32 cpu, struct task_struct *p)
0: (79) r1 = *(u64 *)(r1 +8)          ; R1_w=trusted_ptr_or_null_task_struct(id=1,off=0,imm=0)
; vtime_test = p->scx.dsq_vtime;
1: (79) r1 = *(u64 *)(r1 +1016)
R1 invalid mem access 'trusted_ptr_or_null_'
processed 2 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0
-- END PROG LOAD LOG --
libbpf: prog 'maybe_null_fail_dispatch': failed to load: -13
libbpf: failed to load object 'maybe_null_fail_dsp'
libbpf: failed to load BPF skeleton 'maybe_null_fail_dsp': -13
libbpf: prog 'maybe_null_fail_yield': BPF program load failed: Permission denied
libbpf: prog 'maybe_null_fail_yield': -- BEGIN PROG LOAD LOG --
reg type unsupported for arg#0 function maybe_null_fail_yield#8
0: R1=ctx(off=0,imm=0) R10=fp0
; bool BPF_STRUCT_OPS(maybe_null_fail_yield, struct task_struct *from,
0: (b7) r2 = 3344                     ; R2_w=3344
; bool BPF_STRUCT_OPS(maybe_null_fail_yield, struct task_struct *from,
1: (79) r1 = *(u64 *)(r1 +8)          ; R1_w=trusted_ptr_or_null_task_struct(id=1,off=0,imm=0)
2: (bf) r3 = r1                       ; R1_w=trusted_ptr_or_null_task_struct(id=1,off=0,imm=0) R3_w=trusted_ptr_or_null_task_struct(id=1,off=0,imm=0)
3: (0f) r3 += r2
R3 pointer arithmetic on trusted_ptr_or_null_ prohibited, null-check it first
processed 4 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0
-- END PROG LOAD LOG --
libbpf: prog 'maybe_null_fail_yield': failed to load: -13
libbpf: failed to load object 'maybe_null_fail_yld'
libbpf: failed to load BPF skeleton 'maybe_null_fail_yld': -13
ok 11 maybe_null # 
=====  END  =====
===== START =====
TEST: minimal
DESCRIPTION: Verify we can load a fully minimal scheduler
OUTPUT:
ok 12 minimal # 
=====  END  =====
===== START =====
TEST: prog_run
DESCRIPTION: Verify we can call into a scheduler with BPF_PROG_RUN, and invoke kfuncs
OUTPUT:
ok 13 prog_run # 
=====  END  =====
===== START =====
TEST: reload_loop
DESCRIPTION: Stress test loading and unloading schedulers repeatedly in a tight loop
OUTPUT:
ok 14 reload_loop # 
=====  END  =====
===== START =====
TEST: select_cpu_dfl
DESCRIPTION: Verify the default ops.select_cpu() dispatches tasks when idles cores are found, and skips ops.enqueue()
OUTPUT:
ok 15 select_cpu_dfl # 
=====  END  =====
===== START =====
TEST: select_cpu_dfl_nodispatch
DESCRIPTION: Verify behavior of scx_bpf_select_cpu_dfl() in ops.select_cpu()
OUTPUT:
ok 16 select_cpu_dfl_nodispatch # 
=====  END  =====
===== START =====
TEST: select_cpu_dispatch
DESCRIPTION: Test direct dispatching to built-in DSQs from ops.select_cpu()
OUTPUT:
ok 17 select_cpu_dispatch # 
=====  END  =====
===== START =====
TEST: select_cpu_dispatch_bad_dsq
DESCRIPTION: Verify graceful failure if we direct-dispatch to a bogus DSQ in ops.select_cpu()
OUTPUT:
ok 18 select_cpu_dispatch_bad_dsq # 
=====  END  =====
===== START =====
TEST: select_cpu_dispatch_dbl_dsp
DESCRIPTION: Verify graceful failure if we dispatch twice to a DSQ in ops.select_cpu()
OUTPUT:
ok 19 select_cpu_dispatch_dbl_dsp # 
=====  END  =====
===== START =====
TEST: select_cpu_vtime
DESCRIPTION: Test doing direct vtime-dispatching from ops.select_cpu(), to a non-built-in DSQ
OUTPUT:
ok 20 select_cpu_vtime # 
=====  END  =====
===== START =====
TEST: example
DESCRIPTION: Validate the basic function of the test suite itself
OUTPUT:
ok 21 example # 
=====  END  =====

=============================

RESULTS:

PASSED:  21
SKIPPED: 0
FAILED:  0
```
其中maybe_null测试用例告警是符合预期的。

测试用例测试结果
![输入图片说明](https://foruda.gitee.com/images/1763543032494922635/af7a727c_563961.png "屏幕截图")
![输入图片说明](https://foruda.gitee.com/images/1763543065092926190/c785655b_563961.png "屏幕截图")
![输入图片说明](https://foruda.gitee.com/images/1763543089702362261/1d120c4c_563961.png "屏幕截图")
![输入图片说明](https://foruda.gitee.com/images/1763543134081937982/f04754a4_563961.png "屏幕截图")

GVPopenEuler/kernel

搜索帮助

GVP openEuler/kernel