openEuler-24.03-LTS+patch-6.6.0-6.0.0-rt20.patch启动出现BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48

> _**请尽量提供详细的信息，如缺乏必要的定位信息，则缺陷不会被定位**_

## 环境信息

【OS版本】openEuler-24.03-LTS, openEuler embedded

【内核版本】6.6.0-27.0.0

【硬件平台】飞腾E2000

## 缺陷信息

【问题复现步骤】
1. openEuler embedded构建rt版本内核
2. rt-patch: https://gitee.com/src-openeuler/kernel/blob/master/patch-6.6.0-6.0.0-rt20.patch
2. 启动后出现异常日志

【实际结果】出现BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48

【期望结果】无异常日志出现

【其他相关附件信息】比如 syslog、dmesg、panic、lockup、kdump 信息、图片等

```
[ 16.139209] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48
[ 16.139225] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1420, name: auditd
[ 16.139231] preempt_count: 1, expected: 0
[ 16.139234] RCU nest depth: 0, expected: 0
[ 16.139239] CPU: 0 PID: 1420 Comm: auditd Tainted: G S W 6.6.0-openeuler #1
[ 16.139247] Hardware name: Pe2204 DEMO DDR4 (DT)
[ 16.139251] Call trace:
[ 16.139255] dump_backtrace+0xac/0xd8
[ 16.139274] show_stack+0x18/0x24
[ 16.139282] dump_stack_lvl+0x48/0x60
[ 16.139291] dump_stack+0x18/0x24
[ 16.139297] might_resched+0x128/0x144
[ 16.139305] rt_spin_lock+0x30/0x8c
[ 16.139313] pcpu_alloc+0x15c/0x738
[ 16.139323] alloc_percpu_gfp+0x18/0x24
[ 16.139332] percpu_counter_init_many+0x38/0xe8
[ 16.139343] percpu_counter_switch_to_pcpu_many+0x54/0xb4
[ 16.139351] copy_process+0xae0/0x19dc
[ 16.139358] kernel_clone+0x9c/0x3a4
[ 16.139364] se_sys_clone3+0xb8/0xe4
[ 16.139370] __arm64_sys_clone3+0x14/0x20
[ 16.139376] invoke_syscall+0x74/0x104
[ 16.139385] el0_svc_common.constprop.0+0x74/0xd8
[ 16.139394] do_el0_svc+0x1c/0x28
[ 16.139402] el0_svc+0xb4/0x114
[ 16.139408] el0t_64_sync_handler+0x8c/0x12c
[ 16.139415] el0t_64_sync+0x174/0x178
```

【已分析信息】如已经做过分析和定位，请尽量附上详细的分析结果
欧拉内核在6.6上增加了如下机制：
在单线程进程时，将mm counter默认改为atomic变量，降低内存消耗与性能开销；进程创建第二个线程时，将atomic变量改为percpu变量，降低mm counter的误差范围；
https://gitee.com/openeuler/kernel/issues/I9IA1I
所以在copy_process时，会进新的处理函数：
```c
int percpu_counter_switch_to_pcpu_many(struct percpu_counter *fbc,
                       u32 nr_counters)
{
    static struct lock_class_key __key;
    unsigned long flags;
    bool ret = 0;

if (percpu_counter_initialized(fbc))
        return 0;

preempt_disable();
    local_irq_save(flags);
    if (likely(!percpu_counter_initialized(fbc)))
        ret = __percpu_counter_init_many(fbc, 0,
                    GFP_ATOMIC|__GFP_NOWARN|__GFP_ZERO,
                    nr_counters, &__key, true);
    local_irq_restore(flags);
    preempt_enable();

return ret;
}
```
在这个函数关掉了抢占和irq, 但是在__percpu_counter_init_many()中在pcpu_alloc()会调用普通的spin_lock，于是就会报错。
在仓库中的patch-6.6.0-6.0.0-rt20.patch未对这部分改动进行处理。
已验证revert掉在https://gitee.com/openeuler/kernel/issues/I9IA1I提交的两笔commit, 6.6-rt启动后无`BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48`的异常。希望能提供更合适的解法。

src-openEuler/kernel

内容风险标识

评论 (2)

src-openEuler/kernel .gitee-modal { width: 500px !important; }

内容风险标识