Re: [PATCH] tools/sched_ext: scx_central: start timer from central dispatch

[Zhao Mengmeng]

On 3/27/26 08:58, Tejun Heo wrote:
> Hello, Zhao.
> 
> On Thu, Mar 26, 2026 at 05:15:23PM +0800, Zhao Mengmeng wrote:
>> From: Zhao Mengmeng <zhaomengmeng@xxxxxxxxxx>
>>
>> scx_central currently assumes that ops.init() runs on the selected
>> central CPU and aborts otherwise. This is no longer true, as ops.init()
>> is invoked from the scx_enable_helper thread, which can run on any
>> CPU. As a result, sched_setaffinity() from userspace doesn't work, loading
>> scx_central can fail with:
>>
>> [ 1985.319942] sched_ext: central: scx_central.bpf.c:314: init from non-central CPU
>> [ 1985.320317]    scx_exit+0xa3/0xd0
>> [ 1985.320535]    scx_bpf_error_bstr+0xbd/0x220
>> [ 1985.320840]    bpf_prog_3a445a8163fa8149_central_init+0x103/0x1ba
>> [ 1985.321073]    bpf__sched_ext_ops_init+0x40/0xa8
>> [ 1985.321286]    scx_root_enable_workfn+0x507/0x1650
>> [ 1985.321461]    kthread_worker_fn+0x260/0x940
>> [ 1985.321745]    kthread+0x303/0x3e0
>> [ 1985.321901]    ret_from_fork+0x589/0x7d0
>> [ 1985.322065]    ret_from_fork_asm+0x1a/0x30
>>
>> DEBUG DUMP
>> ===================================================================
>>
>> central: root
>> scx_enable_help[134] triggered exit kind 1025:
>>   scx_bpf_error (scx_central.bpf.c:314: init from non-central CPU)
>>
>> To fix it, initialize the BPF timer from ops.init(), but defer
>> bpf_timer_start() to the first dispatch on the central CPU. This preserves
>> the requirement that the timer be started from the central CPU when
>> BPF_F_TIMER_CPU_PIN is used, without depending on the CPU affinity of
>> the enable path.
> 
> Can you do the followings please?
> 
> - Remove now unnecessary sched_setaffinity() call.
> 
> - Do scx_bpf_kick_cpu(central_cpu, 0) so that we're guaranteed to enter the
>   dispatch path on the central CPU immediately.
> 
> Thanks.
> 
No Problem. Will try it.