Re: [PATCH] sched_ext: Fix lockdep warnings in the ops_dump_init()

["Zqiang" <qiang.zhang@linux.dev>]

> On 12/8/25 11:23, Zqiang wrote:  
> 
> > For builts with PREEMPT\_RT=y kernel, the dump\_lock convert to  
> > sleepable rt-spinlock and not disable-irq, the following lockdep  
> > warning will appear:  
> >   
> > WARNING: CPU: 3 PID: 47 at kernel/sched/ext.c:4130 ops\_dump\_init+0x60/0xa0  
> > Modules linked in:  
> > CPU: 3 UID: 0 PID: 47 Comm: irq\_work/3  
> > RIP: 0010:ops\_dump\_init+0x60/0xa0  
> > Call Trace:  
> > <TASK>  
> > scx\_dump\_state+0x1091/0x1620  
> > ? \_\_schedule+0xf1c/0x3ec0  
> > ? rcu\_is\_watching+0x19/0xb0  
> > ? \_\_pfx\_scx\_dump\_state+0x10/0x10  
> > ? kvm\_sched\_clock\_read+0x15/0x30  
> > ? \_\_lock\_acquire+0x98b/0x1ae0  
> > ? \_\_this\_cpu\_preempt\_check+0x17/0x20  
> > ? insn\_get\_modrm+0x107/0x6e0  
> > scx\_error\_irq\_workfn+0xeb/0x120  
> > irq\_work\_single+0x113/0x260  
> > irq\_work\_run\_list.part.7+0x44/0x70  
> > run\_irq\_workd+0x6b/0x90  
> > ? \_\_pfx\_run\_irq\_workd+0x10/0x10  
> > smpboot\_thread\_fn+0x34c/0xa50  
> > ? trace\_preempt\_on+0x54/0x120  
> > ? \_\_pfx\_smpboot\_thread\_fn+0x10/0x10  
> > kthread+0x411/0x8a0  
> > ? \_\_pfx\_kthread+0x10/0x10  
> > ? rt\_spin\_unlock+0x9f/0x210  
> > ? \_\_pfx\_kthread+0x10/0x10  
> > ret\_from\_fork+0x406/0x500  
> > ? \_\_switch\_to\_asm+0x33/0x70  
> > ? \_\_pfx\_kthread+0x10/0x10  
> > ret\_from\_fork\_asm+0x1a/0x30  
> > </TASK>  
> >   
> > This commit therefore convert dump\_lock to raw spinlock.  
> >   
> > Signed-off-by: Zqiang <qiang.zhang@linux.dev>  
> > \---  
> > kernel/sched/ext.c | 6 +++---  
> > 1 file changed, 3 insertions(+), 3 deletions(-)  
> >   
> > diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c  
> > index 59fa391636c7..2e19e3689f1c 100644  
> > \--- a/kernel/sched/ext.c  
> > +++ b/kernel/sched/ext.c  
> > @@ -4235,7 +4235,7 @@ static void scx\_dump\_task(struct seq\_buf \*s, struct scx\_dump\_ctx \*dctx,  
> >   
> > static void scx\_dump\_state(struct scx\_exit\_info \*ei, size\_t dump\_len)  
> > {  
> > \- static DEFINE\_SPINLOCK(dump\_lock);  
> > \+ static DEFINE\_RAW\_SPINLOCK(dump\_lock);  
> > static const char trunc\_marker\[\] = "\\n\\n~~~~ TRUNCATED ~~~~\\n";  
> > struct scx\_sched \*sch = scx\_root;  
> > struct scx\_dump\_ctx dctx = {  
> > @@ -4251,7 +4251,7 @@ static void scx\_dump\_state(struct scx\_exit\_info \*ei, size\_t dump\_len)  
> > char \*buf;  
> > int cpu;  
> >   
> > \- spin\_lock\_irqsave(&dump\_lock, flags);  
> > \+ raw\_spin\_lock\_irqsave(&dump\_lock, flags);  
> >   
> > seq\_buf\_init(&s, ei->dump, dump\_len);  
> >   
> > @@ -4375,7 +4375,7 @@ static void scx\_dump\_state(struct scx\_exit\_info \*ei, size\_t dump\_len)  
> > memcpy(ei->dump + dump\_len - sizeof(trunc\_marker),  
> > trunc\_marker, sizeof(trunc\_marker));  
> >   
> > \- spin\_unlock\_irqrestore(&dump\_lock, flags);  
> > \+ raw\_spin\_unlock\_irqrestore(&dump\_lock, flags);  
> > }  
> >   
> > static void scx\_error\_irq\_workfn(struct irq\_work \*irq\_work)
> 
>   
> This is already part of Tejun's cgroup sub-scheduler support series:  
> [https://lore.kernel.org/lkml/20250920005931.2753828-39-tj@kernel.org/](https://lore.kernel.org/lkml/20250920005931.2753828-39-tj@kernel.org/)


Thank you for pointing it :),  please this patch.

Thanks
Zqiang