Re: [PATCH v2 1/3] interconnect: debugfs: replace writable string helper

On Thu, Aug 06, 2026 at 11:01:07AM +0200, Greg KH wrote:
> On Thu, Aug 06, 2026 at 04:48:52PM +0800, Yichong Chen wrote:
> > +	mutex_lock(&debugfs_lock);
> > +	copy = kstrdup(*node ?: "", GFP_KERNEL);
> > +	mutex_unlock(&debugfs_lock);
> 
> scoped guard?

Yes, scoped guard would be cleaner here.  I can use it in the next version.

> > -	src = kstrdup(src, GFP_ATOMIC);
> > -	dst = kstrdup(dst, GFP_ATOMIC);
> > -	rcu_read_unlock();
> > +	src = kstrdup(src_node, GFP_KERNEL);
> > +	dst = kstrdup(dst_node, GFP_KERNEL);
> 
> Why is GFP_KERNEL now ok, while GFP_ATOMIC wasn't?  Is this the rcu
> stuff interacting somehow?

Yes.  The old code duplicated the strings while still inside the RCU
read-side critical section, so it had to use GFP_ATOMIC.

After this change, src_node and dst_node are protected by debugfs_lock
instead of RCU.  The duplication is done while holding that mutex, so the
allocation can sleep and GFP_KERNEL should be OK.

I will make this clearer in the changelog when sending the next version.

Thanks,
Yichong