On 13/09/22(Tue) 13:17, Stefan Sperling wrote:
> On Tue, Sep 13, 2022 at 12:34:49PM +0200, Stefan Sperling wrote:
> > The attached profile trace graph suggests a bottleneck is in the open()
> > call in get_file_status(). We might be able to work around that somehow.
> 
> This patch avoids the open() call in favour of lstat(); However, this does
> not improve performance. We spend 37% runtime in either open() or lstat(),
> so this patch makes no difference.

Maybe you could measure the time spent in the syscall with btrace(8) or
ktrace(1). This might confirm the time is spent in kernel.