Re: [PATCH v2] clk: hisilicon: Improve deallocation in error path

["J. Neuschäfer" <j.ne@posteo.net>]

On Wed, Mar 04, 2026 at 10:00:15AM -0500, Brian Masney wrote:
> On Tue, Mar 03, 2026 at 04:25:18PM +0100, J. Neuschäfer via B4 Relay wrote:
> > From: "J. Neuschäfer" <j.ne@posteo.net>
> > 
> > Unmap 'base' if an error occurs after it has been mapped.
> > 
> > Reported-by: Tao Lan <taolan@huawei.com>
> > Closes: https://lore.kernel.org/lkml/ZNlSH+eWV8Sk3FYn@probook/
> > Signed-off-by: J. Neuschäfer <j.ne@posteo.net>
> 
> My understanding is that you need to use your full first name in the
> Signed-off-by.

I have been contributing to the Linux kernel and other projects using my
abbreviated first name for over a year. It has been mostly well received.

> 
> > ---
> > Changes in v2:
> > - Rebase on v7.0-rc2
> > - Update my email address
> > - Reword the commit message somewhat
> > - Link to v1: https://lore.kernel.org/r/20240708-hisi-error-v1-1-727bc7399b63@gmx.net
> > ---
> >  drivers/clk/hisilicon/clk.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/clk/hisilicon/clk.c b/drivers/clk/hisilicon/clk.c
> > index fae65127cd4aa8..08050ff1c8cf9a 100644
> > --- a/drivers/clk/hisilicon/clk.c
> > +++ b/drivers/clk/hisilicon/clk.c
> > @@ -70,7 +70,7 @@ struct hisi_clock_data *hisi_clk_init(struct device_node *np,
> >  
> >  	clk_data = kzalloc_obj(*clk_data);
> >  	if (!clk_data)
> > -		goto err;
> > +		goto err_base;
> >  
> >  	clk_data->base = base;
> >  	clk_table = kzalloc_objs(*clk_table, nr_clks);
> > @@ -83,6 +83,8 @@ struct hisi_clock_data *hisi_clk_init(struct device_node *np,
> >  	return clk_data;
> >  err_data:
> >  	kfree(clk_data);
> > +err_base:
> > +	iounmap(base);
> 
> Use devm_of_iomap() and it will be automatically cleaned up for you.

This would require a larger restructuring of the HiSilicon clock drivers
to pass a struct device * to hisi_clk_init. I'd rather avoid that
because I can't currently test these drivers and because the benefit is
questionable. However, I still had the upstreaming of this patch in my
backlog, and I think it's simple enough to review as is.

For instance, in drivers/clk/hisilicon/clk-hi3620.c:

static void __init hi3620_clk_init(struct device_node *np)
{
	struct hisi_clock_data *clk_data;

	clk_data = hisi_clk_init(np, HI3620_NR_CLKS);
	...
}
CLK_OF_DECLARE(hi3620_clk, "hisilicon,hi3620-clock", hi3620_clk_init);

Init functions registered with CLK_OF_DECLARE don't have access to a
struct device *.


Best regards,
J. Neuschäfer