Re: [PATCH 2/2] clk: shmobile: Fix MSTP clock array initialization

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Valentine,

On Saturday 28 December 2013 12:35:31 Laurent Pinchart wrote:
> On Friday 27 December 2013 22:06:39 Valentine Barshak wrote:
> > The clks member of the clk_onecell_data structure should
> > point to a valid clk array (no NULL entries allowed),
> > and the clk_num should be equal to the number
> > of elements in the clks array.
> > 
> > The MSTP driver fails to satisfy the above conditions.
> > The clks array may contain NULL entries if not all
> > clock-indices are initialized in the device tree.
> > Thus, if the clock indices are interleaved we end up
> > with NULL pointers in-between.
> 
> I don't think that's an issue in practice as long as no reference to a NULL
> clock exists in the device tree, but it should of course be fixed.
> 
> > The other problem is the driver uses maximum clock index
> > as the number of clocks, which is incorrect (less than
> > the actual number of clocks by 1).
> 
> Good catch.
> 
> > Fix the first issue by pre-setting the whole clks array
> > with ERR_PTR(-ENOENT) pointers instead of zeros; and
> > use maximum clkidx + 1 as the number of clocks to fix
> > the other one.
> > 
> > This should make of_clk_src_onecell_get() return the following:
> > * valid clk pointers for all clocks registered;
> > * ERR_PTR(-EINVAL) if (idx >= clk_data->clk_num);
> > * ERR_PTR(-ENOENT) if the clock at the selected index was not
> > 
> >   initialized in the device tree (and was not registered).
> > 
> > Signed-off-by: Valentine Barshak <valentine.barshak@cogentembedded.com>
> > ---
> > 
> >  drivers/clk/shmobile/clk-mstp.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/clk/shmobile/clk-mstp.c
> > b/drivers/clk/shmobile/clk-mstp.c index be7d017..14b91ad 100644
> > --- a/drivers/clk/shmobile/clk-mstp.c
> > +++ b/drivers/clk/shmobile/clk-mstp.c
> > @@ -160,7 +160,7 @@ static void __init cpg_mstp_clocks_init(struct
> > device_node *np)
> >  	unsigned int i;
> > 
> >  	group = kzalloc(sizeof(*group), GFP_KERNEL);
> > -	clks = kzalloc(MSTP_MAX_CLOCKS * sizeof(*clks), GFP_KERNEL);
> > +	clks = kmalloc(MSTP_MAX_CLOCKS * sizeof(*clks), GFP_KERNEL);
> >  	if (group = NULL || clks = NULL) {
> >  		kfree(group);
> >  		kfree(clks);
> > @@ -182,6 +182,10 @@ static void __init cpg_mstp_clocks_init(struct
> > device_node *np) }
> > 
> >  	for (i = 0; i < MSTP_MAX_CLOCKS; ++i) {
> > +		clks[i] = ERR_PTR(-ENOENT);
> > +	}
> 
> No need for brackets here.
> 
> With this fixed,
> 
> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

Could you please resubmit the series with this fixed and the Reviewed-by line 
from Ben picked for patch 1/2, and ask Mike to apply it for v3.14 in the cover 
letter ?

> > +
> > +	for (i = 0; i < MSTP_MAX_CLOCKS; ++i) {
> >  		const char *parent_name;
> >  		const char *name;
> >  		u32 clkidx;
> > @@ -208,7 +212,8 @@ static void __init cpg_mstp_clocks_init(struct
> > device_node *np)
> >  		clks[clkidx] = cpg_mstp_clock_register(name, parent_name,
> >  						       clkidx, group);
> >  		
> >  		if (!IS_ERR(clks[clkidx])) {
> > -			group->data.clk_num = max(group->data.clk_num, clkidx);
> > +			group->data.clk_num = max(group->data.clk_num,
> > +						  clkidx + 1);
> >  			/*
> >  			 * Register a clkdev to let board code retrieve the
> >  			 * clock by name and register aliases for non-DT
-- 
Regards,

Laurent Pinchart