Re: [PATCH 5/5] spi: fix return value check in hspi_probe().

[Dan Carpenter]

On Fri, Jan 11, 2013 at 03:02:02PM +0000, Grant Likely wrote:
> On Thu, 3 Jan 2013 19:04:55 +0300, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> > On Sat, Dec 22, 2012 at 09:56:20AM +0000, Grant Likely wrote:
> > > On Wed, 19 Dec 2012 19:39:14 +0300, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> > > > On Wed, Dec 19, 2012 at 03:11:54PM +0000, Grant Likely wrote:
> > > > > On Tue, 11 Dec 2012 16:36:27 -0800 (PST), Kuninori Morimoto <kuninori.morimoto.gx@xxxxxxxxxxx> wrote:
> > > > > > 
> > > > > > Hi
> > > > > > 
> > > > > > > According to its documentation, clk_get() returns a "valid IS_ERR() condition
> > > > > > > containing errno", so we should call IS_ERR() rather than a NULL check.
> > > > > > > 
> > > > > > > Signed-off-by: Cyril Roelandt <tipecaml@xxxxxxxxx>
> > > > > > 
> > > > > > Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@xxxxxxxxxxx>
> > > > > 
> > > > > Applied, thanks.
> > > > 
> > > > In another thread, we were just talking about who clk_get() can
> > > > return a NULL if !CONFIG_HAVE_CLK.  That might change to match the
> > > > documentation later...  Not sure.
> > > 
> > > So what is the solution here? Will the dummy clk_get() be changed, or is
> > > there more work needed on the drivers?
> > > 
> > 
> > This driver can't function without a clk.  It should select that at
> > build time instead of testing for it at probe.  Just checking for
> > IS_ERR() will NOT cause a NULL dereference so the patch is ok in
> > that respect.  I'm not sure if this can be build without
> > CONFIG_HAVE_CLK.
> 
> Your suggestion is to make this driver depend on CONFIG_HAVE_CLK? I'm
> fine with that, but it doesn't sort out the core infrastructure question
> about whether it is valid for the empty clk_get() to return NULL. Nor is
> it a particularly sustainable solution (ie. if a large number of drivers
> want to do the same thing).
> 

We've spread this conversation out across several different threads.

http://lists.infradead.org/pipermail/linux-arm-kernel/2013-January/139186.html

Basically, drivers should use accessor functions which handle NULL
pointers instead of dereferencing the pointer directly.

regards,
dan carpenter

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/