On Tue, May 15, 2018 at 05:30:15AM +1000, James Kelly wrote:
> Dan,
>
> > On 14 May 2018, at 11:47 pm, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> >
> > On Mon, May 07, 2018 at 11:20:29AM +1000, James Kelly wrote:
> >> +static int clk_wzrd_probe(struct platform_device *pdev)
> >> +{
> >> + int ret;
> >> + struct device *dev = &pdev->dev;
> >> +
> >> + ret = clk_wzrd_get_device_tree_data(dev);
> >> + if (ret)
> >> + return ret;
> >> +
> >> + ret = clk_wzrd_regmap_alloc(dev);
> >> + if (ret)
> >> + return ret;
> >> +
> >> + ret = clk_wzrd_register_ccf(dev);
> >> + if (ret)
> >> + return ret;
> >> +
> >> + return 0;
> >
> > The error handling is a terrible layer violation now... Every
> > allocation function should have a matching free function. But now
> > instead of that if clk_wzrd_register_ccf() fails then it starts cleaning
> > up the clk_wzrd->axi_clk that was allocated in
> > clk_wzrd_get_device_tree_data().
> >
> > regards,
> > dan carpenter
> >
> >
>
> This gets cleaned up in patch 6. There seemed little point in putting a whole lot of
> code in patch 3 only to rip it out again in patch 6. I was trying to keep the patches
> simple.
I later posted that it's actually buggy so it's not allow to introduce
a bug and then fix it.
I really quarrel with the word "simple". As a reviewer it's my job to
spot bugs being introduced. It took me a long time to review this patch
until I spotted the bug. It's better to not take short cuts, but just
write error handling in the standard way.
regards,
dan carpenter
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
