Re: [PATCH V4 net-bugfixs] net/ethernet: Update ret when ptp_clock is ERROR

[Grygorii Strashko <grygorii.strashko@ti.com>]

On 12/11/2020 10:25, Arnd Bergmann wrote:
> On Thu, Nov 12, 2020 at 2:48 AM 王擎 <wangqing@vivo.com> wrote:
>>>> On Wed, Nov 11, 2020 at 03:24:33PM +0200, Grygorii Strashko wrote:
>>>
>>> I don't think v1 builds cleanly folks (not 100% sure, cpts is not
>>> compiled on x86):
>>>
>>>                ret = cpts->ptp_clock ? cpts->ptp_clock : (-ENODEV);
>>>
>>> ptp_clock is a pointer, ret is an integer, right?
>>
>> yeah, I will modify like: ret = cpts->ptp_clock ? PTR_ERR(cpts->ptp_clock) : -ENODEV;
> 
> This is not really getting any better. If Richard is worried about
> Kconfig getting changed here, I would suggest handling the
> case of PTP being disabled by returning an error early on in the
> function, like
> 
> struct am65_cpts *am65_cpts_create(struct device *dev, void __iomem *regs,
>                                     struct device_node *node)
> {
>          struct am65_cpts *cpts;
>          int ret, i;
> 
>          if (!IS_ENABLED(CONFIG_PTP_1588_CLOCK))
>                   return -ENODEV;
> 
> Then you can replace the broken IS_ERR_OR_NULL() path with
> a simpler IS_ERR() case and keep the rest of the function readable.

There is proper blocker in am65-cpts.h #if IS_ENABLED(CONFIG_TI_K3_AM65_CPTS)
and in Makefile and proper dependency in Kconfig.

config TI_K3_AM65_CPTS
	tristate "TI K3 AM65x CPTS"
	depends on ARCH_K3 && OF
	depends on PTP_1588_CLOCK

But, as Richard mentioned [1], ptp_clock_register() may return NULL even if PTP_1588_CLOCK=y
(which I can't confirm neither deny - from the fast look at ptp_clock_register()
  code it seems should not return NULL)

> 
>>> Grygorii, would you mind sending a correct patch in so Wang Qing can
>>> see how it's done? I've been asking for a fixes tag multiple times
>>> already :(
>>
>> I still don't quite understand what a fixes tag means，
>> can you tell me how to do this, thanks.
> 
> This identifies which patch introduced the problem you are fixing
> originally. If you add an alias in your ~/.gitconfig such as
> 
> [alias]
>          fixes = show --format='Fixes: %h (\"%s\")' -s
> 
> then running
> 
> $ git fixes f6bd59526c
> produces this line:
> 
> Fixes: f6bd59526ca5 ("net: ethernet: ti: introduce am654 common
> platform time sync driver")

correct

> 
> which you can add to the changelog, just above the Signed-off-by
> lines.



[1] https://lore.kernel.org/patchwork/patch/1334067/#1529232
-- 
Best regards,
grygorii