Am 2022-04-07 16:12, schrieb Vladimir Oltean:
> On Thu, Apr 07, 2022 at 04:04:20PM +0200, Michael Walle wrote:
>> Am 2022-04-07 15:56, schrieb Vladimir Oltean:
>> > On Thu, Apr 07, 2022 at 03:06:25PM +0200, Michael Walle wrote:
>> > > Due to missing prerequisites the probe of the felix switch might be
>> > > deferred:
>> > > [ 4.435305] mscc_felix 0000:00:00.5: Failed to register DSA
>> > > switch: -517
>> > >
>> > > It's not an error. Use dev_err_probe() to demote the error to a debug
>> > > message. While at it, replace all the dev_err()'s in the probe with
>> > > dev_err_probe().
>> > >
>> > > Signed-off-by: Michael Walle <[email protected]>
>> > > ---
>> >
>> > Please limit the dev_err_probe() to dsa_register_switch(). The resource
>> > that is missing is the DSA master, see of_find_net_device_by_node().
>> > The others cannot possibly return -EPROBE_DEFER.
>>
>> This was my rationale (from the function doc):
>>
>> * Note that it is deemed acceptable to use this function for error
>> * prints during probe even if the @err is known to never be
>> -EPROBE_DEFER.
>> * The benefit compared to a normal dev_err() is the standardized
>> format
>> * of the error code and the fact that the error code is returned.
>>
>> In any case I don't have a strong opinion.
>
> Take this case:
>
> err = -ENOMEM;
> - dev_err(&pdev->dev, "Failed to allocate driver memory\n");
> + dev_err_probe(&pdev->dev, err, "Failed to allocate driver
> memory\n");
>
> (1) there is no need to print ENOMEM if we say "failed to allocate
> memory"
That was actually a brain fart of mine ;) But AFAIK there is no need in
reporting ENOMEM errors anyway.
-michael
> (2) we don't use the return value of dev_err_probe() anyway, we have
> actual teardown to do (pci_disable_device).
> (3) we _surely_ know that -ENOMEM != -EPROBE_DEFER
>>
>> > >
>> > > Should this be a patch with a Fixes tag?
>> >
>> > Whichever way you wish, no preference.
>>
>> I'll limit it to just the one dev_err() and add a Fixes,
>> there might be scripts out there who greps dmesg for errors.
>
> Ok.