Re: [PATCH v2 3/3] phy: phy-rockchip-inno-usb2: Improve error handling while probing

From: Dragan Simic
Date: Wed Aug 21 2024 - 07:03:34 EST

On 2024-08-21 11:17, Heiko Stübner wrote:
Am Mittwoch, 21. August 2024, 11:09:03 CEST schrieb Dragan Simic:
On 2024-08-21 10:44, Heiko Stübner wrote:
> Am Mittwoch, 21. August 2024, 09:37:55 CEST schrieb Dragan Simic:
>> Improve error handling in the probe path by using function
>> dev_err_probe()
>> where appropriate, and by no longer using it rather pointlessly in one
>> place
>> that actually produces a single, hardcoded error code.
>>
>> Signed-off-by: Dragan Simic <dsimic@xxxxxxxxxxx>
>
>> @@ -1375,8 +1372,10 @@ static int rockchip_usb2phy_probe(struct
>> platform_device *pdev)
>> rphy->irq = platform_get_irq_optional(pdev, 0);
>> platform_set_drvdata(pdev, rphy);
>>
>> - if (!phy_cfgs)
>> - return dev_err_probe(dev, -EINVAL, "phy configs are not
>> assigned!\n");
>> + if (!phy_cfgs) {
>> + dev_err(dev, "phy configs are not assigned\n");
>> + return -EINVAL;
>> + }
>>
>> ret = rockchip_usb2phy_extcon_register(rphy);
>> if (ret)
>
> I really don't understand the rationale here. Using dev_err_probe here
> is just fine and with that change you just introduce more lines of code
> for exactly the same functionality?

As we know, dev_err_probe() decides how to log the received error
message
based on the error code it receives, but in this case the error code is
hardcoded as -EINVAL. Thus, in this case it isn't about keeping the LoC
count a bit lower, but about using dev_err() where the resulting outcome
of error logging is aleady known, and where logging the error code
actually
isn't helpful, because it's hardcoded and the logged error message
already
tells everything about the error condition.

In other words, it's about being as precise as possible when deciding
between
dev_err() and dev_err_probe(), in both directions. I hope it makes
sense.

I'd disagree a bit, using one format only creates a way nicer pattern in the
driver, by not mixing different styles.

dev_err_probe documentation seems to agree [0], by stating:

"Using this helper in your probe function is totally fine even if @err is
known to never be -EPROBE_DEFER.
The benefit compared to a normal dev_err() is the standardized format
of the error code, it being emitted symbolically (i.e. you get "EAGAIN"
instead of "-35") and the fact that the error code is returned which allows
more compact error paths."

Yes, I saw that already in the documentation. Though, it might be debatable
does hardcoding the passed error code to some value qualifies as knowing that
it can't be -EPROBE_DEFER. The way I read that part of the documentation is
that using dev_err_probe() is fine without going into the implementation of
the previously invoked function that may fail, and researching can it actually
return -EPROBE_DEFER or not. Also, the invoked function may change at some
point in future and start returning -EPROBE_DEFER, but a hardcoded error code
that's produced locally can't become changed that way.

In addition to that, we already have at least a couple of instances [1][2] in
the same function in which dev_err() is used when the error code is hardcoded,
so there's actually already another pattern to follow.

I know that replacing dev_err_probe() with dev_err() may look strange in a
patch that mostly performs the opposite replacement, but the patch just tries
to be strict and precise, and to follow other examples of how dev_err() is
already used in the same function when the error code is produced locally
instead of being received from another invoked function.

[0] https://elixir.bootlin.com/linux/v6.10.6/source/drivers/base/core.c#L5009
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-usb2.c?h=v6.11-rc4#n1361
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-usb2.c?h=v6.11-rc4#n1369