Re: [PATCH] sdhci: Forward EPROBE_DEFER on vmmc and vqmmc regulators

[Arnd Bergmann <arnd@arndb.de>]

On Monday 07 April 2014 13:18:54 Ben Dooks wrote:
> On 07/04/14 13:16, Ben Dooks wrote:
> > On 07/04/14 13:09, Mike Looijmans wrote:
> >> On 04/07/2014 10:11 AM, Arnd Bergmann wrote:
> >>> On Monday 07 April 2014 08:38:28 Mike Looijmans wrote:
> >>>> index 34aef81..43b90c1 100644
> >>>> --- a/drivers/mmc/host/sdhci.c
> >>>> +++ b/drivers/mmc/host/sdhci.c
> >>>> @@ -2972,6 +2972,8 @@ int sdhci_add_host(struct sdhci_host *host)
> >>>>          host->vqmmc = regulator_get_optional(mmc_dev(mmc), "vqmmc");
> >>>>          if (IS_ERR_OR_NULL(host->vqmmc)) {
> >>>>                  if (PTR_ERR(host->vqmmc) < 0) {
> >>>> +                       if (PTR_ERR(host->vqmmc) == -EPROBE_DEFER)
> >>>> +                               return -EPROBE_DEFER;
> >>>>                          pr_info("%s: no vqmmc regulator found\n",
> >>>>                                  mmc_hostname(mmc));
> >>>>                          host->vqmmc = NULL;
> >>>> @@ -3048,8 +3050,10 @@ int sdhci_add_host(struct sdhci_host *host)
> >>>>          host->vmmc = regulator_get_optional(mmc_dev(mmc), "vmmc");
> >>>>          if (IS_ERR_OR_NULL(host->vmmc)) {
> >>>>                  if (PTR_ERR(host->vmmc) < 0) {
> >>>> -                       pr_info("%s: no vmmc regulator found\n",
> >>>> -                               mmc_hostname(mmc));
> >>>> +                       if (PTR_ERR(host->vmmc) == -EPROBE_DEFER)
> >>>> +                               return -EPROBE_DEFER;
> >>>> +                       pr_info("%s: no vmmc regulator found (%d)\n",
> >>>> +                               mmc_hostname(mmc),
> >>>> PTR_ERR(host->vmmc));
> >>>>                          host->vmmc = NULL;
> >>>>                  }
> >>>
> >>> Please change the code to not use IS_ERR_OR_NULL() instead, getting
> >>> a NULL return value from regulator_get_optional() should not be
> >>> considered a bug, while getting an error return should always
> >>> cause the probe function to fail.
> >
> > Surely it needs to be changed to IS_ERR(), nor IS_ERR_OR_NULL()?
> 
> 
> And by that, I mean the use of "PTR_ERR(host->vmmc) < 0" to be
> changed to IS_ERR(), which I think Arnd was originally complaining
> about.

I mean something like

	host->vqmmc = regulator_get_optional(mmc_dev(mmc), "vqmmc");

	if (IS_ERR(host->vqmmc)) {
		if (PTR_ERR(host->vqmmc) != -EPROBE_DEFER)
			complain(); /* only print a message if we won't retry */
		return ERR_PTR(host->vqmmc); /* never ignore an error */
	}
	/* silently continue if no regulator is defined */

regulator_get_optional() means we can continue if it's not there, but
we should not continue if there is a regulator that we fail to use
for some reason.

As has been found a number of times, any use of IS_ERR_OR_NULL()
means that the person responsible for the code is confused about
what the API is supposed to be.

	Arnd