Re: [PATCH v6] tpm: fix reference counting for struct tpm_chip

From: Jarkko Sakkinen
Date: Sun Feb 21 2021 - 08:27:00 EST


On Sun, Feb 21, 2021 at 11:19:28AM +0100, Lino Sanfilippo wrote:
>
> Hi,
>
> On 19.02.21 at 10:13, Jarkko Sakkinen wrote:
>
> >> + rc = cdev_device_add(&chip->cdevs, &chip->devs);
> >> + if (rc) {
> >> + dev_err(&chip->devs,
> >> + "unable to cdev_device_add() %s, major %d, minor %d, err=%d\n",
> >> + dev_name(&chip->devs), MAJOR(chip->devs.devt),
> >> + MINOR(chip->devs.devt), rc);
> >> + goto out_put_devs;
> >> + }
> >> +
> >> + return 0;
> >> +
> >> +out_put_devs:
> >
> > A nit:
> >
> > 1. You have already del_cdev:
> > 2. Here you use a differing convention with out prefix.
> >
> > I'd suggest that you put err_ to both:
> >
> > 1. err_del_cdev
> > 2. err_put_devs
> >
> > It's quite coherent what we have already:
> >
> > linux-tpmdd on  next took 8s
> > ❯ git grep "^err_.*" drivers/char/tpm/ | wc -l
> > 17
> >
>
>
> The label del_cdev is indeed a bit inconsistent with the rest of the code.
> But AFAICS out_put_devs is not:
> 1. all labels in tpm2-space.c start with out_
> 2. there are more hits for out_ across the whole TPM code (i.e. with the same command
> you used above I get 31 hits for _out) than for err_.
>
> I suggest to rename del_cdev to something like out_del_cdev or maybe out_cdev which
> seems to be even closer to the existing naming scheme for labels.

OK, that makes sense, thanks!

> Regards,
> Lino

/Jarkko