Re: [PATCH RESEND v5] tpm: fix reference counting for struct tpm_chip

[Jarkko Sakkinen <jarkko@kernel.org>]

On Thu, Feb 18, 2021 at 08:13:57PM +0100, Lino Sanfilippo wrote:
> 
> Hi,
> 
> On 17.02.21 at 23:18, Jarkko Sakkinen wrote:
> 
> >> +
> >
> > /*
> >  * Please describe what the heck the function does. No need for full on
> >  * kdoc.
> >  */
> 
> Ok.
> 
> >> +int tpm2_add_device(struct tpm_chip *chip)
> >
> > Please, rename as tpm_devs_add for coherency sake.
> >
> 
> Sorry I confused this and renamed it wrongly. I will fix it in the next
> patch version.
> 
> >> +{
> >> +	int rc;
> >> +
> >> +	device_initialize(&chip->devs);
> >> +	chip->devs.parent = chip->dev.parent;
> >> +	chip->devs.class = tpmrm_class;
> >
> > Empty line. Cannot recall, if I stated before.
> >> +	/* +	 * get extra reference on main device to hold on behalf of devs.
> >> +	 * This holds the chip structure while cdevs is in use. The
> >> +	 * corresponding put is in the tpm_devs_release.
> >> +	 */
> >> +	get_device(&chip->dev);
> >> +	chip->devs.release = tpm_devs_release;
> >> +	chip->devs.devt = MKDEV(MAJOR(tpm_devt),
> >> +					chip->dev_num + TPM_NUM_DEVICES);
> >
> > NAK, same comment as before.
> >
> 
> Thx for the review, I will fix these issues.

Yeah, I mean I'm going to collect this fix anyway after rc1 has been
released so there's a lot of time to polish small details. I.e. I'm
doing a PR for rc2 with the fix included.

> Regards,
> Lino

/Jarkko