Re: [PATCH v2 1/4] tpm: Implement tpm_chip_find() and tpm_chip_put() for other subsystems

[Jason Gunthorpe]

On Thu, Jun 21, 2018 at 01:45:03PM -0400, Stefan Berger wrote:
> On 06/21/2018 01:15 PM, Jarkko Sakkinen wrote:
> >On Wed, Jun 20, 2018 at 04:42:33PM -0400, Stefan Berger wrote:
> >>Implement tpm_chip_find() for other subsystems to find a TPM chip and
> >>get a reference to that chip. Once done with using the chip, the reference
> >>is released using tpm_chip_put().
> >>
> >>Signed-off-by: Stefan Berger <stefanb@xxxxxxxxxxxxxxxxxx>
> >You should sort this out in a way that we don't end up with duplicate
> >functions.
> 
> Do you want me to create a function *like* tpm_chip_find_get() that takes an
> additional parameter whether to get the ops semaphore and have that function
> called by the existing tpm_chip_find_get() and the new tpm_chip_find(). The
> latter would then not get the ops semphore. I didn't want to do this since
> one time the function returns with a lock held and the other time not.

Another option, and I haven't looked, is to revise the callers of
tpm_chip_find_get to not require it to hold the ops semaphore for
them.

Either by giving them an API to do it, or revising the TPM entry
points to do it.

I didn't look, but how did the ops semaphore get grabbed in your
revised patches? They do grab it, right?

Jason