On Mon, Jul 06, 2020 at 05:43:35PM +0300, Jarkko Sakkinen wrote:
> On Mon, Jul 06, 2020 at 05:00:51PM +0300, Jarkko Sakkinen wrote:
> > On Mon, Jul 06, 2020 at 11:46:46AM +0000, peter.hu...@infineon.com wrote:
> > > Hi,
> > > NACK
> > >
> > > > % git --no-pager grep IFX0102 drivers/char/tpm
> > > > drivers/char/tpm/tpm_infineon.c: {"IFX0102", 0},
> > > > drivers/char/tpm/tpm_tis.c: {"IFX0102", 0}, /* Infineon */
> > > > Obviously IFX0102 was added to the HID table for the TCG TIS driver by
> > > > mistake.
> > >
> > > The HID IFX0102 was NOT added by mistake.
> > > Let me explain the history a bit:
> > >
> > > Old SLB 9635 / 9630 TPMs had two ways to interface them
> > > - proprietary 'io' mapped protocol (tpm_infineon) - tis protocol
> > > (tpm_tis)
> > >
> > > Both match the same HID.
> > > However with the emerging of the tis protocol, the io protocol eventually
> > > went away for newer products.
> > > So all TPM1.2 by IFX match the HID0102 and the TCG generic ones PNP0C31
> > >
> > > So basically you break TPM1.2 support for all (newer) Infineon chips if
> > > the platform vendor used the IFX0102 HID as they would speak via
> > > tpm_infineon driver.
> > > The bug must be something different, especially as it only seems to
> > > happen after suspend resume.
> >
> > Peter,
> >
> > Looking at dmesg:
> >
> > 1. tmp_infineon initializes cleanly
> > 2. tpm_tis fails misserably with bunch error messages
> >
> > I'm cool with reverting the patch though. Please send a revert patch and
> > explain this in the commit message because right now what you are saying
> > is completely undocumented.
> >
> > Also, this tpm_infineon issue needs to be fixed properly after the
> > revert.
> >
> > The bugzilla bug is unrelated to this issue but it causes extra harm
> > fixing any bugs and confusion among the users as the bug discussions
> > proves.
> >
> > How do we get the quirks for tpm_tis and tpm_infineon so that they can
> > separate each other?
>
> Also in the revert commit, please add a comment to tpm_tis.c
> about the existing conflict, e.g.
>
> /*
> * Legacy Infineon devices can emit illegit warnings as tpm_tis and
> * tpm_infineon have a conflicting device ID IFX0102.
> */
>
> I'm cool reverting it as long as I get a patch with the required
> premises to do so and proper documentation, because the issue is
> still real.
We do have this kind of thing for MSFT0101 in tpm_tis.c: check_acpi_tpm2().
This is not the root cause for the bugzilla bug, but is a bug itself,
and this kind of behaviour should not exist. It makes fixing real bugs
factors harder when you have drivers putting arbirtrary warnings to
klog.
/Jarkko
