Re: [PATCH] tpm.h: increase poll timings to fix tpm_tis regression

[Jason Gunthorpe <jgg@ziepe.ca>]

On Wed, Jul 11, 2018 at 12:08:53PM -0700, James Bottomley wrote:
> On Wed, 2018-07-11 at 12:21 -0600, Jason Gunthorpe wrote:
> > On Wed, Jul 11, 2018 at 10:11:14AM -0700, James Bottomley wrote:
> > > tpm_tis regressed recently to the point where the TPM being driven
> > > by
> > > it falls off the bus and cannot be contacted after some hours of
> > > use.
> > > This is the failure trace:
> > > 
> > > jejb@jarvis:~> dmesg|grep tpm
> > > [    3.282605] tpm_tis MSFT0101:00: 2.0 TPM (device-id 0xFE, rev-id 
> > > 2)
> > > [14566.626614] tpm tpm0: Operation Timed out
> > > [14566.626621] tpm tpm0: tpm2_load_context: failed with a system
> > > error -62
> > > [14568.626607] tpm tpm0: tpm_try_transmit: tpm_send: error -62
> > > [14570.626594] tpm tpm0: tpm_try_transmit: tpm_send: error -62
> > > [14570.626605] tpm tpm0: tpm2_load_context: failed with a system
> > > error -62
> > > [14572.626526] tpm tpm0: tpm_try_transmit: tpm_send: error -62
> > > [14577.710441] tpm tpm0: tpm_try_transmit: tpm_send: error -62
> > > ...
> > > 
> > > The problem is caused by a change that caused us to poke the TPM
> > > far
> > > more often to see if it's ready.  Apparently something about the
> > > bus
> > > its on and the TPM means that it crashes or falls off the bus if
> > > you
> > > poke it too often and once this happens, only a reboot will recover
> > > it.
> > 
> > I wonder if something about triggering ETIME even once breaks the
> > driver so it can't talk to the chip at all thereafter..
> > 
> > Ie it doesn't abort the command properly and becomes desynced with
> > the TIS execution protocol.
> 
> Yes, I wondered about this, but I don't understand the bus protocol
> well enough.  The tpm-interface:tpm_try_transmit() which throws the
> first ETIME says after we get that we send chip->ops->cancel() which
> tpm_tis simply translates to tpm_tis_ready() which also times out.  Is
> there a bigger hammer I can hit it with?

I don't remember off hand.. But this is, IMHO, a better guess than the
firmware crashes from reading the status register..

Perhaps reducing the timeout to force a ETIME would prove the theory?

Jason