Re: TPM operation times out (very rarely)

[Jonathan McDowell <noodles@earth.li>]

On Sat, Feb 08, 2025 at 10:29:55PM +0200, Jarkko Sakkinen wrote:
> On Fri, Feb 07, 2025 at 09:26:16AM +0000, Jonathan McDowell wrote:
> > > The whole arithmetic with timeout_a/b/c is mostly gibberish and could
> > > be replaced with a single "max" constant without issues (just set it
> > > large enough).
> > > 
> > > They could be all be replaced with let's say 3s timeout in a constant.
> > 
> > This appears to have come up before:
> > 
> > https://lore.kernel.org/linux-integrity/358e89ed2b766d51b5f57abf31ab7a925ac63379.1552348123.git.calvinowens@fb.com/
> > 
> > That patch was deemed overly complex and it was suggested to split it
> > up; I can't find any indication that was ever done which I guess is why
> > the discussion died off.
> 
> Looking back I suggest splitting timeouts and durations into separate
> patches:
> 
> https://lore.kernel.org/linux-integrity/20190312145553.GB6682@linux.intel.com/

Yup, that seems to be where it all died out from what I could find.

> > So just to clarify, this more recent patch is working around a situation
> > where the status register gets stuck and needs a complete retry of the
> > command send - it's an Infineon errata, not something that would be
> > fixed with a longer timeout.
> 
> Hmm... please shout if I ignore something but if we could -ERESTARTSYS
> semantics here that should ignite completely new transmit flow, wouldn't
> it?

Who then handles the ERESTARTSYS though? Part of the issues we've seen
is the failure happens in a context save or load, which is all within
the kernel rather than directly under the control of userspace. I'm
guessing the HMAC changes are likely to hit similar problems. I think
some level of timeout improvement in tpm_transmit is appropriate, if we
can work out what it should be.

Likewise for the Infineon errata I think an internal retry makes sense,
because it could happen for a number of internally generated kernel
reasons as well as a userspace sent command.

J.

-- 
How I wish, how I wish you were here.