Re: [PATCH] tpm: Revert "tpm: fix invalid locking in NONBLOCKING mode"

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2020-05-27 at 18:10 -0700, Alex Guzman wrote:
> On Wed, 2020-05-27 at 23:15 +0300, Jarkko Sakkinen wrote:
> > On Tue, 2020-05-26 at 19:23 +0000, Mario.Limonciello@dell.com
> > wrote:
> > > Thanks, I don't pretend to understand the nuances of this
> > > particular code, but I was hoping that the request to revert got
> > > some attention since Alex's kernel Bugzilla and message a few
> > > months ago to linux integrity weren't.
> > 
> > Removing linux-kernel from CC since this subsystem internal
> > discussion.
> > 
> > Seeing the whole thing first time today.
> > 
> > Bugzilla is the first thing to ignore when busy. It is good as
> > place holder for bugs, but all discussions should happen only in
> > LKML. There's no official requirement to proactively use Bugzilla
> > for anything.
> > 
> > That said I'm happy that people put stuff there so that it gets
> > logged.
> > 
> > For follow-up's use only LKML if it is important to you. Those will
> > get processed.
> > 
> > As far as this goes, if nothing is heard from me, check that you
> > put me as CC to the original email. Otherwise, I might have missed
> > it (by mistake, not by purpose).
> > 
> > Honestly, I'm not sure what point was this patch when there was
> > time to wait for months without response. Why the passivity for all
> > this time?
> > 
> > /Jarkko
> > 
> 
> It largely went quiet because I didn't raise the issue in the mailing
> list again. I pinged back in February (
> https://lore.kernel.org/linux-integrity/CAJ7-
> PMbujee92N1f9xVF8vtXgS49qpe7qHkeWh1Z0R-Rk-Jkaw@mail.gmail.com/) but
> the conversation died out and I was content to simply use the
> last working kernel version and see if the bug was resolved on its
> own.

I think its just a state of knowledge problem: back in February I
didn't know how unusual EPERM errors are in the TPM so the issue just
flew by as a "this is a curious issue with an O_NONBLOCK path" thing,
but thanks to some key stuff I've been doing I now do.  So this time
your EPERM struck me as "that's impossible surely" which is why I dug
into the code to find out where it was coming from ... and sure enough,
it was impossible: it was an untranslated failure return, but at least
it accidentally told me exactly what the real error was.

So the upshot is you got lucky this time around ...

James

> I raised the issue again on the bugtracker a few days ago, leading to
> this follow up here. :)