Re: [PATCH 2/2] tpm2-space: add handling for global session exhaustion

[Jarkko Sakkinen]

On Thu, Jan 26, 2017 at 04:45:52PM -0800, James Bottomley wrote:
> On Thu, 2017-01-26 at 14:56 +0200, Jarkko Sakkinen wrote:
> > On Mon, Jan 23, 2017 at 09:38:33PM -0800, James Bottomley wrote:
> > > In a TPM2, sessions can be globally exhausted once there are
> > > TPM_PT_ACTIVE_SESSION_MAX of them (even if they're all context 
> > > saved). The Strategy for handling this is to keep a global count of 
> > > all the sessions along with their creation time.  Then if we see 
> > > the TPM run out of sessions (via the TPM_RC_SESSION_HANDLES) we 
> > > first wait for one to become free, but if it doesn't, we forcibly 
> > > evict an existing one. The eviction strategy waits until the 
> > > current command is repeated to evict the session which should 
> > > guarantee there is an available slot.
> > > 
> > > On the force eviction case, we make sure that the victim session is 
> > > at least SESSION_TIMEOUT old (currently 2 seconds).  The wait queue
> > > for session slots is a FIFO one, ensuring that once we run out of
> > > sessions, everyone will get a session in a bounded time and once 
> > > they get one, they'll have SESSION_TIMEOUT to use it before it may 
> > > be subject to eviction.
> > > 
> > > Signed-off-by: James Bottomley <
> > > James.Bottomley@xxxxxxxxxxxxxxxxxxxxx>
> > 
> > This is not a proper review yet. Just quick question: why do you need
> > a real time (i.e. created)? Maybe in the force eviction case it would
> > be enough to sleep lets say 500 ms and pick the victim with smallest
> > number? I.e. just have increasing u64 counter instead of real time.
> 
> So that if the oldest session has already been around for > 2s there's
> no need to wait.  In order to guarantee everyone gets a session for at
> least 2s without tracking the age of sessions, you'd have to sleep for
> 2s after you find the oldest session.
> 
> > That would simplify this patch a lot and does not prevent to refine
> > later on if workloads show need for more complex logic.
> 
> An increasing monotonic counter would actually not be much simpler: all
> you could really cut out would be the four lines and two comments at
> the bottom of the for loop in tpm2_session_wait() which check the age
> of the found session.
> 
> James

Right. Thanks for explaining this. I'll check the code with more detail
ASAP.

/Jarkko