Jason Gunthorpe <[email protected]> wrote on 02/16/2016 
11:47:50 PM:


> 
> On Tue, Feb 16, 2016 at 02:26:56PM -0500, Stefan Berger wrote:
> > The re-ordering of the tpm_chip and vtpm_dev release actually solved 
that
> > problem.
> 
> ? You mean you don't see the lock up because there is no error unwind?
> 
> > [My feeling is, something outside of the vtpm driver doesn't unroll 
properly
> > and locks up if there's lots of concurrency].
> 
> Hum, I'd be surprised if this is not caused by something outside 
tpm/vtpm..
> It would be very nice to know, but perhaps not essential. I didn't see
> anything obvious in vtpm either.
> 
> > > One other thing you can look at is ditching the vtpm struct
> > > device. With the last patches I sent the tpm core would only need 
two
> > > more lines to work with a null parent device, which would be even
> > > simpler for vtpm.
> > 
> > Not having a parent device isn't really necessary, except for not 
having it
> > appear in sysfs maybe?
> 
> Dropping the dummy parent out of vtpm makes it work a little more more
> like the rest of the virtual stuff and slightly changes the sysfs
> paths visble to user space. If this is the right way to go it would be
> great to do that at the start. It wasn't worth bringing up before, but
> now that I wrote the parent removal patch it is pretty trivial.

Beside the code to lock the module with the parent device, the 
chip->dev.parent is used in sysfs API calls. Using the chip->dev.kobj 
there in case chip->dev.parent is NULL also seems a way to hang the 
system.

   Stefan



------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
_______________________________________________
tpmdd-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/tpmdd-devel

Reply via email to