Re: [Regression] Commit "tpm: Introduce function to poll for result of self test" breaks suspend

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Monday, January 16, 2012, Stefan Berger wrote:
> On 01/15/2012 06:22 PM, Rafael J. Wysocki wrote:
> > Hi Stefan,
> >
> Hi Rafael,
> > Unfortunately, the following commit from you:
> >
> > commit 68d6e6713fcb2ea6278661aaaf5f1c9c821b3751
> > Author: Stefan Berger<stefanb@linux.vnet.ibm.com>
> > Date:   Fri Nov 11 12:57:04 2011 -0500
> >
> >      tpm: Introduce function to poll for result of self test
> >
> >      This patch introduces a function that runs the TPM_ContinueSelfTest()
> >      function and then polls the TPM to check whether it finished the selftest
> >      and can receive new commands.
> >
> >      Signed-off-by: Stefan Berger<stefanb@linux.vnet.ibm.com>
> >
> > broke suspend on my Toshiba Portege R500.  It causes the tpm driver to
> > return an error code from .suspend() on every attempt to suspend the system.
> >
> 
> Can you show what the error output is?

I get:

[   96.103556] PM: Entering mem sleep
[   96.103606] Suspending console(s) (use no_console_suspend to debug)
[   96.107159] sd 0:0:0:0: [sda] Synchronizing SCSI cache
[   96.107548] sd 0:0:0:0: [sda] Stopping disk
[   96.108436] option: option_instat_callback: error -2
[   96.386200] tpm_inf_pnp 00:0a: saving TPM state
[   96.407630] tpm_inf_pnp 00:0a: Timeout while clearing FIFO
[   96.407634] tpm_inf_pnp 00:0a: error while saving TPM state
[   96.407646] legacy_suspend(): pnp_bus_suspend+0x0/0xa0 returns -5
[   96.407656] PM: Device 00:0a failed to suspend: error -5
[   96.407662] PM: Some devices failed to suspend

100% of the time.

> Strange enough this function is not called during a suspend but during 
> boot and again during resume. Can you try the following after a boot
> 
> cd /sys/devices
> find -name pcrs | xargs cat
> 
> and see whether this displays 24 PCRs.

That doesn't produce any output.

> This sends a couple of commands to the TPM. Then try a suspend. Did this
> help anything?

Nope.  The result is the same as before.

> If yes, can you run above commands again?

It still doesn't produce any output.

> FYI: TPM_ContinueSelfTest is the only TPM function that triggers an 
> asynchronous action of the TPM, here the continuation of its self test. 
> To prevent subsequent commands sent to the TPM from failing due to the 
> ongoing self test, we are now waiting for this command to finish. None 
> of the machines that I tested it with had a TPM that actually did 
> anything during the TPM_ContinueSelfTest() but must have fully completed 
> their selftest during TPM_SelfTest() sent to it by the BIOS.
> 
> > Revertig this commit on top of the current Linus' tree fixes the problem for me.
> >
> > If necessary, I'll collect more information about this problem tomorrow.
> 
> The output of 'caps' would be interesting in the same dir as pcrs:
> 
> cd /sys/devices
> find -name caps | xargs cat

That doesn't show anything too.

However, after reverting the above commit I get:

$ find /sys/devices -name caps | xargs cat
Manufacturer: 0x49465800
TCG version: 1.2
Firmware version: 1.2

and system suspend works, although "find /sys/devices -name pcrs | xargs cat"
still doesn't show anything.

Thanks,
Rafael