On Fri, Jan 13, 2017 at 04:42:30PM -0800, Andrey Pronin wrote: > On Fri, Jan 13, 2017 at 05:28:57PM -0700, Jason Gunthorpe wrote: > > On Fri, Jan 13, 2017 at 04:09:54PM -0800, Andrey Pronin wrote: > > > Resetting TPM while processing a command may lead to issues > > > on the next boot. Ensure that we don't have any ongoing > > > commands, and that no further commands can be sent to the chip > > > by unregistering the device in the shutdown handler. > > > tpm_chip_unregister() waits for the completion of an ongoing > > > command, if any, and then clears out chip->ops and unregisters > > > sysfs entities. > > > > Unregistering in a shutdown handler seems very strange, it also waits > > for userspace things, so I wonder if it could be problematic? > > > > Maybe just use > > > > down_write(&chip->ops_sem); > > chip->ops = NULL; > > up_write(&chip->ops_sem); > > > > In the shutdown handler? > > down_write(&chip->ops_sem) would still wait for completing the initiated > writes, since tpm_write() in tpm-dev.c calls tpm_try_get_ops().
Yes, but that is a timeout limited wait. unregister waits for sysfs files to be closed which is potentially unbounded. > Yes, but it doesn't wait for sysfs > Also, tpm-sysfs.c calls chip->ops directly, so sysfs should be > unregistered first. Yes, sorry, I should have mentioned that.. Maybe that is too much to fix.. > And the last thing, this driver supports TPM 1.2, but if it was a 2.0 > chip, it'd also need to send TPM2_Shutdown(CLEAR) from its shutdown > handler (or get an unorderly shutdown and DA counter increment). I'm confused - doesn't your system reset the TPM when it reboots? Isn't that required so the firmware starts with known PCRs? Doesn't reset trump unorderly shutdown? In any event that seems like an all-chips problem not a chip specific bug fix? > All these things are handled by tpm_chip_unregister(). I thought about > creating a tpm_chip_shutdown routine that could be called from shutdown > handlers of the drivers that need it (and I'd do it for every driver, > especially in 2.0 case). But decided that it's better to reuse the > existing tpm_chip_unregister() that already does what's needed. If for some reason we need this for every driver then this is probably a better approach - but that seems very, very strange to me. Jason