Re: [PATCH v5 0/2] tpm: add support for nonblocking operation

From: Jarkko Sakkinen
Date: Fri Aug 31 2018 - 04:57:30 EST

Are you planning to send v6 soon fixing the minor things in 1/2 (typo
+ change for the commit message)?


On Mon, Aug 13, 2018 at 01:32:48PM -0700, Tadeusz Struk wrote:
> The TCG SAPI specification [1] defines a set of functions, which allow
> applications to use the TPM device in either blocking or non-blocking fashion.
> Each command defined by the specification has a corresponding
> Tss2_Sys_<COMMAND>_Prepare() and Tss2_Sys_<COMMAND>_Complete() call, which
> together with Tss2_Sys_ExecuteAsync() is designed to allow asynchronous
> mode of operation. Currently the TPM driver supports only blocking calls,
> which doesn't allow asynchronous IO operations.
> This patch changes it and adds support for nonblocking write and a new poll
> function to enable applications, which want to take advantage of this feature.
> The new functionality can be tested using standard TPM tools implemented
> in [2], together with modified TCTI from [3], and an example application
> by Philip Tricca [4]. Here is a short description from Philip:
> "The example application `glib-tss2-event` uses a glib main event loop
> to create an RSA 2048 primary key in the TPM2 NULL hierarchy while
> using a glib timer event to time the operation. A GSource object is
> used to generate an event when the FD underlying the tss2 function
> call has data ready. While the application waits for an event indicating
> that the CreatePrimary operation is complete, it counts timer events
> that occur every 100ms. Once the CreatePrimary operation completes the
> number of timer events that occurred is used to make a rough calculation
> of the elapsed time. This value is then printed to the console.
> This takes ~300 lines of C code and requires no management or
> synchronization of threads. The glib GMainContext is "just a poll()
> loop" according to the glib documentation here:
> and so supporting 'poll' is the easiest way to integrate with glib /
> gtk+. This is true of any other event system that relies on 'poll'
> instead of worker threads."
> [1]
> [2]
> [3]
> [4]
> ---
> Changes in v5:
> - Changed the workqueue allocation time back from the first user interface
> open to module init.
> Changes in v4:
> - Changed the way buffer_mutex is handled in nonblocking mode so that
> it is not held when write() returns to user space.
> Changes in v3:
> - Fixed problem reported by 0-dey kbuild test robot around __exitcall.
> It complained because there is a module_exit() in another file already.
> - Added info on example application from Philip
> Changes in v2:
> - Split the change into two separate patches. First patch adds a pointer
> to the space to the struct file_priv to have access to it from the async job.
> This is to avoid memory allocations on every write call. Now everything
> what's needed is in the file_priv struct.
> - Renamed the 'work' member of the timer to avoid confusion.
> Now there are 'timeout_work' and 'async_work'.
> - Removed the global wait queue and moved it to file_priv.
> - Only creating the work queue when the first file is opened.
> Tadeusz Struk (2):
> tpm: add ptr to the tpm_space struct to file_priv
> tpm: add support for nonblocking operation
> drivers/char/tpm/tpm-dev-common.c | 150 +++++++++++++++++++++++++++----------
> drivers/char/tpm/tpm-dev.c | 22 +++--
> drivers/char/tpm/tpm-dev.h | 19 +++--
> drivers/char/tpm/tpm-interface.c | 1
> drivers/char/tpm/tpm.h | 1
> drivers/char/tpm/tpmrm-dev.c | 31 ++++----
> 6 files changed, 152 insertions(+), 72 deletions(-)
> --
> TS