Re: [PATCH] Add TPM hardware enablement driver

[Nish Aravamudan <nish.aravamudan@gmail.com>]

On Wed, 9 Mar 2005 16:42:01 -0800, Greg KH <greg@kroah.com> wrote:
> ChangeSet 1.2035, 2005/03/09 10:12:19-08:00, kjhall@us.ibm.com
> 
> [PATCH] Add TPM hardware enablement driver

<snip>

> +void tpm_time_expired(unsigned long ptr)
> +{
> +       int *exp = (int *) ptr;
> +       *exp = 1;
> +}
> +
> +EXPORT_SYMBOL_GPL(tpm_time_expired);

<snip>

> +       down(&chip->timer_manipulation_mutex);
> +       chip->time_expired = 0;
> +       init_timer(&chip->device_timer);
> +       chip->device_timer.function = tpm_time_expired;
> +       chip->device_timer.expires = jiffies + 2 * 60 * HZ;
> +       chip->device_timer.data = (unsigned long) &chip->time_expired;
> +       add_timer(&chip->device_timer);
> +       up(&chip->timer_manipulation_mutex);
> +
> +       do {
> +               u8 status = inb(chip->vendor->base + 1);
> +               if ((status & chip->vendor->req_complete_mask) ==
> +                   chip->vendor->req_complete_val) {
> +                       down(&chip->timer_manipulation_mutex);
> +                       del_singleshot_timer_sync(&chip->device_timer);
> +                       up(&chip->timer_manipulation_mutex);
> +                       goto out_recv;
> +               }
> +               set_current_state(TASK_UNINTERRUPTIBLE);
> +               schedule_timeout(TPM_TIMEOUT);
> +               rmb();
> +       } while (!chip->time_expired);

<snip>

It seems like this use of schedule_timeout() and the others are a bit
excessive. In this case, a timer is set to go off in 2 hours or so,
with tpm_time_expired() as the callback. tpm_time_expired(), it seems
just takes data and sets it to 1, which in this case is
chip->time_expired (and is similar in the other cases). We then loop
while (!chip->time_expired), which to me means until
chip->device_timer goes off, checking if the request is complete every
5 milliseconds. The chip->device_timer doesn't really do anything,
does it? It just guarantees a maximum time (of 2 hours). Couldn't the
same be achieved with (please excuse the lack of tabs, any real
patches I submit will have them):

unsigned long stop = jiffies + 2 * 60 * HZ;
do {
     u8 status = inb(chip->vendor->base + 1);
     if ((status & chip->vendor->req_complete_mask ==
           chip->vendor->req_complete_val)
               goto out_recv;
     msleep(TPM_TIMEOUT); // TPM_TIMEOUT could now be 5 ms
     rmb();
} while (time_before(jiffies, stop);

I think similar replacements would work in the other locations.

If people agree, I will send patches.

Thanks,
Nish