Re: [PATCH 1/4] TPM: update char dev BKL pushdown

[Rajiv Andrade <srajiv@linux.vnet.ibm.com>]

On Tue, 2008-09-23 at 14:58 -0600, Jonathan Corbet wrote:
> On Tue, 23 Sep 2008 14:19:26 -0300
> Rajiv Andrade <srajiv@linux.vnet.ibm.com> wrote:
> 
> > + * It's assured that the chip will be opened just once,
> > + * by the check of is_open variable, which is protected
> > + * by driver_lock.
> 
> Taking a look at the code, I'm convinced.  BKL removal seems
> appropriate.
> 
> While I was in the neighborhood, though, something caught my eye:
> 
> int tpm_release(struct inode *inode, struct file *file)
> {
> 	struct tpm_chip *chip = file->private_data;
> 
> 	flush_scheduled_work();
> 
> Here you have waited until you've got nothing in the workqueue.
> 
> 	spin_lock(&driver_lock);
> 	file->private_data = NULL;
> 	del_singleshot_timer_sync(&chip->user_read_timer);
> 
> But, until you get here, your timer could have resubmitted a job into
> the workqueue - job which could run after you've freed "chip" and
> forgotten all about it. I think you need either a "don't resubmit" flag,
> or you need to delete the timer first.
> 
> jon

Yes, like in tpm_read(), the timer must be deleted before
flush_scheduled_work(). Since it's a fix to a new issue, I'm going to
submit a another patch.

Thanks,
Rajiv Andrade
IBM Linux Technology Center
Security Development