Re: usage of rcu_dereference_raw

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, Apr 21, 2014 at 05:37:01PM -0400, Pranith Kumar wrote:
> Hi Paul,
> 
> I was trying to see the various uses of rcu_dereference_ functions to
> understand how they were being used. A doubt cropped up while doing this:
> 
> * rcu_dereference_raw(): the documentation explicitly mentions that this
>   should be minimally used as this does no checking of read critical sections
>   and does not implement barriers. But looking at the code, there are various
>   places where this is being used and it is being used I think in a buggy
>   way. For example, in drivers/net/wireless/iwlwifi/dvm/main.c, there is this:
> 
>         kfree(rcu_dereference_raw(priv->noa_data));
> 
>   I can't imagine a scenario in which this is valid. So my question is this:
>   do most of the uses of rcu_dereference_raw() need to be changed to use other
>   dereference functions or are there cases where its usage is valid?

Well, the first call to it from iwl_op_mode_dvm_start() is valid because
that field has never been exposed to readers, so that no other task or
CPU has access to this field.

The second call from iwl_op_mode_dvm_stop() -might- be valid.  For it
to be valid, there must be a grace period between the time that the
field was made inaccessible to readers and the time that iwl_uninit_drv()
was called.  Usually something like synchronize_rcu() waits for the
needed grace period.

Please let me know what you find!

							Thanx, Paul