Re: [PATCH] net: wireless: iwlegacy: Fix possible data races in il4965_send_rxon_assoc()

[Jia-Ju Bai]

Thanks for your reply :)


On 2018/10/4 15:59, Stanislaw Gruszka wrote:
> On Wed, Oct 03, 2018 at 10:07:45PM +0800, Jia-Ju Bai wrote:
> > These possible races are detected by a runtime testing.
> > To fix these races, the mutex lock is used in il4965_send_rxon_assoc()
> > to protect the data.
> Really ? I'm surprised by that, see below.

My runtime testing shows that il4965_send_rxon_assoc() and 
il4965_configure_filter() are concurrently executed.
But after seeing your reply, I need to carefully check whether my 
runtime testing is right, because I think you are right.
In fact, I only monitored the iwl4965 driver, but did not monitor the 
iwlegacy driver, so I will do the testing again with monitoring the 
lwlegacy driver.

> > @@ -1297,6 +1297,7 @@ il4965_send_rxon_assoc(struct il_priv *il)
> >   	const struct il_rxon_cmd *rxon1 = &il->staging;
> >   	const struct il_rxon_cmd *rxon2 = &il->active;
> >   
> > +	mutex_lock(&il->mutex);
> >   	if (rxon1->flags == rxon2->flags &&
> For 4965 driver il4965_send_rxon_assoc() is only called by
> il_mac_bss_info_changed() and il4965_commit_rxon().
>
> il_mac_bss_info_changed() acquire il->mutex and
> callers of  il4965_commit_rxon() acquire il->mutex
> (but I did not check all of them).
>
> So I wonder how this patch did not cause the deadlock ?

Oh, sorry, anyway, my patch will cause double locks...

>    
> Anyway what can be done is adding:
>
> lockdep_assert_held(&il->mutex);
>
> il4965_commit_rxon() to check if we hold the mutex.

I agree.


Best wishes,
Jia-Ju Bai