Re: False warning from Sparse

From: Josh Triplett
Date: Sun Jul 22 2012 - 00:45:20 EST


On Sat, Jul 21, 2012 at 10:52:54PM -0500, Larry Finger wrote:
> I am getting the following false warning from sparse:
>
> CHECK drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c
> drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:1158:13: warning:
> context imbalance in 'rtl92c_dm_refresh_rate_adaptive_mask' -
> different lock contexts for basic block
>
> The only code in that routine that does any locking is the following:
>
> /* Only the PCI card uses sta in the update rate table
> * callback routine */
> if (rtlhal->interface == INTF_PCI) {
> rcu_read_lock();
> sta = ieee80211_find_sta(mac->vif, mac->bssid);
> }
> rtlpriv->cfg->ops->update_rate_tbl(hw, sta,
> p_ra->ratr_state);
>
> p_ra->pre_ratr_state = p_ra->ratr_state;
> if (rtlhal->interface == INTF_PCI)
> rcu_read_unlock();
>
> Does the warning get output because the code cannot assume that
> rtlhal->interface is the same in both if statements?

Correct; sparse does not have enough dataflow analysis to connect the
two if bodies, nor can it verify that nothing between the two changes
rtlhal->interface.

> If that is the
> case, are there any compiler directives that would tell sparse of
> the situation?

No, none of Sparse's directives will let you override this. However,
you could do something like the following:

/* Only the PCI card uses sta in the update rate table
* callback routine */
if (rtlhal->interface == INTF_PCI) {
rcu_read_lock();
sta = ieee80211_find_sta(mac->vif, mac->bssid);
rtlpriv->cfg->ops->update_rate_tbl(hw, sta, p_ra->ratr_state);
p_ra->pre_ratr_state = p_ra->ratr_state;
rcu_read_unlock();
} else {
rtlpriv->cfg->ops->update_rate_tbl(hw, sta, p_ra->ratr_state);
p_ra->pre_ratr_state = p_ra->ratr_state;
}

You could wrap those two lines in a function if you don't want to
duplicate them.

If that solution or a similar refactoring doesn't appeal to you, then
you don't currently have any way of making that context warning go away,
short of improving Sparse to handle this situation better.

- Josh Triplett
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/