[bug report] nl80211/cfg80211: add radar detection command/event

[Dan Carpenter <dan.carpenter@oracle.com>]

Hi wireless devs,

The patch 04f39047af2a: "nl80211/cfg80211: add radar detection
command/event" from Feb 8, 2013, leads to the following static
checker warning:

	net/wireless/chan.c:250 cfg80211_set_chans_dfs_state()
	warn: 'center_freq + bandwidth / 2 - 10' negative user limit promoted to high

net/wireless/chan.c
   242  static void cfg80211_set_chans_dfs_state(struct wiphy *wiphy, u32 center_freq,
   243                                           u32 bandwidth,
   244                                           enum nl80211_dfs_state dfs_state)
   245  {
   246          struct ieee80211_channel *c;
   247          u32 freq;
   248  
   249          for (freq = center_freq - bandwidth/2 + 10;
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   250               freq <= center_freq + bandwidth/2 - 10;
                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This isn't really a big issue but center_freq comes from
nla_get_u32(attrs[NL80211_ATTR_WIPHY_FREQ]) in nl80211_parse_chandef().
Smatch is complaining that there is an issue with the math
over/underflowing.  It just means that we loop for a long time.  It's
not a security problem.  Even without the overflow, we could end up
looping for a long time.

Is center_freq capped somewhere that I haven't seen?

   251               freq += 20) {
   252                  c = ieee80211_get_channel(wiphy, freq);
   253                  if (!c || !(c->flags & IEEE80211_CHAN_RADAR))
   254                          continue;
   255  
   256                  c->dfs_state = dfs_state;
   257                  c->dfs_state_entered = jiffies;
   258          }
   259  }

regards,
dan carpenter