potential null deref in iwlagn_request_scan()?
Smatch complains about iwlagn_request_scan().
drivers/net/wireless/iwlwifi/iwl-agn-lib.c +1354 iwlagn_request_scan(204)
error: we previously assumed 'vif' could be null.
1351 if (!priv->is_internal_short_scan) {
1352 cmd_len = iwl_fill_probe_req(priv,
1353 (struct ieee80211_mgmt *)scan->data,
1354 vif->addr,
^^^^^^^^^
1355 priv->scan_request->ie,
1356 priv->scan_request->ie_len,
1357 IWL_MAX_SCAN_SIZE - sizeof(*scan));
1358 } else {
1359 /* use bcast addr, will not be transmitted but must be valid */This was added in 3a0b9aad0a8166e9f "iwlwifi: use virtual interface address for scan". Prior to that commit the function assumed that vif could be NULL throughout.
I don't know the code well enough to know what to do about this.
Also the same thing for:drivers/net/wireless/iwlwifi/iwl3945-base.c +2963 iwl3945_request_scan(158)
error: we previously assumed 'vif' could be null.
2962 if (!priv->is_internal_short_scan) {
2963 scan->tx_cmd.len = cpu_to_le16(
2964 iwl_fill_probe_req(priv,
2965 (struct ieee80211_mgmt *)scan->data,
2966 vif->addr,
^^^^^^^^^ 2967 priv->scan_request->ie, 2968 priv->scan_request->ie_len, 2969 IWL_MAX_SCAN_SIZE - sizeof(*scan)));
2970 } else {
This is from the above commit as well.
regards,
dan carpenter
PATCH 1/1 iwlwifi: assume vif is NULL for internal scans and non-NULL otherwise
The current practice of checking vif for NULL in one place but not another seems to confuse some static checkers, smatch in particular. Since vif will only be NULL in the case of internal scans, adjust the checks accordingly.
--- drivers/net/wireless/iwlwifi/iwl-agn-lib.c | 5 ++++- drivers/net/wireless/iwlwifi/iwl3945-base.c | 5 ++++- 2 files changed, 8 insertions(+), 2 deletions(-)
drivers/net/wireless/iwlwifi/iwl-agn-lib.c +4 -1
@@ -1234,7 +1234,10 @@ void iwlagn_request_scan(struct iwl_priv *priv, struct ieee80211_vif *vif)IWL_DEBUG_INFO (priv,"Scanning while associated...\n" );spin_lock_irqsave (&priv->lock, flags); - interval = vif ? vif->bss_conf.beacon_int :0 ; +if (priv->is_internal_short_scan) + interval =0 ; +else + interval = vif->bss_conf.beacon_int;spin_unlock_irqrestore (&priv->lock, flags); scan->suspend_time =0 ;
drivers/net/wireless/iwlwifi/iwl3945-base.c +4 -1
@@ -2883,7 +2883,10 @@ void iwl3945_request_scan(struct iwl_priv *priv, struct ieee80211_vif *vif)IWL_DEBUG_INFO (priv,"Scanning while associated...\n" );spin_lock_irqsave (&priv->lock, flags); - interval = vif ? vif->bss_conf.beacon_int :0 ; +if (priv->is_internal_short_scan) + interval =0 ; +else + interval = vif->bss_conf.beacon_int;spin_unlock_irqrestore (&priv->lock, flags); scan->suspend_time =0 ;
Signature · 1 line
1.7.1.1
Quoted 21 lines · Dan Carpenter
Smatch complains about
iwlagn_request_scan().
drivers/net/wireless/iwlwifi/iwl-agn-lib.c+1354iwlagn_request_scan(204)
error: we previously assumed 'vif' could be null.1351 if (!priv->is_internal_short_scan) {
1352cmd_len= iwl_fill_probe_req(priv,
1353 (structieee80211_mgmt*)scan->data,
1354vif->addr,^^^^^^^^^1355 priv->scan_request->ie, 1356 priv->scan_request->ie_len, 1357 IWL_MAX_SCAN_SIZE - sizeof(*scan)); 1358 } else { 1359 /* use bcast addr, will not be transmitted but must be valid */This was added in
3a0b9aad0a8166e9f"iwlwifi: use virtual interface address for scan". Prior to that commit the function assumed that vif could be NULL throughout.
Quoted 5 lines
I don't know the code well enough to know what to do about this.
Also the same thing for:
drivers/net/wireless/iwlwifi/iwl3945-base.c+2963iwl3945_request_scan(158)
error: we previously assumed 'vif' could be null.
I think both of these are more-or-less OK. It looks like the only time vif would be NULL is in the "priv->is_internal_short_scan == true" case. I'll send a patch, but I don't know if it really matters.
John
Signature · 2 lines
John W. Linville Someday the world will need a hero, and you [email protected] might be all we have. Be ready.
Re: [PATCH] iwlwifi: assume vif is NULL for internal scans and non-NULL otherwise
In reply to John W. LinvilleHi John,
Quoted 33 lines · John W. Linville
The current practice of checking vif for NULL in one place but not another seems to confuse some static checkers, smatch in particular. Since vif will only be NULL in the case of internal scans, adjust the checks accordingly.
Signed-off-by John W. Linville <[email protected]>--- drivers/net/wireless/iwlwifi/iwl-agn-lib.c | 5 ++++- drivers/net/wireless/iwlwifi/iwl3945-base.c | 5 ++++- 2 files changed, 8 insertions(+), 2 deletions(-)drivers/net/wireless/iwlwifi/iwl-agn-lib.c +4
-1@@ -1234,7 +1234,10 @@ void iwlagn_request_scan(struct iwl_priv *priv, struct ieee80211_vif *vif)IWL_DEBUG_INFO (priv,"Scanning while associated...\n" );spin_lock_irqsave (&priv->lock, flags); - interval = vif ? vif->bss_conf.beacon_int :0 ; +if (priv->is_internal_short_scan) + interval =0 ; +else + interval = vif->bss_conf.beacon_int;spin_unlock_irqrestore (&priv->lock, flags); scan->suspend_time =0 ;drivers/net/wireless/iwlwifi/iwl3945-base.c +0
-0@@ -2883,7 +2883,10 @@ void iwl3945_request_scan(struct iwl_priv *priv, struct ieee80211_vif *vif)IWL_DEBUG_INFO (priv,"Scanning while associated...\n" );
drivers/net/wireless/iwlwifi/iwl3945-base.c void iwl3945_request_scan(struct iwl_priv *priv, struct ieee80211_vif *vif) · line 2892
Quoted 10 lines
drivers/net/wireless/iwlwifi/iwl3945-base.c +4
-1spin_lock_irqsave (&priv->lock, flags); - interval = vif ? vif->bss_conf.beacon_int :0 ; +if (priv->is_internal_short_scan) + interval =0 ; +else + interval = vif->bss_conf.beacon_int;spin_unlock_irqrestore (&priv->lock, flags); scan->suspend_time =0 ;
Make sense, Thanks.
Wey