Skip to contentSkip to conversation

rtlwifi: Initialize rate_init member of struct rate_control_ops

PATCH In review Catalin Iacob 3 participants 6 messages Feb 11, 2013 Networking drivers (wireless)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 rtlwifi: Initialize rate_init member of struct rate_control_ops

This partially reverts commit 44ba973699b831414c3f8eef68ee5a7fe1208a05.

rate_control_rate_init assumes the rate_init member of
struct rate_control_ops is not NULL therefore not initializing it leads to an oops as soon the driver succesfully associates to an AP.

The removal of rate_update from 44ba973699b831414c3f8eef68ee5a7fe1208a05 is ok because rate_update is checked for NULL before being called.

---
 drivers/net/wireless/rtlwifi/rc.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/net/wireless/rtlwifi/rc.c +7 -0
index 204f46c..6ad8bb7 100644
@@ -217,6 +217,12 @@ static void rtl_tx_status(void *ppriv,
 	}
 }
 
+static void rtl_rate_init(void *ppriv,
+			  struct ieee80211_supported_band *sband,
+			  struct ieee80211_sta *sta, void *priv_sta)
+{
+}
+
 static void *rtl_rate_alloc(struct ieee80211_hw *hw,
 		struct dentry *debugfsdir)
 {
@@ -261,6 +267,7 @@ static struct rate_control_ops rtl_rate_ops = {
 	.free = rtl_rate_free,
 	.alloc_sta = rtl_rate_alloc_sta,
 	.free_sta = rtl_rate_free_sta,
+	.rate_init = rtl_rate_init,
 	.tx_status = rtl_tx_status,
 	.get_rate = rtl_get_rate,
 };
Signature · 1 line
1.8.1.1
maintainer
#1 lore raw
Quoted 9 lines · Catalin Iacob

This partially reverts commit 44ba973699b831414c3f8eef68ee5a7fe1208a05.

rate_control_rate_init assumes the rate_init member of
struct rate_control_ops is not NULL therefore not initializing it leads to an oops as soon the driver succesfully associates to an AP.

The removal of rate_update from 44ba973699b831414c3f8eef68ee5a7fe1208a05 is ok because rate_update is checked for NULL before being called.

OTOH, if the driver supports HT then rate_update is critical to its functioning properly, e.g. for SMPS updates.

johannes

#2 lore raw
In reply to Catalin Iacob
Quoted 33 lines · Catalin Iacob

This partially reverts commit 44ba973699b831414c3f8eef68ee5a7fe1208a05.

rate_control_rate_init assumes the rate_init member of
struct rate_control_ops is not NULL therefore not initializing it leads to an oops as soon the driver succesfully associates to an AP.

The removal of rate_update from 44ba973699b831414c3f8eef68ee5a7fe1208a05 is ok because rate_update is checked for NULL before being called.

---
  drivers/net/wireless/rtlwifi/rc.c | 7 +++++++
  1 file changed, 7 insertions(+)
drivers/net/wireless/rtlwifi/rc.c +6 -0
index 204f46c..6ad8bb7 100644
@@ -217,6 +217,12 @@ static void rtl_tx_status(void *ppriv,
  	}
  }
 
+static void rtl_rate_init(void *ppriv,
+			  struct ieee80211_supported_band *sband,
+			  struct ieee80211_sta *sta, void *priv_sta)
+{
+}
+
  static void *rtl_rate_alloc(struct ieee80211_hw *hw,
  		struct dentry *debugfsdir)
  {

drivers/net/wireless/rtlwifi/rc.c static struct rate_control_ops rtl_rate_ops = { · line 273

Quoted 7 lines
drivers/net/wireless/rtlwifi/rc.c +1 -0
@@ -261,6 +267,7 @@ static struct rate_control_ops rtl_rate_ops = {
  	.free = rtl_rate_free,
  	.alloc_sta = rtl_rate_alloc_sta,
  	.free_sta = rtl_rate_free_sta,
+	.rate_init = rtl_rate_init,
  	.tx_status = rtl_tx_status,
  	.get_rate = rtl_get_rate,
  };

Shouldn't rate_control_rate_init() in net/mac80211/rate.h be changed to protect against the oops? I don't see any value in the client driver having to provide a dummy routine. The rtlwifi family of drivers use their own rate-control mechanism, and get no rate-control info from mac80211.

Larry

maintainer
#3 lore raw
Quoted 9 lines · Larry Finger

drivers/net/wireless/rtlwifi/rc.c static struct rate_control_ops rtl_rate_ops = { · line 273

Quoted 4 lines
drivers/net/wireless/rtlwifi/rc.c +1 -0
+	.rate_init = rtl_rate_init,
  	.tx_status = rtl_tx_status,
  	.get_rate = rtl_get_rate,
  };

Shouldn't rate_control_rate_init() in net/mac80211/rate.h be changed to protect against the oops? I don't see any value in the client driver having to provide a dummy routine. The rtlwifi family of drivers use their own rate-control mechanism, and get no rate-control info from mac80211.

In this case, I see no value in them providing rate control operations at all, and they could just set IEEE80211_HW_HAS_RATE_CONTROL, no?

johannes

#4 lore raw
Quoted 12 lines · Johannes Berg
Quoted 9 lines · Larry Finger

drivers/net/wireless/rtlwifi/rc.c static struct rate_control_ops rtl_rate_ops = { · line 273

Quoted 4 lines
drivers/net/wireless/rtlwifi/rc.c +1 -0
+	.rate_init = rtl_rate_init,
   	.tx_status = rtl_tx_status,
   	.get_rate = rtl_get_rate,
   };

Shouldn't rate_control_rate_init() in net/mac80211/rate.h be changed to protect against the oops? I don't see any value in the client driver having to provide a dummy routine. The rtlwifi family of drivers use their own rate-control mechanism, and get no rate-control info from mac80211.

Quoted 2 lines

In this case, I see no value in them providing rate control operations at all, and they could just set IEEE80211_HW_HAS_RATE_CONTROL, no?

I think that makes sense.

Larry

Acked-by on v1 1/1
#5 lore raw
In reply to Catalin Iacob
Quoted 11 lines · Catalin Iacob

This partially reverts commit 44ba973699b831414c3f8eef68ee5a7fe1208a05.

rate_control_rate_init assumes the rate_init member of
struct rate_control_ops is not NULL therefore not initializing it leads to an oops as soon the driver succesfully associates to an AP.

The removal of rate_update from 44ba973699b831414c3f8eef68ee5a7fe1208a05 is ok because rate_update is checked for NULL before being called.

John,

This patch should be applied. Obviously, I missed the oops in my testing.

The real fix will be to not supply any rate-control operations as suggested by Johannes, but that will take some time.

Larry