From: Emmanuel Grumbach <[email protected]>

iwl_trace_data is somewhat confusing. It returns a bool
that tells if the payload of the skb should be added to
the tx_data event. If it returns false, then the payload
of the skb is added to the tx event.

The purpose is to be able to start tracing with
-e iwlwifi
and record non-data packets only which saves bandwidth.

Since EAPOLs are important, seldom and not real data
packet (despite being WiFi data packets), they are
included in tx event and thus iwl_trace_data returns false
on those. This last part was buggy, and because of that,
all the data packets were included in the tx event.

Fix that.

Fixes: 0c4cb7314d15 ("iwlwifi: tracing: decouple from mac80211")
Signed-off-by: Emmanuel Grumbach <[email protected]>
Signed-off-by: Luca Coelho <[email protected]>
---
 drivers/net/wireless/intel/iwlwifi/iwl-devtrace.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/intel/iwlwifi/iwl-devtrace.h 
b/drivers/net/wireless/intel/iwlwifi/iwl-devtrace.h
index 545d14b0bc92..f5c1127253cb 100644
--- a/drivers/net/wireless/intel/iwlwifi/iwl-devtrace.h
+++ b/drivers/net/wireless/intel/iwlwifi/iwl-devtrace.h
@@ -55,8 +55,8 @@ static inline bool iwl_trace_data(struct sk_buff *skb)
        /* also account for the RFC 1042 header, of course */
        offs += 6;
 
-       return skb->len > offs + 2 &&
-              *(__be16 *)(skb->data + offs) == cpu_to_be16(ETH_P_PAE);
+       return skb->len <= offs + 2 ||
+               *(__be16 *)(skb->data + offs) != cpu_to_be16(ETH_P_PAE);
 }
 
 static inline size_t iwl_rx_trace_len(const struct iwl_trans *trans,
-- 
2.11.0

Reply via email to