Skip to contentSkip to conversation

staging: vt6656: vnt_beacon_xmit use extra_tx_headroom.

PATCH In review Malcolm Priestley 2 participants 3 messages May 16, 2020 – May 18, 2020 Staging

lore
Share & tools
lore
#0 lore raw

PATCH 2/4 staging: vt6656: vnt_beacon_xmit use extra_tx_headroom.

Create room for vnt_tx_short_buf_head in sk_buff and vnt_tx_usb_header.

The struct ieee80211_mgmt is not longer in the header and is at the initial skb->data point.

---
 drivers/staging/vt6656/rxtx.c | 22 +++++++++-------------
 drivers/staging/vt6656/rxtx.h |  5 +----
 2 files changed, 10 insertions(+), 17 deletions(-)
drivers/staging/vt6656/rxtx.c +9 -13
index 792833f8192a..cf194c95df03 100644
@@ -684,8 +684,9 @@ static int vnt_beacon_xmit(struct vnt_private *priv, struct sk_buff *skb)
 
 	spin_unlock_irqrestore(&priv->lock, flags);
 
-	beacon_buffer = (struct vnt_beacon_buffer *)&context->data[0];
-	short_head = &beacon_buffer->short_head;
+	mgmt_hdr = (struct ieee80211_mgmt *)skb->data;
+	short_head = skb_push(skb, sizeof(*short_head));
+	count = skb->len;
 
 	if (priv->bb_type == BB_TYPE_11A) {
 		current_rate = RATE_6M;
@@ -710,10 +711,6 @@ static int vnt_beacon_xmit(struct vnt_private *priv, struct sk_buff *skb)
 			vnt_time_stamp_off(priv, current_rate);
 	}
 
-	/* Generate Beacon Header */
-	mgmt_hdr = &beacon_buffer->mgmt_hdr;
-	memcpy(mgmt_hdr, skb->data, skb->len);
-
 	/* Get Duration */
 	short_head->duration = mgmt_hdr->duration;
 
@@ -732,15 +729,14 @@ static int vnt_beacon_xmit(struct vnt_private *priv, struct sk_buff *skb)
 	if (priv->seq_counter > 0x0fff)
 		priv->seq_counter = 0;
 
-	count = sizeof(struct vnt_tx_short_buf_head) + skb->len;
-
-	beacon_buffer->tx_byte_count = cpu_to_le16(count);
-	beacon_buffer->pkt_no = context->pkt_no;
-	beacon_buffer->type = 0x01;
+	beacon_buffer = skb_push(skb, sizeof(struct vnt_tx_usb_header));
+	beacon_buffer->usb.tx_byte_count = cpu_to_le16(count);
+	beacon_buffer->usb.pkt_no = context->pkt_no;
+	beacon_buffer->usb.type = 0x01;
 
 	context->type = CONTEXT_BEACON_PACKET;
-	context->tx_buffer = &context->data;
-	context->buf_len = count + 4; /* USB header */
+	context->tx_buffer = beacon_buffer;
+	context->buf_len = skb->len;
 
 	spin_lock_irqsave(&priv->lock, flags);
 
drivers/staging/vt6656/rxtx.h +1 -4
index 819b45394673..fd64d0838e34 100644
@@ -180,11 +180,8 @@ struct vnt_tx_short_buf_head {
 } __packed;
 
 struct vnt_beacon_buffer {
-	u8 type;
-	u8 pkt_no;
-	__le16 tx_byte_count;
+	struct vnt_tx_usb_header usb;
 	struct vnt_tx_short_buf_head short_head;
-	struct ieee80211_mgmt mgmt_hdr;
 } __packed;
 
 int vnt_tx_packet(struct vnt_private *priv, struct sk_buff *skb);
Signature · 1 line
2.25.1
#1 lore raw
Quoted 4 lines · Malcolm Priestley

Create room for vnt_tx_short_buf_head in sk_buff and vnt_tx_usb_header.

The struct ieee80211_mgmt is not longer in the header and is at the initial skb->data point.

I feel like the SubmittingPatches guidelines on verb tenses and not saying "this patch" or "I" has got everyone so worried that it's like playing Taboo. Do you mean that the struct moved before or after we aply *this patch*?

Quoted 2 lines

I can't understand the point of this patch at all. Is it a fix or a clean up? If I had to guess from the subjec, I would say it's a performance improvement but I don't know.

regards,
dan carpenter

#2 lore raw
Quoted 10 lines · Dan Carpenter
Quoted 4 lines · Malcolm Priestley

Create room for vnt_tx_short_buf_head in sk_buff and vnt_tx_usb_header.

The struct ieee80211_mgmt is not longer in the header and is at the initial skb->data point.

I feel like the SubmittingPatches guidelines on verb tenses and not saying "this patch" or "I" has got everyone so worried that it's like playing Taboo. Do you mean that the struct moved before or after we aply *this patch*?

The struct has not moved, before skb->data was copied on to the address along with the rest of frame.

So now struct needs to be at skb->data.

Quoted 6 lines
Quoted 2 lines

I can't understand the point of this patch at all. Is it a fix or a clean up? If I had to guess from the subjec, I would say it's a performance improvement but I don't know.

Well there is a performance improvement as there is only one buffer instead of two.

Mainly to bring into line with other drivers in the mac80211 tree there is no need for a secondary buffer in driver.

Regards

Malcolm