Re: [Madwifi-devel] Frame length question

[Mathieu D. <mat...@cr...>]

Oops, I was actually talking about ath_tx_start, where
    pktlen = skb->len;

Since the "round up" effect should be visible in ath_tx_tasklet, that 
does not explain my problem, since I see this 2 bytes mismatch in 
ath_tx_start.
Also, to make sure there was no rounding effect, I also tried the 
following tests (which result in a frame of length multiple of 4 bytes, 
32 bits):

198 bytes data
8 bytes UDP overhead
20 bytes IPv4 header
34 bytes 802.11 data frame overhead [Frame ctrl (2), Dur/ID (2) Addr1(6), Addr2 (6), Addr3 (6), SeqCtrl (2), Addr4 (6), FCS (4)]
____________________________________________
Total 260 bytes

But the value of pktlen is 262 bytes in ath_tx_start.

To convince myself more, I also did the following test:

199 bytes data
8 bytes UDP overhead
20 bytes IPv4 header
34 bytes 802.11 data frame overhead [Frame ctrl (2), Dur/ID (2) Addr1(6), Addr2 (6), Addr3 (6), SeqCtrl (2), Addr4 (6), FCS (4)]
____________________________________________
Total 261 bytes

And here the value of pktlen is 263 bytes in ath_tx_start.

Can anyone explain this 2 bytes mismatch?

Thank you,
Mathieu.

   

Sam Leffler wrote:

>I see no reference to pktlen in ath_tx_tasklet.  If you are saying the value 
>in the h/w descriptor in ath_tx_tasklet is different from the value placed 
>there in ath_tx_start then it looks to be rounded up to a 32-bit boundary and 
>so reflects the actual DMA that was done.  But the frame on the wire should 
>still be whatever ath_tx_start says to use.
>
>	Sam
>  
>