[Prism54-devel] TX path improvement

[Jean-Baptiste Note]

Dear list,

The following patch tries to put some order in the tx path.
There *may* be improvements on the tx path. Could people reporting many
"tx timeout" problems try this ?

I'm also wondering : in the same way as the device can signal to us that
it has offset the data frame in the buffer we presented to it, wouldn't
there be a way to tell the device that the data we present it is offset
in the dma buffer we share with it ? -- this is possible on softmac
devices, and allows copyless TX.

Alternatively, maybe there's a way to inform the upper layers which
alignement we'd like ? I know for instance that it is possible to
request a minimum headroom (which will be usefull for WDS).

-- the aim would be to swith the flag on 
   if   (((long) skb->data & 0x03))

from likely, which is worst-case performance-wise, to likely...

JB

-------------- next part --------------
A non-text attachment was scrubbed...
Name: improve-tx-path.patch
Type: text/x-patch
Size: 7312 bytes
Desc: not available
Url : http://prism54.org/pipermail/prism54-devel/attachments/20050517/9be9e1a9/improve-tx-path.bin
-------------- next part --------------

-- 
Jean-Baptiste Note
+33 (0)6 83 03 42 38
jean-baptiste.note at wanadoo.fr