Dear Eric Nelson,
On 03/13/2012 07:04 AM, Eric Nelson wrote:
ensure that transmit and receive buffers are cache-line aligned
invalidate cache for each packet as receivedupdate receive buffer descriptors one cache line at a time
flush cache before transmittingOriginal patch by Marek: http://lists.denx.de/pipermail/u-boot/2012-February/117695.html
Signed-off-by: Eric Nelsoneric.nelson@boundarydevices.com
V2 addresses some concerns from the ML:
Use readl()/writel() instead of mapped data structure
accesses
Wrong comment style
-&rbd_base[0] == rbd_base removed 'volatile' from fec_send().
<snip>
diff --git a/drivers/net/fec_mxc.c b/drivers/net/fec_mxc.c index 1fdd071..94a3927 100644 --- a/drivers/net/fec_mxc.c +++ b/drivers/net/fec_mxc.c
...
@@ -631,9 +682,11 @@ static void fec_halt(struct eth_device *dev)
- @param[in] length Data count in bytes
- @return 0 on success
*/
-static int fec_send(struct eth_device *dev, volatile void* packet, int length) +static int fec_send(struct eth_device *dev, void *packet, int length)
{
unsigned int status;
I made this change to keep checkpatch happy (it doesn't like volatile), but the declaration of struct eth_device in include/net.h seems to want the volatile:
int (*send) (struct eth_device*, volatile void* packet, int length);
I'd rather have a checkpatch warning than a compiler warning, so I'll fix this in V4...
I believe the volatile is there for a reason (not a reason meaningful for this device though). I believe on some boards, this was used to avoid some cache trouble or such.
Best regards, Marek Vasut