the musb ip on blackfin parts has an anomaly with DMA mode 1 transmits
where if the transfer is not a multiple of the maximum packet size,
the last short packet might be corrupted. to work around this issue,
we transfer the full sized packets using dma mode 1 and then use dma
mode 0 for the last short packet.
so in common code:
--- a/drivers/usb/musb/musb_host.c
+++ b/drivers/usb/musb/musb_host.c
@@ -672,7 +672,10 @@ static bool musb_tx_dma_program(struct dma_controller *dma,
wmb();
if (!dma->channel_program(channel, pkt_size, mode,
- urb->transfer_dma + offset, length)) {
+ urb->transfer_dma + offset,
+ (channel->desired_mode == 0) ? length :
+ length - (length % qh->maxpacket)))
+ {
dma->channel_release(channel);
hw_ep->tx_channel = NULL;
@@ -1750,7 +1753,10 @@ void musb_host_rx(struct musb *musb, u8 epnum)
*/
ret = c->channel_program(
dma, qh->maxpacket,
- dma->desired_mode, buf, length);
+ dma->desired_mode, buf,
+ (dma->desired_mode == 0)
+ ? length
+ : length - (length % qh->maxpacket));
if (!ret) {
c->channel_release(dma);
any ideas on how to get this workaround merged ? another idea i had
was that we could add a hook to the top of dma_channel_program to let
targets muck with the arguments ... the Blackfin code would just need
to do something like this at the start of the func:
+ if (musb_channel->transmit && channel->desired_mode == 1)
+ len = len - (len % packet_sz);
-mike
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
