On Tue, 2010-05-18 at 16:05 +0200, ext Sebastien Jan wrote: > On Tuesday 18 May 2010 11:07:20 Carlos Chinea wrote: > [cut] > > > > + val |= __raw_readl(omap_ssi->sys + > > > > SSI_MPU_ENABLE_REG(port->num, 0)); + __raw_writel(val, > > > > omap_ssi->sys + > > > > SSI_MPU_ENABLE_REG(port->num, 0)); + > > > > + msg->status = HSI_STATUS_COMPLETED; > > > > + msg->actual_len = sg_dma_len(msg->sgt.sgl); > > > > + spin_unlock(&omap_ssi->lock); > > > > +} > > > > > > Don't you need to check the queue related to this transfer at this point, > > > to start the potentially next queued transfer on the same channel? > > > (calling ssi_start_transfer(), like in ssi_pio_complete()?) > > > > No this is done in ssi_pio_complete(). Notice that we do not call the > > complete callback at any point here. We just arm the pio interrupt for > > that channel and transfer direction. AFAIK, this is the SW logic > > expected by the OMAP SSI HW. > > Ok, though I would not expect the interrupt to fire in an Rx scenario as the > fifo would have already been emptied by the DMA for this transfer (unless you > rely on the next transfer initiated by the peer to make the Rx interrupt fire > on this channel?)? No I do not rely on the next RX transfer. I rely on the fact that the GDD(DMA) controller does not reset the RX status bit for that channel when RX transfer is finished. Br, -- Carlos Chinea <carlos.chinea@xxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
