On Sat, Jul 19, 2014 at 12:48:51PM +0200, Guennadi Liakhovetski wrote: > This patch adds a driver for NBPF DMAC IP cores from Renesas, designed for > the AMBA AXI bus. > > +struct nbpf_desc { > + struct dma_async_tx_descriptor async_tx; > + bool user_wait; sounds odd? > +static int nbpf_chan_probe(struct nbpf_device *nbpf, int n) > +{ > + struct dma_device *dma_dev = &nbpf->dma_dev; > + struct nbpf_channel *chan = nbpf->chan + n; > + int ret; > + > + chan->nbpf = nbpf; > + chan->base = nbpf->base + NBPF_REG_CHAN_OFFSET + NBPF_REG_CHAN_SIZE * n; > + INIT_LIST_HEAD(&chan->desc_page); > + spin_lock_init(&chan->lock); > + chan->dma_chan.device = dma_dev; > + dma_cookie_init(&chan->dma_chan); > + nbpf_chan_prepare_default(chan); > + > + dev_dbg(dma_dev->dev, "%s(): channel %d: -> %p\n", __func__, n, chan->base); > + > + snprintf(chan->name, sizeof(chan->name), "nbpf %d", n); > + > + ret = devm_request_threaded_irq(dma_dev->dev, chan->irq, > + nbpf_chan_irq, nbpf_chan_irqt, IRQF_SHARED, > + chan->name, chan); devm_request_irq and friends arent apt here. You are in .remove and get an irq, what prevents that race? You need to run free_irq and synchronize_irq() to ensure everything is proper before freeing up. Also why threaded_irq when dmaengine API mandates you to use tasklet? The callback is supposed to be invoked from a tasklet. Thanks -- ~Vinod -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html