Re: [PATCH v2 2/2] dmaengine: xilinx: xdma: Support cyclic transfers

[Miquel Raynal]

Hi Vinod,

vkoul@xxxxxxxxxx wrote on Wed, 4 Oct 2023 12:59:25 +0530:

> On 03-10-23, 11:02, Miquel Raynal wrote:
> > Hi Vinod,
> > 
> > Thanks for the feedback.
> > 
> > vkoul@xxxxxxxxxx wrote on Thu, 28 Sep 2023 16:24:31 +0530:
> >   
> > > On 22-09-23, 18:20, Miquel Raynal wrote:
> > >   
> > > > @@ -583,7 +690,36 @@ static int xdma_alloc_chan_resources(struct dma_chan *chan)
> > > >  static enum dma_status xdma_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
> > > >  				      struct dma_tx_state *state)
> > > >  {
> > > > -	return dma_cookie_status(chan, cookie, state);
> > > > +	struct xdma_chan *xdma_chan = to_xdma_chan(chan);
> > > > +	struct xdma_desc *desc = NULL;
> > > > +	struct virt_dma_desc *vd;
> > > > +	enum dma_status ret;
> > > > +	unsigned long flags;
> > > > +	unsigned int period_idx;
> > > > +	u32 residue = 0;
> > > > +
> > > > +	ret = dma_cookie_status(chan, cookie, state);
> > > > +	if (ret == DMA_COMPLETE)
> > > > +		return ret;
> > > > +
> > > > +	spin_lock_irqsave(&xdma_chan->vchan.lock, flags);
> > > > +
> > > > +	vd = vchan_find_desc(&xdma_chan->vchan, cookie);
> > > > +	if (vd)
> > > > +		desc = to_xdma_desc(vd);    
> > > 
> > > vd is not used in below check, so should be done after below checks, why
> > > do this for cyclic case?  
> > 
> > I'm not sure I get this comment. vd is my way to get the descriptor,
> > and I need the descriptor to know whether we are in a cyclic transfer
> > or not. If the transfer is not cyclic, I just return the value from
> > dma_cookie_status() like before, otherwise I update the residue based
> > on the content of desc.
> > 
> > Maybe I don't understand what you mean, would you mind explaining it
> > again?  
> 
> Sorry I am not sure what I was thinking, this looks fine, we need the
> lock to get the desc and use it

Ah ok, no problem :) I'll send the v3 with the missing kernel doc line
(kernel test robot report).

Thanks,
Miquèl