Re: [PATCH v2] net: ethernet: ti: davinci_cpdma: fix warning "device driver frees DMA memory with different size"

[Ivan Khoronzhuk]

On Thu, Dec 05, 2019 at 12:48:46PM +0200, Grygorii Strashko wrote:
> On 04/12/2019 22:37, David Miller wrote:
> > From: Grygorii Strashko <grygorii.strashko@xxxxxx>
> > Date: Wed, 4 Dec 2019 18:50:29 +0200
> >
> > > @@ -1018,7 +1018,7 @@ static int cpdma_chan_submit_si(struct submit_info *si)
> > >  	struct cpdma_chan		*chan = si->chan;
> > >  	struct cpdma_ctlr		*ctlr = chan->ctlr;
> > >  	int				len = si->len;
> > > -	int				swlen = len;
> > > +	int				swlen;
> > >  	struct cpdma_desc __iomem	*desc;
> > >  	dma_addr_t			buffer;
> > >  	u32				mode;
> > > @@ -1040,6 +1040,7 @@ static int cpdma_chan_submit_si(struct submit_info *si)
> > >  		chan->stats.runt_transmit_buff++;
> > >  	}
> > > +	swlen = len;
> > >  	mode = CPDMA_DESC_OWNER | CPDMA_DESC_SOP | CPDMA_DESC_EOP;
> > >  	cpdma_desc_to_port(chan, mode, si->directed);
> > > --
> > > 2.17.1
> >
> > Now there is no reason to keep a separate swlen variable.
> >
> > The integral value is always consumed as the length before the descriptor bits
> > are added to it.
> >
> > Therefore you can just use 'len' everywhere in this function now.
>
> Sry, but seems i can't, at least i can't just drop swlen.
>
> Below in this function:
> 	writel_relaxed(0, &desc->hw_next);
> 	writel_relaxed(buffer, &desc->hw_buffer);
> 	writel_relaxed(len, &desc->hw_len);
> 	writel_relaxed(mode | len, &desc->hw_mode);
> ^^ here the "len" should be use
>
> 	writel_relaxed((uintptr_t)si->token, &desc->sw_token);
> 	writel_relaxed(buffer, &desc->sw_buffer);
> 	writel_relaxed(swlen, &desc->sw_len);
> ^^ and here "len"|CPDMA_DMA_EXT_MAP if (si->data_dma) [1]
>
> 	desc_read(desc, sw_len);
>
> so additional if statement has to be added at [1] if "swlen" is dropped
>
> --
> Best regards,
> grygorii

Seems like yes,

And the "swlen" can be avoided like this:

--- a/drivers/net/ethernet/ti/davinci_cpdma.c
+++ b/drivers/net/ethernet/ti/davinci_cpdma.c
@@ -1018,7 +1018,6 @@ static int cpdma_chan_submit_si(struct submit_info *si)
       struct cpdma_chan               *chan = si->chan;
       struct cpdma_ctlr               *ctlr = chan->ctlr;
       int                             len = si->len;
-       int                             swlen = len;
       struct cpdma_desc __iomem       *desc;
       dma_addr_t                      buffer;
       u32                             mode;
@@ -1046,7 +1045,7 @@ static int cpdma_chan_submit_si(struct submit_info *si)
       if (si->data_dma) {
               buffer = si->data_dma;
               dma_sync_single_for_device(ctlr->dev, buffer, len, chan->dir);
-               swlen |= CPDMA_DMA_EXT_MAP;
+               writel_relaxed(len | CPDMA_DMA_EXT_MAP, &desc->sw_len);
       } else {
               buffer = dma_map_single(ctlr->dev, si->data_virt, len, chan->dir);
               ret = dma_mapping_error(ctlr->dev, buffer);
@@ -1054,6 +1053,7 @@ static int cpdma_chan_submit_si(struct submit_info *si)
                       cpdma_desc_free(ctlr->pool, desc, 1);
                       return -EINVAL;
               }
+               writel_relaxed(len, &desc->sw_len);
       }

       /* Relaxed IO accessors can be used here as there is read barrier
@@ -1065,7 +1065,6 @@ static int cpdma_chan_submit_si(struct submit_info *si)
       writel_relaxed(mode | len, &desc->hw_mode);
       writel_relaxed((uintptr_t)si->token, &desc->sw_token);
       writel_relaxed(buffer, &desc->sw_buffer);
-       writel_relaxed(swlen, &desc->sw_len);
       desc_read(desc, sw_len);

       __cpdma_chan_submit(chan, desc);

But not sure what is better.

--
Regards,
Ivan Khoronzhuk