Re: [PATCH] dma-debug: Check scatterlist segments

[Robin Murphy]

On 25/04/18 06:58, Christoph Hellwig wrote:
> This looks interesting.  I suspect it is going to blow up in
> quite a few places, so maybe at least for now it might make sense
> to have a separate config option?

True, it's nice to verify this for 'traditional' dma_map_sg() usage, but 
places where it's just used as an intermediate shorthand for "prepare 
all these pages for arbitrary future DMA" are liable to be noisy with 
false-positives for not respecting the default values when they arguably 
don't matter. I'll respin this as an additional config under 
DMA_API_DEBUG, and then we can debate "default y" or not.

Robin.

> On Tue, Apr 24, 2018 at 05:12:19PM +0100, Robin Murphy wrote:
> > Drivers/subsystems creating scatterlists for DMA should be taking care
> > to respect the scatter-gather limitations of the appropriate device, as
> > described by dma_parms. A DMA API implementation cannot feasibly split
> > a scatterlist into *more* entries than originally passed, so it is not
> > well defined what they should do when given a segment larger than the
> > limit they are also required to respect.
> >
> > Conversely, devices which are less limited than the rather conservative
> > defaults, or indeed have no limitations at all (e.g. GPUs with their own
> > internal MMU), should be encouraged to set appropriate dma_parms, as
> > they may get more efficient DMA mapping performance out of it.
> >
> > Signed-off-by: Robin Murphy <robin.murphy@xxxxxxx>
> > ---
> >   lib/dma-debug.c | 26 ++++++++++++++++++++++++++
> >   1 file changed, 26 insertions(+)
> >
> > diff --git a/lib/dma-debug.c b/lib/dma-debug.c
> > index 7f5cdc1e6b29..9f158941004d 100644
> > --- a/lib/dma-debug.c
> > +++ b/lib/dma-debug.c
> > @@ -1293,6 +1293,30 @@ static void check_sync(struct device *dev,
> >   	put_hash_bucket(bucket, &flags);
> >   }
> >   
> > +static void check_sg_segment(struct device *dev, struct scatterlist *sg)
> > +{
> > +	unsigned int max_seg = dma_get_max_seg_size(dev);
> > +	dma_addr_t start, end, boundary = dma_get_seg_boundary(dev);
> > +
> > +	/*
> > +	 * Either the driver forgot to set dma_parms appropriately, or
> > +	 * whoever generated the list forgot to check them.
> > +	 */
> > +	if (sg->length > max_seg)
> > +		err_printk(dev, NULL, "DMA-API: mapping sg segment longer than device claims to support [len=%u] [max=%u]\n",
> > +			   sg->length, max_seg);
> > +	/*
> > +	 * In some cases this could potentially be the DMA API
> > +	 * implementation's fault, but it would usually imply that
> > +	 * the scatterlist was built inappropriately to begin with.
> > +	 */
> > +	start = sg_dma_address(sg);
> > +	end = start + sg_dma_len(sg) - 1;
> > +	if ((start ^ end) & ~boundary)
> > +		err_printk(dev, NULL, "DMA-API: mapping sg segment across boundary [start=0x%016llx] [end=0x%016llx] [boundary=0x%016llx]\n",
> > +			   start, end, boundary);
> > +}
> > +
> >   void debug_dma_map_page(struct device *dev, struct page *page, size_t offset,
> >   			size_t size, int direction, dma_addr_t dma_addr,
> >   			bool map_single)
> > @@ -1423,6 +1447,8 @@ void debug_dma_map_sg(struct device *dev, struct scatterlist *sg,
> >   			check_for_illegal_area(dev, sg_virt(s), sg_dma_len(s));
> >   		}
> >   
> > +		check_sg_segment(dev, s);
> > +
> >   		add_dma_entry(entry);
> >   	}
> >   }
> > --
> > 2.17.0.dirty
> ---end quoted text---