[Intel-gfx] [PATCH rdma-next v4 2/3] lib/scatterlist: Fix wrong update of orig_nents

Jason Gunthorpe jgg at nvidia.com
Wed Aug 25 13:09:23 UTC 2021


On Wed, Aug 25, 2021 at 07:59:27AM +0300, Maor Gottlieb wrote:
> 
> On 8/24/2021 10:12 PM, Jason Gunthorpe wrote:
> > On Tue, Aug 24, 2021 at 05:25:30PM +0300, Maor Gottlieb wrote:
> > > @@ -514,11 +531,13 @@ struct scatterlist *sg_alloc_append_table_from_pages(struct sg_table *sgt,
> > >   		offset = 0;
> > >   		cur_page = j;
> > >   	}
> > > -	sgt->nents += added_nents;
> > > +	sgt_append->sgt.nents += added_nents;
> > > +	sgt_append->sgt.orig_nents = sgt_append->sgt.nents;
> > > +	sgt_append->prv = s;
> > Why is nents being touched here?
> > 
> > Shouldn't it just be
> > 
> >      sgt_append->sgt.orig_nents += added_nents;
> >      sgt_append->prv = s;
> > 
> > ?
> 
> In general yes, but all the sg_alloc_* functions update both
> orig_nents and nents (to same value). May be drivers that pass nents
> instead of orig_nents to dma_map_sg* (like umem did before this
> patch set). So I prefer to set it here as well, also not to break
> sg_alloc_table_from_pages_segment which use the same path. This kind
> of change could be done very carefully next cycle for all
> sg_alloc_*.

Ok, interesting

Jason


More information about the Intel-gfx mailing list