linux-kernel - Re: [PATCHv4 8/8] videobuf2: handle non-contiguous DMA allocations

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAAFQd5BA4M89nQThm4gQgLp-LWzFfwnYw2HPyAs2QgUVn-ZLNQ@mail.gmail.com>
Date:   Wed, 18 Aug 2021 18:20:40 +0900
From:   Tomasz Figa <tfiga@...omium.org>
To:     Sergey Senozhatsky <senozhatsky@...omium.org>
Cc:     Hans Verkuil <hverkuil-cisco@...all.nl>,
        Dafna Hirschfeld <dafna.hirschfeld@...labora.com>,
        Ricardo Ribalda <ribalda@...omium.org>,
        Christoph Hellwig <hch@....de>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        linux-media@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCHv4 8/8] videobuf2: handle non-contiguous DMA allocations

On Tue, Aug 17, 2021 at 8:57 PM Sergey Senozhatsky
<senozhatsky@...omium.org> wrote:
>
> Hi,
>
> On (21/08/03 12:15), Hans Verkuil wrote:
> > >  static void *vb2_dc_vaddr(struct vb2_buffer *vb, void *buf_priv)
> > >  {
> > >     struct vb2_dc_buf *buf = buf_priv;
> > > -   struct dma_buf_map map;
> > > -   int ret;
> > >
> > > -   if (!buf->vaddr && buf->db_attach) {
> > > -           ret = dma_buf_vmap(buf->db_attach->dmabuf, &map);
> > > -           buf->vaddr = ret ? NULL : map.vaddr;
> > > +   if (buf->vaddr)
> > > +           return buf->vaddr;
> > > +
> > > +   if (buf->db_attach) {
> > > +           struct dma_buf_map map;
> > > +
> > > +           if (!dma_buf_vmap(buf->db_attach->dmabuf, &map))
> > > +                   buf->vaddr = map.vaddr;
> > > +
> > > +           return buf->vaddr;
> > >     }
> > >
> > > +   if (!buf->coherent_mem)
> > > +           buf->vaddr = dma_vmap_noncontiguous(buf->dev, buf->size,
> > > +                                               buf->dma_sgt);
> > >     return buf->vaddr;
> > >  }
> >
> > This function really needs a bunch of comments.
> >
> > What I want to see here specifically is under which circumstances this function
> > can return NULL.
> >
> > - dma_buf_vmap returns an error
> > - for non-coherent memory dma_vmap_noncontiguous returns an error
> > - coherent memory with DMA_ATTR_NO_KERNEL_MAPPING set.
>
> OK, I added some comments.
>
> > In the latter case, if a buffer with coherent memory and DMA_ATTR_NO_KERNEL_MAPPING
> > is exported as a dma_buf, and dma_buf_vmap is called by the importer of this dma-buf,
> > what happens then? I think that in that case dma_buf_vmap should return an error?
>
> Should we error out in vb2_dc_vaddr() in this case?

Yes, because there is no way to create a kernel mapping for buffer
allocated with dma_alloc_coherent() post-factum. Of course it's not
the case for dma_alloc_noncontiguous() for which we can create the
kernel mapping on demand.

>
> ---
>
> diff --git a/drivers/media/common/videobuf2/videobuf2-dma-contig.c b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> index d4089d0b5ec5..e1d8ae1548fa 100644
> --- a/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> +++ b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> @@ -102,6 +102,9 @@ static void *vb2_dc_vaddr(struct vb2_buffer *vb, void *buf_priv)
>         if (buf->db_attach) {
>                 struct dma_buf_map map;
>
> +               if (WARN_ON(buf->attrs & DMA_ATTR_NO_KERNEL_MAPPING))
> +                       return NULL;
> +

Why here? It's the case for buffers imported _into_ vb2, not exported
from vb2 and imported to other users.

>                 if (!dma_buf_vmap(buf->db_attach->dmabuf, &map))
>                         buf->vaddr = map.vaddr;
>
>
> ---
>
>
> [..]
> > > @@ -362,7 +451,7 @@ static int vb2_dc_dmabuf_ops_vmap(struct dma_buf *dbuf, struct dma_buf_map *map)
> > >  {
> > >     struct vb2_dc_buf *buf = dbuf->priv;
> > >
> > > -   dma_buf_map_set_vaddr(map, buf->vaddr);
> > > +   dma_buf_map_set_vaddr(map, vb2_dc_vaddr(buf->vb, buf));
> >
> > vb2_dc_vaddr() can return NULL, shouldn't this function return an error in that case?
>
> Done, thanks.
>
> > BTW, looking at where vb2_plane_vaddr() is called in drivers I notice that most (all?)
> > drivers do not check for NULL. Somewhat scary, to be honest. That's a separate issue, though.
>
> I may have some time in the future and can add missing if-s to the
> drivers.