Re: [PATCHv2] media: videobuf2-dma-contig: fix bad kfree in vb2_dma_contig_clear_max_seg_size

[Ulf Hansson]

On Thu, 28 May 2020 at 11:24, Marek Szyprowski <m.szyprowski@xxxxxxxxxxx> wrote:
>
> Hi Ulf,
>
> On 28.05.2020 11:14, Ulf Hansson wrote:
> > On Wed, 27 May 2020 at 10:23, Tomi Valkeinen <tomi.valkeinen@xxxxxx> wrote:
> >> Commit 9495b7e92f716ab2bd6814fab5e97ab4a39adfdd ("driver core: platform:
> >> Initialize dma_parms for platform devices") in v5.7-rc5 causes
> >> vb2_dma_contig_clear_max_seg_size() to kfree memory that was not
> >> allocated by vb2_dma_contig_set_max_seg_size().
> >>
> >> The assumption in vb2_dma_contig_set_max_seg_size() seems to be that
> >> dev->dma_parms is always NULL when the driver is probed, and the case
> >> where dev->dma_parms has bee initialized by someone else than the driver
> >> (by calling vb2_dma_contig_set_max_seg_size) will cause a failure.
> >>
> >> All the current users of these functions are platform devices, which now
> >> always have dma_parms set by the driver core. To fix the issue for v5.7,
> >> make vb2_dma_contig_set_max_seg_size() return an error if dma_parms is
> >> NULL to be on the safe side, and remove the kfree code from
> >> vb2_dma_contig_clear_max_seg_size().
> >>
> >> For v5.8 we should remove the two functions and move the
> >> dma_set_max_seg_size() calls into the drivers.
> >>
> >> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@xxxxxx>
> >> Fixes: 9495b7e92f71 ("driver core: platform: Initialize dma_parms for platform devices")
> >> Cc: stable@xxxxxxxxxxxxxxx
> > Thanks for fixing this!
> >
> > However, as I tried to point out in v1, don't you need to care about
> > drivers/media/platform/s5p-mfc/s5p_mfc.c, which allocates its own type
> > of struct device (non-platform). No?
>
> I will send a patch for the S5P MFC driver separately. It is not so
> critical, because the mentioned 2port memory case is not used on any
> board with the default exynos_defconfig from mainline. On Exynos4-based
> boards it is only used when IOMMU is disabled. It is mainly for
> S5PV210/S5PC110 boards, which are still not fully functional after
> conversion to device-tree.

Okay, makes sense to me. Feel free to add:

Reviewed-by: Ulf Hansson <ulf.hansson@xxxxxxxxxx>

For the s5p_mfc issue, here's how I would have done it (just pick the
code if you want).

diff --git a/drivers/media/platform/s5p-mfc/s5p_mfc.c
b/drivers/media/platform/s5p-mfc/s5p_mfc.c
index 5c2a23b953a4..7acf2a03812d 100644
--- a/drivers/media/platform/s5p-mfc/s5p_mfc.c
+++ b/drivers/media/platform/s5p-mfc/s5p_mfc.c
@@ -1070,6 +1070,7 @@ static const struct v4l2_file_operations s5p_mfc_fops = {
 /* DMA memory related helper functions */
 static void s5p_mfc_memdev_release(struct device *dev)
 {
+       kfree(dev->dma_parms);
        of_reserved_mem_device_release(dev);
 }

@@ -1090,6 +1091,10 @@ static struct device
*s5p_mfc_alloc_memdev(struct device *dev,
        child->dma_mask = dev->dma_mask;
        child->release = s5p_mfc_memdev_release;

+       child->dma_parms = kzalloc(sizeof(*child->dma_parms), GFP_KERNEL);
+       if (!child->dma_parms)
+               goto err;
+
        /*
         * The memdevs are not proper OF platform devices, so in order for them
         * to be treated as valid DMA masters we need a bit of a hack to force
@@ -1105,6 +1110,7 @@ static struct device
*s5p_mfc_alloc_memdev(struct device *dev,
                device_del(child);
        }

+err:
        put_device(child);
        return NULL;
 }

Kind regards
Uffe