Hi Hans,
On Wed, Apr 24, 2024 at 05:50:56PM +0200, Hans Verkuil wrote:
> On 24/04/2024 17:22, Laurent Pinchart wrote:
> > Set len_routes of struct v4l2_subdev_routing. ENOSPC error code is no
> > longer used, i.e. having room for fewer routes than exist in the
> > configuration is not considered an error anymore.
> >
> > Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > ---
> > utils/v4l2-compliance/v4l2-compliance.cpp | 12 +++++++-----
> > utils/v4l2-compliance/v4l2-test-subdevs.cpp | 19 +++++++++++--------
> > 2 files changed, 18 insertions(+), 13 deletions(-)
> >
> > diff --git a/utils/v4l2-compliance/v4l2-compliance.cpp b/utils/v4l2-compliance/v4l2-compliance.cpp
> > index fd7e7d76e214..144f961842c6 100644
> > --- a/utils/v4l2-compliance/v4l2-compliance.cpp
> > +++ b/utils/v4l2-compliance/v4l2-compliance.cpp
> > @@ -1272,15 +1272,17 @@ void testNode(struct node &node, struct node &node_m2m_cap, struct node &expbuf_
> >
> > for (unsigned which = V4L2_SUBDEV_FORMAT_TRY;
> > which <= V4L2_SUBDEV_FORMAT_ACTIVE; which++) {
> > + struct v4l2_subdev_routing &routing = sd_routing[which];
> >
> > - sd_routing[which].which = which;
> > - sd_routing[which].routes = (uintptr_t)sd_routes[which];
> > - sd_routing[which].num_routes = NUM_ROUTES_MAX;
> > + routing.which = which;
> > + routing.routes = (uintptr_t)sd_routes[which];
> > + routing.len_routes = NUM_ROUTES_MAX;
> > + routing.num_routes = 0;
> >
> > - ret = doioctl(&node, VIDIOC_SUBDEV_G_ROUTING, &sd_routing[which]);
> > + ret = doioctl(&node, VIDIOC_SUBDEV_G_ROUTING, &routing);
> > if (ret) {
> > fail("VIDIOC_SUBDEV_G_ROUTING: failed to get routing\n");
> > - sd_routing[which].num_routes = 0;
> > + routing.num_routes = 0;
> > }
> > }
> > }
> > diff --git a/utils/v4l2-compliance/v4l2-test-subdevs.cpp b/utils/v4l2-compliance/v4l2-test-subdevs.cpp
> > index da304a8caa8a..41eaf77112f0 100644
> > --- a/utils/v4l2-compliance/v4l2-test-subdevs.cpp
> > +++ b/utils/v4l2-compliance/v4l2-test-subdevs.cpp
> > @@ -587,17 +587,15 @@ int testSubDevRouting(struct node *node, unsigned which)
> >
> > routing.which = which;
> > routing.routes = (uintptr_t)&routes;
> > + routing.len_routes = 0;
> > routing.num_routes = 0;
> > memset(routing.reserved, 0xff, sizeof(routing.reserved));
> >
> > - /*
> > - * First test that G_ROUTING either returns success, or ENOSPC and
> > - * updates num_routes.
> > - */
> > + /* First test that G_ROUTING returns success even when len_routes is 0. */
> >
> > ret = doioctl(node, VIDIOC_SUBDEV_G_ROUTING, &routing);
> > - fail_on_test(ret && ret != ENOSPC);
> > - fail_on_test(ret == ENOSPC && routing.num_routes == 0);
> > + fail_on_test(ret);
> > + fail_on_test(routing.num_routes > NUM_ROUTES_MAX);
> > fail_on_test(check_0(routing.reserved, sizeof(routing.reserved)));
> >
> > if (!routing.num_routes)
> > @@ -609,7 +607,8 @@ int testSubDevRouting(struct node *node, unsigned which)
> > */
> >
> > uint32_t num_routes = routing.num_routes;
> > - routing.num_routes = num_routes + 1;
> > + routing.len_routes = NUM_ROUTES_MAX;
> > + routing.num_routes = 0;
> > fail_on_test(doioctl(node, VIDIOC_SUBDEV_G_ROUTING, &routing));
> > fail_on_test(routing.num_routes != num_routes);
> >
> > @@ -633,10 +632,14 @@ int testSubDevRouting(struct node *node, unsigned which)
> > }
> > }
> >
> > - /* Set the same routes back, which should always succeed. */
> > + /*
> > + * Set the same routes back, which should always succeed and report the
> > + * same number of routes.
> > + */
> >
> > memset(routing.reserved, 0xff, sizeof(routing.reserved));
> > fail_on_test(doioctl(node, VIDIOC_SUBDEV_S_ROUTING, &routing));
> > + fail_on_test(routing.num_routes != num_routes);
> > fail_on_test(check_0(routing.reserved, sizeof(routing.reserved)));
>
> I think this needs another test: if S_ROUTING is called with num_routes > len_routes,
> then EINVAL must be returned. This is an important test to make sure the core handles
> this correctly and won't be using more than len_routes routes.
I'll add the test now, that should be easy. I'll push an update to my
branch but won't repost the patches yet, to let reviewers a chance to
check the series.
> >
> > /* Test setting invalid pads */
--
Regards,
Laurent Pinchart
