On Thu, Sep 17, 2020 at 05:07:12PM +0100, Kieran Bingham wrote:
> Hi Laurent,
>
> On 17/09/2020 01:58, Laurent Pinchart wrote:
> > Hi Kieran,
> >
> > Thank you for the patch.
> >
> > On Wed, Sep 16, 2020 at 03:18:15PM +0100, Kieran Bingham wrote:
> >> Perform unbind-bind testing of the VSP devices to validate
> >> successful removal of the drivers.
> >
> > This unbinds all VSP instances, including the ones used by the DU and
> > not exposed through V4L2. What happens to the display ?
>
> Bad things :-)
>
> Is that something that 'can' be fixed?
Only with lots of effort. I wouldn't bother for now to be honest. Could
we however skip unbinding the vspd instances in this test ?
> Or goes into the category of "shouldn't ever happen".
>
> Although I think the reason for adding this test in the first place was
> because of reports of someone testing bind/unbind and hitting failures.
>
> After unbind/bind, the display is left blank. There is no 'issue' until
> something tries to use the display again: where we get:
>
> [ 273.468961] vsp1 fea20000.vsp: failed to reset wpf.1
> [ 273.474156] vsp1 fea20000.vsp: DRM pipeline stop timeout
> [ 273.479595] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000038
> [ 273.488627] Mem abort info:
> [ 273.491570] ESR = 0x96000004
> [ 273.495657] EC = 0x25: DABT (current EL), IL = 32 bits
> [ 273.501618] SET = 0, FnV = 0
> [ 273.505093] EA = 0, S1PTW = 0
> [ 273.509022] Data abort info:
> [ 273.512213] ISV = 0, ISS = 0x00000004
> [ 273.516685] CM = 0, WnR = 0
> [ 273.519891] user pgtable: 4k pages, 48-bit VAs, pgdp=000000072faf0000
> [ 273.527015] [0000000000000038] pgd=0000000000000000, p4d=0000000000000000
> [ 273.534399] Internal error: Oops: 96000004 [#1] PREEMPT SMP
> [ 273.540041] CPU: 0 PID: 1338 Comm: kmstest Not tainted 5.9.0-rc1-arm64-renesas-00264-g939a20a7f8b4 #339
> [ 273.549523] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
> [ 273.557517] pstate: 60000005 (nZCv daif -PAN -UAO BTYPE=--)
> [ 273.563152] pc : vsp1_du_setup_lif+0x3fc/0x4d0
> [ 273.567638] lr : vsp1_du_setup_lif+0x4cc/0x4d0
> [ 273.572121] sp : ffff80001489b900
> [ 273.575465] x29: ffff80001489b900 x28: 000000000000006f
> [ 273.580829] x27: ffff0006fa105080 x26: 0000000000000000
> [ 273.586192] x25: 0000000000000000 x24: ffff0006f0f896c0
> [ 273.591555] x23: 0000000000000378 x22: ffff0006f81c83f8
> [ 273.596917] x21: ffff0006f81c8420 x20: 0000000000000001
> [ 273.602280] x19: 0000000000000000 x18: 0000000000000010
> [ 273.607642] x17: 0000000000000000 x16: 0000000000000000
> [ 273.613004] x15: 0000000000aaaaaa x14: 0000000000000020
> [ 273.618366] x13: 00000000ffffffff x12: ffff800010b68ac8
> [ 273.623728] x11: ffffffffffffffff x10: ffff800010ef8018
> [ 273.629090] x9 : 0000000000000001 x8 : 0000000000000000
> [ 273.634452] x7 : 0000000000000001 x6 : 0000000000000001
> [ 273.639814] x5 : dead000000000100 x4 : dead000000000100
> [ 273.645176] x3 : dead000000000122 x2 : b1d0872bc242a400
> [ 273.650538] x1 : 0000000000000000 x0 : 0000000000000000
> [ 273.655900] Call trace:
> [ 273.658371] vsp1_du_setup_lif+0x3fc/0x4d0
> [ 273.662512] rcar_du_vsp_disable+0x1c/0x28
> [ 273.666648] rcar_du_crtc_atomic_disable+0x2c4/0x3e0
> [ 273.671663] drm_atomic_helper_commit_modeset_disables+0x37c/0x420
> [ 273.677902] rcar_du_atomic_commit_tail+0x90/0xd8
> [ 273.684043] commit_tail+0x9c/0x180
> [ 273.688952] drm_atomic_helper_commit+0x168/0x198
> [ 273.695096] drm_atomic_commit+0x48/0x58
> [ 273.700437] drm_mode_atomic_ioctl+0x9fc/0xb08
> [ 273.706278] drm_ioctl_kernel+0xb8/0x110
> [ 273.711563] drm_ioctl+0x230/0x470
> [ 273.716293] __arm64_sys_ioctl+0x8c/0xd0
> [ 273.721520] el0_svc_common.constprop.0+0x70/0x180
> [ 273.727629] do_el0_svc+0x20/0x80
> [ 273.732219] el0_sync_handler+0x94/0x1b0
> [ 273.737386] el0_sync+0x140/0x180
> [ 273.741943] Code: 8b1c0ed6 f9414ec0 b90262df f901b6df (a9438402)
> [ 273.749313] ---[ end trace 67e85a6fa8c4fc15 ]---
>
> We don't actually have any bind/unbind hooks in the vsp1 so I guess
> that's something for me to investigate next to make sure we shut the
> device down correctly...
>
> >> Signed-off-by: Kieran Bingham <kieran.bingham@xxxxxxxxxxxxxxxx>
> >> Reviewed-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> >> ---
> >>
> >> v2:
> >> - Semi-colons removed
> >> - duplicated tests removed.
> >>
> >> This is the updated patch, I intend to push to master.
> >>
> >> --
> >> Kieran
> >>
> >>
> >> tests/vsp-unit-test-0026.sh | 55 +++++++++++++++++++++++++++++++++++++
> >> 1 file changed, 55 insertions(+)
> >> create mode 100755 tests/vsp-unit-test-0026.sh
> >>
> >> diff --git a/tests/vsp-unit-test-0026.sh b/tests/vsp-unit-test-0026.sh
> >> new file mode 100755
> >> index 000000000000..0e013cec881b
> >> --- /dev/null
> >> +++ b/tests/vsp-unit-test-0026.sh
> >> @@ -0,0 +1,55 @@
> >> +#!/bin/sh
> >> +
> >> +#
> >> +# Test unbinding and binding all VSP1 devices, performing a simple
> >> +# copy test to validate the hardware afterwards.
> >> +#
> >> +
> >> +. ./vsp-lib.sh
> >> +
> >> +features="rpf.0 wpf.0"
> >> +
> >> +vsp1_driver=/sys/bus/platform/drivers/vsp1
> >> +vsps=$(cd /sys/bus/platform/devices/; ls | grep vsp)
> >> +
> >> +unbind_vsp() {
> >> + echo $1 > $vsp1_driver/unbind
> >> +}
> >> +
> >> +bind_vsp() {
> >> + echo $1 > $vsp1_driver/bind
> >> +}
> >> +
> >> +# Input is directly copied to the output. No change in format or size.
> >> +test_copy() {
> >> + local format=$1
> >> + local insize=$2
> >> +
> >> + test_start "simple hardware validation after unbind/bind cycles"
> >> +
> >> + pipe_configure rpf-wpf 0 0
> >> + format_configure rpf-wpf 0 0 $format $insize $format
> >> +
> >> + vsp_runner rpf.0 &
> >> + vsp_runner wpf.0
> >> +
> >> + local result=$(compare_frames)
> >> +
> >> + test_complete $result
> >> +}
> >> +
> >> +test_main() {
> >> + local format
> >> +
> >> + # Unbind and rebind VSPs individually
> >> + for v in $vsps; do
> >> + unbind_vsp $v
> >> + bind_vsp $v
> >> + done
> >> +
> >> + # Perform a simple copy test to validate HW is alive
> >> + test_copy RGB24 128x128
> >> +}
> >> +
> >> +test_init $0 "$features"
> >> +test_run
--
Regards,
Laurent Pinchart
