On Thu, Aug 26, 2021 at 08:37:19AM -0700, Rob Clark wrote:
> On Wed, Aug 25, 2021 at 10:28 PM Petri Latvala <petri.latvala@xxxxxxxxx> wrote:
> >
> > On Wed, Aug 25, 2021 at 04:31:39PM -0700, Rob Clark wrote:
> > > From: Rob Clark <robdclark@xxxxxxxxxxxx>
> > >
> > > Add an initial set of tests for the submit ioctl.
> > >
> > > Signed-off-by: Rob Clark <robdclark@xxxxxxxxxxxx>
> > > ---
> > > tests/meson.build | 1 +
> > > tests/msm_submit.c | 186 +++++++++++++++++++++++++++++++++++++++++++++
> > > 2 files changed, 187 insertions(+)
> > > create mode 100644 tests/msm_submit.c
> > >
> > > diff --git a/tests/meson.build b/tests/meson.build
> > > index 1bdfddbb..ff7c709a 100644
> > > --- a/tests/meson.build
> > > +++ b/tests/meson.build
> > > @@ -107,6 +107,7 @@ test_progs = [
> > > 'vc4_wait_seqno',
> > > 'vgem_basic',
> > > 'vgem_slow',
> > > + 'msm_submit',
> > > ]
> > >
> > > i915_progs = [
> > > diff --git a/tests/msm_submit.c b/tests/msm_submit.c
> > > new file mode 100644
> > > index 00000000..da93c574
> > > --- /dev/null
> > > +++ b/tests/msm_submit.c
> > > @@ -0,0 +1,186 @@
> > > +/*
> > > + * Copyright © 2021 Google, Inc.
> > > + *
> > > + * Permission is hereby granted, free of charge, to any person obtaining a
> > > + * copy of this software and associated documentation files (the "Software"),
> > > + * to deal in the Software without restriction, including without limitation
> > > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > > + * and/or sell copies of the Software, and to permit persons to whom the
> > > + * Software is furnished to do so, subject to the following conditions:
> > > + *
> > > + * The above copyright notice and this permission notice (including the next
> > > + * paragraph) shall be included in all copies or substantial portions of the
> > > + * Software.
> > > + *
> > > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> > > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> > > + * IN THE SOFTWARE.
> > > + */
> > > +
> > > +#include "igt.h"
> > > +#include "igt_msm.h"
> > > +
> > > +igt_main
> > > +{
> > > + struct msm_device *dev;
> > > + struct msm_pipe *pipe;
> > > + struct msm_bo *a, *b;
> > > +
> > > + igt_fixture {
> > > + dev = igt_msm_dev_open();
> >
> > What I replied on 2/3 applies here: If opening the device fails,
> > igt_msm_dev_open() does not return and 'dev' is left uninitialized,
> > those other pointers likewise. Leading to...
> >
> > > + pipe = igt_msm_pipe_open(dev, 0);
> > > + a = igt_msm_bo_new(dev, 0x1000, MSM_BO_WC);
> > > + b = igt_msm_bo_new(dev, 0x1000, MSM_BO_WC);
> > > + }
> > > +
> > > + igt_subtest("empty-submit") {
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = pipe->submitqueue_id,
> > > + };
> > > + do_ioctl(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req);
> > > + }
> > > +
> > > + igt_subtest("invalid-queue-submit") {
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = 0x1234,
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, ENOENT);
> > > + }
> > > +
> > > + igt_subtest("invalid-flags-submit") {
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = 0x1234,
> > > + .queueid = pipe->submitqueue_id,
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, EINVAL);
> > > + }
> > > +
> > > + igt_subtest("invalid-in-fence-submit") {
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe | MSM_SUBMIT_FENCE_FD_IN,
> > > + .queueid = pipe->submitqueue_id,
> > > + .fence_fd = dev->fd, /* This is not a fence fd! */
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, EINVAL);
> > > + }
> > > +
> > > + igt_subtest("invalid-duplicate-bo-submit") {
> > > + struct drm_msm_gem_submit_bo bos[] = {
> > > + [0] = {
> > > + .handle = a->handle,
> > > + .flags = MSM_SUBMIT_BO_READ,
> > > + },
> > > + [1] = {
> > > + .handle = b->handle,
> > > + .flags = MSM_SUBMIT_BO_READ,
> > > + },
> > > + [2] = {
> > > + /* this is invalid.. there should not be two entries
> > > + * for the same bo, instead a single entry w/ all
> > > + * usage flags OR'd together should be used. Kernel
> > > + * should catch this, and return an error code after
> > > + * cleaning up properly (not leaking any bo's)
> > > + */
> > > + .handle = a->handle,
> > > + .flags = MSM_SUBMIT_BO_WRITE,
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = pipe->submitqueue_id,
> > > + .nr_bos = ARRAY_SIZE(bos),
> > > + .bos = VOID2U64(bos),
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, EINVAL);
> > > + }
> > > +
> > > + igt_subtest("invalid-cmd-idx-submit") {
> > > + struct drm_msm_gem_submit_cmd cmds[] = {
> > > + [0] = {
> > > + .type = MSM_SUBMIT_CMD_BUF,
> > > + .submit_idx = 0, /* bos[0] does not exist */
> > > + .size = 4 * 4, /* 4 dwords in cmdbuf */
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = pipe->submitqueue_id,
> > > + .nr_cmds = ARRAY_SIZE(cmds),
> > > + .cmds = VOID2U64(cmds),
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, EINVAL);
> > > + }
> > > +
> > > + igt_subtest("invalid-cmd-type-submit") {
> > > + struct drm_msm_gem_submit_bo bos[] = {
> > > + [0] = {
> > > + .handle = a->handle,
> > > + .flags = MSM_SUBMIT_BO_READ,
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit_cmd cmds[] = {
> > > + [0] = {
> > > + .type = 0x1234,
> > > + .submit_idx = 0,
> > > + .size = 4 * 4, /* 4 dwords in cmdbuf */
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = pipe->submitqueue_id,
> > > + .nr_cmds = ARRAY_SIZE(cmds),
> > > + .cmds = VOID2U64(cmds),
> > > + .nr_bos = ARRAY_SIZE(bos),
> > > + .bos = VOID2U64(bos),
> > > + };
> > > + do_ioctl_err(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req, EINVAL);
> > > + }
> > > +
> > > + igt_subtest("valid-submit") {
> > > + struct drm_msm_gem_submit_bo bos[] = {
> > > + [0] = {
> > > + .handle = a->handle,
> > > + .flags = MSM_SUBMIT_BO_READ,
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit_cmd cmds[] = {
> > > + [0] = {
> > > + .type = MSM_SUBMIT_CMD_BUF,
> > > + .submit_idx = 0,
> > > + .size = 4 * 4, /* 4 dwords in cmdbuf */
> > > + },
> > > + };
> > > + struct drm_msm_gem_submit req = {
> > > + .flags = pipe->pipe,
> > > + .queueid = pipe->submitqueue_id,
> > > + .nr_cmds = ARRAY_SIZE(cmds),
> > > + .cmds = VOID2U64(cmds),
> > > + .nr_bos = ARRAY_SIZE(bos),
> > > + .bos = VOID2U64(bos),
> > > + };
> > > + uint32_t *cmdstream = igt_msm_bo_map(a);
> > > + if (dev->gen >= 5) {
> > > + *(cmdstream++) = pm4_pkt7_hdr(CP_NOP, 3);
> > > + } else {
> > > + *(cmdstream++) = pm4_pkt3_hdr(CP_NOP, 3);
> > > + }
> > > + *(cmdstream++) = 0;
> > > + *(cmdstream++) = 0;
> > > + *(cmdstream++) = 0;
> > > +
> > > + do_ioctl(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req);
> > > + }
> > > +
> > > + igt_fixture {
> > > + igt_msm_bo_free(a);
> > > + igt_msm_bo_free(b);
> > > + igt_msm_pipe_close(pipe);
> > > + igt_msm_dev_close(dev);
> >
> > ... crashes in here.
> >
>
> I did test this on intel as well, and it skips properly.. I think the
> setjmp/longjmp magic just bails completely out so we never try to run
> the cleanup?
Ah, indeed, skip_henceforth prevents entering the latter fixture.
It still looks scary, can I implore to add some defensive layers with
NULL-inits and handling of NULLs in the functions? It might not crash
today...
--
Petri Latvala
