[igt-dev] [PATCH i-g-t] lib/hang_ctx: Make use of dummyload library to create recursive batch

[Chris Wilson]

Quoting Antonio Argenziano (2018-07-09 21:20:00)
> 
> 
> On 09/07/18 12:53, Chris Wilson wrote:
> > Quoting Antonio Argenziano (2018-07-09 19:47:04)
> >> An hanging batch is nothing more than a spinning batch that never gets
> >> stopped, so re-use the routines implemented in dummyload.c.
> >>
> >> Signed-off-by: Antonio Argenziano <antonio.argenziano at intel.com>
> >> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> >> ---
> >>   lib/igt_gt.c | 57 ++++++++-------------------------------------------------
> >>   lib/igt_gt.h |  1 +
> >>   2 files changed, 9 insertions(+), 49 deletions(-)
> >>
> >> diff --git a/lib/igt_gt.c b/lib/igt_gt.c
> >> index 4569fd36..25b7f1c4 100644
> >> --- a/lib/igt_gt.c
> >> +++ b/lib/igt_gt.c
> >> @@ -264,14 +264,10 @@ igt_hang_t igt_hang_ctx(int fd,
> >>                          unsigned flags,
> >>                          uint64_t *offset)
> >>   {
> >> -       struct drm_i915_gem_relocation_entry reloc;
> >> -       struct drm_i915_gem_execbuffer2 execbuf;
> >> -       struct drm_i915_gem_exec_object2 exec;
> >> +       igt_spin_t *spinning_batch;
> >> +       struct igt_spin_factory opts = {};
> >>          struct drm_i915_gem_context_param param;
> >> -       uint32_t b[16];
> >>          unsigned ban;
> >> -       unsigned len;
> >> -       int gen;
> >>   
> >>          igt_require_hang_ring(fd, ring);
> >>   
> >> @@ -295,52 +291,15 @@ igt_hang_t igt_hang_ctx(int fd,
> >>          if ((flags & HANG_ALLOW_BAN) == 0)
> >>                  context_set_ban(fd, ctx, 0);
> >>   
> >> -       memset(&reloc, 0, sizeof(reloc));
> >> -       memset(&exec, 0, sizeof(exec));
> >> -       memset(&execbuf, 0, sizeof(execbuf));
> >> -
> >> -       exec.handle = gem_create(fd, 4096);
> >> -       exec.relocation_count = 1;
> >> -       exec.relocs_ptr = to_user_pointer(&reloc);
> >> -
> >> -       memset(b, 0xc5, sizeof(b));
> >> -
> >> -       len = 0;
> >> -       gen = intel_gen(intel_get_drm_devid(fd));
> >> -       if (gen >= 8) {
> >> -               b[len++] = MI_BATCH_BUFFER_START | 1 << 8 | 1;
> >> -               b[len++] = 0;
> >> -               b[len++] = 0;
> >> -       } else if (gen >= 6) {
> >> -               b[len++] = MI_BATCH_BUFFER_START | 1 << 8;
> >> -               b[len++] = 0;
> >> -       } else {
> >> -               b[len++] = MI_BATCH_BUFFER_START | 2 << 6;
> >> -               b[len] = 0;
> >> -               if (gen < 4) {
> >> -                       b[len] |= 1;
> >> -                       reloc.delta = 1;
> >> -               }
> >> -               len++;
> >> -       }
> >> -       b[len++] = MI_BATCH_BUFFER_END;
> >> -       b[len] = MI_NOOP;
> >> -       gem_write(fd, exec.handle, 0, b, sizeof(b));
> >> -
> >> -       reloc.offset = sizeof(uint32_t);
> >> -       reloc.target_handle = exec.handle;
> >> -       reloc.read_domains = I915_GEM_DOMAIN_COMMAND;
> >> -
> >> -       execbuf.buffers_ptr = to_user_pointer(&exec);
> >> -       execbuf.buffer_count = 1;
> >> -       execbuf.flags = ring;
> >> -       i915_execbuffer2_set_context_id(execbuf, ctx);
> >> -       gem_execbuf(fd, &execbuf);
> >> +       opts.ctx = ctx;
> >> +       opts.engine = ring;
> >> +       opts.flags = IGT_SPIN_FAST; /* Current driver hangs faster */
> > 
> > Be careful, the hanging batch is where we first encountered the
> > traumatic experience from spinning too fast. Use IGT_SPIN_FAST with
> > care, and probably best left to the caller's discretion.
> 
> I'll leave it to the caller to spin faster. I actually wanted to add 
> both hanging batches: fast and slow. My idea would be to have, even if 
> only basics, hang tests for all situations: Moving Head; Still HEAD; 
> Idle engine are the ones that come to mind. We could have a require() 
> for gen version if needed.

It's not terminal (anymore), just certain ops are slow; slow enough to
disrupt the test in some circumstances.
-Chris