[Intel-gfx] [PATCH igt] lib/dummyload: Avoid assertions in lowlevel spin constructor

[Chris Wilson]

Quoting Antonio Argenziano (2018-02-22 21:24:31)
> 
> 
> On 22/02/18 00:50, Chris Wilson wrote:
> > __igt_spin_batch_new() may be used inside a background helper which is
> > competing against the GPU being reset. As such, we cannot even assert
> > that the spin->handle is busy immediately after submission as it may
> > have already been reset by another client writing to i915_wedged.
> > 
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >   lib/igt_dummyload.c | 17 ++++++++++++++---
> >   tests/pm_rps.c      |  4 +++-
> >   2 files changed, 17 insertions(+), 4 deletions(-)
> > 
> > diff --git a/lib/igt_dummyload.c b/lib/igt_dummyload.c
> > index 27eb402b..8ecd1dc8 100644
> > --- a/lib/igt_dummyload.c
> > +++ b/lib/igt_dummyload.c
> > @@ -25,6 +25,7 @@
> >   #include <time.h>
> >   #include <signal.h>
> >   #include <pthread.h>
> > +#include <sys/poll.h>
> >   
> >   #include <i915_drm.h>
> >   
> > @@ -205,7 +206,6 @@ ___igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep,
> >   
> >       spin->out_fence = emit_recursive_batch(spin, fd, ctx, engine, dep,
> >                                              out_fence);
> > -     igt_assert(gem_bo_busy(fd, spin->handle));
> >   
> >       pthread_mutex_lock(&list_lock);
> >       igt_list_add(&spin->link, &spin_list);
> > @@ -238,9 +238,14 @@ __igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep)
> >   igt_spin_t *
> >   igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep)
> >   {
> > +     igt_spin_t *spin;
> > +
> >       igt_require_gem(fd);
> >   
> > -     return __igt_spin_batch_new(fd, ctx, engine, dep);
> > +     spin = __igt_spin_batch_new(fd, ctx, engine, dep);
> > +     igt_assert(gem_bo_busy(fd, spin->handle));
> 
> Should all uses of the low level constructors be changed to this high 
> level one if the test doesn't have the race with i915_wedged?

No. There is the waiting side-effect as well which is more often the
problem.
-Chris