On Wed, Nov 23, 2016 at 05:53:50PM +0200, Abdiel Janulgue wrote:
> +static int at_exit_drm_fd = -1;
> +
> +static void quiescent_gpu_at_exit(int sig)
> +{
> + if (at_exit_drm_fd < 0)
> + return;
> +
> + gem_quiescent_gpu(at_exit_drm_fd);
> + at_exit_drm_fd = -1;
> +}
> +
> +/**
> + * igt_spin_batch_new:
> + * @fd: open i915 drm file descriptor
> + * @engine: Ring to execute batch OR'd with execbuf flags. If value is less
> + * than 0, execute on all available rings.
> + * @dep_handle: handle to a buffer object dependency. If greater than 0, add a
> + * relocation entry to this buffer within the batch.
> + *
> + * Start a recursive batch on a ring. Immediately returns a #igt_spin_t that
> + * contains the batch's handle that can be waited upon. The returned structure
> + * must be passed to igt_spin_batch_free() for post-processing.
> + *
> + * Returns:
> + * Structure with helper internal state for igt_spin_batch_free().
> + */
> +igt_spin_t *
> +igt_spin_batch_new(int fd, int engine, unsigned dep_handle)
> +{
> + igt_spin_t *spin = calloc(1, sizeof(struct igt_spin));
> + uint32_t handle = emit_recursive_batch(fd, engine, dep_handle);
> + igt_assert(gem_bo_busy(fd, handle));
> + at_exit_drm_fd = fd;
> + igt_install_exit_handler(quiescent_gpu_at_exit);
We already do igt_install_exit_handler(quiescent_gpu_at_exit); doing a
plain one ourselves still incurs a GPU hang if the user hits ^C. (And we
are installing one for every new spinner.)
What I meant was that if we kept a list of active spinners and walked
that list from gem_quiescent_gpu() we could leverage the existing
infrastructure to ensure that the GPU was idled as kick as possible
after a failed/interrupted test (without GPU hangs causing havoc).
P.S. imagine running checkpatch.pl and keeping in the habit of using the
kernel CodingStyle.
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
