Hi all,
This patchset tries to address the seqno wrapping issues in driver.
When seqnos wrap, all requests need to be retired and this can fail.
Thus seqno_allocation is needed. We pre allocate seqnos prior to
pushing stuff into rings. Patch from Chris also avoids allocation
of seqno's when gpu_idle() is called for retiring. Without this
we would recurse back into next_request_seqno().
Comments, flames and reviews are all welcome.
Chris Wilson (1):
drm/i915: Wait upon the last request seqno, rather than a future
seqno
Mika Kuoppala (4):
drm/i915: fix debugfs seqno info print to use uint
drm/i915: Add i915_gem_alloc_seqno
drm/i915: Allocate seqno conditionally in i915_add_request
drm/i915: don't return seqno from ring->add_request()
drivers/gpu/drm/i915/i915_debugfs.c | 2 +-
drivers/gpu/drm/i915/i915_drv.h | 1 +
drivers/gpu/drm/i915/i915_gem.c | 102 ++++++++++++++++++++++++----
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 29 +++-----
drivers/gpu/drm/i915/intel_ringbuffer.c | 18 ++---
drivers/gpu/drm/i915/intel_ringbuffer.h | 2 +-
6 files changed, 108 insertions(+), 46 deletions(-)
--
1.7.9.5
