Skip to content

Commit

Permalink
drm/i915: Pull the unconditional GPU cache invalidation into request …
Browse files Browse the repository at this point in the history
…construction

As the request will, in the following patch, implicitly invoke a
context-switch on construction, we should precede that with a GPU TLB
invalidation. Also, even before using GGTT, we always want to invalidate
the TLBs for any updates (as well as the ppgtt invalidates that are
unconditionally applied by execbuf). Since we almost always require the
TLB invalidate, do it unconditionally on request allocation and so we can
remove it from all other paths.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171120102002.22254-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
  • Loading branch information
Chris Wilson committed Nov 20, 2017
1 parent 7c52a22 commit 2113184
Show file tree
Hide file tree
Showing 7 changed files with 20 additions and 37 deletions.
7 changes: 1 addition & 6 deletions drivers/gpu/drm/i915/i915_gem_execbuffer.c
Original file line number Diff line number Diff line change
Expand Up @@ -1111,10 +1111,6 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
if (err)
goto err_request;

err = eb->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto err_request;

err = i915_switch_context(rq);
if (err)
goto err_request;
Expand Down Expand Up @@ -1818,8 +1814,7 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
/* Unconditionally flush any chipset caches (for streaming writes). */
i915_gem_chipset_flush(eb->i915);

/* Unconditionally invalidate GPU caches and TLBs. */
return eb->engine->emit_flush(eb->request, EMIT_INVALIDATE);
return 0;
}

static bool i915_gem_check_execbuffer(struct drm_i915_gem_execbuffer2 *exec)
Expand Down
4 changes: 0 additions & 4 deletions drivers/gpu/drm/i915/i915_gem_render_state.c
Original file line number Diff line number Diff line change
Expand Up @@ -208,10 +208,6 @@ int i915_gem_render_state_emit(struct drm_i915_gem_request *rq)
if (err)
goto err_unpin;

err = engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto err_unpin;

err = engine->emit_bb_start(rq,
so.batch_offset, so.batch_size,
I915_DISPATCH_SECURE);
Expand Down
24 changes: 19 additions & 5 deletions drivers/gpu/drm/i915/i915_gem_request.c
Original file line number Diff line number Diff line change
Expand Up @@ -703,17 +703,31 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
req->reserved_space = MIN_SPACE_FOR_ADD_REQUEST;
GEM_BUG_ON(req->reserved_space < engine->emit_breadcrumb_sz);

ret = engine->request_alloc(req);
if (ret)
goto err_ctx;

/* Record the position of the start of the request so that
/*
* Record the position of the start of the request so that
* should we detect the updated seqno part-way through the
* GPU processing the request, we never over-estimate the
* position of the head.
*/
req->head = req->ring->emit;

/* Unconditionally invalidate GPU caches and TLBs. */
ret = engine->emit_flush(req, EMIT_INVALIDATE);
if (ret)
goto err_ctx;

ret = engine->request_alloc(req);
if (ret) {
/*
* Past the point-of-no-return. Since we may have updated
* global state after partially completing the request alloc,
* we need to commit any commands so far emitted in the
* request to the HW.
*/
__i915_add_request(req, false);
return ERR_PTR(ret);
}

/* Check that we didn't interrupt ourselves with a new request */
GEM_BUG_ON(req->timeline->seqno != req->fence.seqno);
return req;
Expand Down
4 changes: 0 additions & 4 deletions drivers/gpu/drm/i915/selftests/huge_pages.c
Original file line number Diff line number Diff line change
Expand Up @@ -989,10 +989,6 @@ static int gpu_write(struct i915_vma *vma,
i915_vma_unpin(batch);
i915_vma_close(batch);

err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto err_request;

err = i915_switch_context(rq);
if (err)
goto err_request;
Expand Down
4 changes: 0 additions & 4 deletions drivers/gpu/drm/i915/selftests/i915_gem_context.c
Original file line number Diff line number Diff line change
Expand Up @@ -158,10 +158,6 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
goto err_batch;
}

err = engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto err_request;

err = i915_switch_context(rq);
if (err)
goto err_request;
Expand Down
10 changes: 0 additions & 10 deletions drivers/gpu/drm/i915/selftests/i915_gem_request.c
Original file line number Diff line number Diff line change
Expand Up @@ -459,10 +459,6 @@ empty_request(struct intel_engine_cs *engine,
if (IS_ERR(request))
return request;

err = engine->emit_flush(request, EMIT_INVALIDATE);
if (err)
goto out_request;

err = i915_switch_context(request);
if (err)
goto out_request;
Expand Down Expand Up @@ -675,9 +671,6 @@ static int live_all_engines(void *arg)
goto out_request;
}

err = engine->emit_flush(request[id], EMIT_INVALIDATE);
GEM_BUG_ON(err);

err = i915_switch_context(request[id]);
GEM_BUG_ON(err);

Expand Down Expand Up @@ -797,9 +790,6 @@ static int live_sequential_engines(void *arg)
}
}

err = engine->emit_flush(request[id], EMIT_INVALIDATE);
GEM_BUG_ON(err);

err = i915_switch_context(request[id]);
GEM_BUG_ON(err);

Expand Down
4 changes: 0 additions & 4 deletions drivers/gpu/drm/i915/selftests/intel_hangcheck.c
Original file line number Diff line number Diff line change
Expand Up @@ -114,10 +114,6 @@ static int emit_recurse_batch(struct hang *h,
if (err)
goto unpin_vma;

err = rq->engine->emit_flush(rq, EMIT_INVALIDATE);
if (err)
goto unpin_hws;

err = i915_switch_context(rq);
if (err)
goto unpin_hws;
Expand Down

0 comments on commit 2113184

Please sign in to comment.