2014-05-14 08:02:16 -06:00
|
|
|
/*
|
|
|
|
* Copyright © 2014 Intel Corporation
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
* Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
|
|
* IN THE SOFTWARE.
|
|
|
|
*
|
|
|
|
* Authors:
|
|
|
|
* Mika Kuoppala <mika.kuoppala@intel.com>
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "i915_drv.h"
|
|
|
|
#include "intel_renderstate.h"
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
struct intel_render_state {
|
2016-08-02 15:50:36 -06:00
|
|
|
const struct intel_renderstate_rodata *rodata;
|
2016-08-15 03:49:03 -06:00
|
|
|
struct i915_vma *vma;
|
2016-10-28 06:58:31 -06:00
|
|
|
u32 batch_offset;
|
|
|
|
u32 batch_size;
|
|
|
|
u32 aux_offset;
|
|
|
|
u32 aux_size;
|
2016-08-02 15:50:36 -06:00
|
|
|
};
|
|
|
|
|
2014-05-14 08:02:16 -06:00
|
|
|
static const struct intel_renderstate_rodata *
|
2016-10-28 06:58:31 -06:00
|
|
|
render_state_get_rodata(const struct intel_engine_cs *engine)
|
2014-05-14 08:02:16 -06:00
|
|
|
{
|
2016-10-28 06:58:31 -06:00
|
|
|
switch (INTEL_GEN(engine->i915)) {
|
2014-05-14 08:02:16 -06:00
|
|
|
case 6:
|
|
|
|
return &gen6_null_state;
|
|
|
|
case 7:
|
|
|
|
return &gen7_null_state;
|
|
|
|
case 8:
|
|
|
|
return &gen8_null_state;
|
2014-10-23 09:34:28 -06:00
|
|
|
case 9:
|
|
|
|
return &gen9_null_state;
|
2014-05-14 08:02:16 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2015-07-20 03:46:10 -06:00
|
|
|
/*
|
|
|
|
* Macro to add commands to auxiliary batch.
|
|
|
|
* This macro only checks for page overflow before inserting the commands,
|
|
|
|
* this is sufficient as the null state generator makes the final batch
|
|
|
|
* with two passes to build command and state separately. At this point
|
|
|
|
* the size of both are known and it compacts them by relocating the state
|
|
|
|
* right after the commands taking care of aligment so we should sufficient
|
|
|
|
* space below them for adding new commands.
|
|
|
|
*/
|
|
|
|
#define OUT_BATCH(batch, i, val) \
|
|
|
|
do { \
|
2016-10-28 06:58:31 -06:00
|
|
|
if ((i) >= PAGE_SIZE / sizeof(u32)) \
|
|
|
|
goto err; \
|
2015-07-20 03:46:10 -06:00
|
|
|
(batch)[(i)++] = (val); \
|
|
|
|
} while(0)
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
static int render_state_setup(struct intel_render_state *so,
|
|
|
|
struct drm_i915_private *i915)
|
2014-06-10 04:23:33 -06:00
|
|
|
{
|
|
|
|
const struct intel_renderstate_rodata *rodata = so->rodata;
|
2016-10-28 06:58:31 -06:00
|
|
|
struct drm_i915_gem_object *obj = so->vma->obj;
|
2014-06-10 04:23:33 -06:00
|
|
|
unsigned int i = 0, reloc_index = 0;
|
2016-10-28 06:58:31 -06:00
|
|
|
unsigned int needs_clflush;
|
2014-06-10 04:23:33 -06:00
|
|
|
u32 *d;
|
|
|
|
int ret;
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
ret = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush);
|
2014-05-14 08:02:16 -06:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
d = kmap_atomic(i915_gem_object_get_dirty_page(obj, 0));
|
2014-06-10 04:23:33 -06:00
|
|
|
|
2014-05-14 08:02:16 -06:00
|
|
|
while (i < rodata->batch_items) {
|
|
|
|
u32 s = rodata->batch[i];
|
|
|
|
|
2014-06-10 04:23:33 -06:00
|
|
|
if (i * 4 == rodata->reloc[reloc_index]) {
|
2016-08-15 03:49:03 -06:00
|
|
|
u64 r = s + so->vma->node.start;
|
2014-06-10 04:23:33 -06:00
|
|
|
s = lower_32_bits(r);
|
2016-11-03 02:39:46 -06:00
|
|
|
if (HAS_64BIT_RELOC(i915)) {
|
2014-05-14 08:02:16 -06:00
|
|
|
if (i + 1 >= rodata->batch_items ||
|
2016-10-28 06:58:31 -06:00
|
|
|
rodata->batch[i + 1] != 0)
|
|
|
|
goto err;
|
2014-05-14 08:02:16 -06:00
|
|
|
|
2014-06-10 04:23:33 -06:00
|
|
|
d[i++] = s;
|
|
|
|
s = upper_32_bits(r);
|
2014-05-14 08:02:16 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
reloc_index++;
|
|
|
|
}
|
|
|
|
|
2014-06-10 04:23:33 -06:00
|
|
|
d[i++] = s;
|
2014-05-14 08:02:16 -06:00
|
|
|
}
|
2015-07-20 03:46:10 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (rodata->reloc[reloc_index] != -1) {
|
|
|
|
DRM_ERROR("only %d relocs resolved\n", reloc_index);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
so->batch_offset = so->vma->node.start;
|
|
|
|
so->batch_size = rodata->batch_items * sizeof(u32);
|
|
|
|
|
2015-07-20 03:46:10 -06:00
|
|
|
while (i % CACHELINE_DWORDS)
|
|
|
|
OUT_BATCH(d, i, MI_NOOP);
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
so->aux_offset = i * sizeof(u32);
|
2015-07-20 03:46:10 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (HAS_POOLED_EU(i915)) {
|
2016-06-02 23:34:33 -06:00
|
|
|
/*
|
|
|
|
* We always program 3x6 pool config but depending upon which
|
|
|
|
* subslice is disabled HW drops down to appropriate config
|
|
|
|
* shown below.
|
|
|
|
*
|
|
|
|
* In the below table 2x6 config always refers to
|
|
|
|
* fused-down version, native 2x6 is not available and can
|
|
|
|
* be ignored
|
|
|
|
*
|
|
|
|
* SNo subslices config eu pool configuration
|
|
|
|
* -----------------------------------------------------------
|
|
|
|
* 1 3 subslices enabled (3x6) - 0x00777000 (9+9)
|
|
|
|
* 2 ss0 disabled (2x6) - 0x00777000 (3+9)
|
|
|
|
* 3 ss1 disabled (2x6) - 0x00770000 (6+6)
|
|
|
|
* 4 ss2 disabled (2x6) - 0x00007000 (9+3)
|
|
|
|
*/
|
|
|
|
u32 eu_pool_config = 0x00777000;
|
|
|
|
|
|
|
|
OUT_BATCH(d, i, GEN9_MEDIA_POOL_STATE);
|
|
|
|
OUT_BATCH(d, i, GEN9_MEDIA_POOL_ENABLE);
|
|
|
|
OUT_BATCH(d, i, eu_pool_config);
|
|
|
|
OUT_BATCH(d, i, 0);
|
|
|
|
OUT_BATCH(d, i, 0);
|
|
|
|
OUT_BATCH(d, i, 0);
|
|
|
|
}
|
|
|
|
|
2015-07-20 03:46:10 -06:00
|
|
|
OUT_BATCH(d, i, MI_BATCH_BUFFER_END);
|
2016-10-28 06:58:31 -06:00
|
|
|
so->aux_size = i * sizeof(u32) - so->aux_offset;
|
|
|
|
so->aux_offset += so->batch_offset;
|
2015-07-20 03:46:10 -06:00
|
|
|
/*
|
|
|
|
* Since we are sending length, we need to strictly conform to
|
|
|
|
* all requirements. For Gen2 this must be a multiple of 8.
|
|
|
|
*/
|
2016-10-28 06:58:31 -06:00
|
|
|
so->aux_size = ALIGN(so->aux_size, 8);
|
2014-05-14 08:02:16 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (needs_clflush)
|
|
|
|
drm_clflush_virt_range(d, i * sizeof(u32));
|
|
|
|
kunmap_atomic(d);
|
2015-07-17 10:08:51 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
ret = i915_gem_object_set_to_gtt_domain(obj, false);
|
|
|
|
out:
|
|
|
|
i915_gem_obj_finish_shmem_access(obj);
|
2015-07-17 10:08:51 -06:00
|
|
|
return ret;
|
2016-10-28 06:58:31 -06:00
|
|
|
|
|
|
|
err:
|
|
|
|
kunmap_atomic(d);
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
2014-05-14 08:02:16 -06:00
|
|
|
}
|
|
|
|
|
2015-07-20 03:46:10 -06:00
|
|
|
#undef OUT_BATCH
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
int i915_gem_render_state_init(struct intel_engine_cs *engine)
|
2014-05-14 08:02:16 -06:00
|
|
|
{
|
2016-10-28 06:58:31 -06:00
|
|
|
struct intel_render_state *so;
|
|
|
|
const struct intel_renderstate_rodata *rodata;
|
2016-08-15 03:49:03 -06:00
|
|
|
struct drm_i915_gem_object *obj;
|
2014-05-14 08:02:16 -06:00
|
|
|
int ret;
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (engine->id != RCS)
|
|
|
|
return 0;
|
2014-05-21 10:01:06 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
rodata = render_state_get_rodata(engine);
|
|
|
|
if (!rodata)
|
2014-06-10 04:23:33 -06:00
|
|
|
return 0;
|
2014-05-14 08:02:16 -06:00
|
|
|
|
2017-01-10 07:47:34 -07:00
|
|
|
if (rodata->batch_items * 4 > PAGE_SIZE)
|
2016-08-02 15:50:37 -06:00
|
|
|
return -EINVAL;
|
2014-08-21 04:40:54 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
so = kmalloc(sizeof(*so), GFP_KERNEL);
|
|
|
|
if (!so)
|
|
|
|
return -ENOMEM;
|
2014-08-21 04:40:54 -06:00
|
|
|
|
2017-01-10 07:47:34 -07:00
|
|
|
obj = i915_gem_object_create_internal(engine->i915, PAGE_SIZE);
|
2016-10-28 06:58:31 -06:00
|
|
|
if (IS_ERR(obj)) {
|
|
|
|
ret = PTR_ERR(obj);
|
|
|
|
goto err_free;
|
2016-08-15 03:49:03 -06:00
|
|
|
}
|
2014-08-21 04:40:54 -06:00
|
|
|
|
2017-01-16 08:21:30 -07:00
|
|
|
so->vma = i915_vma_instance(obj, &engine->i915->ggtt.base, NULL);
|
2016-10-28 06:58:31 -06:00
|
|
|
if (IS_ERR(so->vma)) {
|
|
|
|
ret = PTR_ERR(so->vma);
|
2016-08-15 03:49:03 -06:00
|
|
|
goto err_obj;
|
2016-10-28 06:58:31 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
so->rodata = rodata;
|
|
|
|
engine->render_state = so;
|
|
|
|
return 0;
|
2016-08-02 15:50:37 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
err_obj:
|
|
|
|
i915_gem_object_put(obj);
|
|
|
|
err_free:
|
|
|
|
kfree(so);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int i915_gem_render_state_emit(struct drm_i915_gem_request *req)
|
|
|
|
{
|
|
|
|
struct intel_render_state *so;
|
|
|
|
int ret;
|
|
|
|
|
2016-10-28 06:58:32 -06:00
|
|
|
lockdep_assert_held(&req->i915->drm.struct_mutex);
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
so = req->engine->render_state;
|
|
|
|
if (!so)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Recreate the page after shrinking */
|
2016-10-28 06:58:35 -06:00
|
|
|
if (!so->vma->obj->mm.pages)
|
2016-10-28 06:58:31 -06:00
|
|
|
so->batch_offset = -1;
|
|
|
|
|
|
|
|
ret = i915_vma_pin(so->vma, 0, 0, PIN_GLOBAL | PIN_HIGH);
|
2016-08-02 15:50:37 -06:00
|
|
|
if (ret)
|
2016-10-28 06:58:31 -06:00
|
|
|
return ret;
|
2014-05-14 08:02:16 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (so->vma->node.start != so->batch_offset) {
|
|
|
|
ret = render_state_setup(so, req->i915);
|
|
|
|
if (ret)
|
|
|
|
goto err_unpin;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = req->engine->emit_bb_start(req,
|
|
|
|
so->batch_offset, so->batch_size,
|
2016-08-02 15:50:27 -06:00
|
|
|
I915_DISPATCH_SECURE);
|
2014-05-14 08:02:16 -06:00
|
|
|
if (ret)
|
2016-08-02 15:50:37 -06:00
|
|
|
goto err_unpin;
|
2014-05-14 08:02:16 -06:00
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
if (so->aux_size > 8) {
|
2016-08-02 15:50:27 -06:00
|
|
|
ret = req->engine->emit_bb_start(req,
|
2016-10-28 06:58:31 -06:00
|
|
|
so->aux_offset, so->aux_size,
|
2016-08-02 15:50:27 -06:00
|
|
|
I915_DISPATCH_SECURE);
|
2015-07-20 03:46:10 -06:00
|
|
|
if (ret)
|
2016-08-02 15:50:37 -06:00
|
|
|
goto err_unpin;
|
2015-07-20 03:46:10 -06:00
|
|
|
}
|
|
|
|
|
2016-10-28 06:58:31 -06:00
|
|
|
i915_vma_move_to_active(so->vma, req, 0);
|
2016-08-02 15:50:37 -06:00
|
|
|
err_unpin:
|
2016-10-28 06:58:31 -06:00
|
|
|
i915_vma_unpin(so->vma);
|
2014-05-14 08:02:16 -06:00
|
|
|
return ret;
|
|
|
|
}
|
2016-10-28 06:58:31 -06:00
|
|
|
|
|
|
|
void i915_gem_render_state_fini(struct intel_engine_cs *engine)
|
|
|
|
{
|
|
|
|
struct intel_render_state *so;
|
|
|
|
struct drm_i915_gem_object *obj;
|
|
|
|
|
|
|
|
so = fetch_and_zero(&engine->render_state);
|
|
|
|
if (!so)
|
|
|
|
return;
|
|
|
|
|
|
|
|
obj = so->vma->obj;
|
|
|
|
|
|
|
|
i915_vma_close(so->vma);
|
|
|
|
__i915_gem_object_release_unless_active(obj);
|
|
|
|
|
|
|
|
kfree(so);
|
|
|
|
}
|