2010-05-22 04:26:39 +08:00
|
|
|
/*
|
|
|
|
* Copyright © 2008-2010 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:
|
|
|
|
* Eric Anholt <eric@anholt.net>
|
|
|
|
* Zou Nan hai <nanhai.zou@intel.com>
|
|
|
|
* Xiang Hai hao<haihao.xiang@intel.com>
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2012-10-03 01:01:07 +08:00
|
|
|
#include <drm/drmP.h>
|
2010-05-22 04:26:39 +08:00
|
|
|
#include "i915_drv.h"
|
2012-10-03 01:01:07 +08:00
|
|
|
#include <drm/i915_drm.h>
|
2010-05-22 04:26:39 +08:00
|
|
|
#include "i915_trace.h"
|
2010-09-19 21:40:43 +08:00
|
|
|
#include "intel_drv.h"
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2011-10-16 16:23:31 +08:00
|
|
|
/*
|
|
|
|
* 965+ support PIPE_CONTROL commands, which provide finer grained control
|
|
|
|
* over cache flushing.
|
|
|
|
*/
|
|
|
|
struct pipe_control {
|
|
|
|
struct drm_i915_gem_object *obj;
|
|
|
|
volatile u32 *cpu_page;
|
|
|
|
u32 gtt_offset;
|
|
|
|
};
|
|
|
|
|
2011-01-21 01:00:10 +08:00
|
|
|
static inline int ring_space(struct intel_ring_buffer *ring)
|
|
|
|
{
|
2012-12-04 00:43:32 +08:00
|
|
|
int space = (ring->head & HEAD_ADDR) - (ring->tail + I915_RING_FREE_SPACE);
|
2011-01-21 01:00:10 +08:00
|
|
|
if (space < 0)
|
|
|
|
space += ring->size;
|
|
|
|
return space;
|
|
|
|
}
|
|
|
|
|
2011-01-05 01:34:02 +08:00
|
|
|
static int
|
2012-04-18 18:12:11 +08:00
|
|
|
gen2_render_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate_domains,
|
|
|
|
u32 flush_domains)
|
|
|
|
{
|
|
|
|
u32 cmd;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
cmd = MI_FLUSH;
|
2012-04-19 22:45:22 +08:00
|
|
|
if (((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER) == 0)
|
2012-04-18 18:12:11 +08:00
|
|
|
cmd |= MI_NO_WRITE_FLUSH;
|
|
|
|
|
|
|
|
if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER)
|
|
|
|
cmd |= MI_READ_FLUSH;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, cmd);
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
gen4_render_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate_domains,
|
|
|
|
u32 flush_domains)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2010-08-07 18:01:22 +08:00
|
|
|
u32 cmd;
|
2011-01-05 01:34:02 +08:00
|
|
|
int ret;
|
2010-08-07 18:01:22 +08:00
|
|
|
|
2011-03-20 06:26:49 +08:00
|
|
|
/*
|
|
|
|
* read/write caches:
|
|
|
|
*
|
|
|
|
* I915_GEM_DOMAIN_RENDER is always invalidated, but is
|
|
|
|
* only flushed if MI_NO_WRITE_FLUSH is unset. On 965, it is
|
|
|
|
* also flushed at 2d versus 3d pipeline switches.
|
|
|
|
*
|
|
|
|
* read-only caches:
|
|
|
|
*
|
|
|
|
* I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if
|
|
|
|
* MI_READ_FLUSH is set, and is always flushed on 965.
|
|
|
|
*
|
|
|
|
* I915_GEM_DOMAIN_COMMAND may not exist?
|
|
|
|
*
|
|
|
|
* I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is
|
|
|
|
* invalidated when MI_EXE_FLUSH is set.
|
|
|
|
*
|
|
|
|
* I915_GEM_DOMAIN_VERTEX, which exists on 965, is
|
|
|
|
* invalidated with every MI_FLUSH.
|
|
|
|
*
|
|
|
|
* TLBs:
|
|
|
|
*
|
|
|
|
* On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND
|
|
|
|
* and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and
|
|
|
|
* I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER
|
|
|
|
* are flushed at any MI_FLUSH.
|
|
|
|
*/
|
|
|
|
|
|
|
|
cmd = MI_FLUSH | MI_NO_WRITE_FLUSH;
|
2012-04-18 18:12:11 +08:00
|
|
|
if ((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER)
|
2011-03-20 06:26:49 +08:00
|
|
|
cmd &= ~MI_NO_WRITE_FLUSH;
|
|
|
|
if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION)
|
|
|
|
cmd |= MI_EXE_FLUSH;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2011-03-20 06:26:49 +08:00
|
|
|
if (invalidate_domains & I915_GEM_DOMAIN_COMMAND &&
|
|
|
|
(IS_G4X(dev) || IS_GEN5(dev)))
|
|
|
|
cmd |= MI_INVALIDATE_ISP;
|
2010-11-30 22:07:47 +08:00
|
|
|
|
2011-03-20 06:26:49 +08:00
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2011-01-05 01:34:02 +08:00
|
|
|
|
2011-03-20 06:26:49 +08:00
|
|
|
intel_ring_emit(ring, cmd);
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
2011-01-05 01:34:02 +08:00
|
|
|
|
|
|
|
return 0;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
|
|
|
|
2011-10-16 16:23:31 +08:00
|
|
|
/**
|
|
|
|
* Emits a PIPE_CONTROL with a non-zero post-sync operation, for
|
|
|
|
* implementing two workarounds on gen6. From section 1.4.7.1
|
|
|
|
* "PIPE_CONTROL" of the Sandy Bridge PRM volume 2 part 1:
|
|
|
|
*
|
|
|
|
* [DevSNB-C+{W/A}] Before any depth stall flush (including those
|
|
|
|
* produced by non-pipelined state commands), software needs to first
|
|
|
|
* send a PIPE_CONTROL with no bits set except Post-Sync Operation !=
|
|
|
|
* 0.
|
|
|
|
*
|
|
|
|
* [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache Flush Enable
|
|
|
|
* =1, a PIPE_CONTROL with any non-zero post-sync-op is required.
|
|
|
|
*
|
|
|
|
* And the workaround for these two requires this workaround first:
|
|
|
|
*
|
|
|
|
* [Dev-SNB{W/A}]: Pipe-control with CS-stall bit set must be sent
|
|
|
|
* BEFORE the pipe-control with a post-sync op and no write-cache
|
|
|
|
* flushes.
|
|
|
|
*
|
|
|
|
* And this last workaround is tricky because of the requirements on
|
|
|
|
* that bit. From section 1.4.7.2.3 "Stall" of the Sandy Bridge PRM
|
|
|
|
* volume 2 part 1:
|
|
|
|
*
|
|
|
|
* "1 of the following must also be set:
|
|
|
|
* - Render Target Cache Flush Enable ([12] of DW1)
|
|
|
|
* - Depth Cache Flush Enable ([0] of DW1)
|
|
|
|
* - Stall at Pixel Scoreboard ([1] of DW1)
|
|
|
|
* - Depth Stall ([13] of DW1)
|
|
|
|
* - Post-Sync Operation ([13] of DW1)
|
|
|
|
* - Notify Enable ([8] of DW1)"
|
|
|
|
*
|
|
|
|
* The cache flushes require the workaround flush that triggered this
|
|
|
|
* one, so we can't use it. Depth stall would trigger the same.
|
|
|
|
* Post-sync nonzero is what triggered this second workaround, so we
|
|
|
|
* can't use that one either. Notify enable is IRQs, which aren't
|
|
|
|
* really our business. That leaves only stall at scoreboard.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
intel_emit_post_sync_nonzero_flush(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
u32 scratch_addr = pc->gtt_offset + 128;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 6);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5));
|
|
|
|
intel_ring_emit(ring, PIPE_CONTROL_CS_STALL |
|
|
|
|
PIPE_CONTROL_STALL_AT_SCOREBOARD);
|
|
|
|
intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */
|
|
|
|
intel_ring_emit(ring, 0); /* low dword */
|
|
|
|
intel_ring_emit(ring, 0); /* high dword */
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 6);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5));
|
|
|
|
intel_ring_emit(ring, PIPE_CONTROL_QW_WRITE);
|
|
|
|
intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
gen6_render_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate_domains, u32 flush_domains)
|
|
|
|
{
|
|
|
|
u32 flags = 0;
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
u32 scratch_addr = pc->gtt_offset + 128;
|
|
|
|
int ret;
|
|
|
|
|
2012-08-18 05:35:42 +08:00
|
|
|
/* Force SNB workarounds for PIPE_CONTROL flushes */
|
|
|
|
ret = intel_emit_post_sync_nonzero_flush(ring);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2011-10-16 16:23:31 +08:00
|
|
|
/* Just flush everything. Experiments have shown that reducing the
|
|
|
|
* number of bits based on the write domains has little performance
|
|
|
|
* impact.
|
|
|
|
*/
|
2012-08-10 17:18:10 +08:00
|
|
|
if (flush_domains) {
|
|
|
|
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
|
|
|
|
flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
|
|
|
|
/*
|
|
|
|
* Ensure that any following seqno writes only happen
|
|
|
|
* when the render cache is indeed flushed.
|
|
|
|
*/
|
2012-06-28 15:48:42 +08:00
|
|
|
flags |= PIPE_CONTROL_CS_STALL;
|
2012-08-10 17:18:10 +08:00
|
|
|
}
|
|
|
|
if (invalidate_domains) {
|
|
|
|
flags |= PIPE_CONTROL_TLB_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;
|
|
|
|
/*
|
|
|
|
* TLB invalidate requires a post-sync write.
|
|
|
|
*/
|
2012-10-26 03:15:47 +08:00
|
|
|
flags |= PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_CS_STALL;
|
2012-08-10 17:18:10 +08:00
|
|
|
}
|
2011-10-16 16:23:31 +08:00
|
|
|
|
2012-07-21 01:02:28 +08:00
|
|
|
ret = intel_ring_begin(ring, 4);
|
2011-10-16 16:23:31 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2012-07-21 01:02:28 +08:00
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4));
|
2011-10-16 16:23:31 +08:00
|
|
|
intel_ring_emit(ring, flags);
|
|
|
|
intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT);
|
2012-07-21 01:02:28 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
2011-10-16 16:23:31 +08:00
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
drm/i915: add workarounds to gen7_render_ring_flush
From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section
detailing the various workarounds:
"[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit
set must be issued before a pipe-control command that has the State
Cache Invalidate bit set."
Note that public Bspec has different numbering, it's Vol2Part1,
Section 1.10.4.1 "PIPE_CONTROL" there.
There's also a second workaround for the PIPE_CONTROL command itself:
"[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not
counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set,
must have a CS_STALL bit set"
For simplicity we simply set the CS_STALL bit on every pipe_control on
gen7+
Note that this massively helps on some hsw machines, together with the
following patch to unconditionally set the CS_STALL bit on every
pipe_control it prevents a gpu hang every few seconds.
This is a regression that has been introduced in the pipe_control
cleanup:
commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Fri Jul 20 18:02:28 2012 +0100
drm/i915: Only apply the SNB pipe control w/a to gen6
It looks like the massive snb pipe_control workaround also papered
over any issues on ivb and hsw.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: squashed both workarounds together, pimped commit message
with Bsepc citations, regression commit citation and changed the
comment in the code a bit to clarify that we unconditionally set
CS_STALL to avoid being hurt by trying to be clever.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-18 05:35:43 +08:00
|
|
|
static int
|
|
|
|
gen7_render_ring_cs_stall_wa(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 4);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4));
|
|
|
|
intel_ring_emit(ring, PIPE_CONTROL_CS_STALL |
|
|
|
|
PIPE_CONTROL_STALL_AT_SCOREBOARD);
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-08-18 05:35:41 +08:00
|
|
|
static int
|
|
|
|
gen7_render_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate_domains, u32 flush_domains)
|
|
|
|
{
|
|
|
|
u32 flags = 0;
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
u32 scratch_addr = pc->gtt_offset + 128;
|
|
|
|
int ret;
|
|
|
|
|
drm/i915: add workarounds to gen7_render_ring_flush
From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section
detailing the various workarounds:
"[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit
set must be issued before a pipe-control command that has the State
Cache Invalidate bit set."
Note that public Bspec has different numbering, it's Vol2Part1,
Section 1.10.4.1 "PIPE_CONTROL" there.
There's also a second workaround for the PIPE_CONTROL command itself:
"[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not
counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set,
must have a CS_STALL bit set"
For simplicity we simply set the CS_STALL bit on every pipe_control on
gen7+
Note that this massively helps on some hsw machines, together with the
following patch to unconditionally set the CS_STALL bit on every
pipe_control it prevents a gpu hang every few seconds.
This is a regression that has been introduced in the pipe_control
cleanup:
commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Fri Jul 20 18:02:28 2012 +0100
drm/i915: Only apply the SNB pipe control w/a to gen6
It looks like the massive snb pipe_control workaround also papered
over any issues on ivb and hsw.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: squashed both workarounds together, pimped commit message
with Bsepc citations, regression commit citation and changed the
comment in the code a bit to clarify that we unconditionally set
CS_STALL to avoid being hurt by trying to be clever.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-18 05:35:43 +08:00
|
|
|
/*
|
|
|
|
* Ensure that any following seqno writes only happen when the render
|
|
|
|
* cache is indeed flushed.
|
|
|
|
*
|
|
|
|
* Workaround: 4th PIPE_CONTROL command (except the ones with only
|
|
|
|
* read-cache invalidate bits set) must have the CS_STALL bit set. We
|
|
|
|
* don't try to be clever and just set it unconditionally.
|
|
|
|
*/
|
|
|
|
flags |= PIPE_CONTROL_CS_STALL;
|
|
|
|
|
2012-08-18 05:35:41 +08:00
|
|
|
/* Just flush everything. Experiments have shown that reducing the
|
|
|
|
* number of bits based on the write domains has little performance
|
|
|
|
* impact.
|
|
|
|
*/
|
|
|
|
if (flush_domains) {
|
|
|
|
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
|
|
|
|
flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
|
|
|
|
}
|
|
|
|
if (invalidate_domains) {
|
|
|
|
flags |= PIPE_CONTROL_TLB_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
|
|
|
|
flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;
|
|
|
|
/*
|
|
|
|
* TLB invalidate requires a post-sync write.
|
|
|
|
*/
|
|
|
|
flags |= PIPE_CONTROL_QW_WRITE;
|
2013-02-15 03:53:51 +08:00
|
|
|
flags |= PIPE_CONTROL_GLOBAL_GTT_IVB;
|
drm/i915: add workarounds to gen7_render_ring_flush
From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section
detailing the various workarounds:
"[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit
set must be issued before a pipe-control command that has the State
Cache Invalidate bit set."
Note that public Bspec has different numbering, it's Vol2Part1,
Section 1.10.4.1 "PIPE_CONTROL" there.
There's also a second workaround for the PIPE_CONTROL command itself:
"[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not
counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set,
must have a CS_STALL bit set"
For simplicity we simply set the CS_STALL bit on every pipe_control on
gen7+
Note that this massively helps on some hsw machines, together with the
following patch to unconditionally set the CS_STALL bit on every
pipe_control it prevents a gpu hang every few seconds.
This is a regression that has been introduced in the pipe_control
cleanup:
commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Fri Jul 20 18:02:28 2012 +0100
drm/i915: Only apply the SNB pipe control w/a to gen6
It looks like the massive snb pipe_control workaround also papered
over any issues on ivb and hsw.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: squashed both workarounds together, pimped commit message
with Bsepc citations, regression commit citation and changed the
comment in the code a bit to clarify that we unconditionally set
CS_STALL to avoid being hurt by trying to be clever.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-18 05:35:43 +08:00
|
|
|
|
|
|
|
/* Workaround: we must issue a pipe_control with CS-stall bit
|
|
|
|
* set before a pipe_control command that has the state cache
|
|
|
|
* invalidate bit set. */
|
|
|
|
gen7_render_ring_cs_stall_wa(ring);
|
2012-08-18 05:35:41 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 4);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4));
|
|
|
|
intel_ring_emit(ring, flags);
|
2013-02-15 03:53:51 +08:00
|
|
|
intel_ring_emit(ring, scratch_addr);
|
2012-08-18 05:35:41 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static void ring_write_tail(struct intel_ring_buffer *ring,
|
2010-10-23 00:02:41 +08:00
|
|
|
u32 value)
|
2010-09-16 10:43:12 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
drm_i915_private_t *dev_priv = ring->dev->dev_private;
|
2010-10-23 00:02:41 +08:00
|
|
|
I915_WRITE_TAIL(ring, value);
|
2010-09-16 10:43:12 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
u32 intel_ring_get_active_head(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
drm_i915_private_t *dev_priv = ring->dev->dev_private;
|
|
|
|
u32 acthd_reg = INTEL_INFO(ring->dev)->gen >= 4 ?
|
2010-09-25 03:14:22 +08:00
|
|
|
RING_ACTHD(ring->mmio_base) : ACTHD;
|
2010-05-21 09:08:55 +08:00
|
|
|
|
|
|
|
return I915_READ(acthd_reg);
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static int init_ring_common(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2012-06-04 17:18:15 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2010-11-09 03:18:58 +08:00
|
|
|
struct drm_i915_gem_object *obj = ring->obj;
|
2012-06-04 17:18:15 +08:00
|
|
|
int ret = 0;
|
2010-05-21 09:08:55 +08:00
|
|
|
u32 head;
|
|
|
|
|
2012-06-04 17:18:15 +08:00
|
|
|
if (HAS_FORCE_WAKE(dev))
|
|
|
|
gen6_gt_force_wake_get(dev_priv);
|
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
/* Stop the ring if it's running. */
|
2010-08-02 23:06:59 +08:00
|
|
|
I915_WRITE_CTL(ring, 0);
|
2010-08-02 23:06:23 +08:00
|
|
|
I915_WRITE_HEAD(ring, 0);
|
2010-10-27 19:18:21 +08:00
|
|
|
ring->write_tail(ring, 0);
|
2010-05-21 09:08:55 +08:00
|
|
|
|
2010-08-02 23:06:23 +08:00
|
|
|
head = I915_READ_HEAD(ring) & HEAD_ADDR;
|
2010-05-21 09:08:55 +08:00
|
|
|
|
|
|
|
/* G45 ring initialization fails to reset head to zero */
|
|
|
|
if (head != 0) {
|
2010-12-06 04:42:33 +08:00
|
|
|
DRM_DEBUG_KMS("%s head not reset to zero "
|
|
|
|
"ctl %08x head %08x tail %08x start %08x\n",
|
|
|
|
ring->name,
|
|
|
|
I915_READ_CTL(ring),
|
|
|
|
I915_READ_HEAD(ring),
|
|
|
|
I915_READ_TAIL(ring),
|
|
|
|
I915_READ_START(ring));
|
2010-05-21 09:08:55 +08:00
|
|
|
|
2010-08-02 23:06:23 +08:00
|
|
|
I915_WRITE_HEAD(ring, 0);
|
2010-05-21 09:08:55 +08:00
|
|
|
|
2010-12-06 04:42:33 +08:00
|
|
|
if (I915_READ_HEAD(ring) & HEAD_ADDR) {
|
|
|
|
DRM_ERROR("failed to set %s head to zero "
|
|
|
|
"ctl %08x head %08x tail %08x start %08x\n",
|
|
|
|
ring->name,
|
|
|
|
I915_READ_CTL(ring),
|
|
|
|
I915_READ_HEAD(ring),
|
|
|
|
I915_READ_TAIL(ring),
|
|
|
|
I915_READ_START(ring));
|
|
|
|
}
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
|
|
|
|
2012-08-07 15:54:14 +08:00
|
|
|
/* Initialize the ring. This must happen _after_ we've cleared the ring
|
|
|
|
* registers with the above sequence (the readback of the HEAD registers
|
|
|
|
* also enforces ordering), otherwise the hw might lose the new ring
|
|
|
|
* register values. */
|
|
|
|
I915_WRITE_START(ring, obj->gtt_offset);
|
2010-08-02 23:06:59 +08:00
|
|
|
I915_WRITE_CTL(ring,
|
2010-11-07 19:45:52 +08:00
|
|
|
((ring->size - PAGE_SIZE) & RING_NR_PAGES)
|
drm/i915: Remove use of the autoreported ringbuffer HEAD position
This is a revert of 6aa56062eaba67adfb247cded244fd877329588d.
This was originally introduced to workaround reads of the ringbuffer
registers returning 0 on SandyBridge causing hangs due to ringbuffer
overflow. The root cause here was reads through the GT powerwell require
the forcewake dance, something we only learnt of later. Now it appears
that reading the reported head position from the HWS is returning
garbage, leading once again to hangs.
For example, on q35 the autoreported head reports:
[ 217.975608] head now 00010000, actual 00010000
[ 436.725613] head now 00200000, actual 00200000
[ 462.956033] head now 00210000, actual 00210010
[ 485.501409] head now 00400000, actual 00400020
[ 508.064280] head now 00410000, actual 00410000
[ 530.576078] head now 00600000, actual 00600020
[ 553.273489] head now 00610000, actual 00610018
which appears reasonably sane. In contrast, if we look at snb:
[ 141.970680] head now 00e10000, actual 00008238
[ 141.974062] head now 02734000, actual 000083c8
[ 141.974425] head now 00e10000, actual 00008488
[ 141.980374] head now 032b5000, actual 000088b8
[ 141.980885] head now 03271000, actual 00008950
[ 142.040628] head now 02101000, actual 00008b40
[ 142.180173] head now 02734000, actual 00009050
[ 142.181090] head now 00000000, actual 00000ae0
[ 142.183737] head now 02734000, actual 00009050
In addition, the automatic reporting of the head position is scheduled
to be defeatured in the future. It has no more utility, remove it.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45492
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
2012-02-08 21:34:13 +08:00
|
|
|
| RING_VALID);
|
2010-05-21 09:08:55 +08:00
|
|
|
|
|
|
|
/* If the head is still not zero, the ring is dead */
|
2012-03-17 00:43:22 +08:00
|
|
|
if (wait_for((I915_READ_CTL(ring) & RING_VALID) != 0 &&
|
|
|
|
I915_READ_START(ring) == obj->gtt_offset &&
|
|
|
|
(I915_READ_HEAD(ring) & HEAD_ADDR) == 0, 50)) {
|
2010-11-09 18:16:56 +08:00
|
|
|
DRM_ERROR("%s initialization failed "
|
|
|
|
"ctl %08x head %08x tail %08x start %08x\n",
|
|
|
|
ring->name,
|
|
|
|
I915_READ_CTL(ring),
|
|
|
|
I915_READ_HEAD(ring),
|
|
|
|
I915_READ_TAIL(ring),
|
|
|
|
I915_READ_START(ring));
|
2012-06-04 17:18:15 +08:00
|
|
|
ret = -EIO;
|
|
|
|
goto out;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
if (!drm_core_check_feature(ring->dev, DRIVER_MODESET))
|
|
|
|
i915_kernel_lost_context(ring->dev);
|
2010-05-21 09:08:55 +08:00
|
|
|
else {
|
2011-01-21 01:00:10 +08:00
|
|
|
ring->head = I915_READ_HEAD(ring);
|
2010-08-02 22:29:44 +08:00
|
|
|
ring->tail = I915_READ_TAIL(ring) & TAIL_ADDR;
|
2011-01-21 01:00:10 +08:00
|
|
|
ring->space = ring_space(ring);
|
2012-05-29 05:33:02 +08:00
|
|
|
ring->last_retired_head = -1;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
2010-12-04 19:30:53 +08:00
|
|
|
|
2012-06-04 17:18:15 +08:00
|
|
|
out:
|
|
|
|
if (HAS_FORCE_WAKE(dev))
|
|
|
|
gen6_gt_force_wake_put(dev_priv);
|
|
|
|
|
|
|
|
return ret;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
static int
|
|
|
|
init_pipe_control(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct pipe_control *pc;
|
|
|
|
struct drm_i915_gem_object *obj;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (ring->private)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
pc = kmalloc(sizeof(*pc), GFP_KERNEL);
|
|
|
|
if (!pc)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
obj = i915_gem_alloc_object(ring->dev, 4096);
|
|
|
|
if (obj == NULL) {
|
|
|
|
DRM_ERROR("Failed to allocate seqno page\n");
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto err;
|
|
|
|
}
|
2011-04-04 16:44:39 +08:00
|
|
|
|
|
|
|
i915_gem_object_set_cache_level(obj, I915_CACHE_LLC);
|
2010-12-15 17:56:50 +08:00
|
|
|
|
2012-08-11 22:41:04 +08:00
|
|
|
ret = i915_gem_object_pin(obj, 4096, true, false);
|
2010-12-15 17:56:50 +08:00
|
|
|
if (ret)
|
|
|
|
goto err_unref;
|
|
|
|
|
|
|
|
pc->gtt_offset = obj->gtt_offset;
|
2013-05-28 17:51:44 +08:00
|
|
|
pc->cpu_page = kmap(sg_page(obj->pages->sgl));
|
|
|
|
if (pc->cpu_page == NULL) {
|
|
|
|
ret = -ENOMEM;
|
2010-12-15 17:56:50 +08:00
|
|
|
goto err_unpin;
|
2013-05-28 17:51:44 +08:00
|
|
|
}
|
2010-12-15 17:56:50 +08:00
|
|
|
|
2013-02-13 04:01:38 +08:00
|
|
|
DRM_DEBUG_DRIVER("%s pipe control offset: 0x%08x\n",
|
|
|
|
ring->name, pc->gtt_offset);
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
pc->obj = obj;
|
|
|
|
ring->private = pc;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err_unpin:
|
|
|
|
i915_gem_object_unpin(obj);
|
|
|
|
err_unref:
|
|
|
|
drm_gem_object_unreference(&obj->base);
|
|
|
|
err:
|
|
|
|
kfree(pc);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
cleanup_pipe_control(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
struct drm_i915_gem_object *obj;
|
|
|
|
|
|
|
|
if (!ring->private)
|
|
|
|
return;
|
|
|
|
|
|
|
|
obj = pc->obj;
|
2012-06-01 22:20:22 +08:00
|
|
|
|
|
|
|
kunmap(sg_page(obj->pages->sgl));
|
2010-12-15 17:56:50 +08:00
|
|
|
i915_gem_object_unpin(obj);
|
|
|
|
drm_gem_object_unreference(&obj->base);
|
|
|
|
|
|
|
|
kfree(pc);
|
|
|
|
ring->private = NULL;
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static int init_render_ring(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2010-12-04 19:30:53 +08:00
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
2010-10-27 19:18:21 +08:00
|
|
|
int ret = init_ring_common(ring);
|
2010-08-30 16:12:42 +08:00
|
|
|
|
2013-01-21 00:11:20 +08:00
|
|
|
if (INTEL_INFO(dev)->gen > 3)
|
2012-04-24 20:04:12 +08:00
|
|
|
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH));
|
2013-01-21 00:11:20 +08:00
|
|
|
|
|
|
|
/* We need to disable the AsyncFlip performance optimisations in order
|
|
|
|
* to use MI_WAIT_FOR_EVENT within the CS. It should already be
|
|
|
|
* programmed to '1' on all products.
|
2013-05-04 01:48:11 +08:00
|
|
|
*
|
|
|
|
* WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv
|
2013-01-21 00:11:20 +08:00
|
|
|
*/
|
|
|
|
if (INTEL_INFO(dev)->gen >= 6)
|
|
|
|
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(ASYNC_FLIP_PERF_DISABLE));
|
|
|
|
|
2013-01-21 00:33:32 +08:00
|
|
|
/* Required for the hardware to program scanline values for waiting */
|
|
|
|
if (INTEL_INFO(dev)->gen == 6)
|
|
|
|
I915_WRITE(GFX_MODE,
|
|
|
|
_MASKED_BIT_ENABLE(GFX_TLB_INVALIDATE_ALWAYS));
|
|
|
|
|
2013-01-21 00:11:20 +08:00
|
|
|
if (IS_GEN7(dev))
|
|
|
|
I915_WRITE(GFX_MODE_GEN7,
|
|
|
|
_MASKED_BIT_DISABLE(GFX_TLB_INVALIDATE_ALWAYS) |
|
|
|
|
_MASKED_BIT_ENABLE(GFX_REPLAY_MODE));
|
2010-10-27 19:18:21 +08:00
|
|
|
|
2011-10-16 16:23:31 +08:00
|
|
|
if (INTEL_INFO(dev)->gen >= 5) {
|
2010-12-15 17:56:50 +08:00
|
|
|
ret = init_pipe_control(ring);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2012-05-08 19:39:59 +08:00
|
|
|
if (IS_GEN6(dev)) {
|
2012-04-28 03:44:41 +08:00
|
|
|
/* From the Sandybridge PRM, volume 1 part 3, page 24:
|
|
|
|
* "If this bit is set, STCunit will have LRA as replacement
|
|
|
|
* policy. [...] This bit must be reset. LRA replacement
|
|
|
|
* policy is not supported."
|
|
|
|
*/
|
|
|
|
I915_WRITE(CACHE_MODE_0,
|
2012-05-08 19:39:59 +08:00
|
|
|
_MASKED_BIT_DISABLE(CM0_STC_EVICT_DISABLE_LRA_SNB));
|
2012-06-05 05:42:50 +08:00
|
|
|
|
|
|
|
/* This is not explicitly set for GEN6, so read the register.
|
|
|
|
* see intel_ring_mi_set_context() for why we care.
|
|
|
|
* TODO: consider explicitly setting the bit for GEN5
|
|
|
|
*/
|
|
|
|
ring->itlb_before_ctx_switch =
|
|
|
|
!!(I915_READ(GFX_MODE) & GFX_TLB_INVALIDATE_ALWAYS);
|
2011-12-13 11:21:58 +08:00
|
|
|
}
|
|
|
|
|
2012-04-24 20:04:12 +08:00
|
|
|
if (INTEL_INFO(dev)->gen >= 6)
|
|
|
|
I915_WRITE(INSTPM, _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING));
|
2011-12-13 11:21:58 +08:00
|
|
|
|
2012-07-25 11:47:31 +08:00
|
|
|
if (HAS_L3_GPU_CACHE(dev))
|
2013-05-29 10:22:29 +08:00
|
|
|
I915_WRITE_IMR(ring, ~GT_RENDER_L3_PARITY_ERROR_INTERRUPT);
|
2012-05-26 07:56:23 +08:00
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
static void render_ring_cleanup(struct intel_ring_buffer *ring)
|
|
|
|
{
|
2012-12-17 23:21:27 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
if (!ring->private)
|
|
|
|
return;
|
|
|
|
|
2012-12-17 23:21:27 +08:00
|
|
|
if (HAS_BROKEN_CS_TLB(dev))
|
|
|
|
drm_gem_object_unreference(to_gem_object(ring->private));
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
cleanup_pipe_control(ring);
|
|
|
|
}
|
|
|
|
|
2010-12-04 19:30:53 +08:00
|
|
|
static void
|
2011-09-15 11:32:47 +08:00
|
|
|
update_mboxes(struct intel_ring_buffer *ring,
|
2012-11-28 00:22:52 +08:00
|
|
|
u32 mmio_offset)
|
2010-12-04 19:30:53 +08:00
|
|
|
{
|
2013-05-29 10:22:18 +08:00
|
|
|
/* NB: In order to be able to do semaphore MBOX updates for varying number
|
|
|
|
* of rings, it's easiest if we round up each individual update to a
|
|
|
|
* multiple of 2 (since ring updates must always be a multiple of 2)
|
|
|
|
* even though the actual update only requires 3 dwords.
|
|
|
|
*/
|
|
|
|
#define MBOX_UPDATE_DWORDS 4
|
2012-11-14 17:15:14 +08:00
|
|
|
intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1));
|
2011-09-15 11:32:47 +08:00
|
|
|
intel_ring_emit(ring, mmio_offset);
|
2012-11-28 00:22:52 +08:00
|
|
|
intel_ring_emit(ring, ring->outstanding_lazy_request);
|
2013-05-29 10:22:18 +08:00
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
2010-12-04 19:30:53 +08:00
|
|
|
}
|
|
|
|
|
2011-09-15 11:32:47 +08:00
|
|
|
/**
|
|
|
|
* gen6_add_request - Update the semaphore mailbox registers
|
|
|
|
*
|
|
|
|
* @ring - ring that is adding a request
|
|
|
|
* @seqno - return seqno stuck into the ring
|
|
|
|
*
|
|
|
|
* Update the mailbox registers in the *other* rings with the current seqno.
|
|
|
|
* This acts like a signal in the canonical semaphore.
|
|
|
|
*/
|
2010-12-04 19:30:53 +08:00
|
|
|
static int
|
2012-11-28 00:22:52 +08:00
|
|
|
gen6_add_request(struct intel_ring_buffer *ring)
|
2010-12-04 19:30:53 +08:00
|
|
|
{
|
2013-05-29 10:22:18 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
|
|
|
struct intel_ring_buffer *useless;
|
|
|
|
int i, ret;
|
2010-12-04 19:30:53 +08:00
|
|
|
|
2013-05-29 10:22:18 +08:00
|
|
|
ret = intel_ring_begin(ring, ((I915_NUM_RINGS-1) *
|
|
|
|
MBOX_UPDATE_DWORDS) +
|
|
|
|
4);
|
2010-12-04 19:30:53 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2013-05-29 10:22:18 +08:00
|
|
|
#undef MBOX_UPDATE_DWORDS
|
2010-12-04 19:30:53 +08:00
|
|
|
|
2013-05-29 10:22:18 +08:00
|
|
|
for_each_ring(useless, dev_priv, i) {
|
|
|
|
u32 mbox_reg = ring->signal_mbox[i];
|
|
|
|
if (mbox_reg != GEN6_NOSYNC)
|
|
|
|
update_mboxes(ring, mbox_reg);
|
|
|
|
}
|
2010-12-04 19:30:53 +08:00
|
|
|
|
|
|
|
intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
|
|
|
|
intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
|
2012-11-28 00:22:52 +08:00
|
|
|
intel_ring_emit(ring, ring->outstanding_lazy_request);
|
2010-12-04 19:30:53 +08:00
|
|
|
intel_ring_emit(ring, MI_USER_INTERRUPT);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-12-10 21:41:48 +08:00
|
|
|
static inline bool i915_gem_has_seqno_wrapped(struct drm_device *dev,
|
|
|
|
u32 seqno)
|
|
|
|
{
|
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
|
|
|
return dev_priv->last_seqno < seqno;
|
|
|
|
}
|
|
|
|
|
2011-09-15 11:32:47 +08:00
|
|
|
/**
|
|
|
|
* intel_ring_sync - sync the waiter to the signaller on seqno
|
|
|
|
*
|
|
|
|
* @waiter - ring that is waiting
|
|
|
|
* @signaller - ring which has, or will signal
|
|
|
|
* @seqno - seqno which the waiter will block on
|
|
|
|
*/
|
|
|
|
static int
|
2012-04-12 04:12:52 +08:00
|
|
|
gen6_ring_sync(struct intel_ring_buffer *waiter,
|
|
|
|
struct intel_ring_buffer *signaller,
|
|
|
|
u32 seqno)
|
2010-12-04 19:30:53 +08:00
|
|
|
{
|
|
|
|
int ret;
|
2011-09-15 11:32:47 +08:00
|
|
|
u32 dw1 = MI_SEMAPHORE_MBOX |
|
|
|
|
MI_SEMAPHORE_COMPARE |
|
|
|
|
MI_SEMAPHORE_REGISTER;
|
2010-12-04 19:30:53 +08:00
|
|
|
|
2012-04-12 02:18:21 +08:00
|
|
|
/* Throughout all of the GEM code, seqno passed implies our current
|
|
|
|
* seqno is >= the last seqno executed. However for hardware the
|
|
|
|
* comparison is strictly greater than.
|
|
|
|
*/
|
|
|
|
seqno -= 1;
|
|
|
|
|
2012-04-12 04:12:52 +08:00
|
|
|
WARN_ON(signaller->semaphore_register[waiter->id] ==
|
|
|
|
MI_SEMAPHORE_SYNC_INVALID);
|
|
|
|
|
2011-09-15 11:32:47 +08:00
|
|
|
ret = intel_ring_begin(waiter, 4);
|
2010-12-04 19:30:53 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2012-12-10 21:41:48 +08:00
|
|
|
/* If seqno wrap happened, omit the wait with no-ops */
|
|
|
|
if (likely(!i915_gem_has_seqno_wrapped(waiter->dev, seqno))) {
|
|
|
|
intel_ring_emit(waiter,
|
|
|
|
dw1 |
|
|
|
|
signaller->semaphore_register[waiter->id]);
|
|
|
|
intel_ring_emit(waiter, seqno);
|
|
|
|
intel_ring_emit(waiter, 0);
|
|
|
|
intel_ring_emit(waiter, MI_NOOP);
|
|
|
|
} else {
|
|
|
|
intel_ring_emit(waiter, MI_NOOP);
|
|
|
|
intel_ring_emit(waiter, MI_NOOP);
|
|
|
|
intel_ring_emit(waiter, MI_NOOP);
|
|
|
|
intel_ring_emit(waiter, MI_NOOP);
|
|
|
|
}
|
2011-09-15 11:32:47 +08:00
|
|
|
intel_ring_advance(waiter);
|
2010-12-04 19:30:53 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
#define PIPE_CONTROL_FLUSH(ring__, addr__) \
|
|
|
|
do { \
|
2011-10-12 05:41:08 +08:00
|
|
|
intel_ring_emit(ring__, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | \
|
|
|
|
PIPE_CONTROL_DEPTH_STALL); \
|
2010-12-15 17:56:50 +08:00
|
|
|
intel_ring_emit(ring__, (addr__) | PIPE_CONTROL_GLOBAL_GTT); \
|
|
|
|
intel_ring_emit(ring__, 0); \
|
|
|
|
intel_ring_emit(ring__, 0); \
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
static int
|
2012-11-28 00:22:52 +08:00
|
|
|
pc_render_add_request(struct intel_ring_buffer *ring)
|
2010-12-15 17:56:50 +08:00
|
|
|
{
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
u32 scratch_addr = pc->gtt_offset + 128;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* For Ironlake, MI_USER_INTERRUPT was deprecated and apparently
|
|
|
|
* incoherent with writes to memory, i.e. completely fubar,
|
|
|
|
* so we need to use PIPE_NOTIFY instead.
|
|
|
|
*
|
|
|
|
* However, we also need to workaround the qword write
|
|
|
|
* incoherence by flushing the 6 PIPE_NOTIFY buffers out to
|
|
|
|
* memory before requesting an interrupt.
|
|
|
|
*/
|
|
|
|
ret = intel_ring_begin(ring, 32);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2011-10-12 05:41:08 +08:00
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
|
2011-10-12 05:41:09 +08:00
|
|
|
PIPE_CONTROL_WRITE_FLUSH |
|
|
|
|
PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);
|
2010-12-15 17:56:50 +08:00
|
|
|
intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
|
2012-11-28 00:22:52 +08:00
|
|
|
intel_ring_emit(ring, ring->outstanding_lazy_request);
|
2010-12-15 17:56:50 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
|
|
|
scratch_addr += 128; /* write to separate cachelines */
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
|
|
|
scratch_addr += 128;
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
|
|
|
scratch_addr += 128;
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
|
|
|
scratch_addr += 128;
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
|
|
|
scratch_addr += 128;
|
|
|
|
PIPE_CONTROL_FLUSH(ring, scratch_addr);
|
2012-02-15 19:25:36 +08:00
|
|
|
|
2011-10-12 05:41:08 +08:00
|
|
|
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
|
2011-10-12 05:41:09 +08:00
|
|
|
PIPE_CONTROL_WRITE_FLUSH |
|
|
|
|
PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
|
2010-12-15 17:56:50 +08:00
|
|
|
PIPE_CONTROL_NOTIFY);
|
|
|
|
intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
|
2012-11-28 00:22:52 +08:00
|
|
|
intel_ring_emit(ring, ring->outstanding_lazy_request);
|
2010-12-15 17:56:50 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-12-14 23:01:25 +08:00
|
|
|
static u32
|
2012-08-09 17:58:30 +08:00
|
|
|
gen6_ring_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
|
2012-12-14 23:01:25 +08:00
|
|
|
{
|
|
|
|
/* Workaround to force correct ordering between irq and seqno writes on
|
|
|
|
* ivb (and maybe also on snb) by reading from a CS register (like
|
|
|
|
* ACTHD) before reading the status page. */
|
2012-08-09 17:58:30 +08:00
|
|
|
if (!lazy_coherency)
|
2012-12-14 23:01:25 +08:00
|
|
|
intel_ring_get_active_head(ring);
|
|
|
|
return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
|
|
|
|
}
|
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
static u32
|
2012-08-09 17:58:30 +08:00
|
|
|
ring_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2010-12-04 19:30:53 +08:00
|
|
|
return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
|
|
|
|
}
|
|
|
|
|
2012-12-19 17:13:05 +08:00
|
|
|
static void
|
|
|
|
ring_set_seqno(struct intel_ring_buffer *ring, u32 seqno)
|
|
|
|
{
|
|
|
|
intel_write_status_page(ring, I915_GEM_HWS_INDEX, seqno);
|
|
|
|
}
|
|
|
|
|
2010-12-15 17:56:50 +08:00
|
|
|
static u32
|
2012-08-09 17:58:30 +08:00
|
|
|
pc_render_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
|
2010-12-15 17:56:50 +08:00
|
|
|
{
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
return pc->cpu_page[0];
|
|
|
|
}
|
|
|
|
|
2012-12-19 17:13:05 +08:00
|
|
|
static void
|
|
|
|
pc_render_set_seqno(struct intel_ring_buffer *ring, u32 seqno)
|
|
|
|
{
|
|
|
|
struct pipe_control *pc = ring->private;
|
|
|
|
pc->cpu_page[0] = seqno;
|
|
|
|
}
|
|
|
|
|
2012-04-12 04:12:54 +08:00
|
|
|
static bool
|
|
|
|
gen5_ring_get_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2012-04-12 04:12:54 +08:00
|
|
|
|
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return false;
|
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (ring->irq_refcount.gt++ == 0) {
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->gt_irq_mask &= ~ring->irq_enable_mask;
|
|
|
|
I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
|
|
|
|
POSTING_READ(GTIMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2012-04-12 04:12:54 +08:00
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
gen5_ring_put_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2012-04-12 04:12:54 +08:00
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (--ring->irq_refcount.gt == 0) {
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->gt_irq_mask |= ring->irq_enable_mask;
|
|
|
|
I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
|
|
|
|
POSTING_READ(GTIMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2012-04-12 04:12:54 +08:00
|
|
|
}
|
|
|
|
|
2010-12-14 00:54:50 +08:00
|
|
|
static bool
|
2012-04-12 04:12:53 +08:00
|
|
|
i9xx_ring_get_irq(struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2011-01-05 06:22:56 +08:00
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-12-14 00:54:50 +08:00
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return false;
|
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (ring->irq_refcount.gt++ == 0) {
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->irq_mask &= ~ring->irq_enable_mask;
|
|
|
|
I915_WRITE(IMR, dev_priv->irq_mask);
|
|
|
|
POSTING_READ(IMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2010-12-14 00:54:50 +08:00
|
|
|
|
|
|
|
return true;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
static void
|
2012-04-12 04:12:53 +08:00
|
|
|
i9xx_ring_put_irq(struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2011-01-05 06:22:56 +08:00
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (--ring->irq_refcount.gt == 0) {
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->irq_mask |= ring->irq_enable_mask;
|
|
|
|
I915_WRITE(IMR, dev_priv->irq_mask);
|
|
|
|
POSTING_READ(IMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2012-04-23 04:13:57 +08:00
|
|
|
static bool
|
|
|
|
i8xx_ring_get_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2012-04-23 04:13:57 +08:00
|
|
|
|
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return false;
|
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (ring->irq_refcount.gt++ == 0) {
|
2012-04-23 04:13:57 +08:00
|
|
|
dev_priv->irq_mask &= ~ring->irq_enable_mask;
|
|
|
|
I915_WRITE16(IMR, dev_priv->irq_mask);
|
|
|
|
POSTING_READ16(IMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2012-04-23 04:13:57 +08:00
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
i8xx_ring_put_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2012-04-23 04:13:57 +08:00
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (--ring->irq_refcount.gt == 0) {
|
2012-04-23 04:13:57 +08:00
|
|
|
dev_priv->irq_mask |= ring->irq_enable_mask;
|
|
|
|
I915_WRITE16(IMR, dev_priv->irq_mask);
|
|
|
|
POSTING_READ16(IMR);
|
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2012-04-23 04:13:57 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
void intel_ring_setup_status_page(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2011-05-07 08:12:35 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2010-10-27 19:18:21 +08:00
|
|
|
drm_i915_private_t *dev_priv = ring->dev->dev_private;
|
2011-05-07 08:12:35 +08:00
|
|
|
u32 mmio = 0;
|
|
|
|
|
|
|
|
/* The ring status page addresses are no longer next to the rest of
|
|
|
|
* the ring registers as of gen7.
|
|
|
|
*/
|
|
|
|
if (IS_GEN7(dev)) {
|
|
|
|
switch (ring->id) {
|
2011-12-14 20:57:00 +08:00
|
|
|
case RCS:
|
2011-05-07 08:12:35 +08:00
|
|
|
mmio = RENDER_HWS_PGA_GEN7;
|
|
|
|
break;
|
2011-12-14 20:57:00 +08:00
|
|
|
case BCS:
|
2011-05-07 08:12:35 +08:00
|
|
|
mmio = BLT_HWS_PGA_GEN7;
|
|
|
|
break;
|
2011-12-14 20:57:00 +08:00
|
|
|
case VCS:
|
2011-05-07 08:12:35 +08:00
|
|
|
mmio = BSD_HWS_PGA_GEN7;
|
|
|
|
break;
|
2013-05-29 10:22:19 +08:00
|
|
|
case VECS:
|
2013-05-29 10:22:23 +08:00
|
|
|
mmio = VEBOX_HWS_PGA_GEN7;
|
|
|
|
break;
|
2011-05-07 08:12:35 +08:00
|
|
|
}
|
|
|
|
} else if (IS_GEN6(ring->dev)) {
|
|
|
|
mmio = RING_HWS_PGA_GEN6(ring->mmio_base);
|
|
|
|
} else {
|
|
|
|
mmio = RING_HWS_PGA(ring->mmio_base);
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
I915_WRITE(mmio, (u32)ring->status_page.gfx_addr);
|
|
|
|
POSTING_READ(mmio);
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
|
|
|
|
2011-01-05 01:34:02 +08:00
|
|
|
static int
|
2010-10-27 19:18:21 +08:00
|
|
|
bsd_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate_domains,
|
|
|
|
u32 flush_domains)
|
2010-05-21 09:08:57 +08:00
|
|
|
{
|
2011-01-05 01:34:02 +08:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring, MI_FLUSH);
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
return 0;
|
2010-05-21 09:08:57 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 23:11:02 +08:00
|
|
|
static int
|
2012-11-28 00:22:52 +08:00
|
|
|
i9xx_add_request(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:57 +08:00
|
|
|
{
|
2010-10-27 23:11:02 +08:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 4);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-08-07 18:01:22 +08:00
|
|
|
|
2010-10-27 23:11:02 +08:00
|
|
|
intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
|
|
|
|
intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
|
2012-11-28 00:22:52 +08:00
|
|
|
intel_ring_emit(ring, ring->outstanding_lazy_request);
|
2010-10-27 23:11:02 +08:00
|
|
|
intel_ring_emit(ring, MI_USER_INTERRUPT);
|
|
|
|
intel_ring_advance(ring);
|
2010-05-21 09:08:57 +08:00
|
|
|
|
2010-10-27 23:11:02 +08:00
|
|
|
return 0;
|
2010-05-21 09:08:57 +08:00
|
|
|
}
|
|
|
|
|
2011-01-05 01:35:21 +08:00
|
|
|
static bool
|
2012-03-30 10:11:27 +08:00
|
|
|
gen6_ring_get_irq(struct intel_ring_buffer *ring)
|
2011-01-05 01:35:21 +08:00
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
2011-01-05 06:22:56 +08:00
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2011-01-05 01:35:21 +08:00
|
|
|
|
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return false;
|
|
|
|
|
2012-12-14 23:01:25 +08:00
|
|
|
/* It looks like we need to prevent the gt from suspending while waiting
|
|
|
|
* for an notifiy irq, otherwise irqs seem to get lost on at least the
|
|
|
|
* blt/bsd rings on ivb. */
|
2012-01-25 21:04:00 +08:00
|
|
|
gen6_gt_force_wake_get(dev_priv);
|
2012-12-14 23:01:25 +08:00
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (ring->irq_refcount.gt++ == 0) {
|
2012-07-25 11:47:31 +08:00
|
|
|
if (HAS_L3_GPU_CACHE(dev) && ring->id == RCS)
|
2013-05-29 10:22:29 +08:00
|
|
|
I915_WRITE_IMR(ring,
|
|
|
|
~(ring->irq_enable_mask |
|
|
|
|
GT_RENDER_L3_PARITY_ERROR_INTERRUPT));
|
2012-05-26 07:56:23 +08:00
|
|
|
else
|
|
|
|
I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->gt_irq_mask &= ~ring->irq_enable_mask;
|
|
|
|
I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
|
|
|
|
POSTING_READ(GTIMR);
|
2011-01-05 01:35:21 +08:00
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2011-01-05 01:35:21 +08:00
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-03-30 10:11:27 +08:00
|
|
|
gen6_ring_put_irq(struct intel_ring_buffer *ring)
|
2011-01-05 01:35:21 +08:00
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
2011-01-05 06:22:56 +08:00
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2012-04-25 04:48:47 +08:00
|
|
|
unsigned long flags;
|
2011-01-05 01:35:21 +08:00
|
|
|
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_lock_irqsave(&dev_priv->irq_lock, flags);
|
2013-05-29 10:22:28 +08:00
|
|
|
if (--ring->irq_refcount.gt == 0) {
|
2012-07-25 11:47:31 +08:00
|
|
|
if (HAS_L3_GPU_CACHE(dev) && ring->id == RCS)
|
2013-05-29 10:22:29 +08:00
|
|
|
I915_WRITE_IMR(ring,
|
|
|
|
~GT_RENDER_L3_PARITY_ERROR_INTERRUPT);
|
2012-05-26 07:56:23 +08:00
|
|
|
else
|
|
|
|
I915_WRITE_IMR(ring, ~0);
|
2012-04-12 04:12:59 +08:00
|
|
|
dev_priv->gt_irq_mask |= ring->irq_enable_mask;
|
|
|
|
I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
|
|
|
|
POSTING_READ(GTIMR);
|
2010-12-04 19:30:53 +08:00
|
|
|
}
|
2012-04-25 04:48:47 +08:00
|
|
|
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
|
2012-12-14 23:01:25 +08:00
|
|
|
|
2012-01-25 21:04:00 +08:00
|
|
|
gen6_gt_force_wake_put(dev_priv);
|
2010-05-21 09:08:57 +08:00
|
|
|
}
|
|
|
|
|
2013-05-29 10:22:30 +08:00
|
|
|
static bool
|
|
|
|
hsw_vebox_get_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&dev_priv->rps.lock, flags);
|
|
|
|
if (ring->irq_refcount.pm++ == 0) {
|
|
|
|
u32 pm_imr = I915_READ(GEN6_PMIMR);
|
|
|
|
I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
|
|
|
|
I915_WRITE(GEN6_PMIMR, pm_imr & ~ring->irq_enable_mask);
|
|
|
|
POSTING_READ(GEN6_PMIMR);
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&dev_priv->rps.lock, flags);
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
hsw_vebox_put_irq(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_device *dev = ring->dev;
|
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&dev_priv->rps.lock, flags);
|
|
|
|
if (--ring->irq_refcount.pm == 0) {
|
|
|
|
u32 pm_imr = I915_READ(GEN6_PMIMR);
|
|
|
|
I915_WRITE_IMR(ring, ~0);
|
|
|
|
I915_WRITE(GEN6_PMIMR, pm_imr | ring->irq_enable_mask);
|
|
|
|
POSTING_READ(GEN6_PMIMR);
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&dev_priv->rps.lock, flags);
|
|
|
|
}
|
|
|
|
|
2010-05-21 09:08:57 +08:00
|
|
|
static int
|
2012-10-17 19:09:54 +08:00
|
|
|
i965_dispatch_execbuffer(struct intel_ring_buffer *ring,
|
|
|
|
u32 offset, u32 length,
|
|
|
|
unsigned flags)
|
2010-05-21 09:08:57 +08:00
|
|
|
{
|
2010-10-27 19:45:26 +08:00
|
|
|
int ret;
|
2010-10-27 19:18:21 +08:00
|
|
|
|
2010-10-27 19:45:26 +08:00
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
intel_ring_emit(ring,
|
2012-04-17 23:38:12 +08:00
|
|
|
MI_BATCH_BUFFER_START |
|
|
|
|
MI_BATCH_GTT |
|
2012-10-17 19:09:54 +08:00
|
|
|
(flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965));
|
2010-11-30 22:10:25 +08:00
|
|
|
intel_ring_emit(ring, offset);
|
2010-10-27 19:18:21 +08:00
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
2010-05-21 09:08:57 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-12-17 23:21:27 +08:00
|
|
|
/* Just userspace ABI convention to limit the wa batch bo to a resonable size */
|
|
|
|
#define I830_BATCH_LIMIT (256*1024)
|
2010-05-21 09:08:55 +08:00
|
|
|
static int
|
2012-04-12 04:12:56 +08:00
|
|
|
i830_dispatch_execbuffer(struct intel_ring_buffer *ring,
|
2012-10-17 19:09:54 +08:00
|
|
|
u32 offset, u32 len,
|
|
|
|
unsigned flags)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-11-30 22:10:25 +08:00
|
|
|
int ret;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-12-17 23:21:27 +08:00
|
|
|
if (flags & I915_DISPATCH_PINNED) {
|
|
|
|
ret = intel_ring_begin(ring, 4);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-12-17 23:21:27 +08:00
|
|
|
intel_ring_emit(ring, MI_BATCH_BUFFER);
|
|
|
|
intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE));
|
|
|
|
intel_ring_emit(ring, offset + len - 8);
|
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
} else {
|
|
|
|
struct drm_i915_gem_object *obj = ring->private;
|
|
|
|
u32 cs_offset = obj->gtt_offset;
|
|
|
|
|
|
|
|
if (len > I830_BATCH_LIMIT)
|
|
|
|
return -ENOSPC;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 9+3);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
/* Blit the batch (which has now all relocs applied) to the stable batch
|
|
|
|
* scratch bo area (so that the CS never stumbles over its tlb
|
|
|
|
* invalidation bug) ... */
|
|
|
|
intel_ring_emit(ring, XY_SRC_COPY_BLT_CMD |
|
|
|
|
XY_SRC_COPY_BLT_WRITE_ALPHA |
|
|
|
|
XY_SRC_COPY_BLT_WRITE_RGB);
|
|
|
|
intel_ring_emit(ring, BLT_DEPTH_32 | BLT_ROP_GXCOPY | 4096);
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_emit(ring, (DIV_ROUND_UP(len, 4096) << 16) | 1024);
|
|
|
|
intel_ring_emit(ring, cs_offset);
|
|
|
|
intel_ring_emit(ring, 0);
|
|
|
|
intel_ring_emit(ring, 4096);
|
|
|
|
intel_ring_emit(ring, offset);
|
|
|
|
intel_ring_emit(ring, MI_FLUSH);
|
|
|
|
|
|
|
|
/* ... and execute it. */
|
|
|
|
intel_ring_emit(ring, MI_BATCH_BUFFER);
|
|
|
|
intel_ring_emit(ring, cs_offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE));
|
|
|
|
intel_ring_emit(ring, cs_offset + len - 8);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
}
|
2010-10-27 19:45:26 +08:00
|
|
|
|
2012-04-12 04:12:56 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
i915_dispatch_execbuffer(struct intel_ring_buffer *ring,
|
2012-10-17 19:09:54 +08:00
|
|
|
u32 offset, u32 len,
|
|
|
|
unsigned flags)
|
2012-04-12 04:12:56 +08:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2012-04-17 23:38:12 +08:00
|
|
|
intel_ring_emit(ring, MI_BATCH_BUFFER_START | MI_BATCH_GTT);
|
2012-10-17 19:09:54 +08:00
|
|
|
intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE));
|
2010-11-30 22:10:25 +08:00
|
|
|
intel_ring_advance(ring);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static void cleanup_status_page(struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-11-09 03:18:58 +08:00
|
|
|
struct drm_i915_gem_object *obj;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
obj = ring->status_page.obj;
|
|
|
|
if (obj == NULL)
|
2010-05-22 04:26:39 +08:00
|
|
|
return;
|
|
|
|
|
2012-06-01 22:20:22 +08:00
|
|
|
kunmap(sg_page(obj->pages->sgl));
|
2010-05-22 04:26:39 +08:00
|
|
|
i915_gem_object_unpin(obj);
|
2010-11-09 03:18:58 +08:00
|
|
|
drm_gem_object_unreference(&obj->base);
|
2010-05-21 09:08:55 +08:00
|
|
|
ring->status_page.obj = NULL;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static int init_status_page(struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2010-11-09 03:18:58 +08:00
|
|
|
struct drm_i915_gem_object *obj;
|
2010-05-22 04:26:39 +08:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
obj = i915_gem_alloc_object(dev, 4096);
|
|
|
|
if (obj == NULL) {
|
|
|
|
DRM_ERROR("Failed to allocate status page\n");
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto err;
|
|
|
|
}
|
2011-04-04 16:44:39 +08:00
|
|
|
|
|
|
|
i915_gem_object_set_cache_level(obj, I915_CACHE_LLC);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-08-11 22:41:04 +08:00
|
|
|
ret = i915_gem_object_pin(obj, 4096, true, false);
|
2010-05-22 04:26:39 +08:00
|
|
|
if (ret != 0) {
|
|
|
|
goto err_unref;
|
|
|
|
}
|
|
|
|
|
2010-11-09 03:18:58 +08:00
|
|
|
ring->status_page.gfx_addr = obj->gtt_offset;
|
2012-06-01 22:20:22 +08:00
|
|
|
ring->status_page.page_addr = kmap(sg_page(obj->pages->sgl));
|
2010-05-21 09:08:55 +08:00
|
|
|
if (ring->status_page.page_addr == NULL) {
|
2012-07-13 14:16:12 +08:00
|
|
|
ret = -ENOMEM;
|
2010-05-22 04:26:39 +08:00
|
|
|
goto err_unpin;
|
|
|
|
}
|
2010-05-21 09:08:55 +08:00
|
|
|
ring->status_page.obj = obj;
|
|
|
|
memset(ring->status_page.page_addr, 0, PAGE_SIZE);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
intel_ring_setup_status_page(ring);
|
2010-05-21 09:08:55 +08:00
|
|
|
DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n",
|
|
|
|
ring->name, ring->status_page.gfx_addr);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err_unpin:
|
|
|
|
i915_gem_object_unpin(obj);
|
|
|
|
err_unref:
|
2010-11-09 03:18:58 +08:00
|
|
|
drm_gem_object_unreference(&obj->base);
|
2010-05-22 04:26:39 +08:00
|
|
|
err:
|
2010-05-21 09:08:55 +08:00
|
|
|
return ret;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2012-11-16 19:43:20 +08:00
|
|
|
static int init_phys_hws_pga(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
struct drm_i915_private *dev_priv = ring->dev->dev_private;
|
|
|
|
u32 addr;
|
|
|
|
|
|
|
|
if (!dev_priv->status_page_dmah) {
|
|
|
|
dev_priv->status_page_dmah =
|
|
|
|
drm_pci_alloc(ring->dev, PAGE_SIZE, PAGE_SIZE);
|
|
|
|
if (!dev_priv->status_page_dmah)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
addr = dev_priv->status_page_dmah->busaddr;
|
|
|
|
if (INTEL_INFO(ring->dev)->gen >= 4)
|
|
|
|
addr |= (dev_priv->status_page_dmah->busaddr >> 28) & 0xf0;
|
|
|
|
I915_WRITE(HWS_PGA, addr);
|
|
|
|
|
|
|
|
ring->status_page.page_addr = dev_priv->status_page_dmah->vaddr;
|
|
|
|
memset(ring->status_page.page_addr, 0, PAGE_SIZE);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-17 05:07:40 +08:00
|
|
|
static int intel_init_ring_buffer(struct drm_device *dev,
|
|
|
|
struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-11-09 03:18:58 +08:00
|
|
|
struct drm_i915_gem_object *obj;
|
2012-06-07 21:55:57 +08:00
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
2010-08-07 18:01:34 +08:00
|
|
|
int ret;
|
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
ring->dev = dev;
|
2010-09-29 23:10:57 +08:00
|
|
|
INIT_LIST_HEAD(&ring->active_list);
|
|
|
|
INIT_LIST_HEAD(&ring->request_list);
|
2012-04-12 04:12:47 +08:00
|
|
|
ring->size = 32 * PAGE_SIZE;
|
2012-11-28 00:22:52 +08:00
|
|
|
memset(ring->sync_seqno, 0, sizeof(ring->sync_seqno));
|
2011-01-05 18:32:24 +08:00
|
|
|
|
2011-03-29 20:19:09 +08:00
|
|
|
init_waitqueue_head(&ring->irq_queue);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
if (I915_NEED_GFX_HWS(dev)) {
|
2010-10-27 19:18:21 +08:00
|
|
|
ret = init_status_page(ring);
|
2010-05-21 09:08:55 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2012-11-16 19:43:20 +08:00
|
|
|
} else {
|
|
|
|
BUG_ON(ring->id != RCS);
|
|
|
|
ret = init_phys_hws_pga(ring);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-11-15 19:32:28 +08:00
|
|
|
obj = NULL;
|
|
|
|
if (!HAS_LLC(dev))
|
|
|
|
obj = i915_gem_object_create_stolen(dev, ring->size);
|
|
|
|
if (obj == NULL)
|
|
|
|
obj = i915_gem_alloc_object(dev, ring->size);
|
2010-05-22 04:26:39 +08:00
|
|
|
if (obj == NULL) {
|
|
|
|
DRM_ERROR("Failed to allocate ringbuffer\n");
|
2010-05-21 09:08:55 +08:00
|
|
|
ret = -ENOMEM;
|
2010-08-07 18:01:34 +08:00
|
|
|
goto err_hws;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2010-11-09 03:18:58 +08:00
|
|
|
ring->obj = obj;
|
2010-05-21 09:08:55 +08:00
|
|
|
|
2012-08-11 22:41:04 +08:00
|
|
|
ret = i915_gem_object_pin(obj, PAGE_SIZE, true, false);
|
2010-08-07 18:01:34 +08:00
|
|
|
if (ret)
|
|
|
|
goto err_unref;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-06-05 00:05:40 +08:00
|
|
|
ret = i915_gem_object_set_to_gtt_domain(obj, true);
|
|
|
|
if (ret)
|
|
|
|
goto err_unpin;
|
|
|
|
|
2012-06-07 21:55:57 +08:00
|
|
|
ring->virtual_start =
|
2013-01-18 04:45:16 +08:00
|
|
|
ioremap_wc(dev_priv->gtt.mappable_base + obj->gtt_offset,
|
2012-06-07 21:55:57 +08:00
|
|
|
ring->size);
|
2012-04-27 05:28:16 +08:00
|
|
|
if (ring->virtual_start == NULL) {
|
2010-05-22 04:26:39 +08:00
|
|
|
DRM_ERROR("Failed to map ringbuffer.\n");
|
2010-05-21 09:08:55 +08:00
|
|
|
ret = -EINVAL;
|
2010-08-07 18:01:34 +08:00
|
|
|
goto err_unpin;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
ret = ring->init(ring);
|
2010-08-07 18:01:34 +08:00
|
|
|
if (ret)
|
|
|
|
goto err_unmap;
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-12-22 22:04:47 +08:00
|
|
|
/* Workaround an erratum on the i830 which causes a hang if
|
|
|
|
* the TAIL pointer points to within the last 2 cachelines
|
|
|
|
* of the buffer.
|
|
|
|
*/
|
|
|
|
ring->effective_size = ring->size;
|
2012-04-09 20:59:46 +08:00
|
|
|
if (IS_I830(ring->dev) || IS_845G(ring->dev))
|
2010-12-22 22:04:47 +08:00
|
|
|
ring->effective_size -= 128;
|
|
|
|
|
2010-10-30 01:15:52 +08:00
|
|
|
return 0;
|
2010-08-07 18:01:34 +08:00
|
|
|
|
|
|
|
err_unmap:
|
2012-04-27 05:28:16 +08:00
|
|
|
iounmap(ring->virtual_start);
|
2010-08-07 18:01:34 +08:00
|
|
|
err_unpin:
|
|
|
|
i915_gem_object_unpin(obj);
|
|
|
|
err_unref:
|
2010-11-09 03:18:58 +08:00
|
|
|
drm_gem_object_unreference(&obj->base);
|
|
|
|
ring->obj = NULL;
|
2010-08-07 18:01:34 +08:00
|
|
|
err_hws:
|
2010-10-27 19:18:21 +08:00
|
|
|
cleanup_status_page(ring);
|
2010-05-21 09:08:55 +08:00
|
|
|
return ret;
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
void intel_cleanup_ring_buffer(struct intel_ring_buffer *ring)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-29 23:18:36 +08:00
|
|
|
struct drm_i915_private *dev_priv;
|
|
|
|
int ret;
|
|
|
|
|
2010-11-09 03:18:58 +08:00
|
|
|
if (ring->obj == NULL)
|
2010-05-22 04:26:39 +08:00
|
|
|
return;
|
|
|
|
|
2010-10-29 23:18:36 +08:00
|
|
|
/* Disable the ring buffer. The ring must be idle at this point */
|
|
|
|
dev_priv = ring->dev->dev_private;
|
2012-11-28 00:22:54 +08:00
|
|
|
ret = intel_ring_idle(ring);
|
2011-01-25 00:35:42 +08:00
|
|
|
if (ret)
|
|
|
|
DRM_ERROR("failed to quiesce %s whilst cleaning up: %d\n",
|
|
|
|
ring->name, ret);
|
|
|
|
|
2010-10-29 23:18:36 +08:00
|
|
|
I915_WRITE_CTL(ring, 0);
|
|
|
|
|
2012-04-27 05:28:16 +08:00
|
|
|
iounmap(ring->virtual_start);
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-11-09 03:18:58 +08:00
|
|
|
i915_gem_object_unpin(ring->obj);
|
|
|
|
drm_gem_object_unreference(&ring->obj->base);
|
|
|
|
ring->obj = NULL;
|
2010-10-27 19:18:21 +08:00
|
|
|
|
2010-11-02 16:31:01 +08:00
|
|
|
if (ring->cleanup)
|
|
|
|
ring->cleanup(ring);
|
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
cleanup_status_page(ring);
|
2010-05-22 04:26:39 +08:00
|
|
|
}
|
|
|
|
|
2012-02-15 19:25:36 +08:00
|
|
|
static int intel_ring_wait_seqno(struct intel_ring_buffer *ring, u32 seqno)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
2012-05-25 06:03:11 +08:00
|
|
|
ret = i915_wait_seqno(ring, seqno);
|
2012-04-27 07:02:58 +08:00
|
|
|
if (!ret)
|
|
|
|
i915_gem_retire_requests_ring(ring);
|
2012-02-15 19:25:36 +08:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int intel_ring_wait_request(struct intel_ring_buffer *ring, int n)
|
|
|
|
{
|
|
|
|
struct drm_i915_gem_request *request;
|
|
|
|
u32 seqno = 0;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
i915_gem_retire_requests_ring(ring);
|
|
|
|
|
|
|
|
if (ring->last_retired_head != -1) {
|
|
|
|
ring->head = ring->last_retired_head;
|
|
|
|
ring->last_retired_head = -1;
|
|
|
|
ring->space = ring_space(ring);
|
|
|
|
if (ring->space >= n)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
list_for_each_entry(request, &ring->request_list, list) {
|
|
|
|
int space;
|
|
|
|
|
|
|
|
if (request->tail == -1)
|
|
|
|
continue;
|
|
|
|
|
2012-12-04 00:43:32 +08:00
|
|
|
space = request->tail - (ring->tail + I915_RING_FREE_SPACE);
|
2012-02-15 19:25:36 +08:00
|
|
|
if (space < 0)
|
|
|
|
space += ring->size;
|
|
|
|
if (space >= n) {
|
|
|
|
seqno = request->seqno;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Consume this request in case we need more space than
|
|
|
|
* is available and so need to prevent a race between
|
|
|
|
* updating last_retired_head and direct reads of
|
|
|
|
* I915_RING_HEAD. It also provides a nice sanity check.
|
|
|
|
*/
|
|
|
|
request->tail = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (seqno == 0)
|
|
|
|
return -ENOSPC;
|
|
|
|
|
|
|
|
ret = intel_ring_wait_seqno(ring, seqno);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
if (WARN_ON(ring->last_retired_head == -1))
|
|
|
|
return -ENOSPC;
|
|
|
|
|
|
|
|
ring->head = ring->last_retired_head;
|
|
|
|
ring->last_retired_head = -1;
|
|
|
|
ring->space = ring_space(ring);
|
|
|
|
if (WARN_ON(ring->space < n))
|
|
|
|
return -ENOSPC;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-11-28 00:22:54 +08:00
|
|
|
static int ring_wait_for_space(struct intel_ring_buffer *ring, int n)
|
2010-05-22 04:26:39 +08:00
|
|
|
{
|
2010-10-27 19:18:21 +08:00
|
|
|
struct drm_device *dev = ring->dev;
|
2010-11-09 17:17:32 +08:00
|
|
|
struct drm_i915_private *dev_priv = dev->dev_private;
|
2010-10-27 19:18:21 +08:00
|
|
|
unsigned long end;
|
2012-02-15 19:25:36 +08:00
|
|
|
int ret;
|
2011-01-21 01:00:10 +08:00
|
|
|
|
2012-02-15 19:25:36 +08:00
|
|
|
ret = intel_ring_wait_request(ring, n);
|
|
|
|
if (ret != -ENOSPC)
|
|
|
|
return ret;
|
|
|
|
|
2011-02-03 19:57:46 +08:00
|
|
|
trace_i915_ring_wait_begin(ring);
|
2012-04-23 22:50:50 +08:00
|
|
|
/* With GEM the hangcheck timer should kick us out of the loop,
|
|
|
|
* leaving it early runs the risk of corrupting GEM state (due
|
|
|
|
* to running on almost untested codepaths). But on resume
|
|
|
|
* timers don't work yet, so prevent a complete hang in that
|
|
|
|
* case by choosing an insanely large timeout. */
|
|
|
|
end = jiffies + 60 * HZ;
|
2011-12-14 20:56:59 +08:00
|
|
|
|
2010-05-21 09:08:55 +08:00
|
|
|
do {
|
2011-01-21 01:00:10 +08:00
|
|
|
ring->head = I915_READ_HEAD(ring);
|
|
|
|
ring->space = ring_space(ring);
|
2010-05-22 04:26:39 +08:00
|
|
|
if (ring->space >= n) {
|
2011-02-03 19:57:46 +08:00
|
|
|
trace_i915_ring_wait_end(ring);
|
2010-05-22 04:26:39 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dev->primary->master) {
|
|
|
|
struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
|
|
|
|
if (master_priv->sarea_priv)
|
|
|
|
master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
|
|
|
|
}
|
2010-05-21 09:08:57 +08:00
|
|
|
|
2010-10-13 17:09:14 +08:00
|
|
|
msleep(1);
|
drm/i915: non-interruptible sleeps can't handle -EAGAIN
So don't return -EAGAIN, even in the case of a gpu hang. Remap it to
-EIO instead. Note that this isn't really an issue with
interruptability, but more that we have quite a few codepaths (mostly
around kms stuff) that simply can't handle any errors and hence not
even -EAGAIN. Instead of adding proper failure paths so that we could
restart these ioctls we've opted for the cheap way out of sleeping
non-interruptibly. Which works everywhere but when the gpu dies,
which this patch fixes.
So essentially interruptible == false means 'wait for the gpu or die
trying'.'
This patch is a bit ugly because intel_ring_begin is all non-interruptible
and hence only returns -EIO. But as the comment in there says,
auditing all the callsites would be a pain.
To avoid duplicating code, reuse i915_gem_check_wedge in __wait_seqno
and intel_wait_ring_buffer. Also use the opportunity to clarify the
different cases in i915_gem_check_wedge a bit with comments.
v2: Don't access dev_priv->mm.interruptible from check_wedge - we
might not hold dev->struct_mutex, making this racy. Instead pass
interruptible in as a parameter. I've noticed this because I've hit a
BUG_ON(!mutex_is_locked) at the top of check_wedge. This has been
added in
commit b4aca0106c466b5a0329318203f65bac2d91b682
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Wed Apr 25 20:50:12 2012 -0700
drm/i915: extract some common olr+wedge code
although that commit is missing any justification for this. I guess
it's just copy&paste, because the same commit add the same BUG_ON
check to check_olr, where it indeed makes sense.
But in check_wedge everything we access is protected by other means,
so this is superflous. And because it now gets in the way (we add a
new caller in __wait_seqno, which can be called without
dev->struct_mutext) let's just remove it.
v3: Group all the i915_gem_check_wedge refactoring into this patch, so
that this patch here is all about not returning -EAGAIN to callsites
that can't handle syscall restarting.
v4: Add clarification what interuptible == fales means in our code,
requested by Ben Widawsky.
v5: Fix EAGAIN mispell noticed by Chris Wilson.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-07-05 04:54:13 +08:00
|
|
|
|
2012-11-15 00:14:05 +08:00
|
|
|
ret = i915_gem_check_wedge(&dev_priv->gpu_error,
|
|
|
|
dev_priv->mm.interruptible);
|
drm/i915: non-interruptible sleeps can't handle -EAGAIN
So don't return -EAGAIN, even in the case of a gpu hang. Remap it to
-EIO instead. Note that this isn't really an issue with
interruptability, but more that we have quite a few codepaths (mostly
around kms stuff) that simply can't handle any errors and hence not
even -EAGAIN. Instead of adding proper failure paths so that we could
restart these ioctls we've opted for the cheap way out of sleeping
non-interruptibly. Which works everywhere but when the gpu dies,
which this patch fixes.
So essentially interruptible == false means 'wait for the gpu or die
trying'.'
This patch is a bit ugly because intel_ring_begin is all non-interruptible
and hence only returns -EIO. But as the comment in there says,
auditing all the callsites would be a pain.
To avoid duplicating code, reuse i915_gem_check_wedge in __wait_seqno
and intel_wait_ring_buffer. Also use the opportunity to clarify the
different cases in i915_gem_check_wedge a bit with comments.
v2: Don't access dev_priv->mm.interruptible from check_wedge - we
might not hold dev->struct_mutex, making this racy. Instead pass
interruptible in as a parameter. I've noticed this because I've hit a
BUG_ON(!mutex_is_locked) at the top of check_wedge. This has been
added in
commit b4aca0106c466b5a0329318203f65bac2d91b682
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Wed Apr 25 20:50:12 2012 -0700
drm/i915: extract some common olr+wedge code
although that commit is missing any justification for this. I guess
it's just copy&paste, because the same commit add the same BUG_ON
check to check_olr, where it indeed makes sense.
But in check_wedge everything we access is protected by other means,
so this is superflous. And because it now gets in the way (we add a
new caller in __wait_seqno, which can be called without
dev->struct_mutext) let's just remove it.
v3: Group all the i915_gem_check_wedge refactoring into this patch, so
that this patch here is all about not returning -EAGAIN to callsites
that can't handle syscall restarting.
v4: Add clarification what interuptible == fales means in our code,
requested by Ben Widawsky.
v5: Fix EAGAIN mispell noticed by Chris Wilson.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-07-05 04:54:13 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-05-21 09:08:55 +08:00
|
|
|
} while (!time_after(jiffies, end));
|
2011-02-03 19:57:46 +08:00
|
|
|
trace_i915_ring_wait_end(ring);
|
2010-05-21 09:08:55 +08:00
|
|
|
return -EBUSY;
|
|
|
|
}
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2012-11-28 00:22:54 +08:00
|
|
|
static int intel_wrap_ring_buffer(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
uint32_t __iomem *virt;
|
|
|
|
int rem = ring->size - ring->tail;
|
|
|
|
|
|
|
|
if (ring->space < rem) {
|
|
|
|
int ret = ring_wait_for_space(ring, rem);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
virt = ring->virtual_start + ring->tail;
|
|
|
|
rem /= 4;
|
|
|
|
while (rem--)
|
|
|
|
iowrite32(MI_NOOP, virt++);
|
|
|
|
|
|
|
|
ring->tail = 0;
|
|
|
|
ring->space = ring_space(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int intel_ring_idle(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
u32 seqno;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* We need to add any requests required to flush the objects and ring */
|
|
|
|
if (ring->outstanding_lazy_request) {
|
|
|
|
ret = i915_add_request(ring, NULL, NULL);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Wait upon the last request to be completed */
|
|
|
|
if (list_empty(&ring->request_list))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
seqno = list_entry(ring->request_list.prev,
|
|
|
|
struct drm_i915_gem_request,
|
|
|
|
list)->seqno;
|
|
|
|
|
|
|
|
return i915_wait_seqno(ring, seqno);
|
|
|
|
}
|
|
|
|
|
2012-11-28 00:22:52 +08:00
|
|
|
static int
|
|
|
|
intel_ring_alloc_seqno(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
if (ring->outstanding_lazy_request)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return i915_gem_get_seqno(ring->dev, &ring->outstanding_lazy_request);
|
|
|
|
}
|
|
|
|
|
2012-12-04 21:12:03 +08:00
|
|
|
static int __intel_ring_begin(struct intel_ring_buffer *ring,
|
|
|
|
int bytes)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (unlikely(ring->tail + bytes > ring->effective_size)) {
|
|
|
|
ret = intel_wrap_ring_buffer(ring);
|
|
|
|
if (unlikely(ret))
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(ring->space < bytes)) {
|
|
|
|
ret = ring_wait_for_space(ring, bytes);
|
|
|
|
if (unlikely(ret))
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
ring->space -= bytes;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-10-27 19:45:26 +08:00
|
|
|
int intel_ring_begin(struct intel_ring_buffer *ring,
|
|
|
|
int num_dwords)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2012-07-05 04:52:50 +08:00
|
|
|
drm_i915_private_t *dev_priv = ring->dev->dev_private;
|
2010-10-27 19:45:26 +08:00
|
|
|
int ret;
|
2010-10-27 19:18:21 +08:00
|
|
|
|
2012-11-15 00:14:05 +08:00
|
|
|
ret = i915_gem_check_wedge(&dev_priv->gpu_error,
|
|
|
|
dev_priv->mm.interruptible);
|
2012-07-05 04:52:50 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2011-01-26 23:55:56 +08:00
|
|
|
|
2012-11-28 00:22:52 +08:00
|
|
|
/* Preallocate the olr before touching the ring */
|
|
|
|
ret = intel_ring_alloc_seqno(ring);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2012-12-04 21:12:03 +08:00
|
|
|
return __intel_ring_begin(ring, num_dwords * sizeof(uint32_t));
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
2010-10-27 19:18:21 +08:00
|
|
|
|
2012-12-19 17:13:06 +08:00
|
|
|
void intel_ring_init_seqno(struct intel_ring_buffer *ring, u32 seqno)
|
2012-12-04 21:12:04 +08:00
|
|
|
{
|
2012-12-19 17:13:06 +08:00
|
|
|
struct drm_i915_private *dev_priv = ring->dev->dev_private;
|
2012-12-04 21:12:04 +08:00
|
|
|
|
|
|
|
BUG_ON(ring->outstanding_lazy_request);
|
|
|
|
|
2012-12-19 17:13:06 +08:00
|
|
|
if (INTEL_INFO(ring->dev)->gen >= 6) {
|
|
|
|
I915_WRITE(RING_SYNC_0(ring->mmio_base), 0);
|
|
|
|
I915_WRITE(RING_SYNC_1(ring->mmio_base), 0);
|
2010-10-27 19:45:26 +08:00
|
|
|
}
|
2010-08-04 22:18:13 +08:00
|
|
|
|
2012-12-19 17:13:06 +08:00
|
|
|
ring->set_seqno(ring, seqno);
|
2013-05-24 22:16:07 +08:00
|
|
|
ring->hangcheck.seqno = seqno;
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
void intel_ring_advance(struct intel_ring_buffer *ring)
|
2010-05-21 09:08:55 +08:00
|
|
|
{
|
2012-05-03 20:48:16 +08:00
|
|
|
struct drm_i915_private *dev_priv = ring->dev->dev_private;
|
|
|
|
|
2010-08-04 22:18:13 +08:00
|
|
|
ring->tail &= ring->size - 1;
|
2012-11-15 00:14:04 +08:00
|
|
|
if (dev_priv->gpu_error.stop_rings & intel_ring_flag(ring))
|
2012-05-03 20:48:16 +08:00
|
|
|
return;
|
2010-10-27 19:18:21 +08:00
|
|
|
ring->write_tail(ring, ring->tail);
|
2010-05-21 09:08:55 +08:00
|
|
|
}
|
2010-05-22 04:26:39 +08:00
|
|
|
|
2010-09-19 21:40:43 +08:00
|
|
|
|
2010-10-27 19:18:21 +08:00
|
|
|
static void gen6_bsd_ring_write_tail(struct intel_ring_buffer *ring,
|
2010-10-23 00:02:41 +08:00
|
|
|
u32 value)
|
2010-09-19 21:40:43 +08:00
|
|
|
{
|
2011-08-17 03:34:10 +08:00
|
|
|
drm_i915_private_t *dev_priv = ring->dev->dev_private;
|
2010-09-19 21:40:43 +08:00
|
|
|
|
|
|
|
/* Every tail move must follow the sequence below */
|
2012-07-06 00:14:01 +08:00
|
|
|
|
|
|
|
/* Disable notification that the ring is IDLE. The GT
|
|
|
|
* will then assume that it is busy and bring it out of rc6.
|
|
|
|
*/
|
2011-08-17 03:34:10 +08:00
|
|
|
I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
|
2012-07-06 00:14:01 +08:00
|
|
|
_MASKED_BIT_ENABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
|
|
|
|
|
|
|
|
/* Clear the context id. Here be magic! */
|
|
|
|
I915_WRITE64(GEN6_BSD_RNCID, 0x0);
|
2011-08-17 03:34:10 +08:00
|
|
|
|
2012-07-06 00:14:01 +08:00
|
|
|
/* Wait for the ring not to be idle, i.e. for it to wake up. */
|
2011-08-17 03:34:10 +08:00
|
|
|
if (wait_for((I915_READ(GEN6_BSD_SLEEP_PSMI_CONTROL) &
|
2012-07-06 00:14:01 +08:00
|
|
|
GEN6_BSD_SLEEP_INDICATOR) == 0,
|
|
|
|
50))
|
|
|
|
DRM_ERROR("timed out waiting for the BSD ring to wake up\n");
|
2011-08-17 03:34:10 +08:00
|
|
|
|
2012-07-06 00:14:01 +08:00
|
|
|
/* Now that the ring is fully powered up, update the tail */
|
2011-08-17 03:34:10 +08:00
|
|
|
I915_WRITE_TAIL(ring, value);
|
2012-07-06 00:14:01 +08:00
|
|
|
POSTING_READ(RING_TAIL(ring->mmio_base));
|
|
|
|
|
|
|
|
/* Let the ring send IDLE messages to the GT again,
|
|
|
|
* and so let it sleep to conserve power when idle.
|
|
|
|
*/
|
2011-08-17 03:34:10 +08:00
|
|
|
I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
|
2012-07-06 00:14:01 +08:00
|
|
|
_MASKED_BIT_DISABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
|
2010-09-19 21:40:43 +08:00
|
|
|
}
|
|
|
|
|
2013-05-29 10:22:21 +08:00
|
|
|
static int gen6_bsd_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate, u32 flush)
|
2010-09-19 21:40:43 +08:00
|
|
|
{
|
2011-02-02 20:13:49 +08:00
|
|
|
uint32_t cmd;
|
2011-01-05 01:34:02 +08:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 4);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2011-02-02 20:13:49 +08:00
|
|
|
cmd = MI_FLUSH_DW;
|
2012-10-27 00:42:42 +08:00
|
|
|
/*
|
|
|
|
* Bspec vol 1c.5 - video engine command streamer:
|
|
|
|
* "If ENABLED, all TLBs will be invalidated once the flush
|
|
|
|
* operation is complete. This bit is only valid when the
|
|
|
|
* Post-Sync Operation field is a value of 1h or 3h."
|
|
|
|
*/
|
2011-02-02 20:13:49 +08:00
|
|
|
if (invalidate & I915_GEM_GPU_DOMAINS)
|
2012-10-27 00:42:42 +08:00
|
|
|
cmd |= MI_INVALIDATE_TLB | MI_INVALIDATE_BSD |
|
|
|
|
MI_FLUSH_DW_STORE_INDEX | MI_FLUSH_DW_OP_STOREDW;
|
2011-02-02 20:13:49 +08:00
|
|
|
intel_ring_emit(ring, cmd);
|
2012-10-27 00:42:42 +08:00
|
|
|
intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT);
|
2011-01-05 01:34:02 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
2011-02-02 20:13:49 +08:00
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
2011-01-05 01:34:02 +08:00
|
|
|
intel_ring_advance(ring);
|
|
|
|
return 0;
|
2010-09-19 21:40:43 +08:00
|
|
|
}
|
|
|
|
|
2012-10-17 19:09:54 +08:00
|
|
|
static int
|
|
|
|
hsw_ring_dispatch_execbuffer(struct intel_ring_buffer *ring,
|
|
|
|
u32 offset, u32 len,
|
|
|
|
unsigned flags)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
intel_ring_emit(ring,
|
|
|
|
MI_BATCH_BUFFER_START | MI_BATCH_PPGTT_HSW |
|
|
|
|
(flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_HSW));
|
|
|
|
/* bit0-7 is the length on GEN6+ */
|
|
|
|
intel_ring_emit(ring, offset);
|
|
|
|
intel_ring_advance(ring);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-09-19 21:40:43 +08:00
|
|
|
static int
|
2010-10-27 19:18:21 +08:00
|
|
|
gen6_ring_dispatch_execbuffer(struct intel_ring_buffer *ring,
|
2012-10-17 19:09:54 +08:00
|
|
|
u32 offset, u32 len,
|
|
|
|
unsigned flags)
|
2010-09-19 21:40:43 +08:00
|
|
|
{
|
2011-08-17 03:34:10 +08:00
|
|
|
int ret;
|
2010-09-20 00:53:44 +08:00
|
|
|
|
2011-08-17 03:34:10 +08:00
|
|
|
ret = intel_ring_begin(ring, 2);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2010-10-27 19:45:26 +08:00
|
|
|
|
2012-10-17 19:09:54 +08:00
|
|
|
intel_ring_emit(ring,
|
|
|
|
MI_BATCH_BUFFER_START |
|
|
|
|
(flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965));
|
2011-08-17 03:34:10 +08:00
|
|
|
/* bit0-7 is the length on GEN6+ */
|
|
|
|
intel_ring_emit(ring, offset);
|
|
|
|
intel_ring_advance(ring);
|
2010-09-20 00:53:44 +08:00
|
|
|
|
2011-08-17 03:34:10 +08:00
|
|
|
return 0;
|
2010-09-19 21:40:43 +08:00
|
|
|
}
|
|
|
|
|
2010-10-19 18:19:32 +08:00
|
|
|
/* Blitter support (SandyBridge+) */
|
|
|
|
|
2013-05-29 10:22:21 +08:00
|
|
|
static int gen6_ring_flush(struct intel_ring_buffer *ring,
|
|
|
|
u32 invalidate, u32 flush)
|
2010-11-02 16:31:01 +08:00
|
|
|
{
|
2011-02-02 20:13:49 +08:00
|
|
|
uint32_t cmd;
|
2011-01-05 01:34:02 +08:00
|
|
|
int ret;
|
|
|
|
|
2011-12-14 20:57:07 +08:00
|
|
|
ret = intel_ring_begin(ring, 4);
|
2011-01-05 01:34:02 +08:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2011-02-02 20:13:49 +08:00
|
|
|
cmd = MI_FLUSH_DW;
|
2012-10-27 00:42:42 +08:00
|
|
|
/*
|
|
|
|
* Bspec vol 1c.3 - blitter engine command streamer:
|
|
|
|
* "If ENABLED, all TLBs will be invalidated once the flush
|
|
|
|
* operation is complete. This bit is only valid when the
|
|
|
|
* Post-Sync Operation field is a value of 1h or 3h."
|
|
|
|
*/
|
2011-02-02 20:13:49 +08:00
|
|
|
if (invalidate & I915_GEM_DOMAIN_RENDER)
|
2012-10-27 00:42:42 +08:00
|
|
|
cmd |= MI_INVALIDATE_TLB | MI_FLUSH_DW_STORE_INDEX |
|
2012-11-04 19:24:47 +08:00
|
|
|
MI_FLUSH_DW_OP_STOREDW;
|
2011-02-02 20:13:49 +08:00
|
|
|
intel_ring_emit(ring, cmd);
|
2012-10-27 00:42:42 +08:00
|
|
|
intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT);
|
2011-01-05 01:34:02 +08:00
|
|
|
intel_ring_emit(ring, 0);
|
2011-02-02 20:13:49 +08:00
|
|
|
intel_ring_emit(ring, MI_NOOP);
|
2011-01-05 01:34:02 +08:00
|
|
|
intel_ring_advance(ring);
|
|
|
|
return 0;
|
2010-11-02 16:31:01 +08:00
|
|
|
}
|
|
|
|
|
2010-09-16 10:43:11 +08:00
|
|
|
int intel_init_render_ring_buffer(struct drm_device *dev)
|
|
|
|
{
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2010-12-04 19:30:53 +08:00
|
|
|
struct intel_ring_buffer *ring = &dev_priv->ring[RCS];
|
2010-09-16 10:43:11 +08:00
|
|
|
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->name = "render ring";
|
|
|
|
ring->id = RCS;
|
|
|
|
ring->mmio_base = RENDER_RING_BASE;
|
|
|
|
|
2010-12-04 19:30:53 +08:00
|
|
|
if (INTEL_INFO(dev)->gen >= 6) {
|
|
|
|
ring->add_request = gen6_add_request;
|
2012-08-18 05:35:41 +08:00
|
|
|
ring->flush = gen7_render_ring_flush;
|
2012-07-21 01:02:28 +08:00
|
|
|
if (INTEL_INFO(dev)->gen == 6)
|
2012-08-18 05:35:42 +08:00
|
|
|
ring->flush = gen6_render_ring_flush;
|
2012-03-30 10:11:27 +08:00
|
|
|
ring->irq_get = gen6_ring_get_irq;
|
|
|
|
ring->irq_put = gen6_ring_put_irq;
|
2013-05-29 10:22:29 +08:00
|
|
|
ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
|
2012-12-14 23:01:25 +08:00
|
|
|
ring->get_seqno = gen6_ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2012-04-12 04:12:52 +08:00
|
|
|
ring->sync_to = gen6_ring_sync;
|
2013-05-29 10:22:17 +08:00
|
|
|
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_INVALID;
|
|
|
|
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_RV;
|
|
|
|
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_RB;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_RVE;
|
2013-05-29 10:22:18 +08:00
|
|
|
ring->signal_mbox[RCS] = GEN6_NOSYNC;
|
|
|
|
ring->signal_mbox[VCS] = GEN6_VRSYNC;
|
|
|
|
ring->signal_mbox[BCS] = GEN6_BRSYNC;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->signal_mbox[VECS] = GEN6_VERSYNC;
|
2010-12-15 17:56:50 +08:00
|
|
|
} else if (IS_GEN5(dev)) {
|
|
|
|
ring->add_request = pc_render_add_request;
|
2012-04-18 18:12:11 +08:00
|
|
|
ring->flush = gen4_render_ring_flush;
|
2010-12-15 17:56:50 +08:00
|
|
|
ring->get_seqno = pc_render_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = pc_render_set_seqno;
|
2012-04-12 04:12:54 +08:00
|
|
|
ring->irq_get = gen5_ring_get_irq;
|
|
|
|
ring->irq_put = gen5_ring_put_irq;
|
2013-05-29 10:22:29 +08:00
|
|
|
ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT |
|
|
|
|
GT_RENDER_PIPECTL_NOTIFY_INTERRUPT;
|
2012-04-12 04:12:48 +08:00
|
|
|
} else {
|
2012-04-12 04:12:57 +08:00
|
|
|
ring->add_request = i9xx_add_request;
|
2012-04-18 18:12:11 +08:00
|
|
|
if (INTEL_INFO(dev)->gen < 4)
|
|
|
|
ring->flush = gen2_render_ring_flush;
|
|
|
|
else
|
|
|
|
ring->flush = gen4_render_ring_flush;
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->get_seqno = ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2012-04-23 04:13:57 +08:00
|
|
|
if (IS_GEN2(dev)) {
|
|
|
|
ring->irq_get = i8xx_ring_get_irq;
|
|
|
|
ring->irq_put = i8xx_ring_put_irq;
|
|
|
|
} else {
|
|
|
|
ring->irq_get = i9xx_ring_get_irq;
|
|
|
|
ring->irq_put = i9xx_ring_put_irq;
|
|
|
|
}
|
2012-04-12 04:12:53 +08:00
|
|
|
ring->irq_enable_mask = I915_USER_INTERRUPT;
|
2010-12-04 19:30:53 +08:00
|
|
|
}
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->write_tail = ring_write_tail;
|
2012-10-17 19:09:54 +08:00
|
|
|
if (IS_HASWELL(dev))
|
|
|
|
ring->dispatch_execbuffer = hsw_ring_dispatch_execbuffer;
|
|
|
|
else if (INTEL_INFO(dev)->gen >= 6)
|
2012-04-12 04:12:56 +08:00
|
|
|
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
|
|
|
|
else if (INTEL_INFO(dev)->gen >= 4)
|
|
|
|
ring->dispatch_execbuffer = i965_dispatch_execbuffer;
|
|
|
|
else if (IS_I830(dev) || IS_845G(dev))
|
|
|
|
ring->dispatch_execbuffer = i830_dispatch_execbuffer;
|
|
|
|
else
|
|
|
|
ring->dispatch_execbuffer = i915_dispatch_execbuffer;
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->init = init_render_ring;
|
|
|
|
ring->cleanup = render_ring_cleanup;
|
|
|
|
|
2012-12-17 23:21:27 +08:00
|
|
|
/* Workaround batchbuffer to combat CS tlb bug. */
|
|
|
|
if (HAS_BROKEN_CS_TLB(dev)) {
|
|
|
|
struct drm_i915_gem_object *obj;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
obj = i915_gem_alloc_object(dev, I830_BATCH_LIMIT);
|
|
|
|
if (obj == NULL) {
|
|
|
|
DRM_ERROR("Failed to allocate batch bo\n");
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = i915_gem_object_pin(obj, 0, true, false);
|
|
|
|
if (ret != 0) {
|
|
|
|
drm_gem_object_unreference(&obj->base);
|
|
|
|
DRM_ERROR("Failed to ping batch bo\n");
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
ring->private = obj;
|
|
|
|
}
|
|
|
|
|
2010-12-04 19:30:53 +08:00
|
|
|
return intel_init_ring_buffer(dev, ring);
|
2010-09-16 10:43:11 +08:00
|
|
|
}
|
|
|
|
|
2011-01-20 17:57:11 +08:00
|
|
|
int intel_render_ring_init_dri(struct drm_device *dev, u64 start, u32 size)
|
|
|
|
{
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
|
|
|
struct intel_ring_buffer *ring = &dev_priv->ring[RCS];
|
2012-11-16 19:43:20 +08:00
|
|
|
int ret;
|
2011-01-20 17:57:11 +08:00
|
|
|
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->name = "render ring";
|
|
|
|
ring->id = RCS;
|
|
|
|
ring->mmio_base = RENDER_RING_BASE;
|
|
|
|
|
2011-01-20 17:57:11 +08:00
|
|
|
if (INTEL_INFO(dev)->gen >= 6) {
|
2012-04-12 04:12:51 +08:00
|
|
|
/* non-kms not supported on gen6+ */
|
|
|
|
return -ENODEV;
|
2011-01-20 17:57:11 +08:00
|
|
|
}
|
2012-04-12 04:12:58 +08:00
|
|
|
|
|
|
|
/* Note: gem is not supported on gen5/ilk without kms (the corresponding
|
|
|
|
* gem_init ioctl returns with -ENODEV). Hence we do not need to set up
|
|
|
|
* the special gen5 functions. */
|
|
|
|
ring->add_request = i9xx_add_request;
|
2012-04-18 18:12:11 +08:00
|
|
|
if (INTEL_INFO(dev)->gen < 4)
|
|
|
|
ring->flush = gen2_render_ring_flush;
|
|
|
|
else
|
|
|
|
ring->flush = gen4_render_ring_flush;
|
2012-04-12 04:12:58 +08:00
|
|
|
ring->get_seqno = ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2012-04-23 04:13:57 +08:00
|
|
|
if (IS_GEN2(dev)) {
|
|
|
|
ring->irq_get = i8xx_ring_get_irq;
|
|
|
|
ring->irq_put = i8xx_ring_put_irq;
|
|
|
|
} else {
|
|
|
|
ring->irq_get = i9xx_ring_get_irq;
|
|
|
|
ring->irq_put = i9xx_ring_put_irq;
|
|
|
|
}
|
2012-04-12 04:12:58 +08:00
|
|
|
ring->irq_enable_mask = I915_USER_INTERRUPT;
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->write_tail = ring_write_tail;
|
2012-04-12 04:12:56 +08:00
|
|
|
if (INTEL_INFO(dev)->gen >= 4)
|
|
|
|
ring->dispatch_execbuffer = i965_dispatch_execbuffer;
|
|
|
|
else if (IS_I830(dev) || IS_845G(dev))
|
|
|
|
ring->dispatch_execbuffer = i830_dispatch_execbuffer;
|
|
|
|
else
|
|
|
|
ring->dispatch_execbuffer = i915_dispatch_execbuffer;
|
2012-04-12 04:12:48 +08:00
|
|
|
ring->init = init_render_ring;
|
|
|
|
ring->cleanup = render_ring_cleanup;
|
2011-01-20 17:57:11 +08:00
|
|
|
|
|
|
|
ring->dev = dev;
|
|
|
|
INIT_LIST_HEAD(&ring->active_list);
|
|
|
|
INIT_LIST_HEAD(&ring->request_list);
|
|
|
|
|
|
|
|
ring->size = size;
|
|
|
|
ring->effective_size = ring->size;
|
2012-10-29 22:59:26 +08:00
|
|
|
if (IS_I830(ring->dev) || IS_845G(ring->dev))
|
2011-01-20 17:57:11 +08:00
|
|
|
ring->effective_size -= 128;
|
|
|
|
|
2012-04-27 05:28:16 +08:00
|
|
|
ring->virtual_start = ioremap_wc(start, size);
|
|
|
|
if (ring->virtual_start == NULL) {
|
2011-01-20 17:57:11 +08:00
|
|
|
DRM_ERROR("can not ioremap virtual address for"
|
|
|
|
" ring buffer\n");
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2012-11-16 19:43:20 +08:00
|
|
|
if (!I915_NEED_GFX_HWS(dev)) {
|
|
|
|
ret = init_phys_hws_pga(ring);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2011-01-20 17:57:11 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-09-16 10:43:11 +08:00
|
|
|
int intel_init_bsd_ring_buffer(struct drm_device *dev)
|
|
|
|
{
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2010-12-04 19:30:53 +08:00
|
|
|
struct intel_ring_buffer *ring = &dev_priv->ring[VCS];
|
2010-09-16 10:43:11 +08:00
|
|
|
|
2012-04-12 04:12:49 +08:00
|
|
|
ring->name = "bsd ring";
|
|
|
|
ring->id = VCS;
|
|
|
|
|
2012-04-12 04:12:55 +08:00
|
|
|
ring->write_tail = ring_write_tail;
|
2012-04-12 04:12:49 +08:00
|
|
|
if (IS_GEN6(dev) || IS_GEN7(dev)) {
|
|
|
|
ring->mmio_base = GEN6_BSD_RING_BASE;
|
2012-04-12 04:12:55 +08:00
|
|
|
/* gen6 bsd needs a special wa for tail updates */
|
|
|
|
if (IS_GEN6(dev))
|
|
|
|
ring->write_tail = gen6_bsd_ring_write_tail;
|
2013-05-29 10:22:21 +08:00
|
|
|
ring->flush = gen6_bsd_ring_flush;
|
2012-04-12 04:12:49 +08:00
|
|
|
ring->add_request = gen6_add_request;
|
|
|
|
ring->get_seqno = gen6_ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2013-05-29 10:22:29 +08:00
|
|
|
ring->irq_enable_mask = GT_BSD_USER_INTERRUPT;
|
2012-04-12 04:12:49 +08:00
|
|
|
ring->irq_get = gen6_ring_get_irq;
|
|
|
|
ring->irq_put = gen6_ring_put_irq;
|
|
|
|
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
|
2012-04-12 04:12:52 +08:00
|
|
|
ring->sync_to = gen6_ring_sync;
|
2013-05-29 10:22:17 +08:00
|
|
|
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_VR;
|
|
|
|
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_INVALID;
|
|
|
|
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_VB;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_VVE;
|
2013-05-29 10:22:18 +08:00
|
|
|
ring->signal_mbox[RCS] = GEN6_RVSYNC;
|
|
|
|
ring->signal_mbox[VCS] = GEN6_NOSYNC;
|
|
|
|
ring->signal_mbox[BCS] = GEN6_BVSYNC;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->signal_mbox[VECS] = GEN6_VEVSYNC;
|
2012-04-12 04:12:49 +08:00
|
|
|
} else {
|
|
|
|
ring->mmio_base = BSD_RING_BASE;
|
|
|
|
ring->flush = bsd_ring_flush;
|
2012-04-12 04:12:57 +08:00
|
|
|
ring->add_request = i9xx_add_request;
|
2012-04-12 04:12:49 +08:00
|
|
|
ring->get_seqno = ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2012-04-12 04:12:54 +08:00
|
|
|
if (IS_GEN5(dev)) {
|
2013-05-29 10:22:29 +08:00
|
|
|
ring->irq_enable_mask = ILK_BSD_USER_INTERRUPT;
|
2012-04-12 04:12:54 +08:00
|
|
|
ring->irq_get = gen5_ring_get_irq;
|
|
|
|
ring->irq_put = gen5_ring_put_irq;
|
|
|
|
} else {
|
2012-04-12 04:12:53 +08:00
|
|
|
ring->irq_enable_mask = I915_BSD_USER_INTERRUPT;
|
2012-04-12 04:12:54 +08:00
|
|
|
ring->irq_get = i9xx_ring_get_irq;
|
|
|
|
ring->irq_put = i9xx_ring_put_irq;
|
|
|
|
}
|
2012-04-12 04:12:56 +08:00
|
|
|
ring->dispatch_execbuffer = i965_dispatch_execbuffer;
|
2012-04-12 04:12:49 +08:00
|
|
|
}
|
|
|
|
ring->init = init_ring_common;
|
|
|
|
|
2010-12-04 19:30:53 +08:00
|
|
|
return intel_init_ring_buffer(dev, ring);
|
2010-09-16 10:43:11 +08:00
|
|
|
}
|
2010-10-19 18:19:32 +08:00
|
|
|
|
|
|
|
int intel_init_blt_ring_buffer(struct drm_device *dev)
|
|
|
|
{
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
2010-12-04 19:30:53 +08:00
|
|
|
struct intel_ring_buffer *ring = &dev_priv->ring[BCS];
|
2010-10-19 18:19:32 +08:00
|
|
|
|
2012-04-12 04:12:50 +08:00
|
|
|
ring->name = "blitter ring";
|
|
|
|
ring->id = BCS;
|
|
|
|
|
|
|
|
ring->mmio_base = BLT_RING_BASE;
|
|
|
|
ring->write_tail = ring_write_tail;
|
2013-05-29 10:22:21 +08:00
|
|
|
ring->flush = gen6_ring_flush;
|
2012-04-12 04:12:50 +08:00
|
|
|
ring->add_request = gen6_add_request;
|
|
|
|
ring->get_seqno = gen6_ring_get_seqno;
|
2012-12-19 17:13:05 +08:00
|
|
|
ring->set_seqno = ring_set_seqno;
|
2013-05-29 10:22:29 +08:00
|
|
|
ring->irq_enable_mask = GT_BLT_USER_INTERRUPT;
|
2012-04-12 04:12:50 +08:00
|
|
|
ring->irq_get = gen6_ring_get_irq;
|
|
|
|
ring->irq_put = gen6_ring_put_irq;
|
|
|
|
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
|
2012-04-12 04:12:52 +08:00
|
|
|
ring->sync_to = gen6_ring_sync;
|
2013-05-29 10:22:17 +08:00
|
|
|
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_BR;
|
|
|
|
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_BV;
|
|
|
|
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_INVALID;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_BVE;
|
2013-05-29 10:22:18 +08:00
|
|
|
ring->signal_mbox[RCS] = GEN6_RBSYNC;
|
|
|
|
ring->signal_mbox[VCS] = GEN6_VBSYNC;
|
|
|
|
ring->signal_mbox[BCS] = GEN6_NOSYNC;
|
2013-05-29 10:22:20 +08:00
|
|
|
ring->signal_mbox[VECS] = GEN6_VEBSYNC;
|
2012-04-12 04:12:50 +08:00
|
|
|
ring->init = init_ring_common;
|
2010-10-19 18:19:32 +08:00
|
|
|
|
2010-12-04 19:30:53 +08:00
|
|
|
return intel_init_ring_buffer(dev, ring);
|
2010-10-19 18:19:32 +08:00
|
|
|
}
|
2012-07-20 19:41:08 +08:00
|
|
|
|
2013-05-29 10:22:23 +08:00
|
|
|
int intel_init_vebox_ring_buffer(struct drm_device *dev)
|
|
|
|
{
|
|
|
|
drm_i915_private_t *dev_priv = dev->dev_private;
|
|
|
|
struct intel_ring_buffer *ring = &dev_priv->ring[VECS];
|
|
|
|
|
|
|
|
ring->name = "video enhancement ring";
|
|
|
|
ring->id = VECS;
|
|
|
|
|
|
|
|
ring->mmio_base = VEBOX_RING_BASE;
|
|
|
|
ring->write_tail = ring_write_tail;
|
|
|
|
ring->flush = gen6_ring_flush;
|
|
|
|
ring->add_request = gen6_add_request;
|
|
|
|
ring->get_seqno = gen6_ring_get_seqno;
|
|
|
|
ring->set_seqno = ring_set_seqno;
|
2013-05-29 10:22:31 +08:00
|
|
|
ring->irq_enable_mask = PM_VEBOX_USER_INTERRUPT |
|
|
|
|
PM_VEBOX_CS_ERROR_INTERRUPT;
|
2013-05-29 10:22:30 +08:00
|
|
|
ring->irq_get = hsw_vebox_get_irq;
|
|
|
|
ring->irq_put = hsw_vebox_put_irq;
|
2013-05-29 10:22:23 +08:00
|
|
|
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
|
|
|
|
ring->sync_to = gen6_ring_sync;
|
|
|
|
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_VER;
|
|
|
|
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_VEV;
|
|
|
|
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_VEB;
|
|
|
|
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_INVALID;
|
|
|
|
ring->signal_mbox[RCS] = GEN6_RVESYNC;
|
|
|
|
ring->signal_mbox[VCS] = GEN6_VVESYNC;
|
|
|
|
ring->signal_mbox[BCS] = GEN6_BVESYNC;
|
|
|
|
ring->signal_mbox[VECS] = GEN6_NOSYNC;
|
|
|
|
ring->init = init_ring_common;
|
|
|
|
|
|
|
|
return intel_init_ring_buffer(dev, ring);
|
|
|
|
}
|
|
|
|
|
2012-07-20 19:41:08 +08:00
|
|
|
int
|
|
|
|
intel_ring_flush_all_caches(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!ring->gpu_caches_dirty)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ret = ring->flush(ring, 0, I915_GEM_GPU_DOMAINS);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
trace_i915_gem_ring_flush(ring, 0, I915_GEM_GPU_DOMAINS);
|
|
|
|
|
|
|
|
ring->gpu_caches_dirty = false;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
intel_ring_invalidate_all_caches(struct intel_ring_buffer *ring)
|
|
|
|
{
|
|
|
|
uint32_t flush_domains;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
flush_domains = 0;
|
|
|
|
if (ring->gpu_caches_dirty)
|
|
|
|
flush_domains = I915_GEM_GPU_DOMAINS;
|
|
|
|
|
|
|
|
ret = ring->flush(ring, I915_GEM_GPU_DOMAINS, flush_domains);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
trace_i915_gem_ring_flush(ring, I915_GEM_GPU_DOMAINS, flush_domains);
|
|
|
|
|
|
|
|
ring->gpu_caches_dirty = false;
|
|
|
|
return 0;
|
|
|
|
}
|