linux/drivers/gpu/drm/i915/intel_ringbuffer.c

2110 lines
55 KiB
C
Raw Normal View History

/*
* 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>
*
*/
#include <drm/drmP.h>
#include "i915_drv.h"
#include <drm/i915_drm.h>
#include "i915_trace.h"
#include "intel_drv.h"
static inline int ring_space(struct intel_ring_buffer *ring)
{
int space = (ring->head & HEAD_ADDR) - (ring->tail + I915_RING_FREE_SPACE);
if (space < 0)
space += ring->size;
return space;
}
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-10 21:16:32 +00:00
void __intel_ring_advance(struct intel_ring_buffer *ring)
{
struct drm_i915_private *dev_priv = ring->dev->dev_private;
ring->tail &= ring->size - 1;
if (dev_priv->gpu_error.stop_rings & intel_ring_flag(ring))
return;
ring->write_tail(ring, ring->tail);
}
static int
gen2_render_ring_flush(struct intel_ring_buffer *ring,
u32 invalidate_domains,
u32 flush_domains)
{
u32 cmd;
int ret;
cmd = MI_FLUSH;
if (((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER) == 0)
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)
{
struct drm_device *dev = ring->dev;
u32 cmd;
int ret;
/*
* 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;
if ((invalidate_domains|flush_domains) & I915_GEM_DOMAIN_RENDER)
cmd &= ~MI_NO_WRITE_FLUSH;
if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION)
cmd |= MI_EXE_FLUSH;
if (invalidate_domains & I915_GEM_DOMAIN_COMMAND &&
(IS_G4X(dev) || IS_GEN5(dev)))
cmd |= MI_INVALIDATE_ISP;
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;
}
/**
* 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)
{
u32 scratch_addr = ring->scratch.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;
u32 scratch_addr = ring->scratch.gtt_offset + 128;
int ret;
/* Force SNB workarounds for PIPE_CONTROL flushes */
ret = intel_emit_post_sync_nonzero_flush(ring);
if (ret)
return ret;
/* 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;
/*
* Ensure that any following seqno writes only happen
* when the render cache is indeed flushed.
*/
flags |= PIPE_CONTROL_CS_STALL;
}
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 | PIPE_CONTROL_CS_STALL;
}
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4));
intel_ring_emit(ring, flags);
intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT);
intel_ring_emit(ring, 0);
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>
2012-08-17 21:35:43 +00:00
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;
}
static int gen7_ring_fbc_flush(struct intel_ring_buffer *ring, u32 value)
{
int ret;
if (!ring->fbc_dirty)
return 0;
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
intel_ring_emit(ring, MI_NOOP);
/* WaFbcNukeOn3DBlt:ivb/hsw */
intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1));
intel_ring_emit(ring, MSG_FBC_REND_STATE);
intel_ring_emit(ring, value);
intel_ring_advance(ring);
ring->fbc_dirty = false;
return 0;
}
static int
gen7_render_ring_flush(struct intel_ring_buffer *ring,
u32 invalidate_domains, u32 flush_domains)
{
u32 flags = 0;
u32 scratch_addr = ring->scratch.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>
2012-08-17 21:35:43 +00:00
/*
* 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;
/* 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;
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>
2012-08-17 21:35:43 +00:00
/* 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);
}
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4));
intel_ring_emit(ring, flags);
intel_ring_emit(ring, scratch_addr);
intel_ring_emit(ring, 0);
intel_ring_advance(ring);
if (flush_domains)
return gen7_ring_fbc_flush(ring, FBC_REND_NUKE);
return 0;
}
static void ring_write_tail(struct intel_ring_buffer *ring,
u32 value)
{
drm_i915_private_t *dev_priv = ring->dev->dev_private;
I915_WRITE_TAIL(ring, value);
}
u32 intel_ring_get_active_head(struct intel_ring_buffer *ring)
{
drm_i915_private_t *dev_priv = ring->dev->dev_private;
u32 acthd_reg = INTEL_INFO(ring->dev)->gen >= 4 ?
RING_ACTHD(ring->mmio_base) : ACTHD;
return I915_READ(acthd_reg);
}
static void ring_setup_phys_status_page(struct intel_ring_buffer *ring)
{
struct drm_i915_private *dev_priv = ring->dev->dev_private;
u32 addr;
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);
}
static int init_ring_common(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
struct drm_i915_gem_object *obj = ring->obj;
int ret = 0;
u32 head;
gen6_gt_force_wake_get(dev_priv);
if (I915_NEED_GFX_HWS(dev))
intel_ring_setup_status_page(ring);
else
ring_setup_phys_status_page(ring);
/* Stop the ring if it's running. */
I915_WRITE_CTL(ring, 0);
I915_WRITE_HEAD(ring, 0);
ring->write_tail(ring, 0);
head = I915_READ_HEAD(ring) & HEAD_ADDR;
/* G45 ring initialization fails to reset head to zero */
if (head != 0) {
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));
I915_WRITE_HEAD(ring, 0);
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));
}
}
/* 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, i915_gem_obj_ggtt_offset(obj));
I915_WRITE_CTL(ring,
((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 13:34:13 +00:00
| RING_VALID);
/* If the head is still not zero, the ring is dead */
if (wait_for((I915_READ_CTL(ring) & RING_VALID) != 0 &&
I915_READ_START(ring) == i915_gem_obj_ggtt_offset(obj) &&
(I915_READ_HEAD(ring) & HEAD_ADDR) == 0, 50)) {
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));
ret = -EIO;
goto out;
}
if (!drm_core_check_feature(ring->dev, DRIVER_MODESET))
i915_kernel_lost_context(ring->dev);
else {
ring->head = I915_READ_HEAD(ring);
ring->tail = I915_READ_TAIL(ring) & TAIL_ADDR;
ring->space = ring_space(ring);
ring->last_retired_head = -1;
}
memset(&ring->hangcheck, 0, sizeof(ring->hangcheck));
out:
gen6_gt_force_wake_put(dev_priv);
return ret;
}
static int
init_pipe_control(struct intel_ring_buffer *ring)
{
int ret;
if (ring->scratch.obj)
return 0;
ring->scratch.obj = i915_gem_alloc_object(ring->dev, 4096);
if (ring->scratch.obj == NULL) {
DRM_ERROR("Failed to allocate seqno page\n");
ret = -ENOMEM;
goto err;
}
i915_gem_object_set_cache_level(ring->scratch.obj, I915_CACHE_LLC);
ret = i915_gem_obj_ggtt_pin(ring->scratch.obj, 4096, true, false);
if (ret)
goto err_unref;
ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(ring->scratch.obj);
ring->scratch.cpu_page = kmap(sg_page(ring->scratch.obj->pages->sgl));
if (ring->scratch.cpu_page == NULL) {
ret = -ENOMEM;
goto err_unpin;
}
DRM_DEBUG_DRIVER("%s pipe control offset: 0x%08x\n",
ring->name, ring->scratch.gtt_offset);
return 0;
err_unpin:
i915_gem_object_unpin(ring->scratch.obj);
err_unref:
drm_gem_object_unreference(&ring->scratch.obj->base);
err:
return ret;
}
static int init_render_ring(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
struct drm_i915_private *dev_priv = dev->dev_private;
int ret = init_ring_common(ring);
if (INTEL_INFO(dev)->gen > 3)
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH));
/* 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.
*
* WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv
*/
if (INTEL_INFO(dev)->gen >= 6)
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(ASYNC_FLIP_PERF_DISABLE));
/* 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));
if (IS_GEN7(dev))
I915_WRITE(GFX_MODE_GEN7,
_MASKED_BIT_DISABLE(GFX_TLB_INVALIDATE_ALWAYS) |
_MASKED_BIT_ENABLE(GFX_REPLAY_MODE));
if (INTEL_INFO(dev)->gen >= 5) {
ret = init_pipe_control(ring);
if (ret)
return ret;
}
if (IS_GEN6(dev)) {
/* 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,
_MASKED_BIT_DISABLE(CM0_STC_EVICT_DISABLE_LRA_SNB));
/* 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);
}
if (INTEL_INFO(dev)->gen >= 6)
I915_WRITE(INSTPM, _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING));
if (HAS_L3_DPF(dev))
I915_WRITE_IMR(ring, ~GT_PARITY_ERROR(dev));
return ret;
}
static void render_ring_cleanup(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
if (ring->scratch.obj == NULL)
return;
if (INTEL_INFO(dev)->gen >= 5) {
kunmap(sg_page(ring->scratch.obj->pages->sgl));
i915_gem_object_unpin(ring->scratch.obj);
}
drm_gem_object_unreference(&ring->scratch.obj->base);
ring->scratch.obj = NULL;
}
static void
update_mboxes(struct intel_ring_buffer *ring,
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
u32 mmio_offset)
{
/* 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
intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1));
intel_ring_emit(ring, mmio_offset);
intel_ring_emit(ring, ring->outstanding_lazy_seqno);
intel_ring_emit(ring, MI_NOOP);
}
/**
* 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.
*/
static int
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
gen6_add_request(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
struct drm_i915_private *dev_priv = dev->dev_private;
struct intel_ring_buffer *useless;
int i, ret;
ret = intel_ring_begin(ring, ((I915_NUM_RINGS-1) *
MBOX_UPDATE_DWORDS) +
4);
if (ret)
return ret;
#undef MBOX_UPDATE_DWORDS
for_each_ring(useless, dev_priv, i) {
u32 mbox_reg = ring->signal_mbox[i];
if (mbox_reg != GEN6_NOSYNC)
update_mboxes(ring, mbox_reg);
}
intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
intel_ring_emit(ring, ring->outstanding_lazy_seqno);
intel_ring_emit(ring, MI_USER_INTERRUPT);
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-10 21:16:32 +00:00
__intel_ring_advance(ring);
return 0;
}
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;
}
/**
* 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
gen6_ring_sync(struct intel_ring_buffer *waiter,
struct intel_ring_buffer *signaller,
u32 seqno)
{
int ret;
u32 dw1 = MI_SEMAPHORE_MBOX |
MI_SEMAPHORE_COMPARE |
MI_SEMAPHORE_REGISTER;
/* 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;
WARN_ON(signaller->semaphore_register[waiter->id] ==
MI_SEMAPHORE_SYNC_INVALID);
ret = intel_ring_begin(waiter, 4);
if (ret)
return ret;
/* 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);
}
intel_ring_advance(waiter);
return 0;
}
#define PIPE_CONTROL_FLUSH(ring__, addr__) \
do { \
intel_ring_emit(ring__, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | \
PIPE_CONTROL_DEPTH_STALL); \
intel_ring_emit(ring__, (addr__) | PIPE_CONTROL_GLOBAL_GTT); \
intel_ring_emit(ring__, 0); \
intel_ring_emit(ring__, 0); \
} while (0)
static int
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
pc_render_add_request(struct intel_ring_buffer *ring)
{
u32 scratch_addr = ring->scratch.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;
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
PIPE_CONTROL_WRITE_FLUSH |
PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);
intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
intel_ring_emit(ring, ring->outstanding_lazy_seqno);
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);
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
PIPE_CONTROL_WRITE_FLUSH |
PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
PIPE_CONTROL_NOTIFY);
intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
intel_ring_emit(ring, ring->outstanding_lazy_seqno);
intel_ring_emit(ring, 0);
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-10 21:16:32 +00:00
__intel_ring_advance(ring);
return 0;
}
2012-12-14 15:01:25 +00:00
static u32
gen6_ring_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
2012-12-14 15:01:25 +00:00
{
/* 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. */
if (!lazy_coherency)
2012-12-14 15:01:25 +00:00
intel_ring_get_active_head(ring);
return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
}
static u32
ring_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
{
return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
}
static void
ring_set_seqno(struct intel_ring_buffer *ring, u32 seqno)
{
intel_write_status_page(ring, I915_GEM_HWS_INDEX, seqno);
}
static u32
pc_render_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency)
{
return ring->scratch.cpu_page[0];
}
static void
pc_render_set_seqno(struct intel_ring_buffer *ring, u32 seqno)
{
ring->scratch.cpu_page[0] = seqno;
}
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;
unsigned long flags;
if (!dev->irq_enabled)
return false;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (ring->irq_refcount++ == 0)
ilk_enable_gt_irq(dev_priv, ring->irq_enable_mask);
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
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;
unsigned long flags;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (--ring->irq_refcount == 0)
ilk_disable_gt_irq(dev_priv, ring->irq_enable_mask);
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
}
static bool
i9xx_ring_get_irq(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
unsigned long flags;
if (!dev->irq_enabled)
return false;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (ring->irq_refcount++ == 0) {
dev_priv->irq_mask &= ~ring->irq_enable_mask;
I915_WRITE(IMR, dev_priv->irq_mask);
POSTING_READ(IMR);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
return true;
}
static void
i9xx_ring_put_irq(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
unsigned long flags;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (--ring->irq_refcount == 0) {
dev_priv->irq_mask |= ring->irq_enable_mask;
I915_WRITE(IMR, dev_priv->irq_mask);
POSTING_READ(IMR);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
}
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;
unsigned long flags;
if (!dev->irq_enabled)
return false;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (ring->irq_refcount++ == 0) {
dev_priv->irq_mask &= ~ring->irq_enable_mask;
I915_WRITE16(IMR, dev_priv->irq_mask);
POSTING_READ16(IMR);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
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;
unsigned long flags;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (--ring->irq_refcount == 0) {
dev_priv->irq_mask |= ring->irq_enable_mask;
I915_WRITE16(IMR, dev_priv->irq_mask);
POSTING_READ16(IMR);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
}
void intel_ring_setup_status_page(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = ring->dev->dev_private;
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) {
case RCS:
mmio = RENDER_HWS_PGA_GEN7;
break;
case BCS:
mmio = BLT_HWS_PGA_GEN7;
break;
case VCS:
mmio = BSD_HWS_PGA_GEN7;
break;
case VECS:
mmio = VEBOX_HWS_PGA_GEN7;
break;
}
} else if (IS_GEN6(ring->dev)) {
mmio = RING_HWS_PGA_GEN6(ring->mmio_base);
} else {
mmio = RING_HWS_PGA(ring->mmio_base);
}
I915_WRITE(mmio, (u32)ring->status_page.gfx_addr);
POSTING_READ(mmio);
/* Flush the TLB for this page */
if (INTEL_INFO(dev)->gen >= 6) {
u32 reg = RING_INSTPM(ring->mmio_base);
I915_WRITE(reg,
_MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE |
INSTPM_SYNC_FLUSH));
if (wait_for((I915_READ(reg) & INSTPM_SYNC_FLUSH) == 0,
1000))
DRM_ERROR("%s: wait for SyncFlush to complete for TLB invalidation timed out\n",
ring->name);
}
}
static int
bsd_ring_flush(struct intel_ring_buffer *ring,
u32 invalidate_domains,
u32 flush_domains)
{
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;
}
static int
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
i9xx_add_request(struct intel_ring_buffer *ring)
{
int ret;
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
intel_ring_emit(ring, ring->outstanding_lazy_seqno);
intel_ring_emit(ring, MI_USER_INTERRUPT);
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-10 21:16:32 +00:00
__intel_ring_advance(ring);
return 0;
}
static bool
gen6_ring_get_irq(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
unsigned long flags;
if (!dev->irq_enabled)
return false;
2012-12-14 15:01:25 +00:00
/* 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. */
gen6_gt_force_wake_get(dev_priv);
2012-12-14 15:01:25 +00:00
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (ring->irq_refcount++ == 0) {
if (HAS_L3_DPF(dev) && ring->id == RCS)
I915_WRITE_IMR(ring,
~(ring->irq_enable_mask |
GT_PARITY_ERROR(dev)));
else
I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
ilk_enable_gt_irq(dev_priv, ring->irq_enable_mask);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
return true;
}
static void
gen6_ring_put_irq(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
drm_i915_private_t *dev_priv = dev->dev_private;
unsigned long flags;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (--ring->irq_refcount == 0) {
if (HAS_L3_DPF(dev) && ring->id == RCS)
I915_WRITE_IMR(ring, ~GT_PARITY_ERROR(dev));
else
I915_WRITE_IMR(ring, ~0);
ilk_disable_gt_irq(dev_priv, ring->irq_enable_mask);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
2012-12-14 15:01:25 +00:00
gen6_gt_force_wake_put(dev_priv);
}
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->irq_lock, flags);
if (ring->irq_refcount++ == 0) {
I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
snb_enable_pm_irq(dev_priv, ring->irq_enable_mask);
}
spin_unlock_irqrestore(&dev_priv->irq_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->irq_lock, flags);
if (--ring->irq_refcount == 0) {
I915_WRITE_IMR(ring, ~0);
snb_disable_pm_irq(dev_priv, ring->irq_enable_mask);
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
}
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
static bool
gen8_ring_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->irq_lock, flags);
if (ring->irq_refcount++ == 0) {
if (HAS_L3_DPF(dev) && ring->id == RCS) {
I915_WRITE_IMR(ring,
~(ring->irq_enable_mask |
GT_RENDER_L3_PARITY_ERROR_INTERRUPT));
} else {
I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
}
POSTING_READ(RING_IMR(ring->mmio_base));
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
return true;
}
static void
gen8_ring_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;
spin_lock_irqsave(&dev_priv->irq_lock, flags);
if (--ring->irq_refcount == 0) {
if (HAS_L3_DPF(dev) && ring->id == RCS) {
I915_WRITE_IMR(ring,
~GT_RENDER_L3_PARITY_ERROR_INTERRUPT);
} else {
I915_WRITE_IMR(ring, ~0);
}
POSTING_READ(RING_IMR(ring->mmio_base));
}
spin_unlock_irqrestore(&dev_priv->irq_lock, flags);
}
static int
i965_dispatch_execbuffer(struct intel_ring_buffer *ring,
u32 offset, u32 length,
unsigned flags)
{
int ret;
ret = intel_ring_begin(ring, 2);
if (ret)
return ret;
intel_ring_emit(ring,
MI_BATCH_BUFFER_START |
MI_BATCH_GTT |
(flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965));
intel_ring_emit(ring, offset);
intel_ring_advance(ring);
return 0;
}
/* Just userspace ABI convention to limit the wa batch bo to a resonable size */
#define I830_BATCH_LIMIT (256*1024)
static int
i830_dispatch_execbuffer(struct intel_ring_buffer *ring,
u32 offset, u32 len,
unsigned flags)
{
int ret;
if (flags & I915_DISPATCH_PINNED) {
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
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 {
u32 cs_offset = ring->scratch.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);
}
return 0;
}
static int
i915_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_GTT);
intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE));
intel_ring_advance(ring);
return 0;
}
static void cleanup_status_page(struct intel_ring_buffer *ring)
{
struct drm_i915_gem_object *obj;
obj = ring->status_page.obj;
if (obj == NULL)
return;
kunmap(sg_page(obj->pages->sgl));
i915_gem_object_unpin(obj);
drm_gem_object_unreference(&obj->base);
ring->status_page.obj = NULL;
}
static int init_status_page(struct intel_ring_buffer *ring)
{
struct drm_device *dev = ring->dev;
struct drm_i915_gem_object *obj;
int ret;
obj = i915_gem_alloc_object(dev, 4096);
if (obj == NULL) {
DRM_ERROR("Failed to allocate status page\n");
ret = -ENOMEM;
goto err;
}
i915_gem_object_set_cache_level(obj, I915_CACHE_LLC);
ret = i915_gem_obj_ggtt_pin(obj, 4096, true, false);
if (ret != 0) {
goto err_unref;
}
ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(obj);
ring->status_page.page_addr = kmap(sg_page(obj->pages->sgl));
if (ring->status_page.page_addr == NULL) {
ret = -ENOMEM;
goto err_unpin;
}
ring->status_page.obj = obj;
memset(ring->status_page.page_addr, 0, PAGE_SIZE);
DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n",
ring->name, ring->status_page.gfx_addr);
return 0;
err_unpin:
i915_gem_object_unpin(obj);
err_unref:
drm_gem_object_unreference(&obj->base);
err:
return ret;
}
static int init_phys_status_page(struct intel_ring_buffer *ring)
{
struct drm_i915_private *dev_priv = ring->dev->dev_private;
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;
}
ring->status_page.page_addr = dev_priv->status_page_dmah->vaddr;
memset(ring->status_page.page_addr, 0, PAGE_SIZE);
return 0;
}
static int intel_init_ring_buffer(struct drm_device *dev,
struct intel_ring_buffer *ring)
{
struct drm_i915_gem_object *obj;
struct drm_i915_private *dev_priv = dev->dev_private;
int ret;
ring->dev = dev;
INIT_LIST_HEAD(&ring->active_list);
INIT_LIST_HEAD(&ring->request_list);
ring->size = 32 * PAGE_SIZE;
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
memset(ring->sync_seqno, 0, sizeof(ring->sync_seqno));
init_waitqueue_head(&ring->irq_queue);
if (I915_NEED_GFX_HWS(dev)) {
ret = init_status_page(ring);
if (ret)
return ret;
} else {
BUG_ON(ring->id != RCS);
ret = init_phys_status_page(ring);
if (ret)
return ret;
}
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);
if (obj == NULL) {
DRM_ERROR("Failed to allocate ringbuffer\n");
ret = -ENOMEM;
goto err_hws;
}
ring->obj = obj;
ret = i915_gem_obj_ggtt_pin(obj, PAGE_SIZE, true, false);
if (ret)
goto err_unref;
ret = i915_gem_object_set_to_gtt_domain(obj, true);
if (ret)
goto err_unpin;
ring->virtual_start =
ioremap_wc(dev_priv->gtt.mappable_base + i915_gem_obj_ggtt_offset(obj),
ring->size);
if (ring->virtual_start == NULL) {
DRM_ERROR("Failed to map ringbuffer.\n");
ret = -EINVAL;
goto err_unpin;
}
ret = ring->init(ring);
if (ret)
goto err_unmap;
/* 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;
if (IS_I830(ring->dev) || IS_845G(ring->dev))
ring->effective_size -= 128;
return 0;
err_unmap:
iounmap(ring->virtual_start);
err_unpin:
i915_gem_object_unpin(obj);
err_unref:
drm_gem_object_unreference(&obj->base);
ring->obj = NULL;
err_hws:
cleanup_status_page(ring);
return ret;
}
void intel_cleanup_ring_buffer(struct intel_ring_buffer *ring)
{
struct drm_i915_private *dev_priv;
int ret;
if (ring->obj == NULL)
return;
/* Disable the ring buffer. The ring must be idle at this point */
dev_priv = ring->dev->dev_private;
ret = intel_ring_idle(ring);
if (ret && !i915_reset_in_progress(&dev_priv->gpu_error))
DRM_ERROR("failed to quiesce %s whilst cleaning up: %d\n",
ring->name, ret);
I915_WRITE_CTL(ring, 0);
iounmap(ring->virtual_start);
i915_gem_object_unpin(ring->obj);
drm_gem_object_unreference(&ring->obj->base);
ring->obj = NULL;
ring->preallocated_lazy_request = NULL;
ring->outstanding_lazy_seqno = 0;
if (ring->cleanup)
ring->cleanup(ring);
cleanup_status_page(ring);
}
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
static int intel_ring_wait_seqno(struct intel_ring_buffer *ring, u32 seqno)
{
int ret;
ret = i915_wait_seqno(ring, seqno);
if (!ret)
i915_gem_retire_requests_ring(ring);
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
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;
space = request->tail - (ring->tail + I915_RING_FREE_SPACE);
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
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;
}
static int ring_wait_for_space(struct intel_ring_buffer *ring, int n)
{
struct drm_device *dev = ring->dev;
struct drm_i915_private *dev_priv = dev->dev_private;
unsigned long end;
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
int ret;
drm/i915: Record the tail at each request and use it to estimate the head By recording the location of every request in the ringbuffer, we know that in order to retire the request the GPU must have finished reading it and so the GPU head is now beyond the tail of the request. We can therefore provide a conservative estimate of where the GPU is reading from in order to avoid having to read back the ring buffer registers when polling for space upon starting a new write into the ringbuffer. A secondary effect is that this allows us to convert intel_ring_buffer_wait() to use i915_wait_request() and so consolidate upon the single function to handle the complicated task of waiting upon the GPU. A necessary precaution is that we need to make that wait uninterruptible to match the existing conditions as all the callers of intel_ring_begin() have not been audited to handle ERESTARTSYS correctly. By using a conservative estimate for the head, and always processing all outstanding requests first, we prevent a race condition between using the estimate and direct reads of I915_RING_HEAD which could result in the value of the head going backwards, and the tail overflowing once again. We are also careful to mark any request that we skip over in order to free space in ring as consumed which provides a self-consistency check. Given sufficient abuse, such as a set of unthrottled GPU bound cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on Sandy Bridge (i5-2520m): firefox-paintball 18927ms -> 15646ms: 1.21x speedup firefox-fishtank 12563ms -> 11278ms: 1.11x speedup which is a mild consolation for the performance those traces achieved from exploiting the buggy autoreported head. v2: Add a few more comments and make request->tail a conservative estimate as suggested by Daniel Vetter. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: resolve conflicts with retirement defering and the lack of the autoreport head removal (that will go in through -fixes).] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-02-15 11:25:36 +00:00
ret = intel_ring_wait_request(ring, n);
if (ret != -ENOSPC)
return ret;
drm/i915: Write RING_TAIL once per-request Ignoring the legacy DRI1 code, and a couple of special cases (to be discussed later), all access to the ring is mediated through requests. The first write to a ring will grab a seqno and mark the ring as having an outstanding_lazy_request. Either through explicitly adding a request after an execbuffer or through an implicit wait (either by the CPU or by a semaphore), that sequence of writes will be terminated with a request. So we can ellide all the intervening writes to the tail register and send the entire command stream to the GPU at once. This will reduce the number of *serialising* writes to the tail register by a factor or 3-5 times (depending upon architecture and number of workarounds, context switches, etc involved). This becomes even more noticeable when the register write is overloaded with a number of debugging tools. The astute reader will wonder if it is then possible to overflow the ring with a single command. It is not. When we start a command sequence to the ring, we check for available space and issue a wait in case we have not. The ring wait will in this case be forced to flush the outstanding register write and then poll the ACTHD for sufficient space to continue. The exception to the rule where everything is inside a request are a few initialisation cases where we may want to write GPU commands via the CS before userspace wakes up and page flips. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-10 21:16:32 +00:00
/* force the tail write in case we have been skipping them */
__intel_ring_advance(ring);
trace_i915_ring_wait_begin(ring);
/* 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;
do {
ring->head = I915_READ_HEAD(ring);
ring->space = ring_space(ring);
if (ring->space >= n) {
trace_i915_ring_wait_end(ring);
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;
}
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>
2012-07-04 20:54:13 +00:00
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>
2012-07-04 20:54:13 +00:00
if (ret)
return ret;
} while (!time_after(jiffies, end));
trace_i915_ring_wait_end(ring);
return -EBUSY;
}
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_seqno) {
ret = i915_add_request(ring, 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);
}
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
static int
intel_ring_alloc_seqno(struct intel_ring_buffer *ring)
{
if (ring->outstanding_lazy_seqno)
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
return 0;
if (ring->preallocated_lazy_request == NULL) {
struct drm_i915_gem_request *request;
request = kmalloc(sizeof(*request), GFP_KERNEL);
if (request == NULL)
return -ENOMEM;
ring->preallocated_lazy_request = request;
}
return i915_gem_get_seqno(ring->dev, &ring->outstanding_lazy_seqno);
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
}
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;
}
int intel_ring_begin(struct intel_ring_buffer *ring,
int num_dwords)
{
drm_i915_private_t *dev_priv = ring->dev->dev_private;
int ret;
ret = i915_gem_check_wedge(&dev_priv->gpu_error,
dev_priv->mm.interruptible);
if (ret)
return ret;
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
/* Preallocate the olr before touching the ring */
ret = intel_ring_alloc_seqno(ring);
if (ret)
return ret;
return __intel_ring_begin(ring, num_dwords * sizeof(uint32_t));
}
void intel_ring_init_seqno(struct intel_ring_buffer *ring, u32 seqno)
{
struct drm_i915_private *dev_priv = ring->dev->dev_private;
BUG_ON(ring->outstanding_lazy_seqno);
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);
if (HAS_VEBOX(ring->dev))
I915_WRITE(RING_SYNC_2(ring->mmio_base), 0);
}
ring->set_seqno(ring, seqno);
ring->hangcheck.seqno = seqno;
}
static void gen6_bsd_ring_write_tail(struct intel_ring_buffer *ring,
u32 value)
{
drm_i915_private_t *dev_priv = ring->dev->dev_private;
/* Every tail move must follow the sequence below */
/* Disable notification that the ring is IDLE. The GT
* will then assume that it is busy and bring it out of rc6.
*/
I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
_MASKED_BIT_ENABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
/* Clear the context id. Here be magic! */
I915_WRITE64(GEN6_BSD_RNCID, 0x0);
/* Wait for the ring not to be idle, i.e. for it to wake up. */
if (wait_for((I915_READ(GEN6_BSD_SLEEP_PSMI_CONTROL) &
GEN6_BSD_SLEEP_INDICATOR) == 0,
50))
DRM_ERROR("timed out waiting for the BSD ring to wake up\n");
/* Now that the ring is fully powered up, update the tail */
I915_WRITE_TAIL(ring, value);
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.
*/
I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
_MASKED_BIT_DISABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
}
static int gen6_bsd_ring_flush(struct intel_ring_buffer *ring,
u32 invalidate, u32 flush)
{
uint32_t cmd;
int ret;
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
cmd = MI_FLUSH_DW;
/*
* 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."
*/
if (invalidate & I915_GEM_GPU_DOMAINS)
cmd |= MI_INVALIDATE_TLB | MI_INVALIDATE_BSD |
MI_FLUSH_DW_STORE_INDEX | MI_FLUSH_DW_OP_STOREDW;
intel_ring_emit(ring, cmd);
intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT);
intel_ring_emit(ring, 0);
intel_ring_emit(ring, MI_NOOP);
intel_ring_advance(ring);
return 0;
}
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;
}
static int
gen6_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 |
(flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965));
/* bit0-7 is the length on GEN6+ */
intel_ring_emit(ring, offset);
intel_ring_advance(ring);
return 0;
}
/* Blitter support (SandyBridge+) */
static int gen6_ring_flush(struct intel_ring_buffer *ring,
u32 invalidate, u32 flush)
{
struct drm_device *dev = ring->dev;
uint32_t cmd;
int ret;
ret = intel_ring_begin(ring, 4);
if (ret)
return ret;
cmd = MI_FLUSH_DW;
/*
* 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."
*/
if (invalidate & I915_GEM_DOMAIN_RENDER)
cmd |= MI_INVALIDATE_TLB | MI_FLUSH_DW_STORE_INDEX |
MI_FLUSH_DW_OP_STOREDW;
intel_ring_emit(ring, cmd);
intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT);
intel_ring_emit(ring, 0);
intel_ring_emit(ring, MI_NOOP);
intel_ring_advance(ring);
if (IS_GEN7(dev) && flush)
return gen7_ring_fbc_flush(ring, FBC_REND_CACHE_CLEAN);
return 0;
}
int intel_init_render_ring_buffer(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
struct intel_ring_buffer *ring = &dev_priv->ring[RCS];
ring->name = "render ring";
ring->id = RCS;
ring->mmio_base = RENDER_RING_BASE;
if (INTEL_INFO(dev)->gen >= 6) {
ring->add_request = gen6_add_request;
ring->flush = gen7_render_ring_flush;
if (INTEL_INFO(dev)->gen == 6)
ring->flush = gen6_render_ring_flush;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
if (INTEL_INFO(dev)->gen >= 8) {
ring->irq_get = gen8_ring_get_irq;
ring->irq_put = gen8_ring_put_irq;
} else {
ring->irq_get = gen6_ring_get_irq;
ring->irq_put = gen6_ring_put_irq;
}
ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
2012-12-14 15:01:25 +00:00
ring->get_seqno = gen6_ring_get_seqno;
ring->set_seqno = ring_set_seqno;
ring->sync_to = gen6_ring_sync;
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_INVALID;
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_RV;
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_RB;
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_RVE;
ring->signal_mbox[RCS] = GEN6_NOSYNC;
ring->signal_mbox[VCS] = GEN6_VRSYNC;
ring->signal_mbox[BCS] = GEN6_BRSYNC;
ring->signal_mbox[VECS] = GEN6_VERSYNC;
} else if (IS_GEN5(dev)) {
ring->add_request = pc_render_add_request;
ring->flush = gen4_render_ring_flush;
ring->get_seqno = pc_render_get_seqno;
ring->set_seqno = pc_render_set_seqno;
ring->irq_get = gen5_ring_get_irq;
ring->irq_put = gen5_ring_put_irq;
ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT |
GT_RENDER_PIPECTL_NOTIFY_INTERRUPT;
} else {
ring->add_request = i9xx_add_request;
if (INTEL_INFO(dev)->gen < 4)
ring->flush = gen2_render_ring_flush;
else
ring->flush = gen4_render_ring_flush;
ring->get_seqno = ring_get_seqno;
ring->set_seqno = ring_set_seqno;
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;
}
ring->irq_enable_mask = I915_USER_INTERRUPT;
}
ring->write_tail = ring_write_tail;
if (IS_HASWELL(dev))
ring->dispatch_execbuffer = hsw_ring_dispatch_execbuffer;
else if (INTEL_INFO(dev)->gen >= 6)
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;
ring->init = init_render_ring;
ring->cleanup = render_ring_cleanup;
/* 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_obj_ggtt_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->scratch.obj = obj;
ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(obj);
}
return intel_init_ring_buffer(dev, ring);
}
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
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];
int ret;
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
ring->name = "render ring";
ring->id = RCS;
ring->mmio_base = RENDER_RING_BASE;
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
if (INTEL_INFO(dev)->gen >= 6) {
/* non-kms not supported on gen6+ */
return -ENODEV;
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
}
/* 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;
if (INTEL_INFO(dev)->gen < 4)
ring->flush = gen2_render_ring_flush;
else
ring->flush = gen4_render_ring_flush;
ring->get_seqno = ring_get_seqno;
ring->set_seqno = ring_set_seqno;
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;
}
ring->irq_enable_mask = I915_USER_INTERRUPT;
ring->write_tail = ring_write_tail;
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;
ring->init = init_render_ring;
ring->cleanup = render_ring_cleanup;
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
ring->dev = dev;
INIT_LIST_HEAD(&ring->active_list);
INIT_LIST_HEAD(&ring->request_list);
ring->size = size;
ring->effective_size = ring->size;
if (IS_I830(ring->dev) || IS_845G(ring->dev))
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
ring->effective_size -= 128;
ring->virtual_start = ioremap_wc(start, size);
if (ring->virtual_start == NULL) {
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
DRM_ERROR("can not ioremap virtual address for"
" ring buffer\n");
return -ENOMEM;
}
if (!I915_NEED_GFX_HWS(dev)) {
ret = init_phys_status_page(ring);
if (ret)
return ret;
}
drm/i915: Initialise ring vfuncs for old DRI paths We weren't setting up the vfunc table when initialising the old DRI ringbuffer, leading to such OOPSes as: BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<(null)>] (null) PGD 10c441067 PUD 1185e5067 PMD 0 Oops: 0010 [#1] PREEMPT SMP last sysfs file: /sys/class/dmi/id/chassis_asset_tag CPU 3 Modules linked in: i915 drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea video backlight output cfbimgblt cfbfillrect autofs4 ipv6 nfs lockd fscache nfs_acl auth_rpcgss sunrpc coretemp hwmon_vid mousedev usbhid hid option usb_wwan snd_hda_codec_via asus_atk0110 atl1e usbserial snd_hda_intel snd_hda_codec firmware_class snd_hwdep snd_pcm snd_seq snd_timer snd_seq_device processor parport_pc thermal snd thermal_sys parport 8250_pnp button rng_core rtc_cmos shpchp hwmon rtc_core ehci_hcd pci_hotplug uhci_hcd soundcore tpm_tis i2c_i801 rtc_lib tpm serio_raw snd_page_alloc tpm_bios i2c_core usbcore psmouse intel_agp sg pcspkr sr_mod evdev cdrom ext3 jbd mbcache dm_mod sd_mod ata_piix libata scsi_mod unix Jan 18 15:49:29 lithui kernel: Pid: 3605, comm: Xorg Not tainted 2.6.36.2 #5 P5KPL-CM/System Product Name RIP: 0010:[<0000000000000000>] [<(null)>] (null) RSP: 0018:ffff8801150d1d40 EFLAGS: 00010202 RAX: 000000000001ffff RBX: ffff88011a011b00 RCX: 000000000001a704 RDX: ffff880118566028 RSI: ffff880118566028 RDI: ffff880117876800 RBP: ffff8801150d1d48 R08: ffff8801195fe300 R09: 00000000c0086444 R10: 0000000000000001 R11: 0000000000003206 R12: ffff880117876800 R13: ffff880118566000 R14: ffff880117876820 R15: ffff8801150d1df8 FS: 00007f1038d456e0(0000) GS:ffff880001780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000001187e7000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process Xorg (pid: 3605, threadinfo ffff8801150d0000, task ffff88011b016e40) Stack: ffffffffa043b8e6 ffff8801150d1d98 ffffffffa041768b dead000000000000 <0> 0000000000000048 00007f1023f2a000 0000000000000044 0000000000000008 <0> ffff88010d26bd80 ffff880117876800 ffff8801150d1df8 ffff8801150d1ea8 Call Trace: [<ffffffffa043b8e6>] ? intel_ring_advance+0x16/0x20 [i915] [<ffffffffa041768b>] i915_irq_emit+0x15b/0x240 [i915] [<ffffffffa03ea7b1>] drm_ioctl+0x1f1/0x460 [drm] [<ffffffffa0417530>] ? i915_irq_emit+0x0/0x240 [i915] [<ffffffff810dd8f1>] ? do_sync_read+0xd1/0x120 [<ffffffff81025b1f>] ? do_page_fault+0x1df/0x3d0 [<ffffffff810ed5c7>] do_vfs_ioctl+0x97/0x550 [<ffffffff8115c2ea>] ? security_file_permission+0x7a/0x90 [<ffffffff810edb19>] sys_ioctl+0x99/0xa0 [<ffffffff810024ab>] system_call_fastpath+0x16/0x1b Code: Bad RIP value. RIP [<(null)>] (null) RSP <ffff8801150d1d40> CR2: 0000000000000000 Reported-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Herbert Xu <herbert@gondor.apana.org.au> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=29153 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23172 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org
2011-01-20 09:57:11 +00:00
return 0;
}
int intel_init_bsd_ring_buffer(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
struct intel_ring_buffer *ring = &dev_priv->ring[VCS];
ring->name = "bsd ring";
ring->id = VCS;
ring->write_tail = ring_write_tail;
if (IS_GEN6(dev) || IS_GEN7(dev)) {
ring->mmio_base = GEN6_BSD_RING_BASE;
/* gen6 bsd needs a special wa for tail updates */
if (IS_GEN6(dev))
ring->write_tail = gen6_bsd_ring_write_tail;
ring->flush = gen6_bsd_ring_flush;
ring->add_request = gen6_add_request;
ring->get_seqno = gen6_ring_get_seqno;
ring->set_seqno = ring_set_seqno;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
if (INTEL_INFO(dev)->gen >= 8) {
ring->irq_enable_mask =
GT_RENDER_USER_INTERRUPT << GEN8_VCS1_IRQ_SHIFT;
ring->irq_get = gen8_ring_get_irq;
ring->irq_put = gen8_ring_put_irq;
} else {
ring->irq_enable_mask = GT_BSD_USER_INTERRUPT;
ring->irq_get = gen6_ring_get_irq;
ring->irq_put = gen6_ring_put_irq;
}
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
ring->sync_to = gen6_ring_sync;
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_VR;
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_INVALID;
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_VB;
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_VVE;
ring->signal_mbox[RCS] = GEN6_RVSYNC;
ring->signal_mbox[VCS] = GEN6_NOSYNC;
ring->signal_mbox[BCS] = GEN6_BVSYNC;
ring->signal_mbox[VECS] = GEN6_VEVSYNC;
} else {
ring->mmio_base = BSD_RING_BASE;
ring->flush = bsd_ring_flush;
ring->add_request = i9xx_add_request;
ring->get_seqno = ring_get_seqno;
ring->set_seqno = ring_set_seqno;
if (IS_GEN5(dev)) {
ring->irq_enable_mask = ILK_BSD_USER_INTERRUPT;
ring->irq_get = gen5_ring_get_irq;
ring->irq_put = gen5_ring_put_irq;
} else {
ring->irq_enable_mask = I915_BSD_USER_INTERRUPT;
ring->irq_get = i9xx_ring_get_irq;
ring->irq_put = i9xx_ring_put_irq;
}
ring->dispatch_execbuffer = i965_dispatch_execbuffer;
}
ring->init = init_ring_common;
return intel_init_ring_buffer(dev, ring);
}
int intel_init_blt_ring_buffer(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
struct intel_ring_buffer *ring = &dev_priv->ring[BCS];
ring->name = "blitter ring";
ring->id = BCS;
ring->mmio_base = BLT_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;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
if (INTEL_INFO(dev)->gen >= 8) {
ring->irq_enable_mask =
GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
ring->irq_get = gen8_ring_get_irq;
ring->irq_put = gen8_ring_put_irq;
} else {
ring->irq_enable_mask = GT_BLT_USER_INTERRUPT;
ring->irq_get = gen6_ring_get_irq;
ring->irq_put = gen6_ring_put_irq;
}
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
ring->sync_to = gen6_ring_sync;
ring->semaphore_register[RCS] = MI_SEMAPHORE_SYNC_BR;
ring->semaphore_register[VCS] = MI_SEMAPHORE_SYNC_BV;
ring->semaphore_register[BCS] = MI_SEMAPHORE_SYNC_INVALID;
ring->semaphore_register[VECS] = MI_SEMAPHORE_SYNC_BVE;
ring->signal_mbox[RCS] = GEN6_RBSYNC;
ring->signal_mbox[VCS] = GEN6_VBSYNC;
ring->signal_mbox[BCS] = GEN6_NOSYNC;
ring->signal_mbox[VECS] = GEN6_VEBSYNC;
ring->init = init_ring_common;
return intel_init_ring_buffer(dev, ring);
}
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;
ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
if (INTEL_INFO(dev)->gen >= 8) {
ring->irq_enable_mask =
(GT_RENDER_USER_INTERRUPT << GEN8_VECS_IRQ_SHIFT) |
GT_RENDER_CS_MASTER_ERROR_INTERRUPT;
ring->irq_get = gen8_ring_get_irq;
ring->irq_put = gen8_ring_put_irq;
} else {
ring->irq_enable_mask = PM_VEBOX_USER_INTERRUPT;
ring->irq_get = hsw_vebox_get_irq;
ring->irq_put = hsw_vebox_put_irq;
}
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);
}
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;
}