linux_dsm_epyc7002/drivers/gpu/drm/i915/gvt/display.c

532 lines
16 KiB
C
Raw Normal View History

/*
* Copyright(c) 2011-2016 Intel Corporation. All rights reserved.
*
* 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:
* Ke Yu
* Zhiyuan Lv <zhiyuan.lv@intel.com>
*
* Contributors:
* Terrence Xu <terrence.xu@intel.com>
* Changbin Du <changbin.du@intel.com>
* Bing Niu <bing.niu@intel.com>
* Zhi Wang <zhi.a.wang@intel.com>
*
*/
#include "i915_drv.h"
#include "gvt.h"
static int get_edp_pipe(struct intel_vgpu *vgpu)
{
u32 data = vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP);
int pipe = -1;
switch (data & TRANS_DDI_EDP_INPUT_MASK) {
case TRANS_DDI_EDP_INPUT_A_ON:
case TRANS_DDI_EDP_INPUT_A_ONOFF:
pipe = PIPE_A;
break;
case TRANS_DDI_EDP_INPUT_B_ONOFF:
pipe = PIPE_B;
break;
case TRANS_DDI_EDP_INPUT_C_ONOFF:
pipe = PIPE_C;
break;
}
return pipe;
}
static int edp_pipe_is_enabled(struct intel_vgpu *vgpu)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
if (!(vgpu_vreg_t(vgpu, PIPECONF(_PIPE_EDP)) & PIPECONF_ENABLE))
return 0;
if (!(vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP) & TRANS_DDI_FUNC_ENABLE))
return 0;
return 1;
}
int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
if (WARN_ON(pipe < PIPE_A || pipe >= I915_MAX_PIPES))
return -EINVAL;
if (vgpu_vreg_t(vgpu, PIPECONF(pipe)) & PIPECONF_ENABLE)
return 1;
if (edp_pipe_is_enabled(vgpu) &&
get_edp_pipe(vgpu) == pipe)
return 1;
return 0;
}
static unsigned char virtual_dp_monitor_edid[GVT_EDID_NUM][EDID_SIZE] = {
{
/* EDID with 1024x768 as its resolution */
/*Header*/
0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00,
/* Vendor & Product Identification */
0x22, 0xf0, 0x54, 0x29, 0x00, 0x00, 0x00, 0x00, 0x04, 0x17,
/* Version & Revision */
0x01, 0x04,
/* Basic Display Parameters & Features */
0xa5, 0x34, 0x20, 0x78, 0x23,
/* Color Characteristics */
0xfc, 0x81, 0xa4, 0x55, 0x4d, 0x9d, 0x25, 0x12, 0x50, 0x54,
/* Established Timings: maximum resolution is 1024x768 */
0x21, 0x08, 0x00,
/* Standard Timings. All invalid */
0x00, 0xc0, 0x00, 0xc0, 0x00, 0x40, 0x00, 0x80, 0x00, 0x00,
0x00, 0x40, 0x00, 0x00, 0x00, 0x01,
/* 18 Byte Data Blocks 1: invalid */
0x00, 0x00, 0x80, 0xa0, 0x70, 0xb0,
0x23, 0x40, 0x30, 0x20, 0x36, 0x00, 0x06, 0x44, 0x21, 0x00, 0x00, 0x1a,
/* 18 Byte Data Blocks 2: invalid */
0x00, 0x00, 0x00, 0xfd, 0x00, 0x18, 0x3c, 0x18, 0x50, 0x11, 0x00, 0x0a,
0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
/* 18 Byte Data Blocks 3: invalid */
0x00, 0x00, 0x00, 0xfc, 0x00, 0x48,
0x50, 0x20, 0x5a, 0x52, 0x32, 0x34, 0x34, 0x30, 0x77, 0x0a, 0x20, 0x20,
/* 18 Byte Data Blocks 4: invalid */
0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 0x4e, 0x34, 0x33, 0x30, 0x34, 0x30,
0x44, 0x58, 0x51, 0x0a, 0x20, 0x20,
/* Extension Block Count */
0x00,
/* Checksum */
0xef,
},
{
/* EDID with 1920x1200 as its resolution */
/*Header*/
0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00,
/* Vendor & Product Identification */
0x22, 0xf0, 0x54, 0x29, 0x00, 0x00, 0x00, 0x00, 0x04, 0x17,
/* Version & Revision */
0x01, 0x04,
/* Basic Display Parameters & Features */
0xa5, 0x34, 0x20, 0x78, 0x23,
/* Color Characteristics */
0xfc, 0x81, 0xa4, 0x55, 0x4d, 0x9d, 0x25, 0x12, 0x50, 0x54,
/* Established Timings: maximum resolution is 1024x768 */
0x21, 0x08, 0x00,
/*
* Standard Timings.
* below new resolutions can be supported:
* 1920x1080, 1280x720, 1280x960, 1280x1024,
* 1440x900, 1600x1200, 1680x1050
*/
0xd1, 0xc0, 0x81, 0xc0, 0x81, 0x40, 0x81, 0x80, 0x95, 0x00,
0xa9, 0x40, 0xb3, 0x00, 0x01, 0x01,
/* 18 Byte Data Blocks 1: max resolution is 1920x1200 */
0x28, 0x3c, 0x80, 0xa0, 0x70, 0xb0,
0x23, 0x40, 0x30, 0x20, 0x36, 0x00, 0x06, 0x44, 0x21, 0x00, 0x00, 0x1a,
/* 18 Byte Data Blocks 2: invalid */
0x00, 0x00, 0x00, 0xfd, 0x00, 0x18, 0x3c, 0x18, 0x50, 0x11, 0x00, 0x0a,
0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
/* 18 Byte Data Blocks 3: invalid */
0x00, 0x00, 0x00, 0xfc, 0x00, 0x48,
0x50, 0x20, 0x5a, 0x52, 0x32, 0x34, 0x34, 0x30, 0x77, 0x0a, 0x20, 0x20,
/* 18 Byte Data Blocks 4: invalid */
0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 0x4e, 0x34, 0x33, 0x30, 0x34, 0x30,
0x44, 0x58, 0x51, 0x0a, 0x20, 0x20,
/* Extension Block Count */
0x00,
/* Checksum */
0x45,
},
};
#define DPCD_HEADER_SIZE 0xb
/* let the virtual display supports DP1.2 */
static u8 dpcd_fix_data[DPCD_HEADER_SIZE] = {
0x12, 0x014, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
};
static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
int pipe;
if (IS_BROXTON(dev_priv)) {
vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &= ~(BXT_DE_PORT_HP_DDIA |
BXT_DE_PORT_HP_DDIB |
BXT_DE_PORT_HP_DDIC);
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
BXT_DE_PORT_HP_DDIA;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
BXT_DE_PORT_HP_DDIB;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
BXT_DE_PORT_HP_DDIC;
}
return;
}
vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTB_HOTPLUG_CPT |
SDE_PORTC_HOTPLUG_CPT |
SDE_PORTD_HOTPLUG_CPT);
if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv) ||
IS_COFFEELAKE(dev_priv)) {
vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTA_HOTPLUG_SPT |
SDE_PORTE_HOTPLUG_SPT);
vgpu_vreg_t(vgpu, SKL_FUSE_STATUS) |=
SKL_FUSE_DOWNLOAD_STATUS |
SKL_FUSE_PG_DIST_STATUS(SKL_PG0) |
SKL_FUSE_PG_DIST_STATUS(SKL_PG1) |
SKL_FUSE_PG_DIST_STATUS(SKL_PG2);
vgpu_vreg_t(vgpu, LCPLL1_CTL) |=
LCPLL_PLL_ENABLE |
LCPLL_PLL_LOCK;
vgpu_vreg_t(vgpu, LCPLL2_CTL) |= LCPLL_PLL_ENABLE;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIB_DETECTED;
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
TRANS_DDI_PORT_MASK);
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
(PORT_B << TRANS_DDI_PORT_SHIFT) |
TRANS_DDI_FUNC_ENABLE);
if (IS_BROADWELL(dev_priv)) {
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) &=
~PORT_CLK_SEL_MASK;
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) |=
PORT_CLK_SEL_LCPLL_810;
}
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) |= DDI_BUF_CTL_ENABLE;
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) &= ~DDI_BUF_IS_IDLE;
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTB_HOTPLUG_CPT;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTC_HOTPLUG_CPT;
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
TRANS_DDI_PORT_MASK);
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
(PORT_C << TRANS_DDI_PORT_SHIFT) |
TRANS_DDI_FUNC_ENABLE);
if (IS_BROADWELL(dev_priv)) {
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) &=
~PORT_CLK_SEL_MASK;
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) |=
PORT_CLK_SEL_LCPLL_810;
}
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) |= DDI_BUF_CTL_ENABLE;
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) &= ~DDI_BUF_IS_IDLE;
vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIC_DETECTED;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_D)) {
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
TRANS_DDI_PORT_MASK);
vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
(PORT_D << TRANS_DDI_PORT_SHIFT) |
TRANS_DDI_FUNC_ENABLE);
if (IS_BROADWELL(dev_priv)) {
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) &=
~PORT_CLK_SEL_MASK;
vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) |=
PORT_CLK_SEL_LCPLL_810;
}
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) |= DDI_BUF_CTL_ENABLE;
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) &= ~DDI_BUF_IS_IDLE;
vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDID_DETECTED;
}
if ((IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv) ||
IS_COFFEELAKE(dev_priv)) &&
intel_vgpu_has_monitor_on_port(vgpu, PORT_E)) {
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTE_HOTPLUG_SPT;
}
if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
if (IS_BROADWELL(dev_priv))
vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
GEN8_PORT_DP_A_HOTPLUG;
else
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTA_HOTPLUG_SPT;
vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_A)) |= DDI_INIT_DISPLAY_DETECTED;
}
/* Clear host CRT status, so guest couldn't detect this host CRT. */
if (IS_BROADWELL(dev_priv))
vgpu_vreg_t(vgpu, PCH_ADPA) &= ~ADPA_CRT_HOTPLUG_MONITOR_MASK;
/* Disable Primary/Sprite/Cursor plane */
for_each_pipe(dev_priv, pipe) {
vgpu_vreg_t(vgpu, DSPCNTR(pipe)) &= ~DISPLAY_PLANE_ENABLE;
vgpu_vreg_t(vgpu, SPRCTL(pipe)) &= ~SPRITE_ENABLE;
vgpu_vreg_t(vgpu, CURCNTR(pipe)) &= ~MCURSOR_MODE;
vgpu_vreg_t(vgpu, CURCNTR(pipe)) |= MCURSOR_MODE_DISABLE;
}
vgpu_vreg_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
}
static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
{
struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
kfree(port->edid);
port->edid = NULL;
kfree(port->dpcd);
port->dpcd = NULL;
}
static int setup_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num,
int type, unsigned int resolution)
{
struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
if (WARN_ON(resolution >= GVT_EDID_NUM))
return -EINVAL;
port->edid = kzalloc(sizeof(*(port->edid)), GFP_KERNEL);
if (!port->edid)
return -ENOMEM;
port->dpcd = kzalloc(sizeof(*(port->dpcd)), GFP_KERNEL);
if (!port->dpcd) {
kfree(port->edid);
return -ENOMEM;
}
memcpy(port->edid->edid_block, virtual_dp_monitor_edid[resolution],
EDID_SIZE);
port->edid->data_valid = true;
memcpy(port->dpcd->data, dpcd_fix_data, DPCD_HEADER_SIZE);
port->dpcd->data_valid = true;
port->dpcd->data[DPCD_SINK_COUNT] = 0x1;
port->type = type;
port->id = resolution;
emulate_monitor_status_change(vgpu);
return 0;
}
/**
* intel_gvt_check_vblank_emulation - check if vblank emulation timer should
* be turned on/off when a virtual pipe is enabled/disabled.
* @gvt: a GVT device
*
* This function is used to turn on/off vblank timer according to currently
* enabled/disabled virtual pipes.
*
*/
void intel_gvt_check_vblank_emulation(struct intel_gvt *gvt)
{
struct intel_gvt_irq *irq = &gvt->irq;
struct intel_vgpu *vgpu;
int pipe, id;
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
int found = false;
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
mutex_lock(&gvt->lock);
for_each_active_vgpu(gvt, vgpu, id) {
for (pipe = 0; pipe < I915_MAX_PIPES; pipe++) {
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
if (pipe_is_enabled(vgpu, pipe)) {
found = true;
break;
}
}
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
if (found)
break;
}
/* all the pipes are disabled */
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
if (!found)
hrtimer_cancel(&irq->vblank_timer.timer);
else
hrtimer_start(&irq->vblank_timer.timer,
ktime_add_ns(ktime_get(), irq->vblank_timer.period),
HRTIMER_MODE_ABS);
mutex_unlock(&gvt->lock);
}
static void emulate_vblank_on_pipe(struct intel_vgpu *vgpu, int pipe)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
struct intel_vgpu_irq *irq = &vgpu->irq;
int vblank_event[] = {
[PIPE_A] = PIPE_A_VBLANK,
[PIPE_B] = PIPE_B_VBLANK,
[PIPE_C] = PIPE_C_VBLANK,
};
int event;
if (pipe < PIPE_A || pipe > PIPE_C)
return;
for_each_set_bit(event, irq->flip_done_event[pipe],
INTEL_GVT_EVENT_MAX) {
clear_bit(event, irq->flip_done_event[pipe]);
if (!pipe_is_enabled(vgpu, pipe))
continue;
intel_vgpu_trigger_virtual_event(vgpu, event);
}
if (pipe_is_enabled(vgpu, pipe)) {
vgpu_vreg_t(vgpu, PIPE_FRMCOUNT_G4X(pipe))++;
intel_vgpu_trigger_virtual_event(vgpu, vblank_event[pipe]);
}
}
static void emulate_vblank(struct intel_vgpu *vgpu)
{
int pipe;
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
mutex_lock(&vgpu->vgpu_lock);
for_each_pipe(vgpu->gvt->dev_priv, pipe)
emulate_vblank_on_pipe(vgpu, pipe);
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
mutex_unlock(&vgpu->vgpu_lock);
}
/**
* intel_gvt_emulate_vblank - trigger vblank events for vGPUs on GVT device
* @gvt: a GVT device
*
* This function is used to trigger vblank interrupts for vGPUs on GVT device
*
*/
void intel_gvt_emulate_vblank(struct intel_gvt *gvt)
{
struct intel_vgpu *vgpu;
int id;
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
mutex_lock(&gvt->lock);
for_each_active_vgpu(gvt, vgpu, id)
emulate_vblank(vgpu);
drm/i915/gvt: Use vgpu_lock to protect per vgpu access The patch set splits out 2 small locks from the original big gvt lock: - vgpu_lock protects per-vGPU data and logic, especially the vGPU trap emulation path. - sched_lock protects gvt scheudler structure, context schedule logic and vGPU's schedule data. Use vgpu_lock to replace the gvt big lock. By doing this, the mmio read/write trap path, vgpu virtual event emulation and other vgpu related process, would be protected under per vgpu_lock. v9: - Change commit author since the patches are improved a lot compared with original version. Original author: Pei Zhang <pei.zhang@intel.com> - Rebase to latest gvt-staging. v8: - Correct coding and comment style. - Rebase to latest gvt-staging. v7: - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock. - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock. v6: - Rebase to latest gvt-staging. v5: - Rebase to latest gvt-staging. - intel_vgpu_page_track_handler should use vgpu_lock. v4: - Rebase to latest gvt-staging. - Protect vgpu->active access with vgpu_lock. - Do not wait gpu idle in vgpu_lock. v3: update to latest code base v2: add gvt->lock in function gvt_check_vblank_emulation Performance comparison on Kabylake platform. - Configuration: Host: Ubuntu 16.04. Guest 1 & 2: Ubuntu 16.04. glmark2 score comparison: - Configuration: Host: glxgears. Guests: glmark2. +--------------------------------+-----------------+ | Setup | glmark2 score | +--------------------------------+-----------------+ | unified lock, iommu=on | 58~62 (avg. 60) | +--------------------------------+-----------------+ | unified lock, iommu=igfx_off | 57~61 (avg. 59) | +--------------------------------+-----------------+ | per-logic lock, iommu=on | 60~68 (avg. 64) | +--------------------------------+-----------------+ | per-logic lock, iommu=igfx_off | 61~67 (avg. 64) | +--------------------------------+-----------------+ lock_stat comparison: - Configuration: Stop lock stat immediately after boot up. Boot 2 VM Guests. Run glmark2 in guests. Start perf lock_stat for 20 seconds and stop again. - Legend: c - contentions; w - waittime-avg +------------+-----------------+-----------+---------------+------------+ | | gvt_lock |sched_lock | vgpu_lock | gtt_lock | + lock type; +-----------------+-----------+---------------+------------+ | iommu set | c | w | c | w | c | w | c | w | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 20697 | 839 |N/A | N/A | N/A | N/A | N/A | N/A | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | unified; | 21838 | 658.15 |N/A | N/A | N/A | N/A | N/A | N/A | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1553 | 1599.96 |9458|429.97| 5846 | 274.33 | 0 | 0.00 | | on | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ | per-logic; | 1911 | 1678.32 |8335|445.16| 5451 | 244.80 | 0 | 0.00 | | igfx_off | | | | | | | | | +------------+-------+---------+----+------+------+--------+-----+------+ Signed-off-by: Pei Zhang <pei.zhang@intel.com> Signed-off-by: Colin Xu <colin.xu@intel.com> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
2018-05-19 11:28:54 +07:00
mutex_unlock(&gvt->lock);
}
/**
* intel_vgpu_emulate_hotplug - trigger hotplug event for vGPU
* @vgpu: a vGPU
* @connected: link state
*
* This function is used to trigger hotplug interrupt for vGPU
*
*/
void intel_vgpu_emulate_hotplug(struct intel_vgpu *vgpu, bool connected)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
/* TODO: add more platforms support */
if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
if (connected) {
vgpu_vreg_t(vgpu, SFUSE_STRAP) |=
SFUSE_STRAP_DDID_DETECTED;
vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
} else {
vgpu_vreg_t(vgpu, SFUSE_STRAP) &=
~SFUSE_STRAP_DDID_DETECTED;
vgpu_vreg_t(vgpu, SDEISR) &= ~SDE_PORTD_HOTPLUG_CPT;
}
vgpu_vreg_t(vgpu, SDEIIR) |= SDE_PORTD_HOTPLUG_CPT;
vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
PORTD_HOTPLUG_STATUS_MASK;
intel_vgpu_trigger_virtual_event(vgpu, DP_D_HOTPLUG);
}
}
/**
* intel_vgpu_clean_display - clean vGPU virtual display emulation
* @vgpu: a vGPU
*
* This function is used to clean vGPU virtual display emulation stuffs
*
*/
void intel_vgpu_clean_display(struct intel_vgpu *vgpu)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv) ||
IS_COFFEELAKE(dev_priv))
clean_virtual_dp_monitor(vgpu, PORT_D);
else
clean_virtual_dp_monitor(vgpu, PORT_B);
}
/**
* intel_vgpu_init_display- initialize vGPU virtual display emulation
* @vgpu: a vGPU
* @resolution: resolution index for intel_vgpu_edid
*
* This function is used to initialize vGPU virtual display emulation stuffs
*
* Returns:
* Zero on success, negative error code if failed.
*
*/
int intel_vgpu_init_display(struct intel_vgpu *vgpu, u64 resolution)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
intel_vgpu_init_i2c_edid(vgpu);
if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv) ||
IS_COFFEELAKE(dev_priv))
return setup_virtual_dp_monitor(vgpu, PORT_D, GVT_DP_D,
resolution);
else
return setup_virtual_dp_monitor(vgpu, PORT_B, GVT_DP_B,
resolution);
}
/**
* intel_vgpu_reset_display- reset vGPU virtual display emulation
* @vgpu: a vGPU
*
* This function is used to reset vGPU virtual display emulation stuffs
*
*/
void intel_vgpu_reset_display(struct intel_vgpu *vgpu)
{
emulate_monitor_status_change(vgpu);
}