linux_dsm_epyc7002/drivers/gpu/drm/i915/i915_drv.h

2724 lines
82 KiB
C
Raw Normal View History

/* i915_drv.h -- Private header for the I915 driver -*- linux-c -*-
*/
/*
*
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
* 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, sub license, 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 NON-INFRINGEMENT.
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
*
*/
#ifndef _I915_DRV_H_
#define _I915_DRV_H_
#include <uapi/drm/i915_drm.h>
#include "i915_reg.h"
#include "intel_bios.h"
#include "intel_ringbuffer.h"
#include <linux/io-mapping.h>
#include <linux/i2c.h>
#include <linux/i2c-algo-bit.h>
#include <drm/intel-gtt.h>
#include <linux/backlight.h>
#include <linux/intel-iommu.h>
#include <linux/kref.h>
drm/i915: irq-drive the dp aux communication At least on the platforms that have a dp aux irq and also have it enabled - vlvhsw should have one, too. But I don't have a machine to test this on. Judging from docs there's no dp aux interrupt for gm45. Also, I only have an ivb cpu edp machine, so the dp aux A code for snb/ilk is untested. For dpcd probing when nothing is connected it slashes about 5ms of cpu time (cpu time is now negligible), which agrees with 3 * 5 400 usec timeouts. A previous version of this patch increases the time required to go through the dp_detect cycle (which includes reading the edid) from around 33 ms to around 40 ms. Experiments indicated that this is purely due to the irq latency - the hw doesn't allow us to queue up dp aux transactions and hence irq latency directly affects throughput. gmbus is much better, there we have a 8 byte buffer, and we get the irq once another 4 bytes can be queued up. But by using the pm_qos interface to request the lowest possible cpu wake-up latency this slowdown completely disappeared. Since all our output detection logic is single-threaded with the mode_config mutex right now anyway, I've decide not ot play fancy and to just reuse the gmbus wait queue. But this would definitely prep the way to run dp detection on different ports in parallel v2: Add a timeout for dp aux transfers when using interrupts - the hw _does_ prevent this with the hw-based 400 usec timeout, but if the irq somehow doesn't arrive we're screwed. Lesson learned while developing this ;-) v3: While at it also convert the busy-loop to wait_for_atomic, so that we don't run the risk of an infinite loop any more. v4: Ensure we have the smallest possible irq latency by using the pm_qos interface. v5: Add a comment to the code to explain why we frob pm_qos. Suggested by Chris Wilson. v6: Disable dp irq for vlv, that's easier than trying to get at docs and hw. v7: Squash in a fix for Haswell that Paulo Zanoni tracked down - the dp aux registers aren't at a fixed offset any more, but can be on the PCH while the DP port is on the cpu die. Reviewed-by: Imre Deak <imre.deak@intel.com> (v6) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-01 19:53:48 +07:00
#include <linux/pm_qos.h>
/* General customization:
*/
#define DRIVER_AUTHOR "Tungsten Graphics, Inc."
#define DRIVER_NAME "i915"
#define DRIVER_DESC "Intel Graphics"
#define DRIVER_DATE "20080730"
enum pipe {
INVALID_PIPE = -1,
PIPE_A = 0,
PIPE_B,
PIPE_C,
_PIPE_EDP,
I915_MAX_PIPES = _PIPE_EDP
};
#define pipe_name(p) ((p) + 'A')
drm/i915: add TRANSCODER_EDP Before Haswell we used to have the CPU pipes and the PCH transcoders. We had the same amount of pipes and transcoders, and there was a 1:1 mapping between them. After Haswell what we used to call CPU pipe was split into CPU pipe and CPU transcoder. So now we have 3 CPU pipes (A, B and C), 4 CPU transcoders (A, B, C and EDP) and 1 PCH transcoder (only used for VGA). For all the outputs except for EDP we have an 1:1 mapping on the CPU pipes and CPU transcoders, so if you're using CPU pipe A you have to use CPU transcoder A. When have an eDP output you have to use transcoder EDP and you can attach this CPU transcoder to any of the 3 CPU pipes. When using VGA you need to select a pair of matching CPU pipes/transcoders (A/A, B/B, C/C) and you also need to enable/use the PCH transcoder. For now we're just creating the cpu_transcoder definitions and setting cpu_transcoder to TRANSCODER_EDP on DDI eDP code, but none of the registers was ported to use transcoder instead of pipe. The goal is to keep the code backwards-compatible since on all cases except when using eDP we must have pipe == cpu_transcoder. V2: Comment the haswell_crtc_off chunk, suggested by Damien Lespiau and Daniel Vetter. We currently need the haswell_crtc_off chunk because TRANSCODER_EDP can be used by any CRTC, so when you stop using it you have to stop saying you're using it, otherwise you may have at some point 2 CRTCs claiming they're using TRANSCODER_EDP (a disabled CRTC and an enabled one), then the HW state readout code will get completely confused. In other words: Imagine the following case: xrandr --output eDP1 --auto --crtc 0 xrandr --output eDP1 --off xrandr --output eDP1 --auto --crtc 2 After the last command you could get a "pipe A assertion failure (expected off, current on)" because CRTC 0 still claims it's using TRANSCODER_EDP, so the HW state readout function will read it (through PIPECONF) and expect it to be off, when it's actually on because it's being used by CRTC 2. So when we make "intel_crtc->cpu_transcoder = intel_crtc->pipe" we make sure we're pointing to our own original CRTC which is certainly not used by any other CRTC. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-10-25 00:59:34 +07:00
enum transcoder {
TRANSCODER_A = 0,
TRANSCODER_B,
TRANSCODER_C,
TRANSCODER_EDP,
I915_MAX_TRANSCODERS
drm/i915: add TRANSCODER_EDP Before Haswell we used to have the CPU pipes and the PCH transcoders. We had the same amount of pipes and transcoders, and there was a 1:1 mapping between them. After Haswell what we used to call CPU pipe was split into CPU pipe and CPU transcoder. So now we have 3 CPU pipes (A, B and C), 4 CPU transcoders (A, B, C and EDP) and 1 PCH transcoder (only used for VGA). For all the outputs except for EDP we have an 1:1 mapping on the CPU pipes and CPU transcoders, so if you're using CPU pipe A you have to use CPU transcoder A. When have an eDP output you have to use transcoder EDP and you can attach this CPU transcoder to any of the 3 CPU pipes. When using VGA you need to select a pair of matching CPU pipes/transcoders (A/A, B/B, C/C) and you also need to enable/use the PCH transcoder. For now we're just creating the cpu_transcoder definitions and setting cpu_transcoder to TRANSCODER_EDP on DDI eDP code, but none of the registers was ported to use transcoder instead of pipe. The goal is to keep the code backwards-compatible since on all cases except when using eDP we must have pipe == cpu_transcoder. V2: Comment the haswell_crtc_off chunk, suggested by Damien Lespiau and Daniel Vetter. We currently need the haswell_crtc_off chunk because TRANSCODER_EDP can be used by any CRTC, so when you stop using it you have to stop saying you're using it, otherwise you may have at some point 2 CRTCs claiming they're using TRANSCODER_EDP (a disabled CRTC and an enabled one), then the HW state readout code will get completely confused. In other words: Imagine the following case: xrandr --output eDP1 --auto --crtc 0 xrandr --output eDP1 --off xrandr --output eDP1 --auto --crtc 2 After the last command you could get a "pipe A assertion failure (expected off, current on)" because CRTC 0 still claims it's using TRANSCODER_EDP, so the HW state readout function will read it (through PIPECONF) and expect it to be off, when it's actually on because it's being used by CRTC 2. So when we make "intel_crtc->cpu_transcoder = intel_crtc->pipe" we make sure we're pointing to our own original CRTC which is certainly not used by any other CRTC. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-10-25 00:59:34 +07:00
};
#define transcoder_name(t) ((t) + 'A')
enum plane {
PLANE_A = 0,
PLANE_B,
PLANE_C,
};
#define plane_name(p) ((p) + 'A')
#define sprite_name(p, s) ((p) * INTEL_INFO(dev)->num_sprites + (s) + 'A')
enum port {
PORT_A = 0,
PORT_B,
PORT_C,
PORT_D,
PORT_E,
I915_MAX_PORTS
};
#define port_name(p) ((p) + 'A')
#define I915_NUM_PHYS_VLV 1
enum dpio_channel {
DPIO_CH0,
DPIO_CH1
};
enum dpio_phy {
DPIO_PHY0,
DPIO_PHY1
};
enum intel_display_power_domain {
POWER_DOMAIN_PIPE_A,
POWER_DOMAIN_PIPE_B,
POWER_DOMAIN_PIPE_C,
POWER_DOMAIN_PIPE_A_PANEL_FITTER,
POWER_DOMAIN_PIPE_B_PANEL_FITTER,
POWER_DOMAIN_PIPE_C_PANEL_FITTER,
POWER_DOMAIN_TRANSCODER_A,
POWER_DOMAIN_TRANSCODER_B,
POWER_DOMAIN_TRANSCODER_C,
POWER_DOMAIN_TRANSCODER_EDP,
POWER_DOMAIN_VGA,
POWER_DOMAIN_AUDIO,
drm/i915: use power get/put instead of set for power on after init Currently we make sure that all power domains are enabled during driver init and turn off unneded ones only after the first modeset. Similarly during suspend we enable all power domains, which will remain on through the following resume until the first modeset. This logic is supported by intel_set_power_well() in the power domain framework. It would be nice to simplify the API, so that we only have get/put functions and make it more explicit on the higher level how this "power well on during init" logic works. This will make it also easier if in the future we want to shorten the time the power wells are on. For this add a new device private flag tracking whether we have the power wells on because of init/suspend and use only intel_display_power_get()/put(). As nothing else uses intel_set_power_well() we can remove it. This also fixes commit 6efdf354ddb186c6604d1692075421e8d2c740e9 Author: Imre Deak <imre.deak@intel.com> Date: Wed Oct 16 17:25:52 2013 +0300 drm/i915: enable only the needed power domains during modeset where removing intel_set_power_well() resulted in not releasing the reference on the power well that was taken during init and thus leaving the power well on all the time. Regression reported by Paulo. v2: - move the init_power_on flag to the power_domains struct (Daniel) v3: - add note about this being a regression fix too (Paulo) Signed-off-by: Imre Deak <imre.deak@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-10-25 21:36:48 +07:00
POWER_DOMAIN_INIT,
POWER_DOMAIN_NUM,
};
#define POWER_DOMAIN_MASK (BIT(POWER_DOMAIN_NUM) - 1)
#define POWER_DOMAIN_PIPE(pipe) ((pipe) + POWER_DOMAIN_PIPE_A)
#define POWER_DOMAIN_PIPE_PANEL_FITTER(pipe) \
((pipe) + POWER_DOMAIN_PIPE_A_PANEL_FITTER)
#define POWER_DOMAIN_TRANSCODER(tran) \
((tran) == TRANSCODER_EDP ? POWER_DOMAIN_TRANSCODER_EDP : \
(tran) + POWER_DOMAIN_TRANSCODER_A)
#define HSW_ALWAYS_ON_POWER_DOMAINS ( \
BIT(POWER_DOMAIN_PIPE_A) | \
BIT(POWER_DOMAIN_TRANSCODER_EDP))
#define BDW_ALWAYS_ON_POWER_DOMAINS ( \
BIT(POWER_DOMAIN_PIPE_A) | \
BIT(POWER_DOMAIN_TRANSCODER_EDP) | \
BIT(POWER_DOMAIN_PIPE_A_PANEL_FITTER))
enum hpd_pin {
HPD_NONE = 0,
HPD_PORT_A = HPD_NONE, /* PORT_A is internal */
HPD_TV = HPD_NONE, /* TV is known to be unreliable */
HPD_CRT,
HPD_SDVO_B,
HPD_SDVO_C,
HPD_PORT_B,
HPD_PORT_C,
HPD_PORT_D,
HPD_NUM_PINS
};
#define I915_GEM_GPU_DOMAINS \
(I915_GEM_DOMAIN_RENDER | \
I915_GEM_DOMAIN_SAMPLER | \
I915_GEM_DOMAIN_COMMAND | \
I915_GEM_DOMAIN_INSTRUCTION | \
I915_GEM_DOMAIN_VERTEX)
#define for_each_pipe(p) for ((p) = 0; (p) < INTEL_INFO(dev)->num_pipes; (p)++)
#define for_each_encoder_on_crtc(dev, __crtc, intel_encoder) \
list_for_each_entry((intel_encoder), &(dev)->mode_config.encoder_list, base.head) \
if ((intel_encoder)->base.crtc == (__crtc))
#define for_each_connector_on_encoder(dev, __encoder, intel_connector) \
list_for_each_entry((intel_connector), &(dev)->mode_config.connector_list, base.head) \
if ((intel_connector)->base.encoder == (__encoder))
struct drm_i915_private;
enum intel_dpll_id {
DPLL_ID_PRIVATE = -1, /* non-shared dpll in use */
/* real shared dpll ids must be >= 0 */
DPLL_ID_PCH_PLL_A,
DPLL_ID_PCH_PLL_B,
};
#define I915_NUM_PLLS 2
struct intel_dpll_hw_state {
uint32_t dpll;
uint32_t dpll_md;
uint32_t fp0;
uint32_t fp1;
};
struct intel_shared_dpll {
int refcount; /* count of number of CRTCs sharing this PLL */
int active; /* count of number of active CRTCs (i.e. DPMS on) */
bool on; /* is the PLL actually active? Disabled during modeset */
const char *name;
/* should match the index in the dev_priv->shared_dplls array */
enum intel_dpll_id id;
struct intel_dpll_hw_state hw_state;
void (*mode_set)(struct drm_i915_private *dev_priv,
struct intel_shared_dpll *pll);
void (*enable)(struct drm_i915_private *dev_priv,
struct intel_shared_dpll *pll);
void (*disable)(struct drm_i915_private *dev_priv,
struct intel_shared_dpll *pll);
bool (*get_hw_state)(struct drm_i915_private *dev_priv,
struct intel_shared_dpll *pll,
struct intel_dpll_hw_state *hw_state);
};
/* Used by dp and fdi links */
struct intel_link_m_n {
uint32_t tu;
uint32_t gmch_m;
uint32_t gmch_n;
uint32_t link_m;
uint32_t link_n;
};
void intel_link_compute_m_n(int bpp, int nlanes,
int pixel_clock, int link_clock,
struct intel_link_m_n *m_n);
struct intel_ddi_plls {
int spll_refcount;
int wrpll1_refcount;
int wrpll2_refcount;
};
/* Interface history:
*
* 1.1: Original.
* 1.2: Add Power Management
* 1.3: Add vblank support
* 1.4: Fix cmdbuffer path, add heap destroy
* 1.5: Add vblank pipe configuration
* 1.6: - New ioctl for scheduling buffer swaps on vertical blank
* - Support vertical blank on secondary display pipe
*/
#define DRIVER_MAJOR 1
#define DRIVER_MINOR 6
#define DRIVER_PATCHLEVEL 0
#define WATCH_LISTS 0
#define WATCH_GTT 0
#define I915_GEM_PHYS_CURSOR_0 1
#define I915_GEM_PHYS_CURSOR_1 2
#define I915_GEM_PHYS_OVERLAY_REGS 3
#define I915_MAX_PHYS_OBJECT (I915_GEM_PHYS_OVERLAY_REGS)
struct drm_i915_gem_phys_object {
int id;
struct page **page_list;
drm_dma_handle_t *handle;
struct drm_i915_gem_object *cur_obj;
};
struct opregion_header;
struct opregion_acpi;
struct opregion_swsci;
struct opregion_asle;
struct intel_opregion {
struct opregion_header __iomem *header;
struct opregion_acpi __iomem *acpi;
struct opregion_swsci __iomem *swsci;
u32 swsci_gbda_sub_functions;
u32 swsci_sbcb_sub_functions;
struct opregion_asle __iomem *asle;
void __iomem *vbt;
u32 __iomem *lid_state;
struct work_struct asle_work;
};
#define OPREGION_SIZE (8*1024)
struct intel_overlay;
struct intel_overlay_error_state;
struct drm_i915_master_private {
drm_local_map_t *sarea;
struct _drm_i915_sarea *sarea_priv;
};
#define I915_FENCE_REG_NONE -1
#define I915_MAX_NUM_FENCES 32
/* 32 fences + sign bit for FENCE_REG_NONE */
#define I915_MAX_NUM_FENCE_BITS 6
struct drm_i915_fence_reg {
struct list_head lru_list;
struct drm_i915_gem_object *obj;
int pin_count;
};
struct sdvo_device_mapping {
u8 initialized;
u8 dvo_port;
u8 slave_addr;
u8 dvo_wiring;
u8 i2c_pin;
u8 ddc_pin;
};
struct intel_display_error_state;
struct drm_i915_error_state {
struct kref ref;
struct timeval time;
/* Generic register state */
u32 eir;
u32 pgtbl_er;
u32 ier;
u32 ccid;
u32 derrmr;
u32 forcewake;
u32 error; /* gen6+ */
u32 err_int; /* gen7 */
u32 done_reg;
u32 gac_eco;
u32 gam_ecochk;
u32 gab_ctl;
u32 gfx_mode;
u32 extra_instdone[I915_NUM_INSTDONE_REG];
u32 pipestat[I915_MAX_PIPES];
u64 fence[I915_MAX_NUM_FENCES];
struct intel_overlay_error_state *overlay;
struct intel_display_error_state *display;
struct drm_i915_error_ring {
bool valid;
/* Software tracked state */
bool waiting;
int hangcheck_score;
enum intel_ring_hangcheck_action hangcheck_action;
int num_requests;
/* our own tracking of ring head and tail */
u32 cpu_ring_head;
u32 cpu_ring_tail;
u32 semaphore_seqno[I915_NUM_RINGS - 1];
/* Register state */
u32 tail;
u32 head;
u32 ctl;
u32 hws;
u32 ipeir;
u32 ipehr;
u32 instdone;
u32 acthd;
u32 bbstate;
u32 instpm;
u32 instps;
u32 seqno;
u64 bbaddr;
u32 fault_reg;
u32 faddr;
u32 rc_psmi; /* sleep state */
u32 semaphore_mboxes[I915_NUM_RINGS - 1];
struct drm_i915_error_object {
int page_count;
u32 gtt_offset;
u32 *pages[0];
} *ringbuffer, *batchbuffer, *ctx, *hws_page;
struct drm_i915_error_request {
long jiffies;
u32 seqno;
u32 tail;
} *requests;
struct {
u32 gfx_mode;
union {
u64 pdp[4];
u32 pp_dir_base;
};
} vm_info;
} ring[I915_NUM_RINGS];
struct drm_i915_error_buffer {
u32 size;
u32 name;
u32 rseqno, wseqno;
u32 gtt_offset;
u32 read_domains;
u32 write_domain;
s32 fence_reg:I915_MAX_NUM_FENCE_BITS;
s32 pinned:2;
u32 tiling:2;
u32 dirty:1;
u32 purgeable:1;
s32 ring:4;
u32 cache_level:3;
} **active_bo, **pinned_bo;
u32 *active_bo_count, *pinned_bo_count;
};
struct intel_connector;
struct intel_crtc_config;
struct intel_crtc;
drm/i915: move find_pll callback to dev_priv->display Now that the DP madness is cleared out, this is all only per-platform. So move it out from the intel clock limits structure. While at it drop the intel prefix on the static functions, call the vtable entry find_dpll (since it's for the display pll) and rip out the now unnecessary forward declarations. Note that the parameters of ->find_dpll are still unchanged, but they eventually need to be moved over to just take in a pipe configuration. But currently a lot of things are still missing from the pipe configuration (reflock, output-specific dpll limits and preferences, downclocked dotclock). So this will happen in a later step. Note that intel_g4x_limit has a peculiar case where it selects intel_limits_i9xx_sdvo as the limit. This is pretty bogus and also not used since the only output types left are DP and native TV-out which both use special pre-tuned dpll values. v2: Re-add comment for the find_pll callback (requested by Paulo) and elaborate on why the transformation is correct for g4x platforms (to clarify a review question from Paulo). Double up on that by adding a WARN as suggested by Paulo Zanoni on irc. v3: Initialize limits to NULL since gcc is now unhappy. v4: v2/3 will blow up with a NULL dereference in ->find_dpll for dp and TV-out ports, spotted by Paulo on irc. So just give up on this madness for now, and leave this to be fixed in a later patch. v5: Since the ever-so-slight change for g4x might result in some dpll parameter computation failing spuriously where before it didn't for ports with preset dpll settings (DP & TV-out) override this. For paranoia also do it in the ilk+ code. Cc: Paulo Zanoni <przanoni@gmail.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-06-04 03:40:22 +07:00
struct intel_limit;
struct dpll;
struct drm_i915_display_funcs {
bool (*fbc_enabled)(struct drm_device *dev);
void (*enable_fbc)(struct drm_crtc *crtc);
void (*disable_fbc)(struct drm_device *dev);
int (*get_display_clock_speed)(struct drm_device *dev);
int (*get_fifo_size)(struct drm_device *dev, int plane);
drm/i915: move find_pll callback to dev_priv->display Now that the DP madness is cleared out, this is all only per-platform. So move it out from the intel clock limits structure. While at it drop the intel prefix on the static functions, call the vtable entry find_dpll (since it's for the display pll) and rip out the now unnecessary forward declarations. Note that the parameters of ->find_dpll are still unchanged, but they eventually need to be moved over to just take in a pipe configuration. But currently a lot of things are still missing from the pipe configuration (reflock, output-specific dpll limits and preferences, downclocked dotclock). So this will happen in a later step. Note that intel_g4x_limit has a peculiar case where it selects intel_limits_i9xx_sdvo as the limit. This is pretty bogus and also not used since the only output types left are DP and native TV-out which both use special pre-tuned dpll values. v2: Re-add comment for the find_pll callback (requested by Paulo) and elaborate on why the transformation is correct for g4x platforms (to clarify a review question from Paulo). Double up on that by adding a WARN as suggested by Paulo Zanoni on irc. v3: Initialize limits to NULL since gcc is now unhappy. v4: v2/3 will blow up with a NULL dereference in ->find_dpll for dp and TV-out ports, spotted by Paulo on irc. So just give up on this madness for now, and leave this to be fixed in a later patch. v5: Since the ever-so-slight change for g4x might result in some dpll parameter computation failing spuriously where before it didn't for ports with preset dpll settings (DP & TV-out) override this. For paranoia also do it in the ilk+ code. Cc: Paulo Zanoni <przanoni@gmail.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-06-04 03:40:22 +07:00
/**
* find_dpll() - Find the best values for the PLL
* @limit: limits for the PLL
* @crtc: current CRTC
* @target: target frequency in kHz
* @refclk: reference clock frequency in kHz
* @match_clock: if provided, @best_clock P divider must
* match the P divider from @match_clock
* used for LVDS downclocking
* @best_clock: best PLL values found
*
* Returns true on success, false on failure.
*/
bool (*find_dpll)(const struct intel_limit *limit,
struct drm_crtc *crtc,
int target, int refclk,
struct dpll *match_clock,
struct dpll *best_clock);
void (*update_wm)(struct drm_crtc *crtc);
void (*update_sprite_wm)(struct drm_plane *plane,
struct drm_crtc *crtc,
uint32_t sprite_width, int pixel_size,
bool enable, bool scaled);
drm/i915: add ->display.modeset_global_resources callback After all relevant pipes are disabled and after we've updated all the state with the staged state, but before we call the per-crtc ->mode_set functions there's a very natural point to set up any shared/global resources like - shared plls (obviously only the setup, the enabling needs to be separately handling with a separate refcount) - global watermark state like the DSPARB on gmch platforms - workaround bits that depend upon the exact global output configuration - enabling the right set of refclocks - enabling/disabling manual power wells. Now for a lot of these things we can't move them into this function yet, most often because we only compute the required information in the per-crtc ->mode_set callback. Which is too late. But due to a bunch of reasons (check-only atomic modeset, fastboot&hw state checks, ...) we need to separate the computation of that state from the actual hw frobbery anyway. So we can move things into this new callback step- by-step. Others can't be moved here (or implemented at all) because our code lacks the smarts to properly update them. E.g. the DSPARB can only be updated when all pipes are disabled, so if we decide to change it's value, we need to disable _all_ pipes. The infrastructure for that is already in place (with the various pipe masks that driver the modeset logic). But again we need to move a few things out of ->mode_set first before we can even implement the correct decision making. In any case, we need to start somewhere, so let's start with the callback: Some small follow-up patches will make immediate good use of it. Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-10-26 15:58:18 +07:00
void (*modeset_global_resources)(struct drm_device *dev);
/* Returns the active state of the crtc, and if the crtc is active,
* fills out the pipe-config with the hw state. */
bool (*get_pipe_config)(struct intel_crtc *,
struct intel_crtc_config *);
int (*crtc_mode_set)(struct drm_crtc *crtc,
int x, int y,
struct drm_framebuffer *old_fb);
void (*crtc_enable)(struct drm_crtc *crtc);
void (*crtc_disable)(struct drm_crtc *crtc);
void (*off)(struct drm_crtc *crtc);
drm/i915: pass ELD to HDMI/DP audio driver Add ELD support for Intel Eaglelake, IbexPeak/Ironlake, SandyBridge/CougarPoint and IvyBridge/PantherPoint chips. ELD (EDID-Like Data) describes to the HDMI/DP audio driver the audio capabilities of the plugged monitor. It's built and passed to audio driver in 2 steps: (1) at get_modes time, parse EDID and save ELD to drm_connector.eld[] (2) at mode_set time, write drm_connector.eld[] to the Transcoder's hw ELD buffer and set the ELD_valid bit to inform HDMI/DP audio driver This patch is tested OK on G45/HDMI, IbexPeak/HDMI and IvyBridge/HDMI+DP. Test scheme: plug in the HDMI/DP monitor, and run cat /proc/asound/card0/eld* to check if the monitor name, HDMI/DP type, etc. show up correctly. Minor imperfection: the GEN5_AUD_CNTL_ST/DIP_Port_Select field always reads 0 (reserved). Without knowing the port number, I worked it around by setting the ELD_valid bit for ALL the three ports. It's tested to not be a problem, because the audio driver will find invalid ELD data and hence rightfully abort, even when it sees the ELD_valid indicator. Thanks to Zhenyu and Pierre-Louis for a lot of valuable help and testing. CC: Zhao Yakui <yakui.zhao@intel.com> CC: Wang Zhenyu <zhenyu.z.wang@intel.com> CC: Jeremy Bush <contractfrombelow@gmail.com> CC: Christopher White <c.white@pulseforce.com> CC: Pierre-Louis Bossart <pierre-louis.bossart@intel.com> CC: Paul Menzel <paulepanter@users.sourceforge.net> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Keith Packard <keithp@keithp.com>
2011-09-05 13:25:34 +07:00
void (*write_eld)(struct drm_connector *connector,
struct drm_crtc *crtc,
struct drm_display_mode *mode);
void (*fdi_link_train)(struct drm_crtc *crtc);
void (*init_clock_gating)(struct drm_device *dev);
int (*queue_flip)(struct drm_device *dev, struct drm_crtc *crtc,
struct drm_framebuffer *fb,
struct drm_i915_gem_object *obj,
uint32_t flags);
int (*update_plane)(struct drm_crtc *crtc, struct drm_framebuffer *fb,
int x, int y);
drm/i915: Fixup hpd irq register setup ordering For GMCH platforms we set up the hpd irq registers in the irq postinstall hook. But since we only enable the irq sources we actually need in PORT_HOTPLUG_EN/STATUS, taking dev_priv->hotplug_supported_mask into account, no hpd interrupt sources is enabled since commit 52d7ecedac3f96fb562cb482c139015372728638 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Sat Dec 1 21:03:22 2012 +0100 drm/i915: reorder setup sequence to have irqs for output setup Wrongly set-up interrupts also lead to broken hw-based load-detection on at least GM45, resulting in ghost VGA/TV-out outputs. To fix this, delay the hotplug register setup until after all outputs are set up, by moving it into a new dev_priv->display.hpd_irq_callback. We might also move the PCH_SPLIT platforms to such a setup eventually. Another funny part is that we need to delay the fbdev initial config probing until after the hpd regs are setup, for otherwise it'll detect ghost outputs. But we can only enable the hpd interrupt handling itself (and the output polling) _after_ that initial scan, due to massive locking brain-damage in the fbdev setup code. Add a big comment to explain this cute little dragon lair. v2: Encapsulate all the fbdev handling by wrapping the move call into intel_fbdev_initial_config in intel_fb.c. Requested by Chris Wilson. v3: Applied bikeshed from Jesse Barnes. v4: Imre Deak noticed that we also need to call intel_hpd_init after the drm_irqinstall calls in the gpu reset and resume paths - otherwise hotplug will be broken. Also improve the comment a bit about why hpd_init needs to be called before we set up the initial fbdev config. Bugzilla: Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54943 Reported-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v3) Reviewed-by: Imre Deak <imre.deak@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-11 20:05:07 +07:00
void (*hpd_irq_setup)(struct drm_device *dev);
/* clock updates for mode set */
/* cursor updates */
/* render clock increase/decrease */
/* display clock increase/decrease */
/* pll clock increase/decrease */
int (*setup_backlight)(struct intel_connector *connector);
uint32_t (*get_backlight)(struct intel_connector *connector);
void (*set_backlight)(struct intel_connector *connector,
uint32_t level);
void (*disable_backlight)(struct intel_connector *connector);
void (*enable_backlight)(struct intel_connector *connector);
};
struct intel_uncore_funcs {
void (*force_wake_get)(struct drm_i915_private *dev_priv,
int fw_engine);
void (*force_wake_put)(struct drm_i915_private *dev_priv,
int fw_engine);
uint8_t (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
uint32_t (*mmio_readl)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
uint64_t (*mmio_readq)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
void (*mmio_writeb)(struct drm_i915_private *dev_priv, off_t offset,
uint8_t val, bool trace);
void (*mmio_writew)(struct drm_i915_private *dev_priv, off_t offset,
uint16_t val, bool trace);
void (*mmio_writel)(struct drm_i915_private *dev_priv, off_t offset,
uint32_t val, bool trace);
void (*mmio_writeq)(struct drm_i915_private *dev_priv, off_t offset,
uint64_t val, bool trace);
};
struct intel_uncore {
spinlock_t lock; /** lock is also taken in irq contexts. */
struct intel_uncore_funcs funcs;
unsigned fifo_count;
unsigned forcewake_count;
unsigned fw_rendercount;
unsigned fw_mediacount;
struct delayed_work force_wake_work;
};
#define DEV_INFO_FOR_EACH_FLAG(func, sep) \
func(is_mobile) sep \
func(is_i85x) sep \
func(is_i915g) sep \
func(is_i945gm) sep \
func(is_g33) sep \
func(need_gfx_hws) sep \
func(is_g4x) sep \
func(is_pineview) sep \
func(is_broadwater) sep \
func(is_crestline) sep \
func(is_ivybridge) sep \
func(is_valleyview) sep \
func(is_haswell) sep \
func(is_preliminary) sep \
func(has_fbc) sep \
func(has_pipe_cxsr) sep \
func(has_hotplug) sep \
func(cursor_needs_physical) sep \
func(has_overlay) sep \
func(overlay_needs_physical) sep \
func(supports_tv) sep \
func(has_llc) sep \
func(has_ddi) sep \
func(has_fpga_dbg)
#define DEFINE_FLAG(name) u8 name:1
#define SEP_SEMICOLON ;
struct intel_device_info {
u32 display_mmio_offset;
u8 num_pipes:3;
u8 num_sprites:2;
u8 gen;
u8 ring_mask; /* Rings supported by the HW */
DEV_INFO_FOR_EACH_FLAG(DEFINE_FLAG, SEP_SEMICOLON);
/* Register offsets for the various display pipes and transcoders */
int pipe_offsets[I915_MAX_TRANSCODERS];
int trans_offsets[I915_MAX_TRANSCODERS];
int dpll_offsets[I915_MAX_PIPES];
int dpll_md_offsets[I915_MAX_PIPES];
int palette_offsets[I915_MAX_PIPES];
};
#undef DEFINE_FLAG
#undef SEP_SEMICOLON
enum i915_cache_level {
I915_CACHE_NONE = 0,
I915_CACHE_LLC, /* also used for snoopable memory on non-LLC */
I915_CACHE_L3_LLC, /* gen7+, L3 sits between the domain specifc
caches, eg sampler/render caches, and the
large Last-Level-Cache. LLC is coherent with
the CPU, but L3 is only visible to the GPU. */
I915_CACHE_WT, /* hsw:gt3e WriteThrough for scanouts */
};
typedef uint32_t gen6_gtt_pte_t;
drm/i915: Create bind/unbind abstraction for VMAs To sum up what goes on here, we abstract the vma binding, similarly to the previous object binding. This helps for distinguishing legacy binding, versus modern binding. To keep the code churn as minimal as possible, I am leaving in insert_entries(). It serves as the per platform pte writing basically. bind_vma and insert_entries do share a lot of similarities, and I did have designs to combine the two, but as mentioned already... too much churn in an already massive patchset. What follows are the 3 commits which existed discretely in the original submissions. Upon rebasing on Broadwell support, it became clear that separation was not good, and only made for more error prone code. Below are the 3 commit messages with all their history. drm/i915: Add bind/unbind object functions to VMA drm/i915: Use the new vm [un]bind functions drm/i915: reduce vm->insert_entries() usage drm/i915: Add bind/unbind object functions to VMA As we plumb the code with more VM information, it has become more obvious that the easiest way to deal with bind and unbind is to simply put the function pointers in the vm, and let those choose the correct way to handle the page table updates. This change allows many places in the code to simply be vm->bind, and not have to worry about distinguishing PPGTT vs GGTT. Notice that this patch has no impact on functionality. I've decided to save the actual change until the next patch because I think it's easier to review that way. I'm happy to squash the two, or let Daniel do it on merge. v2: Make ggtt handle the quirky aliasing ppgtt Add flags to bind object to support above Don't ever call bind/unbind directly for PPGTT until we have real, full PPGTT (use NULLs to assert this) Make sure we rebind the ggtt if there already is a ggtt binding. This happens on set cache levels. Use VMA for bind/unbind (Daniel, Ben) v3: Reorganize ggtt_vma_bind to be more concise and easier to read (Ville). Change logic in unbind to only unbind ggtt when there is a global mapping, and to remove a redundant check if the aliasing ppgtt exists. v4: Make the bind function a bit smarter about the cache levels to avoid unnecessary multiple remaps. "I accept it is a wart, I think unifying the pin_vma / bind_vma could be unified later" (Chris) Removed the git notes, and put version info here. (Daniel) v5: Update the comment to not suck (Chris) v6: Move bind/unbind to the VMA. It makes more sense in the VMA structure (always has, but I was previously lazy). With this change, it will allow us to keep a distinct insert_entries. Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> drm/i915: Use the new vm [un]bind functions Building on the last patch which created the new function pointers in the VM for bind/unbind, here we actually put those new function pointers to use. Split out as a separate patch to aid in review. I'm fine with squashing into the previous patch if people request it. v2: Updated to address the smart ggtt which can do aliasing as needed Make sure we bind to global gtt when mappable and fenceable. I thought we could get away without this initialy, but we cannot. v3: Make the global GTT binding explicitly use the ggtt VM for bind_vma(). While at it, use the new ggtt_vma helper (Chris) At this point the original mailing list thread diverges. ie. v4^: use target_obj instead of obj for gen6 relocate_entry vma->bind_vma() can be called safely during pin. So simply do that instead of the complicated conditionals. Don't restore PPGTT bound objects on resume path Bug fix in resume path for globally bound Bos Properly handle secure dispatch Rebased on vma bind/unbind conversion Signed-off-by: Ben Widawsky <ben@bwidawsk.net> drm/i915: reduce vm->insert_entries() usage FKA: drm/i915: eliminate vm->insert_entries() With bind/unbind function pointers in place, we no longer need insert_entries. We could, and want, to remove clear_range, however it's not totally easy at this point. Since it's used in a couple of place still that don't only deal in objects: setup, ppgtt init, and restore gtt mappings. v2: Don't actually remove insert_entries, just limit its usage. It will be useful when we introduce gen8. It will always be called from the vma bind/unbind. Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v1) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-12-07 05:10:56 +07:00
/**
* A VMA represents a GEM BO that is bound into an address space. Therefore, a
* VMA's presence cannot be guaranteed before binding, or after unbinding the
* object into/from the address space.
*
* To make things as simple as possible (ie. no refcounting), a VMA's lifetime
* will always be <= an objects lifetime. So object refcounting should cover us.
*/
struct i915_vma {
struct drm_mm_node node;
struct drm_i915_gem_object *obj;
struct i915_address_space *vm;
/** This object's place on the active/inactive lists */
struct list_head mm_list;
struct list_head vma_link; /* Link in the object's VMA list */
/** This vma's place in the batchbuffer or on the eviction list */
struct list_head exec_list;
/**
* Used for performing relocations during execbuffer insertion.
*/
struct hlist_node exec_node;
unsigned long exec_handle;
struct drm_i915_gem_exec_object2 *exec_entry;
/**
* How many users have pinned this object in GTT space. The following
* users can each hold at most one reference: pwrite/pread, pin_ioctl
* (via user_pin_count), execbuffer (objects are not allowed multiple
* times for the same batchbuffer), and the framebuffer code. When
* switching/pageflipping, the framebuffer code has at most two buffers
* pinned per crtc.
*
* In the worst case this is 1 + 1 + 1 + 2*2 = 7. That would fit into 3
* bits with absolutely no headroom. So use 4 bits. */
unsigned int pin_count:4;
#define DRM_I915_GEM_OBJECT_MAX_PIN_COUNT 0xf
/** Unmap an object from an address space. This usually consists of
* setting the valid PTE entries to a reserved scratch page. */
void (*unbind_vma)(struct i915_vma *vma);
/* Map an object into an address space with the given cache flags. */
#define GLOBAL_BIND (1<<0)
void (*bind_vma)(struct i915_vma *vma,
enum i915_cache_level cache_level,
u32 flags);
};
struct i915_address_space {
struct drm_mm mm;
struct drm_device *dev;
struct list_head global_link;
unsigned long start; /* Start offset always 0 for dri2 */
size_t total; /* size addr space maps (ex. 2GB for ggtt) */
struct {
dma_addr_t addr;
struct page *page;
} scratch;
/**
* List of objects currently involved in rendering.
*
* Includes buffers having the contents of their GPU caches
* flushed, not necessarily primitives. last_rendering_seqno
* represents when the rendering involved will be completed.
*
* A reference is held on the buffer while on this list.
*/
struct list_head active_list;
/**
* LRU list of objects which are not in the ringbuffer and
* are ready to unbind, but are still in the GTT.
*
* last_rendering_seqno is 0 while an object is in this list.
*
* A reference is not held on the buffer while on this list,
* as merely being GTT-bound shouldn't prevent its being
* freed, and we'll pull it off the list in the free path.
*/
struct list_head inactive_list;
/* FIXME: Need a more generic return type */
gen6_gtt_pte_t (*pte_encode)(dma_addr_t addr,
enum i915_cache_level level,
bool valid); /* Create a valid PTE */
void (*clear_range)(struct i915_address_space *vm,
unsigned int first_entry,
drm/i915: Disable GGTT PTEs on GEN6+ suspend Once the machine gets to a certain point in the suspend process, we expect the GPU to be idle. If it is not, we might corrupt memory. Empirically (with an early version of this patch) we have seen this is not the case. We cannot currently explain why the latent GPU writes occur. In the technical sense, this patch is a workaround in that we have an issue we can't explain, and the patch indirectly solves the issue. However, it's really better than a workaround because we understand why it works, and it really should be a safe thing to do in all cases. The noticeable effect other than the debug messages would be an increase in the suspend time. I have not measure how expensive it actually is. I think it would be good to spend further time to root cause why we're seeing these latent writes, but it shouldn't preclude preventing the fallout. NOTE: It should be safe (and makes some sense IMO) to also keep the VALID bit unset on resume when we clear_range(). I've opted not to do this as properly clearing those bits at some later point would be extra work. v2: Fix bugzilla link Bugzilla: http://bugs.freedesktop.org/show_bug.cgi?id=65496 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=59321 Tested-by: Takashi Iwai <tiwai@suse.de> Tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Tested-By: Todd Previte <tprevite@gmail.com> Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-10-16 23:21:30 +07:00
unsigned int num_entries,
bool use_scratch);
void (*insert_entries)(struct i915_address_space *vm,
struct sg_table *st,
unsigned int first_entry,
enum i915_cache_level cache_level);
void (*cleanup)(struct i915_address_space *vm);
};
/* The Graphics Translation Table is the way in which GEN hardware translates a
* Graphics Virtual Address into a Physical Address. In addition to the normal
* collateral associated with any va->pa translations GEN hardware also has a
* portion of the GTT which can be mapped by the CPU and remain both coherent
* and correct (in cases like swizzling). That region is referred to as GMADR in
* the spec.
*/
struct i915_gtt {
struct i915_address_space base;
size_t stolen_size; /* Total size of stolen memory */
unsigned long mappable_end; /* End offset that we can CPU map */
struct io_mapping *mappable; /* Mapping to our CPU mappable region */
phys_addr_t mappable_base; /* PA of our GMADR */
/** "Graphics Stolen Memory" holds the global PTEs */
void __iomem *gsm;
bool do_idle_maps;
int mtrr;
/* global gtt ops */
int (*gtt_probe)(struct drm_device *dev, size_t *gtt_total,
size_t *stolen, phys_addr_t *mappable_base,
unsigned long *mappable_end);
};
#define gtt_total_entries(gtt) ((gtt).base.total >> PAGE_SHIFT)
struct i915_hw_ppgtt {
struct i915_address_space base;
struct kref ref;
drm/i915: Use drm_mm for PPGTT PDEs When PPGTT support was originally enabled, it was only designed to support 1 PPGTT. It therefore made sense to simply hide the GGTT space required to enable this from the drm_mm allocator. Since we intend to support full PPGTT, which means more than 1, and they can be created and destroyed ad hoc it will be required to use the proper allocation techniques we already have. The first step here is to make the existing single PPGTT use the allocator. The astute observer will notice that we are reserving space in the GGTT for the PDEs for the lifetime of the address space, and would be right to question whether or not this is a good idea. It does not make a difference with this current patch only the aliasing PPGTT (indeed the PDEs should still be hidden from the shrinker). For the future, we are allocating from top to bottom to avoid using the precious "gtt space" The GGTT space at that point should only be used for scanout, HW contexts, ringbuffers, HWSP, PDEs, and a couple of other small buffers (potentially) used by the kernel. Everything else should be mapped into a PPGTT. To put the consumption in more tangible terms, it takes approximately 4 sets of PDEs to equal one 19x10 framebuffer (with no fancy stride or alignment constraints). 3/4 of the total [average] GGTT can be used for PDEs, and hopefully never touch the 1/4 that the framebuffer needs. The astute, and persistent observer might ask about the page tables which are also pinned for the address space. This waste is unfortunate. We use 2MB of memory per address space. We leave wrapping the PDEs as a real GEM object as a TODO. v2: Align PDEs to 64b in GTT Allocate the node dynamically so we can use drm_mm_put_block Now tested on IGT Allocate node at the top to avoid fragmentation (Chris) v3: Use Chris' top down allocator v4: Embed drm_mm_node into ppgtt struct (Jesse) Remove hunks which didn't belong (Jesse) v5: Don't subtract guard page since we now killed the guard page prior to this patch. (Ben) v6: Rebased and removed guard page stuff. Added a chunk to the commit message Allow adding a context to mappable region v7: Undo v3, so we can make the drm patch last in the series Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v4) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> squash: drm/i915: allow PPGTT to use mappable Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-12-07 05:11:07 +07:00
struct drm_mm_node node;
unsigned num_pd_entries;
union {
struct page **pt_pages;
struct page *gen8_pt_pages;
};
struct page *pd_pages;
int num_pd_pages;
int num_pt_pages;
union {
uint32_t pd_offset;
dma_addr_t pd_dma_addr[4];
};
union {
dma_addr_t *pt_dma_addr;
dma_addr_t *gen8_pt_dma_addr[4];
};
drm/i915: Convert execbuf code to use vmas In order to transition more of our code over to using a VMA instead of an <OBJ, VM> pair - we must have the vma accessible at execbuf time. Up until now, we've only had a VMA when actually binding an object. The previous patch helped handle the distinction on bound vs. unbound. This patch will help us catch leaks, and other issues before we actually shuffle a bunch of stuff around. This attempts to convert all the execbuf code to speak in vmas. Since the execbuf code is very self contained it was a nice isolated conversion. The meat of the code is about turning eb_objects into eb_vma, and then wiring up the rest of the code to use vmas instead of obj, vm pairs. Unfortunately, to do this, we must move the exec_list link from the obj structure. This list is reused in the eviction code, so we must also modify the eviction code to make this work. WARNING: This patch makes an already hotly profiled path slower. The cost is unavoidable. In reply to this mail, I will attach the extra data. v2: Release table lock early, and two a 2 phase vma lookup to avoid having to use a GFP_ATOMIC. (Chris) v3: s/obj_exec_list/obj_exec_link/ Updates to address commit 6d2b888569d366beb4be72cacfde41adee2c25e1 Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Wed Aug 7 18:30:54 2013 +0100 drm/i915: List objects allocated from stolen memory in debugfs v4: Use obj = vma->obj for neatness in some places (Chris) need_reloc_mappable() should return false if ppgtt (Chris) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> [danvet: Split out prep patches. Also remove a FIXME comment which is now taken care of.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-14 16:38:36 +07:00
int (*enable)(struct i915_hw_ppgtt *ppgtt);
int (*switch_mm)(struct i915_hw_ppgtt *ppgtt,
struct intel_ring_buffer *ring,
bool synchronous);
void (*debug_dump)(struct i915_hw_ppgtt *ppgtt, struct seq_file *m);
};
struct i915_ctx_hang_stats {
/* This context had batch pending when hang was declared */
unsigned batch_pending;
/* This context had batch active when hang was declared */
unsigned batch_active;
/* Time when this context was last blamed for a GPU reset */
unsigned long guilty_ts;
/* This context is banned to submit more work */
bool banned;
};
/* This must match up with the value previously used for execbuf2.rsvd1. */
#define DEFAULT_CONTEXT_ID 0
struct i915_hw_context {
struct kref ref;
int id;
bool is_initialized;
drm/i915: Do remaps for all contexts On both Ivybridge and Haswell, row remapping information is saved and restored with context. This means, we never actually properly supported the l3 remapping because our sysfs interface is asynchronous (and not tied to any context), and the known faulty HW would be reused by the next context to run. Not that due to the asynchronous nature of the sysfs entry, there is no point modifying the registers for the existing context. Instead we set a flag for all contexts to load the correct remapping information on the next run. Interested clients can use debugfs to determine whether or not the row has been remapped. One could propose at this point that we just do the remapping in the kernel. I guess since we have to maintain the sysfs interface anyway, I'm not sure how useful it is, and I do like keeping the policy in userspace; (it wasn't my original decision to make the interface the way it is, so I'm not attached). v2: Force a context switch when we have a remap on the next switch. (Ville) Don't let userspace use the interface with disabled contexts. v3: Don't force a context switch, just let it nop Improper context slice remap initialization, 1<<1 instead of 1<<i, but I rewrote it to avoid a second round of confusion. Error print moved to error path (All Ville) Added a comment on why the slice remap initialization happens. CC: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-19 09:03:18 +07:00
uint8_t remap_slice;
struct drm_i915_file_private *file_priv;
struct intel_ring_buffer *last_ring;
struct drm_i915_gem_object *obj;
struct i915_ctx_hang_stats hang_stats;
struct i915_address_space *vm;
struct list_head link;
};
struct i915_fbc {
unsigned long size;
unsigned int fb_id;
enum plane plane;
int y;
struct drm_mm_node *compressed_fb;
struct drm_mm_node *compressed_llb;
struct intel_fbc_work {
struct delayed_work work;
struct drm_crtc *crtc;
struct drm_framebuffer *fb;
} *fbc_work;
enum no_fbc_reason {
FBC_OK, /* FBC is enabled */
FBC_UNSUPPORTED, /* FBC is not supported by this chipset */
FBC_NO_OUTPUT, /* no outputs enabled to compress */
FBC_STOLEN_TOO_SMALL, /* not enough space for buffers */
FBC_UNSUPPORTED_MODE, /* interlace or doublescanned mode */
FBC_MODE_TOO_LARGE, /* mode too large for compression */
FBC_BAD_PLANE, /* fbc not supported on plane */
FBC_NOT_TILED, /* buffer not tiled */
FBC_MULTIPLE_PIPES, /* more than one pipe active */
FBC_MODULE_PARAM,
FBC_CHIP_DEFAULT, /* disabled by default on this chip */
} no_fbc_reason;
};
struct i915_psr {
bool sink_support;
bool source_ok;
};
enum intel_pch {
drm/i915: add PCH_NONE to enum intel_pch And rely on the fact that it's 0 to assume that machines without a PCH will have PCH_NONE as dev_priv->pch_type. Just today I finally realized that HAS_PCH_IBX is true for machines without a PCH. IMHO this is totally counter-intuitive and I don't think it's a good idea to assume that we're going to check for HAS_PCH_IBX only after we check for HAS_PCH_SPLIT. I believe that in the future we'll have more PCH types and checks like: if (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev)) will become more and more common. There's a good chance that we may break non-PCH machines by adding these checks in code that runs on all machines. I also believe that the HAS_PCH_SPLIT check will become less common as we add more and more different PCH types. We'll probably start replacing checks like: if (HAS_PCH_SPLIT(dev)) foo(); else bar(); with: if (HAS_PCH_NEW(dev)) baz(); else if (HAS_PCH_OLD(dev) || HAS_PCH_IBX(dev)) foo(); else bar(); and this may break gen 2/3/4. As far as we have investigated, this patch will affect the behavior of intel_hdmi_dpms and intel_dp_link_down on gen 4. In both functions the code inside the HAS_PCH_IBX check is for IBX-specific workarounds, so we should be safe. If we start bisecting gen 2/3/4 bugs to this commit we should consider replacing the HAS_PCH_IBX checks with something else. V2: Improve commit message, list possible side effects and solution. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-07-04 04:48:16 +07:00
PCH_NONE = 0, /* No PCH present */
PCH_IBX, /* Ibexpeak PCH */
PCH_CPT, /* Cougarpoint PCH */
PCH_LPT, /* Lynxpoint PCH */
PCH_NOP,
};
enum intel_sbi_destination {
SBI_ICLK,
SBI_MPHY,
};
#define QUIRK_PIPEA_FORCE (1<<0)
#define QUIRK_LVDS_SSC_DISABLE (1<<1)
#define QUIRK_INVERT_BRIGHTNESS (1<<2)
struct intel_fbdev;
struct intel_fbc_work;
struct intel_gmbus {
struct i2c_adapter adapter;
u32 force_bit;
u32 reg0;
u32 gpio_reg;
struct i2c_algo_bit_data bit_algo;
struct drm_i915_private *dev_priv;
};
struct i915_suspend_saved_registers {
u8 saveLBB;
u32 saveDSPACNTR;
u32 saveDSPBCNTR;
u32 saveDSPARB;
u32 savePIPEACONF;
u32 savePIPEBCONF;
u32 savePIPEASRC;
u32 savePIPEBSRC;
u32 saveFPA0;
u32 saveFPA1;
u32 saveDPLL_A;
u32 saveDPLL_A_MD;
u32 saveHTOTAL_A;
u32 saveHBLANK_A;
u32 saveHSYNC_A;
u32 saveVTOTAL_A;
u32 saveVBLANK_A;
u32 saveVSYNC_A;
u32 saveBCLRPAT_A;
u32 saveTRANSACONF;
u32 saveTRANS_HTOTAL_A;
u32 saveTRANS_HBLANK_A;
u32 saveTRANS_HSYNC_A;
u32 saveTRANS_VTOTAL_A;
u32 saveTRANS_VBLANK_A;
u32 saveTRANS_VSYNC_A;
u32 savePIPEASTAT;
u32 saveDSPASTRIDE;
u32 saveDSPASIZE;
u32 saveDSPAPOS;
u32 saveDSPAADDR;
u32 saveDSPASURF;
u32 saveDSPATILEOFF;
u32 savePFIT_PGM_RATIOS;
u32 saveBLC_HIST_CTL;
u32 saveBLC_PWM_CTL;
u32 saveBLC_PWM_CTL2;
u32 saveBLC_HIST_CTL_B;
u32 saveBLC_CPU_PWM_CTL;
u32 saveBLC_CPU_PWM_CTL2;
u32 saveFPB0;
u32 saveFPB1;
u32 saveDPLL_B;
u32 saveDPLL_B_MD;
u32 saveHTOTAL_B;
u32 saveHBLANK_B;
u32 saveHSYNC_B;
u32 saveVTOTAL_B;
u32 saveVBLANK_B;
u32 saveVSYNC_B;
u32 saveBCLRPAT_B;
u32 saveTRANSBCONF;
u32 saveTRANS_HTOTAL_B;
u32 saveTRANS_HBLANK_B;
u32 saveTRANS_HSYNC_B;
u32 saveTRANS_VTOTAL_B;
u32 saveTRANS_VBLANK_B;
u32 saveTRANS_VSYNC_B;
u32 savePIPEBSTAT;
u32 saveDSPBSTRIDE;
u32 saveDSPBSIZE;
u32 saveDSPBPOS;
u32 saveDSPBADDR;
u32 saveDSPBSURF;
u32 saveDSPBTILEOFF;
u32 saveVGA0;
u32 saveVGA1;
u32 saveVGA_PD;
u32 saveVGACNTRL;
u32 saveADPA;
u32 saveLVDS;
u32 savePP_ON_DELAYS;
u32 savePP_OFF_DELAYS;
u32 saveDVOA;
u32 saveDVOB;
u32 saveDVOC;
u32 savePP_ON;
u32 savePP_OFF;
u32 savePP_CONTROL;
u32 savePP_DIVISOR;
u32 savePFIT_CONTROL;
u32 save_palette_a[256];
u32 save_palette_b[256];
u32 saveFBC_CONTROL;
u32 saveIER;
u32 saveIIR;
u32 saveIMR;
u32 saveDEIER;
u32 saveDEIMR;
u32 saveGTIER;
u32 saveGTIMR;
u32 saveFDI_RXA_IMR;
u32 saveFDI_RXB_IMR;
u32 saveCACHE_MODE_0;
u32 saveMI_ARB_STATE;
u32 saveSWF0[16];
u32 saveSWF1[16];
u32 saveSWF2[3];
u8 saveMSR;
u8 saveSR[8];
u8 saveGR[25];
u8 saveAR_INDEX;
u8 saveAR[21];
u8 saveDACMASK;
u8 saveCR[37];
uint64_t saveFENCE[I915_MAX_NUM_FENCES];
u32 saveCURACNTR;
u32 saveCURAPOS;
u32 saveCURABASE;
u32 saveCURBCNTR;
u32 saveCURBPOS;
u32 saveCURBBASE;
u32 saveCURSIZE;
u32 saveDP_B;
u32 saveDP_C;
u32 saveDP_D;
u32 savePIPEA_GMCH_DATA_M;
u32 savePIPEB_GMCH_DATA_M;
u32 savePIPEA_GMCH_DATA_N;
u32 savePIPEB_GMCH_DATA_N;
u32 savePIPEA_DP_LINK_M;
u32 savePIPEB_DP_LINK_M;
u32 savePIPEA_DP_LINK_N;
u32 savePIPEB_DP_LINK_N;
u32 saveFDI_RXA_CTL;
u32 saveFDI_TXA_CTL;
u32 saveFDI_RXB_CTL;
u32 saveFDI_TXB_CTL;
u32 savePFA_CTL_1;
u32 savePFB_CTL_1;
u32 savePFA_WIN_SZ;
u32 savePFB_WIN_SZ;
u32 savePFA_WIN_POS;
u32 savePFB_WIN_POS;
u32 savePCH_DREF_CONTROL;
u32 saveDISP_ARB_CTL;
u32 savePIPEA_DATA_M1;
u32 savePIPEA_DATA_N1;
u32 savePIPEA_LINK_M1;
u32 savePIPEA_LINK_N1;
u32 savePIPEB_DATA_M1;
u32 savePIPEB_DATA_N1;
u32 savePIPEB_LINK_M1;
u32 savePIPEB_LINK_N1;
u32 saveMCHBAR_RENDER_STANDBY;
u32 savePCH_PORT_HOTPLUG;
};
struct intel_gen6_power_mgmt {
/* work and pm_iir are protected by dev_priv->irq_lock */
struct work_struct work;
u32 pm_iir;
u8 cur_delay;
u8 min_delay;
u8 max_delay;
u8 rpe_delay;
drm/i915: Tweak RPS thresholds to more aggressively downclock After applying wait-boost we often find ourselves stuck at higher clocks than required. The current threshold value requires the GPU to be continuously and completely idle for 313ms before it is dropped by one bin. Conversely, we require the GPU to be busy for an average of 90% over a 84ms period before we upclock. So the current thresholds almost never downclock the GPU, and respond very slowly to sudden demands for more power. It is easy to observe that we currently lock into the wrong bin and both underperform in benchmarks and consume more power than optimal (just by repeating the task and measuring the different results). An alternative approach, as discussed in the bspec, is to use a continuous threshold for upclocking, and an average value for downclocking. This is good for quickly detecting and reacting to state changes within a frame, however it fails with the common throttling method of waiting upon the outstanding frame - at least it is difficult to choose a threshold that works well at 15,000fps and at 60fps. So continue to use average busy/idle loads to determine frequency change. v2: Use 3 power zones to keep frequencies low in steady-state mostly idle (e.g. scrolling, interactive 2D drawing), and frequencies high for demanding games. In between those end-states, we use a fast-reclocking algorithm to converge more quickly on the desired bin. v3: Bug fixes - make sure we reset adj after switching power zones. v4: Tune - drop the continuous busy thresholds as it prevents us from choosing the right frequency for glxgears style swap benchmarks. Instead the goal is to be able to find the right clocks irrespective of the wait-boost. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:57 +07:00
u8 rp1_delay;
u8 rp0_delay;
u8 hw_max;
bool rp_up_masked;
bool rp_down_masked;
drm/i915: Tweak RPS thresholds to more aggressively downclock After applying wait-boost we often find ourselves stuck at higher clocks than required. The current threshold value requires the GPU to be continuously and completely idle for 313ms before it is dropped by one bin. Conversely, we require the GPU to be busy for an average of 90% over a 84ms period before we upclock. So the current thresholds almost never downclock the GPU, and respond very slowly to sudden demands for more power. It is easy to observe that we currently lock into the wrong bin and both underperform in benchmarks and consume more power than optimal (just by repeating the task and measuring the different results). An alternative approach, as discussed in the bspec, is to use a continuous threshold for upclocking, and an average value for downclocking. This is good for quickly detecting and reacting to state changes within a frame, however it fails with the common throttling method of waiting upon the outstanding frame - at least it is difficult to choose a threshold that works well at 15,000fps and at 60fps. So continue to use average busy/idle loads to determine frequency change. v2: Use 3 power zones to keep frequencies low in steady-state mostly idle (e.g. scrolling, interactive 2D drawing), and frequencies high for demanding games. In between those end-states, we use a fast-reclocking algorithm to converge more quickly on the desired bin. v3: Bug fixes - make sure we reset adj after switching power zones. v4: Tune - drop the continuous busy thresholds as it prevents us from choosing the right frequency for glxgears style swap benchmarks. Instead the goal is to be able to find the right clocks irrespective of the wait-boost. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:57 +07:00
int last_adj;
enum { LOW_POWER, BETWEEN, HIGH_POWER } power;
bool enabled;
struct delayed_work delayed_resume_work;
/*
* Protects RPS/RC6 register access and PCU communication.
* Must be taken after struct_mutex if nested.
*/
struct mutex hw_lock;
};
/* defined intel_pm.c */
extern spinlock_t mchdev_lock;
struct intel_ilk_power_mgmt {
u8 cur_delay;
u8 min_delay;
u8 max_delay;
u8 fmax;
u8 fstart;
u64 last_count1;
unsigned long last_time1;
unsigned long chipset_power;
u64 last_count2;
struct timespec last_time2;
unsigned long gfx_power;
u8 corr;
int c_m;
int r_t;
struct drm_i915_gem_object *pwrctx;
struct drm_i915_gem_object *renderctx;
};
/* Power well structure for haswell */
struct i915_power_well {
const char *name;
bool always_on;
/* power well enable/disable usage count */
int count;
unsigned long domains;
void *data;
void (*set)(struct drm_device *dev, struct i915_power_well *power_well,
bool enable);
bool (*is_enabled)(struct drm_device *dev,
struct i915_power_well *power_well);
};
struct i915_power_domains {
drm/i915: use power get/put instead of set for power on after init Currently we make sure that all power domains are enabled during driver init and turn off unneded ones only after the first modeset. Similarly during suspend we enable all power domains, which will remain on through the following resume until the first modeset. This logic is supported by intel_set_power_well() in the power domain framework. It would be nice to simplify the API, so that we only have get/put functions and make it more explicit on the higher level how this "power well on during init" logic works. This will make it also easier if in the future we want to shorten the time the power wells are on. For this add a new device private flag tracking whether we have the power wells on because of init/suspend and use only intel_display_power_get()/put(). As nothing else uses intel_set_power_well() we can remove it. This also fixes commit 6efdf354ddb186c6604d1692075421e8d2c740e9 Author: Imre Deak <imre.deak@intel.com> Date: Wed Oct 16 17:25:52 2013 +0300 drm/i915: enable only the needed power domains during modeset where removing intel_set_power_well() resulted in not releasing the reference on the power well that was taken during init and thus leaving the power well on all the time. Regression reported by Paulo. v2: - move the init_power_on flag to the power_domains struct (Daniel) v3: - add note about this being a regression fix too (Paulo) Signed-off-by: Imre Deak <imre.deak@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-10-25 21:36:48 +07:00
/*
* Power wells needed for initialization at driver init and suspend
* time are on. They are kept on until after the first modeset.
*/
bool init_power_on;
int power_well_count;
drm/i915: use power get/put instead of set for power on after init Currently we make sure that all power domains are enabled during driver init and turn off unneded ones only after the first modeset. Similarly during suspend we enable all power domains, which will remain on through the following resume until the first modeset. This logic is supported by intel_set_power_well() in the power domain framework. It would be nice to simplify the API, so that we only have get/put functions and make it more explicit on the higher level how this "power well on during init" logic works. This will make it also easier if in the future we want to shorten the time the power wells are on. For this add a new device private flag tracking whether we have the power wells on because of init/suspend and use only intel_display_power_get()/put(). As nothing else uses intel_set_power_well() we can remove it. This also fixes commit 6efdf354ddb186c6604d1692075421e8d2c740e9 Author: Imre Deak <imre.deak@intel.com> Date: Wed Oct 16 17:25:52 2013 +0300 drm/i915: enable only the needed power domains during modeset where removing intel_set_power_well() resulted in not releasing the reference on the power well that was taken during init and thus leaving the power well on all the time. Regression reported by Paulo. v2: - move the init_power_on flag to the power_domains struct (Daniel) v3: - add note about this being a regression fix too (Paulo) Signed-off-by: Imre Deak <imre.deak@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-10-25 21:36:48 +07:00
struct mutex lock;
int domain_use_count[POWER_DOMAIN_NUM];
struct i915_power_well *power_wells;
};
struct i915_dri1_state {
unsigned allow_batchbuffer : 1;
u32 __iomem *gfx_hws_cpu_addr;
unsigned int cpp;
int back_offset;
int front_offset;
int current_page;
int page_flipping;
uint32_t counter;
};
struct i915_ums_state {
/**
* Flag if the X Server, and thus DRM, is not currently in
* control of the device.
*
* This is set between LeaveVT and EnterVT. It needs to be
* replaced with a semaphore. It also needs to be
* transitioned away from for kernel modesetting.
*/
int mm_suspended;
};
#define MAX_L3_SLICES 2
struct intel_l3_parity {
u32 *remap_info[MAX_L3_SLICES];
struct work_struct error_work;
int which_slice;
};
struct i915_gem_mm {
/** Memory allocator for GTT stolen memory */
struct drm_mm stolen;
/** List of all objects in gtt_space. Used to restore gtt
* mappings on resume */
struct list_head bound_list;
/**
* List of objects which are not bound to the GTT (thus
* are idle and not used by the GPU) but still have
* (presumably uncached) pages still attached.
*/
struct list_head unbound_list;
/** Usable portion of the GTT for GEM */
unsigned long stolen_base; /* limited to low memory (32-bit) */
/** PPGTT used for aliasing the PPGTT with the GTT */
struct i915_hw_ppgtt *aliasing_ppgtt;
struct shrinker inactive_shrinker;
bool shrinker_no_lock_stealing;
/** LRU list of objects with fence regs on them. */
struct list_head fence_list;
/**
* We leave the user IRQ off as much as possible,
* but this means that requests will finish and never
* be retired once the system goes idle. Set a timer to
* fire periodically while the ring is running. When it
* fires, go retire requests.
*/
struct delayed_work retire_work;
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
/**
* When we detect an idle GPU, we want to turn on
* powersaving features. So once we see that there
* are no more requests outstanding and no more
* arrive within a small period of time, we fire
* off the idle_work.
*/
struct delayed_work idle_work;
/**
* Are we in a non-interruptible section of code like
* modesetting?
*/
bool interruptible;
/** Bit 6 swizzling required for X tiling */
uint32_t bit_6_swizzle_x;
/** Bit 6 swizzling required for Y tiling */
uint32_t bit_6_swizzle_y;
/* storage for physical objects */
struct drm_i915_gem_phys_object *phys_objs[I915_MAX_PHYS_OBJECT];
/* accounting, useful for userland debugging */
spinlock_t object_stat_lock;
size_t object_memory;
u32 object_count;
};
struct drm_i915_error_state_buf {
unsigned bytes;
unsigned size;
int err;
u8 *buf;
loff_t start;
loff_t pos;
};
struct i915_error_state_file_priv {
struct drm_device *dev;
struct drm_i915_error_state *error;
};
struct i915_gpu_error {
/* For hangcheck timer */
#define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */
#define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD)
/* Hang gpu twice in this window and your context gets banned */
#define DRM_I915_CTX_BAN_PERIOD DIV_ROUND_UP(8*DRM_I915_HANGCHECK_PERIOD, 1000)
struct timer_list hangcheck_timer;
/* For reset and error_state handling. */
spinlock_t lock;
/* Protected by the above dev->gpu_error.lock. */
struct drm_i915_error_state *first_error;
struct work_struct work;
unsigned long missed_irq_rings;
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
/**
* State variable controlling the reset flow and count
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
*
* This is a counter which gets incremented when reset is triggered,
* and again when reset has been handled. So odd values (lowest bit set)
* means that reset is in progress and even values that
* (reset_counter >> 1):th reset was successfully completed.
*
* If reset is not completed succesfully, the I915_WEDGE bit is
* set meaning that hardware is terminally sour and there is no
* recovery. All waiters on the reset_queue will be woken when
* that happens.
*
* This counter is used by the wait_seqno code to notice that reset
* event happened and it needs to restart the entire ioctl (since most
* likely the seqno it waited for won't ever signal anytime soon).
drm/i915: create a race-free reset detection With the previous patch the state transition handling of the reset code itself is now (hopefully) race free and solid. But that still leaves out everyone else - with the various lock-free wait paths we have there's the possibility that the reset happens between the point where we read the seqno we should wait on and the actual wait. And if __wait_seqno then never sees the RESET_IN_PROGRESS state, we'll happily wait for a seqno which will in all likelyhood never signal. In practice this is not a big problem since the X server gets constantly interrupted, and can then submit more work (hopefully) to unblock everyone else: As soon as a new seqno write lands, all waiters will unblock. But running the i-g-t reset testcase ZZ_hangman can expose this race, especially on slower hw with fewer cpu cores. Now looking forward to ARB_robustness and friends that's not the best possible behaviour, hence this patch adds a reset_counter to be able to detect any reset, even if a given thread never observed the in-progress state. The important part is to correctly order things: - The write side needs to increment the counter after any seqno gets reset. Hence we need to do that at the end of the reset work, and again wake everyone up. We also need to place a barrier in between any possible seqno changes and the counter increment, since any unlock operations only guarantee that nothing leaks out, but not that at later load operation gets moved ahead. - On the read side we need to ensure that no reset can sneak in and invalidate the seqno. In all cases we can use the one-sided barrier that unlock operations guarantee (of the lock protecting the respective seqno/ring pair) to ensure correct ordering. Hence it is sufficient to place the atomic read before the mutex/spin_unlock and no additional barriers are required. The end-result of all this is that we need to wake up everyone twice in a reset operation: - First, before the reset starts, to get any lockholders of the locks, so that the reset can proceed. - Second, after the reset is completed, to allow waiters to properly and reliably detect the reset condition and bail out. I admit that this entire reset_counter thing smells a bit like overkill, but I think it's justified since it makes it really explicit what the bail-out condition is. And we need a reset counter anyway to implement ARB_robustness, and imo with finer-grained locking on the horizont this is the most resilient scheme I could think of. v2: Drop spurious change in the wait_for_error EXIT_COND - we only need to wait until we leave the reset-in-progress wedged state. v3: Don't play tricks with barriers in the throttle ioctl, the spin_unlock is barrier enough. I've also considered using a little helper to grab the current reset_counter, but then decided that hiding the atomic_read isn't a great idea, since having it explicitly show up in the code is a nice remainder to reviews to check the memory barriers. v4: Add a comment to explain why we need to fall through in __wait_seqno in the end variable assignments. v5: Review from Damien: - s/smb/smp/ in a comment - don't increment the reset counter after we've set it to WEDGED. Now we (again) properly wedge the gpu when the reset fails. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-06 15:01:42 +07:00
*
* This is important for lock-free wait paths, where no contended lock
* naturally enforces the correct ordering between the bail-out of the
* waiter and the gpu reset work code.
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
*/
atomic_t reset_counter;
#define I915_RESET_IN_PROGRESS_FLAG 1
#define I915_WEDGED (1 << 31)
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
/**
* Waitqueue to signal when the reset has completed. Used by clients
* that wait for dev_priv->mm.wedged to settle.
*/
wait_queue_head_t reset_queue;
/* For gpu hang simulation. */
unsigned int stop_rings;
/* For missed irq/seqno simulation. */
unsigned int test_irq_rings;
};
i915: ignore lid open event when resuming i915 driver needs to do modeset when 1. system resumes from sleep 2. lid is opened In PM_SUSPEND_MEM state, all the GPEs are cleared when system resumes, thus it is the i915_resume code does the modeset rather than intel_lid_notify(). But in PM_SUSPEND_FREEZE state, this will be broken because system is still responsive to the lid events. 1. When we close the lid in Freeze state, intel_lid_notify() sets modeset_on_lid. 2. When we reopen the lid, intel_lid_notify() will do a modeset, before the system is resumed. here is the error log, [92146.548074] WARNING: at drivers/gpu/drm/i915/intel_display.c:1028 intel_wait_for_pipe_off+0x184/0x190 [i915]() [92146.548076] Hardware name: VGN-Z540N [92146.548078] pipe_off wait timed out [92146.548167] Modules linked in: hid_generic usbhid hid snd_hda_codec_realtek snd_hda_intel snd_hda_codec parport_pc snd_hwdep ppdev snd_pcm_oss i915 snd_mixer_oss snd_pcm arc4 iwldvm snd_seq_dummy mac80211 snd_seq_oss snd_seq_midi fbcon tileblit font bitblit softcursor drm_kms_helper snd_rawmidi snd_seq_midi_event coretemp drm snd_seq kvm btusb bluetooth snd_timer iwlwifi pcmcia tpm_infineon i2c_algo_bit joydev snd_seq_device intel_agp cfg80211 snd intel_gtt yenta_socket pcmcia_rsrc sony_laptop agpgart microcode psmouse tpm_tis serio_raw mxm_wmi soundcore snd_page_alloc tpm acpi_cpufreq lpc_ich pcmcia_core tpm_bios mperf processor lp parport firewire_ohci firewire_core crc_itu_t sdhci_pci sdhci thermal e1000e [92146.548173] Pid: 4304, comm: kworker/0:0 Tainted: G W 3.8.0-rc3-s0i3-v3-test+ #9 [92146.548175] Call Trace: [92146.548189] [<c10378e2>] warn_slowpath_common+0x72/0xa0 [92146.548227] [<f86398b4>] ? intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548263] [<f86398b4>] ? intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548270] [<c10379b3>] warn_slowpath_fmt+0x33/0x40 [92146.548307] [<f86398b4>] intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548344] [<f86399c2>] intel_disable_pipe+0x102/0x190 [i915] [92146.548380] [<f8639ea4>] ? intel_disable_plane+0x64/0x80 [i915] [92146.548417] [<f8639f7c>] i9xx_crtc_disable+0xbc/0x150 [i915] [92146.548456] [<f863ebee>] intel_crtc_update_dpms+0x5e/0x90 [i915] [92146.548493] [<f86437cf>] intel_modeset_setup_hw_state+0x42f/0x8f0 [i915] [92146.548535] [<f8645b0b>] intel_lid_notify+0x9b/0xc0 [i915] [92146.548543] [<c15610d3>] notifier_call_chain+0x43/0x60 [92146.548550] [<c105d1e1>] __blocking_notifier_call_chain+0x41/0x80 [92146.548556] [<c105d23f>] blocking_notifier_call_chain+0x1f/0x30 [92146.548563] [<c131a684>] acpi_lid_send_state+0x78/0xa4 [92146.548569] [<c131aa9e>] acpi_button_notify+0x3b/0xf1 [92146.548577] [<c12df56a>] ? acpi_os_execute+0x17/0x19 [92146.548582] [<c12e591a>] ? acpi_ec_sync_query+0xa5/0xbc [92146.548589] [<c12e2b82>] acpi_device_notify+0x16/0x18 [92146.548595] [<c12f4904>] acpi_ev_notify_dispatch+0x38/0x4f [92146.548600] [<c12df0e8>] acpi_os_execute_deferred+0x20/0x2b [92146.548607] [<c1051208>] process_one_work+0x128/0x3f0 [92146.548613] [<c1564f73>] ? common_interrupt+0x33/0x38 [92146.548618] [<c104f8c0>] ? wake_up_worker+0x30/0x30 [92146.548624] [<c12df0c8>] ? acpi_os_wait_events_complete+0x1e/0x1e [92146.548629] [<c10524f9>] worker_thread+0x119/0x3b0 [92146.548634] [<c10523e0>] ? manage_workers+0x240/0x240 [92146.548640] [<c1056e84>] kthread+0x94/0xa0 [92146.548647] [<c1060000>] ? ftrace_raw_output_sched_stat_runtime+0x70/0xf0 [92146.548652] [<c15649b7>] ret_from_kernel_thread+0x1b/0x28 [92146.548658] [<c1056df0>] ? kthread_create_on_node+0xc0/0xc0 three different modeset flags are introduced in this patch MODESET_ON_LID_OPEN: do modeset on next lid open event MODESET_DONE: modeset already done MODESET_SUSPENDED: suspended, only do modeset when system is resumed In this way, 1. when lid is closed, MODESET_ON_LID_OPEN is set so that we'll do modeset on next lid open event. 2. when lid is opened, MODESET_DONE is set so that duplicate lid open events will be ignored. 3. when system suspends, MODESET_SUSPENDED is set. In this case, we will not do modeset on any lid events. Plus, locking mechanism is also introduced to avoid racing. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-02-05 14:41:53 +07:00
enum modeset_restore {
MODESET_ON_LID_OPEN,
MODESET_DONE,
MODESET_SUSPENDED,
};
struct ddi_vbt_port_info {
uint8_t hdmi_level_shift;
uint8_t supports_dvi:1;
uint8_t supports_hdmi:1;
uint8_t supports_dp:1;
};
struct intel_vbt_data {
struct drm_display_mode *lfp_lvds_vbt_mode; /* if any */
struct drm_display_mode *sdvo_lvds_vbt_mode; /* if any */
/* Feature bits */
unsigned int int_tv_support:1;
unsigned int lvds_dither:1;
unsigned int lvds_vbt:1;
unsigned int int_crt_support:1;
unsigned int lvds_use_ssc:1;
unsigned int display_clock_mode:1;
unsigned int fdi_rx_polarity_inverted:1;
int lvds_ssc_freq;
unsigned int bios_lvds_val; /* initial [PCH_]LVDS reg val in VBIOS */
/* eDP */
int edp_rate;
int edp_lanes;
int edp_preemphasis;
int edp_vswing;
bool edp_initialized;
bool edp_support;
int edp_bpp;
struct edp_power_seq edp_pps;
struct {
u16 pwm_freq_hz;
bool active_low_pwm;
} backlight;
/* MIPI DSI */
struct {
u16 panel_id;
} dsi;
int crt_ddc_pin;
int child_dev_num;
union child_device_config *child_dev;
struct ddi_vbt_port_info ddi_port_info[I915_MAX_PORTS];
};
enum intel_ddb_partitioning {
INTEL_DDB_PART_1_2,
INTEL_DDB_PART_5_6, /* IVB+ */
};
struct intel_wm_level {
bool enable;
uint32_t pri_val;
uint32_t spr_val;
uint32_t cur_val;
uint32_t fbc_val;
};
struct ilk_wm_values {
uint32_t wm_pipe[3];
uint32_t wm_lp[3];
uint32_t wm_lp_spr[3];
uint32_t wm_linetime[3];
bool enable_fbc_wm;
enum intel_ddb_partitioning partitioning;
};
drm/i915: allow package C8+ states on Haswell (disabled) This patch allows PC8+ states on Haswell. These states can only be reached when all the display outputs are disabled, and they allow some more power savings. The fact that the graphics device is allowing PC8+ doesn't mean that the machine will actually enter PC8+: all the other devices also need to allow PC8+. For now this option is disabled by default. You need i915.allow_pc8=1 if you want it. This patch adds a big comment inside i915_drv.h explaining how it works and how it tracks things. Read it. v2: (this is not really v2, many previous versions were already sent, but they had different names) - Use the new functions to enable/disable GTIMR and GEN6_PMIMR - Rename almost all variables and functions to names suggested by Chris - More WARNs on the IRQ handling code - Also disable PC8 when there's GPU work to do (thanks to Ben for the help on this), so apps can run caster - Enable PC8 on a delayed work function that is delayed for 5 seconds. This makes sure we only enable PC8+ if we're really idle - Make sure we're not in PC8+ when suspending v3: - WARN if IRQs are disabled on __wait_seqno - Replace some DRM_ERRORs with WARNs - Fix calls to restore GT and PM interrupts - Use intel_mark_busy instead of intel_ring_advance to disable PC8 v4: - Use the force_wake, Luke! v5: - Remove the "IIR is not zero" WARNs - Move the force_wake chunk to its own patch - Only restore what's missing from RC6, not everything Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-19 23:18:09 +07:00
/*
* This struct tracks the state needed for the Package C8+ feature.
*
* Package states C8 and deeper are really deep PC states that can only be
* reached when all the devices on the system allow it, so even if the graphics
* device allows PC8+, it doesn't mean the system will actually get to these
* states.
*
* Our driver only allows PC8+ when all the outputs are disabled, the power well
* is disabled and the GPU is idle. When these conditions are met, we manually
* do the other conditions: disable the interrupts, clocks and switch LCPLL
* refclk to Fclk.
*
* When we really reach PC8 or deeper states (not just when we allow it) we lose
* the state of some registers, so when we come back from PC8+ we need to
* restore this state. We don't get into PC8+ if we're not in RC6, so we don't
* need to take care of the registers kept by RC6.
*
* The interrupt disabling is part of the requirements. We can only leave the
* PCH HPD interrupts enabled. If we're in PC8+ and we get another interrupt we
* can lock the machine.
*
* Ideally every piece of our code that needs PC8+ disabled would call
* hsw_disable_package_c8, which would increment disable_count and prevent the
* system from reaching PC8+. But we don't have a symmetric way to do this for
* everything, so we have the requirements_met and gpu_idle variables. When we
* switch requirements_met or gpu_idle to true we decrease disable_count, and
* increase it in the opposite case. The requirements_met variable is true when
* all the CRTCs, encoders and the power well are disabled. The gpu_idle
* variable is true when the GPU is idle.
*
* In addition to everything, we only actually enable PC8+ if disable_count
* stays at zero for at least some seconds. This is implemented with the
* enable_work variable. We do this so we don't enable/disable PC8 dozens of
* consecutive times when all screens are disabled and some background app
* queries the state of our connectors, or we have some application constantly
* waking up to use the GPU. Only after the enable_work function actually
* enables PC8+ the "enable" variable will become true, which means that it can
* be false even if disable_count is 0.
*
* The irqs_disabled variable becomes true exactly after we disable the IRQs and
* goes back to false exactly before we reenable the IRQs. We use this variable
* to check if someone is trying to enable/disable IRQs while they're supposed
* to be disabled. This shouldn't happen and we'll print some error messages in
* case it happens, but if it actually happens we'll also update the variables
* inside struct regsave so when we restore the IRQs they will contain the
* latest expected values.
*
* For more, read "Display Sequences for Package C8" on our documentation.
*/
struct i915_package_c8 {
bool requirements_met;
bool gpu_idle;
bool irqs_disabled;
/* Only true after the delayed work task actually enables it. */
bool enabled;
int disable_count;
struct mutex lock;
struct delayed_work enable_work;
struct {
uint32_t deimr;
uint32_t sdeimr;
uint32_t gtimr;
uint32_t gtier;
uint32_t gen6_pmimr;
} regsave;
};
struct i915_runtime_pm {
bool suspended;
};
enum intel_pipe_crc_source {
INTEL_PIPE_CRC_SOURCE_NONE,
INTEL_PIPE_CRC_SOURCE_PLANE1,
INTEL_PIPE_CRC_SOURCE_PLANE2,
INTEL_PIPE_CRC_SOURCE_PF,
INTEL_PIPE_CRC_SOURCE_PIPE,
/* TV/DP on pre-gen5/vlv can't use the pipe source. */
INTEL_PIPE_CRC_SOURCE_TV,
INTEL_PIPE_CRC_SOURCE_DP_B,
INTEL_PIPE_CRC_SOURCE_DP_C,
INTEL_PIPE_CRC_SOURCE_DP_D,
INTEL_PIPE_CRC_SOURCE_AUTO,
INTEL_PIPE_CRC_SOURCE_MAX,
};
struct intel_pipe_crc_entry {
uint32_t frame;
uint32_t crc[5];
};
#define INTEL_PIPE_CRC_ENTRIES_NR 128
struct intel_pipe_crc {
spinlock_t lock;
bool opened; /* exclusive access to the result file */
struct intel_pipe_crc_entry *entries;
enum intel_pipe_crc_source source;
int head, tail;
wait_queue_head_t wq;
};
typedef struct drm_i915_private {
struct drm_device *dev;
struct kmem_cache *slab;
const struct intel_device_info info;
int relative_constants_mode;
void __iomem *regs;
struct intel_uncore uncore;
struct intel_gmbus gmbus[GMBUS_NUM_PORTS];
drm/i915: use the gmbus irq for waits We need two special things to properly wire this up: - Add another argument to gmbus_wait_hw_status to pass in the correct interrupt bit in gmbus4. - Since we can only get an irq for one of the two events we want, hand-roll the wait_event_timeout code so that we wake up every jiffie and can check for NAKs. This way we also subsume gmbus support for platforms without interrupts (or where those are not yet enabled). The important bit really is to only enable one gmbus interrupt source at the same time - with that piece of lore figured out, this seems to work flawlessly. Ben Widawsky rightfully complained the lack of measurements for the claimed benefits (especially since the first version was actually broken and fell back to bit-banging). Previously reading the 256 byte hdmi EDID takes about 72 ms here. With this patch it's down to 33 ms. Given that transfering the 256 bytes over i2c at wire speed takes 20.5ms alone, the reduction in additional overhead is rather nice. v2: Chris Wilson wondered whether GMBUS4 might contain some set bits when booting up an hence result in some spurious interrupts. Since we clear GMBUS4 after every wait and we do gmbus transfer really early in the setup sequence to detect displays the window is small, but still be paranoid and clear it properly. v3: Clarify the comment that gmbus irq generation can only support one kind of event, why it bothers us and how we work around that limit. Cc: Daniel Kurtz <djkurtz@chromium.org> Reviewed-by: Imre Deak <imre.deak@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-01 19:53:45 +07:00
/** gmbus_mutex protects against concurrent usage of the single hw gmbus
* controller on different i2c buses. */
struct mutex gmbus_mutex;
/**
* Base address of the gmbus and gpio block.
*/
uint32_t gpio_mmio_base;
drm/i915: use the gmbus irq for waits We need two special things to properly wire this up: - Add another argument to gmbus_wait_hw_status to pass in the correct interrupt bit in gmbus4. - Since we can only get an irq for one of the two events we want, hand-roll the wait_event_timeout code so that we wake up every jiffie and can check for NAKs. This way we also subsume gmbus support for platforms without interrupts (or where those are not yet enabled). The important bit really is to only enable one gmbus interrupt source at the same time - with that piece of lore figured out, this seems to work flawlessly. Ben Widawsky rightfully complained the lack of measurements for the claimed benefits (especially since the first version was actually broken and fell back to bit-banging). Previously reading the 256 byte hdmi EDID takes about 72 ms here. With this patch it's down to 33 ms. Given that transfering the 256 bytes over i2c at wire speed takes 20.5ms alone, the reduction in additional overhead is rather nice. v2: Chris Wilson wondered whether GMBUS4 might contain some set bits when booting up an hence result in some spurious interrupts. Since we clear GMBUS4 after every wait and we do gmbus transfer really early in the setup sequence to detect displays the window is small, but still be paranoid and clear it properly. v3: Clarify the comment that gmbus irq generation can only support one kind of event, why it bothers us and how we work around that limit. Cc: Daniel Kurtz <djkurtz@chromium.org> Reviewed-by: Imre Deak <imre.deak@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-01 19:53:45 +07:00
wait_queue_head_t gmbus_wait_queue;
struct pci_dev *bridge_dev;
struct intel_ring_buffer ring[I915_NUM_RINGS];
uint32_t last_seqno, next_seqno;
drm_dma_handle_t *status_page_dmah;
struct resource mch_res;
/* protects the irq masks */
spinlock_t irq_lock;
drm/i915: irq-drive the dp aux communication At least on the platforms that have a dp aux irq and also have it enabled - vlvhsw should have one, too. But I don't have a machine to test this on. Judging from docs there's no dp aux interrupt for gm45. Also, I only have an ivb cpu edp machine, so the dp aux A code for snb/ilk is untested. For dpcd probing when nothing is connected it slashes about 5ms of cpu time (cpu time is now negligible), which agrees with 3 * 5 400 usec timeouts. A previous version of this patch increases the time required to go through the dp_detect cycle (which includes reading the edid) from around 33 ms to around 40 ms. Experiments indicated that this is purely due to the irq latency - the hw doesn't allow us to queue up dp aux transactions and hence irq latency directly affects throughput. gmbus is much better, there we have a 8 byte buffer, and we get the irq once another 4 bytes can be queued up. But by using the pm_qos interface to request the lowest possible cpu wake-up latency this slowdown completely disappeared. Since all our output detection logic is single-threaded with the mode_config mutex right now anyway, I've decide not ot play fancy and to just reuse the gmbus wait queue. But this would definitely prep the way to run dp detection on different ports in parallel v2: Add a timeout for dp aux transfers when using interrupts - the hw _does_ prevent this with the hw-based 400 usec timeout, but if the irq somehow doesn't arrive we're screwed. Lesson learned while developing this ;-) v3: While at it also convert the busy-loop to wait_for_atomic, so that we don't run the risk of an infinite loop any more. v4: Ensure we have the smallest possible irq latency by using the pm_qos interface. v5: Add a comment to the code to explain why we frob pm_qos. Suggested by Chris Wilson. v6: Disable dp irq for vlv, that's easier than trying to get at docs and hw. v7: Squash in a fix for Haswell that Paulo Zanoni tracked down - the dp aux registers aren't at a fixed offset any more, but can be on the PCH while the DP port is on the cpu die. Reviewed-by: Imre Deak <imre.deak@intel.com> (v6) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-01 19:53:48 +07:00
/* To control wakeup latency, e.g. for irq-driven dp aux transfers. */
struct pm_qos_request pm_qos;
/* DPIO indirect register protection */
struct mutex dpio_lock;
/** Cached value of IMR to avoid reads in updating the bitfield */
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 11:07:09 +07:00
union {
u32 irq_mask;
u32 de_irq_mask[I915_MAX_PIPES];
};
u32 gt_irq_mask;
u32 pm_irq_mask;
u32 pipestat_irq_mask[I915_MAX_PIPES];
struct work_struct hotplug_work;
bool enable_hotplug_processing;
2013-04-16 18:36:54 +07:00
struct {
unsigned long hpd_last_jiffies;
int hpd_cnt;
enum {
HPD_ENABLED = 0,
HPD_DISABLED = 1,
HPD_MARK_DISABLED = 2
} hpd_mark;
} hpd_stats[HPD_NUM_PINS];
u32 hpd_event_bits;
struct timer_list hotplug_reenable_timer;
struct i915_fbc fbc;
struct intel_opregion opregion;
struct intel_vbt_data vbt;
/* overlay */
struct intel_overlay *overlay;
/* backlight registers and fields in struct intel_panel */
spinlock_t backlight_lock;
/* LVDS info */
bool no_aux_handshake;
struct drm_i915_fence_reg fence_regs[I915_MAX_NUM_FENCES]; /* assume 965 */
int fence_reg_start; /* 4 if userland hasn't ioctl'd us yet */
int num_fence_regs; /* 8 on pre-965, 16 otherwise */
unsigned int fsb_freq, mem_freq, is_ddr3;
drm/i915: fix hpd work vs. flush_work in the pageflip code deadlock Historically we've run our own driver hotplug handling in our own work-queue, which then launched the drm core hotplug handling in the system workqueue. This is important since we flush our own driver workqueue in the pageflip code while hodling modeset locks, and only the drm hotplug code grabbed these locks. But with commit 69787f7da6b2adc4054357a661aaa1701a9ca76f Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Tue Oct 23 18:23:34 2012 +0000 drm: run the hpd irq event code directly this was changed and now we could deadlock in our flip handler if there's a hotplug work blocking the progress of the crucial unpin works. So this broke the careful deadlock avoidance implemented in commit b4a98e57fc27854b5938fc8b08b68e5e68b91e1f Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Thu Nov 1 09:26:26 2012 +0000 drm/i915: Flush outstanding unpin tasks before pageflipping Since the rule thus far has been that work items on our own workqueue may never grab modeset locks simply restore that rule again. v2: Add a comment to the declaration of dev_priv->wq to warn readers about the tricky implications of using it. Suggested by Chris Wilson. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Stuart Abercrombie <sabercrombie@chromium.org> Reported-by: Stuart Abercrombie <sabercrombie@chromium.org> References: http://permalink.gmane.org/gmane.comp.freedesktop.xorg.drivers.intel/26239 Cc: stable@vger.kernel.org Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: Squash in a comment at the place where we schedule the work. Requested after-the-fact by Chris on irc since the hpd work isn't the only place we botch this.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-02 21:22:25 +07:00
/**
* wq - Driver workqueue for GEM.
*
* NOTE: Work items scheduled here are not allowed to grab any modeset
* locks, for otherwise the flushing done in the pageflip code will
* result in deadlocks.
*/
struct workqueue_struct *wq;
/* Display functions */
struct drm_i915_display_funcs display;
/* PCH chipset type */
enum intel_pch pch_type;
unsigned short pch_id;
unsigned long quirks;
i915: ignore lid open event when resuming i915 driver needs to do modeset when 1. system resumes from sleep 2. lid is opened In PM_SUSPEND_MEM state, all the GPEs are cleared when system resumes, thus it is the i915_resume code does the modeset rather than intel_lid_notify(). But in PM_SUSPEND_FREEZE state, this will be broken because system is still responsive to the lid events. 1. When we close the lid in Freeze state, intel_lid_notify() sets modeset_on_lid. 2. When we reopen the lid, intel_lid_notify() will do a modeset, before the system is resumed. here is the error log, [92146.548074] WARNING: at drivers/gpu/drm/i915/intel_display.c:1028 intel_wait_for_pipe_off+0x184/0x190 [i915]() [92146.548076] Hardware name: VGN-Z540N [92146.548078] pipe_off wait timed out [92146.548167] Modules linked in: hid_generic usbhid hid snd_hda_codec_realtek snd_hda_intel snd_hda_codec parport_pc snd_hwdep ppdev snd_pcm_oss i915 snd_mixer_oss snd_pcm arc4 iwldvm snd_seq_dummy mac80211 snd_seq_oss snd_seq_midi fbcon tileblit font bitblit softcursor drm_kms_helper snd_rawmidi snd_seq_midi_event coretemp drm snd_seq kvm btusb bluetooth snd_timer iwlwifi pcmcia tpm_infineon i2c_algo_bit joydev snd_seq_device intel_agp cfg80211 snd intel_gtt yenta_socket pcmcia_rsrc sony_laptop agpgart microcode psmouse tpm_tis serio_raw mxm_wmi soundcore snd_page_alloc tpm acpi_cpufreq lpc_ich pcmcia_core tpm_bios mperf processor lp parport firewire_ohci firewire_core crc_itu_t sdhci_pci sdhci thermal e1000e [92146.548173] Pid: 4304, comm: kworker/0:0 Tainted: G W 3.8.0-rc3-s0i3-v3-test+ #9 [92146.548175] Call Trace: [92146.548189] [<c10378e2>] warn_slowpath_common+0x72/0xa0 [92146.548227] [<f86398b4>] ? intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548263] [<f86398b4>] ? intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548270] [<c10379b3>] warn_slowpath_fmt+0x33/0x40 [92146.548307] [<f86398b4>] intel_wait_for_pipe_off+0x184/0x190 [i915] [92146.548344] [<f86399c2>] intel_disable_pipe+0x102/0x190 [i915] [92146.548380] [<f8639ea4>] ? intel_disable_plane+0x64/0x80 [i915] [92146.548417] [<f8639f7c>] i9xx_crtc_disable+0xbc/0x150 [i915] [92146.548456] [<f863ebee>] intel_crtc_update_dpms+0x5e/0x90 [i915] [92146.548493] [<f86437cf>] intel_modeset_setup_hw_state+0x42f/0x8f0 [i915] [92146.548535] [<f8645b0b>] intel_lid_notify+0x9b/0xc0 [i915] [92146.548543] [<c15610d3>] notifier_call_chain+0x43/0x60 [92146.548550] [<c105d1e1>] __blocking_notifier_call_chain+0x41/0x80 [92146.548556] [<c105d23f>] blocking_notifier_call_chain+0x1f/0x30 [92146.548563] [<c131a684>] acpi_lid_send_state+0x78/0xa4 [92146.548569] [<c131aa9e>] acpi_button_notify+0x3b/0xf1 [92146.548577] [<c12df56a>] ? acpi_os_execute+0x17/0x19 [92146.548582] [<c12e591a>] ? acpi_ec_sync_query+0xa5/0xbc [92146.548589] [<c12e2b82>] acpi_device_notify+0x16/0x18 [92146.548595] [<c12f4904>] acpi_ev_notify_dispatch+0x38/0x4f [92146.548600] [<c12df0e8>] acpi_os_execute_deferred+0x20/0x2b [92146.548607] [<c1051208>] process_one_work+0x128/0x3f0 [92146.548613] [<c1564f73>] ? common_interrupt+0x33/0x38 [92146.548618] [<c104f8c0>] ? wake_up_worker+0x30/0x30 [92146.548624] [<c12df0c8>] ? acpi_os_wait_events_complete+0x1e/0x1e [92146.548629] [<c10524f9>] worker_thread+0x119/0x3b0 [92146.548634] [<c10523e0>] ? manage_workers+0x240/0x240 [92146.548640] [<c1056e84>] kthread+0x94/0xa0 [92146.548647] [<c1060000>] ? ftrace_raw_output_sched_stat_runtime+0x70/0xf0 [92146.548652] [<c15649b7>] ret_from_kernel_thread+0x1b/0x28 [92146.548658] [<c1056df0>] ? kthread_create_on_node+0xc0/0xc0 three different modeset flags are introduced in this patch MODESET_ON_LID_OPEN: do modeset on next lid open event MODESET_DONE: modeset already done MODESET_SUSPENDED: suspended, only do modeset when system is resumed In this way, 1. when lid is closed, MODESET_ON_LID_OPEN is set so that we'll do modeset on next lid open event. 2. when lid is opened, MODESET_DONE is set so that duplicate lid open events will be ignored. 3. when system suspends, MODESET_SUSPENDED is set. In this case, we will not do modeset on any lid events. Plus, locking mechanism is also introduced to avoid racing. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-02-05 14:41:53 +07:00
enum modeset_restore modeset_restore;
struct mutex modeset_restore_lock;
struct list_head vm_list; /* Global list of all address spaces */
struct i915_gtt gtt; /* VMA representing the global address space */
struct i915_gem_mm mm;
/* Kernel Modesetting */
struct sdvo_device_mapping sdvo_mappings[2];
struct drm_crtc *plane_to_crtc_mapping[I915_MAX_PIPES];
struct drm_crtc *pipe_to_crtc_mapping[I915_MAX_PIPES];
wait_queue_head_t pending_flip_queue;
#ifdef CONFIG_DEBUG_FS
struct intel_pipe_crc pipe_crc[I915_MAX_PIPES];
#endif
int num_shared_dpll;
struct intel_shared_dpll shared_dplls[I915_NUM_PLLS];
struct intel_ddi_plls ddi_plls;
int dpio_phy_iosf_port[I915_NUM_PHYS_VLV];
/* Reclocking support */
bool render_reclock_avail;
bool lvds_downclock_avail;
/* indicates the reduced downclock for LVDS*/
int lvds_downclock;
u16 orig_clock;
bool mchbar_need_disable;
struct intel_l3_parity l3_parity;
/* Cannot be determined by PCIID. You must always read a register. */
size_t ellc_size;
/* gen6+ rps state */
struct intel_gen6_power_mgmt rps;
/* ilk-only ips/rps state. Everything in here is protected by the global
* mchdev_lock in intel_pm.c */
struct intel_ilk_power_mgmt ips;
struct i915_power_domains power_domains;
struct i915_psr psr;
struct i915_gpu_error gpu_error;
struct drm_i915_gem_object *vlv_pctx;
#ifdef CONFIG_DRM_I915_FBDEV
/* list of fbdev register on this device */
struct intel_fbdev *fbdev;
#endif
/*
* The console may be contended at resume, but we don't
* want it to block on it.
*/
struct work_struct console_resume_work;
struct drm_property *broadcast_rgb_property;
struct drm_property *force_audio_property;
drm/i915: preliminary context support Very basic code for context setup/destruction in the driver. Adds the file i915_gem_context.c This file implements HW context support. On gen5+ a HW context consists of an opaque GPU object which is referenced at times of context saves and restores. With RC6 enabled, the context is also referenced as the GPU enters and exists from RC6 (GPU has it's own internal power context, except on gen5). Though something like a context does exist for the media ring, the code only supports contexts for the render ring. In software, there is a distinction between contexts created by the user, and the default HW context. The default HW context is used by GPU clients that do not request setup of their own hardware context. The default context's state is never restored to help prevent programming errors. This would happen if a client ran and piggy-backed off another clients GPU state. The default context only exists to give the GPU some offset to load as the current to invoke a save of the context we actually care about. In fact, the code could likely be constructed, albeit in a more complicated fashion, to never use the default context, though that limits the driver's ability to swap out, and/or destroy other contexts. All other contexts are created as a request by the GPU client. These contexts store GPU state, and thus allow GPU clients to not re-emit state (and potentially query certain state) at any time. The kernel driver makes certain that the appropriate commands are inserted. There are 4 entry points into the contexts, init, fini, open, close. The names are self-explanatory except that init can be called during reset, and also during pm thaw/resume. As we expect our context to be preserved across these events, we do not reinitialize in this case. As Adam Jackson pointed out, The cutoff of 1MB where a HW context is considered too big is arbitrary. The reason for this is even though context sizes are increasing with every generation, they have yet to eclipse even 32k. If we somehow read back way more than that, it probably means BIOS has done something strange, or we're running on a platform that wasn't designed for this. v2: rename load/unload to init/fini (daniel) remove ILK support for get_size() (indirectly daniel) add HAS_HW_CONTEXTS macro to clarify supported platforms (daniel) added comments (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
2012-06-05 04:42:42 +07:00
uint32_t hw_context_size;
struct list_head context_list;
u32 fdi_rx_config;
struct i915_suspend_saved_registers regfile;
struct {
/*
* Raw watermark latency values:
* in 0.1us units for WM0,
* in 0.5us units for WM1+.
*/
/* primary */
uint16_t pri_latency[5];
/* sprite */
uint16_t spr_latency[5];
/* cursor */
uint16_t cur_latency[5];
/* current hardware state */
struct ilk_wm_values hw;
} wm;
drm/i915: allow package C8+ states on Haswell (disabled) This patch allows PC8+ states on Haswell. These states can only be reached when all the display outputs are disabled, and they allow some more power savings. The fact that the graphics device is allowing PC8+ doesn't mean that the machine will actually enter PC8+: all the other devices also need to allow PC8+. For now this option is disabled by default. You need i915.allow_pc8=1 if you want it. This patch adds a big comment inside i915_drv.h explaining how it works and how it tracks things. Read it. v2: (this is not really v2, many previous versions were already sent, but they had different names) - Use the new functions to enable/disable GTIMR and GEN6_PMIMR - Rename almost all variables and functions to names suggested by Chris - More WARNs on the IRQ handling code - Also disable PC8 when there's GPU work to do (thanks to Ben for the help on this), so apps can run caster - Enable PC8 on a delayed work function that is delayed for 5 seconds. This makes sure we only enable PC8+ if we're really idle - Make sure we're not in PC8+ when suspending v3: - WARN if IRQs are disabled on __wait_seqno - Replace some DRM_ERRORs with WARNs - Fix calls to restore GT and PM interrupts - Use intel_mark_busy instead of intel_ring_advance to disable PC8 v4: - Use the force_wake, Luke! v5: - Remove the "IIR is not zero" WARNs - Move the force_wake chunk to its own patch - Only restore what's missing from RC6, not everything Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-19 23:18:09 +07:00
struct i915_package_c8 pc8;
struct i915_runtime_pm pm;
/* Old dri1 support infrastructure, beware the dragons ya fools entering
* here! */
struct i915_dri1_state dri1;
/* Old ums support infrastructure, same warning applies. */
struct i915_ums_state ums;
} drm_i915_private_t;
static inline struct drm_i915_private *to_i915(const struct drm_device *dev)
{
return dev->dev_private;
}
/* Iterate over initialised rings */
#define for_each_ring(ring__, dev_priv__, i__) \
for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
if (((ring__) = &(dev_priv__)->ring[(i__)]), intel_ring_initialized((ring__)))
enum hdmi_force_audio {
HDMI_AUDIO_OFF_DVI = -2, /* no aux data for HDMI-DVI converter */
HDMI_AUDIO_OFF, /* force turn off HDMI audio */
HDMI_AUDIO_AUTO, /* trust EDID */
HDMI_AUDIO_ON, /* force turn on HDMI audio */
};
#define I915_GTT_OFFSET_NONE ((u32)-1)
struct drm_i915_gem_object_ops {
/* Interface between the GEM object and its backing storage.
* get_pages() is called once prior to the use of the associated set
* of pages before to binding them into the GTT, and put_pages() is
* called after we no longer need them. As we expect there to be
* associated cost with migrating pages between the backing storage
* and making them available for the GPU (e.g. clflush), we may hold
* onto the pages after they are no longer referenced by the GPU
* in case they may be used again shortly (for example migrating the
* pages to a different memory domain within the GTT). put_pages()
* will therefore most likely be called when the object itself is
* being released or under memory pressure (where we attempt to
* reap pages for the shrinker).
*/
int (*get_pages)(struct drm_i915_gem_object *);
void (*put_pages)(struct drm_i915_gem_object *);
};
struct drm_i915_gem_object {
struct drm_gem_object base;
const struct drm_i915_gem_object_ops *ops;
/** List of VMAs backed by this object */
struct list_head vma_list;
/** Stolen memory for this object, instead of being backed by shmem. */
struct drm_mm_node *stolen;
struct list_head global_list;
struct list_head ring_list;
/** Used in execbuf to temporarily hold a ref */
struct list_head obj_exec_link;
/**
* This is set if the object is on the active lists (has pending
* rendering and so a non-zero seqno), and is not set if it i s on
* inactive (ready to be unbound) list.
*/
unsigned int active:1;
/**
* This is set if the object has been written to since last bound
* to the GTT
*/
unsigned int dirty:1;
/**
* Fence register bits (if any) for this object. Will be set
* as needed when mapped into the GTT.
* Protected by dev->struct_mutex.
*/
signed int fence_reg:I915_MAX_NUM_FENCE_BITS;
/**
* Advice: are the backing pages purgeable?
*/
unsigned int madv:2;
/**
* Current tiling mode for the object.
*/
unsigned int tiling_mode:2;
/**
* Whether the tiling parameters for the currently associated fence
* register have changed. Note that for the purposes of tracking
* tiling changes we also treat the unfenced register, the register
* slot that the object occupies whilst it executes a fenced
* command (such as BLT on gen2/3), as a "fence".
*/
unsigned int fence_dirty:1;
/**
* Is the object at the current location in the gtt mappable and
* fenceable? Used to avoid costly recalculations.
*/
unsigned int map_and_fenceable:1;
/**
* Whether the current gtt mapping needs to be mappable (and isn't just
* mappable by accident). Track pin and fault separate for a more
* accurate mappable working set.
*/
unsigned int fault_mappable:1;
unsigned int pin_mappable:1;
unsigned int pin_display:1;
/*
* Is the GPU currently using a fence to access this buffer,
*/
unsigned int pending_fenced_gpu_access:1;
unsigned int fenced_gpu_access:1;
unsigned int cache_level:3;
unsigned int has_aliasing_ppgtt_mapping:1;
unsigned int has_global_gtt_mapping:1;
unsigned int has_dma_mapping:1;
struct sg_table *pages;
int pages_pin_count;
i915: add dmabuf/prime buffer sharing support. This adds handle->fd and fd->handle support to i915, this is to allow for offloading of rendering in one direction and outputs in the other. v2 from Daniel Vetter: - fixup conflicts with the prepare/finish gtt prep work. - implement ppgtt binding support. Note that we have squat i-g-t testcoverage for any of the lifetime and access rules dma_buf/prime support brings along. And there are quite a few intricate situations here. Also note that the integration with the existing code is a bit hackish, especially around get_gtt_pages and put_gtt_pages. It imo would be easier with the prep code from Chris Wilson's unbound series, but that is for 3.6. Also note that I didn't bother to put the new prepare/finish gtt hooks to good use by moving the dma_buf_map/unmap_attachment calls in there (like we've originally planned for). Last but not least this patch is only compile-tested, but I've changed very little compared to Dave Airlie's version. So there's a decent chance v2 on drm-next works as well as v1 on 3.4-rc. v3: Right when I've hit sent I've noticed that I've screwed up one obj->sg_list (for dmar support) and obj->sg_table (for prime support) disdinction. We should be able to merge these 2 paths, but that's material for another patch. v4: fix the error reporting bugs pointed out by ickle. v5: fix another error, and stop non-gtt mmaps on shared objects stop pread/pwrite on imported objects, add fake kmap Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-05-10 20:25:09 +07:00
/* prime dma-buf support */
void *dma_buf_vmapping;
int vmapping_count;
struct intel_ring_buffer *ring;
/** Breadcrumb of last rendering to the buffer. */
uint32_t last_read_seqno;
uint32_t last_write_seqno;
/** Breadcrumb of last fenced GPU access to the buffer. */
uint32_t last_fenced_seqno;
/** Current tiling stride for the object, if it's tiled. */
uint32_t stride;
/** References from framebuffers, locks out tiling changes. */
unsigned long framebuffer_references;
/** Record of address bit 17 of each page at last unbind. */
unsigned long *bit_17;
/** User space pin count and filp owning the pin */
unsigned long user_pin_count;
struct drm_file *pin_filp;
/** for phy allocated objects */
struct drm_i915_gem_phys_object *phys_obj;
};
#define to_gem_object(obj) (&((struct drm_i915_gem_object *)(obj))->base)
#define to_intel_bo(x) container_of(x, struct drm_i915_gem_object, base)
/**
* Request queue structure.
*
* The request queue allows us to note sequence numbers that have been emitted
* and may be associated with active buffers to be retired.
*
* By keeping this list, we can avoid having to do questionable
* sequence-number comparisons on buffer last_rendering_seqnos, and associate
* an emission time with seqnos for tracking how far ahead of the GPU we are.
*/
struct drm_i915_gem_request {
/** On Which ring this request was generated */
struct intel_ring_buffer *ring;
/** GEM sequence number associated with this request. */
uint32_t seqno;
/** Position in the ringbuffer of the start of the request */
u32 head;
/** Position in the ringbuffer of the end of the request */
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 18:25:36 +07:00
u32 tail;
/** Context related to this request */
struct i915_hw_context *ctx;
/** Batch buffer related to this request if any */
struct drm_i915_gem_object *batch_obj;
/** Time at which this request was emitted, in jiffies. */
unsigned long emitted_jiffies;
/** global list entry for this request */
struct list_head list;
struct drm_i915_file_private *file_priv;
/** file_priv list entry for this request */
struct list_head client_list;
};
struct drm_i915_file_private {
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
struct drm_i915_private *dev_priv;
struct {
i915: convert struct spinlock to spinlock_t spinlock_t should always be used. LD drivers/gpu/drm/i915/built-in.o CHECK drivers/gpu/drm/i915/i915_drv.c CC [M] drivers/gpu/drm/i915/i915_drv.o CHECK drivers/gpu/drm/i915/i915_dma.c CC [M] drivers/gpu/drm/i915/i915_dma.o CHECK drivers/gpu/drm/i915/i915_irq.c CC [M] drivers/gpu/drm/i915/i915_irq.o CHECK drivers/gpu/drm/i915/i915_debugfs.c drivers/gpu/drm/i915/i915_debugfs.c:558:31: warning: dereference of noderef expression drivers/gpu/drm/i915/i915_debugfs.c:558:39: warning: dereference of noderef expression drivers/gpu/drm/i915/i915_debugfs.c:558:51: warning: dereference of noderef expression drivers/gpu/drm/i915/i915_debugfs.c:558:63: warning: dereference of noderef expression CC [M] drivers/gpu/drm/i915/i915_debugfs.o CHECK drivers/gpu/drm/i915/i915_suspend.c CC [M] drivers/gpu/drm/i915/i915_suspend.o CHECK drivers/gpu/drm/i915/i915_gem.c drivers/gpu/drm/i915/i915_gem.c:3703:14: warning: incorrect type in assignment (different base types) drivers/gpu/drm/i915/i915_gem.c:3703:14: expected unsigned int [unsigned] [usertype] mask drivers/gpu/drm/i915/i915_gem.c:3703:14: got restricted gfp_t drivers/gpu/drm/i915/i915_gem.c:3706:22: warning: invalid assignment: &= drivers/gpu/drm/i915/i915_gem.c:3706:22: left side has type unsigned int drivers/gpu/drm/i915/i915_gem.c:3706:22: right side has type restricted gfp_t drivers/gpu/drm/i915/i915_gem.c:3707:22: warning: invalid assignment: |= drivers/gpu/drm/i915/i915_gem.c:3707:22: left side has type unsigned int drivers/gpu/drm/i915/i915_gem.c:3707:22: right side has type restricted gfp_t drivers/gpu/drm/i915/i915_gem.c:3711:39: warning: incorrect type in argument 2 (different base types) drivers/gpu/drm/i915/i915_gem.c:3711:39: expected restricted gfp_t [usertype] mask drivers/gpu/drm/i915/i915_gem.c:3711:39: got unsigned int [unsigned] [usertype] mask CC [M] drivers/gpu/drm/i915/i915_gem.o CHECK drivers/gpu/drm/i915/i915_gem_context.c CC [M] drivers/gpu/drm/i915/i915_gem_context.o CHECK drivers/gpu/drm/i915/i915_gem_debug.c CC [M] drivers/gpu/drm/i915/i915_gem_debug.o CHECK drivers/gpu/drm/i915/i915_gem_evict.c CC [M] drivers/gpu/drm/i915/i915_gem_evict.o CHECK drivers/gpu/drm/i915/i915_gem_execbuffer.c CC [M] drivers/gpu/drm/i915/i915_gem_execbuffer.o CHECK drivers/gpu/drm/i915/i915_gem_gtt.c CC [M] drivers/gpu/drm/i915/i915_gem_gtt.o CHECK drivers/gpu/drm/i915/i915_gem_stolen.c CC [M] drivers/gpu/drm/i915/i915_gem_stolen.o CHECK drivers/gpu/drm/i915/i915_gem_tiling.c CC [M] drivers/gpu/drm/i915/i915_gem_tiling.o CHECK drivers/gpu/drm/i915/i915_sysfs.c CC [M] drivers/gpu/drm/i915/i915_sysfs.o CHECK drivers/gpu/drm/i915/i915_trace_points.c CC [M] drivers/gpu/drm/i915/i915_trace_points.o CHECK drivers/gpu/drm/i915/intel_display.c drivers/gpu/drm/i915/intel_display.c:1736:9: warning: mixing different enum types drivers/gpu/drm/i915/intel_display.c:1736:9: int enum transcoder versus drivers/gpu/drm/i915/intel_display.c:1736:9: int enum pipe drivers/gpu/drm/i915/intel_display.c:3659:48: warning: mixing different enum types drivers/gpu/drm/i915/intel_display.c:3659:48: int enum pipe versus drivers/gpu/drm/i915/intel_display.c:3659:48: int enum transcoder CC [M] drivers/gpu/drm/i915/intel_display.o CHECK drivers/gpu/drm/i915/intel_crt.c CC [M] drivers/gpu/drm/i915/intel_crt.o CHECK drivers/gpu/drm/i915/intel_lvds.c CC [M] drivers/gpu/drm/i915/intel_lvds.o CHECK drivers/gpu/drm/i915/intel_bios.c drivers/gpu/drm/i915/intel_bios.c:706:60: warning: incorrect type in initializer (different address spaces) drivers/gpu/drm/i915/intel_bios.c:706:60: expected struct vbt_header *vbt drivers/gpu/drm/i915/intel_bios.c:706:60: got void [noderef] <asn:2>*vbt drivers/gpu/drm/i915/intel_bios.c:726:42: warning: incorrect type in argument 1 (different address spaces) drivers/gpu/drm/i915/intel_bios.c:726:42: expected void const *<noident> drivers/gpu/drm/i915/intel_bios.c:726:42: got unsigned char [noderef] [usertype] <asn:2>* drivers/gpu/drm/i915/intel_bios.c:727:40: warning: cast removes address space of expression drivers/gpu/drm/i915/intel_bios.c:738:24: warning: cast removes address space of expression CC [M] drivers/gpu/drm/i915/intel_bios.o CHECK drivers/gpu/drm/i915/intel_ddi.c drivers/gpu/drm/i915/intel_ddi.c:87:6: warning: symbol 'intel_prepare_ddi_buffers' was not declared. Should it be static? drivers/gpu/drm/i915/intel_ddi.c:1036:34: warning: mixing different enum types drivers/gpu/drm/i915/intel_ddi.c:1036:34: int enum pipe versus drivers/gpu/drm/i915/intel_ddi.c:1036:34: int enum transcoder CC [M] drivers/gpu/drm/i915/intel_ddi.o drivers/gpu/drm/i915/intel_ddi.c: In function ‘intel_ddi_setup_hw_pll_state’: drivers/gpu/drm/i915/intel_ddi.c:1129:2: warning: ‘port’ may be used uninitialized in this function [-Wmaybe-uninitialized] drivers/gpu/drm/i915/intel_ddi.c:1111:12: note: ‘port’ was declared here CHECK drivers/gpu/drm/i915/intel_dp.c CC [M] drivers/gpu/drm/i915/intel_dp.o CHECK drivers/gpu/drm/i915/intel_hdmi.c CC [M] drivers/gpu/drm/i915/intel_hdmi.o CHECK drivers/gpu/drm/i915/intel_sdvo.c CC [M] drivers/gpu/drm/i915/intel_sdvo.o CHECK drivers/gpu/drm/i915/intel_modes.c CC [M] drivers/gpu/drm/i915/intel_modes.o CHECK drivers/gpu/drm/i915/intel_panel.c CC [M] drivers/gpu/drm/i915/intel_panel.o CHECK drivers/gpu/drm/i915/intel_pm.c drivers/gpu/drm/i915/intel_pm.c:2173:1: warning: symbol 'mchdev_lock' was not declared. Should it be static? CC [M] drivers/gpu/drm/i915/intel_pm.o CHECK drivers/gpu/drm/i915/intel_i2c.c CC [M] drivers/gpu/drm/i915/intel_i2c.o CHECK drivers/gpu/drm/i915/intel_fb.c CC [M] drivers/gpu/drm/i915/intel_fb.o CHECK drivers/gpu/drm/i915/intel_tv.c CC [M] drivers/gpu/drm/i915/intel_tv.o CHECK drivers/gpu/drm/i915/intel_dvo.c CC [M] drivers/gpu/drm/i915/intel_dvo.o CHECK drivers/gpu/drm/i915/intel_ringbuffer.c CC [M] drivers/gpu/drm/i915/intel_ringbuffer.o CHECK drivers/gpu/drm/i915/intel_overlay.c CC [M] drivers/gpu/drm/i915/intel_overlay.o CHECK drivers/gpu/drm/i915/intel_sprite.c CC [M] drivers/gpu/drm/i915/intel_sprite.o CHECK drivers/gpu/drm/i915/intel_opregion.c CC [M] drivers/gpu/drm/i915/intel_opregion.o CHECK drivers/gpu/drm/i915/dvo_ch7xxx.c CC [M] drivers/gpu/drm/i915/dvo_ch7xxx.o CHECK drivers/gpu/drm/i915/dvo_ch7017.c CC [M] drivers/gpu/drm/i915/dvo_ch7017.o CHECK drivers/gpu/drm/i915/dvo_ivch.c CC [M] drivers/gpu/drm/i915/dvo_ivch.o CHECK drivers/gpu/drm/i915/dvo_tfp410.c CC [M] drivers/gpu/drm/i915/dvo_tfp410.o CHECK drivers/gpu/drm/i915/dvo_sil164.c CC [M] drivers/gpu/drm/i915/dvo_sil164.o CHECK drivers/gpu/drm/i915/dvo_ns2501.c CC [M] drivers/gpu/drm/i915/dvo_ns2501.o CHECK drivers/gpu/drm/i915/i915_gem_dmabuf.c CC [M] drivers/gpu/drm/i915/i915_gem_dmabuf.o CHECK drivers/gpu/drm/i915/i915_ioc32.c CC [M] drivers/gpu/drm/i915/i915_ioc32.o CHECK drivers/gpu/drm/i915/intel_acpi.c CC [M] drivers/gpu/drm/i915/intel_acpi.o LD [M] drivers/gpu/drm/i915/i915.o Building modules, stage 2. MODPOST 1 modules CC drivers/gpu/drm/i915/i915.mod.o LD [M] drivers/gpu/drm/i915/i915.ko Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: intel-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Reported-by: Hauke Mehrtens <hauke@hauke-m.de> Signed-off-by: Luis R. Rodriguez <mcgrof@do-not-panic.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-30 03:45:06 +07:00
spinlock_t lock;
struct list_head request_list;
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
struct delayed_work idle_work;
} mm;
struct idr context_idr;
struct i915_hw_context *private_default_ctx;
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
atomic_t rps_wait_boost;
};
#define INTEL_INFO(dev) (&to_i915(dev)->info)
#define IS_I830(dev) ((dev)->pdev->device == 0x3577)
#define IS_845G(dev) ((dev)->pdev->device == 0x2562)
#define IS_I85X(dev) (INTEL_INFO(dev)->is_i85x)
#define IS_I865G(dev) ((dev)->pdev->device == 0x2572)
#define IS_I915G(dev) (INTEL_INFO(dev)->is_i915g)
#define IS_I915GM(dev) ((dev)->pdev->device == 0x2592)
#define IS_I945G(dev) ((dev)->pdev->device == 0x2772)
#define IS_I945GM(dev) (INTEL_INFO(dev)->is_i945gm)
#define IS_BROADWATER(dev) (INTEL_INFO(dev)->is_broadwater)
#define IS_CRESTLINE(dev) (INTEL_INFO(dev)->is_crestline)
#define IS_GM45(dev) ((dev)->pdev->device == 0x2A42)
#define IS_G4X(dev) (INTEL_INFO(dev)->is_g4x)
#define IS_PINEVIEW_G(dev) ((dev)->pdev->device == 0xa001)
#define IS_PINEVIEW_M(dev) ((dev)->pdev->device == 0xa011)
#define IS_PINEVIEW(dev) (INTEL_INFO(dev)->is_pineview)
#define IS_G33(dev) (INTEL_INFO(dev)->is_g33)
#define IS_IRONLAKE_M(dev) ((dev)->pdev->device == 0x0046)
#define IS_IVYBRIDGE(dev) (INTEL_INFO(dev)->is_ivybridge)
#define IS_IVB_GT1(dev) ((dev)->pdev->device == 0x0156 || \
(dev)->pdev->device == 0x0152 || \
(dev)->pdev->device == 0x015a)
#define IS_SNB_GT1(dev) ((dev)->pdev->device == 0x0102 || \
(dev)->pdev->device == 0x0106 || \
(dev)->pdev->device == 0x010A)
#define IS_VALLEYVIEW(dev) (INTEL_INFO(dev)->is_valleyview)
#define IS_HASWELL(dev) (INTEL_INFO(dev)->is_haswell)
#define IS_BROADWELL(dev) (INTEL_INFO(dev)->gen == 8)
#define IS_MOBILE(dev) (INTEL_INFO(dev)->is_mobile)
#define IS_HSW_EARLY_SDV(dev) (IS_HASWELL(dev) && \
((dev)->pdev->device & 0xFF00) == 0x0C00)
#define IS_BDW_ULT(dev) (IS_BROADWELL(dev) && \
(((dev)->pdev->device & 0xf) == 0x2 || \
((dev)->pdev->device & 0xf) == 0x6 || \
((dev)->pdev->device & 0xf) == 0xe))
#define IS_HSW_ULT(dev) (IS_HASWELL(dev) && \
((dev)->pdev->device & 0xFF00) == 0x0A00)
#define IS_ULT(dev) (IS_HSW_ULT(dev) || IS_BDW_ULT(dev))
#define IS_HSW_GT3(dev) (IS_HASWELL(dev) && \
((dev)->pdev->device & 0x00F0) == 0x0020)
#define IS_PRELIMINARY_HW(intel_info) ((intel_info)->is_preliminary)
/*
* The genX designation typically refers to the render engine, so render
* capability related checks should use IS_GEN, while display and other checks
* have their own (e.g. HAS_PCH_SPLIT for ILK+ display, IS_foo for particular
* chips, etc.).
*/
#define IS_GEN2(dev) (INTEL_INFO(dev)->gen == 2)
#define IS_GEN3(dev) (INTEL_INFO(dev)->gen == 3)
#define IS_GEN4(dev) (INTEL_INFO(dev)->gen == 4)
#define IS_GEN5(dev) (INTEL_INFO(dev)->gen == 5)
#define IS_GEN6(dev) (INTEL_INFO(dev)->gen == 6)
#define IS_GEN7(dev) (INTEL_INFO(dev)->gen == 7)
#define IS_GEN8(dev) (INTEL_INFO(dev)->gen == 8)
#define RENDER_RING (1<<RCS)
#define BSD_RING (1<<VCS)
#define BLT_RING (1<<BCS)
#define VEBOX_RING (1<<VECS)
#define HAS_BSD(dev) (INTEL_INFO(dev)->ring_mask & BSD_RING)
#define HAS_BLT(dev) (INTEL_INFO(dev)->ring_mask & BLT_RING)
#define HAS_VEBOX(dev) (INTEL_INFO(dev)->ring_mask & VEBOX_RING)
#define HAS_LLC(dev) (INTEL_INFO(dev)->has_llc)
#define HAS_WT(dev) (IS_HASWELL(dev) && to_i915(dev)->ellc_size)
#define I915_NEED_GFX_HWS(dev) (INTEL_INFO(dev)->need_gfx_hws)
drm/i915: preliminary context support Very basic code for context setup/destruction in the driver. Adds the file i915_gem_context.c This file implements HW context support. On gen5+ a HW context consists of an opaque GPU object which is referenced at times of context saves and restores. With RC6 enabled, the context is also referenced as the GPU enters and exists from RC6 (GPU has it's own internal power context, except on gen5). Though something like a context does exist for the media ring, the code only supports contexts for the render ring. In software, there is a distinction between contexts created by the user, and the default HW context. The default HW context is used by GPU clients that do not request setup of their own hardware context. The default context's state is never restored to help prevent programming errors. This would happen if a client ran and piggy-backed off another clients GPU state. The default context only exists to give the GPU some offset to load as the current to invoke a save of the context we actually care about. In fact, the code could likely be constructed, albeit in a more complicated fashion, to never use the default context, though that limits the driver's ability to swap out, and/or destroy other contexts. All other contexts are created as a request by the GPU client. These contexts store GPU state, and thus allow GPU clients to not re-emit state (and potentially query certain state) at any time. The kernel driver makes certain that the appropriate commands are inserted. There are 4 entry points into the contexts, init, fini, open, close. The names are self-explanatory except that init can be called during reset, and also during pm thaw/resume. As we expect our context to be preserved across these events, we do not reinitialize in this case. As Adam Jackson pointed out, The cutoff of 1MB where a HW context is considered too big is arbitrary. The reason for this is even though context sizes are increasing with every generation, they have yet to eclipse even 32k. If we somehow read back way more than that, it probably means BIOS has done something strange, or we're running on a platform that wasn't designed for this. v2: rename load/unload to init/fini (daniel) remove ILK support for get_size() (indirectly daniel) add HAS_HW_CONTEXTS macro to clarify supported platforms (daniel) added comments (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
2012-06-05 04:42:42 +07:00
#define HAS_HW_CONTEXTS(dev) (INTEL_INFO(dev)->gen >= 6)
#define HAS_ALIASING_PPGTT(dev) (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
#define HAS_PPGTT(dev) (INTEL_INFO(dev)->gen >= 7 && !IS_VALLEYVIEW(dev) \
&& !IS_BROADWELL(dev))
#define USES_PPGTT(dev) intel_enable_ppgtt(dev, false)
#define USES_FULL_PPGTT(dev) intel_enable_ppgtt(dev, true)
#define HAS_OVERLAY(dev) (INTEL_INFO(dev)->has_overlay)
#define OVERLAY_NEEDS_PHYSICAL(dev) (INTEL_INFO(dev)->overlay_needs_physical)
/* Early gen2 have a totally busted CS tlb and require pinned batches. */
#define HAS_BROKEN_CS_TLB(dev) (IS_I830(dev) || IS_845G(dev))
/* With the 945 and later, Y tiling got adjusted so that it was 32 128-byte
* rows, which changed the alignment requirements and fence programming.
*/
#define HAS_128_BYTE_Y_TILING(dev) (!IS_GEN2(dev) && !(IS_I915G(dev) || \
IS_I915GM(dev)))
#define SUPPORTS_DIGITAL_OUTPUTS(dev) (!IS_GEN2(dev) && !IS_PINEVIEW(dev))
#define SUPPORTS_INTEGRATED_HDMI(dev) (IS_G4X(dev) || IS_GEN5(dev))
#define SUPPORTS_INTEGRATED_DP(dev) (IS_G4X(dev) || IS_GEN5(dev))
#define SUPPORTS_TV(dev) (INTEL_INFO(dev)->supports_tv)
#define I915_HAS_HOTPLUG(dev) (INTEL_INFO(dev)->has_hotplug)
#define HAS_FW_BLC(dev) (INTEL_INFO(dev)->gen > 2)
#define HAS_PIPE_CXSR(dev) (INTEL_INFO(dev)->has_pipe_cxsr)
#define HAS_FBC(dev) (INTEL_INFO(dev)->has_fbc)
#define HAS_IPS(dev) (IS_ULT(dev) || IS_BROADWELL(dev))
#define HAS_DDI(dev) (INTEL_INFO(dev)->has_ddi)
#define HAS_FPGA_DBG_UNCLAIMED(dev) (INTEL_INFO(dev)->has_fpga_dbg)
#define HAS_PSR(dev) (IS_HASWELL(dev) || IS_BROADWELL(dev))
#define HAS_PC8(dev) (IS_HASWELL(dev)) /* XXX HSW:ULX */
#define HAS_RUNTIME_PM(dev) (IS_HASWELL(dev))
#define INTEL_PCH_DEVICE_ID_MASK 0xff00
#define INTEL_PCH_IBX_DEVICE_ID_TYPE 0x3b00
#define INTEL_PCH_CPT_DEVICE_ID_TYPE 0x1c00
#define INTEL_PCH_PPT_DEVICE_ID_TYPE 0x1e00
#define INTEL_PCH_LPT_DEVICE_ID_TYPE 0x8c00
#define INTEL_PCH_LPT_LP_DEVICE_ID_TYPE 0x9c00
#define INTEL_PCH_TYPE(dev) (to_i915(dev)->pch_type)
#define HAS_PCH_LPT(dev) (INTEL_PCH_TYPE(dev) == PCH_LPT)
#define HAS_PCH_CPT(dev) (INTEL_PCH_TYPE(dev) == PCH_CPT)
#define HAS_PCH_IBX(dev) (INTEL_PCH_TYPE(dev) == PCH_IBX)
#define HAS_PCH_NOP(dev) (INTEL_PCH_TYPE(dev) == PCH_NOP)
#define HAS_PCH_SPLIT(dev) (INTEL_PCH_TYPE(dev) != PCH_NONE)
/* DPF == dynamic parity feature */
#define HAS_L3_DPF(dev) (IS_IVYBRIDGE(dev) || IS_HASWELL(dev))
#define NUM_L3_SLICES(dev) (IS_HSW_GT3(dev) ? 2 : HAS_L3_DPF(dev))
#define GT_FREQUENCY_MULTIPLIER 50
#include "i915_trace.h"
extern const struct drm_ioctl_desc i915_ioctls[];
extern int i915_max_ioctl;
vga_switcheroo: initial implementation (v15) Many new laptops now come with 2 gpus, one to be used for low power modes and one for gaming/on-ac applications. These GPUs are typically wired to the laptop panel and VGA ports via a multiplexer unit which is controlled via ACPI methods. 4 combinations of systems typically exist - with 2 ACPI methods. Intel/ATI - Lenovo W500/T500 - use ATPX ACPI method ATI/ATI - some ASUS - use ATPX ACPI Method Intel/Nvidia - - use _DSM ACPI method Nvidia/Nvidia - - use _DSM ACPI method. TODO: This patch adds support for the ATPX method and initial bits for the _DSM methods that need to written by someone with access to the hardware. Add a proper non-debugfs interface - need to get some proper testing first. v2: add power up/down support for both devices on W500 puts i915/radeon into D3 and cuts power to radeon. v3: redo probing methods, no DMI list, drm devices call to register with switcheroo, it tries to find an ATPX method on any device and once there is two devices + ATPX it inits the switcher. v4: ATPX msg handling using buffers - should work on more machines v5: rearchitect after more mjg59 discussion - move ATPX handling to radeon driver. v6: add file headers + initial nouveau bits (to be filled out). v7: merge delayed switcher code. v8: avoid suspend/resume of gpu that is off v9: rearchitect - mjg59 is always right. - move all ATPX code to radeon, should allow simpler DSM also proper ATRM handling v10: add ATRM support for radeon BIOS, add mutex to lock vgasr_priv v11: fix bug in resuming Intel for 2nd time. v12: start fixing up nvidia code blindly. v13: blindly guess at finishing nvidia code v14: remove radeon audio hacks - fix up intel resume more like upstream v15: clean up printks + remove unnecessary igd/dis pointers mount debugfs /sys/kernel/debug/vgaswitcheroo/switch - should exist if ATPX detected + 2 cards. DIS - immediate change to discrete IGD - immediate change to IGD DDIS - delayed change to discrete DIGD - delayed change to IGD ON - turn on not in use OFF - turn off not in use Tested on W500 (Intel/ATI) and T500 (Intel/ATI) Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-01 12:38:10 +07:00
extern int i915_suspend(struct drm_device *dev, pm_message_t state);
extern int i915_resume(struct drm_device *dev);
extern int i915_master_create(struct drm_device *dev, struct drm_master *master);
extern void i915_master_destroy(struct drm_device *dev, struct drm_master *master);
/* i915_params.c */
struct i915_params {
int modeset;
int panel_ignore_lid;
unsigned int powersave;
int semaphores;
unsigned int lvds_downclock;
int lvds_channel_mode;
int panel_use_ssc;
int vbt_sdvo_panel_type;
int enable_rc6;
int enable_fbc;
int enable_ppgtt;
int enable_psr;
unsigned int preliminary_hw_support;
int disable_power_well;
int enable_ips;
int enable_pc8;
int pc8_timeout;
int invert_brightness;
/* leave bools at the end to not create holes */
bool enable_hangcheck;
bool fastboot;
bool prefault_disable;
bool reset;
bool disable_display;
};
extern struct i915_params i915 __read_mostly;
/* i915_dma.c */
void i915_update_dri1_breadcrumb(struct drm_device *dev);
extern void i915_kernel_lost_context(struct drm_device * dev);
extern int i915_driver_load(struct drm_device *, unsigned long flags);
extern int i915_driver_unload(struct drm_device *);
extern int i915_driver_open(struct drm_device *dev, struct drm_file *file_priv);
extern void i915_driver_lastclose(struct drm_device * dev);
extern void i915_driver_preclose(struct drm_device *dev,
struct drm_file *file_priv);
extern void i915_driver_postclose(struct drm_device *dev,
struct drm_file *file_priv);
extern int i915_driver_device_is_agp(struct drm_device * dev);
#ifdef CONFIG_COMPAT
extern long i915_compat_ioctl(struct file *filp, unsigned int cmd,
unsigned long arg);
#endif
extern int i915_emit_box(struct drm_device *dev,
struct drm_clip_rect *box,
int DR1, int DR4);
extern int intel_gpu_reset(struct drm_device *dev);
extern int i915_reset(struct drm_device *dev);
extern unsigned long i915_chipset_val(struct drm_i915_private *dev_priv);
extern unsigned long i915_mch_val(struct drm_i915_private *dev_priv);
extern unsigned long i915_gfx_val(struct drm_i915_private *dev_priv);
extern void i915_update_gfx_val(struct drm_i915_private *dev_priv);
extern void intel_console_resume(struct work_struct *work);
/* i915_irq.c */
void i915_queue_hangcheck(struct drm_device *dev);
void i915_handle_error(struct drm_device *dev, bool wedged);
void gen6_set_pm_mask(struct drm_i915_private *dev_priv, u32 pm_iir,
int new_delay);
extern void intel_irq_init(struct drm_device *dev);
drm/i915: Fixup hpd irq register setup ordering For GMCH platforms we set up the hpd irq registers in the irq postinstall hook. But since we only enable the irq sources we actually need in PORT_HOTPLUG_EN/STATUS, taking dev_priv->hotplug_supported_mask into account, no hpd interrupt sources is enabled since commit 52d7ecedac3f96fb562cb482c139015372728638 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Sat Dec 1 21:03:22 2012 +0100 drm/i915: reorder setup sequence to have irqs for output setup Wrongly set-up interrupts also lead to broken hw-based load-detection on at least GM45, resulting in ghost VGA/TV-out outputs. To fix this, delay the hotplug register setup until after all outputs are set up, by moving it into a new dev_priv->display.hpd_irq_callback. We might also move the PCH_SPLIT platforms to such a setup eventually. Another funny part is that we need to delay the fbdev initial config probing until after the hpd regs are setup, for otherwise it'll detect ghost outputs. But we can only enable the hpd interrupt handling itself (and the output polling) _after_ that initial scan, due to massive locking brain-damage in the fbdev setup code. Add a big comment to explain this cute little dragon lair. v2: Encapsulate all the fbdev handling by wrapping the move call into intel_fbdev_initial_config in intel_fb.c. Requested by Chris Wilson. v3: Applied bikeshed from Jesse Barnes. v4: Imre Deak noticed that we also need to call intel_hpd_init after the drm_irqinstall calls in the gpu reset and resume paths - otherwise hotplug will be broken. Also improve the comment a bit about why hpd_init needs to be called before we set up the initial fbdev config. Bugzilla: Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54943 Reported-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v3) Reviewed-by: Imre Deak <imre.deak@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-12-11 20:05:07 +07:00
extern void intel_hpd_init(struct drm_device *dev);
extern void intel_uncore_sanitize(struct drm_device *dev);
extern void intel_uncore_early_sanitize(struct drm_device *dev);
extern void intel_uncore_init(struct drm_device *dev);
extern void intel_uncore_check_errors(struct drm_device *dev);
extern void intel_uncore_fini(struct drm_device *dev);
void
i915_enable_pipestat(drm_i915_private_t *dev_priv, enum pipe pipe,
u32 status_mask);
void
i915_disable_pipestat(drm_i915_private_t *dev_priv, enum pipe pipe,
u32 status_mask);
/* i915_gem.c */
int i915_gem_init_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_create_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_pread_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_execbuffer(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_execbuffer2(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_pin_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_unpin_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_busy_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
int i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_entervt_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_leavevt_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_set_tiling(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_get_tiling(struct drm_device *dev, void *data,
struct drm_file *file_priv);
int i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
drm/i915: wait render timeout ioctl This helps implement GL_ARB_sync but stops short of allowing full blown sync objects. Finally we can use the new timed seqno waiting function to allow userspace to wait on a buffer object with a timeout. This implements that interface. The IOCTL will take as input a buffer object handle, and a timeout in nanoseconds (flags is currently optional but will likely be used for permutations of flush operations). Users may specify 0 nanoseconds to instantly check. The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any non-zero timeout parameter the wait ioctl will wait for the given number of nanoseconds on an object becoming unbusy. Since the wait itself does so holding struct_mutex the object may become re-busied before this completes. A similar but shorter race condition exists in the busy ioctl. v2: ETIME/ERESTARTSYS instead of changing to EBUSY, and EGAIN (Chris) Flush the object from the gpu write domain (Chris + Daniel) Fix leaked refcount in good case (Chris) Naturally align ioctl struct (Chris) v3: Drop lock after getting seqno to avoid ugly dance (Chris) v4: check for 0 timeout after olr check to allow polling (Chris) v5: Updated the comment. (Chris) v6: Return -ETIME instead of -EBUSY when timeout_ns is 0 (Daniel) Fix the commit message comment to be less ugly (Ben) Add a warning to check the return timespec (Ben) v7: Use DRM_AUTH for the ioctl. (Eugeni) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-05-25 05:03:10 +07:00
int i915_gem_wait_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
void i915_gem_load(struct drm_device *dev);
void *i915_gem_object_alloc(struct drm_device *dev);
void i915_gem_object_free(struct drm_i915_gem_object *obj);
void i915_gem_object_init(struct drm_i915_gem_object *obj,
const struct drm_i915_gem_object_ops *ops);
struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
size_t size);
void i915_init_vm(struct drm_i915_private *dev_priv,
struct i915_address_space *vm);
void i915_gem_free_object(struct drm_gem_object *obj);
void i915_gem_vma_destroy(struct i915_vma *vma);
#define PIN_MAPPABLE 0x1
#define PIN_NONBLOCK 0x2
#define PIN_GLOBAL 0x4
int __must_check i915_gem_object_pin(struct drm_i915_gem_object *obj,
struct i915_address_space *vm,
uint32_t alignment,
unsigned flags);
drm/i915: plumb VM into bind/unbind code As alluded to in several patches, and it will be reiterated later... A VMA is an abstraction for a GEM BO bound into an address space. Therefore it stands to reason, that the existing bind, and unbind are the ones which will be the most impacted. This patch implements this, and updates all callers which weren't already updated in the series (because it was too messy). This patch represents the bulk of an earlier, larger patch. I've pulled out a bunch of things by the request of Daniel. The history is preserved for posterity with the email convention of ">" One big change from the original patch aside from a bunch of cropping is I've created an i915_vma_unbind() function. That is because we always have the VMA anyway, and doing an extra lookup is useful. There is a caveat, we retain an i915_gem_object_ggtt_unbind, for the global cases which might not talk in VMAs. > drm/i915: plumb VM into object operations > > This patch was formerly known as: > "drm/i915: Create VMAs (part 3) - plumbing" > > This patch adds a VM argument, bind/unbind, and the object > offset/size/color getters/setters. It preserves the old ggtt helper > functions because things still need, and will continue to need them. > > Some code will still need to be ported over after this. > > v2: Fix purge to pick an object and unbind all vmas > This was doable because of the global bound list change. > > v3: With the commit to actually pin/unpin pages in place, there is no > longer a need to check if unbind succeeded before calling put_pages(). > Make put_pages only BUG() after checking pin count. > > v4: Rebased on top of the new hangcheck work by Mika > plumbed eb_destroy also > Many checkpatch related fixes > > v5: Very large rebase > > v6: > Change BUG_ON to WARN_ON (Daniel) > Rename vm to ggtt in preallocate stolen, since it is always ggtt when > dealing with stolen memory. (Daniel) > list_for_each will short-circuit already (Daniel) > remove superflous space (Daniel) > Use per object list of vmas (Daniel) > Make obj_bound_any() use obj_bound for each vm (Ben) > s/bind_to_gtt/bind_to_vm/ (Ben) > > Fixed up the inactive shrinker. As Daniel noticed the code could > potentially count the same object multiple times. While it's not > possible in the current case, since 1 object can only ever be bound into > 1 address space thus far - we may as well try to get something more > future proof in place now. With a prep patch before this to switch over > to using the bound list + inactive check, we're now able to carry that > forward for every address space an object is bound into. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> [danvet: Rebase on top of the loss of "drm/i915: Cleanup more of VMA in destroy".] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-08-01 07:00:10 +07:00
int __must_check i915_vma_unbind(struct i915_vma *vma);
int i915_gem_object_put_pages(struct drm_i915_gem_object *obj);
void i915_gem_release_all_mmaps(struct drm_i915_private *dev_priv);
void i915_gem_release_mmap(struct drm_i915_gem_object *obj);
void i915_gem_lastclose(struct drm_device *dev);
int __must_check i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
static inline struct page *i915_gem_object_get_page(struct drm_i915_gem_object *obj, int n)
{
struct sg_page_iter sg_iter;
for_each_sg_page(obj->pages->sgl, &sg_iter, obj->pages->nents, n)
return sg_page_iter_page(&sg_iter);
return NULL;
}
static inline void i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
{
BUG_ON(obj->pages == NULL);
obj->pages_pin_count++;
}
static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
{
BUG_ON(obj->pages_pin_count == 0);
obj->pages_pin_count--;
}
int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
int i915_gem_object_sync(struct drm_i915_gem_object *obj,
struct intel_ring_buffer *to);
void i915_vma_move_to_active(struct i915_vma *vma,
struct intel_ring_buffer *ring);
int i915_gem_dumb_create(struct drm_file *file_priv,
struct drm_device *dev,
struct drm_mode_create_dumb *args);
int i915_gem_mmap_gtt(struct drm_file *file_priv, struct drm_device *dev,
uint32_t handle, uint64_t *offset);
/**
* Returns true if seq1 is later than seq2.
*/
static inline bool
i915_seqno_passed(uint32_t seq1, uint32_t seq2)
{
return (int32_t)(seq1 - seq2) >= 0;
}
int __must_check i915_gem_get_seqno(struct drm_device *dev, u32 *seqno);
int __must_check i915_gem_set_seqno(struct drm_device *dev, u32 seqno);
int __must_check i915_gem_object_get_fence(struct drm_i915_gem_object *obj);
int __must_check i915_gem_object_put_fence(struct drm_i915_gem_object *obj);
static inline bool
i915_gem_object_pin_fence(struct drm_i915_gem_object *obj)
{
if (obj->fence_reg != I915_FENCE_REG_NONE) {
struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
dev_priv->fence_regs[obj->fence_reg].pin_count++;
return true;
} else
return false;
}
static inline void
i915_gem_object_unpin_fence(struct drm_i915_gem_object *obj)
{
if (obj->fence_reg != I915_FENCE_REG_NONE) {
struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
WARN_ON(dev_priv->fence_regs[obj->fence_reg].pin_count <= 0);
dev_priv->fence_regs[obj->fence_reg].pin_count--;
}
}
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
bool i915_gem_retire_requests(struct drm_device *dev);
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 18:25:36 +07:00
void i915_gem_retire_requests_ring(struct intel_ring_buffer *ring);
int __must_check i915_gem_check_wedge(struct i915_gpu_error *error,
drm/i915: non-interruptible sleeps can't handle -EAGAIN So don't return -EAGAIN, even in the case of a gpu hang. Remap it to -EIO instead. Note that this isn't really an issue with interruptability, but more that we have quite a few codepaths (mostly around kms stuff) that simply can't handle any errors and hence not even -EAGAIN. Instead of adding proper failure paths so that we could restart these ioctls we've opted for the cheap way out of sleeping non-interruptibly. Which works everywhere but when the gpu dies, which this patch fixes. So essentially interruptible == false means 'wait for the gpu or die trying'.' This patch is a bit ugly because intel_ring_begin is all non-interruptible and hence only returns -EIO. But as the comment in there says, auditing all the callsites would be a pain. To avoid duplicating code, reuse i915_gem_check_wedge in __wait_seqno and intel_wait_ring_buffer. Also use the opportunity to clarify the different cases in i915_gem_check_wedge a bit with comments. v2: Don't access dev_priv->mm.interruptible from check_wedge - we might not hold dev->struct_mutex, making this racy. Instead pass interruptible in as a parameter. I've noticed this because I've hit a BUG_ON(!mutex_is_locked) at the top of check_wedge. This has been added in commit b4aca0106c466b5a0329318203f65bac2d91b682 Author: Ben Widawsky <ben@bwidawsk.net> Date: Wed Apr 25 20:50:12 2012 -0700 drm/i915: extract some common olr+wedge code although that commit is missing any justification for this. I guess it's just copy&paste, because the same commit add the same BUG_ON check to check_olr, where it indeed makes sense. But in check_wedge everything we access is protected by other means, so this is superflous. And because it now gets in the way (we add a new caller in __wait_seqno, which can be called without dev->struct_mutext) let's just remove it. v3: Group all the i915_gem_check_wedge refactoring into this patch, so that this patch here is all about not returning -EAGAIN to callsites that can't handle syscall restarting. v4: Add clarification what interuptible == fales means in our code, requested by Ben Widawsky. v5: Fix EAGAIN mispell noticed by Chris Wilson. Reviewed-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Tested-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-07-05 03:54:13 +07:00
bool interruptible);
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
static inline bool i915_reset_in_progress(struct i915_gpu_error *error)
{
return unlikely(atomic_read(&error->reset_counter)
& (I915_RESET_IN_PROGRESS_FLAG | I915_WEDGED));
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
}
static inline bool i915_terminally_wedged(struct i915_gpu_error *error)
{
return atomic_read(&error->reset_counter) & I915_WEDGED;
}
static inline u32 i915_reset_count(struct i915_gpu_error *error)
{
return ((atomic_read(&error->reset_counter) & ~I915_WEDGED) + 1) / 2;
drm/i915: clear up wedged transitions We have two important transitions of the wedged state in the current code: - 0 -> 1: This means a hang has been detected, and signals to everyone that they please get of any locks, so that the reset work item can do its job. - 1 -> 0: The reset handler has completed. Now the last transition mixes up two states: "Reset completed and successful" and "Reset failed". To distinguish these two we do some tricks with the reset completion, but I simply could not convince myself that this doesn't race under odd circumstances. Hence split this up, and add a new terminal state indicating that the hw is gone for good. Also add explicit #defines for both states, update comments. v2: Split out the reset handling bugfix for the throttle ioctl. v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase error which prevented this patch from actually compiling. v4: To unify the wedged state with the reset counter, keep the reset-in-progress state just as a flag. The terminally-wedged state is now denoted with a big number. v5: Add a comment to the reset_counter special values explaining that WEDGED & RESET_IN_PROGRESS needs to be true for the code to be correct. v6: Fixup logic errors introduced with the wedged+reset_counter unification. Since WEDGED implies reset-in-progress (in a way we're terminally stuck in the dead-but-reset-not-completed state), we need ensure that we check for this everywhere. The specific bug was in wait_for_error, which would simply have timed out. v7: Extract an inline i915_reset_in_progress helper to make the code more readable. Also annote the reset-in-progress case with an unlikely, to help the compiler optimize the fastpath. Do the same for the terminally wedged case with i915_terminally_wedged. Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-15 23:17:22 +07:00
}
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 18:25:36 +07:00
void i915_gem_reset(struct drm_device *dev);
bool i915_gem_clflush_object(struct drm_i915_gem_object *obj, bool force);
int __must_check i915_gem_object_finish_gpu(struct drm_i915_gem_object *obj);
int __must_check i915_gem_init(struct drm_device *dev);
int __must_check i915_gem_init_hw(struct drm_device *dev);
int i915_gem_l3_remap(struct intel_ring_buffer *ring, int slice);
void i915_gem_init_swizzling(struct drm_device *dev);
void i915_gem_cleanup_ringbuffer(struct drm_device *dev);
int __must_check i915_gpu_idle(struct drm_device *dev);
int __must_check i915_gem_suspend(struct drm_device *dev);
int __i915_add_request(struct intel_ring_buffer *ring,
struct drm_file *file,
struct drm_i915_gem_object *batch_obj,
u32 *seqno);
#define i915_add_request(ring, seqno) \
__i915_add_request(ring, NULL, NULL, seqno)
int __must_check i915_wait_seqno(struct intel_ring_buffer *ring,
uint32_t seqno);
int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
int __must_check
i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj,
bool write);
int __must_check
i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write);
int __must_check
i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
u32 alignment,
struct intel_ring_buffer *pipelined);
void i915_gem_object_unpin_from_display_plane(struct drm_i915_gem_object *obj);
int i915_gem_attach_phys_object(struct drm_device *dev,
struct drm_i915_gem_object *obj,
int id,
int align);
void i915_gem_detach_phys_object(struct drm_device *dev,
struct drm_i915_gem_object *obj);
void i915_gem_free_all_phys_object(struct drm_device *dev);
drm/i915: Boost RPS frequency for CPU stalls If we encounter a situation where the CPU blocks waiting for results from the GPU, give the GPU a kick to boost its the frequency. This should work to reduce user interface stalls and to quickly promote mesa to high frequencies - but the cost is that our requested frequency stalls high (as we do not idle for long enough before rc6 to start reducing frequencies, nor are we aggressive at down clocking an underused GPU). However, this should be mitigated by rc6 itself powering off the GPU when idle, and that energy use is dependent upon the workload of the GPU in addition to its frequency (e.g. the math or sampler functions only consume power when used). Still, this is likely to adversely affect light workloads. In particular, this nearly eliminates the highly noticeable wake-up lag in animations from idle. For example, expose or workspace transitions. (However, given the situation where we fail to downclock, our requested frequency is almost always the maximum, except for Baytrail where we manually downclock upon idling. This often masks the latency of upclocking after being idle, so animations are typically smooth - at the cost of increased power consumption.) Stéphane raised the concern that this will punish good applications and reward bad applications - but due to the nature of how mesa performs its client throttling, I believe all mesa applications will be roughly equally affected. To address this concern, and to prevent applications like compositors from permanently boosting the RPS state, we ratelimit the frequency of the wait-boosts each client recieves. Unfortunately, this techinique is ineffective with Ironlake - which also has dynamic render power states and suffers just as dramatically. For Ironlake, the thermal/power headroom is shared with the CPU through Intelligent Power Sharing and the intel-ips module. This leaves us with no GPU boost frequencies available when coming out of idle, and due to hardware limitations we cannot change the arbitration between the CPU and GPU quickly enough to be effective. v2: Limit each client to receiving a single boost for each active period. Tested by QA to only marginally increase power, and to demonstrably increase throughput in games. No latency measurements yet. v3: Cater for front-buffer rendering with manual throttling. v4: Tidy up. v5: Sadly the compositor needs frequent boosts as it may never idle, but due to its picking mechanism (using ReadPixels) may require frequent waits. Those waits, along with the waits for the vrefresh swap, conspire to keep the GPU at low frequencies despite the interactive latency. To overcome this we ditch the one-boost-per-active-period and just ratelimit the number of wait-boosts each client can receive. Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Kenneth Graunke <kenneth@whitecape.org> Cc: Stéphane Marchesin <stephane.marchesin@gmail.com> Cc: Owen Taylor <otaylor@redhat.com> Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com> Cc: "Zhuang, Lena" <lena.zhuang@intel.com> Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> [danvet: No extern for function prototypes in headers.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-09-25 23:34:56 +07:00
int i915_gem_open(struct drm_device *dev, struct drm_file *file);
void i915_gem_release(struct drm_device *dev, struct drm_file *file);
uint32_t
i915_gem_get_gtt_size(struct drm_device *dev, uint32_t size, int tiling_mode);
uint32_t
i915_gem_get_gtt_alignment(struct drm_device *dev, uint32_t size,
int tiling_mode, bool fenced);
int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
enum i915_cache_level cache_level);
i915: add dmabuf/prime buffer sharing support. This adds handle->fd and fd->handle support to i915, this is to allow for offloading of rendering in one direction and outputs in the other. v2 from Daniel Vetter: - fixup conflicts with the prepare/finish gtt prep work. - implement ppgtt binding support. Note that we have squat i-g-t testcoverage for any of the lifetime and access rules dma_buf/prime support brings along. And there are quite a few intricate situations here. Also note that the integration with the existing code is a bit hackish, especially around get_gtt_pages and put_gtt_pages. It imo would be easier with the prep code from Chris Wilson's unbound series, but that is for 3.6. Also note that I didn't bother to put the new prepare/finish gtt hooks to good use by moving the dma_buf_map/unmap_attachment calls in there (like we've originally planned for). Last but not least this patch is only compile-tested, but I've changed very little compared to Dave Airlie's version. So there's a decent chance v2 on drm-next works as well as v1 on 3.4-rc. v3: Right when I've hit sent I've noticed that I've screwed up one obj->sg_list (for dmar support) and obj->sg_table (for prime support) disdinction. We should be able to merge these 2 paths, but that's material for another patch. v4: fix the error reporting bugs pointed out by ickle. v5: fix another error, and stop non-gtt mmaps on shared objects stop pread/pwrite on imported objects, add fake kmap Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-05-10 20:25:09 +07:00
struct drm_gem_object *i915_gem_prime_import(struct drm_device *dev,
struct dma_buf *dma_buf);
struct dma_buf *i915_gem_prime_export(struct drm_device *dev,
struct drm_gem_object *gem_obj, int flags);
void i915_gem_restore_fences(struct drm_device *dev);
unsigned long i915_gem_obj_offset(struct drm_i915_gem_object *o,
struct i915_address_space *vm);
bool i915_gem_obj_bound_any(struct drm_i915_gem_object *o);
bool i915_gem_obj_bound(struct drm_i915_gem_object *o,
struct i915_address_space *vm);
unsigned long i915_gem_obj_size(struct drm_i915_gem_object *o,
struct i915_address_space *vm);
struct i915_vma *i915_gem_obj_to_vma(struct drm_i915_gem_object *obj,
struct i915_address_space *vm);
struct i915_vma *
i915_gem_obj_lookup_or_create_vma(struct drm_i915_gem_object *obj,
struct i915_address_space *vm);
struct i915_vma *i915_gem_obj_to_ggtt(struct drm_i915_gem_object *obj);
static inline bool i915_gem_obj_is_pinned(struct drm_i915_gem_object *obj) {
struct i915_vma *vma;
list_for_each_entry(vma, &obj->vma_list, vma_link)
if (vma->pin_count > 0)
return true;
return false;
}
/* Some GGTT VM helpers */
#define obj_to_ggtt(obj) \
(&((struct drm_i915_private *)(obj)->base.dev->dev_private)->gtt.base)
static inline bool i915_is_ggtt(struct i915_address_space *vm)
{
struct i915_address_space *ggtt =
&((struct drm_i915_private *)(vm)->dev->dev_private)->gtt.base;
return vm == ggtt;
}
static inline bool i915_gem_obj_ggtt_bound(struct drm_i915_gem_object *obj)
{
return i915_gem_obj_bound(obj, obj_to_ggtt(obj));
}
static inline unsigned long
i915_gem_obj_ggtt_offset(struct drm_i915_gem_object *obj)
{
return i915_gem_obj_offset(obj, obj_to_ggtt(obj));
}
static inline unsigned long
i915_gem_obj_ggtt_size(struct drm_i915_gem_object *obj)
{
return i915_gem_obj_size(obj, obj_to_ggtt(obj));
}
static inline int __must_check
i915_gem_obj_ggtt_pin(struct drm_i915_gem_object *obj,
uint32_t alignment,
unsigned flags)
{
return i915_gem_object_pin(obj, obj_to_ggtt(obj), alignment, flags | PIN_GLOBAL);
}
static inline int
i915_gem_object_ggtt_unbind(struct drm_i915_gem_object *obj)
{
return i915_vma_unbind(i915_gem_obj_to_ggtt(obj));
}
void i915_gem_object_ggtt_unpin(struct drm_i915_gem_object *obj);
drm/i915: preliminary context support Very basic code for context setup/destruction in the driver. Adds the file i915_gem_context.c This file implements HW context support. On gen5+ a HW context consists of an opaque GPU object which is referenced at times of context saves and restores. With RC6 enabled, the context is also referenced as the GPU enters and exists from RC6 (GPU has it's own internal power context, except on gen5). Though something like a context does exist for the media ring, the code only supports contexts for the render ring. In software, there is a distinction between contexts created by the user, and the default HW context. The default HW context is used by GPU clients that do not request setup of their own hardware context. The default context's state is never restored to help prevent programming errors. This would happen if a client ran and piggy-backed off another clients GPU state. The default context only exists to give the GPU some offset to load as the current to invoke a save of the context we actually care about. In fact, the code could likely be constructed, albeit in a more complicated fashion, to never use the default context, though that limits the driver's ability to swap out, and/or destroy other contexts. All other contexts are created as a request by the GPU client. These contexts store GPU state, and thus allow GPU clients to not re-emit state (and potentially query certain state) at any time. The kernel driver makes certain that the appropriate commands are inserted. There are 4 entry points into the contexts, init, fini, open, close. The names are self-explanatory except that init can be called during reset, and also during pm thaw/resume. As we expect our context to be preserved across these events, we do not reinitialize in this case. As Adam Jackson pointed out, The cutoff of 1MB where a HW context is considered too big is arbitrary. The reason for this is even though context sizes are increasing with every generation, they have yet to eclipse even 32k. If we somehow read back way more than that, it probably means BIOS has done something strange, or we're running on a platform that wasn't designed for this. v2: rename load/unload to init/fini (daniel) remove ILK support for get_size() (indirectly daniel) add HAS_HW_CONTEXTS macro to clarify supported platforms (daniel) added comments (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
2012-06-05 04:42:42 +07:00
/* i915_gem_context.c */
#define ctx_to_ppgtt(ctx) container_of((ctx)->vm, struct i915_hw_ppgtt, base)
int __must_check i915_gem_context_init(struct drm_device *dev);
drm/i915: preliminary context support Very basic code for context setup/destruction in the driver. Adds the file i915_gem_context.c This file implements HW context support. On gen5+ a HW context consists of an opaque GPU object which is referenced at times of context saves and restores. With RC6 enabled, the context is also referenced as the GPU enters and exists from RC6 (GPU has it's own internal power context, except on gen5). Though something like a context does exist for the media ring, the code only supports contexts for the render ring. In software, there is a distinction between contexts created by the user, and the default HW context. The default HW context is used by GPU clients that do not request setup of their own hardware context. The default context's state is never restored to help prevent programming errors. This would happen if a client ran and piggy-backed off another clients GPU state. The default context only exists to give the GPU some offset to load as the current to invoke a save of the context we actually care about. In fact, the code could likely be constructed, albeit in a more complicated fashion, to never use the default context, though that limits the driver's ability to swap out, and/or destroy other contexts. All other contexts are created as a request by the GPU client. These contexts store GPU state, and thus allow GPU clients to not re-emit state (and potentially query certain state) at any time. The kernel driver makes certain that the appropriate commands are inserted. There are 4 entry points into the contexts, init, fini, open, close. The names are self-explanatory except that init can be called during reset, and also during pm thaw/resume. As we expect our context to be preserved across these events, we do not reinitialize in this case. As Adam Jackson pointed out, The cutoff of 1MB where a HW context is considered too big is arbitrary. The reason for this is even though context sizes are increasing with every generation, they have yet to eclipse even 32k. If we somehow read back way more than that, it probably means BIOS has done something strange, or we're running on a platform that wasn't designed for this. v2: rename load/unload to init/fini (daniel) remove ILK support for get_size() (indirectly daniel) add HAS_HW_CONTEXTS macro to clarify supported platforms (daniel) added comments (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
2012-06-05 04:42:42 +07:00
void i915_gem_context_fini(struct drm_device *dev);
void i915_gem_context_reset(struct drm_device *dev);
int i915_gem_context_open(struct drm_device *dev, struct drm_file *file);
int i915_gem_context_enable(struct drm_i915_private *dev_priv);
drm/i915: preliminary context support Very basic code for context setup/destruction in the driver. Adds the file i915_gem_context.c This file implements HW context support. On gen5+ a HW context consists of an opaque GPU object which is referenced at times of context saves and restores. With RC6 enabled, the context is also referenced as the GPU enters and exists from RC6 (GPU has it's own internal power context, except on gen5). Though something like a context does exist for the media ring, the code only supports contexts for the render ring. In software, there is a distinction between contexts created by the user, and the default HW context. The default HW context is used by GPU clients that do not request setup of their own hardware context. The default context's state is never restored to help prevent programming errors. This would happen if a client ran and piggy-backed off another clients GPU state. The default context only exists to give the GPU some offset to load as the current to invoke a save of the context we actually care about. In fact, the code could likely be constructed, albeit in a more complicated fashion, to never use the default context, though that limits the driver's ability to swap out, and/or destroy other contexts. All other contexts are created as a request by the GPU client. These contexts store GPU state, and thus allow GPU clients to not re-emit state (and potentially query certain state) at any time. The kernel driver makes certain that the appropriate commands are inserted. There are 4 entry points into the contexts, init, fini, open, close. The names are self-explanatory except that init can be called during reset, and also during pm thaw/resume. As we expect our context to be preserved across these events, we do not reinitialize in this case. As Adam Jackson pointed out, The cutoff of 1MB where a HW context is considered too big is arbitrary. The reason for this is even though context sizes are increasing with every generation, they have yet to eclipse even 32k. If we somehow read back way more than that, it probably means BIOS has done something strange, or we're running on a platform that wasn't designed for this. v2: rename load/unload to init/fini (daniel) remove ILK support for get_size() (indirectly daniel) add HAS_HW_CONTEXTS macro to clarify supported platforms (daniel) added comments (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
2012-06-05 04:42:42 +07:00
void i915_gem_context_close(struct drm_device *dev, struct drm_file *file);
int i915_switch_context(struct intel_ring_buffer *ring,
drm/i915: Get context early in execbuf We need to have the address space when reserving space for the objects. Since the address space and context are tied together, and reserve occurs before context switch (for good reason), we must lookup our context earlier in the process. This leaves some room for optimizations where we no longer need to use ctx_id in certain places. This will be addressed in a subsequent patch. Important tricky bit: Because slow relocations during execbuffer drop struct_mutex Perhaps it would be best to acquire the reference when we get the context, but I'll save that for another day (note I have written the patch before, and I found the changes required to be uglier than this). Note that since we currently access everything via context id, and not the data structure this is fine, though not desirable. The next change attempts to get the context only once via the context ID idr lookup, and as such, the following can happen: CTX-A is created, refcount = 1 CTX-A execbuf, mutex dropped close IOCTL called on CTX-A, refcount = 0 CTX-A resumes in execbuf. v2: Rebased on top of commit b6359918b885da7c7b58c050674278dbd06020ab Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Wed Oct 30 15:44:16 2013 +0200 drm/i915: add i915_get_reset_stats_ioctl v3: Rebased on top of commit 25b3dfc87bff80317d67ddd2cd4cfb91e6fe7d79 Author: Mika Westerberg <mika.westerberg@linux.intel.com> Date: Tue Nov 12 11:57:30 2013 +0200 Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Tue Nov 26 16:14:33 2013 +0200 drm/i915: check context reset stats before relocations Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-12-07 05:11:21 +07:00
struct drm_file *file, struct i915_hw_context *to);
struct i915_hw_context *
i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id);
void i915_gem_context_free(struct kref *ctx_ref);
static inline void i915_gem_context_reference(struct i915_hw_context *ctx)
{
if (ctx->obj && HAS_HW_CONTEXTS(ctx->obj->base.dev))
kref_get(&ctx->ref);
}
static inline void i915_gem_context_unreference(struct i915_hw_context *ctx)
{
if (ctx->obj && HAS_HW_CONTEXTS(ctx->obj->base.dev))
kref_put(&ctx->ref, i915_gem_context_free);
}
static inline bool i915_gem_context_is_default(const struct i915_hw_context *c)
{
return c->id == DEFAULT_CONTEXT_ID;
}
int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
i915: add dmabuf/prime buffer sharing support. This adds handle->fd and fd->handle support to i915, this is to allow for offloading of rendering in one direction and outputs in the other. v2 from Daniel Vetter: - fixup conflicts with the prepare/finish gtt prep work. - implement ppgtt binding support. Note that we have squat i-g-t testcoverage for any of the lifetime and access rules dma_buf/prime support brings along. And there are quite a few intricate situations here. Also note that the integration with the existing code is a bit hackish, especially around get_gtt_pages and put_gtt_pages. It imo would be easier with the prep code from Chris Wilson's unbound series, but that is for 3.6. Also note that I didn't bother to put the new prepare/finish gtt hooks to good use by moving the dma_buf_map/unmap_attachment calls in there (like we've originally planned for). Last but not least this patch is only compile-tested, but I've changed very little compared to Dave Airlie's version. So there's a decent chance v2 on drm-next works as well as v1 on 3.4-rc. v3: Right when I've hit sent I've noticed that I've screwed up one obj->sg_list (for dmar support) and obj->sg_table (for prime support) disdinction. We should be able to merge these 2 paths, but that's material for another patch. v4: fix the error reporting bugs pointed out by ickle. v5: fix another error, and stop non-gtt mmaps on shared objects stop pread/pwrite on imported objects, add fake kmap Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-05-10 20:25:09 +07:00
/* i915_gem_evict.c */
int __must_check i915_gem_evict_something(struct drm_device *dev,
struct i915_address_space *vm,
int min_size,
unsigned alignment,
unsigned cache_level,
unsigned flags);
int i915_gem_evict_vm(struct i915_address_space *vm, bool do_idle);
int i915_gem_evict_everything(struct drm_device *dev);
/* i915_gem_gtt.c */
drm/i915: Disable GGTT PTEs on GEN6+ suspend Once the machine gets to a certain point in the suspend process, we expect the GPU to be idle. If it is not, we might corrupt memory. Empirically (with an early version of this patch) we have seen this is not the case. We cannot currently explain why the latent GPU writes occur. In the technical sense, this patch is a workaround in that we have an issue we can't explain, and the patch indirectly solves the issue. However, it's really better than a workaround because we understand why it works, and it really should be a safe thing to do in all cases. The noticeable effect other than the debug messages would be an increase in the suspend time. I have not measure how expensive it actually is. I think it would be good to spend further time to root cause why we're seeing these latent writes, but it shouldn't preclude preventing the fallout. NOTE: It should be safe (and makes some sense IMO) to also keep the VALID bit unset on resume when we clear_range(). I've opted not to do this as properly clearing those bits at some later point would be extra work. v2: Fix bugzilla link Bugzilla: http://bugs.freedesktop.org/show_bug.cgi?id=65496 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=59321 Tested-by: Takashi Iwai <tiwai@suse.de> Tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Tested-By: Todd Previte <tprevite@gmail.com> Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2013-10-16 23:21:30 +07:00
void i915_check_and_clear_faults(struct drm_device *dev);
void i915_gem_suspend_gtt_mappings(struct drm_device *dev);
void i915_gem_restore_gtt_mappings(struct drm_device *dev);
int __must_check i915_gem_gtt_prepare_object(struct drm_i915_gem_object *obj);
void i915_gem_gtt_finish_object(struct drm_i915_gem_object *obj);
void i915_gem_init_global_gtt(struct drm_device *dev);
void i915_gem_setup_global_gtt(struct drm_device *dev, unsigned long start,
unsigned long mappable_end, unsigned long end);
drm/i915: Stop using AGP layer for GEN6+ As a quick hack we make the old intel_gtt structure mutable so we can fool a bunch of the existing code which depends on elements in that data structure. We can/should try to remove this in a subsequent patch. This should preserve the old gtt init behavior which upon writing these patches seems incorrect. The next patch will fix these things. The one exception is VLV which doesn't have the preserved flush control write behavior. Since we want to do that for all GEN6+ stuff, we'll handle that in a later patch. Mainstream VLV support doesn't actually exist yet anyway. v2: Update the comment to remove the "voodoo" Check that the last pte written matches what we readback v3: actually kill cache_level_to_agp_type since most of the flags will disappear in an upcoming patch v4: v3 was actually not what we wanted (Daniel) Make the ggtt bind assertions better and stricter (Chris) Fix some uncaught errors at gtt init (Chris) Some other random stuff that Chris wanted v5: check for i==0 in gen6_ggtt_bind_object to shut up gcc (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by [v4]: Chris Wilson <chris@chris-wilson.co.uk> [danvet: Make the cache_level -> agp_flags conversion for pre-gen6 a tad more robust by mapping everything != CACHE_NONE to the cached agp flag - we have a 1:1 uncached mapping, but different modes of cacheable (at least on later generations). Suggested by Chris Wilson.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-05 00:21:27 +07:00
int i915_gem_gtt_init(struct drm_device *dev);
static inline void i915_gem_chipset_flush(struct drm_device *dev)
drm/i915: Stop using AGP layer for GEN6+ As a quick hack we make the old intel_gtt structure mutable so we can fool a bunch of the existing code which depends on elements in that data structure. We can/should try to remove this in a subsequent patch. This should preserve the old gtt init behavior which upon writing these patches seems incorrect. The next patch will fix these things. The one exception is VLV which doesn't have the preserved flush control write behavior. Since we want to do that for all GEN6+ stuff, we'll handle that in a later patch. Mainstream VLV support doesn't actually exist yet anyway. v2: Update the comment to remove the "voodoo" Check that the last pte written matches what we readback v3: actually kill cache_level_to_agp_type since most of the flags will disappear in an upcoming patch v4: v3 was actually not what we wanted (Daniel) Make the ggtt bind assertions better and stricter (Chris) Fix some uncaught errors at gtt init (Chris) Some other random stuff that Chris wanted v5: check for i==0 in gen6_ggtt_bind_object to shut up gcc (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by [v4]: Chris Wilson <chris@chris-wilson.co.uk> [danvet: Make the cache_level -> agp_flags conversion for pre-gen6 a tad more robust by mapping everything != CACHE_NONE to the cached agp flag - we have a 1:1 uncached mapping, but different modes of cacheable (at least on later generations). Suggested by Chris Wilson.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-05 00:21:27 +07:00
{
if (INTEL_INFO(dev)->gen < 6)
intel_gtt_chipset_flush();
}
int i915_gem_init_ppgtt(struct drm_device *dev, struct i915_hw_ppgtt *ppgtt);
static inline bool intel_enable_ppgtt(struct drm_device *dev, bool full)
{
if (i915.enable_ppgtt == 0 || !HAS_ALIASING_PPGTT(dev))
return false;
drm/i915: Stop using AGP layer for GEN6+ As a quick hack we make the old intel_gtt structure mutable so we can fool a bunch of the existing code which depends on elements in that data structure. We can/should try to remove this in a subsequent patch. This should preserve the old gtt init behavior which upon writing these patches seems incorrect. The next patch will fix these things. The one exception is VLV which doesn't have the preserved flush control write behavior. Since we want to do that for all GEN6+ stuff, we'll handle that in a later patch. Mainstream VLV support doesn't actually exist yet anyway. v2: Update the comment to remove the "voodoo" Check that the last pte written matches what we readback v3: actually kill cache_level_to_agp_type since most of the flags will disappear in an upcoming patch v4: v3 was actually not what we wanted (Daniel) Make the ggtt bind assertions better and stricter (Chris) Fix some uncaught errors at gtt init (Chris) Some other random stuff that Chris wanted v5: check for i==0 in gen6_ggtt_bind_object to shut up gcc (Ben) Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Reviewed-by [v4]: Chris Wilson <chris@chris-wilson.co.uk> [danvet: Make the cache_level -> agp_flags conversion for pre-gen6 a tad more robust by mapping everything != CACHE_NONE to the cached agp flag - we have a 1:1 uncached mapping, but different modes of cacheable (at least on later generations). Suggested by Chris Wilson.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-05 00:21:27 +07:00
if (i915.enable_ppgtt == 1 && full)
return false;
#ifdef CONFIG_INTEL_IOMMU
/* Disable ppgtt on SNB if VT-d is on. */
if (INTEL_INFO(dev)->gen == 6 && intel_iommu_gfx_mapped) {
DRM_INFO("Disabling PPGTT because VT-d is on\n");
return false;
}
#endif
if (full)
return HAS_PPGTT(dev);
else
return HAS_ALIASING_PPGTT(dev);
}
static inline void ppgtt_release(struct kref *kref)
{
struct i915_hw_ppgtt *ppgtt = container_of(kref, struct i915_hw_ppgtt, ref);
struct drm_device *dev = ppgtt->base.dev;
struct drm_i915_private *dev_priv = dev->dev_private;
struct i915_address_space *vm = &ppgtt->base;
if (ppgtt == dev_priv->mm.aliasing_ppgtt ||
(list_empty(&vm->active_list) && list_empty(&vm->inactive_list))) {
ppgtt->base.cleanup(&ppgtt->base);
return;
}
/*
* Make sure vmas are unbound before we take down the drm_mm
*
* FIXME: Proper refcounting should take care of this, this shouldn't be
* needed at all.
*/
if (!list_empty(&vm->active_list)) {
struct i915_vma *vma;
list_for_each_entry(vma, &vm->active_list, mm_list)
if (WARN_ON(list_empty(&vma->vma_link) ||
list_is_singular(&vma->vma_link)))
break;
i915_gem_evict_vm(&ppgtt->base, true);
} else {
i915_gem_retire_requests(dev);
i915_gem_evict_vm(&ppgtt->base, false);
}
ppgtt->base.cleanup(&ppgtt->base);
}
/* i915_gem_stolen.c */
int i915_gem_init_stolen(struct drm_device *dev);
int i915_gem_stolen_setup_compression(struct drm_device *dev, int size);
void i915_gem_stolen_cleanup_compression(struct drm_device *dev);
void i915_gem_cleanup_stolen(struct drm_device *dev);
struct drm_i915_gem_object *
i915_gem_object_create_stolen(struct drm_device *dev, u32 size);
struct drm_i915_gem_object *
i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,
u32 stolen_offset,
u32 gtt_offset,
u32 size);
void i915_gem_object_release_stolen(struct drm_i915_gem_object *obj);
/* i915_gem_tiling.c */
static inline bool i915_gem_object_needs_bit17_swizzle(struct drm_i915_gem_object *obj)
{
drm_i915_private_t *dev_priv = obj->base.dev->dev_private;
return dev_priv->mm.bit_6_swizzle_x == I915_BIT_6_SWIZZLE_9_10_17 &&
obj->tiling_mode != I915_TILING_NONE;
}
void i915_gem_detect_bit_6_swizzle(struct drm_device *dev);
void i915_gem_object_do_bit_17_swizzle(struct drm_i915_gem_object *obj);
void i915_gem_object_save_bit_17_swizzle(struct drm_i915_gem_object *obj);
/* i915_gem_debug.c */
#if WATCH_LISTS
int i915_verify_lists(struct drm_device *dev);
#else
#define i915_verify_lists(dev) 0
#endif
/* i915_debugfs.c */
int i915_debugfs_init(struct drm_minor *minor);
void i915_debugfs_cleanup(struct drm_minor *minor);
#ifdef CONFIG_DEBUG_FS
void intel_display_crc_init(struct drm_device *dev);
#else
static inline void intel_display_crc_init(struct drm_device *dev) {}
#endif
/* i915_gpu_error.c */
__printf(2, 3)
void i915_error_printf(struct drm_i915_error_state_buf *e, const char *f, ...);
int i915_error_state_to_str(struct drm_i915_error_state_buf *estr,
const struct i915_error_state_file_priv *error);
int i915_error_state_buf_init(struct drm_i915_error_state_buf *eb,
size_t count, loff_t pos);
static inline void i915_error_state_buf_release(
struct drm_i915_error_state_buf *eb)
{
kfree(eb->buf);
}
void i915_capture_error_state(struct drm_device *dev);
void i915_error_state_get(struct drm_device *dev,
struct i915_error_state_file_priv *error_priv);
void i915_error_state_put(struct i915_error_state_file_priv *error_priv);
void i915_destroy_error_state(struct drm_device *dev);
void i915_get_extra_instdone(struct drm_device *dev, uint32_t *instdone);
const char *i915_cache_level_str(int type);
/* i915_suspend.c */
extern int i915_save_state(struct drm_device *dev);
extern int i915_restore_state(struct drm_device *dev);
/* i915_ums.c */
void i915_save_display_reg(struct drm_device *dev);
void i915_restore_display_reg(struct drm_device *dev);
drm/i915: rc6 in sysfs Merge rc6 information into the power group for our device. Until now the i915 driver has not had any sysfs entries (aside from the connector stuff enabled by drm core). Since it seems like we're likely to have more in the future I created a new file for sysfs stubs, as well as the rc6 sysfs functions which don't really belong elsewhere (perhaps i915_suspend, but most of the stuff is in intel_display,c). displays rc6 modes enabled (as a hex mask): cat /sys/class/drm/card0/power/rc6_enable displays #ms GPU has been in rc6 since boot: cat /sys/class/drm/card0/power/rc6_residency_ms displays #ms GPU has been in deep rc6 since boot: cat /sys/class/drm/card0/power/rc6p_residency_ms displays #ms GPU has been in deepest rc6 since boot: cat /sys/class/drm/card0/power/rc6pp_residency_ms Important note: I've seen on SNB that even when RC6 is *not* enabled the rc6 register seems to have a random value in it. I can only guess at the reason reason for this. Those writing tools that utilize this value need to be careful and probably want to scrutinize the value very carefully. v2: use common rc6 residency units to milliseconds for the other RC6 types v3: don't create sysfs files for GEN <= 5 add a rc6_enable to show a mask of enabled rc6 types use unmerge instead of remove for sysfs group squash intel_enable_rc6() extraction into this patch v4: rename sysfs files (Chris) CC: Chris Wilson <chris@chris-wilson.co.uk> CC: Daniel Vetter <daniel.vetter@ffwll.ch>f CC: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: squash in the 64bit division fix by Chris Wilson.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-04-11 11:17:01 +07:00
/* i915_sysfs.c */
void i915_setup_sysfs(struct drm_device *dev_priv);
void i915_teardown_sysfs(struct drm_device *dev_priv);
/* intel_i2c.c */
extern int intel_setup_gmbus(struct drm_device *dev);
extern void intel_teardown_gmbus(struct drm_device *dev);
static inline bool intel_gmbus_is_port_valid(unsigned port)
{
return (port >= GMBUS_PORT_SSC && port <= GMBUS_PORT_DPD);
}
extern struct i2c_adapter *intel_gmbus_get_adapter(
struct drm_i915_private *dev_priv, unsigned port);
extern void intel_gmbus_set_speed(struct i2c_adapter *adapter, int speed);
extern void intel_gmbus_force_bit(struct i2c_adapter *adapter, bool force_bit);
static inline bool intel_gmbus_is_forced_bit(struct i2c_adapter *adapter)
{
return container_of(adapter, struct intel_gmbus, adapter)->force_bit;
}
extern void intel_i2c_reset(struct drm_device *dev);
/* intel_opregion.c */
struct intel_encoder;
extern int intel_opregion_setup(struct drm_device *dev);
#ifdef CONFIG_ACPI
extern void intel_opregion_init(struct drm_device *dev);
extern void intel_opregion_fini(struct drm_device *dev);
extern void intel_opregion_asle_intr(struct drm_device *dev);
extern int intel_opregion_notify_encoder(struct intel_encoder *intel_encoder,
bool enable);
extern int intel_opregion_notify_adapter(struct drm_device *dev,
pci_power_t state);
#else
static inline void intel_opregion_init(struct drm_device *dev) { return; }
static inline void intel_opregion_fini(struct drm_device *dev) { return; }
static inline void intel_opregion_asle_intr(struct drm_device *dev) { return; }
static inline int
intel_opregion_notify_encoder(struct intel_encoder *intel_encoder, bool enable)
{
return 0;
}
static inline int
intel_opregion_notify_adapter(struct drm_device *dev, pci_power_t state)
{
return 0;
}
#endif
/* intel_acpi.c */
#ifdef CONFIG_ACPI
extern void intel_register_dsm_handler(void);
extern void intel_unregister_dsm_handler(void);
#else
static inline void intel_register_dsm_handler(void) { return; }
static inline void intel_unregister_dsm_handler(void) { return; }
#endif /* CONFIG_ACPI */
/* modesetting */
extern void intel_modeset_init_hw(struct drm_device *dev);
extern void intel_modeset_suspend_hw(struct drm_device *dev);
extern void intel_modeset_init(struct drm_device *dev);
extern void intel_modeset_gem_init(struct drm_device *dev);
extern void intel_modeset_cleanup(struct drm_device *dev);
extern void intel_connector_unregister(struct intel_connector *);
extern int intel_modeset_vga_set_state(struct drm_device *dev, bool state);
extern void intel_modeset_setup_hw_state(struct drm_device *dev,
bool force_restore);
extern void i915_redisable_vga(struct drm_device *dev);
extern bool intel_fbc_enabled(struct drm_device *dev);
extern void intel_disable_fbc(struct drm_device *dev);
extern bool ironlake_set_drps(struct drm_device *dev, u8 val);
extern void intel_init_pch_refclk(struct drm_device *dev);
extern void gen6_set_rps(struct drm_device *dev, u8 val);
extern void valleyview_set_rps(struct drm_device *dev, u8 val);
extern int valleyview_rps_max_freq(struct drm_i915_private *dev_priv);
extern int valleyview_rps_min_freq(struct drm_i915_private *dev_priv);
extern void intel_detect_pch(struct drm_device *dev);
extern int intel_trans_dp_port_sel(struct drm_crtc *crtc);
drm/i915: rc6 in sysfs Merge rc6 information into the power group for our device. Until now the i915 driver has not had any sysfs entries (aside from the connector stuff enabled by drm core). Since it seems like we're likely to have more in the future I created a new file for sysfs stubs, as well as the rc6 sysfs functions which don't really belong elsewhere (perhaps i915_suspend, but most of the stuff is in intel_display,c). displays rc6 modes enabled (as a hex mask): cat /sys/class/drm/card0/power/rc6_enable displays #ms GPU has been in rc6 since boot: cat /sys/class/drm/card0/power/rc6_residency_ms displays #ms GPU has been in deep rc6 since boot: cat /sys/class/drm/card0/power/rc6p_residency_ms displays #ms GPU has been in deepest rc6 since boot: cat /sys/class/drm/card0/power/rc6pp_residency_ms Important note: I've seen on SNB that even when RC6 is *not* enabled the rc6 register seems to have a random value in it. I can only guess at the reason reason for this. Those writing tools that utilize this value need to be careful and probably want to scrutinize the value very carefully. v2: use common rc6 residency units to milliseconds for the other RC6 types v3: don't create sysfs files for GEN <= 5 add a rc6_enable to show a mask of enabled rc6 types use unmerge instead of remove for sysfs group squash intel_enable_rc6() extraction into this patch v4: rename sysfs files (Chris) CC: Chris Wilson <chris@chris-wilson.co.uk> CC: Daniel Vetter <daniel.vetter@ffwll.ch>f CC: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> [danvet: squash in the 64bit division fix by Chris Wilson.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-04-11 11:17:01 +07:00
extern int intel_enable_rc6(const struct drm_device *dev);
extern bool i915_semaphore_is_enabled(struct drm_device *dev);
int i915_reg_read_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
int i915_get_reset_stats_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
/* overlay */
extern struct intel_overlay_error_state *intel_overlay_capture_error_state(struct drm_device *dev);
extern void intel_overlay_print_error_state(struct drm_i915_error_state_buf *e,
struct intel_overlay_error_state *error);
extern struct intel_display_error_state *intel_display_capture_error_state(struct drm_device *dev);
extern void intel_display_print_error_state(struct drm_i915_error_state_buf *e,
struct drm_device *dev,
struct intel_display_error_state *error);
/* On SNB platform, before reading ring registers forcewake bit
* must be set to prevent GT core from power down and stale values being
* returned.
*/
void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine);
void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine);
int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u8 mbox, u32 *val);
int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u8 mbox, u32 val);
/* intel_sideband.c */
u32 vlv_punit_read(struct drm_i915_private *dev_priv, u8 addr);
void vlv_punit_write(struct drm_i915_private *dev_priv, u8 addr, u32 val);
u32 vlv_nc_read(struct drm_i915_private *dev_priv, u8 addr);
u32 vlv_gpio_nc_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_gpio_nc_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
u32 vlv_cck_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_cck_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
u32 vlv_ccu_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_ccu_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
u32 vlv_bunit_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_bunit_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
u32 vlv_gps_core_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_gps_core_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
u32 vlv_dpio_read(struct drm_i915_private *dev_priv, enum pipe pipe, int reg);
void vlv_dpio_write(struct drm_i915_private *dev_priv, enum pipe pipe, int reg, u32 val);
u32 intel_sbi_read(struct drm_i915_private *dev_priv, u16 reg,
enum intel_sbi_destination destination);
void intel_sbi_write(struct drm_i915_private *dev_priv, u16 reg, u32 value,
enum intel_sbi_destination destination);
u32 vlv_flisdsi_read(struct drm_i915_private *dev_priv, u32 reg);
void vlv_flisdsi_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
int vlv_gpu_freq(struct drm_i915_private *dev_priv, int val);
int vlv_freq_opcode(struct drm_i915_private *dev_priv, int val);
void vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine);
void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine);
#define FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg) \
(((reg) >= 0x2000 && (reg) < 0x4000) ||\
((reg) >= 0x5000 && (reg) < 0x8000) ||\
((reg) >= 0xB000 && (reg) < 0x12000) ||\
((reg) >= 0x2E000 && (reg) < 0x30000))
#define FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(reg)\
(((reg) >= 0x12000 && (reg) < 0x14000) ||\
((reg) >= 0x22000 && (reg) < 0x24000) ||\
((reg) >= 0x30000 && (reg) < 0x40000))
#define FORCEWAKE_RENDER (1 << 0)
#define FORCEWAKE_MEDIA (1 << 1)
#define FORCEWAKE_ALL (FORCEWAKE_RENDER | FORCEWAKE_MEDIA)
#define I915_READ8(reg) dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true)
#define I915_WRITE8(reg, val) dev_priv->uncore.funcs.mmio_writeb(dev_priv, (reg), (val), true)
#define I915_READ16(reg) dev_priv->uncore.funcs.mmio_readw(dev_priv, (reg), true)
#define I915_WRITE16(reg, val) dev_priv->uncore.funcs.mmio_writew(dev_priv, (reg), (val), true)
#define I915_READ16_NOTRACE(reg) dev_priv->uncore.funcs.mmio_readw(dev_priv, (reg), false)
#define I915_WRITE16_NOTRACE(reg, val) dev_priv->uncore.funcs.mmio_writew(dev_priv, (reg), (val), false)
#define I915_READ(reg) dev_priv->uncore.funcs.mmio_readl(dev_priv, (reg), true)
#define I915_WRITE(reg, val) dev_priv->uncore.funcs.mmio_writel(dev_priv, (reg), (val), true)
#define I915_READ_NOTRACE(reg) dev_priv->uncore.funcs.mmio_readl(dev_priv, (reg), false)
#define I915_WRITE_NOTRACE(reg, val) dev_priv->uncore.funcs.mmio_writel(dev_priv, (reg), (val), false)
#define I915_WRITE64(reg, val) dev_priv->uncore.funcs.mmio_writeq(dev_priv, (reg), (val), true)
#define I915_READ64(reg) dev_priv->uncore.funcs.mmio_readq(dev_priv, (reg), true)
#define POSTING_READ(reg) (void)I915_READ_NOTRACE(reg)
#define POSTING_READ16(reg) (void)I915_READ16_NOTRACE(reg)
/* "Broadcast RGB" property */
#define INTEL_BROADCAST_RGB_AUTO 0
#define INTEL_BROADCAST_RGB_FULL 1
#define INTEL_BROADCAST_RGB_LIMITED 2
static inline uint32_t i915_vgacntrl_reg(struct drm_device *dev)
{
if (HAS_PCH_SPLIT(dev))
return CPU_VGACNTRL;
else if (IS_VALLEYVIEW(dev))
return VLV_VGACNTRL;
else
return VGACNTRL;
}
static inline void __user *to_user_ptr(u64 address)
{
return (void __user *)(uintptr_t)address;
}
static inline unsigned long msecs_to_jiffies_timeout(const unsigned int m)
{
unsigned long j = msecs_to_jiffies(m);
return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1);
}
static inline unsigned long
timespec_to_jiffies_timeout(const struct timespec *value)
{
unsigned long j = timespec_to_jiffies(value);
return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1);
}
/*
* If you need to wait X milliseconds between events A and B, but event B
* doesn't happen exactly after event A, you record the timestamp (jiffies) of
* when event A happened, then just before event B you call this function and
* pass the timestamp as the first argument, and X as the second argument.
*/
static inline void
wait_remaining_ms_from_jiffies(unsigned long timestamp_jiffies, int to_wait_ms)
{
unsigned long target_jiffies, tmp_jiffies, remaining_jiffies;
/*
* Don't re-read the value of "jiffies" every time since it may change
* behind our back and break the math.
*/
tmp_jiffies = jiffies;
target_jiffies = timestamp_jiffies +
msecs_to_jiffies_timeout(to_wait_ms);
if (time_after(target_jiffies, tmp_jiffies)) {
remaining_jiffies = target_jiffies - tmp_jiffies;
while (remaining_jiffies)
remaining_jiffies =
schedule_timeout_uninterruptible(remaining_jiffies);
}
}
#endif