Merge drm/drm-next into drm-intel-gt-next
authorTvrtko Ursulin <tvrtko.ursulin@intel.com>
Tue, 18 Jan 2022 10:54:02 +0000 (10:54 +0000)
committerTvrtko Ursulin <tvrtko.ursulin@intel.com>
Tue, 18 Jan 2022 10:54:02 +0000 (10:54 +0000)
Maarten needs backmerge to account for header file renames/changes which
landed via drm-intel-next and are interfering with his pinning work.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
17 files changed:
1  2 
drivers/gpu/drm/i915/Makefile
drivers/gpu/drm/i915/display/intel_dpt.c
drivers/gpu/drm/i915/display/intel_fbc.c
drivers/gpu/drm/i915/display/intel_plane_initial.c
drivers/gpu/drm/i915/gem/i915_gem_context.c
drivers/gpu/drm/i915/gem/i915_gem_mman.c
drivers/gpu/drm/i915/gt/intel_ggtt.c
drivers/gpu/drm/i915/gt/intel_gt.c
drivers/gpu/drm/i915/gt/intel_gtt.c
drivers/gpu/drm/i915/gt/intel_gtt.h
drivers/gpu/drm/i915/gt/uc/intel_uc.c
drivers/gpu/drm/i915/i915_debugfs.c
drivers/gpu/drm/i915/i915_driver.c
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_module.c
drivers/gpu/drm/i915/i915_request.c
drivers/gpu/drm/i915/i915_vma_types.h

Simple merge
index 9208d52e3cecf98d20e123fe1ac37e187de4f2a9,160fd2bdafe59bdb048f363acb4f02385a73ea1b..c0a973eeb4059a6f6ee341e796a49f058c626571
@@@ -259,168 -423,215 +423,215 @@@ static void g4x_fbc_deactivate(struct i
        }
  }
  
- static bool g4x_fbc_is_active(struct drm_i915_private *dev_priv)
+ static bool g4x_fbc_is_active(struct intel_fbc *fbc)
  {
-       return intel_de_read(dev_priv, DPFC_CONTROL) & DPFC_CTL_EN;
+       return intel_de_read(fbc->i915, DPFC_CONTROL) & DPFC_CTL_EN;
  }
  
- static void i8xx_fbc_recompress(struct drm_i915_private *dev_priv)
+ static bool g4x_fbc_is_compressing(struct intel_fbc *fbc)
  {
-       struct intel_fbc_reg_params *params = &dev_priv->fbc.params;
-       enum i9xx_plane_id i9xx_plane = params->crtc.i9xx_plane;
+       return intel_de_read(fbc->i915, DPFC_STATUS) & DPFC_COMP_SEG_MASK;
+ }
  
-       spin_lock_irq(&dev_priv->uncore.lock);
-       intel_de_write_fw(dev_priv, DSPADDR(i9xx_plane),
-                         intel_de_read_fw(dev_priv, DSPADDR(i9xx_plane)));
-       spin_unlock_irq(&dev_priv->uncore.lock);
+ static void g4x_fbc_program_cfb(struct intel_fbc *fbc)
+ {
+       struct drm_i915_private *i915 = fbc->i915;
+       intel_de_write(i915, DPFC_CB_BASE, fbc->compressed_fb.start);
  }
  
- static void i965_fbc_recompress(struct drm_i915_private *dev_priv)
+ static const struct intel_fbc_funcs g4x_fbc_funcs = {
+       .activate = g4x_fbc_activate,
+       .deactivate = g4x_fbc_deactivate,
+       .is_active = g4x_fbc_is_active,
+       .is_compressing = g4x_fbc_is_compressing,
+       .nuke = i965_fbc_nuke,
+       .program_cfb = g4x_fbc_program_cfb,
+ };
+ static void ilk_fbc_activate(struct intel_fbc *fbc)
  {
-       struct intel_fbc_reg_params *params = &dev_priv->fbc.params;
-       enum i9xx_plane_id i9xx_plane = params->crtc.i9xx_plane;
+       struct intel_fbc_state *fbc_state = &fbc->state;
+       struct drm_i915_private *i915 = fbc->i915;
  
-       spin_lock_irq(&dev_priv->uncore.lock);
-       intel_de_write_fw(dev_priv, DSPSURF(i9xx_plane),
-                         intel_de_read_fw(dev_priv, DSPSURF(i9xx_plane)));
-       spin_unlock_irq(&dev_priv->uncore.lock);
+       intel_de_write(i915, ILK_DPFC_FENCE_YOFF,
+                      fbc_state->fence_y_offset);
+       intel_de_write(i915, ILK_DPFC_CONTROL,
+                      DPFC_CTL_EN | g4x_dpfc_ctl(fbc));
  }
  
- /* This function forces a CFB recompression through the nuke operation. */
- static void snb_fbc_recompress(struct drm_i915_private *dev_priv)
+ static void ilk_fbc_deactivate(struct intel_fbc *fbc)
  {
-       intel_de_write(dev_priv, MSG_FBC_REND_STATE, FBC_REND_NUKE);
-       intel_de_posting_read(dev_priv, MSG_FBC_REND_STATE);
+       struct drm_i915_private *i915 = fbc->i915;
+       u32 dpfc_ctl;
+       /* Disable compression */
+       dpfc_ctl = intel_de_read(i915, ILK_DPFC_CONTROL);
+       if (dpfc_ctl & DPFC_CTL_EN) {
+               dpfc_ctl &= ~DPFC_CTL_EN;
+               intel_de_write(i915, ILK_DPFC_CONTROL, dpfc_ctl);
+       }
+ }
+ static bool ilk_fbc_is_active(struct intel_fbc *fbc)
+ {
+       return intel_de_read(fbc->i915, ILK_DPFC_CONTROL) & DPFC_CTL_EN;
  }
  
- static void intel_fbc_recompress(struct drm_i915_private *dev_priv)
+ static bool ilk_fbc_is_compressing(struct intel_fbc *fbc)
  {
-       struct intel_fbc *fbc = &dev_priv->fbc;
+       return intel_de_read(fbc->i915, ILK_DPFC_STATUS) & DPFC_COMP_SEG_MASK;
+ }
  
-       trace_intel_fbc_nuke(fbc->crtc);
+ static void ilk_fbc_program_cfb(struct intel_fbc *fbc)
+ {
+       struct drm_i915_private *i915 = fbc->i915;
  
-       if (DISPLAY_VER(dev_priv) >= 6)
-               snb_fbc_recompress(dev_priv);
-       else if (DISPLAY_VER(dev_priv) >= 4)
-               i965_fbc_recompress(dev_priv);
-       else
-               i8xx_fbc_recompress(dev_priv);
+       intel_de_write(i915, ILK_DPFC_CB_BASE, fbc->compressed_fb.start);
  }
  
- static void ilk_fbc_activate(struct drm_i915_private *dev_priv)
+ static const struct intel_fbc_funcs ilk_fbc_funcs = {
+       .activate = ilk_fbc_activate,
+       .deactivate = ilk_fbc_deactivate,
+       .is_active = ilk_fbc_is_active,
+       .is_compressing = ilk_fbc_is_compressing,
+       .nuke = i965_fbc_nuke,
+       .program_cfb = ilk_fbc_program_cfb,
+ };
+ static void snb_fbc_program_fence(struct intel_fbc *fbc)
  {
-       struct intel_fbc_reg_params *params = &dev_priv->fbc.params;
-       u32 dpfc_ctl;
+       const struct intel_fbc_state *fbc_state = &fbc->state;
+       struct drm_i915_private *i915 = fbc->i915;
+       u32 ctl = 0;
  
-       dpfc_ctl = DPFC_CTL_PLANE(params->crtc.i9xx_plane);
+       if (fbc_state->fence_id >= 0)
+               ctl = SNB_DPFC_FENCE_EN | SNB_DPFC_FENCENO(fbc_state->fence_id);
  
-       dpfc_ctl |= g4x_dpfc_ctl_limit(dev_priv);
+       intel_de_write(i915, SNB_DPFC_CTL_SA, ctl);
+       intel_de_write(i915, SNB_DPFC_CPU_FENCE_OFFSET, fbc_state->fence_y_offset);
+ }
  
-       if (params->fence_id >= 0) {
-               dpfc_ctl |= DPFC_CTL_FENCE_EN;
-               if (IS_IRONLAKE(dev_priv))
-                       dpfc_ctl |= params->fence_id;
-               if (IS_SANDYBRIDGE(dev_priv)) {
-                       intel_de_write(dev_priv, SNB_DPFC_CTL_SA,
-                                      SNB_CPU_FENCE_ENABLE | params->fence_id);
-                       intel_de_write(dev_priv, DPFC_CPU_FENCE_OFFSET,
-                                      params->fence_y_offset);
-               }
-       } else {
-               if (IS_SANDYBRIDGE(dev_priv)) {
-                       intel_de_write(dev_priv, SNB_DPFC_CTL_SA, 0);
-                       intel_de_write(dev_priv, DPFC_CPU_FENCE_OFFSET, 0);
-               }
-       }
+ static void snb_fbc_activate(struct intel_fbc *fbc)
+ {
+       snb_fbc_program_fence(fbc);
  
-       intel_de_write(dev_priv, ILK_DPFC_FENCE_YOFF,
-                      params->fence_y_offset);
-       /* enable it... */
-       intel_de_write(dev_priv, ILK_DPFC_CONTROL, dpfc_ctl | DPFC_CTL_EN);
+       ilk_fbc_activate(fbc);
  }
  
- static void ilk_fbc_deactivate(struct drm_i915_private *dev_priv)
+ static void snb_fbc_nuke(struct intel_fbc *fbc)
  {
-       u32 dpfc_ctl;
+       struct drm_i915_private *i915 = fbc->i915;
  
-       /* Disable compression */
-       dpfc_ctl = intel_de_read(dev_priv, ILK_DPFC_CONTROL);
-       if (dpfc_ctl & DPFC_CTL_EN) {
-               dpfc_ctl &= ~DPFC_CTL_EN;
-               intel_de_write(dev_priv, ILK_DPFC_CONTROL, dpfc_ctl);
-       }
+       intel_de_write(i915, MSG_FBC_REND_STATE, FBC_REND_NUKE);
+       intel_de_posting_read(i915, MSG_FBC_REND_STATE);
+ }
+ static const struct intel_fbc_funcs snb_fbc_funcs = {
+       .activate = snb_fbc_activate,
+       .deactivate = ilk_fbc_deactivate,
+       .is_active = ilk_fbc_is_active,
+       .is_compressing = ilk_fbc_is_compressing,
+       .nuke = snb_fbc_nuke,
+       .program_cfb = ilk_fbc_program_cfb,
+ };
+ static void glk_fbc_program_cfb_stride(struct intel_fbc *fbc)
+ {
+       const struct intel_fbc_state *fbc_state = &fbc->state;
+       struct drm_i915_private *i915 = fbc->i915;
+       u32 val = 0;
+       if (fbc_state->override_cfb_stride)
+               val |= FBC_STRIDE_OVERRIDE |
+                       FBC_STRIDE(fbc_state->override_cfb_stride / fbc->limit);
+       intel_de_write(i915, GLK_FBC_STRIDE, val);
  }
  
- static bool ilk_fbc_is_active(struct drm_i915_private *dev_priv)
+ static void skl_fbc_program_cfb_stride(struct intel_fbc *fbc)
  {
-       return intel_de_read(dev_priv, ILK_DPFC_CONTROL) & DPFC_CTL_EN;
+       const struct intel_fbc_state *fbc_state = &fbc->state;
+       struct drm_i915_private *i915 = fbc->i915;
+       u32 val = 0;
+       /* Display WA #0529: skl, kbl, bxt. */
+       if (fbc_state->override_cfb_stride)
+               val |= CHICKEN_FBC_STRIDE_OVERRIDE |
+                       CHICKEN_FBC_STRIDE(fbc_state->override_cfb_stride / fbc->limit);
+       intel_de_rmw(i915, CHICKEN_MISC_4,
+                    CHICKEN_FBC_STRIDE_OVERRIDE |
+                    CHICKEN_FBC_STRIDE_MASK, val);
  }
  
- static void gen7_fbc_activate(struct drm_i915_private *dev_priv)
+ static u32 ivb_dpfc_ctl(struct intel_fbc *fbc)
  {
-       struct intel_fbc *fbc = &dev_priv->fbc;
-       const struct intel_fbc_reg_params *params = &fbc->params;
+       const struct intel_fbc_state *fbc_state = &fbc->state;
+       struct drm_i915_private *i915 = fbc->i915;
        u32 dpfc_ctl;
  
-       if (DISPLAY_VER(dev_priv) >= 10) {
-               u32 val = 0;
+       dpfc_ctl = g4x_dpfc_ctl_limit(fbc);
  
-               if (params->override_cfb_stride)
-                       val |= FBC_STRIDE_OVERRIDE |
-                               FBC_STRIDE(params->override_cfb_stride / fbc->limit);
+       if (IS_IVYBRIDGE(i915))
+               dpfc_ctl |= DPFC_CTL_PLANE_IVB(fbc_state->plane->i9xx_plane);
  
-               intel_de_write(dev_priv, GLK_FBC_STRIDE, val);
-       } else if (DISPLAY_VER(dev_priv) == 9) {
-               u32 val = 0;
+       if (fbc_state->fence_id >= 0)
+               dpfc_ctl |= DPFC_CTL_FENCE_EN_IVB;
  
-               /* Display WA #0529: skl, kbl, bxt. */
-               if (params->override_cfb_stride)
-                       val |= CHICKEN_FBC_STRIDE_OVERRIDE |
-                               CHICKEN_FBC_STRIDE(params->override_cfb_stride / fbc->limit);
+       if (fbc->false_color)
+               dpfc_ctl |= DPFC_CTL_FALSE_COLOR;
  
-               intel_de_rmw(dev_priv, CHICKEN_MISC_4,
-                            CHICKEN_FBC_STRIDE_OVERRIDE |
-                            CHICKEN_FBC_STRIDE_MASK, val);
-       }
+       return dpfc_ctl;
+ }
  
-       dpfc_ctl = 0;
-       if (IS_IVYBRIDGE(dev_priv))
-               dpfc_ctl |= IVB_DPFC_CTL_PLANE(params->crtc.i9xx_plane);
-       dpfc_ctl |= g4x_dpfc_ctl_limit(dev_priv);
-       if (params->fence_id >= 0) {
-               dpfc_ctl |= IVB_DPFC_CTL_FENCE_EN;
-               intel_de_write(dev_priv, SNB_DPFC_CTL_SA,
-                              SNB_CPU_FENCE_ENABLE | params->fence_id);
-               intel_de_write(dev_priv, DPFC_CPU_FENCE_OFFSET,
-                              params->fence_y_offset);
-       } else if (to_gt(dev_priv)->ggtt->num_fences) {
-               intel_de_write(dev_priv, SNB_DPFC_CTL_SA, 0);
-               intel_de_write(dev_priv, DPFC_CPU_FENCE_OFFSET, 0);
-       }
+ static void ivb_fbc_activate(struct intel_fbc *fbc)
+ {
+       struct drm_i915_private *i915 = fbc->i915;
+       if (DISPLAY_VER(i915) >= 10)
+               glk_fbc_program_cfb_stride(fbc);
+       else if (DISPLAY_VER(i915) == 9)
+               skl_fbc_program_cfb_stride(fbc);
  
-       if (dev_priv->fbc.false_color)
-               dpfc_ctl |= FBC_CTL_FALSE_COLOR;
 -      if (i915->ggtt.num_fences)
++      if (to_gt(i915)->ggtt->num_fences)
+               snb_fbc_program_fence(fbc);
  
-       intel_de_write(dev_priv, ILK_DPFC_CONTROL, dpfc_ctl | DPFC_CTL_EN);
+       intel_de_write(i915, ILK_DPFC_CONTROL,
+                      DPFC_CTL_EN | ivb_dpfc_ctl(fbc));
  }
  
- static bool intel_fbc_hw_is_active(struct drm_i915_private *dev_priv)
+ static bool ivb_fbc_is_compressing(struct intel_fbc *fbc)
  {
-       if (DISPLAY_VER(dev_priv) >= 5)
-               return ilk_fbc_is_active(dev_priv);
-       else if (IS_GM45(dev_priv))
-               return g4x_fbc_is_active(dev_priv);
-       else
-               return i8xx_fbc_is_active(dev_priv);
+       return intel_de_read(fbc->i915, ILK_DPFC_STATUS2) & DPFC_COMP_SEG_MASK_IVB;
  }
  
- static void intel_fbc_hw_activate(struct drm_i915_private *dev_priv)
+ static void ivb_fbc_set_false_color(struct intel_fbc *fbc,
+                                   bool enable)
  {
-       struct intel_fbc *fbc = &dev_priv->fbc;
+       intel_de_rmw(fbc->i915, ILK_DPFC_CONTROL,
+                    DPFC_CTL_FALSE_COLOR, enable ? DPFC_CTL_FALSE_COLOR : 0);
+ }
  
-       trace_intel_fbc_activate(fbc->crtc);
+ static const struct intel_fbc_funcs ivb_fbc_funcs = {
+       .activate = ivb_fbc_activate,
+       .deactivate = ilk_fbc_deactivate,
+       .is_active = ilk_fbc_is_active,
+       .is_compressing = ivb_fbc_is_compressing,
+       .nuke = snb_fbc_nuke,
+       .program_cfb = ilk_fbc_program_cfb,
+       .set_false_color = ivb_fbc_set_false_color,
+ };
+ static bool intel_fbc_hw_is_active(struct intel_fbc *fbc)
+ {
+       return fbc->funcs->is_active(fbc);
+ }
+ static void intel_fbc_hw_activate(struct intel_fbc *fbc)
+ {
+       trace_intel_fbc_activate(fbc->state.plane);
  
        fbc->active = true;
        fbc->activated = true;
index 91ea2882efda8c2aa4010fbf6281ca77858501fa,5263dda7f8d52dd7b4442edbf6ebee179c7392c3..a1b2761bc16e1d371f3a5de87fc12848be6208b5
@@@ -1257,7 -1280,7 +1284,7 @@@ bool i915_ggtt_resume_vm(struct i915_ad
                        atomic_read(&vma->flags) & I915_VMA_BIND_MASK;
  
                GEM_BUG_ON(!was_bound);
-               vma->ops->bind_vma(&ggtt->vm, NULL, vma->resource,
 -              vma->ops->bind_vma(vm, NULL, vma,
++              vma->ops->bind_vma(vm, NULL, vma->resource,
                                   obj ? obj->cache_level : 0,
                                   was_bound);
                if (obj) { /* only used during resume => exclusive access */
index 3ebf8136e44bd92e8c3b69ad82ccf64d5670363e,f98f0fb21efbe69b3095402c6655e05b7e378879..298ff32c8d0c13ba608b5bb00b5a7f76326cb42d
@@@ -3,7 -3,7 +3,8 @@@
   * Copyright Â© 2019 Intel Corporation
   */
  
 +#include <drm/drm_managed.h>
+ #include <drm/intel-gtt.h>
  
  #include "intel_gt_debugfs.h"
  
Simple merge
Simple merge
Simple merge
Simple merge
index 0000000000000000000000000000000000000000,95174938b16024197c1d0d6e54c2e416f63be729..5f2343389b5e909dbbb5170d6a0e1413e28a1e0f
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1850 +1,1852 @@@
 -      intel_gt_init_hw_early(to_gt(dev_priv), &dev_priv->ggtt);
 -
+ /* i915_drv.c -- i830,i845,i855,i865,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.
+  *
+  */
+ #include <linux/acpi.h>
+ #include <linux/device.h>
+ #include <linux/module.h>
+ #include <linux/oom.h>
+ #include <linux/pci.h>
+ #include <linux/pm.h>
+ #include <linux/pm_runtime.h>
+ #include <linux/pnp.h>
+ #include <linux/slab.h>
+ #include <linux/vga_switcheroo.h>
+ #include <linux/vt.h>
+ #include <drm/drm_aperture.h>
+ #include <drm/drm_atomic_helper.h>
+ #include <drm/drm_ioctl.h>
+ #include <drm/drm_managed.h>
+ #include <drm/drm_probe_helper.h>
+ #include "display/intel_acpi.h"
+ #include "display/intel_bw.h"
+ #include "display/intel_cdclk.h"
+ #include "display/intel_display_types.h"
+ #include "display/intel_dmc.h"
+ #include "display/intel_dp.h"
+ #include "display/intel_dpt.h"
+ #include "display/intel_fbdev.h"
+ #include "display/intel_hotplug.h"
+ #include "display/intel_overlay.h"
+ #include "display/intel_pch_refclk.h"
+ #include "display/intel_pipe_crc.h"
+ #include "display/intel_pps.h"
+ #include "display/intel_sprite.h"
+ #include "display/intel_vga.h"
+ #include "gem/i915_gem_context.h"
+ #include "gem/i915_gem_ioctls.h"
+ #include "gem/i915_gem_mman.h"
+ #include "gem/i915_gem_pm.h"
+ #include "gt/intel_gt.h"
+ #include "gt/intel_gt_pm.h"
+ #include "gt/intel_rc6.h"
+ #include "pxp/intel_pxp_pm.h"
+ #include "i915_debugfs.h"
+ #include "i915_driver.h"
+ #include "i915_drv.h"
+ #include "i915_ioc32.h"
+ #include "i915_irq.h"
+ #include "i915_memcpy.h"
+ #include "i915_perf.h"
+ #include "i915_query.h"
+ #include "i915_suspend.h"
+ #include "i915_switcheroo.h"
+ #include "i915_sysfs.h"
+ #include "i915_vgpu.h"
+ #include "intel_dram.h"
+ #include "intel_gvt.h"
+ #include "intel_memory_region.h"
+ #include "intel_pcode.h"
+ #include "intel_pm.h"
+ #include "intel_region_ttm.h"
+ #include "vlv_suspend.h"
+ static const struct drm_driver i915_drm_driver;
+ static int i915_get_bridge_dev(struct drm_i915_private *dev_priv)
+ {
+       int domain = pci_domain_nr(to_pci_dev(dev_priv->drm.dev)->bus);
+       dev_priv->bridge_dev =
+               pci_get_domain_bus_and_slot(domain, 0, PCI_DEVFN(0, 0));
+       if (!dev_priv->bridge_dev) {
+               drm_err(&dev_priv->drm, "bridge device not found\n");
+               return -EIO;
+       }
+       return 0;
+ }
+ /* Allocate space for the MCH regs if needed, return nonzero on error */
+ static int
+ intel_alloc_mchbar_resource(struct drm_i915_private *dev_priv)
+ {
+       int reg = GRAPHICS_VER(dev_priv) >= 4 ? MCHBAR_I965 : MCHBAR_I915;
+       u32 temp_lo, temp_hi = 0;
+       u64 mchbar_addr;
+       int ret;
+       if (GRAPHICS_VER(dev_priv) >= 4)
+               pci_read_config_dword(dev_priv->bridge_dev, reg + 4, &temp_hi);
+       pci_read_config_dword(dev_priv->bridge_dev, reg, &temp_lo);
+       mchbar_addr = ((u64)temp_hi << 32) | temp_lo;
+       /* If ACPI doesn't have it, assume we need to allocate it ourselves */
+ #ifdef CONFIG_PNP
+       if (mchbar_addr &&
+           pnp_range_reserved(mchbar_addr, mchbar_addr + MCHBAR_SIZE))
+               return 0;
+ #endif
+       /* Get some space for it */
+       dev_priv->mch_res.name = "i915 MCHBAR";
+       dev_priv->mch_res.flags = IORESOURCE_MEM;
+       ret = pci_bus_alloc_resource(dev_priv->bridge_dev->bus,
+                                    &dev_priv->mch_res,
+                                    MCHBAR_SIZE, MCHBAR_SIZE,
+                                    PCIBIOS_MIN_MEM,
+                                    0, pcibios_align_resource,
+                                    dev_priv->bridge_dev);
+       if (ret) {
+               drm_dbg(&dev_priv->drm, "failed bus alloc: %d\n", ret);
+               dev_priv->mch_res.start = 0;
+               return ret;
+       }
+       if (GRAPHICS_VER(dev_priv) >= 4)
+               pci_write_config_dword(dev_priv->bridge_dev, reg + 4,
+                                      upper_32_bits(dev_priv->mch_res.start));
+       pci_write_config_dword(dev_priv->bridge_dev, reg,
+                              lower_32_bits(dev_priv->mch_res.start));
+       return 0;
+ }
+ /* Setup MCHBAR if possible, return true if we should disable it again */
+ static void
+ intel_setup_mchbar(struct drm_i915_private *dev_priv)
+ {
+       int mchbar_reg = GRAPHICS_VER(dev_priv) >= 4 ? MCHBAR_I965 : MCHBAR_I915;
+       u32 temp;
+       bool enabled;
+       if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv))
+               return;
+       dev_priv->mchbar_need_disable = false;
+       if (IS_I915G(dev_priv) || IS_I915GM(dev_priv)) {
+               pci_read_config_dword(dev_priv->bridge_dev, DEVEN, &temp);
+               enabled = !!(temp & DEVEN_MCHBAR_EN);
+       } else {
+               pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
+               enabled = temp & 1;
+       }
+       /* If it's already enabled, don't have to do anything */
+       if (enabled)
+               return;
+       if (intel_alloc_mchbar_resource(dev_priv))
+               return;
+       dev_priv->mchbar_need_disable = true;
+       /* Space is allocated or reserved, so enable it. */
+       if (IS_I915G(dev_priv) || IS_I915GM(dev_priv)) {
+               pci_write_config_dword(dev_priv->bridge_dev, DEVEN,
+                                      temp | DEVEN_MCHBAR_EN);
+       } else {
+               pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg, &temp);
+               pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg, temp | 1);
+       }
+ }
+ static void
+ intel_teardown_mchbar(struct drm_i915_private *dev_priv)
+ {
+       int mchbar_reg = GRAPHICS_VER(dev_priv) >= 4 ? MCHBAR_I965 : MCHBAR_I915;
+       if (dev_priv->mchbar_need_disable) {
+               if (IS_I915G(dev_priv) || IS_I915GM(dev_priv)) {
+                       u32 deven_val;
+                       pci_read_config_dword(dev_priv->bridge_dev, DEVEN,
+                                             &deven_val);
+                       deven_val &= ~DEVEN_MCHBAR_EN;
+                       pci_write_config_dword(dev_priv->bridge_dev, DEVEN,
+                                              deven_val);
+               } else {
+                       u32 mchbar_val;
+                       pci_read_config_dword(dev_priv->bridge_dev, mchbar_reg,
+                                             &mchbar_val);
+                       mchbar_val &= ~1;
+                       pci_write_config_dword(dev_priv->bridge_dev, mchbar_reg,
+                                              mchbar_val);
+               }
+       }
+       if (dev_priv->mch_res.start)
+               release_resource(&dev_priv->mch_res);
+ }
+ static int i915_workqueues_init(struct drm_i915_private *dev_priv)
+ {
+       /*
+        * The i915 workqueue is primarily used for batched retirement of
+        * requests (and thus managing bo) once the task has been completed
+        * by the GPU. i915_retire_requests() is called directly when we
+        * need high-priority retirement, such as waiting for an explicit
+        * bo.
+        *
+        * It is also used for periodic low-priority events, such as
+        * idle-timers and recording error state.
+        *
+        * All tasks on the workqueue are expected to acquire the dev mutex
+        * so there is no point in running more than one instance of the
+        * workqueue at any time.  Use an ordered one.
+        */
+       dev_priv->wq = alloc_ordered_workqueue("i915", 0);
+       if (dev_priv->wq == NULL)
+               goto out_err;
+       dev_priv->hotplug.dp_wq = alloc_ordered_workqueue("i915-dp", 0);
+       if (dev_priv->hotplug.dp_wq == NULL)
+               goto out_free_wq;
+       return 0;
+ out_free_wq:
+       destroy_workqueue(dev_priv->wq);
+ out_err:
+       drm_err(&dev_priv->drm, "Failed to allocate workqueues.\n");
+       return -ENOMEM;
+ }
+ static void i915_workqueues_cleanup(struct drm_i915_private *dev_priv)
+ {
+       destroy_workqueue(dev_priv->hotplug.dp_wq);
+       destroy_workqueue(dev_priv->wq);
+ }
+ /*
+  * We don't keep the workarounds for pre-production hardware, so we expect our
+  * driver to fail on these machines in one way or another. A little warning on
+  * dmesg may help both the user and the bug triagers.
+  *
+  * Our policy for removing pre-production workarounds is to keep the
+  * current gen workarounds as a guide to the bring-up of the next gen
+  * (workarounds have a habit of persisting!). Anything older than that
+  * should be removed along with the complications they introduce.
+  */
+ static void intel_detect_preproduction_hw(struct drm_i915_private *dev_priv)
+ {
+       bool pre = false;
+       pre |= IS_HSW_EARLY_SDV(dev_priv);
+       pre |= IS_SKYLAKE(dev_priv) && INTEL_REVID(dev_priv) < 0x6;
+       pre |= IS_BROXTON(dev_priv) && INTEL_REVID(dev_priv) < 0xA;
+       pre |= IS_KABYLAKE(dev_priv) && INTEL_REVID(dev_priv) < 0x1;
+       pre |= IS_GEMINILAKE(dev_priv) && INTEL_REVID(dev_priv) < 0x3;
+       pre |= IS_ICELAKE(dev_priv) && INTEL_REVID(dev_priv) < 0x7;
+       if (pre) {
+               drm_err(&dev_priv->drm, "This is a pre-production stepping. "
+                         "It may not be fully functional.\n");
+               add_taint(TAINT_MACHINE_CHECK, LOCKDEP_STILL_OK);
+       }
+ }
+ static void sanitize_gpu(struct drm_i915_private *i915)
+ {
+       if (!INTEL_INFO(i915)->gpu_reset_clobbers_display)
+               __intel_gt_reset(to_gt(i915), ALL_ENGINES);
+ }
+ /**
+  * i915_driver_early_probe - setup state not requiring device access
+  * @dev_priv: device private
+  *
+  * Initialize everything that is a "SW-only" state, that is state not
+  * requiring accessing the device or exposing the driver via kernel internal
+  * or userspace interfaces. Example steps belonging here: lock initialization,
+  * system memory allocation, setting up device specific attributes and
+  * function hooks not requiring accessing the device.
+  */
+ static int i915_driver_early_probe(struct drm_i915_private *dev_priv)
+ {
+       int ret = 0;
+       if (i915_inject_probe_failure(dev_priv))
+               return -ENODEV;
+       intel_device_info_subplatform_init(dev_priv);
+       intel_step_init(dev_priv);
+       intel_gt_init_early(to_gt(dev_priv), dev_priv);
+       intel_uncore_mmio_debug_init_early(&dev_priv->mmio_debug);
+       intel_uncore_init_early(&dev_priv->uncore, to_gt(dev_priv));
+       spin_lock_init(&dev_priv->irq_lock);
+       spin_lock_init(&dev_priv->gpu_error.lock);
+       mutex_init(&dev_priv->backlight_lock);
+       mutex_init(&dev_priv->sb_lock);
+       cpu_latency_qos_add_request(&dev_priv->sb_qos, PM_QOS_DEFAULT_VALUE);
+       mutex_init(&dev_priv->audio.mutex);
+       mutex_init(&dev_priv->wm.wm_mutex);
+       mutex_init(&dev_priv->pps_mutex);
+       mutex_init(&dev_priv->hdcp_comp_mutex);
+       i915_memcpy_init_early(dev_priv);
+       intel_runtime_pm_init_early(&dev_priv->runtime_pm);
+       ret = i915_workqueues_init(dev_priv);
+       if (ret < 0)
+               return ret;
+       ret = vlv_suspend_init(dev_priv);
+       if (ret < 0)
+               goto err_workqueues;
+       ret = intel_region_ttm_device_init(dev_priv);
+       if (ret)
+               goto err_ttm;
+       intel_wopcm_init_early(&dev_priv->wopcm);
+       __intel_gt_init_early(to_gt(dev_priv), dev_priv);
+       i915_gem_init_early(dev_priv);
+       /* This must be called before any calls to HAS_PCH_* */
+       intel_detect_pch(dev_priv);
+       intel_pm_setup(dev_priv);
+       ret = intel_power_domains_init(dev_priv);
+       if (ret < 0)
+               goto err_gem;
+       intel_irq_init(dev_priv);
+       intel_init_display_hooks(dev_priv);
+       intel_init_clock_gating_hooks(dev_priv);
+       intel_detect_preproduction_hw(dev_priv);
+       return 0;
+ err_gem:
+       i915_gem_cleanup_early(dev_priv);
+       intel_gt_driver_late_release(to_gt(dev_priv));
+       intel_region_ttm_device_fini(dev_priv);
+ err_ttm:
+       vlv_suspend_cleanup(dev_priv);
+ err_workqueues:
+       i915_workqueues_cleanup(dev_priv);
+       return ret;
+ }
+ /**
+  * i915_driver_late_release - cleanup the setup done in
+  *                           i915_driver_early_probe()
+  * @dev_priv: device private
+  */
+ static void i915_driver_late_release(struct drm_i915_private *dev_priv)
+ {
+       intel_irq_fini(dev_priv);
+       intel_power_domains_cleanup(dev_priv);
+       i915_gem_cleanup_early(dev_priv);
+       intel_gt_driver_late_release(to_gt(dev_priv));
+       intel_region_ttm_device_fini(dev_priv);
+       vlv_suspend_cleanup(dev_priv);
+       i915_workqueues_cleanup(dev_priv);
+       cpu_latency_qos_remove_request(&dev_priv->sb_qos);
+       mutex_destroy(&dev_priv->sb_lock);
+       i915_params_free(&dev_priv->params);
+ }
+ /**
+  * i915_driver_mmio_probe - setup device MMIO
+  * @dev_priv: device private
+  *
+  * Setup minimal device state necessary for MMIO accesses later in the
+  * initialization sequence. The setup here should avoid any other device-wide
+  * side effects or exposing the driver via kernel internal or user space
+  * interfaces.
+  */
+ static int i915_driver_mmio_probe(struct drm_i915_private *dev_priv)
+ {
+       int ret;
+       if (i915_inject_probe_failure(dev_priv))
+               return -ENODEV;
+       ret = i915_get_bridge_dev(dev_priv);
+       if (ret < 0)
+               return ret;
+       ret = intel_uncore_setup_mmio(&dev_priv->uncore);
+       if (ret < 0)
+               goto err_bridge;
+       ret = intel_uncore_init_mmio(&dev_priv->uncore);
+       if (ret)
+               goto err_mmio;
+       /* Try to make sure MCHBAR is enabled before poking at it */
+       intel_setup_mchbar(dev_priv);
+       intel_device_info_runtime_init(dev_priv);
+       ret = intel_gt_init_mmio(to_gt(dev_priv));
+       if (ret)
+               goto err_uncore;
+       /* As early as possible, scrub existing GPU state before clobbering */
+       sanitize_gpu(dev_priv);
+       return 0;
+ err_uncore:
+       intel_teardown_mchbar(dev_priv);
+       intel_uncore_fini_mmio(&dev_priv->uncore);
+ err_mmio:
+       intel_uncore_cleanup_mmio(&dev_priv->uncore);
+ err_bridge:
+       pci_dev_put(dev_priv->bridge_dev);
+       return ret;
+ }
+ /**
+  * i915_driver_mmio_release - cleanup the setup done in i915_driver_mmio_probe()
+  * @dev_priv: device private
+  */
+ static void i915_driver_mmio_release(struct drm_i915_private *dev_priv)
+ {
+       intel_teardown_mchbar(dev_priv);
+       intel_uncore_fini_mmio(&dev_priv->uncore);
+       intel_uncore_cleanup_mmio(&dev_priv->uncore);
+       pci_dev_put(dev_priv->bridge_dev);
+ }
+ static void intel_sanitize_options(struct drm_i915_private *dev_priv)
+ {
+       intel_gvt_sanitize_options(dev_priv);
+ }
+ /**
+  * i915_set_dma_info - set all relevant PCI dma info as configured for the
+  * platform
+  * @i915: valid i915 instance
+  *
+  * Set the dma max segment size, device and coherent masks.  The dma mask set
+  * needs to occur before i915_ggtt_probe_hw.
+  *
+  * A couple of platforms have special needs.  Address them as well.
+  *
+  */
+ static int i915_set_dma_info(struct drm_i915_private *i915)
+ {
+       unsigned int mask_size = INTEL_INFO(i915)->dma_mask_size;
+       int ret;
+       GEM_BUG_ON(!mask_size);
+       /*
+        * We don't have a max segment size, so set it to the max so sg's
+        * debugging layer doesn't complain
+        */
+       dma_set_max_seg_size(i915->drm.dev, UINT_MAX);
+       ret = dma_set_mask(i915->drm.dev, DMA_BIT_MASK(mask_size));
+       if (ret)
+               goto mask_err;
+       /* overlay on gen2 is broken and can't address above 1G */
+       if (GRAPHICS_VER(i915) == 2)
+               mask_size = 30;
+       /*
+        * 965GM sometimes incorrectly writes to hardware status page (HWS)
+        * using 32bit addressing, overwriting memory if HWS is located
+        * above 4GB.
+        *
+        * The documentation also mentions an issue with undefined
+        * behaviour if any general state is accessed within a page above 4GB,
+        * which also needs to be handled carefully.
+        */
+       if (IS_I965G(i915) || IS_I965GM(i915))
+               mask_size = 32;
+       ret = dma_set_coherent_mask(i915->drm.dev, DMA_BIT_MASK(mask_size));
+       if (ret)
+               goto mask_err;
+       return 0;
+ mask_err:
+       drm_err(&i915->drm, "Can't set DMA mask/consistent mask (%d)\n", ret);
+       return ret;
+ }
+ /**
+  * i915_driver_hw_probe - setup state requiring device access
+  * @dev_priv: device private
+  *
+  * Setup state that requires accessing the device, but doesn't require
+  * exposing the driver via kernel internal or userspace interfaces.
+  */
+ static int i915_driver_hw_probe(struct drm_i915_private *dev_priv)
+ {
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       int ret;
+       if (i915_inject_probe_failure(dev_priv))
+               return -ENODEV;
+       if (HAS_PPGTT(dev_priv)) {
+               if (intel_vgpu_active(dev_priv) &&
+                   !intel_vgpu_has_full_ppgtt(dev_priv)) {
+                       i915_report_error(dev_priv,
+                                         "incompatible vGPU found, support for isolated ppGTT required\n");
+                       return -ENXIO;
+               }
+       }
+       if (HAS_EXECLISTS(dev_priv)) {
+               /*
+                * Older GVT emulation depends upon intercepting CSB mmio,
+                * which we no longer use, preferring to use the HWSP cache
+                * instead.
+                */
+               if (intel_vgpu_active(dev_priv) &&
+                   !intel_vgpu_has_hwsp_emulation(dev_priv)) {
+                       i915_report_error(dev_priv,
+                                         "old vGPU host found, support for HWSP emulation required\n");
+                       return -ENXIO;
+               }
+       }
+       intel_sanitize_options(dev_priv);
+       /* needs to be done before ggtt probe */
+       intel_dram_edram_detect(dev_priv);
+       ret = i915_set_dma_info(dev_priv);
+       if (ret)
+               return ret;
+       i915_perf_init(dev_priv);
++      ret = intel_gt_assign_ggtt(to_gt(dev_priv));
++      if (ret)
++              goto err_perf;
++
+       ret = i915_ggtt_probe_hw(dev_priv);
+       if (ret)
+               goto err_perf;
+       ret = drm_aperture_remove_conflicting_pci_framebuffers(pdev, dev_priv->drm.driver);
+       if (ret)
+               goto err_ggtt;
+       ret = i915_ggtt_init_hw(dev_priv);
+       if (ret)
+               goto err_ggtt;
+       ret = intel_memory_regions_hw_probe(dev_priv);
+       if (ret)
+               goto err_ggtt;
 -      i915_ggtt_suspend(&dev_priv->ggtt);
+       ret = intel_gt_probe_lmem(to_gt(dev_priv));
+       if (ret)
+               goto err_mem_regions;
+       ret = i915_ggtt_enable_hw(dev_priv);
+       if (ret) {
+               drm_err(&dev_priv->drm, "failed to enable GGTT\n");
+               goto err_mem_regions;
+       }
+       pci_set_master(pdev);
+       /* On the 945G/GM, the chipset reports the MSI capability on the
+        * integrated graphics even though the support isn't actually there
+        * according to the published specs.  It doesn't appear to function
+        * correctly in testing on 945G.
+        * This may be a side effect of MSI having been made available for PEG
+        * and the registers being closely associated.
+        *
+        * According to chipset errata, on the 965GM, MSI interrupts may
+        * be lost or delayed, and was defeatured. MSI interrupts seem to
+        * get lost on g4x as well, and interrupt delivery seems to stay
+        * properly dead afterwards. So we'll just disable them for all
+        * pre-gen5 chipsets.
+        *
+        * dp aux and gmbus irq on gen4 seems to be able to generate legacy
+        * interrupts even when in MSI mode. This results in spurious
+        * interrupt warnings if the legacy irq no. is shared with another
+        * device. The kernel then disables that interrupt source and so
+        * prevents the other device from working properly.
+        */
+       if (GRAPHICS_VER(dev_priv) >= 5) {
+               if (pci_enable_msi(pdev) < 0)
+                       drm_dbg(&dev_priv->drm, "can't enable MSI");
+       }
+       ret = intel_gvt_init(dev_priv);
+       if (ret)
+               goto err_msi;
+       intel_opregion_setup(dev_priv);
+       ret = intel_pcode_init(dev_priv);
+       if (ret)
+               goto err_msi;
+       /*
+        * Fill the dram structure to get the system dram info. This will be
+        * used for memory latency calculation.
+        */
+       intel_dram_detect(dev_priv);
+       intel_bw_init_hw(dev_priv);
+       return 0;
+ err_msi:
+       if (pdev->msi_enabled)
+               pci_disable_msi(pdev);
+ err_mem_regions:
+       intel_memory_regions_driver_release(dev_priv);
+ err_ggtt:
+       i915_ggtt_driver_release(dev_priv);
+       i915_gem_drain_freed_objects(dev_priv);
+       i915_ggtt_driver_late_release(dev_priv);
+ err_perf:
+       i915_perf_fini(dev_priv);
+       return ret;
+ }
+ /**
+  * i915_driver_hw_remove - cleanup the setup done in i915_driver_hw_probe()
+  * @dev_priv: device private
+  */
+ static void i915_driver_hw_remove(struct drm_i915_private *dev_priv)
+ {
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       i915_perf_fini(dev_priv);
+       if (pdev->msi_enabled)
+               pci_disable_msi(pdev);
+ }
+ /**
+  * i915_driver_register - register the driver with the rest of the system
+  * @dev_priv: device private
+  *
+  * Perform any steps necessary to make the driver available via kernel
+  * internal or userspace interfaces.
+  */
+ static void i915_driver_register(struct drm_i915_private *dev_priv)
+ {
+       struct drm_device *dev = &dev_priv->drm;
+       i915_gem_driver_register(dev_priv);
+       i915_pmu_register(dev_priv);
+       intel_vgpu_register(dev_priv);
+       /* Reveal our presence to userspace */
+       if (drm_dev_register(dev, 0)) {
+               drm_err(&dev_priv->drm,
+                       "Failed to register driver for userspace access!\n");
+               return;
+       }
+       i915_debugfs_register(dev_priv);
+       i915_setup_sysfs(dev_priv);
+       /* Depends on sysfs having been initialized */
+       i915_perf_register(dev_priv);
+       intel_gt_driver_register(to_gt(dev_priv));
+       intel_display_driver_register(dev_priv);
+       intel_power_domains_enable(dev_priv);
+       intel_runtime_pm_enable(&dev_priv->runtime_pm);
+       intel_register_dsm_handler();
+       if (i915_switcheroo_register(dev_priv))
+               drm_err(&dev_priv->drm, "Failed to register vga switcheroo!\n");
+ }
+ /**
+  * i915_driver_unregister - cleanup the registration done in i915_driver_regiser()
+  * @dev_priv: device private
+  */
+ static void i915_driver_unregister(struct drm_i915_private *dev_priv)
+ {
+       i915_switcheroo_unregister(dev_priv);
+       intel_unregister_dsm_handler();
+       intel_runtime_pm_disable(&dev_priv->runtime_pm);
+       intel_power_domains_disable(dev_priv);
+       intel_display_driver_unregister(dev_priv);
+       intel_gt_driver_unregister(to_gt(dev_priv));
+       i915_perf_unregister(dev_priv);
+       i915_pmu_unregister(dev_priv);
+       i915_teardown_sysfs(dev_priv);
+       drm_dev_unplug(&dev_priv->drm);
+       i915_gem_driver_unregister(dev_priv);
+ }
+ void
+ i915_print_iommu_status(struct drm_i915_private *i915, struct drm_printer *p)
+ {
+       drm_printf(p, "iommu: %s\n", enableddisabled(intel_vtd_active(i915)));
+ }
+ static void i915_welcome_messages(struct drm_i915_private *dev_priv)
+ {
+       if (drm_debug_enabled(DRM_UT_DRIVER)) {
+               struct drm_printer p = drm_debug_printer("i915 device info:");
+               drm_printf(&p, "pciid=0x%04x rev=0x%02x platform=%s (subplatform=0x%x) gen=%i\n",
+                          INTEL_DEVID(dev_priv),
+                          INTEL_REVID(dev_priv),
+                          intel_platform_name(INTEL_INFO(dev_priv)->platform),
+                          intel_subplatform(RUNTIME_INFO(dev_priv),
+                                            INTEL_INFO(dev_priv)->platform),
+                          GRAPHICS_VER(dev_priv));
+               intel_device_info_print_static(INTEL_INFO(dev_priv), &p);
+               intel_device_info_print_runtime(RUNTIME_INFO(dev_priv), &p);
+               i915_print_iommu_status(dev_priv, &p);
+               intel_gt_info_print(&to_gt(dev_priv)->info, &p);
+       }
+       if (IS_ENABLED(CONFIG_DRM_I915_DEBUG))
+               drm_info(&dev_priv->drm, "DRM_I915_DEBUG enabled\n");
+       if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
+               drm_info(&dev_priv->drm, "DRM_I915_DEBUG_GEM enabled\n");
+       if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_RUNTIME_PM))
+               drm_info(&dev_priv->drm,
+                        "DRM_I915_DEBUG_RUNTIME_PM enabled\n");
+ }
+ static struct drm_i915_private *
+ i915_driver_create(struct pci_dev *pdev, const struct pci_device_id *ent)
+ {
+       const struct intel_device_info *match_info =
+               (struct intel_device_info *)ent->driver_data;
+       struct intel_device_info *device_info;
+       struct drm_i915_private *i915;
+       i915 = devm_drm_dev_alloc(&pdev->dev, &i915_drm_driver,
+                                 struct drm_i915_private, drm);
+       if (IS_ERR(i915))
+               return i915;
+       pci_set_drvdata(pdev, i915);
+       /* Device parameters start as a copy of module parameters. */
+       i915_params_copy(&i915->params, &i915_modparams);
+       /* Setup the write-once "constant" device info */
+       device_info = mkwrite_device_info(i915);
+       memcpy(device_info, match_info, sizeof(*device_info));
+       RUNTIME_INFO(i915)->device_id = pdev->device;
+       return i915;
+ }
+ /**
+  * i915_driver_probe - setup chip and create an initial config
+  * @pdev: PCI device
+  * @ent: matching PCI ID entry
+  *
+  * The driver probe routine has to do several things:
+  *   - drive output discovery via intel_modeset_init()
+  *   - initialize the memory manager
+  *   - allocate initial config memory
+  *   - setup the DRM framebuffer with the allocated memory
+  */
+ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+ {
+       const struct intel_device_info *match_info =
+               (struct intel_device_info *)ent->driver_data;
+       struct drm_i915_private *i915;
+       int ret;
+       i915 = i915_driver_create(pdev, ent);
+       if (IS_ERR(i915))
+               return PTR_ERR(i915);
+       /* Disable nuclear pageflip by default on pre-ILK */
+       if (!i915->params.nuclear_pageflip && match_info->graphics.ver < 5)
+               i915->drm.driver_features &= ~DRIVER_ATOMIC;
+       /*
+        * Check if we support fake LMEM -- for now we only unleash this for
+        * the live selftests(test-and-exit).
+        */
+ #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+       if (IS_ENABLED(CONFIG_DRM_I915_UNSTABLE_FAKE_LMEM)) {
+               if (GRAPHICS_VER(i915) >= 9 && i915_selftest.live < 0 &&
+                   i915->params.fake_lmem_start) {
+                       mkwrite_device_info(i915)->memory_regions =
+                               REGION_SMEM | REGION_LMEM | REGION_STOLEN_SMEM;
+                       GEM_BUG_ON(!HAS_LMEM(i915));
+               }
+       }
+ #endif
+       ret = pci_enable_device(pdev);
+       if (ret)
+               goto out_fini;
+       ret = i915_driver_early_probe(i915);
+       if (ret < 0)
+               goto out_pci_disable;
+       disable_rpm_wakeref_asserts(&i915->runtime_pm);
+       intel_vgpu_detect(i915);
+       ret = i915_driver_mmio_probe(i915);
+       if (ret < 0)
+               goto out_runtime_pm_put;
+       ret = i915_driver_hw_probe(i915);
+       if (ret < 0)
+               goto out_cleanup_mmio;
+       ret = intel_modeset_init_noirq(i915);
+       if (ret < 0)
+               goto out_cleanup_hw;
+       ret = intel_irq_install(i915);
+       if (ret)
+               goto out_cleanup_modeset;
+       ret = intel_modeset_init_nogem(i915);
+       if (ret)
+               goto out_cleanup_irq;
+       ret = i915_gem_init(i915);
+       if (ret)
+               goto out_cleanup_modeset2;
+       ret = intel_modeset_init(i915);
+       if (ret)
+               goto out_cleanup_gem;
+       i915_driver_register(i915);
+       enable_rpm_wakeref_asserts(&i915->runtime_pm);
+       i915_welcome_messages(i915);
+       i915->do_release = true;
+       return 0;
+ out_cleanup_gem:
+       i915_gem_suspend(i915);
+       i915_gem_driver_remove(i915);
+       i915_gem_driver_release(i915);
+ out_cleanup_modeset2:
+       /* FIXME clean up the error path */
+       intel_modeset_driver_remove(i915);
+       intel_irq_uninstall(i915);
+       intel_modeset_driver_remove_noirq(i915);
+       goto out_cleanup_modeset;
+ out_cleanup_irq:
+       intel_irq_uninstall(i915);
+ out_cleanup_modeset:
+       intel_modeset_driver_remove_nogem(i915);
+ out_cleanup_hw:
+       i915_driver_hw_remove(i915);
+       intel_memory_regions_driver_release(i915);
+       i915_ggtt_driver_release(i915);
+       i915_gem_drain_freed_objects(i915);
+       i915_ggtt_driver_late_release(i915);
+ out_cleanup_mmio:
+       i915_driver_mmio_release(i915);
+ out_runtime_pm_put:
+       enable_rpm_wakeref_asserts(&i915->runtime_pm);
+       i915_driver_late_release(i915);
+ out_pci_disable:
+       pci_disable_device(pdev);
+ out_fini:
+       i915_probe_error(i915, "Device initialization failed (%d)\n", ret);
+       return ret;
+ }
+ void i915_driver_remove(struct drm_i915_private *i915)
+ {
+       disable_rpm_wakeref_asserts(&i915->runtime_pm);
+       i915_driver_unregister(i915);
+       /* Flush any external code that still may be under the RCU lock */
+       synchronize_rcu();
+       i915_gem_suspend(i915);
+       intel_gvt_driver_remove(i915);
+       intel_modeset_driver_remove(i915);
+       intel_irq_uninstall(i915);
+       intel_modeset_driver_remove_noirq(i915);
+       i915_reset_error_state(i915);
+       i915_gem_driver_remove(i915);
+       intel_modeset_driver_remove_nogem(i915);
+       i915_driver_hw_remove(i915);
+       enable_rpm_wakeref_asserts(&i915->runtime_pm);
+ }
+ static void i915_driver_release(struct drm_device *dev)
+ {
+       struct drm_i915_private *dev_priv = to_i915(dev);
+       struct intel_runtime_pm *rpm = &dev_priv->runtime_pm;
+       if (!dev_priv->do_release)
+               return;
+       disable_rpm_wakeref_asserts(rpm);
+       i915_gem_driver_release(dev_priv);
+       intel_memory_regions_driver_release(dev_priv);
+       i915_ggtt_driver_release(dev_priv);
+       i915_gem_drain_freed_objects(dev_priv);
+       i915_ggtt_driver_late_release(dev_priv);
+       i915_driver_mmio_release(dev_priv);
+       enable_rpm_wakeref_asserts(rpm);
+       intel_runtime_pm_driver_release(rpm);
+       i915_driver_late_release(dev_priv);
+ }
+ static int i915_driver_open(struct drm_device *dev, struct drm_file *file)
+ {
+       struct drm_i915_private *i915 = to_i915(dev);
+       int ret;
+       ret = i915_gem_open(i915, file);
+       if (ret)
+               return ret;
+       return 0;
+ }
+ /**
+  * i915_driver_lastclose - clean up after all DRM clients have exited
+  * @dev: DRM device
+  *
+  * Take care of cleaning up after all DRM clients have exited.  In the
+  * mode setting case, we want to restore the kernel's initial mode (just
+  * in case the last client left us in a bad state).
+  *
+  * Additionally, in the non-mode setting case, we'll tear down the GTT
+  * and DMA structures, since the kernel won't be using them, and clea
+  * up any GEM state.
+  */
+ static void i915_driver_lastclose(struct drm_device *dev)
+ {
+       struct drm_i915_private *i915 = to_i915(dev);
+       intel_fbdev_restore_mode(dev);
+       if (HAS_DISPLAY(i915))
+               vga_switcheroo_process_delayed_switch();
+ }
+ static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
+ {
+       struct drm_i915_file_private *file_priv = file->driver_priv;
+       i915_gem_context_close(file);
+       kfree_rcu(file_priv, rcu);
+       /* Catch up with all the deferred frees from "this" client */
+       i915_gem_flush_free_objects(to_i915(dev));
+ }
+ static void intel_suspend_encoders(struct drm_i915_private *dev_priv)
+ {
+       struct drm_device *dev = &dev_priv->drm;
+       struct intel_encoder *encoder;
+       if (!HAS_DISPLAY(dev_priv))
+               return;
+       drm_modeset_lock_all(dev);
+       for_each_intel_encoder(dev, encoder)
+               if (encoder->suspend)
+                       encoder->suspend(encoder);
+       drm_modeset_unlock_all(dev);
+ }
+ static void intel_shutdown_encoders(struct drm_i915_private *dev_priv)
+ {
+       struct drm_device *dev = &dev_priv->drm;
+       struct intel_encoder *encoder;
+       if (!HAS_DISPLAY(dev_priv))
+               return;
+       drm_modeset_lock_all(dev);
+       for_each_intel_encoder(dev, encoder)
+               if (encoder->shutdown)
+                       encoder->shutdown(encoder);
+       drm_modeset_unlock_all(dev);
+ }
+ void i915_driver_shutdown(struct drm_i915_private *i915)
+ {
+       disable_rpm_wakeref_asserts(&i915->runtime_pm);
+       intel_runtime_pm_disable(&i915->runtime_pm);
+       intel_power_domains_disable(i915);
+       i915_gem_suspend(i915);
+       if (HAS_DISPLAY(i915)) {
+               drm_kms_helper_poll_disable(&i915->drm);
+               drm_atomic_helper_shutdown(&i915->drm);
+       }
+       intel_dp_mst_suspend(i915);
+       intel_runtime_pm_disable_interrupts(i915);
+       intel_hpd_cancel_work(i915);
+       intel_suspend_encoders(i915);
+       intel_shutdown_encoders(i915);
+       intel_dmc_ucode_suspend(i915);
+       /*
+        * The only requirement is to reboot with display DC states disabled,
+        * for now leaving all display power wells in the INIT power domain
+        * enabled.
+        *
+        * TODO:
+        * - unify the pci_driver::shutdown sequence here with the
+        *   pci_driver.driver.pm.poweroff,poweroff_late sequence.
+        * - unify the driver remove and system/runtime suspend sequences with
+        *   the above unified shutdown/poweroff sequence.
+        */
+       intel_power_domains_driver_remove(i915);
+       enable_rpm_wakeref_asserts(&i915->runtime_pm);
+       intel_runtime_pm_driver_release(&i915->runtime_pm);
+ }
+ static bool suspend_to_idle(struct drm_i915_private *dev_priv)
+ {
+ #if IS_ENABLED(CONFIG_ACPI_SLEEP)
+       if (acpi_target_system_state() < ACPI_STATE_S3)
+               return true;
+ #endif
+       return false;
+ }
+ static int i915_drm_prepare(struct drm_device *dev)
+ {
+       struct drm_i915_private *i915 = to_i915(dev);
+       /*
+        * NB intel_display_suspend() may issue new requests after we've
+        * ostensibly marked the GPU as ready-to-sleep here. We need to
+        * split out that work and pull it forward so that after point,
+        * the GPU is not woken again.
+        */
+       return i915_gem_backup_suspend(i915);
+ }
+ static int i915_drm_suspend(struct drm_device *dev)
+ {
+       struct drm_i915_private *dev_priv = to_i915(dev);
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       pci_power_t opregion_target_state;
+       disable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       /* We do a lot of poking in a lot of registers, make sure they work
+        * properly. */
+       intel_power_domains_disable(dev_priv);
+       if (HAS_DISPLAY(dev_priv))
+               drm_kms_helper_poll_disable(dev);
+       pci_save_state(pdev);
+       intel_display_suspend(dev);
+       intel_dp_mst_suspend(dev_priv);
+       intel_runtime_pm_disable_interrupts(dev_priv);
+       intel_hpd_cancel_work(dev_priv);
+       intel_suspend_encoders(dev_priv);
+       intel_suspend_hw(dev_priv);
+       /* Must be called before GGTT is suspended. */
+       intel_dpt_suspend(dev_priv);
 -      i915_ggtt_resume(&dev_priv->ggtt);
++      i915_ggtt_suspend(to_gt(dev_priv)->ggtt);
+       i915_save_display(dev_priv);
+       opregion_target_state = suspend_to_idle(dev_priv) ? PCI_D1 : PCI_D3cold;
+       intel_opregion_suspend(dev_priv, opregion_target_state);
+       intel_fbdev_set_suspend(dev, FBINFO_STATE_SUSPENDED, true);
+       dev_priv->suspend_count++;
+       intel_dmc_ucode_suspend(dev_priv);
+       enable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       return 0;
+ }
+ static enum i915_drm_suspend_mode
+ get_suspend_mode(struct drm_i915_private *dev_priv, bool hibernate)
+ {
+       if (hibernate)
+               return I915_DRM_SUSPEND_HIBERNATE;
+       if (suspend_to_idle(dev_priv))
+               return I915_DRM_SUSPEND_IDLE;
+       return I915_DRM_SUSPEND_MEM;
+ }
+ static int i915_drm_suspend_late(struct drm_device *dev, bool hibernation)
+ {
+       struct drm_i915_private *dev_priv = to_i915(dev);
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       struct intel_runtime_pm *rpm = &dev_priv->runtime_pm;
+       int ret;
+       disable_rpm_wakeref_asserts(rpm);
+       i915_gem_suspend_late(dev_priv);
+       intel_uncore_suspend(&dev_priv->uncore);
+       intel_power_domains_suspend(dev_priv,
+                                   get_suspend_mode(dev_priv, hibernation));
+       intel_display_power_suspend_late(dev_priv);
+       ret = vlv_suspend_complete(dev_priv);
+       if (ret) {
+               drm_err(&dev_priv->drm, "Suspend complete failed: %d\n", ret);
+               intel_power_domains_resume(dev_priv);
+               goto out;
+       }
+       /*
+        * FIXME: Temporary hammer to avoid freezing the machine on our DGFX
+        * This should be totally removed when we handle the pci states properly
+        * on runtime PM and on s2idle cases.
+        */
+       if (suspend_to_idle(dev_priv))
+               pci_d3cold_disable(pdev);
+       pci_disable_device(pdev);
+       /*
+        * During hibernation on some platforms the BIOS may try to access
+        * the device even though it's already in D3 and hang the machine. So
+        * leave the device in D0 on those platforms and hope the BIOS will
+        * power down the device properly. The issue was seen on multiple old
+        * GENs with different BIOS vendors, so having an explicit blacklist
+        * is inpractical; apply the workaround on everything pre GEN6. The
+        * platforms where the issue was seen:
+        * Lenovo Thinkpad X301, X61s, X60, T60, X41
+        * Fujitsu FSC S7110
+        * Acer Aspire 1830T
+        */
+       if (!(hibernation && GRAPHICS_VER(dev_priv) < 6))
+               pci_set_power_state(pdev, PCI_D3hot);
+ out:
+       enable_rpm_wakeref_asserts(rpm);
+       if (!dev_priv->uncore.user_forcewake_count)
+               intel_runtime_pm_driver_release(rpm);
+       return ret;
+ }
+ int i915_driver_suspend_switcheroo(struct drm_i915_private *i915,
+                                  pm_message_t state)
+ {
+       int error;
+       if (drm_WARN_ON_ONCE(&i915->drm, state.event != PM_EVENT_SUSPEND &&
+                            state.event != PM_EVENT_FREEZE))
+               return -EINVAL;
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       error = i915_drm_suspend(&i915->drm);
+       if (error)
+               return error;
+       return i915_drm_suspend_late(&i915->drm, false);
+ }
+ static int i915_drm_resume(struct drm_device *dev)
+ {
+       struct drm_i915_private *dev_priv = to_i915(dev);
+       int ret;
+       disable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       ret = intel_pcode_init(dev_priv);
+       if (ret)
+               return ret;
+       sanitize_gpu(dev_priv);
+       ret = i915_ggtt_enable_hw(dev_priv);
+       if (ret)
+               drm_err(&dev_priv->drm, "failed to re-enable GGTT\n");
++      i915_ggtt_resume(to_gt(dev_priv)->ggtt);
+       /* Must be called after GGTT is resumed. */
+       intel_dpt_resume(dev_priv);
+       intel_dmc_ucode_resume(dev_priv);
+       i915_restore_display(dev_priv);
+       intel_pps_unlock_regs_wa(dev_priv);
+       intel_init_pch_refclk(dev_priv);
+       /*
+        * Interrupts have to be enabled before any batches are run. If not the
+        * GPU will hang. i915_gem_init_hw() will initiate batches to
+        * update/restore the context.
+        *
+        * drm_mode_config_reset() needs AUX interrupts.
+        *
+        * Modeset enabling in intel_modeset_init_hw() also needs working
+        * interrupts.
+        */
+       intel_runtime_pm_enable_interrupts(dev_priv);
+       if (HAS_DISPLAY(dev_priv))
+               drm_mode_config_reset(dev);
+       i915_gem_resume(dev_priv);
+       intel_modeset_init_hw(dev_priv);
+       intel_init_clock_gating(dev_priv);
+       intel_hpd_init(dev_priv);
+       /* MST sideband requires HPD interrupts enabled */
+       intel_dp_mst_resume(dev_priv);
+       intel_display_resume(dev);
+       intel_hpd_poll_disable(dev_priv);
+       if (HAS_DISPLAY(dev_priv))
+               drm_kms_helper_poll_enable(dev);
+       intel_opregion_resume(dev_priv);
+       intel_fbdev_set_suspend(dev, FBINFO_STATE_RUNNING, false);
+       intel_power_domains_enable(dev_priv);
+       intel_gvt_resume(dev_priv);
+       enable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       return 0;
+ }
+ static int i915_drm_resume_early(struct drm_device *dev)
+ {
+       struct drm_i915_private *dev_priv = to_i915(dev);
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       int ret;
+       /*
+        * We have a resume ordering issue with the snd-hda driver also
+        * requiring our device to be power up. Due to the lack of a
+        * parent/child relationship we currently solve this with an early
+        * resume hook.
+        *
+        * FIXME: This should be solved with a special hdmi sink device or
+        * similar so that power domains can be employed.
+        */
+       /*
+        * Note that we need to set the power state explicitly, since we
+        * powered off the device during freeze and the PCI core won't power
+        * it back up for us during thaw. Powering off the device during
+        * freeze is not a hard requirement though, and during the
+        * suspend/resume phases the PCI core makes sure we get here with the
+        * device powered on. So in case we change our freeze logic and keep
+        * the device powered we can also remove the following set power state
+        * call.
+        */
+       ret = pci_set_power_state(pdev, PCI_D0);
+       if (ret) {
+               drm_err(&dev_priv->drm,
+                       "failed to set PCI D0 power state (%d)\n", ret);
+               return ret;
+       }
+       /*
+        * Note that pci_enable_device() first enables any parent bridge
+        * device and only then sets the power state for this device. The
+        * bridge enabling is a nop though, since bridge devices are resumed
+        * first. The order of enabling power and enabling the device is
+        * imposed by the PCI core as described above, so here we preserve the
+        * same order for the freeze/thaw phases.
+        *
+        * TODO: eventually we should remove pci_disable_device() /
+        * pci_enable_enable_device() from suspend/resume. Due to how they
+        * depend on the device enable refcount we can't anyway depend on them
+        * disabling/enabling the device.
+        */
+       if (pci_enable_device(pdev))
+               return -EIO;
+       pci_set_master(pdev);
+       pci_d3cold_enable(pdev);
+       disable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       ret = vlv_resume_prepare(dev_priv, false);
+       if (ret)
+               drm_err(&dev_priv->drm,
+                       "Resume prepare failed: %d, continuing anyway\n", ret);
+       intel_uncore_resume_early(&dev_priv->uncore);
+       intel_gt_check_and_clear_faults(to_gt(dev_priv));
+       intel_display_power_resume_early(dev_priv);
+       intel_power_domains_resume(dev_priv);
+       enable_rpm_wakeref_asserts(&dev_priv->runtime_pm);
+       return ret;
+ }
+ int i915_driver_resume_switcheroo(struct drm_i915_private *i915)
+ {
+       int ret;
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       ret = i915_drm_resume_early(&i915->drm);
+       if (ret)
+               return ret;
+       return i915_drm_resume(&i915->drm);
+ }
+ static int i915_pm_prepare(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       if (!i915) {
+               dev_err(kdev, "DRM not initialized, aborting suspend.\n");
+               return -ENODEV;
+       }
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_prepare(&i915->drm);
+ }
+ static int i915_pm_suspend(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       if (!i915) {
+               dev_err(kdev, "DRM not initialized, aborting suspend.\n");
+               return -ENODEV;
+       }
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_suspend(&i915->drm);
+ }
+ static int i915_pm_suspend_late(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       /*
+        * We have a suspend ordering issue with the snd-hda driver also
+        * requiring our device to be power up. Due to the lack of a
+        * parent/child relationship we currently solve this with an late
+        * suspend hook.
+        *
+        * FIXME: This should be solved with a special hdmi sink device or
+        * similar so that power domains can be employed.
+        */
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_suspend_late(&i915->drm, false);
+ }
+ static int i915_pm_poweroff_late(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_suspend_late(&i915->drm, true);
+ }
+ static int i915_pm_resume_early(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_resume_early(&i915->drm);
+ }
+ static int i915_pm_resume(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       if (i915->drm.switch_power_state == DRM_SWITCH_POWER_OFF)
+               return 0;
+       return i915_drm_resume(&i915->drm);
+ }
+ /* freeze: before creating the hibernation_image */
+ static int i915_pm_freeze(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       int ret;
+       if (i915->drm.switch_power_state != DRM_SWITCH_POWER_OFF) {
+               ret = i915_drm_suspend(&i915->drm);
+               if (ret)
+                       return ret;
+       }
+       ret = i915_gem_freeze(i915);
+       if (ret)
+               return ret;
+       return 0;
+ }
+ static int i915_pm_freeze_late(struct device *kdev)
+ {
+       struct drm_i915_private *i915 = kdev_to_i915(kdev);
+       int ret;
+       if (i915->drm.switch_power_state != DRM_SWITCH_POWER_OFF) {
+               ret = i915_drm_suspend_late(&i915->drm, true);
+               if (ret)
+                       return ret;
+       }
+       ret = i915_gem_freeze_late(i915);
+       if (ret)
+               return ret;
+       return 0;
+ }
+ /* thaw: called after creating the hibernation image, but before turning off. */
+ static int i915_pm_thaw_early(struct device *kdev)
+ {
+       return i915_pm_resume_early(kdev);
+ }
+ static int i915_pm_thaw(struct device *kdev)
+ {
+       return i915_pm_resume(kdev);
+ }
+ /* restore: called after loading the hibernation image. */
+ static int i915_pm_restore_early(struct device *kdev)
+ {
+       return i915_pm_resume_early(kdev);
+ }
+ static int i915_pm_restore(struct device *kdev)
+ {
+       return i915_pm_resume(kdev);
+ }
+ static int intel_runtime_suspend(struct device *kdev)
+ {
+       struct drm_i915_private *dev_priv = kdev_to_i915(kdev);
+       struct intel_runtime_pm *rpm = &dev_priv->runtime_pm;
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       int ret;
+       if (drm_WARN_ON_ONCE(&dev_priv->drm, !HAS_RUNTIME_PM(dev_priv)))
+               return -ENODEV;
+       drm_dbg_kms(&dev_priv->drm, "Suspending device\n");
+       disable_rpm_wakeref_asserts(rpm);
+       /*
+        * We are safe here against re-faults, since the fault handler takes
+        * an RPM reference.
+        */
+       i915_gem_runtime_suspend(dev_priv);
+       intel_gt_runtime_suspend(to_gt(dev_priv));
+       intel_runtime_pm_disable_interrupts(dev_priv);
+       intel_uncore_suspend(&dev_priv->uncore);
+       intel_display_power_suspend(dev_priv);
+       ret = vlv_suspend_complete(dev_priv);
+       if (ret) {
+               drm_err(&dev_priv->drm,
+                       "Runtime suspend failed, disabling it (%d)\n", ret);
+               intel_uncore_runtime_resume(&dev_priv->uncore);
+               intel_runtime_pm_enable_interrupts(dev_priv);
+               intel_gt_runtime_resume(to_gt(dev_priv));
+               enable_rpm_wakeref_asserts(rpm);
+               return ret;
+       }
+       enable_rpm_wakeref_asserts(rpm);
+       intel_runtime_pm_driver_release(rpm);
+       if (intel_uncore_arm_unclaimed_mmio_detection(&dev_priv->uncore))
+               drm_err(&dev_priv->drm,
+                       "Unclaimed access detected prior to suspending\n");
+       /*
+        * FIXME: Temporary hammer to avoid freezing the machine on our DGFX
+        * This should be totally removed when we handle the pci states properly
+        * on runtime PM and on s2idle cases.
+        */
+       pci_d3cold_disable(pdev);
+       rpm->suspended = true;
+       /*
+        * FIXME: We really should find a document that references the arguments
+        * used below!
+        */
+       if (IS_BROADWELL(dev_priv)) {
+               /*
+                * On Broadwell, if we use PCI_D1 the PCH DDI ports will stop
+                * being detected, and the call we do at intel_runtime_resume()
+                * won't be able to restore them. Since PCI_D3hot matches the
+                * actual specification and appears to be working, use it.
+                */
+               intel_opregion_notify_adapter(dev_priv, PCI_D3hot);
+       } else {
+               /*
+                * current versions of firmware which depend on this opregion
+                * notification have repurposed the D1 definition to mean
+                * "runtime suspended" vs. what you would normally expect (D3)
+                * to distinguish it from notifications that might be sent via
+                * the suspend path.
+                */
+               intel_opregion_notify_adapter(dev_priv, PCI_D1);
+       }
+       assert_forcewakes_inactive(&dev_priv->uncore);
+       if (!IS_VALLEYVIEW(dev_priv) && !IS_CHERRYVIEW(dev_priv))
+               intel_hpd_poll_enable(dev_priv);
+       drm_dbg_kms(&dev_priv->drm, "Device suspended\n");
+       return 0;
+ }
+ static int intel_runtime_resume(struct device *kdev)
+ {
+       struct drm_i915_private *dev_priv = kdev_to_i915(kdev);
+       struct intel_runtime_pm *rpm = &dev_priv->runtime_pm;
+       struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
+       int ret;
+       if (drm_WARN_ON_ONCE(&dev_priv->drm, !HAS_RUNTIME_PM(dev_priv)))
+               return -ENODEV;
+       drm_dbg_kms(&dev_priv->drm, "Resuming device\n");
+       drm_WARN_ON_ONCE(&dev_priv->drm, atomic_read(&rpm->wakeref_count));
+       disable_rpm_wakeref_asserts(rpm);
+       intel_opregion_notify_adapter(dev_priv, PCI_D0);
+       rpm->suspended = false;
+       pci_d3cold_enable(pdev);
+       if (intel_uncore_unclaimed_mmio(&dev_priv->uncore))
+               drm_dbg(&dev_priv->drm,
+                       "Unclaimed access during suspend, bios?\n");
+       intel_display_power_resume(dev_priv);
+       ret = vlv_resume_prepare(dev_priv, true);
+       intel_uncore_runtime_resume(&dev_priv->uncore);
+       intel_runtime_pm_enable_interrupts(dev_priv);
+       /*
+        * No point of rolling back things in case of an error, as the best
+        * we can do is to hope that things will still work (and disable RPM).
+        */
+       intel_gt_runtime_resume(to_gt(dev_priv));
+       /*
+        * On VLV/CHV display interrupts are part of the display
+        * power well, so hpd is reinitialized from there. For
+        * everyone else do it here.
+        */
+       if (!IS_VALLEYVIEW(dev_priv) && !IS_CHERRYVIEW(dev_priv)) {
+               intel_hpd_init(dev_priv);
+               intel_hpd_poll_disable(dev_priv);
+       }
+       intel_enable_ipc(dev_priv);
+       enable_rpm_wakeref_asserts(rpm);
+       if (ret)
+               drm_err(&dev_priv->drm,
+                       "Runtime resume failed, disabling it (%d)\n", ret);
+       else
+               drm_dbg_kms(&dev_priv->drm, "Device resumed\n");
+       return ret;
+ }
+ const struct dev_pm_ops i915_pm_ops = {
+       /*
+        * S0ix (via system suspend) and S3 event handlers [PMSG_SUSPEND,
+        * PMSG_RESUME]
+        */
+       .prepare = i915_pm_prepare,
+       .suspend = i915_pm_suspend,
+       .suspend_late = i915_pm_suspend_late,
+       .resume_early = i915_pm_resume_early,
+       .resume = i915_pm_resume,
+       /*
+        * S4 event handlers
+        * @freeze, @freeze_late    : called (1) before creating the
+        *                            hibernation image [PMSG_FREEZE] and
+        *                            (2) after rebooting, before restoring
+        *                            the image [PMSG_QUIESCE]
+        * @thaw, @thaw_early       : called (1) after creating the hibernation
+        *                            image, before writing it [PMSG_THAW]
+        *                            and (2) after failing to create or
+        *                            restore the image [PMSG_RECOVER]
+        * @poweroff, @poweroff_late: called after writing the hibernation
+        *                            image, before rebooting [PMSG_HIBERNATE]
+        * @restore, @restore_early : called after rebooting and restoring the
+        *                            hibernation image [PMSG_RESTORE]
+        */
+       .freeze = i915_pm_freeze,
+       .freeze_late = i915_pm_freeze_late,
+       .thaw_early = i915_pm_thaw_early,
+       .thaw = i915_pm_thaw,
+       .poweroff = i915_pm_suspend,
+       .poweroff_late = i915_pm_poweroff_late,
+       .restore_early = i915_pm_restore_early,
+       .restore = i915_pm_restore,
+       /* S0ix (via runtime suspend) event handlers */
+       .runtime_suspend = intel_runtime_suspend,
+       .runtime_resume = intel_runtime_resume,
+ };
+ static const struct file_operations i915_driver_fops = {
+       .owner = THIS_MODULE,
+       .open = drm_open,
+       .release = drm_release_noglobal,
+       .unlocked_ioctl = drm_ioctl,
+       .mmap = i915_gem_mmap,
+       .poll = drm_poll,
+       .read = drm_read,
+       .compat_ioctl = i915_ioc32_compat_ioctl,
+       .llseek = noop_llseek,
+ };
+ static int
+ i915_gem_reject_pin_ioctl(struct drm_device *dev, void *data,
+                         struct drm_file *file)
+ {
+       return -ENODEV;
+ }
+ static const struct drm_ioctl_desc i915_ioctls[] = {
+       DRM_IOCTL_DEF_DRV(I915_INIT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_FLUSH, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_FLIP, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_BATCHBUFFER, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_IRQ_EMIT, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_IRQ_WAIT, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_GETPARAM, i915_getparam_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_SETPARAM, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_ALLOC, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_FREE, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_INIT_HEAP, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_CMDBUFFER, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_DESTROY_HEAP,  drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_SET_VBLANK_PIPE,  drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GET_VBLANK_PIPE,  drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_VBLANK_SWAP, drm_noop, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_HWS_ADDR, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_INIT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER, drm_invalid_op, DRM_AUTH),
+       DRM_IOCTL_DEF_DRV(I915_GEM_EXECBUFFER2_WR, i915_gem_execbuffer2_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_PIN, i915_gem_reject_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_UNPIN, i915_gem_reject_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_SET_CACHING, i915_gem_set_caching_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_GET_CACHING, i915_gem_get_caching_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_THROTTLE, i915_gem_throttle_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_ENTERVT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_LEAVEVT, drm_noop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CREATE, i915_gem_create_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CREATE_EXT, i915_gem_create_ext_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_PREAD, i915_gem_pread_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_PWRITE, i915_gem_pwrite_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_MMAP, i915_gem_mmap_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_MMAP_OFFSET, i915_gem_mmap_offset_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_SET_DOMAIN, i915_gem_set_domain_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_SW_FINISH, i915_gem_sw_finish_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_SET_TILING, i915_gem_set_tiling_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_GET_TILING, i915_gem_get_tiling_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_GET_APERTURE, i915_gem_get_aperture_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GET_PIPE_FROM_CRTC_ID, intel_get_pipe_from_crtc_id_ioctl, 0),
+       DRM_IOCTL_DEF_DRV(I915_GEM_MADVISE, i915_gem_madvise_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_OVERLAY_PUT_IMAGE, intel_overlay_put_image_ioctl, DRM_MASTER),
+       DRM_IOCTL_DEF_DRV(I915_OVERLAY_ATTRS, intel_overlay_attrs_ioctl, DRM_MASTER),
+       DRM_IOCTL_DEF_DRV(I915_SET_SPRITE_COLORKEY, intel_sprite_set_colorkey_ioctl, DRM_MASTER),
+       DRM_IOCTL_DEF_DRV(I915_GET_SPRITE_COLORKEY, drm_noop, DRM_MASTER),
+       DRM_IOCTL_DEF_DRV(I915_GEM_WAIT, i915_gem_wait_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CONTEXT_CREATE_EXT, i915_gem_context_create_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CONTEXT_DESTROY, i915_gem_context_destroy_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_REG_READ, i915_reg_read_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GET_RESET_STATS, i915_gem_context_reset_stats_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_USERPTR, i915_gem_userptr_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CONTEXT_GETPARAM, i915_gem_context_getparam_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_CONTEXT_SETPARAM, i915_gem_context_setparam_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_PERF_OPEN, i915_perf_open_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_PERF_ADD_CONFIG, i915_perf_add_config_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_PERF_REMOVE_CONFIG, i915_perf_remove_config_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_QUERY, i915_query_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_VM_CREATE, i915_gem_vm_create_ioctl, DRM_RENDER_ALLOW),
+       DRM_IOCTL_DEF_DRV(I915_GEM_VM_DESTROY, i915_gem_vm_destroy_ioctl, DRM_RENDER_ALLOW),
+ };
+ static const struct drm_driver i915_drm_driver = {
+       /* Don't use MTRRs here; the Xserver or userspace app should
+        * deal with them for Intel hardware.
+        */
+       .driver_features =
+           DRIVER_GEM |
+           DRIVER_RENDER | DRIVER_MODESET | DRIVER_ATOMIC | DRIVER_SYNCOBJ |
+           DRIVER_SYNCOBJ_TIMELINE,
+       .release = i915_driver_release,
+       .open = i915_driver_open,
+       .lastclose = i915_driver_lastclose,
+       .postclose = i915_driver_postclose,
+       .prime_handle_to_fd = drm_gem_prime_handle_to_fd,
+       .prime_fd_to_handle = drm_gem_prime_fd_to_handle,
+       .gem_prime_import = i915_gem_prime_import,
+       .dumb_create = i915_gem_dumb_create,
+       .dumb_map_offset = i915_gem_dumb_mmap_offset,
+       .ioctls = i915_ioctls,
+       .num_ioctls = ARRAY_SIZE(i915_ioctls),
+       .fops = &i915_driver_fops,
+       .name = DRIVER_NAME,
+       .desc = DRIVER_DESC,
+       .date = DRIVER_DATE,
+       .major = DRIVER_MAJOR,
+       .minor = DRIVER_MINOR,
+       .patchlevel = DRIVER_PATCHLEVEL,
+ };
Simple merge
Simple merge
Simple merge
index 5b7da0ee5d640eaeb54a3338cd803cc1cf4fb3dc,ca575e129ced61f1ee3b41ff0787b98a1541e389..88370dadca820ecfa3d139936183b233fae7e898
@@@ -95,15 -95,22 +95,24 @@@ enum i915_cache_level
   *
   */
  
 +struct i915_vma_resource;
 +
  struct intel_remapped_plane_info {
        /* in gtt pages */
-       u32 offset;
-       u16 width;
-       u16 height;
-       u16 src_stride;
-       u16 dst_stride;
+       u32 offset:31;
+       u32 linear:1;
+       union {
+               /* in gtt pages for !linear */
+               struct {
+                       u16 width;
+                       u16 height;
+                       u16 src_stride;
+                       u16 dst_stride;
+               };
+               /* in gtt pages for linear */
+               u32 size;
+       };
  } __packed;
  
  struct intel_remapped_info {