diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index f36126383d26..d893e4da5dce 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1616,22 +1616,6 @@ out: return ret; } -void i915_gem_release_all_mmaps(struct drm_i915_private *dev_priv) -{ - struct i915_vma *vma; - - /* - * Only the global gtt is relevant for gtt memory mappings, so restrict - * list traversal to objects bound into the global address space. Note - * that the active list should be empty, but better safe than sorry. - */ - WARN_ON(!list_empty(&dev_priv->gtt.base.active_list)); - list_for_each_entry(vma, &dev_priv->gtt.base.active_list, mm_list) - i915_gem_release_mmap(vma->obj); - list_for_each_entry(vma, &dev_priv->gtt.base.inactive_list, mm_list) - i915_gem_release_mmap(vma->obj); -} - /** * i915_gem_release_mmap - remove physical page mappings * @obj: obj in question @@ -1657,6 +1641,15 @@ i915_gem_release_mmap(struct drm_i915_gem_object *obj) obj->fault_mappable = false; } +void +i915_gem_release_all_mmaps(struct drm_i915_private *dev_priv) +{ + struct drm_i915_gem_object *obj; + + list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) + i915_gem_release_mmap(obj); +} + uint32_t i915_gem_get_gtt_size(struct drm_device *dev, uint32_t size, int tiling_mode) { diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c index 3521f998a178..34894b573064 100644 --- a/drivers/gpu/drm/i915/i915_gem_render_state.c +++ b/drivers/gpu/drm/i915/i915_gem_render_state.c @@ -31,7 +31,7 @@ struct i915_render_state { struct drm_i915_gem_object *obj; unsigned long ggtt_offset; - void *batch; + u32 *batch; u32 size; u32 len; }; @@ -80,7 +80,7 @@ free: static void render_state_free(struct i915_render_state *so) { - kunmap(so->batch); + kunmap(kmap_to_page(so->batch)); i915_gem_object_ggtt_unpin(so->obj); drm_gem_object_unreference(&so->obj->base); kfree(so); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 267f069765ad..c05c84f3f091 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -2845,7 +2845,7 @@ static int semaphore_passed(struct intel_engine_cs *ring) { struct drm_i915_private *dev_priv = ring->dev->dev_private; struct intel_engine_cs *signaller; - u32 seqno, ctl; + u32 seqno; ring->hangcheck.deadlock++; @@ -2857,15 +2857,12 @@ static int semaphore_passed(struct intel_engine_cs *ring) if (signaller->hangcheck.deadlock >= I915_NUM_RINGS) return -1; - /* cursory check for an unkickable deadlock */ - ctl = I915_READ_CTL(signaller); - if (ctl & RING_WAIT_SEMAPHORE && semaphore_passed(signaller) < 0) - return -1; - if (i915_seqno_passed(signaller->get_seqno(signaller, false), seqno)) return 1; - if (signaller->hangcheck.deadlock) + /* cursory check for an unkickable deadlock */ + if (I915_READ_CTL(signaller) & RING_WAIT_SEMAPHORE && + semaphore_passed(signaller) < 0) return -1; return 0; diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 0b2471107137..c0ea66192fe0 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -2291,6 +2291,7 @@ static void cik_tiling_mode_table_init(struct radeon_device *rdev) gb_tile_moden = 0; break; } + rdev->config.cik.macrotile_mode_array[reg_offset] = gb_tile_moden; WREG32(GB_MACROTILE_MODE0 + (reg_offset * 4), gb_tile_moden); } } else if (num_pipe_configs == 8) { @@ -7376,6 +7377,7 @@ static inline u32 cik_get_ih_wptr(struct radeon_device *rdev) tmp = RREG32(IH_RB_CNTL); tmp |= IH_WPTR_OVERFLOW_CLEAR; WREG32(IH_RB_CNTL, tmp); + wptr &= ~RB_OVERFLOW; } return (wptr & rdev->ih.ptr_mask); } diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 250bac3935a4..15e4f28015e1 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -4756,6 +4756,7 @@ static u32 evergreen_get_ih_wptr(struct radeon_device *rdev) tmp = RREG32(IH_RB_CNTL); tmp |= IH_WPTR_OVERFLOW_CLEAR; WREG32(IH_RB_CNTL, tmp); + wptr &= ~RB_OVERFLOW; } return (wptr & rdev->ih.ptr_mask); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index c66952d4b00c..3c69f58e46ef 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3795,6 +3795,7 @@ static u32 r600_get_ih_wptr(struct radeon_device *rdev) tmp = RREG32(IH_RB_CNTL); tmp |= IH_WPTR_OVERFLOW_CLEAR; WREG32(IH_RB_CNTL, tmp); + wptr &= ~RB_OVERFLOW; } return (wptr & rdev->ih.ptr_mask); } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index b7204500a9a6..60c47f829122 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -449,6 +449,7 @@ struct radeon_bo_va { /* protected by vm mutex */ struct list_head vm_list; + struct list_head vm_status; /* constant after initialization */ struct radeon_vm *vm; @@ -867,6 +868,9 @@ struct radeon_vm { struct list_head va; unsigned id; + /* BOs freed, but not yet updated in the PT */ + struct list_head freed; + /* contains the page directory */ struct radeon_bo *page_directory; uint64_t pd_gpu_addr; @@ -875,6 +879,8 @@ struct radeon_vm { /* array of page tables, one for each page directory entry */ struct radeon_vm_pt *page_tables; + struct radeon_bo_va *ib_bo_va; + struct mutex mutex; /* last fence for cs using this vm */ struct radeon_fence *fence; @@ -2832,9 +2838,10 @@ void radeon_vm_fence(struct radeon_device *rdev, uint64_t radeon_vm_map_gart(struct radeon_device *rdev, uint64_t addr); int radeon_vm_update_page_directory(struct radeon_device *rdev, struct radeon_vm *vm); +int radeon_vm_clear_freed(struct radeon_device *rdev, + struct radeon_vm *vm); int radeon_vm_bo_update(struct radeon_device *rdev, - struct radeon_vm *vm, - struct radeon_bo *bo, + struct radeon_bo_va *bo_va, struct ttm_mem_reg *mem); void radeon_vm_bo_invalidate(struct radeon_device *rdev, struct radeon_bo *bo); @@ -2847,8 +2854,8 @@ int radeon_vm_bo_set_addr(struct radeon_device *rdev, struct radeon_bo_va *bo_va, uint64_t offset, uint32_t flags); -int radeon_vm_bo_rmv(struct radeon_device *rdev, - struct radeon_bo_va *bo_va); +void radeon_vm_bo_rmv(struct radeon_device *rdev, + struct radeon_bo_va *bo_va); /* audio */ void r600_audio_update_hdmi(struct work_struct *work); diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 71a143461478..ae763f60c8a0 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -461,13 +461,23 @@ static int radeon_bo_vm_update_pte(struct radeon_cs_parser *p, struct radeon_vm *vm) { struct radeon_device *rdev = p->rdev; + struct radeon_bo_va *bo_va; int i, r; r = radeon_vm_update_page_directory(rdev, vm); if (r) return r; - r = radeon_vm_bo_update(rdev, vm, rdev->ring_tmp_bo.bo, + r = radeon_vm_clear_freed(rdev, vm); + if (r) + return r; + + if (vm->ib_bo_va == NULL) { + DRM_ERROR("Tmp BO not in VM!\n"); + return -EINVAL; + } + + r = radeon_vm_bo_update(rdev, vm->ib_bo_va, &rdev->ring_tmp_bo.bo->tbo.mem); if (r) return r; @@ -480,7 +490,13 @@ static int radeon_bo_vm_update_pte(struct radeon_cs_parser *p, continue; bo = p->relocs[i].robj; - r = radeon_vm_bo_update(rdev, vm, bo, &bo->tbo.mem); + bo_va = radeon_vm_bo_find(vm, bo); + if (bo_va == NULL) { + dev_err(rdev->dev, "bo %p not in vm %p\n", bo, vm); + return -EINVAL; + } + + r = radeon_vm_bo_update(rdev, bo_va, &bo->tbo.mem); if (r) return r; } diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 03686fab842d..697add2cd4e3 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -1056,36 +1056,36 @@ static void radeon_check_arguments(struct radeon_device *rdev) if (!radeon_check_pot_argument(radeon_vm_size)) { dev_warn(rdev->dev, "VM size (%d) must be a power of 2\n", radeon_vm_size); - radeon_vm_size = 4096; + radeon_vm_size = 4; } - if (radeon_vm_size < 4) { - dev_warn(rdev->dev, "VM size (%d) to small, min is 4MB\n", + if (radeon_vm_size < 1) { + dev_warn(rdev->dev, "VM size (%d) to small, min is 1GB\n", radeon_vm_size); - radeon_vm_size = 4096; + radeon_vm_size = 4; } /* * Max GPUVM size for Cayman, SI and CI are 40 bits. */ - if (radeon_vm_size > 1024*1024) { - dev_warn(rdev->dev, "VM size (%d) to large, max is 1TB\n", + if (radeon_vm_size > 1024) { + dev_warn(rdev->dev, "VM size (%d) too large, max is 1TB\n", radeon_vm_size); - radeon_vm_size = 4096; + radeon_vm_size = 4; } /* defines number of bits in page table versus page directory, * a page is 4KB so we have 12 bits offset, minimum 9 bits in the * page table and the remaining bits are in the page directory */ if (radeon_vm_block_size < 9) { - dev_warn(rdev->dev, "VM page table size (%d) to small\n", + dev_warn(rdev->dev, "VM page table size (%d) too small\n", radeon_vm_block_size); radeon_vm_block_size = 9; } if (radeon_vm_block_size > 24 || - radeon_vm_size < (1ull << radeon_vm_block_size)) { - dev_warn(rdev->dev, "VM page table size (%d) to large\n", + (radeon_vm_size * 1024) < (1ull << radeon_vm_block_size)) { + dev_warn(rdev->dev, "VM page table size (%d) too large\n", radeon_vm_block_size); radeon_vm_block_size = 9; } @@ -1238,7 +1238,7 @@ int radeon_device_init(struct radeon_device *rdev, /* Adjust VM size here. * Max GPUVM size for cayman+ is 40 bits. */ - rdev->vm_manager.max_pfn = radeon_vm_size << 8; + rdev->vm_manager.max_pfn = radeon_vm_size << 18; /* Set asic functions */ r = radeon_asic_init(rdev); diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index cb1421369e3a..e9e361084249 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -173,7 +173,7 @@ int radeon_dpm = -1; int radeon_aspm = -1; int radeon_runtime_pm = -1; int radeon_hard_reset = 0; -int radeon_vm_size = 4096; +int radeon_vm_size = 4; int radeon_vm_block_size = 9; int radeon_deep_color = 0; @@ -243,7 +243,7 @@ module_param_named(runpm, radeon_runtime_pm, int, 0444); MODULE_PARM_DESC(hard_reset, "PCI config reset (1 = force enable, 0 = disable (default))"); module_param_named(hard_reset, radeon_hard_reset, int, 0444); -MODULE_PARM_DESC(vm_size, "VM address space size in megabytes (default 4GB)"); +MODULE_PARM_DESC(vm_size, "VM address space size in gigabytes (default 4GB)"); module_param_named(vm_size, radeon_vm_size, int, 0444); MODULE_PARM_DESC(vm_block_size, "VM page table size in bits (default 9)"); diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 35d931881b4b..d25ae6acfd5a 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -579,7 +579,7 @@ int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) /* new gpu have virtual address space support */ if (rdev->family >= CHIP_CAYMAN) { struct radeon_fpriv *fpriv; - struct radeon_bo_va *bo_va; + struct radeon_vm *vm; int r; fpriv = kzalloc(sizeof(*fpriv), GFP_KERNEL); @@ -587,7 +587,8 @@ int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) return -ENOMEM; } - r = radeon_vm_init(rdev, &fpriv->vm); + vm = &fpriv->vm; + r = radeon_vm_init(rdev, vm); if (r) { kfree(fpriv); return r; @@ -596,22 +597,23 @@ int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) if (rdev->accel_working) { r = radeon_bo_reserve(rdev->ring_tmp_bo.bo, false); if (r) { - radeon_vm_fini(rdev, &fpriv->vm); + radeon_vm_fini(rdev, vm); kfree(fpriv); return r; } /* map the ib pool buffer read only into * virtual address space */ - bo_va = radeon_vm_bo_add(rdev, &fpriv->vm, - rdev->ring_tmp_bo.bo); - r = radeon_vm_bo_set_addr(rdev, bo_va, RADEON_VA_IB_OFFSET, + vm->ib_bo_va = radeon_vm_bo_add(rdev, vm, + rdev->ring_tmp_bo.bo); + r = radeon_vm_bo_set_addr(rdev, vm->ib_bo_va, + RADEON_VA_IB_OFFSET, RADEON_VM_PAGE_READABLE | RADEON_VM_PAGE_SNOOPED); radeon_bo_unreserve(rdev->ring_tmp_bo.bo); if (r) { - radeon_vm_fini(rdev, &fpriv->vm); + radeon_vm_fini(rdev, vm); kfree(fpriv); return r; } @@ -640,21 +642,19 @@ void radeon_driver_postclose_kms(struct drm_device *dev, /* new gpu have virtual address space support */ if (rdev->family >= CHIP_CAYMAN && file_priv->driver_priv) { struct radeon_fpriv *fpriv = file_priv->driver_priv; - struct radeon_bo_va *bo_va; + struct radeon_vm *vm = &fpriv->vm; int r; if (rdev->accel_working) { r = radeon_bo_reserve(rdev->ring_tmp_bo.bo, false); if (!r) { - bo_va = radeon_vm_bo_find(&fpriv->vm, - rdev->ring_tmp_bo.bo); - if (bo_va) - radeon_vm_bo_rmv(rdev, bo_va); + if (vm->ib_bo_va) + radeon_vm_bo_rmv(rdev, vm->ib_bo_va); radeon_bo_unreserve(rdev->ring_tmp_bo.bo); } } - radeon_vm_fini(rdev, &fpriv->vm); + radeon_vm_fini(rdev, vm); kfree(fpriv); file_priv->driver_priv = NULL; } diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c index eecff6bbd341..725d3669014f 100644 --- a/drivers/gpu/drm/radeon/radeon_vm.c +++ b/drivers/gpu/drm/radeon/radeon_vm.c @@ -332,6 +332,7 @@ struct radeon_bo_va *radeon_vm_bo_add(struct radeon_device *rdev, bo_va->ref_count = 1; INIT_LIST_HEAD(&bo_va->bo_list); INIT_LIST_HEAD(&bo_va->vm_list); + INIT_LIST_HEAD(&bo_va->vm_status); mutex_lock(&vm->mutex); list_add(&bo_va->vm_list, &vm->va); @@ -468,6 +469,19 @@ int radeon_vm_bo_set_addr(struct radeon_device *rdev, head = &tmp->vm_list; } + if (bo_va->soffset) { + /* add a clone of the bo_va to clear the old address */ + tmp = kzalloc(sizeof(struct radeon_bo_va), GFP_KERNEL); + if (!tmp) { + mutex_unlock(&vm->mutex); + return -ENOMEM; + } + tmp->soffset = bo_va->soffset; + tmp->eoffset = bo_va->eoffset; + tmp->vm = vm; + list_add(&tmp->vm_status, &vm->freed); + } + bo_va->soffset = soffset; bo_va->eoffset = eoffset; bo_va->flags = flags; @@ -823,25 +837,19 @@ static void radeon_vm_update_ptes(struct radeon_device *rdev, * Object have to be reserved and mutex must be locked! */ int radeon_vm_bo_update(struct radeon_device *rdev, - struct radeon_vm *vm, - struct radeon_bo *bo, + struct radeon_bo_va *bo_va, struct ttm_mem_reg *mem) { + struct radeon_vm *vm = bo_va->vm; struct radeon_ib ib; - struct radeon_bo_va *bo_va; unsigned nptes, ndw; uint64_t addr; int r; - bo_va = radeon_vm_bo_find(vm, bo); - if (bo_va == NULL) { - dev_err(rdev->dev, "bo %p not in vm %p\n", bo, vm); - return -EINVAL; - } if (!bo_va->soffset) { dev_err(rdev->dev, "bo %p don't has a mapping in vm %p\n", - bo, vm); + bo_va->bo, vm); return -EINVAL; } @@ -868,7 +876,7 @@ int radeon_vm_bo_update(struct radeon_device *rdev, trace_radeon_vm_bo_update(bo_va); - nptes = radeon_bo_ngpu_pages(bo); + nptes = (bo_va->eoffset - bo_va->soffset) / RADEON_GPU_PAGE_SIZE; /* padding, etc. */ ndw = 64; @@ -910,6 +918,34 @@ int radeon_vm_bo_update(struct radeon_device *rdev, return 0; } +/** + * radeon_vm_clear_freed - clear freed BOs in the PT + * + * @rdev: radeon_device pointer + * @vm: requested vm + * + * Make sure all freed BOs are cleared in the PT. + * Returns 0 for success. + * + * PTs have to be reserved and mutex must be locked! + */ +int radeon_vm_clear_freed(struct radeon_device *rdev, + struct radeon_vm *vm) +{ + struct radeon_bo_va *bo_va, *tmp; + int r; + + list_for_each_entry_safe(bo_va, tmp, &vm->freed, vm_status) { + list_del(&bo_va->vm_status); + r = radeon_vm_bo_update(rdev, bo_va, NULL); + kfree(bo_va); + if (r) + return r; + } + return 0; + +} + /** * radeon_vm_bo_rmv - remove a bo to a specific vm * @@ -917,27 +953,27 @@ int radeon_vm_bo_update(struct radeon_device *rdev, * @bo_va: requested bo_va * * Remove @bo_va->bo from the requested vm (cayman+). - * Remove @bo_va->bo from the list of bos associated with the bo_va->vm and - * remove the ptes for @bo_va in the page table. - * Returns 0 for success. * * Object have to be reserved! */ -int radeon_vm_bo_rmv(struct radeon_device *rdev, - struct radeon_bo_va *bo_va) +void radeon_vm_bo_rmv(struct radeon_device *rdev, + struct radeon_bo_va *bo_va) { - int r = 0; + struct radeon_vm *vm = bo_va->vm; - mutex_lock(&bo_va->vm->mutex); - if (bo_va->soffset) - r = radeon_vm_bo_update(rdev, bo_va->vm, bo_va->bo, NULL); - - list_del(&bo_va->vm_list); - mutex_unlock(&bo_va->vm->mutex); list_del(&bo_va->bo_list); - kfree(bo_va); - return r; + mutex_lock(&vm->mutex); + list_del(&bo_va->vm_list); + + if (bo_va->soffset) { + bo_va->bo = NULL; + list_add(&bo_va->vm_status, &vm->freed); + } else { + kfree(bo_va); + } + + mutex_unlock(&vm->mutex); } /** @@ -975,11 +1011,13 @@ int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm) int r; vm->id = 0; + vm->ib_bo_va = NULL; vm->fence = NULL; vm->last_flush = NULL; vm->last_id_use = NULL; mutex_init(&vm->mutex); INIT_LIST_HEAD(&vm->va); + INIT_LIST_HEAD(&vm->freed); pd_size = radeon_vm_directory_size(rdev); pd_entries = radeon_vm_num_pdes(rdev); @@ -1034,7 +1072,8 @@ void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm) kfree(bo_va); } } - + list_for_each_entry_safe(bo_va, tmp, &vm->freed, vm_status) + kfree(bo_va); for (i = 0; i < radeon_vm_num_pdes(rdev); i++) radeon_bo_unref(&vm->page_tables[i].bo); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index eba0225259a4..9e854fd016da 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6103,6 +6103,7 @@ static inline u32 si_get_ih_wptr(struct radeon_device *rdev) tmp = RREG32(IH_RB_CNTL); tmp |= IH_WPTR_OVERFLOW_CLEAR; WREG32(IH_RB_CNTL, tmp); + wptr &= ~RB_OVERFLOW; } return (wptr & rdev->ih.ptr_mask); } diff --git a/drivers/gpu/drm/radeon/trinity_dpm.c b/drivers/gpu/drm/radeon/trinity_dpm.c index 20da6ff183df..32e50be9c4ac 100644 --- a/drivers/gpu/drm/radeon/trinity_dpm.c +++ b/drivers/gpu/drm/radeon/trinity_dpm.c @@ -1874,15 +1874,16 @@ int trinity_dpm_init(struct radeon_device *rdev) for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) pi->at[i] = TRINITY_AT_DFLT; - /* There are stability issues reported on latops with - * bapm installed when switching between AC and battery - * power. At the same time, some desktop boards hang - * if it's not enabled and dpm is enabled. + /* There are stability issues reported on with + * bapm enabled when switching between AC and battery + * power. At the same time, some MSI boards hang + * if it's not enabled and dpm is enabled. Just enable + * it for MSI boards right now. */ - if (rdev->flags & RADEON_IS_MOBILITY) - pi->enable_bapm = false; - else + if (rdev->pdev->subsystem_vendor == 0x1462) pi->enable_bapm = true; + else + pi->enable_bapm = false; pi->enable_nbps_policy = true; pi->enable_sclk_ds = true; pi->enable_gfx_power_gating = true;