From d7d7b947a4fa6d0a82ff2bf0db413edc63738e3a Mon Sep 17 00:00:00 2001 From: Lang Yu Date: Fri, 27 Sep 2024 18:27:46 +0800 Subject: [PATCH 1/6] drm/amdkfd: Fix an eviction fence leak Only creating a new reference for each process instead of each VM. Fixes: 9a1c1339abf9 ("drm/amdkfd: Run restore_workers on freezable WQs") Suggested-by: Felix Kuehling Signed-off-by: Lang Yu Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher (cherry picked from commit 5fa436289483ae56427b0896c31f72361223c758) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 4 ++-- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 7 +++++-- 2 files changed, 7 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index ce5ca304dba9..fa572ba7f9fc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -1439,8 +1439,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, list_add_tail(&vm->vm_list_node, &(vm->process_info->vm_list_head)); vm->process_info->n_vms++; - - *ef = dma_fence_get(&vm->process_info->eviction_fence->base); + if (ef) + *ef = dma_fence_get(&vm->process_info->eviction_fence->base); mutex_unlock(&vm->process_info->lock); return 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index d07acf1b2f93..d665ecdcd12f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -1702,12 +1702,15 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd, ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(dev->adev, avm, &p->kgd_process_info, - &ef); + p->ef ? NULL : &ef); if (ret) { dev_err(dev->adev->dev, "Failed to create process VM object\n"); return ret; } - RCU_INIT_POINTER(p->ef, ef); + + if (!p->ef) + RCU_INIT_POINTER(p->ef, ef); + pdd->drm_priv = drm_file->private_data; ret = kfd_process_device_reserve_ib_mem(pdd); From 8e82d1199564ea0baab7be31f6558406bab4fe1d Mon Sep 17 00:00:00 2001 From: Wu Hoi Pok Date: Fri, 4 Oct 2024 09:05:59 +0800 Subject: [PATCH 2/6] drm/radeon: add late_register for connector The patch is to solve null dereference in 'aux.dev', which is introduced in recent radeon rework. By having 'late_register', the connector should be registered after 'drm_dev_register' automatically, where in before it is the opposite. Fixes: 90985660ba48 ("drm/radeon: remove load callback from kms_driver") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3665 Tested-by: Hans de Goede Suggested-by: Christophe Leroy Signed-off-by: Wu Hoi Pok Signed-off-by: Alex Deucher (cherry picked from commit b4c1ad70e279bacbc772a468033bdecce2f5e0dc) --- drivers/gpu/drm/radeon/atombios_dp.c | 9 ++------- drivers/gpu/drm/radeon/radeon_connectors.c | 17 +++++++++++++++++ 2 files changed, 19 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/radeon/atombios_dp.c b/drivers/gpu/drm/radeon/atombios_dp.c index fca8b08535a5..6328627b7c34 100644 --- a/drivers/gpu/drm/radeon/atombios_dp.c +++ b/drivers/gpu/drm/radeon/atombios_dp.c @@ -228,10 +228,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector) { struct drm_device *dev = radeon_connector->base.dev; struct radeon_device *rdev = dev->dev_private; - int ret; radeon_connector->ddc_bus->rec.hpd = radeon_connector->hpd.hpd; - radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev; radeon_connector->ddc_bus->aux.drm_dev = radeon_connector->base.dev; if (ASIC_IS_DCE5(rdev)) { if (radeon_auxch) @@ -242,11 +240,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector) radeon_connector->ddc_bus->aux.transfer = radeon_dp_aux_transfer_atom; } - ret = drm_dp_aux_register(&radeon_connector->ddc_bus->aux); - if (!ret) - radeon_connector->ddc_bus->has_aux = true; - - WARN(ret, "drm_dp_aux_register() failed with error %d\n", ret); + drm_dp_aux_init(&radeon_connector->ddc_bus->aux); + radeon_connector->ddc_bus->has_aux = true; } /***** general DP utility functions *****/ diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c index 528a8f3677c2..f9c73c55f04f 100644 --- a/drivers/gpu/drm/radeon/radeon_connectors.c +++ b/drivers/gpu/drm/radeon/radeon_connectors.c @@ -1786,6 +1786,20 @@ static enum drm_mode_status radeon_dp_mode_valid(struct drm_connector *connector return MODE_OK; } +static int +radeon_connector_late_register(struct drm_connector *connector) +{ + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + int r = 0; + + if (radeon_connector->ddc_bus->has_aux) { + radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev; + r = drm_dp_aux_register(&radeon_connector->ddc_bus->aux); + } + + return r; +} + static const struct drm_connector_helper_funcs radeon_dp_connector_helper_funcs = { .get_modes = radeon_dp_get_modes, .mode_valid = radeon_dp_mode_valid, @@ -1800,6 +1814,7 @@ static const struct drm_connector_funcs radeon_dp_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; static const struct drm_connector_funcs radeon_edp_connector_funcs = { @@ -1810,6 +1825,7 @@ static const struct drm_connector_funcs radeon_edp_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = { @@ -1820,6 +1836,7 @@ static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; void From d6b9f492e229be1d1bd360c3ac5bee4635bacf99 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 2 Oct 2024 17:27:25 -0400 Subject: [PATCH 3/6] drm/amdgpu: partially revert powerplay `__counted_by` changes Partially revert commit 0ca9f757a0e2 ("drm/amd/pm: powerplay: Add `__counted_by` attribute for flexible arrays") The count attribute for these arrays does not get set until after the arrays are allocated and populated leading to false UBSAN warnings. Fixes: 0ca9f757a0e2 ("drm/amd/pm: powerplay: Add `__counted_by` attribute for flexible arrays") Reviewed-by: Mario Limonciello Reviewed-by: Lijo Lazar Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3662 Signed-off-by: Alex Deucher (cherry picked from commit 8a5ae927b653b43623e55610d2215ee94c027e8c) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h | 26 ++++++++++---------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h index 9118fcddbf11..227bf0e84a13 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h @@ -60,7 +60,7 @@ struct vi_dpm_level { struct vi_dpm_table { uint32_t count; - struct vi_dpm_level dpm_level[] __counted_by(count); + struct vi_dpm_level dpm_level[]; }; #define PCIE_PERF_REQ_REMOVE_REGISTRY 0 @@ -91,7 +91,7 @@ struct phm_set_power_state_input { struct phm_clock_array { uint32_t count; - uint32_t values[] __counted_by(count); + uint32_t values[]; }; struct phm_clock_voltage_dependency_record { @@ -123,7 +123,7 @@ struct phm_acpclock_voltage_dependency_record { struct phm_clock_voltage_dependency_table { uint32_t count; - struct phm_clock_voltage_dependency_record entries[] __counted_by(count); + struct phm_clock_voltage_dependency_record entries[]; }; struct phm_phase_shedding_limits_record { @@ -140,7 +140,7 @@ struct phm_uvd_clock_voltage_dependency_record { struct phm_uvd_clock_voltage_dependency_table { uint8_t count; - struct phm_uvd_clock_voltage_dependency_record entries[] __counted_by(count); + struct phm_uvd_clock_voltage_dependency_record entries[]; }; struct phm_acp_clock_voltage_dependency_record { @@ -150,7 +150,7 @@ struct phm_acp_clock_voltage_dependency_record { struct phm_acp_clock_voltage_dependency_table { uint32_t count; - struct phm_acp_clock_voltage_dependency_record entries[] __counted_by(count); + struct phm_acp_clock_voltage_dependency_record entries[]; }; struct phm_vce_clock_voltage_dependency_record { @@ -161,32 +161,32 @@ struct phm_vce_clock_voltage_dependency_record { struct phm_phase_shedding_limits_table { uint32_t count; - struct phm_phase_shedding_limits_record entries[] __counted_by(count); + struct phm_phase_shedding_limits_record entries[]; }; struct phm_vceclock_voltage_dependency_table { uint8_t count; - struct phm_vceclock_voltage_dependency_record entries[] __counted_by(count); + struct phm_vceclock_voltage_dependency_record entries[]; }; struct phm_uvdclock_voltage_dependency_table { uint8_t count; - struct phm_uvdclock_voltage_dependency_record entries[] __counted_by(count); + struct phm_uvdclock_voltage_dependency_record entries[]; }; struct phm_samuclock_voltage_dependency_table { uint8_t count; - struct phm_samuclock_voltage_dependency_record entries[] __counted_by(count); + struct phm_samuclock_voltage_dependency_record entries[]; }; struct phm_acpclock_voltage_dependency_table { uint32_t count; - struct phm_acpclock_voltage_dependency_record entries[] __counted_by(count); + struct phm_acpclock_voltage_dependency_record entries[]; }; struct phm_vce_clock_voltage_dependency_table { uint8_t count; - struct phm_vce_clock_voltage_dependency_record entries[] __counted_by(count); + struct phm_vce_clock_voltage_dependency_record entries[]; }; @@ -393,7 +393,7 @@ union phm_cac_leakage_record { struct phm_cac_leakage_table { uint32_t count; - union phm_cac_leakage_record entries[] __counted_by(count); + union phm_cac_leakage_record entries[]; }; struct phm_samu_clock_voltage_dependency_record { @@ -404,7 +404,7 @@ struct phm_samu_clock_voltage_dependency_record { struct phm_samu_clock_voltage_dependency_table { uint8_t count; - struct phm_samu_clock_voltage_dependency_record entries[] __counted_by(count); + struct phm_samu_clock_voltage_dependency_record entries[]; }; struct phm_cac_tdp_table { From 0a9906cc45d21e21ca8bb2b98b79fd7c05420fda Mon Sep 17 00:00:00 2001 From: Josip Pavic Date: Tue, 24 Sep 2024 17:25:54 -0400 Subject: [PATCH 4/6] drm/amd/display: Clear update flags after update has been applied [Why] Since the surface/stream update flags aren't cleared after applying updates, those same updates may be applied again in a future call to update surfaces/streams for surfaces/streams that aren't actually part of that update (i.e. applying an update for one surface/stream can trigger unintended programming on a different surface/stream). For example, when an update results in a call to program_front_end_for_ctx, that function may call program_pipe on all pipes. If there are surface update flags that were never cleared on the surface some pipe is attached to, then the same update will be programmed again. [How] Clear the surface and stream update flags after applying the updates. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3441 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3616 Cc: Melissa Wen Reviewed-by: Aric Cyr Signed-off-by: Josip Pavic Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 7671f62c10f2a4c77d89b39fd50fab7f918d6809) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/dc/core/dc.c | 45 ++++++++++++++++++------ 1 file changed, 34 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 5c39390ecbd5..a88f1b6ea64c 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -5065,11 +5065,26 @@ static bool update_planes_and_stream_v3(struct dc *dc, return true; } +static void clear_update_flags(struct dc_surface_update *srf_updates, + int surface_count, struct dc_stream_state *stream) +{ + int i; + + if (stream) + stream->update_flags.raw = 0; + + for (i = 0; i < surface_count; i++) + if (srf_updates[i].surface) + srf_updates[i].surface->update_flags.raw = 0; +} + bool dc_update_planes_and_stream(struct dc *dc, struct dc_surface_update *srf_updates, int surface_count, struct dc_stream_state *stream, struct dc_stream_update *stream_update) { + bool ret = false; + dc_exit_ips_for_hw_access(dc); /* * update planes and stream version 3 separates FULL and FAST updates @@ -5086,10 +5101,16 @@ bool dc_update_planes_and_stream(struct dc *dc, * features as they are now transparent to the new sequence. */ if (dc->ctx->dce_version >= DCN_VERSION_4_01) - return update_planes_and_stream_v3(dc, srf_updates, + ret = update_planes_and_stream_v3(dc, srf_updates, surface_count, stream, stream_update); - return update_planes_and_stream_v2(dc, srf_updates, + else + ret = update_planes_and_stream_v2(dc, srf_updates, surface_count, stream, stream_update); + + if (ret) + clear_update_flags(srf_updates, surface_count, stream); + + return ret; } void dc_commit_updates_for_stream(struct dc *dc, @@ -5099,6 +5120,8 @@ void dc_commit_updates_for_stream(struct dc *dc, struct dc_stream_update *stream_update, struct dc_state *state) { + bool ret = false; + dc_exit_ips_for_hw_access(dc); /* TODO: Since change commit sequence can have a huge impact, * we decided to only enable it for DCN3x. However, as soon as @@ -5106,17 +5129,17 @@ void dc_commit_updates_for_stream(struct dc *dc, * the new sequence for all ASICs. */ if (dc->ctx->dce_version >= DCN_VERSION_4_01) { - update_planes_and_stream_v3(dc, srf_updates, surface_count, + ret = update_planes_and_stream_v3(dc, srf_updates, surface_count, stream, stream_update); - return; - } - if (dc->ctx->dce_version >= DCN_VERSION_3_2) { - update_planes_and_stream_v2(dc, srf_updates, surface_count, + } else if (dc->ctx->dce_version >= DCN_VERSION_3_2) { + ret = update_planes_and_stream_v2(dc, srf_updates, surface_count, stream, stream_update); - return; - } - update_planes_and_stream_v1(dc, srf_updates, surface_count, stream, - stream_update, state); + } else + ret = update_planes_and_stream_v1(dc, srf_updates, surface_count, stream, + stream_update, state); + + if (ret) + clear_update_flags(srf_updates, surface_count, stream); } uint8_t dc_get_current_stream_count(struct dc *dc) From 79bc412ef787cf25773d0ece93f8739ce0e6ac1e Mon Sep 17 00:00:00 2001 From: Hamza Mahfooz Date: Fri, 4 Oct 2024 15:22:57 -0400 Subject: [PATCH 5/6] drm/amd/display: fix hibernate entry for DCN35+ Since, two suspend-resume cycles are required to enter hibernate and, since we only need to enable idle optimizations in the first cycle (which is pretty much equivalent to s2idle). We can check in_s0ix, to prevent the system from entering idle optimizations before it actually enters hibernate (from display's perspective). Also, call dc_set_power_state() before dc_allow_idle_optimizations(), since it's safer to do so because dc_set_power_state() writes to DMUB. Acked-by: Alex Deucher Signed-off-by: Hamza Mahfooz Signed-off-by: Alex Deucher (cherry picked from commit 2fe79508d9c393bb9931b0037c5ecaee09a8dc39) Cc: stable@vger.kernel.org # 6.10+ --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 60c617fcc97e..6b5e2206e687 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -2972,10 +2972,11 @@ static int dm_suspend(void *handle) hpd_rx_irq_work_suspend(dm); - if (adev->dm.dc->caps.ips_support) - dc_allow_idle_optimizations(adev->dm.dc, true); - dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D3); + + if (dm->dc->caps.ips_support && adev->in_s0ix) + dc_allow_idle_optimizations(dm->dc, true); + dc_dmub_srv_set_power_state(dm->dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D3); return 0; From 32fda5650625065b8ae6dbd9967fb572d3835c35 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Mon, 7 Oct 2024 20:30:19 +0200 Subject: [PATCH 6/6] drm/radeon: always set GEM function pointer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make sure to always set the GEM function pointer even for in kernel allocations. This fixes a NULL pointer deref caused by switching to GEM references. Signed-off-by: Christian König Fixes: fd69ef05029f ("drm/radeon: use GEM references instead of TTMs") Acked-by: Alex Deucher Tested-by: Hans de Goede Signed-off-by: Alex Deucher (cherry picked from commit 01b64bc063d014641631867a7e0edd8ac55282d4) --- drivers/gpu/drm/radeon/radeon_gem.c | 3 --- drivers/gpu/drm/radeon/radeon_object.c | 1 + 2 files changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index 9735f4968b86..bf2d4b16dc2a 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -44,8 +44,6 @@ struct sg_table *radeon_gem_prime_get_sg_table(struct drm_gem_object *obj); int radeon_gem_prime_pin(struct drm_gem_object *obj); void radeon_gem_prime_unpin(struct drm_gem_object *obj); -const struct drm_gem_object_funcs radeon_gem_object_funcs; - static vm_fault_t radeon_gem_fault(struct vm_fault *vmf) { struct ttm_buffer_object *bo = vmf->vma->vm_private_data; @@ -132,7 +130,6 @@ int radeon_gem_object_create(struct radeon_device *rdev, unsigned long size, return r; } *obj = &robj->tbo.base; - (*obj)->funcs = &radeon_gem_object_funcs; robj->pid = task_pid_nr(current); mutex_lock(&rdev->gem.mutex); diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index d0e4b43d155c..7672404fdb29 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -151,6 +151,7 @@ int radeon_bo_create(struct radeon_device *rdev, if (bo == NULL) return -ENOMEM; drm_gem_private_object_init(rdev_to_drm(rdev), &bo->tbo.base, size); + bo->tbo.base.funcs = &radeon_gem_object_funcs; bo->rdev = rdev; bo->surface_reg = -1; INIT_LIST_HEAD(&bo->list);