dri-devel October 2017

dri-devel@lists.freedesktop.org

228 participants
1625 discussions

[PATCH hwc] drm_hwcomposer: fix return values for ValidateDisplay
by Rob Herring 09 Oct '17

09 Oct '17

According to the API documentation, if we've set num_types, then the return value should be HWC2::Error::HasChanges. Signed-off-by: Rob Herring <robh(a)kernel.org> --- drmhwctwo.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drmhwctwo.cpp b/drmhwctwo.cpp index 89399bf1ba33..0f258770a8bb 100644 --- a/drmhwctwo.cpp +++ b/drmhwctwo.cpp @@ -701,7 +701,7 @@ HWC2::Error DrmHwcTwo::HwcDisplay::ValidateDisplay(uint32_t *num_types, break; } } - return HWC2::Error::None; + return *num_types ? HWC2::Error::HasChanges : HWC2::Error::None; } HWC2::Error DrmHwcTwo::HwcLayer::SetCursorPosition(int32_t x, int32_t y) { -- 2.11.0

1 0

[PATCH] drm/ttm: don't use compound pages for now
by Christian König 09 Oct '17

09 Oct '17

From: Christian König <christian.koenig(a)amd.com> We need to figure out first how to correctly map them into the CPU page tables. Signed-off-by: Christian König <christian.koenig(a)amd.com> --- drivers/gpu/drm/ttm/ttm_page_alloc_dma.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c index e5ef10d..96ad129 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c @@ -913,6 +913,7 @@ static gfp_t ttm_dma_pool_gfp_flags(struct ttm_dma_tt *ttm_dma, bool huge) if (huge) { gfp_flags |= GFP_TRANSHUGE; gfp_flags &= ~__GFP_MOVABLE; + gfp_flags &= ~__GFP_COMP; } return gfp_flags; -- 2.7.4

1 0

[PATCH v7 0/3] drm/tinydrm: drm_of_find_backlight helper
by Meghana Madhyastha 09 Oct '17

09 Oct '17

Move tinydrm_of_find_backlight to drm_of.c and rename it to drm_of_find_backlight for better organizational structure. Changes in v7: -Move the function definitions/declarations back to drm_of.h and modify the config option for of_find_backlight_by_node in linux/backlight.h to resolve the build errors. Meghana Madhyastha (3): backlight: Add IS_ENABLED(CONFIG_BACKLIGHT_CLASS_DEVICE) drm/tinydrm: Move tinydrm_of_find_backlight into drm_of.c drm/tinydrm: Add devres versions of drm_of_find_backlight drivers/gpu/drm/drm_of.c | 85 ++++++++++++++++++++++++++ drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c | 40 ------------ drivers/gpu/drm/tinydrm/mi0283qt.c | 3 +- include/drm/drm_of.h | 14 +++++ include/drm/tinydrm/tinydrm-helpers.h | 1 - include/linux/backlight.h | 2 +- 6 files changed, 102 insertions(+), 43 deletions(-) -- 2.7.4

5 15

[Bug 93301] ns2_linux32: radeon VM fault / VM_CONTEXT1_PROTECTION_FAULT_ADDR
by bugzilla-daemon＠freedesktop.org 09 Oct '17

09 Oct '17

https://bugs.freedesktop.org/show_bug.cgi?id=93301 Bug ID: 93301 Summary: ns2_linux32: radeon VM fault / VM_CONTEXT1_PROTECTION_FAULT_ADDR Product: Mesa Version: git Hardware: Other OS: All Status: NEW Severity: normal Priority: medium Component: Drivers/Gallium/radeonsi Assignee: dri-devel(a)lists.freedesktop.org Reporter: dev(a)illwieckz.net QA Contact: dri-devel(a)lists.freedesktop.org Created attachment 120427 --> https://bugs.freedesktop.org/attachment.cgi?id=120427&action=edit syslog (Linux 4.3, mesa git, radeon hawaii r9 390X) Hi I (re) tried Natural Selection 2 yesterday, and I it crashed (as expected), and this time the syslog said: ``` Dec 8 07:20:25 gollum kernel: [16588.241364] radeon 0000:01:00.0: GPU fault detected: 146 0x0e65e014 Dec 8 07:20:25 gollum kernel: [16588.241368] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x00007673 Dec 8 07:20:25 gollum kernel: [16588.241369] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x050E0014 Dec 8 07:20:25 gollum kernel: [16588.241370] VM fault (0x04, vmid 2) at page 30323, write from 'CB1' (0x43423100) (224) ---8<--- 120 000+ lines --------------------- Dec 8 07:22:58 gollum kernel: [16741.758495] radeon 0000:01:00.0: GPU fault detected: 146 0x06a59014 Dec 8 07:22:58 gollum kernel: [16741.758506] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x00007673 Dec 8 07:22:58 gollum kernel: [16741.758512] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x05150014 Dec 8 07:22:58 gollum kernel: [16741.758518] VM fault (0x04, vmid 2) at page 30323, write from 'CB6' (0x43423600) (336) Dec 8 07:23:03 gollum kernel: [16746.694729] ns2_linux32[16192]: segfault at 1 ip 00000000f3bf8d12 sp 00000000f0b9adf8 error 4 in libc-2.21.so[f3acc000+1b4000] ``` Well, I'm not expecting to see this game working one day, I own this game since 2013-12-15 (it's almost two years) and it never worked, so these guys proved their product is a joke and I don't expect nothing from them. But perhaps this bug can help you to track things on your side, that's why I report it. When the error happened, NS2 was loading assets since 20min to load a tutorial map. I don't have more logs currently but if you have some ideas of things I can record, I can try to reproduce it. I agree the current log is a bit quiet. I've already seen radeon errors like that before in a dpm issue (see #93288 ) but the current bug is unrelated to dpm (dpm was inactive in the current case). I use a Radeon R9 390X with Mesa git on Linux 4.3 here. -- You are receiving this mail because: You are the assignee for the bug.

1 33

[PATCH hwc] drm_hwcomposer: use upstream DRM_MODE_{ROTATE, REFLECT} defines
by Rob Herring 09 Oct '17

09 Oct '17

The DRM_REFLECT_* and DRM_ROTATE_* defines were not upstream, but now they are. Convert to using the upstream version which are defined as a bit mask rather than a bit position. Signed-off-by: Rob Herring <robh(a)kernel.org> --- drmdisplaycompositor.cpp | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drmdisplaycompositor.cpp b/drmdisplaycompositor.cpp index a07d3588d1f5..036a45e8b88b 100644 --- a/drmdisplaycompositor.cpp +++ b/drmdisplaycompositor.cpp @@ -572,15 +572,15 @@ int DrmDisplayCompositor::CommitFrame(DrmDisplayComposition *display_comp, rotation = 0; if (layer.transform & DrmHwcTransform::kFlipH) - rotation |= 1 << DRM_REFLECT_X; + rotation |= DRM_MODE_REFLECT_X; if (layer.transform & DrmHwcTransform::kFlipV) - rotation |= 1 << DRM_REFLECT_Y; + rotation |= DRM_MODE_REFLECT_Y; if (layer.transform & DrmHwcTransform::kRotate90) - rotation |= 1 << DRM_ROTATE_90; + rotation |= DRM_MODE_ROTATE_90; else if (layer.transform & DrmHwcTransform::kRotate180) - rotation |= 1 << DRM_ROTATE_180; + rotation |= DRM_MODE_ROTATE_180; else if (layer.transform & DrmHwcTransform::kRotate270) - rotation |= 1 << DRM_ROTATE_270; + rotation |= DRM_MODE_ROTATE_270; if (fence_fd < 0) { int prop_id = plane->in_fence_fd_property().id(); -- 2.11.0

3 3

[PATCH hwc] drm_hwcomposer: remove NVIDIA importer
by Rob Herring 09 Oct '17

09 Oct '17

There's no opensource implementation for the NVIDIA gralloc implementation, so remove it as it is not testable. As all of the gralloc perform() operations are specific to it, they can be removed, too. Signed-off-by: Rob Herring <robh(a)kernel.org> --- Android.mk | 5 - drmhwcgralloc.h | 33 ----- platformnv.cpp | 374 -------------------------------------------------------- 3 files changed, 412 deletions(-) delete mode 100644 platformnv.cpp diff --git a/Android.mk b/Android.mk index 99bbcac3e984..d7dc6782e1e0 100644 --- a/Android.mk +++ b/Android.mk @@ -71,7 +71,6 @@ LOCAL_SRC_FILES := \ hwcutils.cpp \ platform.cpp \ platformdrmgeneric.cpp \ - platformnv.cpp \ separate_rects.cpp \ virtualcompositorworker.cpp \ vsyncworker.cpp @@ -80,11 +79,7 @@ LOCAL_CPPFLAGS += \ -DHWC2_USE_CPP11 \ -DHWC2_INCLUDE_STRINGIFICATION -ifeq ($(strip $(BOARD_DRM_HWCOMPOSER_BUFFER_IMPORTER)),nvidia-gralloc) -LOCAL_CPPFLAGS += -DUSE_NVIDIA_IMPORTER -else LOCAL_CPPFLAGS += -DUSE_DRM_GENERIC_IMPORTER -endif LOCAL_MODULE := hwcomposer.drm LOCAL_MODULE_TAGS := optional diff --git a/drmhwcgralloc.h b/drmhwcgralloc.h index c4a42eaf1c54..759746a67949 100644 --- a/drmhwcgralloc.h +++ b/drmhwcgralloc.h @@ -19,39 +19,6 @@ #include <stdint.h> -enum { - /* perform(const struct gralloc_module_t *mod, - * int op, - * int drm_fd, - * buffer_handle_t buffer, - * struct hwc_drm_bo *bo); - */ - GRALLOC_MODULE_PERFORM_DRM_IMPORT = 0xffeeff00, - - /* perform(const struct gralloc_module_t *mod, - * int op, - * buffer_handle_t buffer, - * void (*free_callback)(void *), - * void *priv); - */ - GRALLOC_MODULE_PERFORM_SET_IMPORTER_PRIVATE = 0xffeeff01, - - /* perform(const struct gralloc_module_t *mod, - * int op, - * buffer_handle_t buffer, - * void (*free_callback)(void *), - * void **priv); - */ - GRALLOC_MODULE_PERFORM_GET_IMPORTER_PRIVATE = 0xffeeff02, - - /* perform(const struct gralloc_module_t *mod, - * int op, - * buffer_handle_t buffer, - * int *usage); - */ - GRALLOC_MODULE_PERFORM_GET_USAGE = 0xffeeff03, -}; - typedef struct hwc_drm_bo { uint32_t width; uint32_t height; diff --git a/platformnv.cpp b/platformnv.cpp deleted file mode 100644 index e7b6be3430a8..000000000000 --- a/platformnv.cpp +++ /dev/null @@ -1,374 +0,0 @@ -/* - * Copyright (C) 2015 The Android Open Source Project - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -#define LOG_TAG "hwc-platform-nv" - -#include "drmresources.h" -#include "platform.h" -#include "platformnv.h" - -#include <cinttypes> -#include <stdatomic.h> -#include <drm/drm_fourcc.h> -#include <xf86drm.h> -#include <xf86drmMode.h> - -#include <cutils/log.h> -#include <hardware/gralloc.h> - -#ifndef EGL_NATIVE_HANDLE_ANDROID_NVX -#define EGL_NATIVE_HANDLE_ANDROID_NVX 0x322A -#endif - -namespace android { - -#ifdef USE_NVIDIA_IMPORTER -// static -Importer *Importer::CreateInstance(DrmResources *drm) { - NvImporter *importer = new NvImporter(drm); - if (!importer) - return NULL; - - int ret = importer->Init(); - if (ret) { - ALOGE("Failed to initialize the nv importer %d", ret); - delete importer; - return NULL; - } - return importer; -} -#endif - -NvImporter::NvImporter(DrmResources *drm) : drm_(drm) { -} - -NvImporter::~NvImporter() { -} - -int NvImporter::Init() { - int ret = hw_get_module(GRALLOC_HARDWARE_MODULE_ID, - (const hw_module_t **)&gralloc_); - if (ret) { - ALOGE("Failed to open gralloc module %d", ret); - return ret; - } - - if (strcasecmp(gralloc_->common.author, "NVIDIA")) - ALOGW("Using non-NVIDIA gralloc module: %s/%s\n", gralloc_->common.name, - gralloc_->common.author); - - return 0; -} - - -EGLImageKHR NvImporter::ImportImage(EGLDisplay egl_display, buffer_handle_t handle) { - return eglCreateImageKHR( - egl_display, EGL_NO_CONTEXT, EGL_NATIVE_HANDLE_ANDROID_NVX, - (EGLClientBuffer)handle, NULL /* no attribs */); -} - -int NvImporter::ImportBuffer(buffer_handle_t handle, hwc_drm_bo_t *bo) { - memset(bo, 0, sizeof(hwc_drm_bo_t)); - NvBuffer_t *buf = GrallocGetNvBuffer(handle); - if (buf) { - atomic_fetch_add(&buf->ref, 1); - *bo = buf->bo; - return 0; - } - - buf = new NvBuffer_t(); - if (!buf) { - ALOGE("Failed to allocate new NvBuffer_t"); - return -ENOMEM; - } - buf->bo.priv = buf; - buf->importer = this; - - // We initialize the reference count to 2 since NvGralloc is still using this - // buffer (will be cleared in the NvGrallocRelease), and the other - // reference is for HWC (this ImportBuffer call). - atomic_init(&buf->ref, 2); - - int ret = gralloc_->perform(gralloc_, GRALLOC_MODULE_PERFORM_DRM_IMPORT, - drm_->fd(), handle, &buf->bo); - if (ret) { - ALOGE("GRALLOC_MODULE_PERFORM_DRM_IMPORT failed %d", ret); - delete buf; - return ret; - } - - ret = drmModeAddFB2(drm_->fd(), buf->bo.width, buf->bo.height, buf->bo.format, - buf->bo.gem_handles, buf->bo.pitches, buf->bo.offsets, - &buf->bo.fb_id, 0); - if (ret) { - ALOGE("Failed to add fb %d", ret); - ReleaseBufferImpl(&buf->bo); - delete buf; - return ret; - } - - ret = GrallocSetNvBuffer(handle, buf); - if (ret) { - /* This will happen is persist.tegra.gpu_mapping_cache is 0/off, - * or if NV gralloc runs out of "priv slots" (currently 3 per buffer, - * only one of which should be used by drm_hwcomposer). */ - ALOGE("Failed to register free callback for imported buffer %d", ret); - ReleaseBufferImpl(&buf->bo); - delete buf; - return ret; - } - *bo = buf->bo; - return 0; -} - -int NvImporter::ReleaseBuffer(hwc_drm_bo_t *bo) { - NvBuffer_t *buf = (NvBuffer_t *)bo->priv; - if (!buf) { - ALOGE("Freeing bo %" PRIu32 ", buf is NULL!", bo->fb_id); - return 0; - } - if (atomic_fetch_sub(&buf->ref, 1) > 1) - return 0; - - ReleaseBufferImpl(bo); - delete buf; - return 0; -} - -// static -void NvImporter::NvGrallocRelease(void *nv_buffer) { - NvBuffer_t *buf = (NvBuffer *)nv_buffer; - buf->importer->ReleaseBuffer(&buf->bo); -} - -void NvImporter::ReleaseBufferImpl(hwc_drm_bo_t *bo) { - if (bo->fb_id) { - int ret = drmModeRmFB(drm_->fd(), bo->fb_id); - if (ret) - ALOGE("Failed to rm fb %d", ret); - } - - struct drm_gem_close gem_close; - memset(&gem_close, 0, sizeof(gem_close)); - int num_gem_handles = sizeof(bo->gem_handles) / sizeof(bo->gem_handles[0]); - for (int i = 0; i < num_gem_handles; i++) { - if (!bo->gem_handles[i]) - continue; - - gem_close.handle = bo->gem_handles[i]; - int ret = drmIoctl(drm_->fd(), DRM_IOCTL_GEM_CLOSE, &gem_close); - if (ret) { - ALOGE("Failed to close gem handle %d %d", i, ret); - } else { - /* Clear any duplicate gem handle as well but don't close again */ - for (int j = i + 1; j < num_gem_handles; j++) - if (bo->gem_handles[j] == bo->gem_handles[i]) - bo->gem_handles[j] = 0; - bo->gem_handles[i] = 0; - } - } -} - -NvImporter::NvBuffer_t *NvImporter::GrallocGetNvBuffer(buffer_handle_t handle) { - void *priv = NULL; - int ret = - gralloc_->perform(gralloc_, GRALLOC_MODULE_PERFORM_GET_IMPORTER_PRIVATE, - handle, NvGrallocRelease, &priv); - return ret ? NULL : (NvBuffer_t *)priv; -} - -int NvImporter::GrallocSetNvBuffer(buffer_handle_t handle, NvBuffer_t *buf) { - return gralloc_->perform(gralloc_, - GRALLOC_MODULE_PERFORM_SET_IMPORTER_PRIVATE, handle, - NvGrallocRelease, buf); -} - -#ifdef USE_NVIDIA_IMPORTER -// static -std::unique_ptr<Planner> Planner::CreateInstance(DrmResources *) { - std::unique_ptr<Planner> planner(new Planner); - planner->AddStage<PlanStageNvLimits>(); - planner->AddStage<PlanStageProtectedRotated>(); - planner->AddStage<PlanStageProtected>(); - planner->AddStage<PlanStagePrecomp>(); - planner->AddStage<PlanStageGreedy>(); - return planner; -} -#endif - -static DrmPlane *GetCrtcPrimaryPlane(DrmCrtc *crtc, - std::vector<DrmPlane *> *planes) { - for (auto i = planes->begin(); i != planes->end(); ++i) { - if ((*i)->GetCrtcSupported(*crtc)) { - DrmPlane *plane = *i; - planes->erase(i); - return plane; - } - } - return NULL; -} - -int PlanStageProtectedRotated::ProvisionPlanes( - std::vector<DrmCompositionPlane> *composition, - std::map<size_t, DrmHwcLayer *> &layers, DrmCrtc *crtc, - std::vector<DrmPlane *> *planes) { - int ret; - int protected_zorder = -1; - for (auto i = layers.begin(); i != layers.end();) { - if (!i->second->protected_usage() || !i->second->transform) { - ++i; - continue; - } - - auto primary_iter = planes->begin(); - for (; primary_iter != planes->end(); ++primary_iter) { - if ((*primary_iter)->type() == DRM_PLANE_TYPE_PRIMARY) - break; - } - - // We cheat a little here. Since there can only be one primary plane per - // crtc, we know we'll only hit this case once. So we blindly insert the - // protected content at the beginning of the composition, knowing this path - // won't be taken a second time during the loop. - if (primary_iter != planes->end()) { - composition->emplace(composition->begin(), - DrmCompositionPlane::Type::kLayer, *primary_iter, - crtc, i->first); - planes->erase(primary_iter); - protected_zorder = i->first; - } else { - ALOGE("Could not provision primary plane for protected/rotated layer"); - } - i = layers.erase(i); - } - - if (protected_zorder == -1) - return 0; - - // Add any layers below the protected content to the precomposition since we - // need to punch a hole through them. - for (auto i = layers.begin(); i != layers.end();) { - // Skip layers above the z-order of the protected content - if (i->first > static_cast<size_t>(protected_zorder)) { - ++i; - continue; - } - - // If there's no precomp layer already queued, queue one now. - DrmCompositionPlane *precomp = GetPrecomp(composition); - if (precomp) { - precomp->source_layers().emplace_back(i->first); - } else { - if (planes->size()) { - DrmPlane *precomp_plane = planes->back(); - planes->pop_back(); - composition->emplace_back(DrmCompositionPlane::Type::kPrecomp, - precomp_plane, crtc, i->first); - } else { - ALOGE("Not enough planes to reserve for precomp fb"); - } - } - i = layers.erase(i); - } - return 0; -} - -bool PlanStageNvLimits::CheckLayer(size_t zorder, DrmHwcLayer *layer) { - auto src_w = layer->source_crop.width(); - auto src_h = layer->source_crop.height(); - auto dst_w = layer->display_frame.width(); - auto dst_h = layer->display_frame.height(); - int h_limit = 4; - int v_limit; - - switch (layer->buffer->format) { - case DRM_FORMAT_ARGB8888: - case DRM_FORMAT_ABGR8888: - case DRM_FORMAT_XBGR8888: - case DRM_FORMAT_XRGB8888: - // tegra driver assumes any layer with alpha channel has premult - // blending, avoid handling it this is not the case. This is not an - // issue for bottom-most layer since there's nothing to blend with - if (zorder > 0 && layer->blending != DrmHwcBlending::kPreMult) - return false; - - v_limit = 2; - break; - case DRM_FORMAT_YVU420: - case DRM_FORMAT_YUV420: - case DRM_FORMAT_YUV422: - case DRM_FORMAT_UYVY: - case DRM_FORMAT_YUYV: - case DRM_FORMAT_NV12: - case DRM_FORMAT_NV21: - case DRM_FORMAT_RGB565: - case DRM_FORMAT_BGR565: - v_limit = 4; - break; - default: - v_limit = 2; - break; - } - - if (layer->transform & - (DrmHwcTransform::kRotate90 | DrmHwcTransform::kRotate270)) - std::swap(dst_w, dst_h); - - // check for max supported down scaling - if (((src_w / dst_w) > h_limit) || ((src_h / dst_h) > v_limit)) - return false; - - return true; -} - -int PlanStageNvLimits::ProvisionPlanes( - std::vector<DrmCompositionPlane> *composition, - std::map<size_t, DrmHwcLayer *> &layers, DrmCrtc *crtc, - std::vector<DrmPlane *> *planes) { - int ret; - - for (auto i = layers.begin(); i != layers.end();) { - // Skip layer if supported - if (CheckLayer(i->first, i->second)) { - i++; - continue; - } - - if (i->second->protected_usage()) { - // Drop the layer if unsupported and protected, this will just display - // black in the area of this layer but it's better than failing miserably - i = layers.erase(i); - continue; - } - - // If there's no precomp layer already queued, queue one now. - DrmCompositionPlane *precomp = GetPrecomp(composition); - if (precomp) { - precomp->source_layers().emplace_back(i->first); - } else if (!planes->empty()) { - DrmPlane *precomp_plane = planes->back(); - planes->pop_back(); - composition->emplace_back(DrmCompositionPlane::Type::kPrecomp, - precomp_plane, crtc, i->first); - } else { - ALOGE("Not enough planes to reserve for precomp fb"); - } - i = layers.erase(i); - } - - return 0; -} -} -- 2.11.0

3 4

Fwd: linux video documentation ?
by Brad Walker 09 Oct '17

09 Oct '17

I noticed this email address is listed as the relevant area for the Documentation/devicetree/bindings/video/ directory in the Linux kernel. I have a question about this. I noticed the Documentation/devicetree/bindings/video/ is no longer in the Linux kernel repo. It was in release 4.3.6 but got dropped in release v4.4-rc1. Is there a reason for this? I ask because I am using the "simple-framebuffer" and wanted to possibly make some improvements to the documentation. Thanks for any insight you can provide. -brad w.

2 2

[Bug 103168] [CI] igt@prime_mmap_coherency@write-and-fail - fai - Failed assertion: !(!stale)
by bugzilla-daemon＠freedesktop.org 09 Oct '17

09 Oct '17

https://bugs.freedesktop.org/show_bug.cgi?id=103168 Chris Wilson <chris(a)chris-wilson.co.uk> changed: What |Removed |Added ---------------------------------------------------------------------------- Component|DRM/Intel |IGT QA Contact|intel-gfx-bugs(a)lists.freede | |sktop.org | Assignee|intel-gfx-bugs(a)lists.freede |dri-devel(a)lists.freedesktop |sktop.org |.org --- Comment #1 from Chris Wilson <chris(a)chris-wilson.co.uk> --- Just a silliness in the test to assume it was able to hit the cache coherency issue. -- You are receiving this mail because: You are the assignee for the bug.

1 0

[PATCH hwc] platformdrmgeneric: get the gralloc usage directly from the handle
by Rob Herring 09 Oct '17

09 Oct '17

There's no point in using a gralloc perform() op to retrieve the usage as it is already stored in the handle like width, height, format, etc. So just copy it to the hwc_drm_bo_t and get it directly. Signed-off-by: Rob Herring <robh(a)kernel.org> --- I was a bit quick on the trigger with the previous patch. This one is a dependency for it. drmhwcgralloc.h | 1 + hwcutils.cpp | 8 ++------ platformdrmgeneric.cpp | 1 + 3 files changed, 4 insertions(+), 6 deletions(-) diff --git a/drmhwcgralloc.h b/drmhwcgralloc.h index 765c8979d2df..c4a42eaf1c54 100644 --- a/drmhwcgralloc.h +++ b/drmhwcgralloc.h @@ -56,6 +56,7 @@ typedef struct hwc_drm_bo { uint32_t width; uint32_t height; uint32_t format; /* DRM_FORMAT_* from drm_fourcc.h */ + uint32_t usage; uint32_t pitches[4]; uint32_t offsets[4]; uint32_t gem_handles[4]; diff --git a/hwcutils.cpp b/hwcutils.cpp index 00915757469e..53a7d82889b8 100644 --- a/hwcutils.cpp +++ b/hwcutils.cpp @@ -159,12 +159,8 @@ int DrmHwcLayer::ImportBuffer(Importer *importer, if (ret) return ret; - ret = gralloc->perform(gralloc, GRALLOC_MODULE_PERFORM_GET_USAGE, - handle.get(), &gralloc_buffer_usage); - if (ret) { - ALOGE("Failed to get usage for buffer %p (%d)", handle.get(), ret); - return ret; - } + gralloc_buffer_usage = buffer.operator->()->usage; + return 0; } diff --git a/platformdrmgeneric.cpp b/platformdrmgeneric.cpp index aa3d0faee7c0..741d42b6ac33 100644 --- a/platformdrmgeneric.cpp +++ b/platformdrmgeneric.cpp @@ -116,6 +116,7 @@ int DrmGenericImporter::ImportBuffer(buffer_handle_t handle, hwc_drm_bo_t *bo) { bo->width = gr_handle->width; bo->height = gr_handle->height; bo->format = ConvertHalFormatToDrm(gr_handle->format); + bo->usage = gr_handle->usage; bo->pitches[0] = gr_handle->stride; bo->gem_handles[0] = gem_handle; bo->offsets[0] = 0; -- 2.11.0

2 1

[PATCH v2] drm/atomic: Unref duplicated drm_atomic_state in drm_atomic_helper_resume()
by Jeffy Chen 09 Oct '17

09 Oct '17

Kmemleak reported memory leak after suspend and resume: unreferenced object 0xffffffc0e31d8880 (size 128): comm "bash", pid 181, jiffies 4294763583 (age 24.694s) hex dump (first 32 bytes): 01 00 00 00 00 00 00 00 00 20 a2 eb c0 ff ff ff ......... ...... 01 00 00 00 00 00 00 00 80 87 1d e3 c0 ff ff ff ................ backtrace: [<ffffffc00034bb64>] __save_stack_trace+0x48/0x6c [<ffffffc00034c244>] create_object+0x138/0x254 [<ffffffc0009dd218>] kmemleak_alloc+0x58/0x8c [<ffffffc000346de4>] kmem_cache_alloc_trace+0x188/0x254 [<ffffffc0005af4c0>] drm_atomic_state_alloc+0x3c/0x88 [<ffffffc000591f0c>] drm_atomic_helper_duplicate_state+0x28/0x158 [<ffffffc000592098>] drm_atomic_helper_suspend+0x5c/0xf0 Problem here is that we are duplicating the drm_atomic_state in drm_atomic_helper_suspend(), but not unreference it in the resume path. Fixes: 1494276000db ("drm/atomic-helper: Implement subsystem-level suspend/resume") Signed-off-by: Jeffy Chen <jeffy.chen(a)rock-chips.com> --- Changes in v2: Unref duplicated drm_atomic_state in drm_atomic_helper_resume() instead of specific drivers. drivers/gpu/drm/drm_atomic_helper.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 01c34bc5b5b0..4a262380c631 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -3052,6 +3052,7 @@ int drm_atomic_helper_resume(struct drm_device *dev, drm_modeset_backoff(&ctx); } + drm_atomic_state_put(state); drm_modeset_drop_locks(&ctx); drm_modeset_acquire_fini(&ctx); -- 2.11.0

2 1

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel October 2017