dri-devel April 2021

dri-devel@lists.freedesktop.org

317 participants
715 discussions

[PATCH] drm/ttm: add __user annotation in radeon_ttm_vram_read
by Rasmus Villemoes 01 Apr '21

01 Apr '21

Keep sparse happy by preserving the __user annotation when casting. Reported-by: kernel test robot <lkp(a)intel.com> Signed-off-by: Rasmus Villemoes <linux(a)rasmusvillemoes.dk> --- kernel test robot has already started spamming me due to 9c5743dff. If I don't fix those warnings I'll keep getting those emails for months, so let me do the easy ones. drivers/gpu/drm/radeon/radeon_ttm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index 36150b7f31a90aa1eece..ecfe88b0a35d8f317712 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -1005,7 +1005,7 @@ static ssize_t radeon_ttm_vram_read(struct file *f, char __user *buf, value = RREG32(RADEON_MM_DATA); spin_unlock_irqrestore(&rdev->mmio_idx_lock, flags); - r = put_user(value, (uint32_t *)buf); + r = put_user(value, (uint32_t __user *)buf); if (r) return r; -- 2.23.0

3 2

[PATCH] amd/amdgpu: code refactoring to clean code style a bit
by Bernard Zhao 01 Apr '21

01 Apr '21

Fix checkpatch.pl warning: Too many leading tabs - consider code refactoring WARNING: Too many leading tabs - consider code refactoring + for (j = 0; j < profile->ucLeakageBinNum; j++) { WARNING: Too many leading tabs - consider code refactoring + if (vbios_voltage_id <= leakage_bin[j]) { WARNING: Too many leading tabs - consider code refactoring + for (j = 0; j < profile->ucLeakageBinNum; j++) { WARNING: Too many leading tabs - consider code refactoring + if (vbios_voltage_id <= leakage_bin[j]) { Signed-off-by: Bernard Zhao <bernard(a)vivo.com> --- drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c | 84 ++++++++------------ 1 file changed, 35 insertions(+), 49 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 86add0f4ea4d..9968ff8ddc9c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -1283,65 +1283,51 @@ int amdgpu_atombios_get_leakage_vddc_based_on_leakage_params(struct amdgpu_devic profile = (ATOM_ASIC_PROFILING_INFO_V2_1 *) (adev->mode_info.atom_context->bios + data_offset); - switch (frev) { - case 1: + if ((frev != 2) || (crev != 1)) { + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); return -EINVAL; - case 2: - switch (crev) { - case 1: - if (size < sizeof(ATOM_ASIC_PROFILING_INFO_V2_1)) - return -EINVAL; - leakage_bin = (u16 *) - (adev->mode_info.atom_context->bios + data_offset + - le16_to_cpu(profile->usLeakageBinArrayOffset)); - vddc_id_buf = (u16 *) - (adev->mode_info.atom_context->bios + data_offset + - le16_to_cpu(profile->usElbVDDC_IdArrayOffset)); - vddc_buf = (u16 *) - (adev->mode_info.atom_context->bios + data_offset + - le16_to_cpu(profile->usElbVDDC_LevelArrayOffset)); - vddci_id_buf = (u16 *) - (adev->mode_info.atom_context->bios + data_offset + - le16_to_cpu(profile->usElbVDDCI_IdArrayOffset)); - vddci_buf = (u16 *) - (adev->mode_info.atom_context->bios + data_offset + - le16_to_cpu(profile->usElbVDDCI_LevelArrayOffset)); - - if (profile->ucElbVDDC_Num > 0) { - for (i = 0; i < profile->ucElbVDDC_Num; i++) { - if (vddc_id_buf[i] == virtual_voltage_id) { - for (j = 0; j < profile->ucLeakageBinNum; j++) { - if (vbios_voltage_id <= leakage_bin[j]) { - *vddc = vddc_buf[j * profile->ucElbVDDC_Num + i]; - break; - } - } + } + + if (size < sizeof(ATOM_ASIC_PROFILING_INFO_V2_1)) + return -EINVAL; + + leakage_bin = (u16 *)(adev->mode_info.atom_context->bios + data_offset + + le16_to_cpu(profile->usLeakageBinArrayOffset)); + vddc_id_buf = (u16 *)(adev->mode_info.atom_context->bios + data_offset + + le16_to_cpu(profile->usElbVDDC_IdArrayOffset)); + vddc_buf = (u16 *)(adev->mode_info.atom_context->bios + data_offset + + le16_to_cpu(profile->usElbVDDC_LevelArrayOffset)); + vddci_id_buf = (u16 *)(adev->mode_info.atom_context->bios + data_offset + + le16_to_cpu(profile->usElbVDDCI_IdArrayOffset)); + vddci_buf = (u16 *)(adev->mode_info.atom_context->bios + data_offset + + le16_to_cpu(profile->usElbVDDCI_LevelArrayOffset)); + + if (profile->ucElbVDDC_Num > 0) { + for (i = 0; i < profile->ucElbVDDC_Num; i++) { + if (vddc_id_buf[i] == virtual_voltage_id) { + for (j = 0; j < profile->ucLeakageBinNum; j++) { + if (vbios_voltage_id <= leakage_bin[j]) { + *vddc = vddc_buf[j * profile->ucElbVDDC_Num + i]; break; } } + break; } - if (profile->ucElbVDDCI_Num > 0) { - for (i = 0; i < profile->ucElbVDDCI_Num; i++) { - if (vddci_id_buf[i] == virtual_voltage_id) { - for (j = 0; j < profile->ucLeakageBinNum; j++) { - if (vbios_voltage_id <= leakage_bin[j]) { - *vddci = vddci_buf[j * profile->ucElbVDDCI_Num + i]; - break; - } - } + } + } + + if (profile->ucElbVDDCI_Num > 0) { + for (i = 0; i < profile->ucElbVDDCI_Num; i++) { + if (vddci_id_buf[i] == virtual_voltage_id) { + for (j = 0; j < profile->ucLeakageBinNum; j++) { + if (vbios_voltage_id <= leakage_bin[j]) { + *vddci = vddci_buf[j * profile->ucElbVDDCI_Num + i]; break; } } + break; } - break; - default: - DRM_ERROR("Unknown table version %d, %d\n", frev, crev); - return -EINVAL; } - break; - default: - DRM_ERROR("Unknown table version %d, %d\n", frev, crev); - return -EINVAL; } return 0; -- 2.31.0

2 1

[PATCH] drm/amd: cleanup coding style a bit
by Bernard Zhao 01 Apr '21

01 Apr '21

Fix patch check warning: WARNING: suspect code indent for conditional statements (8, 17) + if (obj && obj->use < 0) { + DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", obj->head.name); WARNING: braces {} are not necessary for single statement blocks + if (obj && obj->use < 0) { + DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", obj->head.name); + } Signed-off-by: Bernard Zhao <bernard(a)vivo.com> --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 1fb2a91ad30a..43d17b72c265 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -449,11 +449,10 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev, static inline void put_obj(struct ras_manager *obj) { - if (obj && --obj->use == 0) + if (obj && (--obj->use == 0)) list_del(&obj->node); - if (obj && obj->use < 0) { - DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", obj->head.name); - } + if (obj && (obj->use < 0)) + DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", obj->head.name); } /* make one obj and return it. */ -- 2.31.0

2 1

[PATCH] drm/amd: use kmalloc_array over kmalloc with multiply
by Bernard Zhao 01 Apr '21

01 Apr '21

Fix patch check warning: WARNING: Prefer kmalloc_array over kmalloc with multiply + buf = kmalloc(MAX_KFIFO_SIZE * sizeof(*buf), GFP_KERNEL); Signed-off-by: Bernard Zhao <bernard(a)vivo.com> --- drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c index 17d1736367ea..246522423559 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c @@ -81,7 +81,7 @@ static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user, struct kfd_smi_client *client = filep->private_data; unsigned char *buf; - buf = kmalloc(MAX_KFIFO_SIZE * sizeof(*buf), GFP_KERNEL); + buf = kmalloc_array(MAX_KFIFO_SIZE, sizeof(*buf), GFP_KERNEL); if (!buf) return -ENOMEM; -- 2.31.0

2 1

[PATCH 1/2] gpu: host1x: Split up client initalization and registration
by Thierry Reding 01 Apr '21

01 Apr '21

From: Thierry Reding <treding(a)nvidia.com> In some cases we may need to initialize the host1x client first before registering it. This commit adds a new helper that will do nothing but the initialization of the data structure. At the same time, the initialization is removed from the registration function. Note, however, that for simplicity we explicitly initialize the client when the host1x_client_register() function is called, as opposed to the low-level __host1x_client_register() function. This allows existing callers to remain unchanged. Signed-off-by: Thierry Reding <treding(a)nvidia.com> --- drivers/gpu/host1x/bus.c | 30 ++++++++++++++++++++++++------ include/linux/host1x.h | 30 ++++++++++++++++++++++++------ 2 files changed, 48 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/host1x/bus.c b/drivers/gpu/host1x/bus.c index 46f69c532b6b..218e3718fd68 100644 --- a/drivers/gpu/host1x/bus.c +++ b/drivers/gpu/host1x/bus.c @@ -735,6 +735,29 @@ void host1x_driver_unregister(struct host1x_driver *driver) } EXPORT_SYMBOL(host1x_driver_unregister); +/** + * __host1x_client_init() - initialize a host1x client + * @client: host1x client + * @key: lock class key for the client-specific mutex + */ +void __host1x_client_init(struct host1x_client *client, struct lock_class_key *key) +{ + INIT_LIST_HEAD(&client->list); + __mutex_init(&client->lock, "host1x client lock", key); + client->usecount = 0; +} +EXPORT_SYMBOL(__host1x_client_init); + +/** + * host1x_client_exit() - uninitialize a host1x client + * @client: host1x client + */ +void host1x_client_exit(struct host1x_client *client) +{ + mutex_destroy(&client->lock); +} +EXPORT_SYMBOL(host1x_client_exit); + /** * __host1x_client_register() - register a host1x client * @client: host1x client @@ -747,16 +770,11 @@ EXPORT_SYMBOL(host1x_driver_unregister); * device and call host1x_device_init(), which will in turn call each client's * &host1x_client_ops.init implementation. */ -int __host1x_client_register(struct host1x_client *client, - struct lock_class_key *key) +int __host1x_client_register(struct host1x_client *client) { struct host1x *host1x; int err; - INIT_LIST_HEAD(&client->list); - __mutex_init(&client->lock, "host1x client lock", key); - client->usecount = 0; - mutex_lock(&devices_lock); list_for_each_entry(host1x, &devices, list) { diff --git a/include/linux/host1x.h b/include/linux/host1x.h index ef13bc69b493..7310ce9c70e0 100644 --- a/include/linux/host1x.h +++ b/include/linux/host1x.h @@ -349,12 +349,30 @@ static inline struct host1x_device *to_host1x_device(struct device *dev) int host1x_device_init(struct host1x_device *device); int host1x_device_exit(struct host1x_device *device); -int __host1x_client_register(struct host1x_client *client, - struct lock_class_key *key); -#define host1x_client_register(class) \ - ({ \ - static struct lock_class_key __key; \ - __host1x_client_register(class, &__key); \ +void __host1x_client_init(struct host1x_client *client, struct lock_class_key *key); +void host1x_client_exit(struct host1x_client *client); + +#define host1x_client_init(client) \ + ({ \ + static struct lock_class_key __key; \ + __host1x_client_init(client, &__key); \ + }) + +int __host1x_client_register(struct host1x_client *client); + +/* + * Note that this wrapper calls __host1x_client_init() for compatibility + * with existing callers. Callers that want to separately initialize and + * register a host1x client must first initialize using either of the + * __host1x_client_init() or host1x_client_init() functions and then use + * the low-level __host1x_client_register() function to avoid the client + * getting reinitialized. + */ +#define host1x_client_register(client) \ + ({ \ + static struct lock_class_key __key; \ + __host1x_client_init(client, &__key); \ + __host1x_client_register(client); \ }) int host1x_client_unregister(struct host1x_client *client); -- 2.30.2

2 2

[git pull] drm fixes for 5.12-rc6
by Dave Airlie 01 Apr '21

01 Apr '21

Hi Linus, Things have settled down in time for Easter, a random smattering of small fixes across a few drivers. I'm guessing though there might be some i915 and misc fixes out there I haven't gotten yet, but since today is a public holiday here, I'm sending this early so I can have the day off, I'll see if more requests come in and decide what to do with them later. Dave. drm-fixes-2021-04-02: drm fixes for 5.12-rc6 amdgpu: - Polaris idle power fix - VM fix - Vangogh S3 fix - Fixes for non-4K page sizes amdkfd: - dqm fence memory corruption fix tegra: - lockdep warning fix - runtine PM reference fix - display controller fix - PLL Fix imx: - memory leak in error path fix - LDB driver channel registration fix - oob array warning in LDB driver exynos - unused header file removal The following changes since commit a5e13c6df0e41702d2b2c77c8ad41677ebb065b3: Linux 5.12-rc5 (2021-03-28 15:48:16 -0700) are available in the Git repository at: git://anongit.freedesktop.org/drm/drm tags/drm-fixes-2021-04-02 for you to fetch changes up to 6fdb8e5aba6a33fe5f1a0bd1bcf0cf2884437ead: Merge tag 'imx-drm-fixes-2021-04-01' of git://git.pengutronix.de/git/pza/linux into drm-fixes (2021-04-02 04:53:16 +1000) ---------------------------------------------------------------- drm fixes for 5.12-rc6 amdgpu: - Polaris idle power fix - VM fix - Vangogh S3 fix - Fixes for non-4K page sizes amdkfd: - dqm fence memory corruption fix tegra: - lockdep warning fix - runtine PM reference fix - display controller fix - PLL Fix imx: - memory leak in error path fix - LDB driver channel registration fix - oob array warning in LDB driver exynos - unused header file removal ---------------------------------------------------------------- Alex Deucher (1): drm/amdgpu/vangogh: don't check for dpm in is_dpm_running when in suspend Arnd Bergmann (1): drm/imx: imx-ldb: fix out of bounds array access warning Dave Airlie (4): Merge tag 'exynos-drm-fixes-for-v5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos into drm-fixes Merge tag 'amd-drm-fixes-5.12-2021-03-31' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes Merge tag 'drm/tegra/for-5.12-rc6' of ssh://git.freedesktop.org/git/tegra/linux into drm-fixes Merge tag 'imx-drm-fixes-2021-04-01' of git://git.pengutronix.de/git/pza/linux into drm-fixes Dmitry Osipenko (1): drm/tegra: dc: Don't set PLL clock to 0Hz Evan Quan (1): drm/amd/pm: no need to force MCLK to highest when no display connected Huacai Chen (1): drm/amdgpu: Set a suitable dev_info.gart_page_size Liu Ying (1): drm/imx: imx-ldb: Register LDB channel1 when it is the only channel to be used Mikko Perttunen (1): gpu: host1x: Use different lock classes for each client Nirmoy Das (1): drm/amdgpu: fix offset calculation in amdgpu_vm_bo_clear_mappings() Pan Bian (1): drm/imx: fix memory leak when fails to init Qu Huang (1): drm/amdkfd: dqm fence memory corruption Thierry Reding (2): drm/tegra: dc: Restore coupling of display controllers drm/tegra: sor: Grab runtime PM reference across reset Tian Tao (1): drm/exynos/decon5433: Remove the unused include statements Xℹ Ruoyao (1): drm/amdgpu: check alignment on CPU page for bo map drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 4 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 10 ++++---- drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c | 2 +- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 6 ++--- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 8 +++--- .../gpu/drm/amd/pm/powerplay/hwmgr/smu7_hwmgr.c | 3 ++- drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c | 5 ++++ drivers/gpu/drm/exynos/exynos5433_drm_decon.c | 1 - drivers/gpu/drm/imx/imx-drm-core.c | 2 +- drivers/gpu/drm/imx/imx-ldb.c | 12 ++++++++- drivers/gpu/drm/tegra/dc.c | 30 ++++++++++------------ drivers/gpu/drm/tegra/sor.c | 7 +++++ drivers/gpu/host1x/bus.c | 10 +++++--- include/linux/host1x.h | 9 ++++++- 18 files changed, 72 insertions(+), 45 deletions(-)

2 1

[PATCH v7 0/8] Add support for SVM atomics in Nouveau
by Alistair Popple 01 Apr '21

01 Apr '21

This is the seventh version of a series to add support to Nouveau for atomic memory operations on OpenCL shared virtual memory (SVM) regions. This version primarily improves readability of the Nouveau fault priority calculation code along with other minor functional and cosmetic improvements listed in the changelogs. Exclusive device access is implemented by adding a new swap entry type (SWAP_DEVICE_EXCLUSIVE) which is similar to a migration entry. The main difference is that on fault the original entry is immediately restored by the fault handler instead of waiting. Restoring the entry triggers calls to MMU notifers which allows a device driver to revoke the atomic access permission from the GPU prior to the CPU finalising the entry. Patches 1 & 2 refactor existing migration and device private entry functions. Patches 3 & 4 rework try_to_unmap_one() by splitting out unrelated functionality into separate functions - try_to_migrate_one() and try_to_munlock_one(). These should not change any functionality, but any help testing would be much appreciated as I have not been able to test every usage of try_to_unmap_one(). Patch 5 contains the bulk of the implementation for device exclusive memory. Patch 6 contains some additions to the HMM selftests to ensure everything works as expected. Patch 7 is a cleanup for the Nouveau SVM implementation. Patch 8 contains the implementation of atomic access for the Nouveau driver. This has been tested using the latest upstream Mesa userspace with a simple OpenCL test program which checks the results of atomic GPU operations on a SVM buffer whilst also writing to the same buffer from the CPU. Alistair Popple (8): mm: Remove special swap entry functions mm/swapops: Rework swap entry manipulation code mm/rmap: Split try_to_munlock from try_to_unmap mm/rmap: Split migration into its own function mm: Device exclusive memory access mm: Selftests for exclusive device memory nouveau/svm: Refactor nouveau_range_fault nouveau/svm: Implement atomic SVM access Documentation/vm/hmm.rst | 19 +- arch/s390/mm/pgtable.c | 2 +- drivers/gpu/drm/nouveau/include/nvif/if000c.h | 1 + drivers/gpu/drm/nouveau/nouveau_svm.c | 156 ++++- drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h | 1 + .../drm/nouveau/nvkm/subdev/mmu/vmmgp100.c | 6 + fs/proc/task_mmu.c | 23 +- include/linux/mmu_notifier.h | 26 +- include/linux/rmap.h | 9 +- include/linux/swap.h | 8 +- include/linux/swapops.h | 123 ++-- lib/test_hmm.c | 126 +++- lib/test_hmm_uapi.h | 2 + mm/debug_vm_pgtable.c | 12 +- mm/hmm.c | 12 +- mm/huge_memory.c | 45 +- mm/hugetlb.c | 10 +- mm/memcontrol.c | 2 +- mm/memory.c | 128 +++- mm/migrate.c | 51 +- mm/mprotect.c | 18 +- mm/page_vma_mapped.c | 15 +- mm/rmap.c | 604 +++++++++++++++--- tools/testing/selftests/vm/hmm-tests.c | 158 +++++ 24 files changed, 1282 insertions(+), 275 deletions(-) -- 2.20.1

4 30

[PATCH] drm/msm: a6xx: fix version check for the A650 SQE microcode
by Dmitry Baryshkov 01 Apr '21

01 Apr '21

I suppose the microcode version check for a650 is incorrect. It checks for the version 1.95, while the firmware released have major version of 0: 0.91 (vulnerable), 0.99 (fixing the issue). Lower version requirements to accept firmware 0.99. Fixes: 8490f02a3ca4 ("drm/msm: a6xx: Make sure the SQE microcode is safe") Cc: Akhil P Oommen <akhilpo(a)codeaurora.org> Cc: Jordan Crouse <jcrouse(a)codeaurora.org> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov(a)linaro.org> --- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c index cb2df8736ca8..896b47dc9c85 100644 --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c @@ -567,17 +567,17 @@ static bool a6xx_ucode_check_version(struct a6xx_gpu *a6xx_gpu, } else { /* * a650 tier targets don't need whereami but still need to be - * equal to or newer than 1.95 for other security fixes + * equal to or newer than 0.95 for other security fixes */ if (adreno_is_a650(adreno_gpu)) { - if ((buf[0] & 0xfff) >= 0x195) { + if ((buf[0] & 0xfff) >= 0x095) { ret = true; goto out; } DRM_DEV_ERROR(&gpu->pdev->dev, "a650 SQE ucode is too old. Have version %x need at least %x\n", - buf[0] & 0xfff, 0x195); + buf[0] & 0xfff, 0x095); } /* -- 2.30.2

3 2

[PATCH] drm/msm: Drop mm_lock in scan loop
by Rob Clark 01 Apr '21

01 Apr '21

From: Rob Clark <robdclark(a)chromium.org> lock_stat + mmm_donut[1] say that this reduces contention on mm_lock significantly (~350x lower waittime-max, and ~100x lower waittime-avg) [1] https://chromium.googlesource.com/chromiumos/platform/microbenchmarks/+/ref… Signed-off-by: Rob Clark <robdclark(a)chromium.org> --- drivers/gpu/drm/msm/msm_gem.c | 2 +- drivers/gpu/drm/msm/msm_gem_shrinker.c | 48 ++++++++++++++++++++++---- 2 files changed, 43 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c index 2ecf7f1cef25..75cea5b801da 100644 --- a/drivers/gpu/drm/msm/msm_gem.c +++ b/drivers/gpu/drm/msm/msm_gem.c @@ -719,7 +719,7 @@ void msm_gem_purge(struct drm_gem_object *obj) put_iova_vmas(obj); msm_obj->madv = __MSM_MADV_PURGED; - mark_unpurgable(msm_obj); + update_inactive(msm_obj); drm_vma_node_unmap(&obj->vma_node, dev->anon_inode->i_mapping); drm_gem_free_mmap_offset(obj); diff --git a/drivers/gpu/drm/msm/msm_gem_shrinker.c b/drivers/gpu/drm/msm/msm_gem_shrinker.c index f3e948af01c5..6bbb15d64861 100644 --- a/drivers/gpu/drm/msm/msm_gem_shrinker.c +++ b/drivers/gpu/drm/msm/msm_gem_shrinker.c @@ -22,26 +22,62 @@ msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) { struct msm_drm_private *priv = container_of(shrinker, struct msm_drm_private, shrinker); - struct msm_gem_object *msm_obj; + struct list_head still_in_list; unsigned long freed = 0; + INIT_LIST_HEAD(&still_in_list); + mutex_lock(&priv->mm_lock); - list_for_each_entry(msm_obj, &priv->inactive_dontneed, mm_list) { - if (freed >= sc->nr_to_scan) + while (freed < sc->nr_to_scan) { + struct msm_gem_object *msm_obj = list_first_entry_or_null( + &priv->inactive_dontneed, typeof(*msm_obj), mm_list); + + if (!msm_obj) break; - /* Use trylock, because we cannot block on a obj that - * might be trying to acquire mm_lock + + /* + * If it is in the process of being freed, msm_gem_free_object + * can be blocked on mm_lock waiting to remove it. So just + * skip it. */ - if (!msm_gem_trylock(&msm_obj->base)) + if (!kref_get_unless_zero(&msm_obj->base.refcount)) continue; + + /* + * Now that we own a reference, we can move it to our own + * private temporary list and drop mm_lock for the rest of + * the loop body, to reduce contention with the retire_submit + * path (which could make more objects available to purge) + */ + list_move_tail(&msm_obj->mm_list, &still_in_list); + + mutex_unlock(&priv->mm_lock); + + /* + * Note that this still needs to be trylock, since we can + * hit shrinker in response to trying to get backing pages + * for this obj (ie. while it's lock is already held) + */ + if (!msm_gem_trylock(&msm_obj->base)) + goto tail; + if (is_purgeable(msm_obj)) { + /* + * This will move the obj out of still_in_list to + * the purged list + */ msm_gem_purge(&msm_obj->base); freed += msm_obj->base.size >> PAGE_SHIFT; } msm_gem_unlock(&msm_obj->base); + +tail: + drm_gem_object_put(&msm_obj->base); + mutex_lock(&priv->mm_lock); } + list_splice_tail(&still_in_list, &priv->inactive_dontneed); mutex_unlock(&priv->mm_lock); if (freed > 0) { -- 2.30.2

1 0

[PATCH 0/4] drm/msm: Shrinker (and related) fixes
by Rob Clark 01 Apr '21

01 Apr '21

From: Rob Clark <robdclark(a)chromium.org> I've been spending some time looking into how things behave under high memory pressure. The first patch is a random cleanup I noticed along the way. The second improves the situation significantly when we are getting shrinker called from many threads in parallel. And the last two are $debugfs/gem fixes I needed so I could monitor the state of GEM objects (ie. how many are active/purgable/purged) while triggering high memory pressure. We could probably go a bit further with dropping the mm_lock in the shrinker->scan() loop, but this is already a pretty big improvement. The next step is probably actually to add support to unpin/evict inactive objects. (We are part way there since we have already de- coupled the iova lifetime from the pages lifetime, but there are a few sharp corners to work through.) Rob Clark (4): drm/msm: Remove unused freed llist node drm/msm: Avoid mutex in shrinker_count() drm/msm: Fix debugfs deadlock drm/msm: Improved debugfs gem stats drivers/gpu/drm/msm/msm_debugfs.c | 14 ++---- drivers/gpu/drm/msm/msm_drv.c | 4 ++ drivers/gpu/drm/msm/msm_drv.h | 10 ++++- drivers/gpu/drm/msm/msm_fb.c | 3 +- drivers/gpu/drm/msm/msm_gem.c | 61 +++++++++++++++++++++----- drivers/gpu/drm/msm/msm_gem.h | 58 +++++++++++++++++++++--- drivers/gpu/drm/msm/msm_gem_shrinker.c | 17 +------ 7 files changed, 122 insertions(+), 45 deletions(-) -- 2.30.2

2 22

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

dri-devel April 2021