From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 1A692EAD0; Fri, 15 Nov 2024 06:50:26 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731653426; cv=none; b=J+B3CKZleGLCYz75sqFxkN9Aq37Ut+TIfaNfDjBxxv+oKnMjnVIeEhlZVO6yLHC1Y77j1gM5qCnu9/kBE1P/deQB9vRvbyTd8ioRvyMJWAou0cw0PVA8WVbE+wSF4wirZo9vL278B0cK8CQuOj0+TUmQ+o54c4lfy+ynErgyiDU= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731653426; c=relaxed/simple; bh=MOuqL6arWIWEShJ/5eNs2Az0DMhchv5KUZ3zEQJ6MXY=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=GQuHsnAqWF8wcYOroQ9OEM3b6ZNwOZObP8cRcL/ZLQ/7Xjx9c23StexoNh+KET2N1X5JA5bZmOjow0mrHM2zfjkmrsj/NM85UVP1sEutRUei2wFRBX/lcrq0ZBjbGLrsGJs7fQtZi3xc7gROGyKxGSGxHvqchlMW3TnnCb9+JxE= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=linuxfoundation.org header.i=@linuxfoundation.org header.b=Njlf+FYW; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=linuxfoundation.org header.i=@linuxfoundation.org header.b="Njlf+FYW" Received: by smtp.kernel.org (Postfix) with ESMTPSA id 6ECACC4CECF; Fri, 15 Nov 2024 06:50:25 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1731653425; bh=MOuqL6arWIWEShJ/5eNs2Az0DMhchv5KUZ3zEQJ6MXY=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=Njlf+FYWOT92JMcIiD7KoHkpwpxnT20I2l5YPfDRWQ1JSmPiwoQ+aHJ1IjZFLPYqJ PMCEVdWjIWbnuZjKX0CL9iperdG85hDJsNZPN2ZUXvFsg5q3ZuJBEJXOobjpvQwX0c WxFb8nFcrpC0vthG+eJapIFodP0f/T07Wc2616IY= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Philip Yang , Felix Kuehling , Alex Deucher , Sasha Levin Subject: [PATCH 6.6 19/48] drm/amdkfd: Accounting pdd vram_usage for svm Date: Fri, 15 Nov 2024 07:38:08 +0100 Message-ID: <20241115063723.658323107@linuxfoundation.org> X-Mailer: git-send-email 2.47.0 In-Reply-To: <20241115063722.962047137@linuxfoundation.org> References: <20241115063722.962047137@linuxfoundation.org> User-Agent: quilt/0.67 X-stable: review X-Patchwork-Hint: ignore Precedence: bulk X-Mailing-List: stable@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit 6.6-stable review patch. If anyone has any objections, please let me know. ------------------ From: Philip Yang [ Upstream commit 68d26c10ef503175df3142db6fcd75dd94860592 ] Process device data pdd->vram_usage is read by rocm-smi via sysfs, this is currently missing the svm_bo usage accounting, so "rocm-smi --showpids" per process VRAM usage report is incorrect. Add pdd->vram_usage accounting when svm_bo allocation and release, change to atomic64_t type because it is updated outside process mutex now. Signed-off-by: Philip Yang Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher (cherry picked from commit 98c0b0efcc11f2a5ddf3ce33af1e48eedf808b04) Signed-off-by: Sasha Levin --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 6 +++--- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 ++-- drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 26 ++++++++++++++++++++++++ 4 files changed, 32 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 19d46be639429..8669677662d0c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -1164,7 +1164,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep, if (flags & KFD_IOC_ALLOC_MEM_FLAGS_AQL_QUEUE_MEM) size >>= 1; - WRITE_ONCE(pdd->vram_usage, pdd->vram_usage + PAGE_ALIGN(size)); + atomic64_add(PAGE_ALIGN(size), &pdd->vram_usage); } mutex_unlock(&p->mutex); @@ -1235,7 +1235,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep, kfd_process_device_remove_obj_handle( pdd, GET_IDR_HANDLE(args->handle)); - WRITE_ONCE(pdd->vram_usage, pdd->vram_usage - size); + atomic64_sub(size, &pdd->vram_usage); err_unlock: err_pdd: @@ -2352,7 +2352,7 @@ static int criu_restore_memory_of_gpu(struct kfd_process_device *pdd, } else if (bo_bucket->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { bo_bucket->restored_offset = offset; /* Update the VRAM usage count */ - WRITE_ONCE(pdd->vram_usage, pdd->vram_usage + bo_bucket->size); + atomic64_add(bo_bucket->size, &pdd->vram_usage); } return 0; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 67204c3dfbb8f..27c9d5c43765a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -765,7 +765,7 @@ struct kfd_process_device { enum kfd_pdd_bound bound; /* VRAM usage */ - uint64_t vram_usage; + atomic64_t vram_usage; struct attribute attr_vram; char vram_filename[MAX_SYSFS_FILENAME_LEN]; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 43f520b379670..6c90231e0aec2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -306,7 +306,7 @@ static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr, } else if (strncmp(attr->name, "vram_", 5) == 0) { struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device, attr_vram); - return snprintf(buffer, PAGE_SIZE, "%llu\n", READ_ONCE(pdd->vram_usage)); + return snprintf(buffer, PAGE_SIZE, "%llu\n", atomic64_read(&pdd->vram_usage)); } else if (strncmp(attr->name, "sdma_", 5) == 0) { struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device, attr_sdma); @@ -1589,7 +1589,7 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev, pdd->bound = PDD_UNBOUND; pdd->already_dequeued = false; pdd->runtime_inuse = false; - pdd->vram_usage = 0; + atomic64_set(&pdd->vram_usage, 0); pdd->sdma_past_activity_counter = 0; pdd->user_gpu_id = dev->id; atomic64_set(&pdd->evict_duration_counter, 0); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index ce76d45549984..6b7c6f45a80a8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -391,6 +391,27 @@ static void svm_range_bo_release(struct kref *kref) spin_lock(&svm_bo->list_lock); } spin_unlock(&svm_bo->list_lock); + + if (mmget_not_zero(svm_bo->eviction_fence->mm)) { + struct kfd_process_device *pdd; + struct kfd_process *p; + struct mm_struct *mm; + + mm = svm_bo->eviction_fence->mm; + /* + * The forked child process takes svm_bo device pages ref, svm_bo could be + * released after parent process is gone. + */ + p = kfd_lookup_process_by_mm(mm); + if (p) { + pdd = kfd_get_process_device_data(svm_bo->node, p); + if (pdd) + atomic64_sub(amdgpu_bo_size(svm_bo->bo), &pdd->vram_usage); + kfd_unref_process(p); + } + mmput(mm); + } + if (!dma_fence_is_signaled(&svm_bo->eviction_fence->base)) /* We're not in the eviction worker. Signal the fence. */ dma_fence_signal(&svm_bo->eviction_fence->base); @@ -518,6 +539,7 @@ int svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange, bool clear) { + struct kfd_process_device *pdd; struct amdgpu_bo_param bp; struct svm_range_bo *svm_bo; struct amdgpu_bo_user *ubo; @@ -609,6 +631,10 @@ svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange, list_add(&prange->svm_bo_list, &svm_bo->range_list); spin_unlock(&svm_bo->list_lock); + pdd = svm_range_get_pdd_by_node(prange, node); + if (pdd) + atomic64_add(amdgpu_bo_size(bo), &pdd->vram_usage); + return 0; reserve_bo_failed: -- 2.43.0