linux-kernel - [PATCH v2 26/34] drm/msm: Pre-allocate vm

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20250319145425.51935-27-robdclark@gmail.com>
Date: Wed, 19 Mar 2025 07:52:38 -0700
From: Rob Clark <robdclark@...il.com>
To: dri-devel@...ts.freedesktop.org
Cc: freedreno@...ts.freedesktop.org,
	linux-arm-msm@...r.kernel.org,
	Rob Clark <robdclark@...omium.org>,
	Rob Clark <robdclark@...il.com>,
	Abhinav Kumar <quic_abhinavk@...cinc.com>,
	Dmitry Baryshkov <lumag@...nel.org>,
	Sean Paul <sean@...rly.run>,
	Marijn Suijten <marijn.suijten@...ainline.org>,
	David Airlie <airlied@...il.com>,
	Simona Vetter <simona@...ll.ch>,
	linux-kernel@...r.kernel.org (open list)
Subject: [PATCH v2 26/34] drm/msm: Pre-allocate vm_bo objects

From: Rob Clark <robdclark@...omium.org>

Use drm_gpuvm_bo_obtain() in the synchronous part of the VM_BIND submit,
to hold a reference to the vm_bo for the duration of the submit.  This
ensures that the vm_bo already exists before the async part of the job,
which is in the fence signalling path (and therefore cannot allocate
memory).

Signed-off-by: Rob Clark <robdclark@...omium.org>
---
 drivers/gpu/drm/msm/msm_gem.h     |  1 +
 drivers/gpu/drm/msm/msm_gem_vma.c | 19 +++++++++++++++++--
 2 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h
index cb76959fa8a8..d2ffaa11ec1a 100644
--- a/drivers/gpu/drm/msm/msm_gem.h
+++ b/drivers/gpu/drm/msm/msm_gem.h
@@ -369,6 +369,7 @@ struct msm_gem_submit {
 		uint32_t flags;
 		uint32_t handle;
 		struct drm_gem_object *obj;
+		struct drm_gpuvm_bo *vm_bo;
 		uint64_t iova;
 		uint64_t bo_offset;
 		uint64_t range;
diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
index 5c7d44b004fb..b1808d95002f 100644
--- a/drivers/gpu/drm/msm/msm_gem_vma.c
+++ b/drivers/gpu/drm/msm/msm_gem_vma.c
@@ -278,8 +278,18 @@ msm_vma_job_prepare(struct msm_gem_submit *submit)
 	for (int i = 0; i < submit->nr_bos; i++) {
 		unsigned op = submit->bos[i].flags & MSM_SUBMIT_BO_OP_MASK;
 
-		if (submit->bos[i].obj)
-			msm_gem_assert_locked(submit->bos[i].obj);
+		if (submit->bos[i].obj) {
+			struct drm_gem_object *obj = submit->bos[i].obj;
+
+			msm_gem_assert_locked(obj);
+
+			/*
+			 * Ensure the vm_bo is already allocated by
+			 * holding a ref until the submit is retired
+			 */
+			submit->bos[i].vm_bo =
+				drm_gpuvm_bo_obtain(submit->vm, obj);
+		}
 
 		/*
 		 * OP_MAP/OP_MAP_NULL has one new VMA for the new mapping,
@@ -309,6 +319,11 @@ msm_vma_job_cleanup(struct msm_gem_submit *submit)
 {
 	struct drm_gpuva *vma;
 
+	for (int i = 0; i < submit->nr_bos; i++) {
+		/* If we're holding an extra ref to the vm_bo, drop it now: */
+		drm_gpuvm_bo_put(submit->bos[i].vm_bo);
+	}
+
 	while (!list_empty(&submit->preallocated_vmas)) {
 		vma = list_first_entry(&submit->preallocated_vmas,
 				       struct drm_gpuva,
-- 
2.48.1