[PATCH V1] accel/amdxdna: Remove mmap and export support for ubuf

[PATCH V1] accel/amdxdna: Remove mmap and export support for ubuf

[Lizhi Hou]

Ubuf pages should not be mmaped or exported. Remove the ubuf mmap callback
and return -EOPNOTSUPP when exporting ubuf objects.

ubuf vmap is also removed for there is not a real use case yet.

Fixes: bd72d4acda10 ("accel/amdxdna: Support user space allocated buffer")
Cc: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
---
 drivers/accel/amdxdna/amdxdna_gem.c  |  9 +++++++-
 drivers/accel/amdxdna/amdxdna_gem.h  |  2 ++
 drivers/accel/amdxdna/amdxdna_ubuf.c | 32 ----------------------------
 3 files changed, 10 insertions(+), 33 deletions(-)

diff --git a/drivers/accel/amdxdna/amdxdna_gem.c b/drivers/accel/amdxdna/amdxdna_gem.c
index 319d2064fafa..6087264ba1b5 100644
--- a/drivers/accel/amdxdna/amdxdna_gem.c
+++ b/drivers/accel/amdxdna/amdxdna_gem.c
@@ -492,6 +492,9 @@ static struct dma_buf *amdxdna_gem_prime_export(struct drm_gem_object *gobj, int
 	struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
 	DEFINE_DMA_BUF_EXPORT_INFO(exp_info);
 
+	if (abo->pri)
+		return ERR_PTR(-EOPNOTSUPP);
+
 	if (abo->dma_buf) {
 		get_dma_buf(abo->dma_buf);
 		return abo->dma_buf;
@@ -716,6 +719,7 @@ amdxdna_gem_create_ubuf_object(struct drm_device *dev, struct amdxdna_drm_create
 {
 	struct amdxdna_dev *xdna = to_xdna_dev(dev);
 	struct amdxdna_drm_va_tbl va_tbl;
+	struct amdxdna_gem_obj *abo;
 	struct drm_gem_object *gobj;
 	struct dma_buf *dma_buf;
 
@@ -742,7 +746,10 @@ amdxdna_gem_create_ubuf_object(struct drm_device *dev, struct amdxdna_drm_create
 
 	dma_buf_put(dma_buf);
 
-	return to_xdna_obj(gobj);
+	abo = to_xdna_obj(gobj);
+	abo->pri = true;
+
+	return abo;
 }
 
 static struct amdxdna_gem_obj *
diff --git a/drivers/accel/amdxdna/amdxdna_gem.h b/drivers/accel/amdxdna/amdxdna_gem.h
index 4fc48a1189d2..162e5499f5e0 100644
--- a/drivers/accel/amdxdna/amdxdna_gem.h
+++ b/drivers/accel/amdxdna/amdxdna_gem.h
@@ -54,6 +54,8 @@ struct amdxdna_gem_obj {
 
 	/* True, if BO is managed by XRT, not application */
 	bool				internal;
+	/* True, if BO is not exportable */
+	bool				pri;
 };
 
 #define to_gobj(obj)    (&(obj)->base.base)
diff --git a/drivers/accel/amdxdna/amdxdna_ubuf.c b/drivers/accel/amdxdna/amdxdna_ubuf.c
index 3769210c55cc..df4ab225fbf9 100644
--- a/drivers/accel/amdxdna/amdxdna_ubuf.c
+++ b/drivers/accel/amdxdna/amdxdna_ubuf.c
@@ -87,42 +87,10 @@ static const struct vm_operations_struct amdxdna_ubuf_vm_ops = {
 	.fault = amdxdna_ubuf_vm_fault,
 };
 
-static int amdxdna_ubuf_mmap(struct dma_buf *dbuf, struct vm_area_struct *vma)
-{
-	struct amdxdna_ubuf_priv *ubuf = dbuf->priv;
-
-	vma->vm_ops = &amdxdna_ubuf_vm_ops;
-	vma->vm_private_data = ubuf;
-	vm_flags_set(vma, VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP);
-
-	return 0;
-}
-
-static int amdxdna_ubuf_vmap(struct dma_buf *dbuf, struct iosys_map *map)
-{
-	struct amdxdna_ubuf_priv *ubuf = dbuf->priv;
-	void *kva;
-
-	kva = vmap(ubuf->pages, ubuf->nr_pages, VM_MAP, PAGE_KERNEL);
-	if (!kva)
-		return -EINVAL;
-
-	iosys_map_set_vaddr(map, kva);
-	return 0;
-}
-
-static void amdxdna_ubuf_vunmap(struct dma_buf *dbuf, struct iosys_map *map)
-{
-	vunmap(map->vaddr);
-}
-
 static const struct dma_buf_ops amdxdna_ubuf_dmabuf_ops = {
 	.map_dma_buf = amdxdna_ubuf_map,
 	.unmap_dma_buf = amdxdna_ubuf_unmap,
 	.release = amdxdna_ubuf_release,
-	.mmap = amdxdna_ubuf_mmap,
-	.vmap = amdxdna_ubuf_vmap,
-	.vunmap = amdxdna_ubuf_vunmap,
 };
 
 static int readonly_va_entry(struct amdxdna_drm_va_entry *va_ent)
-- 
2.34.1

Claude review: accel/amdxdna: Remove mmap and export support for ubuf

[Claude Code Review Bot]

Overall Series Review

Subject: accel/amdxdna: Remove mmap and export support for ubuf
Author: Lizhi Hou <lizhi.hou@amd.com>
Patches: 1
Reviewed: 2026-05-16T08:56:25.802864

---

This is a single patch that removes mmap, vmap, and vunmap support from the `amdxdna_ubuf_dmabuf_ops` and blocks re-export of ubuf-backed GEM objects. The intent is reasonable: ubuf wraps user-pinned pages, and re-exporting or mmapping through the dma-buf is problematic (the user already has direct access to the memory). The patch is small and the deletion side is clean. However, there are two issues worth discussing: dead code left behind, and a poorly named flag.

---
Generated by Claude Code Patch Reviewer

Claude review: accel/amdxdna: Remove mmap and export support for ubuf

[Claude Code Review Bot]

Patch Review

**Dead code: `amdxdna_ubuf_vm_ops` and `amdxdna_ubuf_vm_fault` are now orphaned.**

The patch removes `amdxdna_ubuf_mmap` which was the sole user of `amdxdna_ubuf_vm_ops`, which in turn was the sole user of `amdxdna_ubuf_vm_fault`. After this patch, both `amdxdna_ubuf_vm_fault` (line 72) and `amdxdna_ubuf_vm_ops` (line 86) in `amdxdna_ubuf.c` are dead code. They should be removed in this patch to avoid a compiler warning (`-Wunused-function` for the static `amdxdna_ubuf_vm_fault`). This will actually cause a build failure with `-Werror`, which the kernel uses.

**Flag naming: `pri` is cryptic and misleading.**

The new field:
```c
/* True, if BO is not exportable */
bool				pri;
```
The name `pri` doesn't convey "not exportable." It could be misread as "priority" or "primary" or "private." A name like `no_export` or `unexportable` would be immediately clear. Even `private` (spelled out) would be better. The comment helps, but a self-documenting name costs nothing.

**The export-blocking logic is correct but could be more robust.**

The check in `amdxdna_gem_prime_export`:
```c
if (abo->pri)
    return ERR_PTR(-EOPNOTSUPP);
```
This correctly prevents re-export. The ubuf object is created through `amdxdna_gem_prime_import` and then marked with `abo->pri = true`, which will prevent it from being re-exported through the shmem funcs' `.export = amdxdna_gem_prime_export` path.

**No mmap blocking on the GEM object side.**

The patch removes `amdxdna_ubuf_mmap` from the dma-buf ops, which prevents mmap through the dma-buf fd. However, ubuf objects are imported via `drm_gem_shmem_prime_import_sg_table`, and the resulting GEM shmem object has `.mmap = amdxdna_gem_obj_mmap` set via `amdxdna_gem_shmem_funcs`. The commit message says "ubuf pages should not be mmaped" — should `amdxdna_gem_obj_mmap` also check `abo->pri` and reject the mmap? If the intent is only to block the dma-buf mmap path (since user already has the pages mapped), this should be clarified in the commit message. If GEM mmap should also be blocked, it's missing.

**Summary:**
- **Must fix:** Remove the now-dead `amdxdna_ubuf_vm_fault` and `amdxdna_ubuf_vm_ops` — this will cause a build warning/error.
- **Should fix:** Rename `pri` to something self-documenting (`no_export`, `unexportable`, etc.).
- **Should clarify:** Whether GEM-object-level mmap should also be blocked for ubuf objects, or only dma-buf mmap. If only dma-buf mmap, update the commit message to be precise.

---
Generated by Claude Code Patch Reviewer