From: "Wu, David" <davidwu2@amd.com>
To: amd-gfx@lists.freedesktop.org
Subject: Re: [RFC PATCH 2/3] drm/amdgpu: add AMDGPU_GEM_OP_OPEN_GLOBAL
Date: Wed, 10 Dec 2025 17:06:55 -0500 [thread overview]
Message-ID: <be901b82-54fa-43cc-8a69-16388afc1cc9@amd.com> (raw)
In-Reply-To: <20251202151241.2212-2-christian.koenig@amd.com>
On 12/2/2025 10:12 AM, Christian König wrote:
> Instead of abusing the create IOCTL to open global BO add a new
> AMDGPU_GEM_OP_OPEN_GLOBAL functionality.
>
> The new AMDGPU_GEM_OP_OPEN_GLOBAL functionality expects an enum which tells
> it which global BO to open and copies the information about the BO to
> userspace similar to the AMDGPU_GEM_OP_GET_GEM_CREATE_INFO operation.
>
> The advantage is that we don't start overloading the create IOCTL with
> tons of special cases and opening the global BOs doesn't requires knowing
> the exact size and parameters of it in userspace any more.
>
> Heavily WIP and only compile tested.
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 26 ++++++++++++++++++++-----
> include/uapi/drm/amdgpu_drm.h | 5 ++++-
> 2 files changed, 25 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> index 9b81a6677f90..9e9b94dcb699 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
> @@ -968,22 +968,34 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
> int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
> struct drm_file *filp)
> {
> + struct amdgpu_fpriv *fpriv = filp->driver_priv;
> struct drm_amdgpu_gem_op *args = data;
> struct drm_gem_object *gobj;
> struct amdgpu_vm_bo_base *base;
> struct amdgpu_bo *robj;
> struct drm_exec exec;
> - struct amdgpu_fpriv *fpriv = filp->driver_priv;
> int r;
>
> if (args->padding)
> return -EINVAL;
>
> - gobj = drm_gem_object_lookup(filp, args->handle);
> - if (!gobj)
> - return -ENOENT;
> + if (args->op == AMDGPU_GEM_OP_OPEN_GLOBAL) {
> + switch (args->handle) {
> + case AMDGPU_GEM_GLOBAL_MMIO_REMAP:
> + robj = drm_to_adev(dev)->rmmio_remap.bo;
rmmio_remap.bo could be null - so robj should be checked and return error instead.
David
> + break;
> + default:
> + return -EINVAL;
> + }
> + gobj = &robj->tbo.base;
> + drm_gem_object_get(gobj);
> + } else {
> + gobj = drm_gem_object_lookup(filp, args->handle);
> + if (!gobj)
> + return -ENOENT;
>
> - robj = gem_to_amdgpu_bo(gobj);
> + robj = gem_to_amdgpu_bo(gobj);
> + }
>
> drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
> DRM_EXEC_IGNORE_DUPLICATES, 0);
> @@ -1002,6 +1014,7 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
> }
>
> switch (args->op) {
> + case AMDGPU_GEM_OP_OPEN_GLOBAL:
> case AMDGPU_GEM_OP_GET_GEM_CREATE_INFO: {
> struct drm_amdgpu_gem_create_in info;
> void __user *out = u64_to_user_ptr(args->value);
> @@ -1096,6 +1109,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
> r = -EINVAL;
> }
>
> + if (!r && args->op == AMDGPU_GEM_OP_OPEN_GLOBAL)
> + r = drm_gem_handle_create(filp, gobj, &args->handle);
> +
> drm_gem_object_put(gobj);
> return r;
> out_exec:
> diff --git a/include/uapi/drm/amdgpu_drm.h b/include/uapi/drm/amdgpu_drm.h
> index c1336ed4ff75..6927c864a6d1 100644
> --- a/include/uapi/drm/amdgpu_drm.h
> +++ b/include/uapi/drm/amdgpu_drm.h
> @@ -807,6 +807,9 @@ union drm_amdgpu_wait_fences {
> #define AMDGPU_GEM_OP_GET_GEM_CREATE_INFO 0
> #define AMDGPU_GEM_OP_SET_PLACEMENT 1
> #define AMDGPU_GEM_OP_GET_MAPPING_INFO 2
> +#define AMDGPU_GEM_OP_OPEN_GLOBAL 3
> +
> +#define AMDGPU_GEM_GLOBAL_MMIO_REMAP 0
>
> struct drm_amdgpu_gem_vm_entry {
> /* Start of mapping (in bytes) */
> @@ -824,7 +827,7 @@ struct drm_amdgpu_gem_vm_entry {
>
> /* Sets or returns a value associated with a buffer. */
> struct drm_amdgpu_gem_op {
> - /** GEM object handle */
> + /** GEM object handle or AMDGPU_GEM_GLOBAL_* */
> __u32 handle;
> /** AMDGPU_GEM_OP_* */
> __u32 op;
next prev parent reply other threads:[~2025-12-10 22:07 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-12-02 15:12 [RFC PATCH 1/3] drm/amdgpu: rework MMIO_REMAP BO creation Christian König
2025-12-02 15:12 ` [RFC PATCH 2/3] drm/amdgpu: add AMDGPU_GEM_OP_OPEN_GLOBAL Christian König
2025-12-10 4:11 ` SHANMUGAM, SRINIVASAN
2025-12-10 22:06 ` Wu, David [this message]
2025-12-02 15:12 ` [RFC PATCH 3/3] drm/amdgpu: remove AMDGPU_GEM_DOMAIN_DOORBELL Christian König
2025-12-10 4:22 ` SHANMUGAM, SRINIVASAN
2025-12-10 3:22 ` [RFC PATCH 1/3] drm/amdgpu: rework MMIO_REMAP BO creation SHANMUGAM, SRINIVASAN
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=be901b82-54fa-43cc-8a69-16388afc1cc9@amd.com \
--to=davidwu2@amd.com \
--cc=amd-gfx@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox