Skip to contentSkip to conversation

Add CRIU support for amdgpu dmabuf

PATCH Superseded David Francis 2 participants 7 messages Jun 13, 2025 – Jun 16, 2025

Subsystems (4)

v6 of 16 Superseded Latest is v16 →

lore
Share & tools
lore
#0 lore raw

PATCH v6 1/1 Add CRIU support for amdgpu dmabuf

This patch series adds support for CRIU checkpointing of processes that share memory with the amdgpu dmabuf interface.

This v6 cleans up the locking and moves some of the mapping info code into amdgpu_vm.c.

The mapping flags code is a placeholder awaiting the mapping flags rework.

#1 lore raw

PATCH 1/3 drm: Add DRM prime interfaces to reassign GEM handle

CRIU restore of drm buffer objects requires the ability to create or import a buffer object with a specific gem handle.

Add new drm ioctl DRM_IOCTL_GEM_CHANGE_HANDLE, which takes the gem handle of an object and moves that object to a specified new gem handle.

This ioctl needs to call drm_prime_remove_buf_handle, but that function acquires the prime lock, which the ioctl needs to hold for other purposes.

Make drm_prime_remove_buf_handle not acquire the prime lock, and change its other caller to reflect this.

---
 drivers/gpu/drm/drm_gem.c      | 56 ++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_internal.h |  4 +++
 drivers/gpu/drm/drm_ioctl.c    |  1 +
 drivers/gpu/drm/drm_prime.c    |  6 +---
 include/uapi/drm/drm.h         | 17 +++++++++++
 5 files changed, 79 insertions(+), 5 deletions(-)
drivers/gpu/drm/drm_gem.c +56 -0
index c6240bab3fa5..31fe2f1d1137 100644
@@ -282,7 +282,12 @@ drm_gem_object_release_handle(int id, void *ptr, void *data)
 	if (obj->funcs->close)
 		obj->funcs->close(obj, file_priv);
 
+	mutex_lock(&file_priv->prime.lock);
+
 	drm_prime_remove_buf_handle(&file_priv->prime, id);
+
+	mutex_unlock(&file_priv->prime.lock);
+
 	drm_vma_node_revoke(&obj->vma_node, file_priv);
 
 	drm_gem_object_handle_put_unlocked(obj);
@@ -888,6 +893,57 @@ drm_gem_flink_ioctl(struct drm_device *dev, void *data,
 	return ret;
 }
 
+/**
+ * drm_gem_open_ioctl - implementation of the GEM_CHANGE_HANDLE ioctl
+ * @dev: drm_device
+ * @data: ioctl data
+ * @file_priv: drm file-private structure
+ *
+ * Change the handle of a GEM object to the specified one.
+ * The new handle must be unused. On success the old handle is closed
+ * and all further IOCTL should refer to the new handle only.
+ * Calls to DRM_IOCTL_PRIME_FD_TO_HANDLE will return the new handle.
+ */
+int drm_gem_change_handle_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv)
+{
+	struct drm_gem_change_handle *args = data;
+	struct drm_gem_object *obj;
+	int ret;
+
+	obj = drm_gem_object_lookup(file_priv, args->handle);
+	if (!obj)
+		return -ENOENT;
+
+	if (args->handle == args->new_handle)
+		return 0;
+
+	mutex_lock(&file_priv->prime.lock);
+	spin_lock(&file_priv->table_lock);
+
+	ret = idr_alloc(&file_priv->object_idr, obj, args->new_handle, args->new_handle + 1, GFP_NOWAIT);
+	if (ret < 0)
+		goto out_unlock_table;
+
+	spin_unlock(&file_priv->table_lock);
+
+	ret = drm_prime_add_buf_handle(&file_priv->prime, obj->dma_buf, args->new_handle);
+	if (ret < 0)
+		goto out_unlock_prime;
+
+	drm_prime_remove_buf_handle(&file_priv->prime, args->handle);
+
+	spin_lock(&file_priv->table_lock);
+	idr_remove(&file_priv->object_idr, args->handle);
+
+out_unlock_table:
+	spin_unlock(&file_priv->table_lock);
+out_unlock_prime:
+	mutex_unlock(&file_priv->prime.lock);
+
+	return ret;
+}
+
 /**
  * drm_gem_open_ioctl - implementation of the GEM_OPEN ioctl
  * @dev: drm_device
drivers/gpu/drm/drm_internal.h +4 -0
index b2b6a8e49dda..e9d5cdf7e033 100644
@@ -85,6 +85,8 @@ int drm_prime_fd_to_handle_ioctl(struct drm_device *dev, void *data,
 
 void drm_prime_init_file_private(struct drm_prime_file_private *prime_fpriv);
 void drm_prime_destroy_file_private(struct drm_prime_file_private *prime_fpriv);
+int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
+			     struct dma_buf *dma_buf, uint32_t handle);
 void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 				 uint32_t handle);
 
@@ -168,6 +170,8 @@ int drm_gem_close_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
 int drm_gem_flink_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
+int drm_gem_change_handle_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
 int drm_gem_open_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv);
 void drm_gem_open(struct drm_device *dev, struct drm_file *file_private);
drivers/gpu/drm/drm_ioctl.c +1 -0
index f593dc569d31..d8a24875a7ba 100644
@@ -653,6 +653,7 @@ static const struct drm_ioctl_desc drm_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CLOSE, drm_gem_close_ioctl, DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_FLINK, drm_gem_flink_ioctl, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_OPEN, drm_gem_open_ioctl, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CHANGE_HANDLE, drm_gem_change_handle_ioctl, DRM_RENDER_ALLOW),
 
 	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETRESOURCES, drm_mode_getresources, 0),
 
drivers/gpu/drm/drm_prime.c +1 -5
index bdb51c8f262e..1f2e858e5000 100644
@@ -93,7 +93,7 @@ struct drm_prime_member {
 	struct rb_node handle_rb;
 };
 
-static int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
+int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
 				    struct dma_buf *dma_buf, uint32_t handle)
 {
 	struct drm_prime_member *member;
@@ -190,8 +190,6 @@ void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 {
 	struct rb_node *rb;
 
-	mutex_lock(&prime_fpriv->lock);
-
 	rb = prime_fpriv->handles.rb_node;
 	while (rb) {
 		struct drm_prime_member *member;
@@ -210,8 +208,6 @@ void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 			rb = rb->rb_left;
 		}
 	}
-
-	mutex_unlock(&prime_fpriv->lock);
 }
 
 void drm_prime_init_file_private(struct drm_prime_file_private *prime_fpriv)
include/uapi/drm/drm.h +17 -0
index 7fba37b94401..84c819c171d2 100644
@@ -625,6 +625,15 @@ struct drm_gem_open {
 	__u64 size;
 };
 
+/* DRM_IOCTL_GEM_CHANGE_HANDLE ioctl argument type */
+struct drm_gem_change_handle {
+	/** Current handle of object */
+	__u32 handle;
+
+	/** Handle to change that object to */
+	__u32 new_handle;
+};
+
 /**
  * DRM_CAP_DUMB_BUFFER
  *
@@ -1305,6 +1314,14 @@ extern "C" {
  */
 #define DRM_IOCTL_SET_CLIENT_NAME	DRM_IOWR(0xD1, struct drm_set_client_name)
 
+/**
+ * DRM_IOCTL_GEM_CHANGE_HANDLE - Move an object to a different handle
+ *
+ * Some applications (notably CRIU) need objects to have specific gem handles.
+ * This ioctl changes the object at one gem handle to use a new gem handle.
+ */
+#define DRM_IOCTL_GEM_CHANGE_HANDLE    DRM_IOWR(0xD2, struct drm_gem_change_handle)
+
 /*
  * Device specific ioctls should only be in their respective headers
  * The device specific ioctl range is from 0x40 to 0x9f.
Signature · 1 line
2.34.1
maintainer
#2 lore raw

Re: [PATCH 1/3] drm: Add DRM prime interfaces to reassign GEM handle

Quoted 74 lines · David Francis

CRIU restore of drm buffer objects requires the ability to create or import a buffer object with a specific gem handle.

Add new drm ioctl DRM_IOCTL_GEM_CHANGE_HANDLE, which takes the gem handle of an object and moves that object to a specified new gem handle.

This ioctl needs to call drm_prime_remove_buf_handle, but that function acquires the prime lock, which the ioctl needs to hold for other purposes.

Make drm_prime_remove_buf_handle not acquire the prime lock, and change its other caller to reflect this.

---
 drivers/gpu/drm/drm_gem.c      | 56 ++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_internal.h |  4 +++
 drivers/gpu/drm/drm_ioctl.c    |  1 +
 drivers/gpu/drm/drm_prime.c    |  6 +---
 include/uapi/drm/drm.h         | 17 +++++++++++
 5 files changed, 79 insertions(+), 5 deletions(-)
drivers/gpu/drm/drm_gem.c +35 -0
index c6240bab3fa5..31fe2f1d1137 100644
@@ -282,7 +282,12 @@ drm_gem_object_release_handle(int id, void *ptr, void *data)
 	if (obj->funcs->close)
 		obj->funcs->close(obj, file_priv);
 
+	mutex_lock(&file_priv->prime.lock);
+
 	drm_prime_remove_buf_handle(&file_priv->prime, id);
+
+	mutex_unlock(&file_priv->prime.lock);
+
 	drm_vma_node_revoke(&obj->vma_node, file_priv);
 
 	drm_gem_object_handle_put_unlocked(obj);
@@ -888,6 +893,57 @@ drm_gem_flink_ioctl(struct drm_device *dev, void *data,
 	return ret;
 }
 
+/**
+ * drm_gem_open_ioctl - implementation of the GEM_CHANGE_HANDLE ioctl
+ * @dev: drm_device
+ * @data: ioctl data
+ * @file_priv: drm file-private structure
+ *
+ * Change the handle of a GEM object to the specified one.
+ * The new handle must be unused. On success the old handle is closed
+ * and all further IOCTL should refer to the new handle only.
+ * Calls to DRM_IOCTL_PRIME_FD_TO_HANDLE will return the new handle.
+ */
+int drm_gem_change_handle_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv)
+{
+	struct drm_gem_change_handle *args = data;
+	struct drm_gem_object *obj;
+	int ret;
+
+	obj = drm_gem_object_lookup(file_priv, args->handle);
+	if (!obj)
+		return -ENOENT;
+
+	if (args->handle == args->new_handle)
+		return 0;
+
+	mutex_lock(&file_priv->prime.lock);
+	spin_lock(&file_priv->table_lock);
+
+	ret = idr_alloc(&file_priv->object_idr, obj, args->new_handle, args->new_handle + 1, GFP_NOWAIT);
+	if (ret < 0)

drivers/gpu/drm/drm_gem.c drm_gem_flink_ioctl(struct drm_device *dev, void *data, · line 935

Quoted 10 lines
drivers/gpu/drm/drm_gem.c +9 -0
+		goto out_unlock_table;
+
+	spin_unlock(&file_priv->table_lock);
+
 
+	ret = drm_prime_add_buf_handle(&file_priv->prime, obj->dma_buf, args->new_handle);
+	if (ret < 0)
+		goto out_unlock_prime;
+
+	drm_prime_remove_buf_handle(&file_priv->prime, args->handle);

We need to put this into an "if (obj->dma_buf)", exporting a GEM handle as DMA-buf is only optional.

@Sima any objections to this? Or can you think of a better approach?

Thanks,
Christian.

Quoted 115 lines
drivers/gpu/drm/drm_gem.c +12 -0
+
+	spin_lock(&file_priv->table_lock);
+	idr_remove(&file_priv->object_idr, args->handle);
+
+out_unlock_table:
+	spin_unlock(&file_priv->table_lock);
+out_unlock_prime:
+	mutex_unlock(&file_priv->prime.lock);
+
+	return ret;
+}
+
 /**
  * drm_gem_open_ioctl - implementation of the GEM_OPEN ioctl

* @dev: drm_device

drivers/gpu/drm/drm_internal.h +4 -0
index b2b6a8e49dda..e9d5cdf7e033 100644
@@ -85,6 +85,8 @@ int drm_prime_fd_to_handle_ioctl(struct drm_device *dev, void *data,
 
 void drm_prime_init_file_private(struct drm_prime_file_private *prime_fpriv);
 void drm_prime_destroy_file_private(struct drm_prime_file_private *prime_fpriv);
+int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
+			     struct dma_buf *dma_buf, uint32_t handle);
 void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 				 uint32_t handle);
 
@@ -168,6 +170,8 @@ int drm_gem_close_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
 int drm_gem_flink_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
+int drm_gem_change_handle_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
 int drm_gem_open_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv);
 void drm_gem_open(struct drm_device *dev, struct drm_file *file_private);
drivers/gpu/drm/drm_ioctl.c +1 -0
index f593dc569d31..d8a24875a7ba 100644
@@ -653,6 +653,7 @@ static const struct drm_ioctl_desc drm_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CLOSE, drm_gem_close_ioctl, DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_FLINK, drm_gem_flink_ioctl, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_IOCTL_GEM_OPEN, drm_gem_open_ioctl, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_IOCTL_GEM_CHANGE_HANDLE, drm_gem_change_handle_ioctl, DRM_RENDER_ALLOW),
 
 	DRM_IOCTL_DEF(DRM_IOCTL_MODE_GETRESOURCES, drm_mode_getresources, 0),
 
drivers/gpu/drm/drm_prime.c +1 -5
index bdb51c8f262e..1f2e858e5000 100644
@@ -93,7 +93,7 @@ struct drm_prime_member {
 	struct rb_node handle_rb;
 };
 
-static int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
+int drm_prime_add_buf_handle(struct drm_prime_file_private *prime_fpriv,
 				    struct dma_buf *dma_buf, uint32_t handle)
 {
 	struct drm_prime_member *member;
@@ -190,8 +190,6 @@ void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 {
 	struct rb_node *rb;
 
-	mutex_lock(&prime_fpriv->lock);
-
 	rb = prime_fpriv->handles.rb_node;
 	while (rb) {
 		struct drm_prime_member *member;
@@ -210,8 +208,6 @@ void drm_prime_remove_buf_handle(struct drm_prime_file_private *prime_fpriv,
 			rb = rb->rb_left;
 		}
 	}
-
-	mutex_unlock(&prime_fpriv->lock);
 }
 
 void drm_prime_init_file_private(struct drm_prime_file_private *prime_fpriv)
include/uapi/drm/drm.h +17 -0
index 7fba37b94401..84c819c171d2 100644
@@ -625,6 +625,15 @@ struct drm_gem_open {
 	__u64 size;
 };
 
+/* DRM_IOCTL_GEM_CHANGE_HANDLE ioctl argument type */
+struct drm_gem_change_handle {
+	/** Current handle of object */
+	__u32 handle;
+
+	/** Handle to change that object to */
+	__u32 new_handle;
+};
+
 /**
  * DRM_CAP_DUMB_BUFFER
  *
@@ -1305,6 +1314,14 @@ extern "C" {
  */
 #define DRM_IOCTL_SET_CLIENT_NAME	DRM_IOWR(0xD1, struct drm_set_client_name)
 
+/**
+ * DRM_IOCTL_GEM_CHANGE_HANDLE - Move an object to a different handle
+ *
+ * Some applications (notably CRIU) need objects to have specific gem handles.
+ * This ioctl changes the object at one gem handle to use a new gem handle.
+ */
+#define DRM_IOCTL_GEM_CHANGE_HANDLE    DRM_IOWR(0xD2, struct drm_gem_change_handle)
+
 /*
  * Device specific ioctls should only be in their respective headers
  * The device specific ioctl range is from 0x40 to 0x9f.
#3 lore raw

PATCH 2/3 drm/amdgpu: Adding amdgpu CRIU ioctl

In reply to David Francis

amdgpu CRIU requires an amdgpu CRIU ioctl. This ioctl has a similar interface to the amdkfd CRIU ioctl.

The objects that can be checkpointed and restored are bos and vm mappings. Because a single amdgpu bo can have multiple mappings. the mappings are recorded separately.

The ioctl has two modes: PROCESS_INFO, which sends to the user how many bos and vms to expect, and CHECKPOINT, which copies data about bos and vms into user-provided buffers.

Restore is handled using existing amdgpu and drm ioctls.

The new ioctl lives in a new file amdgpu_criu.c with its own header amdgpu_criu.h

---
 drivers/gpu/drm/amd/amdgpu/Makefile      |   2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c | 234 +++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h |  34 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c  |   3 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   |  11 ++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h   |  10 +
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c |   2 +
 include/uapi/drm/amdgpu_drm.h            |  62 ++++++
 8 files changed, 357 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -1
index 87080c06e5fc..0863edcdd03f 100644
@@ -63,7 +63,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
 	amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \
 	amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \
 	amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
-	amdgpu_fw_attestation.o amdgpu_securedisplay.o \
+	amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_criu.o \
 	amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
 	amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \
 	amdgpu_cper.o amdgpu_userq_fence.o amdgpu_eviction_fence.o
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +234 -0

234 lines not on this page · raw message

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h new file +34 -0
new file mode 100644
index 000000000000..9c196973ed0f
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: MIT */
+/*
+* Copyright 2025 Advanced Micro Devices, Inc.
+*
+* Permission is hereby granted, free of charge, to any person obtaining a
+* copy of this software and associated documentation files (the "Software"),
+* to deal in the Software without restriction, including without limitation
+* the rights to use, copy, modify, merge, publish, distribute, sublicense,
+* and/or sell copies of the Software, and to permit persons to whom the
+* Software is furnished to do so, subject to the following conditions:
+*
+* The above copyright notice and this permission notice shall be included in
+* all copies or substantial portions of the Software.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+* OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+#ifndef __AMDGPU_CRIU_H__
+#define __AMDGPU_CRIU_H__
+
+#include <drm/amdgpu_drm.h>
+
+int amdgpu_criu_bo_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+int amdgpu_criu_mapping_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+
+#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +3 -0
index 4db92e0a60da..5f3de93a665d 100644
@@ -53,6 +53,7 @@
 #include "amdgpu_xgmi.h"
 #include "amdgpu_userq.h"
 #include "amdgpu_userq_fence.h"
+#include "amdgpu_criu.h"
 #include "../amdxcp/amdgpu_xcp_drv.h"
 
 /*
@@ -3021,6 +3022,8 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ, amdgpu_userq_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_SIGNAL, amdgpu_userq_signal_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_WAIT, amdgpu_userq_wait_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_BO_INFO, amdgpu_criu_bo_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_MAPPING_INFO, amdgpu_criu_mapping_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +11 -0
index 3911c78f8282..f803908cf46d 100644
@@ -3156,3 +3156,14 @@ bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo)
 {
 	return bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv;
 }
+
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(struct amdgpu_vm *avm, uint64_t start, uint64_t end)
+{
+	return amdgpu_vm_it_iter_first(&avm->va, start, end);
+}
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end)
+{
+	return amdgpu_vm_it_iter_next(mapping, start, end);
+}
\ No newline at end of file
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +10 -0
index f3ad687125ad..cd7d3940cc7a 100644
@@ -668,4 +668,14 @@ void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm,
 				 struct dma_fence **fence);
 
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(
+	struct amdgpu_vm *avm, uint64_t start, uint64_t end);
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(
+	struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end);
+
+#define amdgpu_vm_it_for_each_entry(avm, mapping, start, end) \
+	for (mapping = amdgpu_vm_it_first_mapping_in_range(avm, start, end); \
+		mapping; \
+		mapping = amdgpu_vm_it_next_mapping_in_range(mapping, start, end))
+
 #endif
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +2 -0
index a2149afa5803..a8cf2d4580cc 100644
@@ -45,6 +45,8 @@
 #include "amdgpu_dma_buf.h"
 #include "kfd_debug.h"
 
+#include "amdgpu_criu.h"
+
 static long kfd_ioctl(struct file *, unsigned int, unsigned long);
 static int kfd_open(struct inode *, struct file *);
 static int kfd_release(struct inode *, struct file *);
include/uapi/drm/amdgpu_drm.h +62 -0

80 lines not on this page · raw message

Signature · 1 line
2.34.1
maintainer
#4 lore raw

Re: [PATCH 2/3] drm/amdgpu: Adding amdgpu CRIU ioctl

Quoted 161 lines · David Francis

amdgpu CRIU requires an amdgpu CRIU ioctl. This ioctl has a similar interface to the amdkfd CRIU ioctl.

The objects that can be checkpointed and restored are bos and vm mappings. Because a single amdgpu bo can have multiple mappings. the mappings are recorded separately.

The ioctl has two modes: PROCESS_INFO, which sends to the user how many bos and vms to expect, and CHECKPOINT, which copies data about bos and vms into user-provided buffers.

Restore is handled using existing amdgpu and drm ioctls.

The new ioctl lives in a new file amdgpu_criu.c with its own header amdgpu_criu.h

---
 drivers/gpu/drm/amd/amdgpu/Makefile      |   2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c | 234 +++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h |  34 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c  |   3 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   |  11 ++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h   |  10 +
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c |   2 +
 include/uapi/drm/amdgpu_drm.h            |  62 ++++++
 8 files changed, 357 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -1
index 87080c06e5fc..0863edcdd03f 100644
@@ -63,7 +63,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
 	amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \
 	amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \
 	amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
-	amdgpu_fw_attestation.o amdgpu_securedisplay.o \
+	amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_criu.o \
 	amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
 	amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \
 	amdgpu_cper.o amdgpu_userq_fence.o amdgpu_eviction_fence.o
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +112 -0
new file mode 100644
index 000000000000..8141ab09698c
@@ -0,0 +1,234 @@
+/* SPDX-License-Identifier: MIT */
+/*
+* Copyright 2025 Advanced Micro Devices, Inc.
+*
+* Permission is hereby granted, free of charge, to any person obtaining a
+* copy of this software and associated documentation files (the "Software"),
+* to deal in the Software without restriction, including without limitation
+* the rights to use, copy, modify, merge, publish, distribute, sublicense,
+* and/or sell copies of the Software, and to permit persons to whom the
+* Software is furnished to do so, subject to the following conditions:
+*
+* The above copyright notice and this permission notice shall be included in
+* all copies or substantial portions of the Software.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+* OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+#include <linux/dma-buf.h>
+#include <linux/hashtable.h>
+#include <linux/mutex.h>
+#include <linux/random.h>
+
+#include <drm/amdgpu_drm.h>
+#include <drm/drm_device.h>
+#include <drm/drm_file.h>
+
+#include "amdgpu_criu.h"
+
+#include <drm/amdgpu_drm.h>
+#include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_gem_ttm_helper.h>
+#include <drm/ttm/ttm_tt.h>
+#include <linux/interval_tree_generic.h>
+
+#include "amdgpu.h"
+#include "amdgpu_display.h"
+#include "amdgpu_dma_buf.h"
+#include "amdgpu_hmm.h"
+#include "amdgpu_xgmi.h"
+
+static uint32_t hardware_flags_to_uapi_flags(struct amdgpu_device *adev, uint64_t pte_flags)
+{
+	uint32_t gem_flags = 0;
+
+	//This function will be replaced by the mapping flags rework
+
+	if (pte_flags & AMDGPU_PTE_EXECUTABLE)
+		gem_flags |= AMDGPU_VM_PAGE_EXECUTABLE;
+	if (pte_flags & AMDGPU_PTE_READABLE)
+		gem_flags |= AMDGPU_VM_PAGE_READABLE;
+	if (pte_flags & AMDGPU_PTE_WRITEABLE)
+		gem_flags |= AMDGPU_VM_PAGE_WRITEABLE;
+	if (pte_flags & AMDGPU_PTE_PRT_FLAG(adev))
+		gem_flags |= AMDGPU_VM_PAGE_PRT;
+	if (pte_flags & AMDGPU_PTE_NOALLOC)
+		gem_flags |= AMDGPU_VM_PAGE_NOALLOC;
+
+	return gem_flags;
+}
+
+
+/**
+ * amdgpu_criu_bo_info_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_criu_bo_info_args
+ * @filp: drm file pointer
+ *
+ * num_bos is set as an input to the size of the bo_buckets array.
+ * num_bos is sent back as output as the number of bos in the process.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_criu_bo_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp)
+{
+	struct drm_amdgpu_criu_bo_info_args *args = data;
+	struct drm_amdgpu_criu_bo_bucket *bo_buckets;
+	struct drm_gem_object *gobj;
+	int id, ret = 0;
+	int bo_index = 0;
+	int num_bos = 0;
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id)
+		num_bos += 1;
+	spin_unlock(&filp->table_lock);
+
+	if (args->num_bos < num_bos) {
+		args->num_bos = num_bos;
+		goto exit;
+	}
+	args->num_bos = num_bos;
+	if (num_bos == 0) {
+		goto exit;
+	}
+
+	bo_buckets = kvzalloc(num_bos * sizeof(*bo_buckets), GFP_KERNEL);
+	if (!bo_buckets) {
+		ret = -ENOMEM;
+		goto free_buckets;
+	}

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 122

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+		struct drm_amdgpu_criu_bo_bucket *bo_bucket;
+
+		bo_bucket = &bo_buckets[bo_index];
+
+		bo_bucket->size = amdgpu_bo_size(bo);
+		bo_bucket->offset = amdgpu_bo_mmap_offset(bo);

Not all BOs have a MMAP offset. Additional to that we already have an IOCTL to query the mmap offset which does the appropriate checks and doesn't potentially crash when there is none.

Quoted 31 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +31 -0
+		bo_bucket->alloc_flags = bo->flags & (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE);
+		bo_bucket->preferred_domains = bo->preferred_domains;
+		bo_bucket->gem_handle = id;
+
+		if (bo->tbo.base.import_attach)
+			bo_bucket->flags |= AMDGPU_CRIU_BO_FLAG_IS_IMPORT;
+
+		bo_index += 1;
+	}
+	spin_unlock(&filp->table_lock);
+
+	ret = copy_to_user((void __user *)args->bo_buckets, bo_buckets, num_bos * sizeof(*bo_buckets));
+	if (ret) {
+		pr_debug("Failed to copy BO information to user\n");
+		ret = -EFAULT;
+	}
+
+free_buckets:
+	kvfree(bo_buckets);
+exit:
+
+	return ret;
+}
+
+/**
+ * amdgpu_criu_bo_info_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_criu_mapping_info_args
+ * @filp: drm file pointer
+ *

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 163

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+ * num_mappings is set as an input to the size of the vm_buckets array.
+ * num_mappings is sent back as output as the number of mappings the bo has.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_criu_mapping_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp)

Please split adding that into a separate patch.

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 174

Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +11 -0
+{
+	struct drm_amdgpu_criu_mapping_info_args *args = data;
+	struct drm_gem_object *gobj = idr_find(&filp->object_idr, args->gem_handle);
+	struct amdgpu_vm *avm = &((struct amdgpu_fpriv *)filp->driver_priv)->vm;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+	struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(avm, bo);
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct drm_amdgpu_criu_vm_bucket *vm_buckets;
+	struct amdgpu_bo_va_mapping *mapping;
+	struct drm_exec exec;
 
+	int vm_index = 0;

That needs a better name.

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 175

Quoted 1 line
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +1 -0
+	int ret = 0;

Don't initialize return variables if you don't need it. That is very bad practice.

Quoted 26 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +24 -0
+
+	vm_buckets = kvzalloc(args->num_mappings * sizeof(*vm_buckets), GFP_KERNEL);
+	if (!vm_buckets) {
+		ret = -ENOMEM;
+		goto free_vms;
+	}
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		if (gobj) {
+			ret = drm_exec_lock_obj(&exec, gobj);
+			drm_exec_retry_on_contention(&exec);
+			if (ret)
+				goto unlock_exec;
+		}
+
+		ret = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			goto unlock_exec;
+	}
+
 
 
+	amdgpu_vm_it_for_each_entry(avm, mapping, 0, U64_MAX) {

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 209

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+		if (mapping->bo_va == bo_va) {
+			if (vm_index < args->num_mappings) {
+				vm_buckets[vm_index].start = mapping->start;
+				vm_buckets[vm_index].last = mapping->last;
+				vm_buckets[vm_index].offset = mapping->offset;
+				vm_buckets[vm_index].flags = hardware_flags_to_uapi_flags(drm_to_adev(dev), mapping->flags);
+			}
+			vm_index += 1;
+		}
+	}

This chunk should go into amdgpu_vm.c And I strongly suggest to not go over the mapping rb tree but rather the list in the bo_va.

Regards,
Christian.

Quoted 225 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +25 -0
+
+	drm_exec_fini(&exec);
+
+	if (vm_index > 0) {
+		if (vm_index <= args->num_mappings) {
+			ret = copy_to_user((void __user *)args->vm_buckets, vm_buckets, vm_index * sizeof(*vm_buckets));
+			if (ret) {
+				pr_debug("Failed to copy BO information to user\n");
+				ret = -EFAULT;
+			}
+		}
+	}
+	args->num_mappings = vm_index;
+
+
+	kvfree(vm_buckets);
+
+	return ret;
+unlock_exec:
+	drm_exec_fini(&exec);
+free_vms:
+	kvfree(vm_buckets);
+
+	return ret;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h new file +34 -0
new file mode 100644
index 000000000000..9c196973ed0f
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: MIT */
+/*
+* Copyright 2025 Advanced Micro Devices, Inc.
+*
+* Permission is hereby granted, free of charge, to any person obtaining a
+* copy of this software and associated documentation files (the "Software"),
+* to deal in the Software without restriction, including without limitation
+* the rights to use, copy, modify, merge, publish, distribute, sublicense,
+* and/or sell copies of the Software, and to permit persons to whom the
+* Software is furnished to do so, subject to the following conditions:
+*
+* The above copyright notice and this permission notice shall be included in
+* all copies or substantial portions of the Software.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+* OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+#ifndef __AMDGPU_CRIU_H__
+#define __AMDGPU_CRIU_H__
+
+#include <drm/amdgpu_drm.h>
+
+int amdgpu_criu_bo_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+int amdgpu_criu_mapping_info_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+
+#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +3 -0
index 4db92e0a60da..5f3de93a665d 100644
@@ -53,6 +53,7 @@
 #include "amdgpu_xgmi.h"
 #include "amdgpu_userq.h"
 #include "amdgpu_userq_fence.h"
+#include "amdgpu_criu.h"
 #include "../amdxcp/amdgpu_xcp_drv.h"
 
 /*
@@ -3021,6 +3022,8 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ, amdgpu_userq_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_SIGNAL, amdgpu_userq_signal_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_WAIT, amdgpu_userq_wait_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_BO_INFO, amdgpu_criu_bo_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_MAPPING_INFO, amdgpu_criu_mapping_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +11 -0
index 3911c78f8282..f803908cf46d 100644
@@ -3156,3 +3156,14 @@ bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo)
 {
 	return bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv;
 }
+
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(struct amdgpu_vm *avm, uint64_t start, uint64_t end)
+{
+	return amdgpu_vm_it_iter_first(&avm->va, start, end);
+}
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end)
+{
+	return amdgpu_vm_it_iter_next(mapping, start, end);
+}
\ No newline at end of file
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +10 -0
index f3ad687125ad..cd7d3940cc7a 100644
@@ -668,4 +668,14 @@ void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm,
 				 struct dma_fence **fence);
 
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(
+	struct amdgpu_vm *avm, uint64_t start, uint64_t end);
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(
+	struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end);
+
+#define amdgpu_vm_it_for_each_entry(avm, mapping, start, end) \
+	for (mapping = amdgpu_vm_it_first_mapping_in_range(avm, start, end); \
+		mapping; \
+		mapping = amdgpu_vm_it_next_mapping_in_range(mapping, start, end))
+
 #endif
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +2 -0
index a2149afa5803..a8cf2d4580cc 100644
@@ -45,6 +45,8 @@
 #include "amdgpu_dma_buf.h"
 #include "kfd_debug.h"
 
+#include "amdgpu_criu.h"
+
 static long kfd_ioctl(struct file *, unsigned int, unsigned long);
 static int kfd_open(struct inode *, struct file *);
 static int kfd_release(struct inode *, struct file *);
include/uapi/drm/amdgpu_drm.h +62 -0
index 45c4fa13499c..16aee825e116 100644
@@ -57,6 +57,10 @@ extern "C" {
 #define DRM_AMDGPU_USERQ		0x16
 #define DRM_AMDGPU_USERQ_SIGNAL		0x17
 #define DRM_AMDGPU_USERQ_WAIT		0x18
+#define DRM_AMDGPU_CRIU_OP		0x19
+
+#define DRM_AMDGPU_CRIU_BO_INFO	0x20
+#define DRM_AMDGPU_CRIU_MAPPING_INFO	0x21
 
 #define DRM_IOCTL_AMDGPU_GEM_CREATE	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_CREATE, union drm_amdgpu_gem_create)
 #define DRM_IOCTL_AMDGPU_GEM_MMAP	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_MMAP, union drm_amdgpu_gem_mmap)
@@ -77,6 +81,10 @@ extern "C" {
 #define DRM_IOCTL_AMDGPU_USERQ		DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ, union drm_amdgpu_userq)
 #define DRM_IOCTL_AMDGPU_USERQ_SIGNAL	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ_SIGNAL, struct drm_amdgpu_userq_signal)
 #define DRM_IOCTL_AMDGPU_USERQ_WAIT	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ_WAIT, struct drm_amdgpu_userq_wait)
+#define DRM_IOCTL_AMDGPU_CRIU_OP	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_OP, struct drm_amdgpu_criu_args)
+
+#define DRM_IOCTL_AMDGPU_CRIU_BO_INFO	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_BO_INFO, struct drm_amdgpu_criu_bo_info_args)
+#define DRM_IOCTL_AMDGPU_CRIU_MAPPING_INFO	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_MAPPING_INFO, struct drm_amdgpu_criu_mapping_info_args)
 
 /**
  * DOC: memory domains
@@ -1626,6 +1634,60 @@ struct drm_color_ctm_3x4 {
 	__u64 matrix[12];
 };
 
+#define AMDGPU_CRIU_BO_FLAG_IS_IMPORT	(1 << 0)
+
+struct drm_amdgpu_criu_bo_info_args {
+    /* IN: Size of bo_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+    __u32   num_bos;
+
+    /* User pointer to array of drm_amdgpu_criu_bo_bucket */
+    __u64   bo_buckets;
+};
+
+struct drm_amdgpu_criu_bo_bucket {
+    /* Size of bo */
+    __u64 size;
+
+    /* Offset of bo in device file */
+    __u64 offset;
+
+    /* GEM_CREATE flags for re-creation of buffer */
+    __u64 alloc_flags;
+
+    /* Pending how to handle this; provides information needed to remake the buffer on restore */
+    __u32 preferred_domains;
+
+    /* Currently just one flag: IS_IMPORT */
+    __u32 flags;
+
+    __u32 gem_handle;
+};
+
+struct drm_amdgpu_criu_mapping_info_args {
+    /* Handle of bo to get mappings of */
+    __u32   gem_handle;
+
+    /* IN: Size of vm_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+    __u32   num_mappings;
+
+    /* User pointer to array of drm_amdgpu_criu_vm_bucket */
+    __u64   vm_buckets;
+};
+
+struct drm_amdgpu_criu_vm_bucket {
+    /* Start of mapping (in number of pages) */
+    __u64 start;
+
+    /* End of mapping (in number of pages) */
+    __u64 last;
+
+    /* Mapping offset */
+    __u64 offset;
+
+    /* flags needed to recreate mapping; still pending how to get these */
+    __u64 flags;
+};
+
 #if defined(__cplusplus)
 }
 #endif
#5 lore raw

Re: [PATCH 2/3] drm/amdgpu: Adding amdgpu CRIU ioctl

[AMD Official Use Only - AMD Internal Distribution Only]

Quoted 12 lines

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 209

Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +11 -0
+     amdgpu_vm_it_for_each_entry(avm, mapping, 0, U64_MAX) {
+             if (mapping->bo_va == bo_va) {
+                     if (vm_index < args->num_mappings) {
+                             vm_buckets[vm_index].start = mapping->start;
+                             vm_buckets[vm_index].last = mapping->last;
+                             vm_buckets[vm_index].offset = mapping->offset;
+                             vm_buckets[vm_index].flags = hardware_flags_to_uapi_flags(drm_to_adev(dev), mapping->flags);
+                     }
+                     vm_index += 1;
+             }
+     }
Quoted 2 lines

This chunk should go into amdgpu_vm.c And I strongly suggest to not go over the mapping rb tree but rather the list in the bo_va.

Do you mean the invalids and valids lists? I remember you having objections to using those in a previous patch.

David
________________________________
From: Koenig, Christian <[email protected]>
Sent: Monday, June 16, 2025 5:01 AM
To: Francis, David <[email protected]>; [email protected] <[email protected]>
Cc: [email protected] <[email protected]>; Kuehling, Felix <[email protected]>; Yat Sin, David <[email protected]>; Freehill, Chris <[email protected]>; [email protected] <[email protected]>; [email protected] <[email protected]>; [email protected] <[email protected]>; [email protected] <[email protected]>; [email protected] <[email protected]>
Subject: Re: [PATCH 2/3] drm/amdgpu: Adding amdgpu CRIU ioctl

Quoted 161 lines · David Francis

amdgpu CRIU requires an amdgpu CRIU ioctl. This ioctl has a similar interface to the amdkfd CRIU ioctl.

The objects that can be checkpointed and restored are bos and vm mappings. Because a single amdgpu bo can have multiple mappings. the mappings are recorded separately.

The ioctl has two modes: PROCESS_INFO, which sends to the user how many bos and vms to expect, and CHECKPOINT, which copies data about bos and vms into user-provided buffers.

Restore is handled using existing amdgpu and drm ioctls.

The new ioctl lives in a new file amdgpu_criu.c with its own header amdgpu_criu.h

---
 drivers/gpu/drm/amd/amdgpu/Makefile      |   2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c | 234 +++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h |  34 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c  |   3 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   |  11 ++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h   |  10 +
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c |   2 +
 include/uapi/drm/amdgpu_drm.h            |  62 ++++++
 8 files changed, 357 insertions(+), 1 deletion(-)
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c
 create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -1
index 87080c06e5fc..0863edcdd03f 100644
@@ -63,7 +63,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
       amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \
       amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \
       amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
-     amdgpu_fw_attestation.o amdgpu_securedisplay.o \
+     amdgpu_fw_attestation.o amdgpu_securedisplay.o amdgpu_criu.o \
       amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
       amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \
       amdgpu_cper.o amdgpu_userq_fence.o amdgpu_eviction_fence.o
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +112 -0
new file mode 100644
index 000000000000..8141ab09698c
@@ -0,0 +1,234 @@
+/* SPDX-License-Identifier: MIT */
+/*
+* Copyright 2025 Advanced Micro Devices, Inc.
+*
+* Permission is hereby granted, free of charge, to any person obtaining a
+* copy of this software and associated documentation files (the "Software"),
+* to deal in the Software without restriction, including without limitation
+* the rights to use, copy, modify, merge, publish, distribute, sublicense,
+* and/or sell copies of the Software, and to permit persons to whom the
+* Software is furnished to do so, subject to the following conditions:
+*
+* The above copyright notice and this permission notice shall be included in
+* all copies or substantial portions of the Software.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+* OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+#include <linux/dma-buf.h>
+#include <linux/hashtable.h>
+#include <linux/mutex.h>
+#include <linux/random.h>
+
+#include <drm/amdgpu_drm.h>
+#include <drm/drm_device.h>
+#include <drm/drm_file.h>
+
+#include "amdgpu_criu.h"
+
+#include <drm/amdgpu_drm.h>
+#include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_gem_ttm_helper.h>
+#include <drm/ttm/ttm_tt.h>
+#include <linux/interval_tree_generic.h>
+
+#include "amdgpu.h"
+#include "amdgpu_display.h"
+#include "amdgpu_dma_buf.h"
+#include "amdgpu_hmm.h"
+#include "amdgpu_xgmi.h"
+
+static uint32_t hardware_flags_to_uapi_flags(struct amdgpu_device *adev, uint64_t pte_flags)
+{
+     uint32_t gem_flags = 0;
+
+     //This function will be replaced by the mapping flags rework
+
+     if (pte_flags & AMDGPU_PTE_EXECUTABLE)
+             gem_flags |= AMDGPU_VM_PAGE_EXECUTABLE;
+     if (pte_flags & AMDGPU_PTE_READABLE)
+             gem_flags |= AMDGPU_VM_PAGE_READABLE;
+     if (pte_flags & AMDGPU_PTE_WRITEABLE)
+             gem_flags |= AMDGPU_VM_PAGE_WRITEABLE;
+     if (pte_flags & AMDGPU_PTE_PRT_FLAG(adev))
+             gem_flags |= AMDGPU_VM_PAGE_PRT;
+     if (pte_flags & AMDGPU_PTE_NOALLOC)
+             gem_flags |= AMDGPU_VM_PAGE_NOALLOC;
+
+     return gem_flags;
+}
+
+
+/**
+ * amdgpu_criu_bo_info_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_criu_bo_info_args
+ * @filp: drm file pointer
+ *
+ * num_bos is set as an input to the size of the bo_buckets array.
+ * num_bos is sent back as output as the number of bos in the process.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_criu_bo_info_ioctl(struct drm_device *dev, void *data,
+                         struct drm_file *filp)
+{
+     struct drm_amdgpu_criu_bo_info_args *args = data;
+     struct drm_amdgpu_criu_bo_bucket *bo_buckets;
+     struct drm_gem_object *gobj;
+     int id, ret = 0;
+     int bo_index = 0;
+     int num_bos = 0;
+
+     spin_lock(&filp->table_lock);
+     idr_for_each_entry(&filp->object_idr, gobj, id)
+             num_bos += 1;
+     spin_unlock(&filp->table_lock);
+
+     if (args->num_bos < num_bos) {
+             args->num_bos = num_bos;
+             goto exit;
+     }
+     args->num_bos = num_bos;
+     if (num_bos == 0) {
+             goto exit;
+     }
+
+     bo_buckets = kvzalloc(num_bos * sizeof(*bo_buckets), GFP_KERNEL);
+     if (!bo_buckets) {
+             ret = -ENOMEM;
+             goto free_buckets;
+     }

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 122

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+
+     spin_lock(&filp->table_lock);
+     idr_for_each_entry(&filp->object_idr, gobj, id) {
+             struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+             struct drm_amdgpu_criu_bo_bucket *bo_bucket;
+
+             bo_bucket = &bo_buckets[bo_index];
+
+             bo_bucket->size = amdgpu_bo_size(bo);
+             bo_bucket->offset = amdgpu_bo_mmap_offset(bo);

Not all BOs have a MMAP offset. Additional to that we already have an IOCTL to query the mmap offset which does the appropriate checks and doesn't potentially crash when there is none.

Quoted 31 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +31 -0
+             bo_bucket->alloc_flags = bo->flags & (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE);
+             bo_bucket->preferred_domains = bo->preferred_domains;
+             bo_bucket->gem_handle = id;
+
+             if (bo->tbo.base.import_attach)
+                     bo_bucket->flags |= AMDGPU_CRIU_BO_FLAG_IS_IMPORT;
+
+             bo_index += 1;
+     }
+     spin_unlock(&filp->table_lock);
+
+     ret = copy_to_user((void __user *)args->bo_buckets, bo_buckets, num_bos * sizeof(*bo_buckets));
+     if (ret) {
+             pr_debug("Failed to copy BO information to user\n");
+             ret = -EFAULT;
+     }
+
+free_buckets:
+     kvfree(bo_buckets);
+exit:
+
+     return ret;
+}
+
+/**
+ * amdgpu_criu_bo_info_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_criu_mapping_info_args
+ * @filp: drm file pointer
+ *

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 163

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+ * num_mappings is set as an input to the size of the vm_buckets array.
+ * num_mappings is sent back as output as the number of mappings the bo has.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_criu_mapping_info_ioctl(struct drm_device *dev, void *data,
+                         struct drm_file *filp)

Please split adding that into a separate patch.

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 174

Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +11 -0
+{
+     struct drm_amdgpu_criu_mapping_info_args *args = data;
+     struct drm_gem_object *gobj = idr_find(&filp->object_idr, args->gem_handle);
+     struct amdgpu_vm *avm = &((struct amdgpu_fpriv *)filp->driver_priv)->vm;
+     struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+     struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(avm, bo);
+     struct amdgpu_fpriv *fpriv = filp->driver_priv;
+     struct drm_amdgpu_criu_vm_bucket *vm_buckets;
+     struct amdgpu_bo_va_mapping *mapping;
+     struct drm_exec exec;
 
+     int vm_index = 0;

That needs a better name.

Quoted 1 line

+ int ret = 0;

Don't initialize return variables if you don't need it. That is very bad practice.

Quoted 26 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +24 -0
+
+     vm_buckets = kvzalloc(args->num_mappings * sizeof(*vm_buckets), GFP_KERNEL);
+     if (!vm_buckets) {
+             ret = -ENOMEM;
+             goto free_vms;
+     }
+
+     drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+                   DRM_EXEC_IGNORE_DUPLICATES, 0);
+     drm_exec_until_all_locked(&exec) {
+             if (gobj) {
+                     ret = drm_exec_lock_obj(&exec, gobj);
+                     drm_exec_retry_on_contention(&exec);
+                     if (ret)
+                             goto unlock_exec;
+             }
+
+             ret = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2);
+             drm_exec_retry_on_contention(&exec);
+             if (ret)
+                     goto unlock_exec;
+     }
+
 
 
+     amdgpu_vm_it_for_each_entry(avm, mapping, 0, U64_MAX) {

drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c line 209

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +10 -0
+             if (mapping->bo_va == bo_va) {
+                     if (vm_index < args->num_mappings) {
+                             vm_buckets[vm_index].start = mapping->start;
+                             vm_buckets[vm_index].last = mapping->last;
+                             vm_buckets[vm_index].offset = mapping->offset;
+                             vm_buckets[vm_index].flags = hardware_flags_to_uapi_flags(drm_to_adev(dev), mapping->flags);
+                     }
+                     vm_index += 1;
+             }
+     }

This chunk should go into amdgpu_vm.c And I strongly suggest to not go over the mapping rb tree but rather the list in the bo_va.

Regards,
Christian.

Quoted 225 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.c new file +25 -0
+
+     drm_exec_fini(&exec);
+
+     if (vm_index > 0) {
+             if (vm_index <= args->num_mappings) {
+                     ret = copy_to_user((void __user *)args->vm_buckets, vm_buckets, vm_index * sizeof(*vm_buckets));
+                     if (ret) {
+                             pr_debug("Failed to copy BO information to user\n");
+                             ret = -EFAULT;
+                     }
+             }
+     }
+     args->num_mappings = vm_index;
+
+
+     kvfree(vm_buckets);
+
+     return ret;
+unlock_exec:
+     drm_exec_fini(&exec);
+free_vms:
+     kvfree(vm_buckets);
+
+     return ret;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_criu.h new file +34 -0
new file mode 100644
index 000000000000..9c196973ed0f
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: MIT */
+/*
+* Copyright 2025 Advanced Micro Devices, Inc.
+*
+* Permission is hereby granted, free of charge, to any person obtaining a
+* copy of this software and associated documentation files (the "Software"),
+* to deal in the Software without restriction, including without limitation
+* the rights to use, copy, modify, merge, publish, distribute, sublicense,
+* and/or sell copies of the Software, and to permit persons to whom the
+* Software is furnished to do so, subject to the following conditions:
+*
+* The above copyright notice and this permission notice shall be included in
+* all copies or substantial portions of the Software.
+*
+* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+* OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+#ifndef __AMDGPU_CRIU_H__
+#define __AMDGPU_CRIU_H__
+
+#include <drm/amdgpu_drm.h>
+
+int amdgpu_criu_bo_info_ioctl(struct drm_device *dev, void *data,
+                         struct drm_file *filp);
+int amdgpu_criu_mapping_info_ioctl(struct drm_device *dev, void *data,
+                         struct drm_file *filp);
+
+#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +3 -0
index 4db92e0a60da..5f3de93a665d 100644
@@ -53,6 +53,7 @@
 #include "amdgpu_xgmi.h"
 #include "amdgpu_userq.h"
 #include "amdgpu_userq_fence.h"
+#include "amdgpu_criu.h"
 #include "../amdxcp/amdgpu_xcp_drv.h"
 
 /*
@@ -3021,6 +3022,8 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
       DRM_IOCTL_DEF_DRV(AMDGPU_USERQ, amdgpu_userq_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
       DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_SIGNAL, amdgpu_userq_signal_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
       DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_WAIT, amdgpu_userq_wait_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+     DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_BO_INFO, amdgpu_criu_bo_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+     DRM_IOCTL_DEF_DRV(AMDGPU_CRIU_MAPPING_INFO, amdgpu_criu_mapping_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +11 -0
index 3911c78f8282..f803908cf46d 100644
@@ -3156,3 +3156,14 @@ bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo)
 {
       return bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv;
 }
+
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(struct amdgpu_vm *avm, uint64_t start, uint64_t end)
+{
+     return amdgpu_vm_it_iter_first(&avm->va, start, end);
+}
+
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end)
+{
+     return amdgpu_vm_it_iter_next(mapping, start, end);
+}
\ No newline at end of file
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +10 -0
index f3ad687125ad..cd7d3940cc7a 100644
@@ -668,4 +668,14 @@ void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
                                struct amdgpu_vm *vm,
                                struct dma_fence **fence);
 
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_first_mapping_in_range(
+     struct amdgpu_vm *avm, uint64_t start, uint64_t end);
+struct amdgpu_bo_va_mapping *amdgpu_vm_it_next_mapping_in_range(
+     struct amdgpu_bo_va_mapping *mapping, uint64_t start, uint64_t end);
+
+#define amdgpu_vm_it_for_each_entry(avm, mapping, start, end) \
+     for (mapping = amdgpu_vm_it_first_mapping_in_range(avm, start, end); \
+             mapping; \
+             mapping = amdgpu_vm_it_next_mapping_in_range(mapping, start, end))
+
 #endif
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +2 -0
index a2149afa5803..a8cf2d4580cc 100644
@@ -45,6 +45,8 @@
 #include "amdgpu_dma_buf.h"
 #include "kfd_debug.h"
 
+#include "amdgpu_criu.h"
+
 static long kfd_ioctl(struct file *, unsigned int, unsigned long);
 static int kfd_open(struct inode *, struct file *);
 static int kfd_release(struct inode *, struct file *);
include/uapi/drm/amdgpu_drm.h +62 -0
index 45c4fa13499c..16aee825e116 100644
@@ -57,6 +57,10 @@ extern "C" {
 #define DRM_AMDGPU_USERQ             0x16
 #define DRM_AMDGPU_USERQ_SIGNAL              0x17
 #define DRM_AMDGPU_USERQ_WAIT                0x18
+#define DRM_AMDGPU_CRIU_OP           0x19
+
+#define DRM_AMDGPU_CRIU_BO_INFO      0x20
+#define DRM_AMDGPU_CRIU_MAPPING_INFO 0x21
 
 #define DRM_IOCTL_AMDGPU_GEM_CREATE  DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_CREATE, union drm_amdgpu_gem_create)
 #define DRM_IOCTL_AMDGPU_GEM_MMAP    DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_MMAP, union drm_amdgpu_gem_mmap)
@@ -77,6 +81,10 @@ extern "C" {
 #define DRM_IOCTL_AMDGPU_USERQ               DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ, union drm_amdgpu_userq)
 #define DRM_IOCTL_AMDGPU_USERQ_SIGNAL        DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ_SIGNAL, struct drm_amdgpu_userq_signal)
 #define DRM_IOCTL_AMDGPU_USERQ_WAIT  DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_USERQ_WAIT, struct drm_amdgpu_userq_wait)
+#define DRM_IOCTL_AMDGPU_CRIU_OP     DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_OP, struct drm_amdgpu_criu_args)
+
+#define DRM_IOCTL_AMDGPU_CRIU_BO_INFO        DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_BO_INFO, struct drm_amdgpu_criu_bo_info_args)
+#define DRM_IOCTL_AMDGPU_CRIU_MAPPING_INFO   DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_CRIU_MAPPING_INFO, struct drm_amdgpu_criu_mapping_info_args)
 
 /**
  * DOC: memory domains
@@ -1626,6 +1634,60 @@ struct drm_color_ctm_3x4 {
       __u64 matrix[12];
 };
 
+#define AMDGPU_CRIU_BO_FLAG_IS_IMPORT        (1 << 0)
+
+struct drm_amdgpu_criu_bo_info_args {
+    /* IN: Size of bo_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+    __u32   num_bos;
+
+    /* User pointer to array of drm_amdgpu_criu_bo_bucket */
+    __u64   bo_buckets;
+};
+
+struct drm_amdgpu_criu_bo_bucket {
+    /* Size of bo */
+    __u64 size;
+
+    /* Offset of bo in device file */
+    __u64 offset;
+
+    /* GEM_CREATE flags for re-creation of buffer */
+    __u64 alloc_flags;
+
+    /* Pending how to handle this; provides information needed to remake the buffer on restore */
+    __u32 preferred_domains;
+
+    /* Currently just one flag: IS_IMPORT */
+    __u32 flags;
+
+    __u32 gem_handle;
+};
+
+struct drm_amdgpu_criu_mapping_info_args {
+    /* Handle of bo to get mappings of */
+    __u32   gem_handle;
+
+    /* IN: Size of vm_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+    __u32   num_mappings;
+
+    /* User pointer to array of drm_amdgpu_criu_vm_bucket */
+    __u64   vm_buckets;
+};
+
+struct drm_amdgpu_criu_vm_bucket {
+    /* Start of mapping (in number of pages) */
+    __u64 start;
+
+    /* End of mapping (in number of pages) */
+    __u64 last;
+
+    /* Mapping offset */
+    __u64 offset;
+
+    /* flags needed to recreate mapping; still pending how to get these */
+    __u64 flags;
+};
+
 #if defined(__cplusplus)
 }
 #endif
#6 lore raw

PATCH 3/3 drm/amdgpu: Allow kfd CRIU with no buffer objects

In reply to David Francis

The kfd CRIU checkpoint ioctl would return an error if trying to checkpoint a process with no kfd buffer objects.

This is a normal case and should not be an error.

---
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +2 -2
index a8cf2d4580cc..9617c696d5b6 100644
@@ -2569,8 +2569,8 @@ static int criu_restore(struct file *filep,
 	pr_debug("CRIU restore (num_devices:%u num_bos:%u num_objects:%u priv_data_size:%llu)\n",
 		 args->num_devices, args->num_bos, args->num_objects, args->priv_data_size);
 
-	if (!args->bos || !args->devices || !args->priv_data || !args->priv_data_size ||
-	    !args->num_devices || !args->num_bos)
+	if ((args->num_bos > 0 && !args->bos) || !args->devices || !args->priv_data ||
+	    !args->priv_data_size || !args->num_devices)
 		return -EINVAL;
 
 	mutex_lock(&p->mutex);
Signature · 1 line
2.34.1