Skip to contentSkip to conversation

Add CRIU support for amdgpu dmabuf

PATCH Superseded David Francis 2 participants 6 messages Aug 7, 2025

Subsystems (4)

v10 of 16 Superseded Latest is v16 →

lore
Share & tools
lore
#0 lore raw

PATCH v10 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 v10 moves the ioctls out of their own fie, and into amdgpu_gem.c

Accompanying CRIU changes:
https://github.com/checkpoint-restore/criu/pull/2613

#1 lore raw

PATCH v10 1/3 drm/amdgpu: Add ioctl to get bo info

Add new ioctl DRM_IOCTL_AMDGPU_GEM_BO_INFO.

This ioctl returns a list of bos with their handles, sizes, and flags and domains.

This ioctl is meant to be used during CRIU checkpoint and provide information needed to reconstruct the bos in CRIU restore.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 73 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h |  2 +
 include/uapi/drm/amdgpu_drm.h           | 33 +++++++++++
 4 files changed, 109 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1 -0
index 4ff3a2eaaf55..9ee4644b508d 100644
@@ -3031,6 +3031,7 @@ 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_GEM_BO_INFO, amdgpu_gem_bo_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +73 -0

79 lines not on this page · raw message

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h +2 -0
index b51e8f95ee86..d33b19fd1488 100644
@@ -67,6 +67,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
+int amdgpu_gem_bo_info_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp);
 
 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
include/uapi/drm/amdgpu_drm.h +33 -0
index bdedbaccf776..1b3d2ba2f556 100644
@@ -57,6 +57,7 @@ extern "C" {
 #define DRM_AMDGPU_USERQ		0x16
 #define DRM_AMDGPU_USERQ_SIGNAL		0x17
 #define DRM_AMDGPU_USERQ_WAIT		0x18
+#define DRM_AMDGPU_GEM_BO_INFO		0x19
 
 #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 +78,7 @@ 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_GEM_BO_INFO	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_BO_INFO, struct drm_amdgpu_gem_bo_info)
 
 /**
  * DOC: memory domains
@@ -811,6 +813,37 @@ struct drm_amdgpu_gem_op {
 	__u64	value;
 };
 
+#define AMDGPU_GEM_BO_INFO_FLAG_IS_IMPORT	(1 << 0)
+
+struct drm_amdgpu_gem_bo_info {
+	/* User pointer to array of drm_amdgpu_criu_bo_bucket */
+	__u64   bo_buckets;
+
+	/* IN: Size of bo_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+	__u32   num_bos;
+
+	__u32 padding;
+};
+
+struct drm_amdgpu_gem_bo_info_bucket {
+	/* Size of bo */
+	__u64 size;
+
+	/* 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;
+
+	/* gem handle of buffer object */
+	__u32 gem_handle;
+
+	__u32 padding;
+};
+
 #define AMDGPU_VA_OP_MAP			1
 #define AMDGPU_VA_OP_UNMAP			2
 #define AMDGPU_VA_OP_CLEAR			3
Signature · 1 line
2.34.1
maintainer
#2 lore raw

Re: [PATCH v10 1/3] drm/amdgpu: Add ioctl to get bo info

Quoted 56 lines · David Francis

Add new ioctl DRM_IOCTL_AMDGPU_GEM_BO_INFO.

This ioctl returns a list of bos with their handles, sizes, and flags and domains.

This ioctl is meant to be used during CRIU checkpoint and provide information needed to reconstruct the bos in CRIU restore.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 73 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h |  2 +
 include/uapi/drm/amdgpu_drm.h           | 33 +++++++++++
 4 files changed, 109 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1 -0
index 4ff3a2eaaf55..9ee4644b508d 100644
@@ -3031,6 +3031,7 @@ 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_GEM_BO_INFO, amdgpu_gem_bo_info_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +19 -0
index e3f65977eeee..ec80b751e697 100644
@@ -1032,6 +1032,79 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 	return r;
 }
 
+/**
+ * drm_amdgpu_gem_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_gem_bo_info_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp)
+{
+	struct drm_amdgpu_gem_bo_info *args = data;

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1063

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +10 -0
+	struct drm_amdgpu_gem_bo_info_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);

Yeah that won't work likes this.

Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +9 -0
+
+	if (args->num_bos < num_bos) {
+		args->num_bos = num_bos;
+		return 0;
+	}
+	args->num_bos = num_bos;
+	if (num_bos == 0)
+		return 0;
+

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1082

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +10 -0
+	bo_buckets = kvcalloc(num_bos, sizeof(*bo_buckets), GFP_KERNEL);
+	if (!bo_buckets)
+		return -ENOMEM;
+
+	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_gem_bo_info_bucket *bo_bucket;
+
+		bo_bucket = &bo_buckets[bo_index];

You are relying on that the table won't change after dropping the spinlock, but that's not guaranteed.

You need some extra check here to not write over the end of the buffer or otherwise you create a really nice API to hijack the kernel.

Quoted 46 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +25 -0
+
+		bo_bucket->size = amdgpu_bo_size(bo);
+		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_GEM_BO_INFO_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;
+	}
+
+	kvfree(bo_buckets);
+
+	return ret;
+}
+
+
 static int amdgpu_gem_align_pitch(struct amdgpu_device *adev,
 				  int width,
 				  int cpp,
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h +2 -0
index b51e8f95ee86..d33b19fd1488 100644
@@ -67,6 +67,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
+int amdgpu_gem_bo_info_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp);
 
 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
include/uapi/drm/amdgpu_drm.h +0 -0
index bdedbaccf776..1b3d2ba2f556 100644

include/uapi/drm/amdgpu_drm.h extern "C" { · line 60

Quoted 4 lines
include/uapi/drm/amdgpu_drm.h +1 -0
@@ -57,6 +57,7 @@ extern "C" {
 #define DRM_AMDGPU_USERQ		0x16
 #define DRM_AMDGPU_USERQ_SIGNAL		0x17
 #define DRM_AMDGPU_USERQ_WAIT		0x18
+#define DRM_AMDGPU_GEM_BO_INFO		0x19

BO_INFO is a bit misleading, maybe name that DRM_AMDGPU_GEM_HANDLES.

Quoted 12 lines
include/uapi/drm/amdgpu_drm.h +1 -0
 #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 +78,7 @@ 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_GEM_BO_INFO	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_BO_INFO, struct drm_amdgpu_gem_bo_info)
 
 /**
  * DOC: memory domains

include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_gem_op { · line 819

Quoted 7 lines
include/uapi/drm/amdgpu_drm.h +4 -0
@@ -811,6 +813,37 @@ struct drm_amdgpu_gem_op {
 	__u64	value;
 };
 
+#define AMDGPU_GEM_BO_INFO_FLAG_IS_IMPORT	(1 << 0)
+
+struct drm_amdgpu_gem_bo_info {
+	/* User pointer to array of drm_amdgpu_criu_bo_bucket */

That structure was renamed.

include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_gem_op { · line 828

Quoted 9 lines
include/uapi/drm/amdgpu_drm.h +9 -0
+	__u64   bo_buckets;
+
+	/* IN: Size of bo_buckets buffer. OUT: Number of bos in process (if larger than size of buffer, must retry) */
+	__u32   num_bos;
+
+	__u32 padding;
+};
+
+struct drm_amdgpu_gem_bo_info_bucket {

Why bucket? Maybe "entry" instead?

Quoted 4 lines
include/uapi/drm/amdgpu_drm.h +4 -0
+	/* Size of bo */
+	__u64 size;
+
+	/* GEM_CREATE flags for re-creation of buffer */

include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_gem_op { · line 842

Quoted 10 lines
include/uapi/drm/amdgpu_drm.h +10 -0
+	__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;
+
+	/* gem handle of buffer object */
+	__u32 gem_handle;

Make that the first member, it's basically the whole reason we have that here.

Regards,
Christian.

Quoted 7 lines
include/uapi/drm/amdgpu_drm.h +4 -0
+
+	__u32 padding;
+};
+
 #define AMDGPU_VA_OP_MAP			1
 #define AMDGPU_VA_OP_UNMAP			2
 #define AMDGPU_VA_OP_CLEAR			3
#3 lore raw

PATCH v10 2/3 drm/amdgpu: Add mapping info option for GEM_OP ioctl

In reply to David Francis

Add new GEM_OP_IOCTL option GET_MAPPING_INFO, which returns a list of mappings associated with a given bo, along with their positions and offsets.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 97 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |  5 ++
 include/uapi/drm/amdgpu_drm.h           | 18 +++++
 3 files changed, 120 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +97 -0
index ec80b751e697..8bdb346f1761 100644
@@ -956,6 +956,100 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	return r;
 }
 
+/**
+ * amdgpu_criu_mapping_info_ioctl - get information about a buffer's mappings
+ *
+ * @gobj: gem object
+ * @args: gem_op arguments
+ * @fpriv: drm file pointer
+ *
+ * num_entries is set as an input to the size of the user-allocated array of
+ * drm_amdgpu_gem_vm_bucket stored at args->value.
+ * num_entries 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.
+ */
+static int amdgpu_gem_list_mappings(struct drm_gem_object *gobj, struct amdgpu_fpriv *fpriv,
+					  struct drm_amdgpu_gem_op *args)
+{
+	struct amdgpu_vm *avm = &fpriv->vm;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+	struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(avm, bo);
+	struct drm_amdgpu_gem_vm_bucket *vm_buckets;
+	struct amdgpu_bo_va_mapping *mapping;
+	struct drm_exec exec;
+	int num_mappings = 0;
+	int ret;
+
+	vm_buckets = kvcalloc(args->num_entries, 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_bo_va_for_each_valid_mapping(bo_va, mapping) {
+		if (num_mappings < args->num_entries) {
+			vm_buckets[num_mappings].start = mapping->start;
+			vm_buckets[num_mappings].last = mapping->last;
+			vm_buckets[num_mappings].offset = mapping->offset;
+			vm_buckets[num_mappings].flags = mapping->flags;
+		}
+		num_mappings += 1;
+	}
+
+	amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) {
+		if (num_mappings < args->num_entries) {
+			vm_buckets[num_mappings].start = mapping->start;
+			vm_buckets[num_mappings].last = mapping->last;
+			vm_buckets[num_mappings].offset = mapping->offset;
+			vm_buckets[num_mappings].flags = mapping->flags;
+		}
+		num_mappings += 1;
+	}
+
+	drm_exec_fini(&exec);
+
+	if (num_mappings > 0) {
+		if (num_mappings <= args->num_entries) {
+			ret = copy_to_user((void __user *)args->value, vm_buckets, num_mappings * sizeof(*vm_buckets));
+			if (ret) {
+				pr_debug("Failed to copy BO information to user\n");
+				ret = -EFAULT;
+			}
+		}
+	}
+	args->num_entries = num_mappings;
+
+	kvfree(vm_buckets);
+
+	return ret;
+unlock_exec:
+	drm_exec_fini(&exec);
+free_vms:
+	kvfree(vm_buckets);
+
+	return ret;
+}
+
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
@@ -1022,6 +1116,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 
 		amdgpu_bo_unreserve(robj);
 		break;
+	case AMDGPU_GEM_OP_GET_MAPPING_INFO:
+		r = amdgpu_gem_list_mappings(gobj, filp->driver_priv, args);
+		break;
 	default:
 		amdgpu_bo_unreserve(robj);
 		r = -EINVAL;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +5 -0
index f9549f6b3d1f..5a63ae490b0e 100644
@@ -668,4 +668,9 @@ void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm,
 				 struct dma_fence **fence);
 
+#define amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &bo_va->valids, list)
+#define amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &bo_va->invalids, list)
+
 #endif
include/uapi/drm/amdgpu_drm.h +18 -0
index 1b3d2ba2f556..079e780dc20d 100644
@@ -802,6 +802,21 @@ 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
+
+struct drm_amdgpu_gem_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 */
+	__u64 flags;
+};
 
 /* Sets or returns a value associated with a buffer. */
 struct drm_amdgpu_gem_op {
@@ -811,6 +826,9 @@ struct drm_amdgpu_gem_op {
 	__u32	op;
 	/** Input or return value */
 	__u64	value;
+	/** For MAPPING_INFO op: number of mappings (in/out) */
+	__u32	num_entries;
+	__u32	padding;
 };
 
 #define AMDGPU_GEM_BO_INFO_FLAG_IS_IMPORT	(1 << 0)
Signature · 1 line
2.34.1
maintainer
#4 lore raw

Re: [PATCH v10 2/3] drm/amdgpu: Add mapping info option for GEM_OP ioctl

Quoted 16 lines · David Francis

Add new GEM_OP_IOCTL option GET_MAPPING_INFO, which returns a list of mappings associated with a given bo, along with their positions and offsets.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 97 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |  5 ++
 include/uapi/drm/amdgpu_drm.h           | 18 +++++
 3 files changed, 120 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +0 -0
index ec80b751e697..8bdb346f1761 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, · line 960

Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +2 -0
@@ -956,6 +956,100 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	return r;
 }
 
+/**
+ * amdgpu_criu_mapping_info_ioctl - get information about a buffer's mappings

The function name was updated.

Quoted 62 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +62 -0
+ *
+ * @gobj: gem object
+ * @args: gem_op arguments
+ * @fpriv: drm file pointer
+ *
+ * num_entries is set as an input to the size of the user-allocated array of
+ * drm_amdgpu_gem_vm_bucket stored at args->value.
+ * num_entries 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.
+ */
+static int amdgpu_gem_list_mappings(struct drm_gem_object *gobj, struct amdgpu_fpriv *fpriv,
+					  struct drm_amdgpu_gem_op *args)
+{
+	struct amdgpu_vm *avm = &fpriv->vm;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+	struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(avm, bo);
+	struct drm_amdgpu_gem_vm_bucket *vm_buckets;
+	struct amdgpu_bo_va_mapping *mapping;
+	struct drm_exec exec;
+	int num_mappings = 0;
+	int ret;
+
+	vm_buckets = kvcalloc(args->num_entries, 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_bo_va_for_each_valid_mapping(bo_va, mapping) {
+		if (num_mappings < args->num_entries) {
+			vm_buckets[num_mappings].start = mapping->start;
+			vm_buckets[num_mappings].last = mapping->last;
+			vm_buckets[num_mappings].offset = mapping->offset;
+			vm_buckets[num_mappings].flags = mapping->flags;
+		}
+		num_mappings += 1;
+	}
+
+	amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) {
+		if (num_mappings < args->num_entries) {
+			vm_buckets[num_mappings].start = mapping->start;
+			vm_buckets[num_mappings].last = mapping->last;

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, · line 1032

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +10 -0
+			vm_buckets[num_mappings].offset = mapping->offset;
+			vm_buckets[num_mappings].flags = mapping->flags;
+		}
+		num_mappings += 1;
+	}
+
+	drm_exec_fini(&exec);
+
+	if (num_mappings > 0) {
+		if (num_mappings <= args->num_entries) {

That can be combined to just one if.

Apart from that looks good to me,
Christian.

Quoted 83 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +23 -0
+			ret = copy_to_user((void __user *)args->value, vm_buckets, num_mappings * sizeof(*vm_buckets));
+			if (ret) {
+				pr_debug("Failed to copy BO information to user\n");
+				ret = -EFAULT;
+			}
+		}
+	}
+	args->num_entries = num_mappings;
+
+	kvfree(vm_buckets);
+
+	return ret;
+unlock_exec:
+	drm_exec_fini(&exec);
+free_vms:
+	kvfree(vm_buckets);
+
+	return ret;
+}
+
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
@@ -1022,6 +1116,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 
 		amdgpu_bo_unreserve(robj);
 		break;
+	case AMDGPU_GEM_OP_GET_MAPPING_INFO:
+		r = amdgpu_gem_list_mappings(gobj, filp->driver_priv, args);
+		break;
 	default:
 		amdgpu_bo_unreserve(robj);
 		r = -EINVAL;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +5 -0
index f9549f6b3d1f..5a63ae490b0e 100644
@@ -668,4 +668,9 @@ void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm,
 				 struct dma_fence **fence);
 
+#define amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &bo_va->valids, list)
+#define amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &bo_va->invalids, list)
+
 #endif
include/uapi/drm/amdgpu_drm.h +18 -0
index 1b3d2ba2f556..079e780dc20d 100644
@@ -802,6 +802,21 @@ 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
+
+struct drm_amdgpu_gem_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 */
+	__u64 flags;
+};
 
 /* Sets or returns a value associated with a buffer. */
 struct drm_amdgpu_gem_op {
@@ -811,6 +826,9 @@ struct drm_amdgpu_gem_op {
 	__u32	op;
 	/** Input or return value */
 	__u64	value;
+	/** For MAPPING_INFO op: number of mappings (in/out) */
+	__u32	num_entries;
+	__u32	padding;
 };
 
 #define AMDGPU_GEM_BO_INFO_FLAG_IS_IMPORT	(1 << 0)
#5 lore raw

PATCH v10 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 828a9ceef1e7..f7f34b710d3e 100644
@@ -2566,8 +2566,8 @@ 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