Add CRIU support for amdgpu dmabuf
Subsystems (4)
PATCH v12 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 v12 fixes some code path and code standard identified by Christian
Accompanying CRIU changes:
https://github.com/checkpoint-restore/criu/pull/2613
PATCH v12 1/3 drm/amdgpu: Add ioctl to get all gem handles for a process
Add new ioctl DRM_IOCTL_AMDGPU_GEM_LIST_HANDLES.
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 | 83 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h | 2 + include/uapi/drm/amdgpu_drm.h | 31 +++++++++ 4 files changed, 117 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1 -0
@@ -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_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), }; static const struct drm_driver amdgpu_kms_driver = {
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +83 -0
89 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h +2 -0
@@ -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_list_handles_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 +31 -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_LIST_HANDLES 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_LIST_HANDLES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_LIST_HANDLES, struct drm_amdgpu_gem_list_handles)/** * DOC: memory domains
@@ -811,6 +813,35 @@ struct drm_amdgpu_gem_op {__u64 value; }; +#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0) + +struct drm_amdgpu_gem_list_handles { +/* User pointer to array of drm_amdgpu_gem_bo_info_entry */ +__u64 entries; + +/* IN: Size of entries buffer. OUT: Number of handles in process (if larger than size of buffer, must retry) */ +__u32 num_entries; + +__u32 padding; +}; + +struct drm_amdgpu_gem_list_handles_entry { +/* gem handle of buffer object */ +__u32 gem_handle; + +/* Currently just one flag: IS_IMPORT */ +__u32 flags; + +/* Size of bo */ +__u64 size; + +/* Preferred domains for GEM_CREATE */ +__u64 preferred_domains; + +/* GEM_CREATE flags for re-creation of buffer */ +__u64 alloc_flags; +}; +#define AMDGPU_VA_OP_MAP 1#define AMDGPU_VA_OP_UNMAP 2#define AMDGPU_VA_OP_CLEAR 3
Signature · 1 line
2.34.1
Re: [PATCH v12 1/3] drm/amdgpu: Add ioctl to get all gem handles for a process
Quoted 89 lines · David Francis
Add new ioctl
DRM_IOCTL_AMDGPU_GEM_LIST_HANDLES.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.
Signed-off-by David Francis <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 83 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h | 2 + include/uapi/drm/amdgpu_drm.h | 31 +++++++++ 4 files changed, 117 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1
-0@@ -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_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), }; static const struct drm_driver amdgpu_kms_driver = {drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +52
-0@@ -1032,6 +1032,89 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,return r; } +/** +* drm_amdgpu_gem_list_handles_ioctl - get information about a process' buffer objects +* +* @dev: drm device pointer +* @data: drm_amdgpu_gem_list_handles +* @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_list_handles_ioctl (struct drm_device *dev,void *data, +struct drm_file *filp) +{ +struct drm_amdgpu_gem_list_handles *args = data; +struct drm_amdgpu_gem_list_handles_entry *bo_entries; +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_entries < num_bos) { + args->num_entries = num_bos; +return 0 ; + } + +if (num_bos ==0 ) { + args->num_entries =0 ; +return 0 ; + } + + bo_entries =kvcalloc (num_bos,sizeof (*bo_entries),GFP_KERNEL ); +if (!bo_entries) +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_list_handles_entry *bo_entry; + +if (bo_index >= num_bos) { + ret = -EAGAIN ; +break ;
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1096
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +10
-0+ } + + bo_entry = &bo_entries[bo_index]; + + bo_entry->size =amdgpu_bo_size (bo); + bo_entry->alloc_flags = bo->flags & (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE ); +/* WIPE_ON_RELEASE is set automatically in the driver; it is not permitted in +* BO creation. In the interest of giving the user exactly the flags they need +* to recreate the BO, clear it. +*/
Ha, curious. What is the reason flags userspace cannot use are specified in the uapi header?
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1101
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +5
-0+ bo_entry->preferred_domains = bo->preferred_domains; + bo_entry->gem_handle = id; + +if (bo->tbo.base.import_attach) + bo_entry->flags |=AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT ;
I had a question regarding this part in v11. Any comment on that?
Anyway, thanks for implementing the other changes I suggested, this one LGTM now.
Regards,
Tvrtko
Quoted 88 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +16
-0+ + bo_index +=1 ; + } +spin_unlock (&filp->table_lock); + + args->num_entries = bo_index; + +if (!ret) + ret =copy_to_user (u64_to_user_ptr (args->entries), bo_entries, num_bos *sizeof (*bo_entries)); + +kvfree (bo_entries); + +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@@ -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_list_handles_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 +31
-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_LIST_HANDLES 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_LIST_HANDLES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_LIST_HANDLES, struct drm_amdgpu_gem_list_handles)/** * DOC: memory domains @@ -811,6 +813,35 @@ struct drm_amdgpu_gem_op {__u64 value; }; +#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0) + +struct drm_amdgpu_gem_list_handles { +/* User pointer to array of drm_amdgpu_gem_bo_info_entry */ +__u64 entries; + +/* IN: Size of entries buffer. OUT: Number of handles in process (if larger than size of buffer, must retry) */ +__u32 num_entries; + +__u32 padding; +}; + +struct drm_amdgpu_gem_list_handles_entry { +/* gem handle of buffer object */ +__u32 gem_handle; + +/* Currently just one flag: IS_IMPORT */ +__u32 flags; + +/* Size of bo */ +__u64 size; + +/* Preferred domains for GEM_CREATE */ +__u64 preferred_domains; + +/* GEM_CREATE flags for re-creation of buffer */ +__u64 alloc_flags; +}; +#define AMDGPU_VA_OP_MAP 1#define AMDGPU_VA_OP_UNMAP 2#define AMDGPU_VA_OP_CLEAR 3
Christian König
maintainerRe: [PATCH v12 1/3] drm/amdgpu: Add ioctl to get all gem handles for a process
Quoted 93 lines · Tvrtko Ursulin
Quoted 91 lines · David Francis
Add new ioctl
DRM_IOCTL_AMDGPU_GEM_LIST_HANDLES.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.
Signed-off-by David Francis <[email protected]>---
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 83 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h | 2 + include/uapi/drm/amdgpu_drm.h | 31 +++++++++ 4 files changed, 117 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1
-0@@ -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_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), }; static const struct drm_driver amdgpu_kms_driver = {drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +0
-0@@ -1032,6 +1032,89 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, return r; }+/**
+ *drm_amdgpu_gem_list_handles_ioctl- get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data:drm_amdgpu_gem_list_handles
+ * @filp: drm file pointer
+ *
+ *num_bosis set as an input to the size of thebo_bucketsarray.
+ *num_bosis 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.
+ */drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +41
-0+int amdgpu_gem_list_handles_ioctl (struct drm_device *dev,void *data, + struct drm_file *filp) +{ + struct drm_amdgpu_gem_list_handles *args = data; + struct drm_amdgpu_gem_list_handles_entry *bo_entries; + 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_entries < num_bos) { + args->num_entries = num_bos; + return 0; + } + + if (num_bos == 0) { + args->num_entries = 0; + return 0; + } + + bo_entries = kvcalloc(num_bos, sizeof(*bo_entries), GFP_KERNEL); + if (!bo_entries) + 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_list_handles_entry *bo_entry; + + if (bo_index >= num_bos) { + ret = -EAGAIN; + break; + } + + bo_entry = &bo_entries[bo_index]; +
Quoted 8 lines
Quoted 6 lines
+
bo_entry->size=amdgpu_bo_size(bo);
+bo_entry->alloc_flags=bo->flags& (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE); + /*WIPE_ON_RELEASEis set automatically in the driver; it is not permitted in + * BO creation. In the interest of giving the user exactly the flags they need + * to recreate the BO, clear it.
+ */Ha, curious. What is the reason flags userspace cannot use are specified in the uapi header?
The comment isn't correct. AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE is perfectly allowed in the create IOCTL.
Regards,
Christian.
Quoted 100 lines
Quoted 6 lines
+
bo_entry->preferred_domains=bo->preferred_domains; +bo_entry->gem_handle= id;
+
+ if (bo->tbo.base.import_attach)
+bo_entry->flags|=AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;I had a question regarding this part in v11. Any comment on that?
Anyway, thanks for implementing the other changes I suggested, this one LGTM now.
Regards,
Tvrtko
Quoted 84 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +16
-0+ + bo_index += 1; + } + spin_unlock(&filp->table_lock); + + args->num_entries = bo_index; + + if (!ret) + ret = copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)); + + kvfree(bo_entries); + + 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@@ -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_list_handles_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 +30
-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_LIST_HANDLES 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_LIST_HANDLES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_LIST_HANDLES, struct drm_amdgpu_gem_list_handles) /*** DOC: memory domains @@ -811,6 +813,35 @@ struct drm_amdgpu_gem_op { __u64 value; }; +#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0) + +struct drm_amdgpu_gem_list_handles { + /* User pointer to array of drm_amdgpu_gem_bo_info_entry */ + __u64 entries; + + /* IN: Size of entries buffer. OUT: Number of handles in process (if larger than size of buffer, must retry) */ + __u32 num_entries; + + __u32 padding; +}; + +struct drm_amdgpu_gem_list_handles_entry { + /* gem handle of buffer object */ + __u32 gem_handle; + + /* Currently just one flag: IS_IMPORT */ + __u32 flags; + + /* Size of bo */ + __u64 size; + + /* Preferred domains for GEM_CREATE */ + __u64 preferred_domains; + + /* GEM_CREATE flags for re-creation of buffer */ + __u64 alloc_flags; +}; + #define AMDGPU_VA_OP_MAP 1 #define AMDGPU_VA_OP_UNMAP 2 #define AMDGPU_VA_OP_CLEAR 3
Re: [PATCH v12 1/3] drm/amdgpu: Add ioctl to get all gem handles for a process
Quoted 103 lines · Christian König
Quoted 93 lines · Tvrtko Ursulin
Quoted 91 lines · David Francis
Add new ioctl
DRM_IOCTL_AMDGPU_GEM_LIST_HANDLES.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.
Signed-off-by David Francis <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 83 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h | 2 + include/uapi/drm/amdgpu_drm.h | 31 +++++++++ 4 files changed, 117 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1
-0@@ -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_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), }; static const struct drm_driver amdgpu_kms_driver = {drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +55
-0@@ -1032,6 +1032,89 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, return r; } +/** +* drm_amdgpu_gem_list_handles_ioctl - get information about a process' buffer objects +* +* @dev: drm device pointer +* @data: drm_amdgpu_gem_list_handles +* @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_list_handles_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct drm_amdgpu_gem_list_handles *args = data; + struct drm_amdgpu_gem_list_handles_entry *bo_entries; + 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_entries < num_bos) { + args->num_entries = num_bos; + return 0; + } + + if (num_bos == 0) { + args->num_entries = 0; + return 0; + } + + bo_entries = kvcalloc(num_bos, sizeof(*bo_entries), GFP_KERNEL); + if (!bo_entries) + 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_list_handles_entry *bo_entry; + + if (bo_index >= num_bos) { + ret = -EAGAIN; + break; + } + + bo_entry = &bo_entries[bo_index]; +Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1096
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +6
-0+ bo_entry->size = amdgpu_bo_size(bo); + bo_entry->alloc_flags = bo->flags & (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE); + /* WIPE_ON_RELEASE is set automatically in the driver; it is not permitted in +* BO creation. In the interest of giving the user exactly the flags they need +* to recreate the BO, clear it. +*/ Ha, curious. What is the reason flags userspace cannot use are specified in the uapi header?
Quoted 1 line
The comment isn't correct.
AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASEis perfectly allowed in the create IOCTL.
I looked after reading the comment and it seemed true:
amdgpu_gem_create_ioctl():
...
uint64_t flags = args->in.domain_flags; ... /* reject invalid gem flags */ if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_NO_CPU_ACCESS | AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_VRAM_CLEARED | AMDGPU_GEM_CREATE_VM_ALWAYS_VALID | AMDGPU_GEM_CREATE_EXPLICIT_SYNC | AMDGPU_GEM_CREATE_ENCRYPTED | AMDGPU_GEM_CREATE_GFX12_DCC | AMDGPU_GEM_CREATE_DISCARDABLE)) return -EINVAL;
Actually, thinking about it more, maybe more flags should be cleared in amdgpu_gem_list_handles_ioctl()? Perhaps the above mask should be defined at a common place, or even internal only flags removed from the uapi header. Or uapi vs internal flags split or something.
Regards,
Tvrtko
Quoted 101 lines
Quoted 100 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1101
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +5
-0+ bo_entry->preferred_domains = bo->preferred_domains; + bo_entry->gem_handle = id; + + if (bo->tbo.base.import_attach) + bo_entry->flags |= AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;I had a question regarding this part in v11. Any comment on that?
Anyway, thanks for implementing the other changes I suggested, this one LGTM now.
Regards,
Tvrtko
Quoted 84 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +16
-0+ + bo_index += 1; + } + spin_unlock(&filp->table_lock); + + args->num_entries = bo_index; + + if (!ret) + ret = copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)); + + kvfree(bo_entries); + + 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@@ -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_list_handles_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 +30
-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_LIST_HANDLES 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_LIST_HANDLES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_LIST_HANDLES, struct drm_amdgpu_gem_list_handles) /*** DOC: memory domains @@ -811,6 +813,35 @@ struct drm_amdgpu_gem_op { __u64 value; }; +#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0) + +struct drm_amdgpu_gem_list_handles { + /* User pointer to array of drm_amdgpu_gem_bo_info_entry */ + __u64 entries; + + /* IN: Size of entries buffer. OUT: Number of handles in process (if larger than size of buffer, must retry) */ + __u32 num_entries; + + __u32 padding; +}; + +struct drm_amdgpu_gem_list_handles_entry { + /* gem handle of buffer object */ + __u32 gem_handle; + + /* Currently just one flag: IS_IMPORT */ + __u32 flags; + + /* Size of bo */ + __u64 size; + + /* Preferred domains for GEM_CREATE */ + __u64 preferred_domains; + + /* GEM_CREATE flags for re-creation of buffer */ + __u64 alloc_flags; +}; + #define AMDGPU_VA_OP_MAP 1 #define AMDGPU_VA_OP_UNMAP 2 #define AMDGPU_VA_OP_CLEAR 3
Christian König
maintainerRe: [PATCH v12 1/3] drm/amdgpu: Add ioctl to get all gem handles for a process
Quoted 105 lines · Tvrtko Ursulin
Quoted 103 lines · Christian König
Quoted 101 lines · Tvrtko Ursulin
Quoted 87 lines · David Francis
Add new ioctl
DRM_IOCTL_AMDGPU_GEM_LIST_HANDLES.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.
Signed-off-by David Francis <[email protected]>---
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 83 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h | 2 + include/uapi/drm/amdgpu_drm.h | 31 +++++++++ 4 files changed, 117 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1
-0@@ -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_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), }; static const struct drm_driver amdgpu_kms_driver = {drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +0
-0@@ -1032,6 +1032,89 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, return r; }+/**
+ *drm_amdgpu_gem_list_handles_ioctl- get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data:drm_amdgpu_gem_list_handles
+ * @filp: drm file pointer
+ *
+ *num_bosis set as an input to the size of thebo_bucketsarray.
+ *num_bosis 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.
+ */drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +37
-0+int amdgpu_gem_list_handles_ioctl (struct drm_device *dev,void *data, + struct drm_file *filp) +{ + struct drm_amdgpu_gem_list_handles *args = data; + struct drm_amdgpu_gem_list_handles_entry *bo_entries; + 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_entries < num_bos) { + args->num_entries = num_bos; + return 0; + } + + if (num_bos == 0) { + args->num_entries = 0; + return 0; + } + + bo_entries = kvcalloc(num_bos, sizeof(*bo_entries), GFP_KERNEL); + if (!bo_entries) + 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_list_handles_entry *bo_entry; + + if (bo_index >= num_bos) { + ret = -EAGAIN; + break;drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, · line 1096
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +10
-0+ } + + bo_entry = &bo_entries[bo_index]; + + bo_entry->size = amdgpu_bo_size(bo); + bo_entry->alloc_flags = bo->flags & (~AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE); + /* WIPE_ON_RELEASE is set automatically in the driver; it is not permitted in +* BO creation. In the interest of giving the user exactly the flags they need +* to recreate the BO, clear it. +*/ Ha, curious. What is the reason flags userspace cannot use are specified in the uapi header?
Quoted 3 lines
Quoted 1 line
The comment isn't correct.
AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASEis perfectly allowed in the create IOCTL.I looked after reading the comment and it seemed true:
Oh, good point. Now I understand the issue. The GEM interface doesn't allowed for the flag on create!
Quoted 17 lines
amdgpu_gem_create_ioctl():
...
uint64_t flags =args->in.domain_flags;
...
/* reject invalid gem flags */
if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |AMDGPU_GEM_CREATE_NO_CPU_ACCESS|AMDGPU_GEM_CREATE_CPU_GTT_USWC|AMDGPU_GEM_CREATE_VRAM_CLEARED|AMDGPU_GEM_CREATE_VM_ALWAYS_VALID|AMDGPU_GEM_CREATE_EXPLICIT_SYNC|AMDGPU_GEM_CREATE_ENCRYPTED|AMDGPU_GEM_CREATE_GFX12_DCC|AMDGPU_GEM_CREATE_DISCARDABLE)) return-EINVAL;
Quoted 1 line
Actually, thinking about it more, maybe more flags should be cleared in
amdgpu_gem_list_handles_ioctl()? Perhaps the above mask should be defined at a common place, or even internal only flags removed from the uapi header. Or uapi vs internal flags split or something.
Yeah, probably good idea. But it should most likely not be in an UAPI header.
The general background is that we have a bunch of flags only the kernel can set and are read only for userspace.
Regards,
Christian.
Quoted 107 lines
Regards,
Tvrtko
Quoted 101 lines
Quoted 100 lines
Quoted 6 lines
+
bo_entry->preferred_domains=bo->preferred_domains; +bo_entry->gem_handle= id;
+
+ if (bo->tbo.base.import_attach)
+bo_entry->flags|=AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;I had a question regarding this part in v11. Any comment on that?
Anyway, thanks for implementing the other changes I suggested, this one LGTM now.
Regards,
Tvrtko
Quoted 84 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +16
-0+ + bo_index += 1; + } + spin_unlock(&filp->table_lock); + + args->num_entries = bo_index; + + if (!ret) + ret = copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)); + + kvfree(bo_entries); + + 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@@ -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_list_handles_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 +30
-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_LIST_HANDLES 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_LIST_HANDLES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_LIST_HANDLES, struct drm_amdgpu_gem_list_handles) /*** DOC: memory domains @@ -811,6 +813,35 @@ struct drm_amdgpu_gem_op { __u64 value; }; +#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0) + +struct drm_amdgpu_gem_list_handles { + /* User pointer to array of drm_amdgpu_gem_bo_info_entry */ + __u64 entries; + + /* IN: Size of entries buffer. OUT: Number of handles in process (if larger than size of buffer, must retry) */ + __u32 num_entries; + + __u32 padding; +}; + +struct drm_amdgpu_gem_list_handles_entry { + /* gem handle of buffer object */ + __u32 gem_handle; + + /* Currently just one flag: IS_IMPORT */ + __u32 flags; + + /* Size of bo */ + __u64 size; + + /* Preferred domains for GEM_CREATE */ + __u64 preferred_domains; + + /* GEM_CREATE flags for re-creation of buffer */ + __u64 alloc_flags; +}; + #define AMDGPU_VA_OP_MAP 1 #define AMDGPU_VA_OP_UNMAP 2 #define AMDGPU_VA_OP_CLEAR 3
PATCH v12 2/3 drm/amdgpu: Add mapping info option for GEM_OP ioctl
In reply to David FrancisAdd 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 | 86 ++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 ++ include/uapi/drm/amdgpu_drm.h | 18 ++++++ 3 files changed, 108 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +85 -1
@@ -885,7 +885,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, goto error; } - r =amdgpu_vm_lock_pd (&fpriv->vm, &exec,2 ); + r =amdgpu_vm_lock_pd (&fpriv->vm, &exec,0 );drm_exec_retry_on_contention (&exec);if (unlikely (r)) goto error;
@@ -956,6 +956,86 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,return r; } +/** +* amdgpu_gem_list_mappings - 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) +return -ENOMEM ; + +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 ; + } + +if (num_mappings >0 && num_mappings <= args->num_entries) + ret =copy_to_user (u64_to_user_ptr (args->value), vm_buckets, num_mappings *sizeof (*vm_buckets)); + + args->num_entries = num_mappings; + +unlock_exec: +drm_exec_fini (&exec); +kvfree (vm_buckets); + +return ret; +} +int amdgpu_gem_op_ioctl (struct drm_device *dev,void *data,struct drm_file *filp) {
@@ -1022,6 +1102,10 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,amdgpu_bo_unreserve (robj);break ; +case AMDGPU_GEM_OP_GET_MAPPING_INFO : +amdgpu_bo_unreserve (robj); + 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
@@ -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
@@ -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_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0)
Signature · 1 line
2.34.1
Re: [PATCH v12 2/3] drm/amdgpu: Add mapping info option for GEM_OP ioctl
Quoted 16 lines · David Francis
Add new
GEM_OP_IOCTLoptionGET_MAPPING_INFO, which returns a list of mappings associated with a given bo, along with their positions and offsets.Signed-off-by David Francis <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 86 ++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 ++ include/uapi/drm/amdgpu_drm.h | 18 ++++++ 3 files changed, 108 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, · line 888
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +1
-1@@ -885,7 +885,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, goto error; } - r =amdgpu_vm_lock_pd (&fpriv->vm, &exec,2 ); + r =amdgpu_vm_lock_pd (&fpriv->vm, &exec,0 );
Why is this new hunk related? Is it becase I asked what are the two reserved slots in amdgpu_gem_list_mappings() for? But that one remained so it is all unclear to me.
Quoted 142 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +84
-0drm_exec_retry_on_contention (&exec);if (unlikely (r)) goto error;@@ -956,6 +956,86 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,return r; } +/** +* amdgpu_gem_list_mappings - 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) +return -ENOMEM ; + +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 ; + } + +if (num_mappings >0 && num_mappings <= args->num_entries) + ret =copy_to_user (u64_to_user_ptr (args->value), vm_buckets, num_mappings *sizeof (*vm_buckets)); + + args->num_entries = num_mappings; + +unlock_exec: +drm_exec_fini (&exec); +kvfree (vm_buckets); + +return ret; +} +int amdgpu_gem_op_ioctl (struct drm_device *dev,void *data,struct drm_file *filp) {@@ -1022,6 +1102,10 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,amdgpu_bo_unreserve (robj);break ; +case AMDGPU_GEM_OP_GET_MAPPING_INFO : +amdgpu_bo_unreserve (robj); + 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@@ -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 +15
-0@@ -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 {
include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_gem_op { · line 831
Quoted 6 lines
include/uapi/drm/amdgpu_drm.h +3
-0@@ -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;
I suppose uapi kerneldoc should explain the value field is also used for AMDGPU_GEM_OP_GET_MAPPING_INFO and for what.
Alternatively union could be used to self document it but not sure if that is acceptable for amdgpu_drm.h? Like:
struct drm_amdgpu_gem_op {
/** GEM object handle */
__u32 handle;
/** AMDGPU_GEM_OP_* */
__u32 op;
union {
/** Input or return value */
__u64 value;
/** AMDGPU_GEM_OP_GET_MAPPING_INFO */
struct {
__u64 entries; /** Pointer to struct drm_amdgpu_gem_vm_bucket array
(out) */
__u64 num_entries; /** Number of mappings (in/out) */
};
};
};Just a thought.
Regards,
Tvrtko
Quoted 3 lines
include/uapi/drm/amdgpu_drm.h +0
-0};#define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0)
Christian König
maintainerRe: [PATCH v12 2/3] drm/amdgpu: Add mapping info option for GEM_OP ioctl
Quoted 13 lines · Tvrtko Ursulin
Quoted 11 lines · David Francis
Add new
GEM_OP_IOCTLoptionGET_MAPPING_INFO, which returns a list of mappings associated with a given bo, along with their positions and offsets.Signed-off-by David Francis <[email protected]>---
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c| 86 ++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 ++ include/uapi/drm/amdgpu_drm.h | 18 ++++++3 files changed, 108 insertions(+), 1 deletion(-)
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, · line 888
Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +1
-0@@ -885,7 +885,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, goto error; } - r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2); + r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 0);Why is this new hunk related? Is it becase I asked what are the two reserved slots in
amdgpu_gem_list_mappings()for? But that one remained so it is all unclear to me.
Yeah that looks like a bug to me. The change was most likely applied to the wrong code.
Quoted 166 lines
Quoted 138 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +0
-0drm_exec_retry_on_contention(&exec); if (unlikely(r)) goto error;@@ -956,6 +956,86 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return r; }+/**
+ *amdgpu_gem_list_mappings- get information about a buffer's mappings
+ *
+ * @gobj: gem object
+ * @args:gem_oparguments
+ * @fpriv: drm file pointer
+ *
+ *num_entriesis set as an input to the size of the user-allocated array of
+ *drm_amdgpu_gem_vm_bucketstored atargs->value.
+ *num_entriesis 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.
+ */drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +64
-0+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) + return -ENOMEM; + + 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; + } + + if (num_mappings > 0 && num_mappings <= args->num_entries) + ret = copy_to_user(u64_to_user_ptr(args->value), vm_buckets, num_mappings * sizeof(*vm_buckets)); + + args->num_entries = num_mappings; + +unlock_exec: + drm_exec_fini(&exec); + kvfree(vm_buckets); + + return ret; +} +int
amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
{drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +4
-0@@ -1022,6 +1102,10 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, amdgpu_bo_unreserve(robj); break; + case AMDGPU_GEM_OP_GET_MAPPING_INFO: + amdgpu_bo_unreserve(robj); + 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 +4
-0@@ -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) + #endifinclude/uapi/drm/amdgpu_drm.h +15
-0@@ -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 {include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_gem_op { · line 831
Quoted 6 lines
include/uapi/drm/amdgpu_drm.h +3
-0@@ -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;I suppose uapi kerneldoc should explain the value field is also used for
AMDGPU_GEM_OP_GET_MAPPING_INFOand for what.Alternatively union could be used to self document it but not sure if that is acceptable for
amdgpu_drm.h? Like:struct
drm_amdgpu_gem_op{
/** GEM object handle */
__u32 handle;
/** AMDGPU_GEM_OP_* */
__u32 op;
union {
/** Input or return value */
__u64 value;/**
AMDGPU_GEM_OP_GET_MAPPING_INFO*/
struct {
__u64entries; /** Pointer to structdrm_amdgpu_gem_vm_bucketarray (out) */__u64num_entries; /** Number of mappings (in/out) */
};
};
};
Quoted 1 line
Just a thought.
Na, that looks quite ugly I think.
Extending the structure at the end is perfectly supported, but with the union you need to be really careful not to mess things up.
Regards,
Christian.
Quoted 8 lines
Regards,
Tvrtko
Quoted 2 lines
include/uapi/drm/amdgpu_drm.h +0
-0}; #define AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT (1 << 0)
PATCH v12 3/3 drm/amdgpu: Allow kfd CRIU with no buffer objects
In reply to David FrancisThe 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
@@ -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