drm/xe/xe_vm: Implement xe_vm_get_property_ioctl
Subsystems (3)
v21 cover drm/xe/xe_vm: Implement xe_vm_get_property_ioctl
Add additional information to each VM so they can report up to the first 50 seen faults. Only pagefaults are saved this way currently, though in the future, all faults should be tracked by the VM for future reporting.
Additionally, of the pagefaults reported, only failed pagefaults are saved this way, as successful pagefaults should recover silently and not need to be reported to userspace.
To allow userspace to access these faults, a new ioctl - xe_vm_get_property_ioct - was created.
v2: (Matt Brost)
- Break full ban list request into a separate property.
- Reformat drm_xe_vm_get_property struct.
- Remove need for drm_xe_faults helper struct.
- Separate data pointer and scalar return value in ioctl.
- Get address type on pagefault report and save it to the pagefault.
- Correctly reject writes to read-only VMAs.
- Miscellaneous formatting fixes.
v3: (Matt Brost)
- Only allow querying of failed pagefaults
v4:
- Remove unnecessary size parameter from helper function, as it is a property of the arguments. (jcavitt)
- Remove unnecessary copy_from_user (Jainxun)
- Set address_precision to 1 (Jainxun)
- Report max size instead of dynamic size for memory allocation purposes. Total memory usage is reported separately.
v5:
- Return int from xe_vm_get_property_size (Shuicheng)
- Fix memory leak (Shuicheng)
- Remove unnecessary size variable (jcavitt)
v6:
- Free vm after use (Shuicheng)
- Compress pf copy logic (Shuicheng)
- Update fault_unsuccessful before storing (Shuicheng)
- Fix old struct name in comments (Shuicheng)
- Keep first 50 pagefaults instead of last 50 (Jianxun)
- Rename ioctl to xe_vm_get_faults_ioctl (jcavitt)
v7:
- Avoid unnecessary execution by checking MAX_PFS earlier (jcavitt)
- Fix double-locking error (jcavitt)
- Assert kmemdump is successful (Shuicheng)
- Repair and move fill_faults break condition (Dan Carpenter)
- Free vm after use (jcavitt)
- Combine assertions (jcavitt)
- Expand size check in xe_vm_get_faults_ioctl (jcavitt)
- Remove return mask from fill_faults, as return is already -EFAULT or 0 (jcavitt)
v8:
- Revert back to using drm_xe_vm_get_property_ioctl
- s/Migrate/Move (Michal)
- s/xe_pagefault/xe_gt_pagefault (Michal)
- Create new header file, xe_gt_pagefault_types.h (Michal)
- Add and fix kernel docs (Michal)
- Rename xe_vm.pfs to xe_vm.faults (jcavitt)
- Store fault data and not pagefault in xe_vm faults list (jcavitt)
- Store address, address type, and address precision per fault (jcavitt)
- Store engine class and instance data per fault (Jianxun)
- Properly handle kzalloc error (Michal W)
- s/MAX_PFS/MAX_FAULTS_SAVED_PER_VM (Michal W)
- Store fault level per fault (Micahl M)
- Apply better copy_to_user logic (jcavitt)
v9:
- More kernel doc fixes (Michal W, Jianxun)
- Better error handling (jcavitt)
v10:
- Convert enums to defines in regs folder (Michal W)
- Move xe_guc_pagefault_desc to regs folder (Michal W)
- Future-proof size logic for zero-size properties (jcavitt)
- Replace address type extern with access type (Jianxun)
- Add fault type to xe_drm_fault (Jianxun)
v11:
- Remove unnecessary switch case logic (Raag)
- Compress size get, size validation, and property fill functions into a single helper function (jcavitt)
- Assert valid size (jcavitt)
- Store pagefaults in non-fault-mode VMs as well (Jianxun)
v12:
- Remove unnecessary else condition
- Correct backwards helper function size logic (jcavitt)
- Fix kernel docs and comments (Michal W)
v13:
- Move xe and user engine class mapping arrays to header (John H)
v14:
- Fix double locking issue (Jianxun)
- Use size_t instead of int (Raag)
- Remove unnecessary includes (jcavitt)
v15:
- Do not report faults from reserved engines (Jianxun)
v16:
- Remove engine class and instance (Ivan)
v17:
- Map access type, fault type, and fault level to user macros (Matt Brost, Ivan)
v18:
- Add uAPI merge request to this cover letter
v19:
- Perform kzalloc outside of lock (Auld)
v20:
- Fix inconsistent use of whitespace in defines
v21:
- Remove unnecessary size assertion (jcavitt)
uAPI: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32987
Jonathan Cavitt (5):
drm/xe/xe_gt_pagefault: Disallow writes to read-only VMAs drm/xe/xe_gt_pagefault: Move pagefault struct to header drm/xe/uapi: Define drm_xe_vm_get_property drm/xe/xe_vm: Add per VM fault info drm/xe/xe_vm: Implement xe_vm_get_property_ioctl
drivers/gpu/drm/xe/regs/xe_pagefault_desc.h | 49 +++++ drivers/gpu/drm/xe/xe_device.c | 3 + drivers/gpu/drm/xe/xe_gt_pagefault.c | 72 ++++---- drivers/gpu/drm/xe/xe_gt_pagefault_types.h | 42 +++++ drivers/gpu/drm/xe/xe_guc_fwif.h | 28 --- drivers/gpu/drm/xe/xe_vm.c | 195 ++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 11 ++ drivers/gpu/drm/xe/xe_vm_types.h | 28 +++ include/uapi/drm/xe_drm.h | 86 +++++++++ 9 files changed, 453 insertions(+), 61 deletions(-) create mode 100644 drivers/gpu/drm/xe/regs/xe_pagefault_desc.h create mode 100644 drivers/gpu/drm/xe/xe_gt_pagefault_types.h
Signature · 1 line
2.43.0
PATCH v21 1/5 drm/xe/xe_gt_pagefault: Disallow writes to read-only VMAs
The page fault handler should reject write/atomic access to read only VMAs. Add code to handle this in handle_pagefault after the VMA lookup.
--- drivers/gpu/drm/xe/xe_gt_pagefault.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/gpu/drm/xe/xe_gt_pagefault.c +5 -0
@@ -237,6 +237,11 @@ static int handle_pagefault(struct xe_gt *gt, struct pagefault *pf) goto unlock_vm; } +if (xe_vma_read_only (vma) && pf->access_type !=ACCESS_TYPE_READ ) { + err = -EPERM ; + goto unlock_vm; + } + atomic =access_is_atomic (pf->access_type);if (xe_vma_is_cpu_addr_mirror (vma))
Signature · 1 line
2.43.0
PATCH v21 2/5 drm/xe/xe_gt_pagefault: Move pagefault struct to header
In reply to Jonathan CavittMove the pagefault struct from xe_gt_pagefault.c to the xe_gt_pagefault_types.h header file, and move the associated enum values into the regs folder under xe_pagefault_desc.h
Since xe_pagefault_desc.h is being initialized here, also move the xe_guc_pagefault_desc hardware formats to the new file.
v2:
- Normalize names for common header (Matt Brost)
v3:
- s/Migrate/Move (Michal W)
- s/xe_pagefault/xe_gt_pagefault (Michal W)
- Create new header file, xe_gt_pagefault_types.h (Michal W)
- Add kernel docs (Michal W)
v4:
- Fix includes usage (Michal W)
- Reference Bspec (Michal W)
v5:
- Convert enums to defines in regs folder (Michal W)
- Move xe_guc_pagefault_desc to regs folder (Michal W)
Bspec: 77412
--- drivers/gpu/drm/xe/regs/xe_pagefault_desc.h | 49 +++++++++++++++++++++ drivers/gpu/drm/xe/xe_gt_pagefault.c | 43 ++++-------------- drivers/gpu/drm/xe/xe_gt_pagefault_types.h | 42 ++++++++++++++++++ drivers/gpu/drm/xe/xe_guc_fwif.h | 28 ------------ 4 files changed, 100 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/xe/regs/xe_pagefault_desc.h create mode 100644 drivers/gpu/drm/xe/xe_gt_pagefault_types.h
drivers/gpu/drm/xe/regs/xe_pagefault_desc.h new file +49 -0
@@ -0,0 +1,49 @@ +/* SPDX-License-Identifier: MIT */ +/* +* Copyright © 2025 Intel Corporation +*/ + +#ifndef _XE_PAGEFAULT_DESC_H_ +#define _XE_PAGEFAULT_DESC_H_ + +#include <linux/bits.h> +#include <linux/types.h> + +struct xe_guc_pagefault_desc { +u32 dw0; +#define PFD_FAULT_LEVEL GENMASK(2, 0) +#define PFD_SRC_ID GENMASK(10, 3) +#define PFD_RSVD_0 GENMASK(17, 11) +#define XE2_PFD_TRVA_FAULT BIT(18) +#define PFD_ENG_INSTANCE GENMASK(24, 19) +#define PFD_ENG_CLASS GENMASK(27, 25) +#define PFD_PDATA_LO GENMASK(31, 28) + +u32 dw1; +#define PFD_PDATA_HI GENMASK(11, 0) +#define PFD_PDATA_HI_SHIFT 4 +#define PFD_ASID GENMASK(31, 12) + +u32 dw2; +#define PFD_ACCESS_TYPE GENMASK(1, 0) +#define PFD_FAULT_TYPE GENMASK(3, 2) +#define PFD_VFID GENMASK(9, 4) +#define PFD_RSVD_1 GENMASK(11, 10) +#define PFD_VIRTUAL_ADDR_LO GENMASK(31, 12) +#define PFD_VIRTUAL_ADDR_LO_SHIFT 12 + +u32 dw3; +#define PFD_VIRTUAL_ADDR_HI GENMASK(31, 0) +#define PFD_VIRTUAL_ADDR_HI_SHIFT 32 +} __packed; + +#define FLT_ACCESS_TYPE_READ 0u +#define FLT_ACCESS_TYPE_WRITE 1u +#define FLT_ACCESS_TYPE_ATOMIC 2u +#define FLT_ACCESS_TYPE_RESERVED 3u + +#define FLT_TYPE_NOT_PRESENT_FAULT 0u +#define FLT_TYPE_WRITE_ACCESS_VIOLATION 1u +#define FLT_TYPE_ATOMIC_ACCESS_VIOLATION 2u + +#endif
drivers/gpu/drm/xe/xe_gt_pagefault.c +9 -34
94 lines not on this page · raw message
drivers/gpu/drm/xe/xe_gt_pagefault_types.h new file +42 -0
42 lines not on this page · raw message
drivers/gpu/drm/xe/xe_guc_fwif.h +0 -28
@@ -290,34 +290,6 @@ enum xe_guc_response_desc_type {FAULT_RESPONSE_DESC }; -struct xe_guc_pagefault_desc { -u32 dw0; -#define PFD_FAULT_LEVEL GENMASK(2, 0) -#define PFD_SRC_ID GENMASK(10, 3) -#define PFD_RSVD_0 GENMASK(17, 11) -#define XE2_PFD_TRVA_FAULT BIT(18) -#define PFD_ENG_INSTANCE GENMASK(24, 19) -#define PFD_ENG_CLASS GENMASK(27, 25) -#define PFD_PDATA_LO GENMASK(31, 28) - -u32 dw1; -#define PFD_PDATA_HI GENMASK(11, 0) -#define PFD_PDATA_HI_SHIFT 4 -#define PFD_ASID GENMASK(31, 12) - -u32 dw2; -#define PFD_ACCESS_TYPE GENMASK(1, 0) -#define PFD_FAULT_TYPE GENMASK(3, 2) -#define PFD_VFID GENMASK(9, 4) -#define PFD_RSVD_1 GENMASK(11, 10) -#define PFD_VIRTUAL_ADDR_LO GENMASK(31, 12) -#define PFD_VIRTUAL_ADDR_LO_SHIFT 12 - -u32 dw3; -#define PFD_VIRTUAL_ADDR_HI GENMASK(31, 0) -#define PFD_VIRTUAL_ADDR_HI_SHIFT 32 -} __packed; -struct xe_guc_pagefault_reply {u32 dw0;#define PFR_VALID BIT(0)
Signature · 1 line
2.43.0
PATCH v21 3/5 drm/xe/uapi: Define drm_xe_vm_get_property
In reply to Jonathan CavittAdd initial declarations for the drm_xe_vm_get_property ioctl.
v2:
- Expand kernel docs for drm_xe_vm_get_property (Jianxun)
v3:
- Remove address type external definitions (Jianxun)
- Add fault type to xe_drm_fault struct (Jianxun)
v4:
- Remove engine class and instance (Ivan)
v5:
- Add declares for fault type, access type, and fault level (Matt Brost, Ivan)
v6:
- Fix inconsistent use of whitespace in defines
--- include/uapi/drm/xe_drm.h | 86 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)
include/uapi/drm/xe_drm.h +86 -0
@@ -81,6 +81,7 @@ extern "C" {* - &DRM_IOCTL_XE_EXEC * - &DRM_IOCTL_XE_WAIT_USER_FENCE * - &DRM_IOCTL_XE_OBSERVATION +* - &DRM_IOCTL_XE_VM_GET_PROPERTY */ /*
@@ -102,6 +103,7 @@ extern "C" {#define DRM_XE_EXEC 0x09#define DRM_XE_WAIT_USER_FENCE 0x0a#define DRM_XE_OBSERVATION 0x0b +#define DRM_XE_VM_GET_PROPERTY 0x0c/* Must be kept compact -- no holes */
@@ -117,6 +119,7 @@ extern "C" {#define DRM_IOCTL_XE_EXEC DRM_IOW(DRM_COMMAND_BASE + DRM_XE_EXEC, struct drm_xe_exec)#define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_WAIT_USER_FENCE, struct drm_xe_wait_user_fence)#define DRM_IOCTL_XE_OBSERVATION DRM_IOW(DRM_COMMAND_BASE + DRM_XE_OBSERVATION, struct drm_xe_observation_param) +#define DRM_IOCTL_XE_VM_GET_PROPERTY DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_VM_GET_PROPERTY, struct drm_xe_vm_get_property)/** * DOC: Xe IOCTL Extensions
@@ -1193,6 +1196,89 @@ struct drm_xe_vm_bind {__u64 reserved[2 ]; }; +/** struct xe_vm_fault - Describes faults for %DRM_XE_VM_GET_PROPERTY_FAULTS */ +struct xe_vm_fault { +/** @address: Address of the fault */ +__u64 address; +/** @address_precision: Precision of faulted address */ +__u32 address_precision; +/** @access_type: Type of address access that resulted in fault */ +#define FAULT_ACCESS_TYPE_READ 0 +#define FAULT_ACCESS_TYPE_WRITE 1 +#define FAULT_ACCESS_TYPE_ATOMIC 2 +__u8 access_type; +/** @fault_type: Type of fault reported */ +#define FAULT_TYPE_NOT_PRESENT 0 +#define FAULT_TYPE_WRITE_ACCESS 1 +#define FAULT_TYPE_ATOMIC_ACCESS 2 +__u8 fault_type; +/** @fault_level: fault level of the fault */ +#define FAULT_LEVEL_PTE 0 +#define FAULT_LEVEL_PDE 1 +#define FAULT_LEVEL_PDP 2 +#define FAULT_LEVEL_PML4 3 +#define FAULT_LEVEL_PML5 4 +__u8 fault_level; +/** @pad: MBZ */ +__u8 pad; +/** @reserved: MBZ */ +__u64 reserved[4 ]; +}; + +/** +* struct drm_xe_vm_get_property - Input of &DRM_IOCTL_XE_VM_GET_PROPERTY +* +* The user provides a VM and a property to query among DRM_XE_VM_GET_PROPERTY_*, +* and sets the values in the vm_id and property members, respectively. This +* determines both the VM to get the property of, as well as the property to +* report. +* +* If size is set to 0, the driver fills it with the required size for the +* requested property. The user is expected here to allocate memory for the +* property structure and to provide a pointer to the allocated memory using the +* data member. For some properties, this may be zero, in which case, the +* value of the property will be saved to the value member and size will remain +* zero on return. +* +* If size is not zero, then the IOCTL will attempt to copy the requested +* property into the data member. +* +* The IOCTL will return -ENOENT if the VM could not be identified from the +* provided VM ID, or -EINVAL if the IOCTL fails for any other reason, such as +* providing an invalid size for the given property or if the property data +* could not be copied to the memory allocated to the data member. +* +* The property member can be: +* - %DRM_XE_VM_GET_PROPERTY_FAULTS +*/ +struct drm_xe_vm_get_property { +/** @extensions: Pointer to the first extension struct, if any */ +__u64 extensions; + +/** @vm_id: The ID of the VM to query the properties of */ +__u32 vm_id; + +#define DRM_XE_VM_GET_PROPERTY_FAULTS 0 +/** @property: property to get */ +__u32 property; + +/** @size: Size to allocate for @data */ +__u32 size; + +/** @pad: MBZ */ +__u32 pad; + +union { +/** @data: Pointer to user-defined array of flexible size and type */ +__u64 data; +/** @value: Return value for scalar queries */ +__u64 value; + }; + +/** @reserved: MBZ */ +__u64 reserved[3 ]; +}; +/** * struct drm_xe_exec_queue_create - Input of &DRM_IOCTL_XE_EXEC_QUEUE_CREATE *
Signature · 1 line
2.43.0
RE: [PATCH v21 3/5] drm/xe/uapi: Define drm_xe_vm_get_property
Quoted 68 lines · Jonathan
Add initial declarations for the
drm_xe_vm_get_propertyioctl.v2:
- Expand kernel docs fordrm_xe_vm_get_property(Jianxun)v3:
- Remove address type external definitions (Jianxun)
- Add fault type toxe_drm_faultstruct (Jianxun)v4:
- Remove engine class and instance (Ivan)v5:
- Add declares for fault type, access type, and fault level (Matt Brost, Ivan)v6:
- Fix inconsistent use of whitespace in definesSigned-off-by Jonathan Cavitt <[email protected]>Cc Zhang Jianxun <[email protected]>Cc Ivan Briano <[email protected]>Cc Matthew Brost <[email protected]>--- include/uapi/drm/xe_drm.h | 86 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)include/uapi/drm/xe_drm.h index renamed from include/uapi/drm/xe_drm.h +0
-0
9c08738c3b91..556fc360a076100644include/uapi/drm/xe_drm.h +2
-0@@ -81,6 +81,7 @@ extern "C" {* - &DRM_IOCTL_XE_EXEC * - &DRM_IOCTL_XE_WAIT_USER_FENCE * - &DRM_IOCTL_XE_OBSERVATION +* - &DRM_IOCTL_XE_VM_GET_PROPERTY */ /* @@ -102,6 +103,7 @@ extern "C" {#define DRM_XE_EXEC 0x09#define DRM_XE_WAIT_USER_FENCE 0x0a#define DRM_XE_OBSERVATION 0x0b +#define DRM_XE_VM_GET_PROPERTY 0x0c/* Must be kept compact -- no holes */ @@ -117,6 +119,7 @@ extern "C" {#define DRM_IOCTL_XE_EXEC DRM_IOW(DRM_COMMAND_BASE + DRM_XE_EXEC, struct drm_xe_exec)#define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR (DRM_COMMAND_BASE +DRM_XE_WAIT_USER_FENCE , struct drm_xe_wait_user_fence)#define DRM_IOCTL_XE_OBSERVATION DRM_IOW(DRM_COMMAND_BASE +
DRM_XE_OBSERVATION, structdrm_xe_observation_param)
+#defineDRM_IOCTL_XE_VM_GET_PROPERTY
DRM_IOWR(DRM_COMMAND_BASE +DRM_XE_VM_GET_PROPERTY,
structdrm_xe_vm_get_property)/**
* DOC: Xe IOCTL Extensionsinclude/uapi/drm/xe_drm.h +0
-0@@ -1193,6 +1196,89 @@ struct drm_xe_vm_bind {__u64 reserved[2 ]; };
include/uapi/drm/xe_drm.h struct drm_xe_vm_bind { · line 1208
Quoted 10 lines
include/uapi/drm/xe_drm.h +10
-0+/** struct xe_vm_fault - Describes faults for +%DRM_XE_VM_GET_PROPERTY_FAULTS */ struct xe_vm_fault { +/** @address: Address of the fault */ +__u64 address; +/** @address_precision: Precision of faulted address */ +__u32 address_precision; +/** @access_type: Type of address access that resulted in fault */ +#define FAULT_ACCESS_TYPE_READ 0 +#define FAULT_ACCESS_TYPE_WRITE 1 +#define FAULT_ACCESS_TYPE_ATOMIC 2
There is the same definition of "FLT_ACCESS_TYPE_READ" in the "regs/xe_pagefault_desc.h" Could we remove the definition in xe_pagefault_desc.h, and change to this definition? Also for the access_type in below.
Shuicheng
Quoted 93 lines
include/uapi/drm/xe_drm.h +77
-0+__u8 access_type; +/** @fault_type: Type of fault reported */ +#define FAULT_TYPE_NOT_PRESENT 0 +#define FAULT_TYPE_WRITE_ACCESS 1 +#define FAULT_TYPE_ATOMIC_ACCESS 2 +__u8 fault_type; +/** @fault_level: fault level of the fault */ +#define FAULT_LEVEL_PTE 0 +#define FAULT_LEVEL_PDE 1 +#define FAULT_LEVEL_PDP 2 +#define FAULT_LEVEL_PML4 3 +#define FAULT_LEVEL_PML5 4 +__u8 fault_level; +/** @pad: MBZ */ +__u8 pad; +/** @reserved: MBZ */ +__u64 reserved[4 ]; +}; + +/** +* struct drm_xe_vm_get_property - Input of +&DRM_IOCTL_XE_VM_GET_PROPERTY +* +* The user provides a VM and a property to query among +DRM_XE_VM_GET_PROPERTY_*, +* and sets the values in the vm_id and property members, respectively. +This +* determines both the VM to get the property of, as well as the +property to +* report. +* +* If size is set to 0, the driver fills it with the required size for +the +* requested property. The user is expected here to allocate memory +for the +* property structure and to provide a pointer to the allocated memory +using the +* data member. For some properties, this may be zero, in which case, +the +* value of the property will be saved to the value member and size +will remain +* zero on return. +* +* If size is not zero, then the IOCTL will attempt to copy the +requested +* property into the data member. +* +* The IOCTL will return -ENOENT if the VM could not be identified from +the +* provided VM ID, or -EINVAL if the IOCTL fails for any other reason, +such as +* providing an invalid size for the given property or if the property +data +* could not be copied to the memory allocated to the data member. +* +* The property member can be: +* - %DRM_XE_VM_GET_PROPERTY_FAULTS +*/ +struct drm_xe_vm_get_property { +/** @extensions: Pointer to the first extension struct, if any */ +__u64 extensions; + +/** @vm_id: The ID of the VM to query the properties of */ +__u32 vm_id; + +#define DRM_XE_VM_GET_PROPERTY_FAULTS 0 +/** @property: property to get */ +__u32 property; + +/** @size: Size to allocate for @data */ +__u32 size; + +/** @pad: MBZ */ +__u32 pad; + +union { +/** @data: Pointer to user-defined array of flexible size and type */
include/uapi/drm/xe_drm.h +9
-0+__u64 data; +/** @value: Return value for scalar queries */ +__u64 value; + }; + +/** @reserved: MBZ */ +__u64 reserved[3 ]; +}; +/** * struct drm_xe_exec_queue_create - Input of
&DRM_IOCTL_XE_EXEC_QUEUE_CREATE
*
--
2.43.0
Matthew Brost
maintainerRe: [PATCH v21 3/5] drm/xe/uapi: Define drm_xe_vm_get_property
Quoted 69 lines · Shuicheng
Quoted 68 lines · Jonathan
Add initial declarations for the
drm_xe_vm_get_propertyioctl.v2:
- Expand kernel docs fordrm_xe_vm_get_property(Jianxun)v3:
- Remove address type external definitions (Jianxun)
- Add fault type toxe_drm_faultstruct (Jianxun)v4:
- Remove engine class and instance (Ivan)v5:
- Add declares for fault type, access type, and fault level (Matt Brost, Ivan)v6:
- Fix inconsistent use of whitespace in definesSigned-off-by Jonathan Cavitt <[email protected]>Cc Zhang Jianxun <[email protected]>Cc Ivan Briano <[email protected]>Cc Matthew Brost <[email protected]>--- include/uapi/drm/xe_drm.h | 86 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)include/uapi/drm/xe_drm.h index renamed from include/uapi/drm/xe_drm.h +0
-0
9c08738c3b91..556fc360a076100644include/uapi/drm/xe_drm.h +2
-0@@ -81,6 +81,7 @@ extern "C" {* - &DRM_IOCTL_XE_EXEC * - &DRM_IOCTL_XE_WAIT_USER_FENCE * - &DRM_IOCTL_XE_OBSERVATION +* - &DRM_IOCTL_XE_VM_GET_PROPERTY */ /* @@ -102,6 +103,7 @@ extern "C" {#define DRM_XE_EXEC 0x09#define DRM_XE_WAIT_USER_FENCE 0x0a#define DRM_XE_OBSERVATION 0x0b +#define DRM_XE_VM_GET_PROPERTY 0x0c/* Must be kept compact -- no holes */ @@ -117,6 +119,7 @@ extern "C" {#define DRM_IOCTL_XE_EXEC DRM_IOW(DRM_COMMAND_BASE + DRM_XE_EXEC, struct drm_xe_exec)#define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR (DRM_COMMAND_BASE +DRM_XE_WAIT_USER_FENCE , struct drm_xe_wait_user_fence)#define DRM_IOCTL_XE_OBSERVATION DRM_IOW(DRM_COMMAND_BASE +
DRM_XE_OBSERVATION, structdrm_xe_observation_param)
+#defineDRM_IOCTL_XE_VM_GET_PROPERTY
DRM_IOWR(DRM_COMMAND_BASE +DRM_XE_VM_GET_PROPERTY,
structdrm_xe_vm_get_property)/**
* DOC: Xe IOCTL Extensionsinclude/uapi/drm/xe_drm.h +0
-0@@ -1193,6 +1196,89 @@ struct drm_xe_vm_bind {__u64 reserved[2 ]; };
Quoted 15 lines
include/uapi/drm/xe_drm.h struct drm_xe_vm_bind { · line 1208
Quoted 10 lines
include/uapi/drm/xe_drm.h +10
-0+/** struct xe_vm_fault - Describes faults for +%DRM_XE_VM_GET_PROPERTY_FAULTS */ struct xe_vm_fault { +/** @address: Address of the fault */ +__u64 address; +/** @address_precision: Precision of faulted address */ +__u32 address_precision; +/** @access_type: Type of address access that resulted in fault */ +#define FAULT_ACCESS_TYPE_READ 0 +#define FAULT_ACCESS_TYPE_WRITE 1 +#define FAULT_ACCESS_TYPE_ATOMIC 2There is the same definition of "
FLT_ACCESS_TYPE_READ" in the "regs/xe_pagefault_desc.h" Could we remove the definition inxe_pagefault_desc.h, and change to this definition? Also for theaccess_typein below.
No — I'm getting a little frustrated here.
I've explained this multiple times, both in the comments on the list and in our internal Teams chat. The uAPI is an abstraction of the hardware — it is not a direct interface to hardware-reported values in user space.
The reason for this is simple: if the underlying hardware changes, a direct interface would break the uAPI. Right now, the values may happen to match, but there’s no guarantee they will continue to do so in the future.
Please, everyone, stop bringing this up.
Matt
Quoted 96 lines
Shuicheng
Quoted 93 lines
include/uapi/drm/xe_drm.h +77
-0+__u8 access_type; +/** @fault_type: Type of fault reported */ +#define FAULT_TYPE_NOT_PRESENT 0 +#define FAULT_TYPE_WRITE_ACCESS 1 +#define FAULT_TYPE_ATOMIC_ACCESS 2 +__u8 fault_type; +/** @fault_level: fault level of the fault */ +#define FAULT_LEVEL_PTE 0 +#define FAULT_LEVEL_PDE 1 +#define FAULT_LEVEL_PDP 2 +#define FAULT_LEVEL_PML4 3 +#define FAULT_LEVEL_PML5 4 +__u8 fault_level; +/** @pad: MBZ */ +__u8 pad; +/** @reserved: MBZ */ +__u64 reserved[4 ]; +}; + +/** +* struct drm_xe_vm_get_property - Input of +&DRM_IOCTL_XE_VM_GET_PROPERTY +* +* The user provides a VM and a property to query among +DRM_XE_VM_GET_PROPERTY_*, +* and sets the values in the vm_id and property members, respectively. +This +* determines both the VM to get the property of, as well as the +property to +* report. +* +* If size is set to 0, the driver fills it with the required size for +the +* requested property. The user is expected here to allocate memory +for the +* property structure and to provide a pointer to the allocated memory +using the +* data member. For some properties, this may be zero, in which case, +the +* value of the property will be saved to the value member and size +will remain +* zero on return. +* +* If size is not zero, then the IOCTL will attempt to copy the +requested +* property into the data member. +* +* The IOCTL will return -ENOENT if the VM could not be identified from +the +* provided VM ID, or -EINVAL if the IOCTL fails for any other reason, +such as +* providing an invalid size for the given property or if the property +data +* could not be copied to the memory allocated to the data member. +* +* The property member can be: +* - %DRM_XE_VM_GET_PROPERTY_FAULTS +*/ +struct drm_xe_vm_get_property { +/** @extensions: Pointer to the first extension struct, if any */ +__u64 extensions; + +/** @vm_id: The ID of the VM to query the properties of */ +__u32 vm_id; + +#define DRM_XE_VM_GET_PROPERTY_FAULTS 0 +/** @property: property to get */ +__u32 property; + +/** @size: Size to allocate for @data */ +__u32 size; + +/** @pad: MBZ */ +__u32 pad; + +union { +/** @data: Pointer to user-defined array of flexible size and type */
include/uapi/drm/xe_drm.h +9
-0+__u64 data; +/** @value: Return value for scalar queries */ +__u64 value; + }; + +/** @reserved: MBZ */ +__u64 reserved[3 ]; +}; +/** * struct drm_xe_exec_queue_create - Input of
&DRM_IOCTL_XE_EXEC_QUEUE_CREATE
*
--
2.43.0
Lin, Shuicheng
Reviewed-byRE: [PATCH v21 3/5] drm/xe/uapi: Define drm_xe_vm_get_property
Quoted 18 lines · Matthew
Quoted 17 lines · Shuicheng
Quoted 16 lines · Jonathan
Add initial declarations for the
drm_xe_vm_get_propertyioctl.v2:
- Expand kernel docs fordrm_xe_vm_get_property(Jianxun)v3:
- Remove address type external definitions (Jianxun)
- Add fault type toxe_drm_faultstruct (Jianxun)v4:
- Remove engine class and instance (Ivan)v5:
- Add declares for fault type, access type, and fault level (Matt Brost, Ivan)
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
v6:
- Fix inconsistent use of whitespace in definesSigned-off-by Jonathan Cavitt <[email protected]>Cc Zhang Jianxun <[email protected]>Cc Ivan Briano <[email protected]>Cc Matthew Brost <[email protected]>---
LGTM.
Quoted 70 lines
Quoted 23 lines
Quoted 23 lines
include/uapi/drm/xe_drm.h | 86 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)include/uapi/drm/xe_drm.h +0
-0index
9c08738c3b91..556fc360a076100644include/uapi/drm/xe_drm.h +2
-0@@ -81,6 +81,7 @@ extern "C" {* - &DRM_IOCTL_XE_EXEC * - &DRM_IOCTL_XE_WAIT_USER_FENCE * - &DRM_IOCTL_XE_OBSERVATION +* - &DRM_IOCTL_XE_VM_GET_PROPERTY */ /* @@ -102,6 +103,7 @@ extern "C" {#define DRM_XE_EXEC 0x09#define DRM_XE_WAIT_USER_FENCE 0x0a#define DRM_XE_OBSERVATION 0x0b +#define DRM_XE_VM_GET_PROPERTY 0x0cQuoted 8 lines
include/uapi/drm/xe_drm.h extern "C" { · line 123
Quoted 8 lines
include/uapi/drm/xe_drm.h +0
-0/* Must be kept compact -- no holes */ @@ -117,6 +119,7 @@ extern "C" {#define DRM_IOCTL_XE_EXEC DRM_IOW(DRM_COMMAND_BASE + DRM_XE_EXEC, struct drm_xe_exec)#define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_WAIT_USER_FENCE,struct
Quoted 5 lines
Quoted 5 lines
drm_xe_wait_user_fence) #defineDRM_IOCTL_XE_OBSERVATION
DRM_IOW(DRM_COMMAND_BASE +DRM_XE_OBSERVATION, structdrm_xe_observation_param)
+#defineDRM_IOCTL_XE_VM_GET_PROPERTY
DRM_IOWR(DRM_COMMAND_BASE +DRM_XE_VM_GET_PROPERTY,struct
Quoted 8 lines
Quoted 8 lines
drm_xe_vm_get_property)/**
* DOC: Xe IOCTL Extensionsinclude/uapi/drm/xe_drm.h +0
-0@@ -1193,6 +1196,89 @@ struct drm_xe_vm_bind {__u64 reserved[2 ]; };Quoted 12 lines
include/uapi/drm/xe_drm.h struct drm_xe_vm_bind { · line 1208
Quoted 10 lines
include/uapi/drm/xe_drm.h +10
-0+/** struct xe_vm_fault - Describes faults for +%DRM_XE_VM_GET_PROPERTY_FAULTS */ struct xe_vm_fault { +/** @address: Address of the fault */ +__u64 address; +/** @address_precision: Precision of faulted address */ +__u32 address_precision; +/** @access_type: Type of address access that resulted in fault */ +#define FAULT_ACCESS_TYPE_READ 0 +#define FAULT_ACCESS_TYPE_WRITE 1 +#define FAULT_ACCESS_TYPE_ATOMIC 2There is the same definition of "
FLT_ACCESS_TYPE_READ" in the"
regs/xe_pagefault_desc.h"Quoted 1 line
Could we remove the definition in
xe_pagefault_desc.h, and change to thisdefinition?
Quoted 2 lines
Also for the
access_typein below.No — I'm getting a little frustrated here.
I've explained this multiple times, both in the comments on the list and in our internal Teams chat. The uAPI is an abstraction of the hardware — it is not a direct interface to hardware-reported values in user space.
Quoted 7 lines
The reason for this is simple: if the underlying hardware changes, a direct interface would break the uAPI. Right now, the values may happen to match, but there’s no guarantee they will continue to do so in the future.
Please, everyone, stop bringing this up.
Matt
OK. Get it. Thanks.
Shuicheng
Quoted 96 lines
Quoted 96 lines
Shuicheng
Quoted 92 lines
include/uapi/drm/xe_drm.h +76
-0+__u8 access_type; +/** @fault_type: Type of fault reported */ +#define FAULT_TYPE_NOT_PRESENT 0 +#define FAULT_TYPE_WRITE_ACCESS 1 +#define FAULT_TYPE_ATOMIC_ACCESS 2 +__u8 fault_type; +/** @fault_level: fault level of the fault */ +#define FAULT_LEVEL_PTE 0 +#define FAULT_LEVEL_PDE 1 +#define FAULT_LEVEL_PDP 2 +#define FAULT_LEVEL_PML4 3 +#define FAULT_LEVEL_PML5 4 +__u8 fault_level; +/** @pad: MBZ */ +__u8 pad; +/** @reserved: MBZ */ +__u64 reserved[4 ]; +}; + +/** +* struct drm_xe_vm_get_property - Input of +&DRM_IOCTL_XE_VM_GET_PROPERTY +* +* The user provides a VM and a property to query among +DRM_XE_VM_GET_PROPERTY_*, +* and sets the values in the vm_id and property members, respectively. +This +* determines both the VM to get the property of, as well as the +property to +* report. +* +* If size is set to 0, the driver fills it with the required size +for the +* requested property. The user is expected here to allocate +memory for the +* property structure and to provide a pointer to the allocated +memory using the +* data member. For some properties, this may be zero, in which +case, the +* value of the property will be saved to the value member and size +will remain +* zero on return. +* +* If size is not zero, then the IOCTL will attempt to copy the +requested +* property into the data member. +* +* The IOCTL will return -ENOENT if the VM could not be identified +from the +* provided VM ID, or -EINVAL if the IOCTL fails for any other +reason, such as +* providing an invalid size for the given property or if the +property data +* could not be copied to the memory allocated to the data member. +* +* The property member can be: +* - %DRM_XE_VM_GET_PROPERTY_FAULTS */ struct +drm_xe_vm_get_property { +/** @extensions: Pointer to the first extension struct, if any */ +__u64 extensions; + +/** @vm_id: The ID of the VM to query the properties of */ +__u32 vm_id; + +#define DRM_XE_VM_GET_PROPERTY_FAULTS 0 +/** @property: property to get */ +__u32 property; + +/** @size: Size to allocate for @data */ +__u32 size; + +/** @pad: MBZ */ +__u32 pad; + +union { +/** @data: Pointer to user-defined array of flexible size and type */
include/uapi/drm/xe_drm.h +9
-0+__u64 data; +/** @value: Return value for scalar queries */ +__u64 value; + }; + +/** @reserved: MBZ */ +__u64 reserved[3 ]; +}; +/** * struct drm_xe_exec_queue_create - Input of
&DRM_IOCTL_XE_EXEC_QUEUE_CREATE
*
--
2.43.0
PATCH v21 4/5 drm/xe/xe_vm: Add per VM fault info
In reply to Jonathan CavittAdd additional information to each VM so they can report up to the first 50 seen faults. Only pagefaults are saved this way currently, though in the future, all faults should be tracked by the VM for future reporting.
Additionally, of the pagefaults reported, only failed pagefaults are saved this way, as successful pagefaults should recover silently and not need to be reported to userspace.
v2:
- Free vm after use (Shuicheng)
- Compress pf copy logic (Shuicheng)
- Update fault_unsuccessful before storing (Shuicheng)
- Fix old struct name in comments (Shuicheng)
- Keep first 50 pagefaults instead of last 50 (Jianxun)
v3:
- Avoid unnecessary execution by checking MAX_PFS earlier (jcavitt)
- Fix double-locking error (jcavitt)
- Assert kmemdump is successful (Shuicheng)
v4:
- Rename xe_vm.pfs to xe_vm.faults (jcavitt)
- Store fault data and not pagefault in xe_vm faults list (jcavitt)
- Store address, address type, and address precision per fault (jcavitt)
- Store engine class and instance data per fault (Jianxun)
- Add and fix kernel docs (Michal W)
- Properly handle kzalloc error (Michal W)
- s/MAX_PFS/MAX_FAULTS_SAVED_PER_VM (Michal W)
- Store fault level per fault (Micahl M)
v5:
- Store fault and access type instead of address type (Jianxun)
v6:
- Store pagefaults in non-fault-mode VMs as well (Jianxun)
v7:
- Fix kernel docs and comments (Michal W)
v8:
- Fix double-locking issue (Jianxun)
v9:
- Do not report faults from reserved engines (Jianxun)
v10:
- Remove engine class and instance (Ivan)
v11:
- Perform kzalloc outside of lock (Auld)
--- drivers/gpu/drm/xe/xe_gt_pagefault.c | 26 ++++++++ drivers/gpu/drm/xe/xe_vm.c | 88 ++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 9 +++ drivers/gpu/drm/xe/xe_vm_types.h | 28 +++++++++ 4 files changed, 151 insertions(+)
drivers/gpu/drm/xe/xe_gt_pagefault.c +26 -0
@@ -345,6 +345,31 @@ int xe_guc_pagefault_handler(struct xe_guc *guc, u32 *msg, u32 len)return full ? -ENOSPC :0 ; } +static void save_pagefault_to_vm (struct xe_device *xe,struct xe_gt_pagefault *pf) +{ +struct xe_vm *vm; + +/* +* Pagefault may be associated to VM that is not in fault mode. +* Perform asid_to_vm behavior, except if vm is not in fault +* mode, return the VM anyways. +*/ +down_read (&xe->usm.lock); + vm =xa_load (&xe->usm.asid_to_vm, pf->asid); +if (vm) +xe_vm_get (vm); +else + vm =ERR_PTR (-EINVAL ); +up_read (&xe->usm.lock); + +if (IS_ERR (vm)) +return ; + +xe_vm_add_fault_entry_pf (vm, pf); + +xe_vm_put (vm); +} +#define USM_QUEUE_MAX_RUNTIME_MS 20static void pf_queue_work_func (struct work_struct *w)
@@ -364,6 +389,7 @@ static void pf_queue_work_func(struct work_struct *w)if (unlikely (ret)) {print_pagefault (xe, &pf); pf.fault_unsuccessful =1 ; +save_pagefault_to_vm (xe, &pf);drm_dbg (&xe->drm,"Fault response: Unsuccessful %d\n" , ret); }
drivers/gpu/drm/xe/xe_vm.c +88 -0
113 lines not on this page · raw message
drivers/gpu/drm/xe/xe_vm.h +9 -0
@@ -12,6 +12,12 @@#include "xe_map.h" #include "xe_vm_types.h" +/** +* MAX_FAULTS_SAVED_PER_VM - Maximum number of faults each vm can store before future +* faults are discarded to prevent memory overuse +*/ +#define MAX_FAULTS_SAVED_PER_VM 50 +struct drm_device ;struct drm_printer ;struct drm_file ;
@@ -22,6 +28,7 @@ struct dma_fence;struct xe_exec_queue ;struct xe_file ; +struct xe_gt_pagefault ;struct xe_sync_entry ;struct xe_svm_range ;struct drm_exec ;
@@ -257,6 +264,8 @@ int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);int xe_vma_userptr_check_repin (struct xe_userptr_vma *uvma); +void xe_vm_add_fault_entry_pf (struct xe_vm *vm,struct xe_gt_pagefault *pf); +bool xe_vm_validate_should_retry (struct drm_exec *exec,int err,ktime_t *end);int xe_vm_lock_vma (struct drm_exec *exec,struct xe_vma *vma);
drivers/gpu/drm/xe/xe_vm_types.h +28 -0
@@ -19,6 +19,7 @@#include "xe_range_fence.h" struct xe_bo ; +struct xe_pagefault ;struct xe_svm_range ;struct xe_sync_entry ;struct xe_user_fence ;
@@ -142,6 +143,23 @@ struct xe_userptr_vma {struct xe_device ; +/** +* struct xe_vm_fault_entry - Elements of vm->faults.list +* @list: link into @xe_vm.faults.list +* @address: address of the fault +* @address_type: type of address access that resulted in fault +* @address_precision: precision of faulted address +* @fault_level: fault level of the fault +*/ +struct xe_vm_fault_entry { +struct list_head list; +u64 address; +u32 address_precision; +u8 access_type; +u8 fault_type; +u8 fault_level; +}; +struct xe_vm {/** @gpuvm: base GPUVM used to track VMAs */ struct drm_gpuvm gpuvm;
@@ -305,6 +323,16 @@ struct xe_vm {bool capture_once; } error_capture; +/** @faults: List of all faults associated with this VM */ +struct { +/** @faults.lock: lock protecting @faults.list */ +spinlock_t lock; +/** @faults.list: list of xe_vm_fault_entry entries */ +struct list_head list; +/** @faults.len: length of @faults.list */ +unsigned int len; + } faults; +/** * @tlb_flush_seqno: Required TLB flush seqno for the next exec. * protected by the vm resv.
Signature · 1 line
2.43.0
RE: [PATCH v21 4/5] drm/xe/xe_vm: Add per VM fault info
Quoted 138 lines · Jonathan
Add additional information to each VM so they can report up to the first 50 seen faults. Only pagefaults are saved this way currently, though in the future, all faults should be tracked by the VM for future reporting.
Additionally, of the pagefaults reported, only failed pagefaults are saved this way, as successful pagefaults should recover silently and not need to be reported to userspace.
v2:
- Free vm after use (Shuicheng)
- Compress pf copy logic (Shuicheng)
- Updatefault_unsuccessfulbefore storing (Shuicheng)
- Fix old struct name in comments (Shuicheng)
- Keep first 50 pagefaults instead of last 50 (Jianxun)v3:
- Avoid unnecessary execution by checkingMAX_PFSearlier (jcavitt)
- Fix double-locking error (jcavitt)
- Assert kmemdump is successful (Shuicheng)v4:
- Rename xe_vm.pfs to xe_vm.faults (jcavitt)
- Store fault data and not pagefault inxe_vmfaults list (jcavitt)
- Store address, address type, and address precision per fault (jcavitt)
- Store engine class and instance data per fault (Jianxun)
- Add and fix kernel docs (Michal W)
- Properly handle kzalloc error (Michal W)
- s/MAX_PFS/MAX_FAULTS_SAVED_PER_VM (Michal W)
- Store fault level per fault (Micahl M)v5:
- Store fault and access type instead of address type (Jianxun)v6:
- Store pagefaults in non-fault-mode VMs as well (Jianxun)v7:
- Fix kernel docs and comments (Michal W)v8:
- Fix double-locking issue (Jianxun)v9:
- Do not report faults from reserved engines (Jianxun)v10:
- Remove engine class and instance (Ivan)v11:
- Perform kzalloc outside of lock (Auld)Signed-off-by Jonathan Cavitt <[email protected]>Suggested-by Matthew Brost <[email protected]>Cc Shuicheng Lin <[email protected]>Cc Jianxun Zhang <[email protected]>Cc Michal Wajdeczko <[email protected]>Cc Michal Mzorek <[email protected]>Cc Ivan Briano <[email protected]>Cc Matthew Auld <[email protected]>--- drivers/gpu/drm/xe/xe_gt_pagefault.c | 26 ++++++++ drivers/gpu/drm/xe/xe_vm.c | 88 ++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 9 +++ drivers/gpu/drm/xe/xe_vm_types.h | 28 +++++++++ 4 files changed, 151 insertions(+)drivers/gpu/drm/xe/xe_gt_pagefault.c +25
-0@@ -345,6 +345,31 @@ int xe_guc_pagefault_handler(struct xe_guc *guc, u32 *msg, u32 len)return full ? -ENOSPC :0 ; } +static void save_pagefault_to_vm (struct xe_device *xe,struct +xe_gt_pagefault *pf) { +struct xe_vm *vm; + +/* +* Pagefault may be associated to VM that is not in fault mode. +* Perform asid_to_vm behavior, except if vm is not in fault +* mode, return the VM anyways. +*/ +down_read (&xe->usm.lock); + vm =xa_load (&xe->usm.asid_to_vm, pf->asid); +if (vm) +xe_vm_get (vm); +else + vm =ERR_PTR (-EINVAL ); +up_read (&xe->usm.lock); + +if (IS_ERR (vm)) +return ; + +xe_vm_add_fault_entry_pf (vm, pf); + +xe_vm_put (vm); +} +#define USM_QUEUE_MAX_RUNTIME_MS 20static void
pf_queue_work_func(struct work_struct *w)@@ -364,6 +389,7
@@ static voidpf_queue_work_func(struct work_struct *w)drivers/gpu/drm/xe/xe_gt_pagefault.c +1
-0if (unlikely (ret)) {print_pagefault (xe, &pf); pf.fault_unsuccessful =1 ; +save_pagefault_to_vm (xe, &pf); drm_dbg(&xe->drm, "Fault response:Unsuccessful %d\n", ret);
drivers/gpu/drm/xe/xe_gt_pagefault.c +0
-0}drivers/gpu/drm/xe/xe_vm.c index renamed from drivers/gpu/drm/xe/xe_vm.c +0
-0
0c69ef6b5ec5..107e397b4987100644drivers/gpu/drm/xe/xe_vm.c +9
-0@@ -27,7 +27,9 @@#include "xe_device.h" #include "xe_drm_client.h" #include "xe_exec_queue.h" +#include "xe_gt.h" #include "xe_gt_pagefault.h" +#include "xe_gt_pagefault_types.h" #include "xe_gt_tlb_invalidation.h" #include "xe_migrate.h" #include "xe_pat.h" @@ -778,6 +780,87 @@ int xe_vm_userptr_check_repin(struct xe_vm *vm) list_empty_careful(&vm->userptr.invalidated)) ? 0 : -EAGAIN ; } +static struct xe_hw_engine * +hw_engine_lookup_class_instance (struct xe_vm *vm, +enum xe_engine_class class, +u16 instance) +{ +struct xe_device *xe = vm->xe; +struct xe_hw_engine *hwe;
drivers/gpu/drm/xe/xe_vm.c int xe_vm_userptr_check_repin(struct xe_vm *vm) · line 799
Quoted 10 lines
drivers/gpu/drm/xe/xe_vm.c +10
-0+enum xe_hw_engine_id id; +struct xe_gt *gt; +u8 gt_id; + +for_each_gt (gt, xe, gt_id) +for_each_hw_engine (hwe, gt, id) +if (hwe->class == class && hwe->instance == instance) +return hwe; +return NULL ; +}
There is a similar function xe_gt_hw_engine() to lookup to hwe. Anyway, I am ok with the function.
Shuicheng
Quoted 139 lines
drivers/gpu/drm/xe/xe_vm.c +7
-0+ +/** +* xe_vm_add_fault_entry_pf() - Add pagefault to vm fault list +* @vm: The VM. +* @pf: The pagefault. +* +* This function takes the data from the pagefault @pf and saves it to @vm- Quoted 1 line
faults.list.
+ *
+ * The function exits silently if the list is full, and reports adrivers/gpu/drm/xe/xe_vm.c +10
-0+warning if the pagefault +* could not be saved to the list. +*/ +void xe_vm_add_fault_entry_pf (struct xe_vm *vm,struct xe_gt_pagefault +*pf) { +struct xe_vm_fault_entry *e =NULL ; +struct xe_hw_engine *hwe; + +/* Do not report faults on reserved engines */ + hwe = hw_engine_lookup_class_instance(vm, pf->engine_class, pf-Quoted 1 line
engine_instance);drivers/gpu/drm/xe/xe_vm.c +48
-0+if (!hwe ||xe_hw_engine_is_reserved (hwe)) +return ; + + e =kzalloc (sizeof (*e),GFP_KERNEL ); +if (!e) { +drm_warn (&vm->xe->drm, +"Could not allocate memory for fault!\n" ); +return ; + } + +spin_lock (&vm->faults.lock); + +/* +* Limit the number of faults in the fault list to prevent +* memory overuse. +*/ +if (vm->faults.len >=MAX_FAULTS_SAVED_PER_VM ) { +kfree (e); + goto out; + } + + e->address = pf->page_addr; + e->address_precision =1 ; + e->access_type = pf->access_type; + e->fault_type = pf->fault_type; + e->fault_level = pf->fault_level; + +list_add_tail (&e->list, &vm->faults.list); + vm->faults.len++; +out: +spin_unlock (&vm->faults.lock); +} + +static void xe_vm_clear_fault_entries (struct xe_vm *vm) { +struct xe_vm_fault_entry *e, *tmp; + +spin_lock (&vm->faults.lock); +list_for_each_entry_safe (e, tmp, &vm->faults.list, list) { +list_del (&e->list); +kfree (e); + } + vm->faults.len =0 ; +spin_unlock (&vm->faults.lock); +} +static int xe_vma_ops_alloc (struct xe_vma_ops *vops,bool array_of_binds) {int i;@@ -1660,6 +1743,9 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) init_rwsem(&vm->userptr.notifier_lock);spin_lock_init (&vm->userptr.invalidated_lock); +INIT_LIST_HEAD (&vm->faults.list); +spin_lock_init (&vm->faults.lock); +ttm_lru_bulk_move_init (&vm->lru_bulk_move);
INIT_WORK(&vm->destroy_work, vm_destroy_work_func); @@ -
1930,6 +2016,8 @@ voidxe_vm_close_and_put(struct xe_vm *vm)drivers/gpu/drm/xe/xe_vm.c +2
-0}up_write (&xe->usm.lock); +xe_vm_clear_fault_entries (vm); +for_each_tile (tile, xe, id)xe_range_fence_tree_fini (&vm->rftree[id]);drivers/gpu/drm/xe/xe_vm.h index renamed from drivers/gpu/drm/xe/xe_vm.h +0
-0
0ef811fc2bde..9bd7e93824da100644drivers/gpu/drm/xe/xe_vm.h +2
-0@@ -12,6 +12,12 @@#include "xe_map.h" #include "xe_vm_types.h" +/** +* MAX_FAULTS_SAVED_PER_VM - Maximum number of faults each vm can store
drivers/gpu/drm/xe/xe_vm.h +8
-0+before future +* faults are discarded to prevent memory overuse */ +#define MAX_FAULTS_SAVED_PER_VM 50 +struct drm_device ;struct drm_printer ;struct drm_file ;@@ -22,6 +28,7 @@ struct dma_fence;struct xe_exec_queue ;struct xe_file ; +struct xe_gt_pagefault ;struct xe_sync_entry ;struct xe_svm_range ;struct drm_exec ;@@ -257,6 +264,8 @@ int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);int xe_vma_userptr_check_repin (struct xe_userptr_vma *uvma); +void xe_vm_add_fault_entry_pf (struct xe_vm *vm,struct xe_gt_pagefault +*pf); + bool xe_vm_validate_should_retry(struct drm_exec *exec, int err, ktime_t*end);
int
xe_vm_lock_vma(struct drm_exec *exec, struct xe_vma *vma); diff --gita/drivers/gpu/drm/xe/xe_vm_types.hb/drivers/gpu/drm/xe/xe_vm_types.hindex1662604c4486..ff67d3c7145d100644drivers/gpu/drm/xe/xe_vm_types.h +1
-0@@ -19,6 +19,7 @@#include "xe_range_fence.h" struct xe_bo ; +struct xe_pagefault ;struct xe_svm_range ;struct xe_sync_entry ;struct xe_user_fence ;
drivers/gpu/drm/xe/xe_vm_types.h struct xe_userptr_vma { · line 150
Quoted 8 lines
drivers/gpu/drm/xe/xe_vm_types.h +5
-0@@ -142,6 +143,23 @@ struct xe_userptr_vma {struct xe_device ; +/** +* struct xe_vm_fault_entry - Elements of vm->faults.list +* @list: link into @xe_vm.faults.list +* @address: address of the fault +* @address_type: type of address access that resulted in fault
Should be access_type? And fault_type is missing. Also, the sequence is different from the structure definition.
Shuicheng
Quoted 33 lines
drivers/gpu/drm/xe/xe_vm_types.h +21
-1+* @address_precision: precision of faulted address +* @fault_level: fault level of the fault */ struct xe_vm_fault_entry +{ +struct list_head list; +u64 address; +u32 address_precision; +u8 access_type; +u8 fault_type; +u8 fault_level; +}; +struct xe_vm {/** @gpuvm: base GPUVM used to track VMAs */ struct drm_gpuvm gpuvm;@@ -305,6 +323,16 @@ struct xe_vm {bool capture_once; } error_capture; +/** @faults: List of all faults associated with this VM */ +struct { +/** @faults.lock: lock protecting @faults.list */ +spinlock_t lock; +/** @faults.list: list of xe_vm_fault_entry entries */ +struct list_head list; +/** @faults.len: length of @faults.list */ +unsigned int len; + } faults; +/** * @tlb_flush_seqno: Required TLB flush seqno for the next exec. * protected by the vm resv. -- 2.43.0
RE: [PATCH v21 4/5] drm/xe/xe_vm: Add per VM fault info
-----Original Message-----
From: Lin, Shuicheng <[email protected]>
Sent: Wednesday, April 23, 2025 8:49 PM
To: Cavitt, Jonathan <[email protected]>; [email protected]
Cc: Gupta, saurabhg <[email protected]>; Zuo, Alex <[email protected]>; [email protected]; Brost, Matthew <[email protected]>; Zhang, Jianxun <[email protected]>; [email protected]; Wajdeczko, Michal <[email protected]>; Mrozek, Michal <[email protected]>; Jadav, Raag <[email protected]>; Harrison, John C <[email protected]>; Briano, Ivan <[email protected]>; Auld, Matthew <[email protected]>
Subject: RE: [PATCH v21 4/5] drm/xe/xe_vm: Add per VM fault info
Quoted 145 lines · Jonathan
Quoted 143 lines
Add additional information to each VM so they can report up to the first 50 seen faults. Only pagefaults are saved this way currently, though in the future, all faults should be tracked by the VM for future reporting.
Additionally, of the pagefaults reported, only failed pagefaults are saved this way, as successful pagefaults should recover silently and not need to be reported to userspace.
v2:
- Free vm after use (Shuicheng)
- Compress pf copy logic (Shuicheng)
- Updatefault_unsuccessfulbefore storing (Shuicheng)
- Fix old struct name in comments (Shuicheng)
- Keep first 50 pagefaults instead of last 50 (Jianxun)v3:
- Avoid unnecessary execution by checkingMAX_PFSearlier (jcavitt)
- Fix double-locking error (jcavitt)
- Assert kmemdump is successful (Shuicheng)v4:
- Rename xe_vm.pfs to xe_vm.faults (jcavitt)
- Store fault data and not pagefault inxe_vmfaults list (jcavitt)
- Store address, address type, and address precision per fault (jcavitt)
- Store engine class and instance data per fault (Jianxun)
- Add and fix kernel docs (Michal W)
- Properly handle kzalloc error (Michal W)
- s/MAX_PFS/MAX_FAULTS_SAVED_PER_VM (Michal W)
- Store fault level per fault (Micahl M)v5:
- Store fault and access type instead of address type (Jianxun)v6:
- Store pagefaults in non-fault-mode VMs as well (Jianxun)v7:
- Fix kernel docs and comments (Michal W)v8:
- Fix double-locking issue (Jianxun)v9:
- Do not report faults from reserved engines (Jianxun)v10:
- Remove engine class and instance (Ivan)v11:
- Perform kzalloc outside of lock (Auld)Signed-off-by Jonathan Cavitt <[email protected]>Suggested-by Matthew Brost <[email protected]>Cc Shuicheng Lin <[email protected]>Cc Jianxun Zhang <[email protected]>Cc Michal Wajdeczko <[email protected]>Cc Michal Mzorek <[email protected]>Cc Ivan Briano <[email protected]>Cc Matthew Auld <[email protected]>--- drivers/gpu/drm/xe/xe_gt_pagefault.c | 26 ++++++++ drivers/gpu/drm/xe/xe_vm.c | 88 ++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 9 +++ drivers/gpu/drm/xe/xe_vm_types.h | 28 +++++++++ 4 files changed, 151 insertions(+)drivers/gpu/drm/xe/xe_gt_pagefault.c +25
-0@@ -345,6 +345,31 @@ int xe_guc_pagefault_handler(struct xe_guc *guc, u32 *msg, u32 len)return full ? -ENOSPC :0 ; } +static void save_pagefault_to_vm (struct xe_device *xe,struct +xe_gt_pagefault *pf) { +struct xe_vm *vm; + +/* +* Pagefault may be associated to VM that is not in fault mode. +* Perform asid_to_vm behavior, except if vm is not in fault +* mode, return the VM anyways. +*/ +down_read (&xe->usm.lock); + vm =xa_load (&xe->usm.asid_to_vm, pf->asid); +if (vm) +xe_vm_get (vm); +else + vm =ERR_PTR (-EINVAL ); +up_read (&xe->usm.lock); + +if (IS_ERR (vm)) +return ; + +xe_vm_add_fault_entry_pf (vm, pf); + +xe_vm_put (vm); +} +#define USM_QUEUE_MAX_RUNTIME_MS 20static void
pf_queue_work_func(struct work_struct *w)@@ -364,6 +389,7
@@ static voidpf_queue_work_func(struct work_struct *w)drivers/gpu/drm/xe/xe_gt_pagefault.c +1
-0if (unlikely (ret)) {print_pagefault (xe, &pf); pf.fault_unsuccessful =1 ; +save_pagefault_to_vm (xe, &pf); drm_dbg(&xe->drm, "Fault response:Unsuccessful %d\n", ret);
drivers/gpu/drm/xe/xe_gt_pagefault.c +0
-0}drivers/gpu/drm/xe/xe_vm.c index renamed from drivers/gpu/drm/xe/xe_vm.c +0
-0
0c69ef6b5ec5..107e397b4987100644drivers/gpu/drm/xe/xe_vm.c +14
-0@@ -27,7 +27,9 @@#include "xe_device.h" #include "xe_drm_client.h" #include "xe_exec_queue.h" +#include "xe_gt.h" #include "xe_gt_pagefault.h" +#include "xe_gt_pagefault_types.h" #include "xe_gt_tlb_invalidation.h" #include "xe_migrate.h" #include "xe_pat.h" @@ -778,6 +780,87 @@ int xe_vm_userptr_check_repin(struct xe_vm *vm) list_empty_careful(&vm->userptr.invalidated)) ? 0 : -EAGAIN ; } +static struct xe_hw_engine * +hw_engine_lookup_class_instance (struct xe_vm *vm, +enum xe_engine_class class, +u16 instance) +{ +struct xe_device *xe = vm->xe; +struct xe_hw_engine *hwe; +enum xe_hw_engine_id id; +struct xe_gt *gt; +u8 gt_id; + +for_each_gt (gt, xe, gt_id)
Quoted 8 lines
drivers/gpu/drm/xe/xe_vm.c int xe_vm_userptr_check_repin(struct xe_vm *vm) · line 799
Quoted 5 lines
drivers/gpu/drm/xe/xe_vm.c +5
-0+for_each_hw_engine (hwe, gt, id) +if (hwe->class == class && hwe->instance == instance) +return hwe; +return NULL ; +}There is a similar function
xe_gt_hw_engine()to lookup to hwe. Anyway, I am ok with the function.
This is correct. The problem is that xe_gt_hw_engine takes a struct xe_gt *gt, which is not provided by the PF structure.
Though, I suppose this wasn't necessary to explain since you're already okay with the function.
Quoted 145 lines
Shuicheng
Quoted 142 lines
drivers/gpu/drm/xe/xe_vm.c +7
-0+ +/** +* xe_vm_add_fault_entry_pf() - Add pagefault to vm fault list +* @vm: The VM. +* @pf: The pagefault. +* +* This function takes the data from the pagefault @pf and saves it to @vm- Quoted 1 line
faults.list.
+ *
+ * The function exits silently if the list is full, and reports adrivers/gpu/drm/xe/xe_vm.c +10
-0+warning if the pagefault +* could not be saved to the list. +*/ +void xe_vm_add_fault_entry_pf (struct xe_vm *vm,struct xe_gt_pagefault +*pf) { +struct xe_vm_fault_entry *e =NULL ; +struct xe_hw_engine *hwe; + +/* Do not report faults on reserved engines */ + hwe = hw_engine_lookup_class_instance(vm, pf->engine_class, pf-Quoted 1 line
engine_instance);drivers/gpu/drm/xe/xe_vm.c +48
-0+if (!hwe ||xe_hw_engine_is_reserved (hwe)) +return ; + + e =kzalloc (sizeof (*e),GFP_KERNEL ); +if (!e) { +drm_warn (&vm->xe->drm, +"Could not allocate memory for fault!\n" ); +return ; + } + +spin_lock (&vm->faults.lock); + +/* +* Limit the number of faults in the fault list to prevent +* memory overuse. +*/ +if (vm->faults.len >=MAX_FAULTS_SAVED_PER_VM ) { +kfree (e); + goto out; + } + + e->address = pf->page_addr; + e->address_precision =1 ; + e->access_type = pf->access_type; + e->fault_type = pf->fault_type; + e->fault_level = pf->fault_level; + +list_add_tail (&e->list, &vm->faults.list); + vm->faults.len++; +out: +spin_unlock (&vm->faults.lock); +} + +static void xe_vm_clear_fault_entries (struct xe_vm *vm) { +struct xe_vm_fault_entry *e, *tmp; + +spin_lock (&vm->faults.lock); +list_for_each_entry_safe (e, tmp, &vm->faults.list, list) { +list_del (&e->list); +kfree (e); + } + vm->faults.len =0 ; +spin_unlock (&vm->faults.lock); +} +static int xe_vma_ops_alloc (struct xe_vma_ops *vops,bool array_of_binds) {int i;@@ -1660,6 +1743,9 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) init_rwsem(&vm->userptr.notifier_lock);spin_lock_init (&vm->userptr.invalidated_lock); +INIT_LIST_HEAD (&vm->faults.list); +spin_lock_init (&vm->faults.lock); +ttm_lru_bulk_move_init (&vm->lru_bulk_move);
INIT_WORK(&vm->destroy_work, vm_destroy_work_func); @@ -
1930,6 +2016,8 @@ voidxe_vm_close_and_put(struct xe_vm *vm)drivers/gpu/drm/xe/xe_vm.c +2
-0}up_write (&xe->usm.lock); +xe_vm_clear_fault_entries (vm); +for_each_tile (tile, xe, id)xe_range_fence_tree_fini (&vm->rftree[id]);drivers/gpu/drm/xe/xe_vm.h index renamed from drivers/gpu/drm/xe/xe_vm.h +0
-0
0ef811fc2bde..9bd7e93824da100644drivers/gpu/drm/xe/xe_vm.h +2
-0@@ -12,6 +12,12 @@#include "xe_map.h" #include "xe_vm_types.h" +/** +* MAX_FAULTS_SAVED_PER_VM - Maximum number of faults each vm can store
drivers/gpu/drm/xe/xe_vm.h +8
-0+before future +* faults are discarded to prevent memory overuse */ +#define MAX_FAULTS_SAVED_PER_VM 50 +struct drm_device ;struct drm_printer ;struct drm_file ;@@ -22,6 +28,7 @@ struct dma_fence;struct xe_exec_queue ;struct xe_file ; +struct xe_gt_pagefault ;struct xe_sync_entry ;struct xe_svm_range ;struct drm_exec ;@@ -257,6 +264,8 @@ int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);int xe_vma_userptr_check_repin (struct xe_userptr_vma *uvma); +void xe_vm_add_fault_entry_pf (struct xe_vm *vm,struct xe_gt_pagefault +*pf); + bool xe_vm_validate_should_retry(struct drm_exec *exec, int err, ktime_t*end);
int
xe_vm_lock_vma(struct drm_exec *exec, struct xe_vma *vma); diff --gita/drivers/gpu/drm/xe/xe_vm_types.hb/drivers/gpu/drm/xe/xe_vm_types.hindex1662604c4486..ff67d3c7145d100644drivers/gpu/drm/xe/xe_vm_types.h +1
-0@@ -19,6 +19,7 @@#include "xe_range_fence.h" struct xe_bo ; +struct xe_pagefault ;struct xe_svm_range ;struct xe_sync_entry ;struct xe_user_fence ;@@ -142,6 +143,23 @@ struct xe_userptr_vma {struct xe_device ;
Quoted 8 lines
drivers/gpu/drm/xe/xe_vm_types.h struct xe_userptr_vma { · line 150
Quoted 5 lines
drivers/gpu/drm/xe/xe_vm_types.h +5
-0+/** +* struct xe_vm_fault_entry - Elements of vm->faults.list +* @list: link into @xe_vm.faults.list +* @address: address of the fault +* @address_type: type of address access that resulted in fault Should be
access_type? Andfault_typeis missing. Also, the sequence is different from the structure definition.
I think this was a fragment from an earlier version of the patch that never got updated to the most recent revision. I'll update these kernel docs shortly.
-Jonathan Cavitt
Quoted 38 lines
Shuicheng
Quoted 33 lines
drivers/gpu/drm/xe/xe_vm_types.h +21
-1+* @address_precision: precision of faulted address +* @fault_level: fault level of the fault */ struct xe_vm_fault_entry +{ +struct list_head list; +u64 address; +u32 address_precision; +u8 access_type; +u8 fault_type; +u8 fault_level; +}; +struct xe_vm {/** @gpuvm: base GPUVM used to track VMAs */ struct drm_gpuvm gpuvm;@@ -305,6 +323,16 @@ struct xe_vm {bool capture_once; } error_capture; +/** @faults: List of all faults associated with this VM */ +struct { +/** @faults.lock: lock protecting @faults.list */ +spinlock_t lock; +/** @faults.list: list of xe_vm_fault_entry entries */ +struct list_head list; +/** @faults.len: length of @faults.list */ +unsigned int len; + } faults; +/** * @tlb_flush_seqno: Required TLB flush seqno for the next exec. * protected by the vm resv. -- 2.43.0
PATCH v21 5/5 drm/xe/xe_vm: Implement xe_vm_get_property_ioctl
In reply to Jonathan CavittAdd support for userspace to request a list of observed faults from a specified VM.
v2:
- Only allow querying of failed pagefaults (Matt Brost)
v3:
- Remove unnecessary size parameter from helper function, as it is a property of the arguments. (jcavitt)
- Remove unnecessary copy_from_user (Jainxun)
- Set address_precision to 1 (Jainxun)
- Report max size instead of dynamic size for memory allocation purposes. Total memory usage is reported separately.
v4:
- Return int from xe_vm_get_property_size (Shuicheng)
- Fix memory leak (Shuicheng)
- Remove unnecessary size variable (jcavitt)
v5:
- Rename ioctl to xe_vm_get_faults_ioctl (jcavitt)
- Update fill_property_pfs to eliminate need for kzalloc (Jianxun)
v6:
- Repair and move fill_faults break condition (Dan Carpenter)
- Free vm after use (jcavitt)
- Combine assertions (jcavitt)
- Expand size check in xe_vm_get_faults_ioctl (jcavitt)
- Remove return mask from fill_faults, as return is already -EFAULT or 0 (jcavitt)
v7:
- Revert back to using xe_vm_get_property_ioctl
- Apply better copy_to_user logic (jcavitt)
v8:
- Fix and clean up error value handling in ioctl (jcavitt)
- Reapply return mask for fill_faults (jcavitt)
v9:
- Future-proof size logic for zero-size properties (jcavitt)
- Add access and fault types (Jianxun)
- Remove address type (Jianxun)
v10:
- Remove unnecessary switch case logic (Raag)
- Compress size get, size validation, and property fill functions into a single helper function (jcavitt)
- Assert valid size (jcavitt)
v11:
- Remove unnecessary else condition
- Correct backwards helper function size logic (jcavitt)
v12:
- Use size_t instead of int (Raag)
v13:
- Remove engine class and instance (Ivan)
v14:
- Map access type, fault type, and fault level to user macros (Matt Brost, Ivan)
v15:
- Remove unnecessary size assertion (jcavitt)
--- drivers/gpu/drm/xe/xe_device.c | 3 + drivers/gpu/drm/xe/xe_vm.c | 107 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 2 + 3 files changed, 112 insertions(+)
drivers/gpu/drm/xe/xe_device.c +3 -0
@@ -196,6 +196,9 @@ static const struct drm_ioctl_desc xe_ioctls[] = {DRM_IOCTL_DEF_DRV (XE_WAIT_USER_FENCE , xe_wait_user_fence_ioctl,DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (XE_OBSERVATION , xe_observation_ioctl,DRM_RENDER_ALLOW ), +DRM_IOCTL_DEF_DRV (XE_VM_GET_PROPERTY , xe_vm_get_property_ioctl, + DRM_RENDER_ALLOW), + };static long xe_drm_ioctl (struct file *file,unsigned int cmd,unsigned long arg)
drivers/gpu/drm/xe/xe_vm.c +107 -0
113 lines not on this page · raw message
drivers/gpu/drm/xe/xe_vm.h +2 -0
@@ -196,6 +196,8 @@ int xe_vm_destroy_ioctl(struct drm_device *dev, void *data, struct drm_file *file);int xe_vm_bind_ioctl (struct drm_device *dev,void *data,struct drm_file *file); +int xe_vm_get_property_ioctl (struct drm_device *dev,void *data, +struct drm_file *file);void xe_vm_close_and_put (struct xe_vm *vm);
Signature · 1 line
2.43.0
Quoted 67 lines · Jonathan
Add support for userspace to request a list of observed faults from a specified VM.
v2:
- Only allow querying of failed pagefaults (Matt Brost)v3:
- Remove unnecessary size parameter from helper function, as it is a property of the arguments. (jcavitt)
- Remove unnecessarycopy_from_user(Jainxun)
- Setaddress_precisionto 1 (Jainxun)
- Report max size instead of dynamic size for memory allocation purposes. Total memory usage is reported separately.v4:
- Return int fromxe_vm_get_property_size(Shuicheng)
- Fix memory leak (Shuicheng)
- Remove unnecessary size variable (jcavitt)v5:
- Rename ioctl toxe_vm_get_faults_ioctl(jcavitt)
- Updatefill_property_pfsto eliminate need for kzalloc (Jianxun)v6:
- Repair and movefill_faultsbreak condition (Dan Carpenter)
- Free vm after use (jcavitt)
- Combine assertions (jcavitt)
- Expand size check inxe_vm_get_faults_ioctl(jcavitt)
- Remove return mask fromfill_faults, as return is already-EFAULTor 0 (jcavitt)v7:
- Revert back to usingxe_vm_get_property_ioctl
- Apply bettercopy_to_userlogic (jcavitt)v8:
- Fix and clean up error value handling in ioctl (jcavitt)
- Reapply return mask forfill_faults(jcavitt)v9:
- Future-proof size logic for zero-size properties (jcavitt)
- Add access and fault types (Jianxun)
- Remove address type (Jianxun)v10:
- Remove unnecessary switch case logic (Raag)
- Compress size get, size validation, and property fill functions into a single helper function (jcavitt)
- Assert valid size (jcavitt)v11:
- Remove unnecessary else condition
- Correct backwards helper function size logic (jcavitt)v12:
- Usesize_tinstead of int (Raag)v13:
- Remove engine class and instance (Ivan)v14:
- Map access type, fault type, and fault level to user macros (Matt Brost, Ivan)v15:
- Remove unnecessary size assertion (jcavitt)
Quoted 6 lines
Signed-off-by Jonathan Cavitt <[email protected]>Suggested-by Matthew Brost <[email protected]>Cc Jainxun Zhang <[email protected]>Cc Shuicheng Lin <[email protected]>Cc Raag Jadav <[email protected]>Cc Ivan Briano <[email protected]>
LGTM.
Quoted 156 lines
--- drivers/gpu/drm/xe/xe_device.c | 3 + drivers/gpu/drm/xe/xe_vm.c | 107 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm.h | 2 + 3 files changed, 112 insertions(+)drivers/gpu/drm/xe/xe_device.c +3
-0@@ -196,6 +196,9 @@ static const struct drm_ioctl_desc xe_ioctls[] = {DRM_IOCTL_DEF_DRV (XE_WAIT_USER_FENCE , xe_wait_user_fence_ioctl,DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (XE_OBSERVATION , xe_observation_ioctl,DRM_RENDER_ALLOW ), +DRM_IOCTL_DEF_DRV (XE_VM_GET_PROPERTY , xe_vm_get_property_ioctl, + DRM_RENDER_ALLOW), +};
static long
xe_drm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)drivers/gpu/drm/xe/xe_vm.c index renamed from drivers/gpu/drm/xe/xe_vm.c +0
-0
107e397b4987..c20ac51d8573100644drivers/gpu/drm/xe/xe_vm.c +102
-0@@ -3600,6 +3600,113 @@ int xe_vm_bind_ioctl(struct drm_device *dev, void *data, struct drm_file *file) return err; } +/* +* Map access type, fault type, and fault level from current bspec +* specification to user spec abstraction. The current mapping is +* 1-to-1, but if there is ever a hardware change, we will need +* this abstraction layer to maintain API stability through the +* hardware change. +*/ +static u8 xe_to_user_access_type (u8 access_type) { +return access_type; +} + +static u8 xe_to_user_fault_type (u8 fault_type) { +return fault_type; +} + +static u8 xe_to_user_fault_level (u8 fault_level) { +return fault_level; +} + +static int fill_faults (struct xe_vm *vm, +struct drm_xe_vm_get_property *args) { +struct xe_vm_fault __user *usr_ptr =u64_to_user_ptr (args->data); +struct xe_vm_fault store = {0 }; +struct xe_vm_fault_entry *entry; +int ret =0 , i =0 , count, entry_size; + + entry_size =sizeof (struct xe_vm_fault ); + count = args->size / entry_size; + +spin_lock (&vm->faults.lock); +list_for_each_entry (entry, &vm->faults.list, list) { +if (i++ == count) +break ; + +memset (&store,0 , entry_size); + + store.address = entry->address; + store.address_precision = entry->address_precision; + + store.access_type =xe_to_user_access_type (entry- >access_type); + store.fault_type =xe_to_user_fault_type (entry->fault_type); + store.fault_level =xe_to_user_fault_level (entry->fault_level); + + ret =copy_to_user (usr_ptr, &store, entry_size); +if (ret) +break ; + + usr_ptr++; + } +spin_unlock (&vm->faults.lock); + +return ret ? -EFAULT :0 ; +} + +static int xe_vm_get_property_helper (struct xe_vm *vm, +struct drm_xe_vm_get_property *args) { +size_t size; + +switch (args->property) { +case DRM_XE_VM_GET_PROPERTY_FAULTS : +spin_lock (&vm->faults.lock); + size =size_mul (sizeof (struct xe_vm_fault ), vm->faults.len); +spin_unlock (&vm->faults.lock); + +if (args->size) +/* +* Number of faults may increase between calls to +* xe_vm_get_property_ioctl, so just report the +* number of faults the user requests if it's less +* than or equal to the number of faults in the VM +* fault array. +*/ +return args->size <= size ?fill_faults (vm, args) : -EINVAL ; + + args->size = size; +return 0 ; + } +return -EINVAL ; +} + +int xe_vm_get_property_ioctl (struct drm_device *drm,void *data, +struct drm_file *file) +{ +struct xe_device *xe =to_xe_device (drm); +struct xe_file *xef =to_xe_file (file); +struct drm_xe_vm_get_property *args = data; +struct xe_vm *vm; +int ret =0 ; + +if (XE_IOCTL_DBG (xe, args->reserved[0 ] || args->reserved[1 ])) +return -EINVAL ; + + vm =xe_vm_lookup (xef, args->vm_id); +if (XE_IOCTL_DBG (xe, !vm)) +return -ENOENT ; + + ret =xe_vm_get_property_helper (vm, args); + +xe_vm_put (vm); +return ret; +} +/** * xe_vm_bind_kernel_bo - bind a kernel BO to a VM * @vm: VM to bind the BO to drivers/gpu/drm/xe/xe_vm.h index renamed from drivers/gpu/drm/xe/xe_vm.h +0
-0
9bd7e93824da..63ec22458e04100644drivers/gpu/drm/xe/xe_vm.h +2
-1@@ -196,6 +196,8 @@ int xe_vm_destroy_ioctl(struct drm_device *dev, void *data, struct drm_file *file);int xe_vm_bind_ioctl (struct drm_device *dev,void *data,struct drm_file *file); +int xe_vm_get_property_ioctl (struct drm_device *dev,void *data, +struct drm_file *file);void xe_vm_close_and_put (struct xe_vm *vm); --2.43.0