On 11/5/2025 4:10 PM, Michał Winiarski wrote:
> Now that it's possible to free the packets - connect the restore
> handling logic with the ring.
> The helpers will also be used in upcoming changes that will start
> producing migration data packets.
>
> Signed-off-by: Michał Winiarski <[email protected]>
> Reviewed-by: Michal Wajdeczko <[email protected]>
> ---
> drivers/gpu/drm/xe/Makefile | 1 +
> drivers/gpu/drm/xe/xe_gt_sriov_pf_control.c | 11 +-
> drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c | 33 ++++-
> drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.h | 1 +
> drivers/gpu/drm/xe/xe_sriov_packet.c | 129 ++++++++++++++++++
> drivers/gpu/drm/xe/xe_sriov_packet.h | 30 ++++
> 6 files changed, 202 insertions(+), 3 deletions(-)
> create mode 100644 drivers/gpu/drm/xe/xe_sriov_packet.c
> create mode 100644 drivers/gpu/drm/xe/xe_sriov_packet.h
>
> diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile
> index cb6b30fd8c2d8..e4b273b025d2a 100644
> --- a/drivers/gpu/drm/xe/Makefile
> +++ b/drivers/gpu/drm/xe/Makefile
> @@ -174,6 +174,7 @@ xe-$(CONFIG_PCI_IOV) += \
> xe_lmtt_2l.o \
> xe_lmtt_ml.o \
> xe_pci_sriov.o \
> + xe_sriov_packet.o \
> xe_sriov_pf.o \
> xe_sriov_pf_control.o \
> xe_sriov_pf_debugfs.o \
> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_control.c
> b/drivers/gpu/drm/xe/xe_gt_sriov_pf_control.c
> index db57f9d9b491f..25937592bc42f 100644
> --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_control.c
> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_control.c
> @@ -18,6 +18,7 @@
> #include "xe_gt_sriov_printk.h"
> #include "xe_guc_ct.h"
> #include "xe_sriov.h"
> +#include "xe_sriov_packet.h"
> #include "xe_sriov_pf_control.h"
> #include "xe_sriov_pf_migration.h"
> #include "xe_sriov_pf_service.h"
> @@ -852,6 +853,8 @@ int xe_gt_sriov_pf_control_resume_vf(struct xe_gt *gt,
> unsigned int vfid)
> static void pf_exit_vf_save_wip(struct xe_gt *gt, unsigned int vfid)
> {
> if (pf_exit_vf_state(gt, vfid, XE_GT_SRIOV_STATE_SAVE_WIP)) {
> + xe_gt_sriov_pf_migration_ring_free(gt, vfid);
> +
> pf_escape_vf_state(gt, vfid,
> XE_GT_SRIOV_STATE_SAVE_PROCESS_DATA);
> pf_escape_vf_state(gt, vfid, XE_GT_SRIOV_STATE_SAVE_WAIT_DATA);
> pf_escape_vf_state(gt, vfid, XE_GT_SRIOV_STATE_SAVE_DATA_DONE);
> @@ -1074,6 +1077,8 @@ int xe_gt_sriov_pf_control_finish_save_vf(struct xe_gt
> *gt, unsigned int vfid)
> static void pf_exit_vf_restore_wip(struct xe_gt *gt, unsigned int vfid)
> {
> if (pf_exit_vf_state(gt, vfid, XE_GT_SRIOV_STATE_RESTORE_WIP)) {
> + xe_gt_sriov_pf_migration_ring_free(gt, vfid);
> +
> pf_escape_vf_state(gt, vfid,
> XE_GT_SRIOV_STATE_RESTORE_PROCESS_DATA);
> pf_escape_vf_state(gt, vfid,
> XE_GT_SRIOV_STATE_RESTORE_WAIT_DATA);
> pf_escape_vf_state(gt, vfid,
> XE_GT_SRIOV_STATE_RESTORE_DATA_DONE);
> @@ -1108,6 +1113,8 @@ static int pf_handle_vf_restore_data(struct xe_gt *gt,
> unsigned int vfid)
>
> xe_gt_sriov_notice(gt, "Skipping VF%u unknown data type: %d\n", vfid,
> data->type);
>
> + xe_sriov_packet_free(data);
> +
> return 0;
> }
>
> @@ -1201,8 +1208,10 @@ int xe_gt_sriov_pf_control_restore_data_done(struct
> xe_gt *gt, unsigned int vfid
> */
> int xe_gt_sriov_pf_control_process_restore_data(struct xe_gt *gt, unsigned
> int vfid)
> {
> - if (!pf_expect_vf_not_state(gt, vfid, XE_GT_SRIOV_STATE_RESTORE_FAILED))
> + if (!pf_expect_vf_not_state(gt, vfid,
> XE_GT_SRIOV_STATE_RESTORE_FAILED)) {
> + xe_gt_sriov_pf_migration_ring_free(gt, vfid);
> return -EIO;
> + }
>
> pf_exit_vf_restore_wait_data(gt, vfid);
>
> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c
> b/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c
> index d9ca9ea737dcc..c492677db28d9 100644
> --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c
> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.c
> @@ -14,6 +14,7 @@
> #include "xe_guc.h"
> #include "xe_guc_ct.h"
> #include "xe_sriov.h"
> +#include "xe_sriov_packet.h"
> #include "xe_sriov_pf_migration.h"
>
> #define XE_GT_SRIOV_PF_MIGRATION_RING_SIZE 5
> @@ -418,6 +419,25 @@ bool xe_gt_sriov_pf_migration_ring_full(struct xe_gt
> *gt, unsigned int vfid)
> return ptr_ring_full(&pf_pick_gt_migration(gt, vfid)->ring);
> }
>
> +/**
> + * xe_gt_sriov_pf_migration_ring_free() - Consume and free all data in
> migration ring
> + * @gt: the &xe_gt
> + * @vfid: the VF identifier
> + */
> +void xe_gt_sriov_pf_migration_ring_free(struct xe_gt *gt, unsigned int vfid)
> +{
> + struct xe_gt_sriov_migration_data *migration = pf_pick_gt_migration(gt,
> vfid);
> + struct xe_sriov_packet *data;
> +
> + if (ptr_ring_empty(&migration->ring))
> + return;
> +
> + xe_gt_sriov_notice(gt, "VF%u unprocessed migration data left in the
> ring!\n", vfid);
> +
> + while ((data = ptr_ring_consume(&migration->ring)))
> + xe_sriov_packet_free(data);
> +}
> +
> /**
> * xe_gt_sriov_pf_migration_save_produce() - Add VF save data packet to
> migration ring.
> * @gt: the &xe_gt
> @@ -544,8 +564,10 @@ xe_gt_sriov_pf_migration_save_consume(struct xe_gt *gt,
> unsigned int vfid)
> data = ptr_ring_consume(&migration->ring);
> if (data) {
> ret = xe_gt_sriov_pf_control_process_save_data(gt, vfid);
> - if (ret)
> + if (ret) {
> + xe_sriov_packet_free(data);
> return ERR_PTR(ret);
> + }
>
> return data;
> }
> @@ -559,11 +581,18 @@ xe_gt_sriov_pf_migration_save_consume(struct xe_gt *gt,
> unsigned int vfid)
> return ERR_PTR(-EAGAIN);
> }
>
> +static void pf_mig_packet_destroy(void *ptr)
it's not a "pf" call, but 'destroy' callback, so maybe:
destroy_pf_packet()
> +{
> + struct xe_sriov_packet *data = ptr;
> +
> + xe_sriov_packet_free(data);
> +}
> +
> static void action_ring_cleanup(void *arg)
> {
> struct ptr_ring *r = arg;
>
> - ptr_ring_cleanup(r, NULL);
> + ptr_ring_cleanup(r, pf_mig_packet_destroy);
> }
>
> /**
> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.h
> b/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.h
> index b29d34a835b71..b4c7b5dc9e914 100644
> --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.h
> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_migration.h
> @@ -17,6 +17,7 @@ int xe_gt_sriov_pf_migration_restore_guc_state(struct xe_gt
> *gt, unsigned int vf
>
> bool xe_gt_sriov_pf_migration_ring_empty(struct xe_gt *gt, unsigned int
> vfid);
> bool xe_gt_sriov_pf_migration_ring_full(struct xe_gt *gt, unsigned int vfid);
> +void xe_gt_sriov_pf_migration_ring_free(struct xe_gt *gt, unsigned int vfid);
>
> int xe_gt_sriov_pf_migration_save_produce(struct xe_gt *gt, unsigned int
> vfid,
> struct xe_sriov_packet *data);
> diff --git a/drivers/gpu/drm/xe/xe_sriov_packet.c
> b/drivers/gpu/drm/xe/xe_sriov_packet.c
> new file mode 100644
> index 0000000000000..2cdcfb2073d00
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_sriov_packet.c
> @@ -0,0 +1,129 @@
> +// SPDX-License-Identifier: MIT
> +/*
> + * Copyright © 2025 Intel Corporation
> + */
> +
> +#include "xe_bo.h"
> +#include "xe_device.h"
> +#include "xe_sriov_packet.h"
> +
> +static bool pkt_needs_bo(struct xe_sriov_packet *data)
> +{
> + return data->type == XE_SRIOV_PACKET_TYPE_VRAM;
> +}
> +
> +/**
> + * xe_sriov_packet_alloc() - Allocate migration data packet
> + * @xe: the &xe_device
> + *
> + * Only allocates the "outer" structure, without initializing the migration
> + * data backing storage.
> + *
> + * Return: Pointer to &xe_sriov_packet on success,
> + * NULL in case of error.
> + */
> +struct xe_sriov_packet *xe_sriov_packet_alloc(struct xe_device *xe)
> +{
> + struct xe_sriov_packet *data;
> +
> + data = kzalloc(sizeof(*data), GFP_KERNEL);
> + if (!data)
> + return NULL;
> +
> + data->xe = xe;
> + data->hdr_remaining = sizeof(data->hdr);
> +
> + return data;
> +}
> +
> +/**
> + * xe_sriov_packet_free() - Free migration data packet.
> + * @data: the &xe_sriov_packet
> + */
> +void xe_sriov_packet_free(struct xe_sriov_packet *data)
> +{
> + if (IS_ERR_OR_NULL(data))
> + return;
> +
> + if (pkt_needs_bo(data))
> + xe_bo_unpin_map_no_vm(data->bo);
> + else
> + kvfree(data->buff);
> +
> + kfree(data);
> +}
> +
> +static int mig_pkt_init(struct xe_sriov_packet *data)
nit: no need for "mig_" prefix, see above pkt_needs_bo()
> +{
> + struct xe_gt *gt = xe_device_get_gt(data->xe, data->gt);
xe_device_get_gt() may return NULL and this will make static code analyzers
unhappy
either we should check here gt again, or maybe store the valid (struct xe_gt *)
in the data,
as we already have (struct xe_device *data.xe)
> +
> + if (data->size == 0)
> + return 0;
> +
> + if (pkt_needs_bo(data)) {
> + struct xe_bo *bo;
> +
> + bo = xe_bo_create_pin_map_novm(data->xe, gt->tile,
> PAGE_ALIGN(data->size),
> + ttm_bo_type_kernel,
> + XE_BO_FLAG_SYSTEM |
> XE_BO_FLAG_PINNED, false);
> + if (IS_ERR(bo))
> + return PTR_ERR(bo);
> +
> + data->bo = bo;
> + data->vaddr = bo->vmap.vaddr;
> + } else {
> + void *buff = kvzalloc(data->size, GFP_KERNEL);
> +
> + if (!buff)
> + return -ENOMEM;
> +
> + data->buff = buff;
> + data->vaddr = buff;
> + }
> +
> + return 0;
> +}
> +
> +#define XE_SRIOV_PACKET_SUPPORTED_VERSION 1
nit: add new line here
> +/**
> + * xe_sriov_packet_init() - Initialize migration packet header and backing
> storage.
> + * @data: the &xe_sriov_packet
> + * @tile_id: tile identifier
> + * @gt_id: GT identifier
> + * @type: &xe_sriov_packet_type
> + * @offset: offset of data packet payload (within wider resource)
> + * @size: size of data packet payload
> + *
> + * Return: 0 on success or a negative error code on failure.
> + */
> +int xe_sriov_packet_init(struct xe_sriov_packet *data, u8 tile_id, u8 gt_id,
> + enum xe_sriov_packet_type type, loff_t offset, size_t
> size)
> +{
> + data->version = XE_SRIOV_PACKET_SUPPORTED_VERSION;
> + data->type = type;
> + data->tile = tile_id;
> + data->gt = gt_id;
shouldn't we validate the tile_id / gt_id here?
we do have valid data->xe here, so we can try:
data->gt = xe_device_get_gt(data->xe, gt_id);
if (data->gt)
return -ENODEV;
data->hdr.gt_id = gt_id;
...
> + data->offset = offset;
> + data->size = size;
> + data->remaining = size;
> +
> + return mig_pkt_init(data);
> +}
> +
> +/**
> + * xe_sriov_packet_init_from_hdr() - Initialize migration packet backing
> storage based on header.
> + * @data: the &xe_sriov_packet
> + *
> + * Header data is expected to be filled prior to calling this function.
maybe add some asserts to enforce that and check data->hdr.size and other
fields?
> + *
> + * Return: 0 on success or a negative error code on failure.
> + */
> +int xe_sriov_packet_init_from_hdr(struct xe_sriov_packet *data)
> +{
> + if (data->version != XE_SRIOV_PACKET_SUPPORTED_VERSION)
> + return -EINVAL;
> +
> + data->remaining = data->size;
> +
> + return mig_pkt_init(data);
> +}
> diff --git a/drivers/gpu/drm/xe/xe_sriov_packet.h
> b/drivers/gpu/drm/xe/xe_sriov_packet.h
> new file mode 100644
> index 0000000000000..2c5a49897d763
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_sriov_packet.h
> @@ -0,0 +1,30 @@
> +/* SPDX-License-Identifier: MIT */
> +/*
> + * Copyright © 2025 Intel Corporation
> + */
> +
> +#ifndef _XE_SRIOV_PACKET_H_
> +#define _XE_SRIOV_PACKET_H_
> +
> +#include <linux/types.h>
> +
> +struct xe_device;
also
struct xe_sriov_packet;
> +
> +enum xe_sriov_packet_type {
> + /* Skipping 0 to catch uninitialized data */
> + XE_SRIOV_PACKET_TYPE_DESCRIPTOR = 1,
> + XE_SRIOV_PACKET_TYPE_TRAILER,
> + XE_SRIOV_PACKET_TYPE_GGTT,
> + XE_SRIOV_PACKET_TYPE_MMIO,
> + XE_SRIOV_PACKET_TYPE_GUC,
> + XE_SRIOV_PACKET_TYPE_VRAM,
> +};
shouldn't this enum be in xe_sriov_packet_types.h too ?
> +
> +struct xe_sriov_packet *xe_sriov_packet_alloc(struct xe_device *xe);
> +void xe_sriov_packet_free(struct xe_sriov_packet *data);
> +
> +int xe_sriov_packet_init(struct xe_sriov_packet *data, u8 tile_id, u8 gt_id,
> + enum xe_sriov_packet_type, loff_t offset, size_t size);
> +int xe_sriov_packet_init_from_hdr(struct xe_sriov_packet *data);
> +
> +#endif
