[PATCH 06/12] accel/ivpu: Implement support for hardware scheduler
Jacek Lawrynowicz
jacek.lawrynowicz at linux.intel.com
Wed May 8 13:21:03 UTC 2024
From: "Wachowski, Karol" <karol.wachowski at intel.com>
Add support for HWS (hardware scheduler). It is disabled by default.
The sched_mode module param can be used to enable it.
Each context has multiple command queues with different priorities and
HWS enables priority based execution on the HW/FW side.
The driver in HWS mode has to send a couple additional messages to
initialize HWS and describe command queue priorities.
Signed-off-by: Wachowski, Karol <karol.wachowski at intel.com>
Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz at linux.intel.com>
---
drivers/accel/ivpu/ivpu_drv.c | 20 ++++-
drivers/accel/ivpu/ivpu_fw.c | 9 ++
drivers/accel/ivpu/ivpu_job.c | 162 ++++++++++++++++++++++++----------
3 files changed, 144 insertions(+), 47 deletions(-)
diff --git a/drivers/accel/ivpu/ivpu_drv.c b/drivers/accel/ivpu/ivpu_drv.c
index db47e7ef6322..49261fa7c5f4 100644
--- a/drivers/accel/ivpu/ivpu_drv.c
+++ b/drivers/accel/ivpu/ivpu_drv.c
@@ -78,7 +78,6 @@ static void file_priv_unbind(struct ivpu_device *vdev, struct ivpu_file_priv *fi
ivpu_dbg(vdev, FILE, "file_priv unbind: ctx %u\n", file_priv->ctx.id);
ivpu_cmdq_release_all_locked(file_priv);
- ivpu_jsm_context_release(vdev, file_priv->ctx.id);
ivpu_bo_unbind_all_bos_from_context(vdev, &file_priv->ctx);
ivpu_mmu_user_context_fini(vdev, &file_priv->ctx);
file_priv->bound = false;
@@ -327,6 +326,21 @@ static int ivpu_wait_for_ready(struct ivpu_device *vdev)
return ret;
}
+static int ivpu_hw_sched_init(struct ivpu_device *vdev)
+{
+ int ret = 0;
+
+ if (vdev->hw->sched_mode == VPU_SCHEDULING_MODE_HW) {
+ ret = ivpu_jsm_hws_setup_priority_bands(vdev);
+ if (ret) {
+ ivpu_err(vdev, "Failed to enable hw scheduler: %d", ret);
+ return ret;
+ }
+ }
+
+ return ret;
+}
+
/**
* ivpu_boot() - Start VPU firmware
* @vdev: VPU device
@@ -360,6 +374,10 @@ int ivpu_boot(struct ivpu_device *vdev)
enable_irq(vdev->irq);
ivpu_hw_irq_enable(vdev);
ivpu_ipc_enable(vdev);
+
+ if (ivpu_fw_is_cold_boot(vdev))
+ return ivpu_hw_sched_init(vdev);
+
return 0;
}
diff --git a/drivers/accel/ivpu/ivpu_fw.c b/drivers/accel/ivpu/ivpu_fw.c
index 29ecf7db238b..6a33a3193d5c 100644
--- a/drivers/accel/ivpu/ivpu_fw.c
+++ b/drivers/accel/ivpu/ivpu_fw.c
@@ -44,6 +44,8 @@
#define IVPU_FW_CHECK_API_VER_LT(vdev, fw_hdr, name, major, minor) \
ivpu_fw_check_api_ver_lt(vdev, fw_hdr, #name, VPU_##name##_API_VER_INDEX, major, minor)
+#define IVPU_FOCUS_PRESENT_TIMER_MS 1000
+
static char *ivpu_firmware;
module_param_named_unsafe(firmware, ivpu_firmware, charp, 0644);
MODULE_PARM_DESC(firmware, "NPU firmware binary in /lib/firmware/..");
@@ -467,6 +469,10 @@ static void ivpu_fw_boot_params_print(struct ivpu_device *vdev, struct vpu_boot_
boot_params->punit_telemetry_sram_size);
ivpu_dbg(vdev, FW_BOOT, "boot_params.vpu_telemetry_enable = 0x%x\n",
boot_params->vpu_telemetry_enable);
+ ivpu_dbg(vdev, FW_BOOT, "boot_params.vpu_scheduling_mode = 0x%x\n",
+ boot_params->vpu_scheduling_mode);
+ ivpu_dbg(vdev, FW_BOOT, "boot_params.vpu_focus_present_timer_ms = %u\n",
+ boot_params->vpu_focus_present_timer_ms);
ivpu_dbg(vdev, FW_BOOT, "boot_params.dvfs_mode = %u\n",
boot_params->dvfs_mode);
ivpu_dbg(vdev, FW_BOOT, "boot_params.d0i3_delayed_entry = %d\n",
@@ -567,6 +573,9 @@ void ivpu_fw_boot_params_setup(struct ivpu_device *vdev, struct vpu_boot_params
boot_params->punit_telemetry_sram_base = ivpu_hw_reg_telemetry_offset_get(vdev);
boot_params->punit_telemetry_sram_size = ivpu_hw_reg_telemetry_size_get(vdev);
boot_params->vpu_telemetry_enable = ivpu_hw_reg_telemetry_enable_get(vdev);
+ boot_params->vpu_scheduling_mode = vdev->hw->sched_mode;
+ if (vdev->hw->sched_mode == VPU_SCHEDULING_MODE_HW)
+ boot_params->vpu_focus_present_timer_ms = IVPU_FOCUS_PRESENT_TIMER_MS;
boot_params->dvfs_mode = vdev->fw->dvfs_mode;
if (!IVPU_WA(disable_d0i3_msg))
boot_params->d0i3_delayed_entry = 1;
diff --git a/drivers/accel/ivpu/ivpu_job.c b/drivers/accel/ivpu/ivpu_job.c
index 3ef9d8022c9c..1d7b4388eb3b 100644
--- a/drivers/accel/ivpu/ivpu_job.c
+++ b/drivers/accel/ivpu/ivpu_job.c
@@ -77,11 +77,10 @@ static void ivpu_preemption_buffers_free(struct ivpu_device *vdev,
ivpu_bo_free(cmdq->secondary_preempt_buf);
}
-static struct ivpu_cmdq *ivpu_cmdq_alloc(struct ivpu_file_priv *file_priv, u16 engine)
+static struct ivpu_cmdq *ivpu_cmdq_alloc(struct ivpu_file_priv *file_priv)
{
struct xa_limit db_xa_limit = {.max = IVPU_MAX_DB, .min = IVPU_MIN_DB};
struct ivpu_device *vdev = file_priv->vdev;
- struct vpu_job_queue_header *jobq_header;
struct ivpu_cmdq *cmdq;
int ret;
@@ -103,16 +102,6 @@ static struct ivpu_cmdq *ivpu_cmdq_alloc(struct ivpu_file_priv *file_priv, u16 e
if (ret)
goto err_free_cmdq_mem;
- cmdq->entry_count = (u32)((ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header)) /
- sizeof(struct vpu_job_queue_entry));
-
- cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
- jobq_header = &cmdq->jobq->header;
- jobq_header->engine_idx = engine;
- jobq_header->head = 0;
- jobq_header->tail = 0;
- wmb(); /* Flush WC buffer for jobq->header */
-
return cmdq;
err_free_cmdq_mem:
@@ -135,33 +124,126 @@ static void ivpu_cmdq_free(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *c
kfree(cmdq);
}
+static int ivpu_hws_cmdq_init(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq, u16 engine,
+ u8 priority)
+{
+ struct ivpu_device *vdev = file_priv->vdev;
+ int ret;
+
+ ret = ivpu_jsm_hws_create_cmdq(vdev, file_priv->ctx.id, file_priv->ctx.id, cmdq->db_id,
+ task_pid_nr(current), engine,
+ cmdq->mem->vpu_addr, ivpu_bo_size(cmdq->mem));
+ if (ret)
+ return ret;
+
+ ret = ivpu_jsm_hws_set_context_sched_properties(vdev, file_priv->ctx.id, cmdq->db_id,
+ priority);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int ivpu_register_db(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
+{
+ struct ivpu_device *vdev = file_priv->vdev;
+ int ret;
+
+ if (vdev->hw->sched_mode == VPU_SCHEDULING_MODE_HW)
+ ret = ivpu_jsm_hws_register_db(vdev, file_priv->ctx.id, cmdq->db_id, cmdq->db_id,
+ cmdq->mem->vpu_addr, ivpu_bo_size(cmdq->mem));
+ else
+ ret = ivpu_jsm_register_db(vdev, file_priv->ctx.id, cmdq->db_id,
+ cmdq->mem->vpu_addr, ivpu_bo_size(cmdq->mem));
+
+ if (!ret)
+ ivpu_dbg(vdev, JOB, "DB %d registered to ctx %d\n", cmdq->db_id, file_priv->ctx.id);
+
+ return ret;
+}
+
+static int
+ivpu_cmdq_init(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq, u16 engine, u8 priority)
+{
+ struct ivpu_device *vdev = file_priv->vdev;
+ struct vpu_job_queue_header *jobq_header;
+ int ret;
+
+ lockdep_assert_held(&file_priv->lock);
+
+ if (cmdq->db_registered)
+ return 0;
+
+ cmdq->entry_count = (u32)((ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header)) /
+ sizeof(struct vpu_job_queue_entry));
+
+ cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
+ jobq_header = &cmdq->jobq->header;
+ jobq_header->engine_idx = engine;
+ jobq_header->head = 0;
+ jobq_header->tail = 0;
+ wmb(); /* Flush WC buffer for jobq->header */
+
+ if (vdev->hw->sched_mode == VPU_SCHEDULING_MODE_HW) {
+ ret = ivpu_hws_cmdq_init(file_priv, cmdq, engine, priority);
+ if (ret)
+ return ret;
+ }
+
+ ret = ivpu_register_db(file_priv, cmdq);
+ if (ret)
+ return ret;
+
+ cmdq->db_registered = true;
+
+ return 0;
+}
+
+static int ivpu_cmdq_fini(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
+{
+ struct ivpu_device *vdev = file_priv->vdev;
+ int ret;
+
+ lockdep_assert_held(&file_priv->lock);
+
+ if (!cmdq->db_registered)
+ return 0;
+
+ cmdq->db_registered = false;
+
+ if (vdev->hw->sched_mode == VPU_SCHEDULING_MODE_HW) {
+ ret = ivpu_jsm_hws_destroy_cmdq(vdev, file_priv->ctx.id, cmdq->db_id);
+ if (!ret)
+ ivpu_dbg(vdev, JOB, "Command queue %d destroyed\n", cmdq->db_id);
+ }
+
+ ret = ivpu_jsm_unregister_db(vdev, cmdq->db_id);
+ if (!ret)
+ ivpu_dbg(vdev, JOB, "DB %d unregistered\n", cmdq->db_id);
+
+ return 0;
+}
+
static struct ivpu_cmdq *ivpu_cmdq_acquire(struct ivpu_file_priv *file_priv, u16 engine,
u8 priority)
{
int cmdq_idx = IVPU_CMDQ_INDEX(engine, priority);
struct ivpu_cmdq *cmdq = file_priv->cmdq[cmdq_idx];
- struct ivpu_device *vdev = file_priv->vdev;
int ret;
lockdep_assert_held(&file_priv->lock);
if (!cmdq) {
- cmdq = ivpu_cmdq_alloc(file_priv, engine);
+ cmdq = ivpu_cmdq_alloc(file_priv);
if (!cmdq)
return NULL;
file_priv->cmdq[cmdq_idx] = cmdq;
}
- if (cmdq->db_registered)
- return cmdq;
-
- ret = ivpu_jsm_register_db(vdev, file_priv->ctx.id, cmdq->db_id,
- cmdq->mem->vpu_addr, ivpu_bo_size(cmdq->mem));
+ ret = ivpu_cmdq_init(file_priv, cmdq, engine, priority);
if (ret)
return NULL;
- cmdq->db_registered = true;
-
return cmdq;
}
@@ -174,9 +256,7 @@ static void ivpu_cmdq_release_locked(struct ivpu_file_priv *file_priv, u16 engin
if (cmdq) {
file_priv->cmdq[cmdq_idx] = NULL;
- if (cmdq->db_registered)
- ivpu_jsm_unregister_db(file_priv->vdev, cmdq->db_id);
-
+ ivpu_cmdq_fini(file_priv, cmdq);
ivpu_cmdq_free(file_priv, cmdq);
}
}
@@ -194,36 +274,27 @@ void ivpu_cmdq_release_all_locked(struct ivpu_file_priv *file_priv)
}
/*
- * Mark the doorbell as unregistered and reset job queue pointers.
+ * Mark the doorbell as unregistered
* This function needs to be called when the VPU hardware is restarted
* and FW loses job queue state. The next time job queue is used it
* will be registered again.
*/
-static void ivpu_cmdq_reset_locked(struct ivpu_file_priv *file_priv, u16 engine, u8 priority)
-{
- int cmdq_idx = IVPU_CMDQ_INDEX(engine, priority);
- struct ivpu_cmdq *cmdq = file_priv->cmdq[cmdq_idx];
-
- lockdep_assert_held(&file_priv->lock);
-
- if (cmdq) {
- cmdq->db_registered = false;
- cmdq->jobq->header.head = 0;
- cmdq->jobq->header.tail = 0;
- wmb(); /* Flush WC buffer for jobq header */
- }
-}
-
-static void ivpu_cmdq_reset_all(struct ivpu_file_priv *file_priv)
+static void ivpu_cmdq_reset(struct ivpu_file_priv *file_priv)
{
u16 engine;
u8 priority;
mutex_lock(&file_priv->lock);
- for (engine = 0; engine < IVPU_NUM_ENGINES; engine++)
- for (priority = 0; priority < IVPU_NUM_PRIORITIES; priority++)
- ivpu_cmdq_reset_locked(file_priv, engine, priority);
+ for (engine = 0; engine < IVPU_NUM_ENGINES; engine++) {
+ for (priority = 0; priority < IVPU_NUM_PRIORITIES; priority++) {
+ int cmdq_idx = IVPU_CMDQ_INDEX(engine, priority);
+ struct ivpu_cmdq *cmdq = file_priv->cmdq[cmdq_idx];
+
+ if (cmdq)
+ cmdq->db_registered = false;
+ }
+ }
mutex_unlock(&file_priv->lock);
}
@@ -236,10 +307,9 @@ void ivpu_cmdq_reset_all_contexts(struct ivpu_device *vdev)
mutex_lock(&vdev->context_list_lock);
xa_for_each(&vdev->context_xa, ctx_id, file_priv)
- ivpu_cmdq_reset_all(file_priv);
+ ivpu_cmdq_reset(file_priv);
mutex_unlock(&vdev->context_list_lock);
-
}
static int ivpu_cmdq_push_job(struct ivpu_cmdq *cmdq, struct ivpu_job *job)
--
2.43.2
More information about the dri-devel
mailing list