From 1de8e0029f59be740b4a02a44e694fab06798081 Mon Sep 17 00:00:00 2001
From: Tomeu Vizoso <tomeu@tomeuvizoso.net>
Date: Fri, 14 Feb 2025 17:55:09 +0100
Subject: [PATCH] fixup: multiple runtime PM
---
drivers/accel/rocket/rocket_core.c | 17 +++++-
drivers/accel/rocket/rocket_device.h | 5 --
drivers/accel/rocket/rocket_drv.c | 80 +++++++---------------------
drivers/accel/rocket/rocket_gem.c | 70 +++++++++++++++---------
drivers/accel/rocket/rocket_job.c | 38 ++++---------
drivers/accel/rocket/rocket_job.h | 2 +-
6 files changed, 89 insertions(+), 123 deletions(-)
diff --git a/drivers/accel/rocket/rocket_core.c b/drivers/accel/rocket/rocket_core.c
index 37447f34dbb98..705ed78e24c55 100644
--- a/drivers/accel/rocket/rocket_core.c
+++ b/drivers/accel/rocket/rocket_core.c
@@ -63,10 +63,9 @@ int rocket_core_init(struct rocket_core *core)
{
struct device *dev = core->dev;
struct platform_device *pdev = to_platform_device(dev);
+ uint32_t version;
int err = 0;
- printk("*** %s: core %px %d\n", __func__, core, core->index);
-
err = rocket_clk_init(core);
if (err) {
dev_err(dev, "clk init failed %d\n", err);
@@ -87,6 +86,20 @@ int rocket_core_init(struct rocket_core *core)
if (err)
return err;
+ pm_runtime_use_autosuspend(dev);
+ pm_runtime_set_autosuspend_delay(dev, 50); /* ~3 frames */
+ pm_runtime_enable(dev);
+
+ err = pm_runtime_get_sync(dev);
+
+ version = rocket_read(core, REG_PC_VERSION);
+ version += rocket_read(core, REG_PC_VERSION_NUM) & 0xffff;
+
+ pm_runtime_mark_last_busy(dev);
+ pm_runtime_put_autosuspend(dev);
+
+ dev_info(dev, "Rockchip NPU core %d version: %d\n", core->index, version);
+
return 0;
}
diff --git a/drivers/accel/rocket/rocket_device.h b/drivers/accel/rocket/rocket_device.h
index 86da2484cc2db..f7e0bfd2a7a3c 100644
--- a/drivers/accel/rocket/rocket_device.h
+++ b/drivers/accel/rocket/rocket_device.h
@@ -22,9 +22,4 @@ int rocket_device_init(struct rocket_device *rdev);
void rocket_device_fini(struct rocket_device *rdev);
void rocket_device_reset(struct rocket_device *rdev);
-static inline struct device *get_top_core_device(struct rocket_device *rdev)
-{
- return rdev->cores[0].dev;
-}
-
#endif
diff --git a/drivers/accel/rocket/rocket_drv.c b/drivers/accel/rocket/rocket_drv.c
index f0ca0417c96a1..e66150630e534 100644
--- a/drivers/accel/rocket/rocket_drv.c
+++ b/drivers/accel/rocket/rocket_drv.c
@@ -6,7 +6,6 @@
#include "linux/clk.h"
#include "linux/component.h"
#include "linux/dma-mapping.h"
-#include "linux/printk.h"
#include "rocket_registers.h"
#include <linux/of.h>
#include <linux/module.h>
@@ -146,45 +145,10 @@ static int rocket_drm_init(struct device *dev, bool with_components)
goto err_drm_dev;
}
- pm_runtime_use_autosuspend(dev);
- pm_runtime_set_autosuspend_delay(dev, 50); /* ~3 frames */
- pm_runtime_enable(dev);
-
- printk("*** %s: %d dev %s usage_count %d links_count %d\n", __func__, 98, dev_name(dev), atomic_read(&dev->power.usage_count), dev->power.links_count);
-
- printk("*** %s: %d dev %s usage_count %d links_count %d\n", __func__, 99, dev_name(dev), atomic_read(&dev->power.usage_count), dev->power.links_count);
- printk("*** %s: dev %s dev->power.runtime_error %d\n", __func__, dev_name(dev), dev->power.runtime_error);
- printk("*** %s: dev %s dev->power.usage_count %d\n", __func__, dev_name(dev), atomic_read(&dev->power.usage_count));
- printk("*** %s: dev %s dev->power.ignore_children %d\n", __func__, dev_name(dev), dev->power.ignore_children);
- printk("*** %s: dev %s dev->power.child_count %d\n", __func__, dev_name(dev), atomic_read(&dev->power.child_count));
- printk("*** %s: dev %s dev->power.deferred_resume %d\n", __func__, dev_name(dev), dev->power.deferred_resume);
- printk("*** %s: dev %s dev->power.runtime_status %d\n", __func__, dev_name(dev), dev->power.runtime_status);
- printk("*** %s: dev %s dev->power.request_pending %d\n", __func__, dev_name(dev), dev->power.request_pending);
- printk("*** %s: dev %s dev->power.request %d\n", __func__, dev_name(dev), dev->power.request);
-
- /*
- * Register the DRM device with the core and the connectors with
- * sysfs
- */
err = drm_dev_register(ddev, 0);
if (err < 0)
goto err_pm_runtime;
- printk("*** %s: before reading\n", __func__);
- err = pm_runtime_get_sync(dev);
-
- for (unsigned int core = 0; core < rdev->num_cores; core++) {
- uint32_t version;
-
- version = rocket_read(&rdev->cores[core], REG_PC_VERSION);
- version += rocket_read(&rdev->cores[core], REG_PC_VERSION_NUM) & 0xffff;
-
- dev_info(dev, "Rockchip NPU core %d version: %d\n", core, version);
- }
-
- pm_runtime_mark_last_busy(dev);
- pm_runtime_put_autosuspend(dev);
-
return 0;
err_pm_runtime:
@@ -234,8 +198,8 @@ static int rocket_core_bind(struct device *dev, struct device *master, void *dat
rdev->cores[core].rdev = rdev;
rdev->cores[core].dev = dev;
rdev->cores[core].index = core;
- // rdev->cores[core].link = device_link_add(dev, rdev->cores[0].dev,
- // DL_FLAG_STATELESS);
+ rdev->cores[core].link = device_link_add(dev, rdev->cores[0].dev,
+ DL_FLAG_STATELESS);
rdev->num_cores++;
@@ -255,7 +219,7 @@ static void rocket_core_unbind(struct device *dev, struct device *master, void *
for (unsigned int core = 0; core < rdev->num_cores; core++) {
if (rdev->cores[core].dev == dev) {
rocket_core_fini(&rdev->cores[core]);
- //device_link_del(rdev->cores[core].link);
+ device_link_del(rdev->cores[core].link);
break;
}
}
@@ -321,23 +285,19 @@ static int rocket_device_runtime_resume(struct device *dev)
{
struct rocket_device *rdev = dev_get_drvdata(dev);
- printk("*** %s: %d dev %s\n", __func__, 1, dev_name(dev));
-
- clk_prepare_enable(rdev->clk_npu);
- clk_prepare_enable(rdev->pclk);
-
for (unsigned int core = 0; core < rdev->num_cores; core++) {
- printk("*** %s: %d core %d\n", __func__, 2, core);
+ if (dev != rdev->cores[core].dev)
+ continue;
- struct iommu_domain *domain = iommu_get_domain_for_dev(rdev->cores[core].dev);
- iommu_attach_device(domain, rdev->cores[core].dev);
+ if (core == 0) {
+ clk_prepare_enable(rdev->clk_npu);
+ clk_prepare_enable(rdev->pclk);
+ }
clk_prepare_enable(rdev->cores[core].a_clk);
clk_prepare_enable(rdev->cores[core].h_clk);
}
- printk("*** %s: %d dev %s\n", __func__, 99, dev_name(dev));
-
return 0;
}
@@ -345,26 +305,22 @@ static int rocket_device_runtime_suspend(struct device *dev)
{
struct rocket_device *rdev = dev_get_drvdata(dev);
- printk("*** %s: %d dev %s\n", __func__, 1, dev_name(dev));
-
- if (!rocket_job_is_idle(rdev))
- return -EBUSY;
-
for (unsigned int core = 0; core < rdev->num_cores; core++) {
- printk("*** %s: %d core %d\n", __func__, 2, core);
+ if (dev != rdev->cores[core].dev)
+ continue;
+
+ if (!rocket_job_is_idle(&rdev->cores[core]))
+ return -EBUSY;
clk_disable_unprepare(rdev->cores[core].a_clk);
clk_disable_unprepare(rdev->cores[core].h_clk);
- struct iommu_domain *domain = iommu_get_domain_for_dev(rdev->cores[core].dev);
- iommu_detach_device(domain, rdev->cores[core].dev);
+ if (core == 0) {
+ clk_disable_unprepare(rdev->pclk);
+ clk_disable_unprepare(rdev->clk_npu);
+ }
}
- clk_disable_unprepare(rdev->pclk);
- clk_disable_unprepare(rdev->clk_npu);
-
- printk("*** %s: %d dev %s\n", __func__, 99, dev_name(dev));
-
return 0;
}
diff --git a/drivers/accel/rocket/rocket_gem.c b/drivers/accel/rocket/rocket_gem.c
index 41a94206bd5d0..b83dbe4c0e0d8 100644
--- a/drivers/accel/rocket/rocket_gem.c
+++ b/drivers/accel/rocket/rocket_gem.c
@@ -9,6 +9,34 @@
#include "rocket_gem.h"
#include "rocket_device.h"
+static void rocket_gem_bo_free(struct drm_gem_object *obj)
+{
+ struct rocket_device *rdev = obj->dev->dev_private;
+ struct rocket_gem_object *bo = to_rocket_bo(obj);
+
+ drm_WARN_ON(obj->dev, bo->base.pages_use_count > 1);
+
+ /* Unmap this object from the IOMMUs for cores > 0 */
+ for (unsigned int core = 1; core < rdev->num_cores; core++)
+ dma_unmap_sgtable(rdev->cores[core].dev, bo->base.sgt,
+ DMA_BIDIRECTIONAL, 0);
+
+ /* This will unmap the pages from the IOMMU linked to core 0*/
+ drm_gem_shmem_free(&bo->base);
+}
+
+static const struct drm_gem_object_funcs rocket_gem_funcs = {
+ .free = rocket_gem_bo_free,
+ .print_info = drm_gem_shmem_object_print_info,
+ .pin = drm_gem_shmem_object_pin,
+ .unpin = drm_gem_shmem_object_unpin,
+ .get_sg_table = drm_gem_shmem_object_get_sg_table,
+ .vmap = drm_gem_shmem_object_vmap,
+ .vunmap = drm_gem_shmem_object_vunmap,
+ .mmap = drm_gem_shmem_object_mmap,
+ .vm_ops = &drm_gem_shmem_vm_ops,
+};
+
/**
* rocket_gem_create_object - Implementation of driver->gem_create_object.
* @dev: DRM device
@@ -25,6 +53,8 @@ struct drm_gem_object *rocket_gem_create_object(struct drm_device *dev, size_t s
if (!obj)
return ERR_PTR(-ENOMEM);
+ obj->base.base.funcs = &rocket_gem_funcs;
+
return &obj->base.base;
}
@@ -38,8 +68,6 @@ int rocket_ioctl_create_bo(struct drm_device *dev, void *data, struct drm_file *
struct sg_table *sgt;
int ret;
- printk("*** %s: %d\n", __func__, 1);
-
shmem_obj = drm_gem_shmem_create(dev, args->size);
if (IS_ERR(shmem_obj))
return PTR_ERR(shmem_obj);
@@ -56,22 +84,23 @@ int rocket_ioctl_create_bo(struct drm_device *dev, void *data, struct drm_file *
if (ret)
goto err;
- printk("*** %s: %d before drm_gem_shmem_get_pages_sgt\n", __func__, 1);
+ /* This will map the pages to the IOMMU linked to core 0 */
sgt = drm_gem_shmem_get_pages_sgt(shmem_obj);
if (IS_ERR(sgt)) {
ret = PTR_ERR(sgt);
goto err;
}
- ret = dma_map_sgtable(rdev->cores[1].dev, sgt, DMA_BIDIRECTIONAL, 0);
- ret = dma_map_sgtable(rdev->cores[2].dev, sgt, DMA_BIDIRECTIONAL, 0);
- printk("*** %s: %d after drm_gem_shmem_get_pages_sgt\n", __func__, 1);
+ /* Map the pages to the IOMMUs linked to the other cores, so all cores can access this BO */
+ for (unsigned int core = 1; core < rdev->num_cores; core++) {
+ ret = dma_map_sgtable(rdev->cores[core].dev, sgt, DMA_BIDIRECTIONAL, 0);
+ if (ret)
+ goto err; /* TODO: Cleanup? */
+ }
args->offset = drm_vma_node_offset_addr(&gem_obj->vma_node);
args->dma_address = sg_dma_address(shmem_obj->sgt->sgl);
- printk("*** %s: %d\n", __func__, 99);
-
return 0;
err:
@@ -100,8 +129,6 @@ int rocket_ioctl_prep_bo(struct drm_device *dev, void *data, struct drm_file *fi
bool write = !!(args->op & ROCKET_PREP_WRITE);
long ret = 0;
- printk("*** %s: %d\n", __func__, 1);
-
if (args->op & ~(ROCKET_PREP_READ | ROCKET_PREP_WRITE))
return -EINVAL;
@@ -116,16 +143,15 @@ int rocket_ioctl_prep_bo(struct drm_device *dev, void *data, struct drm_file *fi
shmem_obj = &to_rocket_bo(gem_obj)->base;
- dma_sync_sgtable_for_cpu(rdev->cores[0].dev, shmem_obj->sgt, rocket_op_to_dma_dir(args->op));
- dma_sync_sgtable_for_cpu(rdev->cores[1].dev, shmem_obj->sgt, rocket_op_to_dma_dir(args->op));
- dma_sync_sgtable_for_cpu(rdev->cores[2].dev, shmem_obj->sgt, rocket_op_to_dma_dir(args->op));
+ for (unsigned int core = 1; core < rdev->num_cores; core++) {
+ dma_sync_sgtable_for_cpu(rdev->cores[core].dev, shmem_obj->sgt,
+ rocket_op_to_dma_dir(args->op));
+ }
to_rocket_bo(gem_obj)->last_cpu_prep_op = args->op;
drm_gem_object_put(gem_obj);
- printk("*** %s: %d\n", __func__, 99);
-
return ret;
}
@@ -137,8 +163,6 @@ int rocket_ioctl_fini_bo(struct drm_device *dev, void *data, struct drm_file *fi
struct drm_gem_shmem_object *shmem_obj;
struct rocket_device *rdev = dev->dev_private;
- printk("*** %s: %d\n", __func__, 1);
-
gem_obj = drm_gem_object_lookup(file, args->handle);
if (!gem_obj)
return -ENOENT;
@@ -148,18 +172,14 @@ int rocket_ioctl_fini_bo(struct drm_device *dev, void *data, struct drm_file *fi
WARN_ON(rkt_obj->last_cpu_prep_op == 0);
- dma_sync_sgtable_for_device(rdev->cores[0].dev, shmem_obj->sgt,
- rocket_op_to_dma_dir(rkt_obj->last_cpu_prep_op));
- dma_sync_sgtable_for_device(rdev->cores[1].dev, shmem_obj->sgt,
- rocket_op_to_dma_dir(rkt_obj->last_cpu_prep_op));
- dma_sync_sgtable_for_device(rdev->cores[2].dev, shmem_obj->sgt,
- rocket_op_to_dma_dir(rkt_obj->last_cpu_prep_op));
+ for (unsigned int core = 1; core < rdev->num_cores; core++) {
+ dma_sync_sgtable_for_device(rdev->cores[core].dev, shmem_obj->sgt,
+ rocket_op_to_dma_dir(rkt_obj->last_cpu_prep_op));
+ }
rkt_obj->last_cpu_prep_op = 0;
drm_gem_object_put(gem_obj);
- printk("*** %s: %d\n", __func__, 99);
-
return 0;
}
diff --git a/drivers/accel/rocket/rocket_job.c b/drivers/accel/rocket/rocket_job.c
index d37668fa590f4..817886c5026f5 100644
--- a/drivers/accel/rocket/rocket_job.c
+++ b/drivers/accel/rocket/rocket_job.c
@@ -304,8 +304,6 @@ static struct dma_fence *rocket_job_run(struct drm_sched_job *sched_job)
struct dma_fence *fence = NULL;
int ret;
- printk("*** %s: %d dev %s job %px\n", __func__, 1, dev_name(core->dev), job);
-
if (unlikely(job->base.s_fence->finished.error))
return NULL;
@@ -323,7 +321,7 @@ static struct dma_fence *rocket_job_run(struct drm_sched_job *sched_job)
dma_fence_put(job->done_fence);
job->done_fence = dma_fence_get(fence);
- ret = pm_runtime_get_sync(get_top_core_device(rdev));
+ ret = pm_runtime_get_sync(core->dev);
if (ret < 0)
return fence;
@@ -340,10 +338,6 @@ static struct dma_fence *rocket_job_run(struct drm_sched_job *sched_job)
static void rocket_job_handle_done(struct rocket_core *core,
struct rocket_job *job)
{
- struct rocket_device *rdev = core->rdev;
-
- printk("*** %s: %d dev %s job %px\n", __func__, 1, dev_name(core->dev), job);
-
if (job->next_task_idx < job->task_count) {
rocket_job_hw_submit(core, job);
return;
@@ -351,16 +345,14 @@ static void rocket_job_handle_done(struct rocket_core *core,
core->in_flight_job = NULL;
dma_fence_signal_locked(job->done_fence);
- pm_runtime_put_autosuspend(get_top_core_device(rdev));
+ pm_runtime_put_autosuspend(core->dev);
}
static void rocket_job_handle_irq(struct rocket_core *core)
{
uint32_t status, raw_status;
- printk("*** %s: %d dev %s\n", __func__, 1, dev_name(core->dev));
-
- pm_runtime_mark_last_busy(get_top_core_device(core->rdev));
+ pm_runtime_mark_last_busy(core->dev);
status = rocket_read(core, REG_PC_INTERRUPT_STATUS);
raw_status = rocket_read(core, REG_PC_INTERRUPT_RAW_STATUS);
@@ -422,14 +414,14 @@ rocket_reset(struct rocket_core *core, struct drm_sched_job *bad)
*/
spin_lock(&core->job_lock);
if (core->in_flight_job)
- pm_runtime_put_noidle(get_top_core_device(core->rdev));
+ pm_runtime_put_noidle(core->dev);
core->in_flight_job = NULL;
spin_unlock(&core->job_lock);
/* Proceed with reset now. */
- pm_runtime_force_suspend(get_top_core_device(core->rdev));
- pm_runtime_force_resume(get_top_core_device(core->rdev));
+ pm_runtime_force_suspend(core->dev);
+ pm_runtime_force_resume(core->dev);
/* GPU has been reset, we can clear the reset pending bit. */
atomic_set(&core->reset.pending, 0);
@@ -457,8 +449,6 @@ static enum drm_gpu_sched_stat rocket_job_timedout(struct drm_sched_job *sched_j
struct rocket_device *rdev = job->rdev;
struct rocket_core *core = sched_to_core(rdev, sched_job->sched);
- printk("*** %s: %d dev %s job %px\n", __func__, 1, dev_name(core->dev), job);
-
/*
* If the GPU managed to complete this jobs fence, the timeout is
* spurious. Bail out.
@@ -538,7 +528,6 @@ int rocket_job_init(struct rocket_core *core)
spin_lock_init(&core->job_lock);
core->irq = platform_get_irq_byname(to_platform_device(core->dev), "npu_irq");
- printk("*** %s: %d dev %s core->irq %d\n", __func__, 1, dev_name(core->dev), core->irq);
if (core->irq < 0)
return core->irq;
@@ -615,15 +604,11 @@ void rocket_job_close(struct rocket_file_priv *rocket_priv)
drm_sched_entity_destroy(entity);
}
-int rocket_job_is_idle(struct rocket_device *rdev)
+int rocket_job_is_idle(struct rocket_core *core)
{
- unsigned int core;
-
- for (core = 0; core < rdev->num_cores; core++) {
- /* If there are any jobs in any HW queue, we're not idle */
- if (atomic_read(&rdev->cores[core].sched.credit_count))
- return false;
- }
+ /* If there are any jobs in this HW queue, we're not idle */
+ if (atomic_read(&core->sched.credit_count))
+ return false;
return true;
}
@@ -635,9 +620,6 @@ static int rocket_ioctl_submit_job(struct drm_device *dev, struct drm_file *file
struct rocket_job *rjob = NULL;
int ret = 0;
- udelay(1000);
- printk("*** %s: %d dev %s\n", __func__, 1, dev_name(dev->dev));
-
if (job->task_count == 0)
return -EINVAL;
diff --git a/drivers/accel/rocket/rocket_job.h b/drivers/accel/rocket/rocket_job.h
index 0c3c90e47d391..2302d376f4761 100644
--- a/drivers/accel/rocket/rocket_job.h
+++ b/drivers/accel/rocket/rocket_job.h
@@ -44,6 +44,6 @@ int rocket_job_init(struct rocket_core *core);
void rocket_job_fini(struct rocket_core *core);
int rocket_job_open(struct rocket_file_priv *rocket_priv);
void rocket_job_close(struct rocket_file_priv *rocket_priv);
-int rocket_job_is_idle(struct rocket_device *rdev);
+int rocket_job_is_idle(struct rocket_core *core);
#endif
\ No newline at end of file
--
GitLab