Re: [RFC PATCH 3/3] remoteproc: Add Renesas rcar driver

[Mathieu Poirier <mathieu.poirier@linaro.org>]

Good morning Julien,

On Wed, Oct 27, 2021 at 09:30:20AM +0200, Julien Massot wrote:
> Renesas Gen3 platform includes a Cortex-r7 processor.
> 
> Both: the application cores (A5x) and the realtime core (CR7)
> share access to the RAM and devices with the same address map,
> so device addresses are equal to the Linux physical addresses.
> 
> In order to initialize this remote processor we need to:
> - power on the realtime core
> - put the firmware in a ram area
> - set the boot address for this firmware (reset vector)
> - Deassert the reset
> 
> This initial driver allows to start and stop the Cortex R7
> processor.
> 
> Signed-off-by: Julien Massot <julien.massot@iot.bzh>
> ---
>  drivers/remoteproc/Kconfig      |  11 ++
>  drivers/remoteproc/Makefile     |   1 +
>  drivers/remoteproc/rcar_rproc.c | 229 ++++++++++++++++++++++++++++++++
>  3 files changed, 241 insertions(+)
>  create mode 100644 drivers/remoteproc/rcar_rproc.c
> 
> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
> index 9a6eedc3994a..3e87eadbaf59 100644
> --- a/drivers/remoteproc/Kconfig
> +++ b/drivers/remoteproc/Kconfig
> @@ -261,6 +261,17 @@ config QCOM_WCNSS_PIL
>  	  verified and booted with the help of the Peripheral Authentication
>  	  System (PAS) in TrustZone.
>  
> +config RCAR_REMOTEPROC
> +	tristate "Renesas RCAR remoteproc support"

It is probably a good idea to include the type of SoC being supported, something
like:

        tristate "Renesas Gen3 RCAR remoteproc support"

That will make it easier to support future RCAR processors that may not share
the same architecture.


> +	depends on ARCH_RENESAS
> +	depends on REMOTEPROC
> +	help
> +	  Say y here to support R-Car realtime processor via the
> +	  remote processor framework. An elf firmware can be loaded
> +	  thanks to sysfs remoteproc entries. The remote processor
> +	  can be started and stopped.
> +	  This can be either built-in or a loadable module.  

Please add the name of the module when compiled as such.

> +
>  config ST_REMOTEPROC
>  	tristate "ST remoteproc support"
>  	depends on ARCH_STI
> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
> index bb26c9e4ef9c..bb290cc08e99 100644
> --- a/drivers/remoteproc/Makefile
> +++ b/drivers/remoteproc/Makefile
> @@ -30,6 +30,7 @@ obj-$(CONFIG_QCOM_SYSMON)		+= qcom_sysmon.o
>  obj-$(CONFIG_QCOM_WCNSS_PIL)		+= qcom_wcnss_pil.o
>  qcom_wcnss_pil-y			+= qcom_wcnss.o
>  qcom_wcnss_pil-y			+= qcom_wcnss_iris.o
> +obj-$(CONFIG_RCAR_REMOTEPROC)		+= rcar_rproc.o
>  obj-$(CONFIG_ST_REMOTEPROC)		+= st_remoteproc.o
>  obj-$(CONFIG_ST_SLIM_REMOTEPROC)	+= st_slim_rproc.o
>  obj-$(CONFIG_STM32_RPROC)		+= stm32_rproc.o
> diff --git a/drivers/remoteproc/rcar_rproc.c b/drivers/remoteproc/rcar_rproc.c
> new file mode 100644
> index 000000000000..ae0bfc8d4e85
> --- /dev/null
> +++ b/drivers/remoteproc/rcar_rproc.c
> @@ -0,0 +1,229 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (C) IoT.bzh 2021
> + */
> +
> +#include <linux/module.h>
> +#include <linux/of_device.h>
> +#include <linux/of_reserved_mem.h>
> +#include <linux/pm_runtime.h>
> +#include <linux/remoteproc.h>
> +#include <linux/reset.h>
> +#include <linux/soc/renesas/rcar-rst.h>
> +
> +#include "remoteproc_internal.h"
> +
> +struct rcar_rproc {
> +	struct device			*dev;
> +	struct rproc			*rproc;
> +	struct reset_control            *rst;
> +};
> +
> +static int rcar_rproc_mem_alloc(struct rproc *rproc,
> +				 struct rproc_mem_entry *mem)
> +{
> +	struct device *dev = rproc->dev.parent;
> +	void *va;
> +
> +	dev_dbg(dev, "map memory: %p+%zx\n", &mem->dma, mem->len);

I think this should be "map memory: %pa+%lx\n" to be consistent with dev_err()
below and the original implementation in stm32_rproc.c.

> +	va = ioremap_wc(mem->dma, mem->len);
> +	if (IS_ERR_OR_NULL(va)) {
> +		dev_err(dev, "Unable to map memory region: %pa+%lx\n",
> +			&mem->dma, mem->len);
> +		return -ENOMEM;
> +	}
> +
> +	/* Update memory entry va */
> +	mem->va = va;
> +
> +	return 0;
> +}
> +
> +static int rcar_rproc_mem_release(struct rproc *rproc,
> +				   struct rproc_mem_entry *mem)
> +{
> +	dev_dbg(rproc->dev.parent, "unmap memory: %pa\n", &mem->dma);
> +	iounmap(mem->va);
> +
> +	return 0;
> +}
> +
> +static int rcar_rproc_prepare(struct rproc *rproc)
> +{
> +	struct device *dev = rproc->dev.parent;
> +	struct device_node *np = dev->of_node;
> +	struct of_phandle_iterator it;
> +	struct rproc_mem_entry *mem;
> +	struct reserved_mem *rmem;
> +	u64 da;
> +
> +	/* Register associated reserved memory regions */
> +	of_phandle_iterator_init(&it, np, "memory-region", NULL, 0);
> +	while (of_phandle_iterator_next(&it) == 0) {
> +
> +		rmem = of_reserved_mem_lookup(it.node);
> +		if (!rmem) {
> +			dev_err(dev, "unable to acquire memory-region\n");
> +			return -EINVAL;
> +		}
> +
> +		/* No need to translate pa to da, R-Car use same map */
> +		da = rmem->base;
> +
> +		mem = rproc_mem_entry_init(dev, NULL,
> +					   (dma_addr_t)rmem->base,
> +					   rmem->size, da,
> +					   rcar_rproc_mem_alloc,
> +					   rcar_rproc_mem_release,
> +					   it.node->name);
> +
> +		if (!mem)
> +			return -ENOMEM;
> +
> +		rproc_add_carveout(rproc, mem);
> +	}
> +
> +	return 0;
> +}
> +
> +static int rcar_rproc_parse_fw(struct rproc *rproc, const struct firmware *fw)
> +{
> +	int ret;
> +
> +	ret = rproc_elf_load_rsc_table(rproc, fw);
> +	if (ret)
> +		dev_info(&rproc->dev, "No resource table in elf\n");

In the above functions rproc->dev.parent is used for output.  I don't have a
strong opinion on which of rproc->dev or rproc->dev.parent is used but I would
like to see consistency throughout the driver.

> +
> +	return 0;
> +}
> +
> +static int rcar_rproc_start(struct rproc *rproc)
> +{
> +	struct rcar_rproc *priv = rproc->priv;
> +	int err;
> +
> +	if (!rproc->bootaddr)
> +		return -EINVAL;
> +
> +	err = rcar_rst_set_rproc_boot_addr(rproc->bootaddr);
> +	if (err) {
> +		dev_err(&rproc->dev, "failed to set rproc boot addr\n");

Same comment as above.

> +		return err;
> +	}
> +
> +	err = reset_control_deassert(priv->rst);
> +	if (err)
> +		dev_err(&rproc->dev, "failed to deassert reset\n");
> +
> +	return err;
> +}
> +
> +static int rcar_rproc_stop(struct rproc *rproc)
> +{
> +	struct rcar_rproc *priv = rproc->priv;
> +	int err;
> +
> +	err = reset_control_assert(priv->rst);
> +	if (err)
> +		dev_err(&rproc->dev, "failed to assert reset\n");
> +
> +	return err;
> +}
> +
> +static struct rproc_ops rcar_rproc_ops = {
> +	.prepare	= rcar_rproc_prepare,
> +	.start		= rcar_rproc_start,
> +	.stop		= rcar_rproc_stop,
> +	.load		= rproc_elf_load_segments,
> +	.parse_fw	= rcar_rproc_parse_fw,
> +	.find_loaded_rsc_table = rproc_elf_find_loaded_rsc_table,
> +	.sanity_check	= rproc_elf_sanity_check,
> +	.get_boot_addr	= rproc_elf_get_boot_addr,
> +
> +};
> +
> +static int rcar_rproc_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct device_node *np = dev->of_node;
> +	struct rcar_rproc *priv;
> +	struct rproc *rproc;
> +	int ret;
> +
> +	rproc = rproc_alloc(dev, np->name, &rcar_rproc_ops,
> +			    NULL, sizeof(*priv));
> +	if (!rproc)
> +		return -ENOMEM;
> +
> +	priv = rproc->priv;
> +	priv->rproc = rproc;

I don't see rcar_rproc::rproc being used anywhere.

> +	priv->dev = dev;
> +
> +	priv->rst = devm_reset_control_get_exclusive(&pdev->dev, NULL);
> +	if (IS_ERR(priv->rst)) {
> +		ret = PTR_ERR(priv->rst);
> +		dev_err(dev, "fail to acquire rproc reset\n");
> +		goto free_rproc;
> +	}
> +
> +	pm_runtime_enable(priv->dev);
> +	ret = pm_runtime_get_sync(priv->dev);

There is no dev_pm_ops for the platform driver nor clocks to manage for this
device - is there something that requires pm_runtime operations to be called?

> +	if (ret) {
> +		dev_err(&rproc->dev, "failed to power up\n");
> +		goto free_rproc;
> +	}
> +
> +	dev_set_drvdata(dev, rproc);
> +
> +	/* Manually start the rproc */
> +	rproc->auto_boot = false;
> +
> +	ret = rproc_add(rproc);
> +	if (ret) {
> +		dev_err(dev, "rproc_add failed\n");
> +		goto pm_disable;
> +	}
> +
> +	return 0;
> +
> +pm_disable:
> +	pm_runtime_disable(priv->dev);
> +free_rproc:
> +	rproc_free(rproc);
> +
> +	return ret;
> +}
> +
> +static int rcar_rproc_remove(struct platform_device *pdev)
> +{
> +	struct rproc *rproc = platform_get_drvdata(pdev);
> +	struct rcar_rproc *priv = rproc->priv;
> +
> +	rproc_del(rproc);
> +	pm_runtime_disable(priv->dev);

As far as I can tell rcar_rproc::dev is not required.  It is only used in
rproc_probe() and rproc_remove() where pdev->dev is available.

> +	rproc_free(rproc);
> +
> +	return 0;
> +}
> +
> +static const struct of_device_id rcar_rproc_of_match[] = {
> +	{ .compatible = "renesas,rcar-cr7" },
> +	{},
> +};
> +
> +MODULE_DEVICE_TABLE(of, rcar_rproc_of_match);
> +
> +static struct platform_driver rcar_rproc_driver = {
> +	.probe = rcar_rproc_probe,
> +	.remove = rcar_rproc_remove,
> +	.driver = {
> +		.name = "rcar-rproc",
> +		.of_match_table = rcar_rproc_of_match,
> +	},
> +};

Thanks,
Mathieu

> +
> +module_platform_driver(rcar_rproc_driver);
> +
> +MODULE_LICENSE("GPL v2");
> +MODULE_DESCRIPTION("Renesas Gen3 R-Car remote processor control driver");
> +MODULE_AUTHOR("Julien Massot <julien.massot@iot.bzh>");
> -- 
> 2.31.1
> 
>