[PATCH 1/4] remoteproc: k3-m4: Introduce K3 remote proc driver for M4 subsystem
Judith Mendez
jm at ti.com
Wed Jan 29 23:27:45 CET 2025
Hello Daniel,
On 1/29/25 7:24 AM, Daniel Schultz wrote:
> Hey,
>
> thanks for sending this driver! Tested it on our AM64x board. See two
> comments below.
>
> On 23.01.25 02:07, Judith Mendez wrote:
>> From: Hari Nagalla <hnagalla at ti.com>
>>
>> Some K3 devices like AM64, AM62 devices have a M4 processor in MCU
>> voltage domain.
>>
>> Add a remote proc driver to support this subsystem to be able to load
>> and boot the M4 core.
>>
>> Signed-off-by: Hari Nagalla <hnagalla at ti.com>
>> [Judith: Fixed comments explaining local resets]
>> Signed-off-by: Judith Mendez <jm at ti.com>
> Tested-by: Daniel Schultz <d.schultz at phytec.de>
>> ---
>> drivers/remoteproc/Kconfig | 10 +
>> drivers/remoteproc/Makefile | 1 +
>> drivers/remoteproc/ti_k3_m4_rproc.c | 411 ++++++++++++++++++++++++++++
>> 3 files changed, 422 insertions(+)
>> create mode 100644 drivers/remoteproc/ti_k3_m4_rproc.c
>>
>> diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
>> index 2790b168b19..e579c4e8371 100644
>> --- a/drivers/remoteproc/Kconfig
>> +++ b/drivers/remoteproc/Kconfig
>> @@ -80,6 +80,16 @@ config REMOTEPROC_TI_K3_R5F
>> on various TI K3 family of SoCs through the remote processor
>> framework.
>>
>> +config REMOTEPROC_TI_K3_M4F
>> + bool "TI K3 M4F remoteproc support"
>> + select REMOTEPROC
>> + depends on ARCH_K3
>> + depends on TI_SCI_PROTOCOL
>> + help
>> + Say y here to support TI's M4F remote processor subsystems
>> + on various TI K3 family of SoCs through the remote processor
>> + framework.
>> +
>
>
> this config is placed after REMOTEPROC_TI_K3_R5F and not ordered
> alphabetically ... not sure how important the order is in those Kconfig
> files.
I can fix it no problem. Not sure if it matters much but will fix
alphabetically.
>
>
>> config REMOTEPROC_TI_POWER
>> bool "Support for TI Power processor"
>> select REMOTEPROC
>> diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile
>> index 3a092b7660e..f81e5009c5e 100644
>> --- a/drivers/remoteproc/Makefile
>> +++ b/drivers/remoteproc/Makefile
>> @@ -13,6 +13,7 @@ obj-$(CONFIG_REMOTEPROC_SANDBOX) += sandbox_testproc.o
>> obj-$(CONFIG_REMOTEPROC_STM32_COPRO) += stm32_copro.o
>> obj-$(CONFIG_REMOTEPROC_TI_K3_ARM64) += ti_k3_arm64_rproc.o
>> obj-$(CONFIG_REMOTEPROC_TI_K3_DSP) += ti_k3_dsp_rproc.o
>> +obj-$(CONFIG_REMOTEPROC_TI_K3_M4F) += ti_k3_m4_rproc.o
>> obj-$(CONFIG_REMOTEPROC_TI_K3_R5F) += ti_k3_r5f_rproc.o
>> obj-$(CONFIG_REMOTEPROC_TI_POWER) += ti_power_proc.o
>> obj-$(CONFIG_REMOTEPROC_TI_PRU) += pru_rproc.o
>> diff --git a/drivers/remoteproc/ti_k3_m4_rproc.c b/drivers/remoteproc/ti_k3_m4_rproc.c
>> new file mode 100644
>> index 00000000000..5a5ae2fa81d
>> --- /dev/null
>> +++ b/drivers/remoteproc/ti_k3_m4_rproc.c
>> @@ -0,0 +1,411 @@
>> +// SPDX-License-Identifier: GPL-2.0+
>> +/*
>> + * Texas Instruments' K3 M4 Remoteproc driver
>> + *
>> + * Copyright (C) 2024 Texas Instruments Incorporated - http://www.ti.com/
>> + * Hari Nagalla <hnagalla at ti.com>
>> + */
>> +
>> +#include <dm.h>
>> +#include <log.h>
>> +#include <malloc.h>
>> +#include <remoteproc.h>
>> +#include <errno.h>
>> +#include <clk.h>
>> +#include <reset.h>
>> +#include <asm/io.h>
>> +#include <power-domain.h>
>> +#include <dm/device_compat.h>
>> +#include <linux/err.h>
>> +#include <linux/sizes.h>
>> +#include <linux/soc/ti/ti_sci_protocol.h>
>> +#include "ti_sci_proc.h"
>> +
>> +/**
>> + * struct k3_m4_mem - internal memory structure
>> + * @cpu_addr: MPU virtual address of the memory region
>> + * @bus_addr: Bus address used to access the memory region
>> + * @dev_addr: Device address from remoteproc view
>> + * @size: Size of the memory region
>> + */
>> +struct k3_m4_mem {
>> + void __iomem *cpu_addr;
>> + phys_addr_t bus_addr;
>> + phys_addr_t dev_addr;
>> + size_t size;
>> +};
>> +
>> +/**
>> + * struct k3_m4_mem_data - memory definitions for m4 remote core
>> + * @name: name for this memory entry
>> + * @dev_addr: device address for the memory entry
>> + */
>> +struct k3_m4_mem_data {
>> + const char *name;
>> + const u32 dev_addr;
>> +};
>> +
>> +/**
>> + * struct k3_m4_boot_data - internal data structure used for boot
>> + * @boot_align_addr: Boot vector address alignment granularity
>> + * @uses_lreset: Flag to denote the need for local reset management
>> + */
>> +struct k3_m4_boot_data {
>> + u32 boot_align_addr;
>> + bool uses_lreset;
>> +};
>> +
>> +/**
>> + * struct k3_m4_privdata - Structure representing Remote processor data.
>> + * @m4_rst: m4 rproc reset control data
>> + * @tsp: Pointer to TISCI proc contrl handle
>> + * @data: Pointer to DSP specific boot data structure
>> + * @mem: Array of available memories
>> + * @num_mem: Number of available memories
>> + */
>> +struct k3_m4_privdata {
>> + struct reset_ctl m4_rst;
>> + struct ti_sci_proc tsp;
>> + struct k3_m4_boot_data *data;
>> + struct k3_m4_mem *mem;
>> + int num_mems;
>> +};
>> +
>> +/*
>> + * The M4 cores have a local reset that affects only the CPU, and a
>> + * generic module reset that powers on the device and allows the M4 internal
>> + * memories to be accessed while the local reset is asserted. This function is
>> + * used to release the global reset on M4F to allow loading into the M4F
>> + * internal RAMs. This helper function is invoked in k3_m4_load() before any
>> + * actual firmware loading and is undone only in k3_m4_stop(). The local reset
>> + * cannot be released on M4 cores until after the firmware images are loaded.
>> + */
>> +static int k3_m4_prepare(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + struct k3_m4_boot_data *data = m4->data;
>> + int ret;
>> +
>> + if (!data->uses_lreset)
>> + return 0;
>> +
>> + ret = ti_sci_proc_power_domain_on(&m4->tsp);
>> + if (ret)
>> + dev_err(dev, "cannot enable internal RAM loading, ret = %d\n",
>> + ret);
>> +
>> + return ret;
>> +}
>> +
>> +/*
>> + * This function is the counterpart to k3_m4_prepare() and is used to assert
>> + * the global reset on M4 cores. This completes the second step of powering
>> + * down the M4 cores. The cores themselves are halted through the local reset
>> + * in first step. This function is invoked in k3_m4_stop() after the local
>> + * reset is asserted.
>> + */
>> +static int k3_m4_unprepare(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + struct k3_m4_boot_data *data = m4->data;
>> +
>> + if (!data->uses_lreset)
>> + return 0;
>> +
>> + return ti_sci_proc_power_domain_off(&m4->tsp);
>> +}
>> +
>> +/**
>> + * k3_m4_load() - Load up the Remote processor image
>> + * @dev: rproc device pointer
>> + * @addr: Address at which image is available
>> + * @size: size of the image
>> + *
>> + * Return: 0 if all goes good, else appropriate error message.
>> + */
>> +static int k3_m4_load(struct udevice *dev, ulong addr, ulong size)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + void *image_addr = (void *)addr;
>> + int ret;
>> +
>> + ret = ti_sci_proc_request(&m4->tsp);
>> + if (ret)
>> + return ret;
>> +
>> + ret = k3_m4_prepare(dev);
>> + if (ret) {
>> + dev_err(dev, "Prepare failed for core %d\n",
>> + m4->tsp.proc_id);
>> + goto proc_release;
>> + }
>> +
>> + ti_secure_image_post_process(&image_addr, &size);
>> +
>> + ret = rproc_elf_load_image(dev, addr, size);
>> + if (ret < 0) {
>> + dev_err(dev, "Loading elf failed %d\n", ret);
>> + goto unprepare;
>> + }
>> +
>> +unprepare:
>> + if (ret)
>> + k3_m4_unprepare(dev);
>> +proc_release:
>> + ti_sci_proc_release(&m4->tsp);
>> + return ret;
>> +}
>> +
>> +/**
>> + * k3_m4_start() - Start the remote processor
>> + * @dev: rproc device pointer
>> + *
>> + * Return: 0 if all went ok, else return appropriate error
>> + */
>> +static int k3_m4_start(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + struct k3_m4_boot_data *data = m4->data;
>> + int ret;
>> +
>> + ret = ti_sci_proc_request(&m4->tsp);
>> + if (ret)
>> + return ret;
>> +
>> + if (!data->uses_lreset) {
>> + ret = ti_sci_proc_power_domain_on(&m4->tsp);
>> + if (ret)
>> + goto proc_release;
>> + }
>> +
>> + ret = reset_deassert(&m4->m4_rst);
>> + if (ret) {
>> + if (!data->uses_lreset)
>> + ti_sci_proc_power_domain_off(&m4->tsp);
>> + }
>> +
>> +proc_release:
>> + ti_sci_proc_release(&m4->tsp);
>> +
>> + return ret;
>> +}
>> +
>> +static int k3_m4_stop(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> +
>> + ti_sci_proc_request(&m4->tsp);
>> + reset_assert(&m4->m4_rst);
>> + ti_sci_proc_power_domain_off(&m4->tsp);
>
>
> Is is safe to call ti_sci_prop_power_domain_off when data->uses_lreset
> is set? The start function has a check included.
uses_lreset here is just a flag used in this driver which is always set
actually, so it does not really do anything. I see how this flag can
cause some confusion, let me cleanup this flag from the driver since it
will always be set.
Thanks for reviewing!
~ Judith
>
> - Daniel
>
>> + ti_sci_proc_release(&m4->tsp);
>> +
>> + return 0;
>> +}
>> +
>> +/**
>> + * k3_m4_init() - Initialize the remote processor
>> + * @dev: rproc device pointer
>> + *
>> + * Return: 0 if all went ok, else return appropriate error
>> + */
>> +static int k3_m4_init(struct udevice *dev)
>> +{
>> + return 0;
>> +}
>> +
>> +static int k3_m4_reset(struct udevice *dev)
>> +{
>> + return 0;
>> +}
>> +
>> +static void *k3_m4_da_to_va(struct udevice *dev, ulong da, ulong len)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + phys_addr_t bus_addr, dev_addr;
>> + void __iomem *va = NULL;
>> + size_t size;
>> + u32 offset;
>> + int i;
>> +
>> + if (len <= 0)
>> + return NULL;
>> +
>> + for (i = 0; i < m4->num_mems; i++) {
>> + bus_addr = m4->mem[i].bus_addr;
>> + dev_addr = m4->mem[i].dev_addr;
>> + size = m4->mem[i].size;
>> +
>> + if (da >= dev_addr && ((da + len) <= (dev_addr + size))) {
>> + offset = da - dev_addr;
>> + va = m4->mem[i].cpu_addr + offset;
>> + return (__force void *)va;
>> + }
>> +
>> + if (da >= bus_addr && (da + len) <= (bus_addr + size)) {
>> + offset = da - bus_addr;
>> + va = m4->mem[i].cpu_addr + offset;
>> + return (__force void *)va;
>> + }
>> + }
>> +
>> + /* Assume it is DDR region and return da */
>> + return map_physmem(da, len, MAP_NOCACHE);
>> +}
>> +
>> +static const struct dm_rproc_ops k3_m4_ops = {
>> + .init = k3_m4_init,
>> + .load = k3_m4_load,
>> + .start = k3_m4_start,
>> + .stop = k3_m4_stop,
>> + .reset = k3_m4_reset,
>> + .device_to_virt = k3_m4_da_to_va,
>> +};
>> +
>> +static int ti_sci_proc_of_to_priv(struct udevice *dev, struct ti_sci_proc *tsp)
>> +{
>> + u32 ids[2];
>> + int ret;
>> +
>> + tsp->sci = ti_sci_get_by_phandle(dev, "ti,sci");
>> + if (IS_ERR(tsp->sci)) {
>> + dev_err(dev, "ti_sci get failed: %ld\n", PTR_ERR(tsp->sci));
>> + return PTR_ERR(tsp->sci);
>> + }
>> +
>> + ret = dev_read_u32_array(dev, "ti,sci-proc-ids", ids, 2);
>> + if (ret) {
>> + dev_err(dev, "Proc IDs not populated %d\n", ret);
>> + return ret;
>> + }
>> +
>> + tsp->ops = &tsp->sci->ops.proc_ops;
>> + tsp->proc_id = ids[0];
>> + tsp->host_id = ids[1];
>> + tsp->dev_id = dev_read_u32_default(dev, "ti,sci-dev-id",
>> + TI_SCI_RESOURCE_NULL);
>> + if (tsp->dev_id == TI_SCI_RESOURCE_NULL) {
>> + dev_err(dev, "Device ID not populated %d\n", ret);
>> + return -ENODEV;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static const struct k3_m4_mem_data am6_m4_mems[] = {
>> + { .name = "iram", .dev_addr = 0x0 },
>> + { .name = "dram", .dev_addr = 0x30000 },
>> +};
>> +
>> +static int k3_m4_of_get_memories(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> + int i;
>> +
>> + m4->num_mems = ARRAY_SIZE(am6_m4_mems);
>> + m4->mem = calloc(m4->num_mems, sizeof(*m4->mem));
>> + if (!m4->mem)
>> + return -ENOMEM;
>> +
>> + for (i = 0; i < m4->num_mems; i++) {
>> + m4->mem[i].bus_addr = dev_read_addr_size_name(dev,
>> + am6_m4_mems[i].name,
>> + (fdt_addr_t *)&m4->mem[i].size);
>> + if (m4->mem[i].bus_addr == FDT_ADDR_T_NONE) {
>> + dev_err(dev, "%s bus address not found\n",
>> + am6_m4_mems[i].name);
>> + return -EINVAL;
>> + }
>> + m4->mem[i].cpu_addr = map_physmem(m4->mem[i].bus_addr,
>> + m4->mem[i].size,
>> + MAP_NOCACHE);
>> + m4->mem[i].dev_addr = am6_m4_mems[i].dev_addr;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +/**
>> + * k3_of_to_priv() - generate private data from device tree
>> + * @dev: corresponding k3 m4 processor device
>> + * @m4: pointer to driver specific private data
>> + *
>> + * Return: 0 if all goes good, else appropriate error message.
>> + */
>> +static int k3_m4_of_to_priv(struct udevice *dev, struct k3_m4_privdata *m4)
>> +{
>> + int ret;
>> +
>> + ret = reset_get_by_index(dev, 0, &m4->m4_rst);
>> + if (ret) {
>> + dev_err(dev, "reset_get() failed: %d\n", ret);
>> + return ret;
>> + }
>> +
>> + ret = ti_sci_proc_of_to_priv(dev, &m4->tsp);
>> + if (ret)
>> + return ret;
>> +
>> + ret = k3_m4_of_get_memories(dev);
>> + if (ret)
>> + return ret;
>> +
>> + m4->data = (struct k3_m4_boot_data *)dev_get_driver_data(dev);
>> +
>> + return 0;
>> +}
>> +
>> +/**
>> + * k3_m4_probe() - Basic probe
>> + * @dev: corresponding k3 remote processor device
>> + *
>> + * Return: 0 if all goes good, else appropriate error message.
>> + */
>> +static int k3_m4_probe(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4;
>> + int ret;
>> +
>> + m4 = dev_get_priv(dev);
>> + ret = k3_m4_of_to_priv(dev, m4);
>> + if (ret)
>> + return ret;
>> +
>> + /*
>> + * The M4 local resets are deasserted by default on Power-On-Reset.
>> + * Assert the local resets to ensure the M4s don't execute bogus code
>> + * in .load() callback when the module reset is released to support
>> + * internal memory loading. This is needed for M4 cores.
>> + */
>> + reset_assert(&m4->m4_rst);
>> +
>> + return 0;
>> +}
>> +
>> +static int k3_m4_remove(struct udevice *dev)
>> +{
>> + struct k3_m4_privdata *m4 = dev_get_priv(dev);
>> +
>> + free(m4->mem);
>> +
>> + return 0;
>> +}
>> +
>> +static const struct k3_m4_boot_data m4_data = {
>> + .boot_align_addr = SZ_1K,
>> + .uses_lreset = true,
>> +};
>> +
>> +static const struct udevice_id k3_m4_ids[] = {
>> + { .compatible = "ti,am64-m4fss", .data = (ulong)&m4_data, },
>> + {}
>> +};
>> +
>> +U_BOOT_DRIVER(k3_m4) = {
>> + .name = "k3_m4",
>> + .of_match = k3_m4_ids,
>> + .id = UCLASS_REMOTEPROC,
>> + .ops = &k3_m4_ops,
>> + .probe = k3_m4_probe,
>> + .remove = k3_m4_remove,
>> + .priv_auto = sizeof(struct k3_m4_privdata),
>> +};
More information about the U-Boot
mailing list