Introduce a new parameter size that accepts elf section size to remoteproc ops callback device_to_virt(). This can enforce more checks on the elf section that is being loaded.

Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- drivers/remoteproc/rproc-elf-loader.c | 3 ++- drivers/remoteproc/sandbox_testproc.c | 4 +++- drivers/remoteproc/stm32_copro.c | 4 +++- include/remoteproc.h | 3 ++- 4 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/remoteproc/rproc-elf-loader.c b/drivers/remoteproc/rproc-elf-loader.c index 67937a7595..7574ba3fb3 100644 --- a/drivers/remoteproc/rproc-elf-loader.c +++ b/drivers/remoteproc/rproc-elf-loader.c @@ -86,7 +86,8 @@ int rproc_elf32_load_image(struct udevice *dev, unsigned long addr) continue;

if (ops->device_to_virt) - dst = ops->device_to_virt(dev, (ulong)dst); + dst = ops->device_to_virt(dev, (ulong)dst, + phdr->p_memsz);

dev_dbg(dev, "Loading phdr %i to 0x%p (%i bytes)\n", i, dst, phdr->p_filesz); diff --git a/drivers/remoteproc/sandbox_testproc.c b/drivers/remoteproc/sandbox_testproc.c index 5f35119ab7..0a8f7871e1 100644 --- a/drivers/remoteproc/sandbox_testproc.c +++ b/drivers/remoteproc/sandbox_testproc.c @@ -306,9 +306,11 @@ static int sandbox_testproc_ping(struct udevice *dev) * sandbox_testproc_device_to_virt() - Convert device address to virtual address * @dev: device to operate upon * @da: device address + * @size: Size of the section * @return converted virtual address */ -static void *sandbox_testproc_device_to_virt(struct udevice *dev, ulong da) +static void *sandbox_testproc_device_to_virt(struct udevice *dev, ulong da, + ulong size) { u64 paddr;

diff --git a/drivers/remoteproc/stm32_copro.c b/drivers/remoteproc/stm32_copro.c index de3b9729f3..ff82313bca 100644 --- a/drivers/remoteproc/stm32_copro.c +++ b/drivers/remoteproc/stm32_copro.c @@ -107,9 +107,11 @@ static int stm32_copro_set_hold_boot(struct udevice *dev, bool hold) * stm32_copro_device_to_virt() - Convert device address to virtual address * @dev: corresponding STM32 remote processor device * @da: device address + * @size: Size of the section * @return converted virtual address */ -static void *stm32_copro_device_to_virt(struct udevice *dev, ulong da) +static void *stm32_copro_device_to_virt(struct udevice *dev, ulong da, + ulong size) { fdt32_t in_addr = cpu_to_be32(da); u64 paddr; diff --git a/include/remoteproc.h b/include/remoteproc.h index 4987194905..81f616a1f9 100644 --- a/include/remoteproc.h +++ b/include/remoteproc.h @@ -122,9 +122,10 @@ struct dm_rproc_ops { * * @dev: Remote proc device * @da: Device address + * @size: Size of the section * @return virtual address. */ - void * (*device_to_virt)(struct udevice *dev, ulong da); + void * (*device_to_virt)(struct udevice *dev, ulong da, ulong size); };

/* Accessor */

Power domain for the remote cores needs to be handled in a right sequence as mandated by the spec. Introduce tisci helper apis that can control power-domains of remote cores. TISCI clients can use this api and control the remote cores power domain instead of hooking it to power-domain layer.

Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- drivers/remoteproc/ti_sci_proc.h | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)

diff --git a/drivers/remoteproc/ti_sci_proc.h b/drivers/remoteproc/ti_sci_proc.h index ccfc39ec88..f8299d1aff 100644 --- a/drivers/remoteproc/ti_sci_proc.h +++ b/drivers/remoteproc/ti_sci_proc.h @@ -19,12 +19,14 @@ * @proc_id: processor id for the consumer remoteproc device * @host_id: host id to pass the control over for this consumer remoteproc * device + * @dev_id: Device ID as identified by system controller. */ struct ti_sci_proc { const struct ti_sci_handle *sci; const struct ti_sci_proc_ops *ops; u8 proc_id; u8 host_id; + u16 dev_id; };

static inline int ti_sci_proc_request(struct ti_sci_proc *tsp) @@ -118,4 +120,29 @@ static inline int ti_sci_proc_set_control(struct ti_sci_proc *tsp, return ret; }

+static inline int ti_sci_proc_power_domain_on(struct ti_sci_proc *tsp) +{ + int ret; + + debug("%s: dev_id = %d\n", __func__, tsp->dev_id); + + ret = tsp->sci->ops.dev_ops.get_device_exclusive(tsp->sci, tsp->dev_id); + if (ret) + pr_err("Power-domain on failed for dev = %d\n", tsp->dev_id); + + return ret; +} + +static inline int ti_sci_proc_power_domain_off(struct ti_sci_proc *tsp) +{ + int ret; + + debug("%s: dev_id = %d\n", __func__, tsp->dev_id); + + ret = tsp->sci->ops.dev_ops.put_device(tsp->sci, tsp->dev_id); + if (ret) + pr_err("Power-domain off failed for dev = %d\n", tsp->dev_id); + + return ret; +} #endif /* REMOTEPROC_TI_SCI_PROC_H */

SoCs with K3 architecture have an integrated Arm Cortex-R5F subsystem that is comprised of dual-core Arm Cortex-R5F processor cores. This R5 subsytem can be configured at boot time to be either run in a LockStep mode or in an Asymmetric Multi Processing (AMP) fashion in Split-mode. This subsystem has each Tightly-Coupled Memory (TCM) internal memories for each core split between two banks - TCMA and TCMB.

Add a remoteproc driver to support this subsystem to be able to load and boot the R5 cores primarily in LockStep mode or split mode.

Signed-off-by: Lokesh Vutla lokeshvutla@ti.com Signed-off-by: Suman Anna s-anna@ti.com --- drivers/remoteproc/Kconfig | 10 + drivers/remoteproc/Makefile | 1 + drivers/remoteproc/ti_k3_r5f_rproc.c | 816 +++++++++++++++++++++++++++ 3 files changed, 827 insertions(+) create mode 100644 drivers/remoteproc/ti_k3_r5f_rproc.c

diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig index f54a245424..c2d59ba000 100644 --- a/drivers/remoteproc/Kconfig +++ b/drivers/remoteproc/Kconfig @@ -52,6 +52,16 @@ config REMOTEPROC_TI_K3_ARM64 on various TI K3 family of SoCs through the remote processor framework.

+config REMOTEPROC_TI_K3_R5F + bool "TI K3 R5F remoteproc support" + select REMOTEPROC + depends on ARCH_K3 + depends on TI_SCI_PROTOCOL + help + Say y here to support TI's R5F remote processor subsystems + on various TI K3 family of SoCs through the remote processor + framework. + config REMOTEPROC_TI_POWER bool "Support for TI Power processor" select REMOTEPROC diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile index 271ba55b09..9d247ba5e7 100644 --- a/drivers/remoteproc/Makefile +++ b/drivers/remoteproc/Makefile @@ -11,4 +11,5 @@ obj-$(CONFIG_K3_SYSTEM_CONTROLLER) += k3_system_controller.o obj-$(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-$(CONFIG_REMOTEPROC_TI_K3_R5F) += ti_k3_r5f_rproc.o obj-$(CONFIG_REMOTEPROC_TI_POWER) += ti_power_proc.o diff --git a/drivers/remoteproc/ti_k3_r5f_rproc.c b/drivers/remoteproc/ti_k3_r5f_rproc.c new file mode 100644 index 0000000000..ae1e4b9e04 --- /dev/null +++ b/drivers/remoteproc/ti_k3_r5f_rproc.c @@ -0,0 +1,816 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Texas Instruments' K3 R5 Remoteproc driver + * + * Copyright (C) 2018-2019 Texas Instruments Incorporated - http://www.ti.com/ + * Lokesh Vutla lokeshvutla@ti.com + */ + +#include <common.h> +#include <dm.h> +#include <remoteproc.h> +#include <errno.h> +#include <clk.h> +#include <reset.h> +#include <asm/io.h> +#include <linux/kernel.h> +#include <linux/soc/ti/ti_sci_protocol.h> +#include "ti_sci_proc.h" + +/* + * R5F's view of this address can either be for ATCM or BTCM with the other + * at address 0x0 based on loczrama signal. + */ +#define K3_R5_TCM_DEV_ADDR 0x41010000 + +/* R5 TI-SCI Processor Configuration Flags */ +#define PROC_BOOT_CFG_FLAG_R5_DBG_EN 0x00000001 +#define PROC_BOOT_CFG_FLAG_R5_DBG_NIDEN 0x00000002 +#define PROC_BOOT_CFG_FLAG_R5_LOCKSTEP 0x00000100 +#define PROC_BOOT_CFG_FLAG_R5_TEINIT 0x00000200 +#define PROC_BOOT_CFG_FLAG_R5_NMFI_EN 0x00000400 +#define PROC_BOOT_CFG_FLAG_R5_TCM_RSTBASE 0x00000800 +#define PROC_BOOT_CFG_FLAG_R5_BTCM_EN 0x00001000 +#define PROC_BOOT_CFG_FLAG_R5_ATCM_EN 0x00002000 +#define PROC_BOOT_CFG_FLAG_GEN_IGN_BOOTVECTOR 0x10000000 + +/* R5 TI-SCI Processor Control Flags */ +#define PROC_BOOT_CTRL_FLAG_R5_CORE_HALT 0x00000001 + +/* R5 TI-SCI Processor Status Flags */ +#define PROC_BOOT_STATUS_FLAG_R5_WFE 0x00000001 +#define PROC_BOOT_STATUS_FLAG_R5_WFI 0x00000002 +#define PROC_BOOT_STATUS_FLAG_R5_CLK_GATED 0x00000004 +#define PROC_BOOT_STATUS_FLAG_R5_LOCKSTEP_PERMITTED 0x00000100 + +#define NR_CORES 2 + +enum cluster_mode { + CLUSTER_MODE_SPLIT = 0, + CLUSTER_MODE_LOCKSTEP, +}; + +/** + * struct k3_r5_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_r5f_mem { + void __iomem *cpu_addr; + phys_addr_t bus_addr; + u32 dev_addr; + size_t size; +}; + +/** + * struct k3_r5f_core - K3 R5 core structure + * @dev: cached device pointer + * @cluster: pointer to the parent cluster. + * @reset: reset control handle + * @tsp: TI-SCI processor control handle + * @mem: Array of available internal memories + * @num_mem: Number of available memories + * @atcm_enable: flag to control ATCM enablement + * @btcm_enable: flag to control BTCM enablement + * @loczrama: flag to dictate which TCM is at device address 0x0 + * @in_use: flag to tell if the core is already in use. + */ +struct k3_r5f_core { + struct udevice *dev; + struct k3_r5f_cluster *cluster; + struct reset_ctl reset; + struct ti_sci_proc tsp; + struct k3_r5f_mem *mem; + int num_mems; + u32 atcm_enable; + u32 btcm_enable; + u32 loczrama; + bool in_use; +}; + +/** + * struct k3_r5f_cluster - K3 R5F Cluster structure + * @mode: Mode to configure the Cluster - Split or LockStep + * @cores: Array of pointers to R5 cores within the cluster + */ +struct k3_r5f_cluster { + enum cluster_mode mode; + struct k3_r5f_core *cores[NR_CORES]; +}; + +static bool is_primary_core(struct k3_r5f_core *core) +{ + return core == core->cluster->cores[0]; +} + +static int k3_r5f_proc_request(struct k3_r5f_core *core) +{ + struct k3_r5f_cluster *cluster = core->cluster; + int i, ret; + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) { + for (i = 0; i < NR_CORES; i++) { + ret = ti_sci_proc_request(&cluster->cores[i]->tsp); + if (ret) + goto proc_release; + } + } else { + ret = ti_sci_proc_request(&core->tsp); + } + + return 0; + +proc_release: + while (i >= 0) { + ti_sci_proc_release(&cluster->cores[i]->tsp); + i--; + } + return ret; +} + +static void k3_r5f_proc_release(struct k3_r5f_core *core) +{ + struct k3_r5f_cluster *cluster = core->cluster; + int i; + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) + for (i = 0; i < NR_CORES; i++) + ti_sci_proc_release(&cluster->cores[i]->tsp); + else + ti_sci_proc_release(&core->tsp); +} + +static int k3_r5f_lockstep_release(struct k3_r5f_cluster *cluster) +{ + int ret, c; + + dev_dbg(dev, "%s\n", __func__); + + for (c = NR_CORES - 1; c >= 0; c--) { + ret = ti_sci_proc_power_domain_on(&cluster->cores[c]->tsp); + if (ret) + goto unroll_module_reset; + } + + /* deassert local reset on all applicable cores */ + for (c = NR_CORES - 1; c >= 0; c--) { + ret = reset_deassert(&cluster->cores[c]->reset); + if (ret) + goto unroll_local_reset; + } + + return 0; + +unroll_local_reset: + while (c < NR_CORES) { + reset_assert(&cluster->cores[c]->reset); + c++; + } + c = 0; +unroll_module_reset: + while (c < NR_CORES) { + ti_sci_proc_power_domain_off(&cluster->cores[c]->tsp); + c++; + } + + return ret; +} + +static int k3_r5f_split_release(struct k3_r5f_core *core) +{ + int ret; + + dev_dbg(dev, "%s\n", __func__); + + ret = ti_sci_proc_power_domain_on(&core->tsp); + if (ret) { + dev_err(core->dev, "module-reset deassert failed, ret = %d\n", + ret); + return ret; + } + + ret = reset_deassert(&core->reset); + if (ret) { + dev_err(core->dev, "local-reset deassert failed, ret = %d\n", + ret); + if (ti_sci_proc_power_domain_off(&core->tsp)) + dev_warn(core->dev, "module-reset assert back failed\n"); + } + + return ret; +} + +static int k3_r5f_prepare(struct udevice *dev) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + struct k3_r5f_cluster *cluster = core->cluster; + int ret = 0; + + dev_dbg(dev, "%s\n", __func__); + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) + ret = k3_r5f_lockstep_release(cluster); + else + ret = k3_r5f_split_release(core); + + if (ret) + dev_err(dev, "Unable to enable cores for TCM loading %d\n", + ret); + + return ret; +} + +static int k3_r5f_core_sanity_check(struct k3_r5f_core *core) +{ + struct k3_r5f_cluster *cluster = core->cluster; + + if (core->in_use) { + dev_err(dev, "Invalid op: Trying to load/start on already running core %d\n", + core->tsp.proc_id); + return -EINVAL; + } + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP && !cluster->cores[1]) { + printf("Secondary core is not probed in this cluster\n"); + return -EAGAIN; + } + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP && !is_primary_core(core)) { + dev_err(dev, "Invalid op: Trying to start secondary core %d in lockstep mode\n", + core->tsp.proc_id); + return -EINVAL; + } + + if (cluster->mode == CLUSTER_MODE_SPLIT && !is_primary_core(core)) { + if (!core->cluster->cores[0]->in_use) { + dev_err(dev, "Invalid seq: Enable primary core before loading secondary core\n"); + return -EINVAL; + } + } + + return 0; +} + +/** + * k3_r5f_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_r5f_load(struct udevice *dev, ulong addr, ulong size) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + u32 boot_vector; + int ret; + + dev_dbg(dev, "%s addr = 0x%lx, size = 0x%lx\n", __func__, addr, size); + + ret = k3_r5f_core_sanity_check(core); + if (ret) + return ret; + + ret = k3_r5f_proc_request(core); + if (ret) + return ret; + + ret = k3_r5f_prepare(dev); + if (ret) { + dev_err(dev, "R5f prepare failed for core %d\n", + core->tsp.proc_id); + goto proc_release; + } + + /* Zero out TCMs so that ECC can be effective on all TCM addresses */ + if (core->atcm_enable) + memset(core->mem[0].cpu_addr, 0x00, core->mem[0].size); + if (core->btcm_enable) + memset(core->mem[1].cpu_addr, 0x00, core->mem[1].size); + + ret = rproc_elf_load_image(dev, addr, size); + if (ret < 0) { + dev_err(dev, "Loading elf failedi %d\n", ret); + goto proc_release; + } + + boot_vector = rproc_elf_get_boot_addr(dev, addr); + + dev_dbg(dev, "%s: Boot vector = 0x%x\n", __func__, boot_vector); + + ret = ti_sci_proc_set_config(&core->tsp, boot_vector, 0, 0); + +proc_release: + k3_r5f_proc_release(core); + + return ret; +} + +static int k3_r5f_core_halt(struct k3_r5f_core *core) +{ + int ret; + + ret = ti_sci_proc_set_control(&core->tsp, + PROC_BOOT_CTRL_FLAG_R5_CORE_HALT, 0); + if (ret) + dev_err(core->dev, "Core %d failed to stop\n", + core->tsp.proc_id); + + return ret; +} + +static int k3_r5f_core_run(struct k3_r5f_core *core) +{ + int ret; + + ret = ti_sci_proc_set_control(&core->tsp, + 0, PROC_BOOT_CTRL_FLAG_R5_CORE_HALT); + if (ret) { + dev_err(core->dev, "Core %d failed to start\n", + core->tsp.proc_id); + return ret; + } + + return 0; +} + +/** + * k3_r5f_start() - Start the remote processor + * @dev: rproc device pointer + * + * Return: 0 if all went ok, else return appropriate error + */ +static int k3_r5f_start(struct udevice *dev) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + struct k3_r5f_cluster *cluster = core->cluster; + int ret, c; + + dev_dbg(dev, "%s\n", __func__); + + ret = k3_r5f_core_sanity_check(core); + if (ret) + return ret; + + ret = k3_r5f_proc_request(core); + if (ret) + return ret; + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) { + if (is_primary_core(core)) { + for (c = NR_CORES - 1; c >= 0; c--) { + ret = k3_r5f_core_run(cluster->cores[c]); + if (ret) + goto unroll_core_run; + } + } else { + dev_err(dev, "Invalid op: Trying to start secondary core %d in lockstep mode\n", + core->tsp.proc_id); + ret = -EINVAL; + goto proc_release; + } + } else { + ret = k3_r5f_core_run(core); + if (ret) + goto proc_release; + } + + core->in_use = true; + + k3_r5f_proc_release(core); + return 0; + +unroll_core_run: + while (c < NR_CORES) { + k3_r5f_core_halt(cluster->cores[c]); + c++; + } +proc_release: + k3_r5f_proc_release(core); + + return ret; +} + +static int k3_r5f_split_reset(struct k3_r5f_core *core) +{ + int ret; + + dev_dbg(dev, "%s\n", __func__); + + if (reset_assert(&core->reset)) + ret = -EINVAL; + + if (ti_sci_proc_power_domain_off(&core->tsp)) + ret = -EINVAL; + + return ret; +} + +static int k3_r5f_lockstep_reset(struct k3_r5f_cluster *cluster) +{ + int ret = 0, c; + + dev_dbg(dev, "%s\n", __func__); + + for (c = 0; c < NR_CORES; c++) + if (reset_assert(&cluster->cores[c]->reset)) + ret = -EINVAL; + + /* disable PSC modules on all applicable cores */ + for (c = 0; c < NR_CORES; c++) + if (ti_sci_proc_power_domain_off(&cluster->cores[c]->tsp)) + ret = -EINVAL; + + return ret; +} + +static int k3_r5f_unprepare(struct udevice *dev) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + struct k3_r5f_cluster *cluster = core->cluster; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) { + if (is_primary_core(core)) + ret = k3_r5f_lockstep_reset(cluster); + } else { + ret = k3_r5f_split_reset(core); + } + + if (ret) + dev_warn(dev, "Unable to enable cores for TCM loading %d\n", + ret); + + return 0; +} + +static int k3_r5f_stop(struct udevice *dev) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + struct k3_r5f_cluster *cluster = core->cluster; + int c, ret; + + dev_dbg(dev, "%s\n", __func__); + + ret = k3_r5f_proc_request(core); + if (ret) + return ret; + + core->in_use = false; + + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) { + if (is_primary_core(core)) { + for (c = 0; c < NR_CORES; c++) + k3_r5f_core_halt(cluster->cores[c]); + } else { + dev_err(dev, "Invalid op: Trying to stop secondary core in lockstep mode\n"); + ret = -EINVAL; + goto proc_release; + } + } else { + k3_r5f_core_halt(core); + } + + ret = k3_r5f_unprepare(dev); +proc_release: + k3_r5f_proc_release(core); + return ret; +} + +static void *k3_r5f_da_to_va(struct udevice *dev, ulong da, ulong size) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + void __iomem *va = NULL; + phys_addr_t bus_addr; + u32 dev_addr, offset; + ulong mem_size; + int i; + + dev_dbg(dev, "%s\n", __func__); + + if (size <= 0) + return NULL; + + for (i = 0; i < core->num_mems; i++) { + bus_addr = core->mem[i].bus_addr; + dev_addr = core->mem[i].dev_addr; + mem_size = core->mem[i].size; + + if (da >= bus_addr && (da + size) <= (bus_addr + mem_size)) { + offset = da - bus_addr; + va = core->mem[i].cpu_addr + offset; + return (__force void *)va; + } + + if (da >= dev_addr && (da + size) <= (dev_addr + mem_size)) { + offset = da - dev_addr; + va = core->mem[i].cpu_addr + offset; + return (__force void *)va; + } + } + + /* Assume it is DDR region and return da */ + return map_physmem(da, size, MAP_NOCACHE); +} + +static int k3_r5f_init(struct udevice *dev) +{ + return 0; +} + +static int k3_r5f_reset(struct udevice *dev) +{ + return 0; +} + +static const struct dm_rproc_ops k3_r5f_rproc_ops = { + .init = k3_r5f_init, + .reset = k3_r5f_reset, + .start = k3_r5f_start, + .stop = k3_r5f_stop, + .load = k3_r5f_load, + .device_to_virt = k3_r5f_da_to_va, +}; + +static int k3_r5f_rproc_configure(struct k3_r5f_core *core) +{ + struct k3_r5f_cluster *cluster = core->cluster; + u32 set_cfg = 0, clr_cfg = 0, cfg, ctrl, sts; + u64 boot_vec = 0; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + ret = ti_sci_proc_request(&core->tsp); + if (ret < 0) + return ret; + + /* Do not touch boot vector now. Load will take care of it. */ + clr_cfg |= PROC_BOOT_CFG_FLAG_GEN_IGN_BOOTVECTOR; + + ret = ti_sci_proc_get_status(&core->tsp, &boot_vec, &cfg, &ctrl, &sts); + if (ret) + goto out; + + /* Sanity check for Lockstep mode */ + if (cluster->mode && is_primary_core(core) && + !(sts & PROC_BOOT_STATUS_FLAG_R5_LOCKSTEP_PERMITTED)) { + dev_err(core->dev, "LockStep mode not permitted on this device\n"); + ret = -EINVAL; + goto out; + } + + /* Primary core only configuration */ + if (is_primary_core(core)) { + /* always enable ARM mode */ + clr_cfg |= PROC_BOOT_CFG_FLAG_R5_TEINIT; + if (cluster->mode == CLUSTER_MODE_LOCKSTEP) + set_cfg |= PROC_BOOT_CFG_FLAG_R5_LOCKSTEP; + else + clr_cfg |= PROC_BOOT_CFG_FLAG_R5_LOCKSTEP; + } + + if (core->atcm_enable) + set_cfg |= PROC_BOOT_CFG_FLAG_R5_ATCM_EN; + else + clr_cfg |= PROC_BOOT_CFG_FLAG_R5_ATCM_EN; + + if (core->btcm_enable) + set_cfg |= PROC_BOOT_CFG_FLAG_R5_BTCM_EN; + else + clr_cfg |= PROC_BOOT_CFG_FLAG_R5_BTCM_EN; + + if (core->loczrama) + set_cfg |= PROC_BOOT_CFG_FLAG_R5_TCM_RSTBASE; + else + clr_cfg |= PROC_BOOT_CFG_FLAG_R5_TCM_RSTBASE; + + ret = k3_r5f_core_halt(core); + if (ret) + goto out; + + ret = ti_sci_proc_set_config(&core->tsp, boot_vec, set_cfg, clr_cfg); +out: + ti_sci_proc_release(&core->tsp); + return ret; +} + +static int ti_sci_proc_of_to_priv(struct udevice *dev, struct ti_sci_proc *tsp) +{ + u32 ids[2]; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + 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 int k3_r5f_of_to_priv(struct k3_r5f_core *core) +{ + int ret; + + dev_dbg(dev, "%s\n", __func__); + + core->atcm_enable = dev_read_u32_default(core->dev, "atcm-enable", 0); + core->btcm_enable = dev_read_u32_default(core->dev, "btcm-enable", 1); + core->loczrama = dev_read_u32_default(core->dev, "loczrama", 1); + + ret = ti_sci_proc_of_to_priv(core->dev, &core->tsp); + if (ret) + return ret; + + ret = reset_get_by_index(core->dev, 0, &core->reset); + if (ret) { + dev_err(core->dev, "Reset lines not available: %d\n", ret); + return ret; + } + + return 0; +} + +static int k3_r5f_core_of_get_memories(struct k3_r5f_core *core) +{ + static const char * const mem_names[] = {"atcm", "btcm"}; + struct udevice *dev = core->dev; + int i; + + dev_dbg(dev, "%s\n", __func__); + + core->num_mems = ARRAY_SIZE(mem_names); + core->mem = calloc(core->num_mems, sizeof(*core->mem)); + if (!core->mem) + return -ENOMEM; + + for (i = 0; i < core->num_mems; i++) { + core->mem[i].bus_addr = dev_read_addr_size_name(dev, + mem_names[i], + (fdt_addr_t *)&core->mem[i].size); + if (core->mem[i].bus_addr == FDT_ADDR_T_NONE) { + dev_err(dev, "%s bus address not found\n", + mem_names[i]); + return -EINVAL; + } + core->mem[i].cpu_addr = map_physmem(core->mem[i].bus_addr, + core->mem[i].size, + MAP_NOCACHE); + if (!strcmp(mem_names[i], "atcm")) { + core->mem[i].dev_addr = core->loczrama ? + 0 : K3_R5_TCM_DEV_ADDR; + } else { + core->mem[i].dev_addr = core->loczrama ? + K3_R5_TCM_DEV_ADDR : 0; + } + + dev_dbg(dev, "memory %8s: bus addr %pa size 0x%zx va %p da 0x%x\n", + mem_names[i], &core->mem[i].bus_addr, + core->mem[i].size, core->mem[i].cpu_addr, + core->mem[i].dev_addr); + } + + return 0; +} + +/** + * k3_r5f_probe() - Basic probe + * @dev: corresponding k3 remote processor device + * + * Return: 0 if all goes good, else appropriate error message. + */ +static int k3_r5f_probe(struct udevice *dev) +{ + struct k3_r5f_cluster *cluster = dev_get_priv(dev->parent); + struct k3_r5f_core *core = dev_get_priv(dev); + bool r_state; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + core->dev = dev; + ret = k3_r5f_of_to_priv(core); + if (ret) + return ret; + + core->cluster = cluster; + /* Assume Primary core gets probed first */ + if (!cluster->cores[0]) + cluster->cores[0] = core; + else + cluster->cores[1] = core; + + ret = k3_r5f_core_of_get_memories(core); + if (ret) { + dev_err(dev, "Rproc getting internal memories failed\n"); + return ret; + } + + ret = core->tsp.sci->ops.dev_ops.is_on(core->tsp.sci, core->tsp.dev_id, + &r_state, &core->in_use); + if (ret) + return ret; + + if (core->in_use) { + dev_info(dev, "Core %d is already in use. No rproc commands work\n", + core->tsp.proc_id); + return 0; + } + + /* Make sure Local reset is asserted. Redundant? */ + reset_assert(&core->reset); + + ret = k3_r5f_rproc_configure(core); + if (ret) { + dev_err(dev, "rproc configure failed %d\n", ret); + return ret; + } + + dev_dbg(dev, "Remoteproc successfully probed\n"); + + return 0; +} + +static int k3_r5f_remove(struct udevice *dev) +{ + struct k3_r5f_core *core = dev_get_priv(dev); + + free(core->mem); + + ti_sci_proc_release(&core->tsp); + + return 0; +} + +static const struct udevice_id k3_r5f_rproc_ids[] = { + { .compatible = "ti,am654-r5f"}, + { .compatible = "ti,j721e-r5f"}, + {} +}; + +U_BOOT_DRIVER(k3_r5f_rproc) = { + .name = "k3_r5f_rproc", + .of_match = k3_r5f_rproc_ids, + .id = UCLASS_REMOTEPROC, + .ops = &k3_r5f_rproc_ops, + .probe = k3_r5f_probe, + .remove = k3_r5f_remove, + .priv_auto_alloc_size = sizeof(struct k3_r5f_core), +}; + +static int k3_r5f_cluster_probe(struct udevice *dev) +{ + struct k3_r5f_cluster *cluster = dev_get_priv(dev); + + dev_dbg(dev, "%s\n", __func__); + + cluster->mode = dev_read_u32_default(dev, "lockstep-mode", + CLUSTER_MODE_LOCKSTEP); + + if (device_get_child_count(dev) != 2) { + dev_err(dev, "Invalid number of R5 cores"); + return -EINVAL; + } + + dev_dbg(dev, "%s: Cluster successfully probed in %s mode\n", + __func__, cluster->mode ? "lockstep" : "split"); + + return 0; +} + +static const struct udevice_id k3_r5fss_ids[] = { + { .compatible = "ti,am654-r5fss"}, + { .compatible = "ti,j721e-r5fss"}, + {} +}; + +U_BOOT_DRIVER(k3_r5fss) = { + .name = "k3_r5fss", + .of_match = k3_r5fss_ids, + .id = UCLASS_MISC, + .probe = k3_r5f_cluster_probe, + .priv_auto_alloc_size = sizeof(struct k3_r5f_cluster), +};

Certain SoCs with K3 architecture have integrated a C66 Corepac DSP subsystem and an advanced C71 DSPs. Introduce a remoteproc driver that that does take care of loading an elf to any of the specified DSPs and start it.

Signed-off-by: Lokesh Vutla lokeshvutla@ti.com Signed-off-by: Suman Anna s-anna@ti.com --- drivers/remoteproc/Kconfig | 10 + drivers/remoteproc/Makefile | 1 + drivers/remoteproc/ti_k3_dsp_rproc.c | 354 +++++++++++++++++++++++++++ 3 files changed, 365 insertions(+) create mode 100644 drivers/remoteproc/ti_k3_dsp_rproc.c

diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig index c2d59ba000..7c2e4804b5 100644 --- a/drivers/remoteproc/Kconfig +++ b/drivers/remoteproc/Kconfig @@ -52,6 +52,16 @@ config REMOTEPROC_TI_K3_ARM64 on various TI K3 family of SoCs through the remote processor framework.

+config REMOTEPROC_TI_K3_DSP + bool "TI K3 C66 and C71 remoteproc support" + select REMOTEPROC + depends on ARCH_K3 + depends on TI_SCI_PROTOCOL + help + Say y here to support TI's C66/C71 remote processor subsystems + on various TI K3 family of SoCs through the remote processor + framework. + config REMOTEPROC_TI_K3_R5F bool "TI K3 R5F remoteproc support" select REMOTEPROC diff --git a/drivers/remoteproc/Makefile b/drivers/remoteproc/Makefile index 9d247ba5e7..69ae7bd1e8 100644 --- a/drivers/remoteproc/Makefile +++ b/drivers/remoteproc/Makefile @@ -11,5 +11,6 @@ obj-$(CONFIG_K3_SYSTEM_CONTROLLER) += k3_system_controller.o obj-$(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_R5F) += ti_k3_r5f_rproc.o obj-$(CONFIG_REMOTEPROC_TI_POWER) += ti_power_proc.o diff --git a/drivers/remoteproc/ti_k3_dsp_rproc.c b/drivers/remoteproc/ti_k3_dsp_rproc.c new file mode 100644 index 0000000000..c5dc6b25da --- /dev/null +++ b/drivers/remoteproc/ti_k3_dsp_rproc.c @@ -0,0 +1,354 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Texas Instruments' K3 DSP Remoteproc driver + * + * Copyright (C) 2018-2019 Texas Instruments Incorporated - http://www.ti.com/ + * Lokesh Vutla lokeshvutla@ti.com + * + */ + +#include <common.h> +#include <dm.h> +#include <remoteproc.h> +#include <errno.h> +#include <clk.h> +#include <reset.h> +#include <asm/io.h> +#include <power-domain.h> +#include <linux/soc/ti/ti_sci_protocol.h> +#include "ti_sci_proc.h" + +#define KEYSTONE_RPROC_LOCAL_ADDRESS_MASK (SZ_16M - 1) + +/** + * struct k3_dsp_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_dsp_mem { + void __iomem *cpu_addr; + phys_addr_t bus_addr; + phys_addr_t dev_addr; + size_t size; +}; + +/** + * struct k3_dsp_privdata - Structure representing Remote processor data. + * @rproc_rst: rproc reset control data + * @tsp: Pointer to TISCI proc contrl handle + * @mem: Array of available memories + * @num_mem: Number of available memories + */ +struct k3_dsp_privdata { + struct reset_ctl dsp_rst; + struct ti_sci_proc tsp; + struct k3_dsp_mem *mem; + int num_mems; +}; + +/** + * k3_dsp_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_dsp_load(struct udevice *dev, ulong addr, ulong size) +{ + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + u32 boot_vector; + int ret; + + dev_dbg(dev, "%s addr = 0x%lx, size = 0x%lx\n", __func__, addr, size); + ret = ti_sci_proc_request(&dsp->tsp); + if (ret) + return ret; + + ret = rproc_elf_load_image(dev, addr, size); + if (ret < 0) { + dev_err(dev, "Loading elf failed %d\n", ret); + goto proc_release; + } + + boot_vector = rproc_elf_get_boot_addr(dev, addr); + + dev_dbg(dev, "%s: Boot vector = 0x%x\n", __func__, boot_vector); + + ret = ti_sci_proc_set_config(&dsp->tsp, boot_vector, 0, 0); +proc_release: + ti_sci_proc_release(&dsp->tsp); + return ret; +} + +/** + * k3_dsp_start() - Start the remote processor + * @dev: rproc device pointer + * + * Return: 0 if all went ok, else return appropriate error + */ +static int k3_dsp_start(struct udevice *dev) +{ + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + int ret; + + dev_dbg(dev, "%s\n", __func__); + + ret = ti_sci_proc_request(&dsp->tsp); + if (ret) + return ret; + /* + * Setting the right clock frequency would have taken care by + * assigned-clock-rates during the device probe. So no need to + * set the frequency again here. + */ + ret = ti_sci_proc_power_domain_on(&dsp->tsp); + if (ret) + goto proc_release; + + ret = reset_deassert(&dsp->dsp_rst); + +proc_release: + ti_sci_proc_release(&dsp->tsp); + + return ret; +} + +static int k3_dsp_stop(struct udevice *dev) +{ + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + + dev_dbg(dev, "%s\n", __func__); + + ti_sci_proc_request(&dsp->tsp); + reset_assert(&dsp->dsp_rst); + ti_sci_proc_power_domain_off(&dsp->tsp); + ti_sci_proc_release(&dsp->tsp); + + return 0; +} + +/** + * k3_dsp_init() - Initialize the remote processor + * @dev: rproc device pointer + * + * Return: 0 if all went ok, else return appropriate error + */ +static int k3_dsp_init(struct udevice *dev) +{ + dev_dbg(dev, "%s\n", __func__); + + return 0; +} + +static int k3_dsp_reset(struct udevice *dev) +{ + dev_dbg(dev, "%s\n", __func__); + + return 0; +} + +static void *k3_dsp_da_to_va(struct udevice *dev, ulong da, ulong len) +{ + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + phys_addr_t bus_addr, dev_addr; + void __iomem *va = NULL; + size_t size; + u32 offset; + int i; + + dev_dbg(dev, "%s\n", __func__); + + if (len <= 0) + return NULL; + + for (i = 0; i < dsp->num_mems; i++) { + bus_addr = dsp->mem[i].bus_addr; + dev_addr = dsp->mem[i].dev_addr; + size = dsp->mem[i].size; + + if (da >= dev_addr && ((da + len) <= (dev_addr + size))) { + offset = da - dev_addr; + va = dsp->mem[i].cpu_addr + offset; + return (__force void *)va; + } + + if (da >= bus_addr && (da + len) <= (bus_addr + size)) { + offset = da - bus_addr; + va = dsp->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_dsp_ops = { + .init = k3_dsp_init, + .load = k3_dsp_load, + .start = k3_dsp_start, + .stop = k3_dsp_stop, + .reset = k3_dsp_reset, + .device_to_virt = k3_dsp_da_to_va, +}; + +static int ti_sci_proc_of_to_priv(struct udevice *dev, struct ti_sci_proc *tsp) +{ + u32 ids[2]; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + 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 int k3_dsp_of_get_memories(struct udevice *dev) +{ + static const char * const mem_names[] = {"l2sram", "l1pram", "l1dram"}; + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + int i; + + dev_dbg(dev, "%s\n", __func__); + + dsp->num_mems = ARRAY_SIZE(mem_names); + dsp->mem = calloc(dsp->num_mems, sizeof(*dsp->mem)); + if (!dsp->mem) + return -ENOMEM; + + for (i = 0; i < dsp->num_mems; i++) { + /* C71 cores only have a L1P Cache, there are no L1P SRAMs */ + if (device_is_compatible(dev, "ti,j721e-c71-dsp") && + !strcmp(mem_names[i], "l1pram")) { + dsp->mem[i].bus_addr = FDT_ADDR_T_NONE; + dsp->mem[i].dev_addr = FDT_ADDR_T_NONE; + dsp->mem[i].cpu_addr = NULL; + dsp->mem[i].size = 0; + continue; + } + + dsp->mem[i].bus_addr = dev_read_addr_size_name(dev, mem_names[i], + (fdt_addr_t *)&dsp->mem[i].size); + if (dsp->mem[i].bus_addr == FDT_ADDR_T_NONE) { + dev_err(dev, "%s bus address not found\n", mem_names[i]); + return -EINVAL; + } + dsp->mem[i].cpu_addr = map_physmem(dsp->mem[i].bus_addr, + dsp->mem[i].size, + MAP_NOCACHE); + dsp->mem[i].dev_addr = dsp->mem[i].bus_addr & + KEYSTONE_RPROC_LOCAL_ADDRESS_MASK; + + dev_dbg(dev, "memory %8s: bus addr %pa size 0x%zx va %p da %pa\n", + mem_names[i], &dsp->mem[i].bus_addr, + dsp->mem[i].size, dsp->mem[i].cpu_addr, + &dsp->mem[i].dev_addr); + } + + return 0; +} + +/** + * k3_of_to_priv() - generate private data from device tree + * @dev: corresponding k3 dsp processor device + * @dsp: pointer to driver specific private data + * + * Return: 0 if all goes good, else appropriate error message. + */ +static int k3_dsp_of_to_priv(struct udevice *dev, struct k3_dsp_privdata *dsp) +{ + int ret; + + dev_dbg(dev, "%s\n", __func__); + + ret = reset_get_by_index(dev, 0, &dsp->dsp_rst); + if (ret) { + dev_err(dev, "reset_get() failed: %d\n", ret); + return ret; + } + + ret = ti_sci_proc_of_to_priv(dev, &dsp->tsp); + if (ret) + return ret; + + ret = k3_dsp_of_get_memories(dev); + if (ret) + return ret; + + return 0; +} + +/** + * k3_dsp_probe() - Basic probe + * @dev: corresponding k3 remote processor device + * + * Return: 0 if all goes good, else appropriate error message. + */ +static int k3_dsp_probe(struct udevice *dev) +{ + struct k3_dsp_privdata *dsp; + int ret; + + dev_dbg(dev, "%s\n", __func__); + + dsp = dev_get_priv(dev); + + ret = k3_dsp_of_to_priv(dev, dsp); + if (ret) { + dev_dbg(dev, "%s: Probe failed with error %d\n", __func__, ret); + return ret; + } + + dev_dbg(dev, "Remoteproc successfully probed\n"); + + return 0; +} + +static int k3_dsp_remove(struct udevice *dev) +{ + struct k3_dsp_privdata *dsp = dev_get_priv(dev); + + free(dsp->mem); + + return 0; +} + +static const struct udevice_id k3_dsp_ids[] = { + { .compatible = "ti,j721e-c66-dsp"}, + { .compatible = "ti,j721e-c71-dsp"}, + {} +}; + +U_BOOT_DRIVER(k3_dsp) = { + .name = "k3_dsp", + .of_match = k3_dsp_ids, + .id = UCLASS_REMOTEPROC, + .ops = &k3_dsp_ops, + .probe = k3_dsp_probe, + .remove = k3_dsp_remove, + .priv_auto_alloc_size = sizeof(struct k3_dsp_privdata), +};

The J721E SoCs have 3 dual-core Arm Cortex-R5F processor (R5FSS) subsystems/clusters. One R5F cluster (MCU_R5FSS0) is present within the MCU domain, and the remaining two clusters are present in the MAIN domain (MAIN_R5FSS0 & MAIN_R5FSS1). Each of these can be configured at boot time to be either run in a LockStep mode or in an Asymmetric Multi Processing (AMP) fashion in Split-mode. These subsystems have 64 KB each Tightly-Coupled Memory (TCM) internal memories for each core split between two banks - ATCM and BTCM (further interleaved into two banks). There are some IP integration differences from standard Arm R5 clusters such as the absence of an ACP port, presence of an additional TI-specific Region Address Translater (RAT) module for translating 32-bit CPU addresses into larger system bus addresses etc.

Add the DT nodes for these two MAIN domain R5F cluster/subsystems, the two R5 cores are each added as child nodes to the corresponding main cluster node. Configure SS0 in split mode an SS1 in lockstep mode, with the ATCMs enabled to allow the R5 cores to execute code from DDR with boot-strapping code from ATCM.

Signed-off-by: Suman Anna s-anna@ti.com Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- arch/arm/dts/k3-j721e-common-proc-board.dts | 4 ++ arch/arm/dts/k3-j721e-main.dtsi | 76 +++++++++++++++++++++ 2 files changed, 80 insertions(+)

diff --git a/arch/arm/dts/k3-j721e-common-proc-board.dts b/arch/arm/dts/k3-j721e-common-proc-board.dts index 070d21cd05..21afdc8ce0 100644 --- a/arch/arm/dts/k3-j721e-common-proc-board.dts +++ b/arch/arm/dts/k3-j721e-common-proc-board.dts @@ -16,6 +16,10 @@ aliases { remoteproc0 = &mcu_r5fss0_core0; remoteproc1 = &mcu_r5fss0_core1; + remoteproc2 = &main_r5fss0_core0; + remoteproc3 = &main_r5fss0_core1; + remoteproc4 = &main_r5fss1_core0; + remoteproc5 = &main_r5fss1_core1; }; };

diff --git a/arch/arm/dts/k3-j721e-main.dtsi b/arch/arm/dts/k3-j721e-main.dtsi index 3445784293..59ca4e5686 100644 --- a/arch/arm/dts/k3-j721e-main.dtsi +++ b/arch/arm/dts/k3-j721e-main.dtsi @@ -228,4 +228,80 @@ ti,trm-icp = <0x8>; dma-coherent; }; + + main_r5fss0: r5fss@5c00000 { + compatible = "ti,j721e-r5fss"; + lockstep-mode = <0>; + #address-cells = <1>; + #size-cells = <1>; + ranges = <0x5c00000 0x00 0x5c00000 0x20000>, + <0x5d00000 0x00 0x5d00000 0x20000>; + power-domains = <&k3_pds 243 TI_SCI_PD_EXCLUSIVE>; + + main_r5fss0_core0: r5f@5c00000 { + compatible = "ti,j721e-r5f"; + reg = <0x5c00000 0x00008000>, + <0x5c10000 0x00008000>; + reg-names = "atcm", "btcm"; + ti,sci = <&dmsc>; + ti,sci-dev-id = <245>; + ti,sci-proc-ids = <0x06 0xFF>; + resets = <&k3_reset 245 1>; + atcm-enable = <1>; + btcm-enable = <1>; + loczrama = <1>; + }; + + main_r5fss0_core1: r5f@5d00000 { + compatible = "ti,j721e-r5f"; + reg = <0x5d00000 0x00008000>, + <0x5d10000 0x00008000>; + reg-names = "atcm", "btcm"; + ti,sci = <&dmsc>; + ti,sci-dev-id = <246>; + ti,sci-proc-ids = <0x07 0xFF>; + resets = <&k3_reset 246 1>; + atcm-enable = <1>; + btcm-enable = <1>; + loczrama = <1>; + }; + }; + + main_r5fss1: r5fss@5e00000 { + compatible = "ti,j721e-r5fss"; + lockstep-mode = <1>; + #address-cells = <1>; + #size-cells = <1>; + ranges = <0x5e00000 0x00 0x5e00000 0x20000>, + <0x5f00000 0x00 0x5f00000 0x20000>; + power-domains = <&k3_pds 244 TI_SCI_PD_EXCLUSIVE>; + + main_r5fss1_core0: r5f@5e00000 { + compatible = "ti,j721e-r5f"; + reg = <0x5e00000 0x00008000>, + <0x5e10000 0x00008000>; + reg-names = "atcm", "btcm"; + ti,sci = <&dmsc>; + ti,sci-dev-id = <247>; + ti,sci-proc-ids = <0x08 0xFF>; + resets = <&k3_reset 247 1>; + atcm-enable = <1>; + btcm-enable = <1>; + loczrama = <1>; + }; + + main_r5fss1_core1: r5f@5f00000 { + compatible = "ti,j721e-r5f"; + reg = <0x5f00000 0x00008000>, + <0x5f10000 0x00008000>; + reg-names = "atcm", "btcm"; + ti,sci = <&dmsc>; + ti,sci-dev-id = <248>; + ti,sci-proc-ids = <0x09 0xFF>; + resets = <&k3_reset 248 1>; + atcm-enable = <1>; + btcm-enable = <1>; + loczrama = <1>; + }; + }; };

The J721E SoCs have a single TMS320C71x DSP Subsystem in the MAIN voltage domain containing the next-generation C711 CPU core. The subsystem has 32 KB of L1D configurable SRAM/Cache and 512 KB of L2 configurable SRAM/Cache. This subsystem has a CMMU but is not used currently. The inter-processor communication between the main A72 cores and the C711 processor is achieved through shared memory and a Mailbox. Add the DT node for this DSP processor sub-system in the common k3-j721e-main.dtsi file.

Signed-off-by: Suman Anna s-anna@ti.com Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- arch/arm/dts/k3-j721e-common-proc-board.dts | 1 + arch/arm/dts/k3-j721e-main.dtsi | 11 +++++++++++ 2 files changed, 12 insertions(+)

diff --git a/arch/arm/dts/k3-j721e-common-proc-board.dts b/arch/arm/dts/k3-j721e-common-proc-board.dts index a548277718..b21f597a80 100644 --- a/arch/arm/dts/k3-j721e-common-proc-board.dts +++ b/arch/arm/dts/k3-j721e-common-proc-board.dts @@ -22,6 +22,7 @@ remoteproc5 = &main_r5fss1_core1; remoteproc6 = &c66_0; remoteproc7 = &c66_1; + remoteproc8 = &c71_0; }; };

diff --git a/arch/arm/dts/k3-j721e-main.dtsi b/arch/arm/dts/k3-j721e-main.dtsi index c3aa0cdcf1..6bd59bac52 100644 --- a/arch/arm/dts/k3-j721e-main.dtsi +++ b/arch/arm/dts/k3-j721e-main.dtsi @@ -328,4 +328,15 @@ ti,sci-proc-ids = <0x04 0xFF>; resets = <&k3_reset 143 1>; }; + + c71_0: dsp@64800000 { + compatible = "ti,j721e-c71-dsp"; + reg = <0x00 0x64800000 0x00 0x00080000>, + <0x00 0x64e00000 0x00 0x0000c000>; + reg-names = "l2sram", "l1dram"; + ti,sci = <&dmsc>; + ti,sci-dev-id = <15>; + ti,sci-proc-ids = <0x30 0xFF>; + resets = <&k3_reset 15 1>; + }; };

From: Suman Anna s-anna@ti.com

Add a new file include/environment/ti/k3_rproc.h that defines common environment variables useful for booting various remote processors from U-Boot. This file is expected to be included in the board config files with the EXTRA_ENV_RPROC_SETTINGS added to CONFIG_EXTRA_ENV_SETTINGS and DEFAULT_RPROCS macro overwritten to include the actual list of processors to be booted.

The 'boot_rprocs' variable just needs to be added to the board's bootcmd to automatically boot the processors, and runtime control can be achieved through the 'dorprocboot' variable.

The variables are currently defined to use MMC as the boot media, and can be expanded in the future to include other boot media. The immediate usage is intended for K3 J721E SoCs.

Signed-off-by: Jean-Jacques Hiblot jjhiblot@ti.com Signed-off-by: Suman Anna s-anna@ti.com --- include/environment/ti/k3_rproc.h | 52 +++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) create mode 100644 include/environment/ti/k3_rproc.h

diff --git a/include/environment/ti/k3_rproc.h b/include/environment/ti/k3_rproc.h new file mode 100644 index 0000000000..3418cb42be --- /dev/null +++ b/include/environment/ti/k3_rproc.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (C) 2019 Texas Instruments Incorporated - http://www.ti.com + * + * rproc environment variable definitions for various TI K3 SoCs. + */ + +#ifndef __TI_RPROC_H +#define __TI_RPROC_H + +/* + * should contain a list of <rproc_id fw_name> tuplies, + * override in board config files with the actual list + */ +#define DEFAULT_RPROCS "" + +#ifdef CONFIG_CMD_REMOTEPROC +#define EXTRA_ENV_RPROC_SETTINGS \ + "dorprocboot=0\0" \ + "boot_rprocs=" \ + "if test ${dorprocboot} -eq 1 && test ${boot} = mmc; then "\ + "rproc init;" \ + "run boot_rprocs_mmc;" \ + "fi;\0" \ + "rproc_load_and_boot_one=" \ + "if load mmc ${bootpart} $loadaddr ${rproc_fw}; then " \ + "if rproc load ${rproc_id} ${loadaddr} ${filesize}; then "\ + "rproc start ${rproc_id};" \ + "fi;" \ + "fi\0" \ + "boot_rprocs_mmc=" \ + "env set rproc_id;" \ + "env set rproc_fw;" \ + "for i in ${rproc_fw_binaries} ; do " \ + "if test -z "${rproc_id}" ; then " \ + "env set rproc_id $i;" \ + "else " \ + "env set rproc_fw $i;" \ + "run rproc_load_and_boot_one;" \ + "env set rproc_id;" \ + "env set rproc_fw;" \ + "fi;" \ + "done\0" \ + "rproc_fw_binaries=" \ + DEFAULT_RPROCS \ + "\0" +#else +#define EXTRA_ENV_RPROC_SETTINGS \ + "boot_rprocs= \0" +#endif /* CONFIG_CMD_REMOTEPROC */ + +#endif /* __TI_RPROC_H */

From: Suman Anna s-anna@ti.com

Add support to boot the MCU domain R5F Core0 remoteproc at U-boot prompt on the AM65x EVM boards by using the 'boot_rprocs' and other env variables defined in the common environment file k3_rproc.h, and updating the 'DEFAULT_RPROCS' macro.

The default configuration is to use the MCU R5F in Split mode, so both the R5F Core0 and Core1 are started before loading and booting the Linux kernel using the following firmware: MCU R5FSS0 Core0 (Split) : 0 /lib/firmware/am65x-mcu-r5f0_0-fw MCU R5FSS0 Core1 (Split) : 1 /lib/firmware/am65x-mcu-r5f0_1-fw

The MCU R5FSS was initially running the R5 SPL in LockStep mode with ATCM disabled, and is actually shutdown to enable it to be reconfigured and booted by either A53 U-Boot or Linux kernel in remoteproc mode and using ATCM.

The MCU R5FSS would need to be reconfigured for Lockstep mode through DT if a fault-tolerant/safety application were to be run on the cluster with the DEFAULT_RPROCS macro updated to remove the Core1 firmware.

Signed-off-by: Suman Anna s-anna@ti.com Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- include/configs/am65x_evm.h | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/include/configs/am65x_evm.h b/include/configs/am65x_evm.h index 6072e4a48c..5c44b22ead 100644 --- a/include/configs/am65x_evm.h +++ b/include/configs/am65x_evm.h @@ -12,6 +12,7 @@ #include <linux/sizes.h> #include <config_distro_bootcmd.h> #include <environment/ti/mmc.h> +#include <environment/ti/k3_rproc.h>

#define CONFIG_ENV_SIZE (128 << 10)

@@ -96,11 +97,19 @@ "${bootdir}/${name_kern}\0" \ "partitions=" PARTS_DEFAULT

+#ifdef DEFAULT_RPROCS +#undef DEFAULT_RPROCS +#endif +#define DEFAULT_RPROCS "" \ + "0 /lib/firmware/am65x-mcu-r5f0_0-fw " \ + "1 /lib/firmware/am65x-mcu-r5f0_1-fw " + /* Incorporate settings into the U-Boot environment */ #define CONFIG_EXTRA_ENV_SETTINGS \ DEFAULT_MMC_TI_ARGS \ EXTRA_ENV_AM65X_BOARD_SETTINGS \ - EXTRA_ENV_AM65X_BOARD_SETTINGS_MMC + EXTRA_ENV_AM65X_BOARD_SETTINGS_MMC \ + EXTRA_ENV_RPROC_SETTINGS

/* MMC ENV related defines */ #ifdef CONFIG_ENV_IS_IN_MMC

From: Suman Anna s-anna@ti.com

The A72 U-boot can support early booting of any of the R5F or C66x or C71x remote processors from U-boot prompt to achieve various system usecases before booting the Linux kernel. Update the default BOOTCOMMAND to provide an automatic and easier way to start various remote processors through added environment variables.

Signed-off-by: Jean-Jacques Hiblot jjhiblot@ti.com Signed-off-by: Suman Anna s-anna@ti.com --- configs/j721e_evm_a72_defconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/configs/j721e_evm_a72_defconfig b/configs/j721e_evm_a72_defconfig index 6e0c3f05fb..56df452fb6 100644 --- a/configs/j721e_evm_a72_defconfig +++ b/configs/j721e_evm_a72_defconfig @@ -17,7 +17,7 @@ CONFIG_SPL_SPI_SUPPORT=y CONFIG_DISTRO_DEFAULTS=y # CONFIG_SYS_MALLOC_CLEAR_ON_INIT is not set CONFIG_SPL_LOAD_FIT=y -CONFIG_BOOTCOMMAND="run findfdt; run envboot; run init_${boot}; run get_kern_${boot}; run get_fdt_${boot}; run get_overlay_${boot}; run run_kern" +CONFIG_BOOTCOMMAND="run findfdt; run envboot; run init_${boot}; run boot_rprocs; run get_kern_${boot}; run get_fdt_${boot}; run get_overlay_${boot}; run run_kern" # CONFIG_DISPLAY_CPUINFO is not set CONFIG_SPL_TEXT_BASE=0x80080000 CONFIG_SPL_SYS_MALLOC_SIMPLE=y

From: Suman Anna s-anna@ti.com

The A53 U-boot can support early booting of the MCU R5F remote processor(s) from U-boot prompt to achieve various system usecases before booting the Linux kernel. Update the default BOOTCOMMAND to provide an automatic and easier way to start the MCU R5F cores through added environment variables.

Signed-off-by: Suman Anna s-anna@ti.com Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- configs/am65x_evm_a53_defconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/configs/am65x_evm_a53_defconfig b/configs/am65x_evm_a53_defconfig index d6a0d8583c..2d91d25b6a 100644 --- a/configs/am65x_evm_a53_defconfig +++ b/configs/am65x_evm_a53_defconfig @@ -16,7 +16,7 @@ CONFIG_DISTRO_DEFAULTS=y # CONFIG_SYS_MALLOC_CLEAR_ON_INIT is not set CONFIG_SPL_LOAD_FIT=y CONFIG_OF_BOARD_SETUP=y -CONFIG_BOOTCOMMAND="run findfdt; run envboot; run init_${boot}; run get_kern_${boot}; run get_fdt_${boot}; run get_overlay_${boot}; run run_kern" +CONFIG_BOOTCOMMAND="run findfdt; run envboot; run init_${boot}; run boot_rprocs; run get_kern_${boot}; run get_fdt_${boot}; run get_overlay_${boot}; run run_kern" # CONFIG_DISPLAY_CPUINFO is not set CONFIG_SPL_TEXT_BASE=0x80080000 CONFIG_SPL_SYS_MALLOC_SIMPLE=y

From: Kedar Chitnis kedarc@ti.com

The A72 U-Boot code loads and boots a number of remote processors including the C71x DSP, both the C66_0 and C66_1 DSPs, and the various Main R5FSS Cores. In order to view the code loaded by the U-Boot by remote cores, U-Boot should configure the memory region with right memory attributes. Right now U-Boot carves out a memory region which is not sufficient for all the images to be loaded. So, increase this carve out region by 256MB.

Signed-off-by: Kedar Chitnis kedarc@ti.com Signed-off-by: Lokesh Vutla lokeshvutla@ti.com --- arch/arm/mach-k3/arm64-mmu.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/arm/mach-k3/arm64-mmu.c b/arch/arm/mach-k3/arm64-mmu.c index 98c5a777e5..7f908eee80 100644 --- a/arch/arm/mach-k3/arm64-mmu.c +++ b/arch/arm/mach-k3/arm64-mmu.c @@ -80,13 +80,13 @@ struct mm_region j721e_mem_map[NR_MMU_REGIONS] = { }, { .virt = 0xa0000000UL, .phys = 0xa0000000UL, - .size = 0x0bc00000UL, + .size = 0x1bc00000UL, .attrs = PTE_BLOCK_MEMTYPE(MT_NORMAL_NC) | PTE_BLOCK_NON_SHARE }, { - .virt = 0xabc00000UL, - .phys = 0xabc00000UL, - .size = 0x54400000UL, + .virt = 0xbbc00000UL, + .phys = 0xbbc00000UL, + .size = 0x44400000UL, .attrs = PTE_BLOCK_MEMTYPE(MT_NORMAL) | PTE_BLOCK_INNER_SHARE }, {