u-boot/arch/arm/mach-k3/common.c
Lokesh Vutla 40109f4d7e arm: mach-k3: Enable WA for R5F deadlock
On K3 devices there are 2 conditions where R5F can deadlock:
1.When software is performing series of store operations to
  cacheable write back/write allocate memory region and later
  on software execute barrier operation (DSB or DMB). R5F may
  hang at the barrier instruction.
2.When software is performing a mix of load and store operations
  within a tight loop and store operations are all writing to
  cacheable write back/write allocates memory regions, R5F may
  hang at one of the load instruction.

To avoid the above two conditions disable linefill optimization
inside Cortex R5F which will make R5F to only issue up to 2 cache
line fills at any point of time.

Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com>
2020-01-03 09:47:11 -05:00

271 lines
6.5 KiB
C

// SPDX-License-Identifier: GPL-2.0+
/*
* K3: Common Architecture initialization
*
* Copyright (C) 2018 Texas Instruments Incorporated - http://www.ti.com/
* Lokesh Vutla <lokeshvutla@ti.com>
*/
#include <common.h>
#include <spl.h>
#include "common.h"
#include <dm.h>
#include <remoteproc.h>
#include <linux/soc/ti/ti_sci_protocol.h>
#include <fdt_support.h>
#include <asm/arch/sys_proto.h>
#include <asm/hardware.h>
#include <asm/io.h>
struct ti_sci_handle *get_ti_sci_handle(void)
{
struct udevice *dev;
int ret;
ret = uclass_get_device_by_driver(UCLASS_FIRMWARE,
DM_GET_DRIVER(ti_sci), &dev);
if (ret)
panic("Failed to get SYSFW (%d)\n", ret);
return (struct ti_sci_handle *)ti_sci_get_handle_from_sysfw(dev);
}
DECLARE_GLOBAL_DATA_PTR;
#ifdef CONFIG_K3_EARLY_CONS
int early_console_init(void)
{
struct udevice *dev;
int ret;
gd->baudrate = CONFIG_BAUDRATE;
ret = uclass_get_device_by_seq(UCLASS_SERIAL, CONFIG_K3_EARLY_CONS_IDX,
&dev);
if (ret) {
printf("Error getting serial dev for early console! (%d)\n",
ret);
return ret;
}
gd->cur_serial_dev = dev;
gd->flags |= GD_FLG_SERIAL_READY;
gd->have_console = 1;
return 0;
}
#endif
#ifdef CONFIG_SYS_K3_SPL_ATF
void __noreturn jump_to_image_no_args(struct spl_image_info *spl_image)
{
struct ti_sci_handle *ti_sci = get_ti_sci_handle();
int ret;
/* Release all the exclusive devices held by SPL before starting ATF */
ti_sci->ops.dev_ops.release_exclusive_devices(ti_sci);
/*
* It is assumed that remoteproc device 1 is the corresponding
* Cortex-A core which runs ATF. Make sure DT reflects the same.
*/
ret = rproc_dev_init(1);
if (ret)
panic("%s: ATF failed to initialize on rproc (%d)\n", __func__,
ret);
ret = rproc_load(1, spl_image->entry_point, 0x200);
if (ret)
panic("%s: ATF failed to load on rproc (%d)\n", __func__, ret);
/* Add an extra newline to differentiate the ATF logs from SPL */
printf("Starting ATF on ARM64 core...\n\n");
ret = rproc_start(1);
if (ret)
panic("%s: ATF failed to start on rproc (%d)\n", __func__, ret);
debug("Releasing resources...\n");
release_resources_for_core_shutdown();
debug("Finalizing core shutdown...\n");
while (1)
asm volatile("wfe");
}
#endif
#if defined(CONFIG_OF_LIBFDT)
int fdt_fixup_msmc_ram(void *blob, char *parent_path, char *node_name)
{
u64 msmc_start = 0, msmc_end = 0, msmc_size, reg[2];
struct ti_sci_handle *ti_sci = get_ti_sci_handle();
int ret, node, subnode, len, prev_node;
u32 range[4], addr, size;
const fdt32_t *sub_reg;
ti_sci->ops.core_ops.query_msmc(ti_sci, &msmc_start, &msmc_end);
msmc_size = msmc_end - msmc_start + 1;
debug("%s: msmc_start = 0x%llx, msmc_size = 0x%llx\n", __func__,
msmc_start, msmc_size);
/* find or create "msmc_sram node */
ret = fdt_path_offset(blob, parent_path);
if (ret < 0)
return ret;
node = fdt_find_or_add_subnode(blob, ret, node_name);
if (node < 0)
return node;
ret = fdt_setprop_string(blob, node, "compatible", "mmio-sram");
if (ret < 0)
return ret;
reg[0] = cpu_to_fdt64(msmc_start);
reg[1] = cpu_to_fdt64(msmc_size);
ret = fdt_setprop(blob, node, "reg", reg, sizeof(reg));
if (ret < 0)
return ret;
fdt_setprop_cell(blob, node, "#address-cells", 1);
fdt_setprop_cell(blob, node, "#size-cells", 1);
range[0] = 0;
range[1] = cpu_to_fdt32(msmc_start >> 32);
range[2] = cpu_to_fdt32(msmc_start & 0xffffffff);
range[3] = cpu_to_fdt32(msmc_size);
ret = fdt_setprop(blob, node, "ranges", range, sizeof(range));
if (ret < 0)
return ret;
subnode = fdt_first_subnode(blob, node);
prev_node = 0;
/* Look for invalid subnodes and delete them */
while (subnode >= 0) {
sub_reg = fdt_getprop(blob, subnode, "reg", &len);
addr = fdt_read_number(sub_reg, 1);
sub_reg++;
size = fdt_read_number(sub_reg, 1);
debug("%s: subnode = %d, addr = 0x%x. size = 0x%x\n", __func__,
subnode, addr, size);
if (addr + size > msmc_size ||
!strncmp(fdt_get_name(blob, subnode, &len), "sysfw", 5) ||
!strncmp(fdt_get_name(blob, subnode, &len), "l3cache", 7)) {
fdt_del_node(blob, subnode);
debug("%s: deleting subnode %d\n", __func__, subnode);
if (!prev_node)
subnode = fdt_first_subnode(blob, node);
else
subnode = fdt_next_subnode(blob, prev_node);
} else {
prev_node = subnode;
subnode = fdt_next_subnode(blob, prev_node);
}
}
return 0;
}
int fdt_disable_node(void *blob, char *node_path)
{
int offs;
int ret;
offs = fdt_path_offset(blob, node_path);
if (offs < 0) {
debug("Node %s not found.\n", node_path);
return 0;
}
ret = fdt_setprop_string(blob, offs, "status", "disabled");
if (ret < 0) {
printf("Could not add status property to node %s: %s\n",
node_path, fdt_strerror(ret));
return ret;
}
return 0;
}
#endif
#ifndef CONFIG_SYSRESET
void reset_cpu(ulong ignored)
{
}
#endif
#if defined(CONFIG_DISPLAY_CPUINFO)
int print_cpuinfo(void)
{
u32 soc, rev;
char *name;
soc = (readl(CTRLMMR_WKUP_JTAG_DEVICE_ID) &
DEVICE_ID_FAMILY_MASK) >> DEVICE_ID_FAMILY_SHIFT;
rev = (readl(CTRLMMR_WKUP_JTAG_ID) &
JTAG_ID_VARIANT_MASK) >> JTAG_ID_VARIANT_SHIFT;
printf("SoC: ");
switch (soc) {
case AM654:
name = "AM654";
break;
case J721E:
name = "J721E";
break;
default:
name = "Unknown Silicon";
};
printf("%s PG ", name);
switch (rev) {
case REV_PG1_0:
name = "1.0";
break;
case REV_PG2_0:
name = "2.0";
break;
default:
name = "Unknown Revision";
};
printf("%s\n", name);
return 0;
}
#endif
#ifdef CONFIG_ARM64
void board_prep_linux(bootm_headers_t *images)
{
debug("Linux kernel Image start = 0x%lx end = 0x%lx\n",
images->os.start, images->os.end);
__asm_flush_dcache_range(images->os.start,
ROUND(images->os.end,
CONFIG_SYS_CACHELINE_SIZE));
}
#endif
#ifdef CONFIG_CPU_V7R
void disable_linefill_optimization(void)
{
u32 actlr;
/*
* On K3 devices there are 2 conditions where R5F can deadlock:
* 1.When software is performing series of store operations to
* cacheable write back/write allocate memory region and later
* on software execute barrier operation (DSB or DMB). R5F may
* hang at the barrier instruction.
* 2.When software is performing a mix of load and store operations
* within a tight loop and store operations are all writing to
* cacheable write back/write allocates memory regions, R5F may
* hang at one of the load instruction.
*
* To avoid the above two conditions disable linefill optimization
* inside Cortex R5F.
*/
asm("mrc p15, 0, %0, c1, c0, 1" : "=r" (actlr));
actlr |= (1 << 13); /* Set DLFO bit */
asm("mcr p15, 0, %0, c1, c0, 1" : : "r" (actlr));
}
#endif