Unverified Commit fba8a867 authored by Nick Kossifidis's avatar Nick Kossifidis Committed by Palmer Dabbelt
Browse files

RISC-V: Add kexec support

This patch adds support for kexec on RISC-V. On SMP systems it depends
on HOTPLUG_CPU in order to be able to bring up all harts after kexec.
It also needs a recent OpenSBI version that supports the HSM extension.
I tested it on riscv64 QEMU on both an smp and a non-smp system.

Signed-off-by: default avatarNick Kossifidis <mick@ics.forth.gr>
Signed-off-by: default avatarPalmer Dabbelt <palmerdabbelt@google.com>
parent d83e682e
......@@ -391,6 +391,21 @@ config RISCV_SBI_V01
This config allows kernel to use SBI v0.1 APIs. This will be
deprecated in future once legacy M-mode software are no longer in use.
config KEXEC
bool "Kexec system call"
depends on MMU
kexec is a system call that implements the ability to shutdown your
current kernel, and to start another kernel. It is like a reboot
but it is independent of the system firmware. And like a reboot
you can start any kernel with it, not just Linux.
The name comes from the similarity to the exec system call.
menu "Boot options"
/* SPDX-License-Identifier: GPL-2.0 */
* Copyright (C) 2019 FORTH-ICS/CARV
* Nick Kossifidis <mick@ics.forth.gr>
#ifndef _RISCV_KEXEC_H
#define _RISCV_KEXEC_H
#include <asm/page.h> /* For PAGE_SIZE */
/* Maximum physical address we can use pages from */
/* Maximum address we can reach in physical address mode */
/* Maximum address we can use for the control code buffer */
/* Reserve a page for the control code buffer */
static inline void
crash_setup_regs(struct pt_regs *newregs,
struct pt_regs *oldregs)
/* Dummy implementation for now */
struct kimage_arch {
unsigned long fdt_addr;
const extern unsigned char riscv_kexec_relocate[];
const extern unsigned int riscv_kexec_relocate_size;
typedef void (*riscv_kexec_do_relocate)(unsigned long first_ind_entry,
unsigned long jump_addr,
unsigned long fdt_addr,
unsigned long hartid,
unsigned long va_pa_off);
......@@ -9,6 +9,10 @@ CFLAGS_REMOVE_patch.o = $(CC_FLAGS_FTRACE)
AFLAGS_kexec_relocate.o := -mcmodel=medany -mno-relax
extra-y += head.o
extra-y += vmlinux.lds
......@@ -54,6 +58,7 @@ obj-$(CONFIG_SMP) += cpu_ops_sbi.o
obj-$(CONFIG_HOTPLUG_CPU) += cpu-hotplug.o
obj-$(CONFIG_KGDB) += kgdb.o
obj-$(CONFIG_KEXEC) += kexec_relocate.o machine_kexec.o
obj-$(CONFIG_JUMP_LABEL) += jump_label.o
/* SPDX-License-Identifier: GPL-2.0 */
* Copyright (C) 2019 FORTH-ICS/CARV
* Nick Kossifidis <mick@ics.forth.gr>
#include <asm/asm.h> /* For RISCV_* and REG_* macros */
#include <asm/csr.h> /* For CSR_* macros */
#include <asm/page.h> /* For PAGE_SIZE */
#include <linux/linkage.h> /* For SYM_* macros */
.section ".rodata"
* s0: Pointer to the current entry
* s1: (const) Phys address to jump to after relocation
* s2: (const) Phys address of the FDT image
* s3: (const) The hartid of the current hart
* s4: Pointer to the destination address for the relocation
* s5: (const) Number of words per page
* s6: (const) 1, used for subtraction
* s7: (const) va_pa_offset, used when switching MMU off
* s8: (const) Physical address of the main loop
* s9: (debug) indirection page counter
* s10: (debug) entry counter
* s11: (debug) copied words counter
mv s0, a0
mv s1, a1
mv s2, a2
mv s3, a3
mv s4, zero
li s6, 1
mv s7, a4
mv s8, zero
mv s9, zero
mv s10, zero
mv s11, zero
/* Disable / cleanup interrupts */
csrw CSR_SIE, zero
csrw CSR_SIP, zero
* When we switch SATP.MODE to "Bare" we'll only
* play with physical addresses. However the first time
* we try to jump somewhere, the offset on the jump
* will be relative to pc which will still be on VA. To
* deal with this we set stvec to the physical address at
* the start of the loop below so that we jump there in
* any case.
la s8, 1f
sub s8, s8, s7
csrw CSR_STVEC, s8
/* Process entries in a loop */
.align 2
addi s10, s10, 1
REG_L t0, 0(s0) /* t0 = *image->entry */
addi s0, s0, RISCV_SZPTR /* image->entry++ */
/* IND_DESTINATION entry ? -> save destination address */
andi t1, t0, 0x1
beqz t1, 2f
andi s4, t0, ~0x1
j 1b
/* IND_INDIRECTION entry ? -> update next entry ptr (PA) */
andi t1, t0, 0x2
beqz t1, 2f
andi s0, t0, ~0x2
addi s9, s9, 1
csrw CSR_SATP, zero
jalr zero, s8, 0
/* IND_DONE entry ? -> jump to done label */
andi t1, t0, 0x4
beqz t1, 2f
j 4f
* IND_SOURCE entry ? -> copy page word by word to the
* destination address we got from IND_DESTINATION
andi t1, t0, 0x8
beqz t1, 1b /* Unknown entry type, ignore it */
andi t0, t0, ~0x8
mv t3, s5 /* i = num words per page */
3: /* copy loop */
REG_L t1, (t0) /* t1 = *src_ptr */
REG_S t1, (s4) /* *dst_ptr = *src_ptr */
addi t0, t0, RISCV_SZPTR /* stc_ptr++ */
addi s4, s4, RISCV_SZPTR /* dst_ptr++ */
sub t3, t3, s6 /* i-- */
addi s11, s11, 1 /* c++ */
beqz t3, 1b /* copy done ? */
j 3b
/* Pass the arguments to the next kernel / Cleanup*/
mv a0, s3
mv a1, s2
mv a2, s1
/* Cleanup */
mv a3, zero
mv a4, zero
mv a5, zero
mv a6, zero
mv a7, zero
mv s0, zero
mv s1, zero
mv s2, zero
mv s3, zero
mv s4, zero
mv s5, zero
mv s6, zero
mv s7, zero
mv s8, zero
mv s9, zero
mv s10, zero
mv s11, zero
mv t0, zero
mv t1, zero
mv t2, zero
mv t3, zero
mv t4, zero
mv t5, zero
mv t6, zero
csrw CSR_SEPC, zero
csrw CSR_SCAUSE, zero
csrw CSR_SSCRATCH, zero
* Make sure the relocated code is visible
* and jump to the new kernel
jalr zero, a2, 0
.section ".rodata"
.long riscv_kexec_relocate_end - riscv_kexec_relocate)
// SPDX-License-Identifier: GPL-2.0
* Copyright (C) 2019 FORTH-ICS/CARV
* Nick Kossifidis <mick@ics.forth.gr>
#include <linux/kexec.h>
#include <asm/kexec.h> /* For riscv_kexec_* symbol defines */
#include <linux/smp.h> /* For smp_send_stop () */
#include <asm/cacheflush.h> /* For local_flush_icache_all() */
#include <asm/barrier.h> /* For smp_wmb() */
#include <asm/page.h> /* For PAGE_MASK */
#include <linux/libfdt.h> /* For fdt_check_header() */
#include <asm/set_memory.h> /* For set_memory_x() */
#include <linux/compiler.h> /* For unreachable() */
#include <linux/cpu.h> /* For cpu_down() */
* kexec_image_info - Print received image details
static void
kexec_image_info(const struct kimage *image)
unsigned long i;
pr_debug("Kexec image info:\n");
pr_debug("\ttype: %d\n", image->type);
pr_debug("\tstart: %lx\n", image->start);
pr_debug("\thead: %lx\n", image->head);
pr_debug("\tnr_segments: %lu\n", image->nr_segments);
for (i = 0; i < image->nr_segments; i++) {
pr_debug("\t segment[%lu]: %016lx - %016lx", i,
image->segment[i].mem + image->segment[i].memsz);
pr_debug("\t\t0x%lx bytes, %lu pages\n",
(unsigned long) image->segment[i].memsz,
(unsigned long) image->segment[i].memsz / PAGE_SIZE);
* machine_kexec_prepare - Initialize kexec
* This function is called from do_kexec_load, when the user has
* provided us with an image to be loaded. Its goal is to validate
* the image and prepare the control code buffer as needed.
* Note that kimage_alloc_init has already been called and the
* control buffer has already been allocated.
machine_kexec_prepare(struct kimage *image)
struct kimage_arch *internal = &image->arch;
struct fdt_header fdt = {0};
void *control_code_buffer = NULL;
unsigned int control_code_buffer_sz = 0;
int i = 0;
if (image->type == KEXEC_TYPE_CRASH) {
pr_warn("Loading a crash kernel is unsupported for now.\n");
return -EINVAL;
/* Find the Flattened Device Tree and save its physical address */
for (i = 0; i < image->nr_segments; i++) {
if (image->segment[i].memsz <= sizeof(fdt))
if (copy_from_user(&fdt, image->segment[i].buf, sizeof(fdt)))
if (fdt_check_header(&fdt))
internal->fdt_addr = (unsigned long) image->segment[i].mem;
if (!internal->fdt_addr) {
pr_err("Device tree not included in the provided image\n");
return -EINVAL;
/* Copy the assembler code for relocation to the control page */
control_code_buffer = page_address(image->control_code_page);
control_code_buffer_sz = page_size(image->control_code_page);
if (unlikely(riscv_kexec_relocate_size > control_code_buffer_sz)) {
pr_err("Relocation code doesn't fit within a control page\n");
return -EINVAL;
memcpy(control_code_buffer, riscv_kexec_relocate,
/* Mark the control page executable */
set_memory_x((unsigned long) control_code_buffer, 1);
return 0;
* machine_kexec_cleanup - Cleanup any leftovers from
* machine_kexec_prepare
* This function is called by kimage_free to handle any arch-specific
* allocations done on machine_kexec_prepare. Since we didn't do any
* allocations there, this is just an empty function. Note that the
* control buffer is freed by kimage_free.
machine_kexec_cleanup(struct kimage *image)
* machine_shutdown - Prepare for a kexec reboot
* This function is called by kernel_kexec just before machine_kexec
* below. Its goal is to prepare the rest of the system (the other
* harts and possibly devices etc) for a kexec reboot.
void machine_shutdown(void)
* No more interrupts on this hart
* until we are back up.
* machine_crash_shutdown - Prepare to kexec after a kernel crash
* This function is called by crash_kexec just before machine_kexec
* below and its goal is similar to machine_shutdown, but in case of
* a kernel crash. Since we don't handle such cases yet, this function
* is empty.
machine_crash_shutdown(struct pt_regs *regs)
* machine_kexec - Jump to the loaded kimage
* This function is called by kernel_kexec which is called by the
* reboot system call when the reboot cmd is LINUX_REBOOT_CMD_KEXEC,
* or by crash_kernel which is called by the kernel's arch-specific
* trap handler in case of a kernel panic. It's the final stage of
* the kexec process where the pre-loaded kimage is ready to be
* executed. We assume at this point that all other harts are
* suspended and this hart will be the new boot hart.
void __noreturn
machine_kexec(struct kimage *image)
struct kimage_arch *internal = &image->arch;
unsigned long jump_addr = (unsigned long) image->start;
unsigned long first_ind_entry = (unsigned long) &image->head;
unsigned long this_hart_id = raw_smp_processor_id();
unsigned long fdt_addr = internal->fdt_addr;
void *control_code_buffer = page_address(image->control_code_page);
riscv_kexec_do_relocate do_relocate = control_code_buffer;
pr_notice("Will call new kernel at %08lx from hart id %lx\n",
jump_addr, this_hart_id);
pr_notice("FDT image at %08lx\n", fdt_addr);
/* Make sure the relocation code is visible to the hart */
/* Jump to the relocation code */
do_relocate(first_ind_entry, jump_addr, fdt_addr,
this_hart_id, va_pa_offset);
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment