4 * Copyright (C) Linaro.
5 * Copyright (C) Huawei Futurewei Technologies.
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
12 #include <linux/interrupt.h>
13 #include <linux/irq.h>
14 #include <linux/kernel.h>
15 #include <linux/kexec.h>
16 #include <linux/page-flags.h>
17 #include <linux/smp.h>
19 #include <asm/cacheflush.h>
20 #include <asm/cpu_ops.h>
21 #include <asm/daifflags.h>
22 #include <asm/memory.h>
24 #include <asm/mmu_context.h>
27 #include "cpu-reset.h"
29 /* Global variables for the arm64_relocate_new_kernel routine. */
30 extern const unsigned char arm64_relocate_new_kernel[];
31 extern const unsigned long arm64_relocate_new_kernel_size;
34 * kexec_image_info - For debugging output.
36 #define kexec_image_info(_i) _kexec_image_info(__func__, __LINE__, _i)
37 static void _kexec_image_info(const char *func, int line,
38 const struct kimage *kimage)
42 pr_debug("%s:%d:\n", func, line);
43 pr_debug(" kexec kimage info:\n");
44 pr_debug(" type: %d\n", kimage->type);
45 pr_debug(" start: %lx\n", kimage->start);
46 pr_debug(" head: %lx\n", kimage->head);
47 pr_debug(" nr_segments: %lu\n", kimage->nr_segments);
49 for (i = 0; i < kimage->nr_segments; i++) {
50 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
52 kimage->segment[i].mem,
53 kimage->segment[i].mem + kimage->segment[i].memsz,
54 kimage->segment[i].memsz,
55 kimage->segment[i].memsz / PAGE_SIZE);
59 void machine_kexec_cleanup(struct kimage *kimage)
61 /* Empty routine needed to avoid build errors. */
65 * machine_kexec_prepare - Prepare for a kexec reboot.
67 * Called from the core kexec code when a kernel image is loaded.
68 * Forbid loading a kexec kernel if we have no way of hotplugging cpus or cpus
69 * are stuck in the kernel. This avoids a panic once we hit machine_kexec().
71 int machine_kexec_prepare(struct kimage *kimage)
73 kexec_image_info(kimage);
75 if (kimage->type != KEXEC_TYPE_CRASH && cpus_are_stuck_in_kernel()) {
76 pr_err("Can't kexec: CPUs are stuck in the kernel.\n");
84 * kexec_list_flush - Helper to flush the kimage list and source pages to PoC.
86 static void kexec_list_flush(struct kimage *kimage)
88 kimage_entry_t *entry;
90 for (entry = &kimage->head; ; entry++) {
94 /* flush the list entries. */
95 __flush_dcache_area(entry, sizeof(kimage_entry_t));
97 flag = *entry & IND_FLAGS;
101 addr = phys_to_virt(*entry & PAGE_MASK);
104 case IND_INDIRECTION:
105 /* Set entry point just before the new list page. */
106 entry = (kimage_entry_t *)addr - 1;
109 /* flush the source pages. */
110 __flush_dcache_area(addr, PAGE_SIZE);
112 case IND_DESTINATION:
121 * kexec_segment_flush - Helper to flush the kimage segments to PoC.
123 static void kexec_segment_flush(const struct kimage *kimage)
127 pr_debug("%s:\n", __func__);
129 for (i = 0; i < kimage->nr_segments; i++) {
130 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
132 kimage->segment[i].mem,
133 kimage->segment[i].mem + kimage->segment[i].memsz,
134 kimage->segment[i].memsz,
135 kimage->segment[i].memsz / PAGE_SIZE);
137 __flush_dcache_area(phys_to_virt(kimage->segment[i].mem),
138 kimage->segment[i].memsz);
143 * machine_kexec - Do the kexec reboot.
145 * Called from the core kexec code for a sys_reboot with LINUX_REBOOT_CMD_KEXEC.
147 void machine_kexec(struct kimage *kimage)
149 phys_addr_t reboot_code_buffer_phys;
150 void *reboot_code_buffer;
151 bool in_kexec_crash = (kimage == kexec_crash_image);
152 bool stuck_cpus = cpus_are_stuck_in_kernel();
155 * New cpus may have become stuck_in_kernel after we loaded the image.
157 BUG_ON(!in_kexec_crash && (stuck_cpus || (num_online_cpus() > 1)));
158 WARN(in_kexec_crash && (stuck_cpus || smp_crash_stop_failed()),
159 "Some CPUs may be stale, kdump will be unreliable.\n");
161 reboot_code_buffer_phys = page_to_phys(kimage->control_code_page);
162 reboot_code_buffer = phys_to_virt(reboot_code_buffer_phys);
164 kexec_image_info(kimage);
166 pr_debug("%s:%d: control_code_page: %p\n", __func__, __LINE__,
167 kimage->control_code_page);
168 pr_debug("%s:%d: reboot_code_buffer_phys: %pa\n", __func__, __LINE__,
169 &reboot_code_buffer_phys);
170 pr_debug("%s:%d: reboot_code_buffer: %p\n", __func__, __LINE__,
172 pr_debug("%s:%d: relocate_new_kernel: %p\n", __func__, __LINE__,
173 arm64_relocate_new_kernel);
174 pr_debug("%s:%d: relocate_new_kernel_size: 0x%lx(%lu) bytes\n",
175 __func__, __LINE__, arm64_relocate_new_kernel_size,
176 arm64_relocate_new_kernel_size);
179 * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use
180 * after the kernel is shut down.
182 memcpy(reboot_code_buffer, arm64_relocate_new_kernel,
183 arm64_relocate_new_kernel_size);
185 /* Flush the reboot_code_buffer in preparation for its execution. */
186 __flush_dcache_area(reboot_code_buffer, arm64_relocate_new_kernel_size);
189 * Although we've killed off the secondary CPUs, we don't update
190 * the online mask if we're handling a crash kernel and consequently
191 * need to avoid flush_icache_range(), which will attempt to IPI
192 * the offline CPUs. Therefore, we must use the __* variant here.
194 __flush_icache_range((uintptr_t)reboot_code_buffer,
195 arm64_relocate_new_kernel_size);
197 /* Flush the kimage list and its buffers. */
198 kexec_list_flush(kimage);
200 /* Flush the new image if already in place. */
201 if ((kimage != kexec_crash_image) && (kimage->head & IND_DONE))
202 kexec_segment_flush(kimage);
209 * cpu_soft_restart will shutdown the MMU, disable data caches, then
210 * transfer control to the reboot_code_buffer which contains a copy of
211 * the arm64_relocate_new_kernel routine. arm64_relocate_new_kernel
212 * uses physical addressing to relocate the new image to its final
213 * position and transfers control to the image entry point when the
214 * relocation is complete.
215 * In kexec case, kimage->start points to purgatory assuming that
216 * kernel entry and dtb address are embedded in purgatory by
217 * userspace (kexec-tools).
218 * In kexec_file case, the kernel starts directly without purgatory.
220 cpu_soft_restart(reboot_code_buffer_phys, kimage->head, kimage->start,
221 #ifdef CONFIG_KEXEC_FILE
222 kimage->arch.dtb_mem);
227 BUG(); /* Should never get here. */
230 static void machine_kexec_mask_interrupts(void)
233 struct irq_desc *desc;
235 for_each_irq_desc(i, desc) {
236 struct irq_chip *chip;
239 chip = irq_desc_get_chip(desc);
244 * First try to remove the active state. If this
245 * fails, try to EOI the interrupt.
247 ret = irq_set_irqchip_state(i, IRQCHIP_STATE_ACTIVE, false);
249 if (ret && irqd_irq_inprogress(&desc->irq_data) &&
251 chip->irq_eoi(&desc->irq_data);
254 chip->irq_mask(&desc->irq_data);
256 if (chip->irq_disable && !irqd_irq_disabled(&desc->irq_data))
257 chip->irq_disable(&desc->irq_data);
262 * machine_crash_shutdown - shutdown non-crashing cpus and save registers
264 void machine_crash_shutdown(struct pt_regs *regs)
268 /* shutdown non-crashing cpus */
269 crash_smp_send_stop();
271 /* for crashing cpu */
272 crash_save_cpu(regs, smp_processor_id());
273 machine_kexec_mask_interrupts();
275 pr_info("Starting crashdump kernel...\n");
278 void arch_kexec_protect_crashkres(void)
282 kexec_segment_flush(kexec_crash_image);
284 for (i = 0; i < kexec_crash_image->nr_segments; i++)
286 __phys_to_virt(kexec_crash_image->segment[i].mem),
287 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 0);
290 void arch_kexec_unprotect_crashkres(void)
294 for (i = 0; i < kexec_crash_image->nr_segments; i++)
296 __phys_to_virt(kexec_crash_image->segment[i].mem),
297 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 1);
300 #ifdef CONFIG_HIBERNATION
302 * To preserve the crash dump kernel image, the relevant memory segments
303 * should be mapped again around the hibernation.
305 void crash_prepare_suspend(void)
307 if (kexec_crash_image)
308 arch_kexec_unprotect_crashkres();
311 void crash_post_resume(void)
313 if (kexec_crash_image)
314 arch_kexec_protect_crashkres();
320 * Return true only if a page is part of reserved memory for crash dump kernel,
321 * but does not hold any data of loaded kernel image.
323 * Note that all the pages in crash dump kernel memory have been initially
324 * marked as Reserved as memory was allocated via memblock_reserve().
326 * In hibernation, the pages which are Reserved and yet "nosave" are excluded
327 * from the hibernation iamge. crash_is_nosave() does thich check for crash
328 * dump kernel and will reduce the total size of hibernation image.
331 bool crash_is_nosave(unsigned long pfn)
339 /* in reserved memory? */
340 addr = __pfn_to_phys(pfn);
341 if ((addr < crashk_res.start) || (crashk_res.end < addr))
344 if (!kexec_crash_image)
347 /* not part of loaded kernel image? */
348 for (i = 0; i < kexec_crash_image->nr_segments; i++)
349 if (addr >= kexec_crash_image->segment[i].mem &&
350 addr < (kexec_crash_image->segment[i].mem +
351 kexec_crash_image->segment[i].memsz))
357 void crash_free_reserved_phys_range(unsigned long begin, unsigned long end)
362 for (addr = begin; addr < end; addr += PAGE_SIZE) {
363 page = phys_to_page(addr);
364 free_reserved_page(page);
367 #endif /* CONFIG_HIBERNATION */