1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) Linaro.
6 * Copyright (C) Huawei Futurewei Technologies.
9 #include <linux/interrupt.h>
10 #include <linux/irq.h>
11 #include <linux/kernel.h>
12 #include <linux/kexec.h>
13 #include <linux/page-flags.h>
14 #include <linux/set_memory.h>
15 #include <linux/smp.h>
17 #include <asm/cacheflush.h>
18 #include <asm/cpu_ops.h>
19 #include <asm/daifflags.h>
20 #include <asm/memory.h>
22 #include <asm/mmu_context.h>
25 #include "cpu-reset.h"
27 /* Global variables for the arm64_relocate_new_kernel routine. */
28 extern const unsigned char arm64_relocate_new_kernel[];
29 extern const unsigned long arm64_relocate_new_kernel_size;
32 * kexec_image_info - For debugging output.
34 #define kexec_image_info(_i) _kexec_image_info(__func__, __LINE__, _i)
35 static void _kexec_image_info(const char *func, int line,
36 const struct kimage *kimage)
40 pr_debug("%s:%d:\n", func, line);
41 pr_debug(" kexec kimage info:\n");
42 pr_debug(" type: %d\n", kimage->type);
43 pr_debug(" start: %lx\n", kimage->start);
44 pr_debug(" head: %lx\n", kimage->head);
45 pr_debug(" nr_segments: %lu\n", kimage->nr_segments);
46 pr_debug(" kern_reloc: %pa\n", &kimage->arch.kern_reloc);
48 for (i = 0; i < kimage->nr_segments; i++) {
49 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
51 kimage->segment[i].mem,
52 kimage->segment[i].mem + kimage->segment[i].memsz,
53 kimage->segment[i].memsz,
54 kimage->segment[i].memsz / PAGE_SIZE);
58 void machine_kexec_cleanup(struct kimage *kimage)
60 /* Empty routine needed to avoid build errors. */
63 int machine_kexec_post_load(struct kimage *kimage)
65 void *reloc_code = page_to_virt(kimage->control_code_page);
67 memcpy(reloc_code, arm64_relocate_new_kernel,
68 arm64_relocate_new_kernel_size);
69 kimage->arch.kern_reloc = __pa(reloc_code);
70 kexec_image_info(kimage);
73 * For execution with the MMU off, reloc_code needs to be cleaned to the
74 * PoC and invalidated from the I-cache.
76 dcache_clean_inval_poc((unsigned long)reloc_code,
77 (unsigned long)reloc_code +
78 arm64_relocate_new_kernel_size);
79 icache_inval_pou((uintptr_t)reloc_code,
80 (uintptr_t)reloc_code +
81 arm64_relocate_new_kernel_size);
87 * machine_kexec_prepare - Prepare for a kexec reboot.
89 * Called from the core kexec code when a kernel image is loaded.
90 * Forbid loading a kexec kernel if we have no way of hotplugging cpus or cpus
91 * are stuck in the kernel. This avoids a panic once we hit machine_kexec().
93 int machine_kexec_prepare(struct kimage *kimage)
95 if (kimage->type != KEXEC_TYPE_CRASH && cpus_are_stuck_in_kernel()) {
96 pr_err("Can't kexec: CPUs are stuck in the kernel.\n");
104 * kexec_list_flush - Helper to flush the kimage list and source pages to PoC.
106 static void kexec_list_flush(struct kimage *kimage)
108 kimage_entry_t *entry;
110 for (entry = &kimage->head; ; entry++) {
114 /* flush the list entries. */
115 dcache_clean_inval_poc((unsigned long)entry,
116 (unsigned long)entry +
117 sizeof(kimage_entry_t));
119 flag = *entry & IND_FLAGS;
120 if (flag == IND_DONE)
123 addr = (unsigned long)phys_to_virt(*entry & PAGE_MASK);
126 case IND_INDIRECTION:
127 /* Set entry point just before the new list page. */
128 entry = (kimage_entry_t *)addr - 1;
131 /* flush the source pages. */
132 dcache_clean_inval_poc(addr, addr + PAGE_SIZE);
134 case IND_DESTINATION:
143 * kexec_segment_flush - Helper to flush the kimage segments to PoC.
145 static void kexec_segment_flush(const struct kimage *kimage)
149 pr_debug("%s:\n", __func__);
151 for (i = 0; i < kimage->nr_segments; i++) {
152 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
154 kimage->segment[i].mem,
155 kimage->segment[i].mem + kimage->segment[i].memsz,
156 kimage->segment[i].memsz,
157 kimage->segment[i].memsz / PAGE_SIZE);
159 dcache_clean_inval_poc(
160 (unsigned long)phys_to_virt(kimage->segment[i].mem),
161 (unsigned long)phys_to_virt(kimage->segment[i].mem) +
162 kimage->segment[i].memsz);
167 * machine_kexec - Do the kexec reboot.
169 * Called from the core kexec code for a sys_reboot with LINUX_REBOOT_CMD_KEXEC.
171 void machine_kexec(struct kimage *kimage)
173 bool in_kexec_crash = (kimage == kexec_crash_image);
174 bool stuck_cpus = cpus_are_stuck_in_kernel();
177 * New cpus may have become stuck_in_kernel after we loaded the image.
179 BUG_ON(!in_kexec_crash && (stuck_cpus || (num_online_cpus() > 1)));
180 WARN(in_kexec_crash && (stuck_cpus || smp_crash_stop_failed()),
181 "Some CPUs may be stale, kdump will be unreliable.\n");
183 /* Flush the kimage list and its buffers. */
184 kexec_list_flush(kimage);
186 /* Flush the new image if already in place. */
187 if ((kimage != kexec_crash_image) && (kimage->head & IND_DONE))
188 kexec_segment_flush(kimage);
195 * cpu_soft_restart will shutdown the MMU, disable data caches, then
196 * transfer control to the kern_reloc which contains a copy of
197 * the arm64_relocate_new_kernel routine. arm64_relocate_new_kernel
198 * uses physical addressing to relocate the new image to its final
199 * position and transfers control to the image entry point when the
200 * relocation is complete.
201 * In kexec case, kimage->start points to purgatory assuming that
202 * kernel entry and dtb address are embedded in purgatory by
203 * userspace (kexec-tools).
204 * In kexec_file case, the kernel starts directly without purgatory.
206 cpu_soft_restart(kimage->arch.kern_reloc, kimage->head, kimage->start,
207 kimage->arch.dtb_mem);
209 BUG(); /* Should never get here. */
212 static void machine_kexec_mask_interrupts(void)
215 struct irq_desc *desc;
217 for_each_irq_desc(i, desc) {
218 struct irq_chip *chip;
221 chip = irq_desc_get_chip(desc);
226 * First try to remove the active state. If this
227 * fails, try to EOI the interrupt.
229 ret = irq_set_irqchip_state(i, IRQCHIP_STATE_ACTIVE, false);
231 if (ret && irqd_irq_inprogress(&desc->irq_data) &&
233 chip->irq_eoi(&desc->irq_data);
236 chip->irq_mask(&desc->irq_data);
238 if (chip->irq_disable && !irqd_irq_disabled(&desc->irq_data))
239 chip->irq_disable(&desc->irq_data);
244 * machine_crash_shutdown - shutdown non-crashing cpus and save registers
246 void machine_crash_shutdown(struct pt_regs *regs)
250 /* shutdown non-crashing cpus */
251 crash_smp_send_stop();
253 /* for crashing cpu */
254 crash_save_cpu(regs, smp_processor_id());
255 machine_kexec_mask_interrupts();
257 pr_info("Starting crashdump kernel...\n");
260 void arch_kexec_protect_crashkres(void)
264 kexec_segment_flush(kexec_crash_image);
266 for (i = 0; i < kexec_crash_image->nr_segments; i++)
268 __phys_to_virt(kexec_crash_image->segment[i].mem),
269 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 0);
272 void arch_kexec_unprotect_crashkres(void)
276 for (i = 0; i < kexec_crash_image->nr_segments; i++)
278 __phys_to_virt(kexec_crash_image->segment[i].mem),
279 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 1);
282 #ifdef CONFIG_HIBERNATION
284 * To preserve the crash dump kernel image, the relevant memory segments
285 * should be mapped again around the hibernation.
287 void crash_prepare_suspend(void)
289 if (kexec_crash_image)
290 arch_kexec_unprotect_crashkres();
293 void crash_post_resume(void)
295 if (kexec_crash_image)
296 arch_kexec_protect_crashkres();
302 * Return true only if a page is part of reserved memory for crash dump kernel,
303 * but does not hold any data of loaded kernel image.
305 * Note that all the pages in crash dump kernel memory have been initially
306 * marked as Reserved as memory was allocated via memblock_reserve().
308 * In hibernation, the pages which are Reserved and yet "nosave" are excluded
309 * from the hibernation iamge. crash_is_nosave() does thich check for crash
310 * dump kernel and will reduce the total size of hibernation image.
313 bool crash_is_nosave(unsigned long pfn)
321 /* in reserved memory? */
322 addr = __pfn_to_phys(pfn);
323 if ((addr < crashk_res.start) || (crashk_res.end < addr))
326 if (!kexec_crash_image)
329 /* not part of loaded kernel image? */
330 for (i = 0; i < kexec_crash_image->nr_segments; i++)
331 if (addr >= kexec_crash_image->segment[i].mem &&
332 addr < (kexec_crash_image->segment[i].mem +
333 kexec_crash_image->segment[i].memsz))
339 void crash_free_reserved_phys_range(unsigned long begin, unsigned long end)
344 for (addr = begin; addr < end; addr += PAGE_SIZE) {
345 page = phys_to_page(addr);
346 free_reserved_page(page);
349 #endif /* CONFIG_HIBERNATION */