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/smp.h>
16 #include <asm/cacheflush.h>
17 #include <asm/cpu_ops.h>
18 #include <asm/daifflags.h>
19 #include <asm/memory.h>
21 #include <asm/mmu_context.h>
24 #include "cpu-reset.h"
26 /* Global variables for the arm64_relocate_new_kernel routine. */
27 extern const unsigned char arm64_relocate_new_kernel[];
28 extern const unsigned long arm64_relocate_new_kernel_size;
31 * kexec_image_info - For debugging output.
33 #define kexec_image_info(_i) _kexec_image_info(__func__, __LINE__, _i)
34 static void _kexec_image_info(const char *func, int line,
35 const struct kimage *kimage)
39 pr_debug("%s:%d:\n", func, line);
40 pr_debug(" kexec kimage info:\n");
41 pr_debug(" type: %d\n", kimage->type);
42 pr_debug(" start: %lx\n", kimage->start);
43 pr_debug(" head: %lx\n", kimage->head);
44 pr_debug(" nr_segments: %lu\n", kimage->nr_segments);
46 for (i = 0; i < kimage->nr_segments; i++) {
47 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
49 kimage->segment[i].mem,
50 kimage->segment[i].mem + kimage->segment[i].memsz,
51 kimage->segment[i].memsz,
52 kimage->segment[i].memsz / PAGE_SIZE);
56 void machine_kexec_cleanup(struct kimage *kimage)
58 /* Empty routine needed to avoid build errors. */
62 * machine_kexec_prepare - Prepare for a kexec reboot.
64 * Called from the core kexec code when a kernel image is loaded.
65 * Forbid loading a kexec kernel if we have no way of hotplugging cpus or cpus
66 * are stuck in the kernel. This avoids a panic once we hit machine_kexec().
68 int machine_kexec_prepare(struct kimage *kimage)
70 kexec_image_info(kimage);
72 if (kimage->type != KEXEC_TYPE_CRASH && cpus_are_stuck_in_kernel()) {
73 pr_err("Can't kexec: CPUs are stuck in the kernel.\n");
81 * kexec_list_flush - Helper to flush the kimage list and source pages to PoC.
83 static void kexec_list_flush(struct kimage *kimage)
85 kimage_entry_t *entry;
87 for (entry = &kimage->head; ; entry++) {
91 /* flush the list entries. */
92 __flush_dcache_area(entry, sizeof(kimage_entry_t));
94 flag = *entry & IND_FLAGS;
98 addr = phys_to_virt(*entry & PAGE_MASK);
101 case IND_INDIRECTION:
102 /* Set entry point just before the new list page. */
103 entry = (kimage_entry_t *)addr - 1;
106 /* flush the source pages. */
107 __flush_dcache_area(addr, PAGE_SIZE);
109 case IND_DESTINATION:
118 * kexec_segment_flush - Helper to flush the kimage segments to PoC.
120 static void kexec_segment_flush(const struct kimage *kimage)
124 pr_debug("%s:\n", __func__);
126 for (i = 0; i < kimage->nr_segments; i++) {
127 pr_debug(" segment[%lu]: %016lx - %016lx, 0x%lx bytes, %lu pages\n",
129 kimage->segment[i].mem,
130 kimage->segment[i].mem + kimage->segment[i].memsz,
131 kimage->segment[i].memsz,
132 kimage->segment[i].memsz / PAGE_SIZE);
134 __flush_dcache_area(phys_to_virt(kimage->segment[i].mem),
135 kimage->segment[i].memsz);
140 * machine_kexec - Do the kexec reboot.
142 * Called from the core kexec code for a sys_reboot with LINUX_REBOOT_CMD_KEXEC.
144 void machine_kexec(struct kimage *kimage)
146 phys_addr_t reboot_code_buffer_phys;
147 void *reboot_code_buffer;
148 bool in_kexec_crash = (kimage == kexec_crash_image);
149 bool stuck_cpus = cpus_are_stuck_in_kernel();
152 * New cpus may have become stuck_in_kernel after we loaded the image.
154 BUG_ON(!in_kexec_crash && (stuck_cpus || (num_online_cpus() > 1)));
155 WARN(in_kexec_crash && (stuck_cpus || smp_crash_stop_failed()),
156 "Some CPUs may be stale, kdump will be unreliable.\n");
158 reboot_code_buffer_phys = page_to_phys(kimage->control_code_page);
159 reboot_code_buffer = phys_to_virt(reboot_code_buffer_phys);
161 kexec_image_info(kimage);
163 pr_debug("%s:%d: control_code_page: %p\n", __func__, __LINE__,
164 kimage->control_code_page);
165 pr_debug("%s:%d: reboot_code_buffer_phys: %pa\n", __func__, __LINE__,
166 &reboot_code_buffer_phys);
167 pr_debug("%s:%d: reboot_code_buffer: %p\n", __func__, __LINE__,
169 pr_debug("%s:%d: relocate_new_kernel: %p\n", __func__, __LINE__,
170 arm64_relocate_new_kernel);
171 pr_debug("%s:%d: relocate_new_kernel_size: 0x%lx(%lu) bytes\n",
172 __func__, __LINE__, arm64_relocate_new_kernel_size,
173 arm64_relocate_new_kernel_size);
176 * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use
177 * after the kernel is shut down.
179 memcpy(reboot_code_buffer, arm64_relocate_new_kernel,
180 arm64_relocate_new_kernel_size);
182 /* Flush the reboot_code_buffer in preparation for its execution. */
183 __flush_dcache_area(reboot_code_buffer, arm64_relocate_new_kernel_size);
186 * Although we've killed off the secondary CPUs, we don't update
187 * the online mask if we're handling a crash kernel and consequently
188 * need to avoid flush_icache_range(), which will attempt to IPI
189 * the offline CPUs. Therefore, we must use the __* variant here.
191 __flush_icache_range((uintptr_t)reboot_code_buffer,
192 (uintptr_t)reboot_code_buffer +
193 arm64_relocate_new_kernel_size);
195 /* Flush the kimage list and its buffers. */
196 kexec_list_flush(kimage);
198 /* Flush the new image if already in place. */
199 if ((kimage != kexec_crash_image) && (kimage->head & IND_DONE))
200 kexec_segment_flush(kimage);
207 * cpu_soft_restart will shutdown the MMU, disable data caches, then
208 * transfer control to the reboot_code_buffer which contains a copy of
209 * the arm64_relocate_new_kernel routine. arm64_relocate_new_kernel
210 * uses physical addressing to relocate the new image to its final
211 * position and transfers control to the image entry point when the
212 * relocation is complete.
213 * In kexec case, kimage->start points to purgatory assuming that
214 * kernel entry and dtb address are embedded in purgatory by
215 * userspace (kexec-tools).
216 * In kexec_file case, the kernel starts directly without purgatory.
218 cpu_soft_restart(reboot_code_buffer_phys, kimage->head, kimage->start,
219 #ifdef CONFIG_KEXEC_FILE
220 kimage->arch.dtb_mem);
225 BUG(); /* Should never get here. */
228 static void machine_kexec_mask_interrupts(void)
231 struct irq_desc *desc;
233 for_each_irq_desc(i, desc) {
234 struct irq_chip *chip;
237 chip = irq_desc_get_chip(desc);
242 * First try to remove the active state. If this
243 * fails, try to EOI the interrupt.
245 ret = irq_set_irqchip_state(i, IRQCHIP_STATE_ACTIVE, false);
247 if (ret && irqd_irq_inprogress(&desc->irq_data) &&
249 chip->irq_eoi(&desc->irq_data);
252 chip->irq_mask(&desc->irq_data);
254 if (chip->irq_disable && !irqd_irq_disabled(&desc->irq_data))
255 chip->irq_disable(&desc->irq_data);
260 * machine_crash_shutdown - shutdown non-crashing cpus and save registers
262 void machine_crash_shutdown(struct pt_regs *regs)
266 /* shutdown non-crashing cpus */
267 crash_smp_send_stop();
269 /* for crashing cpu */
270 crash_save_cpu(regs, smp_processor_id());
271 machine_kexec_mask_interrupts();
273 pr_info("Starting crashdump kernel...\n");
276 void arch_kexec_protect_crashkres(void)
280 kexec_segment_flush(kexec_crash_image);
282 for (i = 0; i < kexec_crash_image->nr_segments; i++)
284 __phys_to_virt(kexec_crash_image->segment[i].mem),
285 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 0);
288 void arch_kexec_unprotect_crashkres(void)
292 for (i = 0; i < kexec_crash_image->nr_segments; i++)
294 __phys_to_virt(kexec_crash_image->segment[i].mem),
295 kexec_crash_image->segment[i].memsz >> PAGE_SHIFT, 1);
298 #ifdef CONFIG_HIBERNATION
300 * To preserve the crash dump kernel image, the relevant memory segments
301 * should be mapped again around the hibernation.
303 void crash_prepare_suspend(void)
305 if (kexec_crash_image)
306 arch_kexec_unprotect_crashkres();
309 void crash_post_resume(void)
311 if (kexec_crash_image)
312 arch_kexec_protect_crashkres();
318 * Return true only if a page is part of reserved memory for crash dump kernel,
319 * but does not hold any data of loaded kernel image.
321 * Note that all the pages in crash dump kernel memory have been initially
322 * marked as Reserved as memory was allocated via memblock_reserve().
324 * In hibernation, the pages which are Reserved and yet "nosave" are excluded
325 * from the hibernation iamge. crash_is_nosave() does thich check for crash
326 * dump kernel and will reduce the total size of hibernation image.
329 bool crash_is_nosave(unsigned long pfn)
337 /* in reserved memory? */
338 addr = __pfn_to_phys(pfn);
339 if ((addr < crashk_res.start) || (crashk_res.end < addr))
342 if (!kexec_crash_image)
345 /* not part of loaded kernel image? */
346 for (i = 0; i < kexec_crash_image->nr_segments; i++)
347 if (addr >= kexec_crash_image->segment[i].mem &&
348 addr < (kexec_crash_image->segment[i].mem +
349 kexec_crash_image->segment[i].memsz))
355 void crash_free_reserved_phys_range(unsigned long begin, unsigned long end)
360 for (addr = begin; addr < end; addr += PAGE_SIZE) {
361 page = phys_to_page(addr);
362 free_reserved_page(page);
365 #endif /* CONFIG_HIBERNATION */