1 // SPDX-License-Identifier: GPL-2.0-only
3 * Architecture specific (i386/x86_64) functions for kexec based crash dumps.
5 * Created by: Hariprasad Nellitheertha (hari@in.ibm.com)
7 * Copyright (C) IBM Corporation, 2004. All rights reserved.
8 * Copyright (C) Red Hat Inc., 2014. All rights reserved.
10 * Vivek Goyal <vgoyal@redhat.com>
14 #define pr_fmt(fmt) "kexec: " fmt
16 #include <linux/types.h>
17 #include <linux/kernel.h>
18 #include <linux/smp.h>
19 #include <linux/reboot.h>
20 #include <linux/kexec.h>
21 #include <linux/delay.h>
22 #include <linux/elf.h>
23 #include <linux/elfcore.h>
24 #include <linux/export.h>
25 #include <linux/slab.h>
26 #include <linux/vmalloc.h>
27 #include <linux/memblock.h>
29 #include <asm/processor.h>
30 #include <asm/hardirq.h>
32 #include <asm/hw_irq.h>
34 #include <asm/e820/types.h>
35 #include <asm/io_apic.h>
37 #include <linux/kdebug.h>
39 #include <asm/reboot.h>
40 #include <asm/intel_pt.h>
41 #include <asm/crash.h>
42 #include <asm/cmdline.h>
44 /* Used while preparing memory map entries for second kernel */
45 struct crash_memmap_data {
46 struct boot_params *params;
51 #if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC)
53 static void kdump_nmi_callback(int cpu, struct pt_regs *regs)
55 crash_save_cpu(regs, cpu);
58 * Disable Intel PT to stop its logging
60 cpu_emergency_stop_pt();
65 void kdump_nmi_shootdown_cpus(void)
67 nmi_shootdown_cpus(kdump_nmi_callback);
72 /* Override the weak function in kernel/panic.c */
73 void crash_smp_send_stop(void)
75 static int cpus_stopped;
80 if (smp_ops.crash_stop_other_cpus)
81 smp_ops.crash_stop_other_cpus();
89 void crash_smp_send_stop(void)
91 /* There are no cpus to shootdown */
95 void native_machine_crash_shutdown(struct pt_regs *regs)
97 /* This function is only called after the system
98 * has panicked or is otherwise in a critical state.
99 * The minimum amount of code to allow a kexec'd kernel
100 * to run successfully needs to happen here.
102 * In practice this means shooting down the other cpus in
105 /* The kernel is broken so disable interrupts */
108 crash_smp_send_stop();
110 cpu_emergency_disable_virtualization();
113 * Disable Intel PT to stop its logging
115 cpu_emergency_stop_pt();
117 #ifdef CONFIG_X86_IO_APIC
118 /* Prevent crash_kexec() from deadlocking on ioapic_lock. */
123 restore_boot_irq_mode();
124 #ifdef CONFIG_HPET_TIMER
127 crash_save_cpu(regs, safe_smp_processor_id());
130 #ifdef CONFIG_KEXEC_FILE
132 static int get_nr_ram_ranges_callback(struct resource *res, void *arg)
134 unsigned int *nr_ranges = arg;
140 /* Gather all the required information to prepare elf headers for ram regions */
141 static struct crash_mem *fill_up_crash_elf_data(void)
143 unsigned int nr_ranges = 0;
144 struct crash_mem *cmem;
146 walk_system_ram_res(0, -1, &nr_ranges, get_nr_ram_ranges_callback);
151 * Exclusion of crash region and/or crashk_low_res may cause
152 * another range split. So add extra two slots here.
155 cmem = vzalloc(struct_size(cmem, ranges, nr_ranges));
159 cmem->max_nr_ranges = nr_ranges;
166 * Look for any unwanted ranges between mstart, mend and remove them. This
167 * might lead to split and split ranges are put in cmem->ranges[] array
169 static int elf_header_exclude_ranges(struct crash_mem *cmem)
173 /* Exclude the low 1M because it is always reserved */
174 ret = crash_exclude_mem_range(cmem, 0, (1<<20)-1);
178 /* Exclude crashkernel region */
179 ret = crash_exclude_mem_range(cmem, crashk_res.start, crashk_res.end);
183 if (crashk_low_res.end)
184 ret = crash_exclude_mem_range(cmem, crashk_low_res.start,
190 static int prepare_elf64_ram_headers_callback(struct resource *res, void *arg)
192 struct crash_mem *cmem = arg;
194 cmem->ranges[cmem->nr_ranges].start = res->start;
195 cmem->ranges[cmem->nr_ranges].end = res->end;
201 /* Prepare elf headers. Return addr and size */
202 static int prepare_elf_headers(struct kimage *image, void **addr,
205 struct crash_mem *cmem;
208 cmem = fill_up_crash_elf_data();
212 ret = walk_system_ram_res(0, -1, cmem, prepare_elf64_ram_headers_callback);
216 /* Exclude unwanted mem ranges */
217 ret = elf_header_exclude_ranges(cmem);
221 /* By default prepare 64bit headers */
222 ret = crash_prepare_elf64_headers(cmem, IS_ENABLED(CONFIG_X86_64), addr, sz);
229 static int add_e820_entry(struct boot_params *params, struct e820_entry *entry)
231 unsigned int nr_e820_entries;
233 nr_e820_entries = params->e820_entries;
234 if (nr_e820_entries >= E820_MAX_ENTRIES_ZEROPAGE)
237 memcpy(¶ms->e820_table[nr_e820_entries], entry, sizeof(struct e820_entry));
238 params->e820_entries++;
242 static int memmap_entry_callback(struct resource *res, void *arg)
244 struct crash_memmap_data *cmd = arg;
245 struct boot_params *params = cmd->params;
246 struct e820_entry ei;
248 ei.addr = res->start;
249 ei.size = resource_size(res);
251 add_e820_entry(params, &ei);
256 static int memmap_exclude_ranges(struct kimage *image, struct crash_mem *cmem,
257 unsigned long long mstart,
258 unsigned long long mend)
260 unsigned long start, end;
262 cmem->ranges[0].start = mstart;
263 cmem->ranges[0].end = mend;
266 /* Exclude elf header region */
267 start = image->elf_load_addr;
268 end = start + image->elf_headers_sz - 1;
269 return crash_exclude_mem_range(cmem, start, end);
272 /* Prepare memory map for crash dump kernel */
273 int crash_setup_memmap_entries(struct kimage *image, struct boot_params *params)
277 struct e820_entry ei;
278 struct crash_memmap_data cmd;
279 struct crash_mem *cmem;
281 cmem = vzalloc(struct_size(cmem, ranges, 1));
285 memset(&cmd, 0, sizeof(struct crash_memmap_data));
289 cmd.type = E820_TYPE_RAM;
290 flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
291 walk_iomem_res_desc(IORES_DESC_NONE, flags, 0, (1<<20)-1, &cmd,
292 memmap_entry_callback);
294 /* Add ACPI tables */
295 cmd.type = E820_TYPE_ACPI;
296 flags = IORESOURCE_MEM | IORESOURCE_BUSY;
297 walk_iomem_res_desc(IORES_DESC_ACPI_TABLES, flags, 0, -1, &cmd,
298 memmap_entry_callback);
300 /* Add ACPI Non-volatile Storage */
301 cmd.type = E820_TYPE_NVS;
302 walk_iomem_res_desc(IORES_DESC_ACPI_NV_STORAGE, flags, 0, -1, &cmd,
303 memmap_entry_callback);
305 /* Add e820 reserved ranges */
306 cmd.type = E820_TYPE_RESERVED;
307 flags = IORESOURCE_MEM;
308 walk_iomem_res_desc(IORES_DESC_RESERVED, flags, 0, -1, &cmd,
309 memmap_entry_callback);
311 /* Add crashk_low_res region */
312 if (crashk_low_res.end) {
313 ei.addr = crashk_low_res.start;
314 ei.size = resource_size(&crashk_low_res);
315 ei.type = E820_TYPE_RAM;
316 add_e820_entry(params, &ei);
319 /* Exclude some ranges from crashk_res and add rest to memmap */
320 ret = memmap_exclude_ranges(image, cmem, crashk_res.start, crashk_res.end);
324 for (i = 0; i < cmem->nr_ranges; i++) {
325 ei.size = cmem->ranges[i].end - cmem->ranges[i].start + 1;
327 /* If entry is less than a page, skip it */
328 if (ei.size < PAGE_SIZE)
330 ei.addr = cmem->ranges[i].start;
331 ei.type = E820_TYPE_RAM;
332 add_e820_entry(params, &ei);
340 int crash_load_segments(struct kimage *image)
343 struct kexec_buf kbuf = { .image = image, .buf_min = 0,
344 .buf_max = ULONG_MAX, .top_down = false };
346 /* Prepare elf headers and add a segment */
347 ret = prepare_elf_headers(image, &kbuf.buffer, &kbuf.bufsz);
351 image->elf_headers = kbuf.buffer;
352 image->elf_headers_sz = kbuf.bufsz;
354 kbuf.memsz = kbuf.bufsz;
355 kbuf.buf_align = ELF_CORE_HEADER_ALIGN;
356 kbuf.mem = KEXEC_BUF_MEM_UNKNOWN;
357 ret = kexec_add_buffer(&kbuf);
360 image->elf_load_addr = kbuf.mem;
361 pr_debug("Loaded ELF headers at 0x%lx bufsz=0x%lx memsz=0x%lx\n",
362 image->elf_load_addr, kbuf.bufsz, kbuf.memsz);
366 #endif /* CONFIG_KEXEC_FILE */