xref: /openbmc/linux/arch/arm64/kernel/setup.c (revision d2574c33)
1 /*
2  * Based on arch/arm/kernel/setup.c
3  *
4  * Copyright (C) 1995-2001 Russell King
5  * Copyright (C) 2012 ARM Ltd.
6  *
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.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include <linux/acpi.h>
21 #include <linux/export.h>
22 #include <linux/kernel.h>
23 #include <linux/stddef.h>
24 #include <linux/ioport.h>
25 #include <linux/delay.h>
26 #include <linux/initrd.h>
27 #include <linux/console.h>
28 #include <linux/cache.h>
29 #include <linux/screen_info.h>
30 #include <linux/init.h>
31 #include <linux/kexec.h>
32 #include <linux/root_dev.h>
33 #include <linux/cpu.h>
34 #include <linux/interrupt.h>
35 #include <linux/smp.h>
36 #include <linux/fs.h>
37 #include <linux/proc_fs.h>
38 #include <linux/memblock.h>
39 #include <linux/of_fdt.h>
40 #include <linux/efi.h>
41 #include <linux/psci.h>
42 #include <linux/sched/task.h>
43 #include <linux/mm.h>
44 
45 #include <asm/acpi.h>
46 #include <asm/fixmap.h>
47 #include <asm/cpu.h>
48 #include <asm/cputype.h>
49 #include <asm/daifflags.h>
50 #include <asm/elf.h>
51 #include <asm/cpufeature.h>
52 #include <asm/cpu_ops.h>
53 #include <asm/kasan.h>
54 #include <asm/numa.h>
55 #include <asm/sections.h>
56 #include <asm/setup.h>
57 #include <asm/smp_plat.h>
58 #include <asm/cacheflush.h>
59 #include <asm/tlbflush.h>
60 #include <asm/traps.h>
61 #include <asm/efi.h>
62 #include <asm/xen/hypervisor.h>
63 #include <asm/mmu_context.h>
64 
65 static int num_standard_resources;
66 static struct resource *standard_resources;
67 
68 phys_addr_t __fdt_pointer __initdata;
69 
70 /*
71  * Standard memory resources
72  */
73 static struct resource mem_res[] = {
74 	{
75 		.name = "Kernel code",
76 		.start = 0,
77 		.end = 0,
78 		.flags = IORESOURCE_SYSTEM_RAM
79 	},
80 	{
81 		.name = "Kernel data",
82 		.start = 0,
83 		.end = 0,
84 		.flags = IORESOURCE_SYSTEM_RAM
85 	}
86 };
87 
88 #define kernel_code mem_res[0]
89 #define kernel_data mem_res[1]
90 
91 /*
92  * The recorded values of x0 .. x3 upon kernel entry.
93  */
94 u64 __cacheline_aligned boot_args[4];
95 
96 void __init smp_setup_processor_id(void)
97 {
98 	u64 mpidr = read_cpuid_mpidr() & MPIDR_HWID_BITMASK;
99 	cpu_logical_map(0) = mpidr;
100 
101 	/*
102 	 * clear __my_cpu_offset on boot CPU to avoid hang caused by
103 	 * using percpu variable early, for example, lockdep will
104 	 * access percpu variable inside lock_release
105 	 */
106 	set_my_cpu_offset(0);
107 	pr_info("Booting Linux on physical CPU 0x%010lx [0x%08x]\n",
108 		(unsigned long)mpidr, read_cpuid_id());
109 }
110 
111 bool arch_match_cpu_phys_id(int cpu, u64 phys_id)
112 {
113 	return phys_id == cpu_logical_map(cpu);
114 }
115 
116 struct mpidr_hash mpidr_hash;
117 /**
118  * smp_build_mpidr_hash - Pre-compute shifts required at each affinity
119  *			  level in order to build a linear index from an
120  *			  MPIDR value. Resulting algorithm is a collision
121  *			  free hash carried out through shifting and ORing
122  */
123 static void __init smp_build_mpidr_hash(void)
124 {
125 	u32 i, affinity, fs[4], bits[4], ls;
126 	u64 mask = 0;
127 	/*
128 	 * Pre-scan the list of MPIDRS and filter out bits that do
129 	 * not contribute to affinity levels, ie they never toggle.
130 	 */
131 	for_each_possible_cpu(i)
132 		mask |= (cpu_logical_map(i) ^ cpu_logical_map(0));
133 	pr_debug("mask of set bits %#llx\n", mask);
134 	/*
135 	 * Find and stash the last and first bit set at all affinity levels to
136 	 * check how many bits are required to represent them.
137 	 */
138 	for (i = 0; i < 4; i++) {
139 		affinity = MPIDR_AFFINITY_LEVEL(mask, i);
140 		/*
141 		 * Find the MSB bit and LSB bits position
142 		 * to determine how many bits are required
143 		 * to express the affinity level.
144 		 */
145 		ls = fls(affinity);
146 		fs[i] = affinity ? ffs(affinity) - 1 : 0;
147 		bits[i] = ls - fs[i];
148 	}
149 	/*
150 	 * An index can be created from the MPIDR_EL1 by isolating the
151 	 * significant bits at each affinity level and by shifting
152 	 * them in order to compress the 32 bits values space to a
153 	 * compressed set of values. This is equivalent to hashing
154 	 * the MPIDR_EL1 through shifting and ORing. It is a collision free
155 	 * hash though not minimal since some levels might contain a number
156 	 * of CPUs that is not an exact power of 2 and their bit
157 	 * representation might contain holes, eg MPIDR_EL1[7:0] = {0x2, 0x80}.
158 	 */
159 	mpidr_hash.shift_aff[0] = MPIDR_LEVEL_SHIFT(0) + fs[0];
160 	mpidr_hash.shift_aff[1] = MPIDR_LEVEL_SHIFT(1) + fs[1] - bits[0];
161 	mpidr_hash.shift_aff[2] = MPIDR_LEVEL_SHIFT(2) + fs[2] -
162 						(bits[1] + bits[0]);
163 	mpidr_hash.shift_aff[3] = MPIDR_LEVEL_SHIFT(3) +
164 				  fs[3] - (bits[2] + bits[1] + bits[0]);
165 	mpidr_hash.mask = mask;
166 	mpidr_hash.bits = bits[3] + bits[2] + bits[1] + bits[0];
167 	pr_debug("MPIDR hash: aff0[%u] aff1[%u] aff2[%u] aff3[%u] mask[%#llx] bits[%u]\n",
168 		mpidr_hash.shift_aff[0],
169 		mpidr_hash.shift_aff[1],
170 		mpidr_hash.shift_aff[2],
171 		mpidr_hash.shift_aff[3],
172 		mpidr_hash.mask,
173 		mpidr_hash.bits);
174 	/*
175 	 * 4x is an arbitrary value used to warn on a hash table much bigger
176 	 * than expected on most systems.
177 	 */
178 	if (mpidr_hash_size() > 4 * num_possible_cpus())
179 		pr_warn("Large number of MPIDR hash buckets detected\n");
180 }
181 
182 static void __init setup_machine_fdt(phys_addr_t dt_phys)
183 {
184 	void *dt_virt = fixmap_remap_fdt(dt_phys);
185 	const char *name;
186 
187 	if (!dt_virt || !early_init_dt_scan(dt_virt)) {
188 		pr_crit("\n"
189 			"Error: invalid device tree blob at physical address %pa (virtual address 0x%p)\n"
190 			"The dtb must be 8-byte aligned and must not exceed 2 MB in size\n"
191 			"\nPlease check your bootloader.",
192 			&dt_phys, dt_virt);
193 
194 		while (true)
195 			cpu_relax();
196 	}
197 
198 	name = of_flat_dt_get_machine_name();
199 	if (!name)
200 		return;
201 
202 	pr_info("Machine model: %s\n", name);
203 	dump_stack_set_arch_desc("%s (DT)", name);
204 }
205 
206 static void __init request_standard_resources(void)
207 {
208 	struct memblock_region *region;
209 	struct resource *res;
210 	unsigned long i = 0;
211 	size_t res_size;
212 
213 	kernel_code.start   = __pa_symbol(_text);
214 	kernel_code.end     = __pa_symbol(__init_begin - 1);
215 	kernel_data.start   = __pa_symbol(_sdata);
216 	kernel_data.end     = __pa_symbol(_end - 1);
217 
218 	num_standard_resources = memblock.memory.cnt;
219 	res_size = num_standard_resources * sizeof(*standard_resources);
220 	standard_resources = memblock_alloc_low(res_size, SMP_CACHE_BYTES);
221 	if (!standard_resources)
222 		panic("%s: Failed to allocate %zu bytes\n", __func__, res_size);
223 
224 	for_each_memblock(memory, region) {
225 		res = &standard_resources[i++];
226 		if (memblock_is_nomap(region)) {
227 			res->name  = "reserved";
228 			res->flags = IORESOURCE_MEM;
229 		} else {
230 			res->name  = "System RAM";
231 			res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
232 		}
233 		res->start = __pfn_to_phys(memblock_region_memory_base_pfn(region));
234 		res->end = __pfn_to_phys(memblock_region_memory_end_pfn(region)) - 1;
235 
236 		request_resource(&iomem_resource, res);
237 
238 		if (kernel_code.start >= res->start &&
239 		    kernel_code.end <= res->end)
240 			request_resource(res, &kernel_code);
241 		if (kernel_data.start >= res->start &&
242 		    kernel_data.end <= res->end)
243 			request_resource(res, &kernel_data);
244 #ifdef CONFIG_KEXEC_CORE
245 		/* Userspace will find "Crash kernel" region in /proc/iomem. */
246 		if (crashk_res.end && crashk_res.start >= res->start &&
247 		    crashk_res.end <= res->end)
248 			request_resource(res, &crashk_res);
249 #endif
250 	}
251 }
252 
253 static int __init reserve_memblock_reserved_regions(void)
254 {
255 	u64 i, j;
256 
257 	for (i = 0; i < num_standard_resources; ++i) {
258 		struct resource *mem = &standard_resources[i];
259 		phys_addr_t r_start, r_end, mem_size = resource_size(mem);
260 
261 		if (!memblock_is_region_reserved(mem->start, mem_size))
262 			continue;
263 
264 		for_each_reserved_mem_region(j, &r_start, &r_end) {
265 			resource_size_t start, end;
266 
267 			start = max(PFN_PHYS(PFN_DOWN(r_start)), mem->start);
268 			end = min(PFN_PHYS(PFN_UP(r_end)) - 1, mem->end);
269 
270 			if (start > mem->end || end < mem->start)
271 				continue;
272 
273 			reserve_region_with_split(mem, start, end, "reserved");
274 		}
275 	}
276 
277 	return 0;
278 }
279 arch_initcall(reserve_memblock_reserved_regions);
280 
281 u64 __cpu_logical_map[NR_CPUS] = { [0 ... NR_CPUS-1] = INVALID_HWID };
282 
283 void __init setup_arch(char **cmdline_p)
284 {
285 	init_mm.start_code = (unsigned long) _text;
286 	init_mm.end_code   = (unsigned long) _etext;
287 	init_mm.end_data   = (unsigned long) _edata;
288 	init_mm.brk	   = (unsigned long) _end;
289 
290 	*cmdline_p = boot_command_line;
291 
292 	early_fixmap_init();
293 	early_ioremap_init();
294 
295 	setup_machine_fdt(__fdt_pointer);
296 
297 	parse_early_param();
298 
299 	/*
300 	 * Unmask asynchronous aborts and fiq after bringing up possible
301 	 * earlycon. (Report possible System Errors once we can report this
302 	 * occurred).
303 	 */
304 	local_daif_restore(DAIF_PROCCTX_NOIRQ);
305 
306 	/*
307 	 * TTBR0 is only used for the identity mapping at this stage. Make it
308 	 * point to zero page to avoid speculatively fetching new entries.
309 	 */
310 	cpu_uninstall_idmap();
311 
312 	xen_early_init();
313 	efi_init();
314 	arm64_memblock_init();
315 
316 	paging_init();
317 
318 	acpi_table_upgrade();
319 
320 	/* Parse the ACPI tables for possible boot-time configuration */
321 	acpi_boot_table_init();
322 
323 	if (acpi_disabled)
324 		unflatten_device_tree();
325 
326 	bootmem_init();
327 
328 	kasan_init();
329 
330 	request_standard_resources();
331 
332 	early_ioremap_reset();
333 
334 	if (acpi_disabled)
335 		psci_dt_init();
336 	else
337 		psci_acpi_init();
338 
339 	cpu_read_bootcpu_ops();
340 	smp_init_cpus();
341 	smp_build_mpidr_hash();
342 
343 	/* Init percpu seeds for random tags after cpus are set up. */
344 	kasan_init_tags();
345 
346 #ifdef CONFIG_ARM64_SW_TTBR0_PAN
347 	/*
348 	 * Make sure init_thread_info.ttbr0 always generates translation
349 	 * faults in case uaccess_enable() is inadvertently called by the init
350 	 * thread.
351 	 */
352 	init_task.thread_info.ttbr0 = __pa_symbol(empty_zero_page);
353 #endif
354 
355 #ifdef CONFIG_VT
356 	conswitchp = &dummy_con;
357 #endif
358 	if (boot_args[1] || boot_args[2] || boot_args[3]) {
359 		pr_err("WARNING: x1-x3 nonzero in violation of boot protocol:\n"
360 			"\tx1: %016llx\n\tx2: %016llx\n\tx3: %016llx\n"
361 			"This indicates a broken bootloader or old kernel\n",
362 			boot_args[1], boot_args[2], boot_args[3]);
363 	}
364 }
365 
366 static int __init topology_init(void)
367 {
368 	int i;
369 
370 	for_each_online_node(i)
371 		register_one_node(i);
372 
373 	for_each_possible_cpu(i) {
374 		struct cpu *cpu = &per_cpu(cpu_data.cpu, i);
375 		cpu->hotpluggable = 1;
376 		register_cpu(cpu, i);
377 	}
378 
379 	return 0;
380 }
381 subsys_initcall(topology_init);
382 
383 /*
384  * Dump out kernel offset information on panic.
385  */
386 static int dump_kernel_offset(struct notifier_block *self, unsigned long v,
387 			      void *p)
388 {
389 	const unsigned long offset = kaslr_offset();
390 
391 	if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && offset > 0) {
392 		pr_emerg("Kernel Offset: 0x%lx from 0x%lx\n",
393 			 offset, KIMAGE_VADDR);
394 		pr_emerg("PHYS_OFFSET: 0x%llx\n", PHYS_OFFSET);
395 	} else {
396 		pr_emerg("Kernel Offset: disabled\n");
397 	}
398 	return 0;
399 }
400 
401 static struct notifier_block kernel_offset_notifier = {
402 	.notifier_call = dump_kernel_offset
403 };
404 
405 static int __init register_kernel_offset_dumper(void)
406 {
407 	atomic_notifier_chain_register(&panic_notifier_list,
408 				       &kernel_offset_notifier);
409 	return 0;
410 }
411 __initcall(register_kernel_offset_dumper);
412