2 * Low-level CPU initialisation
3 * Based on arch/arm/kernel/head.S
5 * Copyright (C) 1994-2002 Russell King
6 * Copyright (C) 2003-2012 ARM Ltd.
7 * Authors: Catalin Marinas <catalin.marinas@arm.com>
8 * Will Deacon <will.deacon@arm.com>
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2 as
12 * published by the Free Software Foundation.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program. If not, see <http://www.gnu.org/licenses/>.
23 #include <linux/linkage.h>
24 #include <linux/init.h>
26 #include <asm/assembler.h>
27 #include <asm/ptrace.h>
28 #include <asm/asm-offsets.h>
29 #include <asm/cputype.h>
30 #include <asm/memory.h>
31 #include <asm/thread_info.h>
32 #include <asm/pgtable-hwdef.h>
33 #include <asm/pgtable.h>
38 * swapper_pg_dir is the virtual address of the initial page table. We place
39 * the page tables 3 * PAGE_SIZE below KERNEL_RAM_VADDR. The idmap_pg_dir has
40 * 2 pages and is placed below swapper_pg_dir.
42 #define KERNEL_RAM_VADDR (PAGE_OFFSET + TEXT_OFFSET)
44 #if (KERNEL_RAM_VADDR & 0xfffff) != 0x80000
45 #error KERNEL_RAM_VADDR must start at 0xXXX80000
48 #define SWAPPER_DIR_SIZE (3 * PAGE_SIZE)
49 #define IDMAP_DIR_SIZE (2 * PAGE_SIZE)
52 .equ swapper_pg_dir, KERNEL_RAM_VADDR - SWAPPER_DIR_SIZE
55 .equ idmap_pg_dir, swapper_pg_dir - IDMAP_DIR_SIZE
57 .macro pgtbl, ttb0, ttb1, phys
58 add \ttb1, \phys, #TEXT_OFFSET - SWAPPER_DIR_SIZE
59 sub \ttb0, \ttb1, #IDMAP_DIR_SIZE
62 #ifdef CONFIG_ARM64_64K_PAGES
63 #define BLOCK_SHIFT PAGE_SHIFT
64 #define BLOCK_SIZE PAGE_SIZE
66 #define BLOCK_SHIFT SECTION_SHIFT
67 #define BLOCK_SIZE SECTION_SIZE
70 #define KERNEL_START KERNEL_RAM_VADDR
71 #define KERNEL_END _end
74 * Initial memory map attributes.
77 #define PTE_FLAGS PTE_TYPE_PAGE | PTE_AF
78 #define PMD_FLAGS PMD_TYPE_SECT | PMD_SECT_AF
80 #define PTE_FLAGS PTE_TYPE_PAGE | PTE_AF | PTE_SHARED
81 #define PMD_FLAGS PMD_TYPE_SECT | PMD_SECT_AF | PMD_SECT_S
84 #ifdef CONFIG_ARM64_64K_PAGES
85 #define MM_MMUFLAGS PTE_ATTRINDX(MT_NORMAL) | PTE_FLAGS
87 #define MM_MMUFLAGS PMD_ATTRINDX(MT_NORMAL) | PMD_FLAGS
91 * Kernel startup entry point.
92 * ---------------------------
94 * The requirements are:
95 * MMU = off, D-cache = off, I-cache = on or off,
96 * x0 = physical address to the FDT blob.
98 * This code is mostly position independent so you call this at
99 * __pa(PAGE_OFFSET + TEXT_OFFSET).
101 * Note that the callee-saved registers are used for storing variables
102 * that are useful before the MMU is enabled. The allocations are described
103 * in the entry routines.
108 * DO NOT MODIFY. Image header expected by Linux boot-loaders.
110 b stext // branch to kernel start, magic
112 .quad TEXT_OFFSET // Image load offset from start of RAM
118 .byte 0x41 // Magic number, "ARM\x64"
125 mov x21, x0 // x21=FDT
126 bl el2_setup // Drop to EL1, w20=cpu_boot_mode
127 bl __calc_phys_offset // x24=PHYS_OFFSET, x28=PHYS_OFFSET-PAGE_OFFSET
128 bl set_cpu_boot_mode_flag
129 mrs x22, midr_el1 // x22=cpuid
131 bl lookup_processor_type
132 mov x23, x0 // x23=current cpu_table
133 cbz x23, __error_p // invalid processor (x23=0)?
135 bl __create_page_tables // x25=TTBR0, x26=TTBR1
137 * The following calls CPU specific code in a position independent
138 * manner. See arch/arm64/mm/proc.S for details. x23 = base of
139 * cpu_info structure selected by lookup_processor_type above.
140 * On return, the CPU will be ready for the MMU to be turned on and
141 * the TCR will have been set.
143 ldr x27, __switch_data // address to jump to after
144 // MMU has been enabled
145 adr lr, __enable_mmu // return (PIC) address
146 ldr x12, [x23, #CPU_INFO_SETUP]
147 add x12, x12, x28 // __virt_to_phys
148 br x12 // initialise processor
152 * If we're fortunate enough to boot at EL2, ensure that the world is
153 * sane before dropping to EL1.
155 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x20 if
156 * booted in EL1 or EL2 respectively.
160 cmp x0, #PSR_MODE_EL2t
161 ccmp x0, #PSR_MODE_EL2h, #0x4, ne
164 CPU_BE( orr x0, x0, #(1 << 25) ) // Set the EE bit for EL2
165 CPU_LE( bic x0, x0, #(1 << 25) ) // Clear the EE bit for EL2
169 CPU_BE( orr x0, x0, #(3 << 24) ) // Set the EE and E0E bits for EL1
170 CPU_LE( bic x0, x0, #(3 << 24) ) // Clear the EE and E0E bits for EL1
172 mov w20, #BOOT_CPU_MODE_EL1 // This cpu booted in EL1
176 /* Hyp configuration. */
177 2: mov x0, #(1 << 31) // 64-bit EL1
180 /* Generic timers. */
182 orr x0, x0, #3 // Enable EL1 physical timers
184 msr cntvoff_el2, xzr // Clear virtual offset
186 /* Populate ID registers. */
193 mov x0, #0x0800 // Set/clear RES{1,0} bits
194 CPU_BE( movk x0, #0x33d0, lsl #16 ) // Set EE and E0E on BE systems
195 CPU_LE( movk x0, #0x30d0, lsl #16 ) // Clear EE and E0E on LE systems
198 /* Coprocessor traps. */
200 msr cptr_el2, x0 // Disable copro. traps to EL2
203 msr hstr_el2, xzr // Disable CP15 traps to EL2
206 /* Stage-2 translation */
209 /* Hypervisor stub */
210 adr x0, __hyp_stub_vectors
214 mov x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
218 mov w20, #BOOT_CPU_MODE_EL2 // This CPU booted in EL2
223 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
224 * in x20. See arch/arm64/include/asm/virt.h for more info.
226 ENTRY(set_cpu_boot_mode_flag)
227 ldr x1, =__boot_cpu_mode // Compute __boot_cpu_mode
229 cmp w20, #BOOT_CPU_MODE_EL2
232 1: str w20, [x1] // This CPU has booted in EL1
234 ENDPROC(set_cpu_boot_mode_flag)
237 * We need to find out the CPU boot mode long after boot, so we need to
238 * store it in a writable variable.
240 * This is not in .bss, because we set it sufficiently early that the boot-time
241 * zeroing of .bss would clobber it.
244 ENTRY(__boot_cpu_mode)
245 .long BOOT_CPU_MODE_EL2
256 .quad secondary_holding_pen_release
259 * This provides a "holding pen" for platforms to hold all secondary
260 * cores are held until we're ready for them to initialise.
262 ENTRY(secondary_holding_pen)
263 bl el2_setup // Drop to EL1, w20=cpu_boot_mode
264 bl __calc_phys_offset // x24=PHYS_OFFSET, x28=PHYS_OFFSET-PAGE_OFFSET
265 bl set_cpu_boot_mode_flag
267 ldr x1, =MPIDR_HWID_BITMASK
275 b.eq secondary_startup
278 ENDPROC(secondary_holding_pen)
281 * Secondary entry point that jumps straight into the kernel. Only to
282 * be used where CPUs are brought online dynamically by the kernel.
284 ENTRY(secondary_entry)
285 bl el2_setup // Drop to EL1
286 bl __calc_phys_offset // x24=PHYS_OFFSET, x28=PHYS_OFFSET-PAGE_OFFSET
287 bl set_cpu_boot_mode_flag
289 ENDPROC(secondary_entry)
291 ENTRY(secondary_startup)
293 * Common entry point for secondary CPUs.
295 mrs x22, midr_el1 // x22=cpuid
297 bl lookup_processor_type
298 mov x23, x0 // x23=current cpu_table
299 cbz x23, __error_p // invalid processor (x23=0)?
301 pgtbl x25, x26, x24 // x25=TTBR0, x26=TTBR1
302 ldr x12, [x23, #CPU_INFO_SETUP]
303 add x12, x12, x28 // __virt_to_phys
304 blr x12 // initialise processor
306 ldr x21, =secondary_data
307 ldr x27, =__secondary_switched // address to jump to after enabling the MMU
309 ENDPROC(secondary_startup)
311 ENTRY(__secondary_switched)
312 ldr x0, [x21] // get secondary_data.stack
315 b secondary_start_kernel
316 ENDPROC(__secondary_switched)
317 #endif /* CONFIG_SMP */
320 * Setup common bits before finally enabling the MMU. Essentially this is just
321 * loading the page table pointer and vector base registers.
323 * On entry to this code, x0 must contain the SCTLR_EL1 value for turning on
329 msr ttbr0_el1, x25 // load TTBR0
330 msr ttbr1_el1, x26 // load TTBR1
333 ENDPROC(__enable_mmu)
336 * Enable the MMU. This completely changes the structure of the visible memory
337 * space. You will not be able to trace execution through this.
339 * x0 = system control register
340 * x27 = *virtual* address to jump to upon completion
342 * other registers depend on the function called upon completion
349 ENDPROC(__turn_mmu_on)
352 * Calculate the start of physical memory.
357 sub x28, x0, x1 // x28 = PHYS_OFFSET - PAGE_OFFSET
358 add x24, x2, x28 // x24 = PHYS_OFFSET
360 ENDPROC(__calc_phys_offset)
367 * Macro to populate the PGD for the corresponding block entry in the next
368 * level (tbl) for the given virtual address.
370 * Preserves: pgd, tbl, virt
371 * Corrupts: tmp1, tmp2
373 .macro create_pgd_entry, pgd, tbl, virt, tmp1, tmp2
374 lsr \tmp1, \virt, #PGDIR_SHIFT
375 and \tmp1, \tmp1, #PTRS_PER_PGD - 1 // PGD index
376 orr \tmp2, \tbl, #3 // PGD entry table type
377 str \tmp2, [\pgd, \tmp1, lsl #3]
381 * Macro to populate block entries in the page table for the start..end
382 * virtual range (inclusive).
384 * Preserves: tbl, flags
385 * Corrupts: phys, start, end, pstate
387 .macro create_block_map, tbl, flags, phys, start, end
388 lsr \phys, \phys, #BLOCK_SHIFT
389 lsr \start, \start, #BLOCK_SHIFT
390 and \start, \start, #PTRS_PER_PTE - 1 // table index
391 orr \phys, \flags, \phys, lsl #BLOCK_SHIFT // table entry
392 lsr \end, \end, #BLOCK_SHIFT
393 and \end, \end, #PTRS_PER_PTE - 1 // table end index
394 9999: str \phys, [\tbl, \start, lsl #3] // store the entry
395 add \start, \start, #1 // next entry
396 add \phys, \phys, #BLOCK_SIZE // next block
402 * Setup the initial page tables. We only setup the barest amount which is
403 * required to get the kernel running. The following sections are required:
404 * - identity mapping to enable the MMU (low address, TTBR0)
405 * - first few MB of the kernel linear mapping to jump to once the MMU has
406 * been enabled, including the FDT blob (TTBR1)
407 * - pgd entry for fixed mappings (TTBR1)
409 __create_page_tables:
410 pgtbl x25, x26, x24 // idmap_pg_dir and swapper_pg_dir addresses
413 * Clear the idmap and swapper page tables.
416 add x6, x26, #SWAPPER_DIR_SIZE
417 1: stp xzr, xzr, [x0], #16
418 stp xzr, xzr, [x0], #16
419 stp xzr, xzr, [x0], #16
420 stp xzr, xzr, [x0], #16
427 * Create the identity mapping.
429 add x0, x25, #PAGE_SIZE // section table address
430 ldr x3, =KERNEL_START
431 add x3, x3, x28 // __pa(KERNEL_START)
432 create_pgd_entry x25, x0, x3, x5, x6
434 mov x5, x3 // __pa(KERNEL_START)
435 add x6, x6, x28 // __pa(KERNEL_END)
436 create_block_map x0, x7, x3, x5, x6
439 * Map the kernel image (starting with PHYS_OFFSET).
441 add x0, x26, #PAGE_SIZE // section table address
443 create_pgd_entry x26, x0, x5, x3, x6
445 mov x3, x24 // phys offset
446 create_block_map x0, x7, x3, x5, x6
449 * Map the FDT blob (maximum 2MB; must be within 512MB of
452 mov x3, x21 // FDT phys address
453 and x3, x3, #~((1 << 21) - 1) // 2MB aligned
455 sub x5, x3, x24 // subtract PHYS_OFFSET
456 tst x5, #~((1 << 29) - 1) // within 512MB?
457 csel x21, xzr, x21, ne // zero the FDT pointer
459 add x5, x5, x6 // __va(FDT blob)
460 add x6, x5, #1 << 21 // 2MB for the FDT blob
461 sub x6, x6, #1 // inclusive range
462 create_block_map x0, x7, x3, x5, x6
465 * Create the pgd entry for the fixed mappings.
467 ldr x5, =FIXADDR_TOP // Fixed mapping virtual address
468 add x0, x26, #2 * PAGE_SIZE // section table address
469 create_pgd_entry x26, x0, x5, x6, x7
471 ENDPROC(__create_page_tables)
475 .type __switch_data, %object
477 .quad __mmap_switched
478 .quad __bss_start // x6
480 .quad processor_id // x4
481 .quad __fdt_pointer // x5
482 .quad memstart_addr // x6
483 .quad init_thread_union + THREAD_START_SP // sp
486 * The following fragment of code is executed with the MMU on in MMU mode, and
487 * uses absolute addresses; this is not position independent.
490 adr x3, __switch_data + 8
492 ldp x6, x7, [x3], #16
495 str xzr, [x6], #8 // Clear BSS
498 ldp x4, x5, [x3], #16
502 str x22, [x4] // Save processor ID
503 str x21, [x5] // Save FDT pointer
504 str x24, [x6] // Save PHYS_OFFSET
507 ENDPROC(__mmap_switched)
510 * Exception handling. Something went wrong and we can't proceed. We ought to
511 * tell the user, but since we don't have any guarantee that we're even
512 * running on the right architecture, we do virtually nothing.
523 * This function gets the processor ID in w0 and searches the cpu_table[] for
524 * a match. It returns a pointer to the struct cpu_info it found. The
525 * cpu_table[] must end with an empty (all zeros) structure.
527 * This routine can be called via C code and it needs to work with the MMU
528 * both disabled and enabled (the offset is calculated automatically).
530 ENTRY(lookup_processor_type)
531 adr x1, __lookup_processor_type_data
533 sub x1, x1, x2 // get offset between VA and PA
534 add x3, x3, x1 // convert VA to PA
536 ldp w5, w6, [x3] // load cpu_id_val and cpu_id_mask
537 cbz w5, 2f // end of list?
541 add x3, x3, #CPU_INFO_SZ
544 mov x3, #0 // unknown processor
548 ENDPROC(lookup_processor_type)
551 .type __lookup_processor_type_data, %object
552 __lookup_processor_type_data:
555 .size __lookup_processor_type_data, . - __lookup_processor_type_data
558 * Determine validity of the x21 FDT pointer.
559 * The dtb must be 8-byte aligned and live in the first 512M of memory.