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
117 mov x21, x0 // x21=FDT
118 bl __calc_phys_offset // x24=PHYS_OFFSET, x28=PHYS_OFFSET-PAGE_OFFSET
119 bl el2_setup // Drop to EL1
120 mrs x22, midr_el1 // x22=cpuid
122 bl lookup_processor_type
123 mov x23, x0 // x23=current cpu_table
124 cbz x23, __error_p // invalid processor (x23=0)?
126 bl __create_page_tables // x25=TTBR0, x26=TTBR1
128 * The following calls CPU specific code in a position independent
129 * manner. See arch/arm64/mm/proc.S for details. x23 = base of
130 * cpu_info structure selected by lookup_processor_type above.
131 * On return, the CPU will be ready for the MMU to be turned on and
132 * the TCR will have been set.
134 ldr x27, __switch_data // address to jump to after
135 // MMU has been enabled
136 adr lr, __enable_mmu // return (PIC) address
137 ldr x12, [x23, #CPU_INFO_SETUP]
138 add x12, x12, x28 // __virt_to_phys
139 br x12 // initialise processor
143 * If we're fortunate enough to boot at EL2, ensure that the world is
144 * sane before dropping to EL1.
148 cmp x0, #PSR_MODE_EL2t
149 ccmp x0, #PSR_MODE_EL2h, #0x4, ne
150 ldr x0, =__boot_cpu_mode // Compute __boot_cpu_mode
153 str wzr, [x0] // Remember we don't have EL2...
156 /* Hyp configuration. */
157 1: ldr w1, =BOOT_CPU_MODE_EL2
158 str w1, [x0, #4] // This CPU has EL2
159 mov x0, #(1 << 31) // 64-bit EL1
162 /* Generic timers. */
164 orr x0, x0, #3 // Enable EL1 physical timers
166 msr cntvoff_el2, xzr // Clear virtual offset
168 /* Populate ID registers. */
175 mov x0, #0x0800 // Set/clear RES{1,0} bits
176 movk x0, #0x30d0, lsl #16
179 /* Coprocessor traps. */
181 msr cptr_el2, x0 // Disable copro. traps to EL2
184 msr hstr_el2, xzr // Disable CP15 traps to EL2
187 /* Stage-2 translation */
190 /* Hypervisor stub */
191 adr x0, __hyp_stub_vectors
195 mov x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
203 * We need to find out the CPU boot mode long after boot, so we need to
204 * store it in a writable variable.
206 * This is not in .bss, because we set it sufficiently early that the boot-time
207 * zeroing of .bss would clobber it.
210 ENTRY(__boot_cpu_mode)
211 .long BOOT_CPU_MODE_EL2
222 .quad secondary_holding_pen_release
226 CCI400_SI3_SNOOP_CONTROL:
233 * This provides a "holding pen" for platforms to hold all secondary
234 * cores are held until we're ready for them to initialise.
236 ENTRY(secondary_holding_pen)
245 ldr w2, CCI400_SI3_SNOOP_CONTROL
249 ldr w2, CCI400_STATUS
259 bl __calc_phys_offset // x24=phys offset
260 bl el2_setup // Drop to EL1
262 ldr x1, =MPIDR_HWID_BITMASK
270 b.eq secondary_startup
273 ENDPROC(secondary_holding_pen)
276 * Secondary entry point that jumps straight into the kernel. Only to
277 * be used where CPUs are brought online dynamically by the kernel.
279 ENTRY(secondary_entry)
280 bl __calc_phys_offset // x2=phys offset
281 bl el2_setup // Drop to EL1
283 ENDPROC(secondary_entry)
285 ENTRY(secondary_startup)
287 * Common entry point for secondary CPUs.
289 mrs x22, midr_el1 // x22=cpuid
291 bl lookup_processor_type
292 mov x23, x0 // x23=current cpu_table
293 cbz x23, __error_p // invalid processor (x23=0)?
295 pgtbl x25, x26, x24 // x25=TTBR0, x26=TTBR1
296 ldr x12, [x23, #CPU_INFO_SETUP]
297 add x12, x12, x28 // __virt_to_phys
298 blr x12 // initialise processor
300 ldr x21, =secondary_data
301 ldr x27, =__secondary_switched // address to jump to after enabling the MMU
303 ENDPROC(secondary_startup)
305 ENTRY(__secondary_switched)
306 ldr x0, [x21] // get secondary_data.stack
309 b secondary_start_kernel
310 ENDPROC(__secondary_switched)
311 #endif /* CONFIG_SMP */
314 * Setup common bits before finally enabling the MMU. Essentially this is just
315 * loading the page table pointer and vector base registers.
317 * On entry to this code, x0 must contain the SCTLR_EL1 value for turning on
323 msr ttbr0_el1, x25 // load TTBR0
324 msr ttbr1_el1, x26 // load TTBR1
327 ENDPROC(__enable_mmu)
330 * Enable the MMU. This completely changes the structure of the visible memory
331 * space. You will not be able to trace execution through this.
333 * x0 = system control register
334 * x27 = *virtual* address to jump to upon completion
336 * other registers depend on the function called upon completion
343 ENDPROC(__turn_mmu_on)
346 * Calculate the start of physical memory.
348 ENTRY(__calc_phys_offset)
351 sub x28, x0, x1 // x28 = PHYS_OFFSET - PAGE_OFFSET
352 add x24, x2, x28 // x24 = PHYS_OFFSET
354 ENDPROC(__calc_phys_offset)
361 * Macro to populate the PGD for the corresponding block entry in the next
362 * level (tbl) for the given virtual address.
364 * Preserves: pgd, tbl, virt
365 * Corrupts: tmp1, tmp2
367 .macro create_pgd_entry, pgd, tbl, virt, tmp1, tmp2
368 lsr \tmp1, \virt, #PGDIR_SHIFT
369 and \tmp1, \tmp1, #PTRS_PER_PGD - 1 // PGD index
370 orr \tmp2, \tbl, #3 // PGD entry table type
371 str \tmp2, [\pgd, \tmp1, lsl #3]
375 * Macro to populate block entries in the page table for the start..end
376 * virtual range (inclusive).
378 * Preserves: tbl, flags
379 * Corrupts: phys, start, end, pstate
381 .macro create_block_map, tbl, flags, phys, start, end, idmap=0
382 lsr \phys, \phys, #BLOCK_SHIFT
384 and \start, \phys, #PTRS_PER_PTE - 1 // table index
386 lsr \start, \start, #BLOCK_SHIFT
387 and \start, \start, #PTRS_PER_PTE - 1 // table index
389 orr \phys, \flags, \phys, lsl #BLOCK_SHIFT // table entry
391 lsr \end, \end, #BLOCK_SHIFT
392 and \end, \end, #PTRS_PER_PTE - 1 // table end index
394 9999: str \phys, [\tbl, \start, lsl #3] // store the entry
396 add \start, \start, #1 // next entry
397 add \phys, \phys, #BLOCK_SIZE // next block
404 * Setup the initial page tables. We only setup the barest amount which is
405 * required to get the kernel running. The following sections are required:
406 * - identity mapping to enable the MMU (low address, TTBR0)
407 * - first few MB of the kernel linear mapping to jump to once the MMU has
408 * been enabled, including the FDT blob (TTBR1)
409 * - UART mapping if CONFIG_EARLY_PRINTK is enabled (TTBR1)
411 __create_page_tables:
412 pgtbl x25, x26, x24 // idmap_pg_dir and swapper_pg_dir addresses
415 * Clear the idmap and swapper page tables.
418 add x6, x26, #SWAPPER_DIR_SIZE
419 1: stp xzr, xzr, [x0], #16
420 stp xzr, xzr, [x0], #16
421 stp xzr, xzr, [x0], #16
422 stp xzr, xzr, [x0], #16
429 * Create the identity mapping.
431 add x0, x25, #PAGE_SIZE // section table address
432 adr x3, __turn_mmu_on // virtual/physical address
433 create_pgd_entry x25, x0, x3, x5, x6
434 create_block_map x0, x7, x3, x5, x5, idmap=1
437 * Map the kernel image (starting with PHYS_OFFSET).
439 add x0, x26, #PAGE_SIZE // section table address
441 create_pgd_entry x26, x0, x5, x3, x6
442 ldr x6, =KERNEL_END - 1
443 mov x3, x24 // phys offset
444 create_block_map x0, x7, x3, x5, x6
447 * Map the FDT blob (maximum 2MB; must be within 512MB of
450 mov x3, x21 // FDT phys address
451 and x3, x3, #~((1 << 21) - 1) // 2MB aligned
453 sub x5, x3, x24 // subtract PHYS_OFFSET
454 tst x5, #~((1 << 29) - 1) // within 512MB?
455 csel x21, xzr, x21, ne // zero the FDT pointer
457 add x5, x5, x6 // __va(FDT blob)
458 add x6, x5, #1 << 21 // 2MB for the FDT blob
459 sub x6, x6, #1 // inclusive range
460 create_block_map x0, x7, x3, x5, x6
462 #ifdef CONFIG_EARLY_PRINTK
464 * Create the pgd entry for the UART mapping. The full mapping is done
465 * later based earlyprintk kernel parameter.
467 ldr x5, =EARLYCON_IOBASE // UART virtual address
468 add x0, x26, #2 * PAGE_SIZE // section table address
469 create_pgd_entry x26, x0, x5, x6, x7
472 ENDPROC(__create_page_tables)
476 .type __switch_data, %object
478 .quad __mmap_switched
479 .quad __data_loc // x4
481 .quad __bss_start // x6
483 .quad processor_id // x4
484 .quad __fdt_pointer // x5
485 .quad memstart_addr // x6
486 .quad init_thread_union + THREAD_START_SP // sp
489 * The following fragment of code is executed with the MMU on in MMU mode, and
490 * uses absolute addresses; this is not position independent.
493 adr x3, __switch_data + 8
495 ldp x4, x5, [x3], #16
496 ldp x6, x7, [x3], #16
497 cmp x4, x5 // Copy data segment if needed
498 1: ccmp x5, x6, #4, ne
506 str xzr, [x6], #8 // Clear BSS
509 ldp x4, x5, [x3], #16
513 str x22, [x4] // Save processor ID
514 str x21, [x5] // Save FDT pointer
515 str x24, [x6] // Save PHYS_OFFSET
518 ENDPROC(__mmap_switched)
521 * Exception handling. Something went wrong and we can't proceed. We ought to
522 * tell the user, but since we don't have any guarantee that we're even
523 * running on the right architecture, we do virtually nothing.
534 * This function gets the processor ID in w0 and searches the cpu_table[] for
535 * a match. It returns a pointer to the struct cpu_info it found. The
536 * cpu_table[] must end with an empty (all zeros) structure.
538 * This routine can be called via C code and it needs to work with the MMU
539 * both disabled and enabled (the offset is calculated automatically).
541 ENTRY(lookup_processor_type)
542 adr x1, __lookup_processor_type_data
544 sub x1, x1, x2 // get offset between VA and PA
545 add x3, x3, x1 // convert VA to PA
547 ldp w5, w6, [x3] // load cpu_id_val and cpu_id_mask
548 cbz w5, 2f // end of list?
552 add x3, x3, #CPU_INFO_SZ
555 mov x3, #0 // unknown processor
559 ENDPROC(lookup_processor_type)
562 .type __lookup_processor_type_data, %object
563 __lookup_processor_type_data:
566 .size __lookup_processor_type_data, . - __lookup_processor_type_data
569 * Determine validity of the x21 FDT pointer.
570 * The dtb must be 8-byte aligned and live in the first 512M of memory.