2 * linux/arch/arm/mm/proc-xscale.S
4 * Author: Nicolas Pitre
5 * Created: November 2000
6 * Copyright: (C) 2000, 2001 MontaVista Software Inc.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
12 * MMU functions for the Intel XScale CPUs
15 * some contributions by Brett Gaines <brett.w.gaines@intel.com>
16 * Copyright 2001 by Intel Corp.
19 * Completely revisited, many important fixes
20 * Nicolas Pitre <nico@cam.org>
23 #include <linux/linkage.h>
24 #include <linux/init.h>
25 #include <asm/assembler.h>
26 #include <asm/procinfo.h>
27 #include <asm/hardware.h>
28 #include <asm/pgtable.h>
30 #include <asm/ptrace.h>
31 #include "proc-macros.S"
34 * This is the maximum size of an area which will be flushed. If the area
35 * is larger than this, then we flush the whole cache
37 #define MAX_AREA_SIZE 32768
40 * the cache line size of the I and D cache
42 #define CACHELINESIZE 32
45 * the size of the data cache
47 #define CACHESIZE 32768
50 * Virtual address used to allocate the cache when flushed
52 * This must be an address range which is _never_ used. It should
53 * apparently have a mapping in the corresponding page table for
54 * compatibility with future CPUs that _could_ require it. For instance we
57 * This must be aligned on a 2*CACHESIZE boundary. The code selects one of
58 * the 2 areas in alternance each time the clean_d_cache macro is used.
59 * Without this the XScale core exhibits cache eviction problems and no one
62 * Reminder: the vector table is located at 0xffff0000-0xffff0fff.
64 #define CLEAN_ADDR 0xfffe0000
67 * This macro is used to wait for a CP15 write and is needed
68 * when we have to ensure that the last operation to the co-pro
69 * was completed before continuing with operation.
72 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15
73 mov \rd, \rd @ wait for completion
74 sub pc, pc, #4 @ flush instruction pipeline
77 .macro cpwait_ret, lr, rd
78 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15
79 sub pc, \lr, \rd, LSR #32 @ wait for completion and
80 @ flush instruction pipeline
84 * This macro cleans the entire dcache using line allocate.
85 * The main loop has been unrolled to reduce loop overhead.
86 * rd and rs are two scratch registers.
88 .macro clean_d_cache, rd, rs
91 eor \rd, \rd, #CACHESIZE
93 add \rs, \rd, #CACHESIZE
94 1: mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
95 add \rd, \rd, #CACHELINESIZE
96 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
97 add \rd, \rd, #CACHELINESIZE
98 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
99 add \rd, \rd, #CACHELINESIZE
100 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
101 add \rd, \rd, #CACHELINESIZE
107 clean_addr: .word CLEAN_ADDR
112 * cpu_xscale_proc_init()
114 * Nothing too exciting at the moment
116 ENTRY(cpu_xscale_proc_init)
120 * cpu_xscale_proc_fin()
122 ENTRY(cpu_xscale_proc_fin)
124 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
126 bl xscale_flush_kern_cache_all @ clean caches
127 mrc p15, 0, r0, c1, c0, 0 @ ctrl register
128 bic r0, r0, #0x1800 @ ...IZ...........
129 bic r0, r0, #0x0006 @ .............CA.
130 mcr p15, 0, r0, c1, c0, 0 @ disable caches
134 * cpu_xscale_reset(loc)
136 * Perform a soft reset of the system. Put the CPU into the
137 * same state as it would be if it had been reset, and branch
138 * to what would be the reset vector.
140 * loc: location to jump to for soft reset
143 ENTRY(cpu_xscale_reset)
144 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
145 msr cpsr_c, r1 @ reset CPSR
146 mrc p15, 0, r1, c1, c0, 0 @ ctrl register
147 bic r1, r1, #0x0086 @ ........B....CA.
148 bic r1, r1, #0x3900 @ ..VIZ..S........
149 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
150 mcr p15, 0, ip, c7, c7, 0 @ invalidate I,D caches & BTB
151 bic r1, r1, #0x0001 @ ...............M
152 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
153 @ CAUTION: MMU turned off from this point. We count on the pipeline
154 @ already containing those two last instructions to survive.
155 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
159 * cpu_xscale_do_idle()
161 * Cause the processor to idle
163 * For now we do nothing but go to idle mode for every case
165 * XScale supports clock switching, but using idle mode support
166 * allows external hardware to react to system state changes.
170 ENTRY(cpu_xscale_do_idle)
172 mcr p14, 0, r0, c7, c0, 0 @ Go to IDLE
175 /* ================================= CACHE ================================ */
178 * flush_user_cache_all()
180 * Invalidate all cache entries in a particular address
183 ENTRY(xscale_flush_user_cache_all)
187 * flush_kern_cache_all()
189 * Clean and invalidate the entire cache.
191 ENTRY(xscale_flush_kern_cache_all)
197 mcrne p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
198 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
202 * flush_user_cache_range(start, end, vm_flags)
204 * Invalidate a range of cache entries in the specified
207 * - start - start address (may not be aligned)
208 * - end - end address (exclusive, may not be aligned)
209 * - vma - vma_area_struct describing address space
212 ENTRY(xscale_flush_user_cache_range)
214 sub r3, r1, r0 @ calculate total size
215 cmp r3, #MAX_AREA_SIZE
216 bhs __flush_whole_cache
219 mcrne p15, 0, r0, c7, c5, 1 @ Invalidate I cache line
220 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
221 mcr p15, 0, r0, c7, c6, 1 @ Invalidate D cache line
222 add r0, r0, #CACHELINESIZE
226 mcrne p15, 0, ip, c7, c5, 6 @ Invalidate BTB
227 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
231 * coherent_kern_range(start, end)
233 * Ensure coherency between the Icache and the Dcache in the
234 * region described by start. If you have non-snooping
235 * Harvard caches, you need to implement this function.
237 * - start - virtual start address
238 * - end - virtual end address
240 * Note: single I-cache line invalidation isn't used here since
241 * it also trashes the mini I-cache used by JTAG debuggers.
243 ENTRY(xscale_coherent_kern_range)
244 bic r0, r0, #CACHELINESIZE - 1
245 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
246 add r0, r0, #CACHELINESIZE
250 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
251 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
255 * coherent_user_range(start, end)
257 * Ensure coherency between the Icache and the Dcache in the
258 * region described by start. If you have non-snooping
259 * Harvard caches, you need to implement this function.
261 * - start - virtual start address
262 * - end - virtual end address
264 ENTRY(xscale_coherent_user_range)
265 bic r0, r0, #CACHELINESIZE - 1
266 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
267 mcr p15, 0, r0, c7, c5, 1 @ Invalidate I cache entry
268 add r0, r0, #CACHELINESIZE
272 mcr p15, 0, r0, c7, c5, 6 @ Invalidate BTB
273 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
277 * flush_kern_dcache_page(void *page)
279 * Ensure no D cache aliasing occurs, either with itself or
282 * - addr - page aligned address
284 ENTRY(xscale_flush_kern_dcache_page)
286 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
287 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
288 add r0, r0, #CACHELINESIZE
292 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
293 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
297 * dma_inv_range(start, end)
299 * Invalidate (discard) the specified virtual address range.
300 * May not write back any entries. If 'start' or 'end'
301 * are not cache line aligned, those lines must be written
304 * - start - virtual start address
305 * - end - virtual end address
307 ENTRY(xscale_dma_inv_range)
308 mrc p15, 0, r2, c0, c0, 0 @ read ID
309 eor r2, r2, #0x69000000
310 eor r2, r2, #0x00052000
312 beq xscale_dma_flush_range
314 tst r0, #CACHELINESIZE - 1
315 bic r0, r0, #CACHELINESIZE - 1
316 mcrne p15, 0, r0, c7, c10, 1 @ clean D entry
317 tst r1, #CACHELINESIZE - 1
318 mcrne p15, 0, r1, c7, c10, 1 @ clean D entry
319 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
320 add r0, r0, #CACHELINESIZE
323 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
327 * dma_clean_range(start, end)
329 * Clean the specified virtual address range.
331 * - start - virtual start address
332 * - end - virtual end address
334 ENTRY(xscale_dma_clean_range)
335 bic r0, r0, #CACHELINESIZE - 1
336 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
337 add r0, r0, #CACHELINESIZE
340 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
344 * dma_flush_range(start, end)
346 * Clean and invalidate the specified virtual address range.
348 * - start - virtual start address
349 * - end - virtual end address
351 ENTRY(xscale_dma_flush_range)
352 bic r0, r0, #CACHELINESIZE - 1
353 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
354 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
355 add r0, r0, #CACHELINESIZE
358 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
361 ENTRY(xscale_cache_fns)
362 .long xscale_flush_kern_cache_all
363 .long xscale_flush_user_cache_all
364 .long xscale_flush_user_cache_range
365 .long xscale_coherent_kern_range
366 .long xscale_coherent_user_range
367 .long xscale_flush_kern_dcache_page
368 .long xscale_dma_inv_range
369 .long xscale_dma_clean_range
370 .long xscale_dma_flush_range
372 ENTRY(cpu_xscale_dcache_clean_area)
373 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
374 add r0, r0, #CACHELINESIZE
375 subs r1, r1, #CACHELINESIZE
379 /* =============================== PageTable ============================== */
381 #define PTE_CACHE_WRITE_ALLOCATE 0
384 * cpu_xscale_switch_mm(pgd)
386 * Set the translation base pointer to be as described by pgd.
388 * pgd: new page tables
391 ENTRY(cpu_xscale_switch_mm)
393 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
394 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
395 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
396 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
400 * cpu_xscale_set_pte(ptep, pte)
402 * Set a PTE and flush it out
404 * Errata 40: must set memory to write-through for user read-only pages.
407 ENTRY(cpu_xscale_set_pte)
408 str r1, [r0], #-2048 @ linux version
411 orr r2, r2, #PTE_TYPE_EXT @ extended page
413 eor r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
415 tst r3, #L_PTE_USER @ User?
416 orrne r2, r2, #PTE_EXT_AP_URO_SRW @ yes -> user r/o, system r/w
418 tst r3, #L_PTE_WRITE | L_PTE_DIRTY @ Write and Dirty?
419 orreq r2, r2, #PTE_EXT_AP_UNO_SRW @ yes -> user n/a, system r/w
420 @ combined with user -> user r/w
423 @ Handle the X bit. We want to set this bit for the minicache
424 @ (U = E = B = W = 0, C = 1) or when write allocate is enabled,
425 @ and we have a writeable, cacheable region. If we ignore the
426 @ U and E bits, we can allow user space to use the minicache as
429 @ X = (C & ~W & ~B) | (C & W & B & write_allocate)
431 eor ip, r1, #L_PTE_CACHEABLE
432 tst ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
433 #if PTE_CACHE_WRITE_ALLOCATE
434 eorne ip, r1, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
435 tstne ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
437 orreq r2, r2, #PTE_EXT_TEX(1)
440 @ Erratum 40: The B bit must be cleared for a user read-only
443 @ B = B & ~(U & C & ~W)
445 and ip, r1, #L_PTE_USER | L_PTE_WRITE | L_PTE_CACHEABLE
446 teq ip, #L_PTE_USER | L_PTE_CACHEABLE
447 biceq r2, r2, #PTE_BUFFERABLE
449 tst r3, #L_PTE_PRESENT | L_PTE_YOUNG @ Present and Young?
450 movne r2, #0 @ no -> fault
452 str r2, [r0] @ hardware version
454 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
455 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
465 .type __xscale_setup, #function
467 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB
468 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
469 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs
471 mov r0, #0 @ initially disallow access to CP0/CP1
473 mov r0, #1 @ Allow access to CP0
475 orr r0, r0, #1 << 6 @ cp6 for IOP3xx and Bulverde
476 orr r0, r0, #1 << 13 @ Its undefined whether this
477 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes
478 mrc p15, 0, r0, c1, c0, 0 @ get control register
479 ldr r5, xscale_cr1_clear
481 ldr r5, xscale_cr1_set
484 .size __xscale_setup, . - __xscale_setup
488 * .RVI ZFRS BLDP WCAM
489 * ..11 1.01 .... .101
492 .type xscale_cr1_clear, #object
493 .type xscale_cr1_set, #object
502 * Purpose : Function pointers used to access above functions - all calls
506 .type xscale_processor_functions, #object
507 ENTRY(xscale_processor_functions)
508 .word v5t_early_abort
509 .word cpu_xscale_proc_init
510 .word cpu_xscale_proc_fin
511 .word cpu_xscale_reset
512 .word cpu_xscale_do_idle
513 .word cpu_xscale_dcache_clean_area
514 .word cpu_xscale_switch_mm
515 .word cpu_xscale_set_pte
516 .size xscale_processor_functions, . - xscale_processor_functions
520 .type cpu_arch_name, #object
523 .size cpu_arch_name, . - cpu_arch_name
525 .type cpu_elf_name, #object
528 .size cpu_elf_name, . - cpu_elf_name
530 .type cpu_80200_name, #object
532 .asciz "XScale-80200"
533 .size cpu_80200_name, . - cpu_80200_name
535 .type cpu_8032x_name, #object
537 .asciz "XScale-IOP8032x Family"
538 .size cpu_8032x_name, . - cpu_8032x_name
540 .type cpu_8033x_name, #object
542 .asciz "XScale-IOP8033x Family"
543 .size cpu_8033x_name, . - cpu_8033x_name
545 .type cpu_pxa250_name, #object
547 .asciz "XScale-PXA250"
548 .size cpu_pxa250_name, . - cpu_pxa250_name
550 .type cpu_pxa210_name, #object
552 .asciz "XScale-PXA210"
553 .size cpu_pxa210_name, . - cpu_pxa210_name
555 .type cpu_ixp42x_name, #object
557 .asciz "XScale-IXP42x Family"
558 .size cpu_ixp42x_name, . - cpu_ixp42x_name
560 .type cpu_ixp46x_name, #object
562 .asciz "XScale-IXP46x Family"
563 .size cpu_ixp46x_name, . - cpu_ixp46x_name
565 .type cpu_ixp2400_name, #object
567 .asciz "XScale-IXP2400"
568 .size cpu_ixp2400_name, . - cpu_ixp2400_name
570 .type cpu_ixp2800_name, #object
572 .asciz "XScale-IXP2800"
573 .size cpu_ixp2800_name, . - cpu_ixp2800_name
575 .type cpu_pxa255_name, #object
577 .asciz "XScale-PXA255"
578 .size cpu_pxa255_name, . - cpu_pxa255_name
580 .type cpu_pxa270_name, #object
582 .asciz "XScale-PXA270"
583 .size cpu_pxa270_name, . - cpu_pxa270_name
587 .section ".proc.info.init", #alloc, #execinstr
589 .type __80200_proc_info,#object
593 .long PMD_TYPE_SECT | \
594 PMD_SECT_BUFFERABLE | \
595 PMD_SECT_CACHEABLE | \
596 PMD_SECT_AP_WRITE | \
601 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
603 .long xscale_processor_functions
605 .long xscale_mc_user_fns
606 .long xscale_cache_fns
607 .size __80200_proc_info, . - __80200_proc_info
609 .type __8032x_proc_info,#object
612 .long 0xfffff5e0 @ mask should accomodate IOP80219 also
613 .long PMD_TYPE_SECT | \
614 PMD_SECT_BUFFERABLE | \
615 PMD_SECT_CACHEABLE | \
616 PMD_SECT_AP_WRITE | \
621 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
623 .long xscale_processor_functions
625 .long xscale_mc_user_fns
626 .long xscale_cache_fns
627 .size __8032x_proc_info, . - __8032x_proc_info
629 .type __8033x_proc_info,#object
633 .long PMD_TYPE_SECT | \
634 PMD_SECT_BUFFERABLE | \
635 PMD_SECT_CACHEABLE | \
636 PMD_SECT_AP_WRITE | \
641 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
643 .long xscale_processor_functions
645 .long xscale_mc_user_fns
646 .long xscale_cache_fns
647 .size __8033x_proc_info, . - __8033x_proc_info
649 .type __pxa250_proc_info,#object
653 .long PMD_TYPE_SECT | \
654 PMD_SECT_BUFFERABLE | \
655 PMD_SECT_CACHEABLE | \
656 PMD_SECT_AP_WRITE | \
661 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
662 .long cpu_pxa250_name
663 .long xscale_processor_functions
665 .long xscale_mc_user_fns
666 .long xscale_cache_fns
667 .size __pxa250_proc_info, . - __pxa250_proc_info
669 .type __pxa210_proc_info,#object
673 .long PMD_TYPE_SECT | \
674 PMD_SECT_BUFFERABLE | \
675 PMD_SECT_CACHEABLE | \
676 PMD_SECT_AP_WRITE | \
681 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
682 .long cpu_pxa210_name
683 .long xscale_processor_functions
685 .long xscale_mc_user_fns
686 .long xscale_cache_fns
687 .size __pxa210_proc_info, . - __pxa210_proc_info
689 .type __ixp2400_proc_info, #object
693 .long PMD_TYPE_SECT | \
694 PMD_SECT_BUFFERABLE | \
695 PMD_SECT_CACHEABLE | \
696 PMD_SECT_AP_WRITE | \
701 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
702 .long cpu_ixp2400_name
703 .long xscale_processor_functions
705 .long xscale_mc_user_fns
706 .long xscale_cache_fns
707 .size __ixp2400_proc_info, . - __ixp2400_proc_info
709 .type __ixp2800_proc_info, #object
713 .long PMD_TYPE_SECT | \
714 PMD_SECT_BUFFERABLE | \
715 PMD_SECT_CACHEABLE | \
716 PMD_SECT_AP_WRITE | \
721 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
722 .long cpu_ixp2800_name
723 .long xscale_processor_functions
725 .long xscale_mc_user_fns
726 .long xscale_cache_fns
727 .size __ixp2800_proc_info, . - __ixp2800_proc_info
729 .type __ixp42x_proc_info, #object
733 .long PMD_TYPE_SECT | \
734 PMD_SECT_BUFFERABLE | \
735 PMD_SECT_CACHEABLE | \
736 PMD_SECT_AP_WRITE | \
741 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
742 .long cpu_ixp42x_name
743 .long xscale_processor_functions
745 .long xscale_mc_user_fns
746 .long xscale_cache_fns
747 .size __ixp42x_proc_info, . - __ixp42x_proc_info
749 .type __ixp46x_proc_info, #object
757 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
758 .long cpu_ixp46x_name
759 .long xscale_processor_functions
761 .long xscale_mc_user_fns
762 .long xscale_cache_fns
763 .size __ixp46x_proc_info, . - __ixp46x_proc_info
765 .type __pxa255_proc_info,#object
769 .long PMD_TYPE_SECT | \
770 PMD_SECT_BUFFERABLE | \
771 PMD_SECT_CACHEABLE | \
772 PMD_SECT_AP_WRITE | \
777 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
778 .long cpu_pxa255_name
779 .long xscale_processor_functions
781 .long xscale_mc_user_fns
782 .long xscale_cache_fns
783 .size __pxa255_proc_info, . - __pxa255_proc_info
785 .type __pxa270_proc_info,#object
789 .long PMD_TYPE_SECT | \
790 PMD_SECT_BUFFERABLE | \
791 PMD_SECT_CACHEABLE | \
792 PMD_SECT_AP_WRITE | \
797 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
798 .long cpu_pxa270_name
799 .long xscale_processor_functions
801 .long xscale_mc_user_fns
802 .long xscale_cache_fns
803 .size __pxa270_proc_info, . - __pxa270_proc_info