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)
247 * coherent_user_range(start, end)
249 * Ensure coherency between the Icache and the Dcache in the
250 * region described by start. If you have non-snooping
251 * Harvard caches, you need to implement this function.
253 * - start - virtual start address
254 * - end - virtual end address
256 * Note: single I-cache line invalidation isn't used here since
257 * it also trashes the mini I-cache used by JTAG debuggers.
259 ENTRY(xscale_coherent_user_range)
260 bic r0, r0, #CACHELINESIZE - 1
261 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
262 add r0, r0, #CACHELINESIZE
266 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
267 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
271 * flush_kern_dcache_page(void *page)
273 * Ensure no D cache aliasing occurs, either with itself or
276 * - addr - page aligned address
278 ENTRY(xscale_flush_kern_dcache_page)
280 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
281 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
282 add r0, r0, #CACHELINESIZE
286 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
287 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
291 * dma_inv_range(start, end)
293 * Invalidate (discard) the specified virtual address range.
294 * May not write back any entries. If 'start' or 'end'
295 * are not cache line aligned, those lines must be written
298 * - start - virtual start address
299 * - end - virtual end address
301 ENTRY(xscale_dma_inv_range)
302 mrc p15, 0, r2, c0, c0, 0 @ read ID
303 eor r2, r2, #0x69000000
304 eor r2, r2, #0x00052000
306 beq xscale_dma_flush_range
308 tst r0, #CACHELINESIZE - 1
309 bic r0, r0, #CACHELINESIZE - 1
310 mcrne p15, 0, r0, c7, c10, 1 @ clean D entry
311 tst r1, #CACHELINESIZE - 1
312 mcrne p15, 0, r1, c7, c10, 1 @ clean D entry
313 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
314 add r0, r0, #CACHELINESIZE
317 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
321 * dma_clean_range(start, end)
323 * Clean the specified virtual address range.
325 * - start - virtual start address
326 * - end - virtual end address
328 ENTRY(xscale_dma_clean_range)
329 bic r0, r0, #CACHELINESIZE - 1
330 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
331 add r0, r0, #CACHELINESIZE
334 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
338 * dma_flush_range(start, end)
340 * Clean and invalidate the specified virtual address range.
342 * - start - virtual start address
343 * - end - virtual end address
345 ENTRY(xscale_dma_flush_range)
346 bic r0, r0, #CACHELINESIZE - 1
347 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
348 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
349 add r0, r0, #CACHELINESIZE
352 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
355 ENTRY(xscale_cache_fns)
356 .long xscale_flush_kern_cache_all
357 .long xscale_flush_user_cache_all
358 .long xscale_flush_user_cache_range
359 .long xscale_coherent_kern_range
360 .long xscale_coherent_user_range
361 .long xscale_flush_kern_dcache_page
362 .long xscale_dma_inv_range
363 .long xscale_dma_clean_range
364 .long xscale_dma_flush_range
366 ENTRY(cpu_xscale_dcache_clean_area)
367 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
368 add r0, r0, #CACHELINESIZE
369 subs r1, r1, #CACHELINESIZE
373 /* =============================== PageTable ============================== */
375 #define PTE_CACHE_WRITE_ALLOCATE 0
378 * cpu_xscale_switch_mm(pgd)
380 * Set the translation base pointer to be as described by pgd.
382 * pgd: new page tables
385 ENTRY(cpu_xscale_switch_mm)
387 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
388 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
389 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
390 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
394 * cpu_xscale_set_pte(ptep, pte)
396 * Set a PTE and flush it out
398 * Errata 40: must set memory to write-through for user read-only pages.
401 ENTRY(cpu_xscale_set_pte)
402 str r1, [r0], #-2048 @ linux version
405 orr r2, r2, #PTE_TYPE_EXT @ extended page
407 eor r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
409 tst r3, #L_PTE_USER @ User?
410 orrne r2, r2, #PTE_EXT_AP_URO_SRW @ yes -> user r/o, system r/w
412 tst r3, #L_PTE_WRITE | L_PTE_DIRTY @ Write and Dirty?
413 orreq r2, r2, #PTE_EXT_AP_UNO_SRW @ yes -> user n/a, system r/w
414 @ combined with user -> user r/w
417 @ Handle the X bit. We want to set this bit for the minicache
418 @ (U = E = B = W = 0, C = 1) or when write allocate is enabled,
419 @ and we have a writeable, cacheable region. If we ignore the
420 @ U and E bits, we can allow user space to use the minicache as
423 @ X = (C & ~W & ~B) | (C & W & B & write_allocate)
425 eor ip, r1, #L_PTE_CACHEABLE
426 tst ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
427 #if PTE_CACHE_WRITE_ALLOCATE
428 eorne ip, r1, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
429 tstne ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
431 orreq r2, r2, #PTE_EXT_TEX(1)
434 @ Erratum 40: The B bit must be cleared for a user read-only
437 @ B = B & ~(U & C & ~W)
439 and ip, r1, #L_PTE_USER | L_PTE_WRITE | L_PTE_CACHEABLE
440 teq ip, #L_PTE_USER | L_PTE_CACHEABLE
441 biceq r2, r2, #PTE_BUFFERABLE
443 tst r3, #L_PTE_PRESENT | L_PTE_YOUNG @ Present and Young?
444 movne r2, #0 @ no -> fault
446 str r2, [r0] @ hardware version
448 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
449 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
459 .type __xscale_setup, #function
461 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB
462 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
463 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs
465 mov r0, #0 @ initially disallow access to CP0/CP1
467 mov r0, #1 @ Allow access to CP0
469 orr r0, r0, #1 << 6 @ cp6 for IOP3xx and Bulverde
470 orr r0, r0, #1 << 13 @ Its undefined whether this
471 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes
472 mrc p15, 0, r0, c1, c0, 0 @ get control register
473 ldr r5, xscale_cr1_clear
475 ldr r5, xscale_cr1_set
478 .size __xscale_setup, . - __xscale_setup
482 * .RVI ZFRS BLDP WCAM
483 * ..11 1.01 .... .101
486 .type xscale_cr1_clear, #object
487 .type xscale_cr1_set, #object
496 * Purpose : Function pointers used to access above functions - all calls
500 .type xscale_processor_functions, #object
501 ENTRY(xscale_processor_functions)
502 .word v5t_early_abort
503 .word cpu_xscale_proc_init
504 .word cpu_xscale_proc_fin
505 .word cpu_xscale_reset
506 .word cpu_xscale_do_idle
507 .word cpu_xscale_dcache_clean_area
508 .word cpu_xscale_switch_mm
509 .word cpu_xscale_set_pte
510 .size xscale_processor_functions, . - xscale_processor_functions
514 .type cpu_arch_name, #object
517 .size cpu_arch_name, . - cpu_arch_name
519 .type cpu_elf_name, #object
522 .size cpu_elf_name, . - cpu_elf_name
524 .type cpu_80200_name, #object
526 .asciz "XScale-80200"
527 .size cpu_80200_name, . - cpu_80200_name
529 .type cpu_8032x_name, #object
531 .asciz "XScale-IOP8032x Family"
532 .size cpu_8032x_name, . - cpu_8032x_name
534 .type cpu_8033x_name, #object
536 .asciz "XScale-IOP8033x Family"
537 .size cpu_8033x_name, . - cpu_8033x_name
539 .type cpu_pxa250_name, #object
541 .asciz "XScale-PXA250"
542 .size cpu_pxa250_name, . - cpu_pxa250_name
544 .type cpu_pxa210_name, #object
546 .asciz "XScale-PXA210"
547 .size cpu_pxa210_name, . - cpu_pxa210_name
549 .type cpu_ixp42x_name, #object
551 .asciz "XScale-IXP42x Family"
552 .size cpu_ixp42x_name, . - cpu_ixp42x_name
554 .type cpu_ixp46x_name, #object
556 .asciz "XScale-IXP46x Family"
557 .size cpu_ixp46x_name, . - cpu_ixp46x_name
559 .type cpu_ixp2400_name, #object
561 .asciz "XScale-IXP2400"
562 .size cpu_ixp2400_name, . - cpu_ixp2400_name
564 .type cpu_ixp2800_name, #object
566 .asciz "XScale-IXP2800"
567 .size cpu_ixp2800_name, . - cpu_ixp2800_name
569 .type cpu_pxa255_name, #object
571 .asciz "XScale-PXA255"
572 .size cpu_pxa255_name, . - cpu_pxa255_name
574 .type cpu_pxa270_name, #object
576 .asciz "XScale-PXA270"
577 .size cpu_pxa270_name, . - cpu_pxa270_name
581 .section ".proc.info.init", #alloc, #execinstr
583 .type __80200_proc_info,#object
587 .long PMD_TYPE_SECT | \
588 PMD_SECT_BUFFERABLE | \
589 PMD_SECT_CACHEABLE | \
590 PMD_SECT_AP_WRITE | \
595 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
597 .long xscale_processor_functions
599 .long xscale_mc_user_fns
600 .long xscale_cache_fns
601 .size __80200_proc_info, . - __80200_proc_info
603 .type __8032x_proc_info,#object
606 .long 0xfffff5e0 @ mask should accomodate IOP80219 also
607 .long PMD_TYPE_SECT | \
608 PMD_SECT_BUFFERABLE | \
609 PMD_SECT_CACHEABLE | \
610 PMD_SECT_AP_WRITE | \
615 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
617 .long xscale_processor_functions
619 .long xscale_mc_user_fns
620 .long xscale_cache_fns
621 .size __8032x_proc_info, . - __8032x_proc_info
623 .type __8033x_proc_info,#object
627 .long PMD_TYPE_SECT | \
628 PMD_SECT_BUFFERABLE | \
629 PMD_SECT_CACHEABLE | \
630 PMD_SECT_AP_WRITE | \
635 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
637 .long xscale_processor_functions
639 .long xscale_mc_user_fns
640 .long xscale_cache_fns
641 .size __8033x_proc_info, . - __8033x_proc_info
643 .type __pxa250_proc_info,#object
647 .long PMD_TYPE_SECT | \
648 PMD_SECT_BUFFERABLE | \
649 PMD_SECT_CACHEABLE | \
650 PMD_SECT_AP_WRITE | \
655 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
656 .long cpu_pxa250_name
657 .long xscale_processor_functions
659 .long xscale_mc_user_fns
660 .long xscale_cache_fns
661 .size __pxa250_proc_info, . - __pxa250_proc_info
663 .type __pxa210_proc_info,#object
667 .long PMD_TYPE_SECT | \
668 PMD_SECT_BUFFERABLE | \
669 PMD_SECT_CACHEABLE | \
670 PMD_SECT_AP_WRITE | \
675 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
676 .long cpu_pxa210_name
677 .long xscale_processor_functions
679 .long xscale_mc_user_fns
680 .long xscale_cache_fns
681 .size __pxa210_proc_info, . - __pxa210_proc_info
683 .type __ixp2400_proc_info, #object
687 .long PMD_TYPE_SECT | \
688 PMD_SECT_BUFFERABLE | \
689 PMD_SECT_CACHEABLE | \
690 PMD_SECT_AP_WRITE | \
695 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
696 .long cpu_ixp2400_name
697 .long xscale_processor_functions
699 .long xscale_mc_user_fns
700 .long xscale_cache_fns
701 .size __ixp2400_proc_info, . - __ixp2400_proc_info
703 .type __ixp2800_proc_info, #object
707 .long PMD_TYPE_SECT | \
708 PMD_SECT_BUFFERABLE | \
709 PMD_SECT_CACHEABLE | \
710 PMD_SECT_AP_WRITE | \
715 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
716 .long cpu_ixp2800_name
717 .long xscale_processor_functions
719 .long xscale_mc_user_fns
720 .long xscale_cache_fns
721 .size __ixp2800_proc_info, . - __ixp2800_proc_info
723 .type __ixp42x_proc_info, #object
727 .long PMD_TYPE_SECT | \
728 PMD_SECT_BUFFERABLE | \
729 PMD_SECT_CACHEABLE | \
730 PMD_SECT_AP_WRITE | \
735 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
736 .long cpu_ixp42x_name
737 .long xscale_processor_functions
739 .long xscale_mc_user_fns
740 .long xscale_cache_fns
741 .size __ixp42x_proc_info, . - __ixp42x_proc_info
743 .type __ixp46x_proc_info, #object
751 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
752 .long cpu_ixp46x_name
753 .long xscale_processor_functions
755 .long xscale_mc_user_fns
756 .long xscale_cache_fns
757 .size __ixp46x_proc_info, . - __ixp46x_proc_info
759 .type __pxa255_proc_info,#object
763 .long PMD_TYPE_SECT | \
764 PMD_SECT_BUFFERABLE | \
765 PMD_SECT_CACHEABLE | \
766 PMD_SECT_AP_WRITE | \
771 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
772 .long cpu_pxa255_name
773 .long xscale_processor_functions
775 .long xscale_mc_user_fns
776 .long xscale_cache_fns
777 .size __pxa255_proc_info, . - __pxa255_proc_info
779 .type __pxa270_proc_info,#object
783 .long PMD_TYPE_SECT | \
784 PMD_SECT_BUFFERABLE | \
785 PMD_SECT_CACHEABLE | \
786 PMD_SECT_AP_WRITE | \
791 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
792 .long cpu_pxa270_name
793 .long xscale_processor_functions
795 .long xscale_mc_user_fns
796 .long xscale_cache_fns
797 .size __pxa270_proc_info, . - __pxa270_proc_info