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/pgtable.h>
29 #include <asm/ptrace.h>
30 #include "proc-macros.S"
33 * This is the maximum size of an area which will be flushed. If the area
34 * is larger than this, then we flush the whole cache
36 #define MAX_AREA_SIZE 32768
39 * the cache line size of the I and D cache
41 #define CACHELINESIZE 32
44 * the size of the data cache
46 #define CACHESIZE 32768
49 * Virtual address used to allocate the cache when flushed
51 * This must be an address range which is _never_ used. It should
52 * apparently have a mapping in the corresponding page table for
53 * compatibility with future CPUs that _could_ require it. For instance we
56 * This must be aligned on a 2*CACHESIZE boundary. The code selects one of
57 * the 2 areas in alternance each time the clean_d_cache macro is used.
58 * Without this the XScale core exhibits cache eviction problems and no one
61 * Reminder: the vector table is located at 0xffff0000-0xffff0fff.
63 #define CLEAN_ADDR 0xfffe0000
66 * This macro is used to wait for a CP15 write and is needed
67 * when we have to ensure that the last operation to the co-pro
68 * was completed before continuing with operation.
71 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15
72 mov \rd, \rd @ wait for completion
73 sub pc, pc, #4 @ flush instruction pipeline
76 .macro cpwait_ret, lr, rd
77 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15
78 sub pc, \lr, \rd, LSR #32 @ wait for completion and
79 @ flush instruction pipeline
83 * This macro cleans the entire dcache using line allocate.
84 * The main loop has been unrolled to reduce loop overhead.
85 * rd and rs are two scratch registers.
87 .macro clean_d_cache, rd, rs
90 eor \rd, \rd, #CACHESIZE
92 add \rs, \rd, #CACHESIZE
93 1: mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
94 add \rd, \rd, #CACHELINESIZE
95 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
96 add \rd, \rd, #CACHELINESIZE
97 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
98 add \rd, \rd, #CACHELINESIZE
99 mcr p15, 0, \rd, c7, c2, 5 @ allocate D cache line
100 add \rd, \rd, #CACHELINESIZE
106 clean_addr: .word CLEAN_ADDR
111 * cpu_xscale_proc_init()
113 * Nothing too exciting at the moment
115 ENTRY(cpu_xscale_proc_init)
119 * cpu_xscale_proc_fin()
121 ENTRY(cpu_xscale_proc_fin)
123 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
125 bl xscale_flush_kern_cache_all @ clean caches
126 mrc p15, 0, r0, c1, c0, 0 @ ctrl register
127 bic r0, r0, #0x1800 @ ...IZ...........
128 bic r0, r0, #0x0006 @ .............CA.
129 mcr p15, 0, r0, c1, c0, 0 @ disable caches
133 * cpu_xscale_reset(loc)
135 * Perform a soft reset of the system. Put the CPU into the
136 * same state as it would be if it had been reset, and branch
137 * to what would be the reset vector.
139 * loc: location to jump to for soft reset
142 ENTRY(cpu_xscale_reset)
143 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
144 msr cpsr_c, r1 @ reset CPSR
145 mrc p15, 0, r1, c1, c0, 0 @ ctrl register
146 bic r1, r1, #0x0086 @ ........B....CA.
147 bic r1, r1, #0x3900 @ ..VIZ..S........
148 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
149 mcr p15, 0, ip, c7, c7, 0 @ invalidate I,D caches & BTB
150 bic r1, r1, #0x0001 @ ...............M
151 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
152 @ CAUTION: MMU turned off from this point. We count on the pipeline
153 @ already containing those two last instructions to survive.
154 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
158 * cpu_xscale_do_idle()
160 * Cause the processor to idle
162 * For now we do nothing but go to idle mode for every case
164 * XScale supports clock switching, but using idle mode support
165 * allows external hardware to react to system state changes.
169 ENTRY(cpu_xscale_do_idle)
171 mcr p14, 0, r0, c7, c0, 0 @ Go to IDLE
174 /* ================================= CACHE ================================ */
177 * flush_user_cache_all()
179 * Invalidate all cache entries in a particular address
182 ENTRY(xscale_flush_user_cache_all)
186 * flush_kern_cache_all()
188 * Clean and invalidate the entire cache.
190 ENTRY(xscale_flush_kern_cache_all)
196 mcrne p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
197 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
201 * flush_user_cache_range(start, end, vm_flags)
203 * Invalidate a range of cache entries in the specified
206 * - start - start address (may not be aligned)
207 * - end - end address (exclusive, may not be aligned)
208 * - vma - vma_area_struct describing address space
211 ENTRY(xscale_flush_user_cache_range)
213 sub r3, r1, r0 @ calculate total size
214 cmp r3, #MAX_AREA_SIZE
215 bhs __flush_whole_cache
218 mcrne p15, 0, r0, c7, c5, 1 @ Invalidate I cache line
219 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
220 mcr p15, 0, r0, c7, c6, 1 @ Invalidate D cache line
221 add r0, r0, #CACHELINESIZE
225 mcrne p15, 0, ip, c7, c5, 6 @ Invalidate BTB
226 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
230 * coherent_kern_range(start, end)
232 * Ensure coherency between the Icache and the Dcache in the
233 * region described by start. If you have non-snooping
234 * Harvard caches, you need to implement this function.
236 * - start - virtual start address
237 * - end - virtual end address
239 * Note: single I-cache line invalidation isn't used here since
240 * it also trashes the mini I-cache used by JTAG debuggers.
242 ENTRY(xscale_coherent_kern_range)
243 bic r0, r0, #CACHELINESIZE - 1
244 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
245 add r0, r0, #CACHELINESIZE
249 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
250 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
254 * coherent_user_range(start, end)
256 * Ensure coherency between the Icache and the Dcache in the
257 * region described by start. If you have non-snooping
258 * Harvard caches, you need to implement this function.
260 * - start - virtual start address
261 * - end - virtual end address
263 ENTRY(xscale_coherent_user_range)
264 bic r0, r0, #CACHELINESIZE - 1
265 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
266 mcr p15, 0, r0, c7, c5, 1 @ Invalidate I cache entry
267 add r0, r0, #CACHELINESIZE
271 mcr p15, 0, r0, c7, c5, 6 @ Invalidate BTB
272 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
276 * flush_kern_dcache_page(void *page)
278 * Ensure no D cache aliasing occurs, either with itself or
281 * - addr - page aligned address
283 ENTRY(xscale_flush_kern_dcache_page)
285 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
286 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
287 add r0, r0, #CACHELINESIZE
291 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
292 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
296 * dma_inv_range(start, end)
298 * Invalidate (discard) the specified virtual address range.
299 * May not write back any entries. If 'start' or 'end'
300 * are not cache line aligned, those lines must be written
303 * - start - virtual start address
304 * - end - virtual end address
306 ENTRY(xscale_dma_inv_range)
307 mrc p15, 0, r2, c0, c0, 0 @ read ID
308 eor r2, r2, #0x69000000
309 eor r2, r2, #0x00052000
311 beq xscale_dma_flush_range
313 tst r0, #CACHELINESIZE - 1
314 bic r0, r0, #CACHELINESIZE - 1
315 mcrne p15, 0, r0, c7, c10, 1 @ clean D entry
316 tst r1, #CACHELINESIZE - 1
317 mcrne p15, 0, r1, c7, c10, 1 @ clean D entry
318 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
319 add r0, r0, #CACHELINESIZE
322 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
326 * dma_clean_range(start, end)
328 * Clean the specified virtual address range.
330 * - start - virtual start address
331 * - end - virtual end address
333 ENTRY(xscale_dma_clean_range)
334 bic r0, r0, #CACHELINESIZE - 1
335 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
336 add r0, r0, #CACHELINESIZE
339 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
343 * dma_flush_range(start, end)
345 * Clean and invalidate the specified virtual address range.
347 * - start - virtual start address
348 * - end - virtual end address
350 ENTRY(xscale_dma_flush_range)
351 bic r0, r0, #CACHELINESIZE - 1
352 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
353 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
354 add r0, r0, #CACHELINESIZE
357 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
360 ENTRY(xscale_cache_fns)
361 .long xscale_flush_kern_cache_all
362 .long xscale_flush_user_cache_all
363 .long xscale_flush_user_cache_range
364 .long xscale_coherent_kern_range
365 .long xscale_coherent_user_range
366 .long xscale_flush_kern_dcache_page
367 .long xscale_dma_inv_range
368 .long xscale_dma_clean_range
369 .long xscale_dma_flush_range
371 ENTRY(cpu_xscale_dcache_clean_area)
372 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
373 add r0, r0, #CACHELINESIZE
374 subs r1, r1, #CACHELINESIZE
378 /* =============================== PageTable ============================== */
380 #define PTE_CACHE_WRITE_ALLOCATE 0
383 * cpu_xscale_switch_mm(pgd)
385 * Set the translation base pointer to be as described by pgd.
387 * pgd: new page tables
390 ENTRY(cpu_xscale_switch_mm)
392 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
393 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
394 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
395 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
399 * cpu_xscale_set_pte(ptep, pte)
401 * Set a PTE and flush it out
403 * Errata 40: must set memory to write-through for user read-only pages.
406 ENTRY(cpu_xscale_set_pte)
407 str r1, [r0], #-2048 @ linux version
410 orr r2, r2, #PTE_TYPE_EXT @ extended page
412 eor r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
414 tst r3, #L_PTE_USER @ User?
415 orrne r2, r2, #PTE_EXT_AP_URO_SRW @ yes -> user r/o, system r/w
417 tst r3, #L_PTE_WRITE | L_PTE_DIRTY @ Write and Dirty?
418 orreq r2, r2, #PTE_EXT_AP_UNO_SRW @ yes -> user n/a, system r/w
419 @ combined with user -> user r/w
422 @ Handle the X bit. We want to set this bit for the minicache
423 @ (U = E = B = W = 0, C = 1) or when write allocate is enabled,
424 @ and we have a writeable, cacheable region. If we ignore the
425 @ U and E bits, we can allow user space to use the minicache as
428 @ X = (C & ~W & ~B) | (C & W & B & write_allocate)
430 eor ip, r1, #L_PTE_CACHEABLE
431 tst ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
432 #if PTE_CACHE_WRITE_ALLOCATE
433 eorne ip, r1, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
434 tstne ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
436 orreq r2, r2, #PTE_EXT_TEX(1)
439 @ Erratum 40: The B bit must be cleared for a user read-only
442 @ B = B & ~(U & C & ~W)
444 and ip, r1, #L_PTE_USER | L_PTE_WRITE | L_PTE_CACHEABLE
445 teq ip, #L_PTE_USER | L_PTE_CACHEABLE
446 biceq r2, r2, #PTE_BUFFERABLE
448 tst r3, #L_PTE_PRESENT | L_PTE_YOUNG @ Present and Young?
449 movne r2, #0 @ no -> fault
451 str r2, [r0] @ hardware version
453 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
454 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
464 .type __xscale_setup, #function
466 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB
467 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
468 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs
470 mov r0, #0 @ initially disallow access to CP0/CP1
472 mov r0, #1 @ Allow access to CP0
474 orr r0, r0, #1 << 6 @ cp6 for IOP3xx and Bulverde
475 orr r0, r0, #1 << 13 @ Its undefined whether this
476 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes
477 mrc p15, 0, r0, c1, c0, 0 @ get control register
478 ldr r5, xscale_cr1_clear
480 ldr r5, xscale_cr1_set
483 .size __xscale_setup, . - __xscale_setup
487 * .RVI ZFRS BLDP WCAM
488 * ..11 1.01 .... .101
491 .type xscale_cr1_clear, #object
492 .type xscale_cr1_set, #object
501 * Purpose : Function pointers used to access above functions - all calls
505 .type xscale_processor_functions, #object
506 ENTRY(xscale_processor_functions)
507 .word v5t_early_abort
508 .word cpu_xscale_proc_init
509 .word cpu_xscale_proc_fin
510 .word cpu_xscale_reset
511 .word cpu_xscale_do_idle
512 .word cpu_xscale_dcache_clean_area
513 .word cpu_xscale_switch_mm
514 .word cpu_xscale_set_pte
515 .size xscale_processor_functions, . - xscale_processor_functions
519 .type cpu_arch_name, #object
522 .size cpu_arch_name, . - cpu_arch_name
524 .type cpu_elf_name, #object
527 .size cpu_elf_name, . - cpu_elf_name
529 .type cpu_80200_name, #object
531 .asciz "XScale-80200"
532 .size cpu_80200_name, . - cpu_80200_name
534 .type cpu_8032x_name, #object
536 .asciz "XScale-IOP8032x Family"
537 .size cpu_8032x_name, . - cpu_8032x_name
539 .type cpu_8033x_name, #object
541 .asciz "XScale-IOP8033x Family"
542 .size cpu_8033x_name, . - cpu_8033x_name
544 .type cpu_pxa250_name, #object
546 .asciz "XScale-PXA250"
547 .size cpu_pxa250_name, . - cpu_pxa250_name
549 .type cpu_pxa210_name, #object
551 .asciz "XScale-PXA210"
552 .size cpu_pxa210_name, . - cpu_pxa210_name
554 .type cpu_ixp42x_name, #object
556 .asciz "XScale-IXP42x Family"
557 .size cpu_ixp42x_name, . - cpu_ixp42x_name
559 .type cpu_ixp46x_name, #object
561 .asciz "XScale-IXP46x Family"
562 .size cpu_ixp46x_name, . - cpu_ixp46x_name
564 .type cpu_ixp2400_name, #object
566 .asciz "XScale-IXP2400"
567 .size cpu_ixp2400_name, . - cpu_ixp2400_name
569 .type cpu_ixp2800_name, #object
571 .asciz "XScale-IXP2800"
572 .size cpu_ixp2800_name, . - cpu_ixp2800_name
574 .type cpu_pxa255_name, #object
576 .asciz "XScale-PXA255"
577 .size cpu_pxa255_name, . - cpu_pxa255_name
579 .type cpu_pxa270_name, #object
581 .asciz "XScale-PXA270"
582 .size cpu_pxa270_name, . - cpu_pxa270_name
586 .section ".proc.info.init", #alloc, #execinstr
588 .type __80200_proc_info,#object
592 .long PMD_TYPE_SECT | \
593 PMD_SECT_BUFFERABLE | \
594 PMD_SECT_CACHEABLE | \
595 PMD_SECT_AP_WRITE | \
600 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
602 .long xscale_processor_functions
604 .long xscale_mc_user_fns
605 .long xscale_cache_fns
606 .size __80200_proc_info, . - __80200_proc_info
608 .type __8032x_proc_info,#object
611 .long 0xfffff5e0 @ mask should accomodate IOP80219 also
612 .long PMD_TYPE_SECT | \
613 PMD_SECT_BUFFERABLE | \
614 PMD_SECT_CACHEABLE | \
615 PMD_SECT_AP_WRITE | \
620 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
622 .long xscale_processor_functions
624 .long xscale_mc_user_fns
625 .long xscale_cache_fns
626 .size __8032x_proc_info, . - __8032x_proc_info
628 .type __8033x_proc_info,#object
632 .long PMD_TYPE_SECT | \
633 PMD_SECT_BUFFERABLE | \
634 PMD_SECT_CACHEABLE | \
635 PMD_SECT_AP_WRITE | \
640 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
642 .long xscale_processor_functions
644 .long xscale_mc_user_fns
645 .long xscale_cache_fns
646 .size __8033x_proc_info, . - __8033x_proc_info
648 .type __pxa250_proc_info,#object
652 .long PMD_TYPE_SECT | \
653 PMD_SECT_BUFFERABLE | \
654 PMD_SECT_CACHEABLE | \
655 PMD_SECT_AP_WRITE | \
660 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
661 .long cpu_pxa250_name
662 .long xscale_processor_functions
664 .long xscale_mc_user_fns
665 .long xscale_cache_fns
666 .size __pxa250_proc_info, . - __pxa250_proc_info
668 .type __pxa210_proc_info,#object
672 .long PMD_TYPE_SECT | \
673 PMD_SECT_BUFFERABLE | \
674 PMD_SECT_CACHEABLE | \
675 PMD_SECT_AP_WRITE | \
680 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
681 .long cpu_pxa210_name
682 .long xscale_processor_functions
684 .long xscale_mc_user_fns
685 .long xscale_cache_fns
686 .size __pxa210_proc_info, . - __pxa210_proc_info
688 .type __ixp2400_proc_info, #object
692 .long PMD_TYPE_SECT | \
693 PMD_SECT_BUFFERABLE | \
694 PMD_SECT_CACHEABLE | \
695 PMD_SECT_AP_WRITE | \
700 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
701 .long cpu_ixp2400_name
702 .long xscale_processor_functions
704 .long xscale_mc_user_fns
705 .long xscale_cache_fns
706 .size __ixp2400_proc_info, . - __ixp2400_proc_info
708 .type __ixp2800_proc_info, #object
712 .long PMD_TYPE_SECT | \
713 PMD_SECT_BUFFERABLE | \
714 PMD_SECT_CACHEABLE | \
715 PMD_SECT_AP_WRITE | \
720 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
721 .long cpu_ixp2800_name
722 .long xscale_processor_functions
724 .long xscale_mc_user_fns
725 .long xscale_cache_fns
726 .size __ixp2800_proc_info, . - __ixp2800_proc_info
728 .type __ixp42x_proc_info, #object
732 .long PMD_TYPE_SECT | \
733 PMD_SECT_BUFFERABLE | \
734 PMD_SECT_CACHEABLE | \
735 PMD_SECT_AP_WRITE | \
740 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
741 .long cpu_ixp42x_name
742 .long xscale_processor_functions
744 .long xscale_mc_user_fns
745 .long xscale_cache_fns
746 .size __ixp42x_proc_info, . - __ixp42x_proc_info
748 .type __ixp46x_proc_info, #object
756 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
757 .long cpu_ixp46x_name
758 .long xscale_processor_functions
760 .long xscale_mc_user_fns
761 .long xscale_cache_fns
762 .size __ixp46x_proc_info, . - __ixp46x_proc_info
764 .type __pxa255_proc_info,#object
768 .long PMD_TYPE_SECT | \
769 PMD_SECT_BUFFERABLE | \
770 PMD_SECT_CACHEABLE | \
771 PMD_SECT_AP_WRITE | \
776 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
777 .long cpu_pxa255_name
778 .long xscale_processor_functions
780 .long xscale_mc_user_fns
781 .long xscale_cache_fns
782 .size __pxa255_proc_info, . - __pxa255_proc_info
784 .type __pxa270_proc_info,#object
788 .long PMD_TYPE_SECT | \
789 PMD_SECT_BUFFERABLE | \
790 PMD_SECT_CACHEABLE | \
791 PMD_SECT_AP_WRITE | \
796 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
797 .long cpu_pxa270_name
798 .long xscale_processor_functions
800 .long xscale_mc_user_fns
801 .long xscale_cache_fns
802 .size __pxa270_proc_info, . - __pxa270_proc_info