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@fluxnic.net>
23 #include <linux/linkage.h>
24 #include <linux/init.h>
25 #include <asm/assembler.h>
26 #include <asm/hwcap.h>
27 #include <asm/pgtable.h>
28 #include <asm/pgtable-hwdef.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)
117 @ enable write buffer coalescing. Some bootloader disable it
118 mrc p15, 0, r1, c1, c0, 1
120 mcr p15, 0, r1, c1, c0, 1
124 * cpu_xscale_proc_fin()
126 ENTRY(cpu_xscale_proc_fin)
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
142 * Beware PXA270 erratum E7.
145 .pushsection .idmap.text, "ax"
146 ENTRY(cpu_xscale_reset)
147 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
148 msr cpsr_c, r1 @ reset CPSR
149 mcr p15, 0, r1, c10, c4, 1 @ unlock I-TLB
150 mcr p15, 0, r1, c8, c5, 0 @ invalidate I-TLB
151 mrc p15, 0, r1, c1, c0, 0 @ ctrl register
152 bic r1, r1, #0x0086 @ ........B....CA.
153 bic r1, r1, #0x3900 @ ..VIZ..S........
154 sub pc, pc, #4 @ flush pipeline
155 @ *** cache line aligned ***
156 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
157 bic r1, r1, #0x0001 @ ...............M
158 mcr p15, 0, ip, c7, c7, 0 @ invalidate I,D caches & BTB
159 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
160 @ CAUTION: MMU turned off from this point. We count on the pipeline
161 @ already containing those two last instructions to survive.
162 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
164 ENDPROC(cpu_xscale_reset)
168 * cpu_xscale_do_idle()
170 * Cause the processor to idle
172 * For now we do nothing but go to idle mode for every case
174 * XScale supports clock switching, but using idle mode support
175 * allows external hardware to react to system state changes.
179 ENTRY(cpu_xscale_do_idle)
181 mcr p14, 0, r0, c7, c0, 0 @ Go to IDLE
184 /* ================================= CACHE ================================ */
189 * Unconditionally clean and invalidate the entire icache.
191 ENTRY(xscale_flush_icache_all)
193 mcr p15, 0, r0, c7, c5, 0 @ invalidate I cache
195 ENDPROC(xscale_flush_icache_all)
198 * flush_user_cache_all()
200 * Invalidate all cache entries in a particular address
203 ENTRY(xscale_flush_user_cache_all)
207 * flush_kern_cache_all()
209 * Clean and invalidate the entire cache.
211 ENTRY(xscale_flush_kern_cache_all)
217 mcrne p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
218 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
222 * flush_user_cache_range(start, end, vm_flags)
224 * Invalidate a range of cache entries in the specified
227 * - start - start address (may not be aligned)
228 * - end - end address (exclusive, may not be aligned)
229 * - vma - vma_area_struct describing address space
232 ENTRY(xscale_flush_user_cache_range)
234 sub r3, r1, r0 @ calculate total size
235 cmp r3, #MAX_AREA_SIZE
236 bhs __flush_whole_cache
239 mcrne p15, 0, r0, c7, c5, 1 @ Invalidate I cache line
240 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
241 mcr p15, 0, r0, c7, c6, 1 @ Invalidate D cache line
242 add r0, r0, #CACHELINESIZE
246 mcrne p15, 0, ip, c7, c5, 6 @ Invalidate BTB
247 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
251 * coherent_kern_range(start, end)
253 * Ensure coherency between the Icache and the Dcache in the
254 * region described by start. If you have non-snooping
255 * Harvard caches, you need to implement this function.
257 * - start - virtual start address
258 * - end - virtual end address
260 * Note: single I-cache line invalidation isn't used here since
261 * it also trashes the mini I-cache used by JTAG debuggers.
263 ENTRY(xscale_coherent_kern_range)
264 bic r0, r0, #CACHELINESIZE - 1
265 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
266 add r0, r0, #CACHELINESIZE
270 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
271 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
275 * coherent_user_range(start, end)
277 * Ensure coherency between the Icache and the Dcache in the
278 * region described by start. If you have non-snooping
279 * Harvard caches, you need to implement this function.
281 * - start - virtual start address
282 * - end - virtual end address
284 ENTRY(xscale_coherent_user_range)
285 bic r0, r0, #CACHELINESIZE - 1
286 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
287 mcr p15, 0, r0, c7, c5, 1 @ Invalidate I cache entry
288 add r0, r0, #CACHELINESIZE
292 mcr p15, 0, r0, c7, c5, 6 @ Invalidate BTB
293 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
297 * flush_kern_dcache_area(void *addr, size_t size)
299 * Ensure no D cache aliasing occurs, either with itself or
302 * - addr - kernel address
303 * - size - region size
305 ENTRY(xscale_flush_kern_dcache_area)
307 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
308 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
309 add r0, r0, #CACHELINESIZE
313 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
314 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
318 * dma_inv_range(start, end)
320 * Invalidate (discard) the specified virtual address range.
321 * May not write back any entries. If 'start' or 'end'
322 * are not cache line aligned, those lines must be written
325 * - start - virtual start address
326 * - end - virtual end address
328 xscale_dma_inv_range:
329 tst r0, #CACHELINESIZE - 1
330 bic r0, r0, #CACHELINESIZE - 1
331 mcrne p15, 0, r0, c7, c10, 1 @ clean D entry
332 tst r1, #CACHELINESIZE - 1
333 mcrne p15, 0, r1, c7, c10, 1 @ clean D entry
334 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
335 add r0, r0, #CACHELINESIZE
338 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
342 * dma_clean_range(start, end)
344 * Clean the specified virtual address range.
346 * - start - virtual start address
347 * - end - virtual end address
349 xscale_dma_clean_range:
350 bic r0, r0, #CACHELINESIZE - 1
351 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
352 add r0, r0, #CACHELINESIZE
355 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
359 * dma_flush_range(start, end)
361 * Clean and invalidate the specified virtual address range.
363 * - start - virtual start address
364 * - end - virtual end address
366 ENTRY(xscale_dma_flush_range)
367 bic r0, r0, #CACHELINESIZE - 1
368 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
369 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
370 add r0, r0, #CACHELINESIZE
373 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
377 * dma_map_area(start, size, dir)
378 * - start - kernel virtual start address
379 * - size - size of region
380 * - dir - DMA direction
382 ENTRY(xscale_dma_map_area)
384 cmp r2, #DMA_TO_DEVICE
385 beq xscale_dma_clean_range
386 bcs xscale_dma_inv_range
387 b xscale_dma_flush_range
388 ENDPROC(xscale_dma_map_area)
391 * dma_map_area(start, size, dir)
392 * - start - kernel virtual start address
393 * - size - size of region
394 * - dir - DMA direction
396 ENTRY(xscale_80200_A0_A1_dma_map_area)
398 teq r2, #DMA_TO_DEVICE
399 beq xscale_dma_clean_range
400 b xscale_dma_flush_range
401 ENDPROC(xscale_80200_A0_A1_dma_map_area)
404 * dma_unmap_area(start, size, dir)
405 * - start - kernel virtual start address
406 * - size - size of region
407 * - dir - DMA direction
409 ENTRY(xscale_dma_unmap_area)
411 ENDPROC(xscale_dma_unmap_area)
413 .globl xscale_flush_kern_cache_louis
414 .equ xscale_flush_kern_cache_louis, xscale_flush_kern_cache_all
416 @ define struct cpu_cache_fns (see <asm/cacheflush.h> and proc-macros.S)
417 define_cache_functions xscale
420 * On stepping A0/A1 of the 80200, invalidating D-cache by line doesn't
421 * clear the dirty bits, which means that if we invalidate a dirty line,
422 * the dirty data can still be written back to external memory later on.
424 * The recommended workaround is to always do a clean D-cache line before
425 * doing an invalidate D-cache line, so on the affected processors,
426 * dma_inv_range() is implemented as dma_flush_range().
428 * See erratum #25 of "Intel 80200 Processor Specification Update",
429 * revision January 22, 2003, available at:
430 * http://www.intel.com/design/iio/specupdt/273415.htm
432 .macro a0_alias basename
433 .globl xscale_80200_A0_A1_\basename
434 .type xscale_80200_A0_A1_\basename , %function
435 .equ xscale_80200_A0_A1_\basename , xscale_\basename
439 * Most of the cache functions are unchanged for these processor revisions.
440 * Export suitable alias symbols for the unchanged functions:
442 a0_alias flush_icache_all
443 a0_alias flush_user_cache_all
444 a0_alias flush_kern_cache_all
445 a0_alias flush_kern_cache_louis
446 a0_alias flush_user_cache_range
447 a0_alias coherent_kern_range
448 a0_alias coherent_user_range
449 a0_alias flush_kern_dcache_area
450 a0_alias dma_flush_range
451 a0_alias dma_unmap_area
453 @ define struct cpu_cache_fns (see <asm/cacheflush.h> and proc-macros.S)
454 define_cache_functions xscale_80200_A0_A1
456 ENTRY(cpu_xscale_dcache_clean_area)
457 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
458 add r0, r0, #CACHELINESIZE
459 subs r1, r1, #CACHELINESIZE
463 /* =============================== PageTable ============================== */
466 * cpu_xscale_switch_mm(pgd)
468 * Set the translation base pointer to be as described by pgd.
470 * pgd: new page tables
473 ENTRY(cpu_xscale_switch_mm)
475 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
476 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
477 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
478 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
482 * cpu_xscale_set_pte_ext(ptep, pte, ext)
484 * Set a PTE and flush it out
486 * Errata 40: must set memory to write-through for user read-only pages.
489 .long 0x00 @ L_PTE_MT_UNCACHED
490 .long PTE_BUFFERABLE @ L_PTE_MT_BUFFERABLE
491 .long PTE_CACHEABLE @ L_PTE_MT_WRITETHROUGH
492 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEBACK
493 .long PTE_EXT_TEX(1) | PTE_BUFFERABLE @ L_PTE_MT_DEV_SHARED
495 .long PTE_EXT_TEX(1) | PTE_CACHEABLE @ L_PTE_MT_MINICACHE
496 .long PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEALLOC
498 .long PTE_BUFFERABLE @ L_PTE_MT_DEV_WC
500 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_DEV_CACHED
501 .long 0x00 @ L_PTE_MT_DEV_NONSHARED
507 ENTRY(cpu_xscale_set_pte_ext)
508 xscale_set_pte_ext_prologue
511 @ Erratum 40: must set memory to write-through for user read-only pages
513 and ip, r1, #(L_PTE_MT_MASK | L_PTE_USER | L_PTE_RDONLY) & ~(4 << 2)
514 teq ip, #L_PTE_MT_WRITEBACK | L_PTE_USER | L_PTE_RDONLY
516 moveq r1, #L_PTE_MT_WRITETHROUGH
517 and r1, r1, #L_PTE_MT_MASK
518 adr ip, cpu_xscale_mt_table
523 xscale_set_pte_ext_epilogue
529 .globl cpu_xscale_suspend_size
530 .equ cpu_xscale_suspend_size, 4 * 6
531 #ifdef CONFIG_ARM_CPU_SUSPEND
532 ENTRY(cpu_xscale_do_suspend)
533 stmfd sp!, {r4 - r9, lr}
534 mrc p14, 0, r4, c6, c0, 0 @ clock configuration, for turbo mode
535 mrc p15, 0, r5, c15, c1, 0 @ CP access reg
536 mrc p15, 0, r6, c13, c0, 0 @ PID
537 mrc p15, 0, r7, c3, c0, 0 @ domain ID
538 mrc p15, 0, r8, c1, c1, 0 @ auxiliary control reg
539 mrc p15, 0, r9, c1, c0, 0 @ control reg
540 bic r4, r4, #2 @ clear frequency change bit
541 stmia r0, {r4 - r9} @ store cp regs
542 ldmfd sp!, {r4 - r9, pc}
543 ENDPROC(cpu_xscale_do_suspend)
545 ENTRY(cpu_xscale_do_resume)
546 ldmia r0, {r4 - r9} @ load cp regs
548 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
549 mcr p15, 0, ip, c7, c7, 0 @ invalidate I & D caches, BTB
550 mcr p14, 0, r4, c6, c0, 0 @ clock configuration, turbo mode.
551 mcr p15, 0, r5, c15, c1, 0 @ CP access reg
552 mcr p15, 0, r6, c13, c0, 0 @ PID
553 mcr p15, 0, r7, c3, c0, 0 @ domain ID
554 mcr p15, 0, r1, c2, c0, 0 @ translation table base addr
555 mcr p15, 0, r8, c1, c1, 0 @ auxiliary control reg
556 mov r0, r9 @ control register
558 ENDPROC(cpu_xscale_do_resume)
563 .type __xscale_setup, #function
565 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB
566 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
567 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs
568 mov r0, #1 << 6 @ cp6 for IOP3xx and Bulverde
569 orr r0, r0, #1 << 13 @ Its undefined whether this
570 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes
574 mrc p15, 0, r0, c1, c0, 0 @ get control register
578 .size __xscale_setup, . - __xscale_setup
582 * .RVI ZFRS BLDP WCAM
583 * ..11 1.01 .... .101
586 .type xscale_crval, #object
588 crval clear=0x00003b07, mmuset=0x00003905, ucset=0x00001900
592 @ define struct processor (see <asm/proc-fns.h> and proc-macros.S)
593 define_processor_functions xscale, dabort=v5t_early_abort, pabort=legacy_pabort, suspend=1
597 string cpu_arch_name, "armv5te"
598 string cpu_elf_name, "v5"
600 string cpu_80200_A0_A1_name, "XScale-80200 A0/A1"
601 string cpu_80200_name, "XScale-80200"
602 string cpu_80219_name, "XScale-80219"
603 string cpu_8032x_name, "XScale-IOP8032x Family"
604 string cpu_8033x_name, "XScale-IOP8033x Family"
605 string cpu_pxa250_name, "XScale-PXA250"
606 string cpu_pxa210_name, "XScale-PXA210"
607 string cpu_ixp42x_name, "XScale-IXP42x Family"
608 string cpu_ixp43x_name, "XScale-IXP43x Family"
609 string cpu_ixp46x_name, "XScale-IXP46x Family"
610 string cpu_ixp2400_name, "XScale-IXP2400"
611 string cpu_ixp2800_name, "XScale-IXP2800"
612 string cpu_pxa255_name, "XScale-PXA255"
613 string cpu_pxa270_name, "XScale-PXA270"
617 .section ".proc.info.init", #alloc, #execinstr
619 .macro xscale_proc_info name:req, cpu_val:req, cpu_mask:req, cpu_name:req, cache
620 .type __\name\()_proc_info,#object
621 __\name\()_proc_info:
624 .long PMD_TYPE_SECT | \
625 PMD_SECT_BUFFERABLE | \
626 PMD_SECT_CACHEABLE | \
627 PMD_SECT_AP_WRITE | \
629 .long PMD_TYPE_SECT | \
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
641 .long xscale_cache_fns
645 .size __\name\()_proc_info, . - __\name\()_proc_info
648 xscale_proc_info 80200_A0_A1, 0x69052000, 0xfffffffe, cpu_80200_name, \
649 cache=xscale_80200_A0_A1_cache_fns
650 xscale_proc_info 80200, 0x69052000, 0xfffffff0, cpu_80200_name
651 xscale_proc_info 80219, 0x69052e20, 0xffffffe0, cpu_80219_name
652 xscale_proc_info 8032x, 0x69052420, 0xfffff7e0, cpu_8032x_name
653 xscale_proc_info 8033x, 0x69054010, 0xfffffd30, cpu_8033x_name
654 xscale_proc_info pxa250, 0x69052100, 0xfffff7f0, cpu_pxa250_name
655 xscale_proc_info pxa210, 0x69052120, 0xfffff3f0, cpu_pxa210_name
656 xscale_proc_info ixp2400, 0x69054190, 0xfffffff0, cpu_ixp2400_name
657 xscale_proc_info ixp2800, 0x690541a0, 0xfffffff0, cpu_ixp2800_name
658 xscale_proc_info ixp42x, 0x690541c0, 0xffffffc0, cpu_ixp42x_name
659 xscale_proc_info ixp43x, 0x69054040, 0xfffffff0, cpu_ixp43x_name
660 xscale_proc_info ixp46x, 0x69054200, 0xffffff00, cpu_ixp46x_name
661 xscale_proc_info pxa255, 0x69052d00, 0xfffffff0, cpu_pxa255_name
662 xscale_proc_info pxa270, 0x69054110, 0xfffffff0, cpu_pxa270_name