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 ENTRY(cpu_xscale_reset)
146 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
147 msr cpsr_c, r1 @ reset CPSR
148 mcr p15, 0, r1, c10, c4, 1 @ unlock I-TLB
149 mcr p15, 0, r1, c8, c5, 0 @ invalidate I-TLB
150 mrc p15, 0, r1, c1, c0, 0 @ ctrl register
151 bic r1, r1, #0x0086 @ ........B....CA.
152 bic r1, r1, #0x3900 @ ..VIZ..S........
153 sub pc, pc, #4 @ flush pipeline
154 @ *** cache line aligned ***
155 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
156 bic r1, r1, #0x0001 @ ...............M
157 mcr p15, 0, ip, c7, c7, 0 @ invalidate I,D caches & BTB
158 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
159 @ CAUTION: MMU turned off from this point. We count on the pipeline
160 @ already containing those two last instructions to survive.
161 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
165 * cpu_xscale_do_idle()
167 * Cause the processor to idle
169 * For now we do nothing but go to idle mode for every case
171 * XScale supports clock switching, but using idle mode support
172 * allows external hardware to react to system state changes.
176 ENTRY(cpu_xscale_do_idle)
178 mcr p14, 0, r0, c7, c0, 0 @ Go to IDLE
181 /* ================================= CACHE ================================ */
186 * Unconditionally clean and invalidate the entire icache.
188 ENTRY(xscale_flush_icache_all)
190 mcr p15, 0, r0, c7, c5, 0 @ invalidate I cache
192 ENDPROC(xscale_flush_icache_all)
195 * flush_user_cache_all()
197 * Invalidate all cache entries in a particular address
200 ENTRY(xscale_flush_user_cache_all)
204 * flush_kern_cache_all()
206 * Clean and invalidate the entire cache.
208 ENTRY(xscale_flush_kern_cache_all)
214 mcrne p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
215 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
219 * flush_user_cache_range(start, end, vm_flags)
221 * Invalidate a range of cache entries in the specified
224 * - start - start address (may not be aligned)
225 * - end - end address (exclusive, may not be aligned)
226 * - vma - vma_area_struct describing address space
229 ENTRY(xscale_flush_user_cache_range)
231 sub r3, r1, r0 @ calculate total size
232 cmp r3, #MAX_AREA_SIZE
233 bhs __flush_whole_cache
236 mcrne p15, 0, r0, c7, c5, 1 @ Invalidate I cache line
237 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line
238 mcr p15, 0, r0, c7, c6, 1 @ Invalidate D cache line
239 add r0, r0, #CACHELINESIZE
243 mcrne p15, 0, ip, c7, c5, 6 @ Invalidate BTB
244 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
248 * coherent_kern_range(start, end)
250 * Ensure coherency between the Icache and the Dcache in the
251 * region described by start. If you have non-snooping
252 * Harvard caches, you need to implement this function.
254 * - start - virtual start address
255 * - end - virtual end address
257 * Note: single I-cache line invalidation isn't used here since
258 * it also trashes the mini I-cache used by JTAG debuggers.
260 ENTRY(xscale_coherent_kern_range)
261 bic r0, r0, #CACHELINESIZE - 1
262 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
263 add r0, r0, #CACHELINESIZE
267 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
268 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
272 * coherent_user_range(start, end)
274 * Ensure coherency between the Icache and the Dcache in the
275 * region described by start. If you have non-snooping
276 * Harvard caches, you need to implement this function.
278 * - start - virtual start address
279 * - end - virtual end address
281 ENTRY(xscale_coherent_user_range)
282 bic r0, r0, #CACHELINESIZE - 1
283 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
284 mcr p15, 0, r0, c7, c5, 1 @ Invalidate I cache entry
285 add r0, r0, #CACHELINESIZE
289 mcr p15, 0, r0, c7, c5, 6 @ Invalidate BTB
290 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
294 * flush_kern_dcache_area(void *addr, size_t size)
296 * Ensure no D cache aliasing occurs, either with itself or
299 * - addr - kernel address
300 * - size - region size
302 ENTRY(xscale_flush_kern_dcache_area)
304 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
305 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
306 add r0, r0, #CACHELINESIZE
310 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB
311 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
315 * dma_inv_range(start, end)
317 * Invalidate (discard) the specified virtual address range.
318 * May not write back any entries. If 'start' or 'end'
319 * are not cache line aligned, those lines must be written
322 * - start - virtual start address
323 * - end - virtual end address
325 xscale_dma_inv_range:
326 tst r0, #CACHELINESIZE - 1
327 bic r0, r0, #CACHELINESIZE - 1
328 mcrne p15, 0, r0, c7, c10, 1 @ clean D entry
329 tst r1, #CACHELINESIZE - 1
330 mcrne p15, 0, r1, c7, c10, 1 @ clean D entry
331 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
332 add r0, r0, #CACHELINESIZE
335 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
339 * dma_clean_range(start, end)
341 * Clean the specified virtual address range.
343 * - start - virtual start address
344 * - end - virtual end address
346 xscale_dma_clean_range:
347 bic r0, r0, #CACHELINESIZE - 1
348 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
349 add r0, r0, #CACHELINESIZE
352 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
356 * dma_flush_range(start, end)
358 * Clean and invalidate the specified virtual address range.
360 * - start - virtual start address
361 * - end - virtual end address
363 ENTRY(xscale_dma_flush_range)
364 bic r0, r0, #CACHELINESIZE - 1
365 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
366 mcr p15, 0, r0, c7, c6, 1 @ invalidate D entry
367 add r0, r0, #CACHELINESIZE
370 mcr p15, 0, r0, c7, c10, 4 @ Drain Write (& Fill) Buffer
374 * dma_map_area(start, size, dir)
375 * - start - kernel virtual start address
376 * - size - size of region
377 * - dir - DMA direction
379 ENTRY(xscale_dma_map_area)
381 cmp r2, #DMA_TO_DEVICE
382 beq xscale_dma_clean_range
383 bcs xscale_dma_inv_range
384 b xscale_dma_flush_range
385 ENDPROC(xscale_dma_map_area)
388 * dma_map_area(start, size, dir)
389 * - start - kernel virtual start address
390 * - size - size of region
391 * - dir - DMA direction
393 ENTRY(xscale_80200_A0_A1_dma_map_area)
395 teq r2, #DMA_TO_DEVICE
396 beq xscale_dma_clean_range
397 b xscale_dma_flush_range
398 ENDPROC(xscale_80200_A0_A1_dma_map_area)
401 * dma_unmap_area(start, size, dir)
402 * - start - kernel virtual start address
403 * - size - size of region
404 * - dir - DMA direction
406 ENTRY(xscale_dma_unmap_area)
408 ENDPROC(xscale_dma_unmap_area)
410 @ define struct cpu_cache_fns (see <asm/cacheflush.h> and proc-macros.S)
411 define_cache_functions xscale
414 * On stepping A0/A1 of the 80200, invalidating D-cache by line doesn't
415 * clear the dirty bits, which means that if we invalidate a dirty line,
416 * the dirty data can still be written back to external memory later on.
418 * The recommended workaround is to always do a clean D-cache line before
419 * doing an invalidate D-cache line, so on the affected processors,
420 * dma_inv_range() is implemented as dma_flush_range().
422 * See erratum #25 of "Intel 80200 Processor Specification Update",
423 * revision January 22, 2003, available at:
424 * http://www.intel.com/design/iio/specupdt/273415.htm
426 .macro a0_alias basename
427 .globl xscale_80200_A0_A1_\basename
428 .type xscale_80200_A0_A1_\basename , %function
429 .equ xscale_80200_A0_A1_\basename , xscale_\basename
433 * Most of the cache functions are unchanged for these processor revisions.
434 * Export suitable alias symbols for the unchanged functions:
436 a0_alias flush_icache_all
437 a0_alias flush_user_cache_all
438 a0_alias flush_kern_cache_all
439 a0_alias flush_user_cache_range
440 a0_alias coherent_kern_range
441 a0_alias coherent_user_range
442 a0_alias flush_kern_dcache_area
443 a0_alias dma_flush_range
444 a0_alias dma_unmap_area
446 @ define struct cpu_cache_fns (see <asm/cacheflush.h> and proc-macros.S)
447 define_cache_functions xscale_80200_A0_A1
449 ENTRY(cpu_xscale_dcache_clean_area)
450 1: mcr p15, 0, r0, c7, c10, 1 @ clean D entry
451 add r0, r0, #CACHELINESIZE
452 subs r1, r1, #CACHELINESIZE
456 /* =============================== PageTable ============================== */
459 * cpu_xscale_switch_mm(pgd)
461 * Set the translation base pointer to be as described by pgd.
463 * pgd: new page tables
466 ENTRY(cpu_xscale_switch_mm)
468 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB
469 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
470 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
471 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
475 * cpu_xscale_set_pte_ext(ptep, pte, ext)
477 * Set a PTE and flush it out
479 * Errata 40: must set memory to write-through for user read-only pages.
482 .long 0x00 @ L_PTE_MT_UNCACHED
483 .long PTE_BUFFERABLE @ L_PTE_MT_BUFFERABLE
484 .long PTE_CACHEABLE @ L_PTE_MT_WRITETHROUGH
485 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEBACK
486 .long PTE_EXT_TEX(1) | PTE_BUFFERABLE @ L_PTE_MT_DEV_SHARED
488 .long PTE_EXT_TEX(1) | PTE_CACHEABLE @ L_PTE_MT_MINICACHE
489 .long PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEALLOC
491 .long PTE_BUFFERABLE @ L_PTE_MT_DEV_WC
493 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_DEV_CACHED
494 .long 0x00 @ L_PTE_MT_DEV_NONSHARED
500 ENTRY(cpu_xscale_set_pte_ext)
501 xscale_set_pte_ext_prologue
504 @ Erratum 40: must set memory to write-through for user read-only pages
506 and ip, r1, #(L_PTE_MT_MASK | L_PTE_USER | L_PTE_RDONLY) & ~(4 << 2)
507 teq ip, #L_PTE_MT_WRITEBACK | L_PTE_USER | L_PTE_RDONLY
509 moveq r1, #L_PTE_MT_WRITETHROUGH
510 and r1, r1, #L_PTE_MT_MASK
511 adr ip, cpu_xscale_mt_table
516 xscale_set_pte_ext_epilogue
522 .globl cpu_xscale_suspend_size
523 .equ cpu_xscale_suspend_size, 4 * 7
524 #ifdef CONFIG_PM_SLEEP
525 ENTRY(cpu_xscale_do_suspend)
526 stmfd sp!, {r4 - r10, lr}
527 mrc p14, 0, r4, c6, c0, 0 @ clock configuration, for turbo mode
528 mrc p15, 0, r5, c15, c1, 0 @ CP access reg
529 mrc p15, 0, r6, c13, c0, 0 @ PID
530 mrc p15, 0, r7, c3, c0, 0 @ domain ID
531 mrc p15, 0, r8, c2, c0, 0 @ translation table base addr
532 mrc p15, 0, r9, c1, c1, 0 @ auxiliary control reg
533 mrc p15, 0, r10, c1, c0, 0 @ control reg
534 bic r4, r4, #2 @ clear frequency change bit
535 stmia r0, {r4 - r10} @ store cp regs
536 ldmfd sp!, {r4 - r10, pc}
537 ENDPROC(cpu_xscale_do_suspend)
539 ENTRY(cpu_xscale_do_resume)
540 ldmia r0, {r4 - r10} @ load cp regs
542 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs
543 mcr p15, 0, ip, c7, c7, 0 @ invalidate I & D caches, BTB
544 mcr p14, 0, r4, c6, c0, 0 @ clock configuration, turbo mode.
545 mcr p15, 0, r5, c15, c1, 0 @ CP access reg
546 mcr p15, 0, r6, c13, c0, 0 @ PID
547 mcr p15, 0, r7, c3, c0, 0 @ domain ID
548 mcr p15, 0, r8, c2, c0, 0 @ translation table base addr
549 mcr p15, 0, r9, c1, c1, 0 @ auxiliary control reg
550 mov r0, r10 @ control register
551 mov r2, r8, lsr #14 @ get TTB0 base
553 ldr r3, =PMD_TYPE_SECT | PMD_SECT_BUFFERABLE | \
554 PMD_SECT_CACHEABLE | PMD_SECT_AP_WRITE
556 ENDPROC(cpu_xscale_do_resume)
561 .type __xscale_setup, #function
563 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB
564 mcr p15, 0, ip, c7, c10, 4 @ Drain Write (& Fill) Buffer
565 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs
566 mov r0, #1 << 6 @ cp6 for IOP3xx and Bulverde
567 orr r0, r0, #1 << 13 @ Its undefined whether this
568 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes
572 mrc p15, 0, r0, c1, c0, 0 @ get control register
576 .size __xscale_setup, . - __xscale_setup
580 * .RVI ZFRS BLDP WCAM
581 * ..11 1.01 .... .101
584 .type xscale_crval, #object
586 crval clear=0x00003b07, mmuset=0x00003905, ucset=0x00001900
590 @ define struct processor (see <asm/proc-fns.h> and proc-macros.S)
591 define_processor_functions xscale, dabort=v5t_early_abort, pabort=legacy_pabort, suspend=1
595 string cpu_arch_name, "armv5te"
596 string cpu_elf_name, "v5"
598 string cpu_80200_A0_A1_name, "XScale-80200 A0/A1"
599 string cpu_80200_name, "XScale-80200"
600 string cpu_80219_name, "XScale-80219"
601 string cpu_8032x_name, "XScale-IOP8032x Family"
602 string cpu_8033x_name, "XScale-IOP8033x Family"
603 string cpu_pxa250_name, "XScale-PXA250"
604 string cpu_pxa210_name, "XScale-PXA210"
605 string cpu_ixp42x_name, "XScale-IXP42x Family"
606 string cpu_ixp43x_name, "XScale-IXP43x Family"
607 string cpu_ixp46x_name, "XScale-IXP46x Family"
608 string cpu_ixp2400_name, "XScale-IXP2400"
609 string cpu_ixp2800_name, "XScale-IXP2800"
610 string cpu_pxa255_name, "XScale-PXA255"
611 string cpu_pxa270_name, "XScale-PXA270"
615 .section ".proc.info.init", #alloc, #execinstr
617 .macro xscale_proc_info name:req, cpu_val:req, cpu_mask:req, cpu_name:req, cache
618 .type __\name\()_proc_info,#object
619 __\name\()_proc_info:
622 .long PMD_TYPE_SECT | \
623 PMD_SECT_BUFFERABLE | \
624 PMD_SECT_CACHEABLE | \
625 PMD_SECT_AP_WRITE | \
627 .long PMD_TYPE_SECT | \
628 PMD_SECT_AP_WRITE | \
633 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
635 .long xscale_processor_functions
637 .long xscale_mc_user_fns
639 .long xscale_cache_fns
643 .size __\name\()_proc_info, . - __\name\()_proc_info
646 xscale_proc_info 80200_A0_A1, 0x69052000, 0xfffffffe, cpu_80200_name, \
647 cache=xscale_80200_A0_A1_cache_fns
648 xscale_proc_info 80200, 0x69052000, 0xfffffff0, cpu_80200_name
649 xscale_proc_info 80219, 0x69052e20, 0xffffffe0, cpu_80219_name
650 xscale_proc_info 8032x, 0x69052420, 0xfffff7e0, cpu_8032x_name
651 xscale_proc_info 8033x, 0x69054010, 0xfffffd30, cpu_8033x_name
652 xscale_proc_info pxa250, 0x69052100, 0xfffff7f0, cpu_pxa250_name
653 xscale_proc_info pxa210, 0x69052120, 0xfffff3f0, cpu_pxa210_name
654 xscale_proc_info ixp2400, 0x69054190, 0xfffffff0, cpu_ixp2400_name
655 xscale_proc_info ixp2800, 0x690541a0, 0xfffffff0, cpu_ixp2800_name
656 xscale_proc_info ixp42x, 0x690541c0, 0xffffffc0, cpu_ixp42x_name
657 xscale_proc_info ixp43x, 0x69054040, 0xfffffff0, cpu_ixp43x_name
658 xscale_proc_info ixp46x, 0x69054200, 0xffffff00, cpu_ixp46x_name
659 xscale_proc_info pxa255, 0x69052d00, 0xfffffff0, cpu_pxa255_name
660 xscale_proc_info pxa270, 0x69054110, 0xfffffff0, cpu_pxa270_name