2 * linux/arch/arm/mm/proc-xsc3.S
4 * Original Author: Matthew Gilbert
5 * Current Maintainer: Lennert Buytenhek <buytenh@wantstofly.org>
7 * Copyright 2004 (C) Intel Corp.
8 * Copyright 2005 (C) MontaVista Software, Inc.
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2 as
12 * published by the Free Software Foundation.
14 * MMU functions for the Intel XScale3 Core (XSC3). The XSC3 core is
15 * an extension to Intel's original XScale core that adds the following
18 * - ARMv6 Supersections
19 * - Low Locality Reference pages (replaces mini-cache)
22 * - Cache coherency if chipset supports it
24 * Based on original XScale code by Nicolas Pitre.
27 #include <linux/linkage.h>
28 #include <linux/init.h>
29 #include <asm/assembler.h>
30 #include <asm/hwcap.h>
31 #include <mach/hardware.h>
32 #include <asm/pgtable.h>
33 #include <asm/pgtable-hwdef.h>
35 #include <asm/ptrace.h>
36 #include "proc-macros.S"
39 * This is the maximum size of an area which will be flushed. If the
40 * area is larger than this, then we flush the whole cache.
42 #define MAX_AREA_SIZE 32768
45 * The cache line size of the L1 I, L1 D and unified L2 cache.
47 #define CACHELINESIZE 32
50 * The size of the L1 D cache.
52 #define CACHESIZE 32768
55 * This macro is used to wait for a CP15 write and is needed when we
56 * have to ensure that the last operation to the coprocessor was
57 * completed before continuing with operation.
59 .macro cpwait_ret, lr, rd
60 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15
61 sub pc, \lr, \rd, LSR #32 @ wait for completion and
62 @ flush instruction pipeline
66 * This macro cleans and invalidates the entire L1 D cache.
69 .macro clean_d_cache rd, rs
72 1: mcr p15, 0, \rd, c7, c14, 2 @ clean/invalidate L1 D line
73 adds \rd, \rd, #0x40000000
82 * cpu_xsc3_proc_init()
84 * Nothing too exciting at the moment
86 ENTRY(cpu_xsc3_proc_init)
92 ENTRY(cpu_xsc3_proc_fin)
94 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
96 bl xsc3_flush_kern_cache_all @ clean caches
97 mrc p15, 0, r0, c1, c0, 0 @ ctrl register
98 bic r0, r0, #0x1800 @ ...IZ...........
99 bic r0, r0, #0x0006 @ .............CA.
100 mcr p15, 0, r0, c1, c0, 0 @ disable caches
104 * cpu_xsc3_reset(loc)
106 * Perform a soft reset of the system. Put the CPU into the
107 * same state as it would be if it had been reset, and branch
108 * to what would be the reset vector.
110 * loc: location to jump to for soft reset
113 ENTRY(cpu_xsc3_reset)
114 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
115 msr cpsr_c, r1 @ reset CPSR
116 mrc p15, 0, r1, c1, c0, 0 @ ctrl register
117 bic r1, r1, #0x3900 @ ..VIZ..S........
118 bic r1, r1, #0x0086 @ ........B....CA.
119 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
120 mcr p15, 0, ip, c7, c7, 0 @ invalidate L1 caches and BTB
121 bic r1, r1, #0x0001 @ ...............M
122 mcr p15, 0, r1, c1, c0, 0 @ ctrl register
123 @ CAUTION: MMU turned off from this point. We count on the pipeline
124 @ already containing those two last instructions to survive.
125 mcr p15, 0, ip, c8, c7, 0 @ invalidate I and D TLBs
131 * Cause the processor to idle
133 * For now we do nothing but go to idle mode for every case
135 * XScale supports clock switching, but using idle mode support
136 * allows external hardware to react to system state changes.
140 ENTRY(cpu_xsc3_do_idle)
142 mcr p14, 0, r0, c7, c0, 0 @ go to idle
145 /* ================================= CACHE ================================ */
148 * flush_user_cache_all()
150 * Invalidate all cache entries in a particular address
153 ENTRY(xsc3_flush_user_cache_all)
157 * flush_kern_cache_all()
159 * Clean and invalidate the entire cache.
161 ENTRY(xsc3_flush_kern_cache_all)
167 mcrne p15, 0, ip, c7, c5, 0 @ invalidate L1 I cache and BTB
168 mcrne p15, 0, ip, c7, c10, 4 @ data write barrier
169 mcrne p15, 0, ip, c7, c5, 4 @ prefetch flush
173 * flush_user_cache_range(start, end, vm_flags)
175 * Invalidate a range of cache entries in the specified
178 * - start - start address (may not be aligned)
179 * - end - end address (exclusive, may not be aligned)
180 * - vma - vma_area_struct describing address space
183 ENTRY(xsc3_flush_user_cache_range)
185 sub r3, r1, r0 @ calculate total size
186 cmp r3, #MAX_AREA_SIZE
187 bhs __flush_whole_cache
190 mcrne p15, 0, r0, c7, c5, 1 @ invalidate L1 I line
191 mcr p15, 0, r0, c7, c14, 1 @ clean/invalidate L1 D line
192 add r0, r0, #CACHELINESIZE
196 mcrne p15, 0, ip, c7, c5, 6 @ invalidate BTB
197 mcrne p15, 0, ip, c7, c10, 4 @ data write barrier
198 mcrne p15, 0, ip, c7, c5, 4 @ prefetch flush
202 * coherent_kern_range(start, end)
204 * Ensure coherency between the I cache and the D cache in the
205 * region described by start. If you have non-snooping
206 * Harvard caches, you need to implement this function.
208 * - start - virtual start address
209 * - end - virtual end address
211 * Note: single I-cache line invalidation isn't used here since
212 * it also trashes the mini I-cache used by JTAG debuggers.
214 ENTRY(xsc3_coherent_kern_range)
216 ENTRY(xsc3_coherent_user_range)
217 bic r0, r0, #CACHELINESIZE - 1
218 1: mcr p15, 0, r0, c7, c10, 1 @ clean L1 D line
219 add r0, r0, #CACHELINESIZE
223 mcr p15, 0, r0, c7, c5, 0 @ invalidate L1 I cache and BTB
224 mcr p15, 0, r0, c7, c10, 4 @ data write barrier
225 mcr p15, 0, r0, c7, c5, 4 @ prefetch flush
229 * flush_kern_dcache_area(void *addr, size_t size)
231 * Ensure no D cache aliasing occurs, either with itself or
234 * - addr - kernel address
235 * - size - region size
237 ENTRY(xsc3_flush_kern_dcache_area)
239 1: mcr p15, 0, r0, c7, c14, 1 @ clean/invalidate L1 D line
240 add r0, r0, #CACHELINESIZE
244 mcr p15, 0, r0, c7, c5, 0 @ invalidate L1 I cache and BTB
245 mcr p15, 0, r0, c7, c10, 4 @ data write barrier
246 mcr p15, 0, r0, c7, c5, 4 @ prefetch flush
250 * dma_inv_range(start, end)
252 * Invalidate (discard) the specified virtual address range.
253 * May not write back any entries. If 'start' or 'end'
254 * are not cache line aligned, those lines must be written
257 * - start - virtual start address
258 * - end - virtual end address
260 ENTRY(xsc3_dma_inv_range)
261 tst r0, #CACHELINESIZE - 1
262 bic r0, r0, #CACHELINESIZE - 1
263 mcrne p15, 0, r0, c7, c10, 1 @ clean L1 D line
264 tst r1, #CACHELINESIZE - 1
265 mcrne p15, 0, r1, c7, c10, 1 @ clean L1 D line
266 1: mcr p15, 0, r0, c7, c6, 1 @ invalidate L1 D line
267 add r0, r0, #CACHELINESIZE
270 mcr p15, 0, r0, c7, c10, 4 @ data write barrier
274 * dma_clean_range(start, end)
276 * Clean the specified virtual address range.
278 * - start - virtual start address
279 * - end - virtual end address
281 ENTRY(xsc3_dma_clean_range)
282 bic r0, r0, #CACHELINESIZE - 1
283 1: mcr p15, 0, r0, c7, c10, 1 @ clean L1 D line
284 add r0, r0, #CACHELINESIZE
287 mcr p15, 0, r0, c7, c10, 4 @ data write barrier
291 * dma_flush_range(start, end)
293 * Clean and invalidate the specified virtual address range.
295 * - start - virtual start address
296 * - end - virtual end address
298 ENTRY(xsc3_dma_flush_range)
299 bic r0, r0, #CACHELINESIZE - 1
300 1: mcr p15, 0, r0, c7, c14, 1 @ clean/invalidate L1 D line
301 add r0, r0, #CACHELINESIZE
304 mcr p15, 0, r0, c7, c10, 4 @ data write barrier
307 ENTRY(xsc3_cache_fns)
308 .long xsc3_flush_kern_cache_all
309 .long xsc3_flush_user_cache_all
310 .long xsc3_flush_user_cache_range
311 .long xsc3_coherent_kern_range
312 .long xsc3_coherent_user_range
313 .long xsc3_flush_kern_dcache_area
314 .long xsc3_dma_inv_range
315 .long xsc3_dma_clean_range
316 .long xsc3_dma_flush_range
318 ENTRY(cpu_xsc3_dcache_clean_area)
319 1: mcr p15, 0, r0, c7, c10, 1 @ clean L1 D line
320 add r0, r0, #CACHELINESIZE
321 subs r1, r1, #CACHELINESIZE
325 /* =============================== PageTable ============================== */
328 * cpu_xsc3_switch_mm(pgd)
330 * Set the translation base pointer to be as described by pgd.
332 * pgd: new page tables
335 ENTRY(cpu_xsc3_switch_mm)
337 mcr p15, 0, ip, c7, c5, 0 @ invalidate L1 I cache and BTB
338 mcr p15, 0, ip, c7, c10, 4 @ data write barrier
339 mcr p15, 0, ip, c7, c5, 4 @ prefetch flush
340 orr r0, r0, #0x18 @ cache the page table in L2
341 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer
342 mcr p15, 0, ip, c8, c7, 0 @ invalidate I and D TLBs
346 * cpu_xsc3_set_pte_ext(ptep, pte, ext)
348 * Set a PTE and flush it out
351 .long 0x00 @ L_PTE_MT_UNCACHED
352 .long PTE_EXT_TEX(1) @ L_PTE_MT_BUFFERABLE
353 .long PTE_EXT_TEX(5) | PTE_CACHEABLE @ L_PTE_MT_WRITETHROUGH
354 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEBACK
355 .long PTE_EXT_TEX(1) | PTE_BUFFERABLE @ L_PTE_MT_DEV_SHARED
357 .long 0x00 @ L_PTE_MT_MINICACHE (not present)
358 .long PTE_EXT_TEX(5) | PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_WRITEALLOC (not present?)
360 .long PTE_EXT_TEX(1) @ L_PTE_MT_DEV_WC
362 .long PTE_CACHEABLE | PTE_BUFFERABLE @ L_PTE_MT_DEV_CACHED
363 .long PTE_EXT_TEX(2) @ L_PTE_MT_DEV_NONSHARED
369 ENTRY(cpu_xsc3_set_pte_ext)
370 xscale_set_pte_ext_prologue
372 tst r1, #L_PTE_SHARED @ shared?
373 and r1, r1, #L_PTE_MT_MASK
374 adr ip, cpu_xsc3_mt_table
376 orrne r2, r2, #PTE_EXT_COHERENT @ interlock: mask in coherent bit
377 bic r2, r2, #0x0c @ clear old C,B bits
380 xscale_set_pte_ext_epilogue
389 .type __xsc3_setup, #function
391 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE
393 mcr p15, 0, ip, c7, c7, 0 @ invalidate L1 caches and BTB
394 mcr p15, 0, ip, c7, c10, 4 @ data write barrier
395 mcr p15, 0, ip, c7, c5, 4 @ prefetch flush
396 mcr p15, 0, ip, c8, c7, 0 @ invalidate I and D TLBs
397 orr r4, r4, #0x18 @ cache the page table in L2
398 mcr p15, 0, r4, c2, c0, 0 @ load page table pointer
400 mov r0, #1 << 6 @ cp6 access for early sched_clock
401 mcr p15, 0, r0, c15, c1, 0 @ write CP access register
403 mrc p15, 0, r0, c1, c0, 1 @ get auxiliary control reg
404 and r0, r0, #2 @ preserve bit P bit setting
405 orr r0, r0, #(1 << 10) @ enable L2 for LLR cache
406 mcr p15, 0, r0, c1, c0, 1 @ set auxiliary control reg
411 #ifdef CONFIG_CACHE_XSC3L2
412 mrc p15, 1, r0, c0, c0, 1 @ get L2 present information
414 orrne r6, r6, #(1 << 26) @ enable L2 if present
417 mrc p15, 0, r0, c1, c0, 0 @ get control register
418 bic r0, r0, r5 @ ..V. ..R. .... ..A.
419 orr r0, r0, r6 @ ..VI Z..S .... .C.M (mmu)
420 @ ...I Z..S .... .... (uc)
423 .size __xsc3_setup, . - __xsc3_setup
425 .type xsc3_crval, #object
427 crval clear=0x04002202, mmuset=0x00003905, ucset=0x00001900
432 * Purpose : Function pointers used to access above functions - all calls
436 .type xsc3_processor_functions, #object
437 ENTRY(xsc3_processor_functions)
438 .word v5t_early_abort
440 .word cpu_xsc3_proc_init
441 .word cpu_xsc3_proc_fin
443 .word cpu_xsc3_do_idle
444 .word cpu_xsc3_dcache_clean_area
445 .word cpu_xsc3_switch_mm
446 .word cpu_xsc3_set_pte_ext
447 .size xsc3_processor_functions, . - xsc3_processor_functions
451 .type cpu_arch_name, #object
454 .size cpu_arch_name, . - cpu_arch_name
456 .type cpu_elf_name, #object
459 .size cpu_elf_name, . - cpu_elf_name
461 .type cpu_xsc3_name, #object
463 .asciz "XScale-V3 based processor"
464 .size cpu_xsc3_name, . - cpu_xsc3_name
468 .section ".proc.info.init", #alloc, #execinstr
470 .type __xsc3_proc_info,#object
474 .long PMD_TYPE_SECT | \
475 PMD_SECT_BUFFERABLE | \
476 PMD_SECT_CACHEABLE | \
477 PMD_SECT_AP_WRITE | \
479 .long PMD_TYPE_SECT | \
480 PMD_SECT_AP_WRITE | \
485 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
487 .long xsc3_processor_functions
489 .long xsc3_mc_user_fns
491 .size __xsc3_proc_info, . - __xsc3_proc_info
493 /* Note: PXA935 changed its implementor ID from Intel to Marvell */
495 .type __xsc3_pxa935_proc_info,#object
496 __xsc3_pxa935_proc_info:
499 .long PMD_TYPE_SECT | \
500 PMD_SECT_BUFFERABLE | \
501 PMD_SECT_CACHEABLE | \
502 PMD_SECT_AP_WRITE | \
504 .long PMD_TYPE_SECT | \
505 PMD_SECT_AP_WRITE | \
510 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
512 .long xsc3_processor_functions
514 .long xsc3_mc_user_fns
516 .size __xsc3_pxa935_proc_info, . - __xsc3_pxa935_proc_info