2 * Copyright (C) 1995-1999 Gary Thomas, Paul Mackerras, Cort Dougan.
4 #ifndef _ASM_POWERPC_PPC_ASM_H
5 #define _ASM_POWERPC_PPC_ASM_H
7 #include <linux/stringify.h>
8 #include <asm/asm-compat.h>
9 #include <asm/processor.h>
10 #include <asm/ppc-opcode.h>
11 #include <asm/firmware.h>
15 #define SZL (BITS_PER_LONG/8)
18 * Stuff for accurate CPU time accounting.
19 * These macros handle transitions between user and system state
20 * in exception entry and exit and accumulate time to the
21 * user_time and system_time fields in the paca.
24 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
25 #define ACCOUNT_CPU_USER_ENTRY(ptr, ra, rb)
26 #define ACCOUNT_CPU_USER_EXIT(ptr, ra, rb)
27 #define ACCOUNT_STOLEN_TIME
29 #define ACCOUNT_CPU_USER_ENTRY(ptr, ra, rb) \
30 MFTB(ra); /* get timebase */ \
31 PPC_LL rb, ACCOUNT_STARTTIME_USER(ptr); \
32 PPC_STL ra, ACCOUNT_STARTTIME(ptr); \
33 subf rb,rb,ra; /* subtract start value */ \
34 PPC_LL ra, ACCOUNT_USER_TIME(ptr); \
35 add ra,ra,rb; /* add on to user time */ \
36 PPC_STL ra, ACCOUNT_USER_TIME(ptr); \
38 #define ACCOUNT_CPU_USER_EXIT(ptr, ra, rb) \
39 MFTB(ra); /* get timebase */ \
40 PPC_LL rb, ACCOUNT_STARTTIME(ptr); \
41 PPC_STL ra, ACCOUNT_STARTTIME_USER(ptr); \
42 subf rb,rb,ra; /* subtract start value */ \
43 PPC_LL ra, ACCOUNT_SYSTEM_TIME(ptr); \
44 add ra,ra,rb; /* add on to system time */ \
45 PPC_STL ra, ACCOUNT_SYSTEM_TIME(ptr)
47 #ifdef CONFIG_PPC_SPLPAR
48 #define ACCOUNT_STOLEN_TIME \
49 BEGIN_FW_FTR_SECTION; \
51 /* from user - see if there are any DTL entries to process */ \
52 ld r10,PACALPPACAPTR(r13); /* get ptr to VPA */ \
53 ld r11,PACA_DTL_RIDX(r13); /* get log read index */ \
54 addi r10,r10,LPPACA_DTLIDX; \
55 LDX_BE r10,0,r10; /* get log write index */ \
58 bl accumulate_stolen_time; \
60 andi. r10,r12,MSR_PR; /* Restore cr0 (coming from user) */ \
62 END_FW_FTR_SECTION_IFSET(FW_FEATURE_SPLPAR)
64 #else /* CONFIG_PPC_SPLPAR */
65 #define ACCOUNT_STOLEN_TIME
67 #endif /* CONFIG_PPC_SPLPAR */
69 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
72 * Macros for storing registers into and loading registers from
76 #define SAVE_GPR(n, base) std n,GPR0+8*(n)(base)
77 #define REST_GPR(n, base) ld n,GPR0+8*(n)(base)
78 #define SAVE_NVGPRS(base) SAVE_8GPRS(14, base); SAVE_10GPRS(22, base)
79 #define REST_NVGPRS(base) REST_8GPRS(14, base); REST_10GPRS(22, base)
81 #define SAVE_GPR(n, base) stw n,GPR0+4*(n)(base)
82 #define REST_GPR(n, base) lwz n,GPR0+4*(n)(base)
83 #define SAVE_NVGPRS(base) SAVE_GPR(13, base); SAVE_8GPRS(14, base); \
85 #define REST_NVGPRS(base) REST_GPR(13, base); REST_8GPRS(14, base); \
89 #define SAVE_2GPRS(n, base) SAVE_GPR(n, base); SAVE_GPR(n+1, base)
90 #define SAVE_4GPRS(n, base) SAVE_2GPRS(n, base); SAVE_2GPRS(n+2, base)
91 #define SAVE_8GPRS(n, base) SAVE_4GPRS(n, base); SAVE_4GPRS(n+4, base)
92 #define SAVE_10GPRS(n, base) SAVE_8GPRS(n, base); SAVE_2GPRS(n+8, base)
93 #define REST_2GPRS(n, base) REST_GPR(n, base); REST_GPR(n+1, base)
94 #define REST_4GPRS(n, base) REST_2GPRS(n, base); REST_2GPRS(n+2, base)
95 #define REST_8GPRS(n, base) REST_4GPRS(n, base); REST_4GPRS(n+4, base)
96 #define REST_10GPRS(n, base) REST_8GPRS(n, base); REST_2GPRS(n+8, base)
98 #define SAVE_FPR(n, base) stfd n,8*TS_FPRWIDTH*(n)(base)
99 #define SAVE_2FPRS(n, base) SAVE_FPR(n, base); SAVE_FPR(n+1, base)
100 #define SAVE_4FPRS(n, base) SAVE_2FPRS(n, base); SAVE_2FPRS(n+2, base)
101 #define SAVE_8FPRS(n, base) SAVE_4FPRS(n, base); SAVE_4FPRS(n+4, base)
102 #define SAVE_16FPRS(n, base) SAVE_8FPRS(n, base); SAVE_8FPRS(n+8, base)
103 #define SAVE_32FPRS(n, base) SAVE_16FPRS(n, base); SAVE_16FPRS(n+16, base)
104 #define REST_FPR(n, base) lfd n,8*TS_FPRWIDTH*(n)(base)
105 #define REST_2FPRS(n, base) REST_FPR(n, base); REST_FPR(n+1, base)
106 #define REST_4FPRS(n, base) REST_2FPRS(n, base); REST_2FPRS(n+2, base)
107 #define REST_8FPRS(n, base) REST_4FPRS(n, base); REST_4FPRS(n+4, base)
108 #define REST_16FPRS(n, base) REST_8FPRS(n, base); REST_8FPRS(n+8, base)
109 #define REST_32FPRS(n, base) REST_16FPRS(n, base); REST_16FPRS(n+16, base)
111 #define SAVE_VR(n,b,base) li b,16*(n); stvx n,base,b
112 #define SAVE_2VRS(n,b,base) SAVE_VR(n,b,base); SAVE_VR(n+1,b,base)
113 #define SAVE_4VRS(n,b,base) SAVE_2VRS(n,b,base); SAVE_2VRS(n+2,b,base)
114 #define SAVE_8VRS(n,b,base) SAVE_4VRS(n,b,base); SAVE_4VRS(n+4,b,base)
115 #define SAVE_16VRS(n,b,base) SAVE_8VRS(n,b,base); SAVE_8VRS(n+8,b,base)
116 #define SAVE_32VRS(n,b,base) SAVE_16VRS(n,b,base); SAVE_16VRS(n+16,b,base)
117 #define REST_VR(n,b,base) li b,16*(n); lvx n,base,b
118 #define REST_2VRS(n,b,base) REST_VR(n,b,base); REST_VR(n+1,b,base)
119 #define REST_4VRS(n,b,base) REST_2VRS(n,b,base); REST_2VRS(n+2,b,base)
120 #define REST_8VRS(n,b,base) REST_4VRS(n,b,base); REST_4VRS(n+4,b,base)
121 #define REST_16VRS(n,b,base) REST_8VRS(n,b,base); REST_8VRS(n+8,b,base)
122 #define REST_32VRS(n,b,base) REST_16VRS(n,b,base); REST_16VRS(n+16,b,base)
124 #ifdef __BIG_ENDIAN__
125 #define STXVD2X_ROT(n,b,base) STXVD2X(n,b,base)
126 #define LXVD2X_ROT(n,b,base) LXVD2X(n,b,base)
128 #define STXVD2X_ROT(n,b,base) XXSWAPD(n,n); \
132 #define LXVD2X_ROT(n,b,base) LXVD2X(n,b,base); \
135 /* Save the lower 32 VSRs in the thread VSR region */
136 #define SAVE_VSR(n,b,base) li b,16*(n); STXVD2X_ROT(n,R##base,R##b)
137 #define SAVE_2VSRS(n,b,base) SAVE_VSR(n,b,base); SAVE_VSR(n+1,b,base)
138 #define SAVE_4VSRS(n,b,base) SAVE_2VSRS(n,b,base); SAVE_2VSRS(n+2,b,base)
139 #define SAVE_8VSRS(n,b,base) SAVE_4VSRS(n,b,base); SAVE_4VSRS(n+4,b,base)
140 #define SAVE_16VSRS(n,b,base) SAVE_8VSRS(n,b,base); SAVE_8VSRS(n+8,b,base)
141 #define SAVE_32VSRS(n,b,base) SAVE_16VSRS(n,b,base); SAVE_16VSRS(n+16,b,base)
142 #define REST_VSR(n,b,base) li b,16*(n); LXVD2X_ROT(n,R##base,R##b)
143 #define REST_2VSRS(n,b,base) REST_VSR(n,b,base); REST_VSR(n+1,b,base)
144 #define REST_4VSRS(n,b,base) REST_2VSRS(n,b,base); REST_2VSRS(n+2,b,base)
145 #define REST_8VSRS(n,b,base) REST_4VSRS(n,b,base); REST_4VSRS(n+4,b,base)
146 #define REST_16VSRS(n,b,base) REST_8VSRS(n,b,base); REST_8VSRS(n+8,b,base)
147 #define REST_32VSRS(n,b,base) REST_16VSRS(n,b,base); REST_16VSRS(n+16,b,base)
150 * b = base register for addressing, o = base offset from register of 1st EVR
151 * n = first EVR, s = scratch
153 #define SAVE_EVR(n,s,b,o) evmergehi s,s,n; stw s,o+4*(n)(b)
154 #define SAVE_2EVRS(n,s,b,o) SAVE_EVR(n,s,b,o); SAVE_EVR(n+1,s,b,o)
155 #define SAVE_4EVRS(n,s,b,o) SAVE_2EVRS(n,s,b,o); SAVE_2EVRS(n+2,s,b,o)
156 #define SAVE_8EVRS(n,s,b,o) SAVE_4EVRS(n,s,b,o); SAVE_4EVRS(n+4,s,b,o)
157 #define SAVE_16EVRS(n,s,b,o) SAVE_8EVRS(n,s,b,o); SAVE_8EVRS(n+8,s,b,o)
158 #define SAVE_32EVRS(n,s,b,o) SAVE_16EVRS(n,s,b,o); SAVE_16EVRS(n+16,s,b,o)
159 #define REST_EVR(n,s,b,o) lwz s,o+4*(n)(b); evmergelo n,s,n
160 #define REST_2EVRS(n,s,b,o) REST_EVR(n,s,b,o); REST_EVR(n+1,s,b,o)
161 #define REST_4EVRS(n,s,b,o) REST_2EVRS(n,s,b,o); REST_2EVRS(n+2,s,b,o)
162 #define REST_8EVRS(n,s,b,o) REST_4EVRS(n,s,b,o); REST_4EVRS(n+4,s,b,o)
163 #define REST_16EVRS(n,s,b,o) REST_8EVRS(n,s,b,o); REST_8EVRS(n+8,s,b,o)
164 #define REST_32EVRS(n,s,b,o) REST_16EVRS(n,s,b,o); REST_16EVRS(n+16,s,b,o)
166 /* Macros to adjust thread priority for hardware multithreading */
167 #define HMT_VERY_LOW or 31,31,31 # very low priority
168 #define HMT_LOW or 1,1,1
169 #define HMT_MEDIUM_LOW or 6,6,6 # medium low priority
170 #define HMT_MEDIUM or 2,2,2
171 #define HMT_MEDIUM_HIGH or 5,5,5 # medium high priority
172 #define HMT_HIGH or 3,3,3
173 #define HMT_EXTRA_HIGH or 7,7,7 # power7 only
180 #define __VCPU_GPR(n) (VCPU_GPRS + (n * ULONG_SIZE))
181 #define VCPU_GPR(n) __VCPU_GPR(__REG_##n)
186 #define STACKFRAMESIZE 256
187 #define __STK_REG(i) (112 + ((i)-14)*8)
188 #define STK_REG(i) __STK_REG(__REG_##i)
190 #ifdef PPC64_ELF_ABI_v2
192 #define __STK_PARAM(i) (32 + ((i)-3)*8)
195 #define __STK_PARAM(i) (48 + ((i)-3)*8)
197 #define STK_PARAM(i) __STK_PARAM(__REG_##i)
199 #ifdef PPC64_ELF_ABI_v2
201 #define _GLOBAL(name) \
203 .type name,@function; \
207 #define _GLOBAL_TOC(name) \
209 .type name,@function; \
212 0: addis r2,r12,(.TOC.-0b)@ha; \
213 addi r2,r2,(.TOC.-0b)@l; \
214 .localentry name,.-name
220 #define XGLUE(a,b) a##b
221 #define GLUE(a,b) XGLUE(a,b)
223 #define _GLOBAL(name) \
226 .globl GLUE(.,name); \
227 .pushsection ".opd","aw"; \
229 .quad GLUE(.,name); \
230 .quad .TOC.@tocbase; \
233 .type GLUE(.,name),@function; \
236 #define _GLOBAL_TOC(name) _GLOBAL(name)
238 #define DOTSYM(a) GLUE(.,a)
249 .stabs __stringify(n:F-1),N_FUN,0,0,n;\
253 #define _GLOBAL_TOC(name) _GLOBAL(name)
258 * __kprobes (the C annotation) puts the symbol into the .kprobes.text
259 * section, which gets emitted at the end of regular text.
261 * _ASM_NOKPROBE_SYMBOL and NOKPROBE_SYMBOL just adds the symbol to
262 * a blacklist. The former is for core kprobe functions/data, the
263 * latter is for those that incdentially must be excluded from probing
264 * and allows them to be linked at more optimal location within text.
266 #ifdef CONFIG_KPROBES
267 #define _ASM_NOKPROBE_SYMBOL(entry) \
268 .pushsection "_kprobe_blacklist","aw"; \
272 #define _ASM_NOKPROBE_SYMBOL(entry)
275 #define FUNC_START(name) _GLOBAL(name)
276 #define FUNC_END(name)
279 * LOAD_REG_IMMEDIATE(rn, expr)
280 * Loads the value of the constant expression 'expr' into register 'rn'
281 * using immediate instructions only. Use this when it's important not
282 * to reference other data (i.e. on ppc64 when the TOC pointer is not
283 * valid) and when 'expr' is a constant or absolute address.
285 * LOAD_REG_ADDR(rn, name)
286 * Loads the address of label 'name' into register 'rn'. Use this when
287 * you don't particularly need immediate instructions only, but you need
288 * the whole address in one register (e.g. it's a structure address and
289 * you want to access various offsets within it). On ppc32 this is
290 * identical to LOAD_REG_IMMEDIATE.
292 * LOAD_REG_ADDR_PIC(rn, name)
293 * Loads the address of label 'name' into register 'run'. Use this when
294 * the kernel doesn't run at the linked or relocated address. Please
295 * note that this macro will clobber the lr register.
297 * LOAD_REG_ADDRBASE(rn, name)
299 * LOAD_REG_ADDRBASE loads part of the address of label 'name' into
300 * register 'rn'. ADDROFF(name) returns the remainder of the address as
301 * a constant expression. ADDROFF(name) is a signed expression < 16 bits
302 * in size, so is suitable for use directly as an offset in load and store
303 * instructions. Use this when loading/storing a single word or less as:
304 * LOAD_REG_ADDRBASE(rX, name)
305 * ld rY,ADDROFF(name)(rX)
308 /* Be careful, this will clobber the lr register. */
309 #define LOAD_REG_ADDR_PIC(reg, name) \
312 addis reg,reg,(name - 0b)@ha; \
313 addi reg,reg,(name - 0b)@l;
316 #ifdef HAVE_AS_ATHIGH
317 #define __AS_ATHIGH high
319 #define __AS_ATHIGH h
321 #define LOAD_REG_IMMEDIATE(reg,expr) \
322 lis reg,(expr)@highest; \
323 ori reg,reg,(expr)@higher; \
324 rldicr reg,reg,32,31; \
325 oris reg,reg,(expr)@__AS_ATHIGH; \
326 ori reg,reg,(expr)@l;
328 #define LOAD_REG_ADDR(reg,name) \
331 #define LOAD_REG_ADDRBASE(reg,name) LOAD_REG_ADDR(reg,name)
332 #define ADDROFF(name) 0
334 /* offsets for stack frame layout */
339 #define LOAD_REG_IMMEDIATE(reg,expr) \
341 addi reg,reg,(expr)@l;
343 #define LOAD_REG_ADDR(reg,name) LOAD_REG_IMMEDIATE(reg, name)
345 #define LOAD_REG_ADDRBASE(reg, name) lis reg,name@ha
346 #define ADDROFF(name) name@l
348 /* offsets for stack frame layout */
353 /* various errata or part fixups */
354 #ifdef CONFIG_PPC601_SYNC_FIX
359 END_FTR_SECTION_IFSET(CPU_FTR_601)
363 END_FTR_SECTION_IFSET(CPU_FTR_601)
367 END_FTR_SECTION_IFSET(CPU_FTR_601)
374 #if defined(CONFIG_PPC_CELL) || defined(CONFIG_PPC_FSL_BOOK3E)
376 90: mfspr dest, SPRN_TBRL; \
377 BEGIN_FTR_SECTION_NESTED(96); \
380 END_FTR_SECTION_NESTED(CPU_FTR_CELL_TB_BUG, CPU_FTR_CELL_TB_BUG, 96)
382 #define MFTB(dest) MFTBL(dest)
385 #ifdef CONFIG_PPC_8xx
386 #define MFTBL(dest) mftb dest
387 #define MFTBU(dest) mftbu dest
389 #define MFTBL(dest) mfspr dest, SPRN_TBRL
390 #define MFTBU(dest) mfspr dest, SPRN_TBRU
395 #else /* CONFIG_SMP */
396 /* tlbsync is not implemented on 601 */
401 END_FTR_SECTION_IFCLR(CPU_FTR_601)
405 #define MTOCRF(FXM, RS) \
406 BEGIN_FTR_SECTION_NESTED(848); \
408 FTR_SECTION_ELSE_NESTED(848); \
410 ALT_FTR_SECTION_END_NESTED_IFCLR(CPU_FTR_NOEXECUTE, 848)
414 * This instruction is not implemented on the PPC 603 or 601; however, on
415 * the 403GCX and 405GP tlbia IS defined and tlbie is not.
416 * All of these instructions exist in the 8xx, they have magical powers,
417 * and they must be used.
420 #if !defined(CONFIG_4xx) && !defined(CONFIG_PPC_8xx)
424 lis r4,KERNELBASE@h; \
434 #ifdef CONFIG_IBM440EP_ERR42
435 #define PPC440EP_ERR42 isync
437 #define PPC440EP_ERR42
440 /* The following stops all load and store data streams associated with stream
441 * ID (ie. streams created explicitly). The embedded and server mnemonics for
442 * dcbt are different so we use machine "power4" here explicitly.
444 #define DCBT_STOP_ALL_STREAM_IDS(scratch) \
446 .machine "power4" ; \
447 lis scratch,0x60000000@h; \
448 dcbt 0,scratch,0b01010; \
452 * toreal/fromreal/tophys/tovirt macros. 32-bit BookE makes them
453 * keep the address intact to be compatible with code shared with
456 * On the other hand, I find it useful to have them behave as expected
457 * by their name (ie always do the addition) on 64-bit BookE
459 #if defined(CONFIG_BOOKE) && !defined(CONFIG_PPC64)
464 * We use addis to ensure compatibility with the "classic" ppc versions of
465 * these macros, which use rs = 0 to get the tophys offset in rd, rather than
466 * converting the address in r0, and so this version has to do that too
467 * (i.e. set register rd to 0 when rs == 0).
469 #define tophys(rd,rs) \
472 #define tovirt(rd,rs) \
475 #elif defined(CONFIG_PPC64)
476 #define toreal(rd) /* we can access c000... in real mode */
479 #define tophys(rd,rs) \
482 #define tovirt(rd,rs) \
484 ori rd,rd,((KERNELBASE>>48)&0xFFFF);\
488 * On APUS (Amiga PowerPC cpu upgrade board), we don't know the
489 * physical base address of RAM at compile time.
491 #define toreal(rd) tophys(rd,rd)
492 #define fromreal(rd) tovirt(rd,rd)
494 #define tophys(rd,rs) \
495 0: addis rd,rs,-PAGE_OFFSET@h; \
496 .section ".vtop_fixup","aw"; \
501 #define tovirt(rd,rs) \
502 0: addis rd,rs,PAGE_OFFSET@h; \
503 .section ".ptov_fixup","aw"; \
509 #ifdef CONFIG_PPC_BOOK3S_64
511 #define MTMSRD(r) mtmsrd r
512 #define MTMSR_EERI(reg) mtmsrd reg,1
517 #define RFI rfi; b . /* Prevent prefetch past rfi */
519 #define MTMSRD(r) mtmsr r
520 #define MTMSR_EERI(reg) mtmsr reg
523 #endif /* __KERNEL__ */
525 /* The boring bits... */
527 /* Condition Register Bit Fields */
540 * General Purpose Registers (GPRs)
542 * The lower case r0-r31 should be used in preference to the upper
543 * case R0-R31 as they provide more error checking in the assembler.
544 * Use R0-31 only when really nessesary.
581 /* Floating Point Registers (FPRs) */
616 /* AltiVec Registers (VPRs) */
651 /* VSX Registers (VSRs) */
718 /* SPE Registers (EVPRs) */
753 /* some stab codes */
760 * Create an endian fixup trampoline
762 * This starts with a "tdi 0,0,0x48" instruction which is
763 * essentially a "trap never", and thus akin to a nop.
765 * The opcode for this instruction read with the wrong endian
766 * however results in a b . + 8
768 * So essentially we use that trick to execute the following
769 * trampoline in "reverse endian" if we are running with the
770 * MSR_LE bit set the "wrong" way for whatever endianness the
771 * kernel is built for.
774 #ifdef CONFIG_PPC_BOOK3E
778 * This version may be used in in HV or non-HV context.
779 * MSR[EE] must be disabled.
781 #define FIXUP_ENDIAN \
782 tdi 0,0,0x48; /* Reverse endian of b . + 8 */ \
783 b 191f; /* Skip trampoline if endian is good */ \
784 .long 0xa600607d; /* mfmsr r11 */ \
785 .long 0x01006b69; /* xori r11,r11,1 */ \
786 .long 0x00004039; /* li r10,0 */ \
787 .long 0x6401417d; /* mtmsrd r10,1 */ \
788 .long 0x05009f42; /* bcl 20,31,$+4 */ \
789 .long 0xa602487d; /* mflr r10 */ \
790 .long 0x14004a39; /* addi r10,r10,20 */ \
791 .long 0xa6035a7d; /* mtsrr0 r10 */ \
792 .long 0xa6037b7d; /* mtsrr1 r11 */ \
793 .long 0x2400004c; /* rfid */ \
797 * This version that may only be used with MSR[HV]=1
798 * - Does not clear MSR[RI], so more robust.
799 * - Slightly smaller and faster.
801 #define FIXUP_ENDIAN_HV \
802 tdi 0,0,0x48; /* Reverse endian of b . + 8 */ \
803 b 191f; /* Skip trampoline if endian is good */ \
804 .long 0xa600607d; /* mfmsr r11 */ \
805 .long 0x01006b69; /* xori r11,r11,1 */ \
806 .long 0x05009f42; /* bcl 20,31,$+4 */ \
807 .long 0xa602487d; /* mflr r10 */ \
808 .long 0x14004a39; /* addi r10,r10,20 */ \
809 .long 0xa64b5a7d; /* mthsrr0 r10 */ \
810 .long 0xa64b7b7d; /* mthsrr1 r11 */ \
811 .long 0x2402004c; /* hrfid */ \
814 #endif /* !CONFIG_PPC_BOOK3E */
816 #endif /* __ASSEMBLY__ */
819 * Helper macro for exception table entries
821 #define EX_TABLE(_fault, _target) \
822 stringify_in_c(.section __ex_table,"a";)\
823 stringify_in_c(.balign 4;) \
824 stringify_in_c(.long (_fault) - . ;) \
825 stringify_in_c(.long (_target) - . ;) \
826 stringify_in_c(.previous)
828 #endif /* _ASM_POWERPC_PPC_ASM_H */