xref: /linux/arch/powerpc/include/asm/ppc_asm.h (revision 4e284e38)
1 /*
2  * Copyright (C) 1995-1999 Gary Thomas, Paul Mackerras, Cort Dougan.
3  */
4 #ifndef _ASM_POWERPC_PPC_ASM_H
5 #define _ASM_POWERPC_PPC_ASM_H
6 
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>
12 #include <asm/feature-fixups.h>
13 #include <asm/extable.h>
14 
15 #ifdef __ASSEMBLY__
16 
17 #define SZL			(BITS_PER_LONG/8)
18 
19 /*
20  * This expands to a sequence of operations with reg incrementing from
21  * start to end inclusive, of this form:
22  *
23  *   op  reg, (offset + (width * reg))(base)
24  *
25  * Note that offset is not the offset of the first operation unless start
26  * is zero (or width is zero).
27  */
28 .macro OP_REGS op, width, start, end, base, offset
29 	.Lreg=\start
30 	.rept (\end - \start + 1)
31 	\op	.Lreg, \offset + \width * .Lreg(\base)
32 	.Lreg=.Lreg+1
33 	.endr
34 .endm
35 
36 /*
37  * This expands to a sequence of register clears for regs start to end
38  * inclusive, of the form:
39  *
40  *   li rN, 0
41  */
42 .macro ZEROIZE_REGS start, end
43 	.Lreg=\start
44 	.rept (\end - \start + 1)
45 	li	.Lreg, 0
46 	.Lreg=.Lreg+1
47 	.endr
48 .endm
49 
50 /*
51  * Macros for storing registers into and loading registers from
52  * exception frames.
53  */
54 #ifdef __powerpc64__
55 #define SAVE_GPRS(start, end, base)	OP_REGS std, 8, start, end, base, GPR0
56 #define REST_GPRS(start, end, base)	OP_REGS ld, 8, start, end, base, GPR0
57 #define SAVE_NVGPRS(base)		SAVE_GPRS(14, 31, base)
58 #define REST_NVGPRS(base)		REST_GPRS(14, 31, base)
59 #else
60 #define SAVE_GPRS(start, end, base)	OP_REGS stw, 4, start, end, base, GPR0
61 #define REST_GPRS(start, end, base)	OP_REGS lwz, 4, start, end, base, GPR0
62 #define SAVE_NVGPRS(base)		SAVE_GPRS(13, 31, base)
63 #define REST_NVGPRS(base)		REST_GPRS(13, 31, base)
64 #endif
65 
66 #define	ZEROIZE_GPRS(start, end)	ZEROIZE_REGS start, end
67 #ifdef __powerpc64__
68 #define	ZEROIZE_NVGPRS()		ZEROIZE_GPRS(14, 31)
69 #else
70 #define	ZEROIZE_NVGPRS()		ZEROIZE_GPRS(13, 31)
71 #endif
72 #define	ZEROIZE_GPR(n)			ZEROIZE_GPRS(n, n)
73 
74 #define SAVE_GPR(n, base)		SAVE_GPRS(n, n, base)
75 #define REST_GPR(n, base)		REST_GPRS(n, n, base)
76 
77 /* macros for handling user register sanitisation */
78 #ifdef CONFIG_INTERRUPT_SANITIZE_REGISTERS
79 #define SANITIZE_SYSCALL_GPRS()			ZEROIZE_GPR(0);		\
80 						ZEROIZE_GPRS(5, 12);	\
81 						ZEROIZE_NVGPRS()
82 #define SANITIZE_GPR(n)				ZEROIZE_GPR(n)
83 #define SANITIZE_GPRS(start, end)		ZEROIZE_GPRS(start, end)
84 #define SANITIZE_NVGPRS()			ZEROIZE_NVGPRS()
85 #define SANITIZE_RESTORE_NVGPRS()		REST_NVGPRS(r1)
86 #define HANDLER_RESTORE_NVGPRS()
87 #else
88 #define SANITIZE_SYSCALL_GPRS()
89 #define SANITIZE_GPR(n)
90 #define SANITIZE_GPRS(start, end)
91 #define SANITIZE_NVGPRS()
92 #define SANITIZE_RESTORE_NVGPRS()
93 #define HANDLER_RESTORE_NVGPRS()		REST_NVGPRS(r1)
94 #endif /* CONFIG_INTERRUPT_SANITIZE_REGISTERS */
95 
96 #define SAVE_FPR(n, base)	stfd	n,8*TS_FPRWIDTH*(n)(base)
97 #define SAVE_2FPRS(n, base)	SAVE_FPR(n, base); SAVE_FPR(n+1, base)
98 #define SAVE_4FPRS(n, base)	SAVE_2FPRS(n, base); SAVE_2FPRS(n+2, base)
99 #define SAVE_8FPRS(n, base)	SAVE_4FPRS(n, base); SAVE_4FPRS(n+4, base)
100 #define SAVE_16FPRS(n, base)	SAVE_8FPRS(n, base); SAVE_8FPRS(n+8, base)
101 #define SAVE_32FPRS(n, base)	SAVE_16FPRS(n, base); SAVE_16FPRS(n+16, base)
102 #define REST_FPR(n, base)	lfd	n,8*TS_FPRWIDTH*(n)(base)
103 #define REST_2FPRS(n, base)	REST_FPR(n, base); REST_FPR(n+1, base)
104 #define REST_4FPRS(n, base)	REST_2FPRS(n, base); REST_2FPRS(n+2, base)
105 #define REST_8FPRS(n, base)	REST_4FPRS(n, base); REST_4FPRS(n+4, base)
106 #define REST_16FPRS(n, base)	REST_8FPRS(n, base); REST_8FPRS(n+8, base)
107 #define REST_32FPRS(n, base)	REST_16FPRS(n, base); REST_16FPRS(n+16, base)
108 
109 #define SAVE_VR(n,b,base)	li b,16*(n);  stvx n,base,b
110 #define SAVE_2VRS(n,b,base)	SAVE_VR(n,b,base); SAVE_VR(n+1,b,base)
111 #define SAVE_4VRS(n,b,base)	SAVE_2VRS(n,b,base); SAVE_2VRS(n+2,b,base)
112 #define SAVE_8VRS(n,b,base)	SAVE_4VRS(n,b,base); SAVE_4VRS(n+4,b,base)
113 #define SAVE_16VRS(n,b,base)	SAVE_8VRS(n,b,base); SAVE_8VRS(n+8,b,base)
114 #define SAVE_32VRS(n,b,base)	SAVE_16VRS(n,b,base); SAVE_16VRS(n+16,b,base)
115 #define REST_VR(n,b,base)	li b,16*(n); lvx n,base,b
116 #define REST_2VRS(n,b,base)	REST_VR(n,b,base); REST_VR(n+1,b,base)
117 #define REST_4VRS(n,b,base)	REST_2VRS(n,b,base); REST_2VRS(n+2,b,base)
118 #define REST_8VRS(n,b,base)	REST_4VRS(n,b,base); REST_4VRS(n+4,b,base)
119 #define REST_16VRS(n,b,base)	REST_8VRS(n,b,base); REST_8VRS(n+8,b,base)
120 #define REST_32VRS(n,b,base)	REST_16VRS(n,b,base); REST_16VRS(n+16,b,base)
121 
122 #ifdef __BIG_ENDIAN__
123 #define STXVD2X_ROT(n,b,base)		STXVD2X(n,b,base)
124 #define LXVD2X_ROT(n,b,base)		LXVD2X(n,b,base)
125 #else
126 #define STXVD2X_ROT(n,b,base)		XXSWAPD(n,n);		\
127 					STXVD2X(n,b,base);	\
128 					XXSWAPD(n,n)
129 
130 #define LXVD2X_ROT(n,b,base)		LXVD2X(n,b,base);	\
131 					XXSWAPD(n,n)
132 #endif
133 /* Save the lower 32 VSRs in the thread VSR region */
134 #define SAVE_VSR(n,b,base)	li b,16*(n);  STXVD2X_ROT(n,R##base,R##b)
135 #define SAVE_2VSRS(n,b,base)	SAVE_VSR(n,b,base); SAVE_VSR(n+1,b,base)
136 #define SAVE_4VSRS(n,b,base)	SAVE_2VSRS(n,b,base); SAVE_2VSRS(n+2,b,base)
137 #define SAVE_8VSRS(n,b,base)	SAVE_4VSRS(n,b,base); SAVE_4VSRS(n+4,b,base)
138 #define SAVE_16VSRS(n,b,base)	SAVE_8VSRS(n,b,base); SAVE_8VSRS(n+8,b,base)
139 #define SAVE_32VSRS(n,b,base)	SAVE_16VSRS(n,b,base); SAVE_16VSRS(n+16,b,base)
140 #define REST_VSR(n,b,base)	li b,16*(n); LXVD2X_ROT(n,R##base,R##b)
141 #define REST_2VSRS(n,b,base)	REST_VSR(n,b,base); REST_VSR(n+1,b,base)
142 #define REST_4VSRS(n,b,base)	REST_2VSRS(n,b,base); REST_2VSRS(n+2,b,base)
143 #define REST_8VSRS(n,b,base)	REST_4VSRS(n,b,base); REST_4VSRS(n+4,b,base)
144 #define REST_16VSRS(n,b,base)	REST_8VSRS(n,b,base); REST_8VSRS(n+8,b,base)
145 #define REST_32VSRS(n,b,base)	REST_16VSRS(n,b,base); REST_16VSRS(n+16,b,base)
146 
147 /*
148  * b = base register for addressing, o = base offset from register of 1st EVR
149  * n = first EVR, s = scratch
150  */
151 #define SAVE_EVR(n,s,b,o)	evmergehi s,s,n; stw s,o+4*(n)(b)
152 #define SAVE_2EVRS(n,s,b,o)	SAVE_EVR(n,s,b,o); SAVE_EVR(n+1,s,b,o)
153 #define SAVE_4EVRS(n,s,b,o)	SAVE_2EVRS(n,s,b,o); SAVE_2EVRS(n+2,s,b,o)
154 #define SAVE_8EVRS(n,s,b,o)	SAVE_4EVRS(n,s,b,o); SAVE_4EVRS(n+4,s,b,o)
155 #define SAVE_16EVRS(n,s,b,o)	SAVE_8EVRS(n,s,b,o); SAVE_8EVRS(n+8,s,b,o)
156 #define SAVE_32EVRS(n,s,b,o)	SAVE_16EVRS(n,s,b,o); SAVE_16EVRS(n+16,s,b,o)
157 #define REST_EVR(n,s,b,o)	lwz s,o+4*(n)(b); evmergelo n,s,n
158 #define REST_2EVRS(n,s,b,o)	REST_EVR(n,s,b,o); REST_EVR(n+1,s,b,o)
159 #define REST_4EVRS(n,s,b,o)	REST_2EVRS(n,s,b,o); REST_2EVRS(n+2,s,b,o)
160 #define REST_8EVRS(n,s,b,o)	REST_4EVRS(n,s,b,o); REST_4EVRS(n+4,s,b,o)
161 #define REST_16EVRS(n,s,b,o)	REST_8EVRS(n,s,b,o); REST_8EVRS(n+8,s,b,o)
162 #define REST_32EVRS(n,s,b,o)	REST_16EVRS(n,s,b,o); REST_16EVRS(n+16,s,b,o)
163 
164 /* Macros to adjust thread priority for hardware multithreading */
165 #define HMT_VERY_LOW	or	31,31,31	# very low priority
166 #define HMT_LOW		or	1,1,1
167 #define HMT_MEDIUM_LOW  or	6,6,6		# medium low priority
168 #define HMT_MEDIUM	or	2,2,2
169 #define HMT_MEDIUM_HIGH or	5,5,5		# medium high priority
170 #define HMT_HIGH	or	3,3,3
171 #define HMT_EXTRA_HIGH	or	7,7,7		# power7 only
172 
173 #ifdef CONFIG_PPC64
174 #define ULONG_SIZE 	8
175 #else
176 #define ULONG_SIZE	4
177 #endif
178 #define __VCPU_GPR(n)	(VCPU_GPRS + (n * ULONG_SIZE))
179 #define VCPU_GPR(n)	__VCPU_GPR(__REG_##n)
180 
181 #ifdef __KERNEL__
182 
183 /*
184  * Used to name C functions called from asm
185  */
186 #ifdef CONFIG_PPC_KERNEL_PCREL
187 #define CFUNC(name) name@notoc
188 #else
189 #define CFUNC(name) name
190 #endif
191 
192 /*
193  * We use __powerpc64__ here because we want the compat VDSO to use the 32-bit
194  * version below in the else case of the ifdef.
195  */
196 #ifdef __powerpc64__
197 
198 #define STACKFRAMESIZE 256
199 #define __STK_REG(i)   (112 + ((i)-14)*8)
200 #define STK_REG(i)     __STK_REG(__REG_##i)
201 
202 #ifdef CONFIG_PPC64_ELF_ABI_V2
203 #define STK_GOT		24
204 #define STK_PARAM_AREA	32
205 #else
206 #define STK_GOT		40
207 #define STK_PARAM_AREA	48
208 #endif
209 
210 #define __STK_PARAM(i)	(STK_PARAM_AREA + ((i)-3)*8)
211 #define STK_PARAM(i)	__STK_PARAM(__REG_##i)
212 
213 #ifdef CONFIG_PPC64_ELF_ABI_V2
214 
215 #define _GLOBAL(name) \
216 	.align 2 ; \
217 	.type name,@function; \
218 	.globl name; \
219 name:
220 
221 #ifdef CONFIG_PPC_KERNEL_PCREL
222 #define _GLOBAL_TOC _GLOBAL
223 #else
224 #define _GLOBAL_TOC(name) \
225 	.align 2 ; \
226 	.type name,@function; \
227 	.globl name; \
228 name: \
229 0:	addis r2,r12,(.TOC.-0b)@ha; \
230 	addi r2,r2,(.TOC.-0b)@l; \
231 	.localentry name,.-name
232 #endif
233 
234 #define DOTSYM(a)	a
235 
236 #else
237 
238 #define XGLUE(a,b) a##b
239 #define GLUE(a,b) XGLUE(a,b)
240 
241 #define _GLOBAL(name) \
242 	.align 2 ; \
243 	.globl name; \
244 	.globl GLUE(.,name); \
245 	.pushsection ".opd","aw"; \
246 name: \
247 	.quad GLUE(.,name); \
248 	.quad .TOC.@tocbase; \
249 	.quad 0; \
250 	.popsection; \
251 	.type GLUE(.,name),@function; \
252 GLUE(.,name):
253 
254 #define _GLOBAL_TOC(name) _GLOBAL(name)
255 
256 #define DOTSYM(a)	GLUE(.,a)
257 
258 #endif
259 
260 #else /* 32-bit */
261 
262 #define _GLOBAL(n)	\
263 	.globl n;	\
264 n:
265 
266 #define _GLOBAL_TOC(name) _GLOBAL(name)
267 
268 #define DOTSYM(a)	a
269 
270 #endif
271 
272 /*
273  * __kprobes (the C annotation) puts the symbol into the .kprobes.text
274  * section, which gets emitted at the end of regular text.
275  *
276  * _ASM_NOKPROBE_SYMBOL and NOKPROBE_SYMBOL just adds the symbol to
277  * a blacklist. The former is for core kprobe functions/data, the
278  * latter is for those that incdentially must be excluded from probing
279  * and allows them to be linked at more optimal location within text.
280  */
281 #ifdef CONFIG_KPROBES
282 #define _ASM_NOKPROBE_SYMBOL(entry)			\
283 	.pushsection "_kprobe_blacklist","aw";		\
284 	PPC_LONG (entry) ;				\
285 	.popsection
286 #else
287 #define _ASM_NOKPROBE_SYMBOL(entry)
288 #endif
289 
290 #define FUNC_START(name)	_GLOBAL(name)
291 #define FUNC_END(name)
292 
293 /*
294  * LOAD_REG_IMMEDIATE(rn, expr)
295  *   Loads the value of the constant expression 'expr' into register 'rn'
296  *   using immediate instructions only.  Use this when it's important not
297  *   to reference other data (i.e. on ppc64 when the TOC pointer is not
298  *   valid) and when 'expr' is a constant or absolute address.
299  *
300  * LOAD_REG_ADDR(rn, name)
301  *   Loads the address of label 'name' into register 'rn'.  Use this when
302  *   you don't particularly need immediate instructions only, but you need
303  *   the whole address in one register (e.g. it's a structure address and
304  *   you want to access various offsets within it).  On ppc32 this is
305  *   identical to LOAD_REG_IMMEDIATE.
306  *
307  * LOAD_REG_ADDR_PIC(rn, name)
308  *   Loads the address of label 'name' into register 'run'. Use this when
309  *   the kernel doesn't run at the linked or relocated address. Please
310  *   note that this macro will clobber the lr register.
311  *
312  * LOAD_REG_ADDRBASE(rn, name)
313  * ADDROFF(name)
314  *   LOAD_REG_ADDRBASE loads part of the address of label 'name' into
315  *   register 'rn'.  ADDROFF(name) returns the remainder of the address as
316  *   a constant expression.  ADDROFF(name) is a signed expression < 16 bits
317  *   in size, so is suitable for use directly as an offset in load and store
318  *   instructions.  Use this when loading/storing a single word or less as:
319  *      LOAD_REG_ADDRBASE(rX, name)
320  *      ld	rY,ADDROFF(name)(rX)
321  */
322 
323 /* Be careful, this will clobber the lr register. */
324 #define LOAD_REG_ADDR_PIC(reg, name)		\
325 	bcl	20,31,$+4;			\
326 0:	mflr	reg;				\
327 	addis	reg,reg,(name - 0b)@ha;		\
328 	addi	reg,reg,(name - 0b)@l;
329 
330 #if defined(__powerpc64__) && defined(HAVE_AS_ATHIGH)
331 #define __AS_ATHIGH high
332 #else
333 #define __AS_ATHIGH h
334 #endif
335 
336 .macro __LOAD_REG_IMMEDIATE_32 r, x
337 	.if (\x) >= 0x8000 || (\x) < -0x8000
338 		lis \r, (\x)@__AS_ATHIGH
339 		.if (\x) & 0xffff != 0
340 			ori \r, \r, (\x)@l
341 		.endif
342 	.else
343 		li \r, (\x)@l
344 	.endif
345 .endm
346 
347 .macro __LOAD_REG_IMMEDIATE r, x
348 	.if (\x) >= 0x80000000 || (\x) < -0x80000000
349 		__LOAD_REG_IMMEDIATE_32 \r, (\x) >> 32
350 		sldi	\r, \r, 32
351 		.if (\x) & 0xffff0000 != 0
352 			oris \r, \r, (\x)@__AS_ATHIGH
353 		.endif
354 		.if (\x) & 0xffff != 0
355 			ori \r, \r, (\x)@l
356 		.endif
357 	.else
358 		__LOAD_REG_IMMEDIATE_32 \r, \x
359 	.endif
360 .endm
361 
362 #ifdef __powerpc64__
363 
364 #ifdef CONFIG_PPC_KERNEL_PCREL
365 #define __LOAD_PACA_TOC(reg)			\
366 	li	reg,-1
367 #else
368 #define __LOAD_PACA_TOC(reg)			\
369 	ld	reg,PACATOC(r13)
370 #endif
371 
372 #define LOAD_PACA_TOC()				\
373 	__LOAD_PACA_TOC(r2)
374 
375 #define LOAD_REG_IMMEDIATE(reg, expr) __LOAD_REG_IMMEDIATE reg, expr
376 
377 #define LOAD_REG_IMMEDIATE_SYM(reg, tmp, expr)	\
378 	lis	tmp, (expr)@highest;		\
379 	lis	reg, (expr)@__AS_ATHIGH;	\
380 	ori	tmp, tmp, (expr)@higher;	\
381 	ori	reg, reg, (expr)@l;		\
382 	rldimi	reg, tmp, 32, 0
383 
384 #ifdef CONFIG_PPC_KERNEL_PCREL
385 #define LOAD_REG_ADDR(reg,name)			\
386 	pla	reg,name@pcrel
387 
388 #else
389 #define LOAD_REG_ADDR(reg,name)			\
390 	addis	reg,r2,name@toc@ha;		\
391 	addi	reg,reg,name@toc@l
392 #endif
393 
394 #ifdef CONFIG_PPC_BOOK3E_64
395 /*
396  * This is used in register-constrained interrupt handlers. Not to be used
397  * by BOOK3S. ld complains with "got/toc optimization is not supported" if r2
398  * is not used for the TOC offset, so use @got(tocreg). If the interrupt
399  * handlers saved r2 instead, LOAD_REG_ADDR could be used.
400  */
401 #define LOAD_REG_ADDR_ALTTOC(reg,tocreg,name)	\
402 	ld	reg,name@got(tocreg)
403 #endif
404 
405 #define LOAD_REG_ADDRBASE(reg,name)	LOAD_REG_ADDR(reg,name)
406 #define ADDROFF(name)			0
407 
408 /* offsets for stack frame layout */
409 #define LRSAVE	16
410 
411 /*
412  * GCC stack frames follow a different pattern on 32 vs 64. This can be used
413  * to make asm frames be consistent with C.
414  */
415 #define PPC_CREATE_STACK_FRAME(size)			\
416 	mflr		r0;				\
417 	std		r0,16(r1);			\
418 	stdu		r1,-(size)(r1)
419 
420 #else /* 32-bit */
421 
422 #define LOAD_REG_IMMEDIATE(reg, expr) __LOAD_REG_IMMEDIATE_32 reg, expr
423 
424 #define LOAD_REG_IMMEDIATE_SYM(reg,expr)		\
425 	lis	reg,(expr)@ha;		\
426 	addi	reg,reg,(expr)@l;
427 
428 #define LOAD_REG_ADDR(reg,name)		LOAD_REG_IMMEDIATE_SYM(reg, name)
429 
430 #define LOAD_REG_ADDRBASE(reg, name)	lis	reg,name@ha
431 #define ADDROFF(name)			name@l
432 
433 /* offsets for stack frame layout */
434 #define LRSAVE	4
435 
436 #define PPC_CREATE_STACK_FRAME(size)			\
437 	stwu		r1,-(size)(r1);			\
438 	mflr		r0;				\
439 	stw		r0,(size+4)(r1)
440 
441 #endif
442 
443 /* various errata or part fixups */
444 #if defined(CONFIG_PPC_CELL) || defined(CONFIG_PPC_E500)
445 #define MFTB(dest)			\
446 90:	mfspr dest, SPRN_TBRL;		\
447 BEGIN_FTR_SECTION_NESTED(96);		\
448 	cmpwi dest,0;			\
449 	beq-  90b;			\
450 END_FTR_SECTION_NESTED(CPU_FTR_CELL_TB_BUG, CPU_FTR_CELL_TB_BUG, 96)
451 #else
452 #define MFTB(dest)			MFTBL(dest)
453 #endif
454 
455 #ifdef CONFIG_PPC_8xx
456 #define MFTBL(dest)			mftb dest
457 #define MFTBU(dest)			mftbu dest
458 #else
459 #define MFTBL(dest)			mfspr dest, SPRN_TBRL
460 #define MFTBU(dest)			mfspr dest, SPRN_TBRU
461 #endif
462 
463 #ifndef CONFIG_SMP
464 #define TLBSYNC
465 #else
466 #define TLBSYNC		tlbsync; sync
467 #endif
468 
469 #ifdef CONFIG_PPC64
470 #define MTOCRF(FXM, RS)			\
471 	BEGIN_FTR_SECTION_NESTED(848);	\
472 	mtcrf	(FXM), RS;		\
473 	FTR_SECTION_ELSE_NESTED(848);	\
474 	mtocrf (FXM), RS;		\
475 	ALT_FTR_SECTION_END_NESTED_IFCLR(CPU_FTR_NOEXECUTE, 848)
476 #endif
477 
478 /*
479  * This instruction is not implemented on the PPC 603 or 601; however, on
480  * the 403GCX and 405GP tlbia IS defined and tlbie is not.
481  * All of these instructions exist in the 8xx, they have magical powers,
482  * and they must be used.
483  */
484 
485 #if !defined(CONFIG_4xx) && !defined(CONFIG_PPC_8xx)
486 #define tlbia					\
487 	li	r4,1024;			\
488 	mtctr	r4;				\
489 	lis	r4,KERNELBASE@h;		\
490 	.machine push;				\
491 	.machine "power4";			\
492 0:	tlbie	r4;				\
493 	.machine pop;				\
494 	addi	r4,r4,0x1000;			\
495 	bdnz	0b
496 #endif
497 
498 
499 #ifdef CONFIG_IBM440EP_ERR42
500 #define PPC440EP_ERR42 isync
501 #else
502 #define PPC440EP_ERR42
503 #endif
504 
505 /* The following stops all load and store data streams associated with stream
506  * ID (ie. streams created explicitly).  The embedded and server mnemonics for
507  * dcbt are different so this must only be used for server.
508  */
509 #define DCBT_BOOK3S_STOP_ALL_STREAM_IDS(scratch)	\
510        lis     scratch,0x60000000@h;			\
511        .machine push;					\
512        .machine power4;					\
513        dcbt    0,scratch,0b01010;			\
514        .machine pop;
515 
516 #define DCBT_SETUP_STREAMS(from, from_parms, to, to_parms, scratch)	\
517 	lis	scratch,0x8000;	/* GO=1 */				\
518 	clrldi	scratch,scratch,32;					\
519 	.machine push;							\
520 	.machine power4;						\
521 	/* setup read stream 0 */					\
522 	dcbt	0,from,0b01000;		/* addr from */			\
523 	dcbt	0,from_parms,0b01010;	/* length and depth from */	\
524 	/* setup write stream 1 */					\
525 	dcbtst	0,to,0b01000;		/* addr to */			\
526 	dcbtst	0,to_parms,0b01010;	/* length and depth to */	\
527 	eieio;								\
528 	dcbt	0,scratch,0b01010;	/* all streams GO */		\
529 	.machine pop;
530 
531 /*
532  * toreal/fromreal/tophys/tovirt macros. 32-bit BookE makes them
533  * keep the address intact to be compatible with code shared with
534  * 32-bit classic.
535  *
536  * On the other hand, I find it useful to have them behave as expected
537  * by their name (ie always do the addition) on 64-bit BookE
538  */
539 #if defined(CONFIG_BOOKE) && !defined(CONFIG_PPC64)
540 #define toreal(rd)
541 #define fromreal(rd)
542 
543 /*
544  * We use addis to ensure compatibility with the "classic" ppc versions of
545  * these macros, which use rs = 0 to get the tophys offset in rd, rather than
546  * converting the address in r0, and so this version has to do that too
547  * (i.e. set register rd to 0 when rs == 0).
548  */
549 #define tophys(rd,rs)				\
550 	addis	rd,rs,0
551 
552 #define tovirt(rd,rs)				\
553 	addis	rd,rs,0
554 
555 #elif defined(CONFIG_PPC64)
556 #define toreal(rd)		/* we can access c000... in real mode */
557 #define fromreal(rd)
558 
559 #define tophys(rd,rs)                           \
560 	clrldi	rd,rs,2
561 
562 #define tovirt(rd,rs)                           \
563 	rotldi	rd,rs,16;			\
564 	ori	rd,rd,((KERNELBASE>>48)&0xFFFF);\
565 	rotldi	rd,rd,48
566 #else
567 #define toreal(rd)	tophys(rd,rd)
568 #define fromreal(rd)	tovirt(rd,rd)
569 
570 #define tophys(rd, rs)	addis	rd, rs, -PAGE_OFFSET@h
571 #define tovirt(rd, rs)	addis	rd, rs, PAGE_OFFSET@h
572 #endif
573 
574 #ifdef CONFIG_PPC_BOOK3S_64
575 #define MTMSRD(r)	mtmsrd	r
576 #define MTMSR_EERI(reg)	mtmsrd	reg,1
577 #else
578 #define MTMSRD(r)	mtmsr	r
579 #define MTMSR_EERI(reg)	mtmsr	reg
580 #endif
581 
582 #endif /* __KERNEL__ */
583 
584 /* The boring bits... */
585 
586 /* Condition Register Bit Fields */
587 
588 #define	cr0	0
589 #define	cr1	1
590 #define	cr2	2
591 #define	cr3	3
592 #define	cr4	4
593 #define	cr5	5
594 #define	cr6	6
595 #define	cr7	7
596 
597 
598 /*
599  * General Purpose Registers (GPRs)
600  *
601  * The lower case r0-r31 should be used in preference to the upper
602  * case R0-R31 as they provide more error checking in the assembler.
603  * Use R0-31 only when really nessesary.
604  */
605 
606 #define	r0	%r0
607 #define	r1	%r1
608 #define	r2	%r2
609 #define	r3	%r3
610 #define	r4	%r4
611 #define	r5	%r5
612 #define	r6	%r6
613 #define	r7	%r7
614 #define	r8	%r8
615 #define	r9	%r9
616 #define	r10	%r10
617 #define	r11	%r11
618 #define	r12	%r12
619 #define	r13	%r13
620 #define	r14	%r14
621 #define	r15	%r15
622 #define	r16	%r16
623 #define	r17	%r17
624 #define	r18	%r18
625 #define	r19	%r19
626 #define	r20	%r20
627 #define	r21	%r21
628 #define	r22	%r22
629 #define	r23	%r23
630 #define	r24	%r24
631 #define	r25	%r25
632 #define	r26	%r26
633 #define	r27	%r27
634 #define	r28	%r28
635 #define	r29	%r29
636 #define	r30	%r30
637 #define	r31	%r31
638 
639 
640 /* Floating Point Registers (FPRs) */
641 
642 #define	fr0	0
643 #define	fr1	1
644 #define	fr2	2
645 #define	fr3	3
646 #define	fr4	4
647 #define	fr5	5
648 #define	fr6	6
649 #define	fr7	7
650 #define	fr8	8
651 #define	fr9	9
652 #define	fr10	10
653 #define	fr11	11
654 #define	fr12	12
655 #define	fr13	13
656 #define	fr14	14
657 #define	fr15	15
658 #define	fr16	16
659 #define	fr17	17
660 #define	fr18	18
661 #define	fr19	19
662 #define	fr20	20
663 #define	fr21	21
664 #define	fr22	22
665 #define	fr23	23
666 #define	fr24	24
667 #define	fr25	25
668 #define	fr26	26
669 #define	fr27	27
670 #define	fr28	28
671 #define	fr29	29
672 #define	fr30	30
673 #define	fr31	31
674 
675 /* AltiVec Registers (VPRs) */
676 
677 #define	v0	0
678 #define	v1	1
679 #define	v2	2
680 #define	v3	3
681 #define	v4	4
682 #define	v5	5
683 #define	v6	6
684 #define	v7	7
685 #define	v8	8
686 #define	v9	9
687 #define	v10	10
688 #define	v11	11
689 #define	v12	12
690 #define	v13	13
691 #define	v14	14
692 #define	v15	15
693 #define	v16	16
694 #define	v17	17
695 #define	v18	18
696 #define	v19	19
697 #define	v20	20
698 #define	v21	21
699 #define	v22	22
700 #define	v23	23
701 #define	v24	24
702 #define	v25	25
703 #define	v26	26
704 #define	v27	27
705 #define	v28	28
706 #define	v29	29
707 #define	v30	30
708 #define	v31	31
709 
710 /* VSX Registers (VSRs) */
711 
712 #define	vs0	0
713 #define	vs1	1
714 #define	vs2	2
715 #define	vs3	3
716 #define	vs4	4
717 #define	vs5	5
718 #define	vs6	6
719 #define	vs7	7
720 #define	vs8	8
721 #define	vs9	9
722 #define	vs10	10
723 #define	vs11	11
724 #define	vs12	12
725 #define	vs13	13
726 #define	vs14	14
727 #define	vs15	15
728 #define	vs16	16
729 #define	vs17	17
730 #define	vs18	18
731 #define	vs19	19
732 #define	vs20	20
733 #define	vs21	21
734 #define	vs22	22
735 #define	vs23	23
736 #define	vs24	24
737 #define	vs25	25
738 #define	vs26	26
739 #define	vs27	27
740 #define	vs28	28
741 #define	vs29	29
742 #define	vs30	30
743 #define	vs31	31
744 #define	vs32	32
745 #define	vs33	33
746 #define	vs34	34
747 #define	vs35	35
748 #define	vs36	36
749 #define	vs37	37
750 #define	vs38	38
751 #define	vs39	39
752 #define	vs40	40
753 #define	vs41	41
754 #define	vs42	42
755 #define	vs43	43
756 #define	vs44	44
757 #define	vs45	45
758 #define	vs46	46
759 #define	vs47	47
760 #define	vs48	48
761 #define	vs49	49
762 #define	vs50	50
763 #define	vs51	51
764 #define	vs52	52
765 #define	vs53	53
766 #define	vs54	54
767 #define	vs55	55
768 #define	vs56	56
769 #define	vs57	57
770 #define	vs58	58
771 #define	vs59	59
772 #define	vs60	60
773 #define	vs61	61
774 #define	vs62	62
775 #define	vs63	63
776 
777 /* SPE Registers (EVPRs) */
778 
779 #define	evr0	0
780 #define	evr1	1
781 #define	evr2	2
782 #define	evr3	3
783 #define	evr4	4
784 #define	evr5	5
785 #define	evr6	6
786 #define	evr7	7
787 #define	evr8	8
788 #define	evr9	9
789 #define	evr10	10
790 #define	evr11	11
791 #define	evr12	12
792 #define	evr13	13
793 #define	evr14	14
794 #define	evr15	15
795 #define	evr16	16
796 #define	evr17	17
797 #define	evr18	18
798 #define	evr19	19
799 #define	evr20	20
800 #define	evr21	21
801 #define	evr22	22
802 #define	evr23	23
803 #define	evr24	24
804 #define	evr25	25
805 #define	evr26	26
806 #define	evr27	27
807 #define	evr28	28
808 #define	evr29	29
809 #define	evr30	30
810 #define	evr31	31
811 
812 #define RFSCV	.long 0x4c0000a4
813 
814 /*
815  * Create an endian fixup trampoline
816  *
817  * This starts with a "tdi 0,0,0x48" instruction which is
818  * essentially a "trap never", and thus akin to a nop.
819  *
820  * The opcode for this instruction read with the wrong endian
821  * however results in a b . + 8
822  *
823  * So essentially we use that trick to execute the following
824  * trampoline in "reverse endian" if we are running with the
825  * MSR_LE bit set the "wrong" way for whatever endianness the
826  * kernel is built for.
827  */
828 
829 #ifdef CONFIG_PPC_BOOK3E_64
830 #define FIXUP_ENDIAN
831 #else
832 /*
833  * This version may be used in HV or non-HV context.
834  * MSR[EE] must be disabled.
835  */
836 #define FIXUP_ENDIAN						   \
837 	tdi   0,0,0x48;	  /* Reverse endian of b . + 8		*/ \
838 	b     191f;	  /* Skip trampoline if endian is good	*/ \
839 	.long 0xa600607d; /* mfmsr r11				*/ \
840 	.long 0x01006b69; /* xori r11,r11,1			*/ \
841 	.long 0x00004039; /* li r10,0				*/ \
842 	.long 0x6401417d; /* mtmsrd r10,1			*/ \
843 	.long 0x05009f42; /* bcl 20,31,$+4			*/ \
844 	.long 0xa602487d; /* mflr r10				*/ \
845 	.long 0x14004a39; /* addi r10,r10,20			*/ \
846 	.long 0xa6035a7d; /* mtsrr0 r10				*/ \
847 	.long 0xa6037b7d; /* mtsrr1 r11				*/ \
848 	.long 0x2400004c; /* rfid				*/ \
849 191:
850 
851 /*
852  * This version that may only be used with MSR[HV]=1
853  * - Does not clear MSR[RI], so more robust.
854  * - Slightly smaller and faster.
855  */
856 #define FIXUP_ENDIAN_HV						   \
857 	tdi   0,0,0x48;	  /* Reverse endian of b . + 8		*/ \
858 	b     191f;	  /* Skip trampoline if endian is good	*/ \
859 	.long 0xa600607d; /* mfmsr r11				*/ \
860 	.long 0x01006b69; /* xori r11,r11,1			*/ \
861 	.long 0x05009f42; /* bcl 20,31,$+4			*/ \
862 	.long 0xa602487d; /* mflr r10				*/ \
863 	.long 0x14004a39; /* addi r10,r10,20			*/ \
864 	.long 0xa64b5a7d; /* mthsrr0 r10			*/ \
865 	.long 0xa64b7b7d; /* mthsrr1 r11			*/ \
866 	.long 0x2402004c; /* hrfid				*/ \
867 191:
868 
869 #endif /* !CONFIG_PPC_BOOK3E_64 */
870 
871 #endif /*  __ASSEMBLY__ */
872 
873 #define SOFT_MASK_TABLE(_start, _end)		\
874 	stringify_in_c(.section __soft_mask_table,"a";)\
875 	stringify_in_c(.balign 8;)		\
876 	stringify_in_c(.llong (_start);)	\
877 	stringify_in_c(.llong (_end);)		\
878 	stringify_in_c(.previous)
879 
880 #define RESTART_TABLE(_start, _end, _target)	\
881 	stringify_in_c(.section __restart_table,"a";)\
882 	stringify_in_c(.balign 8;)		\
883 	stringify_in_c(.llong (_start);)	\
884 	stringify_in_c(.llong (_end);)		\
885 	stringify_in_c(.llong (_target);)	\
886 	stringify_in_c(.previous)
887 
888 #ifdef CONFIG_PPC_E500
889 #define BTB_FLUSH(reg)			\
890 	lis reg,BUCSR_INIT@h;		\
891 	ori reg,reg,BUCSR_INIT@l;	\
892 	mtspr SPRN_BUCSR,reg;		\
893 	isync;
894 #else
895 #define BTB_FLUSH(reg)
896 #endif /* CONFIG_PPC_E500 */
897 
898 #if defined(CONFIG_PPC64_ELF_ABI_V1)
899 #define STACK_FRAME_PARAMS 48
900 #elif defined(CONFIG_PPC64_ELF_ABI_V2)
901 #define STACK_FRAME_PARAMS 32
902 #elif defined(CONFIG_PPC32)
903 #define STACK_FRAME_PARAMS 8
904 #endif
905 
906 #endif /* _ASM_POWERPC_PPC_ASM_H */
907