Commit | Line | Data |
---|---|---|
4c921d4d JS |
1 | /* |
2 | * Copyright (c) 2008 Silicon Graphics, Inc. All Rights Reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU Lesser General Public License as published by | |
6 | * the Free Software Foundation; either version 2.1 of the License, or | |
7 | * (at your option) any later version. | |
8 | * | |
9 | * This program is distributed in the hope that it will be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU Lesser General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU Lesser General Public License | |
15 | * along with this program; if not, write to the Free Software | |
16 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
17 | */ | |
18 | ||
19 | #ifndef __GRU_INSTRUCTIONS_H__ | |
20 | #define __GRU_INSTRUCTIONS_H__ | |
21 | ||
fe5bb6b0 JS |
22 | extern int gru_check_status_proc(void *cb); |
23 | extern int gru_wait_proc(void *cb); | |
24 | extern void gru_wait_abort_proc(void *cb); | |
25 | ||
26 | ||
4c921d4d JS |
27 | |
28 | /* | |
29 | * Architecture dependent functions | |
30 | */ | |
31 | ||
923f7f69 | 32 | #if defined(CONFIG_IA64) |
4c921d4d JS |
33 | #include <linux/compiler.h> |
34 | #include <asm/intrinsics.h> | |
fe5bb6b0 | 35 | #define __flush_cache(p) ia64_fc((unsigned long)p) |
4c921d4d | 36 | /* Use volatile on IA64 to ensure ordering via st4.rel */ |
57ebb034 | 37 | #define gru_ordered_store_ulong(p, v) \ |
4c921d4d JS |
38 | do { \ |
39 | barrier(); \ | |
57ebb034 | 40 | *((volatile unsigned long *)(p)) = v; /* force st.rel */ \ |
4c921d4d | 41 | } while (0) |
923f7f69 | 42 | #elif defined(CONFIG_X86_64) |
4c921d4d | 43 | #define __flush_cache(p) clflush(p) |
57ebb034 | 44 | #define gru_ordered_store_ulong(p, v) \ |
4c921d4d JS |
45 | do { \ |
46 | barrier(); \ | |
57ebb034 | 47 | *(unsigned long *)p = v; \ |
4c921d4d JS |
48 | } while (0) |
49 | #else | |
50 | #error "Unsupported architecture" | |
51 | #endif | |
52 | ||
53 | /* | |
54 | * Control block status and exception codes | |
55 | */ | |
56 | #define CBS_IDLE 0 | |
57 | #define CBS_EXCEPTION 1 | |
58 | #define CBS_ACTIVE 2 | |
59 | #define CBS_CALL_OS 3 | |
60 | ||
61 | /* CB substatus bitmasks */ | |
62 | #define CBSS_MSG_QUEUE_MASK 7 | |
63 | #define CBSS_IMPLICIT_ABORT_ACTIVE_MASK 8 | |
64 | ||
65 | /* CB substatus message queue values (low 3 bits of substatus) */ | |
66 | #define CBSS_NO_ERROR 0 | |
67 | #define CBSS_LB_OVERFLOWED 1 | |
68 | #define CBSS_QLIMIT_REACHED 2 | |
69 | #define CBSS_PAGE_OVERFLOW 3 | |
70 | #define CBSS_AMO_NACKED 4 | |
71 | #define CBSS_PUT_NACKED 5 | |
72 | ||
73 | /* | |
74 | * Structure used to fetch exception detail for CBs that terminate with | |
75 | * CBS_EXCEPTION | |
76 | */ | |
77 | struct control_block_extended_exc_detail { | |
78 | unsigned long cb; | |
79 | int opc; | |
80 | int ecause; | |
81 | int exopc; | |
82 | long exceptdet0; | |
83 | int exceptdet1; | |
cd1334f0 JS |
84 | int cbrstate; |
85 | int cbrexecstatus; | |
4c921d4d JS |
86 | }; |
87 | ||
88 | /* | |
89 | * Instruction formats | |
90 | */ | |
91 | ||
92 | /* | |
93 | * Generic instruction format. | |
94 | * This definition has precise bit field definitions. | |
95 | */ | |
96 | struct gru_instruction_bits { | |
97 | /* DW 0 - low */ | |
98 | unsigned int icmd: 1; | |
99 | unsigned char ima: 3; /* CB_DelRep, unmapped mode */ | |
100 | unsigned char reserved0: 4; | |
101 | unsigned int xtype: 3; | |
102 | unsigned int iaa0: 2; | |
103 | unsigned int iaa1: 2; | |
104 | unsigned char reserved1: 1; | |
105 | unsigned char opc: 8; /* opcode */ | |
106 | unsigned char exopc: 8; /* extended opcode */ | |
107 | /* DW 0 - high */ | |
108 | unsigned int idef2: 22; /* TRi0 */ | |
109 | unsigned char reserved2: 2; | |
110 | unsigned char istatus: 2; | |
111 | unsigned char isubstatus:4; | |
cd1334f0 JS |
112 | unsigned char reserved3: 1; |
113 | unsigned char tlb_fault_color: 1; | |
4c921d4d JS |
114 | /* DW 1 */ |
115 | unsigned long idef4; /* 42 bits: TRi1, BufSize */ | |
116 | /* DW 2-6 */ | |
117 | unsigned long idef1; /* BAddr0 */ | |
118 | unsigned long idef5; /* Nelem */ | |
119 | unsigned long idef6; /* Stride, Operand1 */ | |
120 | unsigned long idef3; /* BAddr1, Value, Operand2 */ | |
121 | unsigned long reserved4; | |
122 | /* DW 7 */ | |
123 | unsigned long avalue; /* AValue */ | |
124 | }; | |
125 | ||
126 | /* | |
127 | * Generic instruction with friendlier names. This format is used | |
128 | * for inline instructions. | |
129 | */ | |
130 | struct gru_instruction { | |
131 | /* DW 0 */ | |
57ebb034 JS |
132 | union { |
133 | unsigned long op64; /* icmd,xtype,iaa0,ima,opc,tri0 */ | |
134 | struct { | |
135 | unsigned int op32; | |
136 | unsigned int tri0; | |
137 | }; | |
138 | }; | |
4c921d4d JS |
139 | unsigned long tri1_bufsize; /* DW 1 */ |
140 | unsigned long baddr0; /* DW 2 */ | |
141 | unsigned long nelem; /* DW 3 */ | |
142 | unsigned long op1_stride; /* DW 4 */ | |
143 | unsigned long op2_value_baddr1; /* DW 5 */ | |
144 | unsigned long reserved0; /* DW 6 */ | |
145 | unsigned long avalue; /* DW 7 */ | |
146 | }; | |
147 | ||
57ebb034 | 148 | /* Some shifts and masks for the low 64 bits of a GRU command */ |
4c921d4d JS |
149 | #define GRU_CB_ICMD_SHFT 0 |
150 | #define GRU_CB_ICMD_MASK 0x1 | |
151 | #define GRU_CB_XTYPE_SHFT 8 | |
152 | #define GRU_CB_XTYPE_MASK 0x7 | |
153 | #define GRU_CB_IAA0_SHFT 11 | |
154 | #define GRU_CB_IAA0_MASK 0x3 | |
155 | #define GRU_CB_IAA1_SHFT 13 | |
156 | #define GRU_CB_IAA1_MASK 0x3 | |
157 | #define GRU_CB_IMA_SHFT 1 | |
158 | #define GRU_CB_IMA_MASK 0x3 | |
159 | #define GRU_CB_OPC_SHFT 16 | |
160 | #define GRU_CB_OPC_MASK 0xff | |
161 | #define GRU_CB_EXOPC_SHFT 24 | |
162 | #define GRU_CB_EXOPC_MASK 0xff | |
57ebb034 JS |
163 | #define GRU_IDEF2_SHFT 32 |
164 | #define GRU_IDEF2_MASK 0x3ffff | |
165 | #define GRU_ISTATUS_SHFT 56 | |
166 | #define GRU_ISTATUS_MASK 0x3 | |
4c921d4d JS |
167 | |
168 | /* GRU instruction opcodes (opc field) */ | |
169 | #define OP_NOP 0x00 | |
170 | #define OP_BCOPY 0x01 | |
171 | #define OP_VLOAD 0x02 | |
172 | #define OP_IVLOAD 0x03 | |
173 | #define OP_VSTORE 0x04 | |
174 | #define OP_IVSTORE 0x05 | |
175 | #define OP_VSET 0x06 | |
176 | #define OP_IVSET 0x07 | |
177 | #define OP_MESQ 0x08 | |
178 | #define OP_GAMXR 0x09 | |
179 | #define OP_GAMIR 0x0a | |
180 | #define OP_GAMIRR 0x0b | |
181 | #define OP_GAMER 0x0c | |
182 | #define OP_GAMERR 0x0d | |
183 | #define OP_BSTORE 0x0e | |
184 | #define OP_VFLUSH 0x0f | |
185 | ||
186 | ||
187 | /* Extended opcodes values (exopc field) */ | |
188 | ||
189 | /* GAMIR - AMOs with implicit operands */ | |
190 | #define EOP_IR_FETCH 0x01 /* Plain fetch of memory */ | |
191 | #define EOP_IR_CLR 0x02 /* Fetch and clear */ | |
192 | #define EOP_IR_INC 0x05 /* Fetch and increment */ | |
193 | #define EOP_IR_DEC 0x07 /* Fetch and decrement */ | |
194 | #define EOP_IR_QCHK1 0x0d /* Queue check, 64 byte msg */ | |
195 | #define EOP_IR_QCHK2 0x0e /* Queue check, 128 byte msg */ | |
196 | ||
197 | /* GAMIRR - Registered AMOs with implicit operands */ | |
198 | #define EOP_IRR_FETCH 0x01 /* Registered fetch of memory */ | |
199 | #define EOP_IRR_CLR 0x02 /* Registered fetch and clear */ | |
200 | #define EOP_IRR_INC 0x05 /* Registered fetch and increment */ | |
201 | #define EOP_IRR_DEC 0x07 /* Registered fetch and decrement */ | |
202 | #define EOP_IRR_DECZ 0x0f /* Registered fetch and decrement, update on zero*/ | |
203 | ||
204 | /* GAMER - AMOs with explicit operands */ | |
205 | #define EOP_ER_SWAP 0x00 /* Exchange argument and memory */ | |
206 | #define EOP_ER_OR 0x01 /* Logical OR with memory */ | |
207 | #define EOP_ER_AND 0x02 /* Logical AND with memory */ | |
208 | #define EOP_ER_XOR 0x03 /* Logical XOR with memory */ | |
209 | #define EOP_ER_ADD 0x04 /* Add value to memory */ | |
210 | #define EOP_ER_CSWAP 0x08 /* Compare with operand2, write operand1 if match*/ | |
211 | #define EOP_ER_CADD 0x0c /* Queue check, operand1*64 byte msg */ | |
212 | ||
213 | /* GAMERR - Registered AMOs with explicit operands */ | |
214 | #define EOP_ERR_SWAP 0x00 /* Exchange argument and memory */ | |
215 | #define EOP_ERR_OR 0x01 /* Logical OR with memory */ | |
216 | #define EOP_ERR_AND 0x02 /* Logical AND with memory */ | |
217 | #define EOP_ERR_XOR 0x03 /* Logical XOR with memory */ | |
218 | #define EOP_ERR_ADD 0x04 /* Add value to memory */ | |
219 | #define EOP_ERR_CSWAP 0x08 /* Compare with operand2, write operand1 if match*/ | |
220 | #define EOP_ERR_EPOLL 0x09 /* Poll for equality */ | |
221 | #define EOP_ERR_NPOLL 0x0a /* Poll for inequality */ | |
222 | ||
223 | /* GAMXR - SGI Arithmetic unit */ | |
224 | #define EOP_XR_CSWAP 0x0b /* Masked compare exchange */ | |
225 | ||
226 | ||
227 | /* Transfer types (xtype field) */ | |
228 | #define XTYPE_B 0x0 /* byte */ | |
229 | #define XTYPE_S 0x1 /* short (2-byte) */ | |
230 | #define XTYPE_W 0x2 /* word (4-byte) */ | |
231 | #define XTYPE_DW 0x3 /* doubleword (8-byte) */ | |
232 | #define XTYPE_CL 0x6 /* cacheline (64-byte) */ | |
233 | ||
234 | ||
235 | /* Instruction access attributes (iaa0, iaa1 fields) */ | |
236 | #define IAA_RAM 0x0 /* normal cached RAM access */ | |
237 | #define IAA_NCRAM 0x2 /* noncoherent RAM access */ | |
238 | #define IAA_MMIO 0x1 /* noncoherent memory-mapped I/O space */ | |
239 | #define IAA_REGISTER 0x3 /* memory-mapped registers, etc. */ | |
240 | ||
241 | ||
242 | /* Instruction mode attributes (ima field) */ | |
243 | #define IMA_MAPPED 0x0 /* Virtual mode */ | |
244 | #define IMA_CB_DELAY 0x1 /* hold read responses until status changes */ | |
245 | #define IMA_UNMAPPED 0x2 /* bypass the TLBs (OS only) */ | |
246 | #define IMA_INTERRUPT 0x4 /* Interrupt when instruction completes */ | |
247 | ||
248 | /* CBE ecause bits */ | |
249 | #define CBE_CAUSE_RI (1 << 0) | |
250 | #define CBE_CAUSE_INVALID_INSTRUCTION (1 << 1) | |
251 | #define CBE_CAUSE_UNMAPPED_MODE_FORBIDDEN (1 << 2) | |
252 | #define CBE_CAUSE_PE_CHECK_DATA_ERROR (1 << 3) | |
253 | #define CBE_CAUSE_IAA_GAA_MISMATCH (1 << 4) | |
254 | #define CBE_CAUSE_DATA_SEGMENT_LIMIT_EXCEPTION (1 << 5) | |
255 | #define CBE_CAUSE_OS_FATAL_TLB_FAULT (1 << 6) | |
256 | #define CBE_CAUSE_EXECUTION_HW_ERROR (1 << 7) | |
257 | #define CBE_CAUSE_TLBHW_ERROR (1 << 8) | |
258 | #define CBE_CAUSE_RA_REQUEST_TIMEOUT (1 << 9) | |
259 | #define CBE_CAUSE_HA_REQUEST_TIMEOUT (1 << 10) | |
260 | #define CBE_CAUSE_RA_RESPONSE_FATAL (1 << 11) | |
261 | #define CBE_CAUSE_RA_RESPONSE_NON_FATAL (1 << 12) | |
262 | #define CBE_CAUSE_HA_RESPONSE_FATAL (1 << 13) | |
263 | #define CBE_CAUSE_HA_RESPONSE_NON_FATAL (1 << 14) | |
264 | #define CBE_CAUSE_ADDRESS_SPACE_DECODE_ERROR (1 << 15) | |
270952a9 JS |
265 | #define CBE_CAUSE_PROTOCOL_STATE_DATA_ERROR (1 << 16) |
266 | #define CBE_CAUSE_RA_RESPONSE_DATA_ERROR (1 << 17) | |
267 | #define CBE_CAUSE_HA_RESPONSE_DATA_ERROR (1 << 18) | |
5658366a | 268 | #define CBE_CAUSE_FORCED_ERROR (1 << 19) |
4c921d4d | 269 | |
cd1334f0 JS |
270 | /* CBE cbrexecstatus bits */ |
271 | #define CBR_EXS_ABORT_OCC_BIT 0 | |
272 | #define CBR_EXS_INT_OCC_BIT 1 | |
273 | #define CBR_EXS_PENDING_BIT 2 | |
274 | #define CBR_EXS_QUEUED_BIT 3 | |
270952a9 | 275 | #define CBR_EXS_TLB_INVAL_BIT 4 |
cd1334f0 | 276 | #define CBR_EXS_EXCEPTION_BIT 5 |
5658366a | 277 | #define CBR_EXS_CB_INT_PENDING_BIT 6 |
cd1334f0 JS |
278 | |
279 | #define CBR_EXS_ABORT_OCC (1 << CBR_EXS_ABORT_OCC_BIT) | |
280 | #define CBR_EXS_INT_OCC (1 << CBR_EXS_INT_OCC_BIT) | |
281 | #define CBR_EXS_PENDING (1 << CBR_EXS_PENDING_BIT) | |
282 | #define CBR_EXS_QUEUED (1 << CBR_EXS_QUEUED_BIT) | |
5658366a | 283 | #define CBR_EXS_TLB_INVAL (1 << CBR_EXS_TLB_INVAL_BIT) |
cd1334f0 | 284 | #define CBR_EXS_EXCEPTION (1 << CBR_EXS_EXCEPTION_BIT) |
5658366a | 285 | #define CBR_EXS_CB_INT_PENDING (1 << CBR_EXS_CB_INT_PENDING_BIT) |
cd1334f0 | 286 | |
4c921d4d JS |
287 | /* |
288 | * Exceptions are retried for the following cases. If any OTHER bits are set | |
289 | * in ecause, the exception is not retryable. | |
290 | */ | |
270952a9 | 291 | #define EXCEPTION_RETRY_BITS (CBE_CAUSE_EXECUTION_HW_ERROR | \ |
4c921d4d | 292 | CBE_CAUSE_TLBHW_ERROR | \ |
270952a9 JS |
293 | CBE_CAUSE_RA_REQUEST_TIMEOUT | \ |
294 | CBE_CAUSE_RA_RESPONSE_NON_FATAL | \ | |
295 | CBE_CAUSE_HA_RESPONSE_NON_FATAL | \ | |
296 | CBE_CAUSE_RA_RESPONSE_DATA_ERROR | \ | |
297 | CBE_CAUSE_HA_RESPONSE_DATA_ERROR \ | |
298 | ) | |
4c921d4d JS |
299 | |
300 | /* Message queue head structure */ | |
301 | union gru_mesqhead { | |
302 | unsigned long val; | |
303 | struct { | |
304 | unsigned int head; | |
305 | unsigned int limit; | |
306 | }; | |
307 | }; | |
308 | ||
309 | ||
310 | /* Generate the low word of a GRU instruction */ | |
57ebb034 JS |
311 | static inline unsigned long |
312 | __opdword(unsigned char opcode, unsigned char exopc, unsigned char xtype, | |
4c921d4d | 313 | unsigned char iaa0, unsigned char iaa1, |
57ebb034 | 314 | unsigned long idef2, unsigned char ima) |
4c921d4d JS |
315 | { |
316 | return (1 << GRU_CB_ICMD_SHFT) | | |
57ebb034 JS |
317 | ((unsigned long)CBS_ACTIVE << GRU_ISTATUS_SHFT) | |
318 | (idef2<< GRU_IDEF2_SHFT) | | |
4c921d4d JS |
319 | (iaa0 << GRU_CB_IAA0_SHFT) | |
320 | (iaa1 << GRU_CB_IAA1_SHFT) | | |
321 | (ima << GRU_CB_IMA_SHFT) | | |
322 | (xtype << GRU_CB_XTYPE_SHFT) | | |
323 | (opcode << GRU_CB_OPC_SHFT) | | |
324 | (exopc << GRU_CB_EXOPC_SHFT); | |
325 | } | |
326 | ||
4c921d4d JS |
327 | /* |
328 | * Architecture specific intrinsics | |
329 | */ | |
330 | static inline void gru_flush_cache(void *p) | |
331 | { | |
332 | __flush_cache(p); | |
333 | } | |
334 | ||
335 | /* | |
57ebb034 | 336 | * Store the lower 64 bits of the command including the "start" bit. Then |
4c921d4d JS |
337 | * start the instruction executing. |
338 | */ | |
57ebb034 | 339 | static inline void gru_start_instruction(struct gru_instruction *ins, unsigned long op64) |
4c921d4d | 340 | { |
57ebb034 | 341 | gru_ordered_store_ulong(ins, op64); |
67bf04a5 | 342 | mb(); |
923f7f69 | 343 | gru_flush_cache(ins); |
4c921d4d JS |
344 | } |
345 | ||
346 | ||
347 | /* Convert "hints" to IMA */ | |
348 | #define CB_IMA(h) ((h) | IMA_UNMAPPED) | |
349 | ||
350 | /* Convert data segment cache line index into TRI0 / TRI1 value */ | |
351 | #define GRU_DINDEX(i) ((i) * GRU_CACHE_LINE_BYTES) | |
352 | ||
353 | /* Inline functions for GRU instructions. | |
354 | * Note: | |
355 | * - nelem and stride are in elements | |
356 | * - tri0/tri1 is in bytes for the beginning of the data segment. | |
357 | */ | |
289750d1 RH |
358 | static inline void gru_vload_phys(void *cb, unsigned long gpa, |
359 | unsigned int tri0, int iaa, unsigned long hints) | |
360 | { | |
361 | struct gru_instruction *ins = (struct gru_instruction *)cb; | |
362 | ||
363 | ins->baddr0 = (long)gpa | ((unsigned long)iaa << 62); | |
364 | ins->nelem = 1; | |
289750d1 | 365 | ins->op1_stride = 1; |
57ebb034 JS |
366 | gru_start_instruction(ins, __opdword(OP_VLOAD, 0, XTYPE_DW, iaa, 0, |
367 | (unsigned long)tri0, CB_IMA(hints))); | |
289750d1 RH |
368 | } |
369 | ||
76148df1 JS |
370 | static inline void gru_vstore_phys(void *cb, unsigned long gpa, |
371 | unsigned int tri0, int iaa, unsigned long hints) | |
372 | { | |
373 | struct gru_instruction *ins = (struct gru_instruction *)cb; | |
374 | ||
375 | ins->baddr0 = (long)gpa | ((unsigned long)iaa << 62); | |
376 | ins->nelem = 1; | |
377 | ins->op1_stride = 1; | |
378 | gru_start_instruction(ins, __opdword(OP_VSTORE, 0, XTYPE_DW, iaa, 0, | |
379 | (unsigned long)tri0, CB_IMA(hints))); | |
380 | } | |
381 | ||
4c921d4d JS |
382 | static inline void gru_vload(void *cb, unsigned long mem_addr, |
383 | unsigned int tri0, unsigned char xtype, unsigned long nelem, | |
384 | unsigned long stride, unsigned long hints) | |
385 | { | |
386 | struct gru_instruction *ins = (struct gru_instruction *)cb; | |
387 | ||
388 | ins->baddr0 = (long)mem_addr; | |
389 | ins->nelem = nelem; | |
4c921d4d | 390 | ins->op1_stride = stride; |
57ebb034 JS |
391 | gru_start_instruction(ins, __opdword(OP_VLOAD, 0, xtype, IAA_RAM, 0, |
392 | (unsigned long)tri0, CB_IMA(hints))); | |
4c921d4d JS |
393 | } |
394 | ||
395 | static inline void gru_vstore(void *cb, unsigned long mem_addr, | |
396 | unsigned int tri0, unsigned char xtype, unsigned long nelem, | |
397 | unsigned long stride, unsigned long hints) | |
398 | { | |
399 | struct gru_instruction *ins = (void *)cb; | |
400 | ||
401 | ins->baddr0 = (long)mem_addr; | |
402 | ins->nelem = nelem; | |
4c921d4d | 403 | ins->op1_stride = stride; |
57ebb034 JS |
404 | gru_start_instruction(ins, __opdword(OP_VSTORE, 0, xtype, IAA_RAM, 0, |
405 | tri0, CB_IMA(hints))); | |
4c921d4d JS |
406 | } |
407 | ||
408 | static inline void gru_ivload(void *cb, unsigned long mem_addr, | |
409 | unsigned int tri0, unsigned int tri1, unsigned char xtype, | |
410 | unsigned long nelem, unsigned long hints) | |
411 | { | |
412 | struct gru_instruction *ins = (void *)cb; | |
413 | ||
414 | ins->baddr0 = (long)mem_addr; | |
415 | ins->nelem = nelem; | |
4c921d4d | 416 | ins->tri1_bufsize = tri1; |
57ebb034 JS |
417 | gru_start_instruction(ins, __opdword(OP_IVLOAD, 0, xtype, IAA_RAM, 0, |
418 | tri0, CB_IMA(hints))); | |
4c921d4d JS |
419 | } |
420 | ||
421 | static inline void gru_ivstore(void *cb, unsigned long mem_addr, | |
422 | unsigned int tri0, unsigned int tri1, | |
423 | unsigned char xtype, unsigned long nelem, unsigned long hints) | |
424 | { | |
425 | struct gru_instruction *ins = (void *)cb; | |
426 | ||
427 | ins->baddr0 = (long)mem_addr; | |
428 | ins->nelem = nelem; | |
4c921d4d | 429 | ins->tri1_bufsize = tri1; |
57ebb034 JS |
430 | gru_start_instruction(ins, __opdword(OP_IVSTORE, 0, xtype, IAA_RAM, 0, |
431 | tri0, CB_IMA(hints))); | |
4c921d4d JS |
432 | } |
433 | ||
434 | static inline void gru_vset(void *cb, unsigned long mem_addr, | |
435 | unsigned long value, unsigned char xtype, unsigned long nelem, | |
436 | unsigned long stride, unsigned long hints) | |
437 | { | |
438 | struct gru_instruction *ins = (void *)cb; | |
439 | ||
440 | ins->baddr0 = (long)mem_addr; | |
441 | ins->op2_value_baddr1 = value; | |
442 | ins->nelem = nelem; | |
443 | ins->op1_stride = stride; | |
57ebb034 JS |
444 | gru_start_instruction(ins, __opdword(OP_VSET, 0, xtype, IAA_RAM, 0, |
445 | 0, CB_IMA(hints))); | |
4c921d4d JS |
446 | } |
447 | ||
448 | static inline void gru_ivset(void *cb, unsigned long mem_addr, | |
449 | unsigned int tri1, unsigned long value, unsigned char xtype, | |
450 | unsigned long nelem, unsigned long hints) | |
451 | { | |
452 | struct gru_instruction *ins = (void *)cb; | |
453 | ||
454 | ins->baddr0 = (long)mem_addr; | |
455 | ins->op2_value_baddr1 = value; | |
456 | ins->nelem = nelem; | |
457 | ins->tri1_bufsize = tri1; | |
57ebb034 JS |
458 | gru_start_instruction(ins, __opdword(OP_IVSET, 0, xtype, IAA_RAM, 0, |
459 | 0, CB_IMA(hints))); | |
4c921d4d JS |
460 | } |
461 | ||
462 | static inline void gru_vflush(void *cb, unsigned long mem_addr, | |
463 | unsigned long nelem, unsigned char xtype, unsigned long stride, | |
464 | unsigned long hints) | |
465 | { | |
466 | struct gru_instruction *ins = (void *)cb; | |
467 | ||
468 | ins->baddr0 = (long)mem_addr; | |
469 | ins->op1_stride = stride; | |
470 | ins->nelem = nelem; | |
57ebb034 JS |
471 | gru_start_instruction(ins, __opdword(OP_VFLUSH, 0, xtype, IAA_RAM, 0, |
472 | 0, CB_IMA(hints))); | |
4c921d4d JS |
473 | } |
474 | ||
475 | static inline void gru_nop(void *cb, int hints) | |
476 | { | |
477 | struct gru_instruction *ins = (void *)cb; | |
478 | ||
57ebb034 | 479 | gru_start_instruction(ins, __opdword(OP_NOP, 0, 0, 0, 0, 0, CB_IMA(hints))); |
4c921d4d JS |
480 | } |
481 | ||
482 | ||
483 | static inline void gru_bcopy(void *cb, const unsigned long src, | |
484 | unsigned long dest, | |
485 | unsigned int tri0, unsigned int xtype, unsigned long nelem, | |
486 | unsigned int bufsize, unsigned long hints) | |
487 | { | |
488 | struct gru_instruction *ins = (void *)cb; | |
489 | ||
490 | ins->baddr0 = (long)src; | |
491 | ins->op2_value_baddr1 = (long)dest; | |
492 | ins->nelem = nelem; | |
4c921d4d | 493 | ins->tri1_bufsize = bufsize; |
57ebb034 JS |
494 | gru_start_instruction(ins, __opdword(OP_BCOPY, 0, xtype, IAA_RAM, |
495 | IAA_RAM, tri0, CB_IMA(hints))); | |
4c921d4d JS |
496 | } |
497 | ||
498 | static inline void gru_bstore(void *cb, const unsigned long src, | |
499 | unsigned long dest, unsigned int tri0, unsigned int xtype, | |
500 | unsigned long nelem, unsigned long hints) | |
501 | { | |
502 | struct gru_instruction *ins = (void *)cb; | |
503 | ||
504 | ins->baddr0 = (long)src; | |
505 | ins->op2_value_baddr1 = (long)dest; | |
506 | ins->nelem = nelem; | |
57ebb034 JS |
507 | gru_start_instruction(ins, __opdword(OP_BSTORE, 0, xtype, 0, IAA_RAM, |
508 | tri0, CB_IMA(hints))); | |
4c921d4d JS |
509 | } |
510 | ||
511 | static inline void gru_gamir(void *cb, int exopc, unsigned long src, | |
512 | unsigned int xtype, unsigned long hints) | |
513 | { | |
514 | struct gru_instruction *ins = (void *)cb; | |
515 | ||
516 | ins->baddr0 = (long)src; | |
57ebb034 JS |
517 | gru_start_instruction(ins, __opdword(OP_GAMIR, exopc, xtype, IAA_RAM, 0, |
518 | 0, CB_IMA(hints))); | |
4c921d4d JS |
519 | } |
520 | ||
521 | static inline void gru_gamirr(void *cb, int exopc, unsigned long src, | |
522 | unsigned int xtype, unsigned long hints) | |
523 | { | |
524 | struct gru_instruction *ins = (void *)cb; | |
525 | ||
526 | ins->baddr0 = (long)src; | |
57ebb034 JS |
527 | gru_start_instruction(ins, __opdword(OP_GAMIRR, exopc, xtype, IAA_RAM, 0, |
528 | 0, CB_IMA(hints))); | |
4c921d4d JS |
529 | } |
530 | ||
531 | static inline void gru_gamer(void *cb, int exopc, unsigned long src, | |
532 | unsigned int xtype, | |
533 | unsigned long operand1, unsigned long operand2, | |
534 | unsigned long hints) | |
535 | { | |
536 | struct gru_instruction *ins = (void *)cb; | |
537 | ||
538 | ins->baddr0 = (long)src; | |
539 | ins->op1_stride = operand1; | |
540 | ins->op2_value_baddr1 = operand2; | |
57ebb034 JS |
541 | gru_start_instruction(ins, __opdword(OP_GAMER, exopc, xtype, IAA_RAM, 0, |
542 | 0, CB_IMA(hints))); | |
4c921d4d JS |
543 | } |
544 | ||
545 | static inline void gru_gamerr(void *cb, int exopc, unsigned long src, | |
546 | unsigned int xtype, unsigned long operand1, | |
547 | unsigned long operand2, unsigned long hints) | |
548 | { | |
549 | struct gru_instruction *ins = (void *)cb; | |
550 | ||
551 | ins->baddr0 = (long)src; | |
552 | ins->op1_stride = operand1; | |
553 | ins->op2_value_baddr1 = operand2; | |
57ebb034 JS |
554 | gru_start_instruction(ins, __opdword(OP_GAMERR, exopc, xtype, IAA_RAM, 0, |
555 | 0, CB_IMA(hints))); | |
4c921d4d JS |
556 | } |
557 | ||
558 | static inline void gru_gamxr(void *cb, unsigned long src, | |
559 | unsigned int tri0, unsigned long hints) | |
560 | { | |
561 | struct gru_instruction *ins = (void *)cb; | |
562 | ||
563 | ins->baddr0 = (long)src; | |
564 | ins->nelem = 4; | |
57ebb034 JS |
565 | gru_start_instruction(ins, __opdword(OP_GAMXR, EOP_XR_CSWAP, XTYPE_DW, |
566 | IAA_RAM, 0, 0, CB_IMA(hints))); | |
4c921d4d JS |
567 | } |
568 | ||
569 | static inline void gru_mesq(void *cb, unsigned long queue, | |
570 | unsigned long tri0, unsigned long nelem, | |
571 | unsigned long hints) | |
572 | { | |
573 | struct gru_instruction *ins = (void *)cb; | |
574 | ||
575 | ins->baddr0 = (long)queue; | |
576 | ins->nelem = nelem; | |
57ebb034 JS |
577 | gru_start_instruction(ins, __opdword(OP_MESQ, 0, XTYPE_CL, IAA_RAM, 0, |
578 | tri0, CB_IMA(hints))); | |
4c921d4d JS |
579 | } |
580 | ||
581 | static inline unsigned long gru_get_amo_value(void *cb) | |
582 | { | |
583 | struct gru_instruction *ins = (void *)cb; | |
584 | ||
585 | return ins->avalue; | |
586 | } | |
587 | ||
588 | static inline int gru_get_amo_value_head(void *cb) | |
589 | { | |
590 | struct gru_instruction *ins = (void *)cb; | |
591 | ||
592 | return ins->avalue & 0xffffffff; | |
593 | } | |
594 | ||
595 | static inline int gru_get_amo_value_limit(void *cb) | |
596 | { | |
597 | struct gru_instruction *ins = (void *)cb; | |
598 | ||
599 | return ins->avalue >> 32; | |
600 | } | |
601 | ||
602 | static inline union gru_mesqhead gru_mesq_head(int head, int limit) | |
603 | { | |
604 | union gru_mesqhead mqh; | |
605 | ||
606 | mqh.head = head; | |
607 | mqh.limit = limit; | |
608 | return mqh; | |
609 | } | |
610 | ||
611 | /* | |
612 | * Get struct control_block_extended_exc_detail for CB. | |
613 | */ | |
614 | extern int gru_get_cb_exception_detail(void *cb, | |
615 | struct control_block_extended_exc_detail *excdet); | |
616 | ||
617 | #define GRU_EXC_STR_SIZE 256 | |
618 | ||
4c921d4d JS |
619 | |
620 | /* | |
621 | * Control block definition for checking status | |
622 | */ | |
623 | struct gru_control_block_status { | |
624 | unsigned int icmd :1; | |
fe5bb6b0 JS |
625 | unsigned int ima :3; |
626 | unsigned int reserved0 :4; | |
627 | unsigned int unused1 :24; | |
4c921d4d JS |
628 | unsigned int unused2 :24; |
629 | unsigned int istatus :2; | |
630 | unsigned int isubstatus :4; | |
fe5bb6b0 | 631 | unsigned int unused3 :2; |
4c921d4d JS |
632 | }; |
633 | ||
634 | /* Get CB status */ | |
635 | static inline int gru_get_cb_status(void *cb) | |
636 | { | |
637 | struct gru_control_block_status *cbs = (void *)cb; | |
638 | ||
639 | return cbs->istatus; | |
640 | } | |
641 | ||
642 | /* Get CB message queue substatus */ | |
643 | static inline int gru_get_cb_message_queue_substatus(void *cb) | |
644 | { | |
645 | struct gru_control_block_status *cbs = (void *)cb; | |
646 | ||
647 | return cbs->isubstatus & CBSS_MSG_QUEUE_MASK; | |
648 | } | |
649 | ||
650 | /* Get CB substatus */ | |
651 | static inline int gru_get_cb_substatus(void *cb) | |
652 | { | |
653 | struct gru_control_block_status *cbs = (void *)cb; | |
654 | ||
655 | return cbs->isubstatus; | |
656 | } | |
657 | ||
9f250114 JS |
658 | /* |
659 | * User interface to check an instruction status. UPM and exceptions | |
660 | * are handled automatically. However, this function does NOT wait | |
661 | * for an active instruction to complete. | |
662 | * | |
4c921d4d JS |
663 | */ |
664 | static inline int gru_check_status(void *cb) | |
665 | { | |
666 | struct gru_control_block_status *cbs = (void *)cb; | |
923f7f69 | 667 | int ret; |
4c921d4d | 668 | |
923f7f69 | 669 | ret = cbs->istatus; |
9f250114 | 670 | if (ret != CBS_ACTIVE) |
4c921d4d JS |
671 | ret = gru_check_status_proc(cb); |
672 | return ret; | |
673 | } | |
674 | ||
9f250114 JS |
675 | /* |
676 | * User interface (via inline function) to wait for an instruction | |
677 | * to complete. Completion status (IDLE or EXCEPTION is returned | |
678 | * to the user. Exception due to hardware errors are automatically | |
679 | * retried before returning an exception. | |
680 | * | |
4c921d4d JS |
681 | */ |
682 | static inline int gru_wait(void *cb) | |
683 | { | |
9f250114 | 684 | return gru_wait_proc(cb); |
4c921d4d JS |
685 | } |
686 | ||
9f250114 JS |
687 | /* |
688 | * Wait for CB to complete. Aborts program if error. (Note: error does NOT | |
4c921d4d JS |
689 | * mean TLB mis - only fatal errors such as memory parity error or user |
690 | * bugs will cause termination. | |
691 | */ | |
692 | static inline void gru_wait_abort(void *cb) | |
693 | { | |
9f250114 | 694 | gru_wait_abort_proc(cb); |
4c921d4d JS |
695 | } |
696 | ||
57ebb034 JS |
697 | /* |
698 | * Get a pointer to the start of a gseg | |
699 | * p - Any valid pointer within the gseg | |
700 | */ | |
701 | static inline void *gru_get_gseg_pointer (void *p) | |
702 | { | |
703 | return (void *)((unsigned long)p & ~(GRU_GSEG_PAGESIZE - 1)); | |
704 | } | |
4c921d4d JS |
705 | |
706 | /* | |
707 | * Get a pointer to a control block | |
708 | * gseg - GSeg address returned from gru_get_thread_gru_segment() | |
709 | * index - index of desired CB | |
710 | */ | |
711 | static inline void *gru_get_cb_pointer(void *gseg, | |
712 | int index) | |
713 | { | |
714 | return gseg + GRU_CB_BASE + index * GRU_HANDLE_STRIDE; | |
715 | } | |
716 | ||
717 | /* | |
718 | * Get a pointer to a cacheline in the data segment portion of a GSeg | |
719 | * gseg - GSeg address returned from gru_get_thread_gru_segment() | |
720 | * index - index of desired cache line | |
721 | */ | |
722 | static inline void *gru_get_data_pointer(void *gseg, int index) | |
723 | { | |
724 | return gseg + GRU_DS_BASE + index * GRU_CACHE_LINE_BYTES; | |
725 | } | |
726 | ||
727 | /* | |
728 | * Convert a vaddr into the tri index within the GSEG | |
729 | * vaddr - virtual address of within gseg | |
730 | */ | |
731 | static inline int gru_get_tri(void *vaddr) | |
732 | { | |
733 | return ((unsigned long)vaddr & (GRU_GSEG_PAGESIZE - 1)) - GRU_DS_BASE; | |
734 | } | |
735 | #endif /* __GRU_INSTRUCTIONS_H__ */ |