gru_instructions.h 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691
  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. #ifndef __GRU_INSTRUCTIONS_H__
  19. #define __GRU_INSTRUCTIONS_H__
  20. extern int gru_check_status_proc(void *cb);
  21. extern int gru_wait_proc(void *cb);
  22. extern void gru_wait_abort_proc(void *cb);
  23. /*
  24. * Architecture dependent functions
  25. */
  26. #if defined(CONFIG_IA64)
  27. #include <linux/compiler.h>
  28. #include <asm/intrinsics.h>
  29. #define __flush_cache(p) ia64_fc((unsigned long)p)
  30. /* Use volatile on IA64 to ensure ordering via st4.rel */
  31. #define gru_ordered_store_int(p, v) \
  32. do { \
  33. barrier(); \
  34. *((volatile int *)(p)) = v; /* force st.rel */ \
  35. } while (0)
  36. #elif defined(CONFIG_X86_64)
  37. #define __flush_cache(p) clflush(p)
  38. #define gru_ordered_store_int(p, v) \
  39. do { \
  40. barrier(); \
  41. *(int *)p = v; \
  42. } while (0)
  43. #else
  44. #error "Unsupported architecture"
  45. #endif
  46. /*
  47. * Control block status and exception codes
  48. */
  49. #define CBS_IDLE 0
  50. #define CBS_EXCEPTION 1
  51. #define CBS_ACTIVE 2
  52. #define CBS_CALL_OS 3
  53. /* CB substatus bitmasks */
  54. #define CBSS_MSG_QUEUE_MASK 7
  55. #define CBSS_IMPLICIT_ABORT_ACTIVE_MASK 8
  56. /* CB substatus message queue values (low 3 bits of substatus) */
  57. #define CBSS_NO_ERROR 0
  58. #define CBSS_LB_OVERFLOWED 1
  59. #define CBSS_QLIMIT_REACHED 2
  60. #define CBSS_PAGE_OVERFLOW 3
  61. #define CBSS_AMO_NACKED 4
  62. #define CBSS_PUT_NACKED 5
  63. /*
  64. * Structure used to fetch exception detail for CBs that terminate with
  65. * CBS_EXCEPTION
  66. */
  67. struct control_block_extended_exc_detail {
  68. unsigned long cb;
  69. int opc;
  70. int ecause;
  71. int exopc;
  72. long exceptdet0;
  73. int exceptdet1;
  74. int cbrstate;
  75. int cbrexecstatus;
  76. };
  77. /*
  78. * Instruction formats
  79. */
  80. /*
  81. * Generic instruction format.
  82. * This definition has precise bit field definitions.
  83. */
  84. struct gru_instruction_bits {
  85. /* DW 0 - low */
  86. unsigned int icmd: 1;
  87. unsigned char ima: 3; /* CB_DelRep, unmapped mode */
  88. unsigned char reserved0: 4;
  89. unsigned int xtype: 3;
  90. unsigned int iaa0: 2;
  91. unsigned int iaa1: 2;
  92. unsigned char reserved1: 1;
  93. unsigned char opc: 8; /* opcode */
  94. unsigned char exopc: 8; /* extended opcode */
  95. /* DW 0 - high */
  96. unsigned int idef2: 22; /* TRi0 */
  97. unsigned char reserved2: 2;
  98. unsigned char istatus: 2;
  99. unsigned char isubstatus:4;
  100. unsigned char reserved3: 1;
  101. unsigned char tlb_fault_color: 1;
  102. /* DW 1 */
  103. unsigned long idef4; /* 42 bits: TRi1, BufSize */
  104. /* DW 2-6 */
  105. unsigned long idef1; /* BAddr0 */
  106. unsigned long idef5; /* Nelem */
  107. unsigned long idef6; /* Stride, Operand1 */
  108. unsigned long idef3; /* BAddr1, Value, Operand2 */
  109. unsigned long reserved4;
  110. /* DW 7 */
  111. unsigned long avalue; /* AValue */
  112. };
  113. /*
  114. * Generic instruction with friendlier names. This format is used
  115. * for inline instructions.
  116. */
  117. struct gru_instruction {
  118. /* DW 0 */
  119. unsigned int op32; /* icmd,xtype,iaa0,ima,opc */
  120. unsigned int tri0;
  121. unsigned long tri1_bufsize; /* DW 1 */
  122. unsigned long baddr0; /* DW 2 */
  123. unsigned long nelem; /* DW 3 */
  124. unsigned long op1_stride; /* DW 4 */
  125. unsigned long op2_value_baddr1; /* DW 5 */
  126. unsigned long reserved0; /* DW 6 */
  127. unsigned long avalue; /* DW 7 */
  128. };
  129. /* Some shifts and masks for the low 32 bits of a GRU command */
  130. #define GRU_CB_ICMD_SHFT 0
  131. #define GRU_CB_ICMD_MASK 0x1
  132. #define GRU_CB_XTYPE_SHFT 8
  133. #define GRU_CB_XTYPE_MASK 0x7
  134. #define GRU_CB_IAA0_SHFT 11
  135. #define GRU_CB_IAA0_MASK 0x3
  136. #define GRU_CB_IAA1_SHFT 13
  137. #define GRU_CB_IAA1_MASK 0x3
  138. #define GRU_CB_IMA_SHFT 1
  139. #define GRU_CB_IMA_MASK 0x3
  140. #define GRU_CB_OPC_SHFT 16
  141. #define GRU_CB_OPC_MASK 0xff
  142. #define GRU_CB_EXOPC_SHFT 24
  143. #define GRU_CB_EXOPC_MASK 0xff
  144. /* GRU instruction opcodes (opc field) */
  145. #define OP_NOP 0x00
  146. #define OP_BCOPY 0x01
  147. #define OP_VLOAD 0x02
  148. #define OP_IVLOAD 0x03
  149. #define OP_VSTORE 0x04
  150. #define OP_IVSTORE 0x05
  151. #define OP_VSET 0x06
  152. #define OP_IVSET 0x07
  153. #define OP_MESQ 0x08
  154. #define OP_GAMXR 0x09
  155. #define OP_GAMIR 0x0a
  156. #define OP_GAMIRR 0x0b
  157. #define OP_GAMER 0x0c
  158. #define OP_GAMERR 0x0d
  159. #define OP_BSTORE 0x0e
  160. #define OP_VFLUSH 0x0f
  161. /* Extended opcodes values (exopc field) */
  162. /* GAMIR - AMOs with implicit operands */
  163. #define EOP_IR_FETCH 0x01 /* Plain fetch of memory */
  164. #define EOP_IR_CLR 0x02 /* Fetch and clear */
  165. #define EOP_IR_INC 0x05 /* Fetch and increment */
  166. #define EOP_IR_DEC 0x07 /* Fetch and decrement */
  167. #define EOP_IR_QCHK1 0x0d /* Queue check, 64 byte msg */
  168. #define EOP_IR_QCHK2 0x0e /* Queue check, 128 byte msg */
  169. /* GAMIRR - Registered AMOs with implicit operands */
  170. #define EOP_IRR_FETCH 0x01 /* Registered fetch of memory */
  171. #define EOP_IRR_CLR 0x02 /* Registered fetch and clear */
  172. #define EOP_IRR_INC 0x05 /* Registered fetch and increment */
  173. #define EOP_IRR_DEC 0x07 /* Registered fetch and decrement */
  174. #define EOP_IRR_DECZ 0x0f /* Registered fetch and decrement, update on zero*/
  175. /* GAMER - AMOs with explicit operands */
  176. #define EOP_ER_SWAP 0x00 /* Exchange argument and memory */
  177. #define EOP_ER_OR 0x01 /* Logical OR with memory */
  178. #define EOP_ER_AND 0x02 /* Logical AND with memory */
  179. #define EOP_ER_XOR 0x03 /* Logical XOR with memory */
  180. #define EOP_ER_ADD 0x04 /* Add value to memory */
  181. #define EOP_ER_CSWAP 0x08 /* Compare with operand2, write operand1 if match*/
  182. #define EOP_ER_CADD 0x0c /* Queue check, operand1*64 byte msg */
  183. /* GAMERR - Registered AMOs with explicit operands */
  184. #define EOP_ERR_SWAP 0x00 /* Exchange argument and memory */
  185. #define EOP_ERR_OR 0x01 /* Logical OR with memory */
  186. #define EOP_ERR_AND 0x02 /* Logical AND with memory */
  187. #define EOP_ERR_XOR 0x03 /* Logical XOR with memory */
  188. #define EOP_ERR_ADD 0x04 /* Add value to memory */
  189. #define EOP_ERR_CSWAP 0x08 /* Compare with operand2, write operand1 if match*/
  190. #define EOP_ERR_EPOLL 0x09 /* Poll for equality */
  191. #define EOP_ERR_NPOLL 0x0a /* Poll for inequality */
  192. /* GAMXR - SGI Arithmetic unit */
  193. #define EOP_XR_CSWAP 0x0b /* Masked compare exchange */
  194. /* Transfer types (xtype field) */
  195. #define XTYPE_B 0x0 /* byte */
  196. #define XTYPE_S 0x1 /* short (2-byte) */
  197. #define XTYPE_W 0x2 /* word (4-byte) */
  198. #define XTYPE_DW 0x3 /* doubleword (8-byte) */
  199. #define XTYPE_CL 0x6 /* cacheline (64-byte) */
  200. /* Instruction access attributes (iaa0, iaa1 fields) */
  201. #define IAA_RAM 0x0 /* normal cached RAM access */
  202. #define IAA_NCRAM 0x2 /* noncoherent RAM access */
  203. #define IAA_MMIO 0x1 /* noncoherent memory-mapped I/O space */
  204. #define IAA_REGISTER 0x3 /* memory-mapped registers, etc. */
  205. /* Instruction mode attributes (ima field) */
  206. #define IMA_MAPPED 0x0 /* Virtual mode */
  207. #define IMA_CB_DELAY 0x1 /* hold read responses until status changes */
  208. #define IMA_UNMAPPED 0x2 /* bypass the TLBs (OS only) */
  209. #define IMA_INTERRUPT 0x4 /* Interrupt when instruction completes */
  210. /* CBE ecause bits */
  211. #define CBE_CAUSE_RI (1 << 0)
  212. #define CBE_CAUSE_INVALID_INSTRUCTION (1 << 1)
  213. #define CBE_CAUSE_UNMAPPED_MODE_FORBIDDEN (1 << 2)
  214. #define CBE_CAUSE_PE_CHECK_DATA_ERROR (1 << 3)
  215. #define CBE_CAUSE_IAA_GAA_MISMATCH (1 << 4)
  216. #define CBE_CAUSE_DATA_SEGMENT_LIMIT_EXCEPTION (1 << 5)
  217. #define CBE_CAUSE_OS_FATAL_TLB_FAULT (1 << 6)
  218. #define CBE_CAUSE_EXECUTION_HW_ERROR (1 << 7)
  219. #define CBE_CAUSE_TLBHW_ERROR (1 << 8)
  220. #define CBE_CAUSE_RA_REQUEST_TIMEOUT (1 << 9)
  221. #define CBE_CAUSE_HA_REQUEST_TIMEOUT (1 << 10)
  222. #define CBE_CAUSE_RA_RESPONSE_FATAL (1 << 11)
  223. #define CBE_CAUSE_RA_RESPONSE_NON_FATAL (1 << 12)
  224. #define CBE_CAUSE_HA_RESPONSE_FATAL (1 << 13)
  225. #define CBE_CAUSE_HA_RESPONSE_NON_FATAL (1 << 14)
  226. #define CBE_CAUSE_ADDRESS_SPACE_DECODE_ERROR (1 << 15)
  227. #define CBE_CAUSE_RESPONSE_DATA_ERROR (1 << 16)
  228. #define CBE_CAUSE_PROTOCOL_STATE_DATA_ERROR (1 << 17)
  229. /* CBE cbrexecstatus bits */
  230. #define CBR_EXS_ABORT_OCC_BIT 0
  231. #define CBR_EXS_INT_OCC_BIT 1
  232. #define CBR_EXS_PENDING_BIT 2
  233. #define CBR_EXS_QUEUED_BIT 3
  234. #define CBR_EXS_TLBHW_BIT 4
  235. #define CBR_EXS_EXCEPTION_BIT 5
  236. #define CBR_EXS_ABORT_OCC (1 << CBR_EXS_ABORT_OCC_BIT)
  237. #define CBR_EXS_INT_OCC (1 << CBR_EXS_INT_OCC_BIT)
  238. #define CBR_EXS_PENDING (1 << CBR_EXS_PENDING_BIT)
  239. #define CBR_EXS_QUEUED (1 << CBR_EXS_QUEUED_BIT)
  240. #define CBR_EXS_TLBHW (1 << CBR_EXS_TLBHW_BIT)
  241. #define CBR_EXS_EXCEPTION (1 << CBR_EXS_EXCEPTION_BIT)
  242. /*
  243. * Exceptions are retried for the following cases. If any OTHER bits are set
  244. * in ecause, the exception is not retryable.
  245. */
  246. #define EXCEPTION_RETRY_BITS (CBE_CAUSE_RESPONSE_DATA_ERROR | \
  247. CBE_CAUSE_RA_REQUEST_TIMEOUT | \
  248. CBE_CAUSE_TLBHW_ERROR | \
  249. CBE_CAUSE_HA_REQUEST_TIMEOUT)
  250. /* Message queue head structure */
  251. union gru_mesqhead {
  252. unsigned long val;
  253. struct {
  254. unsigned int head;
  255. unsigned int limit;
  256. };
  257. };
  258. /* Generate the low word of a GRU instruction */
  259. static inline unsigned int
  260. __opword(unsigned char opcode, unsigned char exopc, unsigned char xtype,
  261. unsigned char iaa0, unsigned char iaa1,
  262. unsigned char ima)
  263. {
  264. return (1 << GRU_CB_ICMD_SHFT) |
  265. (iaa0 << GRU_CB_IAA0_SHFT) |
  266. (iaa1 << GRU_CB_IAA1_SHFT) |
  267. (ima << GRU_CB_IMA_SHFT) |
  268. (xtype << GRU_CB_XTYPE_SHFT) |
  269. (opcode << GRU_CB_OPC_SHFT) |
  270. (exopc << GRU_CB_EXOPC_SHFT);
  271. }
  272. /*
  273. * Architecture specific intrinsics
  274. */
  275. static inline void gru_flush_cache(void *p)
  276. {
  277. __flush_cache(p);
  278. }
  279. /*
  280. * Store the lower 32 bits of the command including the "start" bit. Then
  281. * start the instruction executing.
  282. */
  283. static inline void gru_start_instruction(struct gru_instruction *ins, int op32)
  284. {
  285. gru_ordered_store_int(ins, op32);
  286. gru_flush_cache(ins);
  287. }
  288. /* Convert "hints" to IMA */
  289. #define CB_IMA(h) ((h) | IMA_UNMAPPED)
  290. /* Convert data segment cache line index into TRI0 / TRI1 value */
  291. #define GRU_DINDEX(i) ((i) * GRU_CACHE_LINE_BYTES)
  292. /* Inline functions for GRU instructions.
  293. * Note:
  294. * - nelem and stride are in elements
  295. * - tri0/tri1 is in bytes for the beginning of the data segment.
  296. */
  297. static inline void gru_vload(void *cb, unsigned long mem_addr,
  298. unsigned int tri0, unsigned char xtype, unsigned long nelem,
  299. unsigned long stride, unsigned long hints)
  300. {
  301. struct gru_instruction *ins = (struct gru_instruction *)cb;
  302. ins->baddr0 = (long)mem_addr;
  303. ins->nelem = nelem;
  304. ins->tri0 = tri0;
  305. ins->op1_stride = stride;
  306. gru_start_instruction(ins, __opword(OP_VLOAD, 0, xtype, IAA_RAM, 0,
  307. CB_IMA(hints)));
  308. }
  309. static inline void gru_vstore(void *cb, unsigned long mem_addr,
  310. unsigned int tri0, unsigned char xtype, unsigned long nelem,
  311. unsigned long stride, unsigned long hints)
  312. {
  313. struct gru_instruction *ins = (void *)cb;
  314. ins->baddr0 = (long)mem_addr;
  315. ins->nelem = nelem;
  316. ins->tri0 = tri0;
  317. ins->op1_stride = stride;
  318. gru_start_instruction(ins, __opword(OP_VSTORE, 0, xtype, IAA_RAM, 0,
  319. CB_IMA(hints)));
  320. }
  321. static inline void gru_ivload(void *cb, unsigned long mem_addr,
  322. unsigned int tri0, unsigned int tri1, unsigned char xtype,
  323. unsigned long nelem, unsigned long hints)
  324. {
  325. struct gru_instruction *ins = (void *)cb;
  326. ins->baddr0 = (long)mem_addr;
  327. ins->nelem = nelem;
  328. ins->tri0 = tri0;
  329. ins->tri1_bufsize = tri1;
  330. gru_start_instruction(ins, __opword(OP_IVLOAD, 0, xtype, IAA_RAM, 0,
  331. CB_IMA(hints)));
  332. }
  333. static inline void gru_ivstore(void *cb, unsigned long mem_addr,
  334. unsigned int tri0, unsigned int tri1,
  335. unsigned char xtype, unsigned long nelem, unsigned long hints)
  336. {
  337. struct gru_instruction *ins = (void *)cb;
  338. ins->baddr0 = (long)mem_addr;
  339. ins->nelem = nelem;
  340. ins->tri0 = tri0;
  341. ins->tri1_bufsize = tri1;
  342. gru_start_instruction(ins, __opword(OP_IVSTORE, 0, xtype, IAA_RAM, 0,
  343. CB_IMA(hints)));
  344. }
  345. static inline void gru_vset(void *cb, unsigned long mem_addr,
  346. unsigned long value, unsigned char xtype, unsigned long nelem,
  347. unsigned long stride, unsigned long hints)
  348. {
  349. struct gru_instruction *ins = (void *)cb;
  350. ins->baddr0 = (long)mem_addr;
  351. ins->op2_value_baddr1 = value;
  352. ins->nelem = nelem;
  353. ins->op1_stride = stride;
  354. gru_start_instruction(ins, __opword(OP_VSET, 0, xtype, IAA_RAM, 0,
  355. CB_IMA(hints)));
  356. }
  357. static inline void gru_ivset(void *cb, unsigned long mem_addr,
  358. unsigned int tri1, unsigned long value, unsigned char xtype,
  359. unsigned long nelem, unsigned long hints)
  360. {
  361. struct gru_instruction *ins = (void *)cb;
  362. ins->baddr0 = (long)mem_addr;
  363. ins->op2_value_baddr1 = value;
  364. ins->nelem = nelem;
  365. ins->tri1_bufsize = tri1;
  366. gru_start_instruction(ins, __opword(OP_IVSET, 0, xtype, IAA_RAM, 0,
  367. CB_IMA(hints)));
  368. }
  369. static inline void gru_vflush(void *cb, unsigned long mem_addr,
  370. unsigned long nelem, unsigned char xtype, unsigned long stride,
  371. unsigned long hints)
  372. {
  373. struct gru_instruction *ins = (void *)cb;
  374. ins->baddr0 = (long)mem_addr;
  375. ins->op1_stride = stride;
  376. ins->nelem = nelem;
  377. gru_start_instruction(ins, __opword(OP_VFLUSH, 0, xtype, IAA_RAM, 0,
  378. CB_IMA(hints)));
  379. }
  380. static inline void gru_nop(void *cb, int hints)
  381. {
  382. struct gru_instruction *ins = (void *)cb;
  383. gru_start_instruction(ins, __opword(OP_NOP, 0, 0, 0, 0, CB_IMA(hints)));
  384. }
  385. static inline void gru_bcopy(void *cb, const unsigned long src,
  386. unsigned long dest,
  387. unsigned int tri0, unsigned int xtype, unsigned long nelem,
  388. unsigned int bufsize, unsigned long hints)
  389. {
  390. struct gru_instruction *ins = (void *)cb;
  391. ins->baddr0 = (long)src;
  392. ins->op2_value_baddr1 = (long)dest;
  393. ins->nelem = nelem;
  394. ins->tri0 = tri0;
  395. ins->tri1_bufsize = bufsize;
  396. gru_start_instruction(ins, __opword(OP_BCOPY, 0, xtype, IAA_RAM,
  397. IAA_RAM, CB_IMA(hints)));
  398. }
  399. static inline void gru_bstore(void *cb, const unsigned long src,
  400. unsigned long dest, unsigned int tri0, unsigned int xtype,
  401. unsigned long nelem, unsigned long hints)
  402. {
  403. struct gru_instruction *ins = (void *)cb;
  404. ins->baddr0 = (long)src;
  405. ins->op2_value_baddr1 = (long)dest;
  406. ins->nelem = nelem;
  407. ins->tri0 = tri0;
  408. gru_start_instruction(ins, __opword(OP_BSTORE, 0, xtype, 0, IAA_RAM,
  409. CB_IMA(hints)));
  410. }
  411. static inline void gru_gamir(void *cb, int exopc, unsigned long src,
  412. unsigned int xtype, unsigned long hints)
  413. {
  414. struct gru_instruction *ins = (void *)cb;
  415. ins->baddr0 = (long)src;
  416. gru_start_instruction(ins, __opword(OP_GAMIR, exopc, xtype, IAA_RAM, 0,
  417. CB_IMA(hints)));
  418. }
  419. static inline void gru_gamirr(void *cb, int exopc, unsigned long src,
  420. unsigned int xtype, unsigned long hints)
  421. {
  422. struct gru_instruction *ins = (void *)cb;
  423. ins->baddr0 = (long)src;
  424. gru_start_instruction(ins, __opword(OP_GAMIRR, exopc, xtype, IAA_RAM, 0,
  425. CB_IMA(hints)));
  426. }
  427. static inline void gru_gamer(void *cb, int exopc, unsigned long src,
  428. unsigned int xtype,
  429. unsigned long operand1, unsigned long operand2,
  430. unsigned long hints)
  431. {
  432. struct gru_instruction *ins = (void *)cb;
  433. ins->baddr0 = (long)src;
  434. ins->op1_stride = operand1;
  435. ins->op2_value_baddr1 = operand2;
  436. gru_start_instruction(ins, __opword(OP_GAMER, exopc, xtype, IAA_RAM, 0,
  437. CB_IMA(hints)));
  438. }
  439. static inline void gru_gamerr(void *cb, int exopc, unsigned long src,
  440. unsigned int xtype, unsigned long operand1,
  441. unsigned long operand2, unsigned long hints)
  442. {
  443. struct gru_instruction *ins = (void *)cb;
  444. ins->baddr0 = (long)src;
  445. ins->op1_stride = operand1;
  446. ins->op2_value_baddr1 = operand2;
  447. gru_start_instruction(ins, __opword(OP_GAMERR, exopc, xtype, IAA_RAM, 0,
  448. CB_IMA(hints)));
  449. }
  450. static inline void gru_gamxr(void *cb, unsigned long src,
  451. unsigned int tri0, unsigned long hints)
  452. {
  453. struct gru_instruction *ins = (void *)cb;
  454. ins->baddr0 = (long)src;
  455. ins->nelem = 4;
  456. gru_start_instruction(ins, __opword(OP_GAMXR, EOP_XR_CSWAP, XTYPE_DW,
  457. IAA_RAM, 0, CB_IMA(hints)));
  458. }
  459. static inline void gru_mesq(void *cb, unsigned long queue,
  460. unsigned long tri0, unsigned long nelem,
  461. unsigned long hints)
  462. {
  463. struct gru_instruction *ins = (void *)cb;
  464. ins->baddr0 = (long)queue;
  465. ins->nelem = nelem;
  466. ins->tri0 = tri0;
  467. gru_start_instruction(ins, __opword(OP_MESQ, 0, XTYPE_CL, IAA_RAM, 0,
  468. CB_IMA(hints)));
  469. }
  470. static inline unsigned long gru_get_amo_value(void *cb)
  471. {
  472. struct gru_instruction *ins = (void *)cb;
  473. return ins->avalue;
  474. }
  475. static inline int gru_get_amo_value_head(void *cb)
  476. {
  477. struct gru_instruction *ins = (void *)cb;
  478. return ins->avalue & 0xffffffff;
  479. }
  480. static inline int gru_get_amo_value_limit(void *cb)
  481. {
  482. struct gru_instruction *ins = (void *)cb;
  483. return ins->avalue >> 32;
  484. }
  485. static inline union gru_mesqhead gru_mesq_head(int head, int limit)
  486. {
  487. union gru_mesqhead mqh;
  488. mqh.head = head;
  489. mqh.limit = limit;
  490. return mqh;
  491. }
  492. /*
  493. * Get struct control_block_extended_exc_detail for CB.
  494. */
  495. extern int gru_get_cb_exception_detail(void *cb,
  496. struct control_block_extended_exc_detail *excdet);
  497. #define GRU_EXC_STR_SIZE 256
  498. /*
  499. * Control block definition for checking status
  500. */
  501. struct gru_control_block_status {
  502. unsigned int icmd :1;
  503. unsigned int ima :3;
  504. unsigned int reserved0 :4;
  505. unsigned int unused1 :24;
  506. unsigned int unused2 :24;
  507. unsigned int istatus :2;
  508. unsigned int isubstatus :4;
  509. unsigned int unused3 :2;
  510. };
  511. /* Get CB status */
  512. static inline int gru_get_cb_status(void *cb)
  513. {
  514. struct gru_control_block_status *cbs = (void *)cb;
  515. return cbs->istatus;
  516. }
  517. /* Get CB message queue substatus */
  518. static inline int gru_get_cb_message_queue_substatus(void *cb)
  519. {
  520. struct gru_control_block_status *cbs = (void *)cb;
  521. return cbs->isubstatus & CBSS_MSG_QUEUE_MASK;
  522. }
  523. /* Get CB substatus */
  524. static inline int gru_get_cb_substatus(void *cb)
  525. {
  526. struct gru_control_block_status *cbs = (void *)cb;
  527. return cbs->isubstatus;
  528. }
  529. /* Check the status of a CB. If the CB is in UPM mode, call the
  530. * OS to handle the UPM status.
  531. * Returns the CB status field value (0 for normal completion)
  532. */
  533. static inline int gru_check_status(void *cb)
  534. {
  535. struct gru_control_block_status *cbs = (void *)cb;
  536. int ret;
  537. ret = cbs->istatus;
  538. if (ret == CBS_CALL_OS)
  539. ret = gru_check_status_proc(cb);
  540. return ret;
  541. }
  542. /* Wait for CB to complete.
  543. * Returns the CB status field value (0 for normal completion)
  544. */
  545. static inline int gru_wait(void *cb)
  546. {
  547. struct gru_control_block_status *cbs = (void *)cb;
  548. int ret = cbs->istatus;
  549. if (ret != CBS_IDLE)
  550. ret = gru_wait_proc(cb);
  551. return ret;
  552. }
  553. /* Wait for CB to complete. Aborts program if error. (Note: error does NOT
  554. * mean TLB mis - only fatal errors such as memory parity error or user
  555. * bugs will cause termination.
  556. */
  557. static inline void gru_wait_abort(void *cb)
  558. {
  559. struct gru_control_block_status *cbs = (void *)cb;
  560. if (cbs->istatus != CBS_IDLE)
  561. gru_wait_abort_proc(cb);
  562. }
  563. /*
  564. * Get a pointer to a control block
  565. * gseg - GSeg address returned from gru_get_thread_gru_segment()
  566. * index - index of desired CB
  567. */
  568. static inline void *gru_get_cb_pointer(void *gseg,
  569. int index)
  570. {
  571. return gseg + GRU_CB_BASE + index * GRU_HANDLE_STRIDE;
  572. }
  573. /*
  574. * Get a pointer to a cacheline in the data segment portion of a GSeg
  575. * gseg - GSeg address returned from gru_get_thread_gru_segment()
  576. * index - index of desired cache line
  577. */
  578. static inline void *gru_get_data_pointer(void *gseg, int index)
  579. {
  580. return gseg + GRU_DS_BASE + index * GRU_CACHE_LINE_BYTES;
  581. }
  582. /*
  583. * Convert a vaddr into the tri index within the GSEG
  584. * vaddr - virtual address of within gseg
  585. */
  586. static inline int gru_get_tri(void *vaddr)
  587. {
  588. return ((unsigned long)vaddr & (GRU_GSEG_PAGESIZE - 1)) - GRU_DS_BASE;
  589. }
  590. #endif /* __GRU_INSTRUCTIONS_H__ */