1 /* Target-dependent code for the Toshiba MeP for GDB, the GNU debugger.
3 Copyright (C) 2001-2019 Free Software Foundation, Inc.
5 Contributed by Red Hat, Inc.
7 This file is part of GDB.
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 3 of the License, or
12 (at your option) any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program. If not, see <http://www.gnu.org/licenses/>. */
24 #include "frame-unwind.h"
25 #include "frame-base.h"
36 #include "arch-utils.h"
39 #include "sim-regno.h"
41 #include "trad-frame.h"
42 #include "reggroups.h"
45 #include "prologue-value.h"
46 #include "cgen/bitset.h"
49 /* Get the user's customized MeP coprocessor register names from
51 #include "../opcodes/mep-desc.h"
52 #include "../opcodes/mep-opc.h"
55 /* The gdbarch_tdep structure. */
57 /* A quick recap for GDB hackers not familiar with the whole Toshiba
58 Media Processor story:
60 The MeP media engine is a configureable processor: users can design
61 their own coprocessors, implement custom instructions, adjust cache
62 sizes, select optional standard facilities like add-and-saturate
63 instructions, and so on. Then, they can build custom versions of
64 the GNU toolchain to support their customized chips. The
65 MeP-Integrator program (see utils/mep) takes a GNU toolchain source
66 tree, and a config file pointing to various files provided by the
67 user describing their customizations, and edits the source tree to
68 produce a compiler that can generate their custom instructions, an
69 assembler that can assemble them and recognize their custom
70 register names, and so on.
72 Furthermore, the user can actually specify several of these custom
73 configurations, called 'me_modules', and get a toolchain which can
74 produce code for any of them, given a compiler/assembler switch;
75 you say something like 'gcc -mconfig=mm_max' to generate code for
76 the me_module named 'mm_max'.
78 GDB, in particular, needs to:
80 - use the coprocessor control register names provided by the user
81 in their hardware description, in expressions, 'info register'
82 output, and disassembly,
84 - know the number, names, and types of the coprocessor's
85 general-purpose registers, adjust the 'info all-registers' output
86 accordingly, and print error messages if the user refers to one
89 - allow access to the control bus space only when the configuration
90 actually has a control bus, and recognize which regions of the
91 control bus space are actually populated,
93 - disassemble using the user's provided mnemonics for their custom
96 - recognize whether the $hi and $lo registers are present, and
97 allow access to them only when they are actually there.
99 There are three sources of information about what sort of me_module
100 we're actually dealing with:
102 - A MeP executable file indicates which me_module it was compiled
103 for, and libopcodes has tables describing each module. So, given
104 an executable file, we can find out about the processor it was
107 - There are SID command-line options to select a particular
108 me_module, overriding the one specified in the ELF file. SID
109 provides GDB with a fake read-only register, 'module', which
110 indicates which me_module GDB is communicating with an instance
113 - There are SID command-line options to enable or disable certain
114 optional processor features, overriding the defaults for the
115 selected me_module. The MeP $OPT register indicates which
116 options are present on the current processor. */
121 /* A CGEN cpu descriptor for this BFD architecture and machine.
123 Note: this is *not* customized for any particular me_module; the
124 MeP libopcodes machinery actually puts off module-specific
125 customization until the last minute. So this contains
126 information about all supported me_modules. */
127 CGEN_CPU_DESC cpu_desc;
129 /* The me_module index from the ELF file we used to select this
130 architecture, or CONFIG_NONE if there was none.
132 Note that we should prefer to use the me_module number available
133 via the 'module' register, whenever we're actually talking to a
136 In the absence of live information, we'd like to get the
137 me_module number from the ELF file. But which ELF file: the
138 executable file, the core file, ... ? The answer is, "the last
139 ELF file we used to set the current architecture". Thus, we
140 create a separate instance of the gdbarch structure for each
141 me_module value mep_gdbarch_init sees, and store the me_module
142 value from the ELF file here. */
143 CONFIG_ATTR me_module;
148 /* Getting me_module information from the CGEN tables. */
151 /* Find an entry in the DESC's hardware table whose name begins with
152 PREFIX, and whose ISA mask intersects COPRO_ISA_MASK, but does not
153 intersect with GENERIC_ISA_MASK. If there is no matching entry,
155 static const CGEN_HW_ENTRY *
156 find_hw_entry_by_prefix_and_isa (CGEN_CPU_DESC desc,
158 CGEN_BITSET *copro_isa_mask,
159 CGEN_BITSET *generic_isa_mask)
161 int prefix_len = strlen (prefix);
164 for (i = 0; i < desc->hw_table.num_entries; i++)
166 const CGEN_HW_ENTRY *hw = desc->hw_table.entries[i];
167 if (strncmp (prefix, hw->name, prefix_len) == 0)
169 CGEN_BITSET *hw_isa_mask
171 &CGEN_ATTR_CGEN_HW_ISA_VALUE (CGEN_HW_ATTRS (hw)));
173 if (cgen_bitset_intersect_p (hw_isa_mask, copro_isa_mask)
174 && ! cgen_bitset_intersect_p (hw_isa_mask, generic_isa_mask))
183 /* Find an entry in DESC's hardware table whose type is TYPE. Return
184 zero if there is none. */
185 static const CGEN_HW_ENTRY *
186 find_hw_entry_by_type (CGEN_CPU_DESC desc, CGEN_HW_TYPE type)
190 for (i = 0; i < desc->hw_table.num_entries; i++)
192 const CGEN_HW_ENTRY *hw = desc->hw_table.entries[i];
194 if (hw->type == type)
202 /* Return the CGEN hardware table entry for the coprocessor register
203 set for ME_MODULE, whose name prefix is PREFIX. If ME_MODULE has
204 no such register set, return zero. If ME_MODULE is the generic
205 me_module CONFIG_NONE, return the table entry for the register set
206 whose hardware type is GENERIC_TYPE. */
207 static const CGEN_HW_ENTRY *
208 me_module_register_set (CONFIG_ATTR me_module,
210 CGEN_HW_TYPE generic_type)
212 /* This is kind of tricky, because the hardware table is constructed
213 in a way that isn't very helpful. Perhaps we can fix that, but
214 here's how it works at the moment:
216 The configuration map, `mep_config_map', is indexed by me_module
217 number, and indicates which coprocessor and core ISAs that
218 me_module supports. The 'core_isa' mask includes all the core
219 ISAs, and the 'cop_isa' mask includes all the coprocessor ISAs.
220 The entry for the generic me_module, CONFIG_NONE, has an empty
221 'cop_isa', and its 'core_isa' selects only the standard MeP
224 The CGEN CPU descriptor's hardware table, desc->hw_table, has
225 entries for all the register sets, for all me_modules. Each
226 entry has a mask indicating which ISAs use that register set.
227 So, if an me_module supports some coprocessor ISA, we can find
228 applicable register sets by scanning the hardware table for
229 register sets whose masks include (at least some of) those ISAs.
231 Each hardware table entry also has a name, whose prefix says
232 whether it's a general-purpose ("h-cr") or control ("h-ccr")
233 coprocessor register set. It might be nicer to have an attribute
234 indicating what sort of register set it was, that we could use
235 instead of pattern-matching on the name.
237 When there is no hardware table entry whose mask includes a
238 particular coprocessor ISA and whose name starts with a given
239 prefix, then that means that that coprocessor doesn't have any
240 registers of that type. In such cases, this function must return
243 Coprocessor register sets' masks may or may not include the core
244 ISA for the me_module they belong to. Those generated by a2cgen
245 do, but the sample me_module included in the unconfigured tree,
248 There are generic coprocessor register sets, intended only for
249 use with the generic me_module. Unfortunately, their masks
250 include *all* ISAs --- even those for coprocessors that don't
251 have such register sets. This makes detecting the case where a
252 coprocessor lacks a particular register set more complicated.
254 So, here's the approach we take:
256 - For CONFIG_NONE, we return the generic coprocessor register set.
258 - For any other me_module, we search for a register set whose
259 mask contains any of the me_module's coprocessor ISAs,
260 specifically excluding the generic coprocessor register sets. */
262 CGEN_CPU_DESC desc = gdbarch_tdep (target_gdbarch ())->cpu_desc;
263 const CGEN_HW_ENTRY *hw;
265 if (me_module == CONFIG_NONE)
266 hw = find_hw_entry_by_type (desc, generic_type);
269 CGEN_BITSET *cop = &mep_config_map[me_module].cop_isa;
270 CGEN_BITSET *core = &mep_config_map[me_module].core_isa;
271 CGEN_BITSET *generic = &mep_config_map[CONFIG_NONE].core_isa;
272 CGEN_BITSET *cop_and_core;
274 /* The coprocessor ISAs include the ISA for the specific core which
275 has that coprocessor. */
276 cop_and_core = cgen_bitset_copy (cop);
277 cgen_bitset_union (cop, core, cop_and_core);
278 hw = find_hw_entry_by_prefix_and_isa (desc, prefix, cop_and_core, generic);
285 /* Given a hardware table entry HW representing a register set, return
286 a pointer to the keyword table with all the register names. If HW
287 is NULL, return NULL, to propage the "no such register set" info
289 static CGEN_KEYWORD *
290 register_set_keyword_table (const CGEN_HW_ENTRY *hw)
295 /* Check that HW is actually a keyword table. */
296 gdb_assert (hw->asm_type == CGEN_ASM_KEYWORD);
298 /* The 'asm_data' field of a register set's hardware table entry
299 refers to a keyword table. */
300 return (CGEN_KEYWORD *) hw->asm_data;
304 /* Given a keyword table KEYWORD and a register number REGNUM, return
305 the name of the register, or "" if KEYWORD contains no register
306 whose number is REGNUM. */
308 register_name_from_keyword (CGEN_KEYWORD *keyword_table, int regnum)
310 const CGEN_KEYWORD_ENTRY *entry
311 = cgen_keyword_lookup_value (keyword_table, regnum);
315 char *name = entry->name;
317 /* The CGEN keyword entries for register names include the
318 leading $, which appears in MeP assembly as well as in GDB.
319 But we don't want to return that; GDB core code adds that
331 /* Masks for option bits in the OPT special-purpose register. */
333 MEP_OPT_DIV = 1 << 25, /* 32-bit divide instruction option */
334 MEP_OPT_MUL = 1 << 24, /* 32-bit multiply instruction option */
335 MEP_OPT_BIT = 1 << 23, /* bit manipulation instruction option */
336 MEP_OPT_SAT = 1 << 22, /* saturation instruction option */
337 MEP_OPT_CLP = 1 << 21, /* clip instruction option */
338 MEP_OPT_MIN = 1 << 20, /* min/max instruction option */
339 MEP_OPT_AVE = 1 << 19, /* average instruction option */
340 MEP_OPT_ABS = 1 << 18, /* absolute difference instruction option */
341 MEP_OPT_LDZ = 1 << 16, /* leading zero instruction option */
342 MEP_OPT_VL64 = 1 << 6, /* 64-bit VLIW operation mode option */
343 MEP_OPT_VL32 = 1 << 5, /* 32-bit VLIW operation mode option */
344 MEP_OPT_COP = 1 << 4, /* coprocessor option */
345 MEP_OPT_DSP = 1 << 2, /* DSP option */
346 MEP_OPT_UCI = 1 << 1, /* UCI option */
347 MEP_OPT_DBG = 1 << 0, /* DBG function option */
351 /* Given the option_mask value for a particular entry in
352 mep_config_map, produce the value the processor's OPT register
353 would use to represent the same set of options. */
355 opt_from_option_mask (unsigned int option_mask)
357 /* A table mapping OPT register bits onto CGEN config map option
360 unsigned int opt_bit, option_mask_bit;
362 { MEP_OPT_DIV, 1 << CGEN_INSN_OPTIONAL_DIV_INSN },
363 { MEP_OPT_MUL, 1 << CGEN_INSN_OPTIONAL_MUL_INSN },
364 { MEP_OPT_DIV, 1 << CGEN_INSN_OPTIONAL_DIV_INSN },
365 { MEP_OPT_DBG, 1 << CGEN_INSN_OPTIONAL_DEBUG_INSN },
366 { MEP_OPT_LDZ, 1 << CGEN_INSN_OPTIONAL_LDZ_INSN },
367 { MEP_OPT_ABS, 1 << CGEN_INSN_OPTIONAL_ABS_INSN },
368 { MEP_OPT_AVE, 1 << CGEN_INSN_OPTIONAL_AVE_INSN },
369 { MEP_OPT_MIN, 1 << CGEN_INSN_OPTIONAL_MINMAX_INSN },
370 { MEP_OPT_CLP, 1 << CGEN_INSN_OPTIONAL_CLIP_INSN },
371 { MEP_OPT_SAT, 1 << CGEN_INSN_OPTIONAL_SAT_INSN },
372 { MEP_OPT_UCI, 1 << CGEN_INSN_OPTIONAL_UCI_INSN },
373 { MEP_OPT_DSP, 1 << CGEN_INSN_OPTIONAL_DSP_INSN },
374 { MEP_OPT_COP, 1 << CGEN_INSN_OPTIONAL_CP_INSN },
378 unsigned int opt = 0;
380 for (i = 0; i < (sizeof (bits) / sizeof (bits[0])); i++)
381 if (option_mask & bits[i].option_mask_bit)
382 opt |= bits[i].opt_bit;
388 /* Return the value the $OPT register would use to represent the set
389 of options for ME_MODULE. */
391 me_module_opt (CONFIG_ATTR me_module)
393 return opt_from_option_mask (mep_config_map[me_module].option_mask);
397 /* Return the width of ME_MODULE's coprocessor data bus, in bits.
398 This is either 32 or 64. */
400 me_module_cop_data_bus_width (CONFIG_ATTR me_module)
402 if (mep_config_map[me_module].option_mask
403 & (1 << CGEN_INSN_OPTIONAL_CP64_INSN))
410 /* Return true if ME_MODULE is big-endian, false otherwise. */
412 me_module_big_endian (CONFIG_ATTR me_module)
414 return mep_config_map[me_module].big_endian;
418 /* Return the name of ME_MODULE, or NULL if it has no name. */
420 me_module_name (CONFIG_ATTR me_module)
422 /* The default me_module has "" as its name, but it's easier for our
423 callers to test for NULL. */
424 if (! mep_config_map[me_module].name
425 || mep_config_map[me_module].name[0] == '\0')
428 return mep_config_map[me_module].name;
434 /* The MeP spec defines the following registers:
435 16 general purpose registers (r0-r15)
436 32 control/special registers (csr0-csr31)
437 32 coprocessor general-purpose registers (c0 -- c31)
438 64 coprocessor control registers (ccr0 -- ccr63)
440 For the raw registers, we assign numbers here explicitly, instead
441 of letting the enum assign them for us; the numbers are a matter of
442 external protocol, and shouldn't shift around as things are edited.
444 We access the control/special registers via pseudoregisters, to
445 enforce read-only portions that some registers have.
447 We access the coprocessor general purpose and control registers via
448 pseudoregisters, to make sure they appear in the proper order in
449 the 'info all-registers' command (which uses the register number
450 ordering), and also to allow them to be renamed and resized
451 depending on the me_module in use.
453 The MeP allows coprocessor general-purpose registers to be either
454 32 or 64 bits long, depending on the configuration. Since we don't
455 want the format of the 'g' packet to vary from one core to another,
456 the raw coprocessor GPRs are always 64 bits. GDB doesn't allow the
457 types of registers to change (see the implementation of
458 register_type), so we have four banks of pseudoregisters for the
459 coprocessor gprs --- 32-bit vs. 64-bit, and integer
460 vs. floating-point --- and we show or hide them depending on the
464 MEP_FIRST_RAW_REGNUM = 0,
466 MEP_FIRST_GPR_REGNUM = 0,
480 MEP_FP_REGNUM = MEP_R8_REGNUM,
482 MEP_TP_REGNUM = MEP_R13_REGNUM, /* (r13) Tiny data pointer */
484 MEP_GP_REGNUM = MEP_R14_REGNUM, /* (r14) Global pointer */
486 MEP_SP_REGNUM = MEP_R15_REGNUM, /* (r15) Stack pointer */
487 MEP_LAST_GPR_REGNUM = MEP_R15_REGNUM,
489 /* The raw control registers. These are the values as received via
490 the remote protocol, directly from the target; we only let user
491 code touch the via the pseudoregisters, which enforce read-only
493 MEP_FIRST_RAW_CSR_REGNUM = 16,
494 MEP_RAW_PC_REGNUM = 16, /* Program counter */
495 MEP_RAW_LP_REGNUM = 17, /* Link pointer */
496 MEP_RAW_SAR_REGNUM = 18, /* Raw shift amount */
497 MEP_RAW_CSR3_REGNUM = 19, /* csr3: reserved */
498 MEP_RAW_RPB_REGNUM = 20, /* Raw repeat begin address */
499 MEP_RAW_RPE_REGNUM = 21, /* Repeat end address */
500 MEP_RAW_RPC_REGNUM = 22, /* Repeat count */
501 MEP_RAW_HI_REGNUM = 23, /* Upper 32 bits of result of 64 bit mult/div */
502 MEP_RAW_LO_REGNUM = 24, /* Lower 32 bits of result of 64 bit mult/div */
503 MEP_RAW_CSR9_REGNUM = 25, /* csr3: reserved */
504 MEP_RAW_CSR10_REGNUM = 26, /* csr3: reserved */
505 MEP_RAW_CSR11_REGNUM = 27, /* csr3: reserved */
506 MEP_RAW_MB0_REGNUM = 28, /* Raw modulo begin address 0 */
507 MEP_RAW_ME0_REGNUM = 29, /* Raw modulo end address 0 */
508 MEP_RAW_MB1_REGNUM = 30, /* Raw modulo begin address 1 */
509 MEP_RAW_ME1_REGNUM = 31, /* Raw modulo end address 1 */
510 MEP_RAW_PSW_REGNUM = 32, /* Raw program status word */
511 MEP_RAW_ID_REGNUM = 33, /* Raw processor ID/revision */
512 MEP_RAW_TMP_REGNUM = 34, /* Temporary */
513 MEP_RAW_EPC_REGNUM = 35, /* Exception program counter */
514 MEP_RAW_EXC_REGNUM = 36, /* Raw exception cause */
515 MEP_RAW_CFG_REGNUM = 37, /* Raw processor configuration*/
516 MEP_RAW_CSR22_REGNUM = 38, /* csr3: reserved */
517 MEP_RAW_NPC_REGNUM = 39, /* Nonmaskable interrupt PC */
518 MEP_RAW_DBG_REGNUM = 40, /* Raw debug */
519 MEP_RAW_DEPC_REGNUM = 41, /* Debug exception PC */
520 MEP_RAW_OPT_REGNUM = 42, /* Raw options */
521 MEP_RAW_RCFG_REGNUM = 43, /* Raw local ram config */
522 MEP_RAW_CCFG_REGNUM = 44, /* Raw cache config */
523 MEP_RAW_CSR29_REGNUM = 45, /* csr3: reserved */
524 MEP_RAW_CSR30_REGNUM = 46, /* csr3: reserved */
525 MEP_RAW_CSR31_REGNUM = 47, /* csr3: reserved */
526 MEP_LAST_RAW_CSR_REGNUM = MEP_RAW_CSR31_REGNUM,
528 /* The raw coprocessor general-purpose registers. These are all 64
530 MEP_FIRST_RAW_CR_REGNUM = 48,
531 MEP_LAST_RAW_CR_REGNUM = MEP_FIRST_RAW_CR_REGNUM + 31,
533 MEP_FIRST_RAW_CCR_REGNUM = 80,
534 MEP_LAST_RAW_CCR_REGNUM = MEP_FIRST_RAW_CCR_REGNUM + 63,
536 /* The module number register. This is the index of the me_module
537 of which the current target is an instance. (This is not a real
538 MeP-specified register; it's provided by SID.) */
541 MEP_LAST_RAW_REGNUM = MEP_MODULE_REGNUM,
543 MEP_NUM_RAW_REGS = MEP_LAST_RAW_REGNUM + 1,
545 /* Pseudoregisters. See mep_pseudo_register_read and
546 mep_pseudo_register_write. */
547 MEP_FIRST_PSEUDO_REGNUM = MEP_NUM_RAW_REGS,
549 /* We have a pseudoregister for every control/special register, to
550 implement registers with read-only bits. */
551 MEP_FIRST_CSR_REGNUM = MEP_FIRST_PSEUDO_REGNUM,
552 MEP_PC_REGNUM = MEP_FIRST_CSR_REGNUM, /* Program counter */
553 MEP_LP_REGNUM, /* Link pointer */
554 MEP_SAR_REGNUM, /* shift amount */
555 MEP_CSR3_REGNUM, /* csr3: reserved */
556 MEP_RPB_REGNUM, /* repeat begin address */
557 MEP_RPE_REGNUM, /* Repeat end address */
558 MEP_RPC_REGNUM, /* Repeat count */
559 MEP_HI_REGNUM, /* Upper 32 bits of the result of 64 bit mult/div */
560 MEP_LO_REGNUM, /* Lower 32 bits of the result of 64 bit mult/div */
561 MEP_CSR9_REGNUM, /* csr3: reserved */
562 MEP_CSR10_REGNUM, /* csr3: reserved */
563 MEP_CSR11_REGNUM, /* csr3: reserved */
564 MEP_MB0_REGNUM, /* modulo begin address 0 */
565 MEP_ME0_REGNUM, /* modulo end address 0 */
566 MEP_MB1_REGNUM, /* modulo begin address 1 */
567 MEP_ME1_REGNUM, /* modulo end address 1 */
568 MEP_PSW_REGNUM, /* program status word */
569 MEP_ID_REGNUM, /* processor ID/revision */
570 MEP_TMP_REGNUM, /* Temporary */
571 MEP_EPC_REGNUM, /* Exception program counter */
572 MEP_EXC_REGNUM, /* exception cause */
573 MEP_CFG_REGNUM, /* processor configuration*/
574 MEP_CSR22_REGNUM, /* csr3: reserved */
575 MEP_NPC_REGNUM, /* Nonmaskable interrupt PC */
576 MEP_DBG_REGNUM, /* debug */
577 MEP_DEPC_REGNUM, /* Debug exception PC */
578 MEP_OPT_REGNUM, /* options */
579 MEP_RCFG_REGNUM, /* local ram config */
580 MEP_CCFG_REGNUM, /* cache config */
581 MEP_CSR29_REGNUM, /* csr3: reserved */
582 MEP_CSR30_REGNUM, /* csr3: reserved */
583 MEP_CSR31_REGNUM, /* csr3: reserved */
584 MEP_LAST_CSR_REGNUM = MEP_CSR31_REGNUM,
586 /* The 32-bit integer view of the coprocessor GPR's. */
587 MEP_FIRST_CR32_REGNUM,
588 MEP_LAST_CR32_REGNUM = MEP_FIRST_CR32_REGNUM + 31,
590 /* The 32-bit floating-point view of the coprocessor GPR's. */
591 MEP_FIRST_FP_CR32_REGNUM,
592 MEP_LAST_FP_CR32_REGNUM = MEP_FIRST_FP_CR32_REGNUM + 31,
594 /* The 64-bit integer view of the coprocessor GPR's. */
595 MEP_FIRST_CR64_REGNUM,
596 MEP_LAST_CR64_REGNUM = MEP_FIRST_CR64_REGNUM + 31,
598 /* The 64-bit floating-point view of the coprocessor GPR's. */
599 MEP_FIRST_FP_CR64_REGNUM,
600 MEP_LAST_FP_CR64_REGNUM = MEP_FIRST_FP_CR64_REGNUM + 31,
602 MEP_FIRST_CCR_REGNUM,
603 MEP_LAST_CCR_REGNUM = MEP_FIRST_CCR_REGNUM + 63,
605 MEP_LAST_PSEUDO_REGNUM = MEP_LAST_CCR_REGNUM,
607 MEP_NUM_PSEUDO_REGS = (MEP_LAST_PSEUDO_REGNUM - MEP_LAST_RAW_REGNUM),
609 MEP_NUM_REGS = MEP_NUM_RAW_REGS + MEP_NUM_PSEUDO_REGS
613 #define IN_SET(set, n) \
614 (MEP_FIRST_ ## set ## _REGNUM <= (n) && (n) <= MEP_LAST_ ## set ## _REGNUM)
616 #define IS_GPR_REGNUM(n) (IN_SET (GPR, (n)))
617 #define IS_RAW_CSR_REGNUM(n) (IN_SET (RAW_CSR, (n)))
618 #define IS_RAW_CR_REGNUM(n) (IN_SET (RAW_CR, (n)))
619 #define IS_RAW_CCR_REGNUM(n) (IN_SET (RAW_CCR, (n)))
621 #define IS_CSR_REGNUM(n) (IN_SET (CSR, (n)))
622 #define IS_CR32_REGNUM(n) (IN_SET (CR32, (n)))
623 #define IS_FP_CR32_REGNUM(n) (IN_SET (FP_CR32, (n)))
624 #define IS_CR64_REGNUM(n) (IN_SET (CR64, (n)))
625 #define IS_FP_CR64_REGNUM(n) (IN_SET (FP_CR64, (n)))
626 #define IS_CR_REGNUM(n) (IS_CR32_REGNUM (n) || IS_FP_CR32_REGNUM (n) \
627 || IS_CR64_REGNUM (n) || IS_FP_CR64_REGNUM (n))
628 #define IS_CCR_REGNUM(n) (IN_SET (CCR, (n)))
630 #define IS_RAW_REGNUM(n) (IN_SET (RAW, (n)))
631 #define IS_PSEUDO_REGNUM(n) (IN_SET (PSEUDO, (n)))
633 #define NUM_REGS_IN_SET(set) \
634 (MEP_LAST_ ## set ## _REGNUM - MEP_FIRST_ ## set ## _REGNUM + 1)
636 #define MEP_GPR_SIZE (4) /* Size of a MeP general-purpose register. */
637 #define MEP_PSW_SIZE (4) /* Size of the PSW register. */
638 #define MEP_LP_SIZE (4) /* Size of the LP register. */
641 /* Many of the control/special registers contain bits that cannot be
642 written to; some are entirely read-only. So we present them all as
645 The following table describes the special properties of each CSR. */
646 struct mep_csr_register
648 /* The number of this CSR's raw register. */
651 /* The number of this CSR's pseudoregister. */
654 /* A mask of the bits that are writeable: if a bit is set here, then
655 it can be modified; if the bit is clear, then it cannot. */
656 LONGEST writeable_bits;
660 /* mep_csr_registers[i] describes the i'th CSR.
661 We just list the register numbers here explicitly to help catch
663 #define CSR(name) MEP_RAW_ ## name ## _REGNUM, MEP_ ## name ## _REGNUM
664 struct mep_csr_register mep_csr_registers[] = {
665 { CSR(PC), 0xffffffff }, /* manual says r/o, but we can write it */
666 { CSR(LP), 0xffffffff },
667 { CSR(SAR), 0x0000003f },
668 { CSR(CSR3), 0xffffffff },
669 { CSR(RPB), 0xfffffffe },
670 { CSR(RPE), 0xffffffff },
671 { CSR(RPC), 0xffffffff },
672 { CSR(HI), 0xffffffff },
673 { CSR(LO), 0xffffffff },
674 { CSR(CSR9), 0xffffffff },
675 { CSR(CSR10), 0xffffffff },
676 { CSR(CSR11), 0xffffffff },
677 { CSR(MB0), 0x0000ffff },
678 { CSR(ME0), 0x0000ffff },
679 { CSR(MB1), 0x0000ffff },
680 { CSR(ME1), 0x0000ffff },
681 { CSR(PSW), 0x000003ff },
682 { CSR(ID), 0x00000000 },
683 { CSR(TMP), 0xffffffff },
684 { CSR(EPC), 0xffffffff },
685 { CSR(EXC), 0x000030f0 },
686 { CSR(CFG), 0x00c0001b },
687 { CSR(CSR22), 0xffffffff },
688 { CSR(NPC), 0xffffffff },
689 { CSR(DBG), 0x00000580 },
690 { CSR(DEPC), 0xffffffff },
691 { CSR(OPT), 0x00000000 },
692 { CSR(RCFG), 0x00000000 },
693 { CSR(CCFG), 0x00000000 },
694 { CSR(CSR29), 0xffffffff },
695 { CSR(CSR30), 0xffffffff },
696 { CSR(CSR31), 0xffffffff },
700 /* If R is the number of a raw register, then mep_raw_to_pseudo[R] is
701 the number of the corresponding pseudoregister. Otherwise,
702 mep_raw_to_pseudo[R] == R. */
703 static int mep_raw_to_pseudo[MEP_NUM_REGS];
705 /* If R is the number of a pseudoregister, then mep_pseudo_to_raw[R]
706 is the number of the underlying raw register. Otherwise
707 mep_pseudo_to_raw[R] == R. */
708 static int mep_pseudo_to_raw[MEP_NUM_REGS];
711 mep_init_pseudoregister_maps (void)
715 /* Verify that mep_csr_registers covers all the CSRs, in order. */
716 gdb_assert (ARRAY_SIZE (mep_csr_registers) == NUM_REGS_IN_SET (CSR));
717 gdb_assert (ARRAY_SIZE (mep_csr_registers) == NUM_REGS_IN_SET (RAW_CSR));
719 /* Verify that the raw and pseudo ranges have matching sizes. */
720 gdb_assert (NUM_REGS_IN_SET (RAW_CSR) == NUM_REGS_IN_SET (CSR));
721 gdb_assert (NUM_REGS_IN_SET (RAW_CR) == NUM_REGS_IN_SET (CR32));
722 gdb_assert (NUM_REGS_IN_SET (RAW_CR) == NUM_REGS_IN_SET (CR64));
723 gdb_assert (NUM_REGS_IN_SET (RAW_CCR) == NUM_REGS_IN_SET (CCR));
725 for (i = 0; i < ARRAY_SIZE (mep_csr_registers); i++)
727 struct mep_csr_register *r = &mep_csr_registers[i];
729 gdb_assert (r->pseudo == MEP_FIRST_CSR_REGNUM + i);
730 gdb_assert (r->raw == MEP_FIRST_RAW_CSR_REGNUM + i);
733 /* Set up the initial raw<->pseudo mappings. */
734 for (i = 0; i < MEP_NUM_REGS; i++)
736 mep_raw_to_pseudo[i] = i;
737 mep_pseudo_to_raw[i] = i;
740 /* Add the CSR raw<->pseudo mappings. */
741 for (i = 0; i < ARRAY_SIZE (mep_csr_registers); i++)
743 struct mep_csr_register *r = &mep_csr_registers[i];
745 mep_raw_to_pseudo[r->raw] = r->pseudo;
746 mep_pseudo_to_raw[r->pseudo] = r->raw;
749 /* Add the CR raw<->pseudo mappings. */
750 for (i = 0; i < NUM_REGS_IN_SET (RAW_CR); i++)
752 int raw = MEP_FIRST_RAW_CR_REGNUM + i;
753 int pseudo32 = MEP_FIRST_CR32_REGNUM + i;
754 int pseudofp32 = MEP_FIRST_FP_CR32_REGNUM + i;
755 int pseudo64 = MEP_FIRST_CR64_REGNUM + i;
756 int pseudofp64 = MEP_FIRST_FP_CR64_REGNUM + i;
758 /* Truly, the raw->pseudo mapping depends on the current module.
759 But we use the raw->pseudo mapping when we read the debugging
760 info; at that point, we don't know what module we'll actually
761 be running yet. So, we always supply the 64-bit register
762 numbers; GDB knows how to pick a smaller value out of a
763 larger register properly. */
764 mep_raw_to_pseudo[raw] = pseudo64;
765 mep_pseudo_to_raw[pseudo32] = raw;
766 mep_pseudo_to_raw[pseudofp32] = raw;
767 mep_pseudo_to_raw[pseudo64] = raw;
768 mep_pseudo_to_raw[pseudofp64] = raw;
771 /* Add the CCR raw<->pseudo mappings. */
772 for (i = 0; i < NUM_REGS_IN_SET (CCR); i++)
774 int raw = MEP_FIRST_RAW_CCR_REGNUM + i;
775 int pseudo = MEP_FIRST_CCR_REGNUM + i;
776 mep_raw_to_pseudo[raw] = pseudo;
777 mep_pseudo_to_raw[pseudo] = raw;
783 mep_debug_reg_to_regnum (struct gdbarch *gdbarch, int debug_reg)
785 /* The debug info uses the raw register numbers. */
786 if (debug_reg >= 0 && debug_reg < ARRAY_SIZE (mep_raw_to_pseudo))
787 return mep_raw_to_pseudo[debug_reg];
792 /* Return the size, in bits, of the coprocessor pseudoregister
795 mep_pseudo_cr_size (int pseudo)
797 if (IS_CR32_REGNUM (pseudo)
798 || IS_FP_CR32_REGNUM (pseudo))
800 else if (IS_CR64_REGNUM (pseudo)
801 || IS_FP_CR64_REGNUM (pseudo))
804 gdb_assert_not_reached ("unexpected coprocessor pseudo register");
808 /* If the coprocessor pseudoregister numbered PSEUDO is a
809 floating-point register, return non-zero; if it is an integer
810 register, return zero. */
812 mep_pseudo_cr_is_float (int pseudo)
814 return (IS_FP_CR32_REGNUM (pseudo)
815 || IS_FP_CR64_REGNUM (pseudo));
819 /* Given a coprocessor GPR pseudoregister number, return its index
820 within that register bank. */
822 mep_pseudo_cr_index (int pseudo)
824 if (IS_CR32_REGNUM (pseudo))
825 return pseudo - MEP_FIRST_CR32_REGNUM;
826 else if (IS_FP_CR32_REGNUM (pseudo))
827 return pseudo - MEP_FIRST_FP_CR32_REGNUM;
828 else if (IS_CR64_REGNUM (pseudo))
829 return pseudo - MEP_FIRST_CR64_REGNUM;
830 else if (IS_FP_CR64_REGNUM (pseudo))
831 return pseudo - MEP_FIRST_FP_CR64_REGNUM;
833 gdb_assert_not_reached ("unexpected coprocessor pseudo register");
837 /* Return the me_module index describing the current target.
839 If the current target has registers (e.g., simulator, remote
840 target), then this uses the value of the 'module' register, raw
841 register MEP_MODULE_REGNUM. Otherwise, this retrieves the value
842 from the ELF header's e_flags field of the current executable
845 current_me_module (void)
847 if (target_has_registers)
850 regcache_cooked_read_unsigned (get_current_regcache (),
851 MEP_MODULE_REGNUM, ®val);
852 return (CONFIG_ATTR) regval;
855 return gdbarch_tdep (target_gdbarch ())->me_module;
859 /* Return the set of options for the current target, in the form that
860 the OPT register would use.
862 If the current target has registers (e.g., simulator, remote
863 target), then this is the actual value of the OPT register. If the
864 current target does not have registers (e.g., an executable file),
865 then use the 'module_opt' field we computed when we build the
866 gdbarch object for this module. */
868 current_options (void)
870 if (target_has_registers)
873 regcache_cooked_read_unsigned (get_current_regcache (),
874 MEP_OPT_REGNUM, ®val);
878 return me_module_opt (current_me_module ());
882 /* Return the width of the current me_module's coprocessor data bus,
883 in bits. This is either 32 or 64. */
885 current_cop_data_bus_width (void)
887 return me_module_cop_data_bus_width (current_me_module ());
891 /* Return the keyword table of coprocessor general-purpose register
892 names appropriate for the me_module we're dealing with. */
893 static CGEN_KEYWORD *
894 current_cr_names (void)
896 const CGEN_HW_ENTRY *hw
897 = me_module_register_set (current_me_module (), "h-cr-", HW_H_CR);
899 return register_set_keyword_table (hw);
903 /* Return non-zero if the coprocessor general-purpose registers are
904 floating-point values, zero otherwise. */
906 current_cr_is_float (void)
908 const CGEN_HW_ENTRY *hw
909 = me_module_register_set (current_me_module (), "h-cr-", HW_H_CR);
911 return CGEN_ATTR_CGEN_HW_IS_FLOAT_VALUE (CGEN_HW_ATTRS (hw));
915 /* Return the keyword table of coprocessor control register names
916 appropriate for the me_module we're dealing with. */
917 static CGEN_KEYWORD *
918 current_ccr_names (void)
920 const CGEN_HW_ENTRY *hw
921 = me_module_register_set (current_me_module (), "h-ccr-", HW_H_CCR);
923 return register_set_keyword_table (hw);
928 mep_register_name (struct gdbarch *gdbarch, int regnr)
930 /* General-purpose registers. */
931 static const char *gpr_names[] = {
932 "r0", "r1", "r2", "r3", /* 0 */
933 "r4", "r5", "r6", "r7", /* 4 */
934 "fp", "r9", "r10", "r11", /* 8 */
935 "r12", "tp", "gp", "sp" /* 12 */
938 /* Special-purpose registers. */
939 static const char *csr_names[] = {
940 "pc", "lp", "sar", "", /* 0 csr3: reserved */
941 "rpb", "rpe", "rpc", "hi", /* 4 */
942 "lo", "", "", "", /* 8 csr9-csr11: reserved */
943 "mb0", "me0", "mb1", "me1", /* 12 */
945 "psw", "id", "tmp", "epc", /* 16 */
946 "exc", "cfg", "", "npc", /* 20 csr22: reserved */
947 "dbg", "depc", "opt", "rcfg", /* 24 */
948 "ccfg", "", "", "" /* 28 csr29-csr31: reserved */
951 if (IS_GPR_REGNUM (regnr))
952 return gpr_names[regnr - MEP_R0_REGNUM];
953 else if (IS_CSR_REGNUM (regnr))
955 /* The 'hi' and 'lo' registers are only present on processors
956 that have the 'MUL' or 'DIV' instructions enabled. */
957 if ((regnr == MEP_HI_REGNUM || regnr == MEP_LO_REGNUM)
958 && (! (current_options () & (MEP_OPT_MUL | MEP_OPT_DIV))))
961 return csr_names[regnr - MEP_FIRST_CSR_REGNUM];
963 else if (IS_CR_REGNUM (regnr))
969 /* Does this module have a coprocessor at all? */
970 if (! (current_options () & MEP_OPT_COP))
973 names = current_cr_names ();
975 /* This module's coprocessor has no general-purpose registers. */
978 cr_size = current_cop_data_bus_width ();
979 if (cr_size != mep_pseudo_cr_size (regnr))
980 /* This module's coprocessor's GPR's are of a different size. */
983 cr_is_float = current_cr_is_float ();
984 /* The extra ! operators ensure we get boolean equality, not
986 if (! cr_is_float != ! mep_pseudo_cr_is_float (regnr))
987 /* This module's coprocessor's GPR's are of a different type. */
990 return register_name_from_keyword (names, mep_pseudo_cr_index (regnr));
992 else if (IS_CCR_REGNUM (regnr))
994 /* Does this module have a coprocessor at all? */
995 if (! (current_options () & MEP_OPT_COP))
999 CGEN_KEYWORD *names = current_ccr_names ();
1002 /* This me_module's coprocessor has no control registers. */
1005 return register_name_from_keyword (names, regnr-MEP_FIRST_CCR_REGNUM);
1009 /* It might be nice to give the 'module' register a name, but that
1010 would affect the output of 'info all-registers', which would
1011 disturb the test suites. So we leave it invisible. */
1017 /* Custom register groups for the MeP. */
1018 static struct reggroup *mep_csr_reggroup; /* control/special */
1019 static struct reggroup *mep_cr_reggroup; /* coprocessor general-purpose */
1020 static struct reggroup *mep_ccr_reggroup; /* coprocessor control */
1024 mep_register_reggroup_p (struct gdbarch *gdbarch, int regnum,
1025 struct reggroup *group)
1027 /* Filter reserved or unused register numbers. */
1029 const char *name = mep_register_name (gdbarch, regnum);
1031 if (! name || name[0] == '\0')
1035 /* We could separate the GPRs and the CSRs. Toshiba has approved of
1036 the existing behavior, so we'd want to run that by them. */
1037 if (group == general_reggroup)
1038 return (IS_GPR_REGNUM (regnum)
1039 || IS_CSR_REGNUM (regnum));
1041 /* Everything is in the 'all' reggroup, except for the raw CSR's. */
1042 else if (group == all_reggroup)
1043 return (IS_GPR_REGNUM (regnum)
1044 || IS_CSR_REGNUM (regnum)
1045 || IS_CR_REGNUM (regnum)
1046 || IS_CCR_REGNUM (regnum));
1048 /* All registers should be saved and restored, except for the raw
1051 This is probably right if the coprocessor is something like a
1052 floating-point unit, but would be wrong if the coprocessor is
1053 something that does I/O, where register accesses actually cause
1054 externally-visible actions. But I get the impression that the
1055 coprocessor isn't supposed to do things like that --- you'd use a
1056 hardware engine, perhaps. */
1057 else if (group == save_reggroup || group == restore_reggroup)
1058 return (IS_GPR_REGNUM (regnum)
1059 || IS_CSR_REGNUM (regnum)
1060 || IS_CR_REGNUM (regnum)
1061 || IS_CCR_REGNUM (regnum));
1063 else if (group == mep_csr_reggroup)
1064 return IS_CSR_REGNUM (regnum);
1065 else if (group == mep_cr_reggroup)
1066 return IS_CR_REGNUM (regnum);
1067 else if (group == mep_ccr_reggroup)
1068 return IS_CCR_REGNUM (regnum);
1074 static struct type *
1075 mep_register_type (struct gdbarch *gdbarch, int reg_nr)
1077 /* Coprocessor general-purpose registers may be either 32 or 64 bits
1078 long. So for them, the raw registers are always 64 bits long (to
1079 keep the 'g' packet format fixed), and the pseudoregisters vary
1081 if (IS_RAW_CR_REGNUM (reg_nr))
1082 return builtin_type (gdbarch)->builtin_uint64;
1084 /* Since GDB doesn't allow registers to change type, we have two
1085 banks of pseudoregisters for the coprocessor general-purpose
1086 registers: one that gives a 32-bit view, and one that gives a
1087 64-bit view. We hide or show one or the other depending on the
1089 if (IS_CR_REGNUM (reg_nr))
1091 int size = mep_pseudo_cr_size (reg_nr);
1094 if (mep_pseudo_cr_is_float (reg_nr))
1095 return builtin_type (gdbarch)->builtin_float;
1097 return builtin_type (gdbarch)->builtin_uint32;
1099 else if (size == 64)
1101 if (mep_pseudo_cr_is_float (reg_nr))
1102 return builtin_type (gdbarch)->builtin_double;
1104 return builtin_type (gdbarch)->builtin_uint64;
1107 gdb_assert_not_reached ("unexpected cr size");
1110 /* All other registers are 32 bits long. */
1112 return builtin_type (gdbarch)->builtin_uint32;
1115 static enum register_status
1116 mep_pseudo_cr32_read (struct gdbarch *gdbarch,
1117 readable_regcache *regcache,
1121 enum register_status status;
1122 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
1123 /* Read the raw register into a 64-bit buffer, and then return the
1124 appropriate end of that buffer. */
1125 int rawnum = mep_pseudo_to_raw[cookednum];
1128 gdb_assert (TYPE_LENGTH (register_type (gdbarch, rawnum)) == sizeof (buf64));
1129 gdb_assert (TYPE_LENGTH (register_type (gdbarch, cookednum)) == 4);
1130 status = regcache->raw_read (rawnum, buf64);
1131 if (status == REG_VALID)
1133 /* Slow, but legible. */
1134 store_unsigned_integer (buf, 4, byte_order,
1135 extract_unsigned_integer (buf64, 8, byte_order));
1141 static enum register_status
1142 mep_pseudo_cr64_read (struct gdbarch *gdbarch,
1143 readable_regcache *regcache,
1147 return regcache->raw_read (mep_pseudo_to_raw[cookednum], buf);
1151 static enum register_status
1152 mep_pseudo_register_read (struct gdbarch *gdbarch,
1153 readable_regcache *regcache,
1157 if (IS_CSR_REGNUM (cookednum)
1158 || IS_CCR_REGNUM (cookednum))
1159 return regcache->raw_read (mep_pseudo_to_raw[cookednum], buf);
1160 else if (IS_CR32_REGNUM (cookednum)
1161 || IS_FP_CR32_REGNUM (cookednum))
1162 return mep_pseudo_cr32_read (gdbarch, regcache, cookednum, buf);
1163 else if (IS_CR64_REGNUM (cookednum)
1164 || IS_FP_CR64_REGNUM (cookednum))
1165 return mep_pseudo_cr64_read (gdbarch, regcache, cookednum, buf);
1167 gdb_assert_not_reached ("unexpected pseudo register");
1172 mep_pseudo_csr_write (struct gdbarch *gdbarch,
1173 struct regcache *regcache,
1175 const gdb_byte *buf)
1177 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
1178 int size = register_size (gdbarch, cookednum);
1179 struct mep_csr_register *r
1180 = &mep_csr_registers[cookednum - MEP_FIRST_CSR_REGNUM];
1182 if (r->writeable_bits == 0)
1183 /* A completely read-only register; avoid the read-modify-
1184 write cycle, and juts ignore the entire write. */
1188 /* A partially writeable register; do a read-modify-write cycle. */
1191 ULONGEST mixed_bits;
1193 regcache_raw_read_unsigned (regcache, r->raw, &old_bits);
1194 new_bits = extract_unsigned_integer (buf, size, byte_order);
1195 mixed_bits = ((r->writeable_bits & new_bits)
1196 | (~r->writeable_bits & old_bits));
1197 regcache_raw_write_unsigned (regcache, r->raw, mixed_bits);
1203 mep_pseudo_cr32_write (struct gdbarch *gdbarch,
1204 struct regcache *regcache,
1206 const gdb_byte *buf)
1208 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
1209 /* Expand the 32-bit value into a 64-bit value, and write that to
1210 the pseudoregister. */
1211 int rawnum = mep_pseudo_to_raw[cookednum];
1214 gdb_assert (TYPE_LENGTH (register_type (gdbarch, rawnum)) == sizeof (buf64));
1215 gdb_assert (TYPE_LENGTH (register_type (gdbarch, cookednum)) == 4);
1216 /* Slow, but legible. */
1217 store_unsigned_integer (buf64, 8, byte_order,
1218 extract_unsigned_integer (buf, 4, byte_order));
1219 regcache->raw_write (rawnum, buf64);
1224 mep_pseudo_cr64_write (struct gdbarch *gdbarch,
1225 struct regcache *regcache,
1227 const gdb_byte *buf)
1229 regcache->raw_write (mep_pseudo_to_raw[cookednum], buf);
1234 mep_pseudo_register_write (struct gdbarch *gdbarch,
1235 struct regcache *regcache,
1237 const gdb_byte *buf)
1239 if (IS_CSR_REGNUM (cookednum))
1240 mep_pseudo_csr_write (gdbarch, regcache, cookednum, buf);
1241 else if (IS_CR32_REGNUM (cookednum)
1242 || IS_FP_CR32_REGNUM (cookednum))
1243 mep_pseudo_cr32_write (gdbarch, regcache, cookednum, buf);
1244 else if (IS_CR64_REGNUM (cookednum)
1245 || IS_FP_CR64_REGNUM (cookednum))
1246 mep_pseudo_cr64_write (gdbarch, regcache, cookednum, buf);
1247 else if (IS_CCR_REGNUM (cookednum))
1248 regcache->raw_write (mep_pseudo_to_raw[cookednum], buf);
1250 gdb_assert_not_reached ("unexpected pseudo register");
1258 mep_gdb_print_insn (bfd_vma pc, disassemble_info * info)
1260 struct obj_section * s = find_pc_section (pc);
1262 info->arch = bfd_arch_mep;
1265 /* The libopcodes disassembly code uses the section to find the
1266 BFD, the BFD to find the ELF header, the ELF header to find
1267 the me_module index, and the me_module index to select the
1268 right instructions to print. */
1269 info->section = s->the_bfd_section;
1272 return print_insn_mep (pc, info);
1276 /* Prologue analysis. */
1279 /* The MeP has two classes of instructions: "core" instructions, which
1280 are pretty normal RISC chip stuff, and "coprocessor" instructions,
1281 which are mostly concerned with moving data in and out of
1282 coprocessor registers, and branching on coprocessor condition
1283 codes. There's space in the instruction set for custom coprocessor
1286 Instructions can be 16 or 32 bits long; the top two bits of the
1287 first byte indicate the length. The coprocessor instructions are
1288 mixed in with the core instructions, and there's no easy way to
1289 distinguish them; you have to completely decode them to tell one
1292 The MeP also supports a "VLIW" operation mode, where instructions
1293 always occur in fixed-width bundles. The bundles are either 32
1294 bits or 64 bits long, depending on a fixed configuration flag. You
1295 decode the first part of the bundle as normal; if it's a core
1296 instruction, and there's any space left in the bundle, the
1297 remainder of the bundle is a coprocessor instruction, which will
1298 execute in parallel with the core instruction. If the first part
1299 of the bundle is a coprocessor instruction, it occupies the entire
1302 So, here are all the cases:
1305 Every bundle is four bytes long, and naturally aligned, and can hold
1306 one or two instructions:
1307 - 16-bit core instruction; 16-bit coprocessor instruction
1308 These execute in parallel.
1309 - 32-bit core instruction
1310 - 32-bit coprocessor instruction
1313 Every bundle is eight bytes long, and naturally aligned, and can hold
1314 one or two instructions:
1315 - 16-bit core instruction; 48-bit (!) coprocessor instruction
1316 These execute in parallel.
1317 - 32-bit core instruction; 32-bit coprocessor instruction
1318 These execute in parallel.
1319 - 64-bit coprocessor instruction
1321 Now, the MeP manual doesn't define any 48- or 64-bit coprocessor
1322 instruction, so I don't really know what's up there; perhaps these
1323 are always the user-defined coprocessor instructions. */
1326 /* Return non-zero if PC is in a VLIW code section, zero
1329 mep_pc_in_vliw_section (CORE_ADDR pc)
1331 struct obj_section *s = find_pc_section (pc);
1333 return (s->the_bfd_section->flags & SEC_MEP_VLIW);
1338 /* Set *INSN to the next core instruction at PC, and return the
1339 address of the next instruction.
1341 The MeP instruction encoding is endian-dependent. 16- and 32-bit
1342 instructions are encoded as one or two two-byte parts, and each
1343 part is byte-swapped independently. Thus:
1348 asm ("movu $1, 0x123456");
1349 asm ("sb $1,0x5678($2)");
1350 asm ("clip $1, 19");
1353 compiles to this big-endian code:
1355 0: d1 56 12 34 movu $1,0x123456
1356 4: c1 28 56 78 sb $1,22136($2)
1357 8: f1 01 10 98 clip $1,0x13
1360 and this little-endian code:
1362 0: 56 d1 34 12 movu $1,0x123456
1363 4: 28 c1 78 56 sb $1,22136($2)
1364 8: 01 f1 98 10 clip $1,0x13
1367 Instructions are returned in *INSN in an endian-independent form: a
1368 given instruction always appears in *INSN the same way, regardless
1369 of whether the instruction stream is big-endian or little-endian.
1371 *INSN's most significant 16 bits are the first (i.e., at lower
1372 addresses) 16 bit part of the instruction. Its least significant
1373 16 bits are the second (i.e., higher-addressed) 16 bit part of the
1374 instruction, or zero for a 16-bit instruction. Both 16-bit parts
1375 are fetched using the current endianness.
1377 So, the *INSN values for the instruction sequence above would be
1378 the following, in either endianness:
1380 0xd1561234 movu $1,0x123456
1381 0xc1285678 sb $1,22136($2)
1382 0xf1011098 clip $1,0x13
1385 (In a sense, it would be more natural to return 16-bit instructions
1386 in the least significant 16 bits of *INSN, but that would be
1387 ambiguous. In order to tell whether you're looking at a 16- or a
1388 32-bit instruction, you have to consult the major opcode field ---
1389 the most significant four bits of the instruction's first 16-bit
1390 part. But if we put 16-bit instructions at the least significant
1391 end of *INSN, then you don't know where to find the major opcode
1392 field until you know if it's a 16- or a 32-bit instruction ---
1393 which is where we started.)
1395 If PC points to a core / coprocessor bundle in a VLIW section, set
1396 *INSN to the core instruction, and return the address of the next
1397 bundle. This has the effect of skipping the bundled coprocessor
1398 instruction. That's okay, since coprocessor instructions aren't
1399 significant to prologue analysis --- for the time being,
1403 mep_get_insn (struct gdbarch *gdbarch, CORE_ADDR pc, unsigned long *insn)
1405 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
1406 int pc_in_vliw_section;
1413 /* Are we in a VLIW section? */
1414 pc_in_vliw_section = mep_pc_in_vliw_section (pc);
1415 if (pc_in_vliw_section)
1417 /* Yes, find out which bundle size. */
1418 vliw_mode = current_options () & (MEP_OPT_VL32 | MEP_OPT_VL64);
1420 /* If PC is in a VLIW section, but the current core doesn't say
1421 that it supports either VLIW mode, then we don't have enough
1422 information to parse the instruction stream it contains.
1423 Since the "undifferentiated" standard core doesn't have
1424 either VLIW mode bit set, this could happen.
1426 But it shouldn't be an error to (say) set a breakpoint in a
1427 VLIW section, if you know you'll never reach it. (Perhaps
1428 you have a script that sets a bunch of standard breakpoints.)
1430 So we'll just return zero here, and hope for the best. */
1431 if (! (vliw_mode & (MEP_OPT_VL32 | MEP_OPT_VL64)))
1434 /* If both VL32 and VL64 are set, that's bogus, too. */
1435 if (vliw_mode == (MEP_OPT_VL32 | MEP_OPT_VL64))
1441 read_memory (pc, buf, sizeof (buf));
1442 *insn = extract_unsigned_integer (buf, 2, byte_order) << 16;
1444 /* The major opcode --- the top four bits of the first 16-bit
1445 part --- indicates whether this instruction is 16 or 32 bits
1446 long. All 32-bit instructions have a major opcode whose top
1447 two bits are 11; all the rest are 16-bit instructions. */
1448 if ((*insn & 0xc0000000) == 0xc0000000)
1450 /* Fetch the second 16-bit part of the instruction. */
1451 read_memory (pc + 2, buf, sizeof (buf));
1452 *insn = *insn | extract_unsigned_integer (buf, 2, byte_order);
1455 /* If we're in VLIW code, then the VLIW width determines the address
1456 of the next instruction. */
1459 /* In 32-bit VLIW code, all bundles are 32 bits long. We ignore the
1460 coprocessor half of a core / copro bundle. */
1461 if (vliw_mode == MEP_OPT_VL32)
1464 /* In 64-bit VLIW code, all bundles are 64 bits long. We ignore the
1465 coprocessor half of a core / copro bundle. */
1466 else if (vliw_mode == MEP_OPT_VL64)
1469 /* We'd better be in either core, 32-bit VLIW, or 64-bit VLIW mode. */
1471 gdb_assert_not_reached ("unexpected vliw mode");
1474 /* Otherwise, the top two bits of the major opcode are (again) what
1475 we need to check. */
1476 else if ((*insn & 0xc0000000) == 0xc0000000)
1481 return pc + insn_len;
1485 /* Sign-extend the LEN-bit value N. */
1486 #define SEXT(n, len) ((((int) (n)) ^ (1 << ((len) - 1))) - (1 << ((len) - 1)))
1488 /* Return the LEN-bit field at POS from I. */
1489 #define FIELD(i, pos, len) (((i) >> (pos)) & ((1 << (len)) - 1))
1491 /* Like FIELD, but sign-extend the field's value. */
1492 #define SFIELD(i, pos, len) (SEXT (FIELD ((i), (pos), (len)), (len)))
1495 /* Macros for decoding instructions.
1497 Remember that 16-bit instructions are placed in bits 16..31 of i,
1498 not at the least significant end; this means that the major opcode
1499 field is always in the same place, regardless of the width of the
1500 instruction. As a reminder of this, we show the lower 16 bits of a
1501 16-bit instruction as xxxx_xxxx_xxxx_xxxx. */
1503 /* SB Rn,(Rm) 0000_nnnn_mmmm_1000 */
1504 /* SH Rn,(Rm) 0000_nnnn_mmmm_1001 */
1505 /* SW Rn,(Rm) 0000_nnnn_mmmm_1010 */
1507 /* SW Rn,disp16(Rm) 1100_nnnn_mmmm_1010 dddd_dddd_dddd_dddd */
1508 #define IS_SW(i) (((i) & 0xf00f0000) == 0xc00a0000)
1509 /* SB Rn,disp16(Rm) 1100_nnnn_mmmm_1000 dddd_dddd_dddd_dddd */
1510 #define IS_SB(i) (((i) & 0xf00f0000) == 0xc0080000)
1511 /* SH Rn,disp16(Rm) 1100_nnnn_mmmm_1001 dddd_dddd_dddd_dddd */
1512 #define IS_SH(i) (((i) & 0xf00f0000) == 0xc0090000)
1513 #define SWBH_32_BASE(i) (FIELD (i, 20, 4))
1514 #define SWBH_32_SOURCE(i) (FIELD (i, 24, 4))
1515 #define SWBH_32_OFFSET(i) (SFIELD (i, 0, 16))
1517 /* SW Rn,disp7.align4(SP) 0100_nnnn_0ddd_dd10 xxxx_xxxx_xxxx_xxxx */
1518 #define IS_SW_IMMD(i) (((i) & 0xf0830000) == 0x40020000)
1519 #define SW_IMMD_SOURCE(i) (FIELD (i, 24, 4))
1520 #define SW_IMMD_OFFSET(i) (FIELD (i, 18, 5) << 2)
1522 /* SW Rn,(Rm) 0000_nnnn_mmmm_1010 xxxx_xxxx_xxxx_xxxx */
1523 #define IS_SW_REG(i) (((i) & 0xf00f0000) == 0x000a0000)
1524 #define SW_REG_SOURCE(i) (FIELD (i, 24, 4))
1525 #define SW_REG_BASE(i) (FIELD (i, 20, 4))
1527 /* ADD3 Rl,Rn,Rm 1001_nnnn_mmmm_llll xxxx_xxxx_xxxx_xxxx */
1528 #define IS_ADD3_16_REG(i) (((i) & 0xf0000000) == 0x90000000)
1529 #define ADD3_16_REG_SRC1(i) (FIELD (i, 20, 4)) /* n */
1530 #define ADD3_16_REG_SRC2(i) (FIELD (i, 24, 4)) /* m */
1532 /* ADD3 Rn,Rm,imm16 1100_nnnn_mmmm_0000 iiii_iiii_iiii_iiii */
1533 #define IS_ADD3_32(i) (((i) & 0xf00f0000) == 0xc0000000)
1534 #define ADD3_32_TARGET(i) (FIELD (i, 24, 4))
1535 #define ADD3_32_SOURCE(i) (FIELD (i, 20, 4))
1536 #define ADD3_32_OFFSET(i) (SFIELD (i, 0, 16))
1538 /* ADD3 Rn,SP,imm7.align4 0100_nnnn_0iii_ii00 xxxx_xxxx_xxxx_xxxx */
1539 #define IS_ADD3_16(i) (((i) & 0xf0830000) == 0x40000000)
1540 #define ADD3_16_TARGET(i) (FIELD (i, 24, 4))
1541 #define ADD3_16_OFFSET(i) (FIELD (i, 18, 5) << 2)
1543 /* ADD Rn,imm6 0110_nnnn_iiii_ii00 xxxx_xxxx_xxxx_xxxx */
1544 #define IS_ADD(i) (((i) & 0xf0030000) == 0x60000000)
1545 #define ADD_TARGET(i) (FIELD (i, 24, 4))
1546 #define ADD_OFFSET(i) (SFIELD (i, 18, 6))
1548 /* LDC Rn,imm5 0111_nnnn_iiii_101I xxxx_xxxx_xxxx_xxxx
1550 #define IS_LDC(i) (((i) & 0xf00e0000) == 0x700a0000)
1551 #define LDC_IMM(i) ((FIELD (i, 16, 1) << 4) | FIELD (i, 20, 4))
1552 #define LDC_TARGET(i) (FIELD (i, 24, 4))
1554 /* LW Rn,disp16(Rm) 1100_nnnn_mmmm_1110 dddd_dddd_dddd_dddd */
1555 #define IS_LW(i) (((i) & 0xf00f0000) == 0xc00e0000)
1556 #define LW_TARGET(i) (FIELD (i, 24, 4))
1557 #define LW_BASE(i) (FIELD (i, 20, 4))
1558 #define LW_OFFSET(i) (SFIELD (i, 0, 16))
1560 /* MOV Rn,Rm 0000_nnnn_mmmm_0000 xxxx_xxxx_xxxx_xxxx */
1561 #define IS_MOV(i) (((i) & 0xf00f0000) == 0x00000000)
1562 #define MOV_TARGET(i) (FIELD (i, 24, 4))
1563 #define MOV_SOURCE(i) (FIELD (i, 20, 4))
1565 /* BRA disp12.align2 1011_dddd_dddd_ddd0 xxxx_xxxx_xxxx_xxxx */
1566 #define IS_BRA(i) (((i) & 0xf0010000) == 0xb0000000)
1567 #define BRA_DISP(i) (SFIELD (i, 17, 11) << 1)
1570 /* This structure holds the results of a prologue analysis. */
1573 /* The architecture for which we generated this prologue info. */
1574 struct gdbarch *gdbarch;
1576 /* The offset from the frame base to the stack pointer --- always
1579 Calling this a "size" is a bit misleading, but given that the
1580 stack grows downwards, using offsets for everything keeps one
1581 from going completely sign-crazy: you never change anything's
1582 sign for an ADD instruction; always change the second operand's
1583 sign for a SUB instruction; and everything takes care of
1587 /* Non-zero if this function has initialized the frame pointer from
1588 the stack pointer, zero otherwise. */
1591 /* If has_frame_ptr is non-zero, this is the offset from the frame
1592 base to where the frame pointer points. This is always zero or
1594 int frame_ptr_offset;
1596 /* The address of the first instruction at which the frame has been
1597 set up and the arguments are where the debug info says they are
1598 --- as best as we can tell. */
1599 CORE_ADDR prologue_end;
1601 /* reg_offset[R] is the offset from the CFA at which register R is
1602 saved, or 1 if register R has not been saved. (Real values are
1603 always zero or negative.) */
1604 int reg_offset[MEP_NUM_REGS];
1607 /* Return non-zero if VALUE is an incoming argument register. */
1610 is_arg_reg (pv_t value)
1612 return (value.kind == pvk_register
1613 && MEP_R1_REGNUM <= value.reg && value.reg <= MEP_R4_REGNUM
1617 /* Return non-zero if a store of REG's current value VALUE to ADDR is
1618 probably spilling an argument register to its stack slot in STACK.
1619 Such instructions should be included in the prologue, if possible.
1621 The store is a spill if:
1622 - the value being stored is REG's original value;
1623 - the value has not already been stored somewhere in STACK; and
1624 - ADDR is a stack slot's address (e.g., relative to the original
1625 value of the SP). */
1627 is_arg_spill (struct gdbarch *gdbarch, pv_t value, pv_t addr,
1628 struct pv_area *stack)
1630 return (is_arg_reg (value)
1631 && pv_is_register (addr, MEP_SP_REGNUM)
1632 && ! stack->find_reg (gdbarch, value.reg, 0));
1636 /* Function for finding saved registers in a 'struct pv_area'; we pass
1637 this to pv_area::scan.
1639 If VALUE is a saved register, ADDR says it was saved at a constant
1640 offset from the frame base, and SIZE indicates that the whole
1641 register was saved, record its offset in RESULT_UNTYPED. */
1643 check_for_saved (void *result_untyped, pv_t addr, CORE_ADDR size, pv_t value)
1645 struct mep_prologue *result = (struct mep_prologue *) result_untyped;
1647 if (value.kind == pvk_register
1649 && pv_is_register (addr, MEP_SP_REGNUM)
1650 && size == register_size (result->gdbarch, value.reg))
1651 result->reg_offset[value.reg] = addr.k;
1655 /* Analyze a prologue starting at START_PC, going no further than
1656 LIMIT_PC. Fill in RESULT as appropriate. */
1658 mep_analyze_prologue (struct gdbarch *gdbarch,
1659 CORE_ADDR start_pc, CORE_ADDR limit_pc,
1660 struct mep_prologue *result)
1664 pv_t reg[MEP_NUM_REGS];
1665 CORE_ADDR after_last_frame_setup_insn = start_pc;
1667 memset (result, 0, sizeof (*result));
1668 result->gdbarch = gdbarch;
1670 for (int rn = 0; rn < MEP_NUM_REGS; rn++)
1672 reg[rn] = pv_register (rn, 0);
1673 result->reg_offset[rn] = 1;
1676 pv_area stack (MEP_SP_REGNUM, gdbarch_addr_bit (gdbarch));
1679 while (pc < limit_pc)
1682 pv_t pre_insn_fp, pre_insn_sp;
1684 next_pc = mep_get_insn (gdbarch, pc, &insn);
1686 /* A zero return from mep_get_insn means that either we weren't
1687 able to read the instruction from memory, or that we don't
1688 have enough information to be able to reliably decode it. So
1689 we'll store here and hope for the best. */
1693 /* Note the current values of the SP and FP, so we can tell if
1694 this instruction changed them, below. */
1695 pre_insn_fp = reg[MEP_FP_REGNUM];
1696 pre_insn_sp = reg[MEP_SP_REGNUM];
1700 int rn = ADD_TARGET (insn);
1701 CORE_ADDR imm6 = ADD_OFFSET (insn);
1703 reg[rn] = pv_add_constant (reg[rn], imm6);
1705 else if (IS_ADD3_16 (insn))
1707 int rn = ADD3_16_TARGET (insn);
1708 int imm7 = ADD3_16_OFFSET (insn);
1710 reg[rn] = pv_add_constant (reg[MEP_SP_REGNUM], imm7);
1712 else if (IS_ADD3_32 (insn))
1714 int rn = ADD3_32_TARGET (insn);
1715 int rm = ADD3_32_SOURCE (insn);
1716 int imm16 = ADD3_32_OFFSET (insn);
1718 reg[rn] = pv_add_constant (reg[rm], imm16);
1720 else if (IS_SW_REG (insn))
1722 int rn = SW_REG_SOURCE (insn);
1723 int rm = SW_REG_BASE (insn);
1725 /* If simulating this store would require us to forget
1726 everything we know about the stack frame in the name of
1727 accuracy, it would be better to just quit now. */
1728 if (stack.store_would_trash (reg[rm]))
1731 if (is_arg_spill (gdbarch, reg[rn], reg[rm], &stack))
1732 after_last_frame_setup_insn = next_pc;
1734 stack.store (reg[rm], 4, reg[rn]);
1736 else if (IS_SW_IMMD (insn))
1738 int rn = SW_IMMD_SOURCE (insn);
1739 int offset = SW_IMMD_OFFSET (insn);
1740 pv_t addr = pv_add_constant (reg[MEP_SP_REGNUM], offset);
1742 /* If simulating this store would require us to forget
1743 everything we know about the stack frame in the name of
1744 accuracy, it would be better to just quit now. */
1745 if (stack.store_would_trash (addr))
1748 if (is_arg_spill (gdbarch, reg[rn], addr, &stack))
1749 after_last_frame_setup_insn = next_pc;
1751 stack.store (addr, 4, reg[rn]);
1753 else if (IS_MOV (insn))
1755 int rn = MOV_TARGET (insn);
1756 int rm = MOV_SOURCE (insn);
1760 if (pv_is_register (reg[rm], rm) && is_arg_reg (reg[rm]))
1761 after_last_frame_setup_insn = next_pc;
1763 else if (IS_SB (insn) || IS_SH (insn) || IS_SW (insn))
1765 int rn = SWBH_32_SOURCE (insn);
1766 int rm = SWBH_32_BASE (insn);
1767 int disp = SWBH_32_OFFSET (insn);
1768 int size = (IS_SB (insn) ? 1
1770 : (gdb_assert (IS_SW (insn)), 4));
1771 pv_t addr = pv_add_constant (reg[rm], disp);
1773 if (stack.store_would_trash (addr))
1776 if (is_arg_spill (gdbarch, reg[rn], addr, &stack))
1777 after_last_frame_setup_insn = next_pc;
1779 stack.store (addr, size, reg[rn]);
1781 else if (IS_LDC (insn))
1783 int rn = LDC_TARGET (insn);
1784 int cr = LDC_IMM (insn) + MEP_FIRST_CSR_REGNUM;
1788 else if (IS_LW (insn))
1790 int rn = LW_TARGET (insn);
1791 int rm = LW_BASE (insn);
1792 int offset = LW_OFFSET (insn);
1793 pv_t addr = pv_add_constant (reg[rm], offset);
1795 reg[rn] = stack.fetch (addr, 4);
1797 else if (IS_BRA (insn) && BRA_DISP (insn) > 0)
1799 /* When a loop appears as the first statement of a function
1800 body, gcc 4.x will use a BRA instruction to branch to the
1801 loop condition checking code. This BRA instruction is
1802 marked as part of the prologue. We therefore set next_pc
1803 to this branch target and also stop the prologue scan.
1804 The instructions at and beyond the branch target should
1805 no longer be associated with the prologue.
1807 Note that we only consider forward branches here. We
1808 presume that a forward branch is being used to skip over
1811 A backwards branch is covered by the default case below.
1812 If we were to encounter a backwards branch, that would
1813 most likely mean that we've scanned through a loop body.
1814 We definitely want to stop the prologue scan when this
1815 happens and that is precisely what is done by the default
1817 next_pc = pc + BRA_DISP (insn);
1818 after_last_frame_setup_insn = next_pc;
1822 /* We've hit some instruction we don't know how to simulate.
1823 Strictly speaking, we should set every value we're
1824 tracking to "unknown". But we'll be optimistic, assume
1825 that we have enough information already, and stop
1829 /* If this instruction changed the FP or decreased the SP (i.e.,
1830 allocated more stack space), then this may be a good place to
1831 declare the prologue finished. However, there are some
1834 - If the instruction just changed the FP back to its original
1835 value, then that's probably a restore instruction. The
1836 prologue should definitely end before that.
1838 - If the instruction increased the value of the SP (that is,
1839 shrunk the frame), then it's probably part of a frame
1840 teardown sequence, and the prologue should end before that. */
1842 if (! pv_is_identical (reg[MEP_FP_REGNUM], pre_insn_fp))
1844 if (! pv_is_register_k (reg[MEP_FP_REGNUM], MEP_FP_REGNUM, 0))
1845 after_last_frame_setup_insn = next_pc;
1847 else if (! pv_is_identical (reg[MEP_SP_REGNUM], pre_insn_sp))
1849 /* The comparison of constants looks odd, there, because .k
1850 is unsigned. All it really means is that the new value
1851 is lower than it was before the instruction. */
1852 if (pv_is_register (pre_insn_sp, MEP_SP_REGNUM)
1853 && pv_is_register (reg[MEP_SP_REGNUM], MEP_SP_REGNUM)
1854 && ((pre_insn_sp.k - reg[MEP_SP_REGNUM].k)
1855 < (reg[MEP_SP_REGNUM].k - pre_insn_sp.k)))
1856 after_last_frame_setup_insn = next_pc;
1862 /* Is the frame size (offset, really) a known constant? */
1863 if (pv_is_register (reg[MEP_SP_REGNUM], MEP_SP_REGNUM))
1864 result->frame_size = reg[MEP_SP_REGNUM].k;
1866 /* Was the frame pointer initialized? */
1867 if (pv_is_register (reg[MEP_FP_REGNUM], MEP_SP_REGNUM))
1869 result->has_frame_ptr = 1;
1870 result->frame_ptr_offset = reg[MEP_FP_REGNUM].k;
1873 /* Record where all the registers were saved. */
1874 stack.scan (check_for_saved, (void *) result);
1876 result->prologue_end = after_last_frame_setup_insn;
1881 mep_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR pc)
1884 CORE_ADDR func_addr, func_end;
1885 struct mep_prologue p;
1887 /* Try to find the extent of the function that contains PC. */
1888 if (! find_pc_partial_function (pc, &name, &func_addr, &func_end))
1891 mep_analyze_prologue (gdbarch, pc, func_end, &p);
1892 return p.prologue_end;
1898 constexpr gdb_byte mep_break_insn[] = { 0x70, 0x32 };
1900 typedef BP_MANIPULATION (mep_break_insn) mep_breakpoint;
1903 /* Frames and frame unwinding. */
1906 static struct mep_prologue *
1907 mep_analyze_frame_prologue (struct frame_info *this_frame,
1908 void **this_prologue_cache)
1910 if (! *this_prologue_cache)
1912 CORE_ADDR func_start, stop_addr;
1914 *this_prologue_cache
1915 = FRAME_OBSTACK_ZALLOC (struct mep_prologue);
1917 func_start = get_frame_func (this_frame);
1918 stop_addr = get_frame_pc (this_frame);
1920 /* If we couldn't find any function containing the PC, then
1921 just initialize the prologue cache, but don't do anything. */
1923 stop_addr = func_start;
1925 mep_analyze_prologue (get_frame_arch (this_frame),
1926 func_start, stop_addr,
1927 (struct mep_prologue *) *this_prologue_cache);
1930 return (struct mep_prologue *) *this_prologue_cache;
1934 /* Given the next frame and a prologue cache, return this frame's
1937 mep_frame_base (struct frame_info *this_frame,
1938 void **this_prologue_cache)
1940 struct mep_prologue *p
1941 = mep_analyze_frame_prologue (this_frame, this_prologue_cache);
1943 /* In functions that use alloca, the distance between the stack
1944 pointer and the frame base varies dynamically, so we can't use
1945 the SP plus static information like prologue analysis to find the
1946 frame base. However, such functions must have a frame pointer,
1947 to be able to restore the SP on exit. So whenever we do have a
1948 frame pointer, use that to find the base. */
1949 if (p->has_frame_ptr)
1952 = get_frame_register_unsigned (this_frame, MEP_FP_REGNUM);
1953 return fp - p->frame_ptr_offset;
1958 = get_frame_register_unsigned (this_frame, MEP_SP_REGNUM);
1959 return sp - p->frame_size;
1965 mep_frame_this_id (struct frame_info *this_frame,
1966 void **this_prologue_cache,
1967 struct frame_id *this_id)
1969 *this_id = frame_id_build (mep_frame_base (this_frame, this_prologue_cache),
1970 get_frame_func (this_frame));
1974 static struct value *
1975 mep_frame_prev_register (struct frame_info *this_frame,
1976 void **this_prologue_cache, int regnum)
1978 struct mep_prologue *p
1979 = mep_analyze_frame_prologue (this_frame, this_prologue_cache);
1981 /* There are a number of complications in unwinding registers on the
1982 MeP, having to do with core functions calling VLIW functions and
1985 The least significant bit of the link register, LP.LTOM, is the
1986 VLIW mode toggle bit: it's set if a core function called a VLIW
1987 function, or vice versa, and clear when the caller and callee
1988 were both in the same mode.
1990 So, if we're asked to unwind the PC, then we really want to
1991 unwind the LP and clear the least significant bit. (Real return
1992 addresses are always even.) And if we want to unwind the program
1993 status word (PSW), we need to toggle PSW.OM if LP.LTOM is set.
1995 Tweaking the register values we return in this way means that the
1996 bits in BUFFERP[] are not the same as the bits you'd find at
1997 ADDRP in the inferior, so we make sure lvalp is not_lval when we
1999 if (regnum == MEP_PC_REGNUM)
2001 struct value *value;
2003 value = mep_frame_prev_register (this_frame, this_prologue_cache,
2005 lp = value_as_long (value);
2006 release_value (value);
2008 return frame_unwind_got_constant (this_frame, regnum, lp & ~1);
2012 CORE_ADDR frame_base = mep_frame_base (this_frame, this_prologue_cache);
2013 struct value *value;
2015 /* Our caller's SP is our frame base. */
2016 if (regnum == MEP_SP_REGNUM)
2017 return frame_unwind_got_constant (this_frame, regnum, frame_base);
2019 /* If prologue analysis says we saved this register somewhere,
2020 return a description of the stack slot holding it. */
2021 if (p->reg_offset[regnum] != 1)
2022 value = frame_unwind_got_memory (this_frame, regnum,
2023 frame_base + p->reg_offset[regnum]);
2025 /* Otherwise, presume we haven't changed the value of this
2026 register, and get it from the next frame. */
2028 value = frame_unwind_got_register (this_frame, regnum, regnum);
2030 /* If we need to toggle the operating mode, do so. */
2031 if (regnum == MEP_PSW_REGNUM)
2035 psw = value_as_long (value);
2036 release_value (value);
2038 /* Get the LP's value, too. */
2039 value = get_frame_register_value (this_frame, MEP_LP_REGNUM);
2040 lp = value_as_long (value);
2041 release_value (value);
2043 /* If LP.LTOM is set, then toggle PSW.OM. */
2047 return frame_unwind_got_constant (this_frame, regnum, psw);
2055 static const struct frame_unwind mep_frame_unwind = {
2057 default_frame_unwind_stop_reason,
2059 mep_frame_prev_register,
2061 default_frame_sniffer
2065 /* Return values. */
2069 mep_use_struct_convention (struct type *type)
2071 return (TYPE_LENGTH (type) > MEP_GPR_SIZE);
2076 mep_extract_return_value (struct gdbarch *arch,
2078 struct regcache *regcache,
2081 int byte_order = gdbarch_byte_order (arch);
2083 /* Values that don't occupy a full register appear at the less
2084 significant end of the value. This is the offset to where the
2088 /* Return values > MEP_GPR_SIZE bytes are returned in memory,
2089 pointed to by R0. */
2090 gdb_assert (TYPE_LENGTH (type) <= MEP_GPR_SIZE);
2092 if (byte_order == BFD_ENDIAN_BIG)
2093 offset = MEP_GPR_SIZE - TYPE_LENGTH (type);
2097 /* Return values that do fit in a single register are returned in R0. */
2098 regcache->cooked_read_part (MEP_R0_REGNUM, offset, TYPE_LENGTH (type),
2104 mep_store_return_value (struct gdbarch *arch,
2106 struct regcache *regcache,
2107 const gdb_byte *valbuf)
2109 int byte_order = gdbarch_byte_order (arch);
2111 /* Values that fit in a single register go in R0. */
2112 if (TYPE_LENGTH (type) <= MEP_GPR_SIZE)
2114 /* Values that don't occupy a full register appear at the least
2115 significant end of the value. This is the offset to where the
2119 if (byte_order == BFD_ENDIAN_BIG)
2120 offset = MEP_GPR_SIZE - TYPE_LENGTH (type);
2124 regcache->cooked_write_part (MEP_R0_REGNUM, offset, TYPE_LENGTH (type),
2128 /* Return values larger than a single register are returned in
2129 memory, pointed to by R0. Unfortunately, we can't count on R0
2130 pointing to the return buffer, so we raise an error here. */
2133 GDB cannot set return values larger than four bytes; the Media Processor's\n\
2134 calling conventions do not provide enough information to do this.\n\
2135 Try using the 'return' command with no argument."));
2138 static enum return_value_convention
2139 mep_return_value (struct gdbarch *gdbarch, struct value *function,
2140 struct type *type, struct regcache *regcache,
2141 gdb_byte *readbuf, const gdb_byte *writebuf)
2143 if (mep_use_struct_convention (type))
2148 /* Although the address of the struct buffer gets passed in R1, it's
2149 returned in R0. Fetch R0's value and then read the memory
2151 regcache_raw_read_unsigned (regcache, MEP_R0_REGNUM, &addr);
2152 read_memory (addr, readbuf, TYPE_LENGTH (type));
2156 /* Return values larger than a single register are returned in
2157 memory, pointed to by R0. Unfortunately, we can't count on R0
2158 pointing to the return buffer, so we raise an error here. */
2160 GDB cannot set return values larger than four bytes; the Media Processor's\n\
2161 calling conventions do not provide enough information to do this.\n\
2162 Try using the 'return' command with no argument."));
2164 return RETURN_VALUE_ABI_RETURNS_ADDRESS;
2168 mep_extract_return_value (gdbarch, type, regcache, readbuf);
2170 mep_store_return_value (gdbarch, type, regcache, writebuf);
2172 return RETURN_VALUE_REGISTER_CONVENTION;
2176 /* Inferior calls. */
2180 mep_frame_align (struct gdbarch *gdbarch, CORE_ADDR sp)
2182 /* Require word alignment. */
2187 /* From "lang_spec2.txt":
2189 4.2 Calling conventions
2191 4.2.1 Core register conventions
2193 - Parameters should be evaluated from left to right, and they
2194 should be held in $1,$2,$3,$4 in order. The fifth parameter or
2195 after should be held in the stack. If the size is larger than 4
2196 bytes in the first four parameters, the pointer should be held in
2197 the registers instead. If the size is larger than 4 bytes in the
2198 fifth parameter or after, the pointer should be held in the stack.
2200 - Return value of a function should be held in register $0. If the
2201 size of return value is larger than 4 bytes, $1 should hold the
2202 pointer pointing memory that would hold the return value. In this
2203 case, the first parameter should be held in $2, the second one in
2204 $3, and the third one in $4, and the forth parameter or after
2205 should be held in the stack.
2207 [This doesn't say so, but arguments shorter than four bytes are
2208 passed in the least significant end of a four-byte word when
2209 they're passed on the stack.] */
2212 /* Traverse the list of ARGC arguments ARGV; for every ARGV[i] too
2213 large to fit in a register, save it on the stack, and place its
2214 address in COPY[i]. SP is the initial stack pointer; return the
2215 new stack pointer. */
2217 push_large_arguments (CORE_ADDR sp, int argc, struct value **argv,
2222 for (i = 0; i < argc; i++)
2224 unsigned arg_len = TYPE_LENGTH (value_type (argv[i]));
2226 if (arg_len > MEP_GPR_SIZE)
2228 /* Reserve space for the copy, and then round the SP down, to
2229 make sure it's all aligned properly. */
2230 sp = (sp - arg_len) & -4;
2231 write_memory (sp, value_contents (argv[i]), arg_len);
2241 mep_push_dummy_call (struct gdbarch *gdbarch, struct value *function,
2242 struct regcache *regcache, CORE_ADDR bp_addr,
2243 int argc, struct value **argv, CORE_ADDR sp,
2244 function_call_return_method return_method,
2245 CORE_ADDR struct_addr)
2247 enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
2248 CORE_ADDR *copy = (CORE_ADDR *) alloca (argc * sizeof (copy[0]));
2251 /* The number of the next register available to hold an argument. */
2254 /* The address of the next stack slot available to hold an argument. */
2255 CORE_ADDR arg_stack;
2257 /* The address of the end of the stack area for arguments. This is
2258 just for error checking. */
2259 CORE_ADDR arg_stack_end;
2261 sp = push_large_arguments (sp, argc, argv, copy);
2263 /* Reserve space for the stack arguments, if any. */
2265 if (argc + (struct_addr ? 1 : 0) > 4)
2266 sp -= ((argc + (struct_addr ? 1 : 0)) - 4) * MEP_GPR_SIZE;
2268 arg_reg = MEP_R1_REGNUM;
2271 /* If we're returning a structure by value, push the pointer to the
2272 buffer as the first argument. */
2273 if (return_method == return_method_struct)
2275 regcache_cooked_write_unsigned (regcache, arg_reg, struct_addr);
2279 for (i = 0; i < argc; i++)
2283 /* Arguments that fit in a GPR get expanded to fill the GPR. */
2284 if (TYPE_LENGTH (value_type (argv[i])) <= MEP_GPR_SIZE)
2285 value = extract_unsigned_integer (value_contents (argv[i]),
2286 TYPE_LENGTH (value_type (argv[i])),
2289 /* Arguments too large to fit in a GPR get copied to the stack,
2290 and we pass a pointer to the copy. */
2294 /* We use $1 -- $4 for passing arguments, then use the stack. */
2295 if (arg_reg <= MEP_R4_REGNUM)
2297 regcache_cooked_write_unsigned (regcache, arg_reg, value);
2302 gdb_byte buf[MEP_GPR_SIZE];
2303 store_unsigned_integer (buf, MEP_GPR_SIZE, byte_order, value);
2304 write_memory (arg_stack, buf, MEP_GPR_SIZE);
2305 arg_stack += MEP_GPR_SIZE;
2309 gdb_assert (arg_stack <= arg_stack_end);
2311 /* Set the return address. */
2312 regcache_cooked_write_unsigned (regcache, MEP_LP_REGNUM, bp_addr);
2314 /* Update the stack pointer. */
2315 regcache_cooked_write_unsigned (regcache, MEP_SP_REGNUM, sp);
2321 /* Initialization. */
2324 static struct gdbarch *
2325 mep_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
2327 struct gdbarch *gdbarch;
2328 struct gdbarch_tdep *tdep;
2330 /* Which me_module are we building a gdbarch object for? */
2331 CONFIG_ATTR me_module;
2333 /* If we have a BFD in hand, figure out which me_module it was built
2334 for. Otherwise, use the no-particular-me_module code. */
2337 /* The way to get the me_module code depends on the object file
2338 format. At the moment, we only know how to handle ELF. */
2339 if (bfd_get_flavour (info.abfd) == bfd_target_elf_flavour)
2341 int flag = elf_elfheader (info.abfd)->e_flags & EF_MEP_INDEX_MASK;
2342 me_module = (CONFIG_ATTR) flag;
2345 me_module = CONFIG_NONE;
2348 me_module = CONFIG_NONE;
2350 /* If we're setting the architecture from a file, check the
2351 endianness of the file against that of the me_module. */
2354 /* The negations on either side make the comparison treat all
2355 non-zero (true) values as equal. */
2356 if (! bfd_big_endian (info.abfd) != ! me_module_big_endian (me_module))
2358 const char *module_name = me_module_name (me_module);
2359 const char *module_endianness
2360 = me_module_big_endian (me_module) ? "big" : "little";
2361 const char *file_name = bfd_get_filename (info.abfd);
2362 const char *file_endianness
2363 = bfd_big_endian (info.abfd) ? "big" : "little";
2365 fputc_unfiltered ('\n', gdb_stderr);
2367 warning (_("the MeP module '%s' is %s-endian, but the executable\n"
2368 "%s is %s-endian."),
2369 module_name, module_endianness,
2370 file_name, file_endianness);
2372 warning (_("the selected MeP module is %s-endian, but the "
2374 "%s is %s-endian."),
2375 module_endianness, file_name, file_endianness);
2379 /* Find a candidate among the list of architectures we've created
2380 already. info->bfd_arch_info needs to match, but we also want
2381 the right me_module: the ELF header's e_flags field needs to
2383 for (arches = gdbarch_list_lookup_by_info (arches, &info);
2385 arches = gdbarch_list_lookup_by_info (arches->next, &info))
2386 if (gdbarch_tdep (arches->gdbarch)->me_module == me_module)
2387 return arches->gdbarch;
2389 tdep = XCNEW (struct gdbarch_tdep);
2390 gdbarch = gdbarch_alloc (&info, tdep);
2392 /* Get a CGEN CPU descriptor for this architecture. */
2394 const char *mach_name = info.bfd_arch_info->printable_name;
2395 enum cgen_endian endian = (info.byte_order == BFD_ENDIAN_BIG
2397 : CGEN_ENDIAN_LITTLE);
2399 tdep->cpu_desc = mep_cgen_cpu_open (CGEN_CPU_OPEN_BFDMACH, mach_name,
2400 CGEN_CPU_OPEN_ENDIAN, endian,
2404 tdep->me_module = me_module;
2407 set_gdbarch_num_regs (gdbarch, MEP_NUM_RAW_REGS);
2408 set_gdbarch_pc_regnum (gdbarch, MEP_PC_REGNUM);
2409 set_gdbarch_sp_regnum (gdbarch, MEP_SP_REGNUM);
2410 set_gdbarch_register_name (gdbarch, mep_register_name);
2411 set_gdbarch_register_type (gdbarch, mep_register_type);
2412 set_gdbarch_num_pseudo_regs (gdbarch, MEP_NUM_PSEUDO_REGS);
2413 set_gdbarch_pseudo_register_read (gdbarch, mep_pseudo_register_read);
2414 set_gdbarch_pseudo_register_write (gdbarch, mep_pseudo_register_write);
2415 set_gdbarch_dwarf2_reg_to_regnum (gdbarch, mep_debug_reg_to_regnum);
2416 set_gdbarch_stab_reg_to_regnum (gdbarch, mep_debug_reg_to_regnum);
2418 set_gdbarch_register_reggroup_p (gdbarch, mep_register_reggroup_p);
2419 reggroup_add (gdbarch, all_reggroup);
2420 reggroup_add (gdbarch, general_reggroup);
2421 reggroup_add (gdbarch, save_reggroup);
2422 reggroup_add (gdbarch, restore_reggroup);
2423 reggroup_add (gdbarch, mep_csr_reggroup);
2424 reggroup_add (gdbarch, mep_cr_reggroup);
2425 reggroup_add (gdbarch, mep_ccr_reggroup);
2428 set_gdbarch_print_insn (gdbarch, mep_gdb_print_insn);
2431 set_gdbarch_breakpoint_kind_from_pc (gdbarch, mep_breakpoint::kind_from_pc);
2432 set_gdbarch_sw_breakpoint_from_kind (gdbarch, mep_breakpoint::bp_from_kind);
2433 set_gdbarch_decr_pc_after_break (gdbarch, 0);
2434 set_gdbarch_skip_prologue (gdbarch, mep_skip_prologue);
2436 /* Frames and frame unwinding. */
2437 frame_unwind_append_unwinder (gdbarch, &mep_frame_unwind);
2438 set_gdbarch_inner_than (gdbarch, core_addr_lessthan);
2439 set_gdbarch_frame_args_skip (gdbarch, 0);
2441 /* Return values. */
2442 set_gdbarch_return_value (gdbarch, mep_return_value);
2444 /* Inferior function calls. */
2445 set_gdbarch_frame_align (gdbarch, mep_frame_align);
2446 set_gdbarch_push_dummy_call (gdbarch, mep_push_dummy_call);
2452 _initialize_mep_tdep (void)
2454 mep_csr_reggroup = reggroup_new ("csr", USER_REGGROUP);
2455 mep_cr_reggroup = reggroup_new ("cr", USER_REGGROUP);
2456 mep_ccr_reggroup = reggroup_new ("ccr", USER_REGGROUP);
2458 register_gdbarch_init (bfd_arch_mep, mep_gdbarch_init);
2460 mep_init_pseudoregister_maps ();