njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 1 | |
| 2 | /*--------------------------------------------------------------------*/ |
| 3 | /*--- The core/tool interface. pub_tool_tooliface.h ---*/ |
| 4 | /*--------------------------------------------------------------------*/ |
| 5 | |
| 6 | /* |
| 7 | This file is part of Valgrind, a dynamic binary instrumentation |
| 8 | framework. |
| 9 | |
sewardj | 9ebd6e0 | 2007-01-08 06:01:59 +0000 | [diff] [blame] | 10 | Copyright (C) 2000-2007 Julian Seward |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 11 | jseward@acm.org |
| 12 | |
| 13 | This program is free software; you can redistribute it and/or |
| 14 | modify it under the terms of the GNU General Public License as |
| 15 | published by the Free Software Foundation; either version 2 of the |
| 16 | License, or (at your option) any later version. |
| 17 | |
| 18 | This program is distributed in the hope that it will be useful, but |
| 19 | WITHOUT ANY WARRANTY; without even the implied warranty of |
| 20 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 21 | General Public License for more details. |
| 22 | |
| 23 | You should have received a copy of the GNU General Public License |
| 24 | along with this program; if not, write to the Free Software |
| 25 | Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA |
| 26 | 02111-1307, USA. |
| 27 | |
| 28 | The GNU General Public License is contained in the file COPYING. |
| 29 | */ |
| 30 | |
| 31 | #ifndef __PUB_TOOL_TOOLIFACE_H |
| 32 | #define __PUB_TOOL_TOOLIFACE_H |
| 33 | |
njn | acd885a | 2005-05-16 20:40:51 +0000 | [diff] [blame] | 34 | #include "pub_tool_errormgr.h" // for Error, Supp |
njn | 0fc5cbd | 2006-10-18 21:50:26 +0000 | [diff] [blame] | 35 | #include "libvex.h" // for all Vex stuff |
njn | acd885a | 2005-05-16 20:40:51 +0000 | [diff] [blame] | 36 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 37 | /* ------------------------------------------------------------------ */ |
| 38 | /* The interface version */ |
| 39 | |
| 40 | /* The version number indicates binary-incompatible changes to the |
| 41 | interface; if the core and tool versions don't match, Valgrind |
| 42 | will abort. */ |
dirk | f8126e9 | 2006-11-14 14:32:46 +0000 | [diff] [blame] | 43 | #define VG_CORE_INTERFACE_VERSION 10 |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 44 | |
| 45 | typedef struct _ToolInfo { |
| 46 | Int sizeof_ToolInfo; |
| 47 | Int interface_version; |
| 48 | |
| 49 | /* Initialise tool. Must do the following: |
| 50 | - initialise the `details' struct, via the VG_(details_*)() functions |
| 51 | - register any helpers called by generated code |
| 52 | |
| 53 | May do the following: |
| 54 | - initialise the `needs' struct to indicate certain requirements, via |
| 55 | the VG_(needs_*)() functions |
| 56 | - initialize all the tool's entrypoints via the VG_(init_*)() functions |
| 57 | - register any tool-specific profiling events |
| 58 | - any other tool-specific initialisation |
| 59 | */ |
| 60 | void (*tl_pre_clo_init) ( void ); |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 61 | } ToolInfo; |
| 62 | |
sewardj | 45f4e7c | 2005-09-27 19:20:21 +0000 | [diff] [blame] | 63 | extern const ToolInfo VG_(tool_info); |
| 64 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 65 | /* Every tool must include this macro somewhere, exactly once. */ |
sewardj | 45f4e7c | 2005-09-27 19:20:21 +0000 | [diff] [blame] | 66 | #define VG_DETERMINE_INTERFACE_VERSION(pre_clo_init) \ |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 67 | const ToolInfo VG_(tool_info) = { \ |
| 68 | .sizeof_ToolInfo = sizeof(ToolInfo), \ |
| 69 | .interface_version = VG_CORE_INTERFACE_VERSION, \ |
| 70 | .tl_pre_clo_init = pre_clo_init, \ |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 71 | }; |
| 72 | |
| 73 | /* ------------------------------------------------------------------ */ |
| 74 | /* Basic tool functions */ |
| 75 | |
sewardj | 461df9c | 2006-01-17 02:06:39 +0000 | [diff] [blame] | 76 | /* The tool_instrument function is passed as a callback to |
sewardj | 7ce6239 | 2006-10-15 12:48:18 +0000 | [diff] [blame] | 77 | LibVEX_Translate. VgCallbackClosure carries additional info |
sewardj | 461df9c | 2006-01-17 02:06:39 +0000 | [diff] [blame] | 78 | which the instrumenter might like to know, but which is opaque to |
| 79 | Vex. |
| 80 | */ |
| 81 | typedef |
| 82 | struct { |
| 83 | Addr64 nraddr; /* non-redirected guest address */ |
| 84 | Addr64 readdr; /* redirected guest address */ |
| 85 | ThreadId tid; /* tid requesting translation */ |
| 86 | } |
| 87 | VgCallbackClosure; |
| 88 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 89 | extern void VG_(basic_tool_funcs)( |
| 90 | // Do any initialisation that can only be done after command line |
| 91 | // processing. |
| 92 | void (*post_clo_init)(void), |
| 93 | |
sewardj | 4ba057c | 2005-10-18 12:04:18 +0000 | [diff] [blame] | 94 | // Instrument a basic block. Must be a true function, ie. the same |
| 95 | // input always results in the same output, because basic blocks |
sewardj | 7ce6239 | 2006-10-15 12:48:18 +0000 | [diff] [blame] | 96 | // can be retranslated, unless you're doing something really |
| 97 | // strange. Anyway, the arguments. Mostly they are straightforward |
| 98 | // except for the distinction between redirected and non-redirected |
| 99 | // guest code addresses, which is important to understand. |
| 100 | // |
| 101 | // VgCallBackClosure* closure contains extra arguments passed |
| 102 | // from Valgrind to the instrumenter, which Vex doesn't know about. |
| 103 | // You are free to look inside this structure. |
| 104 | // |
| 105 | // * closure->tid is the ThreadId of the thread requesting the |
| 106 | // translation. Not sure why this is here; perhaps callgrind |
| 107 | // uses it. |
| 108 | // |
| 109 | // * closure->nraddr is the non-redirected guest address of the |
| 110 | // start of the translation. In other words, the translation is |
| 111 | // being constructed because the guest program jumped to |
| 112 | // closure->nraddr but no translation of it was found. |
| 113 | // |
| 114 | // * closure->readdr is the redirected guest address, from which |
| 115 | // the translation was really made. |
| 116 | // |
| 117 | // To clarify this, consider what happens when, in Memcheck, the |
| 118 | // first call to malloc() happens. The guest program will be |
| 119 | // trying to jump to malloc() in libc; hence ->nraddr will contain |
| 120 | // that address. However, Memcheck intercepts and replaces |
| 121 | // malloc, hence ->readdr will be the address of Memcheck's |
| 122 | // malloc replacement in |
| 123 | // coregrind/m_replacemalloc/vg_replacemalloc.c. It follows |
| 124 | // that the first IMark in the translation will be labelled as |
| 125 | // from ->readdr rather than ->nraddr. |
| 126 | // |
| 127 | // Since most functions are not redirected, the majority of the |
| 128 | // time ->nraddr will be the same as ->readdr. However, you |
| 129 | // cannot assume this: if your tool has metadata associated |
| 130 | // with code addresses it will get into deep trouble if it does |
| 131 | // make this assumption. |
| 132 | // |
sewardj | 0b9d74a | 2006-12-24 02:24:11 +0000 | [diff] [blame] | 133 | // IRSB* sb_in is the incoming superblock to be instrumented, |
| 134 | // in flat IR form. |
sewardj | 7ce6239 | 2006-10-15 12:48:18 +0000 | [diff] [blame] | 135 | // |
| 136 | // VexGuestLayout* layout contains limited info on the layout of |
| 137 | // the guest state: where the stack pointer and program counter |
| 138 | // are, and which fields should be regarded as 'always defined'. |
| 139 | // Memcheck uses this. |
| 140 | // |
| 141 | // VexGuestExtents* vge points to a structure which states the |
| 142 | // precise byte ranges of original code from which this translation |
| 143 | // was made (there may be up to three different ranges involved). |
| 144 | // Note again that these are the real addresses from which the code |
| 145 | // came. And so it should be the case that closure->readdr is the |
| 146 | // same as vge->base[0]; indeed Cachegrind contains this assertion. |
| 147 | // |
| 148 | // Tools which associate shadow data with code addresses |
| 149 | // (cachegrind, callgrind) need to be particularly clear about |
| 150 | // whether they are making the association with redirected or |
| 151 | // non-redirected code addresses. Both approaches are viable |
| 152 | // but you do need to understand what's going on. See comments |
| 153 | // below on discard_basic_block_info(). |
| 154 | // |
| 155 | // IRType gWordTy and IRType hWordTy contain the types of native |
| 156 | // words on the guest (simulated) and host (real) CPUs. They will |
| 157 | // by either Ity_I32 or Ity_I64. So far we have never built a |
| 158 | // cross-architecture Valgrind so they should always be the same. |
| 159 | // |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 160 | /* --- Further comments about the IR that your --- */ |
| 161 | /* --- instrumentation function will receive. --- */ |
| 162 | /* |
njn | a3cc29f | 2007-02-05 23:23:55 +0000 | [diff] [blame] | 163 | In the incoming IRSB, the IR for each instruction begins with an |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 164 | IRStmt_IMark, which states the address and length of the |
| 165 | instruction from which this IR came. This makes it easy for |
| 166 | profiling-style tools to know precisely which guest code |
| 167 | addresses are being executed. |
| 168 | |
| 169 | However, before the first IRStmt_IMark, there may be other IR |
| 170 | statements -- a preamble. In most cases this preamble is empty, |
| 171 | but when it isn't, what it contains is some supporting IR that |
| 172 | the JIT uses to ensure control flow works correctly. This |
| 173 | preamble does not modify any architecturally defined guest state |
| 174 | (registers or memory) and so does not contain anything that will |
| 175 | be of interest to your tool. |
| 176 | |
| 177 | You should therefore |
| 178 | |
| 179 | (1) copy any IR preceding the first IMark verbatim to the start |
njn | a3cc29f | 2007-02-05 23:23:55 +0000 | [diff] [blame] | 180 | of the output IRSB. |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 181 | |
| 182 | (2) not try to instrument it or modify it in any way. |
| 183 | |
| 184 | For the record, stuff that may be in the preamble at |
| 185 | present is: |
| 186 | |
| 187 | - A self-modifying-code check has been requested for this block. |
| 188 | The preamble will contain instructions to checksum the block, |
| 189 | compare against the expected value, and exit the dispatcher |
| 190 | requesting a discard (hence forcing a retranslation) if they |
| 191 | don't match. |
| 192 | |
| 193 | - This block is known to be the entry point of a wrapper of some |
sewardj | 358ebea | 2006-10-15 13:47:43 +0000 | [diff] [blame] | 194 | function F. In this case the preamble contains code to write |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 195 | the address of the original F (the fn being wrapped) into a |
| 196 | 'hidden' guest state register _NRADDR. The wrapper can later |
| 197 | read this register using a client request and make a |
| 198 | non-redirected call to it using another client-request-like |
| 199 | magic macro. |
| 200 | |
| 201 | - For platforms that use the AIX ABI (including ppc64-linux), it |
sewardj | 358ebea | 2006-10-15 13:47:43 +0000 | [diff] [blame] | 202 | is necessary to have a preamble even for replacement functions |
| 203 | (not just for wrappers), because it is necessary to switch the |
| 204 | R2 register (constant-pool pointer) to a different value when |
| 205 | swizzling the program counter. |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 206 | |
| 207 | Hence the preamble pushes both R2 and LR (the return address) |
| 208 | on a small 16-entry stack in the guest state and sets R2 to an |
| 209 | appropriate value for the wrapper/replacement fn. LR is then |
| 210 | set so that the wrapper/replacement fn returns to a magic IR |
| 211 | stub which restores R2 and LR and returns. |
| 212 | |
| 213 | It's all hugely ugly and fragile. And it places a stringent |
| 214 | requirement on m_debuginfo to find out the correct R2 (toc |
| 215 | pointer) value for the wrapper/replacement function. So much |
| 216 | so that m_redir will refuse to honour a redirect-to-me request |
| 217 | if it cannot find (by asking m_debuginfo) a plausible R2 value |
| 218 | for 'me'. |
| 219 | |
| 220 | Because this mechanism maintains a shadow stack of (R2,LR) |
| 221 | pairs in the guest state, it will fail if the |
| 222 | wrapper/redirection function, or anything it calls, longjumps |
| 223 | out past the wrapper, because then the magic return stub will |
| 224 | not be run and so the shadow stack will not be popped. So it |
| 225 | will quickly fill up. Fortunately none of this applies to |
| 226 | {x86,amd64,ppc32}-linux; on those platforms, wrappers can |
| 227 | longjump and recurse arbitrarily and everything should work |
| 228 | fine. |
sewardj | f1962d3 | 2006-10-19 13:22:16 +0000 | [diff] [blame] | 229 | |
| 230 | Note that copying the preamble verbatim may cause complications |
| 231 | for your instrumenter if you shadow IR temporaries. See big |
| 232 | comment in MC_(instrument) in memcheck/mc_translate.c for |
| 233 | details. |
sewardj | c87b5ec | 2006-10-15 13:46:18 +0000 | [diff] [blame] | 234 | */ |
sewardj | 0b9d74a | 2006-12-24 02:24:11 +0000 | [diff] [blame] | 235 | IRSB*(*instrument)(VgCallbackClosure* closure, |
| 236 | IRSB* sb_in, |
sewardj | 7ce6239 | 2006-10-15 12:48:18 +0000 | [diff] [blame] | 237 | VexGuestLayout* layout, |
| 238 | VexGuestExtents* vge, |
| 239 | IRType gWordTy, |
| 240 | IRType hWordTy), |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 241 | |
| 242 | // Finish up, print out any results, etc. `exitcode' is program's exit |
| 243 | // code. The shadow can be found with VG_(get_exit_status_shadow)(). |
| 244 | void (*fini)(Int) |
| 245 | ); |
| 246 | |
| 247 | /* ------------------------------------------------------------------ */ |
| 248 | /* Details */ |
| 249 | |
| 250 | /* Default value for avg_translations_sizeB (in bytes), indicating typical |
| 251 | code expansion of about 6:1. */ |
sewardj | 9644cfd | 2006-10-17 02:25:50 +0000 | [diff] [blame] | 252 | #define VG_DEFAULT_TRANS_SIZEB 172 |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 253 | |
| 254 | /* Information used in the startup message. `name' also determines the |
| 255 | string used for identifying suppressions in a suppression file as |
| 256 | belonging to this tool. `version' can be NULL, in which case (not |
| 257 | surprisingly) no version info is printed; this mechanism is designed for |
| 258 | tools distributed with Valgrind that share a version number with |
| 259 | Valgrind. Other tools not distributed as part of Valgrind should |
| 260 | probably have their own version number. */ |
| 261 | extern void VG_(details_name) ( Char* name ); |
| 262 | extern void VG_(details_version) ( Char* version ); |
| 263 | extern void VG_(details_description) ( Char* description ); |
| 264 | extern void VG_(details_copyright_author) ( Char* copyright_author ); |
| 265 | |
| 266 | /* Average size of a translation, in bytes, so that the translation |
| 267 | storage machinery can allocate memory appropriately. Not critical, |
| 268 | setting is optional. */ |
| 269 | extern void VG_(details_avg_translation_sizeB) ( UInt size ); |
| 270 | |
| 271 | /* String printed if an `tl_assert' assertion fails or VG_(tool_panic) |
| 272 | is called. Should probably be an email address. */ |
| 273 | extern void VG_(details_bug_reports_to) ( Char* bug_reports_to ); |
| 274 | |
| 275 | /* ------------------------------------------------------------------ */ |
| 276 | /* Needs */ |
| 277 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 278 | /* Should __libc_freeres() be run? Bugs in it can crash the tool. */ |
| 279 | extern void VG_(needs_libc_freeres) ( void ); |
| 280 | |
| 281 | /* Want to have errors detected by Valgrind's core reported? Includes: |
njn | 0087c50 | 2005-07-01 04:15:36 +0000 | [diff] [blame] | 282 | - pthread API errors (many; eg. unlocking a non-locked mutex) |
| 283 | [currently disabled] |
| 284 | - invalid file descriptors to syscalls like read() and write() |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 285 | - bad signal numbers passed to sigaction() |
| 286 | - attempt to install signal handler for SIGKILL or SIGSTOP */ |
| 287 | extern void VG_(needs_core_errors) ( void ); |
| 288 | |
| 289 | /* Booleans that indicate extra operations are defined; if these are True, |
| 290 | the corresponding template functions (given below) must be defined. A |
| 291 | lot like being a member of a type class. */ |
| 292 | |
| 293 | /* Want to report errors from tool? This implies use of suppressions, too. */ |
| 294 | extern void VG_(needs_tool_errors) ( |
| 295 | // Identify if two errors are equal, or equal enough. `res' indicates how |
| 296 | // close is "close enough". `res' should be passed on as necessary, eg. if |
| 297 | // the Error's `extra' part contains an ExeContext, `res' should be |
| 298 | // passed to VG_(eq_ExeContext)() if the ExeContexts are considered. Other |
| 299 | // than that, probably don't worry about it unless you have lots of very |
| 300 | // similar errors occurring. |
| 301 | Bool (*eq_Error)(VgRes res, Error* e1, Error* e2), |
| 302 | |
| 303 | // Print error context. |
| 304 | void (*pp_Error)(Error* err), |
| 305 | |
| 306 | // Should fill in any details that could be postponed until after the |
| 307 | // decision whether to ignore the error (ie. details not affecting the |
| 308 | // result of VG_(tdict).tool_eq_Error()). This saves time when errors |
| 309 | // are ignored. |
| 310 | // Yuk. |
| 311 | // Return value: must be the size of the `extra' part in bytes -- used by |
| 312 | // the core to make a copy. |
| 313 | UInt (*update_extra)(Error* err), |
| 314 | |
| 315 | // Return value indicates recognition. If recognised, must set skind using |
| 316 | // VG_(set_supp_kind)(). |
| 317 | Bool (*recognised_suppression)(Char* name, Supp* su), |
| 318 | |
| 319 | // Read any extra info for this suppression kind. Most likely for filling |
| 320 | // in the `extra' and `string' parts (with VG_(set_supp_{extra, string})()) |
| 321 | // of a suppression if necessary. Should return False if a syntax error |
| 322 | // occurred, True otherwise. |
| 323 | Bool (*read_extra_suppression_info)(Int fd, Char* buf, Int nBuf, Supp* su), |
| 324 | |
| 325 | // This should just check the kinds match and maybe some stuff in the |
| 326 | // `string' and `extra' field if appropriate (using VG_(get_supp_*)() to |
| 327 | // get the relevant suppression parts). |
| 328 | Bool (*error_matches_suppression)(Error* err, Supp* su), |
| 329 | |
| 330 | // This should return the suppression name, for --gen-suppressions, or NULL |
| 331 | // if that error type cannot be suppressed. This is the inverse of |
| 332 | // VG_(tdict).tool_recognised_suppression(). |
| 333 | Char* (*get_error_name)(Error* err), |
| 334 | |
| 335 | // This should print any extra info for the error, for --gen-suppressions, |
| 336 | // including the newline. This is the inverse of |
| 337 | // VG_(tdict).tool_read_extra_suppression_info(). |
| 338 | void (*print_extra_suppression_info)(Error* err) |
| 339 | ); |
| 340 | |
sewardj | 5155dec | 2005-10-12 10:09:23 +0000 | [diff] [blame] | 341 | /* Is information kept by the tool about specific instructions or |
| 342 | translations? (Eg. for cachegrind there are cost-centres for every |
| 343 | instruction, stored in a per-translation fashion.) If so, the info |
| 344 | may have to be discarded when translations are unloaded (eg. due to |
| 345 | .so unloading, or otherwise at the discretion of m_transtab, eg |
| 346 | when the table becomes too full) to avoid stale information being |
| 347 | reused for new translations. */ |
sewardj | 0b9d74a | 2006-12-24 02:24:11 +0000 | [diff] [blame] | 348 | extern void VG_(needs_superblock_discards) ( |
sewardj | 5155dec | 2005-10-12 10:09:23 +0000 | [diff] [blame] | 349 | // Discard any information that pertains to specific translations |
sewardj | 4ba057c | 2005-10-18 12:04:18 +0000 | [diff] [blame] | 350 | // or instructions within the address range given. There are two |
| 351 | // possible approaches. |
| 352 | // - If info is being stored at a per-translation level, use orig_addr |
| 353 | // to identify which translation is being discarded. Each translation |
| 354 | // will be discarded exactly once. |
sewardj | 7ce6239 | 2006-10-15 12:48:18 +0000 | [diff] [blame] | 355 | // This orig_addr will match the closure->nraddr which was passed to |
| 356 | // to instrument() (see extensive comments above) when this |
| 357 | // translation was made. Note that orig_addr won't necessarily be |
| 358 | // the same as the first address in "extents". |
sewardj | 5155dec | 2005-10-12 10:09:23 +0000 | [diff] [blame] | 359 | // - If info is being stored at a per-instruction level, you can get |
| 360 | // the address range(s) being discarded by stepping through "extents". |
| 361 | // Note that any single instruction may belong to more than one |
| 362 | // translation, and so could be covered by the "extents" of more than |
| 363 | // one call to this function. |
| 364 | // Doing it the first way (as eg. Cachegrind does) is probably easier. |
sewardj | 0b9d74a | 2006-12-24 02:24:11 +0000 | [diff] [blame] | 365 | void (*discard_superblock_info)(Addr64 orig_addr, VexGuestExtents extents) |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 366 | ); |
| 367 | |
| 368 | /* Tool defines its own command line options? */ |
| 369 | extern void VG_(needs_command_line_options) ( |
| 370 | // Return True if option was recognised. Presumably sets some state to |
njn | a007859 | 2007-03-27 06:46:03 +0000 | [diff] [blame] | 371 | // record the option as well. Nb: tools can assume that the argv will |
| 372 | // never disappear. So they can, for example, store a pointer to a string |
| 373 | // within an option, rather than having to make a copy. |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 374 | Bool (*process_cmd_line_option)(Char* argv), |
| 375 | |
| 376 | // Print out command line usage for options for normal tool operation. |
| 377 | void (*print_usage)(void), |
| 378 | |
| 379 | // Print out command line usage for options for debugging the tool. |
| 380 | void (*print_debug_usage)(void) |
| 381 | ); |
| 382 | |
| 383 | /* Tool defines its own client requests? */ |
| 384 | extern void VG_(needs_client_requests) ( |
| 385 | // If using client requests, the number of the first request should be equal |
| 386 | // to VG_USERREQ_TOOL_BASE('X', 'Y'), where 'X' and 'Y' form a suitable two |
| 387 | // character identification for the string. The second and subsequent |
| 388 | // requests should follow. |
| 389 | // |
| 390 | // This function should use the VG_IS_TOOL_USERREQ macro (in |
| 391 | // include/valgrind.h) to first check if it's a request for this tool. Then |
| 392 | // should handle it if it's recognised (and return True), or return False if |
| 393 | // not recognised. arg_block[0] holds the request number, any further args |
| 394 | // from the request are in arg_block[1..]. 'ret' is for the return value... |
| 395 | // it should probably be filled, if only with 0. |
| 396 | Bool (*handle_client_request)(ThreadId tid, UWord* arg_block, UWord* ret) |
| 397 | ); |
| 398 | |
| 399 | /* Tool does stuff before and/or after system calls? */ |
| 400 | // Nb: If either of the pre_ functions malloc() something to return, the |
| 401 | // corresponding post_ function had better free() it! |
| 402 | extern void VG_(needs_syscall_wrapper) ( |
| 403 | void (* pre_syscall)(ThreadId tid, UInt syscallno), |
sewardj | a8d8e23 | 2005-06-07 20:04:56 +0000 | [diff] [blame] | 404 | void (*post_syscall)(ThreadId tid, UInt syscallno, SysRes res) |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 405 | ); |
| 406 | |
| 407 | /* Are tool-state sanity checks performed? */ |
| 408 | // Can be useful for ensuring a tool's correctness. cheap_sanity_check() |
| 409 | // is called very frequently; expensive_sanity_check() is called less |
| 410 | // frequently and can be more involved. |
| 411 | extern void VG_(needs_sanity_checks) ( |
| 412 | Bool(*cheap_sanity_check)(void), |
| 413 | Bool(*expensive_sanity_check)(void) |
| 414 | ); |
| 415 | |
| 416 | /* Do we need to see data symbols? */ |
| 417 | extern void VG_(needs_data_syms) ( void ); |
| 418 | |
njn | 09ca09b | 2005-10-16 17:48:09 +0000 | [diff] [blame] | 419 | /* Does the tool replace malloc() and friends with its own versions? |
| 420 | This has to be combined with the use of a vgpreload_<tool>.so module |
| 421 | or it won't work. See massif/Makefile.am for how to build it. */ |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 422 | // The 'p' prefix avoids GCC complaints about overshadowing global names. |
njn | fc51f8d | 2005-06-21 03:20:17 +0000 | [diff] [blame] | 423 | extern void VG_(needs_malloc_replacement)( |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 424 | void* (*pmalloc) ( ThreadId tid, SizeT n ), |
| 425 | void* (*p__builtin_new) ( ThreadId tid, SizeT n ), |
| 426 | void* (*p__builtin_vec_new) ( ThreadId tid, SizeT n ), |
| 427 | void* (*pmemalign) ( ThreadId tid, SizeT align, SizeT n ), |
| 428 | void* (*pcalloc) ( ThreadId tid, SizeT nmemb, SizeT size1 ), |
| 429 | void (*pfree) ( ThreadId tid, void* p ), |
| 430 | void (*p__builtin_delete) ( ThreadId tid, void* p ), |
| 431 | void (*p__builtin_vec_delete) ( ThreadId tid, void* p ), |
| 432 | void* (*prealloc) ( ThreadId tid, void* p, SizeT new_size ), |
| 433 | SizeT client_malloc_redzone_szB |
| 434 | ); |
| 435 | |
njn | ca54af3 | 2006-04-16 10:25:43 +0000 | [diff] [blame] | 436 | /* Can the tool do XML output? This is a slight misnomer, because the tool |
| 437 | * is not requesting the core to do anything, rather saying "I can handle |
| 438 | * it". */ |
| 439 | extern void VG_(needs_xml_output)( void ); |
| 440 | |
sewardj | 81651dc | 2007-08-28 06:05:20 +0000 | [diff] [blame] | 441 | /* Does the tool want to have one final pass over the IR after tree |
| 442 | building but before instruction selection? If so specify the |
| 443 | function here. */ |
| 444 | extern void VG_(needs_final_IR_tidy_pass) ( IRSB*(*final_tidy)(IRSB*) ); |
| 445 | |
| 446 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 447 | /* ------------------------------------------------------------------ */ |
| 448 | /* Core events to track */ |
| 449 | |
| 450 | /* Part of the core from which this call was made. Useful for determining |
| 451 | what kind of error message should be emitted. */ |
| 452 | typedef |
njn | bb6311c | 2006-12-15 04:37:25 +0000 | [diff] [blame] | 453 | enum { Vg_CoreStartup, Vg_CoreSignal, Vg_CoreSysCall, |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 454 | Vg_CoreTranslate, Vg_CoreClientReq } |
| 455 | CorePart; |
| 456 | |
| 457 | /* Events happening in core to track. To be notified, pass a callback |
| 458 | function to the appropriate function. To ignore an event, don't do |
| 459 | anything (the default is for events to be ignored). |
| 460 | |
| 461 | Note that most events aren't passed a ThreadId. If the event is one called |
| 462 | from generated code (eg. new_mem_stack_*), you can use |
| 463 | VG_(get_running_tid)() to find it. Otherwise, it has to be passed in, |
| 464 | as in pre_mem_read, and so the event signature will require changing. |
| 465 | |
| 466 | Memory events (Nb: to track heap allocation/freeing, a tool must replace |
| 467 | malloc() et al. See above how to do this.) |
| 468 | |
| 469 | These ones occur at startup, upon some signals, and upon some syscalls |
| 470 | */ |
| 471 | void VG_(track_new_mem_startup) (void(*f)(Addr a, SizeT len, |
| 472 | Bool rr, Bool ww, Bool xx)); |
| 473 | void VG_(track_new_mem_stack_signal)(void(*f)(Addr a, SizeT len)); |
| 474 | void VG_(track_new_mem_brk) (void(*f)(Addr a, SizeT len)); |
| 475 | void VG_(track_new_mem_mmap) (void(*f)(Addr a, SizeT len, |
| 476 | Bool rr, Bool ww, Bool xx)); |
| 477 | |
| 478 | void VG_(track_copy_mem_remap) (void(*f)(Addr from, Addr to, SizeT len)); |
| 479 | void VG_(track_change_mem_mprotect) (void(*f)(Addr a, SizeT len, |
| 480 | Bool rr, Bool ww, Bool xx)); |
| 481 | void VG_(track_die_mem_stack_signal)(void(*f)(Addr a, SizeT len)); |
| 482 | void VG_(track_die_mem_brk) (void(*f)(Addr a, SizeT len)); |
| 483 | void VG_(track_die_mem_munmap) (void(*f)(Addr a, SizeT len)); |
| 484 | |
| 485 | /* These ones are called when SP changes. A tool could track these itself |
| 486 | (except for ban_mem_stack) but it's much easier to use the core's help. |
| 487 | |
| 488 | The specialised ones are called in preference to the general one, if they |
| 489 | are defined. These functions are called a lot if they are used, so |
| 490 | specialising can optimise things significantly. If any of the |
| 491 | specialised cases are defined, the general case must be defined too. |
| 492 | |
njn | af839f5 | 2005-06-23 03:27:57 +0000 | [diff] [blame] | 493 | Nb: all the specialised ones must use the VG_REGPARM(n) attribute. |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 494 | */ |
sewardj | f5c8e37 | 2006-02-12 15:42:20 +0000 | [diff] [blame] | 495 | void VG_(track_new_mem_stack_4) (VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 496 | void VG_(track_new_mem_stack_8) (VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 497 | void VG_(track_new_mem_stack_12) (VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 498 | void VG_(track_new_mem_stack_16) (VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 499 | void VG_(track_new_mem_stack_32) (VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 500 | void VG_(track_new_mem_stack_112)(VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 501 | void VG_(track_new_mem_stack_128)(VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 502 | void VG_(track_new_mem_stack_144)(VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 503 | void VG_(track_new_mem_stack_160)(VG_REGPARM(1) void(*f)(Addr new_ESP)); |
| 504 | void VG_(track_new_mem_stack) (void(*f)(Addr a, SizeT len)); |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 505 | |
sewardj | f5c8e37 | 2006-02-12 15:42:20 +0000 | [diff] [blame] | 506 | void VG_(track_die_mem_stack_4) (VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 507 | void VG_(track_die_mem_stack_8) (VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 508 | void VG_(track_die_mem_stack_12) (VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 509 | void VG_(track_die_mem_stack_16) (VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 510 | void VG_(track_die_mem_stack_32) (VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 511 | void VG_(track_die_mem_stack_112)(VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 512 | void VG_(track_die_mem_stack_128)(VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 513 | void VG_(track_die_mem_stack_144)(VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 514 | void VG_(track_die_mem_stack_160)(VG_REGPARM(1) void(*f)(Addr die_ESP)); |
| 515 | void VG_(track_die_mem_stack) (void(*f)(Addr a, SizeT len)); |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 516 | |
| 517 | /* Used for redzone at end of thread stacks */ |
| 518 | void VG_(track_ban_mem_stack) (void(*f)(Addr a, SizeT len)); |
| 519 | |
| 520 | /* These ones occur around syscalls, signal handling, etc */ |
| 521 | void VG_(track_pre_mem_read) (void(*f)(CorePart part, ThreadId tid, |
| 522 | Char* s, Addr a, SizeT size)); |
| 523 | void VG_(track_pre_mem_read_asciiz)(void(*f)(CorePart part, ThreadId tid, |
| 524 | Char* s, Addr a)); |
| 525 | void VG_(track_pre_mem_write) (void(*f)(CorePart part, ThreadId tid, |
| 526 | Char* s, Addr a, SizeT size)); |
| 527 | void VG_(track_post_mem_write) (void(*f)(CorePart part, ThreadId tid, |
| 528 | Addr a, SizeT size)); |
| 529 | |
| 530 | /* Register events. Use VG_(set_shadow_state_area)() to set the shadow regs |
| 531 | for these events. */ |
| 532 | void VG_(track_pre_reg_read) (void(*f)(CorePart part, ThreadId tid, |
| 533 | Char* s, OffT guest_state_offset, |
| 534 | SizeT size)); |
| 535 | void VG_(track_post_reg_write)(void(*f)(CorePart part, ThreadId tid, |
| 536 | OffT guest_state_offset, |
| 537 | SizeT size)); |
| 538 | |
| 539 | /* This one is called for malloc() et al if they are replaced by a tool. */ |
| 540 | void VG_(track_post_reg_write_clientcall_return)( |
| 541 | void(*f)(ThreadId tid, OffT guest_state_offset, SizeT size, Addr f)); |
| 542 | |
| 543 | |
| 544 | /* Scheduler events (not exhaustive) */ |
sewardj | 9756181 | 2006-12-23 01:21:12 +0000 | [diff] [blame] | 545 | |
| 546 | /* Called when 'tid' starts or stops running client code blocks. |
njn | 3e32c87 | 2006-12-24 07:51:17 +0000 | [diff] [blame] | 547 | Gives the total dispatched block count at that event. Note, this is |
| 548 | not the same as 'tid' holding the BigLock (the lock that ensures that |
| 549 | only one thread runs at a time): a thread can hold the lock for other |
| 550 | purposes (making translations, etc) yet not be running client blocks. |
| 551 | Obviously though, a thread must hold the lock in order to run client |
| 552 | code blocks, so the times bracketed by 'thread_run'..'thread_runstate' |
| 553 | are a subset of the times when thread 'tid' holds the cpu lock. |
sewardj | 9756181 | 2006-12-23 01:21:12 +0000 | [diff] [blame] | 554 | */ |
njn | 3e32c87 | 2006-12-24 07:51:17 +0000 | [diff] [blame] | 555 | void VG_(track_start_client_code)( |
| 556 | void(*f)(ThreadId tid, ULong blocks_dispatched) |
| 557 | ); |
| 558 | void VG_(track_stop_client_code)( |
| 559 | void(*f)(ThreadId tid, ULong blocks_dispatched) |
sewardj | 9756181 | 2006-12-23 01:21:12 +0000 | [diff] [blame] | 560 | ); |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 561 | |
| 562 | |
| 563 | /* Thread events (not exhaustive) |
| 564 | |
| 565 | Called during thread create, before the new thread has run any |
| 566 | instructions (or touched any memory). |
| 567 | */ |
| 568 | void VG_(track_post_thread_create)(void(*f)(ThreadId tid, ThreadId child)); |
| 569 | void VG_(track_post_thread_join) (void(*f)(ThreadId joiner, ThreadId joinee)); |
| 570 | |
njn | 3e32c87 | 2006-12-24 07:51:17 +0000 | [diff] [blame] | 571 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 572 | /* Signal events (not exhaustive) |
| 573 | |
| 574 | ... pre_send_signal, post_send_signal ... |
| 575 | |
| 576 | Called before a signal is delivered; `alt_stack' indicates if it is |
| 577 | delivered on an alternative stack. */ |
| 578 | void VG_(track_pre_deliver_signal) (void(*f)(ThreadId tid, Int sigNo, |
| 579 | Bool alt_stack)); |
| 580 | /* Called after a signal is delivered. Nb: unfortunately, if the signal |
| 581 | handler longjmps, this won't be called. */ |
| 582 | void VG_(track_post_deliver_signal)(void(*f)(ThreadId tid, Int sigNo)); |
| 583 | |
njn | 43b9a8a | 2005-05-10 04:37:01 +0000 | [diff] [blame] | 584 | #endif // __PUB_TOOL_TOOLIFACE_H |
| 585 | |
| 586 | /*--------------------------------------------------------------------*/ |
| 587 | /*--- end ---*/ |
| 588 | /*--------------------------------------------------------------------*/ |