DccFrontend.cpp 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414
  1. #include "DccFrontend.h"
  2. #include "dcc.h"
  3. #include "msvc_fixes.h"
  4. #include "project.h"
  5. #include "disassem.h"
  6. #include "CallGraph.h"
  7. #include <QtCore/QFileInfo>
  8. #include <QtCore/QDebug>
  9. #include <cstdio>
  10. class Loader
  11. {
  12. bool loadIntoProject(IProject *);
  13. };
  14. struct PSP { /* PSP structure */
  15. uint16_t int20h; /* interrupt 20h */
  16. uint16_t eof; /* segment, end of allocation block */
  17. uint8_t res1; /* reserved */
  18. uint8_t dosDisp[5]; /* far call to DOS function dispatcher */
  19. uint8_t int22h[4]; /* vector for terminate routine */
  20. uint8_t int23h[4]; /* vector for ctrl+break routine */
  21. uint8_t int24h[4]; /* vector for error routine */
  22. uint8_t res2[22]; /* reserved */
  23. uint16_t segEnv; /* segment address of environment block */
  24. uint8_t res3[34]; /* reserved */
  25. uint8_t int21h[6]; /* opcode for int21h and far return */
  26. uint8_t res4[6]; /* reserved */
  27. uint8_t fcb1[16]; /* default file control block 1 */
  28. uint8_t fcb2[16]; /* default file control block 2 */
  29. uint8_t res5[4]; /* reserved */
  30. uint8_t cmdTail[0x80]; /* command tail and disk transfer area */
  31. };
  32. static struct MZHeader { /* EXE file header */
  33. uint8_t sigLo; /* .EXE signature: 0x4D 0x5A */
  34. uint8_t sigHi;
  35. uint16_t lastPageSize; /* Size of the last page */
  36. uint16_t numPages; /* Number of pages in the file */
  37. uint16_t numReloc; /* Number of relocation items */
  38. uint16_t numParaHeader; /* # of paragraphs in the header */
  39. uint16_t minAlloc; /* Minimum number of paragraphs */
  40. uint16_t maxAlloc; /* Maximum number of paragraphs */
  41. uint16_t initSS; /* Segment displacement of stack */
  42. uint16_t initSP; /* Contents of SP at entry */
  43. uint16_t checkSum; /* Complemented checksum */
  44. uint16_t initIP; /* Contents of IP at entry */
  45. uint16_t initCS; /* Segment displacement of code */
  46. uint16_t relocTabOffset; /* Relocation table offset */
  47. uint16_t overlayNum; /* Overlay number */
  48. } header;
  49. #define EXE_RELOCATION 0x10 /* EXE images rellocated to above PSP */
  50. //static void LoadImage(char *filename);
  51. static void displayMemMap(void);
  52. /****************************************************************************
  53. * displayLoadInfo - Displays low level loader type info.
  54. ***************************************************************************/
  55. void PROG::displayLoadInfo(void)
  56. {
  57. int i;
  58. printf("File type is %s\n", (fCOM)?"COM":"EXE");
  59. if (not fCOM) {
  60. printf("Signature = %02X%02X\n", header.sigLo, header.sigHi);
  61. printf("File size %% 512 = %04X\n", LH(&header.lastPageSize));
  62. printf("File size / 512 = %04X pages\n", LH(&header.numPages));
  63. printf("# relocation items = %04X\n", LH(&header.numReloc));
  64. printf("Offset to load image = %04X paras\n", LH(&header.numParaHeader));
  65. printf("Minimum allocation = %04X paras\n", LH(&header.minAlloc));
  66. printf("Maximum allocation = %04X paras\n", LH(&header.maxAlloc));
  67. }
  68. printf("Load image size = %04" PRIiPTR "\n", cbImage - sizeof(PSP));
  69. printf("Initial SS:SP = %04X:%04X\n", initSS, initSP);
  70. printf("Initial CS:IP = %04X:%04X\n", initCS, initIP);
  71. if (option.VeryVerbose and cReloc)
  72. {
  73. printf("\nRelocation Table\n");
  74. for (i = 0; i < cReloc; i++)
  75. {
  76. printf("%06X -> [%04X]\n", relocTable[i],LH(image() + relocTable[i]));
  77. }
  78. }
  79. printf("\n");
  80. }
  81. /*****************************************************************************
  82. * fill - Fills line for displayMemMap()
  83. ****************************************************************************/
  84. static void fill(int ip, char *bf)
  85. {
  86. PROG &prog(Project::get()->prog);
  87. static uint8_t type[4] = {'.', 'd', 'c', 'x'};
  88. uint8_t i;
  89. for (i = 0; i < 16; i++, ip++)
  90. {
  91. *bf++ = ' ';
  92. *bf++ = (ip < prog.cbImage)? type[(prog.map[ip >> 2] >> ((ip & 3) * 2)) & 3]: ' ';
  93. }
  94. *bf = '\0';
  95. }
  96. /*****************************************************************************
  97. * displayMemMap - Displays the memory bitmap
  98. ****************************************************************************/
  99. static void displayMemMap(void)
  100. {
  101. PROG &prog(Project::get()->prog);
  102. char c, b1[33], b2[33], b3[33];
  103. uint8_t i;
  104. int ip = 0;
  105. printf("\nMemory Map\n");
  106. while (ip < prog.cbImage)
  107. {
  108. fill(ip, b1);
  109. printf("%06X %s\n", ip, b1);
  110. ip += 16;
  111. for (i = 3, c = b1[1]; i < 32 and c == b1[i]; i += 2)
  112. ; /* Check if all same */
  113. if (i > 32)
  114. {
  115. fill(ip, b2); /* Skip until next two are not same */
  116. fill(ip+16, b3);
  117. if (not (strcmp(b1, b2) || strcmp(b1, b3)))
  118. {
  119. printf(" :\n");
  120. do
  121. {
  122. ip += 16;
  123. fill(ip+16, b1);
  124. } while (0==strcmp(b1, b2));
  125. }
  126. }
  127. }
  128. printf("\n");
  129. }
  130. DccFrontend::DccFrontend(QObject *parent) :
  131. QObject(parent)
  132. {
  133. }
  134. /*****************************************************************************
  135. * FrontEnd - invokes the loader, parser, disassembler (if asm1), icode
  136. * rewritter, and displays any useful information.
  137. ****************************************************************************/
  138. bool DccFrontend::FrontEnd ()
  139. {
  140. /* Do depth first flow analysis building call graph and procedure list,
  141. * and attaching the I-code to each procedure */
  142. parse (*Project::get());
  143. if (option.asm1)
  144. {
  145. qWarning() << "dcc: writing assembler file "<<asm1_name<<'\n';
  146. }
  147. /* Search through code looking for impure references and flag them */
  148. Disassembler ds(1);
  149. for(Function &f : Project::get()->pProcList)
  150. {
  151. f.markImpure();
  152. if (option.asm1)
  153. {
  154. ds.disassem(&f);
  155. }
  156. }
  157. if (option.Interact)
  158. {
  159. interactDis(&Project::get()->pProcList.front(), 0); /* Interactive disassembler */
  160. }
  161. /* Converts jump target addresses to icode offsets */
  162. for(Function &f : Project::get()->pProcList)
  163. {
  164. f.bindIcodeOff();
  165. }
  166. /* Print memory bitmap */
  167. if (option.Map)
  168. displayMemMap();
  169. return(true); // we no longer own proj !
  170. }
  171. struct DosLoader {
  172. protected:
  173. void prepareImage(PROG &prog,size_t sz,QFile &fp) {
  174. /* Allocate a block of memory for the program. */
  175. prog.cbImage = sz + sizeof(PSP);
  176. prog.Imagez = new uint8_t [prog.cbImage];
  177. prog.Imagez[0] = 0xCD; /* Fill in PSP int 20h location */
  178. prog.Imagez[1] = 0x20; /* for termination checking */
  179. /* Read in the image past where a PSP would go */
  180. if (sz != fp.read((char *)prog.Imagez + sizeof(PSP),sz))
  181. fatalError(CANNOT_READ, fp.fileName().toLocal8Bit().data());
  182. }
  183. };
  184. struct ComLoader : public DosLoader {
  185. bool canLoad(QFile &fp) {
  186. fp.seek(0);
  187. char sig[2];
  188. if(2==fp.read(sig,2)) {
  189. return not (sig[0] == 0x4D and sig[1] == 0x5A);
  190. }
  191. return false;
  192. }
  193. bool load(PROG &prog,QFile &fp) {
  194. fp.seek(0);
  195. /* COM file
  196. * In this case the load module size is just the file length
  197. */
  198. auto cb = fp.size();
  199. /* COM programs start off with an ORG 100H (to leave room for a PSP)
  200. * This is also the implied start address so if we load the image
  201. * at offset 100H addresses should all line up properly again.
  202. */
  203. prog.initCS = 0;
  204. prog.initIP = 0x100;
  205. prog.initSS = 0;
  206. prog.initSP = 0xFFFE;
  207. prog.cReloc = 0;
  208. prepareImage(prog,cb,fp);
  209. /* Set up memory map */
  210. cb = (prog.cbImage + 3) / 4;
  211. prog.map = (uint8_t *)malloc(cb);
  212. memset(prog.map, BM_UNKNOWN, (size_t)cb);
  213. return true;
  214. }
  215. };
  216. struct ExeLoader : public DosLoader {
  217. bool canLoad(QFile &fp) {
  218. if(fp.size()<sizeof(header))
  219. return false;
  220. MZHeader tmp_header;
  221. fp.seek(0);
  222. fp.read((char *)&tmp_header, sizeof(header));
  223. if(not (tmp_header.sigLo == 0x4D and tmp_header.sigHi == 0x5A))
  224. return false;
  225. /* This is a typical DOS kludge! */
  226. if (LH(&header.relocTabOffset) == 0x40)
  227. {
  228. qDebug() << "Don't understand new EXE format";
  229. return false;
  230. }
  231. return true;
  232. }
  233. bool load(PROG &prog,QFile &fp) {
  234. /* Read rest of header */
  235. fp.seek(0);
  236. if (fp.read((char *)&header, sizeof(header)) != sizeof(header))
  237. return false;
  238. /* Calculate the load module size.
  239. * This is the number of pages in the file
  240. * less the length of the header and reloc table
  241. * less the number of bytes unused on last page
  242. */
  243. uint32_t cb = (uint32_t)LH(&header.numPages) * 512 - (uint32_t)LH(&header.numParaHeader) * 16;
  244. if (header.lastPageSize)
  245. {
  246. cb -= 512 - LH(&header.lastPageSize);
  247. }
  248. /* We quietly ignore minAlloc and maxAlloc since for our
  249. * purposes it doesn't really matter where in real memory
  250. * the program would end up. EXE programs can't really rely on
  251. * their load location so setting the PSP segment to 0 is fine.
  252. * Certainly programs that prod around in DOS or BIOS are going
  253. * to have to load DS from a constant so it'll be pretty
  254. * obvious.
  255. */
  256. prog.initCS = (int16_t)LH(&header.initCS) + EXE_RELOCATION;
  257. prog.initIP = (int16_t)LH(&header.initIP);
  258. prog.initSS = (int16_t)LH(&header.initSS) + EXE_RELOCATION;
  259. prog.initSP = (int16_t)LH(&header.initSP);
  260. prog.cReloc = (int16_t)LH(&header.numReloc);
  261. /* Allocate the relocation table */
  262. if (prog.cReloc)
  263. {
  264. prog.relocTable.resize(prog.cReloc);
  265. fp.seek(LH(&header.relocTabOffset));
  266. /* Read in seg:offset pairs and convert to Image ptrs */
  267. uint8_t buf[4];
  268. for (int i = 0; i < prog.cReloc; i++)
  269. {
  270. fp.read((char *)buf,4);
  271. prog.relocTable[i] = LH(buf) + (((int)LH(buf+2) + EXE_RELOCATION)<<4);
  272. }
  273. }
  274. /* Seek to start of image */
  275. uint32_t start_of_image= LH(&header.numParaHeader) * 16;
  276. fp.seek(start_of_image);
  277. /* Allocate a block of memory for the program. */
  278. prepareImage(prog,cb,fp);
  279. /* Set up memory map */
  280. cb = (prog.cbImage + 3) / 4;
  281. prog.map = (uint8_t *)malloc(cb);
  282. memset(prog.map, BM_UNKNOWN, (size_t)cb);
  283. /* Relocate segment constants */
  284. for(uint32_t v : prog.relocTable) {
  285. uint8_t *p = &prog.Imagez[v];
  286. uint16_t w = (uint16_t)LH(p) + EXE_RELOCATION;
  287. *p++ = (uint8_t)(w & 0x00FF);
  288. *p = (uint8_t)((w & 0xFF00) >> 8);
  289. }
  290. return true;
  291. }
  292. };
  293. /*****************************************************************************
  294. * LoadImage
  295. ****************************************************************************/
  296. bool Project::load()
  297. {
  298. // addTask(loaderSelection,PreCond(BinaryImage))
  299. // addTask(applyLoader,PreCond(Loader))
  300. const char *fname = binary_path().toLocal8Bit().data();
  301. QFile finfo(binary_path());
  302. /* Open the input file */
  303. if(not finfo.open(QFile::ReadOnly)) {
  304. fatalError(CANNOT_OPEN, fname);
  305. }
  306. /* Read in first 2 bytes to check EXE signature */
  307. if (finfo.size()<=2)
  308. {
  309. fatalError(CANNOT_READ, fname);
  310. }
  311. ComLoader com_loader;
  312. ExeLoader exe_loader;
  313. if(exe_loader.canLoad(finfo)) {
  314. prog.fCOM = false;
  315. return exe_loader.load(prog,finfo);
  316. }
  317. if(com_loader.canLoad(finfo)) {
  318. prog.fCOM = true;
  319. return com_loader.load(prog,finfo);
  320. }
  321. return false;
  322. }
  323. uint32_t SynthLab;
  324. /* Parses the program, builds the call graph, and returns the list of
  325. * procedures found */
  326. void DccFrontend::parse(Project &proj)
  327. {
  328. PROG &prog(proj.prog);
  329. STATE state;
  330. /* Set initial state */
  331. state.setState(rES, 0); /* PSP segment */
  332. state.setState(rDS, 0);
  333. state.setState(rCS, prog.initCS);
  334. state.setState(rSS, prog.initSS);
  335. state.setState(rSP, prog.initSP);
  336. state.IP = ((uint32_t)prog.initCS << 4) + prog.initIP;
  337. SynthLab = SYNTHESIZED_MIN;
  338. /* Check for special settings of initial state, based on idioms of the
  339. startup code */
  340. state.checkStartup();
  341. ilFunction start_proc;
  342. /* Make a struct for the initial procedure */
  343. if (prog.offMain != -1)
  344. {
  345. start_proc = proj.createFunction(0,"main");
  346. start_proc->retVal.loc = REG_FRAME;
  347. start_proc->retVal.type = TYPE_WORD_SIGN;
  348. start_proc->retVal.id.regi = rAX;
  349. /* We know where main() is. Start the flow of control from there */
  350. start_proc->procEntry = prog.offMain;
  351. /* In medium and large models, the segment of main may (will?) not be
  352. the same as the initial CS segment (of the startup code) */
  353. state.setState(rCS, prog.segMain);
  354. state.IP = prog.offMain;
  355. }
  356. else
  357. {
  358. start_proc = proj.createFunction(0,"start");
  359. /* Create initial procedure at program start address */
  360. start_proc->procEntry = (uint32_t)state.IP;
  361. }
  362. /* The state info is for the first procedure */
  363. start_proc->state = state;
  364. /* Set up call graph initial node */
  365. proj.callGraph = new CALL_GRAPH;
  366. proj.callGraph->proc = start_proc;
  367. /* This proc needs to be called to set things up for LibCheck(), which
  368. checks a proc to see if it is a know C (etc) library */
  369. prog.bSigs = SetupLibCheck();
  370. //BUG: proj and g_proj are 'live' at this point !
  371. /* Recursively build entire procedure list */
  372. start_proc->FollowCtrl(proj.callGraph, &state);
  373. /* This proc needs to be called to clean things up from SetupLibCheck() */
  374. CleanupLibCheck();
  375. }