1 /* 2 * User-space Probes (UProbes) for x86 3 * 4 * This program is free software; you can redistribute it and/or modify 5 * it under the terms of the GNU General Public License as published by 6 * the Free Software Foundation; either version 2 of the License, or 7 * (at your option) any later version. 8 * 9 * This program is distributed in the hope that it will be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write to the Free Software 16 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. 17 * 18 * Copyright (C) IBM Corporation, 2008-2011 19 * Authors: 20 * Srikar Dronamraju 21 * Jim Keniston 22 */ 23 #include <linux/kernel.h> 24 #include <linux/sched.h> 25 #include <linux/ptrace.h> 26 #include <linux/uprobes.h> 27 #include <linux/uaccess.h> 28 29 #include <linux/kdebug.h> 30 #include <asm/processor.h> 31 #include <asm/insn.h> 32 33 /* Post-execution fixups. */ 34 35 /* Adjust IP back to vicinity of actual insn */ 36 #define UPROBE_FIX_IP 0x01 37 38 /* Adjust the return address of a call insn */ 39 #define UPROBE_FIX_CALL 0x02 40 41 /* Instruction will modify TF, don't change it */ 42 #define UPROBE_FIX_SETF 0x04 43 44 #define UPROBE_FIX_RIP_SI 0x08 45 #define UPROBE_FIX_RIP_DI 0x10 46 #define UPROBE_FIX_RIP_BX 0x20 47 #define UPROBE_FIX_RIP_MASK \ 48 (UPROBE_FIX_RIP_SI | UPROBE_FIX_RIP_DI | UPROBE_FIX_RIP_BX) 49 50 #define UPROBE_TRAP_NR UINT_MAX 51 52 /* Adaptations for mhiramat x86 decoder v14. */ 53 #define OPCODE1(insn) ((insn)->opcode.bytes[0]) 54 #define OPCODE2(insn) ((insn)->opcode.bytes[1]) 55 #define OPCODE3(insn) ((insn)->opcode.bytes[2]) 56 #define MODRM_REG(insn) X86_MODRM_REG((insn)->modrm.value) 57 58 #define W(row, b0, b1, b2, b3, b4, b5, b6, b7, b8, b9, ba, bb, bc, bd, be, bf)\ 59 (((b0##UL << 0x0)|(b1##UL << 0x1)|(b2##UL << 0x2)|(b3##UL << 0x3) | \ 60 (b4##UL << 0x4)|(b5##UL << 0x5)|(b6##UL << 0x6)|(b7##UL << 0x7) | \ 61 (b8##UL << 0x8)|(b9##UL << 0x9)|(ba##UL << 0xa)|(bb##UL << 0xb) | \ 62 (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf)) \ 63 << (row % 32)) 64 65 /* 66 * Good-instruction tables for 32-bit apps. This is non-const and volatile 67 * to keep gcc from statically optimizing it out, as variable_test_bit makes 68 * some versions of gcc to think only *(unsigned long*) is used. 69 * 70 * Opcodes we'll probably never support: 71 * 6c-6f - ins,outs. SEGVs if used in userspace 72 * e4-e7 - in,out imm. SEGVs if used in userspace 73 * ec-ef - in,out acc. SEGVs if used in userspace 74 * cc - int3. SIGTRAP if used in userspace 75 * ce - into. Not used in userspace - no kernel support to make it useful. SEGVs 76 * (why we support bound (62) then? it's similar, and similarly unused...) 77 * f1 - int1. SIGTRAP if used in userspace 78 * f4 - hlt. SEGVs if used in userspace 79 * fa - cli. SEGVs if used in userspace 80 * fb - sti. SEGVs if used in userspace 81 * 82 * Opcodes which need some work to be supported: 83 * 07,17,1f - pop es/ss/ds 84 * Normally not used in userspace, but would execute if used. 85 * Can cause GP or stack exception if tries to load wrong segment descriptor. 86 * We hesitate to run them under single step since kernel's handling 87 * of userspace single-stepping (TF flag) is fragile. 88 * We can easily refuse to support push es/cs/ss/ds (06/0e/16/1e) 89 * on the same grounds that they are never used. 90 * cd - int N. 91 * Used by userspace for "int 80" syscall entry. (Other "int N" 92 * cause GP -> SEGV since their IDT gates don't allow calls from CPL 3). 93 * Not supported since kernel's handling of userspace single-stepping 94 * (TF flag) is fragile. 95 * cf - iret. Normally not used in userspace. Doesn't SEGV unless arguments are bad 96 */ 97 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION) 98 static volatile u32 good_insns_32[256 / 32] = { 99 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 100 /* ---------------------------------------------- */ 101 W(0x00, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1) | /* 00 */ 102 W(0x10, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0) , /* 10 */ 103 W(0x20, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 20 */ 104 W(0x30, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 30 */ 105 W(0x40, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 40 */ 106 W(0x50, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 50 */ 107 W(0x60, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* 60 */ 108 W(0x70, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 70 */ 109 W(0x80, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 80 */ 110 W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */ 111 W(0xa0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* a0 */ 112 W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* b0 */ 113 W(0xc0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* c0 */ 114 W(0xd0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* d0 */ 115 W(0xe0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0) | /* e0 */ 116 W(0xf0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1) /* f0 */ 117 /* ---------------------------------------------- */ 118 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 119 }; 120 #else 121 #define good_insns_32 NULL 122 #endif 123 124 /* Good-instruction tables for 64-bit apps. 125 * 126 * Genuinely invalid opcodes: 127 * 06,07 - formerly push/pop es 128 * 0e - formerly push cs 129 * 16,17 - formerly push/pop ss 130 * 1e,1f - formerly push/pop ds 131 * 27,2f,37,3f - formerly daa/das/aaa/aas 132 * 60,61 - formerly pusha/popa 133 * 62 - formerly bound. EVEX prefix for AVX512 (not yet supported) 134 * 82 - formerly redundant encoding of Group1 135 * 9a - formerly call seg:ofs 136 * ce - formerly into 137 * d4,d5 - formerly aam/aad 138 * d6 - formerly undocumented salc 139 * ea - formerly jmp seg:ofs 140 * 141 * Opcodes we'll probably never support: 142 * 6c-6f - ins,outs. SEGVs if used in userspace 143 * e4-e7 - in,out imm. SEGVs if used in userspace 144 * ec-ef - in,out acc. SEGVs if used in userspace 145 * cc - int3. SIGTRAP if used in userspace 146 * f1 - int1. SIGTRAP if used in userspace 147 * f4 - hlt. SEGVs if used in userspace 148 * fa - cli. SEGVs if used in userspace 149 * fb - sti. SEGVs if used in userspace 150 * 151 * Opcodes which need some work to be supported: 152 * cd - int N. 153 * Used by userspace for "int 80" syscall entry. (Other "int N" 154 * cause GP -> SEGV since their IDT gates don't allow calls from CPL 3). 155 * Not supported since kernel's handling of userspace single-stepping 156 * (TF flag) is fragile. 157 * cf - iret. Normally not used in userspace. Doesn't SEGV unless arguments are bad 158 */ 159 #if defined(CONFIG_X86_64) 160 static volatile u32 good_insns_64[256 / 32] = { 161 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 162 /* ---------------------------------------------- */ 163 W(0x00, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1) | /* 00 */ 164 W(0x10, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) , /* 10 */ 165 W(0x20, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0) | /* 20 */ 166 W(0x30, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0) , /* 30 */ 167 W(0x40, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 40 */ 168 W(0x50, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 50 */ 169 W(0x60, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* 60 */ 170 W(0x70, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 70 */ 171 W(0x80, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 80 */ 172 W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1) , /* 90 */ 173 W(0xa0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* a0 */ 174 W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* b0 */ 175 W(0xc0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* c0 */ 176 W(0xd0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* d0 */ 177 W(0xe0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0) | /* e0 */ 178 W(0xf0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1) /* f0 */ 179 /* ---------------------------------------------- */ 180 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 181 }; 182 #else 183 #define good_insns_64 NULL 184 #endif 185 186 /* Using this for both 64-bit and 32-bit apps. 187 * Opcodes we don't support: 188 * 0f 00 - SLDT/STR/LLDT/LTR/VERR/VERW/-/- group. System insns 189 * 0f 01 - SGDT/SIDT/LGDT/LIDT/SMSW/-/LMSW/INVLPG group. 190 * Also encodes tons of other system insns if mod=11. 191 * Some are in fact non-system: xend, xtest, rdtscp, maybe more 192 * 0f 05 - syscall 193 * 0f 06 - clts (CPL0 insn) 194 * 0f 07 - sysret 195 * 0f 08 - invd (CPL0 insn) 196 * 0f 09 - wbinvd (CPL0 insn) 197 * 0f 0b - ud2 198 * 0f 30 - wrmsr (CPL0 insn) (then why rdmsr is allowed, it's also CPL0 insn?) 199 * 0f 34 - sysenter 200 * 0f 35 - sysexit 201 * 0f 37 - getsec 202 * 0f 78 - vmread (Intel VMX. CPL0 insn) 203 * 0f 79 - vmwrite (Intel VMX. CPL0 insn) 204 * Note: with prefixes, these two opcodes are 205 * extrq/insertq/AVX512 convert vector ops. 206 * 0f ae - group15: [f]xsave,[f]xrstor,[v]{ld,st}mxcsr,clflush[opt], 207 * {rd,wr}{fs,gs}base,{s,l,m}fence. 208 * Why? They are all user-executable. 209 */ 210 static volatile u32 good_2byte_insns[256 / 32] = { 211 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 212 /* ---------------------------------------------- */ 213 W(0x00, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1) | /* 00 */ 214 W(0x10, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 10 */ 215 W(0x20, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 20 */ 216 W(0x30, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1) , /* 30 */ 217 W(0x40, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 40 */ 218 W(0x50, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 50 */ 219 W(0x60, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 60 */ 220 W(0x70, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1) , /* 70 */ 221 W(0x80, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 80 */ 222 W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */ 223 W(0xa0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1) | /* a0 */ 224 W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* b0 */ 225 W(0xc0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* c0 */ 226 W(0xd0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* d0 */ 227 W(0xe0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* e0 */ 228 W(0xf0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) /* f0 */ 229 /* ---------------------------------------------- */ 230 /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */ 231 }; 232 #undef W 233 234 /* 235 * opcodes we may need to refine support for: 236 * 237 * 0f - 2-byte instructions: For many of these instructions, the validity 238 * depends on the prefix and/or the reg field. On such instructions, we 239 * just consider the opcode combination valid if it corresponds to any 240 * valid instruction. 241 * 242 * 8f - Group 1 - only reg = 0 is OK 243 * c6-c7 - Group 11 - only reg = 0 is OK 244 * d9-df - fpu insns with some illegal encodings 245 * f2, f3 - repnz, repz prefixes. These are also the first byte for 246 * certain floating-point instructions, such as addsd. 247 * 248 * fe - Group 4 - only reg = 0 or 1 is OK 249 * ff - Group 5 - only reg = 0-6 is OK 250 * 251 * others -- Do we need to support these? 252 * 253 * 0f - (floating-point?) prefetch instructions 254 * 07, 17, 1f - pop es, pop ss, pop ds 255 * 26, 2e, 36, 3e - es:, cs:, ss:, ds: segment prefixes -- 256 * but 64 and 65 (fs: and gs:) seem to be used, so we support them 257 * 67 - addr16 prefix 258 * ce - into 259 * f0 - lock prefix 260 */ 261 262 /* 263 * TODO: 264 * - Where necessary, examine the modrm byte and allow only valid instructions 265 * in the different Groups and fpu instructions. 266 */ 267 268 static bool is_prefix_bad(struct insn *insn) 269 { 270 int i; 271 272 for (i = 0; i < insn->prefixes.nbytes; i++) { 273 switch (insn->prefixes.bytes[i]) { 274 case 0x26: /* INAT_PFX_ES */ 275 case 0x2E: /* INAT_PFX_CS */ 276 case 0x36: /* INAT_PFX_DS */ 277 case 0x3E: /* INAT_PFX_SS */ 278 case 0xF0: /* INAT_PFX_LOCK */ 279 return true; 280 } 281 } 282 return false; 283 } 284 285 static int uprobe_init_insn(struct arch_uprobe *auprobe, struct insn *insn, bool x86_64) 286 { 287 u32 volatile *good_insns; 288 289 insn_init(insn, auprobe->insn, sizeof(auprobe->insn), x86_64); 290 /* has the side-effect of processing the entire instruction */ 291 insn_get_length(insn); 292 if (WARN_ON_ONCE(!insn_complete(insn))) 293 return -ENOEXEC; 294 295 if (is_prefix_bad(insn)) 296 return -ENOTSUPP; 297 298 if (x86_64) 299 good_insns = good_insns_64; 300 else 301 good_insns = good_insns_32; 302 303 if (test_bit(OPCODE1(insn), (unsigned long *)good_insns)) 304 return 0; 305 306 if (insn->opcode.nbytes == 2) { 307 if (test_bit(OPCODE2(insn), (unsigned long *)good_2byte_insns)) 308 return 0; 309 } 310 311 return -ENOTSUPP; 312 } 313 314 #ifdef CONFIG_X86_64 315 static inline bool is_64bit_mm(struct mm_struct *mm) 316 { 317 return !config_enabled(CONFIG_IA32_EMULATION) || 318 !(mm->context.ia32_compat == TIF_IA32); 319 } 320 /* 321 * If arch_uprobe->insn doesn't use rip-relative addressing, return 322 * immediately. Otherwise, rewrite the instruction so that it accesses 323 * its memory operand indirectly through a scratch register. Set 324 * defparam->fixups accordingly. (The contents of the scratch register 325 * will be saved before we single-step the modified instruction, 326 * and restored afterward). 327 * 328 * We do this because a rip-relative instruction can access only a 329 * relatively small area (+/- 2 GB from the instruction), and the XOL 330 * area typically lies beyond that area. At least for instructions 331 * that store to memory, we can't execute the original instruction 332 * and "fix things up" later, because the misdirected store could be 333 * disastrous. 334 * 335 * Some useful facts about rip-relative instructions: 336 * 337 * - There's always a modrm byte with bit layout "00 reg 101". 338 * - There's never a SIB byte. 339 * - The displacement is always 4 bytes. 340 * - REX.B=1 bit in REX prefix, which normally extends r/m field, 341 * has no effect on rip-relative mode. It doesn't make modrm byte 342 * with r/m=101 refer to register 1101 = R13. 343 */ 344 static void riprel_analyze(struct arch_uprobe *auprobe, struct insn *insn) 345 { 346 u8 *cursor; 347 u8 reg; 348 u8 reg2; 349 350 if (!insn_rip_relative(insn)) 351 return; 352 353 /* 354 * insn_rip_relative() would have decoded rex_prefix, vex_prefix, modrm. 355 * Clear REX.b bit (extension of MODRM.rm field): 356 * we want to encode low numbered reg, not r8+. 357 */ 358 if (insn->rex_prefix.nbytes) { 359 cursor = auprobe->insn + insn_offset_rex_prefix(insn); 360 /* REX byte has 0100wrxb layout, clearing REX.b bit */ 361 *cursor &= 0xfe; 362 } 363 /* 364 * Similar treatment for VEX3 prefix. 365 * TODO: add XOP/EVEX treatment when insn decoder supports them 366 */ 367 if (insn->vex_prefix.nbytes == 3) { 368 /* 369 * vex2: c5 rvvvvLpp (has no b bit) 370 * vex3/xop: c4/8f rxbmmmmm wvvvvLpp 371 * evex: 62 rxbR00mm wvvvv1pp zllBVaaa 372 * (evex will need setting of both b and x since 373 * in non-sib encoding evex.x is 4th bit of MODRM.rm) 374 * Setting VEX3.b (setting because it has inverted meaning): 375 */ 376 cursor = auprobe->insn + insn_offset_vex_prefix(insn) + 1; 377 *cursor |= 0x20; 378 } 379 380 /* 381 * Convert from rip-relative addressing to register-relative addressing 382 * via a scratch register. 383 * 384 * This is tricky since there are insns with modrm byte 385 * which also use registers not encoded in modrm byte: 386 * [i]div/[i]mul: implicitly use dx:ax 387 * shift ops: implicitly use cx 388 * cmpxchg: implicitly uses ax 389 * cmpxchg8/16b: implicitly uses dx:ax and bx:cx 390 * Encoding: 0f c7/1 modrm 391 * The code below thinks that reg=1 (cx), chooses si as scratch. 392 * mulx: implicitly uses dx: mulx r/m,r1,r2 does r1:r2 = dx * r/m. 393 * First appeared in Haswell (BMI2 insn). It is vex-encoded. 394 * Example where none of bx,cx,dx can be used as scratch reg: 395 * c4 e2 63 f6 0d disp32 mulx disp32(%rip),%ebx,%ecx 396 * [v]pcmpistri: implicitly uses cx, xmm0 397 * [v]pcmpistrm: implicitly uses xmm0 398 * [v]pcmpestri: implicitly uses ax, dx, cx, xmm0 399 * [v]pcmpestrm: implicitly uses ax, dx, xmm0 400 * Evil SSE4.2 string comparison ops from hell. 401 * maskmovq/[v]maskmovdqu: implicitly uses (ds:rdi) as destination. 402 * Encoding: 0f f7 modrm, 66 0f f7 modrm, vex-encoded: c5 f9 f7 modrm. 403 * Store op1, byte-masked by op2 msb's in each byte, to (ds:rdi). 404 * AMD says it has no 3-operand form (vex.vvvv must be 1111) 405 * and that it can have only register operands, not mem 406 * (its modrm byte must have mode=11). 407 * If these restrictions will ever be lifted, 408 * we'll need code to prevent selection of di as scratch reg! 409 * 410 * Summary: I don't know any insns with modrm byte which 411 * use SI register implicitly. DI register is used only 412 * by one insn (maskmovq) and BX register is used 413 * only by one too (cmpxchg8b). 414 * BP is stack-segment based (may be a problem?). 415 * AX, DX, CX are off-limits (many implicit users). 416 * SP is unusable (it's stack pointer - think about "pop mem"; 417 * also, rsp+disp32 needs sib encoding -> insn length change). 418 */ 419 420 reg = MODRM_REG(insn); /* Fetch modrm.reg */ 421 reg2 = 0xff; /* Fetch vex.vvvv */ 422 if (insn->vex_prefix.nbytes == 2) 423 reg2 = insn->vex_prefix.bytes[1]; 424 else if (insn->vex_prefix.nbytes == 3) 425 reg2 = insn->vex_prefix.bytes[2]; 426 /* 427 * TODO: add XOP, EXEV vvvv reading. 428 * 429 * vex.vvvv field is in bits 6-3, bits are inverted. 430 * But in 32-bit mode, high-order bit may be ignored. 431 * Therefore, let's consider only 3 low-order bits. 432 */ 433 reg2 = ((reg2 >> 3) & 0x7) ^ 0x7; 434 /* 435 * Register numbering is ax,cx,dx,bx, sp,bp,si,di, r8..r15. 436 * 437 * Choose scratch reg. Order is important: must not select bx 438 * if we can use si (cmpxchg8b case!) 439 */ 440 if (reg != 6 && reg2 != 6) { 441 reg2 = 6; 442 auprobe->defparam.fixups |= UPROBE_FIX_RIP_SI; 443 } else if (reg != 7 && reg2 != 7) { 444 reg2 = 7; 445 auprobe->defparam.fixups |= UPROBE_FIX_RIP_DI; 446 /* TODO (paranoia): force maskmovq to not use di */ 447 } else { 448 reg2 = 3; 449 auprobe->defparam.fixups |= UPROBE_FIX_RIP_BX; 450 } 451 /* 452 * Point cursor at the modrm byte. The next 4 bytes are the 453 * displacement. Beyond the displacement, for some instructions, 454 * is the immediate operand. 455 */ 456 cursor = auprobe->insn + insn_offset_modrm(insn); 457 /* 458 * Change modrm from "00 reg 101" to "10 reg reg2". Example: 459 * 89 05 disp32 mov %eax,disp32(%rip) becomes 460 * 89 86 disp32 mov %eax,disp32(%rsi) 461 */ 462 *cursor = 0x80 | (reg << 3) | reg2; 463 } 464 465 static inline unsigned long * 466 scratch_reg(struct arch_uprobe *auprobe, struct pt_regs *regs) 467 { 468 if (auprobe->defparam.fixups & UPROBE_FIX_RIP_SI) 469 return ®s->si; 470 if (auprobe->defparam.fixups & UPROBE_FIX_RIP_DI) 471 return ®s->di; 472 return ®s->bx; 473 } 474 475 /* 476 * If we're emulating a rip-relative instruction, save the contents 477 * of the scratch register and store the target address in that register. 478 */ 479 static void riprel_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 480 { 481 if (auprobe->defparam.fixups & UPROBE_FIX_RIP_MASK) { 482 struct uprobe_task *utask = current->utask; 483 unsigned long *sr = scratch_reg(auprobe, regs); 484 485 utask->autask.saved_scratch_register = *sr; 486 *sr = utask->vaddr + auprobe->defparam.ilen; 487 } 488 } 489 490 static void riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 491 { 492 if (auprobe->defparam.fixups & UPROBE_FIX_RIP_MASK) { 493 struct uprobe_task *utask = current->utask; 494 unsigned long *sr = scratch_reg(auprobe, regs); 495 496 *sr = utask->autask.saved_scratch_register; 497 } 498 } 499 #else /* 32-bit: */ 500 static inline bool is_64bit_mm(struct mm_struct *mm) 501 { 502 return false; 503 } 504 /* 505 * No RIP-relative addressing on 32-bit 506 */ 507 static void riprel_analyze(struct arch_uprobe *auprobe, struct insn *insn) 508 { 509 } 510 static void riprel_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 511 { 512 } 513 static void riprel_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 514 { 515 } 516 #endif /* CONFIG_X86_64 */ 517 518 struct uprobe_xol_ops { 519 bool (*emulate)(struct arch_uprobe *, struct pt_regs *); 520 int (*pre_xol)(struct arch_uprobe *, struct pt_regs *); 521 int (*post_xol)(struct arch_uprobe *, struct pt_regs *); 522 void (*abort)(struct arch_uprobe *, struct pt_regs *); 523 }; 524 525 static inline int sizeof_long(void) 526 { 527 return is_ia32_task() ? 4 : 8; 528 } 529 530 static int default_pre_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs) 531 { 532 riprel_pre_xol(auprobe, regs); 533 return 0; 534 } 535 536 static int push_ret_address(struct pt_regs *regs, unsigned long ip) 537 { 538 unsigned long new_sp = regs->sp - sizeof_long(); 539 540 if (copy_to_user((void __user *)new_sp, &ip, sizeof_long())) 541 return -EFAULT; 542 543 regs->sp = new_sp; 544 return 0; 545 } 546 547 /* 548 * We have to fix things up as follows: 549 * 550 * Typically, the new ip is relative to the copied instruction. We need 551 * to make it relative to the original instruction (FIX_IP). Exceptions 552 * are return instructions and absolute or indirect jump or call instructions. 553 * 554 * If the single-stepped instruction was a call, the return address that 555 * is atop the stack is the address following the copied instruction. We 556 * need to make it the address following the original instruction (FIX_CALL). 557 * 558 * If the original instruction was a rip-relative instruction such as 559 * "movl %edx,0xnnnn(%rip)", we have instead executed an equivalent 560 * instruction using a scratch register -- e.g., "movl %edx,0xnnnn(%rsi)". 561 * We need to restore the contents of the scratch register 562 * (FIX_RIP_reg). 563 */ 564 static int default_post_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs) 565 { 566 struct uprobe_task *utask = current->utask; 567 568 riprel_post_xol(auprobe, regs); 569 if (auprobe->defparam.fixups & UPROBE_FIX_IP) { 570 long correction = utask->vaddr - utask->xol_vaddr; 571 regs->ip += correction; 572 } else if (auprobe->defparam.fixups & UPROBE_FIX_CALL) { 573 regs->sp += sizeof_long(); /* Pop incorrect return address */ 574 if (push_ret_address(regs, utask->vaddr + auprobe->defparam.ilen)) 575 return -ERESTART; 576 } 577 /* popf; tell the caller to not touch TF */ 578 if (auprobe->defparam.fixups & UPROBE_FIX_SETF) 579 utask->autask.saved_tf = true; 580 581 return 0; 582 } 583 584 static void default_abort_op(struct arch_uprobe *auprobe, struct pt_regs *regs) 585 { 586 riprel_post_xol(auprobe, regs); 587 } 588 589 static struct uprobe_xol_ops default_xol_ops = { 590 .pre_xol = default_pre_xol_op, 591 .post_xol = default_post_xol_op, 592 .abort = default_abort_op, 593 }; 594 595 static bool branch_is_call(struct arch_uprobe *auprobe) 596 { 597 return auprobe->branch.opc1 == 0xe8; 598 } 599 600 #define CASE_COND \ 601 COND(70, 71, XF(OF)) \ 602 COND(72, 73, XF(CF)) \ 603 COND(74, 75, XF(ZF)) \ 604 COND(78, 79, XF(SF)) \ 605 COND(7a, 7b, XF(PF)) \ 606 COND(76, 77, XF(CF) || XF(ZF)) \ 607 COND(7c, 7d, XF(SF) != XF(OF)) \ 608 COND(7e, 7f, XF(ZF) || XF(SF) != XF(OF)) 609 610 #define COND(op_y, op_n, expr) \ 611 case 0x ## op_y: DO((expr) != 0) \ 612 case 0x ## op_n: DO((expr) == 0) 613 614 #define XF(xf) (!!(flags & X86_EFLAGS_ ## xf)) 615 616 static bool is_cond_jmp_opcode(u8 opcode) 617 { 618 switch (opcode) { 619 #define DO(expr) \ 620 return true; 621 CASE_COND 622 #undef DO 623 624 default: 625 return false; 626 } 627 } 628 629 static bool check_jmp_cond(struct arch_uprobe *auprobe, struct pt_regs *regs) 630 { 631 unsigned long flags = regs->flags; 632 633 switch (auprobe->branch.opc1) { 634 #define DO(expr) \ 635 return expr; 636 CASE_COND 637 #undef DO 638 639 default: /* not a conditional jmp */ 640 return true; 641 } 642 } 643 644 #undef XF 645 #undef COND 646 #undef CASE_COND 647 648 static bool branch_emulate_op(struct arch_uprobe *auprobe, struct pt_regs *regs) 649 { 650 unsigned long new_ip = regs->ip += auprobe->branch.ilen; 651 unsigned long offs = (long)auprobe->branch.offs; 652 653 if (branch_is_call(auprobe)) { 654 /* 655 * If it fails we execute this (mangled, see the comment in 656 * branch_clear_offset) insn out-of-line. In the likely case 657 * this should trigger the trap, and the probed application 658 * should die or restart the same insn after it handles the 659 * signal, arch_uprobe_post_xol() won't be even called. 660 * 661 * But there is corner case, see the comment in ->post_xol(). 662 */ 663 if (push_ret_address(regs, new_ip)) 664 return false; 665 } else if (!check_jmp_cond(auprobe, regs)) { 666 offs = 0; 667 } 668 669 regs->ip = new_ip + offs; 670 return true; 671 } 672 673 static int branch_post_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs) 674 { 675 BUG_ON(!branch_is_call(auprobe)); 676 /* 677 * We can only get here if branch_emulate_op() failed to push the ret 678 * address _and_ another thread expanded our stack before the (mangled) 679 * "call" insn was executed out-of-line. Just restore ->sp and restart. 680 * We could also restore ->ip and try to call branch_emulate_op() again. 681 */ 682 regs->sp += sizeof_long(); 683 return -ERESTART; 684 } 685 686 static void branch_clear_offset(struct arch_uprobe *auprobe, struct insn *insn) 687 { 688 /* 689 * Turn this insn into "call 1f; 1:", this is what we will execute 690 * out-of-line if ->emulate() fails. We only need this to generate 691 * a trap, so that the probed task receives the correct signal with 692 * the properly filled siginfo. 693 * 694 * But see the comment in ->post_xol(), in the unlikely case it can 695 * succeed. So we need to ensure that the new ->ip can not fall into 696 * the non-canonical area and trigger #GP. 697 * 698 * We could turn it into (say) "pushf", but then we would need to 699 * divorce ->insn[] and ->ixol[]. We need to preserve the 1st byte 700 * of ->insn[] for set_orig_insn(). 701 */ 702 memset(auprobe->insn + insn_offset_immediate(insn), 703 0, insn->immediate.nbytes); 704 } 705 706 static struct uprobe_xol_ops branch_xol_ops = { 707 .emulate = branch_emulate_op, 708 .post_xol = branch_post_xol_op, 709 }; 710 711 /* Returns -ENOSYS if branch_xol_ops doesn't handle this insn */ 712 static int branch_setup_xol_ops(struct arch_uprobe *auprobe, struct insn *insn) 713 { 714 u8 opc1 = OPCODE1(insn); 715 int i; 716 717 switch (opc1) { 718 case 0xeb: /* jmp 8 */ 719 case 0xe9: /* jmp 32 */ 720 case 0x90: /* prefix* + nop; same as jmp with .offs = 0 */ 721 break; 722 723 case 0xe8: /* call relative */ 724 branch_clear_offset(auprobe, insn); 725 break; 726 727 case 0x0f: 728 if (insn->opcode.nbytes != 2) 729 return -ENOSYS; 730 /* 731 * If it is a "near" conditional jmp, OPCODE2() - 0x10 matches 732 * OPCODE1() of the "short" jmp which checks the same condition. 733 */ 734 opc1 = OPCODE2(insn) - 0x10; 735 default: 736 if (!is_cond_jmp_opcode(opc1)) 737 return -ENOSYS; 738 } 739 740 /* 741 * 16-bit overrides such as CALLW (66 e8 nn nn) are not supported. 742 * Intel and AMD behavior differ in 64-bit mode: Intel ignores 66 prefix. 743 * No one uses these insns, reject any branch insns with such prefix. 744 */ 745 for (i = 0; i < insn->prefixes.nbytes; i++) { 746 if (insn->prefixes.bytes[i] == 0x66) 747 return -ENOTSUPP; 748 } 749 750 auprobe->branch.opc1 = opc1; 751 auprobe->branch.ilen = insn->length; 752 auprobe->branch.offs = insn->immediate.value; 753 754 auprobe->ops = &branch_xol_ops; 755 return 0; 756 } 757 758 /** 759 * arch_uprobe_analyze_insn - instruction analysis including validity and fixups. 760 * @mm: the probed address space. 761 * @arch_uprobe: the probepoint information. 762 * @addr: virtual address at which to install the probepoint 763 * Return 0 on success or a -ve number on error. 764 */ 765 int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long addr) 766 { 767 struct insn insn; 768 u8 fix_ip_or_call = UPROBE_FIX_IP; 769 int ret; 770 771 ret = uprobe_init_insn(auprobe, &insn, is_64bit_mm(mm)); 772 if (ret) 773 return ret; 774 775 ret = branch_setup_xol_ops(auprobe, &insn); 776 if (ret != -ENOSYS) 777 return ret; 778 779 /* 780 * Figure out which fixups default_post_xol_op() will need to perform, 781 * and annotate defparam->fixups accordingly. 782 */ 783 switch (OPCODE1(&insn)) { 784 case 0x9d: /* popf */ 785 auprobe->defparam.fixups |= UPROBE_FIX_SETF; 786 break; 787 case 0xc3: /* ret or lret -- ip is correct */ 788 case 0xcb: 789 case 0xc2: 790 case 0xca: 791 case 0xea: /* jmp absolute -- ip is correct */ 792 fix_ip_or_call = 0; 793 break; 794 case 0x9a: /* call absolute - Fix return addr, not ip */ 795 fix_ip_or_call = UPROBE_FIX_CALL; 796 break; 797 case 0xff: 798 switch (MODRM_REG(&insn)) { 799 case 2: case 3: /* call or lcall, indirect */ 800 fix_ip_or_call = UPROBE_FIX_CALL; 801 break; 802 case 4: case 5: /* jmp or ljmp, indirect */ 803 fix_ip_or_call = 0; 804 break; 805 } 806 /* fall through */ 807 default: 808 riprel_analyze(auprobe, &insn); 809 } 810 811 auprobe->defparam.ilen = insn.length; 812 auprobe->defparam.fixups |= fix_ip_or_call; 813 814 auprobe->ops = &default_xol_ops; 815 return 0; 816 } 817 818 /* 819 * arch_uprobe_pre_xol - prepare to execute out of line. 820 * @auprobe: the probepoint information. 821 * @regs: reflects the saved user state of current task. 822 */ 823 int arch_uprobe_pre_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 824 { 825 struct uprobe_task *utask = current->utask; 826 827 if (auprobe->ops->pre_xol) { 828 int err = auprobe->ops->pre_xol(auprobe, regs); 829 if (err) 830 return err; 831 } 832 833 regs->ip = utask->xol_vaddr; 834 utask->autask.saved_trap_nr = current->thread.trap_nr; 835 current->thread.trap_nr = UPROBE_TRAP_NR; 836 837 utask->autask.saved_tf = !!(regs->flags & X86_EFLAGS_TF); 838 regs->flags |= X86_EFLAGS_TF; 839 if (test_tsk_thread_flag(current, TIF_BLOCKSTEP)) 840 set_task_blockstep(current, false); 841 842 return 0; 843 } 844 845 /* 846 * If xol insn itself traps and generates a signal(Say, 847 * SIGILL/SIGSEGV/etc), then detect the case where a singlestepped 848 * instruction jumps back to its own address. It is assumed that anything 849 * like do_page_fault/do_trap/etc sets thread.trap_nr != -1. 850 * 851 * arch_uprobe_pre_xol/arch_uprobe_post_xol save/restore thread.trap_nr, 852 * arch_uprobe_xol_was_trapped() simply checks that ->trap_nr is not equal to 853 * UPROBE_TRAP_NR == -1 set by arch_uprobe_pre_xol(). 854 */ 855 bool arch_uprobe_xol_was_trapped(struct task_struct *t) 856 { 857 if (t->thread.trap_nr != UPROBE_TRAP_NR) 858 return true; 859 860 return false; 861 } 862 863 /* 864 * Called after single-stepping. To avoid the SMP problems that can 865 * occur when we temporarily put back the original opcode to 866 * single-step, we single-stepped a copy of the instruction. 867 * 868 * This function prepares to resume execution after the single-step. 869 */ 870 int arch_uprobe_post_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 871 { 872 struct uprobe_task *utask = current->utask; 873 bool send_sigtrap = utask->autask.saved_tf; 874 int err = 0; 875 876 WARN_ON_ONCE(current->thread.trap_nr != UPROBE_TRAP_NR); 877 current->thread.trap_nr = utask->autask.saved_trap_nr; 878 879 if (auprobe->ops->post_xol) { 880 err = auprobe->ops->post_xol(auprobe, regs); 881 if (err) { 882 /* 883 * Restore ->ip for restart or post mortem analysis. 884 * ->post_xol() must not return -ERESTART unless this 885 * is really possible. 886 */ 887 regs->ip = utask->vaddr; 888 if (err == -ERESTART) 889 err = 0; 890 send_sigtrap = false; 891 } 892 } 893 /* 894 * arch_uprobe_pre_xol() doesn't save the state of TIF_BLOCKSTEP 895 * so we can get an extra SIGTRAP if we do not clear TF. We need 896 * to examine the opcode to make it right. 897 */ 898 if (send_sigtrap) 899 send_sig(SIGTRAP, current, 0); 900 901 if (!utask->autask.saved_tf) 902 regs->flags &= ~X86_EFLAGS_TF; 903 904 return err; 905 } 906 907 /* callback routine for handling exceptions. */ 908 int arch_uprobe_exception_notify(struct notifier_block *self, unsigned long val, void *data) 909 { 910 struct die_args *args = data; 911 struct pt_regs *regs = args->regs; 912 int ret = NOTIFY_DONE; 913 914 /* We are only interested in userspace traps */ 915 if (regs && !user_mode(regs)) 916 return NOTIFY_DONE; 917 918 switch (val) { 919 case DIE_INT3: 920 if (uprobe_pre_sstep_notifier(regs)) 921 ret = NOTIFY_STOP; 922 923 break; 924 925 case DIE_DEBUG: 926 if (uprobe_post_sstep_notifier(regs)) 927 ret = NOTIFY_STOP; 928 929 default: 930 break; 931 } 932 933 return ret; 934 } 935 936 /* 937 * This function gets called when XOL instruction either gets trapped or 938 * the thread has a fatal signal. Reset the instruction pointer to its 939 * probed address for the potential restart or for post mortem analysis. 940 */ 941 void arch_uprobe_abort_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) 942 { 943 struct uprobe_task *utask = current->utask; 944 945 if (auprobe->ops->abort) 946 auprobe->ops->abort(auprobe, regs); 947 948 current->thread.trap_nr = utask->autask.saved_trap_nr; 949 regs->ip = utask->vaddr; 950 /* clear TF if it was set by us in arch_uprobe_pre_xol() */ 951 if (!utask->autask.saved_tf) 952 regs->flags &= ~X86_EFLAGS_TF; 953 } 954 955 static bool __skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs) 956 { 957 if (auprobe->ops->emulate) 958 return auprobe->ops->emulate(auprobe, regs); 959 return false; 960 } 961 962 bool arch_uprobe_skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs) 963 { 964 bool ret = __skip_sstep(auprobe, regs); 965 if (ret && (regs->flags & X86_EFLAGS_TF)) 966 send_sig(SIGTRAP, current, 0); 967 return ret; 968 } 969 970 unsigned long 971 arch_uretprobe_hijack_return_addr(unsigned long trampoline_vaddr, struct pt_regs *regs) 972 { 973 int rasize = sizeof_long(), nleft; 974 unsigned long orig_ret_vaddr = 0; /* clear high bits for 32-bit apps */ 975 976 if (copy_from_user(&orig_ret_vaddr, (void __user *)regs->sp, rasize)) 977 return -1; 978 979 /* check whether address has been already hijacked */ 980 if (orig_ret_vaddr == trampoline_vaddr) 981 return orig_ret_vaddr; 982 983 nleft = copy_to_user((void __user *)regs->sp, &trampoline_vaddr, rasize); 984 if (likely(!nleft)) 985 return orig_ret_vaddr; 986 987 if (nleft != rasize) { 988 pr_err("uprobe: return address clobbered: pid=%d, %%sp=%#lx, " 989 "%%ip=%#lx\n", current->pid, regs->sp, regs->ip); 990 991 force_sig_info(SIGSEGV, SEND_SIG_FORCED, current); 992 } 993 994 return -1; 995 } 996