1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #include <errno.h> 27 #include <fcntl.h> 28 #include <stdio.h> 29 #include <stdlib.h> 30 #include <strings.h> 31 #include <unistd.h> 32 #include <thread.h> 33 #include <sys/auxv.h> 34 #include <sys/brand.h> 35 #include <sys/inttypes.h> 36 #include <sys/lwp.h> 37 #include <sys/syscall.h> 38 #include <sys/systm.h> 39 #include <sys/utsname.h> 40 #include <sys/systeminfo.h> 41 #include <sys/zone.h> 42 #include <sys/stat.h> 43 #include <sys/mntent.h> 44 #include <sys/ctfs.h> 45 #include <sys/priv.h> 46 #include <sys/acctctl.h> 47 #include <libgen.h> 48 #include <bsm/audit.h> 49 #include <sys/crypto/ioctl.h> 50 #include <sys/fs/zfs.h> 51 #include <sys/zfs_ioctl.h> 52 #include <sys/ucontext.h> 53 #include <sys/mntio.h> 54 #include <sys/mnttab.h> 55 #include <atomic.h> 56 57 #include <s10_brand.h> 58 #include <s10_misc.h> 59 60 /* 61 * Principles of emulation 101. 62 * 63 * 64 * *** Setting errno 65 * 66 * Just don't do it. This emulation library is loaded onto a 67 * seperate link map from the application who's address space we're 68 * running in. We have our own private copy of libc, so there for, 69 * the errno value accessible from here is is also private and changing 70 * it will not affect any errno value that the processes who's address 71 * space we are running in will see. To return an error condition we 72 * should return the negated errno value we'd like the system to return. 73 * For more information about this see the comment in s10_handler(). 74 * Basically, when we return to the caller that initiated the system 75 * call it's their responsibility to set errno. 76 * 77 * 78 * *** Recursion Considerations 79 * 80 * When emulating system calls we need to be very careful about what 81 * library calls we invoke. Library calls should be kept to a minimum. 82 * One issue is that library calls can invoke system calls, so if we're 83 * emulating a system call and we invoke a library call that depends on 84 * that system call we will probably enter a recursive loop, which would 85 * be bad. 86 * 87 * 88 * *** Return Values. 89 * 90 * When declaring new syscall emulation functions, it is very important 91 * to to set the proper RV_* flags in the s10_sysent_table. Upon failure, 92 * syscall emulation fuctions should return an errno value. Upon success 93 * syscall emulation functions should return 0 and set the sysret_t return 94 * value parameters accordingly. 95 * 96 * There are five possible syscall macro wrappers used in the kernel's system 97 * call sysent table. These turn into the following return values: 98 * SYSENT_CL -> SYSENT_C or SYSENT_CI 99 * SYSENT_C SE_64RVAL RV_DEFAULT 100 * SYSENT_CI SE_32RVAL1 RV_DEFAULT 101 * SYSENT_2CI SE_32RVAL1|SE_32RVAL2 RV_32RVAL2 102 * SYSENT_AP SE_64RVAL RV_64RVAL 103 * 104 * 105 * *** Agent lwp considerations 106 * 107 * It is currently impossible to do any emulation for these system call 108 * when they are being invoked on behalf of an agent lwp. To understand why 109 * it's impossible you have to understand how agent lwp syscalls work. 110 * 111 * The agent lwp syscall process works as follows: 112 * 1 The controlling process stops the target. 113 * 2 The controlling process injects an agent lwp which is also stopped. 114 * This agent lwp assumes the userland stack and register values 115 * of another stopped lwp in the current process. 116 * 3 The controlling process configures the agent lwp to start 117 * executing the requested system call. 118 * 4 The controlling process configure /proc to stop the agent lwp when 119 * it enters the requested system call. 120 * 5 The controlling processes allows the agent lwp to start executing. 121 * 6 The agent lwp traps into the kernel to perform the requested system 122 * call and immediately stop. 123 * 7 The controlling process copies all the arguments for the requested 124 * system call onto the agent lwp's stack. 125 * 8 The controlling process configures /proc to stop the agent lwp 126 * when it completes the requested system call. 127 * 9 The controlling processes allows the agent lwp to start executing. 128 * 10 The agent lwp executes the system call and then stop before returning 129 * to userland. 130 * 11 The controlling process copies the return value and return arguments 131 * back from the agent lwps stack. 132 * 12 The controlling process destroys the agent lwp and restarts 133 * the target process. 134 * 135 * The fundamental problem is that when the agent executes the request 136 * system call in step 5, if we're emulating that system call then the 137 * lwp is redirected back to our emulation layer without blocking 138 * in the kernel. But our emulation layer can't access the arguments 139 * for the system call because they haven't been copied to the stack 140 * yet and they still only exist in the controlling processes address 141 * space. This prevents us from being able to do any emulation of 142 * agent lwp system calls. Hence, currently our brand trap interposition 143 * callback (s10_brand_syscall_callback_common) will detect if a system 144 * call is being made by an agent lwp, and if this is the case it will 145 * never redirect the system call to this emulation library. 146 * 147 * In the future, if this proves to be a problem the the easiest solution 148 * would probably be to replace the branded versions of these application 149 * with their native counterparts. Ie, truss, plimit, and pfiles could be 150 * replace with wrapper scripts that execute the native versions of these 151 * applications. In the case of plimit and pfiles this should be pretty 152 * strait forward. Truss would probably be more tricky since it can 153 * execute applications which would be branded applications, so in that 154 * case it might be necessary to create a loadable library which could 155 * be LD_PRELOADed into truss and this library would interpose on the 156 * exec() system call to allow truss to correctly execute branded 157 * processes. It should be pointed out that this solution could work 158 * because "native agent lwps" (ie, agent lwps created by native 159 * processes) can be treated differently from "branded aged lwps" (ie, 160 * agent lwps created by branded processes), since native agent lwps 161 * would presumably be making native system calls and hence not need 162 * any interposition. 163 * 164 */ 165 166 static zoneid_t zoneid; 167 static boolean_t emul_global_zone = B_FALSE; 168 static s10_emul_bitmap_t emul_bitmap; 169 pid_t zone_init_pid; 170 171 /* 172 * S10_FEATURE_IS_PRESENT is a macro that helps facilitate conditional 173 * emulation. For each constant N defined in the s10_emulated_features 174 * enumeration in usr/src/uts/common/brand/solaris10/s10_brand.h, 175 * S10_FEATURE_IS_PRESENT(N) is true iff the feature/backport represented by N 176 * is present in the Solaris 10 image hosted within the zone. In other words, 177 * S10_FEATURE_IS_PRESENT(N) is true iff the file /usr/lib/brand/solaris10/M, 178 * where M is the enum value of N, was present in the zone when the zone booted. 179 * 180 * 181 * *** Sample Usage 182 * 183 * Suppose that you need to backport a fix to Solaris 10 and there is 184 * emulation in place for the fix. Suppose further that the emulation won't be 185 * needed if the fix is backported (i.e., if the fix is present in the hosted 186 * Solaris 10 environment, then the brand won't need the emulation). Then if 187 * you add a constant named "S10_FEATURE_X" to the end of the 188 * s10_emulated_features enumeration that represents the backported fix and 189 * S10_FEATURE_X evaluates to four, then you should create a file named 190 * /usr/lib/brand/solaris10/4 as part of your backport. Additionally, you 191 * should retain the aforementioned emulation but modify it so that it's 192 * performed only when S10_FEATURE_IS_PRESENT(S10_FEATURE_X) is false. Thus the 193 * emulation function should look something like the following: 194 * 195 * static int 196 * my_emul_function(sysret_t *rv, ...) 197 * { 198 * if (S10_FEATURE_IS_PRESENT(S10_FEATURE_X)) { 199 * // Don't emulate 200 * return (__systemcall(rv, ...)); 201 * } else { 202 * // Emulate whatever needs to be emulated when the 203 * // backport isn't present in the Solaris 10 image. 204 * } 205 * } 206 */ 207 #define S10_FEATURE_IS_PRESENT(s10_emulated_features_constant) \ 208 ((emul_bitmap[(s10_emulated_features_constant) >> 3] & \ 209 (1 << ((s10_emulated_features_constant) & 0x7))) != 0) 210 211 #define EMULATE(cb, args) { (sysent_cb_t)(cb), (args) } 212 #define NOSYS EMULATE(s10_unimpl, (0 | RV_DEFAULT)) 213 214 typedef long (*sysent_cb_t)(); 215 typedef struct s10_sysent_table { 216 sysent_cb_t st_callc; 217 uintptr_t st_args; 218 } s10_sysent_table_t; 219 s10_sysent_table_t s10_sysent_table[]; 220 221 #define S10_UTS_RELEASE "5.10" 222 #define S10_UTS_VERSION "Generic_Virtual" 223 224 /*LINTED: static unused*/ 225 static volatile int s10_abort_err; 226 /*LINTED: static unused*/ 227 static volatile const char *s10_abort_msg; 228 /*LINTED: static unused*/ 229 static volatile const char *s10_abort_file; 230 /*LINTED: static unused*/ 231 static volatile int s10_abort_line; 232 233 extern int errno; 234 235 /*ARGSUSED*/ 236 void 237 _s10_abort(int err, const char *msg, const char *file, int line) 238 { 239 sysret_t rval; 240 241 /* Save the error message into convenient globals */ 242 s10_abort_err = err; 243 s10_abort_msg = msg; 244 s10_abort_file = file; 245 s10_abort_line = line; 246 247 /* kill ourselves */ 248 abort(); 249 250 /* If abort() didn't work, try something stronger. */ 251 (void) __systemcall(&rval, SYS_lwp_kill + 1024, _lwp_self(), SIGKILL); 252 } 253 254 static int 255 s10_uucopy(const void *from, void *to, size_t size) 256 { 257 sysret_t rval; 258 259 if (__systemcall(&rval, SYS_uucopy + 1024, from, to, size) != 0) 260 return (EFAULT); 261 return (0); 262 } 263 264 /* 265 * ATTENTION: uucopystr() does NOT ensure that string are null terminated! 266 */ 267 static int 268 s10_uucopystr(const void *from, void *to, size_t size) 269 { 270 sysret_t rval; 271 272 if (__systemcall(&rval, SYS_uucopystr + 1024, from, to, size) != 0) 273 return (EFAULT); 274 return (0); 275 } 276 277 /* 278 * Figures out the PID of init for the zone. Also returns a boolean 279 * indicating whether this process currently has that pid: if so, 280 * then at this moment, we are init. 281 */ 282 static boolean_t 283 get_initpid_info(void) 284 { 285 pid_t pid; 286 sysret_t rval; 287 int err; 288 289 /* 290 * Determine the current process PID and the PID of the zone's init. 291 * We use care not to call getpid() here, because we're not supposed 292 * to call getpid() until after the program is fully linked-- the 293 * first call to getpid() is a signal from the linker to debuggers 294 * that linking has been completed. 295 */ 296 if ((err = __systemcall(&rval, SYS_brand, 297 B_S10_PIDINFO, &pid, &zone_init_pid)) != 0) { 298 s10_abort(err, "Failed to get init's pid"); 299 } 300 301 /* 302 * Note that we need to be cautious with the pid we get back-- 303 * it should not be stashed and used in place of getpid(), since 304 * we might fork(2). So we keep zone_init_pid and toss the pid 305 * we otherwise got. 306 */ 307 if (pid == zone_init_pid) 308 return (B_TRUE); 309 310 return (B_FALSE); 311 } 312 313 /* 314 * This function is defined to be NOSYS but it won't be called from the 315 * the kernel since the NOSYS system calls are not enabled in the kernel. 316 * Thus, the only time this function is called is directly from within the 317 * indirect system call path. 318 */ 319 /*ARGSUSED*/ 320 static long 321 s10_unimpl(sysret_t *rv, uintptr_t p1) 322 { 323 sysret_t rval; 324 325 /* 326 * We'd like to print out some kind of error message here like 327 * "unsupported syscall", but we can't because it's not safe to 328 * assume that stderr or STDERR_FILENO actually points to something 329 * that is a terminal, and if we wrote to those files we could 330 * inadvertantly write to some applications open files, which would 331 * be bad. 332 * 333 * Normally, if an application calls an invalid system call 334 * it get a SIGSYS sent to it. So we'll just go ahead and send 335 * ourselves a signal here. Note that this is far from ideal since 336 * if the application has registered a signal handler, that signal 337 * handler may recieve a ucontext_t as the third parameter to 338 * indicate the context of the process when the signal was 339 * generated, and in this case that context will not be what the 340 * application is expecting. Hence, we should probably create a 341 * brandsys() kernel function that can deliver the signal to us 342 * with the correct ucontext_t. 343 */ 344 (void) __systemcall(&rval, SYS_lwp_kill + 1024, _lwp_self(), SIGSYS); 345 return (ENOSYS); 346 } 347 348 #if defined(__sparc) && !defined(__sparcv9) 349 /* 350 * Yuck. For 32-bit sparc applications, handle indirect system calls. 351 * Note that we declare this interface to use the maximum number of 352 * system call arguments. If we recieve a system call that uses less 353 * arguments, then the additional arguments will be garbage, but they 354 * will also be ignored so that should be ok. 355 */ 356 static long 357 s10_indir(sysret_t *rv, int code, 358 uintptr_t a0, uintptr_t a1, uintptr_t a2, uintptr_t a3, uintptr_t a4, 359 uintptr_t a5, uintptr_t a6, uintptr_t a7) 360 { 361 s10_sysent_table_t *sst = &(s10_sysent_table[code]); 362 363 s10_assert(code < NSYSCALL); 364 switch (sst->st_args & NARGS_MASK) { 365 case 0: 366 return ((sst->st_callc)(rv)); 367 case 1: 368 return ((sst->st_callc)(rv, a0)); 369 case 2: 370 return ((sst->st_callc)(rv, a0, a1)); 371 case 3: 372 return ((sst->st_callc)(rv, a0, a1, a2)); 373 case 4: 374 return ((sst->st_callc)(rv, a0, a1, a2, a3)); 375 case 5: 376 return ((sst->st_callc)(rv, a0, a1, a2, a3, a4)); 377 case 6: 378 return ((sst->st_callc)(rv, rv, a0, a1, a2, a3, a4, a5)); 379 case 7: 380 return ((sst->st_callc)(rv, a0, a1, a2, a3, a4, a5, a6)); 381 case 8: 382 return ((sst->st_callc)(rv, a0, a1, a2, a3, a4, a5, a6, a7)); 383 } 384 s10_abort(0, "invalid entry in s10_sysent_table"); 385 return (EINVAL); 386 } 387 #endif /* __sparc && !__sparcv9 */ 388 389 /* Free the thread-local storage provided my mntfs_get_mntentbuf() */ 390 static void 391 mntfs_free_mntentbuf(void *arg) 392 { 393 struct mntentbuf *embufp = arg; 394 395 if (embufp == NULL) 396 return; 397 if (embufp->mbuf_emp) 398 free(embufp->mbuf_emp); 399 if (embufp->mbuf_buf) 400 free(embufp->mbuf_buf); 401 bzero(embufp, sizeof (struct mntentbuf)); 402 free(embufp); 403 } 404 405 /* Provide the thread-local storage required by mntfs_ioctl() */ 406 static struct mntentbuf * 407 mntfs_get_mntentbuf(size_t size) 408 { 409 static mutex_t keylock; 410 static thread_key_t key; 411 static int once_per_keyname = 0; 412 void *tsd = NULL; 413 struct mntentbuf *embufp; 414 415 /* Create the key. */ 416 if (!once_per_keyname) { 417 (void) mutex_lock(&keylock); 418 if (!once_per_keyname) { 419 if (thr_keycreate(&key, mntfs_free_mntentbuf)) { 420 (void) mutex_unlock(&keylock); 421 return (NULL); 422 } else { 423 once_per_keyname++; 424 } 425 } 426 (void) mutex_unlock(&keylock); 427 } 428 429 /* 430 * The thread-specific datum for this key is the address of a struct 431 * mntentbuf. If this is the first time here then we allocate the struct 432 * and its contents, and associate its address with the thread; if there 433 * are any problems then we abort. 434 */ 435 if (thr_getspecific(key, &tsd)) 436 return (NULL); 437 if (tsd == NULL) { 438 if (!(embufp = calloc(1, sizeof (struct mntentbuf))) || 439 !(embufp->mbuf_emp = malloc(sizeof (struct extmnttab))) || 440 thr_setspecific(key, embufp)) { 441 mntfs_free_mntentbuf(embufp); 442 return (NULL); 443 } 444 } else { 445 embufp = tsd; 446 } 447 448 /* Return the buffer, resizing it if necessary. */ 449 if (size > embufp->mbuf_bufsize) { 450 if (embufp->mbuf_buf) 451 free(embufp->mbuf_buf); 452 if ((embufp->mbuf_buf = malloc(size)) == NULL) { 453 embufp->mbuf_bufsize = 0; 454 return (NULL); 455 } else { 456 embufp->mbuf_bufsize = size; 457 } 458 } 459 return (embufp); 460 } 461 462 /* 463 * The MNTIOC_GETMNTENT command in this release differs from that in Solaris 10. 464 * Previously, the command would copy a pointer to a struct extmnttab to an 465 * address provided as an argument. The pointer would be somewhere within a 466 * mapping already present within the user's address space. In addition, the 467 * text to which the struct's members pointed would also be within a 468 * pre-existing mapping. Now, the user is required to allocate memory for both 469 * the struct and the text buffer, and to pass the address of each within a 470 * struct mntentbuf. In order to conceal these details from a Solaris 10 client 471 * we allocate some thread-local storage in which to create the necessary data 472 * structures; this is static, thread-safe memory that will be cleaned up 473 * without the caller's intervention. 474 * 475 * MNTIOC_GETEXTMNTENT and MNTIOC_GETMNTANY are new in this release; they should 476 * not work for older clients. 477 */ 478 int 479 mntfs_ioctl(sysret_t *rval, int fdes, int cmd, intptr_t arg) 480 { 481 int err; 482 struct stat statbuf; 483 struct mntentbuf *embufp; 484 static size_t bufsize = MNT_LINE_MAX; 485 486 if ((err = __systemcall(rval, SYS_fstat + 1024, fdes, &statbuf)) != 0) 487 return (err); 488 if (strcmp(statbuf.st_fstype, MNTTYPE_MNTFS) != 0) 489 return (__systemcall(rval, SYS_ioctl + 1024, fdes, cmd, arg)); 490 491 if (cmd == MNTIOC_GETEXTMNTENT || cmd == MNTIOC_GETMNTANY) 492 return (EINVAL); 493 494 if ((embufp = mntfs_get_mntentbuf(bufsize)) == NULL) 495 return (ENOMEM); 496 497 /* 498 * MNTIOC_GETEXTMNTENT advances the file pointer once it has 499 * successfully copied out the result to the address provided. We 500 * therefore need to check the user-supplied address now since the 501 * one we'll be providing is guaranteed to work. 502 */ 503 if (s10_uucopy(&embufp->mbuf_emp, (void *)arg, sizeof (void *)) != 0) 504 return (EFAULT); 505 506 /* 507 * Keep retrying for as long as we fail for want of a large enough 508 * buffer. 509 */ 510 for (;;) { 511 if ((err = __systemcall(rval, SYS_ioctl + 1024, fdes, 512 MNTIOC_GETEXTMNTENT, embufp)) != 0) 513 return (err); 514 515 if (rval->sys_rval1 == MNTFS_TOOLONG) { 516 /* The buffer wasn't large enough. */ 517 (void) atomic_swap_ulong((unsigned long *)&bufsize, 518 2 * embufp->mbuf_bufsize); 519 if ((embufp = mntfs_get_mntentbuf(bufsize)) == NULL) 520 return (ENOMEM); 521 } else { 522 break; 523 } 524 } 525 526 if (s10_uucopy(&embufp->mbuf_emp, (void *)arg, sizeof (void *)) != 0) 527 return (EFAULT); 528 529 return (0); 530 } 531 532 /* 533 * Assign the structure member value from the s (source) structure to the 534 * d (dest) structure. 535 */ 536 #define struct_assign(d, s, val) (((d).val) = ((s).val)) 537 538 /* 539 * The CRYPTO_GET_FUNCTION_LIST parameter structure crypto_function_list_t 540 * changed between S10 and Nevada, so we have to emulate the old S10 541 * crypto_function_list_t structure when interposing on the ioctl syscall. 542 */ 543 typedef struct s10_crypto_function_list { 544 boolean_t fl_digest_init; 545 boolean_t fl_digest; 546 boolean_t fl_digest_update; 547 boolean_t fl_digest_key; 548 boolean_t fl_digest_final; 549 550 boolean_t fl_encrypt_init; 551 boolean_t fl_encrypt; 552 boolean_t fl_encrypt_update; 553 boolean_t fl_encrypt_final; 554 555 boolean_t fl_decrypt_init; 556 boolean_t fl_decrypt; 557 boolean_t fl_decrypt_update; 558 boolean_t fl_decrypt_final; 559 560 boolean_t fl_mac_init; 561 boolean_t fl_mac; 562 boolean_t fl_mac_update; 563 boolean_t fl_mac_final; 564 565 boolean_t fl_sign_init; 566 boolean_t fl_sign; 567 boolean_t fl_sign_update; 568 boolean_t fl_sign_final; 569 boolean_t fl_sign_recover_init; 570 boolean_t fl_sign_recover; 571 572 boolean_t fl_verify_init; 573 boolean_t fl_verify; 574 boolean_t fl_verify_update; 575 boolean_t fl_verify_final; 576 boolean_t fl_verify_recover_init; 577 boolean_t fl_verify_recover; 578 579 boolean_t fl_digest_encrypt_update; 580 boolean_t fl_decrypt_digest_update; 581 boolean_t fl_sign_encrypt_update; 582 boolean_t fl_decrypt_verify_update; 583 584 boolean_t fl_seed_random; 585 boolean_t fl_generate_random; 586 587 boolean_t fl_session_open; 588 boolean_t fl_session_close; 589 boolean_t fl_session_login; 590 boolean_t fl_session_logout; 591 592 boolean_t fl_object_create; 593 boolean_t fl_object_copy; 594 boolean_t fl_object_destroy; 595 boolean_t fl_object_get_size; 596 boolean_t fl_object_get_attribute_value; 597 boolean_t fl_object_set_attribute_value; 598 boolean_t fl_object_find_init; 599 boolean_t fl_object_find; 600 boolean_t fl_object_find_final; 601 602 boolean_t fl_key_generate; 603 boolean_t fl_key_generate_pair; 604 boolean_t fl_key_wrap; 605 boolean_t fl_key_unwrap; 606 boolean_t fl_key_derive; 607 608 boolean_t fl_init_token; 609 boolean_t fl_init_pin; 610 boolean_t fl_set_pin; 611 612 boolean_t prov_is_limited; 613 uint32_t prov_hash_threshold; 614 uint32_t prov_hash_limit; 615 } s10_crypto_function_list_t; 616 617 typedef struct s10_crypto_get_function_list { 618 uint_t fl_return_value; 619 crypto_provider_id_t fl_provider_id; 620 s10_crypto_function_list_t fl_list; 621 } s10_crypto_get_function_list_t; 622 623 /* 624 * The structure returned by the CRYPTO_GET_FUNCTION_LIST ioctl on /dev/crypto 625 * increased in size due to: 626 * 6482533 Threshold for HW offload via PKCS11 interface 627 * between S10 and Nevada. This is a relatively simple process of filling 628 * in the S10 structure fields with the Nevada data. 629 * 630 * We stat the device to make sure that the ioctl is meant for /dev/crypto. 631 * 632 */ 633 static int 634 crypto_ioctl(sysret_t *rval, int fdes, int cmd, intptr_t arg) 635 { 636 int err; 637 s10_crypto_get_function_list_t s10_param; 638 crypto_get_function_list_t native_param; 639 static dev_t crypto_dev = (dev_t)-1; 640 struct stat sbuf; 641 642 if (crypto_dev == (dev_t)-1) { 643 if ((err = __systemcall(rval, SYS_stat + 1024, "/dev/crypto", 644 &sbuf)) != 0) 645 goto nonemuioctl; 646 crypto_dev = major(sbuf.st_rdev); 647 } 648 if ((err = __systemcall(rval, SYS_fstat + 1024, fdes, &sbuf)) != 0) 649 return (err); 650 /* Each open fd of /dev/crypto gets a new minor device. */ 651 if (major(sbuf.st_rdev) != crypto_dev) 652 goto nonemuioctl; 653 654 if (s10_uucopy((const void *)arg, &s10_param, sizeof (s10_param)) != 0) 655 return (EFAULT); 656 struct_assign(native_param, s10_param, fl_provider_id); 657 if ((err = __systemcall(rval, SYS_ioctl + 1024, fdes, cmd, 658 &native_param)) != 0) 659 return (err); 660 661 struct_assign(s10_param, native_param, fl_return_value); 662 struct_assign(s10_param, native_param, fl_provider_id); 663 664 struct_assign(s10_param, native_param, fl_list.fl_digest_init); 665 struct_assign(s10_param, native_param, fl_list.fl_digest); 666 struct_assign(s10_param, native_param, fl_list.fl_digest_update); 667 struct_assign(s10_param, native_param, fl_list.fl_digest_key); 668 struct_assign(s10_param, native_param, fl_list.fl_digest_final); 669 670 struct_assign(s10_param, native_param, fl_list.fl_encrypt_init); 671 struct_assign(s10_param, native_param, fl_list.fl_encrypt); 672 struct_assign(s10_param, native_param, fl_list.fl_encrypt_update); 673 struct_assign(s10_param, native_param, fl_list.fl_encrypt_final); 674 675 struct_assign(s10_param, native_param, fl_list.fl_decrypt_init); 676 struct_assign(s10_param, native_param, fl_list.fl_decrypt); 677 struct_assign(s10_param, native_param, fl_list.fl_decrypt_update); 678 struct_assign(s10_param, native_param, fl_list.fl_decrypt_final); 679 680 struct_assign(s10_param, native_param, fl_list.fl_mac_init); 681 struct_assign(s10_param, native_param, fl_list.fl_mac); 682 struct_assign(s10_param, native_param, fl_list.fl_mac_update); 683 struct_assign(s10_param, native_param, fl_list.fl_mac_final); 684 685 struct_assign(s10_param, native_param, fl_list.fl_sign_init); 686 struct_assign(s10_param, native_param, fl_list.fl_sign); 687 struct_assign(s10_param, native_param, fl_list.fl_sign_update); 688 struct_assign(s10_param, native_param, fl_list.fl_sign_final); 689 struct_assign(s10_param, native_param, fl_list.fl_sign_recover_init); 690 struct_assign(s10_param, native_param, fl_list.fl_sign_recover); 691 692 struct_assign(s10_param, native_param, fl_list.fl_verify_init); 693 struct_assign(s10_param, native_param, fl_list.fl_verify); 694 struct_assign(s10_param, native_param, fl_list.fl_verify_update); 695 struct_assign(s10_param, native_param, fl_list.fl_verify_final); 696 struct_assign(s10_param, native_param, fl_list.fl_verify_recover_init); 697 struct_assign(s10_param, native_param, fl_list.fl_verify_recover); 698 699 struct_assign(s10_param, native_param, 700 fl_list.fl_digest_encrypt_update); 701 struct_assign(s10_param, native_param, 702 fl_list.fl_decrypt_digest_update); 703 struct_assign(s10_param, native_param, fl_list.fl_sign_encrypt_update); 704 struct_assign(s10_param, native_param, 705 fl_list.fl_decrypt_verify_update); 706 707 struct_assign(s10_param, native_param, fl_list.fl_seed_random); 708 struct_assign(s10_param, native_param, fl_list.fl_generate_random); 709 710 struct_assign(s10_param, native_param, fl_list.fl_session_open); 711 struct_assign(s10_param, native_param, fl_list.fl_session_close); 712 struct_assign(s10_param, native_param, fl_list.fl_session_login); 713 struct_assign(s10_param, native_param, fl_list.fl_session_logout); 714 715 struct_assign(s10_param, native_param, fl_list.fl_object_create); 716 struct_assign(s10_param, native_param, fl_list.fl_object_copy); 717 struct_assign(s10_param, native_param, fl_list.fl_object_destroy); 718 struct_assign(s10_param, native_param, fl_list.fl_object_get_size); 719 struct_assign(s10_param, native_param, 720 fl_list.fl_object_get_attribute_value); 721 struct_assign(s10_param, native_param, 722 fl_list.fl_object_set_attribute_value); 723 struct_assign(s10_param, native_param, fl_list.fl_object_find_init); 724 struct_assign(s10_param, native_param, fl_list.fl_object_find); 725 struct_assign(s10_param, native_param, fl_list.fl_object_find_final); 726 727 struct_assign(s10_param, native_param, fl_list.fl_key_generate); 728 struct_assign(s10_param, native_param, fl_list.fl_key_generate_pair); 729 struct_assign(s10_param, native_param, fl_list.fl_key_wrap); 730 struct_assign(s10_param, native_param, fl_list.fl_key_unwrap); 731 struct_assign(s10_param, native_param, fl_list.fl_key_derive); 732 733 struct_assign(s10_param, native_param, fl_list.fl_init_token); 734 struct_assign(s10_param, native_param, fl_list.fl_init_pin); 735 struct_assign(s10_param, native_param, fl_list.fl_set_pin); 736 737 struct_assign(s10_param, native_param, fl_list.prov_is_limited); 738 struct_assign(s10_param, native_param, fl_list.prov_hash_threshold); 739 struct_assign(s10_param, native_param, fl_list.prov_hash_limit); 740 741 return (s10_uucopy(&s10_param, (void *)arg, sizeof (s10_param))); 742 743 nonemuioctl: 744 return (__systemcall(rval, SYS_ioctl + 1024, fdes, cmd, arg)); 745 } 746 747 /* 748 * The process contract CT_TGET and CT_TSET parameter structure ct_param_t 749 * changed between S10 and Nevada, so we have to emulate the old S10 750 * ct_param_t structure when interposing on the ioctl syscall. 751 */ 752 typedef struct s10_ct_param { 753 uint32_t ctpm_id; 754 uint32_t ctpm_pad; 755 uint64_t ctpm_value; 756 } s10_ct_param_t; 757 758 /* 759 * We have to emulate process contract ioctls for init(1M) because the 760 * ioctl parameter structure changed between S10 and Nevada. This is 761 * a relatively simple process of filling Nevada structure fields, 762 * shuffling values, and initiating a native system call. 763 * 764 * For now, we'll assume that all consumers of CT_TGET and CT_TSET will 765 * need emulation. We'll issue a stat to make sure that the ioctl 766 * is meant for the contract file system. 767 * 768 */ 769 static int 770 ctfs_ioctl(sysret_t *rval, int fdes, int cmd, intptr_t arg) 771 { 772 int err; 773 s10_ct_param_t s10param; 774 ct_param_t param; 775 struct stat statbuf; 776 777 if ((err = __systemcall(rval, SYS_fstat + 1024, fdes, &statbuf)) != 0) 778 return (err); 779 if (strcmp(statbuf.st_fstype, MNTTYPE_CTFS) != 0) 780 return (__systemcall(rval, SYS_ioctl + 1024, fdes, cmd, arg)); 781 782 if (s10_uucopy((const void *)arg, &s10param, sizeof (s10param)) != 0) 783 return (EFAULT); 784 param.ctpm_id = s10param.ctpm_id; 785 param.ctpm_size = sizeof (uint64_t); 786 param.ctpm_value = &s10param.ctpm_value; 787 if ((err = __systemcall(rval, SYS_ioctl + 1024, fdes, cmd, ¶m)) 788 != 0) 789 return (err); 790 791 if (cmd == CT_TGET) 792 return (s10_uucopy(&s10param, (void *)arg, sizeof (s10param))); 793 794 return (0); 795 } 796 797 typedef struct s10_zfs_cmd { 798 char zc_name[MAXPATHLEN]; 799 char zc_value[MAXPATHLEN * 2]; 800 char zc_string[MAXNAMELEN]; 801 uint64_t zc_guid; 802 uint64_t zc_nvlist_conf; /* really (char *) */ 803 uint64_t zc_nvlist_conf_size; 804 uint64_t zc_nvlist_src; /* really (char *) */ 805 uint64_t zc_nvlist_src_size; 806 uint64_t zc_nvlist_dst; /* really (char *) */ 807 uint64_t zc_nvlist_dst_size; 808 uint64_t zc_cookie; 809 uint64_t zc_objset_type; 810 uint64_t zc_perm_action; 811 uint64_t zc_history; /* really (char *) */ 812 uint64_t zc_history_len; 813 uint64_t zc_history_offset; 814 uint64_t zc_obj; 815 /* Solaris Next added zc_iflags member here */ 816 zfs_share_t zc_share; 817 dmu_objset_stats_t zc_objset_stats; 818 struct drr_begin zc_begin_record; 819 zinject_record_t zc_inject_record; 820 } s10_zfs_cmd_t; 821 822 /* 823 * There is a difference in the zfs_cmd_t ioctl parameter between S10 and 824 * Solaris Next so we need to translate between the two structures when 825 * making ZFS ioctls. 826 */ 827 static int 828 zfs_ioctl(sysret_t *rval, int fdes, int cmd, intptr_t arg) 829 { 830 int err; 831 s10_zfs_cmd_t s10_param; 832 zfs_cmd_t native_param; 833 static dev_t zfs_dev = (dev_t)-1; 834 struct stat sbuf; 835 836 if (zfs_dev == (dev_t)-1) { 837 if ((err = __systemcall(rval, SYS_stat + 1024, "/dev/zfs", 838 &sbuf)) != 0) 839 goto nonemuioctl; 840 zfs_dev = major(sbuf.st_rdev); 841 } 842 if ((err = __systemcall(rval, SYS_fstat + 1024, fdes, &sbuf)) != 0) 843 return (err); 844 if (major(sbuf.st_rdev) != zfs_dev) 845 goto nonemuioctl; 846 847 if (s10_uucopy((const void *)arg, &s10_param, sizeof (s10_param)) != 0) 848 return (EFAULT); 849 850 bcopy((const void *)s10_param.zc_name, (void *)native_param.zc_name, 851 sizeof (s10_param.zc_name)); 852 bcopy((const void *)s10_param.zc_value, (void *)native_param.zc_value, 853 sizeof (s10_param.zc_value)); 854 bcopy((const void *)s10_param.zc_string, (void *)native_param.zc_string, 855 sizeof (s10_param.zc_string)); 856 struct_assign(native_param, s10_param, zc_guid); 857 struct_assign(native_param, s10_param, zc_nvlist_conf); 858 struct_assign(native_param, s10_param, zc_nvlist_conf_size); 859 struct_assign(native_param, s10_param, zc_nvlist_src); 860 struct_assign(native_param, s10_param, zc_nvlist_src_size); 861 struct_assign(native_param, s10_param, zc_nvlist_dst); 862 struct_assign(native_param, s10_param, zc_nvlist_dst_size); 863 struct_assign(native_param, s10_param, zc_cookie); 864 struct_assign(native_param, s10_param, zc_objset_type); 865 struct_assign(native_param, s10_param, zc_perm_action); 866 struct_assign(native_param, s10_param, zc_history); 867 struct_assign(native_param, s10_param, zc_history_len); 868 struct_assign(native_param, s10_param, zc_history_offset); 869 struct_assign(native_param, s10_param, zc_obj); 870 native_param.zc_iflags = 0; 871 struct_assign(native_param, s10_param, zc_share); 872 struct_assign(native_param, s10_param, zc_objset_stats); 873 struct_assign(native_param, s10_param, zc_begin_record); 874 struct_assign(native_param, s10_param, zc_inject_record); 875 876 err = __systemcall(rval, SYS_ioctl + 1024, fdes, cmd, &native_param); 877 878 bcopy((const void *)native_param.zc_name, (void *)s10_param.zc_name, 879 sizeof (s10_param.zc_name)); 880 bcopy((const void *)native_param.zc_value, (void *)s10_param.zc_value, 881 sizeof (s10_param.zc_value)); 882 bcopy((const void *)native_param.zc_string, (void *)s10_param.zc_string, 883 sizeof (s10_param.zc_string)); 884 struct_assign(s10_param, native_param, zc_guid); 885 struct_assign(s10_param, native_param, zc_nvlist_conf); 886 struct_assign(s10_param, native_param, zc_nvlist_conf_size); 887 struct_assign(s10_param, native_param, zc_nvlist_src); 888 struct_assign(s10_param, native_param, zc_nvlist_src_size); 889 struct_assign(s10_param, native_param, zc_nvlist_dst); 890 struct_assign(s10_param, native_param, zc_nvlist_dst_size); 891 struct_assign(s10_param, native_param, zc_cookie); 892 struct_assign(s10_param, native_param, zc_objset_type); 893 struct_assign(s10_param, native_param, zc_perm_action); 894 struct_assign(s10_param, native_param, zc_history); 895 struct_assign(s10_param, native_param, zc_history_len); 896 struct_assign(s10_param, native_param, zc_history_offset); 897 struct_assign(s10_param, native_param, zc_obj); 898 struct_assign(s10_param, native_param, zc_share); 899 struct_assign(s10_param, native_param, zc_objset_stats); 900 struct_assign(s10_param, native_param, zc_begin_record); 901 struct_assign(s10_param, native_param, zc_inject_record); 902 903 (void) s10_uucopy(&s10_param, (void *)arg, sizeof (s10_param)); 904 return (err); 905 906 nonemuioctl: 907 return (__systemcall(rval, SYS_ioctl + 1024, fdes, cmd, arg)); 908 } 909 910 int 911 s10_ioctl(sysret_t *rval, int fdes, int cmd, intptr_t arg) 912 { 913 switch (cmd) { 914 case CRYPTO_GET_FUNCTION_LIST: 915 return (crypto_ioctl(rval, fdes, cmd, arg)); 916 case CT_TGET: 917 /*FALLTHRU*/ 918 case CT_TSET: 919 return (ctfs_ioctl(rval, fdes, cmd, arg)); 920 case MNTIOC_GETMNTENT: 921 /*FALLTHRU*/ 922 case MNTIOC_GETEXTMNTENT: 923 /*FALLTHRU*/ 924 case MNTIOC_GETMNTANY: 925 return (mntfs_ioctl(rval, fdes, cmd, arg)); 926 } 927 928 if ((cmd & 0xff00) == ZFS_IOC) 929 return (zfs_ioctl(rval, fdes, cmd, arg)); 930 931 return (__systemcall(rval, SYS_ioctl + 1024, fdes, cmd, arg)); 932 } 933 934 /* 935 * Unfortunately, pwrite()'s behavior differs between S10 and Nevada when 936 * applied to files opened with O_APPEND. The offset argument is ignored and 937 * the buffer is appended to the target file in S10, whereas the current file 938 * position is ignored in Nevada (i.e., pwrite() acts as though the target file 939 * wasn't opened with O_APPEND). This is a result of the fix for CR 6655660 940 * (pwrite() must ignore the O_APPEND/FAPPEND flag). 941 * 942 * We emulate the old S10 pwrite() behavior by checking whether the target file 943 * was opened with O_APPEND. If it was, then invoke the write() system call 944 * instead of pwrite(); otherwise, invoke the pwrite() system call as usual. 945 */ 946 static int 947 s10_pwrite(sysret_t *rval, int fd, const void *bufferp, size_t num_bytes, 948 off_t offset) 949 { 950 int err; 951 952 if ((err = __systemcall(rval, SYS_fcntl + 1024, fd, F_GETFL)) != 0) 953 return (err); 954 if (rval->sys_rval1 & O_APPEND) 955 return (__systemcall(rval, SYS_write + 1024, fd, bufferp, 956 num_bytes)); 957 return (__systemcall(rval, SYS_pwrite + 1024, fd, bufferp, num_bytes, 958 offset)); 959 } 960 961 #ifndef _LP64 962 /* 963 * This is the large file version of the pwrite() system call for 32-bit 964 * processes. This exists for the same reason that s10_pwrite() exists; see 965 * the comment above s10_pwrite(). 966 */ 967 static int 968 s10_pwrite64(sysret_t *rval, int fd, const void *bufferp, size32_t num_bytes, 969 uint32_t offset_1, uint32_t offset_2) 970 { 971 int err; 972 973 if ((err = __systemcall(rval, SYS_fcntl + 1024, fd, F_GETFL)) != 0) 974 return (err); 975 if (rval->sys_rval1 & O_APPEND) 976 return (__systemcall(rval, SYS_write + 1024, fd, bufferp, 977 num_bytes)); 978 return (__systemcall(rval, SYS_pwrite64 + 1024, fd, bufferp, 979 num_bytes, offset_1, offset_2)); 980 } 981 #endif /* !_LP64 */ 982 983 #define S10_AC_PROC (0x1 << 28) 984 #define S10_AC_TASK (0x2 << 28) 985 #define S10_AC_FLOW (0x4 << 28) 986 #define S10_AC_MODE(x) ((x) & 0xf0000000) 987 #define S10_AC_OPTION(x) ((x) & 0x0fffffff) 988 989 /* 990 * The mode shift, mode mask and option mask for acctctl have changed. The 991 * mode is currently the top full byte and the option is the lower 3 full bytes. 992 */ 993 int 994 s10_acctctl(sysret_t *rval, int cmd, void *buf, size_t bufsz) 995 { 996 int mode = S10_AC_MODE(cmd); 997 int option = S10_AC_OPTION(cmd); 998 999 switch (mode) { 1000 case S10_AC_PROC: 1001 mode = AC_PROC; 1002 break; 1003 case S10_AC_TASK: 1004 mode = AC_TASK; 1005 break; 1006 case S10_AC_FLOW: 1007 mode = AC_FLOW; 1008 break; 1009 default: 1010 return (S10_TRUSS_POINT_3(rval, SYS_acctctl, EINVAL, cmd, buf, 1011 bufsz)); 1012 } 1013 1014 return (__systemcall(rval, SYS_acctctl + 1024, mode | option, buf, 1015 bufsz)); 1016 } 1017 1018 /* 1019 * The Audit Policy parameters have changed due to: 1020 * 6466722 audituser and AUDIT_USER are defined, unused, undocumented and 1021 * should be removed. 1022 * 1023 * In S10 we had the following flag: 1024 * #define AUDIT_USER 0x0040 1025 * which doesn't exist in Solaris Next where the subsequent flags are shifted 1026 * down. For example, in S10 we had: 1027 * #define AUDIT_GROUP 0x0080 1028 * but on Solaris Next we have: 1029 * #define AUDIT_GROUP 0x0040 1030 * AUDIT_GROUP has the value AUDIT_USER had in S10 and all of the subsequent 1031 * bits are also shifted one place. 1032 * 1033 * When we're getting or setting the Audit Policy parameters we need to 1034 * shift the outgoing or incoming bits into their proper positions. Since 1035 * S10_AUDIT_USER was always unused, we always clear that bit on A_GETPOLICY. 1036 * 1037 * The command we care about, BSM_AUDITCTL, passes the most parameters (3), 1038 * so declare this function to take up to 4 args and just pass them on. 1039 * The number of parameters for s10_auditsys needs to be equal to the BSM_* 1040 * subcommand that has the most parameters, since we want to pass all 1041 * parameters through, regardless of which subcommands we interpose on. 1042 * 1043 * Note that the auditsys system call uses the SYSENT_AP macro wrapper instead 1044 * of the more common SYSENT_CI macro. This means the return value is a 1045 * SE_64RVAL so the syscall table uses RV_64RVAL. 1046 */ 1047 1048 #define S10_AUDIT_HMASK 0xffffffc0 1049 #define S10_AUDIT_LMASK 0x3f 1050 1051 int 1052 s10_auditsys(sysret_t *rval, int bsmcmd, intptr_t a0, intptr_t a1, intptr_t a2) 1053 { 1054 int err; 1055 uint_t m; 1056 1057 if (bsmcmd != BSM_AUDITCTL) 1058 return (__systemcall(rval, SYS_auditsys + 1024, bsmcmd, a0, a1, 1059 a2)); 1060 1061 if ((int)a0 == A_GETPOLICY) { 1062 if ((err = __systemcall(rval, SYS_auditsys + 1024, bsmcmd, a0, 1063 &m, a2)) != 0) 1064 return (err); 1065 m = ((m & S10_AUDIT_HMASK) << 1) | (m & S10_AUDIT_LMASK); 1066 if (s10_uucopy(&m, (void *)a1, sizeof (m)) != 0) 1067 return (EFAULT); 1068 return (0); 1069 1070 } else if ((int)a0 == A_SETPOLICY) { 1071 if (s10_uucopy((const void *)a1, &m, sizeof (m)) != 0) 1072 return (EFAULT); 1073 m = ((m >> 1) & S10_AUDIT_HMASK) | (m & S10_AUDIT_LMASK); 1074 return (__systemcall(rval, SYS_auditsys + 1024, bsmcmd, a0, &m, 1075 a2)); 1076 } 1077 1078 return (__systemcall(rval, SYS_auditsys + 1024, bsmcmd, a0, a1, a2)); 1079 } 1080 1081 /* 1082 * Determine whether the executable passed to SYS_exec or SYS_execve is a 1083 * native executable. The s10_npreload.so invokes the B_S10_NATIVE brand 1084 * operation which patches up the processes exec info to eliminate any trace 1085 * of the wrapper. That will make pgrep and other commands that examine 1086 * process' executable names and command-line parameters work properly. 1087 */ 1088 static int 1089 s10_exec_native(sysret_t *rval, const char *fname, const char **argp, 1090 const char **envp) 1091 { 1092 const char *filename = fname; 1093 char path[64]; 1094 int err; 1095 1096 /* Get a copy of the executable we're trying to run */ 1097 path[0] = '\0'; 1098 (void) s10_uucopystr(filename, path, sizeof (path)); 1099 1100 /* Check if we're trying to run a native binary */ 1101 if (strncmp(path, "/.SUNWnative/usr/lib/brand/solaris10/s10_native", 1102 sizeof (path)) != 0) 1103 return (0); 1104 1105 /* Skip the first element in the argv array */ 1106 argp++; 1107 1108 /* 1109 * The the path of the dynamic linker is the second parameter 1110 * of s10_native_exec(). 1111 */ 1112 if (s10_uucopy(argp, &filename, sizeof (char *)) != 0) 1113 return (EFAULT); 1114 1115 /* If an exec call succeeds, it never returns */ 1116 err = __systemcall(rval, SYS_brand + 1024, B_EXEC_NATIVE, filename, 1117 argp, envp, NULL, NULL, NULL); 1118 s10_assert(err != 0); 1119 return (err); 1120 } 1121 1122 /* 1123 * Interpose on the SYS_exec syscall to detect native wrappers. 1124 */ 1125 int 1126 s10_exec(sysret_t *rval, const char *fname, const char **argp) 1127 { 1128 int err; 1129 1130 if ((err = s10_exec_native(rval, fname, argp, NULL)) != 0) 1131 return (err); 1132 1133 /* If an exec call succeeds, it never returns */ 1134 err = __systemcall(rval, SYS_exec + 1024, fname, argp); 1135 s10_assert(err != 0); 1136 return (err); 1137 } 1138 1139 /* 1140 * Interpose on the SYS_execve syscall to detect native wrappers. 1141 */ 1142 int 1143 s10_execve(sysret_t *rval, const char *fname, const char **argp, 1144 const char **envp) 1145 { 1146 int err; 1147 1148 if ((err = s10_exec_native(rval, fname, argp, envp)) != 0) 1149 return (err); 1150 1151 /* If an exec call succeeds, it never returns */ 1152 err = __systemcall(rval, SYS_execve + 1024, fname, argp, envp); 1153 s10_assert(err != 0); 1154 return (err); 1155 } 1156 1157 /* 1158 * S10's issetugid() syscall is now a subcode to privsys(). 1159 */ 1160 static int 1161 s10_issetugid(sysret_t *rval) 1162 { 1163 return (__systemcall(rval, SYS_privsys + 1024, PRIVSYS_ISSETUGID, 1164 0, 0, 0, 0, 0)); 1165 } 1166 1167 /* 1168 * New last arg "block" flag should be zero. The block flag is used by 1169 * the Opensolaris AIO implementation, which is now part of libc. 1170 */ 1171 static int 1172 s10_sigqueue(sysret_t *rval, pid_t pid, int signo, void *value, int si_code) 1173 { 1174 return (__systemcall(rval, SYS_sigqueue + 1024, pid, signo, value, 1175 si_code, 0)); 1176 } 1177 1178 static long 1179 s10_uname(sysret_t *rv, uintptr_t p1) 1180 { 1181 struct utsname un, *unp = (struct utsname *)p1; 1182 int rev, err; 1183 1184 if ((err = __systemcall(rv, SYS_uname + 1024, &un)) != 0) 1185 return (err); 1186 1187 rev = atoi(&un.release[2]); 1188 s10_assert(rev >= 11); 1189 bzero(un.release, _SYS_NMLN); 1190 (void) strlcpy(un.release, S10_UTS_RELEASE, _SYS_NMLN); 1191 bzero(un.version, _SYS_NMLN); 1192 (void) strlcpy(un.version, S10_UTS_VERSION, _SYS_NMLN); 1193 1194 /* copy out the modified uname info */ 1195 return (s10_uucopy(&un, unp, sizeof (un))); 1196 } 1197 1198 int 1199 s10_sysinfo(sysret_t *rv, int command, char *buf, long count) 1200 { 1201 char *value; 1202 int len; 1203 1204 /* 1205 * We must interpose on the sysinfo(2) commands SI_RELEASE and 1206 * SI_VERSION; all others get passed to the native sysinfo(2) 1207 * command. 1208 */ 1209 switch (command) { 1210 case SI_RELEASE: 1211 value = S10_UTS_RELEASE; 1212 break; 1213 1214 case SI_VERSION: 1215 value = S10_UTS_VERSION; 1216 break; 1217 1218 default: 1219 /* 1220 * The default action is to pass the command to the 1221 * native sysinfo(2) syscall. 1222 */ 1223 return (__systemcall(rv, SYS_systeminfo + 1024, 1224 command, buf, count)); 1225 } 1226 1227 len = strlen(value) + 1; 1228 if (count > 0) { 1229 if (s10_uucopystr(value, buf, count) != 0) 1230 return (EFAULT); 1231 1232 /* Assure NULL termination of buf as s10_uucopystr() doesn't. */ 1233 if (len > count && s10_uucopy("\0", buf + (count - 1), 1) != 0) 1234 return (EFAULT); 1235 } 1236 1237 /* 1238 * On success, sysinfo(2) returns the size of buffer required to hold 1239 * the complete value plus its terminating NULL byte. 1240 */ 1241 (void) S10_TRUSS_POINT_3(rv, SYS_systeminfo, 0, command, buf, count); 1242 rv->sys_rval1 = len; 1243 rv->sys_rval2 = 0; 1244 return (0); 1245 } 1246 1247 #ifdef __x86 1248 #ifdef __amd64 1249 /* 1250 * 64-bit x86 LWPs created by SYS_lwp_create start here if they need to set 1251 * their %fs registers to the legacy Solaris 10 selector value. 1252 * 1253 * This function does three things: 1254 * 1255 * 1. Trap to the kernel so that it can set %fs to the legacy Solaris 10 1256 * selector value. 1257 * 2. Read the LWP's true entry point (the entry point supplied by libc 1258 * when SYS_lwp_create was invoked) from %r14. 1259 * 3. Eliminate this function's stack frame and pass control to the LWP's 1260 * true entry point. 1261 * 1262 * See the comment above s10_lwp_create_correct_fs() (see below) for the reason 1263 * why this function exists. 1264 */ 1265 /*ARGSUSED*/ 1266 static void 1267 s10_lwp_create_entry_point(void *ulwp_structp) 1268 { 1269 sysret_t rval; 1270 1271 /* 1272 * The new LWP's %fs register is initially zero, but libc won't 1273 * function correctly when %fs is zero. Change the LWP's %fs register 1274 * via SYS_brand. 1275 */ 1276 (void) __systemcall(&rval, SYS_brand + 1024, B_S10_FSREGCORRECTION); 1277 1278 /* 1279 * Jump to the true entry point, which is stored in %r14. 1280 * Remove our stack frame before jumping so that 1281 * s10_lwp_create_entry_point() won't be seen in stack traces. 1282 * 1283 * NOTE: s10_lwp_create_entry_point() pushes %r12 onto its stack frame 1284 * so that it can use it as a temporary register. We don't restore %r12 1285 * in this assembly block because we don't care about its value (and 1286 * neither does _lwp_start()). Besides, the System V ABI AMD64 1287 * Actirecture Processor Supplement doesn't specify that %r12 should 1288 * have a special value when LWPs start, so we can ignore its value when 1289 * we jump to the true entry point. Furthermore, %r12 is a callee-saved 1290 * register, so the true entry point should push %r12 onto its stack 1291 * before using the register. We ignore %r14 after we read it for 1292 * similar reasons. 1293 * 1294 * NOTE: The compiler will generate a function epilogue for this 1295 * function despite the fact that the LWP will never execute it. 1296 * We could hand-code this entire function in assembly to eliminate 1297 * the epilogue, but the epilogue is only three or four instructions, 1298 * so we wouldn't save much space. Besides, why would we want 1299 * to create yet another ugly, hard-to-maintain assembly function when 1300 * we could write most of it in C? 1301 */ 1302 __asm__ __volatile__( 1303 "movq %0, %%rdi\n\t" /* pass ulwp_structp as arg1 */ 1304 "movq %%rbp, %%rsp\n\t" /* eliminate the stack frame */ 1305 "popq %%rbp\n\t" 1306 "jmp *%%r14\n\t" /* jump to the true entry point */ 1307 : : "r" (ulwp_structp)); 1308 /*NOTREACHED*/ 1309 } 1310 1311 /* 1312 * The S10 libc expects that %fs will be nonzero for new 64-bit x86 LWPs but the 1313 * Nevada kernel clears %fs for such LWPs. Unforunately, new LWPs do not issue 1314 * SYS_lwp_private (see s10_lwp_private() below) after they are created, so 1315 * we must ensure that new LWPs invoke a brand operation that sets %fs to a 1316 * nonzero value immediately after their creation. 1317 * 1318 * The easiest way to do this is to make new LWPs start at a special function, 1319 * s10_lwp_create_entry_point() (see its definition above), that invokes the 1320 * brand operation that corrects %fs. We'll store the entry points of new LWPs 1321 * in their %r14 registers so that s10_lwp_create_entry_point() can find and 1322 * call them after invoking the special brand operation. %r14 is a callee-saved 1323 * register; therefore, any functions invoked by s10_lwp_create_entry_point() 1324 * and all functions dealing with signals (e.g., sigacthandler()) will preserve 1325 * %r14 for s10_lwp_create_entry_point(). 1326 * 1327 * The Nevada kernel can safely work with nonzero %fs values because the kernel 1328 * configures per-thread %fs segment descriptors so that the legacy %fs selector 1329 * value will still work. See the comment in lwp_load() regarding %fs and 1330 * %fsbase in 64-bit x86 processes. 1331 * 1332 * This emulation exists thanks to CRs 6467491 and 6501650. 1333 */ 1334 static int 1335 s10_lwp_create_correct_fs(sysret_t *rval, ucontext_t *ucp, int flags, 1336 id_t *new_lwp) 1337 { 1338 ucontext_t s10_uc; 1339 1340 /* 1341 * Copy the supplied ucontext_t structure to the local stack 1342 * frame and store the new LWP's entry point (the value of %rip 1343 * stored in the ucontext_t) in the new LWP's %r14 register. 1344 * Then make s10_lwp_create_entry_point() the new LWP's entry 1345 * point. 1346 */ 1347 if (s10_uucopy(ucp, &s10_uc, sizeof (s10_uc)) != 0) 1348 return (EFAULT); 1349 s10_uc.uc_mcontext.gregs[REG_R14] = s10_uc.uc_mcontext.gregs[REG_RIP]; 1350 s10_uc.uc_mcontext.gregs[REG_RIP] = (greg_t)s10_lwp_create_entry_point; 1351 1352 /* 1353 * Issue SYS_lwp_create to create the new LWP. We pass the 1354 * modified ucontext_t to make sure that the new LWP starts at 1355 * s10_lwp_create_entry_point(). 1356 */ 1357 return (__systemcall(rval, SYS_lwp_create + 1024, &s10_uc, 1358 flags, new_lwp)); 1359 } 1360 #endif /* __amd64 */ 1361 1362 /* 1363 * This function is invoked on x86 systems when SYS_lwp_create is issued but no 1364 * %fs register correction is necessary. 1365 * 1366 * See the comment above s10_lwp_create_correct_fs() above for more details. 1367 */ 1368 static int 1369 s10_lwp_create(sysret_t *rval, ucontext_t *ucp, int flags, id_t *new_lwp) 1370 { 1371 return (__systemcall(rval, SYS_lwp_create + 1024, ucp, flags, new_lwp)); 1372 } 1373 1374 /* 1375 * SYS_lwp_private is issued by libc_init() to set %fsbase in 64-bit x86 1376 * processes. The Nevada kernel sets %fs to zero but the S10 libc expects 1377 * %fs to be nonzero. We'll pass the issued system call to the kernel untouched 1378 * and invoke a brand operation to set %fs to the legacy S10 selector value. 1379 * 1380 * This emulation exists thanks to CRs 6467491 and 6501650. 1381 */ 1382 static int 1383 s10_lwp_private(sysret_t *rval, int cmd, int which, uintptr_t base) 1384 { 1385 #ifdef __amd64 1386 int err; 1387 1388 /* 1389 * The current LWP's %fs register should be zero. Determine whether the 1390 * Solaris 10 libc with which we're working functions correctly when %fs 1391 * is zero by calling thr_main() after issuing the SYS_lwp_private 1392 * syscall. If thr_main() barfs (returns -1), then change the LWP's %fs 1393 * register via SYS_brand and patch s10_sysent_table so that issuing 1394 * SYS_lwp_create executes s10_lwp_create_correct_fs() rather than the 1395 * default s10_lwp_create(). s10_lwp_create_correct_fs() will 1396 * guarantee that new LWPs will have correct %fs values. 1397 */ 1398 if ((err = __systemcall(rval, SYS_lwp_private + 1024, cmd, which, 1399 base)) != 0) 1400 return (err); 1401 if (thr_main() == -1) { 1402 /* 1403 * SYS_lwp_private is only issued by libc_init(), which is 1404 * executed when libc is first loaded by ld.so.1. Thus we 1405 * are guaranteed to be single-threaded at this point. Even 1406 * if we were multithreaded at this point, writing a 64-bit 1407 * value to the st_callc field of a s10_sysent_table 1408 * entry is guaranteed to be atomic on 64-bit x86 chips 1409 * as long as the field is not split across cache lines 1410 * (It shouldn't be.). See chapter 8, section 1.1 of 1411 * "The Intel 64 and IA32 Architectures Software Developer's 1412 * Manual," Volume 3A for more details. 1413 */ 1414 s10_sysent_table[SYS_lwp_create].st_callc = 1415 (sysent_cb_t)s10_lwp_create_correct_fs; 1416 return (__systemcall(rval, SYS_brand + 1024, 1417 B_S10_FSREGCORRECTION)); 1418 } 1419 return (0); 1420 #else /* !__amd64 */ 1421 return (__systemcall(rval, SYS_lwp_private + 1024, cmd, which, base)); 1422 #endif /* !__amd64 */ 1423 } 1424 #endif /* __x86 */ 1425 1426 /* 1427 * The Opensolaris versions of lwp_mutex_timedlock() and lwp_mutex_trylock() 1428 * add an extra argument to the interfaces, a uintptr_t value for the mutex's 1429 * mutex_owner field. The Solaris 10 libc assigns the mutex_owner field at 1430 * user-level, so we just make the extra argument be zero in both syscalls. 1431 */ 1432 1433 static int 1434 s10_lwp_mutex_timedlock(sysret_t *rval, lwp_mutex_t *lp, timespec_t *tsp) 1435 { 1436 return (__systemcall(rval, SYS_lwp_mutex_timedlock + 1024, lp, tsp, 0)); 1437 } 1438 1439 static int 1440 s10_lwp_mutex_trylock(sysret_t *rval, lwp_mutex_t *lp) 1441 { 1442 return (__systemcall(rval, SYS_lwp_mutex_trylock + 1024, lp, 0)); 1443 } 1444 1445 /* 1446 * If the emul_global_zone flag is set then emulate some aspects of the 1447 * zone system call. In particular, emulate the global zone ID on the 1448 * ZONE_LOOKUP subcommand and emulate some of the global zone attributes 1449 * on the ZONE_GETATTR subcommand. If the flag is not set or we're performing 1450 * some other operation, simply pass the calls through. 1451 */ 1452 int 1453 s10_zone(sysret_t *rval, int cmd, void *arg1, void *arg2, void *arg3, 1454 void *arg4) 1455 { 1456 char *aval; 1457 int len; 1458 zoneid_t zid; 1459 int attr; 1460 char *buf; 1461 size_t bufsize; 1462 1463 /* 1464 * We only emulate the zone syscall for a subset of specific commands, 1465 * otherwise we just pass the call through. 1466 */ 1467 if (!emul_global_zone) 1468 return (__systemcall(rval, SYS_zone + 1024, cmd, arg1, arg2, 1469 arg3, arg4)); 1470 1471 switch (cmd) { 1472 case ZONE_LOOKUP: 1473 (void) S10_TRUSS_POINT_1(rval, SYS_zone, 0, cmd); 1474 rval->sys_rval1 = GLOBAL_ZONEID; 1475 rval->sys_rval2 = 0; 1476 return (0); 1477 1478 case ZONE_GETATTR: 1479 zid = (zoneid_t)(uintptr_t)arg1; 1480 attr = (int)(uintptr_t)arg2; 1481 buf = (char *)arg3; 1482 bufsize = (size_t)arg4; 1483 1484 /* 1485 * If the request is for the global zone then we're emulating 1486 * that, otherwise pass this thru. 1487 */ 1488 if (zid != GLOBAL_ZONEID) 1489 goto passthru; 1490 1491 switch (attr) { 1492 case ZONE_ATTR_NAME: 1493 aval = GLOBAL_ZONENAME; 1494 break; 1495 1496 case ZONE_ATTR_BRAND: 1497 aval = NATIVE_BRAND_NAME; 1498 break; 1499 default: 1500 /* 1501 * We only emulate a subset of the attrs, use the 1502 * real zone id to pass thru the rest. 1503 */ 1504 arg1 = (void *)(uintptr_t)zoneid; 1505 goto passthru; 1506 } 1507 1508 (void) S10_TRUSS_POINT_5(rval, SYS_zone, 0, cmd, zid, attr, 1509 buf, bufsize); 1510 1511 len = strlen(aval) + 1; 1512 if (len > bufsize) 1513 return (ENAMETOOLONG); 1514 1515 if (buf != NULL) { 1516 if (len == 1) { 1517 if (s10_uucopy("\0", buf, 1) != 0) 1518 return (EFAULT); 1519 } else { 1520 if (s10_uucopystr(aval, buf, len) != 0) 1521 return (EFAULT); 1522 1523 /* 1524 * Assure NULL termination of "buf" as 1525 * s10_uucopystr() does NOT. 1526 */ 1527 if (s10_uucopy("\0", buf + (len - 1), 1) != 0) 1528 return (EFAULT); 1529 } 1530 } 1531 1532 rval->sys_rval1 = len; 1533 rval->sys_rval2 = 0; 1534 return (0); 1535 1536 default: 1537 break; 1538 } 1539 1540 passthru: 1541 return (__systemcall(rval, SYS_zone + 1024, cmd, arg1, arg2, arg3, 1542 arg4)); 1543 } 1544 1545 /* 1546 * Close a libc file handle, but don't actually close the underlying 1547 * file descriptor. 1548 */ 1549 static void 1550 s10_close_fh(FILE *file) 1551 { 1552 int fd, fd_new; 1553 1554 if (file == NULL) 1555 return; 1556 1557 if ((fd = fileno(file)) < 0) 1558 return; 1559 1560 fd_new = dup(fd); 1561 if (fd_new == -1) 1562 return; 1563 1564 (void) fclose(file); 1565 (void) dup2(fd_new, fd); 1566 (void) close(fd_new); 1567 } 1568 1569 /*ARGSUSED*/ 1570 int 1571 s10_init(int argc, char *argv[], char *envp[]) 1572 { 1573 sysret_t rval; 1574 s10_brand_reg_t reg; 1575 s10_elf_data_t sed; 1576 auxv_t *ap; 1577 uintptr_t *p; 1578 int i, err; 1579 char *bname; 1580 1581 /* Sanity check our translation table return value codes */ 1582 for (i = 0; i < NSYSCALL; i++) { 1583 s10_sysent_table_t *est = &(s10_sysent_table[i]); 1584 s10_assert(BIT_ONLYONESET(est->st_args & RV_MASK)); 1585 } 1586 1587 /* 1588 * We need to shutdown all libc stdio. libc stdio normally goes to 1589 * file descriptors, but since we're actually part of a another 1590 * process we don't own these file descriptors and we can't make 1591 * any assumptions about their state. 1592 */ 1593 s10_close_fh(stdin); 1594 s10_close_fh(stdout); 1595 s10_close_fh(stderr); 1596 1597 /* 1598 * Cache the pid of the zone's init process and determine if 1599 * we're init(1m) for the zone. Remember: we might be init 1600 * now, but as soon as we fork(2) we won't be. 1601 */ 1602 (void) get_initpid_info(); 1603 1604 /* get the current zoneid */ 1605 err = __systemcall(&rval, SYS_zone, ZONE_LOOKUP, NULL); 1606 s10_assert(err == 0); 1607 zoneid = (zoneid_t)rval.sys_rval1; 1608 1609 /* Get the zone's emulation bitmap. */ 1610 if ((err = __systemcall(&rval, SYS_zone, ZONE_GETATTR, zoneid, 1611 S10_EMUL_BITMAP, emul_bitmap, sizeof (emul_bitmap))) != 0) { 1612 s10_abort(err, "The zone's patch level is unsupported"); 1613 /*NOTREACHED*/ 1614 } 1615 1616 bname = basename(argv[0]); 1617 1618 /* 1619 * In general we want the S10 commands that are zone-aware to continue 1620 * to behave as they normally do within a zone. Since these commands 1621 * are zone-aware, they should continue to "do the right thing". 1622 * However, some zone-aware commands aren't going to work the way 1623 * we expect them to inside the branded zone. In particular, the pkg 1624 * and patch commands will not properly manage all pkgs/patches 1625 * unless the commands think they are running in the global zone. For 1626 * these commands we want to emulate the global zone. 1627 * 1628 * We don't do any emulation for pkgcond since it is typically used 1629 * in pkg/patch postinstall scripts and we want those scripts to do 1630 * the right thing inside a zone. 1631 * 1632 * One issue is the handling of hollow pkgs. Since the pkgs are 1633 * hollow, they won't use pkgcond in their postinstall scripts. These 1634 * pkgs typically are installing drivers so we handle that by 1635 * replacing add_drv and rem_drv in the s10_boot script. 1636 */ 1637 if (strcmp("pkgadd", bname) == 0 || strcmp("pkgrm", bname) == 0 || 1638 strcmp("patchadd", bname) == 0 || strcmp("patchrm", bname) == 0) 1639 emul_global_zone = B_TRUE; 1640 1641 /* 1642 * Register our syscall emulation table with the kernel. 1643 * Note that we don't have to do invoke (syscall_number + 1024) 1644 * until we've actually establised a syscall emulation callback 1645 * handler address, which is what we're doing with this brand 1646 * syscall. 1647 */ 1648 reg.sbr_version = S10_VERSION; 1649 reg.sbr_handler = (caddr_t)s10_handler; 1650 if ((err = __systemcall(&rval, SYS_brand, B_REGISTER, ®)) != 0) { 1651 s10_abort(err, "Failed to brand current process"); 1652 /*NOTREACHED*/ 1653 } 1654 1655 /* Get data about the executable we're running from the kernel. */ 1656 if ((err = __systemcall(&rval, SYS_brand + 1024, 1657 B_ELFDATA, (void *)&sed)) != 0) { 1658 s10_abort(err, 1659 "Failed to get required brand ELF data from the kernel"); 1660 /*NOTREACHED*/ 1661 } 1662 1663 /* 1664 * Find the aux vector on the stack. 1665 */ 1666 p = (uintptr_t *)envp; 1667 while (*p != NULL) 1668 p++; 1669 1670 /* 1671 * p is now pointing at the 0 word after the environ pointers. 1672 * After that is the aux vectors. 1673 * 1674 * The aux vectors are currently pointing to the brand emulation 1675 * library and associated linker. We're going to change them to 1676 * point to the brand executable and associated linker (or to no 1677 * linker for static binaries). This matches the process data 1678 * stored within the kernel and visible from /proc, which was 1679 * all setup in s10_elfexec(). We do this so that when a debugger 1680 * attaches to the process it sees the process as a normal solaris 1681 * process, this brand emulation library and everything on it's 1682 * link map will not be visible, unless our librtld_db plugin 1683 * is used. Note that this is very different from how Linux 1684 * branded processes are implemented within lx branded zones. 1685 * In that situation, the primary linkmap of the process is the 1686 * brand emulation libraries linkmap, not the Linux applications 1687 * linkmap. 1688 * 1689 * We also need to clear the AF_SUN_NOPLM flag from the AT_SUN_AUXFLAGS 1690 * aux vector. This flag told our linker that we don't have a 1691 * primary link map. Now that our linker is done initializing, we 1692 * want to clear this flag before we transfer control to the 1693 * applications copy of the linker, since we want that linker to have 1694 * a primary link map which will be the link map for the application 1695 * we're running. 1696 */ 1697 p++; 1698 for (ap = (auxv_t *)p; ap->a_type != AT_NULL; ap++) { 1699 switch (ap->a_type) { 1700 case AT_BASE: 1701 /* Hide AT_BASE if static binary */ 1702 if (sed.sed_base == NULL) { 1703 ap->a_type = AT_IGNORE; 1704 ap->a_un.a_val = NULL; 1705 } else { 1706 ap->a_un.a_val = sed.sed_base; 1707 } 1708 break; 1709 case AT_ENTRY: 1710 ap->a_un.a_val = sed.sed_entry; 1711 break; 1712 case AT_PHDR: 1713 ap->a_un.a_val = sed.sed_phdr; 1714 break; 1715 case AT_PHENT: 1716 ap->a_un.a_val = sed.sed_phent; 1717 break; 1718 case AT_PHNUM: 1719 ap->a_un.a_val = sed.sed_phnum; 1720 break; 1721 case AT_SUN_AUXFLAGS: 1722 ap->a_un.a_val &= ~AF_SUN_NOPLM; 1723 break; 1724 case AT_SUN_EMULATOR: 1725 /* 1726 * ld.so.1 inspects AT_SUN_EMULATOR to see if 1727 * if it is the linker for the brand emulation 1728 * library. Hide AT_SUN_EMULATOR, as the 1729 * linker we are about to jump to is the linker 1730 * for the binary. 1731 */ 1732 ap->a_type = AT_IGNORE; 1733 ap->a_un.a_val = NULL; 1734 break; 1735 case AT_SUN_LDDATA: 1736 /* Hide AT_SUN_LDDATA if static binary */ 1737 if (sed.sed_lddata == NULL) { 1738 ap->a_type = AT_IGNORE; 1739 ap->a_un.a_val = NULL; 1740 } else { 1741 ap->a_un.a_val = sed.sed_lddata; 1742 } 1743 break; 1744 default: 1745 break; 1746 } 1747 } 1748 1749 s10_runexe(argv, sed.sed_ldentry); 1750 /*NOTREACHED*/ 1751 s10_abort(0, "s10_runexe() returned"); 1752 return (-1); 1753 } 1754 1755 /* 1756 * This table must have at least NSYSCALL entries in it. 1757 * 1758 * The second parameter of each entry in the s10_sysent_table 1759 * contains the number of parameters and flags that describe the 1760 * syscall return value encoding. See the block comments at the 1761 * top of this file for more information about the syscall return 1762 * value flags and when they should be used. 1763 */ 1764 s10_sysent_table_t s10_sysent_table[] = { 1765 #if defined(__sparc) && !defined(__sparcv9) 1766 EMULATE(s10_indir, 9 | RV_64RVAL), /* 0 */ 1767 #else /* !__sparc || __sparcv9 */ 1768 NOSYS, /* 0 */ 1769 #endif /* !__sparc || __sparcv9 */ 1770 NOSYS, /* 1 */ 1771 NOSYS, /* 2 */ 1772 NOSYS, /* 3 */ 1773 NOSYS, /* 4 */ 1774 NOSYS, /* 5 */ 1775 NOSYS, /* 6 */ 1776 NOSYS, /* 7 */ 1777 NOSYS, /* 8 */ 1778 NOSYS, /* 9 */ 1779 NOSYS, /* 10 */ 1780 EMULATE(s10_exec, 2 | RV_DEFAULT), /* 11 */ 1781 NOSYS, /* 12 */ 1782 NOSYS, /* 13 */ 1783 NOSYS, /* 14 */ 1784 NOSYS, /* 15 */ 1785 NOSYS, /* 16 */ 1786 NOSYS, /* 17 */ 1787 NOSYS, /* 18 */ 1788 NOSYS, /* 19 */ 1789 NOSYS, /* 20 */ 1790 NOSYS, /* 21 */ 1791 NOSYS, /* 22 */ 1792 NOSYS, /* 23 */ 1793 NOSYS, /* 24 */ 1794 NOSYS, /* 25 */ 1795 NOSYS, /* 26 */ 1796 NOSYS, /* 27 */ 1797 NOSYS, /* 28 */ 1798 NOSYS, /* 29 */ 1799 NOSYS, /* 30 */ 1800 NOSYS, /* 31 */ 1801 NOSYS, /* 32 */ 1802 NOSYS, /* 33 */ 1803 NOSYS, /* 34 */ 1804 NOSYS, /* 35 */ 1805 NOSYS, /* 36 */ 1806 NOSYS, /* 37 */ 1807 NOSYS, /* 38 */ 1808 NOSYS, /* 39 */ 1809 NOSYS, /* 40 */ 1810 NOSYS, /* 41 */ 1811 NOSYS, /* 42 */ 1812 NOSYS, /* 43 */ 1813 NOSYS, /* 44 */ 1814 NOSYS, /* 45 */ 1815 NOSYS, /* 46 */ 1816 NOSYS, /* 47 */ 1817 NOSYS, /* 48 */ 1818 NOSYS, /* 49 */ 1819 NOSYS, /* 50 */ 1820 NOSYS, /* 51 */ 1821 NOSYS, /* 52 */ 1822 NOSYS, /* 53 */ 1823 EMULATE(s10_ioctl, 3 | RV_DEFAULT), /* 54 */ 1824 NOSYS, /* 55 */ 1825 NOSYS, /* 56 */ 1826 NOSYS, /* 57 */ 1827 NOSYS, /* 58 */ 1828 EMULATE(s10_execve, 3 | RV_DEFAULT), /* 59 */ 1829 NOSYS, /* 60 */ 1830 NOSYS, /* 61 */ 1831 NOSYS, /* 62 */ 1832 NOSYS, /* 63 */ 1833 NOSYS, /* 64 */ 1834 NOSYS, /* 65 */ 1835 NOSYS, /* 66 */ 1836 NOSYS, /* 67 */ 1837 NOSYS, /* 68 */ 1838 NOSYS, /* 69 */ 1839 NOSYS, /* 70 */ 1840 EMULATE(s10_acctctl, 3 | RV_DEFAULT), /* 71 */ 1841 NOSYS, /* 72 */ 1842 NOSYS, /* 73 */ 1843 NOSYS, /* 74 */ 1844 EMULATE(s10_issetugid, 0 | RV_DEFAULT), /* 75 */ 1845 NOSYS, /* 76 */ 1846 NOSYS, /* 77 */ 1847 NOSYS, /* 78 */ 1848 NOSYS, /* 79 */ 1849 NOSYS, /* 80 */ 1850 NOSYS, /* 81 */ 1851 NOSYS, /* 82 */ 1852 NOSYS, /* 83 */ 1853 NOSYS, /* 84 */ 1854 NOSYS, /* 85 */ 1855 NOSYS, /* 86 */ 1856 NOSYS, /* 87 */ 1857 NOSYS, /* 88 */ 1858 NOSYS, /* 89 */ 1859 NOSYS, /* 90 */ 1860 NOSYS, /* 91 */ 1861 NOSYS, /* 92 */ 1862 NOSYS, /* 93 */ 1863 NOSYS, /* 94 */ 1864 NOSYS, /* 95 */ 1865 NOSYS, /* 96 */ 1866 NOSYS, /* 97 */ 1867 NOSYS, /* 98 */ 1868 NOSYS, /* 99 */ 1869 NOSYS, /* 100 */ 1870 NOSYS, /* 101 */ 1871 NOSYS, /* 102 */ 1872 NOSYS, /* 103 */ 1873 NOSYS, /* 104 */ 1874 NOSYS, /* 105 */ 1875 NOSYS, /* 106 */ 1876 NOSYS, /* 107 */ 1877 NOSYS, /* 108 */ 1878 NOSYS, /* 109 */ 1879 NOSYS, /* 110 */ 1880 NOSYS, /* 111 */ 1881 NOSYS, /* 112 */ 1882 NOSYS, /* 113 */ 1883 NOSYS, /* 114 */ 1884 NOSYS, /* 115 */ 1885 NOSYS, /* 116 */ 1886 NOSYS, /* 117 */ 1887 NOSYS, /* 118 */ 1888 NOSYS, /* 119 */ 1889 NOSYS, /* 120 */ 1890 NOSYS, /* 121 */ 1891 NOSYS, /* 122 */ 1892 NOSYS, /* 123 */ 1893 NOSYS, /* 124 */ 1894 NOSYS, /* 125 */ 1895 NOSYS, /* 126 */ 1896 NOSYS, /* 127 */ 1897 NOSYS, /* 128 */ 1898 NOSYS, /* 129 */ 1899 NOSYS, /* 130 */ 1900 NOSYS, /* 131 */ 1901 NOSYS, /* 132 */ 1902 NOSYS, /* 133 */ 1903 NOSYS, /* 134 */ 1904 EMULATE(s10_uname, 1 | RV_DEFAULT), /* 135 */ 1905 NOSYS, /* 136 */ 1906 NOSYS, /* 137 */ 1907 NOSYS, /* 138 */ 1908 EMULATE(s10_sysinfo, 3 | RV_DEFAULT), /* 139 */ 1909 NOSYS, /* 140 */ 1910 NOSYS, /* 141 */ 1911 NOSYS, /* 142 */ 1912 NOSYS, /* 143 */ 1913 NOSYS, /* 144 */ 1914 NOSYS, /* 145 */ 1915 NOSYS, /* 146 */ 1916 NOSYS, /* 147 */ 1917 NOSYS, /* 148 */ 1918 NOSYS, /* 149 */ 1919 NOSYS, /* 150 */ 1920 NOSYS, /* 151 */ 1921 NOSYS, /* 152 */ 1922 NOSYS, /* 153 */ 1923 NOSYS, /* 154 */ 1924 NOSYS, /* 155 */ 1925 NOSYS, /* 156 */ 1926 NOSYS, /* 157 */ 1927 NOSYS, /* 158 */ 1928 #ifdef __x86 1929 EMULATE(s10_lwp_create, 3 | RV_DEFAULT), /* 159 */ 1930 #else /* !__x86 */ 1931 NOSYS, /* 159 */ 1932 #endif /* !__x86 */ 1933 NOSYS, /* 160 */ 1934 NOSYS, /* 161 */ 1935 NOSYS, /* 162 */ 1936 NOSYS, /* 163 */ 1937 NOSYS, /* 164 */ 1938 NOSYS, /* 165 */ 1939 #ifdef __x86 1940 EMULATE(s10_lwp_private, 3 | RV_DEFAULT), /* 166 */ 1941 #else /* !__x86 */ 1942 NOSYS, /* 166 */ 1943 #endif /* !__x86 */ 1944 NOSYS, /* 167 */ 1945 NOSYS, /* 168 */ 1946 NOSYS, /* 169 */ 1947 NOSYS, /* 170 */ 1948 NOSYS, /* 171 */ 1949 NOSYS, /* 172 */ 1950 NOSYS, /* 173 */ 1951 EMULATE(s10_pwrite, 4 | RV_DEFAULT), /* 174 */ 1952 NOSYS, /* 175 */ 1953 NOSYS, /* 176 */ 1954 NOSYS, /* 177 */ 1955 NOSYS, /* 178 */ 1956 NOSYS, /* 179 */ 1957 NOSYS, /* 180 */ 1958 NOSYS, /* 181 */ 1959 NOSYS, /* 182 */ 1960 NOSYS, /* 183 */ 1961 NOSYS, /* 184 */ 1962 NOSYS, /* 185 */ 1963 EMULATE(s10_auditsys, 4 | RV_64RVAL), /* 186 */ 1964 NOSYS, /* 187 */ 1965 NOSYS, /* 188 */ 1966 NOSYS, /* 189 */ 1967 EMULATE(s10_sigqueue, 4 | RV_DEFAULT), /* 190 */ 1968 NOSYS, /* 191 */ 1969 NOSYS, /* 192 */ 1970 NOSYS, /* 193 */ 1971 NOSYS, /* 194 */ 1972 NOSYS, /* 195 */ 1973 NOSYS, /* 196 */ 1974 NOSYS, /* 197 */ 1975 NOSYS, /* 198 */ 1976 NOSYS, /* 199 */ 1977 NOSYS, /* 200 */ 1978 NOSYS, /* 201 */ 1979 NOSYS, /* 202 */ 1980 NOSYS, /* 203 */ 1981 NOSYS, /* 204 */ 1982 NOSYS, /* 205 */ 1983 NOSYS, /* 206 */ 1984 NOSYS, /* 207 */ 1985 NOSYS, /* 208 */ 1986 NOSYS, /* 209 */ 1987 EMULATE(s10_lwp_mutex_timedlock, 2 | RV_DEFAULT), /* 210 */ 1988 NOSYS, /* 211 */ 1989 NOSYS, /* 212 */ 1990 NOSYS, /* 213 */ 1991 NOSYS, /* 214 */ 1992 NOSYS, /* 215 */ 1993 NOSYS, /* 216 */ 1994 NOSYS, /* 217 */ 1995 NOSYS, /* 218 */ 1996 NOSYS, /* 219 */ 1997 NOSYS, /* 220 */ 1998 NOSYS, /* 221 */ 1999 NOSYS, /* 222 */ 2000 #ifdef _LP64 2001 NOSYS, /* 223 */ 2002 #else /* !_LP64 */ 2003 EMULATE(s10_pwrite64, 5 | RV_DEFAULT), /* 223 */ 2004 #endif /* !_LP64 */ 2005 NOSYS, /* 224 */ 2006 NOSYS, /* 225 */ 2007 NOSYS, /* 226 */ 2008 EMULATE(s10_zone, 5 | RV_DEFAULT), /* 227 */ 2009 NOSYS, /* 228 */ 2010 NOSYS, /* 229 */ 2011 NOSYS, /* 230 */ 2012 NOSYS, /* 231 */ 2013 NOSYS, /* 232 */ 2014 NOSYS, /* 233 */ 2015 NOSYS, /* 234 */ 2016 NOSYS, /* 235 */ 2017 NOSYS, /* 236 */ 2018 NOSYS, /* 237 */ 2019 NOSYS, /* 238 */ 2020 NOSYS, /* 239 */ 2021 NOSYS, /* 240 */ 2022 NOSYS, /* 241 */ 2023 NOSYS, /* 242 */ 2024 NOSYS, /* 243 */ 2025 NOSYS, /* 244 */ 2026 NOSYS, /* 245 */ 2027 NOSYS, /* 246 */ 2028 NOSYS, /* 247 */ 2029 NOSYS, /* 248 */ 2030 NOSYS, /* 249 */ 2031 NOSYS, /* 250 */ 2032 EMULATE(s10_lwp_mutex_trylock, 1 | RV_DEFAULT), /* 251 */ 2033 NOSYS, /* 252 */ 2034 NOSYS, /* 253 */ 2035 NOSYS, /* 254 */ 2036 NOSYS /* 255 */ 2037 }; 2038