| 1 | /* $NetBSD: emul.c,v 1.191 2019/06/02 19:41:51 kre Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 2007-2011 Antti Kantee. All Rights Reserved. |
| 5 | * |
| 6 | * Redistribution and use in source and binary forms, with or without |
| 7 | * modification, are permitted provided that the following conditions |
| 8 | * are met: |
| 9 | * 1. Redistributions of source code must retain the above copyright |
| 10 | * notice, this list of conditions and the following disclaimer. |
| 11 | * 2. Redistributions in binary form must reproduce the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer in the |
| 13 | * documentation and/or other materials provided with the distribution. |
| 14 | * |
| 15 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS |
| 16 | * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED |
| 17 | * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
| 18 | * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
| 19 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 20 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 21 | * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 22 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 23 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 24 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 25 | * SUCH DAMAGE. |
| 26 | */ |
| 27 | |
| 28 | #include <sys/cdefs.h> |
| 29 | __KERNEL_RCSID(0, "$NetBSD: emul.c,v 1.191 2019/06/02 19:41:51 kre Exp $" ); |
| 30 | |
| 31 | #include <sys/param.h> |
| 32 | #include <sys/cprng.h> |
| 33 | #include <sys/filedesc.h> |
| 34 | #include <sys/fstrans.h> |
| 35 | #include <sys/kauth.h> |
| 36 | #include <sys/module.h> |
| 37 | #include <sys/reboot.h> |
| 38 | #include <sys/syscall.h> |
| 39 | #include <sys/pserialize.h> |
| 40 | #ifdef LOCKDEBUG |
| 41 | #include <sys/sleepq.h> |
| 42 | #endif |
| 43 | |
| 44 | #include <dev/cons.h> |
| 45 | |
| 46 | #include <rump-sys/kern.h> |
| 47 | |
| 48 | #include <rump/rumpuser.h> |
| 49 | |
| 50 | void (*rump_vfs_fini)(void) = (void *)nullop; |
| 51 | |
| 52 | /* |
| 53 | * physmem is largely unused (except for nmbcluster calculations), |
| 54 | * so pick a default value which suits ZFS. if an application wants |
| 55 | * a very small memory footprint, it can still adjust this before |
| 56 | * calling rump_init() |
| 57 | */ |
| 58 | #define PHYSMEM 512*256 |
| 59 | psize_t physmem = PHYSMEM; |
| 60 | int nkmempages = PHYSMEM/2; /* from le chapeau */ |
| 61 | #undef PHYSMEM |
| 62 | |
| 63 | struct vnode *rootvp; |
| 64 | dev_t rootdev = NODEV; |
| 65 | |
| 66 | const int schedppq = 1; |
| 67 | bool mp_online = false; |
| 68 | struct timespec boottime; |
| 69 | int cold = 1; |
| 70 | int boothowto = AB_SILENT; |
| 71 | struct tty *constty; |
| 72 | |
| 73 | const struct bdevsw *bdevsw0[255]; |
| 74 | const struct bdevsw **bdevsw = bdevsw0; |
| 75 | const int sys_cdevsws = 255; |
| 76 | int max_cdevsws = 255; |
| 77 | |
| 78 | const struct cdevsw *cdevsw0[255]; |
| 79 | const struct cdevsw **cdevsw = cdevsw0; |
| 80 | const int sys_bdevsws = 255; |
| 81 | int max_bdevsws = 255; |
| 82 | |
| 83 | int mem_no = 2; |
| 84 | |
| 85 | device_t booted_device; |
| 86 | device_t booted_wedge; |
| 87 | int booted_partition; |
| 88 | const char *booted_method; |
| 89 | |
| 90 | /* XXX: unused */ |
| 91 | kmutex_t tty_lock; |
| 92 | krwlock_t exec_lock; |
| 93 | |
| 94 | /* sparc doesn't sport constant page size, pretend we have 4k pages */ |
| 95 | #ifdef __sparc__ |
| 96 | int nbpg = 4096; |
| 97 | int pgofset = 4096-1; |
| 98 | int pgshift = 12; |
| 99 | #endif |
| 100 | |
| 101 | /* on sun3 VM_MAX_ADDRESS is a const variable */ |
| 102 | /* XXX: should be moved into rump.c and initialize for sun3 and sun3x? */ |
| 103 | #ifdef sun3 |
| 104 | const vaddr_t kernbase = KERNBASE3; |
| 105 | #endif |
| 106 | |
| 107 | struct loadavg averunnable = { |
| 108 | { 0 * FSCALE, |
| 109 | 1 * FSCALE, |
| 110 | 11 * FSCALE, }, |
| 111 | FSCALE, |
| 112 | }; |
| 113 | |
| 114 | /* |
| 115 | * Include the autogenerated list of auto-loadable syscalls |
| 116 | */ |
| 117 | #include <kern/syscalls_autoload.c> |
| 118 | |
| 119 | struct emul emul_netbsd = { |
| 120 | .e_name = "netbsd-rump" , |
| 121 | .e_sysent = rump_sysent, |
| 122 | .e_nomodbits = rump_sysent_nomodbits, |
| 123 | #ifndef __HAVE_MINIMAL_EMUL |
| 124 | .e_nsysent = SYS_NSYSENT, |
| 125 | #endif |
| 126 | .e_vm_default_addr = uvm_default_mapaddr, |
| 127 | #ifdef __HAVE_SYSCALL_INTERN |
| 128 | .e_syscall_intern = syscall_intern, |
| 129 | #endif |
| 130 | .e_sc_autoload = netbsd_syscalls_autoload, |
| 131 | }; |
| 132 | |
| 133 | cprng_strong_t *kern_cprng; |
| 134 | |
| 135 | /* not used, but need the symbols for pointer comparisons */ |
| 136 | syncobj_t mutex_syncobj, rw_syncobj; |
| 137 | |
| 138 | int |
| 139 | kpause(const char *wmesg, bool intr, int timeo, kmutex_t *mtx) |
| 140 | { |
| 141 | extern int hz; |
| 142 | int rv __diagused; |
| 143 | uint64_t sec, nsec; |
| 144 | |
| 145 | if (mtx) |
| 146 | mutex_exit(mtx); |
| 147 | |
| 148 | sec = timeo / hz; |
| 149 | nsec = (timeo % hz) * (1000000000 / hz); |
| 150 | rv = rumpuser_clock_sleep(RUMPUSER_CLOCK_RELWALL, sec, nsec); |
| 151 | KASSERT(rv == 0); |
| 152 | |
| 153 | if (mtx) |
| 154 | mutex_enter(mtx); |
| 155 | |
| 156 | return 0; |
| 157 | } |
| 158 | |
| 159 | vaddr_t |
| 160 | calc_cache_size(vsize_t vasz, int pct, int va_pct) |
| 161 | { |
| 162 | paddr_t t; |
| 163 | |
| 164 | t = (paddr_t)physmem * pct / 100 * PAGE_SIZE; |
| 165 | if ((vaddr_t)t != t) { |
| 166 | panic("%s: needs tweak" , __func__); |
| 167 | } |
| 168 | return t; |
| 169 | } |
| 170 | |
| 171 | #define RETURN_ADDRESS (uintptr_t)__builtin_return_address(0) |
| 172 | |
| 173 | void |
| 174 | assert_sleepable(void) |
| 175 | { |
| 176 | const char *reason = NULL; |
| 177 | |
| 178 | /* always sleepable, although we should improve this */ |
| 179 | |
| 180 | if (!pserialize_not_in_read_section()) { |
| 181 | reason = "pserialize" ; |
| 182 | } |
| 183 | |
| 184 | if (reason) { |
| 185 | panic("%s: %s caller=%p" , __func__, reason, |
| 186 | (void *)RETURN_ADDRESS); |
| 187 | } |
| 188 | } |
| 189 | |
| 190 | void |
| 191 | module_init_md(void) |
| 192 | { |
| 193 | |
| 194 | /* |
| 195 | * Nothing for now. However, we should load the librump |
| 196 | * symbol table. |
| 197 | */ |
| 198 | } |
| 199 | |
| 200 | /* |
| 201 | * Try to emulate all the MD definitions of DELAY() / delay(). |
| 202 | * Would be nice to fix the #defines in MD headers, but this quicker. |
| 203 | * |
| 204 | * XXX: we'd need a rumpuser_clock_sleep_nowrap() here. Since we |
| 205 | * don't have it in the current hypercall revision, busyloop. |
| 206 | * Note that rather than calibrate a loop delay and work with that, |
| 207 | * get call gettime (which does not block) in a loop to make sure |
| 208 | * we didn't get virtual ghosttime. That might be slightly inaccurate |
| 209 | * for very small delays ... |
| 210 | * |
| 211 | * The other option would be to run a thread in the hypervisor which |
| 212 | * sleeps for us and we can wait for it using rumpuser_cv_wait_nowrap() |
| 213 | * Probably too fussy. Better just wait for hypercall rev 18 ;) |
| 214 | */ |
| 215 | static void |
| 216 | rump_delay(unsigned int us) |
| 217 | { |
| 218 | struct timespec target, tmp; |
| 219 | uint64_t sec, sec_ini, sec_now; |
| 220 | long nsec, nsec_ini, nsec_now; |
| 221 | int loops; |
| 222 | |
| 223 | rumpuser_clock_gettime(RUMPUSER_CLOCK_ABSMONO, &sec_ini, &nsec_ini); |
| 224 | |
| 225 | #ifdef __mac68k__ |
| 226 | sec = us / 1000; |
| 227 | nsec = (us % 1000) * 1000000; |
| 228 | #else |
| 229 | sec = us / 1000000; |
| 230 | nsec = (us % 1000000) * 1000; |
| 231 | #endif |
| 232 | |
| 233 | target.tv_sec = sec_ini; |
| 234 | tmp.tv_sec = sec; |
| 235 | target.tv_nsec = nsec_ini; |
| 236 | tmp.tv_nsec = nsec; |
| 237 | timespecadd(&target, &tmp, &target); |
| 238 | |
| 239 | if (__predict_false(sec != 0)) |
| 240 | printf("WARNING: over 1s delay\n" ); |
| 241 | |
| 242 | for (loops = 0; loops < 1000*1000*100; loops++) { |
| 243 | struct timespec cur; |
| 244 | |
| 245 | rumpuser_clock_gettime(RUMPUSER_CLOCK_ABSMONO, |
| 246 | &sec_now, &nsec_now); |
| 247 | cur.tv_sec = sec_now; |
| 248 | cur.tv_nsec = nsec_now; |
| 249 | if (timespeccmp(&cur, &target, >=)) { |
| 250 | return; |
| 251 | } |
| 252 | } |
| 253 | printf("WARNING: DELAY ESCAPED\n" ); |
| 254 | } |
| 255 | void (*delay_func)(unsigned int) = rump_delay; |
| 256 | __strong_alias(delay,rump_delay); |
| 257 | __strong_alias(_delay,rump_delay); |
| 258 | |
| 259 | /* Weak alias for getcwd_common to be used unless librumpvfs is present. */ |
| 260 | |
| 261 | int rump_getcwd_common(struct vnode *, struct vnode *, char **, char *, |
| 262 | int, int, struct lwp *); |
| 263 | int |
| 264 | rump_getcwd_common(struct vnode *lvp, struct vnode *rvp, char **bpp, char *bufp, |
| 265 | int limit, int flags, struct lwp *l) |
| 266 | { |
| 267 | |
| 268 | return ENOENT; |
| 269 | } |
| 270 | __weak_alias(getcwd_common,rump_getcwd_common); |
| 271 | |
| 272 | /* Weak aliases for fstrans to be used unless librumpvfs is present. */ |
| 273 | |
| 274 | void rump_fstrans_start(struct mount *); |
| 275 | void |
| 276 | rump_fstrans_start(struct mount *mp) |
| 277 | { |
| 278 | |
| 279 | } |
| 280 | __weak_alias(fstrans_start,rump_fstrans_start); |
| 281 | |
| 282 | int rump_fstrans_start_nowait(struct mount *); |
| 283 | int |
| 284 | rump_fstrans_start_nowait(struct mount *mp) |
| 285 | { |
| 286 | |
| 287 | return 0; |
| 288 | } |
| 289 | __weak_alias(fstrans_start_nowait,rump_fstrans_start_nowait); |
| 290 | |
| 291 | void rump_fstrans_start_lazy(struct mount *); |
| 292 | void |
| 293 | rump_fstrans_start_lazy(struct mount *mp) |
| 294 | { |
| 295 | |
| 296 | } |
| 297 | __weak_alias(fstrans_start_lazy,rump_fstrans_start_lazy); |
| 298 | |
| 299 | |
| 300 | void rump_fstrans_done(struct mount *); |
| 301 | void |
| 302 | rump_fstrans_done(struct mount *mp) |
| 303 | { |
| 304 | |
| 305 | } |
| 306 | __weak_alias(fstrans_done,rump_fstrans_done); |
| 307 | |
| 308 | |
| 309 | void rump_fstrans_lwp_dtor(struct lwp *); |
| 310 | void |
| 311 | rump_fstrans_lwp_dtor(struct lwp *l) |
| 312 | { |
| 313 | |
| 314 | } |
| 315 | __weak_alias(fstrans_lwp_dtor,rump_fstrans_lwp_dtor); |
| 316 | |
| 317 | /* |
| 318 | * Provide weak aliases for tty routines used by printf. |
| 319 | * They will be used unless the rumpkern_tty component is present. |
| 320 | */ |
| 321 | |
| 322 | int rump_ttycheckoutq(struct tty *, int); |
| 323 | int |
| 324 | rump_ttycheckoutq(struct tty *tp, int wait) |
| 325 | { |
| 326 | |
| 327 | return 1; |
| 328 | } |
| 329 | __weak_alias(ttycheckoutq,rump_ttycheckoutq); |
| 330 | |
| 331 | int rump_tputchar(int, int, struct tty *); |
| 332 | int |
| 333 | rump_tputchar(int c, int flags, struct tty *tp) |
| 334 | { |
| 335 | |
| 336 | cnputc(c); |
| 337 | return 0; |
| 338 | } |
| 339 | __weak_alias(tputchar,rump_tputchar); |
| 340 | |
| 341 | void |
| 342 | cnputc(int c) |
| 343 | { |
| 344 | |
| 345 | rumpuser_putchar(c); |
| 346 | } |
| 347 | |
| 348 | void |
| 349 | cnflush(void) |
| 350 | { |
| 351 | |
| 352 | /* done */ |
| 353 | } |
| 354 | |
| 355 | void |
| 356 | resettodr(void) |
| 357 | { |
| 358 | |
| 359 | /* setting clocks is not in the jurisdiction of rump kernels */ |
| 360 | } |
| 361 | |
| 362 | #ifdef __HAVE_SYSCALL_INTERN |
| 363 | void |
| 364 | syscall_intern(struct proc *p) |
| 365 | { |
| 366 | |
| 367 | p->p_emuldata = NULL; |
| 368 | } |
| 369 | #endif |
| 370 | |
| 371 | #ifdef LOCKDEBUG |
| 372 | void |
| 373 | turnstile_print(volatile void *obj, void (*pr)(const char *, ...)) |
| 374 | { |
| 375 | |
| 376 | /* nada */ |
| 377 | } |
| 378 | #endif |
| 379 | |
| 380 | void |
| 381 | cpu_reboot(int howto, char *bootstr) |
| 382 | { |
| 383 | int ruhow = 0; |
| 384 | void *finiarg; |
| 385 | |
| 386 | printf("rump kernel halting...\n" ); |
| 387 | |
| 388 | if (!RUMP_LOCALPROC_P(curproc)) |
| 389 | finiarg = RUMP_SPVM2CTL(curproc->p_vmspace); |
| 390 | else |
| 391 | finiarg = NULL; |
| 392 | |
| 393 | /* dump means we really take the dive here */ |
| 394 | if ((howto & RB_DUMP) || panicstr) { |
| 395 | ruhow = RUMPUSER_PANIC; |
| 396 | goto out; |
| 397 | } |
| 398 | |
| 399 | /* try to sync */ |
| 400 | if (!((howto & RB_NOSYNC) || panicstr)) { |
| 401 | rump_vfs_fini(); |
| 402 | } |
| 403 | |
| 404 | doshutdownhooks(); |
| 405 | |
| 406 | /* your wish is my command */ |
| 407 | if (howto & RB_HALT) { |
| 408 | printf("rump kernel halted (with RB_HALT, not exiting)\n" ); |
| 409 | rump_sysproxy_fini(finiarg); |
| 410 | for (;;) { |
| 411 | rumpuser_clock_sleep(RUMPUSER_CLOCK_RELWALL, 10, 0); |
| 412 | } |
| 413 | } |
| 414 | |
| 415 | /* this function is __dead, we must exit */ |
| 416 | out: |
| 417 | rump_sysproxy_fini(finiarg); |
| 418 | rumpuser_exit(ruhow); |
| 419 | } |
| 420 | |
| 421 | const char * |
| 422 | cpu_getmodel(void) |
| 423 | { |
| 424 | |
| 425 | return "rumpcore (virtual)" ; |
| 426 | } |
| 427 | |