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 * Portions Copyright 2006-2008 John Birrell jb@freebsd.org 22 * 23 * $FreeBSD$ 24 * 25 */ 26 27 /* 28 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 29 * Use is subject to license terms. 30 */ 31 32 #include <sys/cdefs.h> 33 #include <sys/param.h> 34 #include <sys/systm.h> 35 #include <sys/conf.h> 36 #include <sys/cpuvar.h> 37 #include <sys/fcntl.h> 38 #include <sys/filio.h> 39 #include <sys/kdb.h> 40 #include <sys/kernel.h> 41 #include <sys/kmem.h> 42 #include <sys/kthread.h> 43 #include <sys/limits.h> 44 #include <sys/linker.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/module.h> 48 #include <sys/mutex.h> 49 #include <sys/poll.h> 50 #include <sys/proc.h> 51 #include <sys/selinfo.h> 52 #include <sys/smp.h> 53 #include <sys/sysproto.h> 54 #include <sys/sysent.h> 55 #include <sys/uio.h> 56 #include <sys/unistd.h> 57 #include <machine/stdarg.h> 58 59 #include <sys/dtrace.h> 60 61 #ifdef LINUX_SYSTRACE 62 #include <linux.h> 63 #include <linux_syscall.h> 64 #include <linux_proto.h> 65 #include <linux_syscallnames.c> 66 #include <linux_systrace.c> 67 extern struct sysent linux_sysent[]; 68 #define DEVNAME "dtrace/linsystrace" 69 #define PROVNAME "linsyscall" 70 #define MAXSYSCALL LINUX_SYS_MAXSYSCALL 71 #define SYSCALLNAMES linux_syscallnames 72 #define SYSENT linux_sysent 73 #else 74 /* 75 * The syscall arguments are processed into a DTrace argument array 76 * using a generated function. See sys/kern/makesyscalls.sh. 77 */ 78 #include <sys/syscall.h> 79 #include <kern/systrace_args.c> 80 #define DEVNAME "dtrace/systrace" 81 #define PROVNAME "syscall" 82 #define MAXSYSCALL SYS_MAXSYSCALL 83 #define SYSCALLNAMES syscallnames 84 #define SYSENT sysent 85 #endif 86 87 #define SYSTRACE_ARTIFICIAL_FRAMES 1 88 89 #define SYSTRACE_SHIFT 16 90 #define SYSTRACE_ISENTRY(x) ((int)(x) >> SYSTRACE_SHIFT) 91 #define SYSTRACE_SYSNUM(x) ((int)(x) & ((1 << SYSTRACE_SHIFT) - 1)) 92 #define SYSTRACE_ENTRY(id) ((1 << SYSTRACE_SHIFT) | (id)) 93 #define SYSTRACE_RETURN(id) (id) 94 95 #if ((1 << SYSTRACE_SHIFT) <= MAXSYSCALL) 96 #error 1 << SYSTRACE_SHIFT must exceed number of system calls 97 #endif 98 99 static d_open_t systrace_open; 100 static int systrace_unload(void); 101 static void systrace_getargdesc(void *, dtrace_id_t, void *, dtrace_argdesc_t *); 102 static void systrace_provide(void *, dtrace_probedesc_t *); 103 static void systrace_destroy(void *, dtrace_id_t, void *); 104 static void systrace_enable(void *, dtrace_id_t, void *); 105 static void systrace_disable(void *, dtrace_id_t, void *); 106 static void systrace_load(void *); 107 108 static struct cdevsw systrace_cdevsw = { 109 .d_version = D_VERSION, 110 .d_open = systrace_open, 111 #ifdef LINUX_SYSTRACE 112 .d_name = "linsystrace", 113 #else 114 .d_name = "systrace", 115 #endif 116 }; 117 118 static union { 119 const char **p_constnames; 120 char **pp_syscallnames; 121 } uglyhack = { SYSCALLNAMES }; 122 123 static dtrace_pattr_t systrace_attr = { 124 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON }, 125 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN }, 126 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA }, 127 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON }, 128 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA }, 129 }; 130 131 static dtrace_pops_t systrace_pops = { 132 systrace_provide, 133 NULL, 134 systrace_enable, 135 systrace_disable, 136 NULL, 137 NULL, 138 systrace_getargdesc, 139 NULL, 140 NULL, 141 systrace_destroy 142 }; 143 144 static struct cdev *systrace_cdev; 145 static dtrace_provider_id_t systrace_id; 146 147 #if !defined(LINUX_SYSTRACE) 148 /* 149 * Probe callback function. 150 * 151 * Note: This function is called for _all_ syscalls, regardless of which sysent 152 * array the syscall comes from. It could be a standard syscall or a 153 * compat syscall from something like Linux. 154 */ 155 static void 156 systrace_probe(u_int32_t id, int sysnum, struct sysent *sysent, void *params, 157 int ret) 158 { 159 int n_args = 0; 160 u_int64_t uargs[8]; 161 162 memset(uargs, 0, sizeof(uargs)); 163 /* 164 * Check if this syscall has an argument conversion function 165 * registered. 166 */ 167 if (params && sysent->sy_systrace_args_func != NULL) { 168 /* 169 * Convert the syscall parameters using the registered 170 * function. 171 */ 172 (*sysent->sy_systrace_args_func)(sysnum, params, uargs, &n_args); 173 } else if (params) { 174 /* 175 * Use the built-in system call argument conversion 176 * function to translate the syscall structure fields 177 * into the array of 64-bit values that DTrace 178 * expects. 179 */ 180 systrace_args(sysnum, params, uargs, &n_args); 181 } else { 182 /* 183 * Since params is NULL, this is a 'return' probe. 184 * Set arg0 and arg1 as the return value of this syscall. 185 */ 186 uargs[0] = uargs[1] = ret; 187 } 188 189 /* Process the probe using the converted argments. */ 190 dtrace_probe(id, uargs[0], uargs[1], uargs[2], uargs[3], uargs[4]); 191 } 192 #endif 193 194 static void 195 systrace_getargdesc(void *arg, dtrace_id_t id, void *parg, dtrace_argdesc_t *desc) 196 { 197 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 198 199 systrace_setargdesc(sysnum, desc->dtargd_ndx, desc->dtargd_native, 200 sizeof(desc->dtargd_native)); 201 202 if (desc->dtargd_native[0] == '\0') 203 desc->dtargd_ndx = DTRACE_ARGNONE; 204 205 return; 206 } 207 208 static void 209 systrace_provide(void *arg, dtrace_probedesc_t *desc) 210 { 211 int i; 212 213 if (desc != NULL) 214 return; 215 216 for (i = 0; i < MAXSYSCALL; i++) { 217 if (dtrace_probe_lookup(systrace_id, NULL, 218 uglyhack.pp_syscallnames[i], "entry") != 0) 219 continue; 220 221 (void) dtrace_probe_create(systrace_id, NULL, uglyhack.pp_syscallnames[i], 222 "entry", SYSTRACE_ARTIFICIAL_FRAMES, 223 (void *)((uintptr_t)SYSTRACE_ENTRY(i))); 224 (void) dtrace_probe_create(systrace_id, NULL, uglyhack.pp_syscallnames[i], 225 "return", SYSTRACE_ARTIFICIAL_FRAMES, 226 (void *)((uintptr_t)SYSTRACE_RETURN(i))); 227 } 228 } 229 230 static void 231 systrace_destroy(void *arg, dtrace_id_t id, void *parg) 232 { 233 #ifdef DEBUG 234 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 235 236 /* 237 * There's nothing to do here but assert that we have actually been 238 * disabled. 239 */ 240 if (SYSTRACE_ISENTRY((uintptr_t)parg)) { 241 ASSERT(sysent[sysnum].sy_entry == 0); 242 } else { 243 ASSERT(sysent[sysnum].sy_return == 0); 244 } 245 #endif 246 } 247 248 static void 249 systrace_enable(void *arg, dtrace_id_t id, void *parg) 250 { 251 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 252 253 if (SYSENT[sysnum].sy_systrace_args_func == NULL) 254 SYSENT[sysnum].sy_systrace_args_func = systrace_args; 255 256 if (SYSTRACE_ISENTRY((uintptr_t)parg)) 257 SYSENT[sysnum].sy_entry = id; 258 else 259 SYSENT[sysnum].sy_return = id; 260 } 261 262 static void 263 systrace_disable(void *arg, dtrace_id_t id, void *parg) 264 { 265 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 266 267 SYSENT[sysnum].sy_entry = 0; 268 SYSENT[sysnum].sy_return = 0; 269 } 270 271 static void 272 systrace_load(void *dummy) 273 { 274 /* Create the /dev/dtrace/systrace entry. */ 275 systrace_cdev = make_dev(&systrace_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, 276 DEVNAME); 277 278 if (dtrace_register(PROVNAME, &systrace_attr, DTRACE_PRIV_USER, 279 NULL, &systrace_pops, NULL, &systrace_id) != 0) 280 return; 281 282 #if !defined(LINUX_SYSTRACE) 283 systrace_probe_func = systrace_probe; 284 #endif 285 } 286 287 288 static int 289 systrace_unload() 290 { 291 int error = 0; 292 293 if ((error = dtrace_unregister(systrace_id)) != 0) 294 return (error); 295 296 #if !defined(LINUX_SYSTRACE) 297 systrace_probe_func = NULL; 298 #endif 299 300 destroy_dev(systrace_cdev); 301 302 return (error); 303 } 304 305 static int 306 systrace_modevent(module_t mod __unused, int type, void *data __unused) 307 { 308 int error = 0; 309 310 switch (type) { 311 case MOD_LOAD: 312 break; 313 314 case MOD_UNLOAD: 315 break; 316 317 case MOD_SHUTDOWN: 318 break; 319 320 default: 321 error = EOPNOTSUPP; 322 break; 323 324 } 325 return (error); 326 } 327 328 static int 329 systrace_open(struct cdev *dev __unused, int oflags __unused, int devtype __unused, struct thread *td __unused) 330 { 331 return (0); 332 } 333 334 SYSINIT(systrace_load, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, systrace_load, NULL); 335 SYSUNINIT(systrace_unload, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, systrace_unload, NULL); 336 337 #ifdef LINUX_SYSTRACE 338 DEV_MODULE(linsystrace, systrace_modevent, NULL); 339 MODULE_VERSION(linsystrace, 1); 340 MODULE_DEPEND(linsystrace, linux, 1, 1, 1); 341 MODULE_DEPEND(linsystrace, systrace, 1, 1, 1); 342 MODULE_DEPEND(linsystrace, dtrace, 1, 1, 1); 343 MODULE_DEPEND(linsystrace, opensolaris, 1, 1, 1); 344 #else 345 DEV_MODULE(systrace, systrace_modevent, NULL); 346 MODULE_VERSION(systrace, 1); 347 MODULE_DEPEND(systrace, dtrace, 1, 1, 1); 348 MODULE_DEPEND(systrace, opensolaris, 1, 1, 1); 349 #endif 350