1 /* $NetBSD: systrace.c,v 1.2 2010/02/21 01:46:34 darran Exp $ */ 2 3 /* 4 * CDDL HEADER START 5 * 6 * The contents of this file are subject to the terms of the 7 * Common Development and Distribution License (the "License"). 8 * You may not use this file except in compliance with the License. 9 * 10 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 11 * or http://www.opensolaris.org/os/licensing. 12 * See the License for the specific language governing permissions 13 * and limitations under the License. 14 * 15 * When distributing Covered Code, include this CDDL HEADER in each 16 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 17 * If applicable, add the following below this CDDL HEADER, with the 18 * fields enclosed by brackets "[]" replaced with your own identifying 19 * information: Portions Copyright [yyyy] [name of copyright owner] 20 * 21 * CDDL HEADER END 22 * 23 * Portions Copyright 2006-2008 John Birrell jb@freebsd.org 24 * 25 * $FreeBSD: src/sys/cddl/dev/systrace/systrace.c,v 1.2.2.1 2009/08/03 08:13:06 kensmith Exp $ 26 * 27 */ 28 29 /* 30 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 31 * Use is subject to license terms. 32 */ 33 34 #include <sys/cdefs.h> 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/conf.h> 38 #include <sys/cpuvar.h> 39 #include <sys/fcntl.h> 40 #include <sys/filio.h> 41 #include <sys/kdb.h> 42 #include <sys/kernel.h> 43 #include <sys/kmem.h> 44 #include <sys/kthread.h> 45 #include <sys/limits.h> 46 #include <sys/linker.h> 47 #include <sys/lock.h> 48 #include <sys/malloc.h> 49 #include <sys/module.h> 50 #include <sys/mutex.h> 51 #include <sys/poll.h> 52 #include <sys/proc.h> 53 #include <sys/selinfo.h> 54 #include <sys/smp.h> 55 #include <sys/sysproto.h> 56 #include <sys/sysent.h> 57 #include <sys/uio.h> 58 #include <sys/unistd.h> 59 #include <machine/stdarg.h> 60 61 #include <sys/dtrace.h> 62 63 #ifdef LINUX_SYSTRACE 64 #include <linux.h> 65 #include <linux_syscall.h> 66 #include <linux_proto.h> 67 #include <linux_syscallnames.c> 68 #include <linux_systrace.c> 69 extern struct sysent linux_sysent[]; 70 #define DEVNAME "dtrace/linsystrace" 71 #define PROVNAME "linsyscall" 72 #define MAXSYSCALL LINUX_SYS_MAXSYSCALL 73 #define SYSCALLNAMES linux_syscallnames 74 #define SYSENT linux_sysent 75 #else 76 /* 77 * The syscall arguments are processed into a DTrace argument array 78 * using a generated function. See sys/kern/makesyscalls.sh. 79 */ 80 #include <sys/syscall.h> 81 #include <kern/systrace_args.c> 82 extern const char *syscallnames[]; 83 #define DEVNAME "dtrace/systrace" 84 #define PROVNAME "syscall" 85 #define MAXSYSCALL SYS_MAXSYSCALL 86 #define SYSCALLNAMES syscallnames 87 #define SYSENT sysent 88 #endif 89 90 #define SYSTRACE_ARTIFICIAL_FRAMES 1 91 92 #define SYSTRACE_SHIFT 16 93 #define SYSTRACE_ISENTRY(x) ((int)(x) >> SYSTRACE_SHIFT) 94 #define SYSTRACE_SYSNUM(x) ((int)(x) & ((1 << SYSTRACE_SHIFT) - 1)) 95 #define SYSTRACE_ENTRY(id) ((1 << SYSTRACE_SHIFT) | (id)) 96 #define SYSTRACE_RETURN(id) (id) 97 98 #if ((1 << SYSTRACE_SHIFT) <= MAXSYSCALL) 99 #error 1 << SYSTRACE_SHIFT must exceed number of system calls 100 #endif 101 102 static d_open_t systrace_open; 103 static int systrace_unload(void); 104 static void systrace_getargdesc(void *, dtrace_id_t, void *, dtrace_argdesc_t *); 105 static void systrace_provide(void *, dtrace_probedesc_t *); 106 static void systrace_destroy(void *, dtrace_id_t, void *); 107 static void systrace_enable(void *, dtrace_id_t, void *); 108 static void systrace_disable(void *, dtrace_id_t, void *); 109 static void systrace_load(void *); 110 111 static struct cdevsw systrace_cdevsw = { 112 .d_version = D_VERSION, 113 .d_open = systrace_open, 114 #ifdef LINUX_SYSTRACE 115 .d_name = "linsystrace", 116 #else 117 .d_name = "systrace", 118 #endif 119 }; 120 121 static union { 122 const char **p_constnames; 123 char **pp_syscallnames; 124 } uglyhack = { SYSCALLNAMES }; 125 126 static dtrace_pattr_t systrace_attr = { 127 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON }, 128 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN }, 129 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA }, 130 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON }, 131 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA }, 132 }; 133 134 static dtrace_pops_t systrace_pops = { 135 systrace_provide, 136 NULL, 137 systrace_enable, 138 systrace_disable, 139 NULL, 140 NULL, 141 systrace_getargdesc, 142 NULL, 143 NULL, 144 systrace_destroy 145 }; 146 147 static struct cdev *systrace_cdev; 148 static dtrace_provider_id_t systrace_id; 149 150 #if !defined(LINUX_SYSTRACE) 151 /* 152 * Probe callback function. 153 * 154 * Note: This function is called for _all_ syscalls, regardless of which sysent 155 * array the syscall comes from. It could be a standard syscall or a 156 * compat syscall from something like Linux. 157 */ 158 static void 159 systrace_probe(u_int32_t id, int sysnum, struct sysent *sysent, void *params) 160 { 161 int n_args = 0; 162 u_int64_t uargs[8]; 163 164 /* 165 * Check if this syscall has an argument conversion function 166 * registered. 167 */ 168 if (sysent->sy_systrace_args_func != NULL) 169 /* 170 * Convert the syscall parameters using the registered 171 * function. 172 */ 173 (*sysent->sy_systrace_args_func)(sysnum, params, uargs, &n_args); 174 else 175 /* 176 * Use the built-in system call argument conversion 177 * function to translate the syscall structure fields 178 * into the array of 64-bit values that DTrace 179 * expects. 180 */ 181 systrace_args(sysnum, params, uargs, &n_args); 182 183 /* Process the probe using the converted argments. */ 184 dtrace_probe(id, uargs[0], uargs[1], uargs[2], uargs[3], uargs[4]); 185 } 186 #endif 187 188 static void 189 systrace_getargdesc(void *arg, dtrace_id_t id, void *parg, dtrace_argdesc_t *desc) 190 { 191 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 192 193 systrace_setargdesc(sysnum, desc->dtargd_ndx, desc->dtargd_native, 194 sizeof(desc->dtargd_native)); 195 196 if (desc->dtargd_native[0] == '\0') 197 desc->dtargd_ndx = DTRACE_ARGNONE; 198 199 return; 200 } 201 202 static void 203 systrace_provide(void *arg, dtrace_probedesc_t *desc) 204 { 205 int i; 206 207 if (desc != NULL) 208 return; 209 210 for (i = 0; i < MAXSYSCALL; i++) { 211 if (dtrace_probe_lookup(systrace_id, NULL, 212 uglyhack.pp_syscallnames[i], "entry") != 0) 213 continue; 214 215 (void) dtrace_probe_create(systrace_id, NULL, uglyhack.pp_syscallnames[i], 216 "entry", SYSTRACE_ARTIFICIAL_FRAMES, 217 (void *)((uintptr_t)SYSTRACE_ENTRY(i))); 218 (void) dtrace_probe_create(systrace_id, NULL, uglyhack.pp_syscallnames[i], 219 "return", SYSTRACE_ARTIFICIAL_FRAMES, 220 (void *)((uintptr_t)SYSTRACE_RETURN(i))); 221 } 222 } 223 224 static void 225 systrace_destroy(void *arg, dtrace_id_t id, void *parg) 226 { 227 #ifdef DEBUG 228 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 229 230 /* 231 * There's nothing to do here but assert that we have actually been 232 * disabled. 233 */ 234 if (SYSTRACE_ISENTRY((uintptr_t)parg)) { 235 ASSERT(sysent[sysnum].sy_entry == 0); 236 } else { 237 ASSERT(sysent[sysnum].sy_return == 0); 238 } 239 #endif 240 } 241 242 static void 243 systrace_enable(void *arg, dtrace_id_t id, void *parg) 244 { 245 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 246 247 if (SYSENT[sysnum].sy_systrace_args_func == NULL) 248 SYSENT[sysnum].sy_systrace_args_func = systrace_args; 249 250 if (SYSTRACE_ISENTRY((uintptr_t)parg)) 251 SYSENT[sysnum].sy_entry = id; 252 else 253 SYSENT[sysnum].sy_return = id; 254 } 255 256 static void 257 systrace_disable(void *arg, dtrace_id_t id, void *parg) 258 { 259 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg); 260 261 SYSENT[sysnum].sy_entry = 0; 262 SYSENT[sysnum].sy_return = 0; 263 } 264 265 static void 266 systrace_load(void *dummy) 267 { 268 /* Create the /dev/dtrace/systrace entry. */ 269 systrace_cdev = make_dev(&systrace_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, 270 DEVNAME); 271 272 if (dtrace_register(PROVNAME, &systrace_attr, DTRACE_PRIV_USER, 273 NULL, &systrace_pops, NULL, &systrace_id) != 0) 274 return; 275 276 #if !defined(LINUX_SYSTRACE) 277 systrace_probe_func = systrace_probe; 278 #endif 279 } 280 281 282 static int 283 systrace_unload() 284 { 285 int error = 0; 286 287 if ((error = dtrace_unregister(systrace_id)) != 0) 288 return (error); 289 290 #if !defined(LINUX_SYSTRACE) 291 systrace_probe_func = NULL; 292 #endif 293 294 destroy_dev(systrace_cdev); 295 296 return (error); 297 } 298 299 static int 300 systrace_modevent(module_t mod __unused, int type, void *data __unused) 301 { 302 int error = 0; 303 304 switch (type) { 305 case MOD_LOAD: 306 break; 307 308 case MOD_UNLOAD: 309 break; 310 311 case MOD_SHUTDOWN: 312 break; 313 314 default: 315 error = EOPNOTSUPP; 316 break; 317 318 } 319 return (error); 320 } 321 322 static int 323 systrace_open(struct cdev *dev __unused, int oflags __unused, int devtype __unused, struct thread *td __unused) 324 { 325 return (0); 326 } 327 328 SYSINIT(systrace_load, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, systrace_load, NULL); 329 SYSUNINIT(systrace_unload, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, systrace_unload, NULL); 330 331 #ifdef LINUX_SYSTRACE 332 DEV_MODULE(linsystrace, systrace_modevent, NULL); 333 MODULE_VERSION(linsystrace, 1); 334 MODULE_DEPEND(linsystrace, linux, 1, 1, 1); 335 MODULE_DEPEND(linsystrace, systrace, 1, 1, 1); 336 MODULE_DEPEND(linsystrace, dtrace, 1, 1, 1); 337 MODULE_DEPEND(linsystrace, opensolaris, 1, 1, 1); 338 #else 339 DEV_MODULE(systrace, systrace_modevent, NULL); 340 MODULE_VERSION(systrace, 1); 341 MODULE_DEPEND(systrace, dtrace, 1, 1, 1); 342 MODULE_DEPEND(systrace, opensolaris, 1, 1, 1); 343 #endif 344