usr/src/uts/common/sys/user.h

   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 2007 Sun Microsystems, Inc.  All rights reserved.
  23  * Use is subject to license terms.
  24  */
  25
  26 /*      Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
  27 /*        All Rights Reserved   */
  28 /*
  29  * Copyright (c) 2018, Joyent, Inc.
  30  */
  31
  32
  33 #ifndef _SYS_USER_H
  34 #define _SYS_USER_H
  35
  36 #include <sys/types.h>
  37 #include <sys/signal.h>
  38
  39 #ifdef  __cplusplus
  40 extern "C" {
  41 #endif
  42
  43 /*
  44  * struct exdata is visible in and out of the kernel. This is because it
  45  * is referenced in <sys/core.h> which doesn't have this kind of magic.
  46  */
  47 struct exdata {
  48         struct vnode    *vp;
  49         size_t  ux_tsize;       /* text size */
  50         size_t  ux_dsize;       /* data size */
  51         size_t  ux_bsize;       /* bss size */
  52         size_t  ux_lsize;       /* lib size */
  53         long    ux_nshlibs;     /* number of shared libs needed */
  54         short   ux_mach;        /* machine type */
  55         short   ux_mag;         /* magic number MUST be here */
  56         off_t   ux_toffset;     /* file offset to raw text */
  57         off_t   ux_doffset;     /* file offset to raw data */
  58         off_t   ux_loffset;     /* file offset to lib sctn */
  59         caddr_t ux_txtorg;      /* start addr of text in mem */
  60         caddr_t ux_datorg;      /* start addr of data in mem */
  61         caddr_t ux_entloc;      /* entry location */
  62 };
  63
  64 #ifdef  __cplusplus
  65 }
  66 #endif
  67
  68 #if defined(_KERNEL) || defined(_KMEMUSER)
  69
  70 #include <sys/param.h>
  71 #include <sys/pcb.h>
  72 #include <sys/siginfo.h>
  73 #include <sys/resource.h>
  74 #include <sys/time.h>
  75 #include <sys/auxv.h>
  76 #include <sys/errno.h>
  77 #include <sys/t_lock.h>
  78 #include <sys/refstr.h>
  79
  80 #ifdef  __cplusplus
  81 extern "C" {
  82 #endif
  83
  84 /*
  85  * Entry in the per-process list of open files.
  86  * Note: only certain fields are copied in flist_grow() and flist_fork().
  87  * This is indicated in brackets in the structure member comments.
  88  */
  89 typedef struct uf_entry {
  90         kmutex_t        uf_lock;        /* per-fd lock [never copied] */
  91         struct file     *uf_file;       /* file pointer [grow, fork] */
  92         struct fpollinfo *uf_fpollinfo; /* poll state [grow] */
  93         int             uf_refcnt;      /* LWPs accessing this file [grow] */
  94         int             uf_alloc;       /* right subtree allocs [grow, fork] */
  95         short           uf_flag;        /* fcntl F_GETFD flags [grow, fork] */
  96         short           uf_busy;        /* file is allocated [grow, fork] */
  97         kcondvar_t      uf_wanted_cv;   /* waiting for setf() [never copied] */
  98         kcondvar_t      uf_closing_cv;  /* waiting for close() [never copied] */
  99         struct portfd   *uf_portfd;     /* associated with port [grow] */
 100         /* Avoid false sharing - pad to coherency granularity (64 bytes) */
 101         char            uf_pad[64 - sizeof (kmutex_t) - 2 * sizeof (void*) -
 102                 2 * sizeof (int) - 2 * sizeof (short) -
 103                 2 * sizeof (kcondvar_t) - sizeof (struct portfd *)];
 104 } uf_entry_t;
 105
 106 /*
 107  * Retired file lists -- see flist_grow() for details.
 108  */
 109 typedef struct uf_rlist {
 110         struct uf_rlist *ur_next;
 111         uf_entry_t      *ur_list;
 112         int             ur_nfiles;
 113 } uf_rlist_t;
 114
 115 /*
 116  * Per-process file information.
 117  */
 118 typedef struct uf_info {
 119         kmutex_t        fi_lock;        /* see below */
 120         int             fi_badfd;       /* bad file descriptor # */
 121         int             fi_action;      /* action to take on bad fd use */
 122         int             fi_nfiles;      /* number of entries in fi_list[] */
 123         uf_entry_t *volatile fi_list;   /* current file list */
 124         uf_rlist_t      *fi_rlist;      /* retired file lists */
 125 } uf_info_t;
 126
 127 /*
 128  * File list locking.
 129  *
 130  * Each process has a list of open files, fi_list, indexed by fd.
 131  * fi_list is an array of uf_entry_t structures, each with its own lock.
 132  * One might think that the correct way to lock a file descriptor would be:
 133  *
 134  *      ufp = fip->fi_list[fd];
 135  *      mutex_enter(&ufp->uf_lock);
 136  *
 137  * However, that construct is only safe if fi_lock is already held.  If not,
 138  * fi_list can change in the window between loading ufp and entering uf_lock.
 139  * The UF_ENTER() macro deals with this possibility.  UF_ENTER(ufp, fip, fd)
 140  * locks fd and sets ufp to fd's uf_entry.  The locking rules are as follows:
 141  *
 142  * (1) fi_lock protects fi_list and fi_nfiles.  It also protects the
 143  *     uf_alloc and uf_busy fields of every fd's ufp; see fd_find() for
 144  *     details on file descriptor allocation.
 145  *
 146  * (2) UF_ENTER(ufp, fip, fd) locks descriptor fd and sets ufp to point
 147  *     to the uf_entry_t for fd.  UF_ENTER() protects all fields in ufp
 148  *     except uf_alloc and uf_busy.  UF_ENTER(ufp, fip, fd) also prevents
 149  *     ufp->uf_alloc, ufp->uf_busy, fip->fi_list and fip->fi_nfiles from
 150  *     changing.
 151  *
 152  * (3) The lock ordering is (1), (2).
 153  *
 154  * (4) Note that fip->fi_list and fip->fi_nfiles cannot change while *any*
 155  *     file list lock is held.  Thus flist_grow() must acquire all such
 156  *     locks -- fi_lock and every fd's uf_lock -- to install a new file list.
 157  */
 158 #define UF_ENTER(ufp, fip, fd)                                  \
 159         for (;;) {                                              \
 160                 uf_entry_t *_flist = (fip)->fi_list;            \
 161                 ufp = &_flist[fd];                              \
 162                 ASSERT((fd) < (fip)->fi_nfiles);                \
 163                 mutex_enter(&ufp->uf_lock);                     \
 164                 if (_flist == (fip)->fi_list)                   \
 165                         break;                                  \
 166                 mutex_exit(&ufp->uf_lock);                      \
 167         }
 168
 169 #define UF_EXIT(ufp)    mutex_exit(&ufp->uf_lock)
 170
 171 #define PSARGSZ         80      /* Space for exec arguments (used by ps(1)) */
 172 #define MAXCOMLEN       16      /* <= MAXNAMLEN, >= sizeof (ac_comm) */
 173
 174 typedef struct {                /* kernel syscall set type */
 175         uint_t  word[9];        /* space for syscall numbers [1..288] */
 176 } k_sysset_t;
 177
 178 /*
 179  * __KERN_NAUXV_IMPL is defined as a convenience sizing mechanism
 180  * for the portions of the kernel that care about aux vectors.
 181  *
 182  * Applications that need to know how many aux vectors the kernel
 183  * supplies should use the proc(4) interface to read /proc/PID/auxv.
 184  *
 185  * This value should not be changed in a patch.
 186  */
 187 #if defined(__sparc)
 188 #define __KERN_NAUXV_IMPL 20
 189 #elif defined(__i386) || defined(__amd64)
 190 #define __KERN_NAUXV_IMPL 25
 191 #endif
 192
 193 struct execsw;
 194
 195 /*
 196  * The user structure; one allocated per process.  Contains all the
 197  * per-process data that doesn't need to be referenced while the
 198  * process is swapped.
 199  */
 200 typedef struct  user {
 201         /*
 202          * These fields are initialized at process creation time and never
 203          * modified.  They can be accessed without acquiring locks.
 204          */
 205         struct execsw *u_execsw;        /* pointer to exec switch entry */
 206         auxv_t  u_auxv[__KERN_NAUXV_IMPL]; /* aux vector from exec */
 207         timestruc_t u_start;            /* hrestime at process start */
 208         clock_t u_ticks;                /* lbolt at process start */
 209         char    u_comm[MAXCOMLEN + 1];  /* executable file name from exec */
 210         char    u_psargs[PSARGSZ];      /* arguments from exec */
 211         int     u_argc;                 /* value of argc passed to main() */
 212         uintptr_t u_argv;               /* value of argv passed to main() */
 213         uintptr_t u_envp;               /* value of envp passed to main() */
 214         uintptr_t u_commpagep;          /* address of mapped comm page */
 215
 216         /*
 217          * These fields are protected by p_lock:
 218          */
 219         struct vnode *u_cdir;           /* current directory */
 220         struct vnode *u_rdir;           /* root directory */
 221         uint64_t u_mem;                 /* accumulated memory usage */
 222         size_t  u_mem_max;              /* peak RSS (K) */
 223         mode_t  u_cmask;                /* mask for file creation */
 224         char    u_acflag;               /* accounting flag */
 225         char    u_systrap;              /* /proc: any syscall mask bits set? */
 226         refstr_t *u_cwd;                /* cached string for cwd */
 227
 228         k_sysset_t u_entrymask;         /* /proc syscall stop-on-entry mask */
 229         k_sysset_t u_exitmask;          /* /proc syscall stop-on-exit mask */
 230         k_sigset_t u_signodefer;        /* signals defered when caught */
 231         k_sigset_t u_sigonstack;        /* signals taken on alternate stack */
 232         k_sigset_t u_sigresethand;      /* signals reset when caught */
 233         k_sigset_t u_sigrestart;        /* signals that restart system calls */
 234         k_sigset_t u_sigmask[MAXSIG];   /* signals held while in catcher */
 235         void    (*u_signal[MAXSIG])();  /* Disposition of signals */
 236
 237         /*
 238          * Resource controls provide the backend for process resource limits,
 239          * the interfaces for which are maintained for compatibility.  To
 240          * preserve the behaviour associated with the RLIM_SAVED_CUR and
 241          * RLIM_SAVED_MAX tokens, we retain the "saved" rlimits.
 242          */
 243         struct rlimit64 u_saved_rlimit[RLIM_NSAVED];
 244
 245         uf_info_t       u_finfo;        /* open file information */
 246 } user_t;
 247
 248 #include <sys/proc.h>                   /* cannot include before user defined */
 249
 250 #ifdef  _KERNEL
 251 #define P_FINFO(p)      (&(p)->p_user.u_finfo)
 252 #endif  /* _KERNEL */
 253
 254 #ifdef  __cplusplus
 255 }
 256 #endif
 257
 258 #else   /* defined(_KERNEL) || defined(_KMEMUSER) */
 259
 260 /*
 261  * Here, we define a fake version of struct user for programs
 262  * (debuggers) that use ptrace() to read and modify the saved
 263  * registers directly in the u-area.  ptrace() has been removed
 264  * from the operating system and now exists as a library function
 265  * in libc, built on the /proc process filesystem.  The ptrace()
 266  * library function provides access only to the members of the
 267  * fake struct user defined here.
 268  *
 269  * User-level programs that must know the real contents of struct
 270  * user will have to define _KMEMUSER before including <sys/user.h>.
 271  * Such programs also become machine specific. Carefully consider
 272  * the consequences of your actions.
 273  */
 274
 275 #include <sys/regset.h>
 276
 277 #ifdef  __cplusplus
 278 extern "C" {
 279 #endif
 280
 281 #define PSARGSZ         80      /* Space for exec arguments (used by ps(1)) */
 282
 283 typedef struct  user {
 284         gregset_t       u_reg;          /* user's saved registers */
 285         greg_t          *u_ar0;         /* address of user's saved R0 */
 286         char    u_psargs[PSARGSZ];      /* arguments from exec */
 287         void    (*u_signal[MAXSIG])();  /* Disposition of signals */
 288         int             u_code;         /* fault code on trap */
 289         caddr_t         u_addr;         /* fault PC on trap */
 290 } user_t;
 291
 292 #ifdef  __cplusplus
 293 }
 294 #endif
 295
 296 #endif  /* defined(_KERNEL) || defined(_KMEMUSER) */
 297
 298 #endif  /* _SYS_USER_H */