linux32_machdep.c revision 176193
11195Srgrimes/*- 22619Srgrimes * Copyright (c) 2004 Tim J. Robbins 337Srgrimes * Copyright (c) 2002 Doug Rabson 437Srgrimes * Copyright (c) 2000 Marcel Moolenaar 537Srgrimes * All rights reserved. 6147Srgrimes * 7147Srgrimes * Redistribution and use in source and binary forms, with or without 8147Srgrimes * modification, are permitted provided that the following conditions 9207Snate * are met: 101734Sjkh * 1. Redistributions of source code must retain the above copyright 111518Sguido * notice, this list of conditions and the following disclaimer 121518Sguido * in this position and unchanged. 131734Sjkh * 2. Redistributions in binary form must reproduce the above copyright 141734Sjkh * notice, this list of conditions and the following disclaimer in the 151734Sjkh * documentation and/or other materials provided with the distribution. 1637Srgrimes * 3. The name of the author may not be used to endorse or promote products 1737Srgrimes * derived from this software without specific prior written permission. 1837Srgrimes * 1937Srgrimes * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 201773Sjkh * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21147Srgrimes * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22147Srgrimes * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 232570Srgrimes * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 242570Srgrimes * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 252570Srgrimes * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 261767Sjkh * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 272570Srgrimes * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 282570Srgrimes * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 292570Srgrimes */ 302570Srgrimes 312570Srgrimes#include <sys/cdefs.h> 322570Srgrimes__FBSDID("$FreeBSD: head/sys/amd64/linux32/linux32_machdep.c 176193 2008-02-11 19:35:03Z jkim $"); 332570Srgrimes 342570Srgrimes#include <sys/param.h> 351782Sjkh#include <sys/kernel.h> 361782Sjkh#include <sys/systm.h> 37491Srgrimes#include <sys/file.h> 3837Srgrimes#include <sys/fcntl.h> 3937Srgrimes#include <sys/clock.h> 4037Srgrimes#include <sys/imgact.h> 4137Srgrimes#include <sys/limits.h> 4237Srgrimes#include <sys/lock.h> 43263Srgrimes#include <sys/malloc.h> 441130Srgrimes#include <sys/mman.h> 451767Sjkh#include <sys/mutex.h> 461126Srgrimes#include <sys/priv.h> 47993Srgrimes#include <sys/proc.h> 48277Srgrimes#include <sys/resource.h> 49277Srgrimes#include <sys/resourcevar.h> 50284Srgrimes#include <sys/sched.h> 51463Srgrimes#include <sys/syscallsubr.h> 521205Srgrimes#include <sys/sysproto.h> 53284Srgrimes#include <sys/unistd.h> 54284Srgrimes 55284Srgrimes#include <machine/frame.h> 56284Srgrimes#include <machine/pcb.h> 57284Srgrimes#include <machine/psl.h> 581285Srgrimes#include <machine/segments.h> 59284Srgrimes#include <machine/specialreg.h> 601767Sjkh 611285Srgrimes#include <vm/vm.h> 622499Sgpalmer#include <vm/pmap.h> 631285Srgrimes#include <vm/vm_extern.h> 64284Srgrimes#include <vm/vm_kern.h> 651371Srgrimes#include <vm/vm_map.h> 661371Srgrimes 671194Srgrimes#include <amd64/linux32/linux.h> 68358Srgrimes#include <amd64/linux32/linux32_proto.h> 69358Srgrimes#include <compat/linux/linux_ipc.h> 70412Salm#include <compat/linux/linux_signal.h> 71358Srgrimes#include <compat/linux/linux_util.h> 72452Srgrimes#include <compat/linux/linux_emul.h> 73358Srgrimes 741194Srgrimesstruct l_old_select_argv { 751767Sjkh l_int nfds; 761194Srgrimes l_uintptr_t readfds; 771194Srgrimes l_uintptr_t writefds; 781194Srgrimes l_uintptr_t exceptfds; 791194Srgrimes l_uintptr_t timeout; 801194Srgrimes} __packed; 811194Srgrimes 821194Srgrimesint 831194Srgrimeslinux_to_bsd_sigaltstack(int lsa) 841194Srgrimes{ 851194Srgrimes int bsa = 0; 861194Srgrimes 871194Srgrimes if (lsa & LINUX_SS_DISABLE) 881194Srgrimes bsa |= SS_DISABLE; 891194Srgrimes if (lsa & LINUX_SS_ONSTACK) 901194Srgrimes bsa |= SS_ONSTACK; 911194Srgrimes return (bsa); 921194Srgrimes} 931194Srgrimes 941243Srgrimesint 95263Srgrimesbsd_to_linux_sigaltstack(int bsa) 96358Srgrimes{ 971194Srgrimes int lsa = 0; 981194Srgrimes 991194Srgrimes if (bsa & SS_DISABLE) 1001773Sjkh lsa |= LINUX_SS_DISABLE; 1012499Sgpalmer if (bsa & SS_ONSTACK) 1021194Srgrimes lsa |= LINUX_SS_ONSTACK; 103452Srgrimes return (lsa); 1041194Srgrimes} 1051194Srgrimes 106358Srgrimes/* 1071194Srgrimes * Custom version of exec_copyin_args() so that we can translate 1081695Scsgr * the pointers. 1091695Scsgr */ 1101695Scsgrstatic int 1111695Scsgrlinux_exec_copyin_args(struct image_args *args, char *fname, 1121695Scsgr enum uio_seg segflg, char **argv, char **envv) 1131695Scsgr{ 1141695Scsgr char *argp, *envp; 1151243Srgrimes u_int32_t *p32, arg; 1161194Srgrimes size_t length; 1171194Srgrimes int error; 1181243Srgrimes 1191243Srgrimes bzero(args, sizeof(*args)); 120284Srgrimes if (argv == NULL) 1212570Srgrimes return (EFAULT); 1222570Srgrimes 123372Srgrimes /* 124372Srgrimes * Allocate temporary demand zeroed space for argument and 125372Srgrimes * environment strings 1262570Srgrimes */ 1272570Srgrimes args->buf = (char *)kmem_alloc_wait(exec_map, 128372Srgrimes PATH_MAX + ARG_MAX + MAXSHELLCMDLEN); 1291126Srgrimes if (args->buf == NULL) 130347Srgrimes return (ENOMEM); 1311767Sjkh args->begin_argv = args->buf; 13237Srgrimes args->endp = args->begin_argv; 1332538Spst args->stringspace = ARG_MAX; 134347Srgrimes 1352538Spst args->fname = args->buf + ARG_MAX; 136355Srgrimes 137372Srgrimes /* 138347Srgrimes * Copy the file name. 139355Srgrimes */ 140347Srgrimes error = (segflg == UIO_SYSSPACE) ? 1412538Spst copystr(fname, args->fname, PATH_MAX, &length) : 1422538Spst copyinstr(fname, args->fname, PATH_MAX, &length); 1432538Spst if (error != 0) 1442538Spst goto err_exit; 1452538Spst 146372Srgrimes /* 147347Srgrimes * extract arguments first 148355Srgrimes */ 149347Srgrimes p32 = (u_int32_t *)argv; 150347Srgrimes for (;;) { 1512538Spst error = copyin(p32++, &arg, sizeof(arg)); 152147Srgrimes if (error) 1531759Sjkh goto err_exit; 1541759Sjkh if (arg == 0) 1551759Sjkh break; 1561759Sjkh argp = PTRIN(arg); 1571759Sjkh error = copyinstr(argp, args->endp, args->stringspace, &length); 1581731Sjkh if (error) { 1591759Sjkh if (error == ENAMETOOLONG) 1601731Sjkh error = E2BIG; 1611759Sjkh 1621759Sjkh goto err_exit; 16337Srgrimes } 1641759Sjkh args->stringspace -= length; 165347Srgrimes args->endp += length; 1661759Sjkh args->argc++; 167347Srgrimes } 1681731Sjkh 1691731Sjkh args->begin_envv = args->endp; 17037Srgrimes 1711731Sjkh /* 17237Srgrimes * extract environment strings 1731731Sjkh */ 17437Srgrimes if (envv) { 1751731Sjkh p32 = (u_int32_t *)envv; 17637Srgrimes for (;;) { 17737Srgrimes error = copyin(p32++, &arg, sizeof(arg)); 17837Srgrimes if (error) 17937Srgrimes goto err_exit; 1801731Sjkh if (arg == 0) 1811731Sjkh break; 1821731Sjkh envp = PTRIN(arg); 1831731Sjkh error = copyinstr(envp, args->endp, args->stringspace, 18437Srgrimes &length); 18537Srgrimes if (error) { 186147Srgrimes if (error == ENAMETOOLONG) 187147Srgrimes error = E2BIG; 18837Srgrimes goto err_exit; 189147Srgrimes } 19037Srgrimes args->stringspace -= length; 19137Srgrimes args->endp += length; 19237Srgrimes args->envc++; 193288Srgrimes } 194288Srgrimes } 195147Srgrimes 19637Srgrimes return (0); 197147Srgrimes 198147Srgrimeserr_exit: 19937Srgrimes kmem_free_wakeup(exec_map, (vm_offset_t)args->buf, 2001759Sjkh PATH_MAX + ARG_MAX + MAXSHELLCMDLEN); 2011759Sjkh args->buf = NULL; 2021759Sjkh return (error); 2031759Sjkh} 204347Srgrimes 2052538Spstint 2062538Spstlinux_execve(struct thread *td, struct linux_execve_args *args) 207347Srgrimes{ 2082538Spst struct image_args eargs; 2091775Sjkh char *path; 210347Srgrimes int error; 2111759Sjkh 212355Srgrimes LCONVPATHEXIST(td, args->path, &path); 213277Srgrimes 2141126Srgrimes#ifdef DEBUG 2151126Srgrimes if (ldebug(execve)) 2161731Sjkh printf(ARGS(execve, "%s"), path); 217238Sroot#endif 2181759Sjkh 2191731Sjkh error = linux_exec_copyin_args(&eargs, path, UIO_SYSSPACE, args->argp, 2201759Sjkh args->envp); 221333Srgrimes free(path, M_TEMP); 2221759Sjkh if (error == 0) 2231759Sjkh error = kern_execve(td, &eargs, NULL); 224168Srgrimes if (error == 0) 225333Srgrimes /* Linux process can execute FreeBSD one, do not attempt 2261759Sjkh * to create emuldata for such process using 2271759Sjkh * linux_proc_init, this leads to a panic on KASSERT 228333Srgrimes * because such process has p->p_emuldata == NULL. 22937Srgrimes */ 2301731Sjkh if (td->td_proc->p_sysent == &elf_linux_sysvec) 2312619Srgrimes error = linux_proc_init(td, 0, 0); 2321782Sjkh return (error); 2332619Srgrimes} 2341782Sjkh 2351731Sjkhstruct iovec32 { 2361731Sjkh u_int32_t iov_base; 2371731Sjkh int iov_len; 2382570Srgrimes}; 2392570Srgrimes 2401731SjkhCTASSERT(sizeof(struct iovec32) == 8); 2412570Srgrimes 2422570Srgrimesstatic int 2431731Sjkhlinux32_copyinuio(struct iovec32 *iovp, u_int iovcnt, struct uio **uiop) 2442570Srgrimes{ 2452570Srgrimes struct iovec32 iov32; 2461731Sjkh struct iovec *iov; 24737Srgrimes struct uio *uio; 2481759Sjkh u_int iovlen; 2491759Sjkh int error, i; 25037Srgrimes 2511759Sjkh *uiop = NULL; 2521759Sjkh if (iovcnt > UIO_MAXIOV) 2531759Sjkh return (EINVAL); 2541759Sjkh iovlen = iovcnt * sizeof(struct iovec); 2551731Sjkh uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK); 256320Srgrimes iov = (struct iovec *)(uio + 1); 2571759Sjkh for (i = 0; i < iovcnt; i++) { 2581731Sjkh error = copyin(&iovp[i], &iov32, sizeof(struct iovec32)); 259320Srgrimes if (error) { 260320Srgrimes free(uio, M_IOV); 261358Srgrimes return (error); 262568Srgrimes } 2631672Sjkh iov[i].iov_base = PTRIN(iov32.iov_base); 2641762Sjkh iov[i].iov_len = iov32.iov_len; 2651027Sache } 2661027Sache uio->uio_iov = iov; 2671731Sjkh uio->uio_iovcnt = iovcnt; 268333Srgrimes uio->uio_segflg = UIO_USERSPACE; 269284Srgrimes uio->uio_offset = -1; 270320Srgrimes uio->uio_resid = 0; 2712570Srgrimes for (i = 0; i < iovcnt; i++) { 272284Srgrimes if (iov->iov_len > INT_MAX - uio->uio_resid) { 273320Srgrimes free(uio, M_IOV); 2741731Sjkh return (EINVAL); 2751731Sjkh } 2761731Sjkh uio->uio_resid += iov->iov_len; 2771762Sjkh iov++; 2781194Srgrimes } 2791194Srgrimes *uiop = uio; 2801194Srgrimes return (0); 2811194Srgrimes} 282320Srgrimes 2831205Srgrimesint 2842570Srgrimeslinux_readv(struct thread *td, struct linux_readv_args *uap) 2851759Sjkh{ 2861731Sjkh struct uio *auio; 287277Srgrimes int error; 2881027Sache 2891027Sache error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); 2901205Srgrimes if (error) 291358Srgrimes return (error); 2921205Srgrimes error = kern_readv(td, uap->fd, auio); 2931782Sjkh free(auio, M_IOV); 294277Srgrimes return (error); 2951205Srgrimes} 2962570Srgrimes 2971759Sjkhint 2981731Sjkhlinux_writev(struct thread *td, struct linux_writev_args *uap) 299320Srgrimes{ 3001027Sache struct uio *auio; 3011027Sache int error; 3021205Srgrimes 303358Srgrimes error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); 3041205Srgrimes if (error) 3051782Sjkh return (error); 306320Srgrimes error = kern_writev(td, uap->fd, auio); 3071285Srgrimes free(auio, M_IOV); 3082570Srgrimes return (error); 3091371Srgrimes} 3101371Srgrimes 3111371Srgrimesstruct l_ipc_kludge { 3121371Srgrimes l_uintptr_t msgp; 3131371Srgrimes l_long msgtyp; 3141285Srgrimes} __packed; 3151731Sjkh 3161285Srgrimesint 3171731Sjkhlinux_ipc(struct thread *td, struct linux_ipc_args *args) 3181731Sjkh{ 3191731Sjkh 3201731Sjkh switch (args->what & 0xFFFF) { 3211285Srgrimes case LINUX_SEMOP: { 3221285Srgrimes struct linux_semop_args a; 3232570Srgrimes 3241759Sjkh a.semid = args->arg1; 3251731Sjkh a.tsops = args->ptr; 3261285Srgrimes a.nsops = args->arg2; 3271285Srgrimes return (linux_semop(td, &a)); 3281285Srgrimes } 3291285Srgrimes case LINUX_SEMGET: { 3301285Srgrimes struct linux_semget_args a; 3311285Srgrimes 3321782Sjkh a.key = args->arg1; 3331285Srgrimes a.nsems = args->arg2; 3341285Srgrimes a.semflg = args->arg3; 3352570Srgrimes return (linux_semget(td, &a)); 3361759Sjkh } 3371731Sjkh case LINUX_SEMCTL: { 3381285Srgrimes struct linux_semctl_args a; 3391285Srgrimes int error; 3401285Srgrimes 3411285Srgrimes a.semid = args->arg1; 3421285Srgrimes a.semnum = args->arg2; 3431285Srgrimes a.cmd = args->arg3; 3441782Sjkh error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 3451285Srgrimes if (error) 3461205Srgrimes return (error); 347568Srgrimes return (linux_semctl(td, &a)); 3481672Sjkh } 349568Srgrimes case LINUX_MSGSND: { 3501027Sache struct linux_msgsnd_args a; 3511027Sache 3521731Sjkh a.msqid = args->arg1; 353333Srgrimes a.msgp = args->ptr; 354284Srgrimes a.msgsz = args->arg2; 355358Srgrimes a.msgflg = args->arg3; 3562570Srgrimes return (linux_msgsnd(td, &a)); 357284Srgrimes } 3581194Srgrimes case LINUX_MSGRCV: { 3591243Srgrimes struct linux_msgrcv_args a; 3601194Srgrimes 3611194Srgrimes a.msqid = args->arg1; 3621731Sjkh a.msgsz = args->arg2; 3631731Sjkh a.msgflg = args->arg3; 3641731Sjkh if ((args->what >> 16) == 0) { 3651731Sjkh struct l_ipc_kludge tmp; 3661731Sjkh int error; 3671762Sjkh 368284Srgrimes if (args->ptr == 0) 3691194Srgrimes return (EINVAL); 3701194Srgrimes error = copyin(args->ptr, &tmp, sizeof(tmp)); 3711194Srgrimes if (error) 3721194Srgrimes return (error); 373358Srgrimes a.msgp = PTRIN(tmp.msgp); 3741194Srgrimes a.msgtyp = tmp.msgtyp; 375358Srgrimes } else { 376358Srgrimes a.msgp = args->ptr; 3771243Srgrimes a.msgtyp = args->arg5; 378333Srgrimes } 379284Srgrimes return (linux_msgrcv(td, &a)); 3801027Sache } 3811027Sache case LINUX_MSGGET: { 3821205Srgrimes struct linux_msgget_args a; 383358Srgrimes 3841205Srgrimes a.key = args->arg1; 3851782Sjkh a.msgflg = args->arg2; 386284Srgrimes return (linux_msgget(td, &a)); 3871205Srgrimes } 388568Srgrimes case LINUX_MSGCTL: { 3891672Sjkh struct linux_msgctl_args a; 390568Srgrimes 3911027Sache a.msqid = args->arg1; 3921027Sache a.cmd = args->arg2; 3931731Sjkh a.buf = args->ptr; 394333Srgrimes return (linux_msgctl(td, &a)); 395284Srgrimes } 396358Srgrimes case LINUX_SHMAT: { 3971769Sjkh struct linux_shmat_args a; 3981769Sjkh 3991769Sjkh a.shmid = args->arg1; 4001773Sjkh a.shmaddr = args->ptr; 4011773Sjkh a.shmflg = args->arg2; 4021773Sjkh a.raddr = PTRIN((l_uint)args->arg3); 4031769Sjkh return (linux_shmat(td, &a)); 4041769Sjkh } 4051769Sjkh case LINUX_SHMDT: { 4061773Sjkh struct linux_shmdt_args a; 4071773Sjkh 4081773Sjkh a.shmaddr = args->ptr; 409284Srgrimes return (linux_shmdt(td, &a)); 410444Srgrimes } 4111194Srgrimes case LINUX_SHMGET: { 4121194Srgrimes struct linux_shmget_args a; 4131773Sjkh 4141769Sjkh a.key = args->arg1; 4151773Sjkh a.size = args->arg2; 4161773Sjkh a.shmflg = args->arg3; 4171731Sjkh return (linux_shmget(td, &a)); 4181731Sjkh } 4191731Sjkh case LINUX_SHMCTL: { 4201731Sjkh struct linux_shmctl_args a; 4211731Sjkh 4221731Sjkh a.shmid = args->arg1; 423284Srgrimes a.cmd = args->arg2; 4241027Sache a.buf = args->ptr; 4251027Sache return (linux_shmctl(td, &a)); 4261205Srgrimes } 427358Srgrimes default: 4281205Srgrimes break; 4291782Sjkh } 430284Srgrimes 431372Srgrimes return (EINVAL); 432372Srgrimes} 433538Srgrimes 4341782Sjkhint 435372Srgrimeslinux_old_select(struct thread *td, struct linux_old_select_args *args) 4361782Sjkh{ 4372570Srgrimes struct l_old_select_argv linux_args; 4381782Sjkh struct linux_select_args newsel; 4392570Srgrimes int error; 4401782Sjkh 4411782Sjkh#ifdef DEBUG 4421782Sjkh if (ldebug(old_select)) 443538Srgrimes printf(ARGS(old_select, "%p"), args->ptr); 444376Srgrimes#endif 4451782Sjkh 4461782Sjkh error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 447376Srgrimes if (error) 448538Srgrimes return (error); 449376Srgrimes 4501782Sjkh newsel.nfds = linux_args.nfds; 4512570Srgrimes newsel.readfds = PTRIN(linux_args.readfds); 452376Srgrimes newsel.writefds = PTRIN(linux_args.writefds); 453538Srgrimes newsel.exceptfds = PTRIN(linux_args.exceptfds); 454538Srgrimes newsel.timeout = PTRIN(linux_args.timeout); 4551782Sjkh return (linux_select(td, &newsel)); 4561782Sjkh} 457538Srgrimes 458538Srgrimesint 459538Srgrimeslinux_fork(struct thread *td, struct linux_fork_args *args) 4601782Sjkh{ 4611782Sjkh int error; 462538Srgrimes struct proc *p2; 463538Srgrimes struct thread *td2; 464538Srgrimes 4651782Sjkh#ifdef DEBUG 4661782Sjkh if (ldebug(fork)) 467538Srgrimes printf(ARGS(fork, "")); 468538Srgrimes#endif 469538Srgrimes 4701782Sjkh if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2)) != 0) 4711782Sjkh return (error); 472538Srgrimes 473538Srgrimes if (error == 0) { 474538Srgrimes td->td_retval[0] = p2->p_pid; 4751782Sjkh td->td_retval[1] = 0; 4761782Sjkh } 4771782Sjkh 478538Srgrimes if (td->td_retval[1] == 1) 479538Srgrimes td->td_retval[0] = 0; 480538Srgrimes error = linux_proc_init(td, td->td_retval[0], 0); 4811782Sjkh if (error) 4821782Sjkh return (error); 483538Srgrimes 484538Srgrimes td2 = FIRST_THREAD_IN_PROC(p2); 485538Srgrimes 4861782Sjkh /* 4871782Sjkh * Make this runnable after we are finished with it. 488538Srgrimes */ 489538Srgrimes thread_lock(td2); 490538Srgrimes TD_SET_CAN_RUN(td2); 4911782Sjkh sched_add(td2, SRQ_BORING); 4921782Sjkh thread_unlock(td2); 493538Srgrimes 494538Srgrimes return (0); 495538Srgrimes} 4961782Sjkh 4971782Sjkhint 498538Srgrimeslinux_vfork(struct thread *td, struct linux_vfork_args *args) 499538Srgrimes{ 500538Srgrimes int error; 5011782Sjkh struct proc *p2; 5021782Sjkh struct thread *td2; 503538Srgrimes 504538Srgrimes#ifdef DEBUG 505538Srgrimes if (ldebug(vfork)) 5061782Sjkh printf(ARGS(vfork, "")); 5071782Sjkh#endif 508538Srgrimes 509538Srgrimes /* Exclude RFPPWAIT */ 510538Srgrimes if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2)) != 0) 5111782Sjkh return (error); 5121782Sjkh if (error == 0) { 513538Srgrimes td->td_retval[0] = p2->p_pid; 5142619Srgrimes td->td_retval[1] = 0; 5151782Sjkh } 5161782Sjkh /* Are we the child? */ 5171782Sjkh if (td->td_retval[1] == 1) 518538Srgrimes td->td_retval[0] = 0; 519372Srgrimes error = linux_proc_init(td, td->td_retval[0], 0); 520372Srgrimes if (error) 521372Srgrimes return (error); 522372Srgrimes 523372Srgrimes PROC_LOCK(p2); 524372Srgrimes p2->p_flag |= P_PPWAIT; 525372Srgrimes PROC_UNLOCK(p2); 526372Srgrimes 527372Srgrimes td2 = FIRST_THREAD_IN_PROC(p2); 528372Srgrimes 529372Srgrimes /* 530372Srgrimes * Make this runnable after we are finished with it. 531372Srgrimes */ 532372Srgrimes thread_lock(td2); 533372Srgrimes TD_SET_CAN_RUN(td2); 534372Srgrimes sched_add(td2, SRQ_BORING); 535372Srgrimes thread_unlock(td2); 536372Srgrimes 537372Srgrimes /* wait for the children to exit, ie. emulate vfork */ 538372Srgrimes PROC_LOCK(p2); 539372Srgrimes while (p2->p_flag & P_PPWAIT) 540372Srgrimes msleep(td->td_proc, &p2->p_mtx, PWAIT, "ppwait", 0); 541372Srgrimes PROC_UNLOCK(p2); 542372Srgrimes 543372Srgrimes return (0); 544372Srgrimes} 545372Srgrimes 546538Srgrimesint 5471782Sjkhlinux_clone(struct thread *td, struct linux_clone_args *args) 548372Srgrimes{ 549372Srgrimes int error, ff = RFPROC | RFSTOPPED; 550147Srgrimes struct proc *p2; 5511731Sjkh struct thread *td2; 5521731Sjkh int exit_signal; 5531731Sjkh struct linux_emuldata *em; 554372Srgrimes 5551731Sjkh#ifdef DEBUG 556372Srgrimes if (ldebug(clone)) { 557410Srgrimes printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, " 558147Srgrimes "child tid: %p"), (unsigned)args->flags, 5591285Srgrimes args->stack, args->parent_tidptr, args->child_tidptr); 5601285Srgrimes } 561372Srgrimes#endif 5621731Sjkh 5631769Sjkh exit_signal = args->flags & 0x000000ff; 5641731Sjkh if (LINUX_SIG_VALID(exit_signal)) { 565372Srgrimes if (exit_signal <= LINUX_SIGTBLSZ) 5662611Srgrimes exit_signal = 5672611Srgrimes linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 5682611Srgrimes } else if (exit_signal != 0) 569372Srgrimes return (EINVAL); 57037Srgrimes 571 if (args->flags & LINUX_CLONE_VM) 572 ff |= RFMEM; 573 if (args->flags & LINUX_CLONE_SIGHAND) 574 ff |= RFSIGSHARE; 575 /* 576 * XXX: In Linux, sharing of fs info (chroot/cwd/umask) 577 * and open files is independant. In FreeBSD, its in one 578 * structure but in reality it does not cause any problems 579 * because both of these flags are usually set together. 580 */ 581 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS))) 582 ff |= RFFDG; 583 584 /* 585 * Attempt to detect when linux_clone(2) is used for creating 586 * kernel threads. Unfortunately despite the existence of the 587 * CLONE_THREAD flag, version of linuxthreads package used in 588 * most popular distros as of beginning of 2005 doesn't make 589 * any use of it. Therefore, this detection relies on 590 * empirical observation that linuxthreads sets certain 591 * combination of flags, so that we can make more or less 592 * precise detection and notify the FreeBSD kernel that several 593 * processes are in fact part of the same threading group, so 594 * that special treatment is necessary for signal delivery 595 * between those processes and fd locking. 596 */ 597 if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS) 598 ff |= RFTHREAD; 599 600 if (args->flags & LINUX_CLONE_PARENT_SETTID) 601 if (args->parent_tidptr == NULL) 602 return (EINVAL); 603 604 error = fork1(td, ff, 0, &p2); 605 if (error) 606 return (error); 607 608 if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD)) { 609 sx_xlock(&proctree_lock); 610 PROC_LOCK(p2); 611 proc_reparent(p2, td->td_proc->p_pptr); 612 PROC_UNLOCK(p2); 613 sx_xunlock(&proctree_lock); 614 } 615 616 /* create the emuldata */ 617 error = linux_proc_init(td, p2->p_pid, args->flags); 618 /* reference it - no need to check this */ 619 em = em_find(p2, EMUL_DOLOCK); 620 KASSERT(em != NULL, ("clone: emuldata not found.\n")); 621 /* and adjust it */ 622 623 if (args->flags & LINUX_CLONE_THREAD) { 624#ifdef notyet 625 PROC_LOCK(p2); 626 p2->p_pgrp = td->td_proc->p_pgrp; 627 PROC_UNLOCK(p2); 628#endif 629 exit_signal = 0; 630 } 631 632 if (args->flags & LINUX_CLONE_CHILD_SETTID) 633 em->child_set_tid = args->child_tidptr; 634 else 635 em->child_set_tid = NULL; 636 637 if (args->flags & LINUX_CLONE_CHILD_CLEARTID) 638 em->child_clear_tid = args->child_tidptr; 639 else 640 em->child_clear_tid = NULL; 641 642 EMUL_UNLOCK(&emul_lock); 643 644 if (args->flags & LINUX_CLONE_PARENT_SETTID) { 645 error = copyout(&p2->p_pid, args->parent_tidptr, 646 sizeof(p2->p_pid)); 647 if (error) 648 printf(LMSG("copyout failed!")); 649 } 650 651 PROC_LOCK(p2); 652 p2->p_sigparent = exit_signal; 653 PROC_UNLOCK(p2); 654 td2 = FIRST_THREAD_IN_PROC(p2); 655 /* 656 * In a case of stack = NULL, we are supposed to COW calling process 657 * stack. This is what normal fork() does, so we just keep tf_rsp arg 658 * intact. 659 */ 660 if (args->stack) 661 td2->td_frame->tf_rsp = PTROUT(args->stack); 662 663 if (args->flags & LINUX_CLONE_SETTLS) { 664 struct user_segment_descriptor sd; 665 struct l_user_desc info; 666 int a[2]; 667 668 error = copyin((void *)td->td_frame->tf_rsi, &info, 669 sizeof(struct l_user_desc)); 670 if (error) { 671 printf(LMSG("copyin failed!")); 672 } else { 673 /* We might copy out the entry_number as GUGS32_SEL. */ 674 info.entry_number = GUGS32_SEL; 675 error = copyout(&info, (void *)td->td_frame->tf_rsi, 676 sizeof(struct l_user_desc)); 677 if (error) 678 printf(LMSG("copyout failed!")); 679 680 a[0] = LINUX_LDT_entry_a(&info); 681 a[1] = LINUX_LDT_entry_b(&info); 682 683 memcpy(&sd, &a, sizeof(a)); 684#ifdef DEBUG 685 if (ldebug(clone)) 686 printf("Segment created in clone with " 687 "CLONE_SETTLS: lobase: %x, hibase: %x, " 688 "lolimit: %x, hilimit: %x, type: %i, " 689 "dpl: %i, p: %i, xx: %i, long: %i, " 690 "def32: %i, gran: %i\n", sd.sd_lobase, 691 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit, 692 sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx, 693 sd.sd_long, sd.sd_def32, sd.sd_gran); 694#endif 695 td2->td_pcb->pcb_gsbase = (register_t)info.base_addr; 696 td2->td_pcb->pcb_gs32sd = sd; 697 td2->td_pcb->pcb_gs32p = &gdt[GUGS32_SEL]; 698 td2->td_pcb->pcb_gs = GSEL(GUGS32_SEL, SEL_UPL); 699 td2->td_pcb->pcb_flags |= PCB_32BIT; 700 } 701 } 702 703#ifdef DEBUG 704 if (ldebug(clone)) 705 printf(LMSG("clone: successful rfork to %d, " 706 "stack %p sig = %d"), (int)p2->p_pid, args->stack, 707 exit_signal); 708#endif 709 if (args->flags & LINUX_CLONE_VFORK) { 710 PROC_LOCK(p2); 711 p2->p_flag |= P_PPWAIT; 712 PROC_UNLOCK(p2); 713 } 714 715 /* 716 * Make this runnable after we are finished with it. 717 */ 718 thread_lock(td2); 719 TD_SET_CAN_RUN(td2); 720 sched_add(td2, SRQ_BORING); 721 thread_unlock(td2); 722 723 td->td_retval[0] = p2->p_pid; 724 td->td_retval[1] = 0; 725 726 if (args->flags & LINUX_CLONE_VFORK) { 727 /* wait for the children to exit, ie. emulate vfork */ 728 PROC_LOCK(p2); 729 while (p2->p_flag & P_PPWAIT) 730 msleep(td->td_proc, &p2->p_mtx, PWAIT, "ppwait", 0); 731 PROC_UNLOCK(p2); 732 } 733 734 return (0); 735} 736 737#define STACK_SIZE (2 * 1024 * 1024) 738#define GUARD_SIZE (4 * PAGE_SIZE) 739 740static int linux_mmap_common(struct thread *, struct l_mmap_argv *); 741 742int 743linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 744{ 745 struct l_mmap_argv linux_args; 746 747#ifdef DEBUG 748 if (ldebug(mmap2)) 749 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"), 750 args->addr, args->len, args->prot, 751 args->flags, args->fd, args->pgoff); 752#endif 753 754 linux_args.addr = PTROUT(args->addr); 755 linux_args.len = args->len; 756 linux_args.prot = args->prot; 757 linux_args.flags = args->flags; 758 linux_args.fd = args->fd; 759 linux_args.pgoff = args->pgoff; 760 761 return (linux_mmap_common(td, &linux_args)); 762} 763 764int 765linux_mmap(struct thread *td, struct linux_mmap_args *args) 766{ 767 int error; 768 struct l_mmap_argv linux_args; 769 770 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 771 if (error) 772 return (error); 773 774#ifdef DEBUG 775 if (ldebug(mmap)) 776 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"), 777 linux_args.addr, linux_args.len, linux_args.prot, 778 linux_args.flags, linux_args.fd, linux_args.pgoff); 779#endif 780 if ((linux_args.pgoff % PAGE_SIZE) != 0) 781 return (EINVAL); 782 linux_args.pgoff /= PAGE_SIZE; 783 784 return (linux_mmap_common(td, &linux_args)); 785} 786 787static int 788linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args) 789{ 790 struct proc *p = td->td_proc; 791 struct mmap_args /* { 792 caddr_t addr; 793 size_t len; 794 int prot; 795 int flags; 796 int fd; 797 long pad; 798 off_t pos; 799 } */ bsd_args; 800 int error; 801 struct file *fp; 802 803 error = 0; 804 bsd_args.flags = 0; 805 fp = NULL; 806 807 /* 808 * Linux mmap(2): 809 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 810 */ 811 if (! ((linux_args->flags & LINUX_MAP_SHARED) ^ 812 (linux_args->flags & LINUX_MAP_PRIVATE))) 813 return (EINVAL); 814 815 if (linux_args->flags & LINUX_MAP_SHARED) 816 bsd_args.flags |= MAP_SHARED; 817 if (linux_args->flags & LINUX_MAP_PRIVATE) 818 bsd_args.flags |= MAP_PRIVATE; 819 if (linux_args->flags & LINUX_MAP_FIXED) 820 bsd_args.flags |= MAP_FIXED; 821 if (linux_args->flags & LINUX_MAP_ANON) 822 bsd_args.flags |= MAP_ANON; 823 else 824 bsd_args.flags |= MAP_NOSYNC; 825 if (linux_args->flags & LINUX_MAP_GROWSDOWN) 826 bsd_args.flags |= MAP_STACK; 827 828 /* 829 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 830 * on Linux/i386. We do this to ensure maximum compatibility. 831 * Linux/ia64 does the same in i386 emulation mode. 832 */ 833 bsd_args.prot = linux_args->prot; 834 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 835 bsd_args.prot |= PROT_READ | PROT_EXEC; 836 837 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 838 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : linux_args->fd; 839 if (bsd_args.fd != -1) { 840 /* 841 * Linux follows Solaris mmap(2) description: 842 * The file descriptor fildes is opened with 843 * read permission, regardless of the 844 * protection options specified. 845 */ 846 847 if ((error = fget(td, bsd_args.fd, &fp)) != 0) 848 return (error); 849 if (fp->f_type != DTYPE_VNODE) { 850 fdrop(fp, td); 851 return (EINVAL); 852 } 853 854 /* Linux mmap() just fails for O_WRONLY files */ 855 if (!(fp->f_flag & FREAD)) { 856 fdrop(fp, td); 857 return (EACCES); 858 } 859 860 fdrop(fp, td); 861 } 862 863 if (linux_args->flags & LINUX_MAP_GROWSDOWN) { 864 /* 865 * The Linux MAP_GROWSDOWN option does not limit auto 866 * growth of the region. Linux mmap with this option 867 * takes as addr the inital BOS, and as len, the initial 868 * region size. It can then grow down from addr without 869 * limit. However, Linux threads has an implicit internal 870 * limit to stack size of STACK_SIZE. Its just not 871 * enforced explicitly in Linux. But, here we impose 872 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 873 * region, since we can do this with our mmap. 874 * 875 * Our mmap with MAP_STACK takes addr as the maximum 876 * downsize limit on BOS, and as len the max size of 877 * the region. It then maps the top SGROWSIZ bytes, 878 * and auto grows the region down, up to the limit 879 * in addr. 880 * 881 * If we don't use the MAP_STACK option, the effect 882 * of this code is to allocate a stack region of a 883 * fixed size of (STACK_SIZE - GUARD_SIZE). 884 */ 885 886 if ((caddr_t)PTRIN(linux_args->addr) + linux_args->len > 887 p->p_vmspace->vm_maxsaddr) { 888 /* 889 * Some Linux apps will attempt to mmap 890 * thread stacks near the top of their 891 * address space. If their TOS is greater 892 * than vm_maxsaddr, vm_map_growstack() 893 * will confuse the thread stack with the 894 * process stack and deliver a SEGV if they 895 * attempt to grow the thread stack past their 896 * current stacksize rlimit. To avoid this, 897 * adjust vm_maxsaddr upwards to reflect 898 * the current stacksize rlimit rather 899 * than the maximum possible stacksize. 900 * It would be better to adjust the 901 * mmap'ed region, but some apps do not check 902 * mmap's return value. 903 */ 904 PROC_LOCK(p); 905 p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK - 906 lim_cur(p, RLIMIT_STACK); 907 PROC_UNLOCK(p); 908 } 909 910 /* 911 * This gives us our maximum stack size and a new BOS. 912 * If we're using VM_STACK, then mmap will just map 913 * the top SGROWSIZ bytes, and let the stack grow down 914 * to the limit at BOS. If we're not using VM_STACK 915 * we map the full stack, since we don't have a way 916 * to autogrow it. 917 */ 918 if (linux_args->len > STACK_SIZE - GUARD_SIZE) { 919 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr); 920 bsd_args.len = linux_args->len; 921 } else { 922 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr) - 923 (STACK_SIZE - GUARD_SIZE - linux_args->len); 924 bsd_args.len = STACK_SIZE - GUARD_SIZE; 925 } 926 } else { 927 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr); 928 bsd_args.len = linux_args->len; 929 } 930 bsd_args.pos = (off_t)linux_args->pgoff * PAGE_SIZE; 931 932#ifdef DEBUG 933 if (ldebug(mmap)) 934 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 935 __func__, 936 (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot, 937 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 938#endif 939 error = mmap(td, &bsd_args); 940#ifdef DEBUG 941 if (ldebug(mmap)) 942 printf("-> %s() return: 0x%x (0x%08x)\n", 943 __func__, error, (u_int)td->td_retval[0]); 944#endif 945 return (error); 946} 947 948int 949linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) 950{ 951 struct mprotect_args bsd_args; 952 953 bsd_args.addr = uap->addr; 954 bsd_args.len = uap->len; 955 bsd_args.prot = uap->prot; 956 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 957 bsd_args.prot |= PROT_READ | PROT_EXEC; 958 return (mprotect(td, &bsd_args)); 959} 960 961int 962linux_iopl(struct thread *td, struct linux_iopl_args *args) 963{ 964 int error; 965 966 if (args->level < 0 || args->level > 3) 967 return (EINVAL); 968 if ((error = priv_check(td, PRIV_IO)) != 0) 969 return (error); 970 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 971 return (error); 972 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) | 973 (args->level * (PSL_IOPL / 3)); 974 975 return (0); 976} 977 978int 979linux_pipe(struct thread *td, struct linux_pipe_args *args) 980{ 981 int pip[2]; 982 int error; 983 register_t reg_rdx; 984 985#ifdef DEBUG 986 if (ldebug(pipe)) 987 printf(ARGS(pipe, "*")); 988#endif 989 990 reg_rdx = td->td_retval[1]; 991 error = pipe(td, 0); 992 if (error) { 993 td->td_retval[1] = reg_rdx; 994 return (error); 995 } 996 997 pip[0] = td->td_retval[0]; 998 pip[1] = td->td_retval[1]; 999 error = copyout(pip, args->pipefds, 2 * sizeof(int)); 1000 if (error) { 1001 td->td_retval[1] = reg_rdx; 1002 return (error); 1003 } 1004 1005 td->td_retval[1] = reg_rdx; 1006 td->td_retval[0] = 0; 1007 return (0); 1008} 1009 1010int 1011linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 1012{ 1013 l_osigaction_t osa; 1014 l_sigaction_t act, oact; 1015 int error; 1016 1017#ifdef DEBUG 1018 if (ldebug(sigaction)) 1019 printf(ARGS(sigaction, "%d, %p, %p"), 1020 args->sig, (void *)args->nsa, (void *)args->osa); 1021#endif 1022 1023 if (args->nsa != NULL) { 1024 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 1025 if (error) 1026 return (error); 1027 act.lsa_handler = osa.lsa_handler; 1028 act.lsa_flags = osa.lsa_flags; 1029 act.lsa_restorer = osa.lsa_restorer; 1030 LINUX_SIGEMPTYSET(act.lsa_mask); 1031 act.lsa_mask.__bits[0] = osa.lsa_mask; 1032 } 1033 1034 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 1035 args->osa ? &oact : NULL); 1036 1037 if (args->osa != NULL && !error) { 1038 osa.lsa_handler = oact.lsa_handler; 1039 osa.lsa_flags = oact.lsa_flags; 1040 osa.lsa_restorer = oact.lsa_restorer; 1041 osa.lsa_mask = oact.lsa_mask.__bits[0]; 1042 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 1043 } 1044 1045 return (error); 1046} 1047 1048/* 1049 * Linux has two extra args, restart and oldmask. We don't use these, 1050 * but it seems that "restart" is actually a context pointer that 1051 * enables the signal to happen with a different register set. 1052 */ 1053int 1054linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 1055{ 1056 sigset_t sigmask; 1057 l_sigset_t mask; 1058 1059#ifdef DEBUG 1060 if (ldebug(sigsuspend)) 1061 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 1062#endif 1063 1064 LINUX_SIGEMPTYSET(mask); 1065 mask.__bits[0] = args->mask; 1066 linux_to_bsd_sigset(&mask, &sigmask); 1067 return (kern_sigsuspend(td, sigmask)); 1068} 1069 1070int 1071linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 1072{ 1073 l_sigset_t lmask; 1074 sigset_t sigmask; 1075 int error; 1076 1077#ifdef DEBUG 1078 if (ldebug(rt_sigsuspend)) 1079 printf(ARGS(rt_sigsuspend, "%p, %d"), 1080 (void *)uap->newset, uap->sigsetsize); 1081#endif 1082 1083 if (uap->sigsetsize != sizeof(l_sigset_t)) 1084 return (EINVAL); 1085 1086 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 1087 if (error) 1088 return (error); 1089 1090 linux_to_bsd_sigset(&lmask, &sigmask); 1091 return (kern_sigsuspend(td, sigmask)); 1092} 1093 1094int 1095linux_pause(struct thread *td, struct linux_pause_args *args) 1096{ 1097 struct proc *p = td->td_proc; 1098 sigset_t sigmask; 1099 1100#ifdef DEBUG 1101 if (ldebug(pause)) 1102 printf(ARGS(pause, "")); 1103#endif 1104 1105 PROC_LOCK(p); 1106 sigmask = td->td_sigmask; 1107 PROC_UNLOCK(p); 1108 return (kern_sigsuspend(td, sigmask)); 1109} 1110 1111int 1112linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 1113{ 1114 stack_t ss, oss; 1115 l_stack_t lss; 1116 int error; 1117 1118#ifdef DEBUG 1119 if (ldebug(sigaltstack)) 1120 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 1121#endif 1122 1123 if (uap->uss != NULL) { 1124 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 1125 if (error) 1126 return (error); 1127 1128 ss.ss_sp = PTRIN(lss.ss_sp); 1129 ss.ss_size = lss.ss_size; 1130 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 1131 } 1132 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 1133 (uap->uoss != NULL) ? &oss : NULL); 1134 if (!error && uap->uoss != NULL) { 1135 lss.ss_sp = PTROUT(oss.ss_sp); 1136 lss.ss_size = oss.ss_size; 1137 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 1138 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 1139 } 1140 1141 return (error); 1142} 1143 1144int 1145linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 1146{ 1147 struct ftruncate_args sa; 1148 1149#ifdef DEBUG 1150 if (ldebug(ftruncate64)) 1151 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 1152 (intmax_t)args->length); 1153#endif 1154 1155 sa.fd = args->fd; 1156 sa.length = args->length; 1157 return ftruncate(td, &sa); 1158} 1159 1160int 1161linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap) 1162{ 1163 struct timeval atv; 1164 l_timeval atv32; 1165 struct timezone rtz; 1166 int error = 0; 1167 1168 if (uap->tp) { 1169 microtime(&atv); 1170 atv32.tv_sec = atv.tv_sec; 1171 atv32.tv_usec = atv.tv_usec; 1172 error = copyout(&atv32, uap->tp, sizeof(atv32)); 1173 } 1174 if (error == 0 && uap->tzp != NULL) { 1175 rtz.tz_minuteswest = tz_minuteswest; 1176 rtz.tz_dsttime = tz_dsttime; 1177 error = copyout(&rtz, uap->tzp, sizeof(rtz)); 1178 } 1179 return (error); 1180} 1181 1182int 1183linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap) 1184{ 1185 l_timeval atv32; 1186 struct timeval atv, *tvp; 1187 struct timezone atz, *tzp; 1188 int error; 1189 1190 if (uap->tp) { 1191 error = copyin(uap->tp, &atv32, sizeof(atv32)); 1192 if (error) 1193 return (error); 1194 atv.tv_sec = atv32.tv_sec; 1195 atv.tv_usec = atv32.tv_usec; 1196 tvp = &atv; 1197 } else 1198 tvp = NULL; 1199 if (uap->tzp) { 1200 error = copyin(uap->tzp, &atz, sizeof(atz)); 1201 if (error) 1202 return (error); 1203 tzp = &atz; 1204 } else 1205 tzp = NULL; 1206 return (kern_settimeofday(td, tvp, tzp)); 1207} 1208 1209int 1210linux_getrusage(struct thread *td, struct linux_getrusage_args *uap) 1211{ 1212 struct l_rusage s32; 1213 struct rusage s; 1214 int error; 1215 1216 error = kern_getrusage(td, uap->who, &s); 1217 if (error != 0) 1218 return (error); 1219 if (uap->rusage != NULL) { 1220 s32.ru_utime.tv_sec = s.ru_utime.tv_sec; 1221 s32.ru_utime.tv_usec = s.ru_utime.tv_usec; 1222 s32.ru_stime.tv_sec = s.ru_stime.tv_sec; 1223 s32.ru_stime.tv_usec = s.ru_stime.tv_usec; 1224 s32.ru_maxrss = s.ru_maxrss; 1225 s32.ru_ixrss = s.ru_ixrss; 1226 s32.ru_idrss = s.ru_idrss; 1227 s32.ru_isrss = s.ru_isrss; 1228 s32.ru_minflt = s.ru_minflt; 1229 s32.ru_majflt = s.ru_majflt; 1230 s32.ru_nswap = s.ru_nswap; 1231 s32.ru_inblock = s.ru_inblock; 1232 s32.ru_oublock = s.ru_oublock; 1233 s32.ru_msgsnd = s.ru_msgsnd; 1234 s32.ru_msgrcv = s.ru_msgrcv; 1235 s32.ru_nsignals = s.ru_nsignals; 1236 s32.ru_nvcsw = s.ru_nvcsw; 1237 s32.ru_nivcsw = s.ru_nivcsw; 1238 error = copyout(&s32, uap->rusage, sizeof(s32)); 1239 } 1240 return (error); 1241} 1242 1243int 1244linux_sched_rr_get_interval(struct thread *td, 1245 struct linux_sched_rr_get_interval_args *uap) 1246{ 1247 struct timespec ts; 1248 struct l_timespec ts32; 1249 int error; 1250 1251 error = kern_sched_rr_get_interval(td, uap->pid, &ts); 1252 if (error != 0) 1253 return (error); 1254 ts32.tv_sec = ts.tv_sec; 1255 ts32.tv_nsec = ts.tv_nsec; 1256 return (copyout(&ts32, uap->interval, sizeof(ts32))); 1257} 1258 1259int 1260linux_set_thread_area(struct thread *td, 1261 struct linux_set_thread_area_args *args) 1262{ 1263 struct l_user_desc info; 1264 struct user_segment_descriptor sd; 1265 int a[2]; 1266 int error; 1267 1268 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 1269 if (error) 1270 return (error); 1271 1272#ifdef DEBUG 1273 if (ldebug(set_thread_area)) 1274 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, " 1275 "%i, %i, %i"), info.entry_number, info.base_addr, 1276 info.limit, info.seg_32bit, info.contents, 1277 info.read_exec_only, info.limit_in_pages, 1278 info.seg_not_present, info.useable); 1279#endif 1280 1281 /* 1282 * Semantics of Linux version: every thread in the system has array 1283 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. 1284 * This syscall loads one of the selected TLS decriptors with a value 1285 * and also loads GDT descriptors 6, 7 and 8 with the content of 1286 * the per-thread descriptors. 1287 * 1288 * Semantics of FreeBSD version: I think we can ignore that Linux has 1289 * three per-thread descriptors and use just the first one. 1290 * The tls_array[] is used only in [gs]et_thread_area() syscalls and 1291 * for loading the GDT descriptors. We use just one GDT descriptor 1292 * for TLS, so we will load just one. 1293 * 1294 * XXX: This doesn't work when a user space process tries to use more 1295 * than one TLS segment. Comment in the Linux source says wine might 1296 * do this. 1297 */ 1298 1299 /* 1300 * GLIBC reads current %gs and call set_thread_area() with it. 1301 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because 1302 * we use these segments. 1303 */ 1304 switch (info.entry_number) { 1305 case GUGS32_SEL: 1306 case GUDATA_SEL: 1307 case 6: 1308 case -1: 1309 info.entry_number = GUGS32_SEL; 1310 break; 1311 default: 1312 return (EINVAL); 1313 } 1314 1315 /* 1316 * We have to copy out the GDT entry we use. 1317 * 1318 * XXX: What if a user space program does not check the return value 1319 * and tries to use 6, 7 or 8? 1320 */ 1321 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 1322 if (error) 1323 return (error); 1324 1325 if (LINUX_LDT_empty(&info)) { 1326 a[0] = 0; 1327 a[1] = 0; 1328 } else { 1329 a[0] = LINUX_LDT_entry_a(&info); 1330 a[1] = LINUX_LDT_entry_b(&info); 1331 } 1332 1333 memcpy(&sd, &a, sizeof(a)); 1334#ifdef DEBUG 1335 if (ldebug(set_thread_area)) 1336 printf("Segment created in set_thread_area: " 1337 "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, " 1338 "type: %i, dpl: %i, p: %i, xx: %i, long: %i, " 1339 "def32: %i, gran: %i\n", 1340 sd.sd_lobase, 1341 sd.sd_hibase, 1342 sd.sd_lolimit, 1343 sd.sd_hilimit, 1344 sd.sd_type, 1345 sd.sd_dpl, 1346 sd.sd_p, 1347 sd.sd_xx, 1348 sd.sd_long, 1349 sd.sd_def32, 1350 sd.sd_gran); 1351#endif 1352 1353 critical_enter(); 1354 td->td_pcb->pcb_gsbase = (register_t)info.base_addr; 1355 td->td_pcb->pcb_gs32sd = gdt[GUGS32_SEL] = sd; 1356 td->td_pcb->pcb_gs32p = &gdt[GUGS32_SEL]; 1357 td->td_pcb->pcb_flags |= PCB_32BIT; 1358 wrmsr(MSR_KGSBASE, td->td_pcb->pcb_gsbase); 1359 critical_exit(); 1360 1361 return (0); 1362} 1363