The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/compat/linux/linux_misc.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 2002 Doug Rabson
    3  * Copyright (c) 1994-1995 Søren Schmidt
    4  * 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  *    in this position and unchanged.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. The name of the author may not be used to endorse or promote products
   16  *    derived from this software without specific prior written permission
   17  *
   18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
   19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
   20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
   21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
   22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
   23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
   24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
   25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
   26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
   27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
   28  */
   29 
   30 #include <sys/cdefs.h>
   31 __FBSDID("$FreeBSD$");
   32 
   33 #include "opt_compat.h"
   34 #include "opt_kdtrace.h"
   35 
   36 #include <sys/param.h>
   37 #include <sys/blist.h>
   38 #include <sys/fcntl.h>
   39 #if defined(__i386__)
   40 #include <sys/imgact_aout.h>
   41 #endif
   42 #include <sys/jail.h>
   43 #include <sys/kernel.h>
   44 #include <sys/limits.h>
   45 #include <sys/lock.h>
   46 #include <sys/malloc.h>
   47 #include <sys/mman.h>
   48 #include <sys/mount.h>
   49 #include <sys/mutex.h>
   50 #include <sys/namei.h>
   51 #include <sys/priv.h>
   52 #include <sys/proc.h>
   53 #include <sys/reboot.h>
   54 #include <sys/racct.h>
   55 #include <sys/resourcevar.h>
   56 #include <sys/sched.h>
   57 #include <sys/sdt.h>
   58 #include <sys/signalvar.h>
   59 #include <sys/stat.h>
   60 #include <sys/syscallsubr.h>
   61 #include <sys/sysctl.h>
   62 #include <sys/sysproto.h>
   63 #include <sys/systm.h>
   64 #include <sys/time.h>
   65 #include <sys/vmmeter.h>
   66 #include <sys/vnode.h>
   67 #include <sys/wait.h>
   68 #include <sys/cpuset.h>
   69 
   70 #include <security/mac/mac_framework.h>
   71 
   72 #include <vm/vm.h>
   73 #include <vm/pmap.h>
   74 #include <vm/vm_kern.h>
   75 #include <vm/vm_map.h>
   76 #include <vm/vm_extern.h>
   77 #include <vm/vm_object.h>
   78 #include <vm/swap_pager.h>
   79 
   80 #ifdef COMPAT_LINUX32
   81 #include <machine/../linux32/linux.h>
   82 #include <machine/../linux32/linux32_proto.h>
   83 #else
   84 #include <machine/../linux/linux.h>
   85 #include <machine/../linux/linux_proto.h>
   86 #endif
   87 
   88 #include <compat/linux/linux_dtrace.h>
   89 #include <compat/linux/linux_file.h>
   90 #include <compat/linux/linux_mib.h>
   91 #include <compat/linux/linux_signal.h>
   92 #include <compat/linux/linux_timer.h>
   93 #include <compat/linux/linux_util.h>
   94 #include <compat/linux/linux_sysproto.h>
   95 #include <compat/linux/linux_emul.h>
   96 #include <compat/linux/linux_misc.h>
   97 
   98 /**
   99  * Special DTrace provider for the linuxulator.
  100  *
  101  * In this file we define the provider for the entire linuxulator. All
  102  * modules (= files of the linuxulator) use it.
  103  *
  104  * We define a different name depending on the emulated bitsize, see
  105  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
  106  *      native bitsize          = linuxulator
  107  *      amd64, 32bit emulation  = linuxulator32
  108  */
  109 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
  110 
  111 int stclohz;                            /* Statistics clock frequency */
  112 
  113 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
  114         RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
  115         RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
  116         RLIMIT_MEMLOCK, RLIMIT_AS 
  117 };
  118 
  119 struct l_sysinfo {
  120         l_long          uptime;         /* Seconds since boot */
  121         l_ulong         loads[3];       /* 1, 5, and 15 minute load averages */
  122 #define LINUX_SYSINFO_LOADS_SCALE 65536
  123         l_ulong         totalram;       /* Total usable main memory size */
  124         l_ulong         freeram;        /* Available memory size */
  125         l_ulong         sharedram;      /* Amount of shared memory */
  126         l_ulong         bufferram;      /* Memory used by buffers */
  127         l_ulong         totalswap;      /* Total swap space size */
  128         l_ulong         freeswap;       /* swap space still available */
  129         l_ushort        procs;          /* Number of current processes */
  130         l_ushort        pads;
  131         l_ulong         totalbig;
  132         l_ulong         freebig;
  133         l_uint          mem_unit;
  134         char            _f[20-2*sizeof(l_long)-sizeof(l_int)];  /* padding */
  135 };
  136 
  137 struct l_pselect6arg {
  138         l_uintptr_t     ss;
  139         l_size_t        ss_len;
  140 };
  141 
  142 static int      linux_utimensat_nsec_valid(l_long);
  143 
  144 
  145 int
  146 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
  147 {
  148         struct l_sysinfo sysinfo;
  149         vm_object_t object;
  150         int i, j;
  151         struct timespec ts;
  152 
  153         bzero(&sysinfo, sizeof(sysinfo));
  154         getnanouptime(&ts);
  155         if (ts.tv_nsec != 0)
  156                 ts.tv_sec++;
  157         sysinfo.uptime = ts.tv_sec;
  158 
  159         /* Use the information from the mib to get our load averages */
  160         for (i = 0; i < 3; i++)
  161                 sysinfo.loads[i] = averunnable.ldavg[i] *
  162                     LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
  163 
  164         sysinfo.totalram = physmem * PAGE_SIZE;
  165         sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
  166 
  167         sysinfo.sharedram = 0;
  168         mtx_lock(&vm_object_list_mtx);
  169         TAILQ_FOREACH(object, &vm_object_list, object_list)
  170                 if (object->shadow_count > 1)
  171                         sysinfo.sharedram += object->resident_page_count;
  172         mtx_unlock(&vm_object_list_mtx);
  173 
  174         sysinfo.sharedram *= PAGE_SIZE;
  175         sysinfo.bufferram = 0;
  176 
  177         swap_pager_status(&i, &j);
  178         sysinfo.totalswap = i * PAGE_SIZE;
  179         sysinfo.freeswap = (i - j) * PAGE_SIZE;
  180 
  181         sysinfo.procs = nprocs;
  182 
  183         /* The following are only present in newer Linux kernels. */
  184         sysinfo.totalbig = 0;
  185         sysinfo.freebig = 0;
  186         sysinfo.mem_unit = 1;
  187 
  188         return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
  189 }
  190 
  191 int
  192 linux_alarm(struct thread *td, struct linux_alarm_args *args)
  193 {
  194         struct itimerval it, old_it;
  195         u_int secs;
  196         int error;
  197 
  198 #ifdef DEBUG
  199         if (ldebug(alarm))
  200                 printf(ARGS(alarm, "%u"), args->secs);
  201 #endif
  202         secs = args->secs;
  203         /*
  204          * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
  205          * to match kern_setitimer()'s limit to avoid error from it.
  206          *
  207          * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
  208          * platforms.
  209          */
  210         if (secs > INT32_MAX / 2)
  211                 secs = INT32_MAX / 2;
  212 
  213         it.it_value.tv_sec = secs;
  214         it.it_value.tv_usec = 0;
  215         timevalclear(&it.it_interval);
  216         error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
  217         KASSERT(error == 0, ("kern_setitimer returns %d", error));
  218 
  219         if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
  220             old_it.it_value.tv_usec >= 500000)
  221                 old_it.it_value.tv_sec++;
  222         td->td_retval[0] = old_it.it_value.tv_sec;
  223         return (0);
  224 }
  225 
  226 int
  227 linux_brk(struct thread *td, struct linux_brk_args *args)
  228 {
  229         struct vmspace *vm = td->td_proc->p_vmspace;
  230         vm_offset_t new, old;
  231         struct obreak_args /* {
  232                 char * nsize;
  233         } */ tmp;
  234 
  235 #ifdef DEBUG
  236         if (ldebug(brk))
  237                 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
  238 #endif
  239         old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
  240         new = (vm_offset_t)args->dsend;
  241         tmp.nsize = (char *)new;
  242         if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
  243                 td->td_retval[0] = (long)new;
  244         else
  245                 td->td_retval[0] = (long)old;
  246 
  247         return (0);
  248 }
  249 
  250 #if defined(__i386__)
  251 /* XXX: what about amd64/linux32? */
  252 
  253 int
  254 linux_uselib(struct thread *td, struct linux_uselib_args *args)
  255 {
  256         struct nameidata ni;
  257         struct vnode *vp;
  258         struct exec *a_out;
  259         struct vattr attr;
  260         vm_offset_t vmaddr;
  261         unsigned long file_offset;
  262         unsigned long bss_size;
  263         char *library;
  264         ssize_t aresid;
  265         int error, locked, writecount;
  266 
  267         LCONVPATHEXIST(td, args->library, &library);
  268 
  269 #ifdef DEBUG
  270         if (ldebug(uselib))
  271                 printf(ARGS(uselib, "%s"), library);
  272 #endif
  273 
  274         a_out = NULL;
  275         locked = 0;
  276         vp = NULL;
  277 
  278         NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
  279             UIO_SYSSPACE, library, td);
  280         error = namei(&ni);
  281         LFREEPATH(library);
  282         if (error)
  283                 goto cleanup;
  284 
  285         vp = ni.ni_vp;
  286         NDFREE(&ni, NDF_ONLY_PNBUF);
  287 
  288         /*
  289          * From here on down, we have a locked vnode that must be unlocked.
  290          * XXX: The code below largely duplicates exec_check_permissions().
  291          */
  292         locked = 1;
  293 
  294         /* Writable? */
  295         error = VOP_GET_WRITECOUNT(vp, &writecount);
  296         if (error != 0)
  297                 goto cleanup;
  298         if (writecount != 0) {
  299                 error = ETXTBSY;
  300                 goto cleanup;
  301         }
  302 
  303         /* Executable? */
  304         error = VOP_GETATTR(vp, &attr, td->td_ucred);
  305         if (error)
  306                 goto cleanup;
  307 
  308         if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
  309             ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
  310                 /* EACCESS is what exec(2) returns. */
  311                 error = ENOEXEC;
  312                 goto cleanup;
  313         }
  314 
  315         /* Sensible size? */
  316         if (attr.va_size == 0) {
  317                 error = ENOEXEC;
  318                 goto cleanup;
  319         }
  320 
  321         /* Can we access it? */
  322         error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
  323         if (error)
  324                 goto cleanup;
  325 
  326         /*
  327          * XXX: This should use vn_open() so that it is properly authorized,
  328          * and to reduce code redundancy all over the place here.
  329          * XXX: Not really, it duplicates far more of exec_check_permissions()
  330          * than vn_open().
  331          */
  332 #ifdef MAC
  333         error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
  334         if (error)
  335                 goto cleanup;
  336 #endif
  337         error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
  338         if (error)
  339                 goto cleanup;
  340 
  341         /* Pull in executable header into exec_map */
  342         error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
  343             VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
  344         if (error)
  345                 goto cleanup;
  346 
  347         /* Is it a Linux binary ? */
  348         if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
  349                 error = ENOEXEC;
  350                 goto cleanup;
  351         }
  352 
  353         /*
  354          * While we are here, we should REALLY do some more checks
  355          */
  356 
  357         /* Set file/virtual offset based on a.out variant. */
  358         switch ((int)(a_out->a_magic & 0xffff)) {
  359         case 0413:                      /* ZMAGIC */
  360                 file_offset = 1024;
  361                 break;
  362         case 0314:                      /* QMAGIC */
  363                 file_offset = 0;
  364                 break;
  365         default:
  366                 error = ENOEXEC;
  367                 goto cleanup;
  368         }
  369 
  370         bss_size = round_page(a_out->a_bss);
  371 
  372         /* Check various fields in header for validity/bounds. */
  373         if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
  374                 error = ENOEXEC;
  375                 goto cleanup;
  376         }
  377 
  378         /* text + data can't exceed file size */
  379         if (a_out->a_data + a_out->a_text > attr.va_size) {
  380                 error = EFAULT;
  381                 goto cleanup;
  382         }
  383 
  384         /*
  385          * text/data/bss must not exceed limits
  386          * XXX - this is not complete. it should check current usage PLUS
  387          * the resources needed by this library.
  388          */
  389         PROC_LOCK(td->td_proc);
  390         if (a_out->a_text > maxtsiz ||
  391             a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
  392             racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
  393             bss_size) != 0) {
  394                 PROC_UNLOCK(td->td_proc);
  395                 error = ENOMEM;
  396                 goto cleanup;
  397         }
  398         PROC_UNLOCK(td->td_proc);
  399 
  400         /*
  401          * Prevent more writers.
  402          * XXX: Note that if any of the VM operations fail below we don't
  403          * clear this flag.
  404          */
  405         VOP_SET_TEXT(vp);
  406 
  407         /*
  408          * Lock no longer needed
  409          */
  410         locked = 0;
  411         VOP_UNLOCK(vp, 0);
  412 
  413         /*
  414          * Check if file_offset page aligned. Currently we cannot handle
  415          * misalinged file offsets, and so we read in the entire image
  416          * (what a waste).
  417          */
  418         if (file_offset & PAGE_MASK) {
  419 #ifdef DEBUG
  420                 printf("uselib: Non page aligned binary %lu\n", file_offset);
  421 #endif
  422                 /* Map text+data read/write/execute */
  423 
  424                 /* a_entry is the load address and is page aligned */
  425                 vmaddr = trunc_page(a_out->a_entry);
  426 
  427                 /* get anon user mapping, read+write+execute */
  428                 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
  429                     &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
  430                     VM_PROT_ALL, VM_PROT_ALL, 0);
  431                 if (error)
  432                         goto cleanup;
  433 
  434                 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
  435                     a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
  436                     td->td_ucred, NOCRED, &aresid, td);
  437                 if (error != 0)
  438                         goto cleanup;
  439                 if (aresid != 0) {
  440                         error = ENOEXEC;
  441                         goto cleanup;
  442                 }
  443         } else {
  444 #ifdef DEBUG
  445                 printf("uselib: Page aligned binary %lu\n", file_offset);
  446 #endif
  447                 /*
  448                  * for QMAGIC, a_entry is 20 bytes beyond the load address
  449                  * to skip the executable header
  450                  */
  451                 vmaddr = trunc_page(a_out->a_entry);
  452 
  453                 /*
  454                  * Map it all into the process's space as a single
  455                  * copy-on-write "data" segment.
  456                  */
  457                 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
  458                     a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
  459                     MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
  460                 if (error)
  461                         goto cleanup;
  462         }
  463 #ifdef DEBUG
  464         printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
  465             ((long *)vmaddr)[1]);
  466 #endif
  467         if (bss_size != 0) {
  468                 /* Calculate BSS start address */
  469                 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
  470                     a_out->a_data;
  471 
  472                 /* allocate some 'anon' space */
  473                 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
  474                     &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
  475                     VM_PROT_ALL, 0);
  476                 if (error)
  477                         goto cleanup;
  478         }
  479 
  480 cleanup:
  481         /* Unlock vnode if needed */
  482         if (locked)
  483                 VOP_UNLOCK(vp, 0);
  484 
  485         /* Release the temporary mapping. */
  486         if (a_out)
  487                 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
  488 
  489         return (error);
  490 }
  491 
  492 #endif  /* __i386__ */
  493 
  494 int
  495 linux_select(struct thread *td, struct linux_select_args *args)
  496 {
  497         l_timeval ltv;
  498         struct timeval tv0, tv1, utv, *tvp;
  499         int error;
  500 
  501 #ifdef DEBUG
  502         if (ldebug(select))
  503                 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
  504                     (void *)args->readfds, (void *)args->writefds,
  505                     (void *)args->exceptfds, (void *)args->timeout);
  506 #endif
  507 
  508         /*
  509          * Store current time for computation of the amount of
  510          * time left.
  511          */
  512         if (args->timeout) {
  513                 if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
  514                         goto select_out;
  515                 utv.tv_sec = ltv.tv_sec;
  516                 utv.tv_usec = ltv.tv_usec;
  517 #ifdef DEBUG
  518                 if (ldebug(select))
  519                         printf(LMSG("incoming timeout (%jd/%ld)"),
  520                             (intmax_t)utv.tv_sec, utv.tv_usec);
  521 #endif
  522 
  523                 if (itimerfix(&utv)) {
  524                         /*
  525                          * The timeval was invalid.  Convert it to something
  526                          * valid that will act as it does under Linux.
  527                          */
  528                         utv.tv_sec += utv.tv_usec / 1000000;
  529                         utv.tv_usec %= 1000000;
  530                         if (utv.tv_usec < 0) {
  531                                 utv.tv_sec -= 1;
  532                                 utv.tv_usec += 1000000;
  533                         }
  534                         if (utv.tv_sec < 0)
  535                                 timevalclear(&utv);
  536                 }
  537                 microtime(&tv0);
  538                 tvp = &utv;
  539         } else
  540                 tvp = NULL;
  541 
  542         error = kern_select(td, args->nfds, args->readfds, args->writefds,
  543             args->exceptfds, tvp, LINUX_NFDBITS);
  544 
  545 #ifdef DEBUG
  546         if (ldebug(select))
  547                 printf(LMSG("real select returns %d"), error);
  548 #endif
  549         if (error)
  550                 goto select_out;
  551 
  552         if (args->timeout) {
  553                 if (td->td_retval[0]) {
  554                         /*
  555                          * Compute how much time was left of the timeout,
  556                          * by subtracting the current time and the time
  557                          * before we started the call, and subtracting
  558                          * that result from the user-supplied value.
  559                          */
  560                         microtime(&tv1);
  561                         timevalsub(&tv1, &tv0);
  562                         timevalsub(&utv, &tv1);
  563                         if (utv.tv_sec < 0)
  564                                 timevalclear(&utv);
  565                 } else
  566                         timevalclear(&utv);
  567 #ifdef DEBUG
  568                 if (ldebug(select))
  569                         printf(LMSG("outgoing timeout (%jd/%ld)"),
  570                             (intmax_t)utv.tv_sec, utv.tv_usec);
  571 #endif
  572                 ltv.tv_sec = utv.tv_sec;
  573                 ltv.tv_usec = utv.tv_usec;
  574                 if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
  575                         goto select_out;
  576         }
  577 
  578 select_out:
  579 #ifdef DEBUG
  580         if (ldebug(select))
  581                 printf(LMSG("select_out -> %d"), error);
  582 #endif
  583         return (error);
  584 }
  585 
  586 int
  587 linux_mremap(struct thread *td, struct linux_mremap_args *args)
  588 {
  589         struct munmap_args /* {
  590                 void *addr;
  591                 size_t len;
  592         } */ bsd_args;
  593         int error = 0;
  594 
  595 #ifdef DEBUG
  596         if (ldebug(mremap))
  597                 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
  598                     (void *)(uintptr_t)args->addr,
  599                     (unsigned long)args->old_len,
  600                     (unsigned long)args->new_len,
  601                     (unsigned long)args->flags);
  602 #endif
  603 
  604         if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
  605                 td->td_retval[0] = 0;
  606                 return (EINVAL);
  607         }
  608 
  609         /*
  610          * Check for the page alignment.
  611          * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
  612          */
  613         if (args->addr & PAGE_MASK) {
  614                 td->td_retval[0] = 0;
  615                 return (EINVAL);
  616         }
  617 
  618         args->new_len = round_page(args->new_len);
  619         args->old_len = round_page(args->old_len);
  620 
  621         if (args->new_len > args->old_len) {
  622                 td->td_retval[0] = 0;
  623                 return (ENOMEM);
  624         }
  625 
  626         if (args->new_len < args->old_len) {
  627                 bsd_args.addr =
  628                     (caddr_t)((uintptr_t)args->addr + args->new_len);
  629                 bsd_args.len = args->old_len - args->new_len;
  630                 error = sys_munmap(td, &bsd_args);
  631         }
  632 
  633         td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
  634         return (error);
  635 }
  636 
  637 #define LINUX_MS_ASYNC       0x0001
  638 #define LINUX_MS_INVALIDATE  0x0002
  639 #define LINUX_MS_SYNC        0x0004
  640 
  641 int
  642 linux_msync(struct thread *td, struct linux_msync_args *args)
  643 {
  644         struct msync_args bsd_args;
  645 
  646         bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
  647         bsd_args.len = (uintptr_t)args->len;
  648         bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
  649 
  650         return (sys_msync(td, &bsd_args));
  651 }
  652 
  653 int
  654 linux_time(struct thread *td, struct linux_time_args *args)
  655 {
  656         struct timeval tv;
  657         l_time_t tm;
  658         int error;
  659 
  660 #ifdef DEBUG
  661         if (ldebug(time))
  662                 printf(ARGS(time, "*"));
  663 #endif
  664 
  665         microtime(&tv);
  666         tm = tv.tv_sec;
  667         if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
  668                 return (error);
  669         td->td_retval[0] = tm;
  670         return (0);
  671 }
  672 
  673 struct l_times_argv {
  674         l_clock_t       tms_utime;
  675         l_clock_t       tms_stime;
  676         l_clock_t       tms_cutime;
  677         l_clock_t       tms_cstime;
  678 };
  679 
  680 
  681 /*
  682  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
  683  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
  684  * auxiliary vector entry.
  685  */
  686 #define CLK_TCK         100
  687 
  688 #define CONVOTCK(r)     (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
  689 #define CONVNTCK(r)     (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
  690 
  691 #define CONVTCK(r)      (linux_kernver(td) >= LINUX_KERNVER_2004000 ?           \
  692                             CONVNTCK(r) : CONVOTCK(r))
  693 
  694 int
  695 linux_times(struct thread *td, struct linux_times_args *args)
  696 {
  697         struct timeval tv, utime, stime, cutime, cstime;
  698         struct l_times_argv tms;
  699         struct proc *p;
  700         int error;
  701 
  702 #ifdef DEBUG
  703         if (ldebug(times))
  704                 printf(ARGS(times, "*"));
  705 #endif
  706 
  707         if (args->buf != NULL) {
  708                 p = td->td_proc;
  709                 PROC_LOCK(p);
  710                 PROC_STATLOCK(p);
  711                 calcru(p, &utime, &stime);
  712                 PROC_STATUNLOCK(p);
  713                 calccru(p, &cutime, &cstime);
  714                 PROC_UNLOCK(p);
  715 
  716                 tms.tms_utime = CONVTCK(utime);
  717                 tms.tms_stime = CONVTCK(stime);
  718 
  719                 tms.tms_cutime = CONVTCK(cutime);
  720                 tms.tms_cstime = CONVTCK(cstime);
  721 
  722                 if ((error = copyout(&tms, args->buf, sizeof(tms))))
  723                         return (error);
  724         }
  725 
  726         microuptime(&tv);
  727         td->td_retval[0] = (int)CONVTCK(tv);
  728         return (0);
  729 }
  730 
  731 int
  732 linux_newuname(struct thread *td, struct linux_newuname_args *args)
  733 {
  734         struct l_new_utsname utsname;
  735         char osname[LINUX_MAX_UTSNAME];
  736         char osrelease[LINUX_MAX_UTSNAME];
  737         char *p;
  738 
  739 #ifdef DEBUG
  740         if (ldebug(newuname))
  741                 printf(ARGS(newuname, "*"));
  742 #endif
  743 
  744         linux_get_osname(td, osname);
  745         linux_get_osrelease(td, osrelease);
  746 
  747         bzero(&utsname, sizeof(utsname));
  748         strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
  749         getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
  750         getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
  751         strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
  752         strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
  753         for (p = utsname.version; *p != '\0'; ++p)
  754                 if (*p == '\n') {
  755                         *p = '\0';
  756                         break;
  757                 }
  758         strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
  759 
  760         return (copyout(&utsname, args->buf, sizeof(utsname)));
  761 }
  762 
  763 struct l_utimbuf {
  764         l_time_t l_actime;
  765         l_time_t l_modtime;
  766 };
  767 
  768 int
  769 linux_utime(struct thread *td, struct linux_utime_args *args)
  770 {
  771         struct timeval tv[2], *tvp;
  772         struct l_utimbuf lut;
  773         char *fname;
  774         int error;
  775 
  776         LCONVPATHEXIST(td, args->fname, &fname);
  777 
  778 #ifdef DEBUG
  779         if (ldebug(utime))
  780                 printf(ARGS(utime, "%s, *"), fname);
  781 #endif
  782 
  783         if (args->times) {
  784                 if ((error = copyin(args->times, &lut, sizeof lut))) {
  785                         LFREEPATH(fname);
  786                         return (error);
  787                 }
  788                 tv[0].tv_sec = lut.l_actime;
  789                 tv[0].tv_usec = 0;
  790                 tv[1].tv_sec = lut.l_modtime;
  791                 tv[1].tv_usec = 0;
  792                 tvp = tv;
  793         } else
  794                 tvp = NULL;
  795 
  796         error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
  797         LFREEPATH(fname);
  798         return (error);
  799 }
  800 
  801 int
  802 linux_utimes(struct thread *td, struct linux_utimes_args *args)
  803 {
  804         l_timeval ltv[2];
  805         struct timeval tv[2], *tvp = NULL;
  806         char *fname;
  807         int error;
  808 
  809         LCONVPATHEXIST(td, args->fname, &fname);
  810 
  811 #ifdef DEBUG
  812         if (ldebug(utimes))
  813                 printf(ARGS(utimes, "%s, *"), fname);
  814 #endif
  815 
  816         if (args->tptr != NULL) {
  817                 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
  818                         LFREEPATH(fname);
  819                         return (error);
  820                 }
  821                 tv[0].tv_sec = ltv[0].tv_sec;
  822                 tv[0].tv_usec = ltv[0].tv_usec;
  823                 tv[1].tv_sec = ltv[1].tv_sec;
  824                 tv[1].tv_usec = ltv[1].tv_usec;
  825                 tvp = tv;
  826         }
  827 
  828         error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
  829         LFREEPATH(fname);
  830         return (error);
  831 }
  832 
  833 static int
  834 linux_utimensat_nsec_valid(l_long nsec)
  835 {
  836 
  837         if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
  838                 return (0);
  839         if (nsec >= 0 && nsec <= 999999999)
  840                 return (0);
  841         return (1);
  842 }
  843 
  844 int 
  845 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
  846 {
  847         struct l_timespec l_times[2];
  848         struct timespec times[2], *timesp = NULL;
  849         char *path = NULL;
  850         int error, dfd, flags = 0;
  851 
  852         dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
  853 
  854 #ifdef DEBUG
  855         if (ldebug(utimensat))
  856                 printf(ARGS(utimensat, "%d, *"), dfd);
  857 #endif
  858 
  859         if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
  860                 return (EINVAL);
  861 
  862         if (args->times != NULL) {
  863                 error = copyin(args->times, l_times, sizeof(l_times));
  864                 if (error != 0)
  865                         return (error);
  866 
  867                 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
  868                     linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
  869                         return (EINVAL);
  870 
  871                 times[0].tv_sec = l_times[0].tv_sec;
  872                 switch (l_times[0].tv_nsec)
  873                 {
  874                 case LINUX_UTIME_OMIT:
  875                         times[0].tv_nsec = UTIME_OMIT;
  876                         break;
  877                 case LINUX_UTIME_NOW:
  878                         times[0].tv_nsec = UTIME_NOW;
  879                         break;
  880                 default:
  881                         times[0].tv_nsec = l_times[0].tv_nsec;
  882                 }
  883 
  884                 times[1].tv_sec = l_times[1].tv_sec;
  885                 switch (l_times[1].tv_nsec)
  886                 {
  887                 case LINUX_UTIME_OMIT:
  888                         times[1].tv_nsec = UTIME_OMIT;
  889                         break;
  890                 case LINUX_UTIME_NOW:
  891                         times[1].tv_nsec = UTIME_NOW;
  892                         break;
  893                 default:
  894                         times[1].tv_nsec = l_times[1].tv_nsec;
  895                         break;
  896                 }
  897                 timesp = times;
  898 
  899                 /* This breaks POSIX, but is what the Linux kernel does
  900                  * _on purpose_ (documented in the man page for utimensat(2)),
  901                  * so we must follow that behaviour. */
  902                 if (times[0].tv_nsec == UTIME_OMIT &&
  903                     times[1].tv_nsec == UTIME_OMIT)
  904                         return (0);
  905         }
  906 
  907         if (args->pathname != NULL)
  908                 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
  909         else if (args->flags != 0)
  910                 return (EINVAL);
  911 
  912         if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
  913                 flags |= AT_SYMLINK_NOFOLLOW;
  914 
  915         if (path == NULL)
  916                 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
  917         else {
  918                 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
  919                         UIO_SYSSPACE, flags);
  920                 LFREEPATH(path);
  921         }
  922 
  923         return (error);
  924 }
  925 
  926 int
  927 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
  928 {
  929         l_timeval ltv[2];
  930         struct timeval tv[2], *tvp = NULL;
  931         char *fname;
  932         int error, dfd;
  933 
  934         dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
  935         LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
  936 
  937 #ifdef DEBUG
  938         if (ldebug(futimesat))
  939                 printf(ARGS(futimesat, "%s, *"), fname);
  940 #endif
  941 
  942         if (args->utimes != NULL) {
  943                 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
  944                         LFREEPATH(fname);
  945                         return (error);
  946                 }
  947                 tv[0].tv_sec = ltv[0].tv_sec;
  948                 tv[0].tv_usec = ltv[0].tv_usec;
  949                 tv[1].tv_sec = ltv[1].tv_sec;
  950                 tv[1].tv_usec = ltv[1].tv_usec;
  951                 tvp = tv;
  952         }
  953 
  954         error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
  955         LFREEPATH(fname);
  956         return (error);
  957 }
  958 
  959 int
  960 linux_common_wait(struct thread *td, int pid, int *status,
  961     int options, struct rusage *ru)
  962 {
  963         int error, tmpstat;
  964 
  965         error = kern_wait(td, pid, &tmpstat, options, ru);
  966         if (error)
  967                 return (error);
  968 
  969         if (status) {
  970                 tmpstat &= 0xffff;
  971                 if (WIFSIGNALED(tmpstat))
  972                         tmpstat = (tmpstat & 0xffffff80) |
  973                             bsd_to_linux_signal(WTERMSIG(tmpstat));
  974                 else if (WIFSTOPPED(tmpstat))
  975                         tmpstat = (tmpstat & 0xffff00ff) |
  976                             (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
  977                 else if (WIFCONTINUED(tmpstat))
  978                         tmpstat = 0xffff;
  979                 error = copyout(&tmpstat, status, sizeof(int));
  980         }
  981 
  982         return (error);
  983 }
  984 
  985 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
  986 int
  987 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
  988 {
  989         struct linux_wait4_args wait4_args;
  990 
  991 #ifdef DEBUG
  992         if (ldebug(waitpid))
  993                 printf(ARGS(waitpid, "%d, %p, %d"),
  994                     args->pid, (void *)args->status, args->options);
  995 #endif
  996 
  997         wait4_args.pid = args->pid;
  998         wait4_args.status = args->status;
  999         wait4_args.options = args->options;
 1000         wait4_args.rusage = NULL;
 1001 
 1002         return (linux_wait4(td, &wait4_args));
 1003 }
 1004 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
 1005 
 1006 int
 1007 linux_wait4(struct thread *td, struct linux_wait4_args *args)
 1008 {
 1009         int error, options;
 1010         struct rusage ru, *rup;
 1011 
 1012 #ifdef DEBUG
 1013         if (ldebug(wait4))
 1014                 printf(ARGS(wait4, "%d, %p, %d, %p"),
 1015                     args->pid, (void *)args->status, args->options,
 1016                     (void *)args->rusage);
 1017 #endif
 1018         if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
 1019             LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
 1020                 return (EINVAL);
 1021 
 1022         options = WEXITED;
 1023         linux_to_bsd_waitopts(args->options, &options);
 1024 
 1025         if (args->rusage != NULL)
 1026                 rup = &ru;
 1027         else
 1028                 rup = NULL;
 1029         error = linux_common_wait(td, args->pid, args->status, options, rup);
 1030         if (error != 0)
 1031                 return (error);
 1032         if (args->rusage != NULL)
 1033                 error = linux_copyout_rusage(&ru, args->rusage);
 1034         return (error);
 1035 }
 1036 
 1037 int
 1038 linux_waitid(struct thread *td, struct linux_waitid_args *args)
 1039 {
 1040         int status, options, sig;
 1041         struct __wrusage wru;
 1042         siginfo_t siginfo;
 1043         l_siginfo_t lsi;
 1044         idtype_t idtype;
 1045         struct proc *p;
 1046         int error;
 1047 
 1048         options = 0;
 1049         linux_to_bsd_waitopts(args->options, &options);
 1050 
 1051         if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
 1052                 return (EINVAL);
 1053         if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
 1054                 return (EINVAL);
 1055 
 1056         switch (args->idtype) {
 1057         case LINUX_P_ALL:
 1058                 idtype = P_ALL;
 1059                 break;
 1060         case LINUX_P_PID:
 1061                 if (args->id <= 0)
 1062                         return (EINVAL);
 1063                 idtype = P_PID;
 1064                 break;
 1065         case LINUX_P_PGID:
 1066                 if (args->id <= 0)
 1067                         return (EINVAL);
 1068                 idtype = P_PGID;
 1069                 break;
 1070         default:
 1071                 return (EINVAL);
 1072         }
 1073 
 1074         error = kern_wait6(td, idtype, args->id, &status, options,
 1075             &wru, &siginfo);
 1076         if (error != 0)
 1077                 return (error);
 1078         if (args->rusage != NULL) {
 1079                 error = linux_copyout_rusage(&wru.wru_children,
 1080                     args->rusage);
 1081                 if (error != 0)
 1082                         return (error);
 1083         }
 1084         if (args->info != NULL) {
 1085                 p = td->td_proc;
 1086                 if (td->td_retval[0] == 0)
 1087                         bzero(&lsi, sizeof(lsi));
 1088                 else {
 1089                         sig = bsd_to_linux_signal(siginfo.si_signo);
 1090                         siginfo_to_lsiginfo(&siginfo, &lsi, sig);
 1091                 }
 1092                 error = copyout(&lsi, args->info, sizeof(lsi));
 1093         }
 1094         td->td_retval[0] = 0;
 1095 
 1096         return (error);
 1097 }
 1098 
 1099 int
 1100 linux_mknod(struct thread *td, struct linux_mknod_args *args)
 1101 {
 1102         char *path;
 1103         int error;
 1104 
 1105         LCONVPATHCREAT(td, args->path, &path);
 1106 
 1107 #ifdef DEBUG
 1108         if (ldebug(mknod))
 1109                 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
 1110                     (uintmax_t)args->dev);
 1111 #endif
 1112 
 1113         switch (args->mode & S_IFMT) {
 1114         case S_IFIFO:
 1115         case S_IFSOCK:
 1116                 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
 1117                 break;
 1118 
 1119         case S_IFCHR:
 1120         case S_IFBLK:
 1121                 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
 1122                     args->dev);
 1123                 break;
 1124 
 1125         case S_IFDIR:
 1126                 error = EPERM;
 1127                 break;
 1128 
 1129         case 0:
 1130                 args->mode |= S_IFREG;
 1131                 /* FALLTHROUGH */
 1132         case S_IFREG:
 1133                 error = kern_open(td, path, UIO_SYSSPACE,
 1134                     O_WRONLY | O_CREAT | O_TRUNC, args->mode);
 1135                 if (error == 0)
 1136                         kern_close(td, td->td_retval[0]);
 1137                 break;
 1138 
 1139         default:
 1140                 error = EINVAL;
 1141                 break;
 1142         }
 1143         LFREEPATH(path);
 1144         return (error);
 1145 }
 1146 
 1147 int
 1148 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
 1149 {
 1150         char *path;
 1151         int error, dfd;
 1152 
 1153         dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
 1154         LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
 1155 
 1156 #ifdef DEBUG
 1157         if (ldebug(mknodat))
 1158                 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
 1159 #endif
 1160 
 1161         switch (args->mode & S_IFMT) {
 1162         case S_IFIFO:
 1163         case S_IFSOCK:
 1164                 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
 1165                 break;
 1166 
 1167         case S_IFCHR:
 1168         case S_IFBLK:
 1169                 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
 1170                     args->dev);
 1171                 break;
 1172 
 1173         case S_IFDIR:
 1174                 error = EPERM;
 1175                 break;
 1176 
 1177         case 0:
 1178                 args->mode |= S_IFREG;
 1179                 /* FALLTHROUGH */
 1180         case S_IFREG:
 1181                 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
 1182                     O_WRONLY | O_CREAT | O_TRUNC, args->mode);
 1183                 if (error == 0)
 1184                         kern_close(td, td->td_retval[0]);
 1185                 break;
 1186 
 1187         default:
 1188                 error = EINVAL;
 1189                 break;
 1190         }
 1191         LFREEPATH(path);
 1192         return (error);
 1193 }
 1194 
 1195 /*
 1196  * UGH! This is just about the dumbest idea I've ever heard!!
 1197  */
 1198 int
 1199 linux_personality(struct thread *td, struct linux_personality_args *args)
 1200 {
 1201         struct linux_pemuldata *pem;
 1202         struct proc *p = td->td_proc;
 1203         uint32_t old;
 1204 
 1205 #ifdef DEBUG
 1206         if (ldebug(personality))
 1207                 printf(ARGS(personality, "%u"), args->per);
 1208 #endif
 1209 
 1210         PROC_LOCK(p);
 1211         pem = pem_find(p);
 1212         old = pem->persona;
 1213         if (args->per != 0xffffffff)
 1214                 pem->persona = args->per;
 1215         PROC_UNLOCK(p);
 1216 
 1217         td->td_retval[0] = old;
 1218         return (0);
 1219 }
 1220 
 1221 struct l_itimerval {
 1222         l_timeval it_interval;
 1223         l_timeval it_value;
 1224 };
 1225 
 1226 #define B2L_ITIMERVAL(bip, lip)                                         \
 1227         (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;          \
 1228         (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;        \
 1229         (bip)->it_value.tv_sec = (lip)->it_value.tv_sec;                \
 1230         (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
 1231 
 1232 int
 1233 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
 1234 {
 1235         int error;
 1236         struct l_itimerval ls;
 1237         struct itimerval aitv, oitv;
 1238 
 1239 #ifdef DEBUG
 1240         if (ldebug(setitimer))
 1241                 printf(ARGS(setitimer, "%p, %p"),
 1242                     (void *)uap->itv, (void *)uap->oitv);
 1243 #endif
 1244 
 1245         if (uap->itv == NULL) {
 1246                 uap->itv = uap->oitv;
 1247                 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
 1248         }
 1249 
 1250         error = copyin(uap->itv, &ls, sizeof(ls));
 1251         if (error != 0)
 1252                 return (error);
 1253         B2L_ITIMERVAL(&aitv, &ls);
 1254 #ifdef DEBUG
 1255         if (ldebug(setitimer)) {
 1256                 printf("setitimer: value: sec: %jd, usec: %ld\n",
 1257                     (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
 1258                 printf("setitimer: interval: sec: %jd, usec: %ld\n",
 1259                     (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
 1260         }
 1261 #endif
 1262         error = kern_setitimer(td, uap->which, &aitv, &oitv);
 1263         if (error != 0 || uap->oitv == NULL)
 1264                 return (error);
 1265         B2L_ITIMERVAL(&ls, &oitv);
 1266 
 1267         return (copyout(&ls, uap->oitv, sizeof(ls)));
 1268 }
 1269 
 1270 int
 1271 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
 1272 {
 1273         int error;
 1274         struct l_itimerval ls;
 1275         struct itimerval aitv;
 1276 
 1277 #ifdef DEBUG
 1278         if (ldebug(getitimer))
 1279                 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
 1280 #endif
 1281         error = kern_getitimer(td, uap->which, &aitv);
 1282         if (error != 0)
 1283                 return (error);
 1284         B2L_ITIMERVAL(&ls, &aitv);
 1285         return (copyout(&ls, uap->itv, sizeof(ls)));
 1286 }
 1287 
 1288 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
 1289 int
 1290 linux_nice(struct thread *td, struct linux_nice_args *args)
 1291 {
 1292         struct setpriority_args bsd_args;
 1293 
 1294         bsd_args.which = PRIO_PROCESS;
 1295         bsd_args.who = 0;               /* current process */
 1296         bsd_args.prio = args->inc;
 1297         return (sys_setpriority(td, &bsd_args));
 1298 }
 1299 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
 1300 
 1301 int
 1302 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
 1303 {
 1304         struct ucred *newcred, *oldcred;
 1305         l_gid_t *linux_gidset;
 1306         gid_t *bsd_gidset;
 1307         int ngrp, error;
 1308         struct proc *p;
 1309 
 1310         ngrp = args->gidsetsize;
 1311         if (ngrp < 0 || ngrp >= ngroups_max + 1)
 1312                 return (EINVAL);
 1313         linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
 1314         error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
 1315         if (error)
 1316                 goto out;
 1317         newcred = crget();
 1318         crextend(newcred, ngrp + 1);
 1319         p = td->td_proc;
 1320         PROC_LOCK(p);
 1321         oldcred = p->p_ucred;
 1322         crcopy(newcred, oldcred);
 1323 
 1324         /*
 1325          * cr_groups[0] holds egid. Setting the whole set from
 1326          * the supplied set will cause egid to be changed too.
 1327          * Keep cr_groups[0] unchanged to prevent that.
 1328          */
 1329 
 1330         if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
 1331                 PROC_UNLOCK(p);
 1332                 crfree(newcred);
 1333                 goto out;
 1334         }
 1335 
 1336         if (ngrp > 0) {
 1337                 newcred->cr_ngroups = ngrp + 1;
 1338 
 1339                 bsd_gidset = newcred->cr_groups;
 1340                 ngrp--;
 1341                 while (ngrp >= 0) {
 1342                         bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
 1343                         ngrp--;
 1344                 }
 1345         } else
 1346                 newcred->cr_ngroups = 1;
 1347 
 1348         setsugid(p);
 1349         proc_set_cred(p, newcred);
 1350         PROC_UNLOCK(p);
 1351         crfree(oldcred);
 1352         error = 0;
 1353 out:
 1354         free(linux_gidset, M_LINUX);
 1355         return (error);
 1356 }
 1357 
 1358 int
 1359 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
 1360 {
 1361         struct ucred *cred;
 1362         l_gid_t *linux_gidset;
 1363         gid_t *bsd_gidset;
 1364         int bsd_gidsetsz, ngrp, error;
 1365 
 1366         cred = td->td_ucred;
 1367         bsd_gidset = cred->cr_groups;
 1368         bsd_gidsetsz = cred->cr_ngroups - 1;
 1369 
 1370         /*
 1371          * cr_groups[0] holds egid. Returning the whole set
 1372          * here will cause a duplicate. Exclude cr_groups[0]
 1373          * to prevent that.
 1374          */
 1375 
 1376         if ((ngrp = args->gidsetsize) == 0) {
 1377                 td->td_retval[0] = bsd_gidsetsz;
 1378                 return (0);
 1379         }
 1380 
 1381         if (ngrp < bsd_gidsetsz)
 1382                 return (EINVAL);
 1383 
 1384         ngrp = 0;
 1385         linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
 1386             M_LINUX, M_WAITOK);
 1387         while (ngrp < bsd_gidsetsz) {
 1388                 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
 1389                 ngrp++;
 1390         }
 1391 
 1392         error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
 1393         free(linux_gidset, M_LINUX);
 1394         if (error)
 1395                 return (error);
 1396 
 1397         td->td_retval[0] = ngrp;
 1398         return (0);
 1399 }
 1400 
 1401 int
 1402 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
 1403 {
 1404         struct rlimit bsd_rlim;
 1405         struct l_rlimit rlim;
 1406         u_int which;
 1407         int error;
 1408 
 1409 #ifdef DEBUG
 1410         if (ldebug(setrlimit))
 1411                 printf(ARGS(setrlimit, "%d, %p"),
 1412                     args->resource, (void *)args->rlim);
 1413 #endif
 1414 
 1415         if (args->resource >= LINUX_RLIM_NLIMITS)
 1416                 return (EINVAL);
 1417 
 1418         which = linux_to_bsd_resource[args->resource];
 1419         if (which == -1)
 1420                 return (EINVAL);
 1421 
 1422         error = copyin(args->rlim, &rlim, sizeof(rlim));
 1423         if (error)
 1424                 return (error);
 1425 
 1426         bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
 1427         bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
 1428         return (kern_setrlimit(td, which, &bsd_rlim));
 1429 }
 1430 
 1431 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
 1432 int
 1433 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
 1434 {
 1435         struct l_rlimit rlim;
 1436         struct proc *p = td->td_proc;
 1437         struct rlimit bsd_rlim;
 1438         u_int which;
 1439 
 1440 #ifdef DEBUG
 1441         if (ldebug(old_getrlimit))
 1442                 printf(ARGS(old_getrlimit, "%d, %p"),
 1443                     args->resource, (void *)args->rlim);
 1444 #endif
 1445 
 1446         if (args->resource >= LINUX_RLIM_NLIMITS)
 1447                 return (EINVAL);
 1448 
 1449         which = linux_to_bsd_resource[args->resource];
 1450         if (which == -1)
 1451                 return (EINVAL);
 1452 
 1453         PROC_LOCK(p);
 1454         lim_rlimit(p, which, &bsd_rlim);
 1455         PROC_UNLOCK(p);
 1456 
 1457 #ifdef COMPAT_LINUX32
 1458         rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
 1459         if (rlim.rlim_cur == UINT_MAX)
 1460                 rlim.rlim_cur = INT_MAX;
 1461         rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
 1462         if (rlim.rlim_max == UINT_MAX)
 1463                 rlim.rlim_max = INT_MAX;
 1464 #else
 1465         rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
 1466         if (rlim.rlim_cur == ULONG_MAX)
 1467                 rlim.rlim_cur = LONG_MAX;
 1468         rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
 1469         if (rlim.rlim_max == ULONG_MAX)
 1470                 rlim.rlim_max = LONG_MAX;
 1471 #endif
 1472         return (copyout(&rlim, args->rlim, sizeof(rlim)));
 1473 }
 1474 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
 1475 
 1476 int
 1477 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
 1478 {
 1479         struct l_rlimit rlim;
 1480         struct proc *p = td->td_proc;
 1481         struct rlimit bsd_rlim;
 1482         u_int which;
 1483 
 1484 #ifdef DEBUG
 1485         if (ldebug(getrlimit))
 1486                 printf(ARGS(getrlimit, "%d, %p"),
 1487                     args->resource, (void *)args->rlim);
 1488 #endif
 1489 
 1490         if (args->resource >= LINUX_RLIM_NLIMITS)
 1491                 return (EINVAL);
 1492 
 1493         which = linux_to_bsd_resource[args->resource];
 1494         if (which == -1)
 1495                 return (EINVAL);
 1496 
 1497         PROC_LOCK(p);
 1498         lim_rlimit(p, which, &bsd_rlim);
 1499         PROC_UNLOCK(p);
 1500 
 1501         rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
 1502         rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
 1503         return (copyout(&rlim, args->rlim, sizeof(rlim)));
 1504 }
 1505 
 1506 int
 1507 linux_sched_setscheduler(struct thread *td,
 1508     struct linux_sched_setscheduler_args *args)
 1509 {
 1510         struct sched_param sched_param;
 1511         struct thread *tdt;
 1512         int error, policy;
 1513 
 1514 #ifdef DEBUG
 1515         if (ldebug(sched_setscheduler))
 1516                 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
 1517                     args->pid, args->policy, (const void *)args->param);
 1518 #endif
 1519 
 1520         switch (args->policy) {
 1521         case LINUX_SCHED_OTHER:
 1522                 policy = SCHED_OTHER;
 1523                 break;
 1524         case LINUX_SCHED_FIFO:
 1525                 policy = SCHED_FIFO;
 1526                 break;
 1527         case LINUX_SCHED_RR:
 1528                 policy = SCHED_RR;
 1529                 break;
 1530         default:
 1531                 return (EINVAL);
 1532         }
 1533 
 1534         error = copyin(args->param, &sched_param, sizeof(sched_param));
 1535         if (error)
 1536                 return (error);
 1537 
 1538         tdt = linux_tdfind(td, args->pid, -1);
 1539         if (tdt == NULL)
 1540                 return (ESRCH);
 1541 
 1542         error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
 1543         PROC_UNLOCK(tdt->td_proc);
 1544         return (error);
 1545 }
 1546 
 1547 int
 1548 linux_sched_getscheduler(struct thread *td,
 1549     struct linux_sched_getscheduler_args *args)
 1550 {
 1551         struct thread *tdt;
 1552         int error, policy;
 1553 
 1554 #ifdef DEBUG
 1555         if (ldebug(sched_getscheduler))
 1556                 printf(ARGS(sched_getscheduler, "%d"), args->pid);
 1557 #endif
 1558 
 1559         tdt = linux_tdfind(td, args->pid, -1);
 1560         if (tdt == NULL)
 1561                 return (ESRCH);
 1562 
 1563         error = kern_sched_getscheduler(td, tdt, &policy);
 1564         PROC_UNLOCK(tdt->td_proc);
 1565 
 1566         switch (policy) {
 1567         case SCHED_OTHER:
 1568                 td->td_retval[0] = LINUX_SCHED_OTHER;
 1569                 break;
 1570         case SCHED_FIFO:
 1571                 td->td_retval[0] = LINUX_SCHED_FIFO;
 1572                 break;
 1573         case SCHED_RR:
 1574                 td->td_retval[0] = LINUX_SCHED_RR;
 1575                 break;
 1576         }
 1577         return (error);
 1578 }
 1579 
 1580 int
 1581 linux_sched_get_priority_max(struct thread *td,
 1582     struct linux_sched_get_priority_max_args *args)
 1583 {
 1584         struct sched_get_priority_max_args bsd;
 1585 
 1586 #ifdef DEBUG
 1587         if (ldebug(sched_get_priority_max))
 1588                 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
 1589 #endif
 1590 
 1591         switch (args->policy) {
 1592         case LINUX_SCHED_OTHER:
 1593                 bsd.policy = SCHED_OTHER;
 1594                 break;
 1595         case LINUX_SCHED_FIFO:
 1596                 bsd.policy = SCHED_FIFO;
 1597                 break;
 1598         case LINUX_SCHED_RR:
 1599                 bsd.policy = SCHED_RR;
 1600                 break;
 1601         default:
 1602                 return (EINVAL);
 1603         }
 1604         return (sys_sched_get_priority_max(td, &bsd));
 1605 }
 1606 
 1607 int
 1608 linux_sched_get_priority_min(struct thread *td,
 1609     struct linux_sched_get_priority_min_args *args)
 1610 {
 1611         struct sched_get_priority_min_args bsd;
 1612 
 1613 #ifdef DEBUG
 1614         if (ldebug(sched_get_priority_min))
 1615                 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
 1616 #endif
 1617 
 1618         switch (args->policy) {
 1619         case LINUX_SCHED_OTHER:
 1620                 bsd.policy = SCHED_OTHER;
 1621                 break;
 1622         case LINUX_SCHED_FIFO:
 1623                 bsd.policy = SCHED_FIFO;
 1624                 break;
 1625         case LINUX_SCHED_RR:
 1626                 bsd.policy = SCHED_RR;
 1627                 break;
 1628         default:
 1629                 return (EINVAL);
 1630         }
 1631         return (sys_sched_get_priority_min(td, &bsd));
 1632 }
 1633 
 1634 #define REBOOT_CAD_ON   0x89abcdef
 1635 #define REBOOT_CAD_OFF  0
 1636 #define REBOOT_HALT     0xcdef0123
 1637 #define REBOOT_RESTART  0x01234567
 1638 #define REBOOT_RESTART2 0xA1B2C3D4
 1639 #define REBOOT_POWEROFF 0x4321FEDC
 1640 #define REBOOT_MAGIC1   0xfee1dead
 1641 #define REBOOT_MAGIC2   0x28121969
 1642 #define REBOOT_MAGIC2A  0x05121996
 1643 #define REBOOT_MAGIC2B  0x16041998
 1644 
 1645 int
 1646 linux_reboot(struct thread *td, struct linux_reboot_args *args)
 1647 {
 1648         struct reboot_args bsd_args;
 1649 
 1650 #ifdef DEBUG
 1651         if (ldebug(reboot))
 1652                 printf(ARGS(reboot, "0x%x"), args->cmd);
 1653 #endif
 1654 
 1655         if (args->magic1 != REBOOT_MAGIC1)
 1656                 return (EINVAL);
 1657 
 1658         switch (args->magic2) {
 1659         case REBOOT_MAGIC2:
 1660         case REBOOT_MAGIC2A:
 1661         case REBOOT_MAGIC2B:
 1662                 break;
 1663         default:
 1664                 return (EINVAL);
 1665         }
 1666 
 1667         switch (args->cmd) {
 1668         case REBOOT_CAD_ON:
 1669         case REBOOT_CAD_OFF:
 1670                 return (priv_check(td, PRIV_REBOOT));
 1671         case REBOOT_HALT:
 1672                 bsd_args.opt = RB_HALT;
 1673                 break;
 1674         case REBOOT_RESTART:
 1675         case REBOOT_RESTART2:
 1676                 bsd_args.opt = 0;
 1677                 break;
 1678         case REBOOT_POWEROFF:
 1679                 bsd_args.opt = RB_POWEROFF;
 1680                 break;
 1681         default:
 1682                 return (EINVAL);
 1683         }
 1684         return (sys_reboot(td, &bsd_args));
 1685 }
 1686 
 1687 
 1688 /*
 1689  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
 1690  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
 1691  * are assumed to be preserved. The following lightweight syscalls fixes
 1692  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
 1693  *
 1694  * linux_getpid() - MP SAFE
 1695  * linux_getgid() - MP SAFE
 1696  * linux_getuid() - MP SAFE
 1697  */
 1698 
 1699 int
 1700 linux_getpid(struct thread *td, struct linux_getpid_args *args)
 1701 {
 1702 
 1703 #ifdef DEBUG
 1704         if (ldebug(getpid))
 1705                 printf(ARGS(getpid, ""));
 1706 #endif
 1707         td->td_retval[0] = td->td_proc->p_pid;
 1708 
 1709         return (0);
 1710 }
 1711 
 1712 int
 1713 linux_gettid(struct thread *td, struct linux_gettid_args *args)
 1714 {
 1715         struct linux_emuldata *em;
 1716 
 1717 #ifdef DEBUG
 1718         if (ldebug(gettid))
 1719                 printf(ARGS(gettid, ""));
 1720 #endif
 1721 
 1722         em = em_find(td);
 1723         KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
 1724 
 1725         td->td_retval[0] = em->em_tid;
 1726 
 1727         return (0);
 1728 }
 1729 
 1730 
 1731 int
 1732 linux_getppid(struct thread *td, struct linux_getppid_args *args)
 1733 {
 1734 
 1735 #ifdef DEBUG
 1736         if (ldebug(getppid))
 1737                 printf(ARGS(getppid, ""));
 1738 #endif
 1739 
 1740         PROC_LOCK(td->td_proc);
 1741         td->td_retval[0] = td->td_proc->p_pptr->p_pid;
 1742         PROC_UNLOCK(td->td_proc);
 1743         return (0);
 1744 }
 1745 
 1746 int
 1747 linux_getgid(struct thread *td, struct linux_getgid_args *args)
 1748 {
 1749 
 1750 #ifdef DEBUG
 1751         if (ldebug(getgid))
 1752                 printf(ARGS(getgid, ""));
 1753 #endif
 1754 
 1755         td->td_retval[0] = td->td_ucred->cr_rgid;
 1756         return (0);
 1757 }
 1758 
 1759 int
 1760 linux_getuid(struct thread *td, struct linux_getuid_args *args)
 1761 {
 1762 
 1763 #ifdef DEBUG
 1764         if (ldebug(getuid))
 1765                 printf(ARGS(getuid, ""));
 1766 #endif
 1767 
 1768         td->td_retval[0] = td->td_ucred->cr_ruid;
 1769         return (0);
 1770 }
 1771 
 1772 
 1773 int
 1774 linux_getsid(struct thread *td, struct linux_getsid_args *args)
 1775 {
 1776         struct getsid_args bsd;
 1777 
 1778 #ifdef DEBUG
 1779         if (ldebug(getsid))
 1780                 printf(ARGS(getsid, "%i"), args->pid);
 1781 #endif
 1782 
 1783         bsd.pid = args->pid;
 1784         return (sys_getsid(td, &bsd));
 1785 }
 1786 
 1787 int
 1788 linux_nosys(struct thread *td, struct nosys_args *ignore)
 1789 {
 1790 
 1791         return (ENOSYS);
 1792 }
 1793 
 1794 int
 1795 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
 1796 {
 1797         struct getpriority_args bsd_args;
 1798         int error;
 1799 
 1800 #ifdef DEBUG
 1801         if (ldebug(getpriority))
 1802                 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
 1803 #endif
 1804 
 1805         bsd_args.which = args->which;
 1806         bsd_args.who = args->who;
 1807         error = sys_getpriority(td, &bsd_args);
 1808         td->td_retval[0] = 20 - td->td_retval[0];
 1809         return (error);
 1810 }
 1811 
 1812 int
 1813 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
 1814 {
 1815         int name[2];
 1816 
 1817 #ifdef DEBUG
 1818         if (ldebug(sethostname))
 1819                 printf(ARGS(sethostname, "*, %i"), args->len);
 1820 #endif
 1821 
 1822         name[0] = CTL_KERN;
 1823         name[1] = KERN_HOSTNAME;
 1824         return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
 1825             args->len, 0, 0));
 1826 }
 1827 
 1828 int
 1829 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
 1830 {
 1831         int name[2];
 1832 
 1833 #ifdef DEBUG
 1834         if (ldebug(setdomainname))
 1835                 printf(ARGS(setdomainname, "*, %i"), args->len);
 1836 #endif
 1837 
 1838         name[0] = CTL_KERN;
 1839         name[1] = KERN_NISDOMAINNAME;
 1840         return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
 1841             args->len, 0, 0));
 1842 }
 1843 
 1844 int
 1845 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
 1846 {
 1847 
 1848 #ifdef DEBUG
 1849         if (ldebug(exit_group))
 1850                 printf(ARGS(exit_group, "%i"), args->error_code);
 1851 #endif
 1852 
 1853         LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
 1854             args->error_code);
 1855 
 1856         /*
 1857          * XXX: we should send a signal to the parent if
 1858          * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
 1859          * as it doesnt occur often.
 1860          */
 1861         exit1(td, W_EXITCODE(args->error_code, 0));
 1862                 /* NOTREACHED */
 1863 }
 1864 
 1865 #define _LINUX_CAPABILITY_VERSION  0x19980330
 1866 
 1867 struct l_user_cap_header {
 1868         l_int   version;
 1869         l_int   pid;
 1870 };
 1871 
 1872 struct l_user_cap_data {
 1873         l_int   effective;
 1874         l_int   permitted;
 1875         l_int   inheritable;
 1876 };
 1877 
 1878 int
 1879 linux_capget(struct thread *td, struct linux_capget_args *args)
 1880 {
 1881         struct l_user_cap_header luch;
 1882         struct l_user_cap_data lucd;
 1883         int error;
 1884 
 1885         if (args->hdrp == NULL)
 1886                 return (EFAULT);
 1887 
 1888         error = copyin(args->hdrp, &luch, sizeof(luch));
 1889         if (error != 0)
 1890                 return (error);
 1891 
 1892         if (luch.version != _LINUX_CAPABILITY_VERSION) {
 1893                 luch.version = _LINUX_CAPABILITY_VERSION;
 1894                 error = copyout(&luch, args->hdrp, sizeof(luch));
 1895                 if (error)
 1896                         return (error);
 1897                 return (EINVAL);
 1898         }
 1899 
 1900         if (luch.pid)
 1901                 return (EPERM);
 1902 
 1903         if (args->datap) {
 1904                 /*
 1905                  * The current implementation doesn't support setting
 1906                  * a capability (it's essentially a stub) so indicate
 1907                  * that no capabilities are currently set or available
 1908                  * to request.
 1909                  */
 1910                 bzero (&lucd, sizeof(lucd));
 1911                 error = copyout(&lucd, args->datap, sizeof(lucd));
 1912         }
 1913 
 1914         return (error);
 1915 }
 1916 
 1917 int
 1918 linux_capset(struct thread *td, struct linux_capset_args *args)
 1919 {
 1920         struct l_user_cap_header luch;
 1921         struct l_user_cap_data lucd;
 1922         int error;
 1923 
 1924         if (args->hdrp == NULL || args->datap == NULL)
 1925                 return (EFAULT);
 1926 
 1927         error = copyin(args->hdrp, &luch, sizeof(luch));
 1928         if (error != 0)
 1929                 return (error);
 1930 
 1931         if (luch.version != _LINUX_CAPABILITY_VERSION) {
 1932                 luch.version = _LINUX_CAPABILITY_VERSION;
 1933                 error = copyout(&luch, args->hdrp, sizeof(luch));
 1934                 if (error)
 1935                         return (error);
 1936                 return (EINVAL);
 1937         }
 1938 
 1939         if (luch.pid)
 1940                 return (EPERM);
 1941 
 1942         error = copyin(args->datap, &lucd, sizeof(lucd));
 1943         if (error != 0)
 1944                 return (error);
 1945 
 1946         /* We currently don't support setting any capabilities. */
 1947         if (lucd.effective || lucd.permitted || lucd.inheritable) {
 1948                 linux_msg(td,
 1949                           "capset effective=0x%x, permitted=0x%x, "
 1950                           "inheritable=0x%x is not implemented",
 1951                           (int)lucd.effective, (int)lucd.permitted,
 1952                           (int)lucd.inheritable);
 1953                 return (EPERM);
 1954         }
 1955 
 1956         return (0);
 1957 }
 1958 
 1959 int
 1960 linux_prctl(struct thread *td, struct linux_prctl_args *args)
 1961 {
 1962         int error = 0, max_size;
 1963         struct proc *p = td->td_proc;
 1964         char comm[LINUX_MAX_COMM_LEN];
 1965         struct linux_emuldata *em;
 1966         int pdeath_signal;
 1967 
 1968 #ifdef DEBUG
 1969         if (ldebug(prctl))
 1970                 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
 1971                     (uintmax_t)args->arg2, (uintmax_t)args->arg3,
 1972                     (uintmax_t)args->arg4, (uintmax_t)args->arg5);
 1973 #endif
 1974 
 1975         switch (args->option) {
 1976         case LINUX_PR_SET_PDEATHSIG:
 1977                 if (!LINUX_SIG_VALID(args->arg2))
 1978                         return (EINVAL);
 1979                 em = em_find(td);
 1980                 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
 1981                 em->pdeath_signal = args->arg2;
 1982                 break;
 1983         case LINUX_PR_GET_PDEATHSIG:
 1984                 em = em_find(td);
 1985                 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
 1986                 pdeath_signal = em->pdeath_signal;
 1987                 error = copyout(&pdeath_signal,
 1988                     (void *)(register_t)args->arg2,
 1989                     sizeof(pdeath_signal));
 1990                 break;
 1991         case LINUX_PR_GET_KEEPCAPS:
 1992                 /*
 1993                  * Indicate that we always clear the effective and
 1994                  * permitted capability sets when the user id becomes
 1995                  * non-zero (actually the capability sets are simply
 1996                  * always zero in the current implementation).
 1997                  */
 1998                 td->td_retval[0] = 0;
 1999                 break;
 2000         case LINUX_PR_SET_KEEPCAPS:
 2001                 /*
 2002                  * Ignore requests to keep the effective and permitted
 2003                  * capability sets when the user id becomes non-zero.
 2004                  */
 2005                 break;
 2006         case LINUX_PR_SET_NAME:
 2007                 /*
 2008                  * To be on the safe side we need to make sure to not
 2009                  * overflow the size a linux program expects. We already
 2010                  * do this here in the copyin, so that we don't need to
 2011                  * check on copyout.
 2012                  */
 2013                 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
 2014                 error = copyinstr((void *)(register_t)args->arg2, comm,
 2015                     max_size, NULL);
 2016 
 2017                 /* Linux silently truncates the name if it is too long. */
 2018                 if (error == ENAMETOOLONG) {
 2019                         /*
 2020                          * XXX: copyinstr() isn't documented to populate the
 2021                          * array completely, so do a copyin() to be on the
 2022                          * safe side. This should be changed in case
 2023                          * copyinstr() is changed to guarantee this.
 2024                          */
 2025                         error = copyin((void *)(register_t)args->arg2, comm,
 2026                             max_size - 1);
 2027                         comm[max_size - 1] = '\0';
 2028                 }
 2029                 if (error)
 2030                         return (error);
 2031 
 2032                 PROC_LOCK(p);
 2033                 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
 2034                 PROC_UNLOCK(p);
 2035                 break;
 2036         case LINUX_PR_GET_NAME:
 2037                 PROC_LOCK(p);
 2038                 strlcpy(comm, p->p_comm, sizeof(comm));
 2039                 PROC_UNLOCK(p);
 2040                 error = copyout(comm, (void *)(register_t)args->arg2,
 2041                     strlen(comm) + 1);
 2042                 break;
 2043         default:
 2044                 error = EINVAL;
 2045                 break;
 2046         }
 2047 
 2048         return (error);
 2049 }
 2050 
 2051 int
 2052 linux_sched_setparam(struct thread *td,
 2053     struct linux_sched_setparam_args *uap)
 2054 {
 2055         struct sched_param sched_param;
 2056         struct thread *tdt;
 2057         int error;
 2058 
 2059 #ifdef DEBUG
 2060         if (ldebug(sched_setparam))
 2061                 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
 2062 #endif
 2063 
 2064         error = copyin(uap->param, &sched_param, sizeof(sched_param));
 2065         if (error)
 2066                 return (error);
 2067 
 2068         tdt = linux_tdfind(td, uap->pid, -1);
 2069         if (tdt == NULL)
 2070                 return (ESRCH);
 2071 
 2072         error = kern_sched_setparam(td, tdt, &sched_param);
 2073         PROC_UNLOCK(tdt->td_proc);
 2074         return (error);
 2075 }
 2076 
 2077 int
 2078 linux_sched_getparam(struct thread *td,
 2079     struct linux_sched_getparam_args *uap)
 2080 {
 2081         struct sched_param sched_param;
 2082         struct thread *tdt;
 2083         int error;
 2084 
 2085 #ifdef DEBUG
 2086         if (ldebug(sched_getparam))
 2087                 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
 2088 #endif
 2089 
 2090         tdt = linux_tdfind(td, uap->pid, -1);
 2091         if (tdt == NULL)
 2092                 return (ESRCH);
 2093 
 2094         error = kern_sched_getparam(td, tdt, &sched_param);
 2095         PROC_UNLOCK(tdt->td_proc);
 2096         if (error == 0)
 2097                 error = copyout(&sched_param, uap->param,
 2098                     sizeof(sched_param));
 2099         return (error);
 2100 }
 2101 
 2102 /*
 2103  * Get affinity of a process.
 2104  */
 2105 int
 2106 linux_sched_getaffinity(struct thread *td,
 2107     struct linux_sched_getaffinity_args *args)
 2108 {
 2109         int error;
 2110         struct thread *tdt;
 2111         struct cpuset_getaffinity_args cga;
 2112 
 2113 #ifdef DEBUG
 2114         if (ldebug(sched_getaffinity))
 2115                 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
 2116                     args->len);
 2117 #endif
 2118         if (args->len < sizeof(cpuset_t))
 2119                 return (EINVAL);
 2120 
 2121         tdt = linux_tdfind(td, args->pid, -1);
 2122         if (tdt == NULL)
 2123                 return (ESRCH);
 2124 
 2125         PROC_UNLOCK(tdt->td_proc);
 2126         cga.level = CPU_LEVEL_WHICH;
 2127         cga.which = CPU_WHICH_TID;
 2128         cga.id = tdt->td_tid;
 2129         cga.cpusetsize = sizeof(cpuset_t);
 2130         cga.mask = (cpuset_t *) args->user_mask_ptr;
 2131 
 2132         if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
 2133                 td->td_retval[0] = sizeof(cpuset_t);
 2134 
 2135         return (error);
 2136 }
 2137 
 2138 /*
 2139  *  Set affinity of a process.
 2140  */
 2141 int
 2142 linux_sched_setaffinity(struct thread *td,
 2143     struct linux_sched_setaffinity_args *args)
 2144 {
 2145         struct cpuset_setaffinity_args csa;
 2146         struct thread *tdt;
 2147 
 2148 #ifdef DEBUG
 2149         if (ldebug(sched_setaffinity))
 2150                 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
 2151                     args->len);
 2152 #endif
 2153         if (args->len < sizeof(cpuset_t))
 2154                 return (EINVAL);
 2155 
 2156         tdt = linux_tdfind(td, args->pid, -1);
 2157         if (tdt == NULL)
 2158                 return (ESRCH);
 2159 
 2160         PROC_UNLOCK(tdt->td_proc);
 2161         csa.level = CPU_LEVEL_WHICH;
 2162         csa.which = CPU_WHICH_TID;
 2163         csa.id = tdt->td_tid;
 2164         csa.cpusetsize = sizeof(cpuset_t);
 2165         csa.mask = (cpuset_t *) args->user_mask_ptr;
 2166 
 2167         return (sys_cpuset_setaffinity(td, &csa));
 2168 }
 2169 
 2170 struct linux_rlimit64 {
 2171         uint64_t        rlim_cur;
 2172         uint64_t        rlim_max;
 2173 };
 2174 
 2175 int
 2176 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
 2177 {
 2178         struct rlimit rlim, nrlim;
 2179         struct linux_rlimit64 lrlim;
 2180         struct proc *p;
 2181         u_int which;
 2182         int flags;
 2183         int error;
 2184 
 2185 #ifdef DEBUG
 2186         if (ldebug(prlimit64))
 2187                 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
 2188                     args->resource, (void *)args->new, (void *)args->old);
 2189 #endif
 2190 
 2191         if (args->resource >= LINUX_RLIM_NLIMITS)
 2192                 return (EINVAL);
 2193 
 2194         which = linux_to_bsd_resource[args->resource];
 2195         if (which == -1)
 2196                 return (EINVAL);
 2197 
 2198         if (args->new != NULL) {
 2199                 /*
 2200                  * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
 2201                  * rlim is unsigned 64-bit. FreeBSD treats negative limits
 2202                  * as INFINITY so we do not need a conversion even.
 2203                  */
 2204                 error = copyin(args->new, &nrlim, sizeof(nrlim));
 2205                 if (error != 0)
 2206                         return (error);
 2207         }
 2208 
 2209         flags = PGET_HOLD | PGET_NOTWEXIT;
 2210         if (args->new != NULL)
 2211                 flags |= PGET_CANDEBUG;
 2212         else
 2213                 flags |= PGET_CANSEE;
 2214         error = pget(args->pid, flags, &p);
 2215         if (error != 0)
 2216                 return (error);
 2217 
 2218         if (args->old != NULL) {
 2219                 PROC_LOCK(p);
 2220                 lim_rlimit(p, which, &rlim);
 2221                 PROC_UNLOCK(p);
 2222                 if (rlim.rlim_cur == RLIM_INFINITY)
 2223                         lrlim.rlim_cur = LINUX_RLIM_INFINITY;
 2224                 else
 2225                         lrlim.rlim_cur = rlim.rlim_cur;
 2226                 if (rlim.rlim_max == RLIM_INFINITY)
 2227                         lrlim.rlim_max = LINUX_RLIM_INFINITY;
 2228                 else
 2229                         lrlim.rlim_max = rlim.rlim_max;
 2230                 error = copyout(&lrlim, args->old, sizeof(lrlim));
 2231                 if (error != 0)
 2232                         goto out;
 2233         }
 2234 
 2235         if (args->new != NULL)
 2236                 error = kern_proc_setrlimit(td, p, which, &nrlim);
 2237 
 2238  out:
 2239         PRELE(p);
 2240         return (error);
 2241 }
 2242 
 2243 int
 2244 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
 2245 {
 2246         struct timeval utv, tv0, tv1, *tvp;
 2247         struct l_pselect6arg lpse6;
 2248         struct l_timespec lts;
 2249         struct timespec uts;
 2250         l_sigset_t l_ss;
 2251         sigset_t *ssp;
 2252         sigset_t ss;
 2253         int error;
 2254 
 2255         ssp = NULL;
 2256         if (args->sig != NULL) {
 2257                 error = copyin(args->sig, &lpse6, sizeof(lpse6));
 2258                 if (error != 0)
 2259                         return (error);
 2260                 if (lpse6.ss_len != sizeof(l_ss))
 2261                         return (EINVAL);
 2262                 if (lpse6.ss != 0) {
 2263                         error = copyin(PTRIN(lpse6.ss), &l_ss,
 2264                             sizeof(l_ss));
 2265                         if (error != 0)
 2266                                 return (error);
 2267                         linux_to_bsd_sigset(&l_ss, &ss);
 2268                         ssp = &ss;
 2269                 }
 2270         }
 2271 
 2272         /*
 2273          * Currently glibc changes nanosecond number to microsecond.
 2274          * This mean losing precision but for now it is hardly seen.
 2275          */
 2276         if (args->tsp != NULL) {
 2277                 error = copyin(args->tsp, &lts, sizeof(lts));
 2278                 if (error != 0)
 2279                         return (error);
 2280                 error = linux_to_native_timespec(&uts, &lts);
 2281                 if (error != 0)
 2282                         return (error);
 2283 
 2284                 TIMESPEC_TO_TIMEVAL(&utv, &uts);
 2285                 if (itimerfix(&utv))
 2286                         return (EINVAL);
 2287 
 2288                 microtime(&tv0);
 2289                 tvp = &utv;
 2290         } else
 2291                 tvp = NULL;
 2292 
 2293         error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
 2294             args->exceptfds, tvp, ssp, LINUX_NFDBITS);
 2295 
 2296         if (error == 0 && args->tsp != NULL) {
 2297                 if (td->td_retval[0] != 0) {
 2298                         /*
 2299                          * Compute how much time was left of the timeout,
 2300                          * by subtracting the current time and the time
 2301                          * before we started the call, and subtracting
 2302                          * that result from the user-supplied value.
 2303                          */
 2304 
 2305                         microtime(&tv1);
 2306                         timevalsub(&tv1, &tv0);
 2307                         timevalsub(&utv, &tv1);
 2308                         if (utv.tv_sec < 0)
 2309                                 timevalclear(&utv);
 2310                 } else
 2311                         timevalclear(&utv);
 2312 
 2313                 TIMEVAL_TO_TIMESPEC(&utv, &uts);
 2314 
 2315                 native_to_linux_timespec(&lts, &uts);
 2316                 error = copyout(&lts, args->tsp, sizeof(lts));
 2317         }
 2318 
 2319         return (error);
 2320 }
 2321 
 2322 int
 2323 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
 2324 {
 2325         struct timespec ts0, ts1;
 2326         struct l_timespec lts;
 2327         struct timespec uts, *tsp;
 2328         l_sigset_t l_ss;
 2329         sigset_t *ssp;
 2330         sigset_t ss;
 2331         int error;
 2332 
 2333         if (args->sset != NULL) {
 2334                 if (args->ssize != sizeof(l_ss))
 2335                         return (EINVAL);
 2336                 error = copyin(args->sset, &l_ss, sizeof(l_ss));
 2337                 if (error)
 2338                         return (error);
 2339                 linux_to_bsd_sigset(&l_ss, &ss);
 2340                 ssp = &ss;
 2341         } else
 2342                 ssp = NULL;
 2343         if (args->tsp != NULL) {
 2344                 error = copyin(args->tsp, &lts, sizeof(lts));
 2345                 if (error)
 2346                         return (error);
 2347                 error = linux_to_native_timespec(&uts, &lts);
 2348                 if (error != 0)
 2349                         return (error);
 2350 
 2351                 nanotime(&ts0);
 2352                 tsp = &uts;
 2353         } else
 2354                 tsp = NULL;
 2355 
 2356         error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
 2357 
 2358         if (error == 0 && args->tsp != NULL) {
 2359                 if (td->td_retval[0]) {
 2360                         nanotime(&ts1);
 2361                         timespecsub(&ts1, &ts0);
 2362                         timespecsub(&uts, &ts1);
 2363                         if (uts.tv_sec < 0)
 2364                                 timespecclear(&uts);
 2365                 } else
 2366                         timespecclear(&uts);
 2367 
 2368                 native_to_linux_timespec(&lts, &uts);
 2369                 error = copyout(&lts, args->tsp, sizeof(lts));
 2370         }
 2371 
 2372         return (error);
 2373 }
 2374 
 2375 #if defined(DEBUG) || defined(KTR)
 2376 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
 2377 
 2378 #ifdef COMPAT_LINUX32
 2379 #define L_MAXSYSCALL    LINUX32_SYS_MAXSYSCALL
 2380 #else
 2381 #define L_MAXSYSCALL    LINUX_SYS_MAXSYSCALL
 2382 #endif
 2383 
 2384 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
 2385 
 2386 static int
 2387 linux_debug(int syscall, int toggle, int global)
 2388 {
 2389 
 2390         if (global) {
 2391                 char c = toggle ? 0 : 0xff;
 2392 
 2393                 memset(linux_debug_map, c, sizeof(linux_debug_map));
 2394                 return (0);
 2395         }
 2396         if (syscall < 0 || syscall >= L_MAXSYSCALL)
 2397                 return (EINVAL);
 2398         if (toggle)
 2399                 clrbit(linux_debug_map, syscall);
 2400         else
 2401                 setbit(linux_debug_map, syscall);
 2402         return (0);
 2403 }
 2404 #undef L_MAXSYSCALL
 2405 
 2406 /*
 2407  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
 2408  *
 2409  *    E.g.: sysctl linux.debug=21.0
 2410  *
 2411  * As a special case, syscall "all" will apply to all syscalls globally.
 2412  */
 2413 #define LINUX_MAX_DEBUGSTR      16
 2414 int
 2415 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
 2416 {
 2417         char value[LINUX_MAX_DEBUGSTR], *p;
 2418         int error, sysc, toggle;
 2419         int global = 0;
 2420 
 2421         value[0] = '\0';
 2422         error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
 2423         if (error || req->newptr == NULL)
 2424                 return (error);
 2425         for (p = value; *p != '\0' && *p != '.'; p++);
 2426         if (*p == '\0')
 2427                 return (EINVAL);
 2428         *p++ = '\0';
 2429         sysc = strtol(value, NULL, 0);
 2430         toggle = strtol(p, NULL, 0);
 2431         if (strcmp(value, "all") == 0)
 2432                 global = 1;
 2433         error = linux_debug(sysc, toggle, global);
 2434         return (error);
 2435 }
 2436 
 2437 #endif /* DEBUG || KTR */
 2438 
 2439 int
 2440 linux_sched_rr_get_interval(struct thread *td,
 2441     struct linux_sched_rr_get_interval_args *uap)
 2442 {
 2443         struct timespec ts;
 2444         struct l_timespec lts;
 2445         struct thread *tdt;
 2446         int error;
 2447 
 2448         /*
 2449          * According to man in case the invalid pid specified
 2450          * EINVAL should be returned.
 2451          */
 2452         if (uap->pid < 0)
 2453                 return (EINVAL);
 2454 
 2455         tdt = linux_tdfind(td, uap->pid, -1);
 2456         if (tdt == NULL)
 2457                 return (ESRCH);
 2458 
 2459         error = kern_sched_rr_get_interval_td(td, tdt, &ts);
 2460         PROC_UNLOCK(tdt->td_proc);
 2461         if (error != 0)
 2462                 return (error);
 2463         native_to_linux_timespec(&lts, &ts);
 2464         return (copyout(&lts, uap->interval, sizeof(lts)));
 2465 }
 2466 
 2467 /*
 2468  * In case when the Linux thread is the initial thread in
 2469  * the thread group thread id is equal to the process id.
 2470  * Glibc depends on this magic (assert in pthread_getattr_np.c).
 2471  */
 2472 struct thread *
 2473 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
 2474 {
 2475         struct linux_emuldata *em;
 2476         struct thread *tdt;
 2477         struct proc *p;
 2478 
 2479         tdt = NULL;
 2480         if (tid == 0 || tid == td->td_tid) {
 2481                 tdt = td;
 2482                 PROC_LOCK(tdt->td_proc);
 2483         } else if (tid > PID_MAX)
 2484                 tdt = tdfind(tid, pid);
 2485         else {
 2486                 /*
 2487                  * Initial thread where the tid equal to the pid.
 2488                  */
 2489                 p = pfind(tid);
 2490                 if (p != NULL) {
 2491                         if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
 2492                                 /*
 2493                                  * p is not a Linuxulator process.
 2494                                  */
 2495                                 PROC_UNLOCK(p);
 2496                                 return (NULL);
 2497                         }
 2498                         FOREACH_THREAD_IN_PROC(p, tdt) {
 2499                                 em = em_find(tdt);
 2500                                 if (tid == em->em_tid)
 2501                                         return (tdt);
 2502                         }
 2503                         PROC_UNLOCK(p);
 2504                 }
 2505                 return (NULL);
 2506         }
 2507 
 2508         return (tdt);
 2509 }
 2510 
 2511 void
 2512 linux_to_bsd_waitopts(int options, int *bsdopts)
 2513 {
 2514 
 2515         if (options & LINUX_WNOHANG)
 2516                 *bsdopts |= WNOHANG;
 2517         if (options & LINUX_WUNTRACED)
 2518                 *bsdopts |= WUNTRACED;
 2519         if (options & LINUX_WEXITED)
 2520                 *bsdopts |= WEXITED;
 2521         if (options & LINUX_WCONTINUED)
 2522                 *bsdopts |= WCONTINUED;
 2523         if (options & LINUX_WNOWAIT)
 2524                 *bsdopts |= WNOWAIT;
 2525 
 2526         if (options & __WCLONE)
 2527                 *bsdopts |= WLINUXCLONE;
 2528 }

Cache object: a2f2a319cc8abad3ad717ea9fd8dcd41


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.