2 * Copyright (c) 2000 Marcel Moolenaar
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/systm.h>
35 #include <sys/fcntl.h>
36 #include <sys/imgact.h>
38 #include <sys/malloc.h>
40 #include <sys/mutex.h>
44 #include <sys/queue.h>
45 #include <sys/resource.h>
46 #include <sys/resourcevar.h>
47 #include <sys/signalvar.h>
48 #include <sys/syscallsubr.h>
49 #include <sys/sysproto.h>
50 #include <sys/unistd.h>
52 #include <sys/sched.h>
54 #include <machine/frame.h>
55 #include <machine/psl.h>
56 #include <machine/segments.h>
57 #include <machine/sysarch.h>
61 #include <vm/vm_map.h>
63 #include <i386/linux/linux.h>
64 #include <i386/linux/linux_proto.h>
65 #include <compat/linux/linux_ipc.h>
66 #include <compat/linux/linux_signal.h>
67 #include <compat/linux/linux_util.h>
68 #include <compat/linux/linux_emul.h>
70 #include <i386/include/pcb.h> /* needed for pcb definition in linux_set_thread_area */
72 #include "opt_posix.h"
74 extern struct sysentvec elf32_freebsd_sysvec; /* defined in i386/i386/elf_machdep.c */
82 l_uint read_exec_only:1;
83 l_uint limit_in_pages:1;
84 l_uint seg_not_present:1;
88 struct l_old_select_argv {
93 struct l_timeval *timeout;
97 linux_to_bsd_sigaltstack(int lsa)
101 if (lsa & LINUX_SS_DISABLE)
103 if (lsa & LINUX_SS_ONSTACK)
109 bsd_to_linux_sigaltstack(int bsa)
113 if (bsa & SS_DISABLE)
114 lsa |= LINUX_SS_DISABLE;
115 if (bsa & SS_ONSTACK)
116 lsa |= LINUX_SS_ONSTACK;
121 linux_execve(struct thread *td, struct linux_execve_args *args)
125 struct image_args eargs;
127 LCONVPATHEXIST(td, args->path, &newpath);
131 printf(ARGS(execve, "%s"), newpath);
134 error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE,
135 args->argp, args->envp);
136 free(newpath, M_TEMP);
138 error = kern_execve(td, &eargs, NULL);
140 /* linux process can exec fbsd one, dont attempt
141 * to create emuldata for such process using
142 * linux_proc_init, this leads to a panic on KASSERT
143 * because such process has p->p_emuldata == NULL
145 if (td->td_proc->p_sysent == &elf_linux_sysvec)
146 error = linux_proc_init(td, 0, 0);
150 struct l_ipc_kludge {
151 struct l_msgbuf *msgp;
156 linux_ipc(struct thread *td, struct linux_ipc_args *args)
159 switch (args->what & 0xFFFF) {
161 struct linux_semop_args a;
163 a.semid = args->arg1;
165 a.nsops = args->arg2;
166 return (linux_semop(td, &a));
169 struct linux_semget_args a;
172 a.nsems = args->arg2;
173 a.semflg = args->arg3;
174 return (linux_semget(td, &a));
177 struct linux_semctl_args a;
180 a.semid = args->arg1;
181 a.semnum = args->arg2;
183 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
186 return (linux_semctl(td, &a));
189 struct linux_msgsnd_args a;
191 a.msqid = args->arg1;
193 a.msgsz = args->arg2;
194 a.msgflg = args->arg3;
195 return (linux_msgsnd(td, &a));
198 struct linux_msgrcv_args a;
200 a.msqid = args->arg1;
201 a.msgsz = args->arg2;
202 a.msgflg = args->arg3;
203 if ((args->what >> 16) == 0) {
204 struct l_ipc_kludge tmp;
207 if (args->ptr == NULL)
209 error = copyin(args->ptr, &tmp, sizeof(tmp));
213 a.msgtyp = tmp.msgtyp;
216 a.msgtyp = args->arg5;
218 return (linux_msgrcv(td, &a));
221 struct linux_msgget_args a;
224 a.msgflg = args->arg2;
225 return (linux_msgget(td, &a));
228 struct linux_msgctl_args a;
230 a.msqid = args->arg1;
233 return (linux_msgctl(td, &a));
236 struct linux_shmat_args a;
238 a.shmid = args->arg1;
239 a.shmaddr = args->ptr;
240 a.shmflg = args->arg2;
241 a.raddr = (l_ulong *)args->arg3;
242 return (linux_shmat(td, &a));
245 struct linux_shmdt_args a;
247 a.shmaddr = args->ptr;
248 return (linux_shmdt(td, &a));
251 struct linux_shmget_args a;
255 a.shmflg = args->arg3;
256 return (linux_shmget(td, &a));
259 struct linux_shmctl_args a;
261 a.shmid = args->arg1;
264 return (linux_shmctl(td, &a));
274 linux_old_select(struct thread *td, struct linux_old_select_args *args)
276 struct l_old_select_argv linux_args;
277 struct linux_select_args newsel;
281 if (ldebug(old_select))
282 printf(ARGS(old_select, "%p"), args->ptr);
285 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
289 newsel.nfds = linux_args.nfds;
290 newsel.readfds = linux_args.readfds;
291 newsel.writefds = linux_args.writefds;
292 newsel.exceptfds = linux_args.exceptfds;
293 newsel.timeout = linux_args.timeout;
294 return (linux_select(td, &newsel));
298 linux_fork(struct thread *td, struct linux_fork_args *args)
306 printf(ARGS(fork, ""));
309 if ((error = fork1(td, RFFDG | RFPROC | RFSTOPPED, 0, &p2)) != 0)
313 td->td_retval[0] = p2->p_pid;
314 td->td_retval[1] = 0;
317 if (td->td_retval[1] == 1)
318 td->td_retval[0] = 0;
319 error = linux_proc_init(td, td->td_retval[0], 0);
323 td2 = FIRST_THREAD_IN_PROC(p2);
326 * Make this runnable after we are finished with it.
330 sched_add(td2, SRQ_BORING);
337 linux_vfork(struct thread *td, struct linux_vfork_args *args)
345 printf(ARGS(vfork, ""));
348 /* exclude RFPPWAIT */
349 if ((error = fork1(td, RFFDG | RFPROC | RFMEM | RFSTOPPED, 0, &p2)) != 0)
352 td->td_retval[0] = p2->p_pid;
353 td->td_retval[1] = 0;
355 /* Are we the child? */
356 if (td->td_retval[1] == 1)
357 td->td_retval[0] = 0;
358 error = linux_proc_init(td, td->td_retval[0], 0);
363 p2->p_flag |= P_PPWAIT;
366 td2 = FIRST_THREAD_IN_PROC(p2);
369 * Make this runnable after we are finished with it.
373 sched_add(td2, SRQ_BORING);
376 /* wait for the children to exit, ie. emulate vfork */
378 while (p2->p_flag & P_PPWAIT)
379 msleep(td->td_proc, &p2->p_mtx, PWAIT, "ppwait", 0);
386 linux_clone(struct thread *td, struct linux_clone_args *args)
388 int error, ff = RFPROC | RFSTOPPED;
392 struct linux_emuldata *em;
396 printf(ARGS(clone, "flags %x, stack %x, parent tid: %x, child tid: %x"),
397 (unsigned int)args->flags, (unsigned int)args->stack,
398 (unsigned int)args->parent_tidptr, (unsigned int)args->child_tidptr);
402 exit_signal = args->flags & 0x000000ff;
403 if (LINUX_SIG_VALID(exit_signal)) {
404 if (exit_signal <= LINUX_SIGTBLSZ)
406 linux_to_bsd_signal[_SIG_IDX(exit_signal)];
407 } else if (exit_signal != 0)
410 if (args->flags & LINUX_CLONE_VM)
412 if (args->flags & LINUX_CLONE_SIGHAND)
415 * XXX: in linux sharing of fs info (chroot/cwd/umask)
416 * and open files is independant. in fbsd its in one
417 * structure but in reality it doesn't cause any problems
418 * because both of these flags are usually set together.
420 if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
424 * Attempt to detect when linux_clone(2) is used for creating
425 * kernel threads. Unfortunately despite the existence of the
426 * CLONE_THREAD flag, version of linuxthreads package used in
427 * most popular distros as of beginning of 2005 doesn't make
428 * any use of it. Therefore, this detection relies on
429 * empirical observation that linuxthreads sets certain
430 * combination of flags, so that we can make more or less
431 * precise detection and notify the FreeBSD kernel that several
432 * processes are in fact part of the same threading group, so
433 * that special treatment is necessary for signal delivery
434 * between those processes and fd locking.
436 if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS)
439 if (args->flags & LINUX_CLONE_PARENT_SETTID)
440 if (args->parent_tidptr == NULL)
443 error = fork1(td, ff, 0, &p2);
447 if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD)) {
448 sx_xlock(&proctree_lock);
450 proc_reparent(p2, td->td_proc->p_pptr);
452 sx_xunlock(&proctree_lock);
455 /* create the emuldata */
456 error = linux_proc_init(td, p2->p_pid, args->flags);
457 /* reference it - no need to check this */
458 em = em_find(p2, EMUL_DOLOCK);
459 KASSERT(em != NULL, ("clone: emuldata not found.\n"));
462 if (args->flags & LINUX_CLONE_THREAD) {
463 /* XXX: linux mangles pgrp and pptr somehow
464 * I think it might be this but I am not sure.
468 p2->p_pgrp = td->td_proc->p_pgrp;
474 if (args->flags & LINUX_CLONE_CHILD_SETTID)
475 em->child_set_tid = args->child_tidptr;
477 em->child_set_tid = NULL;
479 if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
480 em->child_clear_tid = args->child_tidptr;
482 em->child_clear_tid = NULL;
484 EMUL_UNLOCK(&emul_lock);
486 if (args->flags & LINUX_CLONE_PARENT_SETTID) {
487 error = copyout(&p2->p_pid, args->parent_tidptr, sizeof(p2->p_pid));
489 printf(LMSG("copyout failed!"));
493 p2->p_sigparent = exit_signal;
495 td2 = FIRST_THREAD_IN_PROC(p2);
497 * in a case of stack = NULL we are supposed to COW calling process stack
498 * this is what normal fork() does so we just keep the tf_esp arg intact
501 td2->td_frame->tf_esp = (unsigned int)args->stack;
503 if (args->flags & LINUX_CLONE_SETTLS) {
504 struct l_user_desc info;
507 struct segment_descriptor sd;
509 error = copyin((void *)td->td_frame->tf_esi, &info, sizeof(struct l_user_desc));
511 printf(LMSG("copyin failed!"));
514 idx = info.entry_number;
517 * looks like we're getting the idx we returned
518 * in the set_thread_area() syscall
520 if (idx != 6 && idx != 3) {
521 printf(LMSG("resetting idx!"));
525 /* this doesnt happen in practice */
527 /* we might copy out the entry_number as 3 */
528 info.entry_number = 3;
529 error = copyout(&info, (void *) td->td_frame->tf_esi, sizeof(struct l_user_desc));
531 printf(LMSG("copyout failed!"));
534 a[0] = LINUX_LDT_entry_a(&info);
535 a[1] = LINUX_LDT_entry_b(&info);
537 memcpy(&sd, &a, sizeof(a));
540 printf("Segment created in clone with CLONE_SETTLS: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase,
553 td2->td_pcb->pcb_gsd = sd;
554 td2->td_pcb->pcb_gs = GSEL(GUGS_SEL, SEL_UPL);
560 printf(LMSG("clone: successful rfork to %ld, stack %p sig = %d"),
561 (long)p2->p_pid, args->stack, exit_signal);
563 if (args->flags & LINUX_CLONE_VFORK) {
565 p2->p_flag |= P_PPWAIT;
570 * Make this runnable after we are finished with it.
574 sched_add(td2, SRQ_BORING);
577 td->td_retval[0] = p2->p_pid;
578 td->td_retval[1] = 0;
580 if (args->flags & LINUX_CLONE_VFORK) {
581 /* wait for the children to exit, ie. emulate vfork */
583 while (p2->p_flag & P_PPWAIT)
584 msleep(td->td_proc, &p2->p_mtx, PWAIT, "ppwait", 0);
591 #define STACK_SIZE (2 * 1024 * 1024)
592 #define GUARD_SIZE (4 * PAGE_SIZE)
594 static int linux_mmap_common(struct thread *, struct l_mmap_argv *);
597 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
599 struct l_mmap_argv linux_args;
603 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
604 (void *)args->addr, args->len, args->prot,
605 args->flags, args->fd, args->pgoff);
608 linux_args.addr = args->addr;
609 linux_args.len = args->len;
610 linux_args.prot = args->prot;
611 linux_args.flags = args->flags;
612 linux_args.fd = args->fd;
613 linux_args.pgoff = args->pgoff * PAGE_SIZE;
615 return (linux_mmap_common(td, &linux_args));
619 linux_mmap(struct thread *td, struct linux_mmap_args *args)
622 struct l_mmap_argv linux_args;
624 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
630 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
631 (void *)linux_args.addr, linux_args.len, linux_args.prot,
632 linux_args.flags, linux_args.fd, linux_args.pgoff);
635 return (linux_mmap_common(td, &linux_args));
639 linux_mmap_common(struct thread *td, struct l_mmap_argv *linux_args)
641 struct proc *p = td->td_proc;
642 struct mmap_args /* {
660 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE
662 if (! ((linux_args->flags & LINUX_MAP_SHARED) ^
663 (linux_args->flags & LINUX_MAP_PRIVATE)))
666 if (linux_args->flags & LINUX_MAP_SHARED)
667 bsd_args.flags |= MAP_SHARED;
668 if (linux_args->flags & LINUX_MAP_PRIVATE)
669 bsd_args.flags |= MAP_PRIVATE;
670 if (linux_args->flags & LINUX_MAP_FIXED)
671 bsd_args.flags |= MAP_FIXED;
672 if (linux_args->flags & LINUX_MAP_ANON)
673 bsd_args.flags |= MAP_ANON;
675 bsd_args.flags |= MAP_NOSYNC;
676 if (linux_args->flags & LINUX_MAP_GROWSDOWN)
677 bsd_args.flags |= MAP_STACK;
680 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
681 * on Linux/i386. We do this to ensure maximum compatibility.
682 * Linux/ia64 does the same in i386 emulation mode.
684 bsd_args.prot = linux_args->prot;
685 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
686 bsd_args.prot |= PROT_READ | PROT_EXEC;
688 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
689 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : linux_args->fd;
690 if (bsd_args.fd != -1) {
692 * Linux follows Solaris mmap(2) description:
693 * The file descriptor fildes is opened with
694 * read permission, regardless of the
695 * protection options specified.
698 if ((error = fget(td, bsd_args.fd, &fp)) != 0)
700 if (fp->f_type != DTYPE_VNODE) {
705 /* Linux mmap() just fails for O_WRONLY files */
706 if (!(fp->f_flag & FREAD)) {
714 if (linux_args->flags & LINUX_MAP_GROWSDOWN) {
716 * The linux MAP_GROWSDOWN option does not limit auto
717 * growth of the region. Linux mmap with this option
718 * takes as addr the inital BOS, and as len, the initial
719 * region size. It can then grow down from addr without
720 * limit. However, linux threads has an implicit internal
721 * limit to stack size of STACK_SIZE. Its just not
722 * enforced explicitly in linux. But, here we impose
723 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
724 * region, since we can do this with our mmap.
726 * Our mmap with MAP_STACK takes addr as the maximum
727 * downsize limit on BOS, and as len the max size of
728 * the region. It them maps the top SGROWSIZ bytes,
729 * and auto grows the region down, up to the limit
732 * If we don't use the MAP_STACK option, the effect
733 * of this code is to allocate a stack region of a
734 * fixed size of (STACK_SIZE - GUARD_SIZE).
737 if ((caddr_t)PTRIN(linux_args->addr) + linux_args->len >
738 p->p_vmspace->vm_maxsaddr) {
740 * Some linux apps will attempt to mmap
741 * thread stacks near the top of their
742 * address space. If their TOS is greater
743 * than vm_maxsaddr, vm_map_growstack()
744 * will confuse the thread stack with the
745 * process stack and deliver a SEGV if they
746 * attempt to grow the thread stack past their
747 * current stacksize rlimit. To avoid this,
748 * adjust vm_maxsaddr upwards to reflect
749 * the current stacksize rlimit rather
750 * than the maximum possible stacksize.
751 * It would be better to adjust the
752 * mmap'ed region, but some apps do not check
753 * mmap's return value.
756 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
757 lim_cur(p, RLIMIT_STACK);
761 /* This gives us our maximum stack size */
762 if (linux_args->len > STACK_SIZE - GUARD_SIZE)
763 bsd_args.len = linux_args->len;
765 bsd_args.len = STACK_SIZE - GUARD_SIZE;
768 * This gives us a new BOS. If we're using VM_STACK, then
769 * mmap will just map the top SGROWSIZ bytes, and let
770 * the stack grow down to the limit at BOS. If we're
771 * not using VM_STACK we map the full stack, since we
772 * don't have a way to autogrow it.
774 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr) -
777 bsd_args.addr = (caddr_t)PTRIN(linux_args->addr);
778 bsd_args.len = linux_args->len;
780 bsd_args.pos = linux_args->pgoff;
785 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
787 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot,
788 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
790 error = mmap(td, &bsd_args);
793 printf("-> %s() return: 0x%x (0x%08x)\n",
794 __func__, error, (u_int)td->td_retval[0]);
800 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
802 struct mprotect_args bsd_args;
804 bsd_args.addr = uap->addr;
805 bsd_args.len = uap->len;
806 bsd_args.prot = uap->prot;
807 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
808 bsd_args.prot |= PROT_READ | PROT_EXEC;
809 return (mprotect(td, &bsd_args));
813 linux_pipe(struct thread *td, struct linux_pipe_args *args)
820 printf(ARGS(pipe, "*"));
823 reg_edx = td->td_retval[1];
826 td->td_retval[1] = reg_edx;
830 error = copyout(td->td_retval, args->pipefds, 2*sizeof(int));
832 td->td_retval[1] = reg_edx;
836 td->td_retval[1] = reg_edx;
837 td->td_retval[0] = 0;
842 linux_ioperm(struct thread *td, struct linux_ioperm_args *args)
845 struct i386_ioperm_args iia;
847 iia.start = args->start;
848 iia.length = args->length;
849 iia.enable = args->enable;
851 error = i386_set_ioperm(td, &iia);
857 linux_iopl(struct thread *td, struct linux_iopl_args *args)
861 if (args->level < 0 || args->level > 3)
863 if ((error = priv_check(td, PRIV_IO)) != 0)
865 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
867 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) |
868 (args->level * (PSL_IOPL / 3));
873 linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap)
876 struct i386_ldt_args ldt;
877 struct l_descriptor ld;
878 union descriptor desc;
880 if (uap->ptr == NULL)
884 case 0x00: /* read_ldt */
886 ldt.descs = uap->ptr;
887 ldt.num = uap->bytecount / sizeof(union descriptor);
889 error = i386_get_ldt(td, &ldt);
890 td->td_retval[0] *= sizeof(union descriptor);
893 case 0x01: /* write_ldt */
894 case 0x11: /* write_ldt */
895 if (uap->bytecount != sizeof(ld))
898 error = copyin(uap->ptr, &ld, sizeof(ld));
902 ldt.start = ld.entry_number;
905 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff);
906 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
907 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff);
908 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
909 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
912 desc.sd.sd_p = (ld.seg_not_present ^ 1);
914 desc.sd.sd_def32 = ld.seg_32bit;
915 desc.sd.sd_gran = ld.limit_in_pages;
917 error = i386_set_ldt(td, &ldt, &desc);
925 if (error == EOPNOTSUPP) {
926 printf("linux: modify_ldt needs kernel option USER_LDT\n");
934 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
937 l_sigaction_t act, oact;
941 if (ldebug(sigaction))
942 printf(ARGS(sigaction, "%d, %p, %p"),
943 args->sig, (void *)args->nsa, (void *)args->osa);
946 if (args->nsa != NULL) {
947 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
950 act.lsa_handler = osa.lsa_handler;
951 act.lsa_flags = osa.lsa_flags;
952 act.lsa_restorer = osa.lsa_restorer;
953 LINUX_SIGEMPTYSET(act.lsa_mask);
954 act.lsa_mask.__bits[0] = osa.lsa_mask;
957 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
958 args->osa ? &oact : NULL);
960 if (args->osa != NULL && !error) {
961 osa.lsa_handler = oact.lsa_handler;
962 osa.lsa_flags = oact.lsa_flags;
963 osa.lsa_restorer = oact.lsa_restorer;
964 osa.lsa_mask = oact.lsa_mask.__bits[0];
965 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
972 * Linux has two extra args, restart and oldmask. We dont use these,
973 * but it seems that "restart" is actually a context pointer that
974 * enables the signal to happen with a different register set.
977 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
983 if (ldebug(sigsuspend))
984 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
987 LINUX_SIGEMPTYSET(mask);
988 mask.__bits[0] = args->mask;
989 linux_to_bsd_sigset(&mask, &sigmask);
990 return (kern_sigsuspend(td, sigmask));
994 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
1001 if (ldebug(rt_sigsuspend))
1002 printf(ARGS(rt_sigsuspend, "%p, %d"),
1003 (void *)uap->newset, uap->sigsetsize);
1006 if (uap->sigsetsize != sizeof(l_sigset_t))
1009 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
1013 linux_to_bsd_sigset(&lmask, &sigmask);
1014 return (kern_sigsuspend(td, sigmask));
1018 linux_pause(struct thread *td, struct linux_pause_args *args)
1020 struct proc *p = td->td_proc;
1025 printf(ARGS(pause, ""));
1029 sigmask = td->td_sigmask;
1031 return (kern_sigsuspend(td, sigmask));
1035 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
1042 if (ldebug(sigaltstack))
1043 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
1046 if (uap->uss != NULL) {
1047 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
1051 ss.ss_sp = lss.ss_sp;
1052 ss.ss_size = lss.ss_size;
1053 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
1055 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
1056 (uap->uoss != NULL) ? &oss : NULL);
1057 if (!error && uap->uoss != NULL) {
1058 lss.ss_sp = oss.ss_sp;
1059 lss.ss_size = oss.ss_size;
1060 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
1061 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
1068 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
1070 struct ftruncate_args sa;
1073 if (ldebug(ftruncate64))
1074 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
1075 (intmax_t)args->length);
1080 sa.length = args->length;
1081 return ftruncate(td, &sa);
1085 linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args)
1087 struct l_user_desc info;
1091 struct segment_descriptor sd;
1093 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1098 if (ldebug(set_thread_area))
1099 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"),
1105 info.read_exec_only,
1106 info.limit_in_pages,
1107 info.seg_not_present,
1111 idx = info.entry_number;
1113 * Semantics of linux version: every thread in the system has array of
1114 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This
1115 * syscall loads one of the selected tls decriptors with a value and
1116 * also loads GDT descriptors 6, 7 and 8 with the content of the
1117 * per-thread descriptors.
1119 * Semantics of fbsd version: I think we can ignore that linux has 3
1120 * per-thread descriptors and use just the 1st one. The tls_array[]
1121 * is used only in set/get-thread_area() syscalls and for loading the
1122 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so
1123 * we will load just one.
1125 * XXX: this doesn't work when a user space process tries to use more
1126 * than 1 TLS segment. Comment in the linux sources says wine might do
1131 * we support just GLIBC TLS now
1132 * we should let 3 proceed as well because we use this segment so
1133 * if code does two subsequent calls it should succeed
1135 if (idx != 6 && idx != -1 && idx != 3)
1139 * we have to copy out the GDT entry we use
1140 * FreeBSD uses GDT entry #3 for storing %gs so load that
1142 * XXX: what if a user space program doesn't check this value and tries
1145 idx = info.entry_number = 3;
1146 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1150 if (LINUX_LDT_empty(&info)) {
1154 a[0] = LINUX_LDT_entry_a(&info);
1155 a[1] = LINUX_LDT_entry_b(&info);
1158 memcpy(&sd, &a, sizeof(a));
1160 if (ldebug(set_thread_area))
1161 printf("Segment created in set_thread_area: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase,
1173 /* this is taken from i386 version of cpu_set_user_tls() */
1176 td->td_pcb->pcb_gsd = sd;
1177 PCPU_GET(fsgs_gdt)[1] = sd;
1178 load_gs(GSEL(GUGS_SEL, SEL_UPL));
1185 linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args)
1188 struct l_user_desc info;
1191 struct l_desc_struct desc;
1192 struct segment_descriptor sd;
1195 if (ldebug(get_thread_area))
1196 printf(ARGS(get_thread_area, "%p"), args->desc);
1199 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
1203 idx = info.entry_number;
1204 /* XXX: I am not sure if we want 3 to be allowed too. */
1205 if (idx != 6 && idx != 3)
1210 memset(&info, 0, sizeof(info));
1212 sd = PCPU_GET(fsgs_gdt)[1];
1214 memcpy(&desc, &sd, sizeof(desc));
1216 info.entry_number = idx;
1217 info.base_addr = LINUX_GET_BASE(&desc);
1218 info.limit = LINUX_GET_LIMIT(&desc);
1219 info.seg_32bit = LINUX_GET_32BIT(&desc);
1220 info.contents = LINUX_GET_CONTENTS(&desc);
1221 info.read_exec_only = !LINUX_GET_WRITABLE(&desc);
1222 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc);
1223 info.seg_not_present = !LINUX_GET_PRESENT(&desc);
1224 info.useable = LINUX_GET_USEABLE(&desc);
1226 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
1233 /* copied from kern/kern_time.c */
1235 linux_timer_create(struct thread *td, struct linux_timer_create_args *args)
1237 return ktimer_create(td, (struct ktimer_create_args *) args);
1241 linux_timer_settime(struct thread *td, struct linux_timer_settime_args *args)
1243 return ktimer_settime(td, (struct ktimer_settime_args *) args);
1247 linux_timer_gettime(struct thread *td, struct linux_timer_gettime_args *args)
1249 return ktimer_gettime(td, (struct ktimer_gettime_args *) args);
1253 linux_timer_getoverrun(struct thread *td, struct linux_timer_getoverrun_args *args)
1255 return ktimer_getoverrun(td, (struct ktimer_getoverrun_args *) args);
1259 linux_timer_delete(struct thread *td, struct linux_timer_delete_args *args)
1261 return ktimer_delete(td, (struct ktimer_delete_args *) args);
1264 /* XXX: this wont work with module - convert it */
1266 linux_mq_open(struct thread *td, struct linux_mq_open_args *args)
1268 #ifdef P1003_1B_MQUEUE
1269 return kmq_open(td, (struct kmq_open_args *) args);
1276 linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args)
1278 #ifdef P1003_1B_MQUEUE
1279 return kmq_unlink(td, (struct kmq_unlink_args *) args);
1286 linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args)
1288 #ifdef P1003_1B_MQUEUE
1289 return kmq_timedsend(td, (struct kmq_timedsend_args *) args);
1296 linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args)
1298 #ifdef P1003_1B_MQUEUE
1299 return kmq_timedreceive(td, (struct kmq_timedreceive_args *) args);
1306 linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args)
1308 #ifdef P1003_1B_MQUEUE
1309 return kmq_notify(td, (struct kmq_notify_args *) args);
1316 linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args)
1318 #ifdef P1003_1B_MQUEUE
1319 return kmq_setattr(td, (struct kmq_setattr_args *) args);