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.
28 * $FreeBSD: src/sys/i386/linux/linux_machdep.c,v 1.6.2.4 2001/11/05 19:08:23 marcel Exp $
29 * $DragonFly: src/sys/emulation/linux/i386/linux_machdep.c,v 1.23 2007/07/30 17:41:23 pavalos Exp $
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/imgact.h>
35 #include <sys/kern_syscall.h>
38 #include <sys/nlookup.h>
41 #include <sys/resource.h>
42 #include <sys/resourcevar.h>
43 #include <sys/sysproto.h>
44 #include <sys/unistd.h>
47 #include <machine/frame.h>
48 #include <machine/psl.h>
49 #include <machine/segments.h>
50 #include <machine/sysarch.h>
54 #include <vm/vm_map.h>
57 #include "linux_proto.h"
58 #include "../linux_ipc.h"
59 #include "../linux_signal.h"
60 #include "../linux_util.h"
68 l_uint read_exec_only:1;
69 l_uint limit_in_pages:1;
70 l_uint seg_not_present:1;
74 struct l_old_select_argv {
79 struct l_timeval *timeout;
83 linux_to_bsd_sigaltstack(int lsa)
87 if (lsa & LINUX_SS_DISABLE)
89 if (lsa & LINUX_SS_ONSTACK)
95 bsd_to_linux_sigaltstack(int bsa)
100 lsa |= LINUX_SS_DISABLE;
101 if (bsa & SS_ONSTACK)
102 lsa |= LINUX_SS_ONSTACK;
107 sys_linux_execve(struct linux_execve_args *args)
109 struct nlookupdata nd;
110 struct image_args exec_args;
114 error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
119 kprintf(ARGS(execve, "%s"), path);
121 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
123 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
124 args->argp, args->envp);
127 error = kern_execve(&nd, &exec_args);
131 * The syscall result is returned in registers to the new program.
132 * Linux will register %edx as an atexit function and we must be
133 * sure to set it to 0. XXX
136 args->sysmsg_result64 = 0;
138 exec_free_args(&exec_args);
139 linux_free_path(&path);
142 /* We hit a lethal error condition. Let's die now. */
143 exit1(W_EXITCODE(0, SIGABRT));
150 struct l_ipc_kludge {
151 struct l_msgbuf *msgp;
156 sys_linux_ipc(struct linux_ipc_args *args)
160 switch (args->what & 0xFFFF) {
162 struct linux_semop_args a;
164 a.semid = args->arg1;
166 a.nsops = args->arg2;
167 a.sysmsg_lresult = 0;
168 error = linux_semop(&a);
169 args->sysmsg_lresult = a.sysmsg_lresult;
173 struct linux_semget_args a;
176 a.nsems = args->arg2;
177 a.semflg = args->arg3;
178 a.sysmsg_lresult = 0;
179 error = linux_semget(&a);
180 args->sysmsg_lresult = a.sysmsg_lresult;
184 struct linux_semctl_args a;
187 a.semid = args->arg1;
188 a.semnum = args->arg2;
190 a.sysmsg_lresult = 0;
191 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
194 error = linux_semctl(&a);
195 args->sysmsg_lresult = a.sysmsg_lresult;
199 struct linux_msgsnd_args a;
201 a.msqid = args->arg1;
203 a.msgsz = args->arg2;
204 a.msgflg = args->arg3;
205 a.sysmsg_lresult = 0;
206 error = linux_msgsnd(&a);
207 args->sysmsg_lresult = a.sysmsg_lresult;
211 struct linux_msgrcv_args a;
213 a.msqid = args->arg1;
214 a.msgsz = args->arg2;
215 a.msgflg = args->arg3;
216 a.sysmsg_lresult = 0;
217 if ((args->what >> 16) == 0) {
218 struct l_ipc_kludge tmp;
221 if (args->ptr == NULL)
223 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
227 a.msgtyp = tmp.msgtyp;
230 a.msgtyp = args->arg5;
232 error = linux_msgrcv(&a);
233 args->sysmsg_lresult = a.sysmsg_lresult;
237 struct linux_msgget_args a;
240 a.msgflg = args->arg2;
241 a.sysmsg_lresult = 0;
242 error = linux_msgget(&a);
243 args->sysmsg_lresult = a.sysmsg_lresult;
247 struct linux_msgctl_args a;
249 a.msqid = args->arg1;
252 a.sysmsg_lresult = 0;
253 error = linux_msgctl(&a);
254 args->sysmsg_lresult = a.sysmsg_lresult;
258 struct linux_shmat_args a;
260 a.shmid = args->arg1;
261 a.shmaddr = args->ptr;
262 a.shmflg = args->arg2;
263 a.raddr = (l_ulong *)args->arg3;
264 a.sysmsg_lresult = 0;
265 error = linux_shmat(&a);
266 args->sysmsg_lresult = a.sysmsg_lresult;
270 struct linux_shmdt_args a;
272 a.shmaddr = args->ptr;
273 a.sysmsg_lresult = 0;
274 error = linux_shmdt(&a);
275 args->sysmsg_lresult = a.sysmsg_lresult;
279 struct linux_shmget_args a;
283 a.shmflg = args->arg3;
284 a.sysmsg_lresult = 0;
285 error = linux_shmget(&a);
286 args->sysmsg_lresult = a.sysmsg_lresult;
290 struct linux_shmctl_args a;
292 a.shmid = args->arg1;
295 a.sysmsg_lresult = 0;
296 error = linux_shmctl(&a);
297 args->sysmsg_lresult = a.sysmsg_lresult;
308 sys_linux_old_select(struct linux_old_select_args *args)
310 struct l_old_select_argv linux_args;
311 struct linux_select_args newsel;
315 if (ldebug(old_select))
316 kprintf(ARGS(old_select, "%p"), args->ptr);
319 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
323 newsel.sysmsg_result = 0;
324 newsel.nfds = linux_args.nfds;
325 newsel.readfds = linux_args.readfds;
326 newsel.writefds = linux_args.writefds;
327 newsel.exceptfds = linux_args.exceptfds;
328 newsel.timeout = linux_args.timeout;
329 error = sys_linux_select(&newsel);
330 args->sysmsg_result = newsel.sysmsg_result;
335 sys_linux_fork(struct linux_fork_args *args)
341 kprintf(ARGS(fork, ""));
344 if ((error = sys_fork((struct fork_args *)args)) != 0)
347 if (args->sysmsg_result == 1)
348 args->sysmsg_result = 0;
353 sys_linux_exit_group(struct linux_exit_group_args *args)
355 struct exit_args newargs;
358 newargs.sysmsg_result = 0;
359 newargs.rval = args->rval;
360 error = sys_exit(&newargs);
361 args->sysmsg_result = newargs.sysmsg_result;
366 sys_linux_vfork(struct linux_vfork_args *args)
372 kprintf(ARGS(vfork, ""));
375 if ((error = sys_vfork((struct vfork_args *)args)) != 0)
377 /* Are we the child? */
378 if (args->sysmsg_result == 1)
379 args->sysmsg_result = 0;
383 #define CLONE_VM 0x100
384 #define CLONE_FS 0x200
385 #define CLONE_FILES 0x400
386 #define CLONE_SIGHAND 0x800
387 #define CLONE_PID 0x1000
390 sys_linux_clone(struct linux_clone_args *args)
392 int error, ff = RFPROC;
396 struct rfork_args rf_args;
400 kprintf(ARGS(clone, "flags %x, stack %x"),
401 (unsigned int)args->flags, (unsigned int)args->stack);
402 if (args->flags & CLONE_PID)
403 kprintf(LMSG("CLONE_PID not yet supported"));
410 exit_signal = args->flags & 0x000000ff;
411 if (exit_signal >= LINUX_NSIG)
414 if (exit_signal <= LINUX_SIGTBLSZ)
415 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
417 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
420 if (args->flags & CLONE_VM)
422 if (args->flags & CLONE_SIGHAND)
424 if (!(args->flags & CLONE_FILES))
431 rf_args.sysmsg_result = 0;
432 if ((error = sys_rfork(&rf_args)) != 0)
434 args->sysmsg_result = rf_args.sysmsg_result;
436 p2 = pfind(rf_args.sysmsg_result);
440 p2->p_sigparent = exit_signal;
441 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
442 (unsigned int)args->stack;
446 kprintf(LMSG("clone: successful rfork to %ld"),
463 #define STACK_SIZE (2 * 1024 * 1024)
464 #define GUARD_SIZE (4 * PAGE_SIZE)
467 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
468 int linux_flags, int linux_fd, off_t pos, void **res)
470 struct thread *td = curthread;
471 struct proc *p = td->td_proc;
474 int error, flags, len, prot, fd;
477 if (linux_flags & LINUX_MAP_SHARED)
479 if (linux_flags & LINUX_MAP_PRIVATE)
480 flags |= MAP_PRIVATE;
481 if (linux_flags & LINUX_MAP_FIXED)
483 if (linux_flags & LINUX_MAP_ANON) {
488 if (linux_flags & LINUX_MAP_GROWSDOWN) {
490 /* The linux MAP_GROWSDOWN option does not limit auto
491 * growth of the region. Linux mmap with this option
492 * takes as addr the inital BOS, and as len, the initial
493 * region size. It can then grow down from addr without
494 * limit. However, linux threads has an implicit internal
495 * limit to stack size of STACK_SIZE. Its just not
496 * enforced explicitly in linux. But, here we impose
497 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
498 * region, since we can do this with our mmap.
500 * Our mmap with MAP_STACK takes addr as the maximum
501 * downsize limit on BOS, and as len the max size of
502 * the region. It them maps the top SGROWSIZ bytes,
503 * and autgrows the region down, up to the limit
506 * If we don't use the MAP_STACK option, the effect
507 * of this code is to allocate a stack region of a
508 * fixed size of (STACK_SIZE - GUARD_SIZE).
511 /* This gives us TOS */
512 addr = linux_addr + linux_len;
514 if (addr > p->p_vmspace->vm_maxsaddr) {
515 /* Some linux apps will attempt to mmap
516 * thread stacks near the top of their
517 * address space. If their TOS is greater
518 * than vm_maxsaddr, vm_map_growstack()
519 * will confuse the thread stack with the
520 * process stack and deliver a SEGV if they
521 * attempt to grow the thread stack past their
522 * current stacksize rlimit. To avoid this,
523 * adjust vm_maxsaddr upwards to reflect
524 * the current stacksize rlimit rather
525 * than the maximum possible stacksize.
526 * It would be better to adjust the
527 * mmap'ed region, but some apps do not check
528 * mmap's return value.
530 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
531 p->p_rlimit[RLIMIT_STACK].rlim_cur;
534 /* This gives us our maximum stack size */
535 if (linux_len > STACK_SIZE - GUARD_SIZE) {
538 len = STACK_SIZE - GUARD_SIZE;
540 /* This gives us a new BOS. If we're using VM_STACK, then
541 * mmap will just map the top SGROWSIZ bytes, and let
542 * the stack grow down to the limit at BOS. If we're
543 * not using VM_STACK we map the full stack, since we
544 * don't have a way to autogrow it.
552 prot = linux_prot | PROT_READ;
553 if (linux_flags & LINUX_MAP_ANON) {
560 if (ldebug(mmap) || ldebug(mmap2))
561 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
562 addr, len, prot, flags, fd, pos);
564 error = kern_mmap(curproc->p_vmspace, addr, len,
565 prot, flags, fd, pos, &new);
573 sys_linux_mmap(struct linux_mmap_args *args)
575 struct l_mmap_argv linux_args;
578 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
584 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
585 (void *)linux_args.addr, linux_args.len, linux_args.prot,
586 linux_args.flags, linux_args.fd, linux_args.pos);
588 error = linux_mmap_common(linux_args.addr, linux_args.len,
589 linux_args.prot, linux_args.flags, linux_args.fd,
590 linux_args.pos, &args->sysmsg_resultp);
593 kprintf("-> %p\n", args->sysmsg_resultp);
599 sys_linux_mmap2(struct linux_mmap2_args *args)
605 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
606 (void *)args->addr, args->len, args->prot, args->flags,
607 args->fd, args->pgoff);
609 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
610 args->flags, args->fd, args->pgoff * PAGE_SIZE,
611 &args->sysmsg_resultp);
614 kprintf("-> %p\n", args->sysmsg_resultp);
620 sys_linux_pipe(struct linux_pipe_args *args)
624 struct pipe_args bsd_args;
628 kprintf(ARGS(pipe, "*"));
631 reg_edx = args->sysmsg_fds[1];
632 error = sys_pipe(&bsd_args);
634 args->sysmsg_fds[1] = reg_edx;
638 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
640 args->sysmsg_fds[1] = reg_edx;
644 args->sysmsg_fds[1] = reg_edx;
645 args->sysmsg_fds[0] = 0;
650 sys_linux_ioperm(struct linux_ioperm_args *args)
652 struct sysarch_args sa;
653 struct i386_ioperm_args *iia;
657 sg = stackgap_init();
658 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
659 iia->start = args->start;
660 iia->length = args->length;
661 iia->enable = args->enable;
662 sa.sysmsg_resultp = NULL;
663 sa.op = I386_SET_IOPERM;
664 sa.parms = (char *)iia;
665 error = sys_sysarch(&sa);
666 args->sysmsg_resultp = sa.sysmsg_resultp;
671 sys_linux_iopl(struct linux_iopl_args *args)
673 struct thread *td = curthread;
674 struct lwp *lp = td->td_lwp;
679 if (args->level < 0 || args->level > 3)
681 if ((error = priv_check(td, PRIV_ROOT)) != 0)
685 lp->lwp_md.md_regs->tf_eflags =
686 (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
687 (args->level * (PSL_IOPL / 3));
692 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
696 struct sysarch_args args;
697 struct i386_ldt_args *ldt;
698 struct l_descriptor ld;
699 union descriptor *desc;
701 sg = stackgap_init();
703 if (uap->ptr == NULL)
707 case 0x00: /* read_ldt */
708 ldt = stackgap_alloc(&sg, sizeof(*ldt));
710 ldt->descs = uap->ptr;
711 ldt->num = uap->bytecount / sizeof(union descriptor);
712 args.op = I386_GET_LDT;
713 args.parms = (char*)ldt;
714 args.sysmsg_result = 0;
715 error = sys_sysarch(&args);
716 uap->sysmsg_result = args.sysmsg_result *
717 sizeof(union descriptor);
719 case 0x01: /* write_ldt */
720 case 0x11: /* write_ldt */
721 if (uap->bytecount != sizeof(ld))
724 error = copyin(uap->ptr, &ld, sizeof(ld));
728 ldt = stackgap_alloc(&sg, sizeof(*ldt));
729 desc = stackgap_alloc(&sg, sizeof(*desc));
730 ldt->start = ld.entry_number;
733 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
734 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
735 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
736 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
737 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
740 desc->sd.sd_p = (ld.seg_not_present ^ 1);
742 desc->sd.sd_def32 = ld.seg_32bit;
743 desc->sd.sd_gran = ld.limit_in_pages;
744 args.op = I386_SET_LDT;
745 args.parms = (char*)ldt;
746 args.sysmsg_result = 0;
747 error = sys_sysarch(&args);
748 uap->sysmsg_result = args.sysmsg_result;
759 sys_linux_sigaction(struct linux_sigaction_args *args)
762 l_sigaction_t linux_act, linux_oact;
763 struct sigaction act, oact;
767 if (ldebug(sigaction))
768 kprintf(ARGS(sigaction, "%d, %p, %p"),
769 args->sig, (void *)args->nsa, (void *)args->osa);
773 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
776 linux_act.lsa_handler = osa.lsa_handler;
777 linux_act.lsa_flags = osa.lsa_flags;
778 linux_act.lsa_restorer = osa.lsa_restorer;
779 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
780 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
781 linux_to_bsd_sigaction(&linux_act, &act);
784 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
785 args->osa ? &oact : NULL);
787 if (args->osa != NULL && !error) {
788 bsd_to_linux_sigaction(&oact, &linux_oact);
789 osa.lsa_handler = linux_oact.lsa_handler;
790 osa.lsa_flags = linux_oact.lsa_flags;
791 osa.lsa_restorer = linux_oact.lsa_restorer;
792 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
793 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
799 * Linux has two extra args, restart and oldmask. We dont use these,
800 * but it seems that "restart" is actually a context pointer that
801 * enables the signal to happen with a different register set.
804 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
806 l_sigset_t linux_mask;
811 if (ldebug(sigsuspend))
812 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
815 LINUX_SIGEMPTYSET(mask);
816 mask.__bits[0] = args->mask;
817 linux_to_bsd_sigset(&linux_mask, &mask);
819 error = kern_sigsuspend(&mask);
825 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
827 l_sigset_t linux_mask;
832 if (ldebug(rt_sigsuspend))
833 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
834 (void *)uap->newset, uap->sigsetsize);
837 if (uap->sigsetsize != sizeof(l_sigset_t))
840 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
844 linux_to_bsd_sigset(&linux_mask, &mask);
846 error = kern_sigsuspend(&mask);
852 sys_linux_pause(struct linux_pause_args *args)
854 struct thread *td = curthread;
855 struct lwp *lp = td->td_lwp;
861 kprintf(ARGS(pause, ""));
864 mask = lp->lwp_sigmask;
866 error = kern_sigsuspend(&mask);
872 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
879 if (ldebug(sigaltstack))
880 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
884 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
888 ss.ss_sp = linux_ss.ss_sp;
889 ss.ss_size = linux_ss.ss_size;
890 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
893 error = kern_sigaltstack(uap->uss ? &ss : NULL,
894 uap->uoss ? &oss : NULL);
896 if (error == 0 && uap->uoss) {
897 linux_ss.ss_sp = oss.ss_sp;
898 linux_ss.ss_size = oss.ss_size;
899 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
900 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));