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 1.15 2004/11/12 00:09:19 dillon 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>
40 #include <sys/resource.h>
41 #include <sys/resourcevar.h>
42 #include <sys/sysproto.h>
43 #include <sys/unistd.h>
45 #include <machine/frame.h>
46 #include <machine/psl.h>
47 #include <machine/segments.h>
48 #include <machine/sysarch.h>
52 #include <vm/vm_map.h>
55 #include "linux_proto.h"
56 #include "../linux_ipc.h"
57 #include "../linux_signal.h"
58 #include "../linux_util.h"
66 l_uint read_exec_only:1;
67 l_uint limit_in_pages:1;
68 l_uint seg_not_present:1;
72 struct l_old_select_argv {
77 struct l_timeval *timeout;
81 linux_to_bsd_sigaltstack(int lsa)
85 if (lsa & LINUX_SS_DISABLE)
87 if (lsa & LINUX_SS_ONSTACK)
93 bsd_to_linux_sigaltstack(int bsa)
98 lsa |= LINUX_SS_DISABLE;
100 lsa |= LINUX_SS_ONSTACK;
105 linux_execve(struct linux_execve_args *args)
107 struct nlookupdata nd;
108 struct image_args exec_args;
112 error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
117 printf(ARGS(execve, "%s"), path);
119 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
121 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
122 args->argp, args->envp);
125 error = kern_execve(&nd, &exec_args);
129 * The syscall result is returned in registers to the new program.
130 * Linux will register %edx as an atexit function and we must be
131 * sure to set it to 0. XXX
134 args->sysmsg_result64 = 0;
136 exec_free_args(&exec_args);
137 linux_free_path(&path);
141 struct l_ipc_kludge {
142 struct l_msgbuf *msgp;
147 linux_ipc(struct linux_ipc_args *args)
151 switch (args->what & 0xFFFF) {
153 struct linux_semop_args a;
155 a.semid = args->arg1;
157 a.nsops = args->arg2;
158 a.sysmsg_lresult = 0;
159 error = linux_semop(&a);
160 args->sysmsg_lresult = a.sysmsg_lresult;
164 struct linux_semget_args a;
167 a.nsems = args->arg2;
168 a.semflg = args->arg3;
169 a.sysmsg_lresult = 0;
170 error = linux_semget(&a);
171 args->sysmsg_lresult = a.sysmsg_lresult;
175 struct linux_semctl_args a;
178 a.semid = args->arg1;
179 a.semnum = args->arg2;
181 a.sysmsg_lresult = 0;
182 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
185 error = linux_semctl(&a);
186 args->sysmsg_lresult = a.sysmsg_lresult;
190 struct linux_msgsnd_args a;
192 a.msqid = args->arg1;
194 a.msgsz = args->arg2;
195 a.msgflg = args->arg3;
196 a.sysmsg_lresult = 0;
197 error = linux_msgsnd(&a);
198 args->sysmsg_lresult = a.sysmsg_lresult;
202 struct linux_msgrcv_args a;
204 a.msqid = args->arg1;
205 a.msgsz = args->arg2;
206 a.msgflg = args->arg3;
207 a.sysmsg_lresult = 0;
208 if ((args->what >> 16) == 0) {
209 struct l_ipc_kludge tmp;
212 if (args->ptr == NULL)
214 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
218 a.msgtyp = tmp.msgtyp;
221 a.msgtyp = args->arg5;
223 error = linux_msgrcv(&a);
224 args->sysmsg_lresult = a.sysmsg_lresult;
228 struct linux_msgget_args a;
231 a.msgflg = args->arg2;
232 a.sysmsg_lresult = 0;
233 error = linux_msgget(&a);
234 args->sysmsg_lresult = a.sysmsg_lresult;
238 struct linux_msgctl_args a;
240 a.msqid = args->arg1;
243 a.sysmsg_lresult = 0;
244 error = linux_msgctl(&a);
245 args->sysmsg_lresult = a.sysmsg_lresult;
249 struct linux_shmat_args a;
251 a.shmid = args->arg1;
252 a.shmaddr = args->ptr;
253 a.shmflg = args->arg2;
254 a.raddr = (l_ulong *)args->arg3;
255 a.sysmsg_lresult = 0;
256 error = linux_shmat(&a);
257 args->sysmsg_lresult = a.sysmsg_lresult;
261 struct linux_shmdt_args a;
263 a.shmaddr = args->ptr;
264 a.sysmsg_lresult = 0;
265 error = linux_shmdt(&a);
266 args->sysmsg_lresult = a.sysmsg_lresult;
270 struct linux_shmget_args a;
274 a.shmflg = args->arg3;
275 a.sysmsg_lresult = 0;
276 error = linux_shmget(&a);
277 args->sysmsg_lresult = a.sysmsg_lresult;
281 struct linux_shmctl_args a;
283 a.shmid = args->arg1;
286 a.sysmsg_lresult = 0;
287 error = linux_shmctl(&a);
288 args->sysmsg_lresult = a.sysmsg_lresult;
299 linux_old_select(struct linux_old_select_args *args)
301 struct l_old_select_argv linux_args;
302 struct linux_select_args newsel;
306 if (ldebug(old_select))
307 printf(ARGS(old_select, "%p"), args->ptr);
310 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
314 newsel.sysmsg_result = 0;
315 newsel.nfds = linux_args.nfds;
316 newsel.readfds = linux_args.readfds;
317 newsel.writefds = linux_args.writefds;
318 newsel.exceptfds = linux_args.exceptfds;
319 newsel.timeout = linux_args.timeout;
320 error = linux_select(&newsel);
321 args->sysmsg_result = newsel.sysmsg_result;
326 linux_fork(struct linux_fork_args *args)
332 printf(ARGS(fork, ""));
335 if ((error = fork((struct fork_args *)args)) != 0)
338 if (args->sysmsg_result == 1)
339 args->sysmsg_result = 0;
344 linux_vfork(struct linux_vfork_args *args)
350 printf(ARGS(vfork, ""));
353 if ((error = vfork((struct vfork_args *)args)) != 0)
355 /* Are we the child? */
356 if (args->sysmsg_result == 1)
357 args->sysmsg_result = 0;
361 #define CLONE_VM 0x100
362 #define CLONE_FS 0x200
363 #define CLONE_FILES 0x400
364 #define CLONE_SIGHAND 0x800
365 #define CLONE_PID 0x1000
368 linux_clone(struct linux_clone_args *args)
370 int error, ff = RFPROC;
374 struct rfork_args rf_args;
378 printf(ARGS(clone, "flags %x, stack %x"),
379 (unsigned int)args->flags, (unsigned int)args->stack);
380 if (args->flags & CLONE_PID)
381 printf(LMSG("CLONE_PID not yet supported"));
388 exit_signal = args->flags & 0x000000ff;
389 if (exit_signal >= LINUX_NSIG)
392 if (exit_signal <= LINUX_SIGTBLSZ)
393 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
395 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
398 if (args->flags & CLONE_VM)
400 if (args->flags & CLONE_SIGHAND)
402 if (!(args->flags & CLONE_FILES))
409 rf_args.sysmsg_result = 0;
410 if ((error = rfork(&rf_args)) != 0)
412 args->sysmsg_result = rf_args.sysmsg_result;
414 p2 = pfind(rf_args.sysmsg_result);
418 p2->p_sigparent = exit_signal;
419 p2->p_md.md_regs->tf_esp = (unsigned int)args->stack;
423 printf(LMSG("clone: successful rfork to %ld"),
440 #define STACK_SIZE (2 * 1024 * 1024)
441 #define GUARD_SIZE (4 * PAGE_SIZE)
444 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
445 int linux_flags, int linux_fd, off_t pos, void **res)
447 struct thread *td = curthread;
448 struct proc *p = td->td_proc;
451 int error, flags, len, prot, fd;
454 if (linux_flags & LINUX_MAP_SHARED)
456 if (linux_flags & LINUX_MAP_PRIVATE)
457 flags |= MAP_PRIVATE;
458 if (linux_flags & LINUX_MAP_FIXED)
460 if (linux_flags & LINUX_MAP_ANON) {
465 if (linux_flags & LINUX_MAP_GROWSDOWN) {
467 /* The linux MAP_GROWSDOWN option does not limit auto
468 * growth of the region. Linux mmap with this option
469 * takes as addr the inital BOS, and as len, the initial
470 * region size. It can then grow down from addr without
471 * limit. However, linux threads has an implicit internal
472 * limit to stack size of STACK_SIZE. Its just not
473 * enforced explicitly in linux. But, here we impose
474 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
475 * region, since we can do this with our mmap.
477 * Our mmap with MAP_STACK takes addr as the maximum
478 * downsize limit on BOS, and as len the max size of
479 * the region. It them maps the top SGROWSIZ bytes,
480 * and autgrows the region down, up to the limit
483 * If we don't use the MAP_STACK option, the effect
484 * of this code is to allocate a stack region of a
485 * fixed size of (STACK_SIZE - GUARD_SIZE).
488 /* This gives us TOS */
489 addr = linux_addr + linux_len;
491 if (addr > p->p_vmspace->vm_maxsaddr) {
492 /* Some linux apps will attempt to mmap
493 * thread stacks near the top of their
494 * address space. If their TOS is greater
495 * than vm_maxsaddr, vm_map_growstack()
496 * will confuse the thread stack with the
497 * process stack and deliver a SEGV if they
498 * attempt to grow the thread stack past their
499 * current stacksize rlimit. To avoid this,
500 * adjust vm_maxsaddr upwards to reflect
501 * the current stacksize rlimit rather
502 * than the maximum possible stacksize.
503 * It would be better to adjust the
504 * mmap'ed region, but some apps do not check
505 * mmap's return value.
507 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
508 p->p_rlimit[RLIMIT_STACK].rlim_cur;
511 /* This gives us our maximum stack size */
512 if (linux_len > STACK_SIZE - GUARD_SIZE) {
515 len = STACK_SIZE - GUARD_SIZE;
517 /* This gives us a new BOS. If we're using VM_STACK, then
518 * mmap will just map the top SGROWSIZ bytes, and let
519 * the stack grow down to the limit at BOS. If we're
520 * not using VM_STACK we map the full stack, since we
521 * don't have a way to autogrow it.
529 prot = linux_prot | PROT_READ;
530 if (linux_flags & LINUX_MAP_ANON) {
537 if (ldebug(mmap) || ldebug(mmap2))
538 printf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
539 addr, len, prot, flags, fd, pos);
541 error = kern_mmap(addr, len, prot, flags, fd, pos, &new);
549 linux_mmap(struct linux_mmap_args *args)
551 struct l_mmap_argv linux_args;
554 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
560 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
561 (void *)linux_args.addr, linux_args.len, linux_args.prot,
562 linux_args.flags, linux_args.fd, linux_args.pos);
564 error = linux_mmap_common(linux_args.addr, linux_args.len,
565 linux_args.prot, linux_args.flags, linux_args.fd,
566 linux_args.pos, &args->sysmsg_resultp);
569 printf("-> %p\n", args->sysmsg_resultp);
575 linux_mmap2(struct linux_mmap2_args *args)
581 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
582 (void *)args->addr, args->len, args->prot, args->flags,
583 args->fd, args->pgoff);
585 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
586 args->flags, args->fd, args->pgoff * PAGE_SIZE,
587 &args->sysmsg_resultp);
590 printf("-> %p\n", args->sysmsg_resultp);
596 linux_pipe(struct linux_pipe_args *args)
600 struct pipe_args bsd_args;
604 printf(ARGS(pipe, "*"));
607 reg_edx = args->sysmsg_fds[1];
608 error = pipe(&bsd_args);
610 args->sysmsg_fds[1] = reg_edx;
614 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
616 args->sysmsg_fds[1] = reg_edx;
620 args->sysmsg_fds[1] = reg_edx;
621 args->sysmsg_fds[0] = 0;
626 linux_ioperm(struct linux_ioperm_args *args)
628 struct sysarch_args sa;
629 struct i386_ioperm_args *iia;
633 sg = stackgap_init();
634 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
635 iia->start = args->start;
636 iia->length = args->length;
637 iia->enable = args->enable;
638 sa.sysmsg_resultp = NULL;
639 sa.op = I386_SET_IOPERM;
640 sa.parms = (char *)iia;
641 error = sysarch(&sa);
642 args->sysmsg_resultp = sa.sysmsg_resultp;
647 linux_iopl(struct linux_iopl_args *args)
649 struct thread *td = curthread;
650 struct proc *p = td->td_proc;
655 if (args->level < 0 || args->level > 3)
657 if ((error = suser(td)) != 0)
661 p->p_md.md_regs->tf_eflags = (p->p_md.md_regs->tf_eflags & ~PSL_IOPL) |
662 (args->level * (PSL_IOPL / 3));
667 linux_modify_ldt(struct linux_modify_ldt_args *uap)
671 struct sysarch_args args;
672 struct i386_ldt_args *ldt;
673 struct l_descriptor ld;
674 union descriptor *desc;
676 sg = stackgap_init();
678 if (uap->ptr == NULL)
682 case 0x00: /* read_ldt */
683 ldt = stackgap_alloc(&sg, sizeof(*ldt));
685 ldt->descs = uap->ptr;
686 ldt->num = uap->bytecount / sizeof(union descriptor);
687 args.op = I386_GET_LDT;
688 args.parms = (char*)ldt;
689 args.sysmsg_result = 0;
690 error = sysarch(&args);
691 uap->sysmsg_result = args.sysmsg_result *
692 sizeof(union descriptor);
694 case 0x01: /* write_ldt */
695 case 0x11: /* write_ldt */
696 if (uap->bytecount != sizeof(ld))
699 error = copyin(uap->ptr, &ld, sizeof(ld));
703 ldt = stackgap_alloc(&sg, sizeof(*ldt));
704 desc = stackgap_alloc(&sg, sizeof(*desc));
705 ldt->start = ld.entry_number;
708 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
709 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
710 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
711 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
712 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
715 desc->sd.sd_p = (ld.seg_not_present ^ 1);
717 desc->sd.sd_def32 = ld.seg_32bit;
718 desc->sd.sd_gran = ld.limit_in_pages;
719 args.op = I386_SET_LDT;
720 args.parms = (char*)ldt;
721 args.sysmsg_result = 0;
722 error = sysarch(&args);
723 uap->sysmsg_result = args.sysmsg_result;
734 linux_sigaction(struct linux_sigaction_args *args)
737 l_sigaction_t linux_act, linux_oact;
738 struct sigaction act, oact;
742 if (ldebug(sigaction))
743 printf(ARGS(sigaction, "%d, %p, %p"),
744 args->sig, (void *)args->nsa, (void *)args->osa);
748 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
751 linux_act.lsa_handler = osa.lsa_handler;
752 linux_act.lsa_flags = osa.lsa_flags;
753 linux_act.lsa_restorer = osa.lsa_restorer;
754 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
755 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
756 linux_to_bsd_sigaction(&linux_act, &act);
759 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
760 args->osa ? &oact : NULL);
762 if (args->osa != NULL && !error) {
763 bsd_to_linux_sigaction(&oact, &linux_oact);
764 osa.lsa_handler = linux_oact.lsa_handler;
765 osa.lsa_flags = linux_oact.lsa_flags;
766 osa.lsa_restorer = linux_oact.lsa_restorer;
767 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
768 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
774 * Linux has two extra args, restart and oldmask. We dont use these,
775 * but it seems that "restart" is actually a context pointer that
776 * enables the signal to happen with a different register set.
779 linux_sigsuspend(struct linux_sigsuspend_args *args)
781 l_sigset_t linux_mask;
786 if (ldebug(sigsuspend))
787 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
790 LINUX_SIGEMPTYSET(mask);
791 mask.__bits[0] = args->mask;
792 linux_to_bsd_sigset(&linux_mask, &mask);
794 error = kern_sigsuspend(&mask);
800 linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
802 l_sigset_t linux_mask;
807 if (ldebug(rt_sigsuspend))
808 printf(ARGS(rt_sigsuspend, "%p, %d"),
809 (void *)uap->newset, uap->sigsetsize);
812 if (uap->sigsetsize != sizeof(l_sigset_t))
815 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
819 linux_to_bsd_sigset(&linux_mask, &mask);
821 error = kern_sigsuspend(&mask);
827 linux_pause(struct linux_pause_args *args)
829 struct thread *td = curthread;
830 struct proc *p = td->td_proc;
836 printf(ARGS(pause, ""));
841 error = kern_sigsuspend(&mask);
847 linux_sigaltstack(struct linux_sigaltstack_args *uap)
854 if (ldebug(sigaltstack))
855 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
859 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
863 ss.ss_sp = linux_ss.ss_sp;
864 ss.ss_size = linux_ss.ss_size;
865 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
868 error = kern_sigaltstack(uap->uss ? &ss : NULL,
869 uap->uoss ? &oss : NULL);
871 if (error == 0 && uap->uoss) {
872 linux_ss.ss_sp = oss.ss_sp;
873 linux_ss.ss_size = oss.ss_size;
874 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
875 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));