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>
56 #include <sys/mplock2.h>
59 #include "linux_proto.h"
60 #include "../linux_ipc.h"
61 #include "../linux_signal.h"
62 #include "../linux_util.h"
70 l_uint read_exec_only:1;
71 l_uint limit_in_pages:1;
72 l_uint seg_not_present:1;
76 struct l_old_select_argv {
81 struct l_timeval *timeout;
85 linux_to_bsd_sigaltstack(int lsa)
89 if (lsa & LINUX_SS_DISABLE)
91 if (lsa & LINUX_SS_ONSTACK)
97 bsd_to_linux_sigaltstack(int bsa)
101 if (bsa & SS_DISABLE)
102 lsa |= LINUX_SS_DISABLE;
103 if (bsa & SS_ONSTACK)
104 lsa |= LINUX_SS_ONSTACK;
112 sys_linux_execve(struct linux_execve_args *args)
114 struct nlookupdata nd;
115 struct image_args exec_args;
119 error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
124 kprintf(ARGS(execve, "%s"), path);
127 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
128 bzero(&exec_args, sizeof(exec_args));
130 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
131 args->argp, args->envp);
134 error = kern_execve(&nd, &exec_args);
138 * The syscall result is returned in registers to the new program.
139 * Linux will register %edx as an atexit function and we must be
140 * sure to set it to 0. XXX
143 args->sysmsg_result64 = 0;
145 exec_free_args(&exec_args);
146 linux_free_path(&path);
149 /* We hit a lethal error condition. Let's die now. */
150 exit1(W_EXITCODE(0, SIGABRT));
158 struct l_ipc_kludge {
159 struct l_msgbuf *msgp;
167 sys_linux_ipc(struct linux_ipc_args *args)
173 switch (args->what & 0xFFFF) {
175 struct linux_semop_args a;
177 a.semid = args->arg1;
179 a.nsops = args->arg2;
180 a.sysmsg_lresult = 0;
181 error = linux_semop(&a);
182 args->sysmsg_lresult = a.sysmsg_lresult;
186 struct linux_semget_args a;
189 a.nsems = args->arg2;
190 a.semflg = args->arg3;
191 a.sysmsg_lresult = 0;
192 error = linux_semget(&a);
193 args->sysmsg_lresult = a.sysmsg_lresult;
197 struct linux_semctl_args a;
200 a.semid = args->arg1;
201 a.semnum = args->arg2;
203 a.sysmsg_lresult = 0;
204 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
207 error = linux_semctl(&a);
208 args->sysmsg_lresult = a.sysmsg_lresult;
212 struct linux_msgsnd_args a;
214 a.msqid = args->arg1;
216 a.msgsz = args->arg2;
217 a.msgflg = args->arg3;
218 a.sysmsg_lresult = 0;
219 error = linux_msgsnd(&a);
220 args->sysmsg_lresult = a.sysmsg_lresult;
224 struct linux_msgrcv_args a;
226 a.msqid = args->arg1;
227 a.msgsz = args->arg2;
228 a.msgflg = args->arg3;
229 a.sysmsg_lresult = 0;
230 if ((args->what >> 16) == 0) {
231 struct l_ipc_kludge tmp;
234 if (args->ptr == NULL) {
238 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
242 a.msgtyp = tmp.msgtyp;
245 a.msgtyp = args->arg5;
247 error = linux_msgrcv(&a);
248 args->sysmsg_lresult = a.sysmsg_lresult;
252 struct linux_msgget_args a;
255 a.msgflg = args->arg2;
256 a.sysmsg_lresult = 0;
257 error = linux_msgget(&a);
258 args->sysmsg_lresult = a.sysmsg_lresult;
262 struct linux_msgctl_args a;
264 a.msqid = args->arg1;
267 a.sysmsg_lresult = 0;
268 error = linux_msgctl(&a);
269 args->sysmsg_lresult = a.sysmsg_lresult;
273 struct linux_shmat_args a;
275 a.shmid = args->arg1;
276 a.shmaddr = args->ptr;
277 a.shmflg = args->arg2;
278 a.raddr = (l_ulong *)args->arg3;
279 a.sysmsg_lresult = 0;
280 error = linux_shmat(&a);
281 args->sysmsg_lresult = a.sysmsg_lresult;
285 struct linux_shmdt_args a;
287 a.shmaddr = args->ptr;
288 a.sysmsg_lresult = 0;
289 error = linux_shmdt(&a);
290 args->sysmsg_lresult = a.sysmsg_lresult;
294 struct linux_shmget_args a;
298 a.shmflg = args->arg3;
299 a.sysmsg_lresult = 0;
300 error = linux_shmget(&a);
301 args->sysmsg_lresult = a.sysmsg_lresult;
305 struct linux_shmctl_args a;
307 a.shmid = args->arg1;
310 a.sysmsg_lresult = 0;
311 error = linux_shmctl(&a);
312 args->sysmsg_lresult = a.sysmsg_lresult;
327 sys_linux_old_select(struct linux_old_select_args *args)
329 struct l_old_select_argv linux_args;
330 struct linux_select_args newsel;
334 if (ldebug(old_select))
335 kprintf(ARGS(old_select, "%p"), args->ptr);
338 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
342 newsel.sysmsg_iresult = 0;
343 newsel.nfds = linux_args.nfds;
344 newsel.readfds = linux_args.readfds;
345 newsel.writefds = linux_args.writefds;
346 newsel.exceptfds = linux_args.exceptfds;
347 newsel.timeout = linux_args.timeout;
348 error = sys_linux_select(&newsel);
349 args->sysmsg_iresult = newsel.sysmsg_iresult;
357 sys_linux_fork(struct linux_fork_args *args)
363 kprintf(ARGS(fork, ""));
366 if ((error = sys_fork((struct fork_args *)args)) != 0)
369 if (args->sysmsg_iresult == 1)
370 args->sysmsg_iresult = 0;
378 sys_linux_exit_group(struct linux_exit_group_args *args)
380 struct exit_args newargs;
383 newargs.sysmsg_iresult = 0;
384 newargs.rval = args->rval;
385 error = sys_exit(&newargs);
386 args->sysmsg_iresult = newargs.sysmsg_iresult;
394 sys_linux_vfork(struct linux_vfork_args *args)
400 kprintf(ARGS(vfork, ""));
403 if ((error = sys_vfork((struct vfork_args *)args)) != 0)
405 /* Are we the child? */
406 if (args->sysmsg_iresult == 1)
407 args->sysmsg_iresult = 0;
411 #define CLONE_VM 0x100
412 #define CLONE_FS 0x200
413 #define CLONE_FILES 0x400
414 #define CLONE_SIGHAND 0x800
415 #define CLONE_PID 0x1000
421 sys_linux_clone(struct linux_clone_args *args)
423 int error, ff = RFPROC;
424 struct proc *p2 = NULL;
427 struct rfork_args rf_args;
431 kprintf(ARGS(clone, "flags %x, stack %x"),
432 (unsigned int)args->flags, (unsigned int)args->stack);
433 if (args->flags & CLONE_PID)
434 kprintf(LMSG("CLONE_PID not yet supported"));
441 exit_signal = args->flags & 0x000000ff;
442 if (exit_signal >= LINUX_NSIG)
445 if (exit_signal <= LINUX_SIGTBLSZ)
446 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
448 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
451 if (args->flags & CLONE_VM)
453 if (args->flags & CLONE_SIGHAND)
455 if (!(args->flags & CLONE_FILES))
462 rf_args.sysmsg_iresult = 0;
464 if ((error = sys_rfork(&rf_args)) == 0) {
465 args->sysmsg_iresult = rf_args.sysmsg_iresult;
467 p2 = pfind(rf_args.sysmsg_iresult);
473 p2->p_sigparent = exit_signal;
474 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
475 (unsigned long)args->stack;
479 kprintf(LMSG("clone: successful rfork to %ld"),
496 #define STACK_SIZE (2 * 1024 * 1024)
497 #define GUARD_SIZE (4 * PAGE_SIZE)
503 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
504 int linux_flags, int linux_fd, off_t pos, void **res)
506 struct thread *td = curthread;
507 struct proc *p = td->td_proc;
510 int error, flags, len, prot, fd;
513 if (linux_flags & LINUX_MAP_SHARED)
515 if (linux_flags & LINUX_MAP_PRIVATE)
516 flags |= MAP_PRIVATE;
517 if (linux_flags & LINUX_MAP_FIXED)
519 if (linux_flags & LINUX_MAP_ANON) {
527 if (linux_flags & LINUX_MAP_GROWSDOWN) {
529 /* The linux MAP_GROWSDOWN option does not limit auto
530 * growth of the region. Linux mmap with this option
531 * takes as addr the inital BOS, and as len, the initial
532 * region size. It can then grow down from addr without
533 * limit. However, linux threads has an implicit internal
534 * limit to stack size of STACK_SIZE. Its just not
535 * enforced explicitly in linux. But, here we impose
536 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
537 * region, since we can do this with our mmap.
539 * Our mmap with MAP_STACK takes addr as the maximum
540 * downsize limit on BOS, and as len the max size of
541 * the region. It them maps the top SGROWSIZ bytes,
542 * and autgrows the region down, up to the limit
545 * If we don't use the MAP_STACK option, the effect
546 * of this code is to allocate a stack region of a
547 * fixed size of (STACK_SIZE - GUARD_SIZE).
550 /* This gives us TOS */
551 addr = linux_addr + linux_len;
553 if (addr > p->p_vmspace->vm_maxsaddr) {
554 /* Some linux apps will attempt to mmap
555 * thread stacks near the top of their
556 * address space. If their TOS is greater
557 * than vm_maxsaddr, vm_map_growstack()
558 * will confuse the thread stack with the
559 * process stack and deliver a SEGV if they
560 * attempt to grow the thread stack past their
561 * current stacksize rlimit. To avoid this,
562 * adjust vm_maxsaddr upwards to reflect
563 * the current stacksize rlimit rather
564 * than the maximum possible stacksize.
565 * It would be better to adjust the
566 * mmap'ed region, but some apps do not check
567 * mmap's return value.
569 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
570 p->p_rlimit[RLIMIT_STACK].rlim_cur;
573 /* This gives us our maximum stack size */
574 if (linux_len > STACK_SIZE - GUARD_SIZE) {
577 len = STACK_SIZE - GUARD_SIZE;
579 /* This gives us a new BOS. If we're using VM_STACK, then
580 * mmap will just map the top SGROWSIZ bytes, and let
581 * the stack grow down to the limit at BOS. If we're
582 * not using VM_STACK we map the full stack, since we
583 * don't have a way to autogrow it.
591 prot = linux_prot | PROT_READ;
592 if (linux_flags & LINUX_MAP_ANON) {
599 if (ldebug(mmap) || ldebug(mmap2))
600 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
601 addr, len, prot, flags, fd, pos);
603 error = kern_mmap(curproc->p_vmspace, addr, len,
604 prot, flags, fd, pos, &new);
616 sys_linux_mmap(struct linux_mmap_args *args)
618 struct l_mmap_argv linux_args;
621 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
627 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
628 (void *)linux_args.addr, linux_args.len, linux_args.prot,
629 linux_args.flags, linux_args.fd, linux_args.pos);
631 error = linux_mmap_common(linux_args.addr, linux_args.len,
632 linux_args.prot, linux_args.flags, linux_args.fd,
633 linux_args.pos, &args->sysmsg_resultp);
636 kprintf("-> %p\n", args->sysmsg_resultp);
645 sys_linux_mmap2(struct linux_mmap2_args *args)
651 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
652 (void *)args->addr, args->len, args->prot, args->flags,
653 args->fd, args->pgoff);
655 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
656 args->flags, args->fd, args->pgoff * PAGE_SIZE,
657 &args->sysmsg_resultp);
660 kprintf("-> %p\n", args->sysmsg_resultp);
669 sys_linux_pipe(struct linux_pipe_args *args)
673 struct pipe_args bsd_args;
677 kprintf(ARGS(pipe, "*"));
680 reg_edx = args->sysmsg_fds[1];
681 error = sys_pipe(&bsd_args);
683 args->sysmsg_fds[1] = reg_edx;
687 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
689 args->sysmsg_fds[1] = reg_edx;
693 args->sysmsg_fds[1] = reg_edx;
694 args->sysmsg_fds[0] = 0;
702 sys_linux_ioperm(struct linux_ioperm_args *args)
704 struct sysarch_args sa;
705 struct i386_ioperm_args *iia;
709 sg = stackgap_init();
710 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
711 iia->start = args->start;
712 iia->length = args->length;
713 iia->enable = args->enable;
714 sa.sysmsg_resultp = NULL;
715 sa.op = I386_SET_IOPERM;
716 sa.parms = (char *)iia;
717 error = sys_sysarch(&sa);
718 args->sysmsg_resultp = sa.sysmsg_resultp;
726 sys_linux_iopl(struct linux_iopl_args *args)
728 struct thread *td = curthread;
729 struct lwp *lp = td->td_lwp;
732 if (args->level < 0 || args->level > 3)
734 if ((error = priv_check(td, PRIV_ROOT)) != 0)
738 lp->lwp_md.md_regs->tf_eflags =
739 (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
740 (args->level * (PSL_IOPL / 3));
748 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
752 struct sysarch_args args;
753 struct i386_ldt_args *ldt;
754 struct l_descriptor ld;
755 union descriptor *desc;
757 sg = stackgap_init();
759 if (uap->ptr == NULL)
763 case 0x00: /* read_ldt */
764 ldt = stackgap_alloc(&sg, sizeof(*ldt));
766 ldt->descs = uap->ptr;
767 ldt->num = uap->bytecount / sizeof(union descriptor);
768 args.op = I386_GET_LDT;
769 args.parms = (char*)ldt;
770 args.sysmsg_iresult = 0;
771 error = sys_sysarch(&args);
772 uap->sysmsg_iresult = args.sysmsg_iresult *
773 sizeof(union descriptor);
775 case 0x01: /* write_ldt */
776 case 0x11: /* write_ldt */
777 if (uap->bytecount != sizeof(ld))
780 error = copyin(uap->ptr, &ld, sizeof(ld));
784 ldt = stackgap_alloc(&sg, sizeof(*ldt));
785 desc = stackgap_alloc(&sg, sizeof(*desc));
786 ldt->start = ld.entry_number;
789 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
790 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
791 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
792 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
793 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
796 desc->sd.sd_p = (ld.seg_not_present ^ 1);
798 desc->sd.sd_def32 = ld.seg_32bit;
799 desc->sd.sd_gran = ld.limit_in_pages;
800 args.op = I386_SET_LDT;
801 args.parms = (char*)ldt;
802 args.sysmsg_iresult = 0;
803 error = sys_sysarch(&args);
804 uap->sysmsg_iresult = args.sysmsg_iresult;
818 sys_linux_sigaction(struct linux_sigaction_args *args)
821 l_sigaction_t linux_act, linux_oact;
822 struct sigaction act, oact;
826 if (ldebug(sigaction))
827 kprintf(ARGS(sigaction, "%d, %p, %p"),
828 args->sig, (void *)args->nsa, (void *)args->osa);
832 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
835 linux_act.lsa_handler = osa.lsa_handler;
836 linux_act.lsa_flags = osa.lsa_flags;
837 linux_act.lsa_restorer = osa.lsa_restorer;
838 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
839 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
840 linux_to_bsd_sigaction(&linux_act, &act);
844 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
845 args->osa ? &oact : NULL);
848 if (args->osa != NULL && !error) {
849 bsd_to_linux_sigaction(&oact, &linux_oact);
850 osa.lsa_handler = linux_oact.lsa_handler;
851 osa.lsa_flags = linux_oact.lsa_flags;
852 osa.lsa_restorer = linux_oact.lsa_restorer;
853 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
854 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
860 * Linux has two extra args, restart and oldmask. We dont use these,
861 * but it seems that "restart" is actually a context pointer that
862 * enables the signal to happen with a different register set.
867 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
869 l_sigset_t linux_mask;
874 if (ldebug(sigsuspend))
875 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
878 LINUX_SIGEMPTYSET(mask);
879 mask.__bits[0] = args->mask;
880 linux_to_bsd_sigset(&linux_mask, &mask);
883 error = kern_sigsuspend(&mask);
893 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
895 l_sigset_t linux_mask;
900 if (ldebug(rt_sigsuspend))
901 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
902 (void *)uap->newset, uap->sigsetsize);
905 if (uap->sigsetsize != sizeof(l_sigset_t))
908 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
912 linux_to_bsd_sigset(&linux_mask, &mask);
915 error = kern_sigsuspend(&mask);
925 sys_linux_pause(struct linux_pause_args *args)
927 struct thread *td = curthread;
928 struct lwp *lp = td->td_lwp;
934 kprintf(ARGS(pause, ""));
937 mask = lp->lwp_sigmask;
940 error = kern_sigsuspend(&mask);
950 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
957 if (ldebug(sigaltstack))
958 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
962 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
966 ss.ss_sp = linux_ss.ss_sp;
967 ss.ss_size = linux_ss.ss_size;
968 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
972 error = kern_sigaltstack(uap->uss ? &ss : NULL,
973 uap->uoss ? &oss : NULL);
976 if (error == 0 && uap->uoss) {
977 linux_ss.ss_sp = oss.ss_sp;
978 linux_ss.ss_size = oss.ss_size;
979 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
980 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));