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;
110 sys_linux_execve(struct linux_execve_args *args)
112 struct nlookupdata nd;
113 struct image_args exec_args;
117 error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
122 kprintf(ARGS(execve, "%s"), path);
125 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
126 bzero(&exec_args, sizeof(exec_args));
128 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
129 args->argp, args->envp);
132 error = kern_execve(&nd, &exec_args);
136 * The syscall result is returned in registers to the new program.
137 * Linux will register %edx as an atexit function and we must be
138 * sure to set it to 0. XXX
141 args->sysmsg_result64 = 0;
143 exec_free_args(&exec_args);
144 linux_free_path(&path);
147 /* We hit a lethal error condition. Let's die now. */
148 exit1(W_EXITCODE(0, SIGABRT));
156 struct l_ipc_kludge {
157 struct l_msgbuf *msgp;
165 sys_linux_ipc(struct linux_ipc_args *args)
171 switch (args->what & 0xFFFF) {
173 struct linux_semop_args a;
175 a.semid = args->arg1;
177 a.nsops = args->arg2;
178 a.sysmsg_lresult = 0;
179 error = linux_semop(&a);
180 args->sysmsg_lresult = a.sysmsg_lresult;
184 struct linux_semget_args a;
187 a.nsems = args->arg2;
188 a.semflg = args->arg3;
189 a.sysmsg_lresult = 0;
190 error = linux_semget(&a);
191 args->sysmsg_lresult = a.sysmsg_lresult;
195 struct linux_semctl_args a;
198 a.semid = args->arg1;
199 a.semnum = args->arg2;
201 a.sysmsg_lresult = 0;
202 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
205 error = linux_semctl(&a);
206 args->sysmsg_lresult = a.sysmsg_lresult;
210 struct linux_msgsnd_args a;
212 a.msqid = args->arg1;
214 a.msgsz = args->arg2;
215 a.msgflg = args->arg3;
216 a.sysmsg_lresult = 0;
217 error = linux_msgsnd(&a);
218 args->sysmsg_lresult = a.sysmsg_lresult;
222 struct linux_msgrcv_args a;
224 a.msqid = args->arg1;
225 a.msgsz = args->arg2;
226 a.msgflg = args->arg3;
227 a.sysmsg_lresult = 0;
228 if ((args->what >> 16) == 0) {
229 struct l_ipc_kludge tmp;
232 if (args->ptr == NULL) {
236 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
240 a.msgtyp = tmp.msgtyp;
243 a.msgtyp = args->arg5;
245 error = linux_msgrcv(&a);
246 args->sysmsg_lresult = a.sysmsg_lresult;
250 struct linux_msgget_args a;
253 a.msgflg = args->arg2;
254 a.sysmsg_lresult = 0;
255 error = linux_msgget(&a);
256 args->sysmsg_lresult = a.sysmsg_lresult;
260 struct linux_msgctl_args a;
262 a.msqid = args->arg1;
265 a.sysmsg_lresult = 0;
266 error = linux_msgctl(&a);
267 args->sysmsg_lresult = a.sysmsg_lresult;
271 struct linux_shmat_args a;
273 a.shmid = args->arg1;
274 a.shmaddr = args->ptr;
275 a.shmflg = args->arg2;
276 a.raddr = (l_ulong *)args->arg3;
277 a.sysmsg_lresult = 0;
278 error = linux_shmat(&a);
279 args->sysmsg_lresult = a.sysmsg_lresult;
283 struct linux_shmdt_args a;
285 a.shmaddr = args->ptr;
286 a.sysmsg_lresult = 0;
287 error = linux_shmdt(&a);
288 args->sysmsg_lresult = a.sysmsg_lresult;
292 struct linux_shmget_args a;
296 a.shmflg = args->arg3;
297 a.sysmsg_lresult = 0;
298 error = linux_shmget(&a);
299 args->sysmsg_lresult = a.sysmsg_lresult;
303 struct linux_shmctl_args a;
305 a.shmid = args->arg1;
308 a.sysmsg_lresult = 0;
309 error = linux_shmctl(&a);
310 args->sysmsg_lresult = a.sysmsg_lresult;
325 sys_linux_old_select(struct linux_old_select_args *args)
327 struct l_old_select_argv linux_args;
328 struct linux_select_args newsel;
332 if (ldebug(old_select))
333 kprintf(ARGS(old_select, "%p"), args->ptr);
336 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
340 newsel.sysmsg_iresult = 0;
341 newsel.nfds = linux_args.nfds;
342 newsel.readfds = linux_args.readfds;
343 newsel.writefds = linux_args.writefds;
344 newsel.exceptfds = linux_args.exceptfds;
345 newsel.timeout = linux_args.timeout;
346 error = sys_linux_select(&newsel);
347 args->sysmsg_iresult = newsel.sysmsg_iresult;
355 sys_linux_fork(struct linux_fork_args *args)
361 kprintf(ARGS(fork, ""));
364 if ((error = sys_fork((struct fork_args *)args)) != 0)
367 if (args->sysmsg_iresult == 1)
368 args->sysmsg_iresult = 0;
376 sys_linux_exit_group(struct linux_exit_group_args *args)
378 struct exit_args newargs;
381 newargs.sysmsg_iresult = 0;
382 newargs.rval = args->rval;
383 error = sys_exit(&newargs);
384 args->sysmsg_iresult = newargs.sysmsg_iresult;
392 sys_linux_vfork(struct linux_vfork_args *args)
398 kprintf(ARGS(vfork, ""));
401 if ((error = sys_vfork((struct vfork_args *)args)) != 0)
403 /* Are we the child? */
404 if (args->sysmsg_iresult == 1)
405 args->sysmsg_iresult = 0;
409 #define CLONE_VM 0x100
410 #define CLONE_FS 0x200
411 #define CLONE_FILES 0x400
412 #define CLONE_SIGHAND 0x800
413 #define CLONE_PID 0x1000
419 sys_linux_clone(struct linux_clone_args *args)
421 int error, ff = RFPROC;
425 struct rfork_args rf_args;
429 kprintf(ARGS(clone, "flags %x, stack %x"),
430 (unsigned int)args->flags, (unsigned int)args->stack);
431 if (args->flags & CLONE_PID)
432 kprintf(LMSG("CLONE_PID not yet supported"));
439 exit_signal = args->flags & 0x000000ff;
440 if (exit_signal >= LINUX_NSIG)
443 if (exit_signal <= LINUX_SIGTBLSZ)
444 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
446 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
449 if (args->flags & CLONE_VM)
451 if (args->flags & CLONE_SIGHAND)
453 if (!(args->flags & CLONE_FILES))
460 rf_args.sysmsg_iresult = 0;
462 if ((error = sys_rfork(&rf_args)) == 0) {
463 args->sysmsg_iresult = rf_args.sysmsg_iresult;
465 p2 = pfind(rf_args.sysmsg_iresult);
471 p2->p_sigparent = exit_signal;
472 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
473 (unsigned long)args->stack;
477 kprintf(LMSG("clone: successful rfork to %ld"),
494 #define STACK_SIZE (2 * 1024 * 1024)
495 #define GUARD_SIZE (4 * PAGE_SIZE)
501 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
502 int linux_flags, int linux_fd, off_t pos, void **res)
504 struct thread *td = curthread;
505 struct proc *p = td->td_proc;
508 int error, flags, len, prot, fd;
511 if (linux_flags & LINUX_MAP_SHARED)
513 if (linux_flags & LINUX_MAP_PRIVATE)
514 flags |= MAP_PRIVATE;
515 if (linux_flags & LINUX_MAP_FIXED)
517 if (linux_flags & LINUX_MAP_ANON) {
525 if (linux_flags & LINUX_MAP_GROWSDOWN) {
527 /* The linux MAP_GROWSDOWN option does not limit auto
528 * growth of the region. Linux mmap with this option
529 * takes as addr the inital BOS, and as len, the initial
530 * region size. It can then grow down from addr without
531 * limit. However, linux threads has an implicit internal
532 * limit to stack size of STACK_SIZE. Its just not
533 * enforced explicitly in linux. But, here we impose
534 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
535 * region, since we can do this with our mmap.
537 * Our mmap with MAP_STACK takes addr as the maximum
538 * downsize limit on BOS, and as len the max size of
539 * the region. It them maps the top SGROWSIZ bytes,
540 * and autgrows the region down, up to the limit
543 * If we don't use the MAP_STACK option, the effect
544 * of this code is to allocate a stack region of a
545 * fixed size of (STACK_SIZE - GUARD_SIZE).
548 /* This gives us TOS */
549 addr = linux_addr + linux_len;
551 if (addr > p->p_vmspace->vm_maxsaddr) {
552 /* Some linux apps will attempt to mmap
553 * thread stacks near the top of their
554 * address space. If their TOS is greater
555 * than vm_maxsaddr, vm_map_growstack()
556 * will confuse the thread stack with the
557 * process stack and deliver a SEGV if they
558 * attempt to grow the thread stack past their
559 * current stacksize rlimit. To avoid this,
560 * adjust vm_maxsaddr upwards to reflect
561 * the current stacksize rlimit rather
562 * than the maximum possible stacksize.
563 * It would be better to adjust the
564 * mmap'ed region, but some apps do not check
565 * mmap's return value.
567 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
568 p->p_rlimit[RLIMIT_STACK].rlim_cur;
571 /* This gives us our maximum stack size */
572 if (linux_len > STACK_SIZE - GUARD_SIZE) {
575 len = STACK_SIZE - GUARD_SIZE;
577 /* This gives us a new BOS. If we're using VM_STACK, then
578 * mmap will just map the top SGROWSIZ bytes, and let
579 * the stack grow down to the limit at BOS. If we're
580 * not using VM_STACK we map the full stack, since we
581 * don't have a way to autogrow it.
589 prot = linux_prot | PROT_READ;
590 if (linux_flags & LINUX_MAP_ANON) {
597 if (ldebug(mmap) || ldebug(mmap2))
598 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
599 addr, len, prot, flags, fd, pos);
601 error = kern_mmap(curproc->p_vmspace, addr, len,
602 prot, flags, fd, pos, &new);
614 sys_linux_mmap(struct linux_mmap_args *args)
616 struct l_mmap_argv linux_args;
619 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
625 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
626 (void *)linux_args.addr, linux_args.len, linux_args.prot,
627 linux_args.flags, linux_args.fd, linux_args.pos);
629 error = linux_mmap_common(linux_args.addr, linux_args.len,
630 linux_args.prot, linux_args.flags, linux_args.fd,
631 linux_args.pos, &args->sysmsg_resultp);
634 kprintf("-> %p\n", args->sysmsg_resultp);
643 sys_linux_mmap2(struct linux_mmap2_args *args)
649 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
650 (void *)args->addr, args->len, args->prot, args->flags,
651 args->fd, args->pgoff);
653 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
654 args->flags, args->fd, args->pgoff * PAGE_SIZE,
655 &args->sysmsg_resultp);
658 kprintf("-> %p\n", args->sysmsg_resultp);
667 sys_linux_pipe(struct linux_pipe_args *args)
671 struct pipe_args bsd_args;
675 kprintf(ARGS(pipe, "*"));
678 reg_edx = args->sysmsg_fds[1];
679 error = sys_pipe(&bsd_args);
681 args->sysmsg_fds[1] = reg_edx;
685 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
687 args->sysmsg_fds[1] = reg_edx;
691 args->sysmsg_fds[1] = reg_edx;
692 args->sysmsg_fds[0] = 0;
700 sys_linux_ioperm(struct linux_ioperm_args *args)
702 struct sysarch_args sa;
703 struct i386_ioperm_args *iia;
707 sg = stackgap_init();
708 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
709 iia->start = args->start;
710 iia->length = args->length;
711 iia->enable = args->enable;
712 sa.sysmsg_resultp = NULL;
713 sa.op = I386_SET_IOPERM;
714 sa.parms = (char *)iia;
715 error = sys_sysarch(&sa);
716 args->sysmsg_resultp = sa.sysmsg_resultp;
724 sys_linux_iopl(struct linux_iopl_args *args)
726 struct thread *td = curthread;
727 struct lwp *lp = td->td_lwp;
730 if (args->level < 0 || args->level > 3)
732 if ((error = priv_check(td, PRIV_ROOT)) != 0)
736 lp->lwp_md.md_regs->tf_eflags =
737 (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
738 (args->level * (PSL_IOPL / 3));
746 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
750 struct sysarch_args args;
751 struct i386_ldt_args *ldt;
752 struct l_descriptor ld;
753 union descriptor *desc;
755 sg = stackgap_init();
757 if (uap->ptr == NULL)
761 case 0x00: /* read_ldt */
762 ldt = stackgap_alloc(&sg, sizeof(*ldt));
764 ldt->descs = uap->ptr;
765 ldt->num = uap->bytecount / sizeof(union descriptor);
766 args.op = I386_GET_LDT;
767 args.parms = (char*)ldt;
768 args.sysmsg_iresult = 0;
769 error = sys_sysarch(&args);
770 uap->sysmsg_iresult = args.sysmsg_iresult *
771 sizeof(union descriptor);
773 case 0x01: /* write_ldt */
774 case 0x11: /* write_ldt */
775 if (uap->bytecount != sizeof(ld))
778 error = copyin(uap->ptr, &ld, sizeof(ld));
782 ldt = stackgap_alloc(&sg, sizeof(*ldt));
783 desc = stackgap_alloc(&sg, sizeof(*desc));
784 ldt->start = ld.entry_number;
787 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
788 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
789 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
790 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
791 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
794 desc->sd.sd_p = (ld.seg_not_present ^ 1);
796 desc->sd.sd_def32 = ld.seg_32bit;
797 desc->sd.sd_gran = ld.limit_in_pages;
798 args.op = I386_SET_LDT;
799 args.parms = (char*)ldt;
800 args.sysmsg_iresult = 0;
801 error = sys_sysarch(&args);
802 uap->sysmsg_iresult = args.sysmsg_iresult;
816 sys_linux_sigaction(struct linux_sigaction_args *args)
819 l_sigaction_t linux_act, linux_oact;
820 struct sigaction act, oact;
824 if (ldebug(sigaction))
825 kprintf(ARGS(sigaction, "%d, %p, %p"),
826 args->sig, (void *)args->nsa, (void *)args->osa);
830 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
833 linux_act.lsa_handler = osa.lsa_handler;
834 linux_act.lsa_flags = osa.lsa_flags;
835 linux_act.lsa_restorer = osa.lsa_restorer;
836 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
837 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
838 linux_to_bsd_sigaction(&linux_act, &act);
842 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
843 args->osa ? &oact : NULL);
846 if (args->osa != NULL && !error) {
847 bsd_to_linux_sigaction(&oact, &linux_oact);
848 osa.lsa_handler = linux_oact.lsa_handler;
849 osa.lsa_flags = linux_oact.lsa_flags;
850 osa.lsa_restorer = linux_oact.lsa_restorer;
851 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
852 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
858 * Linux has two extra args, restart and oldmask. We dont use these,
859 * but it seems that "restart" is actually a context pointer that
860 * enables the signal to happen with a different register set.
865 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
867 l_sigset_t linux_mask;
872 if (ldebug(sigsuspend))
873 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
876 LINUX_SIGEMPTYSET(mask);
877 mask.__bits[0] = args->mask;
878 linux_to_bsd_sigset(&linux_mask, &mask);
881 error = kern_sigsuspend(&mask);
891 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
893 l_sigset_t linux_mask;
898 if (ldebug(rt_sigsuspend))
899 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
900 (void *)uap->newset, uap->sigsetsize);
903 if (uap->sigsetsize != sizeof(l_sigset_t))
906 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
910 linux_to_bsd_sigset(&linux_mask, &mask);
913 error = kern_sigsuspend(&mask);
923 sys_linux_pause(struct linux_pause_args *args)
925 struct thread *td = curthread;
926 struct lwp *lp = td->td_lwp;
932 kprintf(ARGS(pause, ""));
935 mask = lp->lwp_sigmask;
938 error = kern_sigsuspend(&mask);
948 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
955 if (ldebug(sigaltstack))
956 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
960 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
964 ss.ss_sp = linux_ss.ss_sp;
965 ss.ss_size = linux_ss.ss_size;
966 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
970 error = kern_sigaltstack(uap->uss ? &ss : NULL,
971 uap->uoss ? &oss : NULL);
974 if (error == 0 && uap->uoss) {
975 linux_ss.ss_sp = oss.ss_sp;
976 linux_ss.ss_size = oss.ss_size;
977 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
978 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));