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>
40 #include <sys/resource.h>
41 #include <sys/resourcevar.h>
42 #include <sys/sysproto.h>
43 #include <sys/unistd.h>
46 #include <machine/frame.h>
47 #include <machine/psl.h>
48 #include <machine/segments.h>
49 #include <machine/sysarch.h>
53 #include <vm/vm_map.h>
56 #include "linux_proto.h"
57 #include "../linux_ipc.h"
58 #include "../linux_signal.h"
59 #include "../linux_util.h"
67 l_uint read_exec_only:1;
68 l_uint limit_in_pages:1;
69 l_uint seg_not_present:1;
73 struct l_old_select_argv {
78 struct l_timeval *timeout;
82 linux_to_bsd_sigaltstack(int lsa)
86 if (lsa & LINUX_SS_DISABLE)
88 if (lsa & LINUX_SS_ONSTACK)
94 bsd_to_linux_sigaltstack(int bsa)
99 lsa |= LINUX_SS_DISABLE;
100 if (bsa & SS_ONSTACK)
101 lsa |= LINUX_SS_ONSTACK;
106 sys_linux_execve(struct linux_execve_args *args)
108 struct nlookupdata nd;
109 struct image_args exec_args;
113 error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
118 kprintf(ARGS(execve, "%s"), path);
120 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
122 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
123 args->argp, args->envp);
126 error = kern_execve(&nd, &exec_args);
130 * The syscall result is returned in registers to the new program.
131 * Linux will register %edx as an atexit function and we must be
132 * sure to set it to 0. XXX
135 args->sysmsg_result64 = 0;
137 exec_free_args(&exec_args);
138 linux_free_path(&path);
141 /* We hit a lethal error condition. Let's die now. */
142 exit1(W_EXITCODE(0, SIGABRT));
149 struct l_ipc_kludge {
150 struct l_msgbuf *msgp;
155 sys_linux_ipc(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 a.sysmsg_lresult = 0;
167 error = linux_semop(&a);
168 args->sysmsg_lresult = a.sysmsg_lresult;
172 struct linux_semget_args a;
175 a.nsems = args->arg2;
176 a.semflg = args->arg3;
177 a.sysmsg_lresult = 0;
178 error = linux_semget(&a);
179 args->sysmsg_lresult = a.sysmsg_lresult;
183 struct linux_semctl_args a;
186 a.semid = args->arg1;
187 a.semnum = args->arg2;
189 a.sysmsg_lresult = 0;
190 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
193 error = linux_semctl(&a);
194 args->sysmsg_lresult = a.sysmsg_lresult;
198 struct linux_msgsnd_args a;
200 a.msqid = args->arg1;
202 a.msgsz = args->arg2;
203 a.msgflg = args->arg3;
204 a.sysmsg_lresult = 0;
205 error = linux_msgsnd(&a);
206 args->sysmsg_lresult = a.sysmsg_lresult;
210 struct linux_msgrcv_args a;
212 a.msqid = args->arg1;
213 a.msgsz = args->arg2;
214 a.msgflg = args->arg3;
215 a.sysmsg_lresult = 0;
216 if ((args->what >> 16) == 0) {
217 struct l_ipc_kludge tmp;
220 if (args->ptr == NULL)
222 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
226 a.msgtyp = tmp.msgtyp;
229 a.msgtyp = args->arg5;
231 error = linux_msgrcv(&a);
232 args->sysmsg_lresult = a.sysmsg_lresult;
236 struct linux_msgget_args a;
239 a.msgflg = args->arg2;
240 a.sysmsg_lresult = 0;
241 error = linux_msgget(&a);
242 args->sysmsg_lresult = a.sysmsg_lresult;
246 struct linux_msgctl_args a;
248 a.msqid = args->arg1;
251 a.sysmsg_lresult = 0;
252 error = linux_msgctl(&a);
253 args->sysmsg_lresult = a.sysmsg_lresult;
257 struct linux_shmat_args a;
259 a.shmid = args->arg1;
260 a.shmaddr = args->ptr;
261 a.shmflg = args->arg2;
262 a.raddr = (l_ulong *)args->arg3;
263 a.sysmsg_lresult = 0;
264 error = linux_shmat(&a);
265 args->sysmsg_lresult = a.sysmsg_lresult;
269 struct linux_shmdt_args a;
271 a.shmaddr = args->ptr;
272 a.sysmsg_lresult = 0;
273 error = linux_shmdt(&a);
274 args->sysmsg_lresult = a.sysmsg_lresult;
278 struct linux_shmget_args a;
282 a.shmflg = args->arg3;
283 a.sysmsg_lresult = 0;
284 error = linux_shmget(&a);
285 args->sysmsg_lresult = a.sysmsg_lresult;
289 struct linux_shmctl_args a;
291 a.shmid = args->arg1;
294 a.sysmsg_lresult = 0;
295 error = linux_shmctl(&a);
296 args->sysmsg_lresult = a.sysmsg_lresult;
307 sys_linux_old_select(struct linux_old_select_args *args)
309 struct l_old_select_argv linux_args;
310 struct linux_select_args newsel;
314 if (ldebug(old_select))
315 kprintf(ARGS(old_select, "%p"), args->ptr);
318 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
322 newsel.sysmsg_result = 0;
323 newsel.nfds = linux_args.nfds;
324 newsel.readfds = linux_args.readfds;
325 newsel.writefds = linux_args.writefds;
326 newsel.exceptfds = linux_args.exceptfds;
327 newsel.timeout = linux_args.timeout;
328 error = sys_linux_select(&newsel);
329 args->sysmsg_result = newsel.sysmsg_result;
334 sys_linux_fork(struct linux_fork_args *args)
340 kprintf(ARGS(fork, ""));
343 if ((error = sys_fork((struct fork_args *)args)) != 0)
346 if (args->sysmsg_result == 1)
347 args->sysmsg_result = 0;
352 sys_linux_exit_group(struct linux_exit_group_args *args)
354 struct exit_args newargs;
357 newargs.sysmsg_result = 0;
358 newargs.rval = args->rval;
359 error = sys_exit(&newargs);
360 args->sysmsg_result = newargs.sysmsg_result;
365 sys_linux_vfork(struct linux_vfork_args *args)
371 kprintf(ARGS(vfork, ""));
374 if ((error = sys_vfork((struct vfork_args *)args)) != 0)
376 /* Are we the child? */
377 if (args->sysmsg_result == 1)
378 args->sysmsg_result = 0;
382 #define CLONE_VM 0x100
383 #define CLONE_FS 0x200
384 #define CLONE_FILES 0x400
385 #define CLONE_SIGHAND 0x800
386 #define CLONE_PID 0x1000
389 sys_linux_clone(struct linux_clone_args *args)
391 int error, ff = RFPROC;
395 struct rfork_args rf_args;
399 kprintf(ARGS(clone, "flags %x, stack %x"),
400 (unsigned int)args->flags, (unsigned int)args->stack);
401 if (args->flags & CLONE_PID)
402 kprintf(LMSG("CLONE_PID not yet supported"));
409 exit_signal = args->flags & 0x000000ff;
410 if (exit_signal >= LINUX_NSIG)
413 if (exit_signal <= LINUX_SIGTBLSZ)
414 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
416 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
419 if (args->flags & CLONE_VM)
421 if (args->flags & CLONE_SIGHAND)
423 if (!(args->flags & CLONE_FILES))
430 rf_args.sysmsg_result = 0;
431 if ((error = sys_rfork(&rf_args)) != 0)
433 args->sysmsg_result = rf_args.sysmsg_result;
435 p2 = pfind(rf_args.sysmsg_result);
439 p2->p_sigparent = exit_signal;
440 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
441 (unsigned int)args->stack;
445 kprintf(LMSG("clone: successful rfork to %ld"),
462 #define STACK_SIZE (2 * 1024 * 1024)
463 #define GUARD_SIZE (4 * PAGE_SIZE)
466 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
467 int linux_flags, int linux_fd, off_t pos, void **res)
469 struct thread *td = curthread;
470 struct proc *p = td->td_proc;
473 int error, flags, len, prot, fd;
476 if (linux_flags & LINUX_MAP_SHARED)
478 if (linux_flags & LINUX_MAP_PRIVATE)
479 flags |= MAP_PRIVATE;
480 if (linux_flags & LINUX_MAP_FIXED)
482 if (linux_flags & LINUX_MAP_ANON) {
487 if (linux_flags & LINUX_MAP_GROWSDOWN) {
489 /* The linux MAP_GROWSDOWN option does not limit auto
490 * growth of the region. Linux mmap with this option
491 * takes as addr the inital BOS, and as len, the initial
492 * region size. It can then grow down from addr without
493 * limit. However, linux threads has an implicit internal
494 * limit to stack size of STACK_SIZE. Its just not
495 * enforced explicitly in linux. But, here we impose
496 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
497 * region, since we can do this with our mmap.
499 * Our mmap with MAP_STACK takes addr as the maximum
500 * downsize limit on BOS, and as len the max size of
501 * the region. It them maps the top SGROWSIZ bytes,
502 * and autgrows the region down, up to the limit
505 * If we don't use the MAP_STACK option, the effect
506 * of this code is to allocate a stack region of a
507 * fixed size of (STACK_SIZE - GUARD_SIZE).
510 /* This gives us TOS */
511 addr = linux_addr + linux_len;
513 if (addr > p->p_vmspace->vm_maxsaddr) {
514 /* Some linux apps will attempt to mmap
515 * thread stacks near the top of their
516 * address space. If their TOS is greater
517 * than vm_maxsaddr, vm_map_growstack()
518 * will confuse the thread stack with the
519 * process stack and deliver a SEGV if they
520 * attempt to grow the thread stack past their
521 * current stacksize rlimit. To avoid this,
522 * adjust vm_maxsaddr upwards to reflect
523 * the current stacksize rlimit rather
524 * than the maximum possible stacksize.
525 * It would be better to adjust the
526 * mmap'ed region, but some apps do not check
527 * mmap's return value.
529 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
530 p->p_rlimit[RLIMIT_STACK].rlim_cur;
533 /* This gives us our maximum stack size */
534 if (linux_len > STACK_SIZE - GUARD_SIZE) {
537 len = STACK_SIZE - GUARD_SIZE;
539 /* This gives us a new BOS. If we're using VM_STACK, then
540 * mmap will just map the top SGROWSIZ bytes, and let
541 * the stack grow down to the limit at BOS. If we're
542 * not using VM_STACK we map the full stack, since we
543 * don't have a way to autogrow it.
551 prot = linux_prot | PROT_READ;
552 if (linux_flags & LINUX_MAP_ANON) {
559 if (ldebug(mmap) || ldebug(mmap2))
560 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
561 addr, len, prot, flags, fd, pos);
563 error = kern_mmap(curproc->p_vmspace, addr, len,
564 prot, flags, fd, pos, &new);
572 sys_linux_mmap(struct linux_mmap_args *args)
574 struct l_mmap_argv linux_args;
577 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
583 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
584 (void *)linux_args.addr, linux_args.len, linux_args.prot,
585 linux_args.flags, linux_args.fd, linux_args.pos);
587 error = linux_mmap_common(linux_args.addr, linux_args.len,
588 linux_args.prot, linux_args.flags, linux_args.fd,
589 linux_args.pos, &args->sysmsg_resultp);
592 kprintf("-> %p\n", args->sysmsg_resultp);
598 sys_linux_mmap2(struct linux_mmap2_args *args)
604 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
605 (void *)args->addr, args->len, args->prot, args->flags,
606 args->fd, args->pgoff);
608 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
609 args->flags, args->fd, args->pgoff * PAGE_SIZE,
610 &args->sysmsg_resultp);
613 kprintf("-> %p\n", args->sysmsg_resultp);
619 sys_linux_pipe(struct linux_pipe_args *args)
623 struct pipe_args bsd_args;
627 kprintf(ARGS(pipe, "*"));
630 reg_edx = args->sysmsg_fds[1];
631 error = sys_pipe(&bsd_args);
633 args->sysmsg_fds[1] = reg_edx;
637 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
639 args->sysmsg_fds[1] = reg_edx;
643 args->sysmsg_fds[1] = reg_edx;
644 args->sysmsg_fds[0] = 0;
649 sys_linux_ioperm(struct linux_ioperm_args *args)
651 struct sysarch_args sa;
652 struct i386_ioperm_args *iia;
656 sg = stackgap_init();
657 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
658 iia->start = args->start;
659 iia->length = args->length;
660 iia->enable = args->enable;
661 sa.sysmsg_resultp = NULL;
662 sa.op = I386_SET_IOPERM;
663 sa.parms = (char *)iia;
664 error = sys_sysarch(&sa);
665 args->sysmsg_resultp = sa.sysmsg_resultp;
670 sys_linux_iopl(struct linux_iopl_args *args)
672 struct thread *td = curthread;
673 struct lwp *lp = td->td_lwp;
678 if (args->level < 0 || args->level > 3)
680 if ((error = suser(td)) != 0)
684 lp->lwp_md.md_regs->tf_eflags =
685 (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
686 (args->level * (PSL_IOPL / 3));
691 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
695 struct sysarch_args args;
696 struct i386_ldt_args *ldt;
697 struct l_descriptor ld;
698 union descriptor *desc;
700 sg = stackgap_init();
702 if (uap->ptr == NULL)
706 case 0x00: /* read_ldt */
707 ldt = stackgap_alloc(&sg, sizeof(*ldt));
709 ldt->descs = uap->ptr;
710 ldt->num = uap->bytecount / sizeof(union descriptor);
711 args.op = I386_GET_LDT;
712 args.parms = (char*)ldt;
713 args.sysmsg_result = 0;
714 error = sys_sysarch(&args);
715 uap->sysmsg_result = args.sysmsg_result *
716 sizeof(union descriptor);
718 case 0x01: /* write_ldt */
719 case 0x11: /* write_ldt */
720 if (uap->bytecount != sizeof(ld))
723 error = copyin(uap->ptr, &ld, sizeof(ld));
727 ldt = stackgap_alloc(&sg, sizeof(*ldt));
728 desc = stackgap_alloc(&sg, sizeof(*desc));
729 ldt->start = ld.entry_number;
732 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
733 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
734 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
735 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
736 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
739 desc->sd.sd_p = (ld.seg_not_present ^ 1);
741 desc->sd.sd_def32 = ld.seg_32bit;
742 desc->sd.sd_gran = ld.limit_in_pages;
743 args.op = I386_SET_LDT;
744 args.parms = (char*)ldt;
745 args.sysmsg_result = 0;
746 error = sys_sysarch(&args);
747 uap->sysmsg_result = args.sysmsg_result;
758 sys_linux_sigaction(struct linux_sigaction_args *args)
761 l_sigaction_t linux_act, linux_oact;
762 struct sigaction act, oact;
766 if (ldebug(sigaction))
767 kprintf(ARGS(sigaction, "%d, %p, %p"),
768 args->sig, (void *)args->nsa, (void *)args->osa);
772 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
775 linux_act.lsa_handler = osa.lsa_handler;
776 linux_act.lsa_flags = osa.lsa_flags;
777 linux_act.lsa_restorer = osa.lsa_restorer;
778 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
779 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
780 linux_to_bsd_sigaction(&linux_act, &act);
783 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
784 args->osa ? &oact : NULL);
786 if (args->osa != NULL && !error) {
787 bsd_to_linux_sigaction(&oact, &linux_oact);
788 osa.lsa_handler = linux_oact.lsa_handler;
789 osa.lsa_flags = linux_oact.lsa_flags;
790 osa.lsa_restorer = linux_oact.lsa_restorer;
791 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
792 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
798 * Linux has two extra args, restart and oldmask. We dont use these,
799 * but it seems that "restart" is actually a context pointer that
800 * enables the signal to happen with a different register set.
803 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
805 l_sigset_t linux_mask;
810 if (ldebug(sigsuspend))
811 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
814 LINUX_SIGEMPTYSET(mask);
815 mask.__bits[0] = args->mask;
816 linux_to_bsd_sigset(&linux_mask, &mask);
818 error = kern_sigsuspend(&mask);
824 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
826 l_sigset_t linux_mask;
831 if (ldebug(rt_sigsuspend))
832 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
833 (void *)uap->newset, uap->sigsetsize);
836 if (uap->sigsetsize != sizeof(l_sigset_t))
839 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
843 linux_to_bsd_sigset(&linux_mask, &mask);
845 error = kern_sigsuspend(&mask);
851 sys_linux_pause(struct linux_pause_args *args)
853 struct thread *td = curthread;
854 struct lwp *lp = td->td_lwp;
860 kprintf(ARGS(pause, ""));
863 mask = lp->lwp_sigmask;
865 error = kern_sigsuspend(&mask);
871 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
878 if (ldebug(sigaltstack))
879 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
883 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
887 ss.ss_sp = linux_ss.ss_sp;
888 ss.ss_size = linux_ss.ss_size;
889 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
892 error = kern_sigaltstack(uap->uss ? &ss : NULL,
893 uap->uoss ? &oss : NULL);
895 if (error == 0 && uap->uoss) {
896 linux_ss.ss_sp = oss.ss_sp;
897 linux_ss.ss_size = oss.ss_size;
898 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
899 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));