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);
122 bzero(&exec_args, sizeof(exec_args));
124 error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
125 args->argp, args->envp);
128 error = kern_execve(&nd, &exec_args);
132 * The syscall result is returned in registers to the new program.
133 * Linux will register %edx as an atexit function and we must be
134 * sure to set it to 0. XXX
137 args->sysmsg_result64 = 0;
139 exec_free_args(&exec_args);
140 linux_free_path(&path);
143 /* We hit a lethal error condition. Let's die now. */
144 exit1(W_EXITCODE(0, SIGABRT));
151 struct l_ipc_kludge {
152 struct l_msgbuf *msgp;
157 sys_linux_ipc(struct linux_ipc_args *args)
161 switch (args->what & 0xFFFF) {
163 struct linux_semop_args a;
165 a.semid = args->arg1;
167 a.nsops = args->arg2;
168 a.sysmsg_lresult = 0;
169 error = linux_semop(&a);
170 args->sysmsg_lresult = a.sysmsg_lresult;
174 struct linux_semget_args a;
177 a.nsems = args->arg2;
178 a.semflg = args->arg3;
179 a.sysmsg_lresult = 0;
180 error = linux_semget(&a);
181 args->sysmsg_lresult = a.sysmsg_lresult;
185 struct linux_semctl_args a;
188 a.semid = args->arg1;
189 a.semnum = args->arg2;
191 a.sysmsg_lresult = 0;
192 error = copyin((caddr_t)args->ptr, &a.arg, sizeof(a.arg));
195 error = linux_semctl(&a);
196 args->sysmsg_lresult = a.sysmsg_lresult;
200 struct linux_msgsnd_args a;
202 a.msqid = args->arg1;
204 a.msgsz = args->arg2;
205 a.msgflg = args->arg3;
206 a.sysmsg_lresult = 0;
207 error = linux_msgsnd(&a);
208 args->sysmsg_lresult = a.sysmsg_lresult;
212 struct linux_msgrcv_args a;
214 a.msqid = args->arg1;
215 a.msgsz = args->arg2;
216 a.msgflg = args->arg3;
217 a.sysmsg_lresult = 0;
218 if ((args->what >> 16) == 0) {
219 struct l_ipc_kludge tmp;
222 if (args->ptr == NULL)
224 error = copyin((caddr_t)args->ptr, &tmp, sizeof(tmp));
228 a.msgtyp = tmp.msgtyp;
231 a.msgtyp = args->arg5;
233 error = linux_msgrcv(&a);
234 args->sysmsg_lresult = a.sysmsg_lresult;
238 struct linux_msgget_args a;
241 a.msgflg = args->arg2;
242 a.sysmsg_lresult = 0;
243 error = linux_msgget(&a);
244 args->sysmsg_lresult = a.sysmsg_lresult;
248 struct linux_msgctl_args a;
250 a.msqid = args->arg1;
253 a.sysmsg_lresult = 0;
254 error = linux_msgctl(&a);
255 args->sysmsg_lresult = a.sysmsg_lresult;
259 struct linux_shmat_args a;
261 a.shmid = args->arg1;
262 a.shmaddr = args->ptr;
263 a.shmflg = args->arg2;
264 a.raddr = (l_ulong *)args->arg3;
265 a.sysmsg_lresult = 0;
266 error = linux_shmat(&a);
267 args->sysmsg_lresult = a.sysmsg_lresult;
271 struct linux_shmdt_args a;
273 a.shmaddr = args->ptr;
274 a.sysmsg_lresult = 0;
275 error = linux_shmdt(&a);
276 args->sysmsg_lresult = a.sysmsg_lresult;
280 struct linux_shmget_args a;
284 a.shmflg = args->arg3;
285 a.sysmsg_lresult = 0;
286 error = linux_shmget(&a);
287 args->sysmsg_lresult = a.sysmsg_lresult;
291 struct linux_shmctl_args a;
293 a.shmid = args->arg1;
296 a.sysmsg_lresult = 0;
297 error = linux_shmctl(&a);
298 args->sysmsg_lresult = a.sysmsg_lresult;
309 sys_linux_old_select(struct linux_old_select_args *args)
311 struct l_old_select_argv linux_args;
312 struct linux_select_args newsel;
316 if (ldebug(old_select))
317 kprintf(ARGS(old_select, "%p"), args->ptr);
320 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
324 newsel.sysmsg_result = 0;
325 newsel.nfds = linux_args.nfds;
326 newsel.readfds = linux_args.readfds;
327 newsel.writefds = linux_args.writefds;
328 newsel.exceptfds = linux_args.exceptfds;
329 newsel.timeout = linux_args.timeout;
330 error = sys_linux_select(&newsel);
331 args->sysmsg_result = newsel.sysmsg_result;
336 sys_linux_fork(struct linux_fork_args *args)
342 kprintf(ARGS(fork, ""));
345 if ((error = sys_fork((struct fork_args *)args)) != 0)
348 if (args->sysmsg_result == 1)
349 args->sysmsg_result = 0;
354 sys_linux_exit_group(struct linux_exit_group_args *args)
356 struct exit_args newargs;
359 newargs.sysmsg_result = 0;
360 newargs.rval = args->rval;
361 error = sys_exit(&newargs);
362 args->sysmsg_result = newargs.sysmsg_result;
367 sys_linux_vfork(struct linux_vfork_args *args)
373 kprintf(ARGS(vfork, ""));
376 if ((error = sys_vfork((struct vfork_args *)args)) != 0)
378 /* Are we the child? */
379 if (args->sysmsg_result == 1)
380 args->sysmsg_result = 0;
384 #define CLONE_VM 0x100
385 #define CLONE_FS 0x200
386 #define CLONE_FILES 0x400
387 #define CLONE_SIGHAND 0x800
388 #define CLONE_PID 0x1000
391 sys_linux_clone(struct linux_clone_args *args)
393 int error, ff = RFPROC;
397 struct rfork_args rf_args;
401 kprintf(ARGS(clone, "flags %x, stack %x"),
402 (unsigned int)args->flags, (unsigned int)args->stack);
403 if (args->flags & CLONE_PID)
404 kprintf(LMSG("CLONE_PID not yet supported"));
411 exit_signal = args->flags & 0x000000ff;
412 if (exit_signal >= LINUX_NSIG)
415 if (exit_signal <= LINUX_SIGTBLSZ)
416 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)];
418 /* RFTHREAD probably not necessary here, but it shouldn't hurt */
421 if (args->flags & CLONE_VM)
423 if (args->flags & CLONE_SIGHAND)
425 if (!(args->flags & CLONE_FILES))
432 rf_args.sysmsg_result = 0;
433 if ((error = sys_rfork(&rf_args)) != 0)
435 args->sysmsg_result = rf_args.sysmsg_result;
437 p2 = pfind(rf_args.sysmsg_result);
441 p2->p_sigparent = exit_signal;
442 ONLY_LWP_IN_PROC(p2)->lwp_md.md_regs->tf_esp =
443 (unsigned int)args->stack;
447 kprintf(LMSG("clone: successful rfork to %ld"),
464 #define STACK_SIZE (2 * 1024 * 1024)
465 #define GUARD_SIZE (4 * PAGE_SIZE)
468 linux_mmap_common(caddr_t linux_addr, size_t linux_len, int linux_prot,
469 int linux_flags, int linux_fd, off_t pos, void **res)
471 struct thread *td = curthread;
472 struct proc *p = td->td_proc;
475 int error, flags, len, prot, fd;
478 if (linux_flags & LINUX_MAP_SHARED)
480 if (linux_flags & LINUX_MAP_PRIVATE)
481 flags |= MAP_PRIVATE;
482 if (linux_flags & LINUX_MAP_FIXED)
484 if (linux_flags & LINUX_MAP_ANON) {
489 if (linux_flags & LINUX_MAP_GROWSDOWN) {
491 /* The linux MAP_GROWSDOWN option does not limit auto
492 * growth of the region. Linux mmap with this option
493 * takes as addr the inital BOS, and as len, the initial
494 * region size. It can then grow down from addr without
495 * limit. However, linux threads has an implicit internal
496 * limit to stack size of STACK_SIZE. Its just not
497 * enforced explicitly in linux. But, here we impose
498 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
499 * region, since we can do this with our mmap.
501 * Our mmap with MAP_STACK takes addr as the maximum
502 * downsize limit on BOS, and as len the max size of
503 * the region. It them maps the top SGROWSIZ bytes,
504 * and autgrows the region down, up to the limit
507 * If we don't use the MAP_STACK option, the effect
508 * of this code is to allocate a stack region of a
509 * fixed size of (STACK_SIZE - GUARD_SIZE).
512 /* This gives us TOS */
513 addr = linux_addr + linux_len;
515 if (addr > p->p_vmspace->vm_maxsaddr) {
516 /* Some linux apps will attempt to mmap
517 * thread stacks near the top of their
518 * address space. If their TOS is greater
519 * than vm_maxsaddr, vm_map_growstack()
520 * will confuse the thread stack with the
521 * process stack and deliver a SEGV if they
522 * attempt to grow the thread stack past their
523 * current stacksize rlimit. To avoid this,
524 * adjust vm_maxsaddr upwards to reflect
525 * the current stacksize rlimit rather
526 * than the maximum possible stacksize.
527 * It would be better to adjust the
528 * mmap'ed region, but some apps do not check
529 * mmap's return value.
531 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
532 p->p_rlimit[RLIMIT_STACK].rlim_cur;
535 /* This gives us our maximum stack size */
536 if (linux_len > STACK_SIZE - GUARD_SIZE) {
539 len = STACK_SIZE - GUARD_SIZE;
541 /* This gives us a new BOS. If we're using VM_STACK, then
542 * mmap will just map the top SGROWSIZ bytes, and let
543 * the stack grow down to the limit at BOS. If we're
544 * not using VM_STACK we map the full stack, since we
545 * don't have a way to autogrow it.
553 prot = linux_prot | PROT_READ;
554 if (linux_flags & LINUX_MAP_ANON) {
561 if (ldebug(mmap) || ldebug(mmap2))
562 kprintf("-> (%p, %d, %d, 0x%08x, %d, %lld)\n",
563 addr, len, prot, flags, fd, pos);
565 error = kern_mmap(curproc->p_vmspace, addr, len,
566 prot, flags, fd, pos, &new);
574 sys_linux_mmap(struct linux_mmap_args *args)
576 struct l_mmap_argv linux_args;
579 error = copyin((caddr_t)args->ptr, &linux_args, sizeof(linux_args));
585 kprintf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"),
586 (void *)linux_args.addr, linux_args.len, linux_args.prot,
587 linux_args.flags, linux_args.fd, linux_args.pos);
589 error = linux_mmap_common(linux_args.addr, linux_args.len,
590 linux_args.prot, linux_args.flags, linux_args.fd,
591 linux_args.pos, &args->sysmsg_resultp);
594 kprintf("-> %p\n", args->sysmsg_resultp);
600 sys_linux_mmap2(struct linux_mmap2_args *args)
606 kprintf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"),
607 (void *)args->addr, args->len, args->prot, args->flags,
608 args->fd, args->pgoff);
610 error = linux_mmap_common((void *)args->addr, args->len, args->prot,
611 args->flags, args->fd, args->pgoff * PAGE_SIZE,
612 &args->sysmsg_resultp);
615 kprintf("-> %p\n", args->sysmsg_resultp);
621 sys_linux_pipe(struct linux_pipe_args *args)
625 struct pipe_args bsd_args;
629 kprintf(ARGS(pipe, "*"));
632 reg_edx = args->sysmsg_fds[1];
633 error = sys_pipe(&bsd_args);
635 args->sysmsg_fds[1] = reg_edx;
639 error = copyout(bsd_args.sysmsg_fds, args->pipefds, 2*sizeof(int));
641 args->sysmsg_fds[1] = reg_edx;
645 args->sysmsg_fds[1] = reg_edx;
646 args->sysmsg_fds[0] = 0;
651 sys_linux_ioperm(struct linux_ioperm_args *args)
653 struct sysarch_args sa;
654 struct i386_ioperm_args *iia;
658 sg = stackgap_init();
659 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args));
660 iia->start = args->start;
661 iia->length = args->length;
662 iia->enable = args->enable;
663 sa.sysmsg_resultp = NULL;
664 sa.op = I386_SET_IOPERM;
665 sa.parms = (char *)iia;
666 error = sys_sysarch(&sa);
667 args->sysmsg_resultp = sa.sysmsg_resultp;
672 sys_linux_iopl(struct linux_iopl_args *args)
674 struct thread *td = curthread;
675 struct lwp *lp = td->td_lwp;
680 if (args->level < 0 || args->level > 3)
682 if ((error = priv_check(td, PRIV_ROOT)) != 0)
686 lp->lwp_md.md_regs->tf_eflags =
687 (lp->lwp_md.md_regs->tf_eflags & ~PSL_IOPL) |
688 (args->level * (PSL_IOPL / 3));
693 sys_linux_modify_ldt(struct linux_modify_ldt_args *uap)
697 struct sysarch_args args;
698 struct i386_ldt_args *ldt;
699 struct l_descriptor ld;
700 union descriptor *desc;
702 sg = stackgap_init();
704 if (uap->ptr == NULL)
708 case 0x00: /* read_ldt */
709 ldt = stackgap_alloc(&sg, sizeof(*ldt));
711 ldt->descs = uap->ptr;
712 ldt->num = uap->bytecount / sizeof(union descriptor);
713 args.op = I386_GET_LDT;
714 args.parms = (char*)ldt;
715 args.sysmsg_result = 0;
716 error = sys_sysarch(&args);
717 uap->sysmsg_result = args.sysmsg_result *
718 sizeof(union descriptor);
720 case 0x01: /* write_ldt */
721 case 0x11: /* write_ldt */
722 if (uap->bytecount != sizeof(ld))
725 error = copyin(uap->ptr, &ld, sizeof(ld));
729 ldt = stackgap_alloc(&sg, sizeof(*ldt));
730 desc = stackgap_alloc(&sg, sizeof(*desc));
731 ldt->start = ld.entry_number;
734 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff);
735 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16;
736 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff);
737 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24;
738 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) |
741 desc->sd.sd_p = (ld.seg_not_present ^ 1);
743 desc->sd.sd_def32 = ld.seg_32bit;
744 desc->sd.sd_gran = ld.limit_in_pages;
745 args.op = I386_SET_LDT;
746 args.parms = (char*)ldt;
747 args.sysmsg_result = 0;
748 error = sys_sysarch(&args);
749 uap->sysmsg_result = args.sysmsg_result;
760 sys_linux_sigaction(struct linux_sigaction_args *args)
763 l_sigaction_t linux_act, linux_oact;
764 struct sigaction act, oact;
768 if (ldebug(sigaction))
769 kprintf(ARGS(sigaction, "%d, %p, %p"),
770 args->sig, (void *)args->nsa, (void *)args->osa);
774 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
777 linux_act.lsa_handler = osa.lsa_handler;
778 linux_act.lsa_flags = osa.lsa_flags;
779 linux_act.lsa_restorer = osa.lsa_restorer;
780 LINUX_SIGEMPTYSET(linux_act.lsa_mask);
781 linux_act.lsa_mask.__bits[0] = osa.lsa_mask;
782 linux_to_bsd_sigaction(&linux_act, &act);
785 error = kern_sigaction(args->sig, args->nsa ? &act : NULL,
786 args->osa ? &oact : NULL);
788 if (args->osa != NULL && !error) {
789 bsd_to_linux_sigaction(&oact, &linux_oact);
790 osa.lsa_handler = linux_oact.lsa_handler;
791 osa.lsa_flags = linux_oact.lsa_flags;
792 osa.lsa_restorer = linux_oact.lsa_restorer;
793 osa.lsa_mask = linux_oact.lsa_mask.__bits[0];
794 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
800 * Linux has two extra args, restart and oldmask. We dont use these,
801 * but it seems that "restart" is actually a context pointer that
802 * enables the signal to happen with a different register set.
805 sys_linux_sigsuspend(struct linux_sigsuspend_args *args)
807 l_sigset_t linux_mask;
812 if (ldebug(sigsuspend))
813 kprintf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
816 LINUX_SIGEMPTYSET(mask);
817 mask.__bits[0] = args->mask;
818 linux_to_bsd_sigset(&linux_mask, &mask);
820 error = kern_sigsuspend(&mask);
826 sys_linux_rt_sigsuspend(struct linux_rt_sigsuspend_args *uap)
828 l_sigset_t linux_mask;
833 if (ldebug(rt_sigsuspend))
834 kprintf(ARGS(rt_sigsuspend, "%p, %d"),
835 (void *)uap->newset, uap->sigsetsize);
838 if (uap->sigsetsize != sizeof(l_sigset_t))
841 error = copyin(uap->newset, &linux_mask, sizeof(l_sigset_t));
845 linux_to_bsd_sigset(&linux_mask, &mask);
847 error = kern_sigsuspend(&mask);
853 sys_linux_pause(struct linux_pause_args *args)
855 struct thread *td = curthread;
856 struct lwp *lp = td->td_lwp;
862 kprintf(ARGS(pause, ""));
865 mask = lp->lwp_sigmask;
867 error = kern_sigsuspend(&mask);
873 sys_linux_sigaltstack(struct linux_sigaltstack_args *uap)
880 if (ldebug(sigaltstack))
881 kprintf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
885 error = copyin(uap->uss, &linux_ss, sizeof(l_stack_t));
889 ss.ss_sp = linux_ss.ss_sp;
890 ss.ss_size = linux_ss.ss_size;
891 ss.ss_flags = linux_to_bsd_sigaltstack(linux_ss.ss_flags);
894 error = kern_sigaltstack(uap->uss ? &ss : NULL,
895 uap->uoss ? &oss : NULL);
897 if (error == 0 && uap->uoss) {
898 linux_ss.ss_sp = oss.ss_sp;
899 linux_ss.ss_size = oss.ss_size;
900 linux_ss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
901 error = copyout(&linux_ss, uap->uoss, sizeof(l_stack_t));