2 * Copyright (c) 2006 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * $DragonFly: src/sys/vm/vm_vmspace.c,v 1.14 2007/08/15 03:15:07 dillon Exp $
37 #include <sys/param.h>
38 #include <sys/kernel.h>
39 #include <sys/systm.h>
40 #include <sys/sysproto.h>
41 #include <sys/kern_syscall.h>
44 #include <sys/malloc.h>
45 #include <sys/sysctl.h>
46 #include <sys/vkernel.h>
47 #include <sys/vmspace.h>
49 #include <vm/vm_extern.h>
52 #include <machine/vmparam.h>
54 #include <sys/spinlock2.h>
55 #include <sys/sysref2.h>
57 static struct vmspace_entry *vkernel_find_vmspace(struct vkernel_proc *vkp,
59 static void vmspace_entry_delete(struct vmspace_entry *ve,
60 struct vkernel_proc *vkp);
62 static MALLOC_DEFINE(M_VKERNEL, "vkernel", "VKernel structures");
65 * vmspace_create (void *id, int type, void *data)
67 * Create a VMSPACE under the control of the caller with the specified id.
68 * An id of NULL cannot be used. The type and data fields must currently
71 * The vmspace starts out completely empty. Memory may be mapped into the
72 * VMSPACE with vmspace_mmap() and MAP_VPAGETABLE section(s) controlled
73 * with vmspace_mcontrol().
78 sys_vmspace_create(struct vmspace_create_args *uap)
80 struct vmspace_entry *ve;
81 struct vkernel_proc *vkp;
84 if (vkernel_enable == 0)
88 * Create a virtual kernel side-structure for the process if one
92 if ((vkp = curproc->p_vkernel) == NULL) {
93 vkp = kmalloc(sizeof(*vkp), M_VKERNEL, M_WAITOK|M_ZERO);
95 spin_init(&vkp->spin);
97 curproc->p_vkernel = vkp;
101 * Create a new VMSPACE
103 * XXX race if kmalloc blocks
105 if (vkernel_find_vmspace(vkp, uap->id)) {
109 ve = kmalloc(sizeof(struct vmspace_entry), M_VKERNEL, M_WAITOK|M_ZERO);
110 ve->vmspace = vmspace_alloc(VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
112 pmap_pinit2(vmspace_pmap(ve->vmspace));
113 RB_INSERT(vmspace_rb_tree, &vkp->root, ve);
121 * vmspace_destroy (void *id)
128 sys_vmspace_destroy(struct vmspace_destroy_args *uap)
130 struct vkernel_proc *vkp;
131 struct vmspace_entry *ve;
135 if ((vkp = curproc->p_vkernel) == NULL) {
139 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
147 vmspace_entry_delete(ve, vkp);
155 * vmspace_ctl (void *id, int cmd, struct trapframe *tframe,
156 * struct vextframe *vframe);
158 * Transfer control to a VMSPACE. Control is returned after the specified
159 * number of microseconds or if a page fault, signal, trap, or system call
160 * occurs. The context is updated as appropriate.
165 sys_vmspace_ctl(struct vmspace_ctl_args *uap)
167 struct vkernel_proc *vkp;
168 struct vkernel_lwp *vklp;
169 struct vmspace_entry *ve;
175 lp = curthread->td_lwp;
179 if ((vkp = p->p_vkernel) == NULL) {
183 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
189 * Signal mailbox interlock
191 if (p->p_flag & P_MAILBOX) {
192 p->p_flag &= ~P_MAILBOX;
198 case VMSPACE_CTL_RUN:
200 * Save the caller's register context, swap VM spaces, and
201 * install the passed register context. Return with
202 * EJUSTRETURN so the syscall code doesn't adjust the context.
204 atomic_add_int(&ve->refs, 1);
205 framesz = sizeof(struct trapframe);
206 if ((vklp = lp->lwp_vkernel) == NULL) {
207 vklp = kmalloc(sizeof(*vklp), M_VKERNEL,
209 lp->lwp_vkernel = vklp;
211 vklp->user_trapframe = uap->tframe;
212 vklp->user_vextframe = uap->vframe;
213 bcopy(uap->sysmsg_frame, &vklp->save_trapframe, framesz);
214 bcopy(&curthread->td_tls, &vklp->save_vextframe.vx_tls,
215 sizeof(vklp->save_vextframe.vx_tls));
216 error = copyin(uap->tframe, uap->sysmsg_frame, framesz);
218 error = copyin(&uap->vframe->vx_tls, &curthread->td_tls, sizeof(struct savetls));
220 error = cpu_sanitize_frame(uap->sysmsg_frame);
222 error = cpu_sanitize_tls(&curthread->td_tls);
224 bcopy(&vklp->save_trapframe, uap->sysmsg_frame, framesz);
225 bcopy(&vklp->save_vextframe.vx_tls, &curthread->td_tls,
226 sizeof(vklp->save_vextframe.vx_tls));
228 atomic_subtract_int(&ve->refs, 1);
231 pmap_setlwpvm(lp, ve->vmspace);
233 set_vkernel_fp(uap->sysmsg_frame);
247 * vmspace_mmap(id, addr, len, prot, flags, fd, offset)
249 * map memory within a VMSPACE. This function is just like a normal mmap()
250 * but operates on the vmspace's memory map. Most callers use this to create
251 * a MAP_VPAGETABLE mapping.
256 sys_vmspace_mmap(struct vmspace_mmap_args *uap)
258 struct vkernel_proc *vkp;
259 struct vmspace_entry *ve;
263 if ((vkp = curproc->p_vkernel) == NULL) {
267 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
271 error = kern_mmap(ve->vmspace, uap->addr, uap->len,
272 uap->prot, uap->flags,
273 uap->fd, uap->offset, &uap->sysmsg_resultp);
280 * vmspace_munmap(id, addr, len)
282 * unmap memory within a VMSPACE.
287 sys_vmspace_munmap(struct vmspace_munmap_args *uap)
289 struct vkernel_proc *vkp;
290 struct vmspace_entry *ve;
293 vm_size_t size, pageoff;
298 if ((vkp = curproc->p_vkernel) == NULL) {
302 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
308 * Copied from sys_munmap()
310 addr = (vm_offset_t)uap->addr;
313 pageoff = (addr & PAGE_MASK);
316 size = (vm_size_t)round_page(size);
317 if (size < uap->len) { /* wrap */
321 tmpaddr = addr + size; /* workaround gcc4 opt */
322 if (tmpaddr < addr) { /* wrap */
331 if (VM_MAX_USER_ADDRESS > 0 && tmpaddr > VM_MAX_USER_ADDRESS) {
335 if (VM_MIN_USER_ADDRESS > 0 && addr < VM_MIN_USER_ADDRESS) {
339 map = &ve->vmspace->vm_map;
340 if (!vm_map_check_protection(map, addr, tmpaddr, VM_PROT_NONE)) {
344 vm_map_remove(map, addr, addr + size);
352 * vmspace_pread(id, buf, nbyte, flags, offset)
354 * Read data from a vmspace. The number of bytes read is returned or
355 * -1 if an unrecoverable error occured. If the number of bytes read is
356 * less then the request size, a page fault occured in the VMSPACE which
357 * the caller must resolve in order to proceed.
359 * (not implemented yet)
364 sys_vmspace_pread(struct vmspace_pread_args *uap)
366 struct vkernel_proc *vkp;
367 struct vmspace_entry *ve;
371 if ((vkp = curproc->p_vkernel) == NULL) {
375 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
386 * vmspace_pwrite(id, buf, nbyte, flags, offset)
388 * Write data to a vmspace. The number of bytes written is returned or
389 * -1 if an unrecoverable error occured. If the number of bytes written is
390 * less then the request size, a page fault occured in the VMSPACE which
391 * the caller must resolve in order to proceed.
393 * (not implemented yet)
398 sys_vmspace_pwrite(struct vmspace_pwrite_args *uap)
400 struct vkernel_proc *vkp;
401 struct vmspace_entry *ve;
405 if ((vkp = curproc->p_vkernel) == NULL) {
409 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
420 * vmspace_mcontrol(id, addr, len, behav, value)
422 * madvise/mcontrol support for a vmspace.
427 sys_vmspace_mcontrol(struct vmspace_mcontrol_args *uap)
429 struct vkernel_proc *vkp;
430 struct vmspace_entry *ve;
431 vm_offset_t start, end;
432 vm_offset_t tmpaddr = (vm_offset_t)uap->addr + uap->len;
436 if ((vkp = curproc->p_vkernel) == NULL) {
440 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL) {
446 * This code is basically copied from sys_mcontrol()
448 if (uap->behav < 0 || uap->behav > MADV_CONTROL_END) {
453 if (tmpaddr < (vm_offset_t)uap->addr) {
457 if (VM_MAX_USER_ADDRESS > 0 && tmpaddr > VM_MAX_USER_ADDRESS) {
461 if (VM_MIN_USER_ADDRESS > 0 && uap->addr < VM_MIN_USER_ADDRESS) {
466 start = trunc_page((vm_offset_t) uap->addr);
467 end = round_page(tmpaddr);
469 error = vm_map_madvise(&ve->vmspace->vm_map, start, end,
470 uap->behav, uap->value);
477 * Red black tree functions
479 static int rb_vmspace_compare(struct vmspace_entry *, struct vmspace_entry *);
480 RB_GENERATE(vmspace_rb_tree, vmspace_entry, rb_entry, rb_vmspace_compare);
482 /* a->start is address, and the only field has to be initialized */
484 rb_vmspace_compare(struct vmspace_entry *a, struct vmspace_entry *b)
486 if ((char *)a->id < (char *)b->id)
488 else if ((char *)a->id > (char *)b->id)
495 rb_vmspace_delete(struct vmspace_entry *ve, void *data)
497 struct vkernel_proc *vkp = data;
499 KKASSERT(ve->refs == 0);
500 vmspace_entry_delete(ve, vkp);
505 * Remove a vmspace_entry from the RB tree and destroy it. We have to clean
506 * up the pmap, the vm_map, then destroy the vmspace.
510 vmspace_entry_delete(struct vmspace_entry *ve, struct vkernel_proc *vkp)
512 RB_REMOVE(vmspace_rb_tree, &vkp->root, ve);
514 pmap_remove_pages(vmspace_pmap(ve->vmspace),
515 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
516 vm_map_remove(&ve->vmspace->vm_map,
517 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
518 sysref_put(&ve->vmspace->vm_sysref);
519 kfree(ve, M_VKERNEL);
524 struct vmspace_entry *
525 vkernel_find_vmspace(struct vkernel_proc *vkp, void *id)
527 struct vmspace_entry *ve;
528 struct vmspace_entry key;
531 ve = RB_FIND(vmspace_rb_tree, &vkp->root, &key);
536 * Manage vkernel refs, used by the kernel when fork()ing or exit()ing
540 vkernel_inherit(struct proc *p1, struct proc *p2)
542 struct vkernel_proc *vkp;
545 KKASSERT(vkp->refs > 0);
546 atomic_add_int(&vkp->refs, 1);
551 vkernel_exit(struct proc *p)
553 struct vkernel_proc *vkp;
559 * Restore the original VM context if we are killed while running
562 * This isn't supposed to happen. What is supposed to happen is
563 * that the process should enter vkernel_trap() before the handling
566 RB_FOREACH(lp, lwp_rb_tree, &p->p_lwp_tree) {
567 vkernel_lwp_exit(lp);
571 * Dereference the common area
574 KKASSERT(vkp->refs > 0);
575 spin_lock_wr(&vkp->spin);
576 if (--vkp->refs == 0)
578 spin_unlock_wr(&vkp->spin);
581 RB_SCAN(vmspace_rb_tree, &vkp->root, NULL,
582 rb_vmspace_delete, vkp);
583 kfree(vkp, M_VKERNEL);
588 vkernel_lwp_exit(struct lwp *lp)
590 struct vkernel_lwp *vklp;
591 struct vmspace_entry *ve;
593 if ((vklp = lp->lwp_vkernel) != NULL) {
594 if ((ve = vklp->ve) != NULL) {
595 kprintf("Warning, pid %d killed with "
596 "active VC!\n", lp->lwp_proc->p_pid);
598 pmap_setlwpvm(lp, lp->lwp_proc->p_vmspace);
600 KKASSERT(ve->refs > 0);
601 atomic_subtract_int(&ve->refs, 1);
603 lp->lwp_vkernel = NULL;
604 kfree(vklp, M_VKERNEL);
609 * A VM space under virtual kernel control trapped out or made a system call
610 * or otherwise needs to return control to the virtual kernel context.
613 vkernel_trap(struct lwp *lp, struct trapframe *frame)
615 struct proc *p = lp->lwp_proc;
616 struct vmspace_entry *ve;
617 struct vkernel_lwp *vklp;
621 * Which vmspace entry was running?
623 vklp = lp->lwp_vkernel;
626 KKASSERT(ve != NULL);
629 * Switch the LWP vmspace back to the virtual kernel's VM space.
632 pmap_setlwpvm(lp, p->p_vmspace);
633 KKASSERT(ve->refs > 0);
634 atomic_subtract_int(&ve->refs, 1);
637 * Copy the emulated process frame to the virtual kernel process.
638 * The emulated process cannot change TLS descriptors so don't
639 * bother saving them, we already have a copy.
641 * Restore the virtual kernel's saved context so the virtual kernel
642 * process can resume.
644 error = copyout(frame, vklp->user_trapframe, sizeof(*frame));
645 bcopy(&vklp->save_trapframe, frame, sizeof(*frame));
646 bcopy(&vklp->save_vextframe.vx_tls, &curthread->td_tls,
647 sizeof(vklp->save_vextframe.vx_tls));