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().
76 sys_vmspace_create(struct vmspace_create_args *uap)
78 struct vmspace_entry *ve;
79 struct vkernel_proc *vkp;
81 if (vkernel_enable == 0)
85 * Create a virtual kernel side-structure for the process if one
88 if ((vkp = curproc->p_vkernel) == NULL) {
89 vkp = kmalloc(sizeof(*vkp), M_VKERNEL, M_WAITOK|M_ZERO);
91 spin_init(&vkp->spin);
93 curproc->p_vkernel = vkp;
97 * Create a new VMSPACE
99 if (vkernel_find_vmspace(vkp, uap->id))
101 ve = kmalloc(sizeof(struct vmspace_entry), M_VKERNEL, M_WAITOK|M_ZERO);
102 ve->vmspace = vmspace_alloc(VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
104 pmap_pinit2(vmspace_pmap(ve->vmspace));
105 RB_INSERT(vmspace_rb_tree, &vkp->root, ve);
110 * vmspace_destroy (void *id)
115 sys_vmspace_destroy(struct vmspace_destroy_args *uap)
117 struct vkernel_proc *vkp;
118 struct vmspace_entry *ve;
120 if ((vkp = curproc->p_vkernel) == NULL)
122 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
126 vmspace_entry_delete(ve, vkp);
131 * vmspace_ctl (void *id, int cmd, struct trapframe *tframe,
132 * struct vextframe *vframe);
134 * Transfer control to a VMSPACE. Control is returned after the specified
135 * number of microseconds or if a page fault, signal, trap, or system call
136 * occurs. The context is updated as appropriate.
139 sys_vmspace_ctl(struct vmspace_ctl_args *uap)
141 struct vkernel_proc *vkp;
142 struct vkernel_lwp *vklp;
143 struct vmspace_entry *ve;
149 lp = curthread->td_lwp;
152 if ((vkp = p->p_vkernel) == NULL)
154 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
158 * Signal mailbox interlock
160 if (p->p_flag & P_MAILBOX) {
161 p->p_flag &= ~P_MAILBOX;
166 case VMSPACE_CTL_RUN:
168 * Save the caller's register context, swap VM spaces, and
169 * install the passed register context. Return with
170 * EJUSTRETURN so the syscall code doesn't adjust the context.
172 atomic_add_int(&ve->refs, 1);
173 framesz = sizeof(struct trapframe);
174 if ((vklp = lp->lwp_vkernel) == NULL) {
175 vklp = kmalloc(sizeof(*vklp), M_VKERNEL,
177 lp->lwp_vkernel = vklp;
179 vklp->user_trapframe = uap->tframe;
180 vklp->user_vextframe = uap->vframe;
181 bcopy(uap->sysmsg_frame, &vklp->save_trapframe, framesz);
182 bcopy(&curthread->td_tls, &vklp->save_vextframe.vx_tls,
183 sizeof(vklp->save_vextframe.vx_tls));
184 error = copyin(uap->tframe, uap->sysmsg_frame, framesz);
186 error = copyin(&uap->vframe->vx_tls, &curthread->td_tls, sizeof(struct savetls));
188 error = cpu_sanitize_frame(uap->sysmsg_frame);
190 error = cpu_sanitize_tls(&curthread->td_tls);
192 bcopy(&vklp->save_trapframe, uap->sysmsg_frame, framesz);
193 bcopy(&vklp->save_vextframe.vx_tls, &curthread->td_tls,
194 sizeof(vklp->save_vextframe.vx_tls));
196 atomic_subtract_int(&ve->refs, 1);
199 pmap_setlwpvm(lp, ve->vmspace);
201 set_vkernel_fp(uap->sysmsg_frame);
213 * vmspace_mmap(id, addr, len, prot, flags, fd, offset)
215 * map memory within a VMSPACE. This function is just like a normal mmap()
216 * but operates on the vmspace's memory map. Most callers use this to create
217 * a MAP_VPAGETABLE mapping.
220 sys_vmspace_mmap(struct vmspace_mmap_args *uap)
222 struct vkernel_proc *vkp;
223 struct vmspace_entry *ve;
226 if ((vkp = curproc->p_vkernel) == NULL)
228 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
230 error = kern_mmap(ve->vmspace, uap->addr, uap->len,
231 uap->prot, uap->flags,
232 uap->fd, uap->offset, &uap->sysmsg_resultp);
237 * vmspace_munmap(id, addr, len)
239 * unmap memory within a VMSPACE.
242 sys_vmspace_munmap(struct vmspace_munmap_args *uap)
244 struct vkernel_proc *vkp;
245 struct vmspace_entry *ve;
248 vm_size_t size, pageoff;
251 if ((vkp = curproc->p_vkernel) == NULL)
253 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
257 * Copied from sys_munmap()
259 addr = (vm_offset_t)uap->addr;
262 pageoff = (addr & PAGE_MASK);
265 size = (vm_size_t)round_page(size);
266 if (size < uap->len) /* wrap */
268 tmpaddr = addr + size; /* workaround gcc4 opt */
269 if (tmpaddr < addr) /* wrap */
274 if (VM_MAX_USER_ADDRESS > 0 && tmpaddr > VM_MAX_USER_ADDRESS)
276 if (VM_MIN_USER_ADDRESS > 0 && addr < VM_MIN_USER_ADDRESS)
278 map = &ve->vmspace->vm_map;
279 if (!vm_map_check_protection(map, addr, tmpaddr, VM_PROT_NONE))
281 vm_map_remove(map, addr, addr + size);
286 * vmspace_pread(id, buf, nbyte, flags, offset)
288 * Read data from a vmspace. The number of bytes read is returned or
289 * -1 if an unrecoverable error occured. If the number of bytes read is
290 * less then the request size, a page fault occured in the VMSPACE which
291 * the caller must resolve in order to proceed.
294 sys_vmspace_pread(struct vmspace_pread_args *uap)
296 struct vkernel_proc *vkp;
297 struct vmspace_entry *ve;
299 if ((vkp = curproc->p_vkernel) == NULL)
301 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
307 * vmspace_pwrite(id, buf, nbyte, flags, offset)
309 * Write data to a vmspace. The number of bytes written is returned or
310 * -1 if an unrecoverable error occured. If the number of bytes written is
311 * less then the request size, a page fault occured in the VMSPACE which
312 * the caller must resolve in order to proceed.
315 sys_vmspace_pwrite(struct vmspace_pwrite_args *uap)
317 struct vkernel_proc *vkp;
318 struct vmspace_entry *ve;
320 if ((vkp = curproc->p_vkernel) == NULL)
322 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
328 * vmspace_mcontrol(id, addr, len, behav, value)
330 * madvise/mcontrol support for a vmspace.
333 sys_vmspace_mcontrol(struct vmspace_mcontrol_args *uap)
335 struct vkernel_proc *vkp;
336 struct vmspace_entry *ve;
337 vm_offset_t start, end;
338 vm_offset_t tmpaddr = (vm_offset_t)uap->addr + uap->len;
340 if ((vkp = curproc->p_vkernel) == NULL)
342 if ((ve = vkernel_find_vmspace(vkp, uap->id)) == NULL)
346 * This code is basically copied from sys_mcontrol()
348 if (uap->behav < 0 || uap->behav > MADV_CONTROL_END)
351 if (tmpaddr < (vm_offset_t)uap->addr)
353 if (VM_MAX_USER_ADDRESS > 0 && tmpaddr > VM_MAX_USER_ADDRESS)
355 if (VM_MIN_USER_ADDRESS > 0 && uap->addr < VM_MIN_USER_ADDRESS)
358 start = trunc_page((vm_offset_t) uap->addr);
359 end = round_page(tmpaddr);
361 return (vm_map_madvise(&ve->vmspace->vm_map, start, end,
362 uap->behav, uap->value));
366 * Red black tree functions
368 static int rb_vmspace_compare(struct vmspace_entry *, struct vmspace_entry *);
369 RB_GENERATE(vmspace_rb_tree, vmspace_entry, rb_entry, rb_vmspace_compare);
371 /* a->start is address, and the only field has to be initialized */
373 rb_vmspace_compare(struct vmspace_entry *a, struct vmspace_entry *b)
375 if ((char *)a->id < (char *)b->id)
377 else if ((char *)a->id > (char *)b->id)
384 rb_vmspace_delete(struct vmspace_entry *ve, void *data)
386 struct vkernel_proc *vkp = data;
388 KKASSERT(ve->refs == 0);
389 vmspace_entry_delete(ve, vkp);
394 * Remove a vmspace_entry from the RB tree and destroy it. We have to clean
395 * up the pmap, the vm_map, then destroy the vmspace.
399 vmspace_entry_delete(struct vmspace_entry *ve, struct vkernel_proc *vkp)
401 RB_REMOVE(vmspace_rb_tree, &vkp->root, ve);
403 pmap_remove_pages(vmspace_pmap(ve->vmspace),
404 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
405 vm_map_remove(&ve->vmspace->vm_map,
406 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
407 sysref_put(&ve->vmspace->vm_sysref);
408 kfree(ve, M_VKERNEL);
413 struct vmspace_entry *
414 vkernel_find_vmspace(struct vkernel_proc *vkp, void *id)
416 struct vmspace_entry *ve;
417 struct vmspace_entry key;
420 ve = RB_FIND(vmspace_rb_tree, &vkp->root, &key);
425 * Manage vkernel refs, used by the kernel when fork()ing or exit()ing
429 vkernel_inherit(struct proc *p1, struct proc *p2)
431 struct vkernel_proc *vkp;
434 KKASSERT(vkp->refs > 0);
435 atomic_add_int(&vkp->refs, 1);
440 vkernel_exit(struct proc *p)
442 struct vkernel_proc *vkp;
448 * Restore the original VM context if we are killed while running
451 * This isn't supposed to happen. What is supposed to happen is
452 * that the process should enter vkernel_trap() before the handling
455 RB_FOREACH(lp, lwp_rb_tree, &p->p_lwp_tree) {
456 vkernel_lwp_exit(lp);
460 * Dereference the common area
463 KKASSERT(vkp->refs > 0);
464 spin_lock_wr(&vkp->spin);
465 if (--vkp->refs == 0)
467 spin_unlock_wr(&vkp->spin);
470 RB_SCAN(vmspace_rb_tree, &vkp->root, NULL,
471 rb_vmspace_delete, vkp);
472 kfree(vkp, M_VKERNEL);
477 vkernel_lwp_exit(struct lwp *lp)
479 struct vkernel_lwp *vklp;
480 struct vmspace_entry *ve;
482 if ((vklp = lp->lwp_vkernel) != NULL) {
483 if ((ve = vklp->ve) != NULL) {
484 kprintf("Warning, pid %d killed with "
485 "active VC!\n", lp->lwp_proc->p_pid);
487 pmap_setlwpvm(lp, lp->lwp_proc->p_vmspace);
489 KKASSERT(ve->refs > 0);
490 atomic_subtract_int(&ve->refs, 1);
492 lp->lwp_vkernel = NULL;
493 kfree(vklp, M_VKERNEL);
498 * A VM space under virtual kernel control trapped out or made a system call
499 * or otherwise needs to return control to the virtual kernel context.
502 vkernel_trap(struct lwp *lp, struct trapframe *frame)
504 struct proc *p = lp->lwp_proc;
505 struct vmspace_entry *ve;
506 struct vkernel_lwp *vklp;
510 * Which vmspace entry was running?
512 vklp = lp->lwp_vkernel;
515 KKASSERT(ve != NULL);
518 * Switch the LWP vmspace back to the virtual kernel's VM space.
521 pmap_setlwpvm(lp, p->p_vmspace);
522 KKASSERT(ve->refs > 0);
523 atomic_subtract_int(&ve->refs, 1);
526 * Copy the emulated process frame to the virtual kernel process.
527 * The emulated process cannot change TLS descriptors so don't
528 * bother saving them, we already have a copy.
530 * Restore the virtual kernel's saved context so the virtual kernel
531 * process can resume.
533 error = copyout(frame, vklp->user_trapframe, sizeof(*frame));
534 bcopy(&vklp->save_trapframe, frame, sizeof(*frame));
535 bcopy(&vklp->save_vextframe.vx_tls, &curthread->td_tls,
536 sizeof(vklp->save_vextframe.vx_tls));