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 1.10 2007/01/14 07:59:09 dillon Exp $
38 #include <sys/param.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/sysproto.h>
42 #include <sys/kern_syscall.h>
45 #include <sys/malloc.h>
46 #include <sys/sysctl.h>
47 #include <sys/vkernel.h>
48 #include <sys/vmspace.h>
49 #include <sys/spinlock2.h>
51 #include <vm/vm_extern.h>
55 #include <machine/vmparam.h>
57 static struct vmspace_entry *vkernel_find_vmspace(struct vkernel_common *vc,
59 static void vmspace_entry_delete(struct vmspace_entry *ve,
60 struct vkernel_common *vc);
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 vkernel_common *vc;
79 struct vmspace_entry *ve;
82 if (vkernel_enable == 0)
86 * Create a virtual kernel side-structure for the process if one
89 if ((vk = curproc->p_vkernel) == NULL) {
90 vk = kmalloc(sizeof(*vk), M_VKERNEL, M_WAITOK|M_ZERO);
91 vc = kmalloc(sizeof(*vc), M_VKERNEL, M_WAITOK|M_ZERO);
93 spin_init(&vc->vc_spin);
94 RB_INIT(&vc->vc_root);
96 curproc->p_vkernel = vk;
101 * Create a new VMSPACE
103 if (vkernel_find_vmspace(vc, uap->id))
105 ve = kmalloc(sizeof(struct vmspace_entry), M_VKERNEL, M_WAITOK|M_ZERO);
106 ve->vmspace = vmspace_alloc(VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
108 pmap_pinit2(vmspace_pmap(ve->vmspace));
109 RB_INSERT(vmspace_rb_tree, &vc->vc_root, ve);
114 * vmspace_destroy (void *id)
119 sys_vmspace_destroy(struct vmspace_destroy_args *uap)
121 struct vkernel_common *vc;
122 struct vmspace_entry *ve;
125 if ((vk = curproc->p_vkernel) == NULL)
128 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
132 vmspace_entry_delete(ve, vc);
137 * vmspace_ctl (void *id, int cmd, struct trapframe *tframe,
138 * struct vextframe *vframe);
140 * Transfer control to a VMSPACE. Control is returned after the specified
141 * number of microseconds or if a page fault, signal, trap, or system call
142 * occurs. The context is updated as appropriate.
145 sys_vmspace_ctl(struct vmspace_ctl_args *uap)
147 struct vkernel_common *vc;
148 struct vmspace_entry *ve;
154 if ((vk = curproc->p_vkernel) == NULL)
157 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
161 * Signal mailbox interlock
163 if (curproc->p_flag & P_MAILBOX) {
164 curproc->p_flag &= ~P_MAILBOX;
169 case VMSPACE_CTL_RUN:
171 * Save the caller's register context, swap VM spaces, and
172 * install the passed register context. Return with
173 * EJUSTRETURN so the syscall code doesn't adjust the context.
177 framesz = sizeof(struct trapframe);
179 vk->vk_save_vmspace = p->p_vmspace;
180 vk->vk_user_trapframe = uap->tframe;
181 vk->vk_user_vextframe = uap->vframe;
182 bcopy(uap->sysmsg_frame, &vk->vk_save_trapframe, framesz);
183 bcopy(&curthread->td_tls, &vk->vk_save_vextframe.vx_tls,
184 sizeof(vk->vk_save_vextframe.vx_tls));
185 error = copyin(uap->tframe, uap->sysmsg_frame, framesz);
187 error = copyin(&uap->vframe->vx_tls, &curthread->td_tls, sizeof(struct savetls));
189 error = cpu_sanitize_frame(uap->sysmsg_frame);
191 error = cpu_sanitize_tls(&curthread->td_tls);
193 bcopy(&vk->vk_save_trapframe, uap->sysmsg_frame, framesz);
194 bcopy(&vk->vk_save_vextframe.vx_tls, &curthread->td_tls,
195 sizeof(vk->vk_save_vextframe.vx_tls));
197 vk->vk_current = NULL;
198 vk->vk_save_vmspace = NULL;
202 p->p_vmspace = ve->vmspace;
205 set_vkernel_fp(uap->sysmsg_frame);
217 * vmspace_mmap(id, addr, len, prot, flags, fd, offset)
219 * map memory within a VMSPACE. This function is just like a normal mmap()
220 * but operates on the vmspace's memory map. Most callers use this to create
221 * a MAP_VPAGETABLE mapping.
224 sys_vmspace_mmap(struct vmspace_mmap_args *uap)
226 struct vkernel_common *vc;
227 struct vmspace_entry *ve;
231 if ((vk = curproc->p_vkernel) == NULL)
234 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
236 error = kern_mmap(ve->vmspace, uap->addr, uap->len,
237 uap->prot, uap->flags,
238 uap->fd, uap->offset, &uap->sysmsg_resultp);
243 * vmspace_munmap(id, addr, len)
245 * unmap memory within a VMSPACE.
248 sys_vmspace_munmap(struct vmspace_munmap_args *uap)
250 struct vkernel_common *vc;
251 struct vmspace_entry *ve;
254 vm_size_t size, pageoff;
257 if ((vk = curproc->p_vkernel) == NULL)
260 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
264 * Copied from sys_munmap()
266 addr = (vm_offset_t)uap->addr;
269 pageoff = (addr & PAGE_MASK);
272 size = (vm_size_t)round_page(size);
273 if (addr + size < addr)
278 if (VM_MAX_USER_ADDRESS > 0 && addr + size > VM_MAX_USER_ADDRESS)
280 if (VM_MIN_USER_ADDRESS > 0 && addr < VM_MIN_USER_ADDRESS)
282 map = &ve->vmspace->vm_map;
283 if (!vm_map_check_protection(map, addr, addr + size, VM_PROT_NONE))
285 vm_map_remove(map, addr, addr + size);
290 * vmspace_pread(id, buf, nbyte, flags, offset)
292 * Read data from a vmspace. The number of bytes read is returned or
293 * -1 if an unrecoverable error occured. If the number of bytes read is
294 * less then the request size, a page fault occured in the VMSPACE which
295 * the caller must resolve in order to proceed.
298 sys_vmspace_pread(struct vmspace_pread_args *uap)
300 struct vkernel_common *vc;
301 struct vmspace_entry *ve;
304 if ((vk = curproc->p_vkernel) == NULL)
307 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
313 * vmspace_pwrite(id, buf, nbyte, flags, offset)
315 * Write data to a vmspace. The number of bytes written is returned or
316 * -1 if an unrecoverable error occured. If the number of bytes written is
317 * less then the request size, a page fault occured in the VMSPACE which
318 * the caller must resolve in order to proceed.
321 sys_vmspace_pwrite(struct vmspace_pwrite_args *uap)
323 struct vkernel_common *vc;
324 struct vmspace_entry *ve;
327 if ((vk = curproc->p_vkernel) == NULL)
330 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
336 * vmspace_mcontrol(id, addr, len, behav, value)
338 * madvise/mcontrol support for a vmspace.
341 sys_vmspace_mcontrol(struct vmspace_mcontrol_args *uap)
343 struct vkernel_common *vc;
344 struct vmspace_entry *ve;
346 vm_offset_t start, end;
348 if ((vk = curproc->p_vkernel) == NULL)
351 if ((ve = vkernel_find_vmspace(vc, uap->id)) == NULL)
355 * This code is basically copied from sys_mcontrol()
357 if (uap->behav < 0 || uap->behav > MADV_CONTROL_END)
360 if (VM_MAX_USER_ADDRESS > 0 &&
361 ((vm_offset_t) uap->addr + uap->len) > VM_MAX_USER_ADDRESS)
363 if (VM_MIN_USER_ADDRESS > 0 && uap->addr < VM_MIN_USER_ADDRESS)
365 if (((vm_offset_t) uap->addr + uap->len) < (vm_offset_t) uap->addr)
368 start = trunc_page((vm_offset_t) uap->addr);
369 end = round_page((vm_offset_t) uap->addr + uap->len);
371 return (vm_map_madvise(&ve->vmspace->vm_map, start, end,
372 uap->behav, uap->value));
376 * Red black tree functions
378 static int rb_vmspace_compare(struct vmspace_entry *, struct vmspace_entry *);
379 RB_GENERATE(vmspace_rb_tree, vmspace_entry, rb_entry, rb_vmspace_compare);
381 /* a->start is address, and the only field has to be initialized */
383 rb_vmspace_compare(struct vmspace_entry *a, struct vmspace_entry *b)
385 if ((char *)a->id < (char *)b->id)
387 else if ((char *)a->id > (char *)b->id)
394 rb_vmspace_delete(struct vmspace_entry *ve, void *data)
396 struct vkernel_common *vc = data;
398 KKASSERT(ve->refs == 0);
399 vmspace_entry_delete(ve, vc);
404 * Remove a vmspace_entry from the RB tree and destroy it. We have to clean
405 * up the pmap, the vm_map, then destroy the vmspace.
409 vmspace_entry_delete(struct vmspace_entry *ve, struct vkernel_common *vc)
411 RB_REMOVE(vmspace_rb_tree, &vc->vc_root, ve);
413 pmap_remove_pages(vmspace_pmap(ve->vmspace),
414 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
415 vm_map_remove(&ve->vmspace->vm_map,
416 VM_MIN_USER_ADDRESS, VM_MAX_USER_ADDRESS);
417 vmspace_free(ve->vmspace);
418 kfree(ve, M_VKERNEL);
423 struct vmspace_entry *
424 vkernel_find_vmspace(struct vkernel_common *vc, void *id)
426 struct vmspace_entry *ve;
427 struct vmspace_entry key;
430 ve = RB_FIND(vmspace_rb_tree, &vc->vc_root, &key);
435 * Manage vkernel refs, used by the kernel when fork()ing or exit()ing
439 vkernel_inherit(struct proc *p1, struct proc *p2)
441 struct vkernel_common *vc;
446 KKASSERT(vc->vc_refs > 0);
447 atomic_add_int(&vc->vc_refs, 1);
448 vk = kmalloc(sizeof(*vk), M_VKERNEL, M_WAITOK|M_ZERO);
454 vkernel_exit(struct proc *p)
456 struct vkernel_common *vc;
457 struct vmspace_entry *ve;
464 vk->vk_common = NULL;
467 * Restore the original VM context if we are killed while running
470 * This isn't supposed to happen. What is supposed to happen is
471 * that the process should enter vkernel_trap() before the handling
474 if ((ve = vk->vk_current) != NULL) {
475 kprintf("Killed with active VC, notify kernel list\n");
477 db_print_backtrace();
479 vk->vk_current = NULL;
481 p->p_vmspace = vk->vk_save_vmspace;
483 vk->vk_save_vmspace = NULL;
484 KKASSERT(ve->refs > 0);
489 * Dereference the common area
491 KKASSERT(vc->vc_refs > 0);
492 spin_lock_wr(&vc->vc_spin);
493 if (--vc->vc_refs == 0)
495 spin_unlock_wr(&vc->vc_spin);
498 RB_SCAN(vmspace_rb_tree, &vc->vc_root, NULL,
499 rb_vmspace_delete, vc);
500 kfree(vc, M_VKERNEL);
502 kfree(vk, M_VKERNEL);
506 * A VM space under virtual kernel control trapped out or made a system call
507 * or otherwise needs to return control to the virtual kernel context.
510 vkernel_trap(struct proc *p, struct trapframe *frame)
512 struct vmspace_entry *ve;
517 * Which vmspace entry was running?
521 vk->vk_current = NULL;
522 KKASSERT(ve != NULL);
525 * Switch the process context back to the virtual kernel's VM space.
528 p->p_vmspace = vk->vk_save_vmspace;
530 vk->vk_save_vmspace = NULL;
531 KKASSERT(ve->refs > 0);
535 * Copy the emulated process frame to the virtual kernel process.
536 * The emulated process cannot change TLS descriptors so don't
537 * bother saving them, we already have a copy.
539 * Restore the virtual kernel's saved context so the virtual kernel
540 * process can resume.
542 error = copyout(frame, vk->vk_user_trapframe, sizeof(*frame));
543 bcopy(&vk->vk_save_trapframe, frame, sizeof(*frame));
544 bcopy(&vk->vk_save_vextframe.vx_tls, &curthread->td_tls,
545 sizeof(vk->vk_save_vextframe.vx_tls));