2 * Copyright (c) 1982, 1986 The Regents of the University of California.
3 * Copyright (c) 1989, 1990 William Jolitz
4 * Copyright (c) 1994 John Dyson
7 * This code is derived from software contributed to Berkeley by
8 * the Systems Programming Group of the University of Utah Computer
9 * Science Department, and William Jolitz.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * from: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91
40 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$
44 * Copyright (c) 1994, 1995, 1996 Carnegie-Mellon University.
45 * All rights reserved.
47 * Author: Chris G. Demetriou
49 * Permission to use, copy, modify and distribute this software and
50 * its documentation is hereby granted, provided that both the copyright
51 * notice and this permission notice appear in all copies of the
52 * software, derivative works or modified versions, and any portions
53 * thereof, and that both notices appear in supporting documentation.
55 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
56 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
57 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
59 * Carnegie Mellon requests users of this software to return to
61 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
62 * School of Computer Science
63 * Carnegie Mellon University
64 * Pittsburgh PA 15213-3890
66 * any improvements or extensions that they make and grant Carnegie the
67 * rights to redistribute these changes.
70 #include <sys/param.h>
71 #include <sys/systm.h>
73 #include <sys/malloc.h>
78 #include <sys/mutex.h>
79 #include <sys/vnode.h>
80 #include <sys/vmmeter.h>
81 #include <sys/kernel.h>
83 #include <sys/sf_buf.h>
84 #include <sys/syscall.h>
85 #include <sys/sysctl.h>
86 #include <sys/sysent.h>
87 #include <sys/unistd.h>
89 #include <machine/cpu.h>
90 #include <machine/fpu.h>
91 #include <machine/frame.h>
92 #include <machine/md_var.h>
93 #include <machine/pcb.h>
95 #include <dev/ofw/openfirm.h>
98 #include <vm/vm_param.h>
99 #include <vm/vm_kern.h>
100 #include <vm/vm_page.h>
101 #include <vm/vm_map.h>
102 #include <vm/vm_extern.h>
105 * On systems without a direct mapped region (e.g. PPC64),
106 * we use the same code as the Book E implementation. Since
107 * we need to have runtime detection of this, define some machinery
108 * for sf_bufs in this case, and ignore it on systems with direct maps.
112 #define NSFBUFS (512 + maxusers * 16)
115 static void sf_buf_init(void *arg);
116 SYSINIT(sock_sf, SI_SUB_MBUF, SI_ORDER_ANY, sf_buf_init, NULL);
118 LIST_HEAD(sf_head, sf_buf);
120 /* A hash table of active sendfile(2) buffers */
121 static struct sf_head *sf_buf_active;
122 static u_long sf_buf_hashmask;
124 #define SF_BUF_HASH(m) (((m) - vm_page_array) & sf_buf_hashmask)
126 static TAILQ_HEAD(, sf_buf) sf_buf_freelist;
127 static u_int sf_buf_alloc_want;
130 * A lock used to synchronize access to the hash table and free list
132 static struct mtx sf_buf_lock;
136 * Finish a fork operation, with process p2 nearly set up.
137 * Copy and update the pcb, set up the stack so that the child
138 * ready to run and return to user mode.
141 cpu_fork(struct thread *td1, struct proc *p2, struct thread *td2, int flags)
144 struct trapframe *tf;
145 struct callframe *cf;
148 KASSERT(td1 == curthread || td1 == &thread0,
149 ("cpu_fork: p1 not curproc and not proc0"));
150 CTR3(KTR_PROC, "cpu_fork: called td1=%08x p2=%08x flags=%x", (u_int)td1, (u_int)p2, flags);
152 if ((flags & RFPROC) == 0)
157 pcb = (struct pcb *)((td2->td_kstack +
158 td2->td_kstack_pages * PAGE_SIZE - sizeof(struct pcb)) & ~0x2fU);
162 bcopy(td1->td_pcb, pcb, sizeof(struct pcb));
165 * Create a fresh stack for the new process.
166 * Copy the trap frame for the return to user mode as if from a
167 * syscall. This copies most of the user mode register values.
169 tf = (struct trapframe *)pcb - 1;
170 bcopy(td1->td_frame, tf, sizeof(*tf));
172 /* Set up trap frame. */
173 tf->fixreg[FIRSTARG] = 0;
174 tf->fixreg[FIRSTARG + 1] = 0;
175 tf->cr &= ~0x10000000;
179 cf = (struct callframe *)tf - 1;
180 memset(cf, 0, sizeof(struct callframe));
181 cf->cf_func = (register_t)fork_return;
182 cf->cf_arg0 = (register_t)td2;
183 cf->cf_arg1 = (register_t)tf;
185 pcb->pcb_sp = (register_t)cf;
186 pcb->pcb_lr = (register_t)fork_trampoline;
187 pcb->pcb_cpu.aim.usr = kernel_pmap->pm_sr[USER_SR];
189 /* Setup to release spin count in fork_exit(). */
190 td2->td_md.md_spinlock_count = 1;
191 td2->td_md.md_saved_msr = PSL_KERNSET;
194 * Now cpu_switch() can schedule the new process.
199 * Intercept the return address from a freshly forked process that has NOT
200 * been scheduled yet.
202 * This is needed to make kernel threads stay in kernel mode.
205 cpu_set_fork_handler(td, func, arg)
207 void (*func)(void *);
210 struct callframe *cf;
212 CTR4(KTR_PROC, "%s called with td=%08x func=%08x arg=%08x",
213 __func__, (u_int)td, (u_int)func, (u_int)arg);
215 cf = (struct callframe *)td->td_pcb->pcb_sp;
217 cf->cf_func = (register_t)func;
218 cf->cf_arg0 = (register_t)arg;
223 register struct thread *td;
228 * Reset back to firmware.
237 * Allocate a pool of sf_bufs (sendfile(2) or "super-fast" if you prefer. :-))
240 sf_buf_init(void *arg)
242 struct sf_buf *sf_bufs;
246 /* Don't bother on systems with a direct map */
252 TUNABLE_INT_FETCH("kern.ipc.nsfbufs", &nsfbufs);
254 sf_buf_active = hashinit(nsfbufs, M_TEMP, &sf_buf_hashmask);
255 TAILQ_INIT(&sf_buf_freelist);
256 sf_base = kmem_alloc_nofault(kernel_map, nsfbufs * PAGE_SIZE);
257 sf_bufs = malloc(nsfbufs * sizeof(struct sf_buf), M_TEMP, M_NOWAIT | M_ZERO);
259 for (i = 0; i < nsfbufs; i++) {
260 sf_bufs[i].kva = sf_base + i * PAGE_SIZE;
261 TAILQ_INSERT_TAIL(&sf_buf_freelist, &sf_bufs[i], free_entry);
263 sf_buf_alloc_want = 0;
264 mtx_init(&sf_buf_lock, "sf_buf", NULL, MTX_DEF);
268 * Get an sf_buf from the freelist. Will block if none are available.
271 sf_buf_alloc(struct vm_page *m, int flags)
273 struct sf_head *hash_list;
278 /* Shortcut the direct mapped case */
280 return ((struct sf_buf *)m);
283 hash_list = &sf_buf_active[SF_BUF_HASH(m)];
284 mtx_lock(&sf_buf_lock);
285 LIST_FOREACH(sf, hash_list, list_entry) {
288 if (sf->ref_count == 1) {
289 TAILQ_REMOVE(&sf_buf_freelist, sf, free_entry);
291 nsfbufspeak = imax(nsfbufspeak, nsfbufsused);
297 while ((sf = TAILQ_FIRST(&sf_buf_freelist)) == NULL) {
298 if (flags & SFB_NOWAIT)
302 mbstat.sf_allocwait++;
303 error = msleep(&sf_buf_freelist, &sf_buf_lock,
304 (flags & SFB_CATCH) ? PCATCH | PVM : PVM, "sfbufa", 0);
308 * If we got a signal, don't risk going back to sleep.
314 TAILQ_REMOVE(&sf_buf_freelist, sf, free_entry);
316 LIST_REMOVE(sf, list_entry);
318 LIST_INSERT_HEAD(hash_list, sf, list_entry);
322 nsfbufspeak = imax(nsfbufspeak, nsfbufsused);
323 pmap_qenter(sf->kva, &sf->m, 1);
325 mtx_unlock(&sf_buf_lock);
330 * Detatch mapped page and release resources back to the system.
332 * Remove a reference from the given sf_buf, adding it to the free
333 * list when its reference count reaches zero. A freed sf_buf still,
334 * however, retains its virtual-to-physical mapping until it is
335 * recycled or reactivated by sf_buf_alloc(9).
338 sf_buf_free(struct sf_buf *sf)
343 mtx_lock(&sf_buf_lock);
345 if (sf->ref_count == 0) {
346 TAILQ_INSERT_TAIL(&sf_buf_freelist, sf, free_entry);
349 if (sf_buf_alloc_want > 0)
350 wakeup_one(&sf_buf_freelist);
352 mtx_unlock(&sf_buf_lock);
356 * Software interrupt handler for queued VM system processing.
361 #if 0 /* XXX: Don't have busdma stuff yet */
362 if (busdma_swi_pending != 0)
368 * Tell whether this address is in some physical memory region.
369 * Currently used by the kernel coredump code in order to avoid
370 * dumping the ``ISA memory hole'' which could cause indefinite hangs,
371 * or other unpredictable behaviour.
376 is_physical_memory(addr)
380 * stuff other tests for known memory-mapped devices (PCI?)
388 * Threading functions
391 cpu_thread_exit(struct thread *td)
396 cpu_thread_clean(struct thread *td)
401 cpu_thread_alloc(struct thread *td)
405 pcb = (struct pcb *)((td->td_kstack + td->td_kstack_pages * PAGE_SIZE -
406 sizeof(struct pcb)) & ~0x2fU);
408 td->td_frame = (struct trapframe *)pcb - 1;
412 cpu_thread_free(struct thread *td)
417 cpu_thread_swapin(struct thread *td)
422 cpu_thread_swapout(struct thread *td)
427 cpu_set_syscall_retval(struct thread *td, int error)
430 struct trapframe *tf;
433 if (error == EJUSTRETURN)
439 if (tf->fixreg[0] == SYS___syscall) {
440 int code = tf->fixreg[FIRSTARG + 1];
441 if (p->p_sysent->sv_mask)
442 code &= p->p_sysent->sv_mask;
443 fixup = (code != SYS_freebsd6_lseek && code != SYS_lseek) ?
452 * 64-bit return, 32-bit syscall. Fixup byte order
454 tf->fixreg[FIRSTARG] = 0;
455 tf->fixreg[FIRSTARG + 1] = td->td_retval[0];
457 tf->fixreg[FIRSTARG] = td->td_retval[0];
458 tf->fixreg[FIRSTARG + 1] = td->td_retval[1];
460 tf->cr &= ~0x10000000; /* XXX: Magic number */
464 * Set user's pc back to redo the system call.
469 if (p->p_sysent->sv_errsize) {
470 error = (error < p->p_sysent->sv_errsize) ?
471 p->p_sysent->sv_errtbl[error] : -1;
473 tf->fixreg[FIRSTARG] = error;
474 tf->cr |= 0x10000000; /* XXX: Magic number */
480 cpu_set_upcall(struct thread *td, struct thread *td0)
483 struct trapframe *tf;
484 struct callframe *cf;
488 /* Copy the upcall pcb */
489 bcopy(td0->td_pcb, pcb2, sizeof(*pcb2));
491 /* Create a stack for the new thread */
493 bcopy(td0->td_frame, tf, sizeof(struct trapframe));
494 tf->fixreg[FIRSTARG] = 0;
495 tf->fixreg[FIRSTARG + 1] = 0;
496 tf->cr &= ~0x10000000;
498 /* Set registers for trampoline to user mode. */
499 cf = (struct callframe *)tf - 1;
500 memset(cf, 0, sizeof(struct callframe));
501 cf->cf_func = (register_t)fork_return;
502 cf->cf_arg0 = (register_t)td;
503 cf->cf_arg1 = (register_t)tf;
505 pcb2->pcb_sp = (register_t)cf;
506 pcb2->pcb_lr = (register_t)fork_trampoline;
507 pcb2->pcb_cpu.aim.usr = kernel_pmap->pm_sr[USER_SR];
509 /* Setup to release spin count in fork_exit(). */
510 td->td_md.md_spinlock_count = 1;
511 td->td_md.md_saved_msr = PSL_KERNSET;
515 cpu_set_upcall_kse(struct thread *td, void (*entry)(void *), void *arg,
518 struct trapframe *tf;
522 /* align stack and alloc space for frame ptr and saved LR */
523 sp = ((uint32_t)stack->ss_sp + stack->ss_size - sizeof(uint64_t)) &
525 bzero(tf, sizeof(struct trapframe));
527 tf->fixreg[1] = (register_t)sp;
528 tf->fixreg[3] = (register_t)arg;
529 tf->srr0 = (register_t)entry;
530 tf->srr1 = PSL_MBO | PSL_USERSET | PSL_FE_DFLT;
531 td->td_pcb->pcb_flags = 0;
533 td->td_retval[0] = (register_t)entry;
534 td->td_retval[1] = 0;
538 cpu_set_user_tls(struct thread *td, void *tls_base)
541 td->td_frame->fixreg[2] = (register_t)tls_base + 0x7008;