vm_machdep.c revision 127086
136865Sdfr/*-
236865Sdfr * Copyright (c) 1982, 1986 The Regents of the University of California.
336865Sdfr * Copyright (c) 1989, 1990 William Jolitz
436865Sdfr * Copyright (c) 1994 John Dyson
536865Sdfr * All rights reserved.
636865Sdfr *
736865Sdfr * This code is derived from software contributed to Berkeley by
836865Sdfr * the Systems Programming Group of the University of Utah Computer
936865Sdfr * Science Department, and William Jolitz.
1036865Sdfr *
1136865Sdfr * Redistribution and use in source and binary forms, with or without
1236865Sdfr * modification, are permitted provided that the following conditions
1336865Sdfr * are met:
1436865Sdfr * 1. Redistributions of source code must retain the above copyright
1536865Sdfr *    notice, this list of conditions and the following disclaimer.
1636865Sdfr * 2. Redistributions in binary form must reproduce the above copyright
1736865Sdfr *    notice, this list of conditions and the following disclaimer in the
1836865Sdfr *    documentation and/or other materials provided with the distribution.
1936865Sdfr * 3. All advertising materials mentioning features or use of this software
2036865Sdfr *    must display the following acknowledgement:
2136865Sdfr *	This product includes software developed by the University of
2236865Sdfr *	California, Berkeley and its contributors.
2336865Sdfr * 4. Neither the name of the University nor the names of its contributors
2436865Sdfr *    may be used to endorse or promote products derived from this software
2536865Sdfr *    without specific prior written permission.
2636865Sdfr *
2736865Sdfr * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
2836865Sdfr * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
2936865Sdfr * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3036865Sdfr * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3136865Sdfr * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3236865Sdfr * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3336865Sdfr * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3436865Sdfr * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3536865Sdfr * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3636865Sdfr * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3736865Sdfr * SUCH DAMAGE.
3836865Sdfr *
3936865Sdfr *	from: @(#)vm_machdep.c	7.3 (Berkeley) 5/13/91
4036865Sdfr *	Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$
4150477Speter * $FreeBSD: head/sys/powerpc/aim/vm_machdep.c 127086 2004-03-16 19:04:28Z alc $
4236865Sdfr */
4336865Sdfr/*
4436865Sdfr * Copyright (c) 1994, 1995, 1996 Carnegie-Mellon University.
4536865Sdfr * All rights reserved.
4636865Sdfr *
4736865Sdfr * Author: Chris G. Demetriou
48126474Sgrehan *
4936865Sdfr * Permission to use, copy, modify and distribute this software and
5036865Sdfr * its documentation is hereby granted, provided that both the copyright
5136865Sdfr * notice and this permission notice appear in all copies of the
5236865Sdfr * software, derivative works or modified versions, and any portions
5336865Sdfr * thereof, and that both notices appear in supporting documentation.
54126474Sgrehan *
55126474Sgrehan * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
56126474Sgrehan * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
5736865Sdfr * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
58126474Sgrehan *
5936865Sdfr * Carnegie Mellon requests users of this software to return to
6036865Sdfr *
6136865Sdfr *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
6236865Sdfr *  School of Computer Science
6336865Sdfr *  Carnegie Mellon University
6436865Sdfr *  Pittsburgh PA 15213-3890
6536865Sdfr *
6636865Sdfr * any improvements or extensions that they make and grant Carnegie the
6736865Sdfr * rights to redistribute these changes.
6836865Sdfr */
6936865Sdfr
70118239Speter#include "opt_kstack_pages.h"
71118239Speter
7236865Sdfr#include <sys/param.h>
7336865Sdfr#include <sys/systm.h>
7436865Sdfr#include <sys/proc.h>
7536865Sdfr#include <sys/malloc.h>
7660041Sphk#include <sys/bio.h>
7736865Sdfr#include <sys/buf.h>
7885201Smp#include <sys/ktr.h>
7985201Smp#include <sys/lock.h>
8067365Sjhb#include <sys/mutex.h>
8136865Sdfr#include <sys/vnode.h>
8236865Sdfr#include <sys/vmmeter.h>
8336865Sdfr#include <sys/kernel.h>
84119563Salc#include <sys/mbuf.h>
85122780Salc#include <sys/sf_buf.h>
8636865Sdfr#include <sys/sysctl.h>
8754207Speter#include <sys/unistd.h>
8836865Sdfr
8936865Sdfr#include <machine/clock.h>
9036865Sdfr#include <machine/cpu.h>
9141499Sdfr#include <machine/fpu.h>
9297397Sbenno#include <machine/frame.h>
9336865Sdfr#include <machine/md_var.h>
9436865Sdfr
9578342Sbenno#include <dev/ofw/openfirm.h>
9678342Sbenno
9736865Sdfr#include <vm/vm.h>
9836865Sdfr#include <vm/vm_param.h>
9936865Sdfr#include <vm/vm_kern.h>
10036865Sdfr#include <vm/vm_page.h>
10136865Sdfr#include <vm/vm_map.h>
10236865Sdfr#include <vm/vm_extern.h>
10336865Sdfr
10436865Sdfr#include <sys/user.h>
10536865Sdfr
106119563Salcstatic void	sf_buf_init(void *arg);
107119563SalcSYSINIT(sock_sf, SI_SUB_MBUF, SI_ORDER_ANY, sf_buf_init, NULL)
108119563Salc
10936865Sdfr/*
110119563Salc * Expanded sf_freelist head. Really an SLIST_HEAD() in disguise, with the
111119563Salc * sf_freelist head with the sf_lock mutex.
112119563Salc */
113119563Salcstatic struct {
114119563Salc	SLIST_HEAD(, sf_buf) sf_head;
115119563Salc	struct mtx sf_lock;
116119563Salc} sf_freelist;
117119563Salc
118119563Salcstatic u_int	sf_buf_alloc_want;
119119563Salc
120119563Salc/*
12136865Sdfr * Finish a fork operation, with process p2 nearly set up.
12236865Sdfr * Copy and update the pcb, set up the stack so that the child
12336865Sdfr * ready to run and return to user mode.
12436865Sdfr */
12536865Sdfrvoid
12690361Sjuliancpu_fork(struct thread *td1, struct proc *p2, struct thread *td2, int flags)
12736865Sdfr{
12884977Sbenno	struct	proc *p1;
12984977Sbenno	struct	trapframe *tf;
13084977Sbenno	struct	callframe *cf;
13190643Sbenno	struct	pcb *pcb;
13284977Sbenno
13391467Sbenno	KASSERT(td1 == curthread || td1 == &thread0,
13490643Sbenno	    ("cpu_fork: p1 not curproc and not proc0"));
13584977Sbenno	CTR3(KTR_PROC, "cpu_fork: called td1=%08x p2=%08x flags=%x", (u_int)td1, (u_int)p2, flags);
13684977Sbenno
13784977Sbenno	if ((flags & RFPROC) == 0)
13884977Sbenno		return;
13984977Sbenno
14084977Sbenno	p1 = td1->td_proc;
14184977Sbenno
14291467Sbenno	pcb = (struct pcb *)((td2->td_kstack + KSTACK_PAGES * PAGE_SIZE -
14391467Sbenno	    sizeof(struct pcb)) & ~0x2fU);
14490643Sbenno	td2->td_pcb = pcb;
14584977Sbenno
14690643Sbenno	/* Copy the pcb */
14790643Sbenno	bcopy(td1->td_pcb, pcb, sizeof(struct pcb));
14884977Sbenno
14984977Sbenno	/*
15090643Sbenno	 * Create a fresh stack for the new process.
15184977Sbenno	 * Copy the trap frame for the return to user mode as if from a
15284977Sbenno	 * syscall.  This copies most of the user mode register values.
15384977Sbenno	 */
15490643Sbenno	tf = (struct trapframe *)pcb - 1;
15590643Sbenno	bcopy(td1->td_frame, tf, sizeof(*tf));
15684977Sbenno
15797397Sbenno	/* Set up trap frame. */
15897397Sbenno	tf->fixreg[FIRSTARG] = 0;
15997397Sbenno	tf->fixreg[FIRSTARG + 1] = 0;
16097397Sbenno	tf->cr &= ~0x10000000;
16190643Sbenno
16290643Sbenno	td2->td_frame = tf;
16390643Sbenno
16490643Sbenno	cf = (struct callframe *)tf - 1;
16591467Sbenno	cf->cf_func = (register_t)fork_return;
16691467Sbenno	cf->cf_arg0 = (register_t)td2;
16791467Sbenno	cf->cf_arg1 = (register_t)tf;
16884977Sbenno
16991467Sbenno	pcb->pcb_sp = (register_t)cf;
17091467Sbenno	pcb->pcb_lr = (register_t)fork_trampoline;
171105611Sgrehan	pcb->pcb_usr = kernel_pmap->pm_sr[USER_SR];
17284977Sbenno
17390643Sbenno	/*
17490643Sbenno 	 * Now cpu_switch() can schedule the new process.
17590643Sbenno	 */
17636865Sdfr}
17736865Sdfr
17836865Sdfr/*
17936865Sdfr * Intercept the return address from a freshly forked process that has NOT
18036865Sdfr * been scheduled yet.
18136865Sdfr *
18236865Sdfr * This is needed to make kernel threads stay in kernel mode.
18336865Sdfr */
18436865Sdfrvoid
18583682Smpcpu_set_fork_handler(td, func, arg)
18683682Smp	struct thread *td;
18792842Salfred	void (*func)(void *);
18848391Speter	void *arg;
18936865Sdfr{
19084977Sbenno	struct	callframe *cf;
19184977Sbenno
19284977Sbenno	CTR3(KTR_PROC, "cpu_set_fork_handler: called with td=%08x func=%08x arg=%08x",
19384977Sbenno	    (u_int)td, (u_int)func, (u_int)arg);
19484977Sbenno
19591467Sbenno	cf = (struct callframe *)td->td_pcb->pcb_sp;
19684977Sbenno
19791467Sbenno	cf->cf_func = (register_t)func;
19891467Sbenno	cf->cf_arg0 = (register_t)arg;
19936865Sdfr}
20036865Sdfr
20136865Sdfr/*
20236865Sdfr * cpu_exit is called as the last action during exit.
20336865Sdfr * We release the address space of the process, block interrupts,
20436865Sdfr * and call switch_exit.  switch_exit switches to proc0's PCB and stack,
20536865Sdfr * then jumps into the middle of cpu_switch, as if it were switching
20636865Sdfr * from proc0.
20736865Sdfr */
20836865Sdfrvoid
20983682Smpcpu_exit(td)
21083682Smp	register struct thread *td;
21136865Sdfr{
21236865Sdfr}
21336865Sdfr
21436865Sdfrvoid
21598765Sjakecpu_sched_exit(td)
21698765Sjake	register struct thread *td;
21798765Sjake{
21898765Sjake}
21998765Sjake
22083276Speter/* Temporary helper */
22183276Spetervoid
222118893Sgrehancpu_throw(struct thread *old, struct thread *new)
22383276Speter{
22483276Speter
225118893Sgrehan	cpu_switch(old, new);
22683276Speter	panic("cpu_throw() didn't");
22736865Sdfr}
22836865Sdfr
22936865Sdfr/*
23065557Sjasone * Reset back to firmware.
23136865Sdfr */
23236865Sdfrvoid
23336865Sdfrcpu_reset()
23436865Sdfr{
23578342Sbenno	OF_exit();
23636865Sdfr}
23736865Sdfr
23836865Sdfr/*
239119563Salc * Allocate a pool of sf_bufs (sendfile(2) or "super-fast" if you prefer. :-))
240119563Salc */
241119563Salcstatic void
242119563Salcsf_buf_init(void *arg)
243119563Salc{
244119563Salc	struct sf_buf *sf_bufs;
245119563Salc	vm_offset_t sf_base;
246119563Salc	int i;
247119563Salc
248119563Salc	mtx_init(&sf_freelist.sf_lock, "sf_bufs list lock", NULL, MTX_DEF);
249119563Salc	SLIST_INIT(&sf_freelist.sf_head);
250119563Salc	sf_base = kmem_alloc_nofault(kernel_map, nsfbufs * PAGE_SIZE);
251119563Salc	sf_bufs = malloc(nsfbufs * sizeof(struct sf_buf), M_TEMP,
252119563Salc	    M_NOWAIT | M_ZERO);
253119563Salc	for (i = 0; i < nsfbufs; i++) {
254119563Salc		sf_bufs[i].kva = sf_base + i * PAGE_SIZE;
255119563Salc		SLIST_INSERT_HEAD(&sf_freelist.sf_head, &sf_bufs[i], free_list);
256119563Salc	}
257119563Salc	sf_buf_alloc_want = 0;
258119563Salc}
259119563Salc
260119563Salc/*
261119563Salc * Get an sf_buf from the freelist. Will block if none are available.
262119563Salc */
263119563Salcstruct sf_buf *
264119563Salcsf_buf_alloc(struct vm_page *m)
265119563Salc{
266119563Salc	struct sf_buf *sf;
267119563Salc	int error;
268119563Salc
269119563Salc	mtx_lock(&sf_freelist.sf_lock);
270119563Salc	while ((sf = SLIST_FIRST(&sf_freelist.sf_head)) == NULL) {
271119563Salc		sf_buf_alloc_want++;
272123929Ssilby		mbstat.sf_allocwait++;
273119563Salc		error = msleep(&sf_freelist, &sf_freelist.sf_lock, PVM|PCATCH,
274119563Salc		    "sfbufa", 0);
275119563Salc		sf_buf_alloc_want--;
276119563Salc
277119563Salc		/*
278126474Sgrehan		 * If we got a signal, don't risk going back to sleep.
279119563Salc		 */
280119563Salc		if (error)
281119563Salc			break;
282119563Salc	}
283119563Salc	if (sf != NULL) {
284119563Salc		SLIST_REMOVE_HEAD(&sf_freelist.sf_head, free_list);
285119563Salc		sf->m = m;
286123884Ssilby		nsfbufsused++;
287123920Ssilby		nsfbufspeak = imax(nsfbufspeak, nsfbufsused);
288119563Salc		pmap_qenter(sf->kva, &sf->m, 1);
289119563Salc	}
290119563Salc	mtx_unlock(&sf_freelist.sf_lock);
291119563Salc	return (sf);
292119563Salc}
293119563Salc
294119563Salc/*
295127086Salc * Release resources back to the system.
296119563Salc */
297119563Salcvoid
298127086Salcsf_buf_free(struct sf_buf *sf)
299119563Salc{
300119563Salc
301127086Salc	pmap_qremove(sf->kva, 1);
302119563Salc	mtx_lock(&sf_freelist.sf_lock);
303119563Salc	SLIST_INSERT_HEAD(&sf_freelist.sf_head, sf, free_list);
304123884Ssilby	nsfbufsused--;
305119563Salc	if (sf_buf_alloc_want > 0)
306119563Salc		wakeup_one(&sf_freelist);
307119563Salc	mtx_unlock(&sf_freelist.sf_lock);
308119563Salc}
309119563Salc
310119563Salc/*
31136865Sdfr * Software interrupt handler for queued VM system processing.
312126474Sgrehan */
313126474Sgrehanvoid
314126474Sgrehanswi_vm(void *dummy)
315126474Sgrehan{
31678342Sbenno#if 0 /* XXX: Don't have busdma stuff yet */
31736865Sdfr	if (busdma_swi_pending != 0)
31836865Sdfr		busdma_swi();
31978342Sbenno#endif
32036865Sdfr}
32136865Sdfr
32236865Sdfr/*
32336865Sdfr * Tell whether this address is in some physical memory region.
32436865Sdfr * Currently used by the kernel coredump code in order to avoid
32536865Sdfr * dumping the ``ISA memory hole'' which could cause indefinite hangs,
32636865Sdfr * or other unpredictable behaviour.
32736865Sdfr */
32836865Sdfr
32936865Sdfr
33036865Sdfrint
33136865Sdfris_physical_memory(addr)
33236865Sdfr	vm_offset_t addr;
33336865Sdfr{
33436865Sdfr	/*
33536865Sdfr	 * stuff other tests for known memory-mapped devices (PCI?)
33636865Sdfr	 * here
33736865Sdfr	 */
33836865Sdfr
33936865Sdfr	return 1;
34036865Sdfr}
34199659Sbenno
34299659Sbenno/*
34399659Sbenno * KSE functions
34499659Sbenno */
34599659Sbennovoid
346126474Sgrehancpu_thread_exit(struct thread *td)
34799659Sbenno{
34899659Sbenno}
34999659Sbenno
35099659Sbennovoid
351126474Sgrehancpu_thread_clean(struct thread *td)
352107180Smux{
353107180Smux}
354107180Smux
355107180Smuxvoid
35699659Sbennocpu_thread_setup(struct thread *td)
35799659Sbenno{
358126474Sgrehan	struct pcb *pcb;
359126474Sgrehan
360126474Sgrehan	pcb = (struct pcb *)((td->td_kstack + KSTACK_PAGES * PAGE_SIZE -
361126474Sgrehan	    sizeof(struct pcb)) & ~0x2fU);
362126474Sgrehan	td->td_pcb = pcb;
363126474Sgrehan	td->td_frame = (struct trapframe *)pcb - 1;
364119004Smarcel}
36599659Sbenno
366119004Smarcelvoid
367119004Smarcelcpu_thread_swapin(struct thread *td)
368119004Smarcel{
36999659Sbenno}
37099659Sbenno
37199659Sbennovoid
372119004Smarcelcpu_thread_swapout(struct thread *td)
373119004Smarcel{
374119004Smarcel}
375119004Smarcel
376119004Smarcelvoid
377115858Smarcelcpu_set_upcall(struct thread *td, struct thread *td0)
37899659Sbenno{
379126474Sgrehan	struct pcb *pcb2;
380126474Sgrehan	struct trapframe *tf;
381126474Sgrehan	struct callframe *cf;
382126474Sgrehan
383126474Sgrehan	pcb2 = td->td_pcb;
384126474Sgrehan
385126474Sgrehan	/* Copy the upcall pcb */
386126474Sgrehan	bcopy(td0->td_pcb, pcb2, sizeof(*pcb2));
387126474Sgrehan
388126474Sgrehan	/* Create a stack for the new thread */
389126474Sgrehan	tf = td->td_frame;
390126474Sgrehan	bcopy(td0->td_frame, tf, sizeof(struct trapframe));
391126474Sgrehan	tf->fixreg[FIRSTARG] = 0;
392126474Sgrehan	tf->fixreg[FIRSTARG + 1] = 0;
393126474Sgrehan	tf->cr &= ~0x10000000;
394126474Sgrehan
395126474Sgrehan	/* Set registers for trampoline to user mode. */
396126474Sgrehan	cf = (struct callframe *)tf - 1;
397126474Sgrehan	cf->cf_func = (register_t)fork_return;
398126474Sgrehan	cf->cf_arg0 = (register_t)td;
399126474Sgrehan	cf->cf_arg1 = (register_t)tf;
400126474Sgrehan
401126474Sgrehan	pcb2->pcb_sp = (register_t)cf;
402126474Sgrehan	pcb2->pcb_lr = (register_t)fork_trampoline;
403126474Sgrehan	pcb2->pcb_usr = kernel_pmap->pm_sr[USER_SR];
40499659Sbenno}
40599659Sbenno
40699659Sbennovoid
407111028Sjeffcpu_set_upcall_kse(struct thread *td, struct kse_upcall *ku)
40899659Sbenno{
409126474Sgrehan        struct trapframe *tf;
410126474Sgrehan        uint32_t sp;
411126474Sgrehan
412126474Sgrehan	tf = td->td_frame;
413126474Sgrehan	/* align stack and alloc space for frame ptr and saved LR */
414126474Sgrehan        sp = ((uint32_t)ku->ku_stack.ss_sp + ku->ku_stack.ss_size
415126474Sgrehan		- 2*sizeof(u_int32_t)) & ~0x1f;
416126474Sgrehan	bzero(tf, sizeof(struct trapframe));
417126474Sgrehan
418126474Sgrehan	tf->fixreg[1] = (register_t)sp;
419126474Sgrehan        tf->fixreg[3] = (register_t)ku->ku_mailbox;
420126474Sgrehan        tf->srr0 = (register_t)ku->ku_func;
421126474Sgrehan        tf->srr1 = PSL_MBO | PSL_USERSET | PSL_FE_DFLT;
422126474Sgrehan        td->td_pcb->pcb_flags = 0;
423126474Sgrehan
424126474Sgrehan        td->td_retval[0] = (register_t)ku->ku_func;
425126474Sgrehan        td->td_retval[1] = 0;
42699659Sbenno}
427