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>
76 #include <sys/vnode.h>
77 #include <sys/vmmeter.h>
78 #include <sys/kernel.h>
79 #include <sys/sysctl.h>
80 #include <sys/unistd.h>
82 #include <machine/clock.h>
83 #include <machine/cpu.h>
84 #include <machine/fpu.h>
85 #include <machine/md_var.h>
86 #include <machine/prom.h>
87 #include <machine/mutex.h>
90 #include <vm/vm_param.h>
92 #include <vm/vm_kern.h>
93 #include <vm/vm_page.h>
94 #include <vm/vm_map.h>
95 #include <vm/vm_extern.h>
100 * quick version of vm_fault
103 vm_fault_quick(v, prot)
108 if (prot & VM_PROT_WRITE)
109 r = subyte(v, fubyte(v));
116 * Finish a fork operation, with process p2 nearly set up.
117 * Copy and update the pcb, set up the stack so that the child
118 * ready to run and return to user mode.
121 cpu_fork(p1, p2, flags)
122 register struct proc *p1, *p2;
125 if ((flags & RFPROC) == 0)
128 p2->p_md.md_tf = p1->p_md.md_tf;
129 p2->p_md.md_flags = p1->p_md.md_flags & (MDP_FPUSED | MDP_UAC_MASK);
132 * Cache the physical address of the pcb, so we can
135 p2->p_md.md_pcbpaddr = (void*)vtophys((vm_offset_t)&p2->p_addr->u_pcb);
138 * Copy floating point state from the FP chip to the PCB
139 * if this process has state stored there.
141 alpha_fpstate_save(p1, 0);
144 * Copy pcb and stack from proc p1 to p2. We do this as
145 * cheaply as possible, copying only the active part of the
146 * stack. The stack and pcb need to agree. Make sure that the
147 * new process has FEN disabled.
149 p2->p_addr->u_pcb = p1->p_addr->u_pcb;
150 p2->p_addr->u_pcb.pcb_hw.apcb_usp = alpha_pal_rdusp();
151 p2->p_addr->u_pcb.pcb_hw.apcb_flags &= ~ALPHA_PCB_FLAGS_FEN;
154 * Set the floating point state.
156 if ((p2->p_addr->u_pcb.pcb_fp_control & IEEE_INHERIT) == 0) {
157 p2->p_addr->u_pcb.pcb_fp_control = 0;
158 p2->p_addr->u_pcb.pcb_fp.fpr_cr = (FPCR_DYN_NORMAL
159 | FPCR_INVD | FPCR_DZED
160 | FPCR_OVFD | FPCR_INED
165 * Arrange for a non-local goto when the new process
166 * is started, to resume here, returning nonzero from setjmp.
170 panic("cpu_fork: curproc");
171 alpha_fpstate_check(p1);
175 * create the child's kernel stack, from scratch.
178 struct user *up = p2->p_addr;
179 struct trapframe *p2tf;
182 * Pick a stack pointer, leaving room for a trapframe;
183 * copy trapframe from parent so return to user mode
184 * will be to right address, with correct registers.
186 p2tf = p2->p_md.md_tf = (struct trapframe *)
187 ((char *)p2->p_addr + USPACE - sizeof(struct trapframe));
188 bcopy(p1->p_md.md_tf, p2->p_md.md_tf,
189 sizeof(struct trapframe));
192 * Set up return-value registers as fork() libc stub expects.
194 p2tf->tf_regs[FRAME_V0] = 0; /* child's pid (linux) */
195 p2tf->tf_regs[FRAME_A3] = 0; /* no error */
196 p2tf->tf_regs[FRAME_A4] = 1; /* is child (FreeBSD) */
199 * Arrange for continuation at child_return(), which
200 * will return to exception_return(). Note that the child
201 * process doesn't stay in the kernel for long!
203 * This is an inlined version of cpu_set_kpc.
205 up->u_pcb.pcb_hw.apcb_ksp = (u_int64_t)p2tf;
206 up->u_pcb.pcb_context[0] =
207 (u_int64_t)child_return; /* s0: pc */
208 up->u_pcb.pcb_context[1] =
209 (u_int64_t)exception_return; /* s1: ra */
210 up->u_pcb.pcb_context[2] = (u_long) p2; /* s2: a0 */
211 up->u_pcb.pcb_context[7] =
212 (u_int64_t)switch_trampoline; /* ra: assembly magic */
217 * Intercept the return address from a freshly forked process that has NOT
218 * been scheduled yet.
220 * This is needed to make kernel threads stay in kernel mode.
223 cpu_set_fork_handler(p, func, arg)
225 void (*func) __P((void *));
229 * Note that the trap frame follows the args, so the function
230 * is really called like this: func(arg, frame);
232 p->p_addr->u_pcb.pcb_context[0] = (u_long) func;
233 p->p_addr->u_pcb.pcb_context[2] = (u_long) arg;
237 * cpu_exit is called as the last action during exit.
238 * We release the address space of the process, block interrupts,
239 * and call switch_exit. switch_exit switches to proc0's PCB and stack,
240 * then jumps into the middle of cpu_switch, as if it were switching
245 register struct proc *p;
247 alpha_fpstate_drop(p);
250 mtx_enter(&sched_lock, MTX_SPIN);
251 mtx_exit(&Giant, MTX_DEF);
261 /* drop per-process resources */
262 pmap_dispose_proc(p);
264 /* and clean-out the vmspace */
265 vmspace_free(p->p_vmspace);
269 * Dump the machine specific header information at the start of a core dump.
272 cpu_coredump(p, vp, cred)
278 return (vn_rdwr(UIO_WRITE, vp, (caddr_t) p->p_addr, ctob(UPAGES),
279 (off_t)0, UIO_SYSSPACE, IO_NODELOCKED|IO_UNIT, cred, (int *)NULL,
285 setredzone(pte, vaddr)
289 /* eventually do this by setting up an expand-down stack segment
290 for ss0: selector, allowing stack access down to top of u.
291 this means though that protection violations need to be handled
292 thru a double fault exception that must do an integral task
293 switch to a known good context, within which a dump can be
294 taken. a sensible scheme might be to save the initial context
295 used by sched (that has physical memory mapped 1:1 at bottom)
296 and take the dump while still in mapped mode */
301 * Map an IO request into kernel virtual address space.
303 * All requests are (re)mapped into kernel VA space.
304 * Notice that we use b_bufsize for the size of the buffer
305 * to be mapped. b_bcount might be modified by the driver.
309 register struct buf *bp;
311 register caddr_t addr, v, kva;
314 if ((bp->b_flags & B_PHYS) == 0)
317 for (v = bp->b_saveaddr, addr = (caddr_t)trunc_page(bp->b_data);
318 addr < bp->b_data + bp->b_bufsize;
319 addr += PAGE_SIZE, v += PAGE_SIZE) {
321 * Do the vm_fault if needed; do the copy-on-write thing
322 * when reading stuff off device into memory.
325 (bp->b_iocmd == BIO_READ)?(VM_PROT_READ|VM_PROT_WRITE):VM_PROT_READ);
326 pa = trunc_page(pmap_kextract((vm_offset_t) addr));
328 panic("vmapbuf: page not present");
329 vm_page_hold(PHYS_TO_VM_PAGE(pa));
330 pmap_kenter((vm_offset_t) v, pa);
333 kva = bp->b_saveaddr;
334 bp->b_saveaddr = bp->b_data;
335 bp->b_data = kva + (((vm_offset_t) bp->b_data) & PAGE_MASK);
339 * Free the io map PTEs associated with this IO operation.
340 * We also invalidate the TLB entries and restore the original b_addr.
344 register struct buf *bp;
346 register caddr_t addr;
349 if ((bp->b_flags & B_PHYS) == 0)
352 for (addr = (caddr_t)trunc_page(bp->b_data);
353 addr < bp->b_data + bp->b_bufsize;
355 pa = trunc_page(pmap_kextract((vm_offset_t) addr));
356 pmap_kremove((vm_offset_t) addr);
357 vm_page_unhold(PHYS_TO_VM_PAGE(pa));
360 bp->b_data = bp->b_saveaddr;
364 * Reset back to firmware.
379 rv = vm_map_growstack (p, sp);
380 if (rv != KERN_SUCCESS)
387 static int cnt_prezero;
389 SYSCTL_INT(_machdep, OID_AUTO, cnt_prezero, CTLFLAG_RD, &cnt_prezero, 0, "");
392 * Implement the pre-zeroed page mechanism.
393 * This routine is called from the idle loop.
396 #define ZIDLE_LO(v) ((v) * 2 / 3)
397 #define ZIDLE_HI(v) ((v) * 4 / 5)
402 static int free_rover;
403 static int zero_state;
408 * Attempt to maintain approximately 1/2 of our free pages in a
409 * PG_ZERO'd state. Add some hysteresis to (attempt to) avoid
410 * generally zeroing a page when the system is near steady-state.
411 * Otherwise we might get 'flutter' during disk I/O / IPC or
412 * fast sleeps. We also do not want to be continuously zeroing
413 * pages because doing so may flush our L1 and L2 caches too much.
416 if (zero_state && vm_page_zero_count >= ZIDLE_LO(cnt.v_free_count))
418 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count))
422 if (KLOCK_ENTER(M_TRY)) {
425 m = vm_page_list_find(PQ_FREE, free_rover, FALSE);
427 if (m != NULL && (m->flags & PG_ZERO) == 0) {
428 vm_page_queues[m->queue].lcnt--;
429 TAILQ_REMOVE(&vm_page_queues[m->queue].pl, m, pageq);
435 pmap_zero_page(VM_PAGE_TO_PHYS(m));
440 vm_page_flag_set(m, PG_ZERO);
441 m->queue = PQ_FREE + m->pc;
442 vm_page_queues[m->queue].lcnt++;
443 TAILQ_INSERT_TAIL(&vm_page_queues[m->queue].pl, m,
445 ++vm_page_zero_count;
447 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count))
450 free_rover = (free_rover + PQ_PRIME2) & PQ_L2_MASK;
463 * Software interrupt handler for queued VM system processing.
468 if (busdma_swi_pending != 0)
473 * Tell whether this address is in some physical memory region.
474 * Currently used by the kernel coredump code in order to avoid
475 * dumping the ``ISA memory hole'' which could cause indefinite hangs,
476 * or other unpredictable behaviour.
481 is_physical_memory(addr)
485 * stuff other tests for known memory-mapped devices (PCI?)