1/*- 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 5 * All rights reserved. 6 * 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. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 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. 26 * 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 37 * SUCH DAMAGE. 38 * 39 * from: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91 40 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$ 41 * $FreeBSD$ 42 */ 43/*- 44 * Copyright (c) 1994, 1995, 1996 Carnegie-Mellon University. 45 * All rights reserved. 46 * 47 * Author: Chris G. Demetriou 48 * 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. 54 * 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. 58 * 59 * Carnegie Mellon requests users of this software to return to 60 * 61 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 62 * School of Computer Science 63 * Carnegie Mellon University 64 * Pittsburgh PA 15213-3890 65 * 66 * any improvements or extensions that they make and grant Carnegie the 67 * rights to redistribute these changes. 68 */ 69 70#include <sys/param.h> 71#include <sys/systm.h> 72#include <sys/proc.h> 73#include <sys/malloc.h> 74#include <sys/bio.h> 75#include <sys/buf.h> 76#include <sys/ktr.h> 77#include <sys/lock.h> 78#include <sys/mutex.h> 79#include <sys/vnode.h> 80#include <sys/vmmeter.h> 81#include <sys/kernel.h> 82#include <sys/mbuf.h> 83#include <sys/sf_buf.h> 84#include <sys/sysctl.h> 85#include <sys/sysent.h> 86#include <sys/unistd.h> 87 88#include <machine/cpu.h> 89#include <machine/fpu.h> 90#include <machine/frame.h> 91#include <machine/md_var.h> 92#include <machine/pcb.h> 93 94#include <dev/ofw/openfirm.h> 95 96#include <vm/vm.h> 97#include <vm/vm_param.h> 98#include <vm/vm_kern.h> 99#include <vm/vm_page.h> 100#include <vm/vm_map.h> 101#include <vm/vm_extern.h> 102 103/* 104 * On systems without a direct mapped region (e.g. PPC64), 105 * we use the same code as the Book E implementation. Since 106 * we need to have runtime detection of this, define some machinery 107 * for sf_bufs in this case, and ignore it on systems with direct maps. 108 */ 109 110#ifndef NSFBUFS 111#define NSFBUFS (512 + maxusers * 16) 112#endif 113 114static int nsfbufs; 115static int nsfbufspeak; 116static int nsfbufsused; 117 118SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufs, CTLFLAG_RDTUN, &nsfbufs, 0, 119 "Maximum number of sendfile(2) sf_bufs available"); 120SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufspeak, CTLFLAG_RD, &nsfbufspeak, 0, 121 "Number of sendfile(2) sf_bufs at peak usage"); 122SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufsused, CTLFLAG_RD, &nsfbufsused, 0, 123 "Number of sendfile(2) sf_bufs in use"); 124 125static void sf_buf_init(void *arg); 126SYSINIT(sock_sf, SI_SUB_MBUF, SI_ORDER_ANY, sf_buf_init, NULL); 127 128LIST_HEAD(sf_head, sf_buf); 129 130/* A hash table of active sendfile(2) buffers */ 131static struct sf_head *sf_buf_active; 132static u_long sf_buf_hashmask; 133 134#define SF_BUF_HASH(m) (((m) - vm_page_array) & sf_buf_hashmask) 135 136static TAILQ_HEAD(, sf_buf) sf_buf_freelist; 137static u_int sf_buf_alloc_want; 138 139/* 140 * A lock used to synchronize access to the hash table and free list 141 */ 142static struct mtx sf_buf_lock; 143 144#ifdef __powerpc64__ 145extern uintptr_t tocbase; 146#endif 147 148 149/* 150 * Finish a fork operation, with process p2 nearly set up. 151 * Copy and update the pcb, set up the stack so that the child 152 * ready to run and return to user mode. 153 */ 154void 155cpu_fork(struct thread *td1, struct proc *p2, struct thread *td2, int flags) 156{ 157 struct trapframe *tf; 158 struct callframe *cf; 159 struct pcb *pcb; 160 161 KASSERT(td1 == curthread || td1 == &thread0, 162 ("cpu_fork: p1 not curproc and not proc0")); 163 CTR3(KTR_PROC, "cpu_fork: called td1=%p p2=%p flags=%x", 164 td1, p2, flags); 165 166 if ((flags & RFPROC) == 0) 167 return; 168 169 pcb = (struct pcb *)((td2->td_kstack + 170 td2->td_kstack_pages * PAGE_SIZE - sizeof(struct pcb)) & ~0x2fUL); 171 td2->td_pcb = pcb; 172 173 /* Copy the pcb */ 174 bcopy(td1->td_pcb, pcb, sizeof(struct pcb)); 175 176 /* 177 * Create a fresh stack for the new process. 178 * Copy the trap frame for the return to user mode as if from a 179 * syscall. This copies most of the user mode register values. 180 */ 181 tf = (struct trapframe *)pcb - 1; 182 bcopy(td1->td_frame, tf, sizeof(*tf)); 183 184 /* Set up trap frame. */ 185 tf->fixreg[FIRSTARG] = 0; 186 tf->fixreg[FIRSTARG + 1] = 0; 187 tf->cr &= ~0x10000000; 188 189 td2->td_frame = tf; 190 191 cf = (struct callframe *)tf - 1; 192 memset(cf, 0, sizeof(struct callframe)); 193 #ifdef __powerpc64__ 194 cf->cf_toc = tocbase; 195 #endif 196 cf->cf_func = (register_t)fork_return; 197 cf->cf_arg0 = (register_t)td2; 198 cf->cf_arg1 = (register_t)tf; 199 200 pcb->pcb_sp = (register_t)cf; 201 KASSERT(pcb->pcb_sp % 16 == 0, ("stack misaligned")); 202 #ifdef __powerpc64__ 203 pcb->pcb_lr = ((register_t *)fork_trampoline)[0]; 204 pcb->pcb_toc = ((register_t *)fork_trampoline)[1]; 205 #else 206 pcb->pcb_lr = (register_t)fork_trampoline; 207 #endif 208 #ifdef AIM 209 pcb->pcb_cpu.aim.usr_vsid = 0; 210 #endif 211 212 /* Setup to release spin count in fork_exit(). */ 213 td2->td_md.md_spinlock_count = 1; 214 td2->td_md.md_saved_msr = PSL_KERNSET; 215 216 /* 217 * Now cpu_switch() can schedule the new process. 218 */ 219} 220 221/* 222 * Intercept the return address from a freshly forked process that has NOT 223 * been scheduled yet. 224 * 225 * This is needed to make kernel threads stay in kernel mode. 226 */ 227void 228cpu_set_fork_handler(struct thread *td, void (*func)(void *), void *arg) 229{ 230 struct callframe *cf; 231 232 CTR4(KTR_PROC, "%s called with td=%p func=%p arg=%p", 233 __func__, td, func, arg); 234 235 cf = (struct callframe *)td->td_pcb->pcb_sp; 236 237 cf->cf_func = (register_t)func; 238 cf->cf_arg0 = (register_t)arg; 239} 240 241void 242cpu_exit(struct thread *td) 243{ 244 245} 246 247/* 248 * Allocate a pool of sf_bufs (sendfile(2) or "super-fast" if you prefer. :-)) 249 */ 250static void 251sf_buf_init(void *arg) 252{ 253 struct sf_buf *sf_bufs; 254 vm_offset_t sf_base; 255 int i; 256 257 /* Don't bother on systems with a direct map */ 258 if (hw_direct_map) 259 return; 260 261 nsfbufs = NSFBUFS; 262 TUNABLE_INT_FETCH("kern.ipc.nsfbufs", &nsfbufs); 263 264 sf_buf_active = hashinit(nsfbufs, M_TEMP, &sf_buf_hashmask); 265 TAILQ_INIT(&sf_buf_freelist); 266 sf_base = kva_alloc(nsfbufs * PAGE_SIZE); 267 sf_bufs = malloc(nsfbufs * sizeof(struct sf_buf), M_TEMP, 268 M_NOWAIT | M_ZERO); 269 270 for (i = 0; i < nsfbufs; i++) { 271 sf_bufs[i].kva = sf_base + i * PAGE_SIZE; 272 TAILQ_INSERT_TAIL(&sf_buf_freelist, &sf_bufs[i], free_entry); 273 } 274 sf_buf_alloc_want = 0; 275 mtx_init(&sf_buf_lock, "sf_buf", NULL, MTX_DEF); 276} 277 278/* 279 * Get an sf_buf from the freelist. Will block if none are available. 280 */ 281struct sf_buf * 282sf_buf_alloc(struct vm_page *m, int flags) 283{ 284 struct sf_head *hash_list; 285 struct sf_buf *sf; 286 int error; 287 288 if (hw_direct_map) { 289 /* Shortcut the direct mapped case */ 290 return ((struct sf_buf *)m); 291 } 292 293 hash_list = &sf_buf_active[SF_BUF_HASH(m)]; 294 mtx_lock(&sf_buf_lock); 295 LIST_FOREACH(sf, hash_list, list_entry) { 296 if (sf->m == m) { 297 sf->ref_count++; 298 if (sf->ref_count == 1) { 299 TAILQ_REMOVE(&sf_buf_freelist, sf, free_entry); 300 nsfbufsused++; 301 nsfbufspeak = imax(nsfbufspeak, nsfbufsused); 302 } 303 goto done; 304 } 305 } 306 307 while ((sf = TAILQ_FIRST(&sf_buf_freelist)) == NULL) { 308 if (flags & SFB_NOWAIT) 309 goto done; 310 311 sf_buf_alloc_want++; 312 SFSTAT_INC(sf_allocwait); 313 error = msleep(&sf_buf_freelist, &sf_buf_lock, 314 (flags & SFB_CATCH) ? PCATCH | PVM : PVM, "sfbufa", 0); 315 sf_buf_alloc_want--; 316 317 /* 318 * If we got a signal, don't risk going back to sleep. 319 */ 320 if (error) 321 goto done; 322 } 323 324 TAILQ_REMOVE(&sf_buf_freelist, sf, free_entry); 325 if (sf->m != NULL) 326 LIST_REMOVE(sf, list_entry); 327 328 LIST_INSERT_HEAD(hash_list, sf, list_entry); 329 sf->ref_count = 1; 330 sf->m = m; 331 nsfbufsused++; 332 nsfbufspeak = imax(nsfbufspeak, nsfbufsused); 333 pmap_qenter(sf->kva, &sf->m, 1); 334done: 335 mtx_unlock(&sf_buf_lock); 336 return (sf); 337} 338 339/* 340 * Detach mapped page and release resources back to the system. 341 * 342 * Remove a reference from the given sf_buf, adding it to the free 343 * list when its reference count reaches zero. A freed sf_buf still, 344 * however, retains its virtual-to-physical mapping until it is 345 * recycled or reactivated by sf_buf_alloc(9). 346 */ 347void 348sf_buf_free(struct sf_buf *sf) 349{ 350 if (hw_direct_map) 351 return; 352 353 mtx_lock(&sf_buf_lock); 354 sf->ref_count--; 355 if (sf->ref_count == 0) { 356 TAILQ_INSERT_TAIL(&sf_buf_freelist, sf, free_entry); 357 nsfbufsused--; 358 359 if (sf_buf_alloc_want > 0) 360 wakeup(&sf_buf_freelist); 361 } 362 mtx_unlock(&sf_buf_lock); 363} 364 365/* 366 * Software interrupt handler for queued VM system processing. 367 */ 368void 369swi_vm(void *dummy) 370{ 371 372 if (busdma_swi_pending != 0) 373 busdma_swi(); 374} 375 376/* 377 * Tell whether this address is in some physical memory region. 378 * Currently used by the kernel coredump code in order to avoid 379 * dumping the ``ISA memory hole'' which could cause indefinite hangs, 380 * or other unpredictable behaviour. 381 */ 382int 383is_physical_memory(vm_offset_t addr) 384{ 385 386 /* 387 * stuff other tests for known memory-mapped devices (PCI?) 388 * here 389 */ 390 return (1); 391} 392 393/* 394 * CPU threading functions related to the VM layer. These could be used 395 * to map the SLB bits required for the kernel stack instead of forcing a 396 * fixed-size KVA. 397 */ 398 399void 400cpu_thread_swapin(struct thread *td) 401{ 402 403} 404 405void 406cpu_thread_swapout(struct thread *td) 407{ 408 409} 410 411