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/platform/vkernel/platform/init.c,v 1.17 2007/01/10 03:04:28 corecode Exp $
37 #include <sys/types.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
43 #include <sys/random.h>
44 #include <sys/vkernel.h>
47 #include <sys/msgbuf.h>
48 #include <sys/vmspace.h>
49 #include <vm/vm_page.h>
51 #include <machine/globaldata.h>
52 #include <machine/tls.h>
53 #include <machine/md_var.h>
54 #include <machine/vmparam.h>
65 vm_paddr_t phys_avail[16];
67 vm_paddr_t Maxmem_bytes;
73 vm_offset_t virtual_start;
74 vm_offset_t virtual_end;
75 vm_offset_t kernel_vm_end;
76 vm_offset_t crashdumpmap;
77 vm_offset_t clean_sva;
78 vm_offset_t clean_eva;
79 struct msgbuf *msgbufp;
82 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
83 u_int cpu_feature; /* XXX */
84 u_int tsc_present; /* XXX */
86 struct privatespace *CPU_prvspace;
88 static struct trapframe proc0_tf;
89 static void *proc0paddr;
91 static void init_sys_memory(char *imageFile);
92 static void init_kern_memory(void);
93 static void init_globaldata(void);
94 static void init_vkernel(void);
95 static void init_rootdevice(char *imageFile);
96 static void usage(const char *ctl);
99 * Kernel startup for virtual kernels - standard main()
102 main(int ac, char **av)
104 char *memImageFile = NULL;
105 char *rootImageFile = NULL;
114 while ((c = getopt(ac, av, "vm:r:e:")) != -1) {
118 * name=value:name=value:name=value...
121 kern_envp = malloc(n + 2);
122 for (i = 0; i < n; ++i) {
123 if (optarg[i] == ':')
126 kern_envp[i] = optarg[i];
135 memImageFile = optarg;
138 rootImageFile = optarg;
141 Maxmem_bytes = strtoull(optarg, &suffix, 0);
158 usage("Bad maxmem option");
167 init_sys_memory(memImageFile);
171 init_rootdevice(rootImageFile);
179 * Initialize system memory. This is the virtual kernel's 'RAM'.
183 init_sys_memory(char *imageFile)
190 * Figure out the system memory image size. If an image file was
191 * specified and -m was not specified, use the image file's size.
194 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
195 Maxmem_bytes = (vm_paddr_t)st.st_size;
196 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
198 err(1, "Cannot create new memory file %s unless "
199 "system memory size is specified with -m",
205 * Maxmem must be known at this time
207 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
208 err(1, "Bad maxmem specification: 32MB minimum, "
209 "multiples of %dMB only",
210 SEG_SIZE / 1024 / 1024);
215 * Generate an image file name if necessary, then open/create the
216 * file exclusively locked. Do not allow multiple virtual kernels
217 * to use the same image file.
219 if (imageFile == NULL) {
220 for (i = 0; i < 1000000; ++i) {
221 asprintf(&imageFile, "/var/vkernel/memimg.%06d", i);
223 O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
224 if (fd < 0 && errno == EWOULDBLOCK) {
231 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
233 printf("Using memory file: %s\n", imageFile);
234 if (fd < 0 || fstat(fd, &st) < 0) {
235 err(1, "Unable to open/create %s: %s",
236 imageFile, strerror(errno));
241 * Truncate or extend the file as necessary.
243 if (st.st_size > Maxmem_bytes) {
244 ftruncate(fd, Maxmem_bytes);
245 } else if (st.st_size < Maxmem_bytes) {
247 off_t off = st.st_size & ~SEG_MASK;
249 kprintf("%s: Reserving blocks for memory image\n", imageFile);
250 zmem = malloc(SEG_SIZE);
251 bzero(zmem, SEG_SIZE);
252 lseek(fd, off, SEEK_SET);
253 while (off < Maxmem_bytes) {
254 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
255 err(1, "Unable to reserve blocks for memory image");
261 err(1, "Unable to reserve blocks for memory image");
265 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
269 * Initialize kernel memory. This reserves kernel virtual memory by using
274 init_kern_memory(void)
282 * Memory map our kernel virtual memory space. Note that the
283 * kernel image itself is not made part of this memory for the
286 * The memory map must be segment-aligned so we can properly
289 base = mmap((void *)0x40000000, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
290 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE, MemImageFd, 0);
291 if (base == MAP_FAILED) {
292 err(1, "Unable to mmap() kernel virtual memory!");
295 KvaStart = (vm_offset_t)base;
296 KvaSize = KERNEL_KVA_SIZE;
297 KvaEnd = KvaStart + KvaSize;
300 * Create a top-level page table self-mapping itself.
302 * Initialize the page directory at physical page index 0 to point
303 * to an array of page table pages starting at physical page index 1
305 lseek(MemImageFd, 0L, 0);
306 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
307 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
308 write(MemImageFd, &pte, sizeof(pte));
312 * Initialize the PTEs in the page table pages required to map the
313 * page table itself. This includes mapping the page directory page
314 * at the base so we go one more loop then normal.
316 lseek(MemImageFd, PAGE_SIZE, 0);
317 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
318 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
319 write(MemImageFd, &pte, sizeof(pte));
323 * Initialize remaining PTEs to 0. We may be reusing a memory image
324 * file. This is approximately a megabyte.
326 i = (KERNEL_KVA_SIZE / PAGE_SIZE - i) * sizeof(pte);
327 zero = malloc(PAGE_SIZE);
329 write(MemImageFd, zero, (i > PAGE_SIZE) ? PAGE_SIZE : i);
330 i = i - ((i > PAGE_SIZE) ? PAGE_SIZE : i);
335 * Enable the page table and calculate pointers to our self-map
336 * for easy kernel page table manipulation.
338 * KernelPTA must be offset so we can do direct VA translations
340 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
341 0 | VPTE_R | VPTE_W | VPTE_V);
342 KernelPTD = (vpte_t *)base; /* pg directory */
343 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
344 KernelPTA -= KvaStart >> PAGE_SHIFT;
347 * phys_avail[] represents unallocated physical memory. MI code
348 * will use phys_avail[] to create the vm_page array.
350 phys_avail[0] = PAGE_SIZE +
351 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
352 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
353 phys_avail[1] = Maxmem_bytes;
356 * (virtual_start, virtual_end) represent unallocated kernel virtual
357 * memory. MI code will create kernel_map using these parameters.
359 virtual_start = KvaStart + PAGE_SIZE +
360 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
361 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
362 virtual_end = KvaStart + KERNEL_KVA_SIZE;
365 * Because we just pre-allocate the entire page table the demark used
366 * to determine when KVM must be grown is just set to the end of
367 * KVM. pmap_growkernel() simply panics.
369 kernel_vm_end = virtual_end;
372 * Allocate space for process 0's UAREA.
374 proc0paddr = (void *)virtual_start;
375 for (i = 0; i < UPAGES; ++i) {
376 pmap_kenter_quick(virtual_start, phys_avail[0]);
377 virtual_start += PAGE_SIZE;
378 phys_avail[0] += PAGE_SIZE;
384 crashdumpmap = virtual_start;
385 virtual_start += MAXDUMPPGS * PAGE_SIZE;
388 * msgbufp maps the system message buffer
390 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
391 msgbufp = (void *)virtual_start;
392 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
393 pmap_kenter_quick(virtual_start, phys_avail[0]);
394 virtual_start += PAGE_SIZE;
395 phys_avail[0] += PAGE_SIZE;
397 msgbufinit(msgbufp, MSGBUF_SIZE);
400 * used by kern_memio for /dev/mem access
402 ptvmmap = (caddr_t)virtual_start;
403 virtual_start += PAGE_SIZE;
406 * Bootstrap the kernel_pmap
412 * Map the per-cpu globaldata for cpu #0. Allocate the space using
413 * virtual_start and phys_avail[0]
417 init_globaldata(void)
424 * Reserve enough KVA to cover possible cpus. This is a considerable
425 * amount of KVA since the privatespace structure includes two
426 * whole page table mappings.
428 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
429 CPU_prvspace = (void *)virtual_start;
430 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
433 * Allocate enough physical memory to cover the mdglobaldata
434 * portion of the space and the idle stack and map the pages
435 * into KVA. For cpu #0 only.
437 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
439 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
440 pmap_kenter_quick(va, pa);
441 phys_avail[0] += PAGE_SIZE;
443 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
445 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
446 pmap_kenter_quick(va, pa);
447 phys_avail[0] += PAGE_SIZE;
451 * Setup the %gs for cpu #0. The mycpu macro works after this
454 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
458 * Initialize very low level systems including thread0, proc0, etc.
464 struct mdglobaldata *gd;
466 gd = &CPU_prvspace[0].mdglobaldata;
467 bzero(gd, sizeof(*gd));
469 gd->mi.gd_curthread = &thread0;
470 thread0.td_gd = &gd->mi;
474 gd->mi.gd_prvspace = &CPU_prvspace[0];
475 mi_gdinit(&gd->mi, 0);
477 mi_proc0init(&gd->mi, proc0paddr);
478 proc0.p_lwp.lwp_md.md_regs = &proc0_tf;
483 #if 0 /* #ifdef DDB */
485 if (boothowto & RB_KDB)
486 Debugger("Boot flags requested debugger");
489 initializecpu(); /* Initialize CPU registers */
491 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
495 * Map the message buffer
497 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
498 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
499 msgbufinit(msgbufp, MSGBUF_SIZE);
502 thread0.td_pcb_cr3 ... MMU
503 proc0.p_lwp.lwp_md.md_regs = &proc0_tf;
508 * The root filesystem path for the virtual kernel is optional. If specified
509 * it points to a filesystem image.
513 init_rootdevice(char *imageFile)
518 RootImageFd = open(imageFile, O_RDWR, 0644);
519 if (RootImageFd < 0 || fstat(RootImageFd, &st) < 0) {
520 err(1, "Unable to open/create %s: %s",
521 imageFile, strerror(errno));
524 rootdevnames[0] = "ufs:vkd0a";
530 usage(const char *ctl)
538 kprintf("cpu reset\n");
545 kprintf("cpu halt\n");
547 __asm__ __volatile("hlt");