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 1.8 2007/01/06 19:40:55 dillon 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 <vm/vm_page.h>
50 #include <machine/globaldata.h>
51 #include <machine/tls.h>
52 #include <machine/md_var.h>
53 #include <machine/vmparam.h>
64 vm_paddr_t phys_avail[16];
66 vm_paddr_t Maxmem_bytes;
72 vm_offset_t virtual_start;
73 vm_offset_t virtual_end;
74 vm_offset_t kernel_vm_end;
75 vm_offset_t crashdumpmap;
76 vm_offset_t clean_sva;
77 vm_offset_t clean_eva;
78 struct msgbuf *msgbufp;
81 vpte_t *KernelPTA; /* Warning: Offset for direct VA translation */
82 u_int cpu_feature; /* XXX */
83 u_int tsc_present; /* XXX */
85 struct privatespace *CPU_prvspace;
87 static void *proc0paddr;
89 static void init_sys_memory(char *imageFile);
90 static void init_kern_memory(void);
91 static void init_globaldata(void);
92 static void init_vkernel(void);
93 static void init_rootdevice(char *imageFile);
94 static void usage(const char *ctl);
97 * Kernel startup for virtual kernels - standard main()
100 main(int ac, char **av)
102 char *memImageFile = NULL;
103 char *rootImageFile = NULL;
110 while ((c = getopt(ac, av, "vm:")) != -1) {
116 memImageFile = optarg;
119 rootImageFile = optarg;
122 Maxmem_bytes = strtoull(optarg, &suffix, 0);
139 usage("Bad maxmem option");
148 init_sys_memory(memImageFile);
152 init_rootdevice(rootImageFile);
159 * Initialize system memory. This is the virtual kernel's 'RAM'.
163 init_sys_memory(char *imageFile)
169 * Figure out the system memory image size. If an image file was
170 * specified and -m was not specified, use the image file's size.
173 if (imageFile && stat(imageFile, &st) == 0 && Maxmem_bytes == 0)
174 Maxmem_bytes = (vm_paddr_t)st.st_size;
175 if ((imageFile == NULL || stat(imageFile, &st) < 0) &&
177 err(1, "Cannot create new memory file %s unless "
178 "system memory size is specified with -m",
184 * Maxmem must be known at this time
186 if (Maxmem_bytes < 32 * 1024 * 1024 || (Maxmem_bytes & SEG_MASK)) {
187 err(1, "Bad maxmem specification: 32MB minimum, "
188 "multiples of %dMB only",
189 SEG_SIZE / 1024 / 1024);
194 * Generate an image file name if necessary, then open/create the
195 * file exclusively locked. Do not allow multiple virtual kernels
196 * to use the same image file.
198 if (imageFile == NULL)
199 asprintf(&imageFile, "/var/vkernel/image.%05d", (int)getpid());
200 fd = open(imageFile, O_RDWR|O_CREAT|O_EXLOCK|O_NONBLOCK, 0644);
201 if (fd < 0 || fstat(fd, &st) < 0) {
202 err(1, "Unable to open/create %s: %s",
203 imageFile, strerror(errno));
208 * Truncate or extend the file as necessary.
210 if (st.st_size > Maxmem_bytes) {
211 ftruncate(fd, Maxmem_bytes);
212 } else if (st.st_size < Maxmem_bytes) {
214 off_t off = st.st_size & ~SEG_MASK;
216 kprintf("%s: Reserving blocks for memory image\n", imageFile);
217 zmem = malloc(SEG_SIZE);
218 bzero(zmem, SEG_SIZE);
220 while (off < Maxmem_bytes) {
221 if (write(fd, zmem, SEG_SIZE) != SEG_SIZE) {
222 err(1, "Unable to reserve blocks for memory image");
228 err(1, "Unable to reserve blocks for memory image");
232 Maxmem = Maxmem_bytes >> PAGE_SHIFT;
236 * Initialize kernel memory. This reserves kernel virtual memory by using
241 init_kern_memory(void)
248 * Memory map our kernel virtual memory space. Note that the
249 * kernel image itself is not made part of this memory for the
252 * The memory map must be segment-aligned so we can properly
255 base = mmap((void *)0x40000000, KERNEL_KVA_SIZE, PROT_READ|PROT_WRITE,
256 MAP_FILE|MAP_SHARED|MAP_VPAGETABLE, MemImageFd, 0);
257 if (base == MAP_FAILED) {
258 err(1, "Unable to mmap() kernel virtual memory!");
261 KvaStart = (vm_offset_t)base;
262 KvaSize = KERNEL_KVA_SIZE;
263 KvaEnd = KvaStart + KvaSize;
266 * Create a top-level page table self-mapping itself.
268 * Initialize the page directory at physical page index 0 to point
269 * to an array of page table pages starting at physical page index 1
271 lseek(MemImageFd, 0L, 0);
272 for (i = 0; i < KERNEL_KVA_SIZE / SEG_SIZE; ++i) {
273 pte = ((i + 1) * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
274 write(MemImageFd, &pte, sizeof(pte));
278 * Initialize the PTEs in the page table pages required to map the
279 * page table itself. This includes mapping the page directory page
280 * at the base so we go one more loop then normal.
282 lseek(MemImageFd, PAGE_SIZE, 0);
283 for (i = 0; i <= KERNEL_KVA_SIZE / SEG_SIZE * sizeof(vpte_t); ++i) {
284 pte = (i * PAGE_SIZE) | VPTE_V | VPTE_R | VPTE_W;
285 write(MemImageFd, &pte, sizeof(pte));
289 * Enable the page table and calculate pointers to our self-map
290 * for easy kernel page table manipulation.
292 * KernelPTA must be offset so we can do direct VA translations
294 mcontrol(base, KERNEL_KVA_SIZE, MADV_SETMAP,
295 0 | VPTE_R | VPTE_W | VPTE_V);
296 KernelPTD = (vpte_t *)base; /* pg directory */
297 KernelPTA = (vpte_t *)((char *)base + PAGE_SIZE); /* pg table pages */
298 KernelPTA -= KvaStart >> PAGE_SHIFT;
301 * phys_avail[] represents unallocated physical memory. MI code
302 * will use phys_avail[] to create the vm_page array.
304 phys_avail[0] = PAGE_SIZE +
305 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
306 phys_avail[0] = (phys_avail[0] + PAGE_MASK) & ~(vm_paddr_t)PAGE_MASK;
307 phys_avail[1] = Maxmem_bytes;
310 * (virtual_start, virtual_end) represent unallocated kernel virtual
311 * memory. MI code will create kernel_map using these parameters.
313 virtual_start = KvaStart + PAGE_SIZE +
314 KERNEL_KVA_SIZE / PAGE_SIZE * sizeof(vpte_t);
315 virtual_start = (virtual_start + PAGE_MASK) & ~(vm_offset_t)PAGE_MASK;
316 virtual_end = KvaStart + KERNEL_KVA_SIZE;
319 * Because we just pre-allocate the entire page table the demark used
320 * to determine when KVM must be grown is just set to the end of
321 * KVM. pmap_growkernel() simply panics.
323 kernel_vm_end = virtual_end;
326 * Allocate space for process 0's UAREA.
328 proc0paddr = (void *)virtual_start;
329 for (i = 0; i < UPAGES; ++i) {
330 pmap_kenter_quick(virtual_start, phys_avail[0]);
331 virtual_start += PAGE_SIZE;
332 phys_avail[0] += PAGE_SIZE;
338 crashdumpmap = virtual_start;
339 virtual_start += MAXDUMPPGS * PAGE_SIZE;
342 * msgbufp maps the system message buffer
344 assert((MSGBUF_SIZE & PAGE_MASK) == 0);
345 msgbufp = (void *)virtual_start;
346 for (i = 0; i < (MSGBUF_SIZE >> PAGE_SHIFT); ++i) {
347 pmap_kenter_quick(virtual_start, phys_avail[0]);
348 virtual_start += PAGE_SIZE;
349 phys_avail[0] += PAGE_SIZE;
351 msgbufinit(msgbufp, MSGBUF_SIZE);
354 * used by kern_memio for /dev/mem access
356 ptvmmap = (caddr_t)virtual_start;
357 virtual_start += PAGE_SIZE;
360 * Bootstrap the kernel_pmap
366 * Map the per-cpu globaldata for cpu #0. Allocate the space using
367 * virtual_start and phys_avail[0]
371 init_globaldata(void)
378 * Reserve enough KVA to cover possible cpus. This is a considerable
379 * amount of KVA since the privatespace structure includes two
380 * whole page table mappings.
382 virtual_start = (virtual_start + SEG_MASK) & ~(vm_offset_t)SEG_MASK;
383 CPU_prvspace = (void *)virtual_start;
384 virtual_start += sizeof(struct privatespace) * SMP_MAXCPU;
387 * Allocate enough physical memory to cover the mdglobaldata
388 * portion of the space and the idle stack and map the pages
389 * into KVA. For cpu #0 only.
391 for (i = 0; i < sizeof(struct mdglobaldata); i += PAGE_SIZE) {
393 va = (vm_offset_t)&CPU_prvspace[0].mdglobaldata + i;
394 pmap_kenter_quick(va, pa);
395 phys_avail[0] += PAGE_SIZE;
397 for (i = 0; i < sizeof(CPU_prvspace[0].idlestack); i += PAGE_SIZE) {
399 va = (vm_offset_t)&CPU_prvspace[0].idlestack + i;
400 pmap_kenter_quick(va, pa);
401 phys_avail[0] += PAGE_SIZE;
405 * Setup the %gs for cpu #0. The mycpu macro works after this
408 tls_set_fs(&CPU_prvspace[0], sizeof(struct privatespace));
412 * Initialize very low level systems including thread0, proc0, etc.
418 struct mdglobaldata *gd;
420 gd = &CPU_prvspace[0].mdglobaldata;
421 bzero(gd, sizeof(*gd));
423 gd->mi.gd_curthread = &thread0;
424 thread0.td_gd = &gd->mi;
428 gd->mi.gd_prvspace = &CPU_prvspace[0];
429 mi_gdinit(&gd->mi, 0);
431 mi_proc0init(&gd->mi, proc0paddr);
435 #if 0 /* #ifdef DDB */
437 if (boothowto & RB_KDB)
438 Debugger("Boot flags requested debugger");
441 initializecpu(); /* Initialize CPU registers */
443 init_param2((phys_avail[1] - phys_avail[0]) / PAGE_SIZE);
447 * Map the message buffer
449 for (off = 0; off < round_page(MSGBUF_SIZE); off += PAGE_SIZE)
450 pmap_kenter((vm_offset_t)msgbufp + off, avail_end + off);
451 msgbufinit(msgbufp, MSGBUF_SIZE);
454 thread0.td_pcb_cr3 ... MMU
455 proc0.p_lwp.lwp_md.md_regs = &proc0_tf;
460 * The root filesystem path for the virtual kernel is optional. If specified
461 * it points to a filesystem image.
465 init_rootdevice(char *imageFile)
471 usage(const char *ctl)
479 kprintf("cpu reset\n");
486 kprintf("cpu halt\n");
488 __asm__ __volatile("hlt");