2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2014 Ian Lepore <ian@freebsd.org>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
36 * Routines for describing and initializing anything related to physical memory.
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/physmem.h>
43 #include <vm/vm_param.h>
44 #include <vm/vm_page.h>
45 #include <vm/vm_phys.h>
46 #include <machine/md_var.h>
49 * These structures are used internally to keep track of regions of physical
50 * ram, and regions within the physical ram that need to be excluded. An
51 * exclusion region can be excluded from crash dumps, from the vm pool of pages
52 * that can be allocated, or both, depending on the exclusion flags associated
56 #define MAX_HWCNT 32 /* ACPI needs more regions */
64 #define MAX_PHYS_ADDR 0xFFFFFFFFull
65 #elif defined(__aarch64__) || defined(__riscv)
66 #define MAX_PHYS_ADDR 0xFFFFFFFFFFFFFFFFull
75 static struct region hwregions[MAX_HWCNT];
76 static struct region exregions[MAX_EXCNT];
82 * realmem is the total number of hardware pages, excluded or not.
83 * Maxmem is one greater than the last physical page number.
89 * Print the contents of the physical and excluded region tables using the
90 * provided printf-like output function (which will be either printf or
94 physmem_dump_tables(int (*prfunc)(const char *, ...))
98 const unsigned int mbyte = 1024 * 1024;
100 prfunc("Physical memory chunk(s):\n");
101 for (i = 0; i < hwcnt; ++i) {
102 addr = hwregions[i].addr;
103 size = hwregions[i].size;
104 prfunc(" 0x%08jx - 0x%08jx, %5ju MB (%7ju pages)\n", addr,
105 addr + size - 1, size / mbyte, size / PAGE_SIZE);
108 prfunc("Excluded memory regions:\n");
109 for (i = 0; i < excnt; ++i) {
110 addr = exregions[i].addr;
111 size = exregions[i].size;
112 flags = exregions[i].flags;
113 prfunc(" 0x%08jx - 0x%08jx, %5ju MB (%7ju pages) %s %s\n",
114 addr, addr + size - 1, size / mbyte, size / PAGE_SIZE,
115 (flags & EXFLAG_NOALLOC) ? "NoAlloc" : "",
116 (flags & EXFLAG_NODUMP) ? "NoDump" : "");
120 prfunc("Avail lists:\n");
121 for (i = 0; phys_avail[i] != 0; ++i) {
122 prfunc(" phys_avail[%d] 0x%08x\n", i, phys_avail[i]);
124 for (i = 0; dump_avail[i] != 0; ++i) {
125 prfunc(" dump_avail[%d] 0x%08x\n", i, dump_avail[i]);
131 * Print the contents of the static mapping table. Used for bootverbose.
134 physmem_print_tables(void)
137 physmem_dump_tables(printf);
141 * Walk the list of hardware regions, processing it against the list of
142 * exclusions that contain the given exflags, and generating an "avail list".
144 * Updates the value at *pavail with the sum of all pages in all hw regions.
146 * Returns the number of pages of non-excluded memory added to the avail list.
149 regions_to_avail(vm_paddr_t *avail, uint32_t exflags, size_t maxavail,
150 long *pavail, long *prealmem)
152 size_t acnt, exi, hwi;
153 uint64_t end, start, xend, xstart;
154 long availmem, totalmem;
155 const struct region *exp, *hwp;
160 for (hwi = 0, hwp = hwregions; hwi < hwcnt; ++hwi, ++hwp) {
162 end = hwp->size + start;
163 totalmem += atop((vm_offset_t)(end - start));
164 for (exi = 0, exp = exregions; exi < excnt; ++exi, ++exp) {
166 * If the excluded region does not match given flags,
167 * continue checking with the next excluded region.
169 if ((exp->flags & exflags) == 0)
172 xend = exp->size + xstart;
174 * If the excluded region ends before this hw region,
175 * continue checking with the next excluded region.
180 * If the excluded region begins after this hw region
181 * we're done because both lists are sorted.
186 * If the excluded region completely covers this hw
187 * region, shrink this hw region to zero size.
189 if ((start >= xstart) && (end <= xend)) {
195 * If the excluded region falls wholly within this hw
196 * region without abutting or overlapping the beginning
197 * or end, create an available entry from the leading
198 * fragment, then adjust the start of this hw region to
199 * the end of the excluded region, and continue checking
200 * the next excluded region because another exclusion
201 * could affect the remainder of this hw region.
203 if ((xstart > start) && (xend < end)) {
205 avail[acnt - 1] == (vm_paddr_t)start) {
206 avail[acnt - 1] = (vm_paddr_t)xstart;
208 avail[acnt++] = (vm_paddr_t)start;
209 avail[acnt++] = (vm_paddr_t)xstart;
211 availmem += atop((vm_offset_t)(xstart - start));
216 * We know the excluded region overlaps either the start
217 * or end of this hardware region (but not both), trim
218 * the excluded portion off the appropriate end.
226 * If the trimming actions above left a non-zero size, create an
227 * available entry for it.
230 if (acnt > 0 && avail[acnt - 1] == (vm_paddr_t)start) {
231 avail[acnt - 1] = (vm_paddr_t)end;
233 avail[acnt++] = (vm_paddr_t)start;
234 avail[acnt++] = (vm_paddr_t)end;
236 availmem += atop((vm_offset_t)(end - start));
238 if (acnt >= maxavail)
239 panic("Not enough space in the dump/phys_avail arrays");
244 if (prealmem != NULL)
245 *prealmem = totalmem;
250 * Insertion-sort a new entry into a regions list; sorted by start address.
253 insert_region(struct region *regions, size_t rcnt, vm_paddr_t addr,
254 vm_size_t size, uint32_t flags)
257 struct region *ep, *rp;
260 for (i = 0, rp = regions; i < rcnt; ++i, ++rp) {
261 if (rp->addr == addr && rp->size == size) /* Pure dup. */
263 if (flags == rp->flags) {
264 if (addr + size == rp->addr) {
268 } else if (rp->addr + rp->size == addr) {
273 if (addr < rp->addr) {
274 bcopy(rp, rp + 1, (ep - rp) * sizeof(*rp));
287 * Add a hardware memory region.
290 physmem_hardware_region(uint64_t pa, uint64_t sz)
295 * Filter out the page at PA 0x00000000. The VM can't handle it, as
296 * pmap_extract() == 0 means failure.
303 } else if (pa > MAX_PHYS_ADDR) {
304 /* This range is past usable memory, ignore it */
309 * Also filter out the page at the end of the physical address space --
310 * if addr is non-zero and addr+size is zero we wrapped to the next byte
311 * beyond what vm_paddr_t can express. That leads to a NULL pointer
312 * deref early in startup; work around it by leaving the last page out.
314 * XXX This just in: subtract out a whole megabyte, not just 1 page.
315 * Reducing the size by anything less than 1MB results in the NULL
316 * pointer deref in _vm_map_lock_read(). Better to give up a megabyte
317 * than leave some folks with an unusable system while we investigate.
319 if ((pa + sz) > (MAX_PHYS_ADDR - 1024 * 1024)) {
320 sz = MAX_PHYS_ADDR - pa + 1;
321 if (sz <= 1024 * 1024)
327 * Round the starting address up to a page boundary, and truncate the
328 * ending page down to a page boundary.
330 adj = round_page(pa) - pa;
332 sz = trunc_page(sz - adj);
334 if (sz > 0 && hwcnt < nitems(hwregions))
335 hwcnt = insert_region(hwregions, hwcnt, pa, sz, 0);
339 * Add an exclusion region.
342 physmem_exclude_region(vm_paddr_t pa, vm_size_t sz, uint32_t exflags)
347 * Truncate the starting address down to a page boundary, and round the
348 * ending page up to a page boundary.
350 adj = pa - trunc_page(pa);
352 sz = round_page(sz + adj);
354 if (excnt >= nitems(exregions))
355 panic("failed to exclude region %#jx-%#jx", (uintmax_t)pa,
356 (uintmax_t)(pa + sz));
357 excnt = insert_region(exregions, excnt, pa, sz, exflags);
361 physmem_avail(vm_paddr_t *avail, size_t maxavail)
364 return (regions_to_avail(avail, EXFLAG_NOALLOC, maxavail, NULL, NULL));
368 * Process all the regions added earlier into the global avail lists.
370 * Updates the kernel global 'physmem' with the number of physical pages
371 * available for use (all pages not in any exclusion region).
373 * Updates the kernel global 'Maxmem' with the page number one greater then the
374 * last page of physical memory in the system.
377 physmem_init_kernel_globals(void)
381 regions_to_avail(dump_avail, EXFLAG_NODUMP, PHYS_AVAIL_ENTRIES, NULL,
383 nextidx = regions_to_avail(phys_avail, EXFLAG_NOALLOC,
384 PHYS_AVAIL_ENTRIES, &physmem, &realmem);
386 panic("No memory entries in phys_avail");
387 Maxmem = atop(phys_avail[nextidx - 1]);
393 DB_SHOW_COMMAND(physmem, db_show_physmem)
396 physmem_dump_tables(db_printf);