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("$FreeBSD$");
38 * Routines for describing and initializing anything related to physical memory.
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/physmem.h>
48 #include <vm/vm_param.h>
49 #include <vm/vm_page.h>
50 #include <vm/vm_phys.h>
51 #include <vm/vm_dumpset.h>
52 #include <machine/md_var.h>
60 * These structures are used internally to keep track of regions of physical
61 * ram, and regions within the physical ram that need to be excluded. An
62 * exclusion region can be excluded from crash dumps, from the vm pool of pages
63 * that can be allocated, or both, depending on the exclusion flags associated
67 #define MAX_HWCNT 32 /* ACPI needs more regions */
75 #define MAX_PHYS_ADDR 0xFFFFFFFFull
76 #elif defined(__aarch64__) || defined(__amd64__) || defined(__riscv)
77 #define MAX_PHYS_ADDR 0xFFFFFFFFFFFFFFFFull
86 static struct region hwregions[MAX_HWCNT];
87 static struct region exregions[MAX_EXCNT];
93 * realmem is the total number of hardware pages, excluded or not.
94 * Maxmem is one greater than the last physical page number.
101 panic(const char *fmt, ...)
106 vfprintf(stderr, fmt, va);
107 fprintf(stderr, "\n");
114 * Print the contents of the physical and excluded region tables using the
115 * provided printf-like output function (which will be either printf or
119 physmem_dump_tables(int (*prfunc)(const char *, ...))
123 uintmax_t addr, size;
124 const unsigned int mbyte = 1024 * 1024;
126 prfunc("Physical memory chunk(s):\n");
127 for (i = 0; i < hwcnt; ++i) {
128 addr = hwregions[i].addr;
129 size = hwregions[i].size;
130 prfunc(" 0x%08jx - 0x%08jx, %5ju MB (%7ju pages)\n", addr,
131 addr + size - 1, size / mbyte, size / PAGE_SIZE);
134 prfunc("Excluded memory regions:\n");
135 for (i = 0; i < excnt; ++i) {
136 addr = exregions[i].addr;
137 size = exregions[i].size;
138 flags = exregions[i].flags;
139 prfunc(" 0x%08jx - 0x%08jx, %5ju MB (%7ju pages) %s %s\n",
140 addr, addr + size - 1, size / mbyte, size / PAGE_SIZE,
141 (flags & EXFLAG_NOALLOC) ? "NoAlloc" : "",
142 (flags & EXFLAG_NODUMP) ? "NoDump" : "");
146 prfunc("Avail lists:\n");
147 for (i = 0; phys_avail[i] != 0; ++i) {
148 prfunc(" phys_avail[%d] 0x%08x\n", i, phys_avail[i]);
150 for (i = 0; dump_avail[i] != 0; ++i) {
151 prfunc(" dump_avail[%d] 0x%08x\n", i, dump_avail[i]);
157 * Print the contents of the static mapping table. Used for bootverbose.
160 physmem_print_tables(void)
163 physmem_dump_tables(printf);
167 * Walk the list of hardware regions, processing it against the list of
168 * exclusions that contain the given exflags, and generating an "avail list".
170 * If maxphyssz is not zero it sets upper limit, in bytes, for the total
171 * "avail list" size. Walk stops once the limit is reached and the last region
172 * is cut short if necessary.
174 * Updates the value at *pavail with the sum of all pages in all hw regions.
176 * Returns the number of pages of non-excluded memory added to the avail list.
179 regions_to_avail(vm_paddr_t *avail, uint32_t exflags, size_t maxavail,
180 uint64_t maxphyssz, long *pavail, long *prealmem)
182 size_t acnt, exi, hwi;
183 uint64_t adj, end, start, xend, xstart;
184 long availmem, totalmem;
185 const struct region *exp, *hwp;
192 for (hwi = 0, hwp = hwregions; hwi < hwcnt; ++hwi, ++hwp) {
193 adj = round_page(hwp->addr) - hwp->addr;
194 start = round_page(hwp->addr);
195 end = trunc_page(hwp->size + adj) + start;
196 totalmem += atop((vm_offset_t)(end - start));
197 for (exi = 0, exp = exregions; exi < excnt; ++exi, ++exp) {
199 * If the excluded region does not match given flags,
200 * continue checking with the next excluded region.
202 if ((exp->flags & exflags) == 0)
205 xend = exp->size + xstart;
207 * If the excluded region ends before this hw region,
208 * continue checking with the next excluded region.
213 * If the excluded region begins after this hw region
214 * we're done because both lists are sorted.
219 * If the excluded region completely covers this hw
220 * region, shrink this hw region to zero size.
222 if ((start >= xstart) && (end <= xend)) {
228 * If the excluded region falls wholly within this hw
229 * region without abutting or overlapping the beginning
230 * or end, create an available entry from the leading
231 * fragment, then adjust the start of this hw region to
232 * the end of the excluded region, and continue checking
233 * the next excluded region because another exclusion
234 * could affect the remainder of this hw region.
236 if ((xstart > start) && (xend < end)) {
238 if ((maxphyssz != 0) &&
239 (availsz + xstart - start > maxphyssz)) {
240 xstart = maxphyssz + start - availsz;
245 avail[acnt - 1] == (vm_paddr_t)start) {
246 avail[acnt - 1] = (vm_paddr_t)xstart;
248 avail[acnt++] = (vm_paddr_t)start;
249 avail[acnt++] = (vm_paddr_t)xstart;
251 availsz += (xstart - start);
252 availmem += atop((vm_offset_t)(xstart - start));
257 * We know the excluded region overlaps either the start
258 * or end of this hardware region (but not both), trim
259 * the excluded portion off the appropriate end.
267 * If the trimming actions above left a non-zero size, create an
268 * available entry for it.
271 if ((maxphyssz != 0) &&
272 (availsz + end - start > maxphyssz)) {
273 end = maxphyssz + start - availsz;
278 if (acnt > 0 && avail[acnt - 1] == (vm_paddr_t)start) {
279 avail[acnt - 1] = (vm_paddr_t)end;
281 avail[acnt++] = (vm_paddr_t)start;
282 avail[acnt++] = (vm_paddr_t)end;
284 availsz += end - start;
285 availmem += atop((vm_offset_t)(end - start));
287 if (acnt >= maxavail)
288 panic("Not enough space in the dump/phys_avail arrays");
293 if (prealmem != NULL)
294 *prealmem = totalmem;
299 * Insertion-sort a new entry into a regions list; sorted by start address.
302 insert_region(struct region *regions, size_t rcnt, vm_paddr_t addr,
303 vm_size_t size, uint32_t flags)
306 struct region *ep, *rp;
309 for (i = 0, rp = regions; i < rcnt; ++i, ++rp) {
310 if (rp->addr == addr && rp->size == size) /* Pure dup. */
312 if (flags == rp->flags) {
313 if (addr + size == rp->addr) {
317 } else if (rp->addr + rp->size == addr) {
322 if (addr < rp->addr) {
323 bcopy(rp, rp + 1, (ep - rp) * sizeof(*rp));
336 * Add a hardware memory region.
339 physmem_hardware_region(uint64_t pa, uint64_t sz)
342 * Filter out the page at PA 0x00000000. The VM can't handle it, as
343 * pmap_extract() == 0 means failure.
350 } else if (pa > MAX_PHYS_ADDR) {
351 /* This range is past usable memory, ignore it */
356 * Also filter out the page at the end of the physical address space --
357 * if addr is non-zero and addr+size is zero we wrapped to the next byte
358 * beyond what vm_paddr_t can express. That leads to a NULL pointer
359 * deref early in startup; work around it by leaving the last page out.
361 * XXX This just in: subtract out a whole megabyte, not just 1 page.
362 * Reducing the size by anything less than 1MB results in the NULL
363 * pointer deref in _vm_map_lock_read(). Better to give up a megabyte
364 * than leave some folks with an unusable system while we investigate.
366 if ((pa + sz) > (MAX_PHYS_ADDR - 1024 * 1024)) {
367 sz = MAX_PHYS_ADDR - pa + 1;
368 if (sz <= 1024 * 1024)
373 if (sz > 0 && hwcnt < nitems(hwregions))
374 hwcnt = insert_region(hwregions, hwcnt, pa, sz, 0);
378 * Add an exclusion region.
381 physmem_exclude_region(vm_paddr_t pa, vm_size_t sz, uint32_t exflags)
386 * Truncate the starting address down to a page boundary, and round the
387 * ending page up to a page boundary.
389 adj = pa - trunc_page(pa);
391 sz = round_page(sz + adj);
393 if (excnt >= nitems(exregions))
394 panic("failed to exclude region %#jx-%#jx", (uintmax_t)pa,
395 (uintmax_t)(pa + sz));
396 excnt = insert_region(exregions, excnt, pa, sz, exflags);
400 physmem_avail(vm_paddr_t *avail, size_t maxavail)
403 return (regions_to_avail(avail, EXFLAG_NOALLOC, maxavail, 0, NULL, NULL));
408 * Process all the regions added earlier into the global avail lists.
410 * Updates the kernel global 'physmem' with the number of physical pages
411 * available for use (all pages not in any exclusion region).
413 * Updates the kernel global 'Maxmem' with the page number one greater then the
414 * last page of physical memory in the system.
417 physmem_init_kernel_globals(void)
423 TUNABLE_ULONG_FETCH("hw.physmem", &hwphyssz);
425 regions_to_avail(dump_avail, EXFLAG_NODUMP, PHYS_AVAIL_ENTRIES,
426 hwphyssz, NULL, NULL);
427 nextidx = regions_to_avail(phys_avail, EXFLAG_NOALLOC,
428 PHYS_AVAIL_ENTRIES, hwphyssz, &physmem, &realmem);
430 panic("No memory entries in phys_avail");
431 Maxmem = atop(phys_avail[nextidx - 1]);
438 DB_SHOW_COMMAND(physmem, db_show_physmem)
441 physmem_dump_tables(db_printf);