// SPDX-License-Identifier: GPL-2.0-only /* * Low level x86 E820 memory map handling functions. * * The firmware and bootloader passes us the "E820 table", which is the primary * physical memory layout description available about x86 systems. * * The kernel takes the E820 memory layout and optionally modifies it with * quirks and other tweaks, and feeds that into the generic Linux memory * allocation code routines via a platform independent interface (memblock, etc.).
*/ #include <linux/crash_dump.h> #include <linux/memblock.h> #include <linux/suspend.h> #include <linux/acpi.h> #include <linux/firmware-map.h> #include <linux/sort.h> #include <linux/memory_hotplug.h>
#include <asm/e820/api.h> #include <asm/setup.h>
/* * We organize the E820 table into three main data structures: * * - 'e820_table_firmware': the original firmware version passed to us by the * bootloader - not modified by the kernel. It is composed of two parts: * the first 128 E820 memory entries in boot_params.e820_table and the remaining * (if any) entries of the SETUP_E820_EXT nodes. We use this to: * * - the hibernation code uses it to generate a kernel-independent CRC32 * checksum of the physical memory layout of a system. * * - 'e820_table_kexec': a slightly modified (by the kernel) firmware version * passed to us by the bootloader - the major difference between * e820_table_firmware[] and this one is that e820_table_kexec[] * might be modified by the kexec itself to fake an mptable. * We use this to: * * - kexec, which is a bootloader in disguise, uses the original E820 * layout to pass to the kexec-ed kernel. This way the original kernel * can have a restricted E820 map while the kexec()-ed kexec-kernel * can have access to full memory - etc. * * Export the memory layout via /sys/firmware/memmap. kexec-tools uses * the entries to create an E820 table for the kexec kernel. * * kexec_file_load in-kernel code uses the table for the kexec kernel. * * - 'e820_table': this is the main E820 table that is massaged by the * low level x86 platform code, or modified by boot parameters, before * passed on to higher level MM layers. * * Once the E820 map has been converted to the standard Linux memory layout * information its role stops - modifying it has no effect and does not get * re-propagated. So its main role is a temporary bootstrap storage of firmware * specific memory layout data during early bootup.
*/ staticstruct e820_table e820_table_init __initdata; staticstruct e820_table e820_table_kexec_init __initdata; staticstruct e820_table e820_table_firmware_init __initdata;
/* For PCI or other memory-mapped resources */ unsignedlong pci_mem_start = 0xaeedbabe; #ifdef CONFIG_PCI
EXPORT_SYMBOL(pci_mem_start); #endif
/* * This function checks if any part of the range <start,end> is mapped * with type.
*/ staticbool _e820__mapped_any(struct e820_table *table,
u64 start, u64 end, enum e820_type type)
{ int i;
for (i = 0; i < table->nr_entries; i++) { struct e820_entry *entry = &table->entries[i];
if (type && entry->type != type) continue; if (entry->addr >= end || entry->addr + entry->size <= start) continue; returntrue;
} returnfalse;
}
/* * This function checks if the entire <start,end> range is mapped with 'type'. * * Note: this function only works correctly once the E820 table is sorted and * not-overlapping (at least for the range specified), which is the case normally.
*/ staticstruct e820_entry *__e820__mapped_all(u64 start, u64 end, enum e820_type type)
{ int i;
for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i];
if (type && entry->type != type) continue;
/* Is the region (part) in overlap with the current region? */ if (entry->addr >= end || entry->addr + entry->size <= start) continue;
/* * If the region is at the beginning of <start,end> we move * 'start' to the end of the region since it's ok until there
*/ if (entry->addr <= start)
start = entry->addr + entry->size;
/* * If 'start' is now at or beyond 'end', we're done, full * coverage of the desired range exists:
*/ if (start >= end) return entry;
}
return NULL;
}
/* * This function checks if the entire range <start,end> is mapped with type.
*/ bool __init e820__mapped_all(u64 start, u64 end, enum e820_type type)
{ return __e820__mapped_all(start, end, type);
}
/* * This function returns the type associated with the range <start,end>.
*/ int e820__get_entry_type(u64 start, u64 end)
{ struct e820_entry *entry = __e820__mapped_all(start, end, 0);
return entry ? entry->type : -EINVAL;
}
/* * Add a memory region to the kernel E820 map.
*/ staticvoid __init __e820__range_add(struct e820_table *table, u64 start, u64 size, enum e820_type type)
{ int x = table->nr_entries;
if (x >= ARRAY_SIZE(table->entries)) {
pr_err("too many entries; ignoring [mem %#010llx-%#010llx]\n",
start, start + size - 1); return;
}
/* * Sanitize an E820 map. * * Some E820 layouts include overlapping entries. The following * replaces the original E820 map with a new one, removing overlaps, * and resolving conflicting memory types in favor of highest * numbered type. * * The input parameter 'entries' points to an array of 'struct * e820_entry' which on entry has elements in the range [0, *nr_entries) * valid, and which has space for up to max_nr_entries entries. * On return, the resulting sanitized E820 map entries will be in * overwritten in the same location, starting at 'entries'. * * The integer pointed to by nr_entries must be valid on entry (the * current number of valid entries located at 'entries'). If the * sanitizing succeeds the *nr_entries will be updated with the new * number of valid entries (something no more than max_nr_entries). * * The return value from e820__update_table() is zero if it * successfully 'sanitized' the map entries passed in, and is -1 * if it did nothing, which can happen if either of (1) it was * only passed one map entry, or (2) any of the input map entries * were invalid (start + size < start, meaning that the size was * so big the described memory range wrapped around through zero.) * * Visually we're performing the following * (1,2,3,4 = memory types)... * * Sample memory map (w/overlaps): * ____22__________________ * ______________________4_ * ____1111________________ * _44_____________________ * 11111111________________ * ____________________33__ * ___________44___________ * __________33333_________ * ______________22________ * ___________________2222_ * _________111111111______ * _____________________11_ * _________________4______ * * Sanitized equivalent (no overlap): * 1_______________________ * _44_____________________ * ___1____________________ * ____22__________________ * ______11________________ * _________1______________ * __________3_____________ * ___________44___________ * _____________33_________ * _______________2________ * ________________1_______ * _________________4______ * ___________________2____ * ____________________33__ * ______________________4_
*/ struct change_member { /* Pointer to the original entry: */ struct e820_entry *entry; /* Address for this change point: */ unsignedlonglong addr;
};
/* * Inputs are pointers to two elements of change_point[]. If their * addresses are not equal, their difference dominates. If the addresses * are equal, then consider one that represents the end of its region * to be greater than one that does not.
*/ if (ap->addr != bp->addr) return ap->addr > bp->addr ? 1 : -1;
staticbool e820_nomerge(enum e820_type type)
{ /* * These types may indicate distinct platform ranges aligned to * numa node, protection domain, performance domain, or other * boundaries. Do not merge them.
*/ if (type == E820_TYPE_PRAM) returntrue; if (type == E820_TYPE_SOFT_RESERVED) returntrue; returnfalse;
}
/* If there's only one memory region, don't bother: */ if (table->nr_entries < 2) return -1;
BUG_ON(table->nr_entries > max_nr_entries);
/* Bail out if we find any unreasonable addresses in the map: */ for (i = 0; i < table->nr_entries; i++) { if (entries[i].addr + entries[i].size < entries[i].addr) return -1;
}
/* Create pointers for initial change-point information (for sorting): */ for (i = 0; i < 2 * table->nr_entries; i++)
change_point[i] = &change_point_list[i];
/* * Record all known change-points (starting and ending addresses), * omitting empty memory regions:
*/
chg_idx = 0; for (i = 0; i < table->nr_entries; i++) { if (entries[i].size != 0) {
change_point[chg_idx]->addr = entries[i].addr;
change_point[chg_idx++]->entry = &entries[i];
change_point[chg_idx]->addr = entries[i].addr + entries[i].size;
change_point[chg_idx++]->entry = &entries[i];
}
}
chg_nr = chg_idx;
/* Sort change-point list by memory addresses (low -> high): */
sort(change_point, chg_nr, sizeof(*change_point), cpcompare, NULL);
/* Create a new memory map, removing overlaps: */
overlap_entries = 0; /* Number of entries in the overlap table */
new_nr_entries = 0; /* Index for creating new map entries */
last_type = 0; /* Start with undefined memory type */
last_addr = 0; /* Start with 0 as last starting address */
/* Loop through change-points, determining effect on the new map: */ for (chg_idx = 0; chg_idx < chg_nr; chg_idx++) { /* Keep track of all overlapping entries */ if (change_point[chg_idx]->addr == change_point[chg_idx]->entry->addr) { /* Add map entry to overlap list (> 1 entry implies an overlap) */
overlap_list[overlap_entries++] = change_point[chg_idx]->entry;
} else { /* Remove entry from list (order independent, so swap with last): */ for (i = 0; i < overlap_entries; i++) { if (overlap_list[i] == change_point[chg_idx]->entry)
overlap_list[i] = overlap_list[overlap_entries-1];
}
overlap_entries--;
} /* * If there are overlapping entries, decide which * "type" to use (larger value takes precedence -- * 1=usable, 2,3,4,4+=unusable)
*/
current_type = 0; for (i = 0; i < overlap_entries; i++) { if (overlap_list[i]->type > current_type)
current_type = overlap_list[i]->type;
}
/* Continue building up new map based on this information: */ if (current_type != last_type || e820_nomerge(current_type)) { if (last_type) {
new_entries[new_nr_entries].size = change_point[chg_idx]->addr - last_addr; /* Move forward only if the new size was non-zero: */ if (new_entries[new_nr_entries].size != 0) /* No more space left for new entries? */ if (++new_nr_entries >= max_nr_entries) break;
} if (current_type) {
new_entries[new_nr_entries].addr = change_point[chg_idx]->addr;
new_entries[new_nr_entries].type = current_type;
last_addr = change_point[chg_idx]->addr;
}
last_type = current_type;
}
}
/* Copy the new entries into the original location: */
memcpy(entries, new_entries, new_nr_entries*sizeof(*entries));
table->nr_entries = new_nr_entries;
while (nr_entries) {
u64 start = entry->addr;
u64 size = entry->size;
u64 end = start + size - 1;
u32 type = entry->type;
/* Ignore the entry on 64-bit overflow: */ if (start > end && likely(size)) return -1;
e820__range_add(start, size, type);
entry++;
nr_entries--;
} return 0;
}
/* * Copy the BIOS E820 map into a safe place. * * Sanity-check it while we're at it.. * * If we're lucky and live on a modern system, the setup code * will have given us a memory map that we can use to properly * set up memory. If we aren't, we'll fake a memory map.
*/ staticint __init append_e820_table(struct boot_e820_entry *entries, u32 nr_entries)
{ /* Only one memory region (or negative)? Ignore it */ if (nr_entries < 2) return -1;
/* * Left range could be head or tail, so need to update * its size first:
*/
entry->size -= final_end - final_start; if (entry->addr < final_start) continue;
/* * Left range could be head or tail, so need to update * the size first:
*/
entry->size -= final_end - final_start; if (entry->addr < final_start) continue;
/* * Search for a gap in the E820 memory space from 0 to MAX_GAP_END (4GB).
*/ staticint __init e820_search_gap(unsignedlong *gapstart, unsignedlong *gapsize)
{ unsignedlonglong last = MAX_GAP_END; int i = e820_table->nr_entries; int found = 0;
while (--i >= 0) { unsignedlonglong start = e820_table->entries[i].addr; unsignedlonglong end = start + e820_table->entries[i].size;
/* * Since "last" is at most 4GB, we know we'll * fit in 32 bits if this condition is true:
*/ if (last > end) { unsignedlong gap = last - end;
if (gap >= *gapsize) {
*gapsize = gap;
*gapstart = end;
found = 1;
}
} if (start < last)
last = start;
} return found;
}
/* * Search for the biggest gap in the low 32 bits of the E820 * memory space. We pass this space to the PCI subsystem, so * that it can assign MMIO resources for hotplug or * unconfigured devices in. * * Hopefully the BIOS let enough space left.
*/
__init void e820__setup_pci_gap(void)
{ unsignedlong gapstart, gapsize; int found;
gapsize = 0x400000;
found = e820_search_gap(&gapstart, &gapsize);
if (!found) { #ifdef CONFIG_X86_64
gapstart = (max_pfn << PAGE_SHIFT) + 1024*1024;
pr_err("Cannot find an available gap in the 32-bit address range\n");
pr_err("PCI devices with unassigned 32-bit BARs may not work!\n"); #else
gapstart = 0x10000000; #endif
}
pr_info("[mem %#010lx-%#010lx] available for PCI devices\n",
gapstart, gapstart + gapsize - 1);
}
/* * Called late during init, in free_initmem(). * * Initial e820_table and e820_table_kexec are largish __initdata arrays. * * Copy them to a (usually much smaller) dynamically allocated area that is * sized precisely after the number of e820 entries. * * This is done after we've performed all the fixes and tweaks to the tables. * All functions which modify them are __init functions, which won't exist * after free_initmem().
*/
__init void e820__reallocate_tables(void)
{ struct e820_table *n; int size;
/* * Because of the small fixed size of struct boot_params, only the first * 128 E820 memory entries are passed to the kernel via boot_params.e820_table, * the remaining (if any) entries are passed via the SETUP_E820_EXT node of * struct setup_data, which is parsed here.
*/ void __init e820__memory_setup_extended(u64 phys_addr, u32 data_len)
{ int entries; struct boot_e820_entry *extmap; struct setup_data *sdata;
/* * Find the ranges of physical addresses that do not correspond to * E820 RAM areas and register the corresponding pages as 'nosave' for * hibernation (32-bit) or software suspend and suspend to RAM (64-bit). * * This function requires the E820 map to be sorted and without any * overlapping entries.
*/ void __init e820__register_nosave_regions(unsignedlong limit_pfn)
{ int i;
u64 last_addr = 0;
for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i];
if (entry->type != E820_TYPE_RAM) continue;
if (last_addr < entry->addr)
register_nosave_region(PFN_DOWN(last_addr), PFN_UP(entry->addr));
#ifdef CONFIG_ACPI /* * Register ACPI NVS memory regions, so that we can save/restore them during * hibernation and the subsequent resume:
*/ staticint __init e820__register_nvs_regions(void)
{ int i;
for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i];
if (entry->type == E820_TYPE_NVS)
acpi_nvs_register(entry->addr, entry->size);
}
/* * Allocate the requested number of bytes with the requested alignment * and return (the physical address) to the caller. Also register this * range in the 'kexec' E820 table as a reserved range. * * This allows kexec to fake a new mptable, as if it came from the real * system.
*/
u64 __init e820__memblock_alloc_reserved(u64 size, u64 align)
{
u64 addr;
addr = memblock_phys_alloc(size, align); if (addr) {
e820__range_update_table(e820_table_kexec, addr, size, E820_TYPE_RAM, E820_TYPE_RESERVED);
pr_info("update e820_table_kexec for e820__memblock_alloc_reserved()\n");
e820__update_table_kexec();
}
/* * Find the highest page frame number we have available
*/ staticunsignedlong __init e820__end_ram_pfn(unsignedlong limit_pfn)
{ int i; unsignedlong last_pfn = 0; unsignedlong max_arch_pfn = MAX_ARCH_PFN;
for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i]; unsignedlong start_pfn; unsignedlong end_pfn;
if (entry->type != E820_TYPE_RAM &&
entry->type != E820_TYPE_ACPI) continue;
/* * Called after parse_early_param(), after early parameters (such as mem=) * have been processed, in which case we already have an E820 table filled in * via the parameter callback function(s), but it's not sorted and printed yet:
*/ void __init e820__finish_early_params(void)
{ if (userdef) { if (e820__update_table(e820_table) < 0)
early_panic("Invalid user supplied memory map");
staticconstchar *__init e820_type_to_string(struct e820_entry *entry)
{ switch (entry->type) { case E820_TYPE_RAM: return"System RAM"; case E820_TYPE_ACPI: return"ACPI Tables"; case E820_TYPE_NVS: return"ACPI Non-volatile Storage"; case E820_TYPE_UNUSABLE: return"Unusable memory"; case E820_TYPE_PRAM: return"Persistent Memory (legacy)"; case E820_TYPE_PMEM: return"Persistent Memory"; case E820_TYPE_RESERVED: return"Reserved"; case E820_TYPE_SOFT_RESERVED: return"Soft Reserved"; default: return"Unknown E820 type";
}
}
staticunsignedlong __init e820_type_to_iomem_type(struct e820_entry *entry)
{ switch (entry->type) { case E820_TYPE_RAM: return IORESOURCE_SYSTEM_RAM; case E820_TYPE_ACPI: /* Fall-through: */ case E820_TYPE_NVS: /* Fall-through: */ case E820_TYPE_UNUSABLE: /* Fall-through: */ case E820_TYPE_PRAM: /* Fall-through: */ case E820_TYPE_PMEM: /* Fall-through: */ case E820_TYPE_RESERVED: /* Fall-through: */ case E820_TYPE_SOFT_RESERVED: /* Fall-through: */ default: return IORESOURCE_MEM;
}
}
staticunsignedlong __init e820_type_to_iores_desc(struct e820_entry *entry)
{ switch (entry->type) { case E820_TYPE_ACPI: return IORES_DESC_ACPI_TABLES; case E820_TYPE_NVS: return IORES_DESC_ACPI_NV_STORAGE; case E820_TYPE_PMEM: return IORES_DESC_PERSISTENT_MEMORY; case E820_TYPE_PRAM: return IORES_DESC_PERSISTENT_MEMORY_LEGACY; case E820_TYPE_RESERVED: return IORES_DESC_RESERVED; case E820_TYPE_SOFT_RESERVED: return IORES_DESC_SOFT_RESERVED; case E820_TYPE_RAM: /* Fall-through: */ case E820_TYPE_UNUSABLE: /* Fall-through: */ default: return IORES_DESC_NONE;
}
}
staticbool __init do_mark_busy(enum e820_type type, struct resource *res)
{ /* this is the legacy bios/dos rom-shadow + mmio region */ if (res->start < (1ULL<<20)) returntrue;
/* * Treat persistent memory and other special memory ranges like * device memory, i.e. reserve it for exclusive use of a driver
*/ switch (type) { case E820_TYPE_RESERVED: case E820_TYPE_SOFT_RESERVED: case E820_TYPE_PRAM: case E820_TYPE_PMEM: returnfalse; case E820_TYPE_RAM: case E820_TYPE_ACPI: case E820_TYPE_NVS: case E820_TYPE_UNUSABLE: default: returntrue;
}
}
/* * Mark E820 reserved areas as busy for the resource manager:
*/
/* * Don't register the region that could be conflicted with * PCI device BAR resources and insert them later in * pcibios_resource_survey():
*/ if (do_mark_busy(entry->type, res)) {
res->flags |= IORESOURCE_BUSY;
insert_resource(&iomem_resource, res);
}
res++;
}
/* Expose the kexec e820 table to the sysfs. */ for (i = 0; i < e820_table_kexec->nr_entries; i++) { struct e820_entry *entry = e820_table_kexec->entries + i;
/* * How much should we pad the end of RAM, depending on where it is?
*/ staticunsignedlong __init ram_alignment(resource_size_t pos)
{ unsignedlong mb = pos >> 20;
/* To 64kB in the first megabyte */ if (!mb) return 64*1024;
/* To 1MB in the first 16MB */ if (mb < 16) return 1024*1024;
/* To 64MB for anything above that */ return 64*1024*1024;
}
#define MAX_RESOURCE_SIZE ((resource_size_t)-1)
void __init e820__reserve_resources_late(void)
{ int i; struct resource *res;
res = e820_res; for (i = 0; i < e820_table->nr_entries; i++) { if (!res->parent && res->end)
insert_resource_expand_to_fit(&iomem_resource, res);
res++;
}
/* * Try to bump up RAM regions to reasonable boundaries, to * avoid stolen RAM:
*/ for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i];
u64 start, end;
if (entry->type != E820_TYPE_RAM) continue;
start = entry->addr + entry->size;
end = round_up(start, ram_alignment(start)) - 1; if (end > MAX_RESOURCE_SIZE)
end = MAX_RESOURCE_SIZE; if (start >= end) continue;
/* * Pass the firmware (bootloader) E820 map to the kernel and process it:
*/ char *__init e820__memory_setup_default(void)
{ char *who = "BIOS-e820";
/* * Try to copy the BIOS-supplied E820-map. * * Otherwise fake a memory map; one section from 0k->640k, * the next section from 1mb->appropriate_mem_k
*/ if (append_e820_table(boot_params.e820_table, boot_params.e820_entries) < 0) {
u64 mem_size;
/* Compare results from other methods and take the one that gives more RAM: */ if (boot_params.alt_mem_k < boot_params.screen_info.ext_mem_k) {
mem_size = boot_params.screen_info.ext_mem_k;
who = "BIOS-88";
} else {
mem_size = boot_params.alt_mem_k;
who = "BIOS-e801";
}
/* We just appended a lot of ranges, sanitize the table: */
e820__update_table(e820_table);
return who;
}
/* * Calls e820__memory_setup_default() in essence to pick up the firmware/bootloader * E820 map - with an optional platform quirk available for virtual platforms * to override this method of boot environment processing:
*/ void __init e820__memory_setup(void)
{ char *who;
/* This is a firmware interface ABI - make sure we don't break it: */
BUILD_BUG_ON(sizeof(struct boot_e820_entry) != 20);
void __init e820__memblock_setup(void)
{ int i;
u64 end;
#ifdef CONFIG_MEMORY_HOTPLUG /* * Memory used by the kernel cannot be hot-removed because Linux * cannot migrate the kernel pages. When memory hotplug is * enabled, we should prevent memblock from allocating memory * for the kernel. * * ACPI SRAT records all hotpluggable memory ranges. But before * SRAT is parsed, we don't know about it. * * The kernel image is loaded into memory at very early time. We * cannot prevent this anyway. So on NUMA system, we set any * node the kernel resides in as un-hotpluggable. * * Since on modern servers, one node could have double-digit * gigabytes memory, we can assume the memory around the kernel * image is also un-hotpluggable. So before SRAT is parsed, just * allocate memory near the kernel image to try the best to keep * the kernel away from hotpluggable memory.
*/ if (movable_node_is_enabled())
memblock_set_bottom_up(true); #endif
/* * At this point only the first megabyte is mapped for sure, the * rest of the memory cannot be used for memblock resizing
*/
memblock_set_current_limit(ISA_END_ADDRESS);
/* * The bootstrap memblock region count maximum is 128 entries * (INIT_MEMBLOCK_REGIONS), but EFI might pass us more E820 entries * than that - so allow memblock resizing. * * This is safe, because this call happens pretty late during x86 setup, * so we know about reserved memory regions already. (This is important * so that memblock resizing does no stomp over reserved areas.)
*/
memblock_allow_resize();
for (i = 0; i < e820_table->nr_entries; i++) { struct e820_entry *entry = &e820_table->entries[i];
end = entry->addr + entry->size; if (end != (resource_size_t)end) continue;
if (entry->type == E820_TYPE_SOFT_RESERVED)
memblock_reserve(entry->addr, entry->size);
if (entry->type != E820_TYPE_RAM) continue;
memblock_add(entry->addr, entry->size);
}
/* * At this point memblock is only allowed to allocate from memory * below 1M (aka ISA_END_ADDRESS) up until direct map is completely set * up in init_mem_mapping(). * * KHO kernels are special and use only scratch memory for memblock * allocations, but memory below 1M is ignored by kernel after early * boot and cannot be naturally marked as scratch. * * To allow allocation of the real-mode trampoline and a few (if any) * other very early allocations from below 1M forcibly mark the memory * below 1M as scratch. * * After real mode trampoline is allocated, we clear that scratch * marking.
*/
memblock_mark_kho_scratch(0, SZ_1M);
/* * 32-bit systems are limited to 4BG of memory even with HIGHMEM and * to even less without it. * Discard memory after max_pfn - the actual limit detected at runtime.
*/ if (IS_ENABLED(CONFIG_X86_32))
memblock_remove(PFN_PHYS(max_pfn), -1);
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.