Aren/toaruos
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
79bd90e44c
toaruos/kernel/arch/x86_64/mmu.c

1213 lines
38 KiB
C
Raw Blame History

/**
* @file kernel/arch/x86_64/mmu.c
* @brief Memory management facilities for x86-64
*
* Frame allocation and mapping routines for x86-64.
*
* @copyright
* This file is part of ToaruOS and is released under the terms
* of the NCSA / University of Illinois License - see LICENSE.md
* Copyright (C) 2021 K. Lange
*/
#include <stdint.h>
#include <kernel/assert.h>
#include <kernel/string.h>
#include <kernel/printf.h>
#include <kernel/process.h>
#include <kernel/spinlock.h>
#include <kernel/misc.h>
#include <kernel/mmu.h>
#include <kernel/arch/x86_64/pml.h>
extern void arch_tlb_shootdown(uintptr_t);
/**
* bitmap page allocator for 4KiB pages
*/
static volatile uint32_t *frames;
static size_t nframes;
static size_t total_memory = 0;
static size_t unavailable_memory = 0;
static uint8_t * mem_refcounts = NULL;
#define PAGE_SHIFT 12
#define PAGE_SIZE 0x1000UL
#define PAGE_SIZE_MASK 0xFFFFffffFFFFf000UL
#define PAGE_LOW_MASK 0x0000000000000FFFUL
#define LARGE_PAGE_SIZE 0x200000UL
#define KERNEL_HEAP_START 0xFFFFff0000000000UL
#define MMIO_BASE_START 0xffffff1fc0000000UL
#define HIGH_MAP_REGION 0xffffff8000000000UL
#define MODULE_BASE_START 0xffffffff80000000UL
/* These are actually defined in the shm layer... */
#define USER_SHM_LOW 0x0000000200000000UL
#define USER_SHM_HIGH 0x0000000400000000UL
#define USER_DEVICE_MAP 0x0000000100000000UL
#define USER_PML_ACCESS 0x07
#define KERNEL_PML_ACCESS 0x03
#define LARGE_PAGE_BIT 0x80
#define PDP_MASK 0x3fffffffUL
#define PD_MASK 0x1fffffUL
#define PT_MASK PAGE_LOW_MASK
#define ENTRY_MASK 0x1FF
#define PHYS_MASK 0x7fffffffffUL
#define CANONICAL_MASK 0xFFFFffffFFFFUL
#define INDEX_FROM_BIT(b) ((b) >> 5)
#define OFFSET_FROM_BIT(b) ((b) & 0x1F)
/**
* @brief Mark a physical page frame as in use.
*
* Sets the bitmap allocator bit for a frame.
*
* @param frame_addr Address of the frame (not index!)
*/
void mmu_frame_set(uintptr_t frame_addr) {
/* If the frame is within bounds... */
if (frame_addr < nframes * PAGE_SIZE) {
uint64_t frame = frame_addr >> 12;
uint64_t index = INDEX_FROM_BIT(frame);
uint32_t offset = OFFSET_FROM_BIT(frame);
frames[index] |= ((uint32_t)1 << offset);
asm ("" ::: "memory");
}
}
static uintptr_t lowest_available = 0;
/**
* @brief Mark a physical page frame as available.
*
* Clears the bitmap allocator bit for a frame.
*
* @param frame_addr Address of the frame (not index!)
*/
void mmu_frame_clear(uintptr_t frame_addr) {
/* If the frame is within bounds... */
if (frame_addr < nframes * PAGE_SIZE) {
uint64_t frame = frame_addr >> PAGE_SHIFT;
uint64_t index = INDEX_FROM_BIT(frame);
uint32_t offset = OFFSET_FROM_BIT(frame);
frames[index] &= ~((uint32_t)1 << offset);
asm ("" ::: "memory");
if (frame < lowest_available) lowest_available = frame;
}
}
/**
* @brief Determine if a physical page is available for use.
*
* @param frame_addr Address of the frame (not index!)
* @returns 0 if available, 1 otherwise.
*/
int mmu_frame_test(uintptr_t frame_addr) {
if (!(frame_addr < nframes * PAGE_SIZE)) return 1;
uint64_t frame = frame_addr >> PAGE_SHIFT;
uint64_t index = INDEX_FROM_BIT(frame);
uint32_t offset = OFFSET_FROM_BIT(frame);
asm ("" ::: "memory");
return !!(frames[index] & ((uint32_t)1 << offset));
}
static spin_lock_t frame_alloc_lock = { 0 };
static spin_lock_t kheap_lock = { 0 };
static spin_lock_t mmio_space_lock = { 0 };
static spin_lock_t module_space_lock = { 0 };
/**
* @brief Find the first range of @p n contiguous frames.
*
* If a large enough region could not be found, results are fatal.
*/
uintptr_t mmu_first_n_frames(int n) {
for (uint64_t i = 0; i < nframes * PAGE_SIZE; i += PAGE_SIZE) {
int bad = 0;
for (int j = 0; j < n; ++j) {
if (mmu_frame_test(i + PAGE_SIZE * j)) {
bad = j + 1;
}
}
if (!bad) {
return i / PAGE_SIZE;
}
}
arch_fatal_prepare();
dprintf("Failed to allocate %d contiguous frames.\n", n);
arch_dump_traceback();
arch_fatal();
return (uintptr_t)-1;
}
/**
* @brief Find the first available frame from the bitmap.
*/
uintptr_t mmu_first_frame(void) {
uintptr_t i, j;
for (i = INDEX_FROM_BIT(lowest_available); i < INDEX_FROM_BIT(nframes); ++i) {
if (frames[i] != (uint32_t)-1) {
for (j = 0; j < (sizeof(uint32_t)*8); ++j) {
uint32_t testFrame = (uint32_t)1 << j;
if (!(frames[i] & testFrame)) {
uintptr_t out = (i << 5) + j;
lowest_available = out + 1;
return out;
}
}
}
}
arch_fatal_prepare();
dprintf("Out of memory.\n");
arch_dump_traceback();
arch_fatal();
return (uintptr_t)-1;
}
/**
* @brief Set the flags for a page, and allocate a frame for it if needed.
*
* Sets the page bits based on the the value of @p flags.
* If @p page->bits.page is unset, a new frame will be allocated.
*/
void mmu_frame_allocate(union PML * page, unsigned int flags) {
if (page->bits.page == 0) {
spin_lock(frame_alloc_lock);
uintptr_t index = mmu_first_frame();
mmu_frame_set(index << PAGE_SHIFT);
page->bits.page = index;
spin_unlock(frame_alloc_lock);
}
page->bits.size = 0;
page->bits.present = 1;
page->bits.writable = (flags & MMU_FLAG_WRITABLE) ? 1 : 0;
page->bits.user = (flags & MMU_FLAG_KERNEL) ? 0 : 1;
page->bits.nocache = (flags & MMU_FLAG_NOCACHE) ? 1 : 0;
page->bits.writethrough = (flags & MMU_FLAG_WRITETHROUGH) ? 1 : 0;
page->bits.size = (flags & MMU_FLAG_SPEC) ? 1 : 0;
page->bits.nx = (flags & MMU_FLAG_NOEXECUTE) ? 1 : 0;
}
/**
* @brief Map the given page to the requested physical address.
*/
void mmu_frame_map_address(union PML * page, unsigned int flags, uintptr_t physAddr) {
mmu_frame_set(physAddr);
page->bits.page = physAddr >> PAGE_SHIFT;
mmu_frame_allocate(page, flags);
}
/* Initial memory maps loaded by boostrap */
#define _pagemap __attribute__((aligned(PAGE_SIZE))) = {0}
union PML init_page_region[3][512] _pagemap;
union PML high_base_pml[512] _pagemap;
union PML heap_base_pml[512] _pagemap;
union PML heap_base_pd[512] _pagemap;
union PML heap_base_pt[512*3] _pagemap;
union PML low_base_pmls[34][512] _pagemap;
union PML twom_high_pds[64][512] _pagemap;
/**
* @brief Maps a frame address to a virtual address.
*
* Returns the virtual address within the general-purpose
* identity mapping region for the given physical frame address.
* This address is not suitable for some operations, such as MMIO.
*/
void * mmu_map_from_physical(uintptr_t frameaddress) {
return (void*)(frameaddress | HIGH_MAP_REGION);
}
union PML * mmu_get_page_other(union PML * root, uintptr_t virtAddr) {
uintptr_t realBits = virtAddr & CANONICAL_MASK;
uintptr_t pageAddr = realBits >> PAGE_SHIFT;
unsigned int pml4_entry = (pageAddr >> 27) & ENTRY_MASK;
unsigned int pdp_entry = (pageAddr >> 18) & ENTRY_MASK;
unsigned int pd_entry = (pageAddr >> 9) & ENTRY_MASK;
unsigned int pt_entry = (pageAddr) & ENTRY_MASK;
/* Get the PML4 entry for this address */
if (!root[pml4_entry].bits.present) {
return NULL;
}
union PML * pdp = mmu_map_from_physical((uintptr_t)root[pml4_entry].bits.page << PAGE_SHIFT);
if (!pdp[pdp_entry].bits.present) {
return NULL;
}
if (pdp[pdp_entry].bits.size) {
return NULL;
}
union PML * pd = mmu_map_from_physical((uintptr_t)pdp[pdp_entry].bits.page << PAGE_SHIFT);
if (!pd[pd_entry].bits.present) {
return NULL;
}
if (pd[pd_entry].bits.size) {
return NULL;
}
union PML * pt = mmu_map_from_physical((uintptr_t)pd[pd_entry].bits.page << PAGE_SHIFT);
return (union PML *)&pt[pt_entry];
}
/**
* @brief Find the physical address at a given virtual address.
*
* Calculates the physical address of the page backing the virtual
* address @p virtAddr. If no page is mapped, a negative value
* is returned indicating which level of the page directory is
* unmapped from -1 (no PDP) to -4 (page not present in table).
*/
uintptr_t mmu_map_to_physical(union PML * root, uintptr_t virtAddr) {
uintptr_t realBits = virtAddr & CANONICAL_MASK;
uintptr_t pageAddr = realBits >> PAGE_SHIFT;
unsigned int pml4_entry = (pageAddr >> 27) & ENTRY_MASK;
unsigned int pdp_entry = (pageAddr >> 18) & ENTRY_MASK;
unsigned int pd_entry = (pageAddr >> 9) & ENTRY_MASK;
unsigned int pt_entry = (pageAddr) & ENTRY_MASK;
/* Get the PML4 entry for this address */
if (!root[pml4_entry].bits.present) return (uintptr_t)-1;
union PML * pdp = mmu_map_from_physical((uintptr_t)root[pml4_entry].bits.page << PAGE_SHIFT);
if (!pdp[pdp_entry].bits.present) return (uintptr_t)-2;
if (pdp[pdp_entry].bits.size) return ((uintptr_t)pdp[pdp_entry].bits.page << PAGE_SHIFT) | (virtAddr & PDP_MASK);
union PML * pd = mmu_map_from_physical((uintptr_t)pdp[pdp_entry].bits.page << PAGE_SHIFT);
if (!pd[pd_entry].bits.present) return (uintptr_t)-3;
if (pd[pd_entry].bits.size) return ((uintptr_t)pd[pd_entry].bits.page << PAGE_SHIFT) | (virtAddr & PD_MASK);
union PML * pt = mmu_map_from_physical((uintptr_t)pd[pd_entry].bits.page << PAGE_SHIFT);
if (!pt[pt_entry].bits.present) return (uintptr_t)-4;
return ((uintptr_t)pt[pt_entry].bits.page << PAGE_SHIFT) | (virtAddr & PT_MASK);
}
/**
* @brief Obtain the page entry for a virtual address.
*
* Digs into the current page directory to obtain the page entry
* for a requested address @p virtAddr. If new intermediary directories
* need to be allocated and @p flags has @c MMU_GET_MAKE set, they
* will be allocated with the user access bits set. Otherwise,
* NULL will be returned. If the requested virtual address is within
* a large page, NULL will be returned.
*
* @param virtAddr Canonical virtual address offset.
* @param flags See @c MMU_GET_MAKE
* @returns the requested page entry, or NULL if doing so required allocating
* an intermediary paging level and @p flags did not have @c MMU_GET_MAKE set.
*/
union PML * mmu_get_page(uintptr_t virtAddr, int flags) {
uintptr_t realBits = virtAddr & CANONICAL_MASK;
uintptr_t pageAddr = realBits >> PAGE_SHIFT;
unsigned int pml4_entry = (pageAddr >> 27) & ENTRY_MASK;
unsigned int pdp_entry = (pageAddr >> 18) & ENTRY_MASK;
unsigned int pd_entry = (pageAddr >> 9) & ENTRY_MASK;
unsigned int pt_entry = (pageAddr) & ENTRY_MASK;
union PML * root = this_core->current_pml;
/* Get the PML4 entry for this address */
if (!root[pml4_entry].bits.present) {
if (!(flags & MMU_GET_MAKE)) goto _noentry;
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
/* zero it */
memset(mmu_map_from_physical(newPage), 0, PAGE_SIZE);
root[pml4_entry].raw = (newPage) | USER_PML_ACCESS;
}
union PML * pdp = mmu_map_from_physical((uintptr_t)root[pml4_entry].bits.page << PAGE_SHIFT);
if (!pdp[pdp_entry].bits.present) {
if (!(flags & MMU_GET_MAKE)) goto _noentry;
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
/* zero it */
memset(mmu_map_from_physical(newPage), 0, PAGE_SIZE);
pdp[pdp_entry].raw = (newPage) | USER_PML_ACCESS;
}
if (pdp[pdp_entry].bits.size) {
printf("Warning: Tried to get page for a 1GiB page!\n");
return NULL;
}
union PML * pd = mmu_map_from_physical((uintptr_t)pdp[pdp_entry].bits.page << PAGE_SHIFT);
if (!pd[pd_entry].bits.present) {
if (!(flags & MMU_GET_MAKE)) goto _noentry;
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
/* zero it */
memset(mmu_map_from_physical(newPage), 0, PAGE_SIZE);
pd[pd_entry].raw = (newPage) | USER_PML_ACCESS;
}
if (pd[pd_entry].bits.size) {
printf("Warning: Tried to get page for a 2MiB page!\n");
return NULL;
}
union PML * pt = mmu_map_from_physical((uintptr_t)pd[pd_entry].bits.page << PAGE_SHIFT);
return (union PML *)&pt[pt_entry];
_noentry:
printf("no entry for requested page\n");
return NULL;
}
/**
* @brief Increment the reference count for a physical page of memory.
*
* We allow up to 255 references to a page, so that we can track individual
* page reference counts in a big @c uint8_t array. If there are already
* that many references (that's a lot of forks!) we give up and do a regular
* copy of the page and the new copy is writable.
*
* @param frame Physical page index
* @returns 1 if there are already too many references to this page, 0 otherwise.
*/
int refcount_inc(uintptr_t frame) {
if (frame >= nframes) {
arch_fatal_prepare();
dprintf("%zu (inc, bad frame)\n", frame);
arch_dump_traceback();
arch_fatal();
}
if (mem_refcounts[frame] == 255) return 1;
mem_refcounts[frame]++;
return 0;
}
/**
* @brief Decrement the reference count for a physical page of memory.
*
* Panics if @p frame is invalid or has a zero reference count.
*
* @param frame Physical page index
* @returns the resulting reference count.
*/
uint8_t refcount_dec(uintptr_t frame) {
if (frame >= nframes) {
arch_fatal_prepare();
dprintf("%zu (dec, bad frame)\n", frame);
arch_dump_traceback();
arch_fatal();
}
if (mem_refcounts[frame] == 0) {
arch_fatal_prepare();
dprintf("%zu (dec, frame has no references)\n", frame);
arch_dump_traceback();
arch_fatal();
}
mem_refcounts[frame]--;
return mem_refcounts[frame];
}
/**
* @brief Handle user pages in mmu_clone
*
* Copies and updates reference counts for pages across forks.
* If a page was writable in the source directory, it will be marked
* read-only and have reference counts initialized for COW.
*
* If a page was already read-only, its reference count will
* be incremented for the new directory.
*
* @param pt_in Existing page table.
* @param pt_out New directory's page table.
* @param l Index into both page tables for this page.
* @param address Virtual address being referenced.
* @returns 0, generally
*/
int copy_page_maybe(union PML * pt_in, union PML * pt_out, size_t l, uintptr_t address) {
/* Can we cow the current page? */
spin_lock(frame_alloc_lock);
/* Is the page writable? */
if (pt_in[l].bits.writable) {
/* Then we need to initialize the refcounts */
if (mem_refcounts[pt_in[l].bits.page] != 0) {
arch_fatal_prepare();
dprintf("%#zx (page=%u) refcount = %u\n",
address, pt_in[l].bits.page, mem_refcounts[pt_in[l].bits.page]);
arch_dump_traceback();
arch_fatal();
return 1;
}
mem_refcounts[pt_in[l].bits.page] = 2;
pt_in[l].bits.writable = 0;
pt_in[l].bits.cow_pending = 1;
pt_out[l].raw = pt_in[l].raw;
asm ("" ::: "memory");
mmu_invalidate(address);
spin_unlock(frame_alloc_lock);
return 0;
}
/* Can we make a new reference? */
if (refcount_inc(pt_in[l].bits.page)) {
/* There are too many references to fit in our refcount table, so just make a new page. */
char * page_in = mmu_map_from_physical((uintptr_t)pt_in[l].bits.page << PAGE_SHIFT);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
char * page_out = mmu_map_from_physical(newPage);
memcpy(page_out,page_in,PAGE_SIZE);
pt_out[l].raw = 0;
pt_out[l].bits.present = 1;
pt_out[l].bits.user = 1;
pt_out[l].bits.page = newPage >> PAGE_SHIFT;
pt_out[l].bits.writable = 1;
pt_out[l].bits.cow_pending = 0;
asm ("" ::: "memory");
} else {
pt_out[l].raw = pt_in[l].raw;
}
spin_unlock(frame_alloc_lock);
return 0;
}
/**
* @brief When freeing a directory, handle individual user pages.
*
* If @p pt_in references a writable user page, we know we can
* free it immediately as it is the only reference to that page.
*
* Otherwise, we need to decrement the reference counts for read-only
* pages, as they are shared COW entries. Only if this was the last
* reference (refcount drops to 0) can we then proceed to free the
* underlying page.
*
* @param pt_in Start of page table
* @param l Offset into page table for this page
* @param address Virtual address being freed (was used for debugging)
* @returns 0, generally
*/
int free_page_maybe(union PML * pt_in, size_t l, uintptr_t address) {
if (pt_in[l].bits.writable) {
assert(mem_refcounts[pt_in[l].bits.page] == 0);
mmu_frame_clear((uintptr_t)pt_in[l].bits.page << PAGE_SHIFT);
return 0;
}
/* No more references */
if (refcount_dec(pt_in[l].bits.page) == 0) {
mmu_frame_clear((uintptr_t)pt_in[l].bits.page << PAGE_SHIFT);
}
return 0;
}
/**
* @brief Create a new address space with the same contents of an existing one.
*
* Allocates all of the necessary intermediary directory levels for a new address space
* and also copies data from the existing address space.
*
* TODO: This doesn't do any CoW and it's kinda complicated.
*
* @param from The directory to clone, or NULL to clone the kernel map.
* @returns a pointer to the new page directory, suitable for mapping to a physical address.
*/
union PML * mmu_clone(union PML * from) {
/* Clone the current PMLs... */
if (!from) from = this_core->current_pml;
/* First get a page for ourselves. */
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
union PML * pml4_out = mmu_map_from_physical(newPage);
/* Zero bottom half */
memset(&pml4_out[0], 0, 256 * sizeof(union PML));
/* Copy top half */
memcpy(&pml4_out[256], &from[256], 256 * sizeof(union PML));
/* Copy PDPs */
for (size_t i = 0; i < 256; ++i) {
if (from[i].bits.present) {
union PML * pdp_in = mmu_map_from_physical((uintptr_t)from[i].bits.page << PAGE_SHIFT);
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
union PML * pdp_out = mmu_map_from_physical(newPage);
memset(pdp_out, 0, 512 * sizeof(union PML));
pml4_out[i].raw = (newPage) | USER_PML_ACCESS;
/* Copy the PDs */
for (size_t j = 0; j < 512; ++j) {
if (pdp_in[j].bits.present) {
union PML * pd_in = mmu_map_from_physical((uintptr_t)pdp_in[j].bits.page << PAGE_SHIFT);
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
union PML * pd_out = mmu_map_from_physical(newPage);
memset(pd_out, 0, 512 * sizeof(union PML));
pdp_out[j].raw = (newPage) | USER_PML_ACCESS;
/* Now copy the PTs */
for (size_t k = 0; k < 512; ++k) {
if (pd_in[k].bits.present) {
union PML * pt_in = mmu_map_from_physical((uintptr_t)pd_in[k].bits.page << PAGE_SHIFT);
spin_lock(frame_alloc_lock);
uintptr_t newPage = mmu_first_frame() << PAGE_SHIFT;
mmu_frame_set(newPage);
spin_unlock(frame_alloc_lock);
union PML * pt_out = mmu_map_from_physical(newPage);
memset(pt_out, 0, 512 * sizeof(union PML));
pd_out[k].raw = (newPage) | USER_PML_ACCESS;
/* Now, finally, copy pages */
for (size_t l = 0; l < 512; ++l) {
uintptr_t address = ((i << (9 * 3 + 12)) | (j << (9*2 + 12)) | (k << (9 + 12)) | (l << PAGE_SHIFT));
if (address >= USER_DEVICE_MAP && address <= USER_SHM_HIGH) continue;
if (pt_in[l].bits.present) {
if (pt_in[l].bits.user) {
copy_page_maybe(pt_in, pt_out, l, address);
} else {
/* If it's not a user page, just copy directly */
pt_out[l].raw = pt_in[l].raw;
}
} /* Else, mmap'd files? */
}
}
}
}
}
}
}
return pml4_out;
}
/**
* @brief Allocate one physical page.
*
* @returns a frame index, not an address
*/
uintptr_t mmu_allocate_a_frame(void) {
spin_lock(frame_alloc_lock);
uintptr_t index = mmu_first_frame();
mmu_frame_set(index << PAGE_SHIFT);
spin_unlock(frame_alloc_lock);
return index;
}
/**
* @brief Allocate a number of contiguous physical pages.
*
* @returns a frame index, not an address
*/
uintptr_t mmu_allocate_n_frames(int n) {
spin_lock(frame_alloc_lock);
uintptr_t index = mmu_first_n_frames(n);
for (int i = 0; i < n; ++i) {
mmu_frame_set((index+i) << PAGE_SHIFT);
}
spin_unlock(frame_alloc_lock);
return index;
}
/**
* @brief Scans a directory to calculate how many user pages are in use.
*
* Calculates how many pages a userspace application has mapped, between
* its general memory space and stack. Excludes shared mappings, such
* as SHM or mapped devices.
*
* TODO: This can probably be reduced to check a smaller range, but as we
* currently stick the user stack at the top of the low half of the
* address space we just scan everything and exclude shared memory...
*
* @param from Top-level page directory to scan.
*/
size_t mmu_count_user(union PML * from) {
size_t out = 0;
for (size_t i = 0; i < 256; ++i) {
if (from[i].bits.present) {
union PML * pdp_in = mmu_map_from_physical((uintptr_t)from[i].bits.page << PAGE_SHIFT);
for (size_t j = 0; j < 512; ++j) {
if (pdp_in[j].bits.present) {
union PML * pd_in = mmu_map_from_physical((uintptr_t)pdp_in[j].bits.page << PAGE_SHIFT);
for (size_t k = 0; k < 512; ++k) {
if (pd_in[k].bits.present) {
union PML * pt_in = mmu_map_from_physical((uintptr_t)pd_in[k].bits.page << PAGE_SHIFT);
for (size_t l = 0; l < 512; ++l) {
/* Calculate final address to skip SHM */
uintptr_t address = ((i << (9 * 3 + 12)) | (j << (9*2 + 12)) | (k << (9 + 12)) | (l << PAGE_SHIFT));
if (address >= USER_DEVICE_MAP && address <= USER_SHM_HIGH) continue;
if (pt_in[l].bits.present) {
if (pt_in[l].bits.user) {
out++;
}
}
}
}
}
}
}
}
}
return out;
}
/**
* @brief Scans a directory to calculate how many shared memory pages are in use.
*
* At the moment, we only ever map shared pages to a specific region, so we just figure
* out how many present pages are in that region and that's the answer.
*
* @param from Top-level page directory to scan.
*/
size_t mmu_count_shm(union PML * from) {
size_t out = 0;
if (from[0].bits.present) {
union PML * pdp_in = mmu_map_from_physical((uintptr_t)from[0].bits.page << PAGE_SHIFT);
/* [0,8,0,0] through [0,15,511,511] map to our current SHM mapping region;
* if you change the bounds of that region, be sure to update this! */
for (size_t j = 8; j < 16; ++j) {
if (pdp_in[j].bits.present) {
union PML * pd_in = mmu_map_from_physical((uintptr_t)pdp_in[j].bits.page << PAGE_SHIFT);
for (size_t k = 0; k < 512; ++k) {
if (pd_in[k].bits.present) {
union PML * pt_in = mmu_map_from_physical((uintptr_t)pd_in[k].bits.page << PAGE_SHIFT);
for (size_t l = 0; l < 512; ++l) {
if (pt_in[l].bits.present) {
if (pt_in[l].bits.user) {
out++;
}
}
}
}
}
}
}
}
return out;
}
/**
* @brief Return the total amount of usable memory.
*
* @returns the total amount of usable memory in KiB.
*/
size_t mmu_total_memory(void) {
return total_memory;
}
/**
* @brief Return the amount of used memory.
*
* Calculates the number of pages currently marked as allocated.
* Multiplies it by 4 because pages are 4KiB.
*
* @returns the amount of memory in use in KiB.
*/
size_t mmu_used_memory(void) {
size_t ret = 0;
size_t i, j;
for (i = 0; i < INDEX_FROM_BIT(nframes); ++i) {
for (j = 0; j < 32; ++j) {
uint32_t testFrame = (uint32_t)0x1 << j;
if (frames[i] & testFrame) {
ret++;
}
}
}
return ret * 4 - unavailable_memory;
}
/**
* @brief Relinquish pages owned by a top-level directory.
*
* Frees the underlying pages for a page directory within the lower (user) region.
* Does not free kernel pages, as those are generally shared in the lower region.
*
* @param from Virtual pointer to top-level directory.
*/
void mmu_free(union PML * from) {
if (!from) {
printf("can't clear NULL directory\n");
return;
}
spin_lock(frame_alloc_lock);
for (size_t i = 0; i < 256; ++i) {
if (from[i].bits.present) {
union PML * pdp_in = mmu_map_from_physical((uintptr_t)from[i].bits.page << PAGE_SHIFT);
for (size_t j = 0; j < 512; ++j) {
if (pdp_in[j].bits.present) {
union PML * pd_in = mmu_map_from_physical((uintptr_t)pdp_in[j].bits.page << PAGE_SHIFT);
for (size_t k = 0; k < 512; ++k) {
if (pd_in[k].bits.present) {
union PML * pt_in = mmu_map_from_physical((uintptr_t)pd_in[k].bits.page << PAGE_SHIFT);
for (size_t l = 0; l < 512; ++l) {
uintptr_t address = ((i << (9 * 3 + 12)) | (j << (9*2 + 12)) | (k << (9 + 12)) | (l << PAGE_SHIFT));
/* Do not free shared mappings; SHM subsystem does that for SHM, devices don't need it. */
if (address >= USER_DEVICE_MAP && address <= USER_SHM_HIGH) continue;
if (pt_in[l].bits.present) {
/* Free only user pages */
if (pt_in[l].bits.user) {
free_page_maybe(pt_in,l,address);
}
}
}
mmu_frame_clear((uintptr_t)pd_in[k].bits.page << PAGE_SHIFT);
}
}
mmu_frame_clear((uintptr_t)pdp_in[j].bits.page << PAGE_SHIFT);
}
}
mmu_frame_clear((uintptr_t)from[i].bits.page << PAGE_SHIFT);
}
}
mmu_frame_clear((((uintptr_t)from) & PHYS_MASK));
spin_unlock(frame_alloc_lock);
}
union PML * mmu_get_kernel_directory(void) {
return mmu_map_from_physical((uintptr_t)&init_page_region[0]);
}
/**
* @brief Switch the active page directory for this core.
*
* Generally called during task creation and switching to change
* the active page directory of a core. Updates @c this_core->current_pml.
*
* x86-64: Loads a given PML into CR3.
*
* @param new_pml Either the physical address or the shadow mapping virtual address
* of the new PML4 directory to switch into, general obtained from
* a process struct; if NULL is passed, the initial kernel directory
* will be used and no userspace mappings will be present.
*/
void mmu_set_directory(union PML * new_pml) {
if (!new_pml) new_pml = mmu_map_from_physical((uintptr_t)&init_page_region[0]);
this_core->current_pml = new_pml;
asm volatile (
"movq %0, %%cr3"
: : "r"((uintptr_t)new_pml & PHYS_MASK));
}
/**
* @brief Mark a virtual address's mappings as invalid in the TLB.
*
* Generally should be called when a mapping is relinquished, as this is what
* the TLB caches, but is also called in a bunch of places where we're just mapping
* new pages...
*
* @param addr Virtual address in the current address space to invalidate.
*/
void mmu_invalidate(uintptr_t addr) {
asm volatile (
"invlpg (%0)"
: : "r"(addr));
arch_tlb_shootdown(addr);
}
static char * heapStart = NULL;
extern char end[];
/**
* @brief Prepare virtual page mappings for use by the kernel.
*
* Called during early boot to switch from the loader/bootstrap mappings
* to ones suitable for general use. Sets up the bitmap allocator, high
* identity mapping, kernel heap, and various mid-level structures to
* ensure that future kernelspace mappings apply to all kernel threads.
*
* @param memsize The maximum accessible physical address.
* @param firstFreePage The address of the first frame the kernel may use for new allocations.
*/
void mmu_init(size_t memsize, uintptr_t firstFreePage) {
this_core->current_pml = (union PML *)&init_page_region[0];
/**
* Enable WP bit, which will cause kernel writes to
* non-writable pages to trigger page faults. We use
* this to perform COW mappings for user processes if
* they passed an unmapped region to a system call, though
* this should be handled by @see mmu_validate_user_pointer
* before we get to that point...
*/
asm volatile (
"movq %%cr0, %%rax\n"
"orq $0x10000, %%rax\n"
"movq %%rax, %%cr0\n"
: : : "rax");
/* Map the high base PDP */
init_page_region[0][511].raw = (uintptr_t)&high_base_pml | KERNEL_PML_ACCESS;
init_page_region[0][510].raw = (uintptr_t)&heap_base_pml | KERNEL_PML_ACCESS;
/* Identity map from -128GB in the boot PML using 2MiB pages */
for (size_t i = 0; i < 64; ++i) {
high_base_pml[i].raw = (uintptr_t)&twom_high_pds[i] | KERNEL_PML_ACCESS;
for (uintptr_t j = 0; j < 512; ++j) {
twom_high_pds[i][j].raw = ((i << 30) + (j << 21)) | LARGE_PAGE_BIT | KERNEL_PML_ACCESS;
}
}
/* Map low base PDP */
low_base_pmls[0][0].raw = (uintptr_t)&low_base_pmls[1] | USER_PML_ACCESS;
/* How much memory do we need to map low for our *kernel* to fit? */
uintptr_t endPtr = ((uintptr_t)&end + PAGE_LOW_MASK) & PAGE_SIZE_MASK;
/* How many pages does that need? */
size_t lowPages = endPtr >> PAGE_SHIFT;
/* And how many 512-page blocks does that fit in? */
size_t pdCount = (lowPages + ENTRY_MASK) >> 9;
for (size_t j = 0; j < pdCount; ++j) {
low_base_pmls[1][j].raw = (uintptr_t)&low_base_pmls[2+j] | KERNEL_PML_ACCESS;
for (int i = 0; i < 512; ++i) {
low_base_pmls[2+j][i].raw = (uintptr_t)(LARGE_PAGE_SIZE * j + PAGE_SIZE * i) | KERNEL_PML_ACCESS;
}
}
/* Unmap null */
low_base_pmls[2][0].raw = 0;
/* Now map our new low base */
init_page_region[0][0].raw = (uintptr_t)&low_base_pmls[0] | USER_PML_ACCESS;
/* Set up the page allocator bitmap... */
nframes = (memsize >> 12);
size_t bytesOfFrames = INDEX_FROM_BIT(nframes * 8);
bytesOfFrames = (bytesOfFrames + PAGE_LOW_MASK) & PAGE_SIZE_MASK;
firstFreePage = (firstFreePage + PAGE_LOW_MASK) & PAGE_SIZE_MASK;
size_t pagesOfFrames = bytesOfFrames >> 12;
/* Set up heap map for that... */
heap_base_pml[0].raw = (uintptr_t)&heap_base_pd | KERNEL_PML_ACCESS;
heap_base_pd[0].raw = (uintptr_t)&heap_base_pt[0] | KERNEL_PML_ACCESS;
heap_base_pd[1].raw = (uintptr_t)&heap_base_pt[512] | KERNEL_PML_ACCESS;
heap_base_pd[2].raw = (uintptr_t)&heap_base_pt[1024] | KERNEL_PML_ACCESS;
if (pagesOfFrames > 512*3) {
printf("Warning: Too much available memory for current setup. Need %zu pages to represent allocation bitmap.\n", pagesOfFrames);
}
for (size_t i = 0; i < pagesOfFrames; i++) {
heap_base_pt[i].raw = (firstFreePage + (i << 12)) | KERNEL_PML_ACCESS;
}
asm volatile ("" : : : "memory");
this_core->current_pml = mmu_map_from_physical((uintptr_t)this_core->current_pml);
asm volatile ("" : : : "memory");
/* We are now in the new stuff. */
frames = (void*)((uintptr_t)KERNEL_HEAP_START);
memset((void*)frames, 0xFF, bytesOfFrames);
extern void mboot_unmark_valid_memory(void);
mboot_unmark_valid_memory();
/* Don't trust anything but our own bitmap... */
size_t unavail = 0, avail = 0;
for (size_t i = 0; i < INDEX_FROM_BIT(nframes); ++i) {
for (size_t j = 0; j < 32; ++j) {
uint32_t testFrame = (uint32_t)0x1 << j;
if (frames[i] & testFrame) {
unavail++;
} else {
avail++;
}
}
}
total_memory = avail * 4;
unavailable_memory = unavail * 4;
/* Now mark everything up to (firstFreePage + bytesOfFrames) as in use */
for (uintptr_t i = 0; i < firstFreePage + bytesOfFrames; i += PAGE_SIZE) {
mmu_frame_set(i);
}
heapStart = (char*)KERNEL_HEAP_START + bytesOfFrames;
/* Then, uh, make a bunch of space for page counts? */
size_t size_of_refcounts = (nframes & PAGE_LOW_MASK) ? (nframes + PAGE_SIZE - (nframes & PAGE_LOW_MASK)) : nframes;
mem_refcounts = sbrk(size_of_refcounts);
memset(mem_refcounts, 0, size_of_refcounts);
}
/**
* @brief Allocate space in the kernel virtual heap.
*
* Called by the kernel heap allocator to obtain space for new heap allocations.
*
* @warning Not to be confused with sys_sbrk
*
* @param bytes Bytes to allocate. Must be a multiple of PAGE_SIZE.
* @returns The previous address of the break point, after which @p bytes may now be used.
*/
void * sbrk(size_t bytes) {
if (!heapStart) {
arch_fatal_prepare();
printf("sbrk: Called before heap was ready.\n");
arch_dump_traceback();
arch_fatal();
}
if (!bytes) {
/* Skip lock acquisition if we just wanted to know where the break was. */
return heapStart;
}
if (bytes & PAGE_LOW_MASK) {
arch_fatal_prepare();
printf("sbrk: Size must be multiple of 4096, was %#zx\n", bytes);
arch_dump_traceback();
arch_fatal();
}
if (bytes > 0x1F00000) {
arch_fatal_prepare();
printf("sbrk: Size must be within a reasonable bound, was %#zx\n", bytes);
arch_dump_traceback();
arch_fatal();
}
spin_lock(kheap_lock);
void * out = heapStart;
for (uintptr_t p = (uintptr_t)out; p < (uintptr_t)out + bytes; p += PAGE_SIZE) {
union PML * page = mmu_get_page(p, MMU_GET_MAKE);
mmu_frame_allocate(page, MMU_FLAG_WRITABLE | MMU_FLAG_KERNEL);
}
//memset(out, 0xAA, bytes);
heapStart += bytes;
spin_unlock(kheap_lock);
return out;
}
static uintptr_t mmio_base_address = MMIO_BASE_START;
/**
* @brief Obtain a writethrough region mapped to the given physical address.
*
* For use by device drivers to obtain mappings suitable for MMIO accesses. Note that the
* virtual address space for these mappings can not be reclaimed, so drivers should keep
* them around or use the other MMU facilities to repurpose them.
*
* @param physical_address Physical memory offset of the destination MMIO space.
* @param size Size of the requested space, which must be a multiple of PAGE_SIZE.
* @returns a virtual address suitable for MMIO accesses.
*/
void * mmu_map_mmio_region(uintptr_t physical_address, size_t size) {
if (size & PAGE_LOW_MASK) {
arch_fatal_prepare();
printf("mmu_map_mmio_region: MMIO region size must be multiple of 4096 bytes, was %#zx.\n", size);
arch_dump_traceback();
arch_fatal();
}
spin_lock(mmio_space_lock);
void * out = (void*)mmio_base_address;
for (size_t i = 0; i < size; i += PAGE_SIZE) {
union PML * p = mmu_get_page(mmio_base_address + i, MMU_GET_MAKE);
mmu_frame_map_address(p, MMU_FLAG_KERNEL | MMU_FLAG_WRITABLE | MMU_FLAG_NOCACHE | MMU_FLAG_WRITETHROUGH, physical_address + i);
}
mmio_base_address += size;
spin_unlock(mmio_space_lock);
return out;
}
static uintptr_t module_base_address = MODULE_BASE_START;
/**
* @brief Obtain space to load a module in the -2GiB region.
*
* This should really start immediately after the kernel, but we don't
* yet load the kernel in the -2GiB region... it might also be worthwhile
* to implement some ASLR here, especially given that we're loading
* relocatable ELF object files and can stick them anywhere.
*
* @param size How much space to allocate, will be rounded up to page size.
* @returns Start of the allocated address space.
*/
void * mmu_map_module(size_t size) {
if (size & PAGE_LOW_MASK) {
size += (PAGE_LOW_MASK + 1) - (size & PAGE_LOW_MASK);
}
spin_lock(module_space_lock);
void * out = (void*)module_base_address;
for (size_t i = 0; i < size; i += PAGE_SIZE) {
union PML * p = mmu_get_page(module_base_address + i, MMU_GET_MAKE);
mmu_frame_allocate(p, MMU_FLAG_KERNEL | MMU_FLAG_WRITABLE);
}
module_base_address += size;
spin_unlock(module_space_lock);
return out;
}
/**
* @brief Free pages allocated for kernel modules.
*
* This rather blindly unmaps pages.
*
* @param start_address Start of mapping to unmap.
* @param size Size of mapping to unmap.
*/
void mmu_unmap_module(uintptr_t start_address, size_t size) {
if ((size & PAGE_LOW_MASK) || (start_address & PAGE_LOW_MASK)) {
arch_fatal_prepare();
printf("mmu_unmap_module start and size must be multiple of page size %#zx:%#zx.\n", start_address, size);
arch_dump_traceback();
arch_fatal();
}
spin_lock(module_space_lock);
uintptr_t end_address = start_address + size;
/* Unmap all pages we just allocated */
for (uintptr_t i = start_address; i < end_address; i += 0x1000) {
union PML * p = mmu_get_page(i, 0);
mmu_frame_clear(p->bits.page << 12);
}
/* Reset module base address if it was at the end, to avoid wasting address space */
if (end_address == module_base_address) {
module_base_address = start_address;
}
spin_unlock(module_space_lock);
}
/**
* @brief Swap a COW page for a writable copy.
*
* Examines @p address to determine if it is a pending
* COW page that has been marked read-only. If it is,
* it will be exchanged for a writable page. If it is
* the last read-only reference to a page, it will be
* marked writable without introducing a new backing page.
*
* @param address Virtual address that triggered the fault.
* @returns 0 if this was a valid and completed COW operation, 1 otherwise.
*/
int mmu_copy_on_write(uintptr_t address) {
union PML * page = mmu_get_page(address,0);
/* Was this address pending a cow? */
if (!page->bits.cow_pending) {
/* No, go back and trigger and a SIGSEGV */
return 1;
}
spin_lock(frame_alloc_lock);
/* Is this the last reference to this page? */
uint8_t refs = refcount_dec(page->bits.page);
if (refs == 0) {
/* Then we can just mark it writable. */
page->bits.writable = 1;
page->bits.cow_pending = 0;
asm ("" ::: "memory");
mmu_invalidate(address);
spin_unlock(frame_alloc_lock);
return 0;
}
/* Allocate a new writable page */
uintptr_t faulting_frame = page->bits.page;
uintptr_t fresh_frame = mmu_first_frame();
mmu_frame_set(fresh_frame << PAGE_SHIFT);
/* Copy the read-only page into the new writable page */
char * page_in = mmu_map_from_physical(faulting_frame << PAGE_SHIFT);
char * page_out = mmu_map_from_physical(fresh_frame << PAGE_SHIFT);
memcpy(page_out, page_in, 4096);
/* And swap out the page table entry. */
page->bits.page = fresh_frame;
page->bits.writable = 1;
page->bits.cow_pending = 0;
spin_unlock(frame_alloc_lock);
asm ("" ::: "memory");
mmu_invalidate(address);
return 0;
}
/**
* @brief Check if the current user process can access address space.
*
* Thoroughly examines page table entries to determine if a user process
* can access the memory at @p addr through @p size bytes.
*
* @p flags can be set to @c MMU_PTR_NULL if @c NULL address should trigger
* a failure, @c MMU_PTR_WRITE if the process must have write access.
*
* @param addr Address to start checking from.
* @param size Size after @p addr to check.
* @param flags Control what constitutes a failure.
* @returns 0 on failure, 1 if process has access.
*/
int mmu_validate_user_pointer(void * addr, size_t size, int flags) {
if (addr == NULL && !(flags & MMU_PTR_NULL)) return 0;
if (size > 0x800000000000) return 0;
uintptr_t base = (uintptr_t)addr;
uintptr_t end = size ? (base + (size - 1)) : base;
/* Get start page, end page */
uintptr_t page_base = base >> 12;
uintptr_t page_end = end >> 12;
for (uintptr_t page = page_base; page <= page_end; ++page) {
if ((page & 0xffff800000000) != 0 && (page & 0xffff800000000) != 0xffff800000000) return 0;
union PML * page_entry = mmu_get_page_other(this_core->current_process->thread.page_directory->directory, page << 12);
if (!page_entry) return 0;
if (!page_entry->bits.present) return 0;
if (!page_entry->bits.user) return 0;
if (!page_entry->bits.writable && (flags & MMU_PTR_WRITE)) {
if (mmu_copy_on_write((uintptr_t)(page << 12))) return 0;
}
}
return 1;
}
Reference in New Issue View Git Blame Copy Permalink