/*
* These routines also need to handle stuff like marking pages dirty
* and/or accessed for architectures that don't do it in hardware (most
* RISC architectures). The early dirtying is also good on the i386.
*
* There is also a hook called "update_mmu_cache()" that architectures
* with external mmu caches can use to update those (ie the Sparc or
* PowerPC hashed page tables that act as extended TLBs).
*
* Note the "page_table_lock". It is to protect against kswapd removing
* pages from under us. Note that kswapd only ever _removes_ pages, never
* adds them. As such, once we have noticed that the page is not present,
* we can drop the lock early.
*
* The adding of pages is protected by the MM semaphore (which we hold),
* so we don't need to worry about a page being suddenly been added into
* our VM.
*/
static inline int handle_pte_fault(struct mm_struct *mm,
struct vm_area_struct * vma, unsigned long address,
int write_access, pte_t * pte)
{
pte_t entry;
/*
* We need the page table lock to synchronize with kswapd
* and the SMP-safe atomic PTE updates.
*/
spin_lock(&mm->page_table_lock);
entry = *pte;
if (!pte_present(entry)) {
/*
* If it truly wasn't present, we know that kswapd
* and the PTE updates will not touch it later. So
* drop the lock.
*/
spin_unlock(&mm->page_table_lock);
if (pte_none(entry))
return do_no_page(mm, vma, address, write_access, pte);
return do_swap_page(mm, vma, address, pte, pte_to_swp_entry(entry), write_access);
}
if (write_access) {
if (!pte_write(entry))
return do_wp_page(mm, vma, address, pte, entry);
entry = pte_mkdirty(entry);
}
entry = pte_mkyoung(entry);
establish_pte(vma, address, pte, entry);
spin_unlock(&mm->page_table_lock);
return 1;
}
/*
* This routine puts a long into any process space by following the page
* tables. NOTE! You should check that the long isn't on a page boundary,
* and that it is in the task area before calling this: this routine does
* no checking.
*
* Now keeps R/W state of page so that a text page stays readonly
* even if a debugger scribbles breakpoints into it. -M.U-
*/
static void put_long(struct task_struct * tsk, struct vm_area_struct * vma, unsigned long addr,
unsigned long data)
{
pgd_t *pgdir;
pmd_t *pgmiddle;
pte_t *pgtable;
unsigned long page;
repeat:
pgdir = pgd_offset(vma->vm_mm, addr);
if (!pgd_present(*pgdir)) {
do_no_page(tsk, vma, addr, 1);
goto repeat;
}
if (pgd_bad(*pgdir)) {
printk("ptrace: bad page directory %08lx\n", pgd_val(*pgdir));
pgd_clear(pgdir);
return;
}
pgmiddle = pmd_offset(pgdir,addr);
if (pmd_none(*pgmiddle)) {
do_no_page(tsk, vma, addr, 1);
goto repeat;
}
if (pmd_bad(*pgmiddle)) {
printk("ptrace: bad page directory %08lx\n",
pmd_val(*pgmiddle));
pmd_clear(pgmiddle);
return;
}
pgtable = pte_offset(pgmiddle, addr);
if (!pte_present(*pgtable)) {
do_no_page(tsk, vma, addr, 1);
goto repeat;
}
page = pte_page(*pgtable);
if (!pte_write(*pgtable)) {
do_wp_page(tsk, vma, addr, 2);
goto repeat;
}
/* this is a hack for non-kernel-mapped video buffers and similar */
if (page < high_memory) {
*(unsigned long *) (page + (addr & ~PAGE_MASK)) = data;
flush_page_to_ram (page);
}
/* we're bypassing pagetables, so we have to set the dirty bit ourselves */
/* this should also re-instate whatever read-only mode there was before */
*pgtable = pte_mkdirty(mk_pte(page, vma->vm_page_prot));
flush_tlb_all();
}
/*
* the function for the no-page and wp_page
*/
void handle_pte_fault(struct vm_area_struct *vma, unsigned long address, pte_t *pte,int write_access)
{
// no page present
if(!pte_present(*pte))
{
do_no_page(vma,address,write_access);
return;
}
*pte = pte_mkyoung(*pte);
if(!write_access)
{
*pte = pte_mkdirty(*pte);
return;
}
// write the shared page
do_wp_page(vma, address, write_access);
}
/*
* The above separate functions for the no-page and wp-page
* cases will go away (they mostly do the same thing anyway),
* and we'll instead use only a general "handle_mm_fault()".
*
* These routines also need to handle stuff like marking pages dirty
* and/or accessed for architectures that don't do it in hardware (most
* RISC architectures). The early dirtying is also good on the i386.
*
* There is also a hook called "update_mmu_cache()" that architectures
* with external mmu caches can use to update those (ie the Sparc or
* PowerPC hashed page tables that act as extended TLBs).
*/
static inline void handle_pte_fault(struct vm_area_struct * vma, unsigned long address,
int write_access, pte_t * pte)
{
if (!pte_present(*pte)) {
do_no_page(current, vma, address, write_access);
return;
}
set_pte(pte, pte_mkyoung(*pte));
flush_tlb_page(vma, address);
if (!write_access)
return;
if (pte_write(*pte)) {
set_pte(pte, pte_mkdirty(*pte));
flush_tlb_page(vma, address);
return;
}
do_wp_page(current, vma, address, write_access);
}
/*
* This routine puts a long into any process space by following the page
* tables. NOTE! You should check that the long isn't on a page boundary,
* and that it is in the task area before calling this: this routine does
* no checking.
*
* Now keeps R/W state of page so that a text page stays readonly
* even if a debugger scribbles breakpoints into it. -M.U-
*/
static void put_long(struct vm_area_struct * vma, unsigned long addr,
unsigned long data)
{
pgd_t *pgdir;
pte_t *pgtable;
unsigned long page;
repeat:
pgdir = PAGE_DIR_OFFSET(vma->vm_task, addr);
if (!pgd_present(*pgdir)) {
do_no_page(vma, addr, 1);
goto repeat;
}
if (pgd_bad(*pgdir)) {
printk("ptrace: bad page directory %08lx\n", pgd_val(*pgdir));
pgd_clear(pgdir);
return;
}
pgtable = (pte_t *) (PAGE_PTR(addr) + pgd_page(*pgdir));
if (!pte_present(*pgtable)) {
do_no_page(vma, addr, 1);
goto repeat;
}
page = pte_page(*pgtable);
if (!pte_write(*pgtable)) {
do_wp_page(vma, addr, 1);
goto repeat;
}
/* this is a hack for non-kernel-mapped video buffers and similar */
if (page < high_memory) {
page += addr & ~PAGE_MASK;
*(unsigned long *) page = data;
}
/* we're bypassing pagetables, so we have to set the dirty bit ourselves */
/* this should also re-instate whatever read-only mode there was before */
*pgtable = pte_mkdirty(mk_pte(page, vma->vm_page_prot));
invalidate();
}
/*
* This isn't really reliable by any means..
*/
int mem_mmap(struct inode * inode, struct file * file,
struct vm_area_struct * vma)
{
struct task_struct *tsk;
pgd_t *src_dir, *dest_dir;
pmd_t *src_middle, *dest_middle;
pte_t *src_table, *dest_table;
unsigned long stmp, dtmp, mapnr;
struct vm_area_struct *src_vma = NULL;
/* Get the source's task information */
tsk = get_task(inode->i_ino >> 16);
if (!tsk)
return -ESRCH;
/* Ensure that we have a valid source area. (Has to be mmap'ed and
have valid page information.) We can't map shared memory at the
moment because working out the vm_area_struct & nattach stuff isn't
worth it. */
src_vma = tsk->mm->mmap;
stmp = vma->vm_offset;
while (stmp < vma->vm_offset + (vma->vm_end - vma->vm_start)) {
while (src_vma && stmp > src_vma->vm_end)
src_vma = src_vma->vm_next;
if (!src_vma || (src_vma->vm_flags & VM_SHM))
return -EINVAL;
src_dir = pgd_offset(tsk->mm, stmp);
if (pgd_none(*src_dir))
return -EINVAL;
if (pgd_bad(*src_dir)) {
printk("Bad source page dir entry %08lx\n", pgd_val(*src_dir));
return -EINVAL;
}
src_middle = pmd_offset(src_dir, stmp);
if (pmd_none(*src_middle))
return -EINVAL;
if (pmd_bad(*src_middle)) {
printk("Bad source page middle entry %08lx\n", pmd_val(*src_middle));
return -EINVAL;
}
src_table = pte_offset(src_middle, stmp);
if (pte_none(*src_table))
return -EINVAL;
if (stmp < src_vma->vm_start) {
if (!(src_vma->vm_flags & VM_GROWSDOWN))
return -EINVAL;
if (src_vma->vm_end - stmp > current->rlim[RLIMIT_STACK].rlim_cur)
return -EINVAL;
}
stmp += PAGE_SIZE;
}
src_vma = tsk->mm->mmap;
stmp = vma->vm_offset;
dtmp = vma->vm_start;
flush_cache_range(vma->vm_mm, vma->vm_start, vma->vm_end);
flush_cache_range(src_vma->vm_mm, src_vma->vm_start, src_vma->vm_end);
while (dtmp < vma->vm_end) {
while (src_vma && stmp > src_vma->vm_end)
src_vma = src_vma->vm_next;
src_dir = pgd_offset(tsk->mm, stmp);
src_middle = pmd_offset(src_dir, stmp);
src_table = pte_offset(src_middle, stmp);
dest_dir = pgd_offset(current->mm, dtmp);
dest_middle = pmd_alloc(dest_dir, dtmp);
if (!dest_middle)
return -ENOMEM;
dest_table = pte_alloc(dest_middle, dtmp);
if (!dest_table)
return -ENOMEM;
if (!pte_present(*src_table))
do_no_page(tsk, src_vma, stmp, 1);
if ((vma->vm_flags & VM_WRITE) && !pte_write(*src_table))
do_wp_page(tsk, src_vma, stmp, 1);
set_pte(src_table, pte_mkdirty(*src_table));
set_pte(dest_table, *src_table);
mapnr = MAP_NR(pte_page(*src_table));
if (mapnr < MAP_NR(high_memory))
mem_map[mapnr].count++;
stmp += PAGE_SIZE;
dtmp += PAGE_SIZE;
}
flush_tlb_range(vma->vm_mm, vma->vm_start, vma->vm_end);
flush_tlb_range(src_vma->vm_mm, src_vma->vm_start, src_vma->vm_end);
return 0;
}
/*
* Ugly, ugly, but the goto's result in better assembly..
*/
int verify_area(int type, const void * addr, unsigned long size)
{
struct vm_area_struct * vma;
unsigned long start = (unsigned long) addr;
/* If the current user space is mapped to kernel space (for the
* case where we use a fake user buffer with get_fs/set_fs()) we
* don't expect to find the address in the user vm map.
*/
if (!size || get_fs() == KERNEL_DS)
return 0;
vma = find_vma(current->mm, start);
if (!vma)
goto bad_area;
if (vma->vm_start > start)
goto check_stack;
good_area:
if (type == VERIFY_WRITE)
goto check_write;
for (;;) {
struct vm_area_struct * next;
if (!(vma->vm_flags & VM_READ))
goto bad_area;
if (vma->vm_end - start >= size)
return 0;
next = vma->vm_next;
if (!next || vma->vm_end != next->vm_start)
goto bad_area;
vma = next;
}
check_write:
if (!(vma->vm_flags & VM_WRITE))
goto bad_area;
if (!wp_works_ok)
goto check_wp_fault_by_hand;
for (;;) {
if (vma->vm_end - start >= size)
break;
if (!vma->vm_next || vma->vm_end != vma->vm_next->vm_start)
goto bad_area;
vma = vma->vm_next;
if (!(vma->vm_flags & VM_WRITE))
goto bad_area;
}
return 0;
check_wp_fault_by_hand:
size--;
size += start & ~PAGE_MASK;
size >>= PAGE_SHIFT;
start &= PAGE_MASK;
for (;;) {
do_wp_page(current, vma, start, 1);
if (!size)
break;
size--;
start += PAGE_SIZE;
if (start < vma->vm_end)
continue;
vma = vma->vm_next;
if (!vma || vma->vm_start != start)
goto bad_area;
if (!(vma->vm_flags & VM_WRITE))
goto bad_area;;
}
return 0;
check_stack:
if (!(vma->vm_flags & VM_GROWSDOWN))
goto bad_area;
if (expand_stack(vma, start) == 0)
goto good_area;
bad_area:
return -EFAULT;
}