2 * sys_ia32.c: Conversion between 32bit and 64bit native syscalls. Derived from sys_sparc32.c.
4 * Copyright (C) 2000 VA Linux Co
5 * Copyright (C) 2000 Don Dugger <n0ano@valinux.com>
6 * Copyright (C) 1999 Arun Sharma <arun.sharma@intel.com>
7 * Copyright (C) 1997,1998 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
8 * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
9 * Copyright (C) 2000-2003, 2005 Hewlett-Packard Co
10 * David Mosberger-Tang <davidm@hpl.hp.com>
11 * Copyright (C) 2004 Gordon Jin <gordon.jin@intel.com>
13 * These routines maintain argument size conversion between 32bit and 64bit
17 #include <linux/kernel.h>
18 #include <linux/syscalls.h>
19 #include <linux/sysctl.h>
20 #include <linux/sched.h>
22 #include <linux/file.h>
23 #include <linux/signal.h>
24 #include <linux/resource.h>
25 #include <linux/times.h>
26 #include <linux/utsname.h>
27 #include <linux/smp.h>
28 #include <linux/smp_lock.h>
29 #include <linux/sem.h>
30 #include <linux/msg.h>
32 #include <linux/shm.h>
33 #include <linux/slab.h>
34 #include <linux/uio.h>
35 #include <linux/nfs_fs.h>
36 #include <linux/quota.h>
37 #include <linux/syscalls.h>
38 #include <linux/sunrpc/svc.h>
39 #include <linux/nfsd/nfsd.h>
40 #include <linux/nfsd/cache.h>
41 #include <linux/nfsd/xdr.h>
42 #include <linux/nfsd/syscall.h>
43 #include <linux/poll.h>
44 #include <linux/eventpoll.h>
45 #include <linux/personality.h>
46 #include <linux/ptrace.h>
47 #include <linux/stat.h>
48 #include <linux/ipc.h>
49 #include <linux/capability.h>
50 #include <linux/compat.h>
51 #include <linux/vfs.h>
52 #include <linux/mman.h>
53 #include <linux/mutex.h>
55 #include <asm/intrinsics.h>
56 #include <asm/types.h>
57 #include <asm/uaccess.h>
58 #include <asm/unistd.h>
68 # define DBG(fmt...) printk(KERN_DEBUG fmt)
73 #define ROUND_UP(x,a) ((__typeof__(x))(((unsigned long)(x) + ((a) - 1)) & ~((a) - 1)))
75 #define OFFSET4K(a) ((a) & 0xfff)
76 #define PAGE_START(addr) ((addr) & PAGE_MASK)
77 #define MINSIGSTKSZ_IA32 2048
79 #define high2lowuid(uid) ((uid) > 65535 ? 65534 : (uid))
80 #define high2lowgid(gid) ((gid) > 65535 ? 65534 : (gid))
83 * Anything that modifies or inspects ia32 user virtual memory must hold this semaphore
86 /* XXX make per-mm: */
87 static DEFINE_MUTEX(ia32_mmap_mutex);
90 sys32_execve (char __user *name, compat_uptr_t __user *argv, compat_uptr_t __user *envp,
95 unsigned long old_map_base, old_task_size, tssd;
97 filename = getname(name);
98 error = PTR_ERR(filename);
102 old_map_base = current->thread.map_base;
103 old_task_size = current->thread.task_size;
104 tssd = ia64_get_kr(IA64_KR_TSSD);
106 /* we may be exec'ing a 64-bit process: reset map base, task-size, and io-base: */
107 current->thread.map_base = DEFAULT_MAP_BASE;
108 current->thread.task_size = DEFAULT_TASK_SIZE;
109 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
110 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
112 error = compat_do_execve(filename, argv, envp, regs);
116 /* oops, execve failed, switch back to old values... */
117 ia64_set_kr(IA64_KR_IO_BASE, IA32_IOBASE);
118 ia64_set_kr(IA64_KR_TSSD, tssd);
119 current->thread.map_base = old_map_base;
120 current->thread.task_size = old_task_size;
126 int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
130 if ((u64) stat->size > MAX_NON_LFS ||
131 !old_valid_dev(stat->dev) ||
132 !old_valid_dev(stat->rdev))
135 if (clear_user(ubuf, sizeof(*ubuf)))
138 err = __put_user(old_encode_dev(stat->dev), &ubuf->st_dev);
139 err |= __put_user(stat->ino, &ubuf->st_ino);
140 err |= __put_user(stat->mode, &ubuf->st_mode);
141 err |= __put_user(stat->nlink, &ubuf->st_nlink);
142 err |= __put_user(high2lowuid(stat->uid), &ubuf->st_uid);
143 err |= __put_user(high2lowgid(stat->gid), &ubuf->st_gid);
144 err |= __put_user(old_encode_dev(stat->rdev), &ubuf->st_rdev);
145 err |= __put_user(stat->size, &ubuf->st_size);
146 err |= __put_user(stat->atime.tv_sec, &ubuf->st_atime);
147 err |= __put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec);
148 err |= __put_user(stat->mtime.tv_sec, &ubuf->st_mtime);
149 err |= __put_user(stat->mtime.tv_nsec, &ubuf->st_mtime_nsec);
150 err |= __put_user(stat->ctime.tv_sec, &ubuf->st_ctime);
151 err |= __put_user(stat->ctime.tv_nsec, &ubuf->st_ctime_nsec);
152 err |= __put_user(stat->blksize, &ubuf->st_blksize);
153 err |= __put_user(stat->blocks, &ubuf->st_blocks);
157 #if PAGE_SHIFT > IA32_PAGE_SHIFT
161 get_page_prot (struct vm_area_struct *vma, unsigned long addr)
165 if (!vma || vma->vm_start > addr)
168 if (vma->vm_flags & VM_READ)
170 if (vma->vm_flags & VM_WRITE)
172 if (vma->vm_flags & VM_EXEC)
178 * Map a subpage by creating an anonymous page that contains the union of the old page and
182 mmap_subpage (struct file *file, unsigned long start, unsigned long end, int prot, int flags,
187 unsigned long ret = 0;
188 struct vm_area_struct *vma = find_vma(current->mm, start);
189 int old_prot = get_page_prot(vma, start);
191 DBG("mmap_subpage(file=%p,start=0x%lx,end=0x%lx,prot=%x,flags=%x,off=0x%llx)\n",
192 file, start, end, prot, flags, off);
195 /* Optimize the case where the old mmap and the new mmap are both anonymous */
196 if ((old_prot & PROT_WRITE) && (flags & MAP_ANONYMOUS) && !vma->vm_file) {
197 if (clear_user((void __user *) start, end - start)) {
204 page = (void *) get_zeroed_page(GFP_KERNEL);
209 copy_from_user(page, (void __user *) PAGE_START(start), PAGE_SIZE);
211 down_write(¤t->mm->mmap_sem);
213 ret = do_mmap(NULL, PAGE_START(start), PAGE_SIZE, prot | PROT_WRITE,
214 flags | MAP_FIXED | MAP_ANONYMOUS, 0);
216 up_write(¤t->mm->mmap_sem);
218 if (IS_ERR((void *) ret))
222 /* copy back the old page contents. */
223 if (offset_in_page(start))
224 copy_to_user((void __user *) PAGE_START(start), page,
225 offset_in_page(start));
226 if (offset_in_page(end))
227 copy_to_user((void __user *) end, page + offset_in_page(end),
228 PAGE_SIZE - offset_in_page(end));
231 if (!(flags & MAP_ANONYMOUS)) {
232 /* read the file contents */
233 inode = file->f_dentry->d_inode;
234 if (!inode->i_fop || !file->f_op->read
235 || ((*file->f_op->read)(file, (char __user *) start, end - start, &off) < 0))
243 if (!(prot & PROT_WRITE))
244 ret = sys_mprotect(PAGE_START(start), PAGE_SIZE, prot | old_prot);
247 free_page((unsigned long) page);
251 /* SLAB cache for partial_page structures */
252 kmem_cache_t *partial_page_cachep;
255 * init partial_page_list.
256 * return 0 means kmalloc fail.
258 struct partial_page_list*
259 ia32_init_pp_list(void)
261 struct partial_page_list *p;
263 if ((p = kmalloc(sizeof(*p), GFP_KERNEL)) == NULL)
268 atomic_set(&p->pp_count, 1);
273 * Search for the partial page with @start in partial page list @ppl.
274 * If finds the partial page, return the found partial page.
275 * Else, return 0 and provide @pprev, @rb_link, @rb_parent to
276 * be used by later __ia32_insert_pp().
278 static struct partial_page *
279 __ia32_find_pp(struct partial_page_list *ppl, unsigned int start,
280 struct partial_page **pprev, struct rb_node ***rb_link,
281 struct rb_node **rb_parent)
283 struct partial_page *pp;
284 struct rb_node **__rb_link, *__rb_parent, *rb_prev;
287 if (pp && pp->base == start)
290 __rb_link = &ppl->ppl_rb.rb_node;
291 rb_prev = __rb_parent = NULL;
294 __rb_parent = *__rb_link;
295 pp = rb_entry(__rb_parent, struct partial_page, pp_rb);
297 if (pp->base == start) {
300 } else if (pp->base < start) {
301 rb_prev = __rb_parent;
302 __rb_link = &__rb_parent->rb_right;
304 __rb_link = &__rb_parent->rb_left;
308 *rb_link = __rb_link;
309 *rb_parent = __rb_parent;
312 *pprev = rb_entry(rb_prev, struct partial_page, pp_rb);
317 * insert @pp into @ppl.
320 __ia32_insert_pp(struct partial_page_list *ppl, struct partial_page *pp,
321 struct partial_page *prev, struct rb_node **rb_link,
322 struct rb_node *rb_parent)
326 pp->next = prev->next;
331 pp->next = rb_entry(rb_parent,
332 struct partial_page, pp_rb);
338 rb_link_node(&pp->pp_rb, rb_parent, rb_link);
339 rb_insert_color(&pp->pp_rb, &ppl->ppl_rb);
345 * delete @pp from partial page list @ppl.
348 __ia32_delete_pp(struct partial_page_list *ppl, struct partial_page *pp,
349 struct partial_page *prev)
352 prev->next = pp->next;
353 if (ppl->pp_hint == pp)
356 ppl->pp_head = pp->next;
357 if (ppl->pp_hint == pp)
358 ppl->pp_hint = pp->next;
360 rb_erase(&pp->pp_rb, &ppl->ppl_rb);
361 kmem_cache_free(partial_page_cachep, pp);
364 static struct partial_page *
365 __pp_prev(struct partial_page *pp)
367 struct rb_node *prev = rb_prev(&pp->pp_rb);
369 return rb_entry(prev, struct partial_page, pp_rb);
375 * Delete partial pages with address between @start and @end.
376 * @start and @end are page aligned.
379 __ia32_delete_pp_range(unsigned int start, unsigned int end)
381 struct partial_page *pp, *prev;
382 struct rb_node **rb_link, *rb_parent;
387 pp = __ia32_find_pp(current->thread.ppl, start, &prev,
388 &rb_link, &rb_parent);
390 prev = __pp_prev(pp);
395 pp = current->thread.ppl->pp_head;
398 while (pp && pp->base < end) {
399 struct partial_page *tmp = pp->next;
400 __ia32_delete_pp(current->thread.ppl, pp, prev);
406 * Set the range between @start and @end in bitmap.
407 * @start and @end should be IA32 page aligned and in the same IA64 page.
410 __ia32_set_pp(unsigned int start, unsigned int end, int flags)
412 struct partial_page *pp, *prev;
413 struct rb_node ** rb_link, *rb_parent;
414 unsigned int pstart, start_bit, end_bit, i;
416 pstart = PAGE_START(start);
417 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
418 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
420 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
421 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
422 &rb_link, &rb_parent);
424 for (i = start_bit; i < end_bit; i++)
425 set_bit(i, &pp->bitmap);
427 * Check: if this partial page has been set to a full page,
430 if (find_first_zero_bit(&pp->bitmap, sizeof(pp->bitmap)*8) >=
431 PAGE_SIZE/IA32_PAGE_SIZE) {
432 __ia32_delete_pp(current->thread.ppl, pp, __pp_prev(pp));
438 * MAP_FIXED may lead to overlapping mmap.
439 * In this case, the requested mmap area may already mmaped as a full
440 * page. So check vma before adding a new partial page.
442 if (flags & MAP_FIXED) {
443 struct vm_area_struct *vma = find_vma(current->mm, pstart);
444 if (vma && vma->vm_start <= pstart)
448 /* new a partial_page */
449 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
454 for (i=start_bit; i<end_bit; i++)
455 set_bit(i, &(pp->bitmap));
457 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
462 * @start and @end should be IA32 page aligned, but don't need to be in the
463 * same IA64 page. Split @start and @end to make sure they're in the same IA64
464 * page, then call __ia32_set_pp().
467 ia32_set_pp(unsigned int start, unsigned int end, int flags)
469 down_write(¤t->mm->mmap_sem);
470 if (flags & MAP_FIXED) {
472 * MAP_FIXED may lead to overlapping mmap. When this happens,
473 * a series of complete IA64 pages results in deletion of
474 * old partial pages in that range.
476 __ia32_delete_pp_range(PAGE_ALIGN(start), PAGE_START(end));
479 if (end < PAGE_ALIGN(start)) {
480 __ia32_set_pp(start, end, flags);
482 if (offset_in_page(start))
483 __ia32_set_pp(start, PAGE_ALIGN(start), flags);
484 if (offset_in_page(end))
485 __ia32_set_pp(PAGE_START(end), end, flags);
487 up_write(¤t->mm->mmap_sem);
491 * Unset the range between @start and @end in bitmap.
492 * @start and @end should be IA32 page aligned and in the same IA64 page.
493 * After doing that, if the bitmap is 0, then free the page and return 1,
495 * If not find the partial page in the list, then
496 * If the vma exists, then the full page is set to a partial page;
497 * Else return -ENOMEM.
500 __ia32_unset_pp(unsigned int start, unsigned int end)
502 struct partial_page *pp, *prev;
503 struct rb_node ** rb_link, *rb_parent;
504 unsigned int pstart, start_bit, end_bit, i;
505 struct vm_area_struct *vma;
507 pstart = PAGE_START(start);
508 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
509 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
511 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
513 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
514 &rb_link, &rb_parent);
516 for (i = start_bit; i < end_bit; i++)
517 clear_bit(i, &pp->bitmap);
518 if (pp->bitmap == 0) {
519 __ia32_delete_pp(current->thread.ppl, pp, __pp_prev(pp));
525 vma = find_vma(current->mm, pstart);
526 if (!vma || vma->vm_start > pstart) {
530 /* new a partial_page */
531 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
536 for (i = 0; i < start_bit; i++)
537 set_bit(i, &(pp->bitmap));
538 for (i = end_bit; i < PAGE_SIZE / IA32_PAGE_SIZE; i++)
539 set_bit(i, &(pp->bitmap));
541 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
546 * Delete pp between PAGE_ALIGN(start) and PAGE_START(end) by calling
547 * __ia32_delete_pp_range(). Unset possible partial pages by calling
549 * The returned value see __ia32_unset_pp().
552 ia32_unset_pp(unsigned int *startp, unsigned int *endp)
554 unsigned int start = *startp, end = *endp;
557 down_write(¤t->mm->mmap_sem);
559 __ia32_delete_pp_range(PAGE_ALIGN(start), PAGE_START(end));
561 if (end < PAGE_ALIGN(start)) {
562 ret = __ia32_unset_pp(start, end);
564 *startp = PAGE_START(start);
565 *endp = PAGE_ALIGN(end);
568 /* to shortcut sys_munmap() in sys32_munmap() */
569 *startp = PAGE_START(start);
570 *endp = PAGE_START(end);
573 if (offset_in_page(start)) {
574 ret = __ia32_unset_pp(start, PAGE_ALIGN(start));
576 *startp = PAGE_START(start);
578 *startp = PAGE_ALIGN(start);
582 if (offset_in_page(end)) {
583 ret = __ia32_unset_pp(PAGE_START(end), end);
585 *endp = PAGE_ALIGN(end);
587 *endp = PAGE_START(end);
592 up_write(¤t->mm->mmap_sem);
597 * Compare the range between @start and @end with bitmap in partial page.
598 * @start and @end should be IA32 page aligned and in the same IA64 page.
601 __ia32_compare_pp(unsigned int start, unsigned int end)
603 struct partial_page *pp, *prev;
604 struct rb_node ** rb_link, *rb_parent;
605 unsigned int pstart, start_bit, end_bit, size;
606 unsigned int first_bit, next_zero_bit; /* the first range in bitmap */
608 pstart = PAGE_START(start);
610 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
611 &rb_link, &rb_parent);
615 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
616 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
617 size = sizeof(pp->bitmap) * 8;
618 first_bit = find_first_bit(&pp->bitmap, size);
619 next_zero_bit = find_next_zero_bit(&pp->bitmap, size, first_bit);
620 if ((start_bit < first_bit) || (end_bit > next_zero_bit)) {
621 /* exceeds the first range in bitmap */
623 } else if ((start_bit == first_bit) && (end_bit == next_zero_bit)) {
624 first_bit = find_next_bit(&pp->bitmap, size, next_zero_bit);
625 if ((next_zero_bit < first_bit) && (first_bit < size))
626 return 1; /* has next range */
628 return 0; /* no next range */
634 * @start and @end should be IA32 page aligned, but don't need to be in the
635 * same IA64 page. Split @start and @end to make sure they're in the same IA64
636 * page, then call __ia32_compare_pp().
638 * Take this as example: the range is the 1st and 2nd 4K page.
639 * Return 0 if they fit bitmap exactly, i.e. bitmap = 00000011;
640 * Return 1 if the range doesn't cover whole bitmap, e.g. bitmap = 00001111;
641 * Return -ENOMEM if the range exceeds the bitmap, e.g. bitmap = 00000001 or
645 ia32_compare_pp(unsigned int *startp, unsigned int *endp)
647 unsigned int start = *startp, end = *endp;
650 down_write(¤t->mm->mmap_sem);
652 if (end < PAGE_ALIGN(start)) {
653 retval = __ia32_compare_pp(start, end);
655 *startp = PAGE_START(start);
656 *endp = PAGE_ALIGN(end);
659 if (offset_in_page(start)) {
660 retval = __ia32_compare_pp(start,
663 *startp = PAGE_START(start);
667 if (offset_in_page(end)) {
668 retval = __ia32_compare_pp(PAGE_START(end), end);
670 *endp = PAGE_ALIGN(end);
675 up_write(¤t->mm->mmap_sem);
680 __ia32_drop_pp_list(struct partial_page_list *ppl)
682 struct partial_page *pp = ppl->pp_head;
685 struct partial_page *next = pp->next;
686 kmem_cache_free(partial_page_cachep, pp);
694 ia32_drop_partial_page_list(struct task_struct *task)
696 struct partial_page_list* ppl = task->thread.ppl;
698 if (ppl && atomic_dec_and_test(&ppl->pp_count))
699 __ia32_drop_pp_list(ppl);
703 * Copy current->thread.ppl to ppl (already initialized).
706 __ia32_copy_pp_list(struct partial_page_list *ppl)
708 struct partial_page *pp, *tmp, *prev;
709 struct rb_node **rb_link, *rb_parent;
713 ppl->ppl_rb = RB_ROOT;
714 rb_link = &ppl->ppl_rb.rb_node;
718 for (pp = current->thread.ppl->pp_head; pp; pp = pp->next) {
719 tmp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
723 __ia32_insert_pp(ppl, tmp, prev, rb_link, rb_parent);
725 rb_link = &tmp->pp_rb.rb_right;
726 rb_parent = &tmp->pp_rb;
732 ia32_copy_partial_page_list(struct task_struct *p, unsigned long clone_flags)
736 if (clone_flags & CLONE_VM) {
737 atomic_inc(¤t->thread.ppl->pp_count);
738 p->thread.ppl = current->thread.ppl;
740 p->thread.ppl = ia32_init_pp_list();
743 down_write(¤t->mm->mmap_sem);
745 retval = __ia32_copy_pp_list(p->thread.ppl);
747 up_write(¤t->mm->mmap_sem);
754 emulate_mmap (struct file *file, unsigned long start, unsigned long len, int prot, int flags,
757 unsigned long tmp, end, pend, pstart, ret, is_congruent, fudge = 0;
762 pstart = PAGE_START(start);
763 pend = PAGE_ALIGN(end);
765 if (flags & MAP_FIXED) {
766 ia32_set_pp((unsigned int)start, (unsigned int)end, flags);
767 if (start > pstart) {
768 if (flags & MAP_SHARED)
770 "%s(%d): emulate_mmap() can't share head (addr=0x%lx)\n",
771 current->comm, current->pid, start);
772 ret = mmap_subpage(file, start, min(PAGE_ALIGN(start), end), prot, flags,
774 if (IS_ERR((void *) ret))
781 if (flags & MAP_SHARED)
783 "%s(%d): emulate_mmap() can't share tail (end=0x%lx)\n",
784 current->comm, current->pid, end);
785 ret = mmap_subpage(file, max(start, PAGE_START(end)), end, prot, flags,
786 (off + len) - offset_in_page(end));
787 if (IS_ERR((void *) ret))
795 * If a start address was specified, use it if the entire rounded out area
798 if (start && !pstart)
799 fudge = 1; /* handle case of mapping to range (0,PAGE_SIZE) */
800 tmp = arch_get_unmapped_area(file, pstart - fudge, pend - pstart, 0, flags);
803 start = pstart + offset_in_page(off); /* make start congruent with off */
805 pend = PAGE_ALIGN(end);
809 poff = off + (pstart - start); /* note: (pstart - start) may be negative */
810 is_congruent = (flags & MAP_ANONYMOUS) || (offset_in_page(poff) == 0);
812 if ((flags & MAP_SHARED) && !is_congruent)
813 printk(KERN_INFO "%s(%d): emulate_mmap() can't share contents of incongruent mmap "
814 "(addr=0x%lx,off=0x%llx)\n", current->comm, current->pid, start, off);
816 DBG("mmap_body: mapping [0x%lx-0x%lx) %s with poff 0x%llx\n", pstart, pend,
817 is_congruent ? "congruent" : "not congruent", poff);
819 down_write(¤t->mm->mmap_sem);
821 if (!(flags & MAP_ANONYMOUS) && is_congruent)
822 ret = do_mmap(file, pstart, pend - pstart, prot, flags | MAP_FIXED, poff);
824 ret = do_mmap(NULL, pstart, pend - pstart,
825 prot | ((flags & MAP_ANONYMOUS) ? 0 : PROT_WRITE),
826 flags | MAP_FIXED | MAP_ANONYMOUS, 0);
828 up_write(¤t->mm->mmap_sem);
830 if (IS_ERR((void *) ret))
834 /* read the file contents */
835 inode = file->f_dentry->d_inode;
836 if (!inode->i_fop || !file->f_op->read
837 || ((*file->f_op->read)(file, (char __user *) pstart, pend - pstart, &poff)
840 sys_munmap(pstart, pend - pstart);
843 if (!(prot & PROT_WRITE) && sys_mprotect(pstart, pend - pstart, prot) < 0)
847 if (!(flags & MAP_FIXED))
848 ia32_set_pp((unsigned int)start, (unsigned int)end, flags);
853 #endif /* PAGE_SHIFT > IA32_PAGE_SHIFT */
855 static inline unsigned int
856 get_prot32 (unsigned int prot)
858 if (prot & PROT_WRITE)
859 /* on x86, PROT_WRITE implies PROT_READ which implies PROT_EEC */
860 prot |= PROT_READ | PROT_WRITE | PROT_EXEC;
861 else if (prot & (PROT_READ | PROT_EXEC))
862 /* on x86, there is no distinction between PROT_READ and PROT_EXEC */
863 prot |= (PROT_READ | PROT_EXEC);
869 ia32_do_mmap (struct file *file, unsigned long addr, unsigned long len, int prot, int flags,
872 DBG("ia32_do_mmap(file=%p,addr=0x%lx,len=0x%lx,prot=%x,flags=%x,offset=0x%llx)\n",
873 file, addr, len, prot, flags, offset);
875 if (file && (!file->f_op || !file->f_op->mmap))
878 len = IA32_PAGE_ALIGN(len);
882 if (len > IA32_PAGE_OFFSET || addr > IA32_PAGE_OFFSET - len)
884 if (flags & MAP_FIXED)
890 if (OFFSET4K(offset))
893 prot = get_prot32(prot);
895 #if PAGE_SHIFT > IA32_PAGE_SHIFT
896 mutex_lock(&ia32_mmap_mutex);
898 addr = emulate_mmap(file, addr, len, prot, flags, offset);
900 mutex_unlock(&ia32_mmap_mutex);
902 down_write(¤t->mm->mmap_sem);
904 addr = do_mmap(file, addr, len, prot, flags, offset);
906 up_write(¤t->mm->mmap_sem);
908 DBG("ia32_do_mmap: returning 0x%lx\n", addr);
913 * Linux/i386 didn't use to be able to handle more than 4 system call parameters, so these
914 * system calls used a memory block for parameter passing..
917 struct mmap_arg_struct {
927 sys32_mmap (struct mmap_arg_struct __user *arg)
929 struct mmap_arg_struct a;
930 struct file *file = NULL;
934 if (copy_from_user(&a, arg, sizeof(a)))
937 if (OFFSET4K(a.offset))
942 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
943 if (!(flags & MAP_ANONYMOUS)) {
949 addr = ia32_do_mmap(file, a.addr, a.len, a.prot, flags, a.offset);
957 sys32_mmap2 (unsigned int addr, unsigned int len, unsigned int prot, unsigned int flags,
958 unsigned int fd, unsigned int pgoff)
960 struct file *file = NULL;
961 unsigned long retval;
963 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
964 if (!(flags & MAP_ANONYMOUS)) {
970 retval = ia32_do_mmap(file, addr, len, prot, flags,
971 (unsigned long) pgoff << IA32_PAGE_SHIFT);
979 sys32_munmap (unsigned int start, unsigned int len)
981 unsigned int end = start + len;
984 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
985 ret = sys_munmap(start, end - start);
990 end = IA32_PAGE_ALIGN(end);
994 ret = ia32_unset_pp(&start, &end);
1001 mutex_lock(&ia32_mmap_mutex);
1002 ret = sys_munmap(start, end - start);
1003 mutex_unlock(&ia32_mmap_mutex);
1008 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1011 * When mprotect()ing a partial page, we set the permission to the union of the old
1012 * settings and the new settings. In other words, it's only possible to make access to a
1013 * partial page less restrictive.
1016 mprotect_subpage (unsigned long address, int new_prot)
1019 struct vm_area_struct *vma;
1021 if (new_prot == PROT_NONE)
1022 return 0; /* optimize case where nothing changes... */
1023 vma = find_vma(current->mm, address);
1024 old_prot = get_page_prot(vma, address);
1025 return sys_mprotect(address, PAGE_SIZE, new_prot | old_prot);
1028 #endif /* PAGE_SHIFT > IA32_PAGE_SHIFT */
1031 sys32_mprotect (unsigned int start, unsigned int len, int prot)
1033 unsigned int end = start + len;
1034 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1038 prot = get_prot32(prot);
1040 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1041 return sys_mprotect(start, end - start, prot);
1043 if (OFFSET4K(start))
1046 end = IA32_PAGE_ALIGN(end);
1050 retval = ia32_compare_pp(&start, &end);
1055 mutex_lock(&ia32_mmap_mutex);
1057 if (offset_in_page(start)) {
1058 /* start address is 4KB aligned but not page aligned. */
1059 retval = mprotect_subpage(PAGE_START(start), prot);
1063 start = PAGE_ALIGN(start);
1065 goto out; /* retval is already zero... */
1068 if (offset_in_page(end)) {
1069 /* end address is 4KB aligned but not page aligned. */
1070 retval = mprotect_subpage(PAGE_START(end), prot);
1074 end = PAGE_START(end);
1076 retval = sys_mprotect(start, end - start, prot);
1079 mutex_unlock(&ia32_mmap_mutex);
1085 sys32_mremap (unsigned int addr, unsigned int old_len, unsigned int new_len,
1086 unsigned int flags, unsigned int new_addr)
1090 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1091 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1093 unsigned int old_end, new_end;
1098 old_len = IA32_PAGE_ALIGN(old_len);
1099 new_len = IA32_PAGE_ALIGN(new_len);
1100 old_end = addr + old_len;
1101 new_end = addr + new_len;
1106 if ((flags & MREMAP_FIXED) && (OFFSET4K(new_addr)))
1109 if (old_len >= new_len) {
1110 ret = sys32_munmap(addr + new_len, old_len - new_len);
1111 if (ret && old_len != new_len)
1114 if (!(flags & MREMAP_FIXED) || (new_addr == addr))
1119 addr = PAGE_START(addr);
1120 old_len = PAGE_ALIGN(old_end) - addr;
1121 new_len = PAGE_ALIGN(new_end) - addr;
1123 mutex_lock(&ia32_mmap_mutex);
1124 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1125 mutex_unlock(&ia32_mmap_mutex);
1127 if ((ret >= 0) && (old_len < new_len)) {
1128 /* mremap expanded successfully */
1129 ia32_set_pp(old_end, new_end, flags);
1136 sys32_pipe (int __user *fd)
1141 retval = do_pipe(fds);
1144 if (copy_to_user(fd, fds, sizeof(fds)))
1151 get_tv32 (struct timeval *o, struct compat_timeval __user *i)
1153 return (!access_ok(VERIFY_READ, i, sizeof(*i)) ||
1154 (__get_user(o->tv_sec, &i->tv_sec) | __get_user(o->tv_usec, &i->tv_usec)));
1158 put_tv32 (struct compat_timeval __user *o, struct timeval *i)
1160 return (!access_ok(VERIFY_WRITE, o, sizeof(*o)) ||
1161 (__put_user(i->tv_sec, &o->tv_sec) | __put_user(i->tv_usec, &o->tv_usec)));
1164 asmlinkage unsigned long
1165 sys32_alarm (unsigned int seconds)
1167 return alarm_setitimer(seconds);
1170 /* Translations due to time_t size differences. Which affects all
1171 sorts of things, like timeval and itimerval. */
1173 extern struct timezone sys_tz;
1176 sys32_gettimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1180 do_gettimeofday(&ktv);
1181 if (put_tv32(tv, &ktv))
1185 if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
1192 sys32_settimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1195 struct timespec kts;
1196 struct timezone ktz;
1199 if (get_tv32(&ktv, tv))
1201 kts.tv_sec = ktv.tv_sec;
1202 kts.tv_nsec = ktv.tv_usec * 1000;
1205 if (copy_from_user(&ktz, tz, sizeof(ktz)))
1209 return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL);
1212 struct getdents32_callback {
1213 struct compat_dirent __user *current_dir;
1214 struct compat_dirent __user *previous;
1219 struct readdir32_callback {
1220 struct old_linux32_dirent __user * dirent;
1225 filldir32 (void *__buf, const char *name, int namlen, loff_t offset, ino_t ino,
1226 unsigned int d_type)
1228 struct compat_dirent __user * dirent;
1229 struct getdents32_callback * buf = (struct getdents32_callback *) __buf;
1230 int reclen = ROUND_UP(offsetof(struct compat_dirent, d_name) + namlen + 1, 4);
1232 buf->error = -EINVAL; /* only used if we fail.. */
1233 if (reclen > buf->count)
1235 buf->error = -EFAULT; /* only used if we fail.. */
1236 dirent = buf->previous;
1238 if (put_user(offset, &dirent->d_off))
1240 dirent = buf->current_dir;
1241 buf->previous = dirent;
1242 if (put_user(ino, &dirent->d_ino)
1243 || put_user(reclen, &dirent->d_reclen)
1244 || copy_to_user(dirent->d_name, name, namlen)
1245 || put_user(0, dirent->d_name + namlen))
1247 dirent = (struct compat_dirent __user *) ((char __user *) dirent + reclen);
1248 buf->current_dir = dirent;
1249 buf->count -= reclen;
1254 sys32_getdents (unsigned int fd, struct compat_dirent __user *dirent, unsigned int count)
1257 struct compat_dirent __user * lastdirent;
1258 struct getdents32_callback buf;
1266 buf.current_dir = dirent;
1267 buf.previous = NULL;
1271 error = vfs_readdir(file, filldir32, &buf);
1275 lastdirent = buf.previous;
1278 if (put_user(file->f_pos, &lastdirent->d_off))
1280 error = count - buf.count;
1290 fillonedir32 (void * __buf, const char * name, int namlen, loff_t offset, ino_t ino,
1291 unsigned int d_type)
1293 struct readdir32_callback * buf = (struct readdir32_callback *) __buf;
1294 struct old_linux32_dirent __user * dirent;
1299 dirent = buf->dirent;
1300 if (put_user(ino, &dirent->d_ino)
1301 || put_user(offset, &dirent->d_offset)
1302 || put_user(namlen, &dirent->d_namlen)
1303 || copy_to_user(dirent->d_name, name, namlen)
1304 || put_user(0, dirent->d_name + namlen))
1310 sys32_readdir (unsigned int fd, void __user *dirent, unsigned int count)
1314 struct readdir32_callback buf;
1322 buf.dirent = dirent;
1324 error = vfs_readdir(file, fillonedir32, &buf);
1332 struct sel_arg_struct {
1341 sys32_old_select (struct sel_arg_struct __user *arg)
1343 struct sel_arg_struct a;
1345 if (copy_from_user(&a, arg, sizeof(a)))
1347 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1348 compat_ptr(a.exp), compat_ptr(a.tvp));
1354 #define SEMTIMEDOP 4
1365 sys32_ipc(u32 call, int first, int second, int third, u32 ptr, u32 fifth)
1369 version = call >> 16; /* hack for backward compatibility */
1375 return compat_sys_semtimedop(first, compat_ptr(ptr),
1376 second, compat_ptr(fifth));
1377 /* else fall through for normal semop() */
1379 /* struct sembuf is the same on 32 and 64bit :)) */
1380 return sys_semtimedop(first, compat_ptr(ptr), second,
1383 return sys_semget(first, second, third);
1385 return compat_sys_semctl(first, second, third, compat_ptr(ptr));
1388 return compat_sys_msgsnd(first, second, third, compat_ptr(ptr));
1390 return compat_sys_msgrcv(first, second, fifth, third, version, compat_ptr(ptr));
1392 return sys_msgget((key_t) first, second);
1394 return compat_sys_msgctl(first, second, compat_ptr(ptr));
1397 return compat_sys_shmat(first, second, third, version, compat_ptr(ptr));
1400 return sys_shmdt(compat_ptr(ptr));
1402 return sys_shmget(first, (unsigned)second, third);
1404 return compat_sys_shmctl(first, second, compat_ptr(ptr));
1413 compat_sys_wait4 (compat_pid_t pid, compat_uint_t * stat_addr, int options,
1414 struct compat_rusage *ru);
1417 sys32_waitpid (int pid, unsigned int *stat_addr, int options)
1419 return compat_sys_wait4(pid, stat_addr, options, NULL);
1423 ia32_peek (struct task_struct *child, unsigned long addr, unsigned int *val)
1428 copied = access_process_vm(child, addr, val, sizeof(*val), 0);
1429 return (copied != sizeof(ret)) ? -EIO : 0;
1433 ia32_poke (struct task_struct *child, unsigned long addr, unsigned int val)
1436 if (access_process_vm(child, addr, &val, sizeof(val), 1) != sizeof(val))
1442 * The order in which registers are stored in the ptrace regs structure
1455 #define PT_ORIG_EAX 11
1463 getreg (struct task_struct *child, int regno)
1465 struct pt_regs *child_regs;
1467 child_regs = task_pt_regs(child);
1468 switch (regno / sizeof(int)) {
1469 case PT_EBX: return child_regs->r11;
1470 case PT_ECX: return child_regs->r9;
1471 case PT_EDX: return child_regs->r10;
1472 case PT_ESI: return child_regs->r14;
1473 case PT_EDI: return child_regs->r15;
1474 case PT_EBP: return child_regs->r13;
1475 case PT_EAX: return child_regs->r8;
1476 case PT_ORIG_EAX: return child_regs->r1; /* see dispatch_to_ia32_handler() */
1477 case PT_EIP: return child_regs->cr_iip;
1478 case PT_UESP: return child_regs->r12;
1479 case PT_EFL: return child->thread.eflag;
1480 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1482 case PT_CS: return __USER_CS;
1484 printk(KERN_ERR "ia32.getreg(): unknown register %d\n", regno);
1491 putreg (struct task_struct *child, int regno, unsigned int value)
1493 struct pt_regs *child_regs;
1495 child_regs = task_pt_regs(child);
1496 switch (regno / sizeof(int)) {
1497 case PT_EBX: child_regs->r11 = value; break;
1498 case PT_ECX: child_regs->r9 = value; break;
1499 case PT_EDX: child_regs->r10 = value; break;
1500 case PT_ESI: child_regs->r14 = value; break;
1501 case PT_EDI: child_regs->r15 = value; break;
1502 case PT_EBP: child_regs->r13 = value; break;
1503 case PT_EAX: child_regs->r8 = value; break;
1504 case PT_ORIG_EAX: child_regs->r1 = value; break;
1505 case PT_EIP: child_regs->cr_iip = value; break;
1506 case PT_UESP: child_regs->r12 = value; break;
1507 case PT_EFL: child->thread.eflag = value; break;
1508 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1509 if (value != __USER_DS)
1511 "ia32.putreg: attempt to set invalid segment register %d = %x\n",
1515 if (value != __USER_CS)
1517 "ia32.putreg: attempt to to set invalid segment register %d = %x\n",
1521 printk(KERN_ERR "ia32.putreg: unknown register %d\n", regno);
1527 put_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1528 struct switch_stack *swp, int tos)
1530 struct _fpreg_ia32 *f;
1533 f = (struct _fpreg_ia32 *)(((unsigned long)buf + 15) & ~15);
1534 if ((regno += tos) >= 8)
1538 ia64f2ia32f(f, &ptp->f8);
1541 ia64f2ia32f(f, &ptp->f9);
1544 ia64f2ia32f(f, &ptp->f10);
1547 ia64f2ia32f(f, &ptp->f11);
1553 ia64f2ia32f(f, &swp->f12 + (regno - 4));
1556 copy_to_user(reg, f, sizeof(*reg));
1560 get_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1561 struct switch_stack *swp, int tos)
1564 if ((regno += tos) >= 8)
1568 copy_from_user(&ptp->f8, reg, sizeof(*reg));
1571 copy_from_user(&ptp->f9, reg, sizeof(*reg));
1574 copy_from_user(&ptp->f10, reg, sizeof(*reg));
1577 copy_from_user(&ptp->f11, reg, sizeof(*reg));
1583 copy_from_user(&swp->f12 + (regno - 4), reg, sizeof(*reg));
1590 save_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1592 struct switch_stack *swp;
1593 struct pt_regs *ptp;
1596 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1599 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1600 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1601 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1602 __put_user(tsk->thread.fir, &save->fip);
1603 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1604 __put_user(tsk->thread.fdr, &save->foo);
1605 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1608 * Stack frames start with 16-bytes of temp space
1610 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1611 ptp = task_pt_regs(tsk);
1612 tos = (tsk->thread.fsr >> 11) & 7;
1613 for (i = 0; i < 8; i++)
1614 put_fpreg(i, &save->st_space[i], ptp, swp, tos);
1619 restore_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1621 struct switch_stack *swp;
1622 struct pt_regs *ptp;
1624 unsigned int fsrlo, fsrhi, num32;
1626 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1629 __get_user(num32, (unsigned int __user *)&save->cwd);
1630 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1631 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1632 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1633 num32 = (fsrhi << 16) | fsrlo;
1634 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1635 __get_user(num32, (unsigned int __user *)&save->fip);
1636 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1637 __get_user(num32, (unsigned int __user *)&save->foo);
1638 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1641 * Stack frames start with 16-bytes of temp space
1643 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1644 ptp = task_pt_regs(tsk);
1645 tos = (tsk->thread.fsr >> 11) & 7;
1646 for (i = 0; i < 8; i++)
1647 get_fpreg(i, &save->st_space[i], ptp, swp, tos);
1652 save_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1654 struct switch_stack *swp;
1655 struct pt_regs *ptp;
1657 unsigned long mxcsr=0;
1658 unsigned long num128[2];
1660 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1663 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1664 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1665 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1666 __put_user(tsk->thread.fir, &save->fip);
1667 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1668 __put_user(tsk->thread.fdr, &save->foo);
1669 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1672 * Stack frames start with 16-bytes of temp space
1674 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1675 ptp = task_pt_regs(tsk);
1676 tos = (tsk->thread.fsr >> 11) & 7;
1677 for (i = 0; i < 8; i++)
1678 put_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1680 mxcsr = ((tsk->thread.fcr>>32) & 0xff80) | ((tsk->thread.fsr>>32) & 0x3f);
1681 __put_user(mxcsr & 0xffff, &save->mxcsr);
1682 for (i = 0; i < 8; i++) {
1683 memcpy(&(num128[0]), &(swp->f16) + i*2, sizeof(unsigned long));
1684 memcpy(&(num128[1]), &(swp->f17) + i*2, sizeof(unsigned long));
1685 copy_to_user(&save->xmm_space[0] + 4*i, num128, sizeof(struct _xmmreg_ia32));
1691 restore_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1693 struct switch_stack *swp;
1694 struct pt_regs *ptp;
1696 unsigned int fsrlo, fsrhi, num32;
1698 unsigned long num64;
1699 unsigned long num128[2];
1701 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1704 __get_user(num32, (unsigned int __user *)&save->cwd);
1705 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1706 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1707 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1708 num32 = (fsrhi << 16) | fsrlo;
1709 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1710 __get_user(num32, (unsigned int __user *)&save->fip);
1711 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1712 __get_user(num32, (unsigned int __user *)&save->foo);
1713 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1716 * Stack frames start with 16-bytes of temp space
1718 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1719 ptp = task_pt_regs(tsk);
1720 tos = (tsk->thread.fsr >> 11) & 7;
1721 for (i = 0; i < 8; i++)
1722 get_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1724 __get_user(mxcsr, (unsigned int __user *)&save->mxcsr);
1725 num64 = mxcsr & 0xff10;
1726 tsk->thread.fcr = (tsk->thread.fcr & (~0xff1000000000UL)) | (num64<<32);
1727 num64 = mxcsr & 0x3f;
1728 tsk->thread.fsr = (tsk->thread.fsr & (~0x3f00000000UL)) | (num64<<32);
1730 for (i = 0; i < 8; i++) {
1731 copy_from_user(num128, &save->xmm_space[0] + 4*i, sizeof(struct _xmmreg_ia32));
1732 memcpy(&(swp->f16) + i*2, &(num128[0]), sizeof(unsigned long));
1733 memcpy(&(swp->f17) + i*2, &(num128[1]), sizeof(unsigned long));
1739 sys32_ptrace (int request, pid_t pid, unsigned int addr, unsigned int data)
1741 struct task_struct *child;
1742 unsigned int value, tmp;
1746 if (request == PTRACE_TRACEME) {
1747 ret = ptrace_traceme();
1751 child = ptrace_get_task_struct(pid);
1752 if (IS_ERR(child)) {
1753 ret = PTR_ERR(child);
1757 if (request == PTRACE_ATTACH) {
1758 ret = sys_ptrace(request, pid, addr, data);
1762 ret = ptrace_check_attach(child, request == PTRACE_KILL);
1767 case PTRACE_PEEKTEXT:
1768 case PTRACE_PEEKDATA: /* read word at location addr */
1769 ret = ia32_peek(child, addr, &value);
1771 ret = put_user(value, (unsigned int __user *) compat_ptr(data));
1776 case PTRACE_POKETEXT:
1777 case PTRACE_POKEDATA: /* write the word at location addr */
1778 ret = ia32_poke(child, addr, data);
1781 case PTRACE_PEEKUSR: /* read word at addr in USER area */
1783 if ((addr & 3) || addr > 17*sizeof(int))
1786 tmp = getreg(child, addr);
1787 if (!put_user(tmp, (unsigned int __user *) compat_ptr(data)))
1791 case PTRACE_POKEUSR: /* write word at addr in USER area */
1793 if ((addr & 3) || addr > 17*sizeof(int))
1796 putreg(child, addr, data);
1800 case IA32_PTRACE_GETREGS:
1801 if (!access_ok(VERIFY_WRITE, compat_ptr(data), 17*sizeof(int))) {
1805 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1806 put_user(getreg(child, i), (unsigned int __user *) compat_ptr(data));
1807 data += sizeof(int);
1812 case IA32_PTRACE_SETREGS:
1813 if (!access_ok(VERIFY_READ, compat_ptr(data), 17*sizeof(int))) {
1817 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1818 get_user(tmp, (unsigned int __user *) compat_ptr(data));
1819 putreg(child, i, tmp);
1820 data += sizeof(int);
1825 case IA32_PTRACE_GETFPREGS:
1826 ret = save_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1830 case IA32_PTRACE_GETFPXREGS:
1831 ret = save_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1835 case IA32_PTRACE_SETFPREGS:
1836 ret = restore_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1840 case IA32_PTRACE_SETFPXREGS:
1841 ret = restore_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1845 case PTRACE_GETEVENTMSG:
1846 ret = put_user(child->ptrace_message, (unsigned int __user *) compat_ptr(data));
1849 case PTRACE_SYSCALL: /* continue, stop after next syscall */
1850 case PTRACE_CONT: /* restart after signal. */
1852 case PTRACE_SINGLESTEP: /* execute chile for one instruction */
1853 case PTRACE_DETACH: /* detach a process */
1854 ret = sys_ptrace(request, pid, addr, data);
1858 ret = ptrace_request(child, request, addr, data);
1863 put_task_struct(child);
1871 unsigned int ss_flags;
1872 unsigned int ss_size;
1876 sys32_sigaltstack (ia32_stack_t __user *uss32, ia32_stack_t __user *uoss32,
1877 long arg2, long arg3, long arg4, long arg5, long arg6,
1878 long arg7, struct pt_regs pt)
1883 mm_segment_t old_fs = get_fs();
1886 if (copy_from_user(&buf32, uss32, sizeof(ia32_stack_t)))
1888 uss.ss_sp = (void __user *) (long) buf32.ss_sp;
1889 uss.ss_flags = buf32.ss_flags;
1890 /* MINSIGSTKSZ is different for ia32 vs ia64. We lie here to pass the
1891 check and set it to the user requested value later */
1892 if ((buf32.ss_flags != SS_DISABLE) && (buf32.ss_size < MINSIGSTKSZ_IA32)) {
1896 uss.ss_size = MINSIGSTKSZ;
1899 ret = do_sigaltstack(uss32 ? (stack_t __user *) &uss : NULL,
1900 (stack_t __user *) &uoss, pt.r12);
1901 current->sas_ss_size = buf32.ss_size;
1907 buf32.ss_sp = (long __user) uoss.ss_sp;
1908 buf32.ss_flags = uoss.ss_flags;
1909 buf32.ss_size = uoss.ss_size;
1910 if (copy_to_user(uoss32, &buf32, sizeof(ia32_stack_t)))
1919 current->state = TASK_INTERRUPTIBLE;
1921 return -ERESTARTNOHAND;
1925 sys32_msync (unsigned int start, unsigned int len, int flags)
1929 if (OFFSET4K(start))
1931 addr = PAGE_START(start);
1932 return sys_msync(addr, len + (start - addr), flags);
1938 unsigned int oldval;
1939 unsigned int oldlenp;
1940 unsigned int newval;
1941 unsigned int newlen;
1942 unsigned int __unused[4];
1945 #ifdef CONFIG_SYSCTL
1947 sys32_sysctl (struct sysctl32 __user *args)
1949 struct sysctl32 a32;
1950 mm_segment_t old_fs = get_fs ();
1951 void __user *oldvalp, *newvalp;
1956 if (copy_from_user(&a32, args, sizeof(a32)))
1960 * We need to pre-validate these because we have to disable address checking
1961 * before calling do_sysctl() because of OLDLEN but we can't run the risk of the
1962 * user specifying bad addresses here. Well, since we're dealing with 32 bit
1963 * addresses, we KNOW that access_ok() will always succeed, so this is an
1964 * expensive NOP, but so what...
1966 namep = (int __user *) compat_ptr(a32.name);
1967 oldvalp = compat_ptr(a32.oldval);
1968 newvalp = compat_ptr(a32.newval);
1970 if ((oldvalp && get_user(oldlen, (int __user *) compat_ptr(a32.oldlenp)))
1971 || !access_ok(VERIFY_WRITE, namep, 0)
1972 || !access_ok(VERIFY_WRITE, oldvalp, 0)
1973 || !access_ok(VERIFY_WRITE, newvalp, 0))
1978 ret = do_sysctl(namep, a32.nlen, oldvalp, (size_t __user *) &oldlen,
1979 newvalp, (size_t) a32.newlen);
1983 if (oldvalp && put_user (oldlen, (int __user *) compat_ptr(a32.oldlenp)))
1991 sys32_newuname (struct new_utsname __user *name)
1993 int ret = sys_newuname(name);
1996 if (copy_to_user(name->machine, "i686\0\0\0", 8))
2002 sys32_getresuid16 (u16 __user *ruid, u16 __user *euid, u16 __user *suid)
2006 mm_segment_t old_fs = get_fs();
2009 ret = sys_getresuid((uid_t __user *) &a, (uid_t __user *) &b, (uid_t __user *) &c);
2012 if (put_user(a, ruid) || put_user(b, euid) || put_user(c, suid))
2018 sys32_getresgid16 (u16 __user *rgid, u16 __user *egid, u16 __user *sgid)
2022 mm_segment_t old_fs = get_fs();
2025 ret = sys_getresgid((gid_t __user *) &a, (gid_t __user *) &b, (gid_t __user *) &c);
2031 return put_user(a, rgid) | put_user(b, egid) | put_user(c, sgid);
2035 sys32_lseek (unsigned int fd, int offset, unsigned int whence)
2037 /* Sign-extension of "offset" is important here... */
2038 return sys_lseek(fd, offset, whence);
2042 groups16_to_user(short __user *grouplist, struct group_info *group_info)
2047 for (i = 0; i < group_info->ngroups; i++) {
2048 group = (short)GROUP_AT(group_info, i);
2049 if (put_user(group, grouplist+i))
2057 groups16_from_user(struct group_info *group_info, short __user *grouplist)
2062 for (i = 0; i < group_info->ngroups; i++) {
2063 if (get_user(group, grouplist+i))
2065 GROUP_AT(group_info, i) = (gid_t)group;
2072 sys32_getgroups16 (int gidsetsize, short __user *grouplist)
2079 get_group_info(current->group_info);
2080 i = current->group_info->ngroups;
2082 if (i > gidsetsize) {
2086 if (groups16_to_user(grouplist, current->group_info)) {
2092 put_group_info(current->group_info);
2097 sys32_setgroups16 (int gidsetsize, short __user *grouplist)
2099 struct group_info *group_info;
2102 if (!capable(CAP_SETGID))
2104 if ((unsigned)gidsetsize > NGROUPS_MAX)
2107 group_info = groups_alloc(gidsetsize);
2110 retval = groups16_from_user(group_info, grouplist);
2112 put_group_info(group_info);
2116 retval = set_current_groups(group_info);
2117 put_group_info(group_info);
2123 sys32_truncate64 (unsigned int path, unsigned int len_lo, unsigned int len_hi)
2125 return sys_truncate(compat_ptr(path), ((unsigned long) len_hi << 32) | len_lo);
2129 sys32_ftruncate64 (int fd, unsigned int len_lo, unsigned int len_hi)
2131 return sys_ftruncate(fd, ((unsigned long) len_hi << 32) | len_lo);
2135 putstat64 (struct stat64 __user *ubuf, struct kstat *kbuf)
2140 if (clear_user(ubuf, sizeof(*ubuf)))
2143 hdev = huge_encode_dev(kbuf->dev);
2144 err = __put_user(hdev, (u32 __user*)&ubuf->st_dev);
2145 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_dev) + 1);
2146 err |= __put_user(kbuf->ino, &ubuf->__st_ino);
2147 err |= __put_user(kbuf->ino, &ubuf->st_ino_lo);
2148 err |= __put_user(kbuf->ino >> 32, &ubuf->st_ino_hi);
2149 err |= __put_user(kbuf->mode, &ubuf->st_mode);
2150 err |= __put_user(kbuf->nlink, &ubuf->st_nlink);
2151 err |= __put_user(kbuf->uid, &ubuf->st_uid);
2152 err |= __put_user(kbuf->gid, &ubuf->st_gid);
2153 hdev = huge_encode_dev(kbuf->rdev);
2154 err = __put_user(hdev, (u32 __user*)&ubuf->st_rdev);
2155 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_rdev) + 1);
2156 err |= __put_user(kbuf->size, &ubuf->st_size_lo);
2157 err |= __put_user((kbuf->size >> 32), &ubuf->st_size_hi);
2158 err |= __put_user(kbuf->atime.tv_sec, &ubuf->st_atime);
2159 err |= __put_user(kbuf->atime.tv_nsec, &ubuf->st_atime_nsec);
2160 err |= __put_user(kbuf->mtime.tv_sec, &ubuf->st_mtime);
2161 err |= __put_user(kbuf->mtime.tv_nsec, &ubuf->st_mtime_nsec);
2162 err |= __put_user(kbuf->ctime.tv_sec, &ubuf->st_ctime);
2163 err |= __put_user(kbuf->ctime.tv_nsec, &ubuf->st_ctime_nsec);
2164 err |= __put_user(kbuf->blksize, &ubuf->st_blksize);
2165 err |= __put_user(kbuf->blocks, &ubuf->st_blocks);
2170 sys32_stat64 (char __user *filename, struct stat64 __user *statbuf)
2173 long ret = vfs_stat(filename, &s);
2175 ret = putstat64(statbuf, &s);
2180 sys32_lstat64 (char __user *filename, struct stat64 __user *statbuf)
2183 long ret = vfs_lstat(filename, &s);
2185 ret = putstat64(statbuf, &s);
2190 sys32_fstat64 (unsigned int fd, struct stat64 __user *statbuf)
2193 long ret = vfs_fstat(fd, &s);
2195 ret = putstat64(statbuf, &s);
2217 sys32_sysinfo (struct sysinfo32 __user *info)
2222 mm_segment_t old_fs = get_fs();
2225 ret = sys_sysinfo((struct sysinfo __user *) &s);
2227 /* Check to see if any memory value is too large for 32-bit and
2228 * scale down if needed.
2230 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2231 while (s.mem_unit < PAGE_SIZE) {
2235 s.totalram >>= bitcount;
2236 s.freeram >>= bitcount;
2237 s.sharedram >>= bitcount;
2238 s.bufferram >>= bitcount;
2239 s.totalswap >>= bitcount;
2240 s.freeswap >>= bitcount;
2241 s.totalhigh >>= bitcount;
2242 s.freehigh >>= bitcount;
2245 if (!access_ok(VERIFY_WRITE, info, sizeof(*info)))
2248 err = __put_user(s.uptime, &info->uptime);
2249 err |= __put_user(s.loads[0], &info->loads[0]);
2250 err |= __put_user(s.loads[1], &info->loads[1]);
2251 err |= __put_user(s.loads[2], &info->loads[2]);
2252 err |= __put_user(s.totalram, &info->totalram);
2253 err |= __put_user(s.freeram, &info->freeram);
2254 err |= __put_user(s.sharedram, &info->sharedram);
2255 err |= __put_user(s.bufferram, &info->bufferram);
2256 err |= __put_user(s.totalswap, &info->totalswap);
2257 err |= __put_user(s.freeswap, &info->freeswap);
2258 err |= __put_user(s.procs, &info->procs);
2259 err |= __put_user (s.totalhigh, &info->totalhigh);
2260 err |= __put_user (s.freehigh, &info->freehigh);
2261 err |= __put_user (s.mem_unit, &info->mem_unit);
2268 sys32_sched_rr_get_interval (pid_t pid, struct compat_timespec __user *interval)
2270 mm_segment_t old_fs = get_fs();
2275 ret = sys_sched_rr_get_interval(pid, (struct timespec __user *) &t);
2277 if (put_compat_timespec(&t, interval))
2283 sys32_pread (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2285 return sys_pread64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2289 sys32_pwrite (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2291 return sys_pwrite64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2295 sys32_sendfile (int out_fd, int in_fd, int __user *offset, unsigned int count)
2297 mm_segment_t old_fs = get_fs();
2301 if (offset && get_user(of, offset))
2305 ret = sys_sendfile(out_fd, in_fd, offset ? (off_t __user *) &of : NULL, count);
2308 if (offset && put_user(of, offset))
2315 sys32_personality (unsigned int personality)
2319 if (current->personality == PER_LINUX32 && personality == PER_LINUX)
2320 personality = PER_LINUX32;
2321 ret = sys_personality(personality);
2322 if (ret == PER_LINUX32)
2327 asmlinkage unsigned long
2328 sys32_brk (unsigned int brk)
2330 unsigned long ret, obrk;
2331 struct mm_struct *mm = current->mm;
2336 clear_user(compat_ptr(ret), PAGE_ALIGN(ret) - ret);
2340 /* Structure for ia32 emulation on ia64 */
2341 struct epoll_event32
2348 sys32_epoll_ctl(int epfd, int op, int fd, struct epoll_event32 __user *event)
2350 mm_segment_t old_fs = get_fs();
2351 struct epoll_event event64;
2355 if (!access_ok(VERIFY_READ, event, sizeof(struct epoll_event32)))
2358 __get_user(event64.events, &event->events);
2359 __get_user(data_halfword, &event->data[0]);
2360 event64.data = data_halfword;
2361 __get_user(data_halfword, &event->data[1]);
2362 event64.data |= (u64)data_halfword << 32;
2365 error = sys_epoll_ctl(epfd, op, fd, (struct epoll_event __user *) &event64);
2372 sys32_epoll_wait(int epfd, struct epoll_event32 __user * events, int maxevents,
2375 struct epoll_event *events64 = NULL;
2376 mm_segment_t old_fs = get_fs();
2377 int numevents, size;
2379 int do_free_pages = 0;
2381 if (maxevents <= 0) {
2385 /* Verify that the area passed by the user is writeable */
2386 if (!access_ok(VERIFY_WRITE, events, maxevents * sizeof(struct epoll_event32)))
2390 * Allocate space for the intermediate copy. If the space needed
2391 * is large enough to cause kmalloc to fail, then try again with
2394 size = maxevents * sizeof(struct epoll_event);
2395 events64 = kmalloc(size, GFP_KERNEL);
2396 if (events64 == NULL) {
2397 events64 = (struct epoll_event *)
2398 __get_free_pages(GFP_KERNEL, get_order(size));
2399 if (events64 == NULL)
2404 /* Do the system call */
2405 set_fs(KERNEL_DS); /* copy_to/from_user should work on kernel mem*/
2406 numevents = sys_epoll_wait(epfd, (struct epoll_event __user *) events64,
2407 maxevents, timeout);
2410 /* Don't modify userspace memory if we're returning an error */
2411 if (numevents > 0) {
2412 /* Translate the 64-bit structures back into the 32-bit
2414 for (evt_idx = 0; evt_idx < numevents; evt_idx++) {
2415 __put_user(events64[evt_idx].events,
2416 &events[evt_idx].events);
2417 __put_user((u32)events64[evt_idx].data,
2418 &events[evt_idx].data[0]);
2419 __put_user((u32)(events64[evt_idx].data >> 32),
2420 &events[evt_idx].data[1]);
2425 free_pages((unsigned long) events64, get_order(size));
2432 * Get a yet unused TLS descriptor index.
2437 struct thread_struct *t = ¤t->thread;
2440 for (idx = 0; idx < GDT_ENTRY_TLS_ENTRIES; idx++)
2441 if (desc_empty(t->tls_array + idx))
2442 return idx + GDT_ENTRY_TLS_MIN;
2447 * Set a given TLS descriptor:
2450 sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2452 struct thread_struct *t = ¤t->thread;
2453 struct ia32_user_desc info;
2454 struct desc_struct *desc;
2457 if (copy_from_user(&info, u_info, sizeof(info)))
2459 idx = info.entry_number;
2462 * index -1 means the kernel should try to find and allocate an empty descriptor:
2465 idx = get_free_idx();
2468 if (put_user(idx, &u_info->entry_number))
2472 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2475 desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN;
2477 cpu = smp_processor_id();
2479 if (LDT_empty(&info)) {
2483 desc->a = LDT_entry_a(&info);
2484 desc->b = LDT_entry_b(&info);
2491 * Get the current Thread-Local Storage area:
2494 #define GET_BASE(desc) ( \
2495 (((desc)->a >> 16) & 0x0000ffff) | \
2496 (((desc)->b << 16) & 0x00ff0000) | \
2497 ( (desc)->b & 0xff000000) )
2499 #define GET_LIMIT(desc) ( \
2500 ((desc)->a & 0x0ffff) | \
2501 ((desc)->b & 0xf0000) )
2503 #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
2504 #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
2505 #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
2506 #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
2507 #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
2508 #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
2511 sys32_get_thread_area (struct ia32_user_desc __user *u_info)
2513 struct ia32_user_desc info;
2514 struct desc_struct *desc;
2517 if (get_user(idx, &u_info->entry_number))
2519 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2522 desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
2524 info.entry_number = idx;
2525 info.base_addr = GET_BASE(desc);
2526 info.limit = GET_LIMIT(desc);
2527 info.seg_32bit = GET_32BIT(desc);
2528 info.contents = GET_CONTENTS(desc);
2529 info.read_exec_only = !GET_WRITABLE(desc);
2530 info.limit_in_pages = GET_LIMIT_PAGES(desc);
2531 info.seg_not_present = !GET_PRESENT(desc);
2532 info.useable = GET_USEABLE(desc);
2534 if (copy_to_user(u_info, &info, sizeof(info)))
2539 long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
2540 __u32 len_low, __u32 len_high, int advice)
2542 return sys_fadvise64_64(fd,
2543 (((u64)offset_high)<<32) | offset_low,
2544 (((u64)len_high)<<32) | len_low,
2548 #ifdef NOTYET /* UNTESTED FOR IA64 FROM HERE DOWN */
2550 asmlinkage long sys32_setreuid(compat_uid_t ruid, compat_uid_t euid)
2554 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2555 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2556 return sys_setreuid(sruid, seuid);
2560 sys32_setresuid(compat_uid_t ruid, compat_uid_t euid,
2563 uid_t sruid, seuid, ssuid;
2565 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2566 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2567 ssuid = (suid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)suid);
2568 return sys_setresuid(sruid, seuid, ssuid);
2572 sys32_setregid(compat_gid_t rgid, compat_gid_t egid)
2576 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2577 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2578 return sys_setregid(srgid, segid);
2582 sys32_setresgid(compat_gid_t rgid, compat_gid_t egid,
2585 gid_t srgid, segid, ssgid;
2587 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2588 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2589 ssgid = (sgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)sgid);
2590 return sys_setresgid(srgid, segid, ssgid);