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)
131 if ((u64) stat->size > MAX_NON_LFS ||
132 !old_valid_dev(stat->dev) ||
133 !old_valid_dev(stat->rdev))
137 if (sizeof(ino) < sizeof(stat->ino) && ino != stat->ino)
140 if (clear_user(ubuf, sizeof(*ubuf)))
143 err = __put_user(old_encode_dev(stat->dev), &ubuf->st_dev);
144 err |= __put_user(ino, &ubuf->st_ino);
145 err |= __put_user(stat->mode, &ubuf->st_mode);
146 err |= __put_user(stat->nlink, &ubuf->st_nlink);
147 err |= __put_user(high2lowuid(stat->uid), &ubuf->st_uid);
148 err |= __put_user(high2lowgid(stat->gid), &ubuf->st_gid);
149 err |= __put_user(old_encode_dev(stat->rdev), &ubuf->st_rdev);
150 err |= __put_user(stat->size, &ubuf->st_size);
151 err |= __put_user(stat->atime.tv_sec, &ubuf->st_atime);
152 err |= __put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec);
153 err |= __put_user(stat->mtime.tv_sec, &ubuf->st_mtime);
154 err |= __put_user(stat->mtime.tv_nsec, &ubuf->st_mtime_nsec);
155 err |= __put_user(stat->ctime.tv_sec, &ubuf->st_ctime);
156 err |= __put_user(stat->ctime.tv_nsec, &ubuf->st_ctime_nsec);
157 err |= __put_user(stat->blksize, &ubuf->st_blksize);
158 err |= __put_user(stat->blocks, &ubuf->st_blocks);
162 #if PAGE_SHIFT > IA32_PAGE_SHIFT
166 get_page_prot (struct vm_area_struct *vma, unsigned long addr)
170 if (!vma || vma->vm_start > addr)
173 if (vma->vm_flags & VM_READ)
175 if (vma->vm_flags & VM_WRITE)
177 if (vma->vm_flags & VM_EXEC)
183 * Map a subpage by creating an anonymous page that contains the union of the old page and
187 mmap_subpage (struct file *file, unsigned long start, unsigned long end, int prot, int flags,
192 unsigned long ret = 0;
193 struct vm_area_struct *vma = find_vma(current->mm, start);
194 int old_prot = get_page_prot(vma, start);
196 DBG("mmap_subpage(file=%p,start=0x%lx,end=0x%lx,prot=%x,flags=%x,off=0x%llx)\n",
197 file, start, end, prot, flags, off);
200 /* Optimize the case where the old mmap and the new mmap are both anonymous */
201 if ((old_prot & PROT_WRITE) && (flags & MAP_ANONYMOUS) && !vma->vm_file) {
202 if (clear_user((void __user *) start, end - start)) {
209 page = (void *) get_zeroed_page(GFP_KERNEL);
214 copy_from_user(page, (void __user *) PAGE_START(start), PAGE_SIZE);
216 down_write(¤t->mm->mmap_sem);
218 ret = do_mmap(NULL, PAGE_START(start), PAGE_SIZE, prot | PROT_WRITE,
219 flags | MAP_FIXED | MAP_ANONYMOUS, 0);
221 up_write(¤t->mm->mmap_sem);
223 if (IS_ERR((void *) ret))
227 /* copy back the old page contents. */
228 if (offset_in_page(start))
229 copy_to_user((void __user *) PAGE_START(start), page,
230 offset_in_page(start));
231 if (offset_in_page(end))
232 copy_to_user((void __user *) end, page + offset_in_page(end),
233 PAGE_SIZE - offset_in_page(end));
236 if (!(flags & MAP_ANONYMOUS)) {
237 /* read the file contents */
238 inode = file->f_dentry->d_inode;
239 if (!inode->i_fop || !file->f_op->read
240 || ((*file->f_op->read)(file, (char __user *) start, end - start, &off) < 0))
248 if (!(prot & PROT_WRITE))
249 ret = sys_mprotect(PAGE_START(start), PAGE_SIZE, prot | old_prot);
252 free_page((unsigned long) page);
256 /* SLAB cache for partial_page structures */
257 kmem_cache_t *partial_page_cachep;
260 * init partial_page_list.
261 * return 0 means kmalloc fail.
263 struct partial_page_list*
264 ia32_init_pp_list(void)
266 struct partial_page_list *p;
268 if ((p = kmalloc(sizeof(*p), GFP_KERNEL)) == NULL)
273 atomic_set(&p->pp_count, 1);
278 * Search for the partial page with @start in partial page list @ppl.
279 * If finds the partial page, return the found partial page.
280 * Else, return 0 and provide @pprev, @rb_link, @rb_parent to
281 * be used by later __ia32_insert_pp().
283 static struct partial_page *
284 __ia32_find_pp(struct partial_page_list *ppl, unsigned int start,
285 struct partial_page **pprev, struct rb_node ***rb_link,
286 struct rb_node **rb_parent)
288 struct partial_page *pp;
289 struct rb_node **__rb_link, *__rb_parent, *rb_prev;
292 if (pp && pp->base == start)
295 __rb_link = &ppl->ppl_rb.rb_node;
296 rb_prev = __rb_parent = NULL;
299 __rb_parent = *__rb_link;
300 pp = rb_entry(__rb_parent, struct partial_page, pp_rb);
302 if (pp->base == start) {
305 } else if (pp->base < start) {
306 rb_prev = __rb_parent;
307 __rb_link = &__rb_parent->rb_right;
309 __rb_link = &__rb_parent->rb_left;
313 *rb_link = __rb_link;
314 *rb_parent = __rb_parent;
317 *pprev = rb_entry(rb_prev, struct partial_page, pp_rb);
322 * insert @pp into @ppl.
325 __ia32_insert_pp(struct partial_page_list *ppl, struct partial_page *pp,
326 struct partial_page *prev, struct rb_node **rb_link,
327 struct rb_node *rb_parent)
331 pp->next = prev->next;
336 pp->next = rb_entry(rb_parent,
337 struct partial_page, pp_rb);
343 rb_link_node(&pp->pp_rb, rb_parent, rb_link);
344 rb_insert_color(&pp->pp_rb, &ppl->ppl_rb);
350 * delete @pp from partial page list @ppl.
353 __ia32_delete_pp(struct partial_page_list *ppl, struct partial_page *pp,
354 struct partial_page *prev)
357 prev->next = pp->next;
358 if (ppl->pp_hint == pp)
361 ppl->pp_head = pp->next;
362 if (ppl->pp_hint == pp)
363 ppl->pp_hint = pp->next;
365 rb_erase(&pp->pp_rb, &ppl->ppl_rb);
366 kmem_cache_free(partial_page_cachep, pp);
369 static struct partial_page *
370 __pp_prev(struct partial_page *pp)
372 struct rb_node *prev = rb_prev(&pp->pp_rb);
374 return rb_entry(prev, struct partial_page, pp_rb);
380 * Delete partial pages with address between @start and @end.
381 * @start and @end are page aligned.
384 __ia32_delete_pp_range(unsigned int start, unsigned int end)
386 struct partial_page *pp, *prev;
387 struct rb_node **rb_link, *rb_parent;
392 pp = __ia32_find_pp(current->thread.ppl, start, &prev,
393 &rb_link, &rb_parent);
395 prev = __pp_prev(pp);
400 pp = current->thread.ppl->pp_head;
403 while (pp && pp->base < end) {
404 struct partial_page *tmp = pp->next;
405 __ia32_delete_pp(current->thread.ppl, pp, prev);
411 * Set the range between @start and @end in bitmap.
412 * @start and @end should be IA32 page aligned and in the same IA64 page.
415 __ia32_set_pp(unsigned int start, unsigned int end, int flags)
417 struct partial_page *pp, *prev;
418 struct rb_node ** rb_link, *rb_parent;
419 unsigned int pstart, start_bit, end_bit, i;
421 pstart = PAGE_START(start);
422 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
423 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
425 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
426 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
427 &rb_link, &rb_parent);
429 for (i = start_bit; i < end_bit; i++)
430 set_bit(i, &pp->bitmap);
432 * Check: if this partial page has been set to a full page,
435 if (find_first_zero_bit(&pp->bitmap, sizeof(pp->bitmap)*8) >=
436 PAGE_SIZE/IA32_PAGE_SIZE) {
437 __ia32_delete_pp(current->thread.ppl, pp, __pp_prev(pp));
443 * MAP_FIXED may lead to overlapping mmap.
444 * In this case, the requested mmap area may already mmaped as a full
445 * page. So check vma before adding a new partial page.
447 if (flags & MAP_FIXED) {
448 struct vm_area_struct *vma = find_vma(current->mm, pstart);
449 if (vma && vma->vm_start <= pstart)
453 /* new a partial_page */
454 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
459 for (i=start_bit; i<end_bit; i++)
460 set_bit(i, &(pp->bitmap));
462 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
467 * @start and @end should be IA32 page aligned, but don't need to be in the
468 * same IA64 page. Split @start and @end to make sure they're in the same IA64
469 * page, then call __ia32_set_pp().
472 ia32_set_pp(unsigned int start, unsigned int end, int flags)
474 down_write(¤t->mm->mmap_sem);
475 if (flags & MAP_FIXED) {
477 * MAP_FIXED may lead to overlapping mmap. When this happens,
478 * a series of complete IA64 pages results in deletion of
479 * old partial pages in that range.
481 __ia32_delete_pp_range(PAGE_ALIGN(start), PAGE_START(end));
484 if (end < PAGE_ALIGN(start)) {
485 __ia32_set_pp(start, end, flags);
487 if (offset_in_page(start))
488 __ia32_set_pp(start, PAGE_ALIGN(start), flags);
489 if (offset_in_page(end))
490 __ia32_set_pp(PAGE_START(end), end, flags);
492 up_write(¤t->mm->mmap_sem);
496 * Unset the range between @start and @end in bitmap.
497 * @start and @end should be IA32 page aligned and in the same IA64 page.
498 * After doing that, if the bitmap is 0, then free the page and return 1,
500 * If not find the partial page in the list, then
501 * If the vma exists, then the full page is set to a partial page;
502 * Else return -ENOMEM.
505 __ia32_unset_pp(unsigned int start, unsigned int end)
507 struct partial_page *pp, *prev;
508 struct rb_node ** rb_link, *rb_parent;
509 unsigned int pstart, start_bit, end_bit, i;
510 struct vm_area_struct *vma;
512 pstart = PAGE_START(start);
513 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
514 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
516 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
518 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
519 &rb_link, &rb_parent);
521 for (i = start_bit; i < end_bit; i++)
522 clear_bit(i, &pp->bitmap);
523 if (pp->bitmap == 0) {
524 __ia32_delete_pp(current->thread.ppl, pp, __pp_prev(pp));
530 vma = find_vma(current->mm, pstart);
531 if (!vma || vma->vm_start > pstart) {
535 /* new a partial_page */
536 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
541 for (i = 0; i < start_bit; i++)
542 set_bit(i, &(pp->bitmap));
543 for (i = end_bit; i < PAGE_SIZE / IA32_PAGE_SIZE; i++)
544 set_bit(i, &(pp->bitmap));
546 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
551 * Delete pp between PAGE_ALIGN(start) and PAGE_START(end) by calling
552 * __ia32_delete_pp_range(). Unset possible partial pages by calling
554 * The returned value see __ia32_unset_pp().
557 ia32_unset_pp(unsigned int *startp, unsigned int *endp)
559 unsigned int start = *startp, end = *endp;
562 down_write(¤t->mm->mmap_sem);
564 __ia32_delete_pp_range(PAGE_ALIGN(start), PAGE_START(end));
566 if (end < PAGE_ALIGN(start)) {
567 ret = __ia32_unset_pp(start, end);
569 *startp = PAGE_START(start);
570 *endp = PAGE_ALIGN(end);
573 /* to shortcut sys_munmap() in sys32_munmap() */
574 *startp = PAGE_START(start);
575 *endp = PAGE_START(end);
578 if (offset_in_page(start)) {
579 ret = __ia32_unset_pp(start, PAGE_ALIGN(start));
581 *startp = PAGE_START(start);
583 *startp = PAGE_ALIGN(start);
587 if (offset_in_page(end)) {
588 ret = __ia32_unset_pp(PAGE_START(end), end);
590 *endp = PAGE_ALIGN(end);
592 *endp = PAGE_START(end);
597 up_write(¤t->mm->mmap_sem);
602 * Compare the range between @start and @end with bitmap in partial page.
603 * @start and @end should be IA32 page aligned and in the same IA64 page.
606 __ia32_compare_pp(unsigned int start, unsigned int end)
608 struct partial_page *pp, *prev;
609 struct rb_node ** rb_link, *rb_parent;
610 unsigned int pstart, start_bit, end_bit, size;
611 unsigned int first_bit, next_zero_bit; /* the first range in bitmap */
613 pstart = PAGE_START(start);
615 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
616 &rb_link, &rb_parent);
620 start_bit = (start % PAGE_SIZE) / IA32_PAGE_SIZE;
621 end_bit = (end % PAGE_SIZE) / IA32_PAGE_SIZE;
622 size = sizeof(pp->bitmap) * 8;
623 first_bit = find_first_bit(&pp->bitmap, size);
624 next_zero_bit = find_next_zero_bit(&pp->bitmap, size, first_bit);
625 if ((start_bit < first_bit) || (end_bit > next_zero_bit)) {
626 /* exceeds the first range in bitmap */
628 } else if ((start_bit == first_bit) && (end_bit == next_zero_bit)) {
629 first_bit = find_next_bit(&pp->bitmap, size, next_zero_bit);
630 if ((next_zero_bit < first_bit) && (first_bit < size))
631 return 1; /* has next range */
633 return 0; /* no next range */
639 * @start and @end should be IA32 page aligned, but don't need to be in the
640 * same IA64 page. Split @start and @end to make sure they're in the same IA64
641 * page, then call __ia32_compare_pp().
643 * Take this as example: the range is the 1st and 2nd 4K page.
644 * Return 0 if they fit bitmap exactly, i.e. bitmap = 00000011;
645 * Return 1 if the range doesn't cover whole bitmap, e.g. bitmap = 00001111;
646 * Return -ENOMEM if the range exceeds the bitmap, e.g. bitmap = 00000001 or
650 ia32_compare_pp(unsigned int *startp, unsigned int *endp)
652 unsigned int start = *startp, end = *endp;
655 down_write(¤t->mm->mmap_sem);
657 if (end < PAGE_ALIGN(start)) {
658 retval = __ia32_compare_pp(start, end);
660 *startp = PAGE_START(start);
661 *endp = PAGE_ALIGN(end);
664 if (offset_in_page(start)) {
665 retval = __ia32_compare_pp(start,
668 *startp = PAGE_START(start);
672 if (offset_in_page(end)) {
673 retval = __ia32_compare_pp(PAGE_START(end), end);
675 *endp = PAGE_ALIGN(end);
680 up_write(¤t->mm->mmap_sem);
685 __ia32_drop_pp_list(struct partial_page_list *ppl)
687 struct partial_page *pp = ppl->pp_head;
690 struct partial_page *next = pp->next;
691 kmem_cache_free(partial_page_cachep, pp);
699 ia32_drop_partial_page_list(struct task_struct *task)
701 struct partial_page_list* ppl = task->thread.ppl;
703 if (ppl && atomic_dec_and_test(&ppl->pp_count))
704 __ia32_drop_pp_list(ppl);
708 * Copy current->thread.ppl to ppl (already initialized).
711 __ia32_copy_pp_list(struct partial_page_list *ppl)
713 struct partial_page *pp, *tmp, *prev;
714 struct rb_node **rb_link, *rb_parent;
718 ppl->ppl_rb = RB_ROOT;
719 rb_link = &ppl->ppl_rb.rb_node;
723 for (pp = current->thread.ppl->pp_head; pp; pp = pp->next) {
724 tmp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
728 __ia32_insert_pp(ppl, tmp, prev, rb_link, rb_parent);
730 rb_link = &tmp->pp_rb.rb_right;
731 rb_parent = &tmp->pp_rb;
737 ia32_copy_partial_page_list(struct task_struct *p, unsigned long clone_flags)
741 if (clone_flags & CLONE_VM) {
742 atomic_inc(¤t->thread.ppl->pp_count);
743 p->thread.ppl = current->thread.ppl;
745 p->thread.ppl = ia32_init_pp_list();
748 down_write(¤t->mm->mmap_sem);
750 retval = __ia32_copy_pp_list(p->thread.ppl);
752 up_write(¤t->mm->mmap_sem);
759 emulate_mmap (struct file *file, unsigned long start, unsigned long len, int prot, int flags,
762 unsigned long tmp, end, pend, pstart, ret, is_congruent, fudge = 0;
767 pstart = PAGE_START(start);
768 pend = PAGE_ALIGN(end);
770 if (flags & MAP_FIXED) {
771 ia32_set_pp((unsigned int)start, (unsigned int)end, flags);
772 if (start > pstart) {
773 if (flags & MAP_SHARED)
775 "%s(%d): emulate_mmap() can't share head (addr=0x%lx)\n",
776 current->comm, current->pid, start);
777 ret = mmap_subpage(file, start, min(PAGE_ALIGN(start), end), prot, flags,
779 if (IS_ERR((void *) ret))
786 if (flags & MAP_SHARED)
788 "%s(%d): emulate_mmap() can't share tail (end=0x%lx)\n",
789 current->comm, current->pid, end);
790 ret = mmap_subpage(file, max(start, PAGE_START(end)), end, prot, flags,
791 (off + len) - offset_in_page(end));
792 if (IS_ERR((void *) ret))
800 * If a start address was specified, use it if the entire rounded out area
803 if (start && !pstart)
804 fudge = 1; /* handle case of mapping to range (0,PAGE_SIZE) */
805 tmp = arch_get_unmapped_area(file, pstart - fudge, pend - pstart, 0, flags);
808 start = pstart + offset_in_page(off); /* make start congruent with off */
810 pend = PAGE_ALIGN(end);
814 poff = off + (pstart - start); /* note: (pstart - start) may be negative */
815 is_congruent = (flags & MAP_ANONYMOUS) || (offset_in_page(poff) == 0);
817 if ((flags & MAP_SHARED) && !is_congruent)
818 printk(KERN_INFO "%s(%d): emulate_mmap() can't share contents of incongruent mmap "
819 "(addr=0x%lx,off=0x%llx)\n", current->comm, current->pid, start, off);
821 DBG("mmap_body: mapping [0x%lx-0x%lx) %s with poff 0x%llx\n", pstart, pend,
822 is_congruent ? "congruent" : "not congruent", poff);
824 down_write(¤t->mm->mmap_sem);
826 if (!(flags & MAP_ANONYMOUS) && is_congruent)
827 ret = do_mmap(file, pstart, pend - pstart, prot, flags | MAP_FIXED, poff);
829 ret = do_mmap(NULL, pstart, pend - pstart,
830 prot | ((flags & MAP_ANONYMOUS) ? 0 : PROT_WRITE),
831 flags | MAP_FIXED | MAP_ANONYMOUS, 0);
833 up_write(¤t->mm->mmap_sem);
835 if (IS_ERR((void *) ret))
839 /* read the file contents */
840 inode = file->f_dentry->d_inode;
841 if (!inode->i_fop || !file->f_op->read
842 || ((*file->f_op->read)(file, (char __user *) pstart, pend - pstart, &poff)
845 sys_munmap(pstart, pend - pstart);
848 if (!(prot & PROT_WRITE) && sys_mprotect(pstart, pend - pstart, prot) < 0)
852 if (!(flags & MAP_FIXED))
853 ia32_set_pp((unsigned int)start, (unsigned int)end, flags);
858 #endif /* PAGE_SHIFT > IA32_PAGE_SHIFT */
860 static inline unsigned int
861 get_prot32 (unsigned int prot)
863 if (prot & PROT_WRITE)
864 /* on x86, PROT_WRITE implies PROT_READ which implies PROT_EEC */
865 prot |= PROT_READ | PROT_WRITE | PROT_EXEC;
866 else if (prot & (PROT_READ | PROT_EXEC))
867 /* on x86, there is no distinction between PROT_READ and PROT_EXEC */
868 prot |= (PROT_READ | PROT_EXEC);
874 ia32_do_mmap (struct file *file, unsigned long addr, unsigned long len, int prot, int flags,
877 DBG("ia32_do_mmap(file=%p,addr=0x%lx,len=0x%lx,prot=%x,flags=%x,offset=0x%llx)\n",
878 file, addr, len, prot, flags, offset);
880 if (file && (!file->f_op || !file->f_op->mmap))
883 len = IA32_PAGE_ALIGN(len);
887 if (len > IA32_PAGE_OFFSET || addr > IA32_PAGE_OFFSET - len)
889 if (flags & MAP_FIXED)
895 if (OFFSET4K(offset))
898 prot = get_prot32(prot);
900 #if PAGE_SHIFT > IA32_PAGE_SHIFT
901 mutex_lock(&ia32_mmap_mutex);
903 addr = emulate_mmap(file, addr, len, prot, flags, offset);
905 mutex_unlock(&ia32_mmap_mutex);
907 down_write(¤t->mm->mmap_sem);
909 addr = do_mmap(file, addr, len, prot, flags, offset);
911 up_write(¤t->mm->mmap_sem);
913 DBG("ia32_do_mmap: returning 0x%lx\n", addr);
918 * Linux/i386 didn't use to be able to handle more than 4 system call parameters, so these
919 * system calls used a memory block for parameter passing..
922 struct mmap_arg_struct {
932 sys32_mmap (struct mmap_arg_struct __user *arg)
934 struct mmap_arg_struct a;
935 struct file *file = NULL;
939 if (copy_from_user(&a, arg, sizeof(a)))
942 if (OFFSET4K(a.offset))
947 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
948 if (!(flags & MAP_ANONYMOUS)) {
954 addr = ia32_do_mmap(file, a.addr, a.len, a.prot, flags, a.offset);
962 sys32_mmap2 (unsigned int addr, unsigned int len, unsigned int prot, unsigned int flags,
963 unsigned int fd, unsigned int pgoff)
965 struct file *file = NULL;
966 unsigned long retval;
968 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
969 if (!(flags & MAP_ANONYMOUS)) {
975 retval = ia32_do_mmap(file, addr, len, prot, flags,
976 (unsigned long) pgoff << IA32_PAGE_SHIFT);
984 sys32_munmap (unsigned int start, unsigned int len)
986 unsigned int end = start + len;
989 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
990 ret = sys_munmap(start, end - start);
995 end = IA32_PAGE_ALIGN(end);
999 ret = ia32_unset_pp(&start, &end);
1006 mutex_lock(&ia32_mmap_mutex);
1007 ret = sys_munmap(start, end - start);
1008 mutex_unlock(&ia32_mmap_mutex);
1013 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1016 * When mprotect()ing a partial page, we set the permission to the union of the old
1017 * settings and the new settings. In other words, it's only possible to make access to a
1018 * partial page less restrictive.
1021 mprotect_subpage (unsigned long address, int new_prot)
1024 struct vm_area_struct *vma;
1026 if (new_prot == PROT_NONE)
1027 return 0; /* optimize case where nothing changes... */
1028 vma = find_vma(current->mm, address);
1029 old_prot = get_page_prot(vma, address);
1030 return sys_mprotect(address, PAGE_SIZE, new_prot | old_prot);
1033 #endif /* PAGE_SHIFT > IA32_PAGE_SHIFT */
1036 sys32_mprotect (unsigned int start, unsigned int len, int prot)
1038 unsigned int end = start + len;
1039 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1043 prot = get_prot32(prot);
1045 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1046 return sys_mprotect(start, end - start, prot);
1048 if (OFFSET4K(start))
1051 end = IA32_PAGE_ALIGN(end);
1055 retval = ia32_compare_pp(&start, &end);
1060 mutex_lock(&ia32_mmap_mutex);
1062 if (offset_in_page(start)) {
1063 /* start address is 4KB aligned but not page aligned. */
1064 retval = mprotect_subpage(PAGE_START(start), prot);
1068 start = PAGE_ALIGN(start);
1070 goto out; /* retval is already zero... */
1073 if (offset_in_page(end)) {
1074 /* end address is 4KB aligned but not page aligned. */
1075 retval = mprotect_subpage(PAGE_START(end), prot);
1079 end = PAGE_START(end);
1081 retval = sys_mprotect(start, end - start, prot);
1084 mutex_unlock(&ia32_mmap_mutex);
1090 sys32_mremap (unsigned int addr, unsigned int old_len, unsigned int new_len,
1091 unsigned int flags, unsigned int new_addr)
1095 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1096 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1098 unsigned int old_end, new_end;
1103 old_len = IA32_PAGE_ALIGN(old_len);
1104 new_len = IA32_PAGE_ALIGN(new_len);
1105 old_end = addr + old_len;
1106 new_end = addr + new_len;
1111 if ((flags & MREMAP_FIXED) && (OFFSET4K(new_addr)))
1114 if (old_len >= new_len) {
1115 ret = sys32_munmap(addr + new_len, old_len - new_len);
1116 if (ret && old_len != new_len)
1119 if (!(flags & MREMAP_FIXED) || (new_addr == addr))
1124 addr = PAGE_START(addr);
1125 old_len = PAGE_ALIGN(old_end) - addr;
1126 new_len = PAGE_ALIGN(new_end) - addr;
1128 mutex_lock(&ia32_mmap_mutex);
1129 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1130 mutex_unlock(&ia32_mmap_mutex);
1132 if ((ret >= 0) && (old_len < new_len)) {
1133 /* mremap expanded successfully */
1134 ia32_set_pp(old_end, new_end, flags);
1141 sys32_pipe (int __user *fd)
1146 retval = do_pipe(fds);
1149 if (copy_to_user(fd, fds, sizeof(fds)))
1156 get_tv32 (struct timeval *o, struct compat_timeval __user *i)
1158 return (!access_ok(VERIFY_READ, i, sizeof(*i)) ||
1159 (__get_user(o->tv_sec, &i->tv_sec) | __get_user(o->tv_usec, &i->tv_usec)));
1163 put_tv32 (struct compat_timeval __user *o, struct timeval *i)
1165 return (!access_ok(VERIFY_WRITE, o, sizeof(*o)) ||
1166 (__put_user(i->tv_sec, &o->tv_sec) | __put_user(i->tv_usec, &o->tv_usec)));
1169 asmlinkage unsigned long
1170 sys32_alarm (unsigned int seconds)
1172 return alarm_setitimer(seconds);
1175 /* Translations due to time_t size differences. Which affects all
1176 sorts of things, like timeval and itimerval. */
1178 extern struct timezone sys_tz;
1181 sys32_gettimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1185 do_gettimeofday(&ktv);
1186 if (put_tv32(tv, &ktv))
1190 if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
1197 sys32_settimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1200 struct timespec kts;
1201 struct timezone ktz;
1204 if (get_tv32(&ktv, tv))
1206 kts.tv_sec = ktv.tv_sec;
1207 kts.tv_nsec = ktv.tv_usec * 1000;
1210 if (copy_from_user(&ktz, tz, sizeof(ktz)))
1214 return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL);
1217 struct getdents32_callback {
1218 struct compat_dirent __user *current_dir;
1219 struct compat_dirent __user *previous;
1224 struct readdir32_callback {
1225 struct old_linux32_dirent __user * dirent;
1230 filldir32 (void *__buf, const char *name, int namlen, loff_t offset, u64 ino,
1231 unsigned int d_type)
1233 struct compat_dirent __user * dirent;
1234 struct getdents32_callback * buf = (struct getdents32_callback *) __buf;
1235 int reclen = ROUND_UP(offsetof(struct compat_dirent, d_name) + namlen + 1, 4);
1238 buf->error = -EINVAL; /* only used if we fail.. */
1239 if (reclen > buf->count)
1242 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino)
1244 buf->error = -EFAULT; /* only used if we fail.. */
1245 dirent = buf->previous;
1247 if (put_user(offset, &dirent->d_off))
1249 dirent = buf->current_dir;
1250 buf->previous = dirent;
1251 if (put_user(d_ino, &dirent->d_ino)
1252 || put_user(reclen, &dirent->d_reclen)
1253 || copy_to_user(dirent->d_name, name, namlen)
1254 || put_user(0, dirent->d_name + namlen))
1256 dirent = (struct compat_dirent __user *) ((char __user *) dirent + reclen);
1257 buf->current_dir = dirent;
1258 buf->count -= reclen;
1263 sys32_getdents (unsigned int fd, struct compat_dirent __user *dirent, unsigned int count)
1266 struct compat_dirent __user * lastdirent;
1267 struct getdents32_callback buf;
1275 buf.current_dir = dirent;
1276 buf.previous = NULL;
1280 error = vfs_readdir(file, filldir32, &buf);
1284 lastdirent = buf.previous;
1287 if (put_user(file->f_pos, &lastdirent->d_off))
1289 error = count - buf.count;
1299 fillonedir32 (void * __buf, const char * name, int namlen, loff_t offset, u64 ino,
1300 unsigned int d_type)
1302 struct readdir32_callback * buf = (struct readdir32_callback *) __buf;
1303 struct old_linux32_dirent __user * dirent;
1309 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino)
1312 dirent = buf->dirent;
1313 if (put_user(d_ino, &dirent->d_ino)
1314 || put_user(offset, &dirent->d_offset)
1315 || put_user(namlen, &dirent->d_namlen)
1316 || copy_to_user(dirent->d_name, name, namlen)
1317 || put_user(0, dirent->d_name + namlen))
1323 sys32_readdir (unsigned int fd, void __user *dirent, unsigned int count)
1327 struct readdir32_callback buf;
1335 buf.dirent = dirent;
1337 error = vfs_readdir(file, fillonedir32, &buf);
1345 struct sel_arg_struct {
1354 sys32_old_select (struct sel_arg_struct __user *arg)
1356 struct sel_arg_struct a;
1358 if (copy_from_user(&a, arg, sizeof(a)))
1360 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1361 compat_ptr(a.exp), compat_ptr(a.tvp));
1367 #define SEMTIMEDOP 4
1378 sys32_ipc(u32 call, int first, int second, int third, u32 ptr, u32 fifth)
1382 version = call >> 16; /* hack for backward compatibility */
1388 return compat_sys_semtimedop(first, compat_ptr(ptr),
1389 second, compat_ptr(fifth));
1390 /* else fall through for normal semop() */
1392 /* struct sembuf is the same on 32 and 64bit :)) */
1393 return sys_semtimedop(first, compat_ptr(ptr), second,
1396 return sys_semget(first, second, third);
1398 return compat_sys_semctl(first, second, third, compat_ptr(ptr));
1401 return compat_sys_msgsnd(first, second, third, compat_ptr(ptr));
1403 return compat_sys_msgrcv(first, second, fifth, third, version, compat_ptr(ptr));
1405 return sys_msgget((key_t) first, second);
1407 return compat_sys_msgctl(first, second, compat_ptr(ptr));
1410 return compat_sys_shmat(first, second, third, version, compat_ptr(ptr));
1413 return sys_shmdt(compat_ptr(ptr));
1415 return sys_shmget(first, (unsigned)second, third);
1417 return compat_sys_shmctl(first, second, compat_ptr(ptr));
1426 compat_sys_wait4 (compat_pid_t pid, compat_uint_t * stat_addr, int options,
1427 struct compat_rusage *ru);
1430 sys32_waitpid (int pid, unsigned int *stat_addr, int options)
1432 return compat_sys_wait4(pid, stat_addr, options, NULL);
1436 * The order in which registers are stored in the ptrace regs structure
1449 #define PT_ORIG_EAX 11
1457 getreg (struct task_struct *child, int regno)
1459 struct pt_regs *child_regs;
1461 child_regs = task_pt_regs(child);
1462 switch (regno / sizeof(int)) {
1463 case PT_EBX: return child_regs->r11;
1464 case PT_ECX: return child_regs->r9;
1465 case PT_EDX: return child_regs->r10;
1466 case PT_ESI: return child_regs->r14;
1467 case PT_EDI: return child_regs->r15;
1468 case PT_EBP: return child_regs->r13;
1469 case PT_EAX: return child_regs->r8;
1470 case PT_ORIG_EAX: return child_regs->r1; /* see dispatch_to_ia32_handler() */
1471 case PT_EIP: return child_regs->cr_iip;
1472 case PT_UESP: return child_regs->r12;
1473 case PT_EFL: return child->thread.eflag;
1474 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1476 case PT_CS: return __USER_CS;
1478 printk(KERN_ERR "ia32.getreg(): unknown register %d\n", regno);
1485 putreg (struct task_struct *child, int regno, unsigned int value)
1487 struct pt_regs *child_regs;
1489 child_regs = task_pt_regs(child);
1490 switch (regno / sizeof(int)) {
1491 case PT_EBX: child_regs->r11 = value; break;
1492 case PT_ECX: child_regs->r9 = value; break;
1493 case PT_EDX: child_regs->r10 = value; break;
1494 case PT_ESI: child_regs->r14 = value; break;
1495 case PT_EDI: child_regs->r15 = value; break;
1496 case PT_EBP: child_regs->r13 = value; break;
1497 case PT_EAX: child_regs->r8 = value; break;
1498 case PT_ORIG_EAX: child_regs->r1 = value; break;
1499 case PT_EIP: child_regs->cr_iip = value; break;
1500 case PT_UESP: child_regs->r12 = value; break;
1501 case PT_EFL: child->thread.eflag = value; break;
1502 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1503 if (value != __USER_DS)
1505 "ia32.putreg: attempt to set invalid segment register %d = %x\n",
1509 if (value != __USER_CS)
1511 "ia32.putreg: attempt to to set invalid segment register %d = %x\n",
1515 printk(KERN_ERR "ia32.putreg: unknown register %d\n", regno);
1521 put_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1522 struct switch_stack *swp, int tos)
1524 struct _fpreg_ia32 *f;
1527 f = (struct _fpreg_ia32 *)(((unsigned long)buf + 15) & ~15);
1528 if ((regno += tos) >= 8)
1532 ia64f2ia32f(f, &ptp->f8);
1535 ia64f2ia32f(f, &ptp->f9);
1538 ia64f2ia32f(f, &ptp->f10);
1541 ia64f2ia32f(f, &ptp->f11);
1547 ia64f2ia32f(f, &swp->f12 + (regno - 4));
1550 copy_to_user(reg, f, sizeof(*reg));
1554 get_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1555 struct switch_stack *swp, int tos)
1558 if ((regno += tos) >= 8)
1562 copy_from_user(&ptp->f8, reg, sizeof(*reg));
1565 copy_from_user(&ptp->f9, reg, sizeof(*reg));
1568 copy_from_user(&ptp->f10, reg, sizeof(*reg));
1571 copy_from_user(&ptp->f11, reg, sizeof(*reg));
1577 copy_from_user(&swp->f12 + (regno - 4), reg, sizeof(*reg));
1584 save_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1586 struct switch_stack *swp;
1587 struct pt_regs *ptp;
1590 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1593 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1594 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1595 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1596 __put_user(tsk->thread.fir, &save->fip);
1597 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1598 __put_user(tsk->thread.fdr, &save->foo);
1599 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1602 * Stack frames start with 16-bytes of temp space
1604 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1605 ptp = task_pt_regs(tsk);
1606 tos = (tsk->thread.fsr >> 11) & 7;
1607 for (i = 0; i < 8; i++)
1608 put_fpreg(i, &save->st_space[i], ptp, swp, tos);
1613 restore_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1615 struct switch_stack *swp;
1616 struct pt_regs *ptp;
1618 unsigned int fsrlo, fsrhi, num32;
1620 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1623 __get_user(num32, (unsigned int __user *)&save->cwd);
1624 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1625 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1626 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1627 num32 = (fsrhi << 16) | fsrlo;
1628 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1629 __get_user(num32, (unsigned int __user *)&save->fip);
1630 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1631 __get_user(num32, (unsigned int __user *)&save->foo);
1632 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1635 * Stack frames start with 16-bytes of temp space
1637 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1638 ptp = task_pt_regs(tsk);
1639 tos = (tsk->thread.fsr >> 11) & 7;
1640 for (i = 0; i < 8; i++)
1641 get_fpreg(i, &save->st_space[i], ptp, swp, tos);
1646 save_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1648 struct switch_stack *swp;
1649 struct pt_regs *ptp;
1651 unsigned long mxcsr=0;
1652 unsigned long num128[2];
1654 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1657 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1658 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1659 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1660 __put_user(tsk->thread.fir, &save->fip);
1661 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1662 __put_user(tsk->thread.fdr, &save->foo);
1663 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1666 * Stack frames start with 16-bytes of temp space
1668 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1669 ptp = task_pt_regs(tsk);
1670 tos = (tsk->thread.fsr >> 11) & 7;
1671 for (i = 0; i < 8; i++)
1672 put_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1674 mxcsr = ((tsk->thread.fcr>>32) & 0xff80) | ((tsk->thread.fsr>>32) & 0x3f);
1675 __put_user(mxcsr & 0xffff, &save->mxcsr);
1676 for (i = 0; i < 8; i++) {
1677 memcpy(&(num128[0]), &(swp->f16) + i*2, sizeof(unsigned long));
1678 memcpy(&(num128[1]), &(swp->f17) + i*2, sizeof(unsigned long));
1679 copy_to_user(&save->xmm_space[0] + 4*i, num128, sizeof(struct _xmmreg_ia32));
1685 restore_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1687 struct switch_stack *swp;
1688 struct pt_regs *ptp;
1690 unsigned int fsrlo, fsrhi, num32;
1692 unsigned long num64;
1693 unsigned long num128[2];
1695 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1698 __get_user(num32, (unsigned int __user *)&save->cwd);
1699 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1700 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1701 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1702 num32 = (fsrhi << 16) | fsrlo;
1703 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1704 __get_user(num32, (unsigned int __user *)&save->fip);
1705 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1706 __get_user(num32, (unsigned int __user *)&save->foo);
1707 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1710 * Stack frames start with 16-bytes of temp space
1712 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1713 ptp = task_pt_regs(tsk);
1714 tos = (tsk->thread.fsr >> 11) & 7;
1715 for (i = 0; i < 8; i++)
1716 get_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1718 __get_user(mxcsr, (unsigned int __user *)&save->mxcsr);
1719 num64 = mxcsr & 0xff10;
1720 tsk->thread.fcr = (tsk->thread.fcr & (~0xff1000000000UL)) | (num64<<32);
1721 num64 = mxcsr & 0x3f;
1722 tsk->thread.fsr = (tsk->thread.fsr & (~0x3f00000000UL)) | (num64<<32);
1724 for (i = 0; i < 8; i++) {
1725 copy_from_user(num128, &save->xmm_space[0] + 4*i, sizeof(struct _xmmreg_ia32));
1726 memcpy(&(swp->f16) + i*2, &(num128[0]), sizeof(unsigned long));
1727 memcpy(&(swp->f17) + i*2, &(num128[1]), sizeof(unsigned long));
1734 sys32_ptrace (int request, pid_t pid, unsigned int addr, unsigned int data)
1736 struct task_struct *child;
1737 unsigned int value, tmp;
1741 if (request == PTRACE_TRACEME) {
1742 ret = ptrace_traceme();
1746 child = ptrace_get_task_struct(pid);
1747 if (IS_ERR(child)) {
1748 ret = PTR_ERR(child);
1752 if (request == PTRACE_ATTACH) {
1753 ret = sys_ptrace(request, pid, addr, data);
1757 ret = ptrace_check_attach(child, request == PTRACE_KILL);
1762 case PTRACE_PEEKTEXT:
1763 case PTRACE_PEEKDATA: /* read word at location addr */
1764 ret = ia32_peek(child, addr, &value);
1766 ret = put_user(value, (unsigned int __user *) compat_ptr(data));
1771 case PTRACE_POKETEXT:
1772 case PTRACE_POKEDATA: /* write the word at location addr */
1773 ret = ia32_poke(child, addr, data);
1776 case PTRACE_PEEKUSR: /* read word at addr in USER area */
1778 if ((addr & 3) || addr > 17*sizeof(int))
1781 tmp = getreg(child, addr);
1782 if (!put_user(tmp, (unsigned int __user *) compat_ptr(data)))
1786 case PTRACE_POKEUSR: /* write word at addr in USER area */
1788 if ((addr & 3) || addr > 17*sizeof(int))
1791 putreg(child, addr, data);
1795 case IA32_PTRACE_GETREGS:
1796 if (!access_ok(VERIFY_WRITE, compat_ptr(data), 17*sizeof(int))) {
1800 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1801 put_user(getreg(child, i), (unsigned int __user *) compat_ptr(data));
1802 data += sizeof(int);
1807 case IA32_PTRACE_SETREGS:
1808 if (!access_ok(VERIFY_READ, compat_ptr(data), 17*sizeof(int))) {
1812 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1813 get_user(tmp, (unsigned int __user *) compat_ptr(data));
1814 putreg(child, i, tmp);
1815 data += sizeof(int);
1820 case IA32_PTRACE_GETFPREGS:
1821 ret = save_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1825 case IA32_PTRACE_GETFPXREGS:
1826 ret = save_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1830 case IA32_PTRACE_SETFPREGS:
1831 ret = restore_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1835 case IA32_PTRACE_SETFPXREGS:
1836 ret = restore_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1841 case PTRACE_GETEVENTMSG:
1842 ret = put_user(child->ptrace_message, (unsigned int __user *) compat_ptr(data));
1846 case PTRACE_SYSCALL: /* continue, stop after next syscall */
1847 case PTRACE_CONT: /* restart after signal. */
1849 case PTRACE_SINGLESTEP: /* execute chile for one instruction */
1850 case PTRACE_DETACH: /* detach a process */
1851 ret = sys_ptrace(request, pid, addr, data);
1855 ret = ptrace_request(child, request, addr, data);
1860 put_task_struct(child);
1869 unsigned int ss_flags;
1870 unsigned int ss_size;
1874 sys32_sigaltstack (ia32_stack_t __user *uss32, ia32_stack_t __user *uoss32,
1875 long arg2, long arg3, long arg4, long arg5, long arg6,
1876 long arg7, struct pt_regs pt)
1881 mm_segment_t old_fs = get_fs();
1884 if (copy_from_user(&buf32, uss32, sizeof(ia32_stack_t)))
1886 uss.ss_sp = (void __user *) (long) buf32.ss_sp;
1887 uss.ss_flags = buf32.ss_flags;
1888 /* MINSIGSTKSZ is different for ia32 vs ia64. We lie here to pass the
1889 check and set it to the user requested value later */
1890 if ((buf32.ss_flags != SS_DISABLE) && (buf32.ss_size < MINSIGSTKSZ_IA32)) {
1894 uss.ss_size = MINSIGSTKSZ;
1897 ret = do_sigaltstack(uss32 ? (stack_t __user *) &uss : NULL,
1898 (stack_t __user *) &uoss, pt.r12);
1899 current->sas_ss_size = buf32.ss_size;
1905 buf32.ss_sp = (long __user) uoss.ss_sp;
1906 buf32.ss_flags = uoss.ss_flags;
1907 buf32.ss_size = uoss.ss_size;
1908 if (copy_to_user(uoss32, &buf32, sizeof(ia32_stack_t)))
1917 current->state = TASK_INTERRUPTIBLE;
1919 return -ERESTARTNOHAND;
1923 sys32_msync (unsigned int start, unsigned int len, int flags)
1927 if (OFFSET4K(start))
1929 addr = PAGE_START(start);
1930 return sys_msync(addr, len + (start - addr), flags);
1936 unsigned int oldval;
1937 unsigned int oldlenp;
1938 unsigned int newval;
1939 unsigned int newlen;
1940 unsigned int __unused[4];
1943 #ifdef CONFIG_SYSCTL
1945 sys32_sysctl (struct sysctl32 __user *args)
1947 struct sysctl32 a32;
1948 mm_segment_t old_fs = get_fs ();
1949 void __user *oldvalp, *newvalp;
1954 if (copy_from_user(&a32, args, sizeof(a32)))
1958 * We need to pre-validate these because we have to disable address checking
1959 * before calling do_sysctl() because of OLDLEN but we can't run the risk of the
1960 * user specifying bad addresses here. Well, since we're dealing with 32 bit
1961 * addresses, we KNOW that access_ok() will always succeed, so this is an
1962 * expensive NOP, but so what...
1964 namep = (int __user *) compat_ptr(a32.name);
1965 oldvalp = compat_ptr(a32.oldval);
1966 newvalp = compat_ptr(a32.newval);
1968 if ((oldvalp && get_user(oldlen, (int __user *) compat_ptr(a32.oldlenp)))
1969 || !access_ok(VERIFY_WRITE, namep, 0)
1970 || !access_ok(VERIFY_WRITE, oldvalp, 0)
1971 || !access_ok(VERIFY_WRITE, newvalp, 0))
1976 ret = do_sysctl(namep, a32.nlen, oldvalp, (size_t __user *) &oldlen,
1977 newvalp, (size_t) a32.newlen);
1981 if (oldvalp && put_user (oldlen, (int __user *) compat_ptr(a32.oldlenp)))
1989 sys32_newuname (struct new_utsname __user *name)
1991 int ret = sys_newuname(name);
1994 if (copy_to_user(name->machine, "i686\0\0\0", 8))
2000 sys32_getresuid16 (u16 __user *ruid, u16 __user *euid, u16 __user *suid)
2004 mm_segment_t old_fs = get_fs();
2007 ret = sys_getresuid((uid_t __user *) &a, (uid_t __user *) &b, (uid_t __user *) &c);
2010 if (put_user(a, ruid) || put_user(b, euid) || put_user(c, suid))
2016 sys32_getresgid16 (u16 __user *rgid, u16 __user *egid, u16 __user *sgid)
2020 mm_segment_t old_fs = get_fs();
2023 ret = sys_getresgid((gid_t __user *) &a, (gid_t __user *) &b, (gid_t __user *) &c);
2029 return put_user(a, rgid) | put_user(b, egid) | put_user(c, sgid);
2033 sys32_lseek (unsigned int fd, int offset, unsigned int whence)
2035 /* Sign-extension of "offset" is important here... */
2036 return sys_lseek(fd, offset, whence);
2040 groups16_to_user(short __user *grouplist, struct group_info *group_info)
2045 for (i = 0; i < group_info->ngroups; i++) {
2046 group = (short)GROUP_AT(group_info, i);
2047 if (put_user(group, grouplist+i))
2055 groups16_from_user(struct group_info *group_info, short __user *grouplist)
2060 for (i = 0; i < group_info->ngroups; i++) {
2061 if (get_user(group, grouplist+i))
2063 GROUP_AT(group_info, i) = (gid_t)group;
2070 sys32_getgroups16 (int gidsetsize, short __user *grouplist)
2077 get_group_info(current->group_info);
2078 i = current->group_info->ngroups;
2080 if (i > gidsetsize) {
2084 if (groups16_to_user(grouplist, current->group_info)) {
2090 put_group_info(current->group_info);
2095 sys32_setgroups16 (int gidsetsize, short __user *grouplist)
2097 struct group_info *group_info;
2100 if (!capable(CAP_SETGID))
2102 if ((unsigned)gidsetsize > NGROUPS_MAX)
2105 group_info = groups_alloc(gidsetsize);
2108 retval = groups16_from_user(group_info, grouplist);
2110 put_group_info(group_info);
2114 retval = set_current_groups(group_info);
2115 put_group_info(group_info);
2121 sys32_truncate64 (unsigned int path, unsigned int len_lo, unsigned int len_hi)
2123 return sys_truncate(compat_ptr(path), ((unsigned long) len_hi << 32) | len_lo);
2127 sys32_ftruncate64 (int fd, unsigned int len_lo, unsigned int len_hi)
2129 return sys_ftruncate(fd, ((unsigned long) len_hi << 32) | len_lo);
2133 putstat64 (struct stat64 __user *ubuf, struct kstat *kbuf)
2138 if (clear_user(ubuf, sizeof(*ubuf)))
2141 hdev = huge_encode_dev(kbuf->dev);
2142 err = __put_user(hdev, (u32 __user*)&ubuf->st_dev);
2143 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_dev) + 1);
2144 err |= __put_user(kbuf->ino, &ubuf->__st_ino);
2145 err |= __put_user(kbuf->ino, &ubuf->st_ino_lo);
2146 err |= __put_user(kbuf->ino >> 32, &ubuf->st_ino_hi);
2147 err |= __put_user(kbuf->mode, &ubuf->st_mode);
2148 err |= __put_user(kbuf->nlink, &ubuf->st_nlink);
2149 err |= __put_user(kbuf->uid, &ubuf->st_uid);
2150 err |= __put_user(kbuf->gid, &ubuf->st_gid);
2151 hdev = huge_encode_dev(kbuf->rdev);
2152 err = __put_user(hdev, (u32 __user*)&ubuf->st_rdev);
2153 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_rdev) + 1);
2154 err |= __put_user(kbuf->size, &ubuf->st_size_lo);
2155 err |= __put_user((kbuf->size >> 32), &ubuf->st_size_hi);
2156 err |= __put_user(kbuf->atime.tv_sec, &ubuf->st_atime);
2157 err |= __put_user(kbuf->atime.tv_nsec, &ubuf->st_atime_nsec);
2158 err |= __put_user(kbuf->mtime.tv_sec, &ubuf->st_mtime);
2159 err |= __put_user(kbuf->mtime.tv_nsec, &ubuf->st_mtime_nsec);
2160 err |= __put_user(kbuf->ctime.tv_sec, &ubuf->st_ctime);
2161 err |= __put_user(kbuf->ctime.tv_nsec, &ubuf->st_ctime_nsec);
2162 err |= __put_user(kbuf->blksize, &ubuf->st_blksize);
2163 err |= __put_user(kbuf->blocks, &ubuf->st_blocks);
2168 sys32_stat64 (char __user *filename, struct stat64 __user *statbuf)
2171 long ret = vfs_stat(filename, &s);
2173 ret = putstat64(statbuf, &s);
2178 sys32_lstat64 (char __user *filename, struct stat64 __user *statbuf)
2181 long ret = vfs_lstat(filename, &s);
2183 ret = putstat64(statbuf, &s);
2188 sys32_fstat64 (unsigned int fd, struct stat64 __user *statbuf)
2191 long ret = vfs_fstat(fd, &s);
2193 ret = putstat64(statbuf, &s);
2215 sys32_sysinfo (struct sysinfo32 __user *info)
2220 mm_segment_t old_fs = get_fs();
2223 ret = sys_sysinfo((struct sysinfo __user *) &s);
2225 /* Check to see if any memory value is too large for 32-bit and
2226 * scale down if needed.
2228 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2229 while (s.mem_unit < PAGE_SIZE) {
2233 s.totalram >>= bitcount;
2234 s.freeram >>= bitcount;
2235 s.sharedram >>= bitcount;
2236 s.bufferram >>= bitcount;
2237 s.totalswap >>= bitcount;
2238 s.freeswap >>= bitcount;
2239 s.totalhigh >>= bitcount;
2240 s.freehigh >>= bitcount;
2243 if (!access_ok(VERIFY_WRITE, info, sizeof(*info)))
2246 err = __put_user(s.uptime, &info->uptime);
2247 err |= __put_user(s.loads[0], &info->loads[0]);
2248 err |= __put_user(s.loads[1], &info->loads[1]);
2249 err |= __put_user(s.loads[2], &info->loads[2]);
2250 err |= __put_user(s.totalram, &info->totalram);
2251 err |= __put_user(s.freeram, &info->freeram);
2252 err |= __put_user(s.sharedram, &info->sharedram);
2253 err |= __put_user(s.bufferram, &info->bufferram);
2254 err |= __put_user(s.totalswap, &info->totalswap);
2255 err |= __put_user(s.freeswap, &info->freeswap);
2256 err |= __put_user(s.procs, &info->procs);
2257 err |= __put_user (s.totalhigh, &info->totalhigh);
2258 err |= __put_user (s.freehigh, &info->freehigh);
2259 err |= __put_user (s.mem_unit, &info->mem_unit);
2266 sys32_sched_rr_get_interval (pid_t pid, struct compat_timespec __user *interval)
2268 mm_segment_t old_fs = get_fs();
2273 ret = sys_sched_rr_get_interval(pid, (struct timespec __user *) &t);
2275 if (put_compat_timespec(&t, interval))
2281 sys32_pread (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2283 return sys_pread64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2287 sys32_pwrite (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2289 return sys_pwrite64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2293 sys32_sendfile (int out_fd, int in_fd, int __user *offset, unsigned int count)
2295 mm_segment_t old_fs = get_fs();
2299 if (offset && get_user(of, offset))
2303 ret = sys_sendfile(out_fd, in_fd, offset ? (off_t __user *) &of : NULL, count);
2306 if (offset && put_user(of, offset))
2313 sys32_personality (unsigned int personality)
2317 if (current->personality == PER_LINUX32 && personality == PER_LINUX)
2318 personality = PER_LINUX32;
2319 ret = sys_personality(personality);
2320 if (ret == PER_LINUX32)
2325 asmlinkage unsigned long
2326 sys32_brk (unsigned int brk)
2328 unsigned long ret, obrk;
2329 struct mm_struct *mm = current->mm;
2334 clear_user(compat_ptr(ret), PAGE_ALIGN(ret) - ret);
2338 /* Structure for ia32 emulation on ia64 */
2339 struct epoll_event32
2346 sys32_epoll_ctl(int epfd, int op, int fd, struct epoll_event32 __user *event)
2348 mm_segment_t old_fs = get_fs();
2349 struct epoll_event event64;
2353 if (!access_ok(VERIFY_READ, event, sizeof(struct epoll_event32)))
2356 __get_user(event64.events, &event->events);
2357 __get_user(data_halfword, &event->data[0]);
2358 event64.data = data_halfword;
2359 __get_user(data_halfword, &event->data[1]);
2360 event64.data |= (u64)data_halfword << 32;
2363 error = sys_epoll_ctl(epfd, op, fd, (struct epoll_event __user *) &event64);
2370 sys32_epoll_wait(int epfd, struct epoll_event32 __user * events, int maxevents,
2373 struct epoll_event *events64 = NULL;
2374 mm_segment_t old_fs = get_fs();
2375 int numevents, size;
2377 int do_free_pages = 0;
2379 if (maxevents <= 0) {
2383 /* Verify that the area passed by the user is writeable */
2384 if (!access_ok(VERIFY_WRITE, events, maxevents * sizeof(struct epoll_event32)))
2388 * Allocate space for the intermediate copy. If the space needed
2389 * is large enough to cause kmalloc to fail, then try again with
2392 size = maxevents * sizeof(struct epoll_event);
2393 events64 = kmalloc(size, GFP_KERNEL);
2394 if (events64 == NULL) {
2395 events64 = (struct epoll_event *)
2396 __get_free_pages(GFP_KERNEL, get_order(size));
2397 if (events64 == NULL)
2402 /* Do the system call */
2403 set_fs(KERNEL_DS); /* copy_to/from_user should work on kernel mem*/
2404 numevents = sys_epoll_wait(epfd, (struct epoll_event __user *) events64,
2405 maxevents, timeout);
2408 /* Don't modify userspace memory if we're returning an error */
2409 if (numevents > 0) {
2410 /* Translate the 64-bit structures back into the 32-bit
2412 for (evt_idx = 0; evt_idx < numevents; evt_idx++) {
2413 __put_user(events64[evt_idx].events,
2414 &events[evt_idx].events);
2415 __put_user((u32)events64[evt_idx].data,
2416 &events[evt_idx].data[0]);
2417 __put_user((u32)(events64[evt_idx].data >> 32),
2418 &events[evt_idx].data[1]);
2423 free_pages((unsigned long) events64, get_order(size));
2430 * Get a yet unused TLS descriptor index.
2435 struct thread_struct *t = ¤t->thread;
2438 for (idx = 0; idx < GDT_ENTRY_TLS_ENTRIES; idx++)
2439 if (desc_empty(t->tls_array + idx))
2440 return idx + GDT_ENTRY_TLS_MIN;
2445 * Set a given TLS descriptor:
2448 sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2450 struct thread_struct *t = ¤t->thread;
2451 struct ia32_user_desc info;
2452 struct desc_struct *desc;
2455 if (copy_from_user(&info, u_info, sizeof(info)))
2457 idx = info.entry_number;
2460 * index -1 means the kernel should try to find and allocate an empty descriptor:
2463 idx = get_free_idx();
2466 if (put_user(idx, &u_info->entry_number))
2470 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2473 desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN;
2475 cpu = smp_processor_id();
2477 if (LDT_empty(&info)) {
2481 desc->a = LDT_entry_a(&info);
2482 desc->b = LDT_entry_b(&info);
2489 * Get the current Thread-Local Storage area:
2492 #define GET_BASE(desc) ( \
2493 (((desc)->a >> 16) & 0x0000ffff) | \
2494 (((desc)->b << 16) & 0x00ff0000) | \
2495 ( (desc)->b & 0xff000000) )
2497 #define GET_LIMIT(desc) ( \
2498 ((desc)->a & 0x0ffff) | \
2499 ((desc)->b & 0xf0000) )
2501 #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
2502 #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
2503 #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
2504 #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
2505 #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
2506 #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
2509 sys32_get_thread_area (struct ia32_user_desc __user *u_info)
2511 struct ia32_user_desc info;
2512 struct desc_struct *desc;
2515 if (get_user(idx, &u_info->entry_number))
2517 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2520 desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
2522 info.entry_number = idx;
2523 info.base_addr = GET_BASE(desc);
2524 info.limit = GET_LIMIT(desc);
2525 info.seg_32bit = GET_32BIT(desc);
2526 info.contents = GET_CONTENTS(desc);
2527 info.read_exec_only = !GET_WRITABLE(desc);
2528 info.limit_in_pages = GET_LIMIT_PAGES(desc);
2529 info.seg_not_present = !GET_PRESENT(desc);
2530 info.useable = GET_USEABLE(desc);
2532 if (copy_to_user(u_info, &info, sizeof(info)))
2537 long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
2538 __u32 len_low, __u32 len_high, int advice)
2540 return sys_fadvise64_64(fd,
2541 (((u64)offset_high)<<32) | offset_low,
2542 (((u64)len_high)<<32) | len_low,
2546 #ifdef NOTYET /* UNTESTED FOR IA64 FROM HERE DOWN */
2548 asmlinkage long sys32_setreuid(compat_uid_t ruid, compat_uid_t euid)
2552 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2553 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2554 return sys_setreuid(sruid, seuid);
2558 sys32_setresuid(compat_uid_t ruid, compat_uid_t euid,
2561 uid_t sruid, seuid, ssuid;
2563 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2564 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2565 ssuid = (suid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)suid);
2566 return sys_setresuid(sruid, seuid, ssuid);
2570 sys32_setregid(compat_gid_t rgid, compat_gid_t egid)
2574 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2575 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2576 return sys_setregid(srgid, segid);
2580 sys32_setresgid(compat_gid_t rgid, compat_gid_t egid,
2583 gid_t srgid, segid, ssgid;
2585 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2586 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2587 ssgid = (sgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)sgid);
2588 return sys_setresgid(srgid, segid, ssgid);