Linux-2.6.12-rc2
[linux-2.6/kvm.git] / arch / ia64 / ia32 / sys_ia32.c
blob247a21c64aea33bb41ad5aa591323abcfb81aaaf
1 /*
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
14 * environment.
17 #include <linux/config.h>
18 #include <linux/kernel.h>
19 #include <linux/syscalls.h>
20 #include <linux/sysctl.h>
21 #include <linux/sched.h>
22 #include <linux/fs.h>
23 #include <linux/file.h>
24 #include <linux/signal.h>
25 #include <linux/resource.h>
26 #include <linux/times.h>
27 #include <linux/utsname.h>
28 #include <linux/timex.h>
29 #include <linux/smp.h>
30 #include <linux/smp_lock.h>
31 #include <linux/sem.h>
32 #include <linux/msg.h>
33 #include <linux/mm.h>
34 #include <linux/shm.h>
35 #include <linux/slab.h>
36 #include <linux/uio.h>
37 #include <linux/nfs_fs.h>
38 #include <linux/quota.h>
39 #include <linux/sunrpc/svc.h>
40 #include <linux/nfsd/nfsd.h>
41 #include <linux/nfsd/cache.h>
42 #include <linux/nfsd/xdr.h>
43 #include <linux/nfsd/syscall.h>
44 #include <linux/poll.h>
45 #include <linux/eventpoll.h>
46 #include <linux/personality.h>
47 #include <linux/ptrace.h>
48 #include <linux/stat.h>
49 #include <linux/ipc.h>
50 #include <linux/compat.h>
51 #include <linux/vfs.h>
52 #include <linux/mman.h>
54 #include <asm/intrinsics.h>
55 #include <asm/semaphore.h>
56 #include <asm/types.h>
57 #include <asm/uaccess.h>
58 #include <asm/unistd.h>
60 #include "ia32priv.h"
62 #include <net/scm.h>
63 #include <net/sock.h>
65 #define DEBUG 0
67 #if DEBUG
68 # define DBG(fmt...) printk(KERN_DEBUG fmt)
69 #else
70 # define DBG(fmt...)
71 #endif
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
84 * while doing so.
86 /* XXX make per-mm: */
87 static DECLARE_MUTEX(ia32_mmap_sem);
89 asmlinkage long
90 sys32_execve (char __user *name, compat_uptr_t __user *argv, compat_uptr_t __user *envp,
91 struct pt_regs *regs)
93 long error;
94 char *filename;
95 unsigned long old_map_base, old_task_size, tssd;
97 filename = getname(name);
98 error = PTR_ERR(filename);
99 if (IS_ERR(filename))
100 return error;
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);
113 putname(filename);
115 if (error < 0) {
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;
123 return error;
126 int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
128 int err;
130 if ((u64) stat->size > MAX_NON_LFS ||
131 !old_valid_dev(stat->dev) ||
132 !old_valid_dev(stat->rdev))
133 return -EOVERFLOW;
135 if (clear_user(ubuf, sizeof(*ubuf)))
136 return -EFAULT;
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);
154 return err;
157 #if PAGE_SHIFT > IA32_PAGE_SHIFT
160 static int
161 get_page_prot (struct vm_area_struct *vma, unsigned long addr)
163 int prot = 0;
165 if (!vma || vma->vm_start > addr)
166 return 0;
168 if (vma->vm_flags & VM_READ)
169 prot |= PROT_READ;
170 if (vma->vm_flags & VM_WRITE)
171 prot |= PROT_WRITE;
172 if (vma->vm_flags & VM_EXEC)
173 prot |= PROT_EXEC;
174 return prot;
178 * Map a subpage by creating an anonymous page that contains the union of the old page and
179 * the subpage.
181 static unsigned long
182 mmap_subpage (struct file *file, unsigned long start, unsigned long end, int prot, int flags,
183 loff_t off)
185 void *page = NULL;
186 struct inode *inode;
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)) {
198 ret = -EFAULT;
199 goto out;
201 goto skip_mmap;
204 page = (void *) get_zeroed_page(GFP_KERNEL);
205 if (!page)
206 return -ENOMEM;
208 if (old_prot)
209 copy_from_user(page, (void __user *) PAGE_START(start), PAGE_SIZE);
211 down_write(&current->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(&current->mm->mmap_sem);
218 if (IS_ERR((void *) ret))
219 goto out;
221 if (old_prot) {
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))
237 ret = -EINVAL;
238 goto out;
242 skip_mmap:
243 if (!(prot & PROT_WRITE))
244 ret = sys_mprotect(PAGE_START(start), PAGE_SIZE, prot | old_prot);
245 out:
246 if (page)
247 free_page((unsigned long) page);
248 return ret;
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)
264 return p;
265 p->pp_head = NULL;
266 p->ppl_rb = RB_ROOT;
267 p->pp_hint = NULL;
268 atomic_set(&p->pp_count, 1);
269 return p;
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;
286 pp = ppl->pp_hint;
287 if (pp && pp->base == start)
288 return pp;
290 __rb_link = &ppl->ppl_rb.rb_node;
291 rb_prev = __rb_parent = NULL;
293 while (*__rb_link) {
294 __rb_parent = *__rb_link;
295 pp = rb_entry(__rb_parent, struct partial_page, pp_rb);
297 if (pp->base == start) {
298 ppl->pp_hint = pp;
299 return pp;
300 } else if (pp->base < start) {
301 rb_prev = __rb_parent;
302 __rb_link = &__rb_parent->rb_right;
303 } else {
304 __rb_link = &__rb_parent->rb_left;
308 *rb_link = __rb_link;
309 *rb_parent = __rb_parent;
310 *pprev = NULL;
311 if (rb_prev)
312 *pprev = rb_entry(rb_prev, struct partial_page, pp_rb);
313 return NULL;
317 * insert @pp into @ppl.
319 static void
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)
324 /* link list */
325 if (prev) {
326 pp->next = prev->next;
327 prev->next = pp;
328 } else {
329 ppl->pp_head = pp;
330 if (rb_parent)
331 pp->next = rb_entry(rb_parent,
332 struct partial_page, pp_rb);
333 else
334 pp->next = NULL;
337 /* link rb */
338 rb_link_node(&pp->pp_rb, rb_parent, rb_link);
339 rb_insert_color(&pp->pp_rb, &ppl->ppl_rb);
341 ppl->pp_hint = pp;
345 * delete @pp from partial page list @ppl.
347 static void
348 __ia32_delete_pp(struct partial_page_list *ppl, struct partial_page *pp,
349 struct partial_page *prev)
351 if (prev) {
352 prev->next = pp->next;
353 if (ppl->pp_hint == pp)
354 ppl->pp_hint = prev;
355 } else {
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);
368 if (prev)
369 return rb_entry(prev, struct partial_page, pp_rb);
370 else
371 return NULL;
375 * Delete partial pages with address between @start and @end.
376 * @start and @end are page aligned.
378 static void
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;
384 if (start >= end)
385 return;
387 pp = __ia32_find_pp(current->thread.ppl, start, &prev,
388 &rb_link, &rb_parent);
389 if (pp)
390 prev = __pp_prev(pp);
391 else {
392 if (prev)
393 pp = prev->next;
394 else
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);
401 pp = tmp;
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.
409 static int
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;
419 if (end_bit == 0)
420 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
421 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
422 &rb_link, &rb_parent);
423 if (pp) {
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,
428 * then delete it.
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));
434 return 0;
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)
445 return 0;
448 /* new a partial_page */
449 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
450 if (!pp)
451 return -ENOMEM;
452 pp->base = pstart;
453 pp->bitmap = 0;
454 for (i=start_bit; i<end_bit; i++)
455 set_bit(i, &(pp->bitmap));
456 pp->next = NULL;
457 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
458 return 0;
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().
466 static void
467 ia32_set_pp(unsigned int start, unsigned int end, int flags)
469 down_write(&current->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);
481 } else {
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(&current->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,
494 * else return 0;
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.
499 static int
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;
510 if (end_bit == 0)
511 end_bit = PAGE_SIZE / IA32_PAGE_SIZE;
513 pp = __ia32_find_pp(current->thread.ppl, pstart, &prev,
514 &rb_link, &rb_parent);
515 if (pp) {
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));
520 return 1;
522 return 0;
525 vma = find_vma(current->mm, pstart);
526 if (!vma || vma->vm_start > pstart) {
527 return -ENOMEM;
530 /* new a partial_page */
531 pp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
532 if (!pp)
533 return -ENOMEM;
534 pp->base = pstart;
535 pp->bitmap = 0;
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));
540 pp->next = NULL;
541 __ia32_insert_pp(current->thread.ppl, pp, prev, rb_link, rb_parent);
542 return 0;
546 * Delete pp between PAGE_ALIGN(start) and PAGE_START(end) by calling
547 * __ia32_delete_pp_range(). Unset possible partial pages by calling
548 * __ia32_unset_pp().
549 * The returned value see __ia32_unset_pp().
551 static int
552 ia32_unset_pp(unsigned int *startp, unsigned int *endp)
554 unsigned int start = *startp, end = *endp;
555 int ret = 0;
557 down_write(&current->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);
563 if (ret == 1) {
564 *startp = PAGE_START(start);
565 *endp = PAGE_ALIGN(end);
567 if (ret == 0) {
568 /* to shortcut sys_munmap() in sys32_munmap() */
569 *startp = PAGE_START(start);
570 *endp = PAGE_START(end);
572 } else {
573 if (offset_in_page(start)) {
574 ret = __ia32_unset_pp(start, PAGE_ALIGN(start));
575 if (ret == 1)
576 *startp = PAGE_START(start);
577 if (ret == 0)
578 *startp = PAGE_ALIGN(start);
579 if (ret < 0)
580 goto out;
582 if (offset_in_page(end)) {
583 ret = __ia32_unset_pp(PAGE_START(end), end);
584 if (ret == 1)
585 *endp = PAGE_ALIGN(end);
586 if (ret == 0)
587 *endp = PAGE_START(end);
591 out:
592 up_write(&current->mm->mmap_sem);
593 return ret;
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.
600 static int
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);
612 if (!pp)
613 return 1;
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 */
622 return -ENOMEM;
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 */
627 else
628 return 0; /* no next range */
629 } else
630 return 1;
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
642 * bitmap = 00000101.
644 static int
645 ia32_compare_pp(unsigned int *startp, unsigned int *endp)
647 unsigned int start = *startp, end = *endp;
648 int retval = 0;
650 down_write(&current->mm->mmap_sem);
652 if (end < PAGE_ALIGN(start)) {
653 retval = __ia32_compare_pp(start, end);
654 if (retval == 0) {
655 *startp = PAGE_START(start);
656 *endp = PAGE_ALIGN(end);
658 } else {
659 if (offset_in_page(start)) {
660 retval = __ia32_compare_pp(start,
661 PAGE_ALIGN(start));
662 if (retval == 0)
663 *startp = PAGE_START(start);
664 if (retval < 0)
665 goto out;
667 if (offset_in_page(end)) {
668 retval = __ia32_compare_pp(PAGE_START(end), end);
669 if (retval == 0)
670 *endp = PAGE_ALIGN(end);
674 out:
675 up_write(&current->mm->mmap_sem);
676 return retval;
679 static void
680 __ia32_drop_pp_list(struct partial_page_list *ppl)
682 struct partial_page *pp = ppl->pp_head;
684 while (pp) {
685 struct partial_page *next = pp->next;
686 kmem_cache_free(partial_page_cachep, pp);
687 pp = next;
690 kfree(ppl);
693 void
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).
705 static int
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;
711 ppl->pp_head = NULL;
712 ppl->pp_hint = NULL;
713 ppl->ppl_rb = RB_ROOT;
714 rb_link = &ppl->ppl_rb.rb_node;
715 rb_parent = NULL;
716 prev = NULL;
718 for (pp = current->thread.ppl->pp_head; pp; pp = pp->next) {
719 tmp = kmem_cache_alloc(partial_page_cachep, GFP_KERNEL);
720 if (!tmp)
721 return -ENOMEM;
722 *tmp = *pp;
723 __ia32_insert_pp(ppl, tmp, prev, rb_link, rb_parent);
724 prev = tmp;
725 rb_link = &tmp->pp_rb.rb_right;
726 rb_parent = &tmp->pp_rb;
728 return 0;
732 ia32_copy_partial_page_list(struct task_struct *p, unsigned long clone_flags)
734 int retval = 0;
736 if (clone_flags & CLONE_VM) {
737 atomic_inc(&current->thread.ppl->pp_count);
738 p->thread.ppl = current->thread.ppl;
739 } else {
740 p->thread.ppl = ia32_init_pp_list();
741 if (!p->thread.ppl)
742 return -ENOMEM;
743 down_write(&current->mm->mmap_sem);
745 retval = __ia32_copy_pp_list(p->thread.ppl);
747 up_write(&current->mm->mmap_sem);
750 return retval;
753 static unsigned long
754 emulate_mmap (struct file *file, unsigned long start, unsigned long len, int prot, int flags,
755 loff_t off)
757 unsigned long tmp, end, pend, pstart, ret, is_congruent, fudge = 0;
758 struct inode *inode;
759 loff_t poff;
761 end = start + len;
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)
769 printk(KERN_INFO
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,
773 off);
774 if (IS_ERR((void *) ret))
775 return ret;
776 pstart += PAGE_SIZE;
777 if (pstart >= pend)
778 goto out; /* done */
780 if (end < pend) {
781 if (flags & MAP_SHARED)
782 printk(KERN_INFO
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))
788 return ret;
789 pend -= PAGE_SIZE;
790 if (pstart >= pend)
791 goto out; /* done */
793 } else {
795 * If a start address was specified, use it if the entire rounded out area
796 * is available.
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);
801 if (tmp != pstart) {
802 pstart = tmp;
803 start = pstart + offset_in_page(off); /* make start congruent with off */
804 end = start + len;
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(&current->mm->mmap_sem);
821 if (!(flags & MAP_ANONYMOUS) && is_congruent)
822 ret = do_mmap(file, pstart, pend - pstart, prot, flags | MAP_FIXED, poff);
823 else
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(&current->mm->mmap_sem);
830 if (IS_ERR((void *) ret))
831 return ret;
833 if (!is_congruent) {
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)
838 < 0))
840 sys_munmap(pstart, pend - pstart);
841 return -EINVAL;
843 if (!(prot & PROT_WRITE) && sys_mprotect(pstart, pend - pstart, prot) < 0)
844 return -EINVAL;
847 if (!(flags & MAP_FIXED))
848 ia32_set_pp((unsigned int)start, (unsigned int)end, flags);
849 out:
850 return start;
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);
865 return prot;
868 unsigned long
869 ia32_do_mmap (struct file *file, unsigned long addr, unsigned long len, int prot, int flags,
870 loff_t offset)
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))
876 return -ENODEV;
878 len = IA32_PAGE_ALIGN(len);
879 if (len == 0)
880 return addr;
882 if (len > IA32_PAGE_OFFSET || addr > IA32_PAGE_OFFSET - len)
884 if (flags & MAP_FIXED)
885 return -ENOMEM;
886 else
887 return -EINVAL;
890 if (OFFSET4K(offset))
891 return -EINVAL;
893 prot = get_prot32(prot);
895 #if PAGE_SHIFT > IA32_PAGE_SHIFT
896 down(&ia32_mmap_sem);
898 addr = emulate_mmap(file, addr, len, prot, flags, offset);
900 up(&ia32_mmap_sem);
901 #else
902 down_write(&current->mm->mmap_sem);
904 addr = do_mmap(file, addr, len, prot, flags, offset);
906 up_write(&current->mm->mmap_sem);
907 #endif
908 DBG("ia32_do_mmap: returning 0x%lx\n", addr);
909 return 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 {
918 unsigned int addr;
919 unsigned int len;
920 unsigned int prot;
921 unsigned int flags;
922 unsigned int fd;
923 unsigned int offset;
926 asmlinkage long
927 sys32_mmap (struct mmap_arg_struct __user *arg)
929 struct mmap_arg_struct a;
930 struct file *file = NULL;
931 unsigned long addr;
932 int flags;
934 if (copy_from_user(&a, arg, sizeof(a)))
935 return -EFAULT;
937 if (OFFSET4K(a.offset))
938 return -EINVAL;
940 flags = a.flags;
942 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
943 if (!(flags & MAP_ANONYMOUS)) {
944 file = fget(a.fd);
945 if (!file)
946 return -EBADF;
949 addr = ia32_do_mmap(file, a.addr, a.len, a.prot, flags, a.offset);
951 if (file)
952 fput(file);
953 return addr;
956 asmlinkage long
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)) {
965 file = fget(fd);
966 if (!file)
967 return -EBADF;
970 retval = ia32_do_mmap(file, addr, len, prot, flags,
971 (unsigned long) pgoff << IA32_PAGE_SHIFT);
973 if (file)
974 fput(file);
975 return retval;
978 asmlinkage long
979 sys32_munmap (unsigned int start, unsigned int len)
981 unsigned int end = start + len;
982 long ret;
984 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
985 ret = sys_munmap(start, end - start);
986 #else
987 if (OFFSET4K(start))
988 return -EINVAL;
990 end = IA32_PAGE_ALIGN(end);
991 if (start >= end)
992 return -EINVAL;
994 ret = ia32_unset_pp(&start, &end);
995 if (ret < 0)
996 return ret;
998 if (start >= end)
999 return 0;
1001 down(&ia32_mmap_sem);
1003 ret = sys_munmap(start, end - start);
1005 up(&ia32_mmap_sem);
1006 #endif
1007 return ret;
1010 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1013 * When mprotect()ing a partial page, we set the permission to the union of the old
1014 * settings and the new settings. In other words, it's only possible to make access to a
1015 * partial page less restrictive.
1017 static long
1018 mprotect_subpage (unsigned long address, int new_prot)
1020 int old_prot;
1021 struct vm_area_struct *vma;
1023 if (new_prot == PROT_NONE)
1024 return 0; /* optimize case where nothing changes... */
1025 vma = find_vma(current->mm, address);
1026 old_prot = get_page_prot(vma, address);
1027 return sys_mprotect(address, PAGE_SIZE, new_prot | old_prot);
1030 #endif /* PAGE_SHIFT > IA32_PAGE_SHIFT */
1032 asmlinkage long
1033 sys32_mprotect (unsigned int start, unsigned int len, int prot)
1035 unsigned int end = start + len;
1036 #if PAGE_SHIFT > IA32_PAGE_SHIFT
1037 long retval = 0;
1038 #endif
1040 prot = get_prot32(prot);
1042 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1043 return sys_mprotect(start, end - start, prot);
1044 #else
1045 if (OFFSET4K(start))
1046 return -EINVAL;
1048 end = IA32_PAGE_ALIGN(end);
1049 if (end < start)
1050 return -EINVAL;
1052 retval = ia32_compare_pp(&start, &end);
1054 if (retval < 0)
1055 return retval;
1057 down(&ia32_mmap_sem);
1059 if (offset_in_page(start)) {
1060 /* start address is 4KB aligned but not page aligned. */
1061 retval = mprotect_subpage(PAGE_START(start), prot);
1062 if (retval < 0)
1063 goto out;
1065 start = PAGE_ALIGN(start);
1066 if (start >= end)
1067 goto out; /* retval is already zero... */
1070 if (offset_in_page(end)) {
1071 /* end address is 4KB aligned but not page aligned. */
1072 retval = mprotect_subpage(PAGE_START(end), prot);
1073 if (retval < 0)
1074 goto out;
1076 end = PAGE_START(end);
1078 retval = sys_mprotect(start, end - start, prot);
1080 out:
1081 up(&ia32_mmap_sem);
1082 return retval;
1083 #endif
1086 asmlinkage long
1087 sys32_mremap (unsigned int addr, unsigned int old_len, unsigned int new_len,
1088 unsigned int flags, unsigned int new_addr)
1090 long ret;
1092 #if PAGE_SHIFT <= IA32_PAGE_SHIFT
1093 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1094 #else
1095 unsigned int old_end, new_end;
1097 if (OFFSET4K(addr))
1098 return -EINVAL;
1100 old_len = IA32_PAGE_ALIGN(old_len);
1101 new_len = IA32_PAGE_ALIGN(new_len);
1102 old_end = addr + old_len;
1103 new_end = addr + new_len;
1105 if (!new_len)
1106 return -EINVAL;
1108 if ((flags & MREMAP_FIXED) && (OFFSET4K(new_addr)))
1109 return -EINVAL;
1111 if (old_len >= new_len) {
1112 ret = sys32_munmap(addr + new_len, old_len - new_len);
1113 if (ret && old_len != new_len)
1114 return ret;
1115 ret = addr;
1116 if (!(flags & MREMAP_FIXED) || (new_addr == addr))
1117 return ret;
1118 old_len = new_len;
1121 addr = PAGE_START(addr);
1122 old_len = PAGE_ALIGN(old_end) - addr;
1123 new_len = PAGE_ALIGN(new_end) - addr;
1125 down(&ia32_mmap_sem);
1127 ret = sys_mremap(addr, old_len, new_len, flags, new_addr);
1129 up(&ia32_mmap_sem);
1131 if ((ret >= 0) && (old_len < new_len)) {
1132 /* mremap expanded successfully */
1133 ia32_set_pp(old_end, new_end, flags);
1135 #endif
1136 return ret;
1139 asmlinkage long
1140 sys32_pipe (int __user *fd)
1142 int retval;
1143 int fds[2];
1145 retval = do_pipe(fds);
1146 if (retval)
1147 goto out;
1148 if (copy_to_user(fd, fds, sizeof(fds)))
1149 retval = -EFAULT;
1150 out:
1151 return retval;
1154 static inline long
1155 get_tv32 (struct timeval *o, struct compat_timeval __user *i)
1157 return (!access_ok(VERIFY_READ, i, sizeof(*i)) ||
1158 (__get_user(o->tv_sec, &i->tv_sec) | __get_user(o->tv_usec, &i->tv_usec)));
1161 static inline long
1162 put_tv32 (struct compat_timeval __user *o, struct timeval *i)
1164 return (!access_ok(VERIFY_WRITE, o, sizeof(*o)) ||
1165 (__put_user(i->tv_sec, &o->tv_sec) | __put_user(i->tv_usec, &o->tv_usec)));
1168 asmlinkage unsigned long
1169 sys32_alarm (unsigned int seconds)
1171 struct itimerval it_new, it_old;
1172 unsigned int oldalarm;
1174 it_new.it_interval.tv_sec = it_new.it_interval.tv_usec = 0;
1175 it_new.it_value.tv_sec = seconds;
1176 it_new.it_value.tv_usec = 0;
1177 do_setitimer(ITIMER_REAL, &it_new, &it_old);
1178 oldalarm = it_old.it_value.tv_sec;
1179 /* ehhh.. We can't return 0 if we have an alarm pending.. */
1180 /* And we'd better return too much than too little anyway */
1181 if (it_old.it_value.tv_usec)
1182 oldalarm++;
1183 return oldalarm;
1186 /* Translations due to time_t size differences. Which affects all
1187 sorts of things, like timeval and itimerval. */
1189 extern struct timezone sys_tz;
1191 asmlinkage long
1192 sys32_gettimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1194 if (tv) {
1195 struct timeval ktv;
1196 do_gettimeofday(&ktv);
1197 if (put_tv32(tv, &ktv))
1198 return -EFAULT;
1200 if (tz) {
1201 if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
1202 return -EFAULT;
1204 return 0;
1207 asmlinkage long
1208 sys32_settimeofday (struct compat_timeval __user *tv, struct timezone __user *tz)
1210 struct timeval ktv;
1211 struct timespec kts;
1212 struct timezone ktz;
1214 if (tv) {
1215 if (get_tv32(&ktv, tv))
1216 return -EFAULT;
1217 kts.tv_sec = ktv.tv_sec;
1218 kts.tv_nsec = ktv.tv_usec * 1000;
1220 if (tz) {
1221 if (copy_from_user(&ktz, tz, sizeof(ktz)))
1222 return -EFAULT;
1225 return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL);
1228 struct getdents32_callback {
1229 struct compat_dirent __user *current_dir;
1230 struct compat_dirent __user *previous;
1231 int count;
1232 int error;
1235 struct readdir32_callback {
1236 struct old_linux32_dirent __user * dirent;
1237 int count;
1240 static int
1241 filldir32 (void *__buf, const char *name, int namlen, loff_t offset, ino_t ino,
1242 unsigned int d_type)
1244 struct compat_dirent __user * dirent;
1245 struct getdents32_callback * buf = (struct getdents32_callback *) __buf;
1246 int reclen = ROUND_UP(offsetof(struct compat_dirent, d_name) + namlen + 1, 4);
1248 buf->error = -EINVAL; /* only used if we fail.. */
1249 if (reclen > buf->count)
1250 return -EINVAL;
1251 buf->error = -EFAULT; /* only used if we fail.. */
1252 dirent = buf->previous;
1253 if (dirent)
1254 if (put_user(offset, &dirent->d_off))
1255 return -EFAULT;
1256 dirent = buf->current_dir;
1257 buf->previous = dirent;
1258 if (put_user(ino, &dirent->d_ino)
1259 || put_user(reclen, &dirent->d_reclen)
1260 || copy_to_user(dirent->d_name, name, namlen)
1261 || put_user(0, dirent->d_name + namlen))
1262 return -EFAULT;
1263 dirent = (struct compat_dirent __user *) ((char __user *) dirent + reclen);
1264 buf->current_dir = dirent;
1265 buf->count -= reclen;
1266 return 0;
1269 asmlinkage long
1270 sys32_getdents (unsigned int fd, struct compat_dirent __user *dirent, unsigned int count)
1272 struct file * file;
1273 struct compat_dirent __user * lastdirent;
1274 struct getdents32_callback buf;
1275 int error;
1277 error = -EBADF;
1278 file = fget(fd);
1279 if (!file)
1280 goto out;
1282 buf.current_dir = dirent;
1283 buf.previous = NULL;
1284 buf.count = count;
1285 buf.error = 0;
1287 error = vfs_readdir(file, filldir32, &buf);
1288 if (error < 0)
1289 goto out_putf;
1290 error = buf.error;
1291 lastdirent = buf.previous;
1292 if (lastdirent) {
1293 error = -EINVAL;
1294 if (put_user(file->f_pos, &lastdirent->d_off))
1295 goto out_putf;
1296 error = count - buf.count;
1299 out_putf:
1300 fput(file);
1301 out:
1302 return error;
1305 static int
1306 fillonedir32 (void * __buf, const char * name, int namlen, loff_t offset, ino_t ino,
1307 unsigned int d_type)
1309 struct readdir32_callback * buf = (struct readdir32_callback *) __buf;
1310 struct old_linux32_dirent __user * dirent;
1312 if (buf->count)
1313 return -EINVAL;
1314 buf->count++;
1315 dirent = buf->dirent;
1316 if (put_user(ino, &dirent->d_ino)
1317 || put_user(offset, &dirent->d_offset)
1318 || put_user(namlen, &dirent->d_namlen)
1319 || copy_to_user(dirent->d_name, name, namlen)
1320 || put_user(0, dirent->d_name + namlen))
1321 return -EFAULT;
1322 return 0;
1325 asmlinkage long
1326 sys32_readdir (unsigned int fd, void __user *dirent, unsigned int count)
1328 int error;
1329 struct file * file;
1330 struct readdir32_callback buf;
1332 error = -EBADF;
1333 file = fget(fd);
1334 if (!file)
1335 goto out;
1337 buf.count = 0;
1338 buf.dirent = dirent;
1340 error = vfs_readdir(file, fillonedir32, &buf);
1341 if (error >= 0)
1342 error = buf.count;
1343 fput(file);
1344 out:
1345 return error;
1348 struct sel_arg_struct {
1349 unsigned int n;
1350 unsigned int inp;
1351 unsigned int outp;
1352 unsigned int exp;
1353 unsigned int tvp;
1356 asmlinkage long
1357 sys32_old_select (struct sel_arg_struct __user *arg)
1359 struct sel_arg_struct a;
1361 if (copy_from_user(&a, arg, sizeof(a)))
1362 return -EFAULT;
1363 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1364 compat_ptr(a.exp), compat_ptr(a.tvp));
1367 #define SEMOP 1
1368 #define SEMGET 2
1369 #define SEMCTL 3
1370 #define SEMTIMEDOP 4
1371 #define MSGSND 11
1372 #define MSGRCV 12
1373 #define MSGGET 13
1374 #define MSGCTL 14
1375 #define SHMAT 21
1376 #define SHMDT 22
1377 #define SHMGET 23
1378 #define SHMCTL 24
1380 asmlinkage long
1381 sys32_ipc(u32 call, int first, int second, int third, u32 ptr, u32 fifth)
1383 int version;
1385 version = call >> 16; /* hack for backward compatibility */
1386 call &= 0xffff;
1388 switch (call) {
1389 case SEMTIMEDOP:
1390 if (fifth)
1391 return compat_sys_semtimedop(first, compat_ptr(ptr),
1392 second, compat_ptr(fifth));
1393 /* else fall through for normal semop() */
1394 case SEMOP:
1395 /* struct sembuf is the same on 32 and 64bit :)) */
1396 return sys_semtimedop(first, compat_ptr(ptr), second,
1397 NULL);
1398 case SEMGET:
1399 return sys_semget(first, second, third);
1400 case SEMCTL:
1401 return compat_sys_semctl(first, second, third, compat_ptr(ptr));
1403 case MSGSND:
1404 return compat_sys_msgsnd(first, second, third, compat_ptr(ptr));
1405 case MSGRCV:
1406 return compat_sys_msgrcv(first, second, fifth, third, version, compat_ptr(ptr));
1407 case MSGGET:
1408 return sys_msgget((key_t) first, second);
1409 case MSGCTL:
1410 return compat_sys_msgctl(first, second, compat_ptr(ptr));
1412 case SHMAT:
1413 return compat_sys_shmat(first, second, third, version, compat_ptr(ptr));
1414 break;
1415 case SHMDT:
1416 return sys_shmdt(compat_ptr(ptr));
1417 case SHMGET:
1418 return sys_shmget(first, (unsigned)second, third);
1419 case SHMCTL:
1420 return compat_sys_shmctl(first, second, compat_ptr(ptr));
1422 default:
1423 return -ENOSYS;
1425 return -EINVAL;
1428 asmlinkage long
1429 compat_sys_wait4 (compat_pid_t pid, compat_uint_t * stat_addr, int options,
1430 struct compat_rusage *ru);
1432 asmlinkage long
1433 sys32_waitpid (int pid, unsigned int *stat_addr, int options)
1435 return compat_sys_wait4(pid, stat_addr, options, NULL);
1438 static unsigned int
1439 ia32_peek (struct task_struct *child, unsigned long addr, unsigned int *val)
1441 size_t copied;
1442 unsigned int ret;
1444 copied = access_process_vm(child, addr, val, sizeof(*val), 0);
1445 return (copied != sizeof(ret)) ? -EIO : 0;
1448 static unsigned int
1449 ia32_poke (struct task_struct *child, unsigned long addr, unsigned int val)
1452 if (access_process_vm(child, addr, &val, sizeof(val), 1) != sizeof(val))
1453 return -EIO;
1454 return 0;
1458 * The order in which registers are stored in the ptrace regs structure
1460 #define PT_EBX 0
1461 #define PT_ECX 1
1462 #define PT_EDX 2
1463 #define PT_ESI 3
1464 #define PT_EDI 4
1465 #define PT_EBP 5
1466 #define PT_EAX 6
1467 #define PT_DS 7
1468 #define PT_ES 8
1469 #define PT_FS 9
1470 #define PT_GS 10
1471 #define PT_ORIG_EAX 11
1472 #define PT_EIP 12
1473 #define PT_CS 13
1474 #define PT_EFL 14
1475 #define PT_UESP 15
1476 #define PT_SS 16
1478 static unsigned int
1479 getreg (struct task_struct *child, int regno)
1481 struct pt_regs *child_regs;
1483 child_regs = ia64_task_regs(child);
1484 switch (regno / sizeof(int)) {
1485 case PT_EBX: return child_regs->r11;
1486 case PT_ECX: return child_regs->r9;
1487 case PT_EDX: return child_regs->r10;
1488 case PT_ESI: return child_regs->r14;
1489 case PT_EDI: return child_regs->r15;
1490 case PT_EBP: return child_regs->r13;
1491 case PT_EAX: return child_regs->r8;
1492 case PT_ORIG_EAX: return child_regs->r1; /* see dispatch_to_ia32_handler() */
1493 case PT_EIP: return child_regs->cr_iip;
1494 case PT_UESP: return child_regs->r12;
1495 case PT_EFL: return child->thread.eflag;
1496 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1497 return __USER_DS;
1498 case PT_CS: return __USER_CS;
1499 default:
1500 printk(KERN_ERR "ia32.getreg(): unknown register %d\n", regno);
1501 break;
1503 return 0;
1506 static void
1507 putreg (struct task_struct *child, int regno, unsigned int value)
1509 struct pt_regs *child_regs;
1511 child_regs = ia64_task_regs(child);
1512 switch (regno / sizeof(int)) {
1513 case PT_EBX: child_regs->r11 = value; break;
1514 case PT_ECX: child_regs->r9 = value; break;
1515 case PT_EDX: child_regs->r10 = value; break;
1516 case PT_ESI: child_regs->r14 = value; break;
1517 case PT_EDI: child_regs->r15 = value; break;
1518 case PT_EBP: child_regs->r13 = value; break;
1519 case PT_EAX: child_regs->r8 = value; break;
1520 case PT_ORIG_EAX: child_regs->r1 = value; break;
1521 case PT_EIP: child_regs->cr_iip = value; break;
1522 case PT_UESP: child_regs->r12 = value; break;
1523 case PT_EFL: child->thread.eflag = value; break;
1524 case PT_DS: case PT_ES: case PT_FS: case PT_GS: case PT_SS:
1525 if (value != __USER_DS)
1526 printk(KERN_ERR
1527 "ia32.putreg: attempt to set invalid segment register %d = %x\n",
1528 regno, value);
1529 break;
1530 case PT_CS:
1531 if (value != __USER_CS)
1532 printk(KERN_ERR
1533 "ia32.putreg: attempt to to set invalid segment register %d = %x\n",
1534 regno, value);
1535 break;
1536 default:
1537 printk(KERN_ERR "ia32.putreg: unknown register %d\n", regno);
1538 break;
1542 static void
1543 put_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1544 struct switch_stack *swp, int tos)
1546 struct _fpreg_ia32 *f;
1547 char buf[32];
1549 f = (struct _fpreg_ia32 *)(((unsigned long)buf + 15) & ~15);
1550 if ((regno += tos) >= 8)
1551 regno -= 8;
1552 switch (regno) {
1553 case 0:
1554 ia64f2ia32f(f, &ptp->f8);
1555 break;
1556 case 1:
1557 ia64f2ia32f(f, &ptp->f9);
1558 break;
1559 case 2:
1560 ia64f2ia32f(f, &ptp->f10);
1561 break;
1562 case 3:
1563 ia64f2ia32f(f, &ptp->f11);
1564 break;
1565 case 4:
1566 case 5:
1567 case 6:
1568 case 7:
1569 ia64f2ia32f(f, &swp->f12 + (regno - 4));
1570 break;
1572 copy_to_user(reg, f, sizeof(*reg));
1575 static void
1576 get_fpreg (int regno, struct _fpreg_ia32 __user *reg, struct pt_regs *ptp,
1577 struct switch_stack *swp, int tos)
1580 if ((regno += tos) >= 8)
1581 regno -= 8;
1582 switch (regno) {
1583 case 0:
1584 copy_from_user(&ptp->f8, reg, sizeof(*reg));
1585 break;
1586 case 1:
1587 copy_from_user(&ptp->f9, reg, sizeof(*reg));
1588 break;
1589 case 2:
1590 copy_from_user(&ptp->f10, reg, sizeof(*reg));
1591 break;
1592 case 3:
1593 copy_from_user(&ptp->f11, reg, sizeof(*reg));
1594 break;
1595 case 4:
1596 case 5:
1597 case 6:
1598 case 7:
1599 copy_from_user(&swp->f12 + (regno - 4), reg, sizeof(*reg));
1600 break;
1602 return;
1606 save_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1608 struct switch_stack *swp;
1609 struct pt_regs *ptp;
1610 int i, tos;
1612 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1613 return -EFAULT;
1615 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1616 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1617 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1618 __put_user(tsk->thread.fir, &save->fip);
1619 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1620 __put_user(tsk->thread.fdr, &save->foo);
1621 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1624 * Stack frames start with 16-bytes of temp space
1626 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1627 ptp = ia64_task_regs(tsk);
1628 tos = (tsk->thread.fsr >> 11) & 7;
1629 for (i = 0; i < 8; i++)
1630 put_fpreg(i, &save->st_space[i], ptp, swp, tos);
1631 return 0;
1634 static int
1635 restore_ia32_fpstate (struct task_struct *tsk, struct ia32_user_i387_struct __user *save)
1637 struct switch_stack *swp;
1638 struct pt_regs *ptp;
1639 int i, tos;
1640 unsigned int fsrlo, fsrhi, num32;
1642 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1643 return(-EFAULT);
1645 __get_user(num32, (unsigned int __user *)&save->cwd);
1646 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1647 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1648 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1649 num32 = (fsrhi << 16) | fsrlo;
1650 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1651 __get_user(num32, (unsigned int __user *)&save->fip);
1652 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1653 __get_user(num32, (unsigned int __user *)&save->foo);
1654 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1657 * Stack frames start with 16-bytes of temp space
1659 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1660 ptp = ia64_task_regs(tsk);
1661 tos = (tsk->thread.fsr >> 11) & 7;
1662 for (i = 0; i < 8; i++)
1663 get_fpreg(i, &save->st_space[i], ptp, swp, tos);
1664 return 0;
1668 save_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1670 struct switch_stack *swp;
1671 struct pt_regs *ptp;
1672 int i, tos;
1673 unsigned long mxcsr=0;
1674 unsigned long num128[2];
1676 if (!access_ok(VERIFY_WRITE, save, sizeof(*save)))
1677 return -EFAULT;
1679 __put_user(tsk->thread.fcr & 0xffff, &save->cwd);
1680 __put_user(tsk->thread.fsr & 0xffff, &save->swd);
1681 __put_user((tsk->thread.fsr>>16) & 0xffff, &save->twd);
1682 __put_user(tsk->thread.fir, &save->fip);
1683 __put_user((tsk->thread.fir>>32) & 0xffff, &save->fcs);
1684 __put_user(tsk->thread.fdr, &save->foo);
1685 __put_user((tsk->thread.fdr>>32) & 0xffff, &save->fos);
1688 * Stack frames start with 16-bytes of temp space
1690 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1691 ptp = ia64_task_regs(tsk);
1692 tos = (tsk->thread.fsr >> 11) & 7;
1693 for (i = 0; i < 8; i++)
1694 put_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1696 mxcsr = ((tsk->thread.fcr>>32) & 0xff80) | ((tsk->thread.fsr>>32) & 0x3f);
1697 __put_user(mxcsr & 0xffff, &save->mxcsr);
1698 for (i = 0; i < 8; i++) {
1699 memcpy(&(num128[0]), &(swp->f16) + i*2, sizeof(unsigned long));
1700 memcpy(&(num128[1]), &(swp->f17) + i*2, sizeof(unsigned long));
1701 copy_to_user(&save->xmm_space[0] + 4*i, num128, sizeof(struct _xmmreg_ia32));
1703 return 0;
1706 static int
1707 restore_ia32_fpxstate (struct task_struct *tsk, struct ia32_user_fxsr_struct __user *save)
1709 struct switch_stack *swp;
1710 struct pt_regs *ptp;
1711 int i, tos;
1712 unsigned int fsrlo, fsrhi, num32;
1713 int mxcsr;
1714 unsigned long num64;
1715 unsigned long num128[2];
1717 if (!access_ok(VERIFY_READ, save, sizeof(*save)))
1718 return(-EFAULT);
1720 __get_user(num32, (unsigned int __user *)&save->cwd);
1721 tsk->thread.fcr = (tsk->thread.fcr & (~0x1f3f)) | (num32 & 0x1f3f);
1722 __get_user(fsrlo, (unsigned int __user *)&save->swd);
1723 __get_user(fsrhi, (unsigned int __user *)&save->twd);
1724 num32 = (fsrhi << 16) | fsrlo;
1725 tsk->thread.fsr = (tsk->thread.fsr & (~0xffffffff)) | num32;
1726 __get_user(num32, (unsigned int __user *)&save->fip);
1727 tsk->thread.fir = (tsk->thread.fir & (~0xffffffff)) | num32;
1728 __get_user(num32, (unsigned int __user *)&save->foo);
1729 tsk->thread.fdr = (tsk->thread.fdr & (~0xffffffff)) | num32;
1732 * Stack frames start with 16-bytes of temp space
1734 swp = (struct switch_stack *)(tsk->thread.ksp + 16);
1735 ptp = ia64_task_regs(tsk);
1736 tos = (tsk->thread.fsr >> 11) & 7;
1737 for (i = 0; i < 8; i++)
1738 get_fpreg(i, (struct _fpreg_ia32 __user *)&save->st_space[4*i], ptp, swp, tos);
1740 __get_user(mxcsr, (unsigned int __user *)&save->mxcsr);
1741 num64 = mxcsr & 0xff10;
1742 tsk->thread.fcr = (tsk->thread.fcr & (~0xff1000000000UL)) | (num64<<32);
1743 num64 = mxcsr & 0x3f;
1744 tsk->thread.fsr = (tsk->thread.fsr & (~0x3f00000000UL)) | (num64<<32);
1746 for (i = 0; i < 8; i++) {
1747 copy_from_user(num128, &save->xmm_space[0] + 4*i, sizeof(struct _xmmreg_ia32));
1748 memcpy(&(swp->f16) + i*2, &(num128[0]), sizeof(unsigned long));
1749 memcpy(&(swp->f17) + i*2, &(num128[1]), sizeof(unsigned long));
1751 return 0;
1754 asmlinkage long
1755 sys32_ptrace (int request, pid_t pid, unsigned int addr, unsigned int data)
1757 struct task_struct *child;
1758 unsigned int value, tmp;
1759 long i, ret;
1761 lock_kernel();
1762 if (request == PTRACE_TRACEME) {
1763 ret = sys_ptrace(request, pid, addr, data);
1764 goto out;
1767 ret = -ESRCH;
1768 read_lock(&tasklist_lock);
1769 child = find_task_by_pid(pid);
1770 if (child)
1771 get_task_struct(child);
1772 read_unlock(&tasklist_lock);
1773 if (!child)
1774 goto out;
1775 ret = -EPERM;
1776 if (pid == 1) /* no messing around with init! */
1777 goto out_tsk;
1779 if (request == PTRACE_ATTACH) {
1780 ret = sys_ptrace(request, pid, addr, data);
1781 goto out_tsk;
1784 ret = ptrace_check_attach(child, request == PTRACE_KILL);
1785 if (ret < 0)
1786 goto out_tsk;
1788 switch (request) {
1789 case PTRACE_PEEKTEXT:
1790 case PTRACE_PEEKDATA: /* read word at location addr */
1791 ret = ia32_peek(child, addr, &value);
1792 if (ret == 0)
1793 ret = put_user(value, (unsigned int __user *) compat_ptr(data));
1794 else
1795 ret = -EIO;
1796 goto out_tsk;
1798 case PTRACE_POKETEXT:
1799 case PTRACE_POKEDATA: /* write the word at location addr */
1800 ret = ia32_poke(child, addr, data);
1801 goto out_tsk;
1803 case PTRACE_PEEKUSR: /* read word at addr in USER area */
1804 ret = -EIO;
1805 if ((addr & 3) || addr > 17*sizeof(int))
1806 break;
1808 tmp = getreg(child, addr);
1809 if (!put_user(tmp, (unsigned int __user *) compat_ptr(data)))
1810 ret = 0;
1811 break;
1813 case PTRACE_POKEUSR: /* write word at addr in USER area */
1814 ret = -EIO;
1815 if ((addr & 3) || addr > 17*sizeof(int))
1816 break;
1818 putreg(child, addr, data);
1819 ret = 0;
1820 break;
1822 case IA32_PTRACE_GETREGS:
1823 if (!access_ok(VERIFY_WRITE, compat_ptr(data), 17*sizeof(int))) {
1824 ret = -EIO;
1825 break;
1827 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1828 put_user(getreg(child, i), (unsigned int __user *) compat_ptr(data));
1829 data += sizeof(int);
1831 ret = 0;
1832 break;
1834 case IA32_PTRACE_SETREGS:
1835 if (!access_ok(VERIFY_READ, compat_ptr(data), 17*sizeof(int))) {
1836 ret = -EIO;
1837 break;
1839 for (i = 0; i < (int) (17*sizeof(int)); i += sizeof(int) ) {
1840 get_user(tmp, (unsigned int __user *) compat_ptr(data));
1841 putreg(child, i, tmp);
1842 data += sizeof(int);
1844 ret = 0;
1845 break;
1847 case IA32_PTRACE_GETFPREGS:
1848 ret = save_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1849 compat_ptr(data));
1850 break;
1852 case IA32_PTRACE_GETFPXREGS:
1853 ret = save_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1854 compat_ptr(data));
1855 break;
1857 case IA32_PTRACE_SETFPREGS:
1858 ret = restore_ia32_fpstate(child, (struct ia32_user_i387_struct __user *)
1859 compat_ptr(data));
1860 break;
1862 case IA32_PTRACE_SETFPXREGS:
1863 ret = restore_ia32_fpxstate(child, (struct ia32_user_fxsr_struct __user *)
1864 compat_ptr(data));
1865 break;
1867 case PTRACE_GETEVENTMSG:
1868 ret = put_user(child->ptrace_message, (unsigned int __user *) compat_ptr(data));
1869 break;
1871 case PTRACE_SYSCALL: /* continue, stop after next syscall */
1872 case PTRACE_CONT: /* restart after signal. */
1873 case PTRACE_KILL:
1874 case PTRACE_SINGLESTEP: /* execute chile for one instruction */
1875 case PTRACE_DETACH: /* detach a process */
1876 ret = sys_ptrace(request, pid, addr, data);
1877 break;
1879 default:
1880 ret = ptrace_request(child, request, addr, data);
1881 break;
1884 out_tsk:
1885 put_task_struct(child);
1886 out:
1887 unlock_kernel();
1888 return ret;
1891 typedef struct {
1892 unsigned int ss_sp;
1893 unsigned int ss_flags;
1894 unsigned int ss_size;
1895 } ia32_stack_t;
1897 asmlinkage long
1898 sys32_sigaltstack (ia32_stack_t __user *uss32, ia32_stack_t __user *uoss32,
1899 long arg2, long arg3, long arg4, long arg5, long arg6,
1900 long arg7, struct pt_regs pt)
1902 stack_t uss, uoss;
1903 ia32_stack_t buf32;
1904 int ret;
1905 mm_segment_t old_fs = get_fs();
1907 if (uss32) {
1908 if (copy_from_user(&buf32, uss32, sizeof(ia32_stack_t)))
1909 return -EFAULT;
1910 uss.ss_sp = (void __user *) (long) buf32.ss_sp;
1911 uss.ss_flags = buf32.ss_flags;
1912 /* MINSIGSTKSZ is different for ia32 vs ia64. We lie here to pass the
1913 check and set it to the user requested value later */
1914 if ((buf32.ss_flags != SS_DISABLE) && (buf32.ss_size < MINSIGSTKSZ_IA32)) {
1915 ret = -ENOMEM;
1916 goto out;
1918 uss.ss_size = MINSIGSTKSZ;
1920 set_fs(KERNEL_DS);
1921 ret = do_sigaltstack(uss32 ? (stack_t __user *) &uss : NULL,
1922 (stack_t __user *) &uoss, pt.r12);
1923 current->sas_ss_size = buf32.ss_size;
1924 set_fs(old_fs);
1925 out:
1926 if (ret < 0)
1927 return(ret);
1928 if (uoss32) {
1929 buf32.ss_sp = (long __user) uoss.ss_sp;
1930 buf32.ss_flags = uoss.ss_flags;
1931 buf32.ss_size = uoss.ss_size;
1932 if (copy_to_user(uoss32, &buf32, sizeof(ia32_stack_t)))
1933 return -EFAULT;
1935 return ret;
1938 asmlinkage int
1939 sys32_pause (void)
1941 current->state = TASK_INTERRUPTIBLE;
1942 schedule();
1943 return -ERESTARTNOHAND;
1946 asmlinkage int
1947 sys32_msync (unsigned int start, unsigned int len, int flags)
1949 unsigned int addr;
1951 if (OFFSET4K(start))
1952 return -EINVAL;
1953 addr = PAGE_START(start);
1954 return sys_msync(addr, len + (start - addr), flags);
1957 struct sysctl32 {
1958 unsigned int name;
1959 int nlen;
1960 unsigned int oldval;
1961 unsigned int oldlenp;
1962 unsigned int newval;
1963 unsigned int newlen;
1964 unsigned int __unused[4];
1967 #ifdef CONFIG_SYSCTL
1968 asmlinkage long
1969 sys32_sysctl (struct sysctl32 __user *args)
1971 struct sysctl32 a32;
1972 mm_segment_t old_fs = get_fs ();
1973 void __user *oldvalp, *newvalp;
1974 size_t oldlen;
1975 int __user *namep;
1976 long ret;
1978 if (copy_from_user(&a32, args, sizeof(a32)))
1979 return -EFAULT;
1982 * We need to pre-validate these because we have to disable address checking
1983 * before calling do_sysctl() because of OLDLEN but we can't run the risk of the
1984 * user specifying bad addresses here. Well, since we're dealing with 32 bit
1985 * addresses, we KNOW that access_ok() will always succeed, so this is an
1986 * expensive NOP, but so what...
1988 namep = (int __user *) compat_ptr(a32.name);
1989 oldvalp = compat_ptr(a32.oldval);
1990 newvalp = compat_ptr(a32.newval);
1992 if ((oldvalp && get_user(oldlen, (int __user *) compat_ptr(a32.oldlenp)))
1993 || !access_ok(VERIFY_WRITE, namep, 0)
1994 || !access_ok(VERIFY_WRITE, oldvalp, 0)
1995 || !access_ok(VERIFY_WRITE, newvalp, 0))
1996 return -EFAULT;
1998 set_fs(KERNEL_DS);
1999 lock_kernel();
2000 ret = do_sysctl(namep, a32.nlen, oldvalp, (size_t __user *) &oldlen,
2001 newvalp, (size_t) a32.newlen);
2002 unlock_kernel();
2003 set_fs(old_fs);
2005 if (oldvalp && put_user (oldlen, (int __user *) compat_ptr(a32.oldlenp)))
2006 return -EFAULT;
2008 return ret;
2010 #endif
2012 asmlinkage long
2013 sys32_newuname (struct new_utsname __user *name)
2015 int ret = sys_newuname(name);
2017 if (!ret)
2018 if (copy_to_user(name->machine, "i686\0\0\0", 8))
2019 ret = -EFAULT;
2020 return ret;
2023 asmlinkage long
2024 sys32_getresuid16 (u16 __user *ruid, u16 __user *euid, u16 __user *suid)
2026 uid_t a, b, c;
2027 int ret;
2028 mm_segment_t old_fs = get_fs();
2030 set_fs(KERNEL_DS);
2031 ret = sys_getresuid((uid_t __user *) &a, (uid_t __user *) &b, (uid_t __user *) &c);
2032 set_fs(old_fs);
2034 if (put_user(a, ruid) || put_user(b, euid) || put_user(c, suid))
2035 return -EFAULT;
2036 return ret;
2039 asmlinkage long
2040 sys32_getresgid16 (u16 __user *rgid, u16 __user *egid, u16 __user *sgid)
2042 gid_t a, b, c;
2043 int ret;
2044 mm_segment_t old_fs = get_fs();
2046 set_fs(KERNEL_DS);
2047 ret = sys_getresgid((gid_t __user *) &a, (gid_t __user *) &b, (gid_t __user *) &c);
2048 set_fs(old_fs);
2050 if (ret)
2051 return ret;
2053 return put_user(a, rgid) | put_user(b, egid) | put_user(c, sgid);
2056 asmlinkage long
2057 sys32_lseek (unsigned int fd, int offset, unsigned int whence)
2059 /* Sign-extension of "offset" is important here... */
2060 return sys_lseek(fd, offset, whence);
2063 static int
2064 groups16_to_user(short __user *grouplist, struct group_info *group_info)
2066 int i;
2067 short group;
2069 for (i = 0; i < group_info->ngroups; i++) {
2070 group = (short)GROUP_AT(group_info, i);
2071 if (put_user(group, grouplist+i))
2072 return -EFAULT;
2075 return 0;
2078 static int
2079 groups16_from_user(struct group_info *group_info, short __user *grouplist)
2081 int i;
2082 short group;
2084 for (i = 0; i < group_info->ngroups; i++) {
2085 if (get_user(group, grouplist+i))
2086 return -EFAULT;
2087 GROUP_AT(group_info, i) = (gid_t)group;
2090 return 0;
2093 asmlinkage long
2094 sys32_getgroups16 (int gidsetsize, short __user *grouplist)
2096 int i;
2098 if (gidsetsize < 0)
2099 return -EINVAL;
2101 get_group_info(current->group_info);
2102 i = current->group_info->ngroups;
2103 if (gidsetsize) {
2104 if (i > gidsetsize) {
2105 i = -EINVAL;
2106 goto out;
2108 if (groups16_to_user(grouplist, current->group_info)) {
2109 i = -EFAULT;
2110 goto out;
2113 out:
2114 put_group_info(current->group_info);
2115 return i;
2118 asmlinkage long
2119 sys32_setgroups16 (int gidsetsize, short __user *grouplist)
2121 struct group_info *group_info;
2122 int retval;
2124 if (!capable(CAP_SETGID))
2125 return -EPERM;
2126 if ((unsigned)gidsetsize > NGROUPS_MAX)
2127 return -EINVAL;
2129 group_info = groups_alloc(gidsetsize);
2130 if (!group_info)
2131 return -ENOMEM;
2132 retval = groups16_from_user(group_info, grouplist);
2133 if (retval) {
2134 put_group_info(group_info);
2135 return retval;
2138 retval = set_current_groups(group_info);
2139 put_group_info(group_info);
2141 return retval;
2144 asmlinkage long
2145 sys32_truncate64 (unsigned int path, unsigned int len_lo, unsigned int len_hi)
2147 return sys_truncate(compat_ptr(path), ((unsigned long) len_hi << 32) | len_lo);
2150 asmlinkage long
2151 sys32_ftruncate64 (int fd, unsigned int len_lo, unsigned int len_hi)
2153 return sys_ftruncate(fd, ((unsigned long) len_hi << 32) | len_lo);
2156 static int
2157 putstat64 (struct stat64 __user *ubuf, struct kstat *kbuf)
2159 int err;
2160 u64 hdev;
2162 if (clear_user(ubuf, sizeof(*ubuf)))
2163 return -EFAULT;
2165 hdev = huge_encode_dev(kbuf->dev);
2166 err = __put_user(hdev, (u32 __user*)&ubuf->st_dev);
2167 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_dev) + 1);
2168 err |= __put_user(kbuf->ino, &ubuf->__st_ino);
2169 err |= __put_user(kbuf->ino, &ubuf->st_ino_lo);
2170 err |= __put_user(kbuf->ino >> 32, &ubuf->st_ino_hi);
2171 err |= __put_user(kbuf->mode, &ubuf->st_mode);
2172 err |= __put_user(kbuf->nlink, &ubuf->st_nlink);
2173 err |= __put_user(kbuf->uid, &ubuf->st_uid);
2174 err |= __put_user(kbuf->gid, &ubuf->st_gid);
2175 hdev = huge_encode_dev(kbuf->rdev);
2176 err = __put_user(hdev, (u32 __user*)&ubuf->st_rdev);
2177 err |= __put_user(hdev >> 32, ((u32 __user*)&ubuf->st_rdev) + 1);
2178 err |= __put_user(kbuf->size, &ubuf->st_size_lo);
2179 err |= __put_user((kbuf->size >> 32), &ubuf->st_size_hi);
2180 err |= __put_user(kbuf->atime.tv_sec, &ubuf->st_atime);
2181 err |= __put_user(kbuf->atime.tv_nsec, &ubuf->st_atime_nsec);
2182 err |= __put_user(kbuf->mtime.tv_sec, &ubuf->st_mtime);
2183 err |= __put_user(kbuf->mtime.tv_nsec, &ubuf->st_mtime_nsec);
2184 err |= __put_user(kbuf->ctime.tv_sec, &ubuf->st_ctime);
2185 err |= __put_user(kbuf->ctime.tv_nsec, &ubuf->st_ctime_nsec);
2186 err |= __put_user(kbuf->blksize, &ubuf->st_blksize);
2187 err |= __put_user(kbuf->blocks, &ubuf->st_blocks);
2188 return err;
2191 asmlinkage long
2192 sys32_stat64 (char __user *filename, struct stat64 __user *statbuf)
2194 struct kstat s;
2195 long ret = vfs_stat(filename, &s);
2196 if (!ret)
2197 ret = putstat64(statbuf, &s);
2198 return ret;
2201 asmlinkage long
2202 sys32_lstat64 (char __user *filename, struct stat64 __user *statbuf)
2204 struct kstat s;
2205 long ret = vfs_lstat(filename, &s);
2206 if (!ret)
2207 ret = putstat64(statbuf, &s);
2208 return ret;
2211 asmlinkage long
2212 sys32_fstat64 (unsigned int fd, struct stat64 __user *statbuf)
2214 struct kstat s;
2215 long ret = vfs_fstat(fd, &s);
2216 if (!ret)
2217 ret = putstat64(statbuf, &s);
2218 return ret;
2221 struct sysinfo32 {
2222 s32 uptime;
2223 u32 loads[3];
2224 u32 totalram;
2225 u32 freeram;
2226 u32 sharedram;
2227 u32 bufferram;
2228 u32 totalswap;
2229 u32 freeswap;
2230 u16 procs;
2231 u16 pad;
2232 u32 totalhigh;
2233 u32 freehigh;
2234 u32 mem_unit;
2235 char _f[8];
2238 asmlinkage long
2239 sys32_sysinfo (struct sysinfo32 __user *info)
2241 struct sysinfo s;
2242 long ret, err;
2243 int bitcount = 0;
2244 mm_segment_t old_fs = get_fs();
2246 set_fs(KERNEL_DS);
2247 ret = sys_sysinfo((struct sysinfo __user *) &s);
2248 set_fs(old_fs);
2249 /* Check to see if any memory value is too large for 32-bit and
2250 * scale down if needed.
2252 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2253 while (s.mem_unit < PAGE_SIZE) {
2254 s.mem_unit <<= 1;
2255 bitcount++;
2257 s.totalram >>= bitcount;
2258 s.freeram >>= bitcount;
2259 s.sharedram >>= bitcount;
2260 s.bufferram >>= bitcount;
2261 s.totalswap >>= bitcount;
2262 s.freeswap >>= bitcount;
2263 s.totalhigh >>= bitcount;
2264 s.freehigh >>= bitcount;
2267 if (!access_ok(VERIFY_WRITE, info, sizeof(*info)))
2268 return -EFAULT;
2270 err = __put_user(s.uptime, &info->uptime);
2271 err |= __put_user(s.loads[0], &info->loads[0]);
2272 err |= __put_user(s.loads[1], &info->loads[1]);
2273 err |= __put_user(s.loads[2], &info->loads[2]);
2274 err |= __put_user(s.totalram, &info->totalram);
2275 err |= __put_user(s.freeram, &info->freeram);
2276 err |= __put_user(s.sharedram, &info->sharedram);
2277 err |= __put_user(s.bufferram, &info->bufferram);
2278 err |= __put_user(s.totalswap, &info->totalswap);
2279 err |= __put_user(s.freeswap, &info->freeswap);
2280 err |= __put_user(s.procs, &info->procs);
2281 err |= __put_user (s.totalhigh, &info->totalhigh);
2282 err |= __put_user (s.freehigh, &info->freehigh);
2283 err |= __put_user (s.mem_unit, &info->mem_unit);
2284 if (err)
2285 return -EFAULT;
2286 return ret;
2289 asmlinkage long
2290 sys32_sched_rr_get_interval (pid_t pid, struct compat_timespec __user *interval)
2292 mm_segment_t old_fs = get_fs();
2293 struct timespec t;
2294 long ret;
2296 set_fs(KERNEL_DS);
2297 ret = sys_sched_rr_get_interval(pid, (struct timespec __user *) &t);
2298 set_fs(old_fs);
2299 if (put_compat_timespec(&t, interval))
2300 return -EFAULT;
2301 return ret;
2304 asmlinkage long
2305 sys32_pread (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2307 return sys_pread64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2310 asmlinkage long
2311 sys32_pwrite (unsigned int fd, void __user *buf, unsigned int count, u32 pos_lo, u32 pos_hi)
2313 return sys_pwrite64(fd, buf, count, ((unsigned long) pos_hi << 32) | pos_lo);
2316 asmlinkage long
2317 sys32_sendfile (int out_fd, int in_fd, int __user *offset, unsigned int count)
2319 mm_segment_t old_fs = get_fs();
2320 long ret;
2321 off_t of;
2323 if (offset && get_user(of, offset))
2324 return -EFAULT;
2326 set_fs(KERNEL_DS);
2327 ret = sys_sendfile(out_fd, in_fd, offset ? (off_t __user *) &of : NULL, count);
2328 set_fs(old_fs);
2330 if (!ret && offset && put_user(of, offset))
2331 return -EFAULT;
2333 return ret;
2336 asmlinkage long
2337 sys32_personality (unsigned int personality)
2339 long ret;
2341 if (current->personality == PER_LINUX32 && personality == PER_LINUX)
2342 personality = PER_LINUX32;
2343 ret = sys_personality(personality);
2344 if (ret == PER_LINUX32)
2345 ret = PER_LINUX;
2346 return ret;
2349 asmlinkage unsigned long
2350 sys32_brk (unsigned int brk)
2352 unsigned long ret, obrk;
2353 struct mm_struct *mm = current->mm;
2355 obrk = mm->brk;
2356 ret = sys_brk(brk);
2357 if (ret < obrk)
2358 clear_user(compat_ptr(ret), PAGE_ALIGN(ret) - ret);
2359 return ret;
2363 * Exactly like fs/open.c:sys_open(), except that it doesn't set the O_LARGEFILE flag.
2365 asmlinkage long
2366 sys32_open (const char __user * filename, int flags, int mode)
2368 char * tmp;
2369 int fd, error;
2371 tmp = getname(filename);
2372 fd = PTR_ERR(tmp);
2373 if (!IS_ERR(tmp)) {
2374 fd = get_unused_fd();
2375 if (fd >= 0) {
2376 struct file *f = filp_open(tmp, flags, mode);
2377 error = PTR_ERR(f);
2378 if (IS_ERR(f))
2379 goto out_error;
2380 fd_install(fd, f);
2382 out:
2383 putname(tmp);
2385 return fd;
2387 out_error:
2388 put_unused_fd(fd);
2389 fd = error;
2390 goto out;
2393 /* Structure for ia32 emulation on ia64 */
2394 struct epoll_event32
2396 u32 events;
2397 u32 data[2];
2400 asmlinkage long
2401 sys32_epoll_ctl(int epfd, int op, int fd, struct epoll_event32 __user *event)
2403 mm_segment_t old_fs = get_fs();
2404 struct epoll_event event64;
2405 int error;
2406 u32 data_halfword;
2408 if (!access_ok(VERIFY_READ, event, sizeof(struct epoll_event32)))
2409 return -EFAULT;
2411 __get_user(event64.events, &event->events);
2412 __get_user(data_halfword, &event->data[0]);
2413 event64.data = data_halfword;
2414 __get_user(data_halfword, &event->data[1]);
2415 event64.data |= (u64)data_halfword << 32;
2417 set_fs(KERNEL_DS);
2418 error = sys_epoll_ctl(epfd, op, fd, (struct epoll_event __user *) &event64);
2419 set_fs(old_fs);
2421 return error;
2424 asmlinkage long
2425 sys32_epoll_wait(int epfd, struct epoll_event32 __user * events, int maxevents,
2426 int timeout)
2428 struct epoll_event *events64 = NULL;
2429 mm_segment_t old_fs = get_fs();
2430 int error, numevents, size;
2431 int evt_idx;
2432 int do_free_pages = 0;
2434 if (maxevents <= 0) {
2435 return -EINVAL;
2438 /* Verify that the area passed by the user is writeable */
2439 if (!access_ok(VERIFY_WRITE, events, maxevents * sizeof(struct epoll_event32)))
2440 return -EFAULT;
2443 * Allocate space for the intermediate copy. If the space needed
2444 * is large enough to cause kmalloc to fail, then try again with
2445 * __get_free_pages.
2447 size = maxevents * sizeof(struct epoll_event);
2448 events64 = kmalloc(size, GFP_KERNEL);
2449 if (events64 == NULL) {
2450 events64 = (struct epoll_event *)
2451 __get_free_pages(GFP_KERNEL, get_order(size));
2452 if (events64 == NULL)
2453 return -ENOMEM;
2454 do_free_pages = 1;
2457 /* Do the system call */
2458 set_fs(KERNEL_DS); /* copy_to/from_user should work on kernel mem*/
2459 numevents = sys_epoll_wait(epfd, (struct epoll_event __user *) events64,
2460 maxevents, timeout);
2461 set_fs(old_fs);
2463 /* Don't modify userspace memory if we're returning an error */
2464 if (numevents > 0) {
2465 /* Translate the 64-bit structures back into the 32-bit
2466 structures */
2467 for (evt_idx = 0; evt_idx < numevents; evt_idx++) {
2468 __put_user(events64[evt_idx].events,
2469 &events[evt_idx].events);
2470 __put_user((u32)events64[evt_idx].data,
2471 &events[evt_idx].data[0]);
2472 __put_user((u32)(events64[evt_idx].data >> 32),
2473 &events[evt_idx].data[1]);
2477 if (do_free_pages)
2478 free_pages((unsigned long) events64, get_order(size));
2479 else
2480 kfree(events64);
2481 return numevents;
2485 * Get a yet unused TLS descriptor index.
2487 static int
2488 get_free_idx (void)
2490 struct thread_struct *t = &current->thread;
2491 int idx;
2493 for (idx = 0; idx < GDT_ENTRY_TLS_ENTRIES; idx++)
2494 if (desc_empty(t->tls_array + idx))
2495 return idx + GDT_ENTRY_TLS_MIN;
2496 return -ESRCH;
2500 * Set a given TLS descriptor:
2502 asmlinkage int
2503 sys32_set_thread_area (struct ia32_user_desc __user *u_info)
2505 struct thread_struct *t = &current->thread;
2506 struct ia32_user_desc info;
2507 struct desc_struct *desc;
2508 int cpu, idx;
2510 if (copy_from_user(&info, u_info, sizeof(info)))
2511 return -EFAULT;
2512 idx = info.entry_number;
2515 * index -1 means the kernel should try to find and allocate an empty descriptor:
2517 if (idx == -1) {
2518 idx = get_free_idx();
2519 if (idx < 0)
2520 return idx;
2521 if (put_user(idx, &u_info->entry_number))
2522 return -EFAULT;
2525 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2526 return -EINVAL;
2528 desc = t->tls_array + idx - GDT_ENTRY_TLS_MIN;
2530 cpu = smp_processor_id();
2532 if (LDT_empty(&info)) {
2533 desc->a = 0;
2534 desc->b = 0;
2535 } else {
2536 desc->a = LDT_entry_a(&info);
2537 desc->b = LDT_entry_b(&info);
2539 load_TLS(t, cpu);
2540 return 0;
2544 * Get the current Thread-Local Storage area:
2547 #define GET_BASE(desc) ( \
2548 (((desc)->a >> 16) & 0x0000ffff) | \
2549 (((desc)->b << 16) & 0x00ff0000) | \
2550 ( (desc)->b & 0xff000000) )
2552 #define GET_LIMIT(desc) ( \
2553 ((desc)->a & 0x0ffff) | \
2554 ((desc)->b & 0xf0000) )
2556 #define GET_32BIT(desc) (((desc)->b >> 22) & 1)
2557 #define GET_CONTENTS(desc) (((desc)->b >> 10) & 3)
2558 #define GET_WRITABLE(desc) (((desc)->b >> 9) & 1)
2559 #define GET_LIMIT_PAGES(desc) (((desc)->b >> 23) & 1)
2560 #define GET_PRESENT(desc) (((desc)->b >> 15) & 1)
2561 #define GET_USEABLE(desc) (((desc)->b >> 20) & 1)
2563 asmlinkage int
2564 sys32_get_thread_area (struct ia32_user_desc __user *u_info)
2566 struct ia32_user_desc info;
2567 struct desc_struct *desc;
2568 int idx;
2570 if (get_user(idx, &u_info->entry_number))
2571 return -EFAULT;
2572 if (idx < GDT_ENTRY_TLS_MIN || idx > GDT_ENTRY_TLS_MAX)
2573 return -EINVAL;
2575 desc = current->thread.tls_array + idx - GDT_ENTRY_TLS_MIN;
2577 info.entry_number = idx;
2578 info.base_addr = GET_BASE(desc);
2579 info.limit = GET_LIMIT(desc);
2580 info.seg_32bit = GET_32BIT(desc);
2581 info.contents = GET_CONTENTS(desc);
2582 info.read_exec_only = !GET_WRITABLE(desc);
2583 info.limit_in_pages = GET_LIMIT_PAGES(desc);
2584 info.seg_not_present = !GET_PRESENT(desc);
2585 info.useable = GET_USEABLE(desc);
2587 if (copy_to_user(u_info, &info, sizeof(info)))
2588 return -EFAULT;
2589 return 0;
2592 asmlinkage long
2593 sys32_timer_create(u32 clock, struct compat_sigevent __user *se32, timer_t __user *timer_id)
2595 struct sigevent se;
2596 mm_segment_t oldfs;
2597 timer_t t;
2598 long err;
2600 if (se32 == NULL)
2601 return sys_timer_create(clock, NULL, timer_id);
2603 if (get_compat_sigevent(&se, se32))
2604 return -EFAULT;
2606 if (!access_ok(VERIFY_WRITE,timer_id,sizeof(timer_t)))
2607 return -EFAULT;
2609 oldfs = get_fs();
2610 set_fs(KERNEL_DS);
2611 err = sys_timer_create(clock, (struct sigevent __user *) &se, (timer_t __user *) &t);
2612 set_fs(oldfs);
2614 if (!err)
2615 err = __put_user (t, timer_id);
2617 return err;
2620 long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
2621 __u32 len_low, __u32 len_high, int advice)
2623 return sys_fadvise64_64(fd,
2624 (((u64)offset_high)<<32) | offset_low,
2625 (((u64)len_high)<<32) | len_low,
2626 advice);
2629 #ifdef NOTYET /* UNTESTED FOR IA64 FROM HERE DOWN */
2631 asmlinkage long sys32_setreuid(compat_uid_t ruid, compat_uid_t euid)
2633 uid_t sruid, seuid;
2635 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2636 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2637 return sys_setreuid(sruid, seuid);
2640 asmlinkage long
2641 sys32_setresuid(compat_uid_t ruid, compat_uid_t euid,
2642 compat_uid_t suid)
2644 uid_t sruid, seuid, ssuid;
2646 sruid = (ruid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)ruid);
2647 seuid = (euid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)euid);
2648 ssuid = (suid == (compat_uid_t)-1) ? ((uid_t)-1) : ((uid_t)suid);
2649 return sys_setresuid(sruid, seuid, ssuid);
2652 asmlinkage long
2653 sys32_setregid(compat_gid_t rgid, compat_gid_t egid)
2655 gid_t srgid, segid;
2657 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2658 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2659 return sys_setregid(srgid, segid);
2662 asmlinkage long
2663 sys32_setresgid(compat_gid_t rgid, compat_gid_t egid,
2664 compat_gid_t sgid)
2666 gid_t srgid, segid, ssgid;
2668 srgid = (rgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)rgid);
2669 segid = (egid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)egid);
2670 ssgid = (sgid == (compat_gid_t)-1) ? ((gid_t)-1) : ((gid_t)sgid);
2671 return sys_setresgid(srgid, segid, ssgid);
2674 /* Handle adjtimex compatibility. */
2676 struct timex32 {
2677 u32 modes;
2678 s32 offset, freq, maxerror, esterror;
2679 s32 status, constant, precision, tolerance;
2680 struct compat_timeval time;
2681 s32 tick;
2682 s32 ppsfreq, jitter, shift, stabil;
2683 s32 jitcnt, calcnt, errcnt, stbcnt;
2684 s32 :32; s32 :32; s32 :32; s32 :32;
2685 s32 :32; s32 :32; s32 :32; s32 :32;
2686 s32 :32; s32 :32; s32 :32; s32 :32;
2689 extern int do_adjtimex(struct timex *);
2691 asmlinkage long
2692 sys32_adjtimex(struct timex32 *utp)
2694 struct timex txc;
2695 int ret;
2697 memset(&txc, 0, sizeof(struct timex));
2699 if(get_user(txc.modes, &utp->modes) ||
2700 __get_user(txc.offset, &utp->offset) ||
2701 __get_user(txc.freq, &utp->freq) ||
2702 __get_user(txc.maxerror, &utp->maxerror) ||
2703 __get_user(txc.esterror, &utp->esterror) ||
2704 __get_user(txc.status, &utp->status) ||
2705 __get_user(txc.constant, &utp->constant) ||
2706 __get_user(txc.precision, &utp->precision) ||
2707 __get_user(txc.tolerance, &utp->tolerance) ||
2708 __get_user(txc.time.tv_sec, &utp->time.tv_sec) ||
2709 __get_user(txc.time.tv_usec, &utp->time.tv_usec) ||
2710 __get_user(txc.tick, &utp->tick) ||
2711 __get_user(txc.ppsfreq, &utp->ppsfreq) ||
2712 __get_user(txc.jitter, &utp->jitter) ||
2713 __get_user(txc.shift, &utp->shift) ||
2714 __get_user(txc.stabil, &utp->stabil) ||
2715 __get_user(txc.jitcnt, &utp->jitcnt) ||
2716 __get_user(txc.calcnt, &utp->calcnt) ||
2717 __get_user(txc.errcnt, &utp->errcnt) ||
2718 __get_user(txc.stbcnt, &utp->stbcnt))
2719 return -EFAULT;
2721 ret = do_adjtimex(&txc);
2723 if(put_user(txc.modes, &utp->modes) ||
2724 __put_user(txc.offset, &utp->offset) ||
2725 __put_user(txc.freq, &utp->freq) ||
2726 __put_user(txc.maxerror, &utp->maxerror) ||
2727 __put_user(txc.esterror, &utp->esterror) ||
2728 __put_user(txc.status, &utp->status) ||
2729 __put_user(txc.constant, &utp->constant) ||
2730 __put_user(txc.precision, &utp->precision) ||
2731 __put_user(txc.tolerance, &utp->tolerance) ||
2732 __put_user(txc.time.tv_sec, &utp->time.tv_sec) ||
2733 __put_user(txc.time.tv_usec, &utp->time.tv_usec) ||
2734 __put_user(txc.tick, &utp->tick) ||
2735 __put_user(txc.ppsfreq, &utp->ppsfreq) ||
2736 __put_user(txc.jitter, &utp->jitter) ||
2737 __put_user(txc.shift, &utp->shift) ||
2738 __put_user(txc.stabil, &utp->stabil) ||
2739 __put_user(txc.jitcnt, &utp->jitcnt) ||
2740 __put_user(txc.calcnt, &utp->calcnt) ||
2741 __put_user(txc.errcnt, &utp->errcnt) ||
2742 __put_user(txc.stbcnt, &utp->stbcnt))
2743 ret = -EFAULT;
2745 return ret;
2747 #endif /* NOTYET */