aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorDavid Vrabel <david.vrabel@csr.com>2009-01-07 05:45:25 -0500
committerDavid Vrabel <david.vrabel@csr.com>2009-01-07 05:45:25 -0500
commita23e66f3b8cfdedec14541e71ef29a754870a20c (patch)
treef8ac23572982e92e6f8ae09c4039db627bdf53ee /fs
parent04c470adb01c62bb9bd663cfc4875cf0a4eb01ab (diff)
parentede6f5aea054d3fb67c78857f7abdee602302043 (diff)
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6 into for-upstream
Diffstat (limited to 'fs')
-rw-r--r--fs/Kconfig17
-rw-r--r--fs/autofs4/autofs_i.h2
-rw-r--r--fs/autofs4/dev-ioctl.c75
-rw-r--r--fs/autofs4/expire.c4
-rw-r--r--fs/autofs4/inode.c14
-rw-r--r--fs/autofs4/waitq.c8
-rw-r--r--fs/bfs/inode.c45
-rw-r--r--fs/binfmt_misc.c2
-rw-r--r--fs/block_dev.c2
-rw-r--r--fs/buffer.c1
-rw-r--r--fs/char_dev.c2
-rw-r--r--fs/compat.c6
-rw-r--r--fs/direct-io.c13
-rw-r--r--fs/ecryptfs/crypto.c514
-rw-r--r--fs/ecryptfs/ecryptfs_kernel.h105
-rw-r--r--fs/ecryptfs/file.c30
-rw-r--r--fs/ecryptfs/inode.c297
-rw-r--r--fs/ecryptfs/keystore.c651
-rw-r--r--fs/ecryptfs/main.c126
-rw-r--r--fs/ecryptfs/messaging.c4
-rw-r--r--fs/ecryptfs/miscdev.c18
-rw-r--r--fs/exec.c29
-rw-r--r--fs/ext4/ext4.h6
-rw-r--r--fs/ext4/inode.c2
-rw-r--r--fs/fs-writeback.c92
-rw-r--r--fs/fuse/control.c6
-rw-r--r--fs/fuse/dev.c113
-rw-r--r--fs/fuse/dir.c48
-rw-r--r--fs/fuse/file.c457
-rw-r--r--fs/fuse/fuse_i.h83
-rw-r--r--fs/fuse/inode.c157
-rw-r--r--fs/hugetlbfs/inode.c12
-rw-r--r--fs/inode.c12
-rw-r--r--fs/minix/dir.c2
-rw-r--r--fs/mpage.c6
-rw-r--r--fs/ncpfs/getopt.c1
-rw-r--r--fs/partitions/check.c11
-rw-r--r--fs/proc/task_mmu.c8
-rw-r--r--fs/select.c76
-rw-r--r--fs/sync.c2
-rw-r--r--fs/ubifs/super.c9
41 files changed, 2395 insertions, 673 deletions
diff --git a/fs/Kconfig b/fs/Kconfig
index f9b6e2979aaa..32883589ee54 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -721,7 +721,20 @@ config CONFIGFS_FS
721 721
722endmenu 722endmenu
723 723
724menu "Miscellaneous filesystems" 724menuconfig MISC_FILESYSTEMS
725 bool "Miscellaneous filesystems"
726 default y
727 ---help---
728 Say Y here to get to see options for various miscellaneous
729 filesystems, such as filesystems that came from other
730 operating systems.
731
732 This option alone does not add any kernel code.
733
734 If you say N, all options in this submenu will be skipped and
735 disabled; if unsure, say Y here.
736
737if MISC_FILESYSTEMS
725 738
726config ADFS_FS 739config ADFS_FS
727 tristate "ADFS file system support (EXPERIMENTAL)" 740 tristate "ADFS file system support (EXPERIMENTAL)"
@@ -1091,7 +1104,7 @@ config UFS_DEBUG
1091 Y here. This will result in _many_ additional debugging messages to be 1104 Y here. This will result in _many_ additional debugging messages to be
1092 written to the system log. 1105 written to the system log.
1093 1106
1094endmenu 1107endif # MISC_FILESYSTEMS
1095 1108
1096menuconfig NETWORK_FILESYSTEMS 1109menuconfig NETWORK_FILESYSTEMS
1097 bool "Network File Systems" 1110 bool "Network File Systems"
diff --git a/fs/autofs4/autofs_i.h b/fs/autofs4/autofs_i.h
index e0f16da00e54..a76803108d06 100644
--- a/fs/autofs4/autofs_i.h
+++ b/fs/autofs4/autofs_i.h
@@ -25,8 +25,6 @@
25#define AUTOFS_DEV_IOCTL_IOC_FIRST (AUTOFS_DEV_IOCTL_VERSION) 25#define AUTOFS_DEV_IOCTL_IOC_FIRST (AUTOFS_DEV_IOCTL_VERSION)
26#define AUTOFS_DEV_IOCTL_IOC_COUNT (AUTOFS_IOC_COUNT - 11) 26#define AUTOFS_DEV_IOCTL_IOC_COUNT (AUTOFS_IOC_COUNT - 11)
27 27
28#define AUTOFS_TYPE_TRIGGER (AUTOFS_TYPE_DIRECT|AUTOFS_TYPE_OFFSET)
29
30#include <linux/kernel.h> 28#include <linux/kernel.h>
31#include <linux/slab.h> 29#include <linux/slab.h>
32#include <linux/time.h> 30#include <linux/time.h>
diff --git a/fs/autofs4/dev-ioctl.c b/fs/autofs4/dev-ioctl.c
index 63b7c7afe8df..025e105bffea 100644
--- a/fs/autofs4/dev-ioctl.c
+++ b/fs/autofs4/dev-ioctl.c
@@ -124,7 +124,7 @@ static inline void free_dev_ioctl(struct autofs_dev_ioctl *param)
124 124
125/* 125/*
126 * Check sanity of parameter control fields and if a path is present 126 * Check sanity of parameter control fields and if a path is present
127 * check that it has a "/" and is terminated. 127 * check that it is terminated and contains at least one "/".
128 */ 128 */
129static int validate_dev_ioctl(int cmd, struct autofs_dev_ioctl *param) 129static int validate_dev_ioctl(int cmd, struct autofs_dev_ioctl *param)
130{ 130{
@@ -138,15 +138,16 @@ static int validate_dev_ioctl(int cmd, struct autofs_dev_ioctl *param)
138 } 138 }
139 139
140 if (param->size > sizeof(*param)) { 140 if (param->size > sizeof(*param)) {
141 err = check_name(param->path); 141 err = invalid_str(param->path,
142 (void *) ((size_t) param + param->size));
142 if (err) { 143 if (err) {
143 AUTOFS_WARN("invalid path supplied for cmd(0x%08x)", 144 AUTOFS_WARN(
144 cmd); 145 "path string terminator missing for cmd(0x%08x)",
146 cmd);
145 goto out; 147 goto out;
146 } 148 }
147 149
148 err = invalid_str(param->path, 150 err = check_name(param->path);
149 (void *) ((size_t) param + param->size));
150 if (err) { 151 if (err) {
151 AUTOFS_WARN("invalid path supplied for cmd(0x%08x)", 152 AUTOFS_WARN("invalid path supplied for cmd(0x%08x)",
152 cmd); 153 cmd);
@@ -180,7 +181,7 @@ static int autofs_dev_ioctl_protover(struct file *fp,
180 struct autofs_sb_info *sbi, 181 struct autofs_sb_info *sbi,
181 struct autofs_dev_ioctl *param) 182 struct autofs_dev_ioctl *param)
182{ 183{
183 param->arg1 = sbi->version; 184 param->protover.version = sbi->version;
184 return 0; 185 return 0;
185} 186}
186 187
@@ -189,7 +190,7 @@ static int autofs_dev_ioctl_protosubver(struct file *fp,
189 struct autofs_sb_info *sbi, 190 struct autofs_sb_info *sbi,
190 struct autofs_dev_ioctl *param) 191 struct autofs_dev_ioctl *param)
191{ 192{
192 param->arg1 = sbi->sub_version; 193 param->protosubver.sub_version = sbi->sub_version;
193 return 0; 194 return 0;
194} 195}
195 196
@@ -335,13 +336,13 @@ static int autofs_dev_ioctl_openmount(struct file *fp,
335 int err, fd; 336 int err, fd;
336 337
337 /* param->path has already been checked */ 338 /* param->path has already been checked */
338 if (!param->arg1) 339 if (!param->openmount.devid)
339 return -EINVAL; 340 return -EINVAL;
340 341
341 param->ioctlfd = -1; 342 param->ioctlfd = -1;
342 343
343 path = param->path; 344 path = param->path;
344 devid = param->arg1; 345 devid = param->openmount.devid;
345 346
346 err = 0; 347 err = 0;
347 fd = autofs_dev_ioctl_open_mountpoint(path, devid); 348 fd = autofs_dev_ioctl_open_mountpoint(path, devid);
@@ -373,7 +374,7 @@ static int autofs_dev_ioctl_ready(struct file *fp,
373{ 374{
374 autofs_wqt_t token; 375 autofs_wqt_t token;
375 376
376 token = (autofs_wqt_t) param->arg1; 377 token = (autofs_wqt_t) param->ready.token;
377 return autofs4_wait_release(sbi, token, 0); 378 return autofs4_wait_release(sbi, token, 0);
378} 379}
379 380
@@ -388,8 +389,8 @@ static int autofs_dev_ioctl_fail(struct file *fp,
388 autofs_wqt_t token; 389 autofs_wqt_t token;
389 int status; 390 int status;
390 391
391 token = (autofs_wqt_t) param->arg1; 392 token = (autofs_wqt_t) param->fail.token;
392 status = param->arg2 ? param->arg2 : -ENOENT; 393 status = param->fail.status ? param->fail.status : -ENOENT;
393 return autofs4_wait_release(sbi, token, status); 394 return autofs4_wait_release(sbi, token, status);
394} 395}
395 396
@@ -412,10 +413,10 @@ static int autofs_dev_ioctl_setpipefd(struct file *fp,
412 int pipefd; 413 int pipefd;
413 int err = 0; 414 int err = 0;
414 415
415 if (param->arg1 == -1) 416 if (param->setpipefd.pipefd == -1)
416 return -EINVAL; 417 return -EINVAL;
417 418
418 pipefd = param->arg1; 419 pipefd = param->setpipefd.pipefd;
419 420
420 mutex_lock(&sbi->wq_mutex); 421 mutex_lock(&sbi->wq_mutex);
421 if (!sbi->catatonic) { 422 if (!sbi->catatonic) {
@@ -457,8 +458,8 @@ static int autofs_dev_ioctl_timeout(struct file *fp,
457{ 458{
458 unsigned long timeout; 459 unsigned long timeout;
459 460
460 timeout = param->arg1; 461 timeout = param->timeout.timeout;
461 param->arg1 = sbi->exp_timeout / HZ; 462 param->timeout.timeout = sbi->exp_timeout / HZ;
462 sbi->exp_timeout = timeout * HZ; 463 sbi->exp_timeout = timeout * HZ;
463 return 0; 464 return 0;
464} 465}
@@ -489,7 +490,7 @@ static int autofs_dev_ioctl_requester(struct file *fp,
489 path = param->path; 490 path = param->path;
490 devid = sbi->sb->s_dev; 491 devid = sbi->sb->s_dev;
491 492
492 param->arg1 = param->arg2 = -1; 493 param->requester.uid = param->requester.gid = -1;
493 494
494 /* Get nameidata of the parent directory */ 495 /* Get nameidata of the parent directory */
495 err = path_lookup(path, LOOKUP_PARENT, &nd); 496 err = path_lookup(path, LOOKUP_PARENT, &nd);
@@ -505,8 +506,8 @@ static int autofs_dev_ioctl_requester(struct file *fp,
505 err = 0; 506 err = 0;
506 autofs4_expire_wait(nd.path.dentry); 507 autofs4_expire_wait(nd.path.dentry);
507 spin_lock(&sbi->fs_lock); 508 spin_lock(&sbi->fs_lock);
508 param->arg1 = ino->uid; 509 param->requester.uid = ino->uid;
509 param->arg2 = ino->gid; 510 param->requester.gid = ino->gid;
510 spin_unlock(&sbi->fs_lock); 511 spin_unlock(&sbi->fs_lock);
511 } 512 }
512 513
@@ -529,10 +530,10 @@ static int autofs_dev_ioctl_expire(struct file *fp,
529 int err = -EAGAIN; 530 int err = -EAGAIN;
530 int how; 531 int how;
531 532
532 how = param->arg1; 533 how = param->expire.how;
533 mnt = fp->f_path.mnt; 534 mnt = fp->f_path.mnt;
534 535
535 if (sbi->type & AUTOFS_TYPE_TRIGGER) 536 if (autofs_type_trigger(sbi->type))
536 dentry = autofs4_expire_direct(sbi->sb, mnt, sbi, how); 537 dentry = autofs4_expire_direct(sbi->sb, mnt, sbi, how);
537 else 538 else
538 dentry = autofs4_expire_indirect(sbi->sb, mnt, sbi, how); 539 dentry = autofs4_expire_indirect(sbi->sb, mnt, sbi, how);
@@ -565,9 +566,9 @@ static int autofs_dev_ioctl_askumount(struct file *fp,
565 struct autofs_sb_info *sbi, 566 struct autofs_sb_info *sbi,
566 struct autofs_dev_ioctl *param) 567 struct autofs_dev_ioctl *param)
567{ 568{
568 param->arg1 = 0; 569 param->askumount.may_umount = 0;
569 if (may_umount(fp->f_path.mnt)) 570 if (may_umount(fp->f_path.mnt))
570 param->arg1 = 1; 571 param->askumount.may_umount = 1;
571 return 0; 572 return 0;
572} 573}
573 574
@@ -600,6 +601,7 @@ static int autofs_dev_ioctl_ismountpoint(struct file *fp,
600 struct nameidata nd; 601 struct nameidata nd;
601 const char *path; 602 const char *path;
602 unsigned int type; 603 unsigned int type;
604 unsigned int devid, magic;
603 int err = -ENOENT; 605 int err = -ENOENT;
604 606
605 if (param->size <= sizeof(*param)) { 607 if (param->size <= sizeof(*param)) {
@@ -608,13 +610,13 @@ static int autofs_dev_ioctl_ismountpoint(struct file *fp,
608 } 610 }
609 611
610 path = param->path; 612 path = param->path;
611 type = param->arg1; 613 type = param->ismountpoint.in.type;
612 614
613 param->arg1 = 0; 615 param->ismountpoint.out.devid = devid = 0;
614 param->arg2 = 0; 616 param->ismountpoint.out.magic = magic = 0;
615 617
616 if (!fp || param->ioctlfd == -1) { 618 if (!fp || param->ioctlfd == -1) {
617 if (type == AUTOFS_TYPE_ANY) { 619 if (autofs_type_any(type)) {
618 struct super_block *sb; 620 struct super_block *sb;
619 621
620 err = path_lookup(path, LOOKUP_FOLLOW, &nd); 622 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
@@ -622,7 +624,7 @@ static int autofs_dev_ioctl_ismountpoint(struct file *fp,
622 goto out; 624 goto out;
623 625
624 sb = nd.path.dentry->d_sb; 626 sb = nd.path.dentry->d_sb;
625 param->arg1 = new_encode_dev(sb->s_dev); 627 devid = new_encode_dev(sb->s_dev);
626 } else { 628 } else {
627 struct autofs_info *ino; 629 struct autofs_info *ino;
628 630
@@ -635,38 +637,41 @@ static int autofs_dev_ioctl_ismountpoint(struct file *fp,
635 goto out_release; 637 goto out_release;
636 638
637 ino = autofs4_dentry_ino(nd.path.dentry); 639 ino = autofs4_dentry_ino(nd.path.dentry);
638 param->arg1 = autofs4_get_dev(ino->sbi); 640 devid = autofs4_get_dev(ino->sbi);
639 } 641 }
640 642
641 err = 0; 643 err = 0;
642 if (nd.path.dentry->d_inode && 644 if (nd.path.dentry->d_inode &&
643 nd.path.mnt->mnt_root == nd.path.dentry) { 645 nd.path.mnt->mnt_root == nd.path.dentry) {
644 err = 1; 646 err = 1;
645 param->arg2 = nd.path.dentry->d_inode->i_sb->s_magic; 647 magic = nd.path.dentry->d_inode->i_sb->s_magic;
646 } 648 }
647 } else { 649 } else {
648 dev_t devid = new_encode_dev(sbi->sb->s_dev); 650 dev_t dev = autofs4_get_dev(sbi);
649 651
650 err = path_lookup(path, LOOKUP_PARENT, &nd); 652 err = path_lookup(path, LOOKUP_PARENT, &nd);
651 if (err) 653 if (err)
652 goto out; 654 goto out;
653 655
654 err = autofs_dev_ioctl_find_super(&nd, devid); 656 err = autofs_dev_ioctl_find_super(&nd, dev);
655 if (err) 657 if (err)
656 goto out_release; 658 goto out_release;
657 659
658 param->arg1 = autofs4_get_dev(sbi); 660 devid = dev;
659 661
660 err = have_submounts(nd.path.dentry); 662 err = have_submounts(nd.path.dentry);
661 663
662 if (nd.path.mnt->mnt_mountpoint != nd.path.mnt->mnt_root) { 664 if (nd.path.mnt->mnt_mountpoint != nd.path.mnt->mnt_root) {
663 if (follow_down(&nd.path.mnt, &nd.path.dentry)) { 665 if (follow_down(&nd.path.mnt, &nd.path.dentry)) {
664 struct inode *inode = nd.path.dentry->d_inode; 666 struct inode *inode = nd.path.dentry->d_inode;
665 param->arg2 = inode->i_sb->s_magic; 667 magic = inode->i_sb->s_magic;
666 } 668 }
667 } 669 }
668 } 670 }
669 671
672 param->ismountpoint.out.devid = devid;
673 param->ismountpoint.out.magic = magic;
674
670out_release: 675out_release:
671 path_put(&nd.path); 676 path_put(&nd.path);
672out: 677out:
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index 4b6fb3f628c0..e3bd50776f9e 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -63,7 +63,7 @@ static int autofs4_mount_busy(struct vfsmount *mnt, struct dentry *dentry)
63 struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb); 63 struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
64 64
65 /* This is an autofs submount, we can't expire it */ 65 /* This is an autofs submount, we can't expire it */
66 if (sbi->type == AUTOFS_TYPE_INDIRECT) 66 if (autofs_type_indirect(sbi->type))
67 goto done; 67 goto done;
68 68
69 /* 69 /*
@@ -490,7 +490,7 @@ int autofs4_expire_multi(struct super_block *sb, struct vfsmount *mnt,
490 if (arg && get_user(do_now, arg)) 490 if (arg && get_user(do_now, arg))
491 return -EFAULT; 491 return -EFAULT;
492 492
493 if (sbi->type & AUTOFS_TYPE_TRIGGER) 493 if (autofs_type_trigger(sbi->type))
494 dentry = autofs4_expire_direct(sb, mnt, sbi, do_now); 494 dentry = autofs4_expire_direct(sb, mnt, sbi, do_now);
495 else 495 else
496 dentry = autofs4_expire_indirect(sb, mnt, sbi, do_now); 496 dentry = autofs4_expire_indirect(sb, mnt, sbi, do_now);
diff --git a/fs/autofs4/inode.c b/fs/autofs4/inode.c
index cfc23e53b6f4..716e12b627b2 100644
--- a/fs/autofs4/inode.c
+++ b/fs/autofs4/inode.c
@@ -197,9 +197,9 @@ static int autofs4_show_options(struct seq_file *m, struct vfsmount *mnt)
197 seq_printf(m, ",minproto=%d", sbi->min_proto); 197 seq_printf(m, ",minproto=%d", sbi->min_proto);
198 seq_printf(m, ",maxproto=%d", sbi->max_proto); 198 seq_printf(m, ",maxproto=%d", sbi->max_proto);
199 199
200 if (sbi->type & AUTOFS_TYPE_OFFSET) 200 if (autofs_type_offset(sbi->type))
201 seq_printf(m, ",offset"); 201 seq_printf(m, ",offset");
202 else if (sbi->type & AUTOFS_TYPE_DIRECT) 202 else if (autofs_type_direct(sbi->type))
203 seq_printf(m, ",direct"); 203 seq_printf(m, ",direct");
204 else 204 else
205 seq_printf(m, ",indirect"); 205 seq_printf(m, ",indirect");
@@ -284,13 +284,13 @@ static int parse_options(char *options, int *pipefd, uid_t *uid, gid_t *gid,
284 *maxproto = option; 284 *maxproto = option;
285 break; 285 break;
286 case Opt_indirect: 286 case Opt_indirect:
287 *type = AUTOFS_TYPE_INDIRECT; 287 set_autofs_type_indirect(type);
288 break; 288 break;
289 case Opt_direct: 289 case Opt_direct:
290 *type = AUTOFS_TYPE_DIRECT; 290 set_autofs_type_direct(type);
291 break; 291 break;
292 case Opt_offset: 292 case Opt_offset:
293 *type = AUTOFS_TYPE_OFFSET; 293 set_autofs_type_offset(type);
294 break; 294 break;
295 default: 295 default:
296 return 1; 296 return 1;
@@ -338,7 +338,7 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
338 sbi->sb = s; 338 sbi->sb = s;
339 sbi->version = 0; 339 sbi->version = 0;
340 sbi->sub_version = 0; 340 sbi->sub_version = 0;
341 sbi->type = AUTOFS_TYPE_INDIRECT; 341 set_autofs_type_indirect(&sbi->type);
342 sbi->min_proto = 0; 342 sbi->min_proto = 0;
343 sbi->max_proto = 0; 343 sbi->max_proto = 0;
344 mutex_init(&sbi->wq_mutex); 344 mutex_init(&sbi->wq_mutex);
@@ -380,7 +380,7 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
380 } 380 }
381 381
382 root_inode->i_fop = &autofs4_root_operations; 382 root_inode->i_fop = &autofs4_root_operations;
383 root_inode->i_op = sbi->type & AUTOFS_TYPE_TRIGGER ? 383 root_inode->i_op = autofs_type_trigger(sbi->type) ?
384 &autofs4_direct_root_inode_operations : 384 &autofs4_direct_root_inode_operations :
385 &autofs4_indirect_root_inode_operations; 385 &autofs4_indirect_root_inode_operations;
386 386
diff --git a/fs/autofs4/waitq.c b/fs/autofs4/waitq.c
index e02cc8ae5eb3..eeb246845909 100644
--- a/fs/autofs4/waitq.c
+++ b/fs/autofs4/waitq.c
@@ -337,7 +337,7 @@ int autofs4_wait(struct autofs_sb_info *sbi, struct dentry *dentry,
337 * is very similar for indirect mounts except only dentrys 337 * is very similar for indirect mounts except only dentrys
338 * in the root of the autofs file system may be negative. 338 * in the root of the autofs file system may be negative.
339 */ 339 */
340 if (sbi->type & AUTOFS_TYPE_TRIGGER) 340 if (autofs_type_trigger(sbi->type))
341 return -ENOENT; 341 return -ENOENT;
342 else if (!IS_ROOT(dentry->d_parent)) 342 else if (!IS_ROOT(dentry->d_parent))
343 return -ENOENT; 343 return -ENOENT;
@@ -348,7 +348,7 @@ int autofs4_wait(struct autofs_sb_info *sbi, struct dentry *dentry,
348 return -ENOMEM; 348 return -ENOMEM;
349 349
350 /* If this is a direct mount request create a dummy name */ 350 /* If this is a direct mount request create a dummy name */
351 if (IS_ROOT(dentry) && sbi->type & AUTOFS_TYPE_TRIGGER) 351 if (IS_ROOT(dentry) && autofs_type_trigger(sbi->type))
352 qstr.len = sprintf(name, "%p", dentry); 352 qstr.len = sprintf(name, "%p", dentry);
353 else { 353 else {
354 qstr.len = autofs4_getpath(sbi, dentry, &name); 354 qstr.len = autofs4_getpath(sbi, dentry, &name);
@@ -406,11 +406,11 @@ int autofs4_wait(struct autofs_sb_info *sbi, struct dentry *dentry,
406 type = autofs_ptype_expire_multi; 406 type = autofs_ptype_expire_multi;
407 } else { 407 } else {
408 if (notify == NFY_MOUNT) 408 if (notify == NFY_MOUNT)
409 type = (sbi->type & AUTOFS_TYPE_TRIGGER) ? 409 type = autofs_type_trigger(sbi->type) ?
410 autofs_ptype_missing_direct : 410 autofs_ptype_missing_direct :
411 autofs_ptype_missing_indirect; 411 autofs_ptype_missing_indirect;
412 else 412 else
413 type = (sbi->type & AUTOFS_TYPE_TRIGGER) ? 413 type = autofs_type_trigger(sbi->type) ?
414 autofs_ptype_expire_direct : 414 autofs_ptype_expire_direct :
415 autofs_ptype_expire_indirect; 415 autofs_ptype_expire_indirect;
416 } 416 }
diff --git a/fs/bfs/inode.c b/fs/bfs/inode.c
index 0ed57b5ee012..cc4062d12ca2 100644
--- a/fs/bfs/inode.c
+++ b/fs/bfs/inode.c
@@ -213,6 +213,9 @@ static void bfs_put_super(struct super_block *s)
213{ 213{
214 struct bfs_sb_info *info = BFS_SB(s); 214 struct bfs_sb_info *info = BFS_SB(s);
215 215
216 if (!info)
217 return;
218
216 brelse(info->si_sbh); 219 brelse(info->si_sbh);
217 mutex_destroy(&info->bfs_lock); 220 mutex_destroy(&info->bfs_lock);
218 kfree(info->si_imap); 221 kfree(info->si_imap);
@@ -327,6 +330,7 @@ static int bfs_fill_super(struct super_block *s, void *data, int silent)
327 unsigned i, imap_len; 330 unsigned i, imap_len;
328 struct bfs_sb_info *info; 331 struct bfs_sb_info *info;
329 long ret = -EINVAL; 332 long ret = -EINVAL;
333 unsigned long i_sblock, i_eblock, i_eoff, s_size;
330 334
331 info = kzalloc(sizeof(*info), GFP_KERNEL); 335 info = kzalloc(sizeof(*info), GFP_KERNEL);
332 if (!info) 336 if (!info)
@@ -350,6 +354,12 @@ static int bfs_fill_super(struct super_block *s, void *data, int silent)
350 354
351 s->s_magic = BFS_MAGIC; 355 s->s_magic = BFS_MAGIC;
352 info->si_sbh = bh; 356 info->si_sbh = bh;
357
358 if (le32_to_cpu(bfs_sb->s_start) > le32_to_cpu(bfs_sb->s_end)) {
359 printf("Superblock is corrupted\n");
360 goto out;
361 }
362
353 info->si_lasti = (le32_to_cpu(bfs_sb->s_start) - BFS_BSIZE) / 363 info->si_lasti = (le32_to_cpu(bfs_sb->s_start) - BFS_BSIZE) /
354 sizeof(struct bfs_inode) 364 sizeof(struct bfs_inode)
355 + BFS_ROOT_INO - 1; 365 + BFS_ROOT_INO - 1;
@@ -380,6 +390,18 @@ static int bfs_fill_super(struct super_block *s, void *data, int silent)
380 - le32_to_cpu(bfs_sb->s_start)) >> BFS_BSIZE_BITS; 390 - le32_to_cpu(bfs_sb->s_start)) >> BFS_BSIZE_BITS;
381 info->si_freei = 0; 391 info->si_freei = 0;
382 info->si_lf_eblk = 0; 392 info->si_lf_eblk = 0;
393
394 /* can we read the last block? */
395 bh = sb_bread(s, info->si_blocks - 1);
396 if (!bh) {
397 printf("Last block not available: %lu\n", info->si_blocks - 1);
398 iput(inode);
399 ret = -EIO;
400 kfree(info->si_imap);
401 goto out;
402 }
403 brelse(bh);
404
383 bh = NULL; 405 bh = NULL;
384 for (i = BFS_ROOT_INO; i <= info->si_lasti; i++) { 406 for (i = BFS_ROOT_INO; i <= info->si_lasti; i++) {
385 struct bfs_inode *di; 407 struct bfs_inode *di;
@@ -397,6 +419,29 @@ static int bfs_fill_super(struct super_block *s, void *data, int silent)
397 419
398 di = (struct bfs_inode *)bh->b_data + off; 420 di = (struct bfs_inode *)bh->b_data + off;
399 421
422 /* test if filesystem is not corrupted */
423
424 i_eoff = le32_to_cpu(di->i_eoffset);
425 i_sblock = le32_to_cpu(di->i_sblock);
426 i_eblock = le32_to_cpu(di->i_eblock);
427 s_size = le32_to_cpu(bfs_sb->s_end);
428
429 if (i_sblock > info->si_blocks ||
430 i_eblock > info->si_blocks ||
431 i_sblock > i_eblock ||
432 i_eoff > s_size ||
433 i_sblock * BFS_BSIZE > i_eoff) {
434
435 printf("Inode 0x%08x corrupted\n", i);
436
437 brelse(bh);
438 s->s_root = NULL;
439 kfree(info->si_imap);
440 kfree(info);
441 s->s_fs_info = NULL;
442 return -EIO;
443 }
444
400 if (!di->i_ino) { 445 if (!di->i_ino) {
401 info->si_freei++; 446 info->si_freei++;
402 continue; 447 continue;
diff --git a/fs/binfmt_misc.c b/fs/binfmt_misc.c
index e1158cb4fbd6..c4e83537ead7 100644
--- a/fs/binfmt_misc.c
+++ b/fs/binfmt_misc.c
@@ -649,7 +649,7 @@ static const struct file_operations bm_register_operations = {
649static ssize_t 649static ssize_t
650bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos) 650bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
651{ 651{
652 char *s = enabled ? "enabled" : "disabled"; 652 char *s = enabled ? "enabled\n" : "disabled\n";
653 653
654 return simple_read_from_buffer(buf, nbytes, ppos, s, strlen(s)); 654 return simple_read_from_buffer(buf, nbytes, ppos, s, strlen(s));
655} 655}
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 349a26c10001..b957717e25ab 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1262,7 +1262,7 @@ EXPORT_SYMBOL(ioctl_by_bdev);
1262 1262
1263/** 1263/**
1264 * lookup_bdev - lookup a struct block_device by name 1264 * lookup_bdev - lookup a struct block_device by name
1265 * @path: special file representing the block device 1265 * @pathname: special file representing the block device
1266 * 1266 *
1267 * Get a reference to the blockdevice at @pathname in the current 1267 * Get a reference to the blockdevice at @pathname in the current
1268 * namespace if possible and return it. Return ERR_PTR(error) 1268 * namespace if possible and return it. Return ERR_PTR(error)
diff --git a/fs/buffer.c b/fs/buffer.c
index a13f09b696f7..c26da785938a 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -2022,7 +2022,6 @@ int block_write_begin(struct file *file, struct address_space *mapping,
2022 if (pos + len > inode->i_size) 2022 if (pos + len > inode->i_size)
2023 vmtruncate(inode, inode->i_size); 2023 vmtruncate(inode, inode->i_size);
2024 } 2024 }
2025 goto out;
2026 } 2025 }
2027 2026
2028out: 2027out:
diff --git a/fs/char_dev.c b/fs/char_dev.c
index 700697a72618..38f71222a552 100644
--- a/fs/char_dev.c
+++ b/fs/char_dev.c
@@ -120,7 +120,7 @@ __register_chrdev_region(unsigned int major, unsigned int baseminor,
120 cd->major = major; 120 cd->major = major;
121 cd->baseminor = baseminor; 121 cd->baseminor = baseminor;
122 cd->minorct = minorct; 122 cd->minorct = minorct;
123 strncpy(cd->name,name, 64); 123 strlcpy(cd->name, name, sizeof(cd->name));
124 124
125 i = major_to_index(major); 125 i = major_to_index(major);
126 126
diff --git a/fs/compat.c b/fs/compat.c
index d1ece79b6411..30f2faa22f5c 100644
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -1187,6 +1187,9 @@ compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec, unsign
1187 ret = compat_do_readv_writev(READ, file, vec, vlen, &file->f_pos); 1187 ret = compat_do_readv_writev(READ, file, vec, vlen, &file->f_pos);
1188 1188
1189out: 1189out:
1190 if (ret > 0)
1191 add_rchar(current, ret);
1192 inc_syscr(current);
1190 fput(file); 1193 fput(file);
1191 return ret; 1194 return ret;
1192} 1195}
@@ -1210,6 +1213,9 @@ compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec, unsig
1210 ret = compat_do_readv_writev(WRITE, file, vec, vlen, &file->f_pos); 1213 ret = compat_do_readv_writev(WRITE, file, vec, vlen, &file->f_pos);
1211 1214
1212out: 1215out:
1216 if (ret > 0)
1217 add_wchar(current, ret);
1218 inc_syscw(current);
1213 fput(file); 1219 fput(file);
1214 return ret; 1220 return ret;
1215} 1221}
diff --git a/fs/direct-io.c b/fs/direct-io.c
index af0558dbe8b7..b6d43908ff7a 100644
--- a/fs/direct-io.c
+++ b/fs/direct-io.c
@@ -1209,6 +1209,19 @@ __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode,
1209 retval = direct_io_worker(rw, iocb, inode, iov, offset, 1209 retval = direct_io_worker(rw, iocb, inode, iov, offset,
1210 nr_segs, blkbits, get_block, end_io, dio); 1210 nr_segs, blkbits, get_block, end_io, dio);
1211 1211
1212 /*
1213 * In case of error extending write may have instantiated a few
1214 * blocks outside i_size. Trim these off again for DIO_LOCKING.
1215 * NOTE: DIO_NO_LOCK/DIO_OWN_LOCK callers have to handle this by
1216 * it's own meaner.
1217 */
1218 if (unlikely(retval < 0 && (rw & WRITE))) {
1219 loff_t isize = i_size_read(inode);
1220
1221 if (end > isize && dio_lock_type == DIO_LOCKING)
1222 vmtruncate(inode, isize);
1223 }
1224
1212 if (rw == READ && dio_lock_type == DIO_LOCKING) 1225 if (rw == READ && dio_lock_type == DIO_LOCKING)
1213 release_i_mutex = 0; 1226 release_i_mutex = 0;
1214 1227
diff --git a/fs/ecryptfs/crypto.c b/fs/ecryptfs/crypto.c
index 6046239465a1..c01e043670e2 100644
--- a/fs/ecryptfs/crypto.c
+++ b/fs/ecryptfs/crypto.c
@@ -175,8 +175,8 @@ out:
175 * 175 *
176 * Returns zero on success; non-zero on error. 176 * Returns zero on success; non-zero on error.
177 */ 177 */
178static int ecryptfs_derive_iv(char *iv, struct ecryptfs_crypt_stat *crypt_stat, 178int ecryptfs_derive_iv(char *iv, struct ecryptfs_crypt_stat *crypt_stat,
179 loff_t offset) 179 loff_t offset)
180{ 180{
181 int rc = 0; 181 int rc = 0;
182 char dst[MD5_DIGEST_SIZE]; 182 char dst[MD5_DIGEST_SIZE];
@@ -924,6 +924,15 @@ static void ecryptfs_copy_mount_wide_flags_to_inode_flags(
924 crypt_stat->flags |= ECRYPTFS_METADATA_IN_XATTR; 924 crypt_stat->flags |= ECRYPTFS_METADATA_IN_XATTR;
925 if (mount_crypt_stat->flags & ECRYPTFS_ENCRYPTED_VIEW_ENABLED) 925 if (mount_crypt_stat->flags & ECRYPTFS_ENCRYPTED_VIEW_ENABLED)
926 crypt_stat->flags |= ECRYPTFS_VIEW_AS_ENCRYPTED; 926 crypt_stat->flags |= ECRYPTFS_VIEW_AS_ENCRYPTED;
927 if (mount_crypt_stat->flags & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES) {
928 crypt_stat->flags |= ECRYPTFS_ENCRYPT_FILENAMES;
929 if (mount_crypt_stat->flags
930 & ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK)
931 crypt_stat->flags |= ECRYPTFS_ENCFN_USE_MOUNT_FNEK;
932 else if (mount_crypt_stat->flags
933 & ECRYPTFS_GLOBAL_ENCFN_USE_FEK)
934 crypt_stat->flags |= ECRYPTFS_ENCFN_USE_FEK;
935 }
927} 936}
928 937
929static int ecryptfs_copy_mount_wide_sigs_to_inode_sigs( 938static int ecryptfs_copy_mount_wide_sigs_to_inode_sigs(
@@ -1060,7 +1069,8 @@ struct ecryptfs_flag_map_elem {
1060static struct ecryptfs_flag_map_elem ecryptfs_flag_map[] = { 1069static struct ecryptfs_flag_map_elem ecryptfs_flag_map[] = {
1061 {0x00000001, ECRYPTFS_ENABLE_HMAC}, 1070 {0x00000001, ECRYPTFS_ENABLE_HMAC},
1062 {0x00000002, ECRYPTFS_ENCRYPTED}, 1071 {0x00000002, ECRYPTFS_ENCRYPTED},
1063 {0x00000004, ECRYPTFS_METADATA_IN_XATTR} 1072 {0x00000004, ECRYPTFS_METADATA_IN_XATTR},
1073 {0x00000008, ECRYPTFS_ENCRYPT_FILENAMES}
1064}; 1074};
1065 1075
1066/** 1076/**
@@ -1149,19 +1159,20 @@ ecryptfs_cipher_code_str_map[] = {
1149 1159
1150/** 1160/**
1151 * ecryptfs_code_for_cipher_string 1161 * ecryptfs_code_for_cipher_string
1152 * @crypt_stat: The cryptographic context 1162 * @cipher_name: The string alias for the cipher
1163 * @key_bytes: Length of key in bytes; used for AES code selection
1153 * 1164 *
1154 * Returns zero on no match, or the cipher code on match 1165 * Returns zero on no match, or the cipher code on match
1155 */ 1166 */
1156u8 ecryptfs_code_for_cipher_string(struct ecryptfs_crypt_stat *crypt_stat) 1167u8 ecryptfs_code_for_cipher_string(char *cipher_name, size_t key_bytes)
1157{ 1168{
1158 int i; 1169 int i;
1159 u8 code = 0; 1170 u8 code = 0;
1160 struct ecryptfs_cipher_code_str_map_elem *map = 1171 struct ecryptfs_cipher_code_str_map_elem *map =
1161 ecryptfs_cipher_code_str_map; 1172 ecryptfs_cipher_code_str_map;
1162 1173
1163 if (strcmp(crypt_stat->cipher, "aes") == 0) { 1174 if (strcmp(cipher_name, "aes") == 0) {
1164 switch (crypt_stat->key_size) { 1175 switch (key_bytes) {
1165 case 16: 1176 case 16:
1166 code = RFC2440_CIPHER_AES_128; 1177 code = RFC2440_CIPHER_AES_128;
1167 break; 1178 break;
@@ -1173,7 +1184,7 @@ u8 ecryptfs_code_for_cipher_string(struct ecryptfs_crypt_stat *crypt_stat)
1173 } 1184 }
1174 } else { 1185 } else {
1175 for (i = 0; i < ARRAY_SIZE(ecryptfs_cipher_code_str_map); i++) 1186 for (i = 0; i < ARRAY_SIZE(ecryptfs_cipher_code_str_map); i++)
1176 if (strcmp(crypt_stat->cipher, map[i].cipher_str) == 0){ 1187 if (strcmp(cipher_name, map[i].cipher_str) == 0) {
1177 code = map[i].cipher_code; 1188 code = map[i].cipher_code;
1178 break; 1189 break;
1179 } 1190 }
@@ -1212,6 +1223,8 @@ int ecryptfs_read_and_validate_header_region(char *data,
1212 &(ecryptfs_inode_to_private(ecryptfs_inode)->crypt_stat); 1223 &(ecryptfs_inode_to_private(ecryptfs_inode)->crypt_stat);
1213 int rc; 1224 int rc;
1214 1225
1226 if (crypt_stat->extent_size == 0)
1227 crypt_stat->extent_size = ECRYPTFS_DEFAULT_EXTENT_SIZE;
1215 rc = ecryptfs_read_lower(data, 0, crypt_stat->extent_size, 1228 rc = ecryptfs_read_lower(data, 0, crypt_stat->extent_size,
1216 ecryptfs_inode); 1229 ecryptfs_inode);
1217 if (rc) { 1230 if (rc) {
@@ -1221,7 +1234,6 @@ int ecryptfs_read_and_validate_header_region(char *data,
1221 } 1234 }
1222 if (!contains_ecryptfs_marker(data + ECRYPTFS_FILE_SIZE_BYTES)) { 1235 if (!contains_ecryptfs_marker(data + ECRYPTFS_FILE_SIZE_BYTES)) {
1223 rc = -EINVAL; 1236 rc = -EINVAL;
1224 ecryptfs_printk(KERN_DEBUG, "Valid marker not found\n");
1225 } 1237 }
1226out: 1238out:
1227 return rc; 1239 return rc;
@@ -1628,95 +1640,95 @@ out:
1628} 1640}
1629 1641
1630/** 1642/**
1631 * ecryptfs_encode_filename - converts a plaintext file name to cipher text 1643 * ecryptfs_encrypt_filename - encrypt filename
1632 * @crypt_stat: The crypt_stat struct associated with the file anem to encode
1633 * @name: The plaintext name
1634 * @length: The length of the plaintext
1635 * @encoded_name: The encypted name
1636 * 1644 *
1637 * Encrypts and encodes a filename into something that constitutes a 1645 * CBC-encrypts the filename. We do not want to encrypt the same
1638 * valid filename for a filesystem, with printable characters. 1646 * filename with the same key and IV, which may happen with hard
1647 * links, so we prepend random bits to each filename.
1639 * 1648 *
1640 * We assume that we have a properly initialized crypto context, 1649 * Returns zero on success; non-zero otherwise
1641 * pointed to by crypt_stat->tfm.
1642 *
1643 * TODO: Implement filename decoding and decryption here, in place of
1644 * memcpy. We are keeping the framework around for now to (1)
1645 * facilitate testing of the components needed to implement filename
1646 * encryption and (2) to provide a code base from which other
1647 * developers in the community can easily implement this feature.
1648 *
1649 * Returns the length of encoded filename; negative if error
1650 */ 1650 */
1651int 1651static int
1652ecryptfs_encode_filename(struct ecryptfs_crypt_stat *crypt_stat, 1652ecryptfs_encrypt_filename(struct ecryptfs_filename *filename,
1653 const char *name, int length, char **encoded_name) 1653 struct ecryptfs_crypt_stat *crypt_stat,
1654 struct ecryptfs_mount_crypt_stat *mount_crypt_stat)
1654{ 1655{
1655 int error = 0; 1656 int rc = 0;
1656 1657
1657 (*encoded_name) = kmalloc(length + 2, GFP_KERNEL); 1658 filename->encrypted_filename = NULL;
1658 if (!(*encoded_name)) { 1659 filename->encrypted_filename_size = 0;
1659 error = -ENOMEM; 1660 if ((crypt_stat && (crypt_stat->flags & ECRYPTFS_ENCFN_USE_MOUNT_FNEK))
1661 || (mount_crypt_stat && (mount_crypt_stat->flags
1662 & ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK))) {
1663 size_t packet_size;
1664 size_t remaining_bytes;
1665
1666 rc = ecryptfs_write_tag_70_packet(
1667 NULL, NULL,
1668 &filename->encrypted_filename_size,
1669 mount_crypt_stat, NULL,
1670 filename->filename_size);
1671 if (rc) {
1672 printk(KERN_ERR "%s: Error attempting to get packet "
1673 "size for tag 72; rc = [%d]\n", __func__,
1674 rc);
1675 filename->encrypted_filename_size = 0;
1676 goto out;
1677 }
1678 filename->encrypted_filename =
1679 kmalloc(filename->encrypted_filename_size, GFP_KERNEL);
1680 if (!filename->encrypted_filename) {
1681 printk(KERN_ERR "%s: Out of memory whilst attempting "
1682 "to kmalloc [%zd] bytes\n", __func__,
1683 filename->encrypted_filename_size);
1684 rc = -ENOMEM;
1685 goto out;
1686 }
1687 remaining_bytes = filename->encrypted_filename_size;
1688 rc = ecryptfs_write_tag_70_packet(filename->encrypted_filename,
1689 &remaining_bytes,
1690 &packet_size,
1691 mount_crypt_stat,
1692 filename->filename,
1693 filename->filename_size);
1694 if (rc) {
1695 printk(KERN_ERR "%s: Error attempting to generate "
1696 "tag 70 packet; rc = [%d]\n", __func__,
1697 rc);
1698 kfree(filename->encrypted_filename);
1699 filename->encrypted_filename = NULL;
1700 filename->encrypted_filename_size = 0;
1701 goto out;
1702 }
1703 filename->encrypted_filename_size = packet_size;
1704 } else {
1705 printk(KERN_ERR "%s: No support for requested filename "
1706 "encryption method in this release\n", __func__);
1707 rc = -ENOTSUPP;
1660 goto out; 1708 goto out;
1661 } 1709 }
1662 /* TODO: Filename encryption is a scheduled feature for a
1663 * future version of eCryptfs. This function is here only for
1664 * the purpose of providing a framework for other developers
1665 * to easily implement filename encryption. Hint: Replace this
1666 * memcpy() with a call to encrypt and encode the
1667 * filename, the set the length accordingly. */
1668 memcpy((void *)(*encoded_name), (void *)name, length);
1669 (*encoded_name)[length] = '\0';
1670 error = length + 1;
1671out: 1710out:
1672 return error; 1711 return rc;
1673} 1712}
1674 1713
1675/** 1714static int ecryptfs_copy_filename(char **copied_name, size_t *copied_name_size,
1676 * ecryptfs_decode_filename - converts the cipher text name to plaintext 1715 const char *name, size_t name_size)
1677 * @crypt_stat: The crypt_stat struct associated with the file
1678 * @name: The filename in cipher text
1679 * @length: The length of the cipher text name
1680 * @decrypted_name: The plaintext name
1681 *
1682 * Decodes and decrypts the filename.
1683 *
1684 * We assume that we have a properly initialized crypto context,
1685 * pointed to by crypt_stat->tfm.
1686 *
1687 * TODO: Implement filename decoding and decryption here, in place of
1688 * memcpy. We are keeping the framework around for now to (1)
1689 * facilitate testing of the components needed to implement filename
1690 * encryption and (2) to provide a code base from which other
1691 * developers in the community can easily implement this feature.
1692 *
1693 * Returns the length of decoded filename; negative if error
1694 */
1695int
1696ecryptfs_decode_filename(struct ecryptfs_crypt_stat *crypt_stat,
1697 const char *name, int length, char **decrypted_name)
1698{ 1716{
1699 int error = 0; 1717 int rc = 0;
1700 1718
1701 (*decrypted_name) = kmalloc(length + 2, GFP_KERNEL); 1719 (*copied_name) = kmalloc((name_size + 2), GFP_KERNEL);
1702 if (!(*decrypted_name)) { 1720 if (!(*copied_name)) {
1703 error = -ENOMEM; 1721 rc = -ENOMEM;
1704 goto out; 1722 goto out;
1705 } 1723 }
1706 /* TODO: Filename encryption is a scheduled feature for a 1724 memcpy((void *)(*copied_name), (void *)name, name_size);
1707 * future version of eCryptfs. This function is here only for 1725 (*copied_name)[(name_size)] = '\0'; /* Only for convenience
1708 * the purpose of providing a framework for other developers
1709 * to easily implement filename encryption. Hint: Replace this
1710 * memcpy() with a call to decode and decrypt the
1711 * filename, the set the length accordingly. */
1712 memcpy((void *)(*decrypted_name), (void *)name, length);
1713 (*decrypted_name)[length + 1] = '\0'; /* Only for convenience
1714 * in printing out the 1726 * in printing out the
1715 * string in debug 1727 * string in debug
1716 * messages */ 1728 * messages */
1717 error = length; 1729 (*copied_name_size) = (name_size + 1);
1718out: 1730out:
1719 return error; 1731 return rc;
1720} 1732}
1721 1733
1722/** 1734/**
@@ -1740,7 +1752,7 @@ ecryptfs_process_key_cipher(struct crypto_blkcipher **key_tfm,
1740 *key_tfm = NULL; 1752 *key_tfm = NULL;
1741 if (*key_size > ECRYPTFS_MAX_KEY_BYTES) { 1753 if (*key_size > ECRYPTFS_MAX_KEY_BYTES) {
1742 rc = -EINVAL; 1754 rc = -EINVAL;
1743 printk(KERN_ERR "Requested key size is [%Zd] bytes; maximum " 1755 printk(KERN_ERR "Requested key size is [%zd] bytes; maximum "
1744 "allowable is [%d]\n", *key_size, ECRYPTFS_MAX_KEY_BYTES); 1756 "allowable is [%d]\n", *key_size, ECRYPTFS_MAX_KEY_BYTES);
1745 goto out; 1757 goto out;
1746 } 1758 }
@@ -1765,7 +1777,7 @@ ecryptfs_process_key_cipher(struct crypto_blkcipher **key_tfm,
1765 get_random_bytes(dummy_key, *key_size); 1777 get_random_bytes(dummy_key, *key_size);
1766 rc = crypto_blkcipher_setkey(*key_tfm, dummy_key, *key_size); 1778 rc = crypto_blkcipher_setkey(*key_tfm, dummy_key, *key_size);
1767 if (rc) { 1779 if (rc) {
1768 printk(KERN_ERR "Error attempting to set key of size [%Zd] for " 1780 printk(KERN_ERR "Error attempting to set key of size [%zd] for "
1769 "cipher [%s]; rc = [%d]\n", *key_size, cipher_name, rc); 1781 "cipher [%s]; rc = [%d]\n", *key_size, cipher_name, rc);
1770 rc = -EINVAL; 1782 rc = -EINVAL;
1771 goto out; 1783 goto out;
@@ -1910,3 +1922,341 @@ out:
1910 mutex_unlock(&key_tfm_list_mutex); 1922 mutex_unlock(&key_tfm_list_mutex);
1911 return rc; 1923 return rc;
1912} 1924}
1925
1926/* 64 characters forming a 6-bit target field */
1927static unsigned char *portable_filename_chars = ("-.0123456789ABCD"
1928 "EFGHIJKLMNOPQRST"
1929 "UVWXYZabcdefghij"
1930 "klmnopqrstuvwxyz");
1931
1932/* We could either offset on every reverse map or just pad some 0x00's
1933 * at the front here */
1934static const unsigned char filename_rev_map[] = {
1935 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 7 */
1936 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 15 */
1937 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 23 */
1938 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 31 */
1939 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 39 */
1940 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, /* 47 */
1941 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, /* 55 */
1942 0x0A, 0x0B, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* 63 */
1943 0x00, 0x0C, 0x0D, 0x0E, 0x0F, 0x10, 0x11, 0x12, /* 71 */
1944 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1A, /* 79 */
1945 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20, 0x21, 0x22, /* 87 */
1946 0x23, 0x24, 0x25, 0x00, 0x00, 0x00, 0x00, 0x00, /* 95 */
1947 0x00, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, /* 103 */
1948 0x2D, 0x2E, 0x2F, 0x30, 0x31, 0x32, 0x33, 0x34, /* 111 */
1949 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B, 0x3C, /* 119 */
1950 0x3D, 0x3E, 0x3F
1951};
1952
1953/**
1954 * ecryptfs_encode_for_filename
1955 * @dst: Destination location for encoded filename
1956 * @dst_size: Size of the encoded filename in bytes
1957 * @src: Source location for the filename to encode
1958 * @src_size: Size of the source in bytes
1959 */
1960void ecryptfs_encode_for_filename(unsigned char *dst, size_t *dst_size,
1961 unsigned char *src, size_t src_size)
1962{
1963 size_t num_blocks;
1964 size_t block_num = 0;
1965 size_t dst_offset = 0;
1966 unsigned char last_block[3];
1967
1968 if (src_size == 0) {
1969 (*dst_size) = 0;
1970 goto out;
1971 }
1972 num_blocks = (src_size / 3);
1973 if ((src_size % 3) == 0) {
1974 memcpy(last_block, (&src[src_size - 3]), 3);
1975 } else {
1976 num_blocks++;
1977 last_block[2] = 0x00;
1978 switch (src_size % 3) {
1979 case 1:
1980 last_block[0] = src[src_size - 1];
1981 last_block[1] = 0x00;
1982 break;
1983 case 2:
1984 last_block[0] = src[src_size - 2];
1985 last_block[1] = src[src_size - 1];
1986 }
1987 }
1988 (*dst_size) = (num_blocks * 4);
1989 if (!dst)
1990 goto out;
1991 while (block_num < num_blocks) {
1992 unsigned char *src_block;
1993 unsigned char dst_block[4];
1994
1995 if (block_num == (num_blocks - 1))
1996 src_block = last_block;
1997 else
1998 src_block = &src[block_num * 3];
1999 dst_block[0] = ((src_block[0] >> 2) & 0x3F);
2000 dst_block[1] = (((src_block[0] << 4) & 0x30)
2001 | ((src_block[1] >> 4) & 0x0F));
2002 dst_block[2] = (((src_block[1] << 2) & 0x3C)
2003 | ((src_block[2] >> 6) & 0x03));
2004 dst_block[3] = (src_block[2] & 0x3F);
2005 dst[dst_offset++] = portable_filename_chars[dst_block[0]];
2006 dst[dst_offset++] = portable_filename_chars[dst_block[1]];
2007 dst[dst_offset++] = portable_filename_chars[dst_block[2]];
2008 dst[dst_offset++] = portable_filename_chars[dst_block[3]];
2009 block_num++;
2010 }
2011out:
2012 return;
2013}
2014
2015/**
2016 * ecryptfs_decode_from_filename
2017 * @dst: If NULL, this function only sets @dst_size and returns. If
2018 * non-NULL, this function decodes the encoded octets in @src
2019 * into the memory that @dst points to.
2020 * @dst_size: Set to the size of the decoded string.
2021 * @src: The encoded set of octets to decode.
2022 * @src_size: The size of the encoded set of octets to decode.
2023 */
2024static void
2025ecryptfs_decode_from_filename(unsigned char *dst, size_t *dst_size,
2026 const unsigned char *src, size_t src_size)
2027{
2028 u8 current_bit_offset = 0;
2029 size_t src_byte_offset = 0;
2030 size_t dst_byte_offset = 0;
2031
2032 if (dst == NULL) {
2033 /* Not exact; conservatively long. Every block of 4
2034 * encoded characters decodes into a block of 3
2035 * decoded characters. This segment of code provides
2036 * the caller with the maximum amount of allocated
2037 * space that @dst will need to point to in a
2038 * subsequent call. */
2039 (*dst_size) = (((src_size + 1) * 3) / 4);
2040 goto out;
2041 }
2042 while (src_byte_offset < src_size) {
2043 unsigned char src_byte =
2044 filename_rev_map[(int)src[src_byte_offset]];
2045
2046 switch (current_bit_offset) {
2047 case 0:
2048 dst[dst_byte_offset] = (src_byte << 2);
2049 current_bit_offset = 6;
2050 break;
2051 case 6:
2052 dst[dst_byte_offset++] |= (src_byte >> 4);
2053 dst[dst_byte_offset] = ((src_byte & 0xF)
2054 << 4);
2055 current_bit_offset = 4;
2056 break;
2057 case 4:
2058 dst[dst_byte_offset++] |= (src_byte >> 2);
2059 dst[dst_byte_offset] = (src_byte << 6);
2060 current_bit_offset = 2;
2061 break;
2062 case 2:
2063 dst[dst_byte_offset++] |= (src_byte);
2064 dst[dst_byte_offset] = 0;
2065 current_bit_offset = 0;
2066 break;
2067 }
2068 src_byte_offset++;
2069 }
2070 (*dst_size) = dst_byte_offset;
2071out:
2072 return;
2073}
2074
2075/**
2076 * ecryptfs_encrypt_and_encode_filename - converts a plaintext file name to cipher text
2077 * @crypt_stat: The crypt_stat struct associated with the file anem to encode
2078 * @name: The plaintext name
2079 * @length: The length of the plaintext
2080 * @encoded_name: The encypted name
2081 *
2082 * Encrypts and encodes a filename into something that constitutes a
2083 * valid filename for a filesystem, with printable characters.
2084 *
2085 * We assume that we have a properly initialized crypto context,
2086 * pointed to by crypt_stat->tfm.
2087 *
2088 * Returns zero on success; non-zero on otherwise
2089 */
2090int ecryptfs_encrypt_and_encode_filename(
2091 char **encoded_name,
2092 size_t *encoded_name_size,
2093 struct ecryptfs_crypt_stat *crypt_stat,
2094 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
2095 const char *name, size_t name_size)
2096{
2097 size_t encoded_name_no_prefix_size;
2098 int rc = 0;
2099
2100 (*encoded_name) = NULL;
2101 (*encoded_name_size) = 0;
2102 if ((crypt_stat && (crypt_stat->flags & ECRYPTFS_ENCRYPT_FILENAMES))
2103 || (mount_crypt_stat && (mount_crypt_stat->flags
2104 & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES))) {
2105 struct ecryptfs_filename *filename;
2106
2107 filename = kzalloc(sizeof(*filename), GFP_KERNEL);
2108 if (!filename) {
2109 printk(KERN_ERR "%s: Out of memory whilst attempting "
2110 "to kzalloc [%zd] bytes\n", __func__,
2111 sizeof(*filename));
2112 rc = -ENOMEM;
2113 goto out;
2114 }
2115 filename->filename = (char *)name;
2116 filename->filename_size = name_size;
2117 rc = ecryptfs_encrypt_filename(filename, crypt_stat,
2118 mount_crypt_stat);
2119 if (rc) {
2120 printk(KERN_ERR "%s: Error attempting to encrypt "
2121 "filename; rc = [%d]\n", __func__, rc);
2122 kfree(filename);
2123 goto out;
2124 }
2125 ecryptfs_encode_for_filename(
2126 NULL, &encoded_name_no_prefix_size,
2127 filename->encrypted_filename,
2128 filename->encrypted_filename_size);
2129 if ((crypt_stat && (crypt_stat->flags
2130 & ECRYPTFS_ENCFN_USE_MOUNT_FNEK))
2131 || (mount_crypt_stat
2132 && (mount_crypt_stat->flags
2133 & ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK)))
2134 (*encoded_name_size) =
2135 (ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE
2136 + encoded_name_no_prefix_size);
2137 else
2138 (*encoded_name_size) =
2139 (ECRYPTFS_FEK_ENCRYPTED_FILENAME_PREFIX_SIZE
2140 + encoded_name_no_prefix_size);
2141 (*encoded_name) = kmalloc((*encoded_name_size) + 1, GFP_KERNEL);
2142 if (!(*encoded_name)) {
2143 printk(KERN_ERR "%s: Out of memory whilst attempting "
2144 "to kzalloc [%zd] bytes\n", __func__,
2145 (*encoded_name_size));
2146 rc = -ENOMEM;
2147 kfree(filename->encrypted_filename);
2148 kfree(filename);
2149 goto out;
2150 }
2151 if ((crypt_stat && (crypt_stat->flags
2152 & ECRYPTFS_ENCFN_USE_MOUNT_FNEK))
2153 || (mount_crypt_stat
2154 && (mount_crypt_stat->flags
2155 & ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK))) {
2156 memcpy((*encoded_name),
2157 ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX,
2158 ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE);
2159 ecryptfs_encode_for_filename(
2160 ((*encoded_name)
2161 + ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE),
2162 &encoded_name_no_prefix_size,
2163 filename->encrypted_filename,
2164 filename->encrypted_filename_size);
2165 (*encoded_name_size) =
2166 (ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE
2167 + encoded_name_no_prefix_size);
2168 (*encoded_name)[(*encoded_name_size)] = '\0';
2169 (*encoded_name_size)++;
2170 } else {
2171 rc = -ENOTSUPP;
2172 }
2173 if (rc) {
2174 printk(KERN_ERR "%s: Error attempting to encode "
2175 "encrypted filename; rc = [%d]\n", __func__,
2176 rc);
2177 kfree((*encoded_name));
2178 (*encoded_name) = NULL;
2179 (*encoded_name_size) = 0;
2180 }
2181 kfree(filename->encrypted_filename);
2182 kfree(filename);
2183 } else {
2184 rc = ecryptfs_copy_filename(encoded_name,
2185 encoded_name_size,
2186 name, name_size);
2187 }
2188out:
2189 return rc;
2190}
2191
2192/**
2193 * ecryptfs_decode_and_decrypt_filename - converts the encoded cipher text name to decoded plaintext
2194 * @plaintext_name: The plaintext name
2195 * @plaintext_name_size: The plaintext name size
2196 * @ecryptfs_dir_dentry: eCryptfs directory dentry
2197 * @name: The filename in cipher text
2198 * @name_size: The cipher text name size
2199 *
2200 * Decrypts and decodes the filename.
2201 *
2202 * Returns zero on error; non-zero otherwise
2203 */
2204int ecryptfs_decode_and_decrypt_filename(char **plaintext_name,
2205 size_t *plaintext_name_size,
2206 struct dentry *ecryptfs_dir_dentry,
2207 const char *name, size_t name_size)
2208{
2209 char *decoded_name;
2210 size_t decoded_name_size;
2211 size_t packet_size;
2212 int rc = 0;
2213
2214 if ((name_size > ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE)
2215 && (strncmp(name, ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX,
2216 ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE) == 0)) {
2217 struct ecryptfs_mount_crypt_stat *mount_crypt_stat =
2218 &ecryptfs_superblock_to_private(
2219 ecryptfs_dir_dentry->d_sb)->mount_crypt_stat;
2220 const char *orig_name = name;
2221 size_t orig_name_size = name_size;
2222
2223 name += ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE;
2224 name_size -= ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE;
2225 ecryptfs_decode_from_filename(NULL, &decoded_name_size,
2226 name, name_size);
2227 decoded_name = kmalloc(decoded_name_size, GFP_KERNEL);
2228 if (!decoded_name) {
2229 printk(KERN_ERR "%s: Out of memory whilst attempting "
2230 "to kmalloc [%zd] bytes\n", __func__,
2231 decoded_name_size);
2232 rc = -ENOMEM;
2233 goto out;
2234 }
2235 ecryptfs_decode_from_filename(decoded_name, &decoded_name_size,
2236 name, name_size);
2237 rc = ecryptfs_parse_tag_70_packet(plaintext_name,
2238 plaintext_name_size,
2239 &packet_size,
2240 mount_crypt_stat,
2241 decoded_name,
2242 decoded_name_size);
2243 if (rc) {
2244 printk(KERN_INFO "%s: Could not parse tag 70 packet "
2245 "from filename; copying through filename "
2246 "as-is\n", __func__);
2247 rc = ecryptfs_copy_filename(plaintext_name,
2248 plaintext_name_size,
2249 orig_name, orig_name_size);
2250 goto out_free;
2251 }
2252 } else {
2253 rc = ecryptfs_copy_filename(plaintext_name,
2254 plaintext_name_size,
2255 name, name_size);
2256 goto out;
2257 }
2258out_free:
2259 kfree(decoded_name);
2260out:
2261 return rc;
2262}
diff --git a/fs/ecryptfs/ecryptfs_kernel.h b/fs/ecryptfs/ecryptfs_kernel.h
index a75026d35d16..c11fc95714ab 100644
--- a/fs/ecryptfs/ecryptfs_kernel.h
+++ b/fs/ecryptfs/ecryptfs_kernel.h
@@ -51,12 +51,16 @@
51#define ECRYPTFS_VERSIONING_XATTR 0x00000010 51#define ECRYPTFS_VERSIONING_XATTR 0x00000010
52#define ECRYPTFS_VERSIONING_MULTKEY 0x00000020 52#define ECRYPTFS_VERSIONING_MULTKEY 0x00000020
53#define ECRYPTFS_VERSIONING_DEVMISC 0x00000040 53#define ECRYPTFS_VERSIONING_DEVMISC 0x00000040
54#define ECRYPTFS_VERSIONING_HMAC 0x00000080
55#define ECRYPTFS_VERSIONING_FILENAME_ENCRYPTION 0x00000100
56#define ECRYPTFS_VERSIONING_GCM 0x00000200
54#define ECRYPTFS_VERSIONING_MASK (ECRYPTFS_VERSIONING_PASSPHRASE \ 57#define ECRYPTFS_VERSIONING_MASK (ECRYPTFS_VERSIONING_PASSPHRASE \
55 | ECRYPTFS_VERSIONING_PLAINTEXT_PASSTHROUGH \ 58 | ECRYPTFS_VERSIONING_PLAINTEXT_PASSTHROUGH \
56 | ECRYPTFS_VERSIONING_PUBKEY \ 59 | ECRYPTFS_VERSIONING_PUBKEY \
57 | ECRYPTFS_VERSIONING_XATTR \ 60 | ECRYPTFS_VERSIONING_XATTR \
58 | ECRYPTFS_VERSIONING_MULTKEY \ 61 | ECRYPTFS_VERSIONING_MULTKEY \
59 | ECRYPTFS_VERSIONING_DEVMISC) 62 | ECRYPTFS_VERSIONING_DEVMISC \
63 | ECRYPTFS_VERSIONING_FILENAME_ENCRYPTION)
60#define ECRYPTFS_MAX_PASSWORD_LENGTH 64 64#define ECRYPTFS_MAX_PASSWORD_LENGTH 64
61#define ECRYPTFS_MAX_PASSPHRASE_BYTES ECRYPTFS_MAX_PASSWORD_LENGTH 65#define ECRYPTFS_MAX_PASSPHRASE_BYTES ECRYPTFS_MAX_PASSWORD_LENGTH
62#define ECRYPTFS_SALT_SIZE 8 66#define ECRYPTFS_SALT_SIZE 8
@@ -199,6 +203,7 @@ ecryptfs_get_key_payload_data(struct key *key)
199#define ECRYPTFS_DEFAULT_CIPHER "aes" 203#define ECRYPTFS_DEFAULT_CIPHER "aes"
200#define ECRYPTFS_DEFAULT_KEY_BYTES 16 204#define ECRYPTFS_DEFAULT_KEY_BYTES 16
201#define ECRYPTFS_DEFAULT_HASH "md5" 205#define ECRYPTFS_DEFAULT_HASH "md5"
206#define ECRYPTFS_TAG_70_DIGEST ECRYPTFS_DEFAULT_HASH
202#define ECRYPTFS_TAG_1_PACKET_TYPE 0x01 207#define ECRYPTFS_TAG_1_PACKET_TYPE 0x01
203#define ECRYPTFS_TAG_3_PACKET_TYPE 0x8C 208#define ECRYPTFS_TAG_3_PACKET_TYPE 0x8C
204#define ECRYPTFS_TAG_11_PACKET_TYPE 0xED 209#define ECRYPTFS_TAG_11_PACKET_TYPE 0xED
@@ -206,30 +211,64 @@ ecryptfs_get_key_payload_data(struct key *key)
206#define ECRYPTFS_TAG_65_PACKET_TYPE 0x41 211#define ECRYPTFS_TAG_65_PACKET_TYPE 0x41
207#define ECRYPTFS_TAG_66_PACKET_TYPE 0x42 212#define ECRYPTFS_TAG_66_PACKET_TYPE 0x42
208#define ECRYPTFS_TAG_67_PACKET_TYPE 0x43 213#define ECRYPTFS_TAG_67_PACKET_TYPE 0x43
214#define ECRYPTFS_TAG_70_PACKET_TYPE 0x46 /* FNEK-encrypted filename
215 * as dentry name */
216#define ECRYPTFS_TAG_71_PACKET_TYPE 0x47 /* FNEK-encrypted filename in
217 * metadata */
218#define ECRYPTFS_TAG_72_PACKET_TYPE 0x48 /* FEK-encrypted filename as
219 * dentry name */
220#define ECRYPTFS_TAG_73_PACKET_TYPE 0x49 /* FEK-encrypted filename as
221 * metadata */
222/* Constraint: ECRYPTFS_FILENAME_MIN_RANDOM_PREPEND_BYTES >=
223 * ECRYPTFS_MAX_IV_BYTES */
224#define ECRYPTFS_FILENAME_MIN_RANDOM_PREPEND_BYTES 16
225#define ECRYPTFS_NON_NULL 0x42 /* A reasonable substitute for NULL */
209#define MD5_DIGEST_SIZE 16 226#define MD5_DIGEST_SIZE 16
227#define ECRYPTFS_TAG_70_DIGEST_SIZE MD5_DIGEST_SIZE
228#define ECRYPTFS_FEK_ENCRYPTED_FILENAME_PREFIX "ECRYPTFS_FEK_ENCRYPTED."
229#define ECRYPTFS_FEK_ENCRYPTED_FILENAME_PREFIX_SIZE 23
230#define ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX "ECRYPTFS_FNEK_ENCRYPTED."
231#define ECRYPTFS_FNEK_ENCRYPTED_FILENAME_PREFIX_SIZE 24
232#define ECRYPTFS_ENCRYPTED_DENTRY_NAME_LEN (18 + 1 + 4 + 1 + 32)
210 233
211struct ecryptfs_key_sig { 234struct ecryptfs_key_sig {
212 struct list_head crypt_stat_list; 235 struct list_head crypt_stat_list;
213 char keysig[ECRYPTFS_SIG_SIZE_HEX]; 236 char keysig[ECRYPTFS_SIG_SIZE_HEX];
214}; 237};
215 238
239struct ecryptfs_filename {
240 struct list_head crypt_stat_list;
241#define ECRYPTFS_FILENAME_CONTAINS_DECRYPTED 0x00000001
242 u32 flags;
243 u32 seq_no;
244 char *filename;
245 char *encrypted_filename;
246 size_t filename_size;
247 size_t encrypted_filename_size;
248 char fnek_sig[ECRYPTFS_SIG_SIZE_HEX];
249 char dentry_name[ECRYPTFS_ENCRYPTED_DENTRY_NAME_LEN + 1];
250};
251
216/** 252/**
217 * This is the primary struct associated with each encrypted file. 253 * This is the primary struct associated with each encrypted file.
218 * 254 *
219 * TODO: cache align/pack? 255 * TODO: cache align/pack?
220 */ 256 */
221struct ecryptfs_crypt_stat { 257struct ecryptfs_crypt_stat {
222#define ECRYPTFS_STRUCT_INITIALIZED 0x00000001 258#define ECRYPTFS_STRUCT_INITIALIZED 0x00000001
223#define ECRYPTFS_POLICY_APPLIED 0x00000002 259#define ECRYPTFS_POLICY_APPLIED 0x00000002
224#define ECRYPTFS_NEW_FILE 0x00000004 260#define ECRYPTFS_NEW_FILE 0x00000004
225#define ECRYPTFS_ENCRYPTED 0x00000008 261#define ECRYPTFS_ENCRYPTED 0x00000008
226#define ECRYPTFS_SECURITY_WARNING 0x00000010 262#define ECRYPTFS_SECURITY_WARNING 0x00000010
227#define ECRYPTFS_ENABLE_HMAC 0x00000020 263#define ECRYPTFS_ENABLE_HMAC 0x00000020
228#define ECRYPTFS_ENCRYPT_IV_PAGES 0x00000040 264#define ECRYPTFS_ENCRYPT_IV_PAGES 0x00000040
229#define ECRYPTFS_KEY_VALID 0x00000080 265#define ECRYPTFS_KEY_VALID 0x00000080
230#define ECRYPTFS_METADATA_IN_XATTR 0x00000100 266#define ECRYPTFS_METADATA_IN_XATTR 0x00000100
231#define ECRYPTFS_VIEW_AS_ENCRYPTED 0x00000200 267#define ECRYPTFS_VIEW_AS_ENCRYPTED 0x00000200
232#define ECRYPTFS_KEY_SET 0x00000400 268#define ECRYPTFS_KEY_SET 0x00000400
269#define ECRYPTFS_ENCRYPT_FILENAMES 0x00000800
270#define ECRYPTFS_ENCFN_USE_MOUNT_FNEK 0x00001000
271#define ECRYPTFS_ENCFN_USE_FEK 0x00002000
233 u32 flags; 272 u32 flags;
234 unsigned int file_version; 273 unsigned int file_version;
235 size_t iv_bytes; 274 size_t iv_bytes;
@@ -332,13 +371,20 @@ struct ecryptfs_mount_crypt_stat {
332#define ECRYPTFS_XATTR_METADATA_ENABLED 0x00000002 371#define ECRYPTFS_XATTR_METADATA_ENABLED 0x00000002
333#define ECRYPTFS_ENCRYPTED_VIEW_ENABLED 0x00000004 372#define ECRYPTFS_ENCRYPTED_VIEW_ENABLED 0x00000004
334#define ECRYPTFS_MOUNT_CRYPT_STAT_INITIALIZED 0x00000008 373#define ECRYPTFS_MOUNT_CRYPT_STAT_INITIALIZED 0x00000008
374#define ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES 0x00000010
375#define ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK 0x00000020
376#define ECRYPTFS_GLOBAL_ENCFN_USE_FEK 0x00000040
335 u32 flags; 377 u32 flags;
336 struct list_head global_auth_tok_list; 378 struct list_head global_auth_tok_list;
337 struct mutex global_auth_tok_list_mutex; 379 struct mutex global_auth_tok_list_mutex;
338 size_t num_global_auth_toks; 380 size_t num_global_auth_toks;
339 size_t global_default_cipher_key_size; 381 size_t global_default_cipher_key_size;
382 size_t global_default_fn_cipher_key_bytes;
340 unsigned char global_default_cipher_name[ECRYPTFS_MAX_CIPHER_NAME_SIZE 383 unsigned char global_default_cipher_name[ECRYPTFS_MAX_CIPHER_NAME_SIZE
341 + 1]; 384 + 1];
385 unsigned char global_default_fn_cipher_name[
386 ECRYPTFS_MAX_CIPHER_NAME_SIZE + 1];
387 char global_default_fnek_sig[ECRYPTFS_SIG_SIZE_HEX + 1];
342}; 388};
343 389
344/* superblock private data. */ 390/* superblock private data. */
@@ -571,13 +617,22 @@ struct ecryptfs_open_req {
571int ecryptfs_interpose(struct dentry *hidden_dentry, 617int ecryptfs_interpose(struct dentry *hidden_dentry,
572 struct dentry *this_dentry, struct super_block *sb, 618 struct dentry *this_dentry, struct super_block *sb,
573 u32 flags); 619 u32 flags);
620int ecryptfs_lookup_and_interpose_lower(struct dentry *ecryptfs_dentry,
621 struct dentry *lower_dentry,
622 struct ecryptfs_crypt_stat *crypt_stat,
623 struct inode *ecryptfs_dir_inode,
624 struct nameidata *ecryptfs_nd);
625int ecryptfs_decode_and_decrypt_filename(char **decrypted_name,
626 size_t *decrypted_name_size,
627 struct dentry *ecryptfs_dentry,
628 const char *name, size_t name_size);
574int ecryptfs_fill_zeros(struct file *file, loff_t new_length); 629int ecryptfs_fill_zeros(struct file *file, loff_t new_length);
575int ecryptfs_decode_filename(struct ecryptfs_crypt_stat *crypt_stat, 630int ecryptfs_encrypt_and_encode_filename(
576 const char *name, int length, 631 char **encoded_name,
577 char **decrypted_name); 632 size_t *encoded_name_size,
578int ecryptfs_encode_filename(struct ecryptfs_crypt_stat *crypt_stat, 633 struct ecryptfs_crypt_stat *crypt_stat,
579 const char *name, int length, 634 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
580 char **encoded_name); 635 const char *name, size_t name_size);
581struct dentry *ecryptfs_lower_dentry(struct dentry *this_dentry); 636struct dentry *ecryptfs_lower_dentry(struct dentry *this_dentry);
582void ecryptfs_dump_hex(char *data, int bytes); 637void ecryptfs_dump_hex(char *data, int bytes);
583int virt_to_scatterlist(const void *addr, int size, struct scatterlist *sg, 638int virt_to_scatterlist(const void *addr, int size, struct scatterlist *sg,
@@ -599,7 +654,7 @@ int ecryptfs_read_and_validate_header_region(char *data,
599 struct inode *ecryptfs_inode); 654 struct inode *ecryptfs_inode);
600int ecryptfs_read_and_validate_xattr_region(char *page_virt, 655int ecryptfs_read_and_validate_xattr_region(char *page_virt,
601 struct dentry *ecryptfs_dentry); 656 struct dentry *ecryptfs_dentry);
602u8 ecryptfs_code_for_cipher_string(struct ecryptfs_crypt_stat *crypt_stat); 657u8 ecryptfs_code_for_cipher_string(char *cipher_name, size_t key_bytes);
603int ecryptfs_cipher_code_to_string(char *str, u8 cipher_code); 658int ecryptfs_cipher_code_to_string(char *str, u8 cipher_code);
604void ecryptfs_set_default_sizes(struct ecryptfs_crypt_stat *crypt_stat); 659void ecryptfs_set_default_sizes(struct ecryptfs_crypt_stat *crypt_stat);
605int ecryptfs_generate_key_packet_set(char *dest_base, 660int ecryptfs_generate_key_packet_set(char *dest_base,
@@ -694,5 +749,17 @@ int ecryptfs_privileged_open(struct file **lower_file,
694 struct vfsmount *lower_mnt, 749 struct vfsmount *lower_mnt,
695 const struct cred *cred); 750 const struct cred *cred);
696int ecryptfs_init_persistent_file(struct dentry *ecryptfs_dentry); 751int ecryptfs_init_persistent_file(struct dentry *ecryptfs_dentry);
752int
753ecryptfs_write_tag_70_packet(char *dest, size_t *remaining_bytes,
754 size_t *packet_size,
755 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
756 char *filename, size_t filename_size);
757int
758ecryptfs_parse_tag_70_packet(char **filename, size_t *filename_size,
759 size_t *packet_size,
760 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
761 char *data, size_t max_packet_size);
762int ecryptfs_derive_iv(char *iv, struct ecryptfs_crypt_stat *crypt_stat,
763 loff_t offset);
697 764
698#endif /* #ifndef ECRYPTFS_KERNEL_H */ 765#endif /* #ifndef ECRYPTFS_KERNEL_H */
diff --git a/fs/ecryptfs/file.c b/fs/ecryptfs/file.c
index 713834371229..9e944057001b 100644
--- a/fs/ecryptfs/file.c
+++ b/fs/ecryptfs/file.c
@@ -77,27 +77,27 @@ struct ecryptfs_getdents_callback {
77 77
78/* Inspired by generic filldir in fs/readdir.c */ 78/* Inspired by generic filldir in fs/readdir.c */
79static int 79static int
80ecryptfs_filldir(void *dirent, const char *name, int namelen, loff_t offset, 80ecryptfs_filldir(void *dirent, const char *lower_name, int lower_namelen,
81 u64 ino, unsigned int d_type) 81 loff_t offset, u64 ino, unsigned int d_type)
82{ 82{
83 struct ecryptfs_crypt_stat *crypt_stat;
84 struct ecryptfs_getdents_callback *buf = 83 struct ecryptfs_getdents_callback *buf =
85 (struct ecryptfs_getdents_callback *)dirent; 84 (struct ecryptfs_getdents_callback *)dirent;
85 size_t name_size;
86 char *name;
86 int rc; 87 int rc;
87 int decoded_length;
88 char *decoded_name;
89 88
90 crypt_stat = ecryptfs_dentry_to_private(buf->dentry)->crypt_stat;
91 buf->filldir_called++; 89 buf->filldir_called++;
92 decoded_length = ecryptfs_decode_filename(crypt_stat, name, namelen, 90 rc = ecryptfs_decode_and_decrypt_filename(&name, &name_size,
93 &decoded_name); 91 buf->dentry, lower_name,
94 if (decoded_length < 0) { 92 lower_namelen);
95 rc = decoded_length; 93 if (rc) {
94 printk(KERN_ERR "%s: Error attempting to decode and decrypt "
95 "filename [%s]; rc = [%d]\n", __func__, lower_name,
96 rc);
96 goto out; 97 goto out;
97 } 98 }
98 rc = buf->filldir(buf->dirent, decoded_name, decoded_length, offset, 99 rc = buf->filldir(buf->dirent, name, name_size, offset, ino, d_type);
99 ino, d_type); 100 kfree(name);
100 kfree(decoded_name);
101 if (rc >= 0) 101 if (rc >= 0)
102 buf->entries_written++; 102 buf->entries_written++;
103out: 103out:
@@ -106,8 +106,8 @@ out:
106 106
107/** 107/**
108 * ecryptfs_readdir 108 * ecryptfs_readdir
109 * @file: The ecryptfs file struct 109 * @file: The eCryptfs directory file
110 * @dirent: Directory entry 110 * @dirent: Directory entry handle
111 * @filldir: The filldir callback function 111 * @filldir: The filldir callback function
112 */ 112 */
113static int ecryptfs_readdir(struct file *file, void *dirent, filldir_t filldir) 113static int ecryptfs_readdir(struct file *file, void *dirent, filldir_t filldir)
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index 0111906a8877..5697899a168d 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -52,8 +52,7 @@ static void unlock_dir(struct dentry *dir)
52/** 52/**
53 * ecryptfs_create_underlying_file 53 * ecryptfs_create_underlying_file
54 * @lower_dir_inode: inode of the parent in the lower fs of the new file 54 * @lower_dir_inode: inode of the parent in the lower fs of the new file
55 * @lower_dentry: New file's dentry in the lower fs 55 * @dentry: New file's dentry
56 * @ecryptfs_dentry: New file's dentry in ecryptfs
57 * @mode: The mode of the new file 56 * @mode: The mode of the new file
58 * @nd: nameidata of ecryptfs' parent's dentry & vfsmount 57 * @nd: nameidata of ecryptfs' parent's dentry & vfsmount
59 * 58 *
@@ -228,8 +227,7 @@ ecryptfs_create(struct inode *directory_inode, struct dentry *ecryptfs_dentry,
228{ 227{
229 int rc; 228 int rc;
230 229
231 /* ecryptfs_do_create() calls ecryptfs_interpose(), which opens 230 /* ecryptfs_do_create() calls ecryptfs_interpose() */
232 * the crypt_stat->lower_file (persistent file) */
233 rc = ecryptfs_do_create(directory_inode, ecryptfs_dentry, mode, nd); 231 rc = ecryptfs_do_create(directory_inode, ecryptfs_dentry, mode, nd);
234 if (unlikely(rc)) { 232 if (unlikely(rc)) {
235 ecryptfs_printk(KERN_WARNING, "Failed to create file in" 233 ecryptfs_printk(KERN_WARNING, "Failed to create file in"
@@ -244,141 +242,91 @@ out:
244} 242}
245 243
246/** 244/**
247 * ecryptfs_lookup 245 * ecryptfs_lookup_and_interpose_lower - Perform a lookup
248 * @dir: inode
249 * @dentry: The dentry
250 * @nd: nameidata, may be NULL
251 *
252 * Find a file on disk. If the file does not exist, then we'll add it to the
253 * dentry cache and continue on to read it from the disk.
254 */ 246 */
255static struct dentry *ecryptfs_lookup(struct inode *dir, struct dentry *dentry, 247int ecryptfs_lookup_and_interpose_lower(struct dentry *ecryptfs_dentry,
256 struct nameidata *nd) 248 struct dentry *lower_dentry,
249 struct ecryptfs_crypt_stat *crypt_stat,
250 struct inode *ecryptfs_dir_inode,
251 struct nameidata *ecryptfs_nd)
257{ 252{
258 int rc = 0;
259 struct dentry *lower_dir_dentry; 253 struct dentry *lower_dir_dentry;
260 struct dentry *lower_dentry;
261 struct vfsmount *lower_mnt; 254 struct vfsmount *lower_mnt;
262 char *encoded_name; 255 struct inode *lower_inode;
263 int encoded_namelen;
264 struct ecryptfs_crypt_stat *crypt_stat = NULL;
265 struct ecryptfs_mount_crypt_stat *mount_crypt_stat; 256 struct ecryptfs_mount_crypt_stat *mount_crypt_stat;
266 char *page_virt = NULL; 257 char *page_virt = NULL;
267 struct inode *lower_inode;
268 u64 file_size; 258 u64 file_size;
259 int rc = 0;
269 260
270 lower_dir_dentry = ecryptfs_dentry_to_lower(dentry->d_parent); 261 lower_dir_dentry = lower_dentry->d_parent;
271 dentry->d_op = &ecryptfs_dops; 262 lower_mnt = mntget(ecryptfs_dentry_to_lower_mnt(
272 if ((dentry->d_name.len == 1 && !strcmp(dentry->d_name.name, ".")) 263 ecryptfs_dentry->d_parent));
273 || (dentry->d_name.len == 2
274 && !strcmp(dentry->d_name.name, ".."))) {
275 d_drop(dentry);
276 goto out;
277 }
278 encoded_namelen = ecryptfs_encode_filename(crypt_stat,
279 dentry->d_name.name,
280 dentry->d_name.len,
281 &encoded_name);
282 if (encoded_namelen < 0) {
283 rc = encoded_namelen;
284 d_drop(dentry);
285 goto out;
286 }
287 ecryptfs_printk(KERN_DEBUG, "encoded_name = [%s]; encoded_namelen "
288 "= [%d]\n", encoded_name, encoded_namelen);
289 lower_dentry = lookup_one_len(encoded_name, lower_dir_dentry,
290 encoded_namelen - 1);
291 kfree(encoded_name);
292 if (IS_ERR(lower_dentry)) {
293 ecryptfs_printk(KERN_ERR, "ERR from lower_dentry\n");
294 rc = PTR_ERR(lower_dentry);
295 d_drop(dentry);
296 goto out;
297 }
298 lower_mnt = mntget(ecryptfs_dentry_to_lower_mnt(dentry->d_parent));
299 ecryptfs_printk(KERN_DEBUG, "lower_dentry = [%p]; lower_dentry->"
300 "d_name.name = [%s]\n", lower_dentry,
301 lower_dentry->d_name.name);
302 lower_inode = lower_dentry->d_inode; 264 lower_inode = lower_dentry->d_inode;
303 fsstack_copy_attr_atime(dir, lower_dir_dentry->d_inode); 265 fsstack_copy_attr_atime(ecryptfs_dir_inode, lower_dir_dentry->d_inode);
304 BUG_ON(!atomic_read(&lower_dentry->d_count)); 266 BUG_ON(!atomic_read(&lower_dentry->d_count));
305 ecryptfs_set_dentry_private(dentry, 267 ecryptfs_set_dentry_private(ecryptfs_dentry,
306 kmem_cache_alloc(ecryptfs_dentry_info_cache, 268 kmem_cache_alloc(ecryptfs_dentry_info_cache,
307 GFP_KERNEL)); 269 GFP_KERNEL));
308 if (!ecryptfs_dentry_to_private(dentry)) { 270 if (!ecryptfs_dentry_to_private(ecryptfs_dentry)) {
309 rc = -ENOMEM; 271 rc = -ENOMEM;
310 ecryptfs_printk(KERN_ERR, "Out of memory whilst attempting " 272 printk(KERN_ERR "%s: Out of memory whilst attempting "
311 "to allocate ecryptfs_dentry_info struct\n"); 273 "to allocate ecryptfs_dentry_info struct\n",
274 __func__);
312 goto out_dput; 275 goto out_dput;
313 } 276 }
314 ecryptfs_set_dentry_lower(dentry, lower_dentry); 277 ecryptfs_set_dentry_lower(ecryptfs_dentry, lower_dentry);
315 ecryptfs_set_dentry_lower_mnt(dentry, lower_mnt); 278 ecryptfs_set_dentry_lower_mnt(ecryptfs_dentry, lower_mnt);
316 if (!lower_dentry->d_inode) { 279 if (!lower_dentry->d_inode) {
317 /* We want to add because we couldn't find in lower */ 280 /* We want to add because we couldn't find in lower */
318 d_add(dentry, NULL); 281 d_add(ecryptfs_dentry, NULL);
319 goto out; 282 goto out;
320 } 283 }
321 rc = ecryptfs_interpose(lower_dentry, dentry, dir->i_sb, 284 rc = ecryptfs_interpose(lower_dentry, ecryptfs_dentry,
322 ECRYPTFS_INTERPOSE_FLAG_D_ADD); 285 ecryptfs_dir_inode->i_sb, 1);
323 if (rc) { 286 if (rc) {
324 ecryptfs_printk(KERN_ERR, "Error interposing\n"); 287 printk(KERN_ERR "%s: Error interposing; rc = [%d]\n",
288 __func__, rc);
325 goto out; 289 goto out;
326 } 290 }
327 if (S_ISDIR(lower_inode->i_mode)) { 291 if (S_ISDIR(lower_inode->i_mode))
328 ecryptfs_printk(KERN_DEBUG, "Is a directory; returning\n");
329 goto out; 292 goto out;
330 } 293 if (S_ISLNK(lower_inode->i_mode))
331 if (S_ISLNK(lower_inode->i_mode)) {
332 ecryptfs_printk(KERN_DEBUG, "Is a symlink; returning\n");
333 goto out; 294 goto out;
334 } 295 if (special_file(lower_inode->i_mode))
335 if (special_file(lower_inode->i_mode)) {
336 ecryptfs_printk(KERN_DEBUG, "Is a special file; returning\n");
337 goto out; 296 goto out;
338 } 297 if (!ecryptfs_nd)
339 if (!nd) {
340 ecryptfs_printk(KERN_DEBUG, "We have a NULL nd, just leave"
341 "as we *think* we are about to unlink\n");
342 goto out; 298 goto out;
343 }
344 /* Released in this function */ 299 /* Released in this function */
345 page_virt = kmem_cache_zalloc(ecryptfs_header_cache_2, 300 page_virt = kmem_cache_zalloc(ecryptfs_header_cache_2, GFP_USER);
346 GFP_USER);
347 if (!page_virt) { 301 if (!page_virt) {
302 printk(KERN_ERR "%s: Cannot kmem_cache_zalloc() a page\n",
303 __func__);
348 rc = -ENOMEM; 304 rc = -ENOMEM;
349 ecryptfs_printk(KERN_ERR,
350 "Cannot ecryptfs_kmalloc a page\n");
351 goto out; 305 goto out;
352 } 306 }
353 crypt_stat = &ecryptfs_inode_to_private(dentry->d_inode)->crypt_stat; 307 if (!ecryptfs_inode_to_private(ecryptfs_dentry->d_inode)->lower_file) {
354 if (!(crypt_stat->flags & ECRYPTFS_POLICY_APPLIED)) 308 rc = ecryptfs_init_persistent_file(ecryptfs_dentry);
355 ecryptfs_set_default_sizes(crypt_stat);
356 if (!ecryptfs_inode_to_private(dentry->d_inode)->lower_file) {
357 rc = ecryptfs_init_persistent_file(dentry);
358 if (rc) { 309 if (rc) {
359 printk(KERN_ERR "%s: Error attempting to initialize " 310 printk(KERN_ERR "%s: Error attempting to initialize "
360 "the persistent file for the dentry with name " 311 "the persistent file for the dentry with name "
361 "[%s]; rc = [%d]\n", __func__, 312 "[%s]; rc = [%d]\n", __func__,
362 dentry->d_name.name, rc); 313 ecryptfs_dentry->d_name.name, rc);
363 goto out; 314 goto out_free_kmem;
364 } 315 }
365 } 316 }
366 rc = ecryptfs_read_and_validate_header_region(page_virt, 317 rc = ecryptfs_read_and_validate_header_region(page_virt,
367 dentry->d_inode); 318 ecryptfs_dentry->d_inode);
368 if (rc) { 319 if (rc) {
369 rc = ecryptfs_read_and_validate_xattr_region(page_virt, dentry); 320 rc = ecryptfs_read_and_validate_xattr_region(page_virt,
321 ecryptfs_dentry);
370 if (rc) { 322 if (rc) {
371 printk(KERN_DEBUG "Valid metadata not found in header "
372 "region or xattr region; treating file as "
373 "unencrypted\n");
374 rc = 0; 323 rc = 0;
375 kmem_cache_free(ecryptfs_header_cache_2, page_virt); 324 goto out_free_kmem;
376 goto out;
377 } 325 }
378 crypt_stat->flags |= ECRYPTFS_METADATA_IN_XATTR; 326 crypt_stat->flags |= ECRYPTFS_METADATA_IN_XATTR;
379 } 327 }
380 mount_crypt_stat = &ecryptfs_superblock_to_private( 328 mount_crypt_stat = &ecryptfs_superblock_to_private(
381 dentry->d_sb)->mount_crypt_stat; 329 ecryptfs_dentry->d_sb)->mount_crypt_stat;
382 if (mount_crypt_stat->flags & ECRYPTFS_ENCRYPTED_VIEW_ENABLED) { 330 if (mount_crypt_stat->flags & ECRYPTFS_ENCRYPTED_VIEW_ENABLED) {
383 if (crypt_stat->flags & ECRYPTFS_METADATA_IN_XATTR) 331 if (crypt_stat->flags & ECRYPTFS_METADATA_IN_XATTR)
384 file_size = (crypt_stat->num_header_bytes_at_front 332 file_size = (crypt_stat->num_header_bytes_at_front
@@ -388,14 +336,103 @@ static struct dentry *ecryptfs_lookup(struct inode *dir, struct dentry *dentry,
388 } else { 336 } else {
389 file_size = get_unaligned_be64(page_virt); 337 file_size = get_unaligned_be64(page_virt);
390 } 338 }
391 i_size_write(dentry->d_inode, (loff_t)file_size); 339 i_size_write(ecryptfs_dentry->d_inode, (loff_t)file_size);
340out_free_kmem:
392 kmem_cache_free(ecryptfs_header_cache_2, page_virt); 341 kmem_cache_free(ecryptfs_header_cache_2, page_virt);
393 goto out; 342 goto out;
394
395out_dput: 343out_dput:
396 dput(lower_dentry); 344 dput(lower_dentry);
397 d_drop(dentry); 345 d_drop(ecryptfs_dentry);
346out:
347 return rc;
348}
349
350/**
351 * ecryptfs_lookup
352 * @ecryptfs_dir_inode: The eCryptfs directory inode
353 * @ecryptfs_dentry: The eCryptfs dentry that we are looking up
354 * @ecryptfs_nd: nameidata; may be NULL
355 *
356 * Find a file on disk. If the file does not exist, then we'll add it to the
357 * dentry cache and continue on to read it from the disk.
358 */
359static struct dentry *ecryptfs_lookup(struct inode *ecryptfs_dir_inode,
360 struct dentry *ecryptfs_dentry,
361 struct nameidata *ecryptfs_nd)
362{
363 char *encrypted_and_encoded_name = NULL;
364 size_t encrypted_and_encoded_name_size;
365 struct ecryptfs_crypt_stat *crypt_stat = NULL;
366 struct ecryptfs_mount_crypt_stat *mount_crypt_stat = NULL;
367 struct ecryptfs_inode_info *inode_info;
368 struct dentry *lower_dir_dentry, *lower_dentry;
369 int rc = 0;
370
371 ecryptfs_dentry->d_op = &ecryptfs_dops;
372 if ((ecryptfs_dentry->d_name.len == 1
373 && !strcmp(ecryptfs_dentry->d_name.name, "."))
374 || (ecryptfs_dentry->d_name.len == 2
375 && !strcmp(ecryptfs_dentry->d_name.name, ".."))) {
376 goto out_d_drop;
377 }
378 lower_dir_dentry = ecryptfs_dentry_to_lower(ecryptfs_dentry->d_parent);
379 lower_dentry = lookup_one_len(ecryptfs_dentry->d_name.name,
380 lower_dir_dentry,
381 ecryptfs_dentry->d_name.len);
382 if (IS_ERR(lower_dentry)) {
383 rc = PTR_ERR(lower_dentry);
384 printk(KERN_ERR "%s: lookup_one_len() returned [%d] on "
385 "lower_dentry = [%s]\n", __func__, rc,
386 ecryptfs_dentry->d_name.name);
387 goto out_d_drop;
388 }
389 if (lower_dentry->d_inode)
390 goto lookup_and_interpose;
391 inode_info = ecryptfs_inode_to_private(ecryptfs_dentry->d_inode);
392 if (inode_info) {
393 crypt_stat = &inode_info->crypt_stat;
394 /* TODO: lock for crypt_stat comparison */
395 if (!(crypt_stat->flags & ECRYPTFS_POLICY_APPLIED))
396 ecryptfs_set_default_sizes(crypt_stat);
397 }
398 if (crypt_stat)
399 mount_crypt_stat = crypt_stat->mount_crypt_stat;
400 else
401 mount_crypt_stat = &ecryptfs_superblock_to_private(
402 ecryptfs_dentry->d_sb)->mount_crypt_stat;
403 if (!(crypt_stat && (crypt_stat->flags & ECRYPTFS_ENCRYPT_FILENAMES))
404 && !(mount_crypt_stat && (mount_crypt_stat->flags
405 & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES)))
406 goto lookup_and_interpose;
407 dput(lower_dentry);
408 rc = ecryptfs_encrypt_and_encode_filename(
409 &encrypted_and_encoded_name, &encrypted_and_encoded_name_size,
410 crypt_stat, mount_crypt_stat, ecryptfs_dentry->d_name.name,
411 ecryptfs_dentry->d_name.len);
412 if (rc) {
413 printk(KERN_ERR "%s: Error attempting to encrypt and encode "
414 "filename; rc = [%d]\n", __func__, rc);
415 goto out_d_drop;
416 }
417 lower_dentry = lookup_one_len(encrypted_and_encoded_name,
418 lower_dir_dentry,
419 encrypted_and_encoded_name_size - 1);
420 if (IS_ERR(lower_dentry)) {
421 rc = PTR_ERR(lower_dentry);
422 printk(KERN_ERR "%s: lookup_one_len() returned [%d] on "
423 "lower_dentry = [%s]\n", __func__, rc,
424 encrypted_and_encoded_name);
425 goto out_d_drop;
426 }
427lookup_and_interpose:
428 rc = ecryptfs_lookup_and_interpose_lower(ecryptfs_dentry, lower_dentry,
429 crypt_stat, ecryptfs_dir_inode,
430 ecryptfs_nd);
431 goto out;
432out_d_drop:
433 d_drop(ecryptfs_dentry);
398out: 434out:
435 kfree(encrypted_and_encoded_name);
399 return ERR_PTR(rc); 436 return ERR_PTR(rc);
400} 437}
401 438
@@ -466,19 +503,21 @@ static int ecryptfs_symlink(struct inode *dir, struct dentry *dentry,
466 struct dentry *lower_dentry; 503 struct dentry *lower_dentry;
467 struct dentry *lower_dir_dentry; 504 struct dentry *lower_dir_dentry;
468 char *encoded_symname; 505 char *encoded_symname;
469 int encoded_symlen; 506 size_t encoded_symlen;
470 struct ecryptfs_crypt_stat *crypt_stat = NULL; 507 struct ecryptfs_mount_crypt_stat *mount_crypt_stat = NULL;
471 508
472 lower_dentry = ecryptfs_dentry_to_lower(dentry); 509 lower_dentry = ecryptfs_dentry_to_lower(dentry);
473 dget(lower_dentry); 510 dget(lower_dentry);
474 lower_dir_dentry = lock_parent(lower_dentry); 511 lower_dir_dentry = lock_parent(lower_dentry);
475 encoded_symlen = ecryptfs_encode_filename(crypt_stat, symname, 512 mount_crypt_stat = &ecryptfs_superblock_to_private(
476 strlen(symname), 513 dir->i_sb)->mount_crypt_stat;
477 &encoded_symname); 514 rc = ecryptfs_encrypt_and_encode_filename(&encoded_symname,
478 if (encoded_symlen < 0) { 515 &encoded_symlen,
479 rc = encoded_symlen; 516 NULL,
517 mount_crypt_stat, symname,
518 strlen(symname));
519 if (rc)
480 goto out_lock; 520 goto out_lock;
481 }
482 rc = vfs_symlink(lower_dir_dentry->d_inode, lower_dentry, 521 rc = vfs_symlink(lower_dir_dentry->d_inode, lower_dentry,
483 encoded_symname); 522 encoded_symname);
484 kfree(encoded_symname); 523 kfree(encoded_symname);
@@ -602,52 +641,54 @@ out_lock:
602} 641}
603 642
604static int 643static int
605ecryptfs_readlink(struct dentry *dentry, char __user * buf, int bufsiz) 644ecryptfs_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
606{ 645{
607 int rc;
608 struct dentry *lower_dentry;
609 char *decoded_name;
610 char *lower_buf; 646 char *lower_buf;
611 mm_segment_t old_fs; 647 struct dentry *lower_dentry;
612 struct ecryptfs_crypt_stat *crypt_stat; 648 struct ecryptfs_crypt_stat *crypt_stat;
649 char *plaintext_name;
650 size_t plaintext_name_size;
651 mm_segment_t old_fs;
652 int rc;
613 653
614 lower_dentry = ecryptfs_dentry_to_lower(dentry); 654 lower_dentry = ecryptfs_dentry_to_lower(dentry);
615 if (!lower_dentry->d_inode->i_op->readlink) { 655 if (!lower_dentry->d_inode->i_op->readlink) {
616 rc = -EINVAL; 656 rc = -EINVAL;
617 goto out; 657 goto out;
618 } 658 }
659 crypt_stat = &ecryptfs_inode_to_private(dentry->d_inode)->crypt_stat;
619 /* Released in this function */ 660 /* Released in this function */
620 lower_buf = kmalloc(bufsiz, GFP_KERNEL); 661 lower_buf = kmalloc(bufsiz, GFP_KERNEL);
621 if (lower_buf == NULL) { 662 if (lower_buf == NULL) {
622 ecryptfs_printk(KERN_ERR, "Out of memory\n"); 663 printk(KERN_ERR "%s: Out of memory whilst attempting to "
664 "kmalloc [%d] bytes\n", __func__, bufsiz);
623 rc = -ENOMEM; 665 rc = -ENOMEM;
624 goto out; 666 goto out;
625 } 667 }
626 old_fs = get_fs(); 668 old_fs = get_fs();
627 set_fs(get_ds()); 669 set_fs(get_ds());
628 ecryptfs_printk(KERN_DEBUG, "Calling readlink w/ "
629 "lower_dentry->d_name.name = [%s]\n",
630 lower_dentry->d_name.name);
631 rc = lower_dentry->d_inode->i_op->readlink(lower_dentry, 670 rc = lower_dentry->d_inode->i_op->readlink(lower_dentry,
632 (char __user *)lower_buf, 671 (char __user *)lower_buf,
633 bufsiz); 672 bufsiz);
634 set_fs(old_fs); 673 set_fs(old_fs);
635 if (rc >= 0) { 674 if (rc >= 0) {
636 crypt_stat = NULL; 675 rc = ecryptfs_decode_and_decrypt_filename(&plaintext_name,
637 rc = ecryptfs_decode_filename(crypt_stat, lower_buf, rc, 676 &plaintext_name_size,
638 &decoded_name); 677 dentry, lower_buf,
639 if (rc == -ENOMEM) 678 rc);
679 if (rc) {
680 printk(KERN_ERR "%s: Error attempting to decode and "
681 "decrypt filename; rc = [%d]\n", __func__,
682 rc);
640 goto out_free_lower_buf; 683 goto out_free_lower_buf;
641 if (rc > 0) {
642 ecryptfs_printk(KERN_DEBUG, "Copying [%d] bytes "
643 "to userspace: [%*s]\n", rc,
644 decoded_name);
645 if (copy_to_user(buf, decoded_name, rc))
646 rc = -EFAULT;
647 } 684 }
648 kfree(decoded_name); 685 rc = copy_to_user(buf, plaintext_name, plaintext_name_size);
649 fsstack_copy_attr_atime(dentry->d_inode, 686 if (rc)
650 lower_dentry->d_inode); 687 rc = -EFAULT;
688 else
689 rc = plaintext_name_size;
690 kfree(plaintext_name);
691 fsstack_copy_attr_atime(dentry->d_inode, lower_dentry->d_inode);
651 } 692 }
652out_free_lower_buf: 693out_free_lower_buf:
653 kfree(lower_buf); 694 kfree(lower_buf);
@@ -669,8 +710,6 @@ static void *ecryptfs_follow_link(struct dentry *dentry, struct nameidata *nd)
669 } 710 }
670 old_fs = get_fs(); 711 old_fs = get_fs();
671 set_fs(get_ds()); 712 set_fs(get_ds());
672 ecryptfs_printk(KERN_DEBUG, "Calling readlink w/ "
673 "dentry->d_name.name = [%s]\n", dentry->d_name.name);
674 rc = dentry->d_inode->i_op->readlink(dentry, (char __user *)buf, len); 713 rc = dentry->d_inode->i_op->readlink(dentry, (char __user *)buf, len);
675 set_fs(old_fs); 714 set_fs(old_fs);
676 if (rc < 0) 715 if (rc < 0)
diff --git a/fs/ecryptfs/keystore.c b/fs/ecryptfs/keystore.c
index 0d713b691941..ff539420cc6f 100644
--- a/fs/ecryptfs/keystore.c
+++ b/fs/ecryptfs/keystore.c
@@ -358,7 +358,7 @@ parse_tag_67_packet(struct ecryptfs_key_record *key_rec,
358 /* verify that everything through the encrypted FEK size is present */ 358 /* verify that everything through the encrypted FEK size is present */
359 if (message_len < 4) { 359 if (message_len < 4) {
360 rc = -EIO; 360 rc = -EIO;
361 printk(KERN_ERR "%s: message_len is [%Zd]; minimum acceptable " 361 printk(KERN_ERR "%s: message_len is [%zd]; minimum acceptable "
362 "message length is [%d]\n", __func__, message_len, 4); 362 "message length is [%d]\n", __func__, message_len, 4);
363 goto out; 363 goto out;
364 } 364 }
@@ -385,13 +385,13 @@ parse_tag_67_packet(struct ecryptfs_key_record *key_rec,
385 i += data_len; 385 i += data_len;
386 if (message_len < (i + key_rec->enc_key_size)) { 386 if (message_len < (i + key_rec->enc_key_size)) {
387 rc = -EIO; 387 rc = -EIO;
388 printk(KERN_ERR "%s: message_len [%Zd]; max len is [%Zd]\n", 388 printk(KERN_ERR "%s: message_len [%zd]; max len is [%zd]\n",
389 __func__, message_len, (i + key_rec->enc_key_size)); 389 __func__, message_len, (i + key_rec->enc_key_size));
390 goto out; 390 goto out;
391 } 391 }
392 if (key_rec->enc_key_size > ECRYPTFS_MAX_ENCRYPTED_KEY_BYTES) { 392 if (key_rec->enc_key_size > ECRYPTFS_MAX_ENCRYPTED_KEY_BYTES) {
393 rc = -EIO; 393 rc = -EIO;
394 printk(KERN_ERR "%s: Encrypted key_size [%Zd] larger than " 394 printk(KERN_ERR "%s: Encrypted key_size [%zd] larger than "
395 "the maximum key size [%d]\n", __func__, 395 "the maximum key size [%d]\n", __func__,
396 key_rec->enc_key_size, 396 key_rec->enc_key_size,
397 ECRYPTFS_MAX_ENCRYPTED_KEY_BYTES); 397 ECRYPTFS_MAX_ENCRYPTED_KEY_BYTES);
@@ -403,6 +403,580 @@ out:
403} 403}
404 404
405static int 405static int
406ecryptfs_find_global_auth_tok_for_sig(
407 struct ecryptfs_global_auth_tok **global_auth_tok,
408 struct ecryptfs_mount_crypt_stat *mount_crypt_stat, char *sig)
409{
410 struct ecryptfs_global_auth_tok *walker;
411 int rc = 0;
412
413 (*global_auth_tok) = NULL;
414 mutex_lock(&mount_crypt_stat->global_auth_tok_list_mutex);
415 list_for_each_entry(walker,
416 &mount_crypt_stat->global_auth_tok_list,
417 mount_crypt_stat_list) {
418 if (memcmp(walker->sig, sig, ECRYPTFS_SIG_SIZE_HEX) == 0) {
419 (*global_auth_tok) = walker;
420 goto out;
421 }
422 }
423 rc = -EINVAL;
424out:
425 mutex_unlock(&mount_crypt_stat->global_auth_tok_list_mutex);
426 return rc;
427}
428
429/**
430 * ecryptfs_find_auth_tok_for_sig
431 * @auth_tok: Set to the matching auth_tok; NULL if not found
432 * @crypt_stat: inode crypt_stat crypto context
433 * @sig: Sig of auth_tok to find
434 *
435 * For now, this function simply looks at the registered auth_tok's
436 * linked off the mount_crypt_stat, so all the auth_toks that can be
437 * used must be registered at mount time. This function could
438 * potentially try a lot harder to find auth_tok's (e.g., by calling
439 * out to ecryptfsd to dynamically retrieve an auth_tok object) so
440 * that static registration of auth_tok's will no longer be necessary.
441 *
442 * Returns zero on no error; non-zero on error
443 */
444static int
445ecryptfs_find_auth_tok_for_sig(
446 struct ecryptfs_auth_tok **auth_tok,
447 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
448 char *sig)
449{
450 struct ecryptfs_global_auth_tok *global_auth_tok;
451 int rc = 0;
452
453 (*auth_tok) = NULL;
454 if (ecryptfs_find_global_auth_tok_for_sig(&global_auth_tok,
455 mount_crypt_stat, sig)) {
456 struct key *auth_tok_key;
457
458 rc = ecryptfs_keyring_auth_tok_for_sig(&auth_tok_key, auth_tok,
459 sig);
460 } else
461 (*auth_tok) = global_auth_tok->global_auth_tok;
462 return rc;
463}
464
465/**
466 * write_tag_70_packet can gobble a lot of stack space. We stuff most
467 * of the function's parameters in a kmalloc'd struct to help reduce
468 * eCryptfs' overall stack usage.
469 */
470struct ecryptfs_write_tag_70_packet_silly_stack {
471 u8 cipher_code;
472 size_t max_packet_size;
473 size_t packet_size_len;
474 size_t block_aligned_filename_size;
475 size_t block_size;
476 size_t i;
477 size_t j;
478 size_t num_rand_bytes;
479 struct mutex *tfm_mutex;
480 char *block_aligned_filename;
481 struct ecryptfs_auth_tok *auth_tok;
482 struct scatterlist src_sg;
483 struct scatterlist dst_sg;
484 struct blkcipher_desc desc;
485 char iv[ECRYPTFS_MAX_IV_BYTES];
486 char hash[ECRYPTFS_TAG_70_DIGEST_SIZE];
487 char tmp_hash[ECRYPTFS_TAG_70_DIGEST_SIZE];
488 struct hash_desc hash_desc;
489 struct scatterlist hash_sg;
490};
491
492/**
493 * write_tag_70_packet - Write encrypted filename (EFN) packet against FNEK
494 * @filename: NULL-terminated filename string
495 *
496 * This is the simplest mechanism for achieving filename encryption in
497 * eCryptfs. It encrypts the given filename with the mount-wide
498 * filename encryption key (FNEK) and stores it in a packet to @dest,
499 * which the callee will encode and write directly into the dentry
500 * name.
501 */
502int
503ecryptfs_write_tag_70_packet(char *dest, size_t *remaining_bytes,
504 size_t *packet_size,
505 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
506 char *filename, size_t filename_size)
507{
508 struct ecryptfs_write_tag_70_packet_silly_stack *s;
509 int rc = 0;
510
511 s = kmalloc(sizeof(*s), GFP_KERNEL);
512 if (!s) {
513 printk(KERN_ERR "%s: Out of memory whilst trying to kmalloc "
514 "[%zd] bytes of kernel memory\n", __func__, sizeof(*s));
515 goto out;
516 }
517 s->desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
518 (*packet_size) = 0;
519 rc = ecryptfs_get_tfm_and_mutex_for_cipher_name(
520 &s->desc.tfm,
521 &s->tfm_mutex, mount_crypt_stat->global_default_fn_cipher_name);
522 if (unlikely(rc)) {
523 printk(KERN_ERR "Internal error whilst attempting to get "
524 "tfm and mutex for cipher name [%s]; rc = [%d]\n",
525 mount_crypt_stat->global_default_fn_cipher_name, rc);
526 goto out;
527 }
528 mutex_lock(s->tfm_mutex);
529 s->block_size = crypto_blkcipher_blocksize(s->desc.tfm);
530 /* Plus one for the \0 separator between the random prefix
531 * and the plaintext filename */
532 s->num_rand_bytes = (ECRYPTFS_FILENAME_MIN_RANDOM_PREPEND_BYTES + 1);
533 s->block_aligned_filename_size = (s->num_rand_bytes + filename_size);
534 if ((s->block_aligned_filename_size % s->block_size) != 0) {
535 s->num_rand_bytes += (s->block_size
536 - (s->block_aligned_filename_size
537 % s->block_size));
538 s->block_aligned_filename_size = (s->num_rand_bytes
539 + filename_size);
540 }
541 /* Octet 0: Tag 70 identifier
542 * Octets 1-N1: Tag 70 packet size (includes cipher identifier
543 * and block-aligned encrypted filename size)
544 * Octets N1-N2: FNEK sig (ECRYPTFS_SIG_SIZE)
545 * Octet N2-N3: Cipher identifier (1 octet)
546 * Octets N3-N4: Block-aligned encrypted filename
547 * - Consists of a minimum number of random characters, a \0
548 * separator, and then the filename */
549 s->max_packet_size = (1 /* Tag 70 identifier */
550 + 3 /* Max Tag 70 packet size */
551 + ECRYPTFS_SIG_SIZE /* FNEK sig */
552 + 1 /* Cipher identifier */
553 + s->block_aligned_filename_size);
554 if (dest == NULL) {
555 (*packet_size) = s->max_packet_size;
556 goto out_unlock;
557 }
558 if (s->max_packet_size > (*remaining_bytes)) {
559 printk(KERN_WARNING "%s: Require [%zd] bytes to write; only "
560 "[%zd] available\n", __func__, s->max_packet_size,
561 (*remaining_bytes));
562 rc = -EINVAL;
563 goto out_unlock;
564 }
565 s->block_aligned_filename = kzalloc(s->block_aligned_filename_size,
566 GFP_KERNEL);
567 if (!s->block_aligned_filename) {
568 printk(KERN_ERR "%s: Out of kernel memory whilst attempting to "
569 "kzalloc [%zd] bytes\n", __func__,
570 s->block_aligned_filename_size);
571 rc = -ENOMEM;
572 goto out_unlock;
573 }
574 s->i = 0;
575 dest[s->i++] = ECRYPTFS_TAG_70_PACKET_TYPE;
576 rc = ecryptfs_write_packet_length(&dest[s->i],
577 (ECRYPTFS_SIG_SIZE
578 + 1 /* Cipher code */
579 + s->block_aligned_filename_size),
580 &s->packet_size_len);
581 if (rc) {
582 printk(KERN_ERR "%s: Error generating tag 70 packet "
583 "header; cannot generate packet length; rc = [%d]\n",
584 __func__, rc);
585 goto out_free_unlock;
586 }
587 s->i += s->packet_size_len;
588 ecryptfs_from_hex(&dest[s->i],
589 mount_crypt_stat->global_default_fnek_sig,
590 ECRYPTFS_SIG_SIZE);
591 s->i += ECRYPTFS_SIG_SIZE;
592 s->cipher_code = ecryptfs_code_for_cipher_string(
593 mount_crypt_stat->global_default_fn_cipher_name,
594 mount_crypt_stat->global_default_fn_cipher_key_bytes);
595 if (s->cipher_code == 0) {
596 printk(KERN_WARNING "%s: Unable to generate code for "
597 "cipher [%s] with key bytes [%zd]\n", __func__,
598 mount_crypt_stat->global_default_fn_cipher_name,
599 mount_crypt_stat->global_default_fn_cipher_key_bytes);
600 rc = -EINVAL;
601 goto out_free_unlock;
602 }
603 dest[s->i++] = s->cipher_code;
604 rc = ecryptfs_find_auth_tok_for_sig(
605 &s->auth_tok, mount_crypt_stat,
606 mount_crypt_stat->global_default_fnek_sig);
607 if (rc) {
608 printk(KERN_ERR "%s: Error attempting to find auth tok for "
609 "fnek sig [%s]; rc = [%d]\n", __func__,
610 mount_crypt_stat->global_default_fnek_sig, rc);
611 goto out_free_unlock;
612 }
613 /* TODO: Support other key modules than passphrase for
614 * filename encryption */
615 BUG_ON(s->auth_tok->token_type != ECRYPTFS_PASSWORD);
616 sg_init_one(
617 &s->hash_sg,
618 (u8 *)s->auth_tok->token.password.session_key_encryption_key,
619 s->auth_tok->token.password.session_key_encryption_key_bytes);
620 s->hash_desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
621 s->hash_desc.tfm = crypto_alloc_hash(ECRYPTFS_TAG_70_DIGEST, 0,
622 CRYPTO_ALG_ASYNC);
623 if (IS_ERR(s->hash_desc.tfm)) {
624 rc = PTR_ERR(s->hash_desc.tfm);
625 printk(KERN_ERR "%s: Error attempting to "
626 "allocate hash crypto context; rc = [%d]\n",
627 __func__, rc);
628 goto out_free_unlock;
629 }
630 rc = crypto_hash_init(&s->hash_desc);
631 if (rc) {
632 printk(KERN_ERR
633 "%s: Error initializing crypto hash; rc = [%d]\n",
634 __func__, rc);
635 goto out_release_free_unlock;
636 }
637 rc = crypto_hash_update(
638 &s->hash_desc, &s->hash_sg,
639 s->auth_tok->token.password.session_key_encryption_key_bytes);
640 if (rc) {
641 printk(KERN_ERR
642 "%s: Error updating crypto hash; rc = [%d]\n",
643 __func__, rc);
644 goto out_release_free_unlock;
645 }
646 rc = crypto_hash_final(&s->hash_desc, s->hash);
647 if (rc) {
648 printk(KERN_ERR
649 "%s: Error finalizing crypto hash; rc = [%d]\n",
650 __func__, rc);
651 goto out_release_free_unlock;
652 }
653 for (s->j = 0; s->j < (s->num_rand_bytes - 1); s->j++) {
654 s->block_aligned_filename[s->j] =
655 s->hash[(s->j % ECRYPTFS_TAG_70_DIGEST_SIZE)];
656 if ((s->j % ECRYPTFS_TAG_70_DIGEST_SIZE)
657 == (ECRYPTFS_TAG_70_DIGEST_SIZE - 1)) {
658 sg_init_one(&s->hash_sg, (u8 *)s->hash,
659 ECRYPTFS_TAG_70_DIGEST_SIZE);
660 rc = crypto_hash_init(&s->hash_desc);
661 if (rc) {
662 printk(KERN_ERR
663 "%s: Error initializing crypto hash; "
664 "rc = [%d]\n", __func__, rc);
665 goto out_release_free_unlock;
666 }
667 rc = crypto_hash_update(&s->hash_desc, &s->hash_sg,
668 ECRYPTFS_TAG_70_DIGEST_SIZE);
669 if (rc) {
670 printk(KERN_ERR
671 "%s: Error updating crypto hash; "
672 "rc = [%d]\n", __func__, rc);
673 goto out_release_free_unlock;
674 }
675 rc = crypto_hash_final(&s->hash_desc, s->tmp_hash);
676 if (rc) {
677 printk(KERN_ERR
678 "%s: Error finalizing crypto hash; "
679 "rc = [%d]\n", __func__, rc);
680 goto out_release_free_unlock;
681 }
682 memcpy(s->hash, s->tmp_hash,
683 ECRYPTFS_TAG_70_DIGEST_SIZE);
684 }
685 if (s->block_aligned_filename[s->j] == '\0')
686 s->block_aligned_filename[s->j] = ECRYPTFS_NON_NULL;
687 }
688 memcpy(&s->block_aligned_filename[s->num_rand_bytes], filename,
689 filename_size);
690 rc = virt_to_scatterlist(s->block_aligned_filename,
691 s->block_aligned_filename_size, &s->src_sg, 1);
692 if (rc != 1) {
693 printk(KERN_ERR "%s: Internal error whilst attempting to "
694 "convert filename memory to scatterlist; "
695 "expected rc = 1; got rc = [%d]. "
696 "block_aligned_filename_size = [%zd]\n", __func__, rc,
697 s->block_aligned_filename_size);
698 goto out_release_free_unlock;
699 }
700 rc = virt_to_scatterlist(&dest[s->i], s->block_aligned_filename_size,
701 &s->dst_sg, 1);
702 if (rc != 1) {
703 printk(KERN_ERR "%s: Internal error whilst attempting to "
704 "convert encrypted filename memory to scatterlist; "
705 "expected rc = 1; got rc = [%d]. "
706 "block_aligned_filename_size = [%zd]\n", __func__, rc,
707 s->block_aligned_filename_size);
708 goto out_release_free_unlock;
709 }
710 /* The characters in the first block effectively do the job
711 * of the IV here, so we just use 0's for the IV. Note the
712 * constraint that ECRYPTFS_FILENAME_MIN_RANDOM_PREPEND_BYTES
713 * >= ECRYPTFS_MAX_IV_BYTES. */
714 memset(s->iv, 0, ECRYPTFS_MAX_IV_BYTES);
715 s->desc.info = s->iv;
716 rc = crypto_blkcipher_setkey(
717 s->desc.tfm,
718 s->auth_tok->token.password.session_key_encryption_key,
719 mount_crypt_stat->global_default_fn_cipher_key_bytes);
720 if (rc < 0) {
721 printk(KERN_ERR "%s: Error setting key for crypto context; "
722 "rc = [%d]. s->auth_tok->token.password.session_key_"
723 "encryption_key = [0x%p]; mount_crypt_stat->"
724 "global_default_fn_cipher_key_bytes = [%zd]\n", __func__,
725 rc,
726 s->auth_tok->token.password.session_key_encryption_key,
727 mount_crypt_stat->global_default_fn_cipher_key_bytes);
728 goto out_release_free_unlock;
729 }
730 rc = crypto_blkcipher_encrypt_iv(&s->desc, &s->dst_sg, &s->src_sg,
731 s->block_aligned_filename_size);
732 if (rc) {
733 printk(KERN_ERR "%s: Error attempting to encrypt filename; "
734 "rc = [%d]\n", __func__, rc);
735 goto out_release_free_unlock;
736 }
737 s->i += s->block_aligned_filename_size;
738 (*packet_size) = s->i;
739 (*remaining_bytes) -= (*packet_size);
740out_release_free_unlock:
741 crypto_free_hash(s->hash_desc.tfm);
742out_free_unlock:
743 memset(s->block_aligned_filename, 0, s->block_aligned_filename_size);
744 kfree(s->block_aligned_filename);
745out_unlock:
746 mutex_unlock(s->tfm_mutex);
747out:
748 kfree(s);
749 return rc;
750}
751
752struct ecryptfs_parse_tag_70_packet_silly_stack {
753 u8 cipher_code;
754 size_t max_packet_size;
755 size_t packet_size_len;
756 size_t parsed_tag_70_packet_size;
757 size_t block_aligned_filename_size;
758 size_t block_size;
759 size_t i;
760 struct mutex *tfm_mutex;
761 char *decrypted_filename;
762 struct ecryptfs_auth_tok *auth_tok;
763 struct scatterlist src_sg;
764 struct scatterlist dst_sg;
765 struct blkcipher_desc desc;
766 char fnek_sig_hex[ECRYPTFS_SIG_SIZE_HEX + 1];
767 char iv[ECRYPTFS_MAX_IV_BYTES];
768 char cipher_string[ECRYPTFS_MAX_CIPHER_NAME_SIZE];
769};
770
771/**
772 * parse_tag_70_packet - Parse and process FNEK-encrypted passphrase packet
773 * @filename: This function kmalloc's the memory for the filename
774 * @filename_size: This function sets this to the amount of memory
775 * kmalloc'd for the filename
776 * @packet_size: This function sets this to the the number of octets
777 * in the packet parsed
778 * @mount_crypt_stat: The mount-wide cryptographic context
779 * @data: The memory location containing the start of the tag 70
780 * packet
781 * @max_packet_size: The maximum legal size of the packet to be parsed
782 * from @data
783 *
784 * Returns zero on success; non-zero otherwise
785 */
786int
787ecryptfs_parse_tag_70_packet(char **filename, size_t *filename_size,
788 size_t *packet_size,
789 struct ecryptfs_mount_crypt_stat *mount_crypt_stat,
790 char *data, size_t max_packet_size)
791{
792 struct ecryptfs_parse_tag_70_packet_silly_stack *s;
793 int rc = 0;
794
795 (*packet_size) = 0;
796 (*filename_size) = 0;
797 (*filename) = NULL;
798 s = kmalloc(sizeof(*s), GFP_KERNEL);
799 if (!s) {
800 printk(KERN_ERR "%s: Out of memory whilst trying to kmalloc "
801 "[%zd] bytes of kernel memory\n", __func__, sizeof(*s));
802 goto out;
803 }
804 s->desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
805 if (max_packet_size < (1 + 1 + ECRYPTFS_SIG_SIZE + 1 + 1)) {
806 printk(KERN_WARNING "%s: max_packet_size is [%zd]; it must be "
807 "at least [%d]\n", __func__, max_packet_size,
808 (1 + 1 + ECRYPTFS_SIG_SIZE + 1 + 1));
809 rc = -EINVAL;
810 goto out;
811 }
812 /* Octet 0: Tag 70 identifier
813 * Octets 1-N1: Tag 70 packet size (includes cipher identifier
814 * and block-aligned encrypted filename size)
815 * Octets N1-N2: FNEK sig (ECRYPTFS_SIG_SIZE)
816 * Octet N2-N3: Cipher identifier (1 octet)
817 * Octets N3-N4: Block-aligned encrypted filename
818 * - Consists of a minimum number of random numbers, a \0
819 * separator, and then the filename */
820 if (data[(*packet_size)++] != ECRYPTFS_TAG_70_PACKET_TYPE) {
821 printk(KERN_WARNING "%s: Invalid packet tag [0x%.2x]; must be "
822 "tag [0x%.2x]\n", __func__,
823 data[((*packet_size) - 1)], ECRYPTFS_TAG_70_PACKET_TYPE);
824 rc = -EINVAL;
825 goto out;
826 }
827 rc = ecryptfs_parse_packet_length(&data[(*packet_size)],
828 &s->parsed_tag_70_packet_size,
829 &s->packet_size_len);
830 if (rc) {
831 printk(KERN_WARNING "%s: Error parsing packet length; "
832 "rc = [%d]\n", __func__, rc);
833 goto out;
834 }
835 s->block_aligned_filename_size = (s->parsed_tag_70_packet_size
836 - ECRYPTFS_SIG_SIZE - 1);
837 if ((1 + s->packet_size_len + s->parsed_tag_70_packet_size)
838 > max_packet_size) {
839 printk(KERN_WARNING "%s: max_packet_size is [%zd]; real packet "
840 "size is [%zd]\n", __func__, max_packet_size,
841 (1 + s->packet_size_len + 1
842 + s->block_aligned_filename_size));
843 rc = -EINVAL;
844 goto out;
845 }
846 (*packet_size) += s->packet_size_len;
847 ecryptfs_to_hex(s->fnek_sig_hex, &data[(*packet_size)],
848 ECRYPTFS_SIG_SIZE);
849 s->fnek_sig_hex[ECRYPTFS_SIG_SIZE_HEX] = '\0';
850 (*packet_size) += ECRYPTFS_SIG_SIZE;
851 s->cipher_code = data[(*packet_size)++];
852 rc = ecryptfs_cipher_code_to_string(s->cipher_string, s->cipher_code);
853 if (rc) {
854 printk(KERN_WARNING "%s: Cipher code [%d] is invalid\n",
855 __func__, s->cipher_code);
856 goto out;
857 }
858 rc = ecryptfs_get_tfm_and_mutex_for_cipher_name(&s->desc.tfm,
859 &s->tfm_mutex,
860 s->cipher_string);
861 if (unlikely(rc)) {
862 printk(KERN_ERR "Internal error whilst attempting to get "
863 "tfm and mutex for cipher name [%s]; rc = [%d]\n",
864 s->cipher_string, rc);
865 goto out;
866 }
867 mutex_lock(s->tfm_mutex);
868 rc = virt_to_scatterlist(&data[(*packet_size)],
869 s->block_aligned_filename_size, &s->src_sg, 1);
870 if (rc != 1) {
871 printk(KERN_ERR "%s: Internal error whilst attempting to "
872 "convert encrypted filename memory to scatterlist; "
873 "expected rc = 1; got rc = [%d]. "
874 "block_aligned_filename_size = [%zd]\n", __func__, rc,
875 s->block_aligned_filename_size);
876 goto out_unlock;
877 }
878 (*packet_size) += s->block_aligned_filename_size;
879 s->decrypted_filename = kmalloc(s->block_aligned_filename_size,
880 GFP_KERNEL);
881 if (!s->decrypted_filename) {
882 printk(KERN_ERR "%s: Out of memory whilst attempting to "
883 "kmalloc [%zd] bytes\n", __func__,
884 s->block_aligned_filename_size);
885 rc = -ENOMEM;
886 goto out_unlock;
887 }
888 rc = virt_to_scatterlist(s->decrypted_filename,
889 s->block_aligned_filename_size, &s->dst_sg, 1);
890 if (rc != 1) {
891 printk(KERN_ERR "%s: Internal error whilst attempting to "
892 "convert decrypted filename memory to scatterlist; "
893 "expected rc = 1; got rc = [%d]. "
894 "block_aligned_filename_size = [%zd]\n", __func__, rc,
895 s->block_aligned_filename_size);
896 goto out_free_unlock;
897 }
898 /* The characters in the first block effectively do the job of
899 * the IV here, so we just use 0's for the IV. Note the
900 * constraint that ECRYPTFS_FILENAME_MIN_RANDOM_PREPEND_BYTES
901 * >= ECRYPTFS_MAX_IV_BYTES. */
902 memset(s->iv, 0, ECRYPTFS_MAX_IV_BYTES);
903 s->desc.info = s->iv;
904 rc = ecryptfs_find_auth_tok_for_sig(&s->auth_tok, mount_crypt_stat,
905 s->fnek_sig_hex);
906 if (rc) {
907 printk(KERN_ERR "%s: Error attempting to find auth tok for "
908 "fnek sig [%s]; rc = [%d]\n", __func__, s->fnek_sig_hex,
909 rc);
910 goto out_free_unlock;
911 }
912 /* TODO: Support other key modules than passphrase for
913 * filename encryption */
914 BUG_ON(s->auth_tok->token_type != ECRYPTFS_PASSWORD);
915 rc = crypto_blkcipher_setkey(
916 s->desc.tfm,
917 s->auth_tok->token.password.session_key_encryption_key,
918 mount_crypt_stat->global_default_fn_cipher_key_bytes);
919 if (rc < 0) {
920 printk(KERN_ERR "%s: Error setting key for crypto context; "
921 "rc = [%d]. s->auth_tok->token.password.session_key_"
922 "encryption_key = [0x%p]; mount_crypt_stat->"
923 "global_default_fn_cipher_key_bytes = [%zd]\n", __func__,
924 rc,
925 s->auth_tok->token.password.session_key_encryption_key,
926 mount_crypt_stat->global_default_fn_cipher_key_bytes);
927 goto out_free_unlock;
928 }
929 rc = crypto_blkcipher_decrypt_iv(&s->desc, &s->dst_sg, &s->src_sg,
930 s->block_aligned_filename_size);
931 if (rc) {
932 printk(KERN_ERR "%s: Error attempting to decrypt filename; "
933 "rc = [%d]\n", __func__, rc);
934 goto out_free_unlock;
935 }
936 s->i = 0;
937 while (s->decrypted_filename[s->i] != '\0'
938 && s->i < s->block_aligned_filename_size)
939 s->i++;
940 if (s->i == s->block_aligned_filename_size) {
941 printk(KERN_WARNING "%s: Invalid tag 70 packet; could not "
942 "find valid separator between random characters and "
943 "the filename\n", __func__);
944 rc = -EINVAL;
945 goto out_free_unlock;
946 }
947 s->i++;
948 (*filename_size) = (s->block_aligned_filename_size - s->i);
949 if (!((*filename_size) > 0 && (*filename_size < PATH_MAX))) {
950 printk(KERN_WARNING "%s: Filename size is [%zd], which is "
951 "invalid\n", __func__, (*filename_size));
952 rc = -EINVAL;
953 goto out_free_unlock;
954 }
955 (*filename) = kmalloc(((*filename_size) + 1), GFP_KERNEL);
956 if (!(*filename)) {
957 printk(KERN_ERR "%s: Out of memory whilst attempting to "
958 "kmalloc [%zd] bytes\n", __func__,
959 ((*filename_size) + 1));
960 rc = -ENOMEM;
961 goto out_free_unlock;
962 }
963 memcpy((*filename), &s->decrypted_filename[s->i], (*filename_size));
964 (*filename)[(*filename_size)] = '\0';
965out_free_unlock:
966 kfree(s->decrypted_filename);
967out_unlock:
968 mutex_unlock(s->tfm_mutex);
969out:
970 if (rc) {
971 (*packet_size) = 0;
972 (*filename_size) = 0;
973 (*filename) = NULL;
974 }
975 kfree(s);
976 return rc;
977}
978
979static int
406ecryptfs_get_auth_tok_sig(char **sig, struct ecryptfs_auth_tok *auth_tok) 980ecryptfs_get_auth_tok_sig(char **sig, struct ecryptfs_auth_tok *auth_tok)
407{ 981{
408 int rc = 0; 982 int rc = 0;
@@ -897,30 +1471,6 @@ out:
897 return rc; 1471 return rc;
898} 1472}
899 1473
900static int
901ecryptfs_find_global_auth_tok_for_sig(
902 struct ecryptfs_global_auth_tok **global_auth_tok,
903 struct ecryptfs_mount_crypt_stat *mount_crypt_stat, char *sig)
904{
905 struct ecryptfs_global_auth_tok *walker;
906 int rc = 0;
907
908 (*global_auth_tok) = NULL;
909 mutex_lock(&mount_crypt_stat->global_auth_tok_list_mutex);
910 list_for_each_entry(walker,
911 &mount_crypt_stat->global_auth_tok_list,
912 mount_crypt_stat_list) {
913 if (memcmp(walker->sig, sig, ECRYPTFS_SIG_SIZE_HEX) == 0) {
914 (*global_auth_tok) = walker;
915 goto out;
916 }
917 }
918 rc = -EINVAL;
919out:
920 mutex_unlock(&mount_crypt_stat->global_auth_tok_list_mutex);
921 return rc;
922}
923
924/** 1474/**
925 * ecryptfs_verify_version 1475 * ecryptfs_verify_version
926 * @version: The version number to confirm 1476 * @version: The version number to confirm
@@ -990,43 +1540,6 @@ out:
990} 1540}
991 1541
992/** 1542/**
993 * ecryptfs_find_auth_tok_for_sig
994 * @auth_tok: Set to the matching auth_tok; NULL if not found
995 * @crypt_stat: inode crypt_stat crypto context
996 * @sig: Sig of auth_tok to find
997 *
998 * For now, this function simply looks at the registered auth_tok's
999 * linked off the mount_crypt_stat, so all the auth_toks that can be
1000 * used must be registered at mount time. This function could
1001 * potentially try a lot harder to find auth_tok's (e.g., by calling
1002 * out to ecryptfsd to dynamically retrieve an auth_tok object) so
1003 * that static registration of auth_tok's will no longer be necessary.
1004 *
1005 * Returns zero on no error; non-zero on error
1006 */
1007static int
1008ecryptfs_find_auth_tok_for_sig(
1009 struct ecryptfs_auth_tok **auth_tok,
1010 struct ecryptfs_crypt_stat *crypt_stat, char *sig)
1011{
1012 struct ecryptfs_mount_crypt_stat *mount_crypt_stat =
1013 crypt_stat->mount_crypt_stat;
1014 struct ecryptfs_global_auth_tok *global_auth_tok;
1015 int rc = 0;
1016
1017 (*auth_tok) = NULL;
1018 if (ecryptfs_find_global_auth_tok_for_sig(&global_auth_tok,
1019 mount_crypt_stat, sig)) {
1020 struct key *auth_tok_key;
1021
1022 rc = ecryptfs_keyring_auth_tok_for_sig(&auth_tok_key, auth_tok,
1023 sig);
1024 } else
1025 (*auth_tok) = global_auth_tok->global_auth_tok;
1026 return rc;
1027}
1028
1029/**
1030 * decrypt_passphrase_encrypted_session_key - Decrypt the session key with the given auth_tok. 1543 * decrypt_passphrase_encrypted_session_key - Decrypt the session key with the given auth_tok.
1031 * @auth_tok: The passphrase authentication token to use to encrypt the FEK 1544 * @auth_tok: The passphrase authentication token to use to encrypt the FEK
1032 * @crypt_stat: The cryptographic context 1545 * @crypt_stat: The cryptographic context
@@ -1256,7 +1769,8 @@ find_next_matching_auth_tok:
1256 rc = -EINVAL; 1769 rc = -EINVAL;
1257 goto out_wipe_list; 1770 goto out_wipe_list;
1258 } 1771 }
1259 ecryptfs_find_auth_tok_for_sig(&matching_auth_tok, crypt_stat, 1772 ecryptfs_find_auth_tok_for_sig(&matching_auth_tok,
1773 crypt_stat->mount_crypt_stat,
1260 candidate_auth_tok_sig); 1774 candidate_auth_tok_sig);
1261 if (matching_auth_tok) { 1775 if (matching_auth_tok) {
1262 found_auth_tok = 1; 1776 found_auth_tok = 1;
@@ -1336,7 +1850,9 @@ pki_encrypt_session_key(struct ecryptfs_auth_tok *auth_tok,
1336 int rc; 1850 int rc;
1337 1851
1338 rc = write_tag_66_packet(auth_tok->token.private_key.signature, 1852 rc = write_tag_66_packet(auth_tok->token.private_key.signature,
1339 ecryptfs_code_for_cipher_string(crypt_stat), 1853 ecryptfs_code_for_cipher_string(
1854 crypt_stat->cipher,
1855 crypt_stat->key_size),
1340 crypt_stat, &payload, &payload_len); 1856 crypt_stat, &payload, &payload_len);
1341 if (rc) { 1857 if (rc) {
1342 ecryptfs_printk(KERN_ERR, "Error generating tag 66 packet\n"); 1858 ecryptfs_printk(KERN_ERR, "Error generating tag 66 packet\n");
@@ -1696,7 +2212,8 @@ encrypted_session_key_set:
1696 dest[(*packet_size)++] = 0x04; /* version 4 */ 2212 dest[(*packet_size)++] = 0x04; /* version 4 */
1697 /* TODO: Break from RFC2440 so that arbitrary ciphers can be 2213 /* TODO: Break from RFC2440 so that arbitrary ciphers can be
1698 * specified with strings */ 2214 * specified with strings */
1699 cipher_code = ecryptfs_code_for_cipher_string(crypt_stat); 2215 cipher_code = ecryptfs_code_for_cipher_string(crypt_stat->cipher,
2216 crypt_stat->key_size);
1700 if (cipher_code == 0) { 2217 if (cipher_code == 0) {
1701 ecryptfs_printk(KERN_WARNING, "Unable to generate code for " 2218 ecryptfs_printk(KERN_WARNING, "Unable to generate code for "
1702 "cipher [%s]\n", crypt_stat->cipher); 2219 "cipher [%s]\n", crypt_stat->cipher);
diff --git a/fs/ecryptfs/main.c b/fs/ecryptfs/main.c
index fd630713c5c7..789cf2e1be1e 100644
--- a/fs/ecryptfs/main.c
+++ b/fs/ecryptfs/main.c
@@ -206,7 +206,9 @@ enum { ecryptfs_opt_sig, ecryptfs_opt_ecryptfs_sig,
206 ecryptfs_opt_cipher, ecryptfs_opt_ecryptfs_cipher, 206 ecryptfs_opt_cipher, ecryptfs_opt_ecryptfs_cipher,
207 ecryptfs_opt_ecryptfs_key_bytes, 207 ecryptfs_opt_ecryptfs_key_bytes,
208 ecryptfs_opt_passthrough, ecryptfs_opt_xattr_metadata, 208 ecryptfs_opt_passthrough, ecryptfs_opt_xattr_metadata,
209 ecryptfs_opt_encrypted_view, ecryptfs_opt_err }; 209 ecryptfs_opt_encrypted_view, ecryptfs_opt_fnek_sig,
210 ecryptfs_opt_fn_cipher, ecryptfs_opt_fn_cipher_key_bytes,
211 ecryptfs_opt_err };
210 212
211static const match_table_t tokens = { 213static const match_table_t tokens = {
212 {ecryptfs_opt_sig, "sig=%s"}, 214 {ecryptfs_opt_sig, "sig=%s"},
@@ -217,6 +219,9 @@ static const match_table_t tokens = {
217 {ecryptfs_opt_passthrough, "ecryptfs_passthrough"}, 219 {ecryptfs_opt_passthrough, "ecryptfs_passthrough"},
218 {ecryptfs_opt_xattr_metadata, "ecryptfs_xattr_metadata"}, 220 {ecryptfs_opt_xattr_metadata, "ecryptfs_xattr_metadata"},
219 {ecryptfs_opt_encrypted_view, "ecryptfs_encrypted_view"}, 221 {ecryptfs_opt_encrypted_view, "ecryptfs_encrypted_view"},
222 {ecryptfs_opt_fnek_sig, "ecryptfs_fnek_sig=%s"},
223 {ecryptfs_opt_fn_cipher, "ecryptfs_fn_cipher=%s"},
224 {ecryptfs_opt_fn_cipher_key_bytes, "ecryptfs_fn_key_bytes=%u"},
220 {ecryptfs_opt_err, NULL} 225 {ecryptfs_opt_err, NULL}
221}; 226};
222 227
@@ -281,8 +286,11 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
281 int rc = 0; 286 int rc = 0;
282 int sig_set = 0; 287 int sig_set = 0;
283 int cipher_name_set = 0; 288 int cipher_name_set = 0;
289 int fn_cipher_name_set = 0;
284 int cipher_key_bytes; 290 int cipher_key_bytes;
285 int cipher_key_bytes_set = 0; 291 int cipher_key_bytes_set = 0;
292 int fn_cipher_key_bytes;
293 int fn_cipher_key_bytes_set = 0;
286 struct ecryptfs_mount_crypt_stat *mount_crypt_stat = 294 struct ecryptfs_mount_crypt_stat *mount_crypt_stat =
287 &ecryptfs_superblock_to_private(sb)->mount_crypt_stat; 295 &ecryptfs_superblock_to_private(sb)->mount_crypt_stat;
288 substring_t args[MAX_OPT_ARGS]; 296 substring_t args[MAX_OPT_ARGS];
@@ -290,7 +298,12 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
290 char *sig_src; 298 char *sig_src;
291 char *cipher_name_dst; 299 char *cipher_name_dst;
292 char *cipher_name_src; 300 char *cipher_name_src;
301 char *fn_cipher_name_dst;
302 char *fn_cipher_name_src;
303 char *fnek_dst;
304 char *fnek_src;
293 char *cipher_key_bytes_src; 305 char *cipher_key_bytes_src;
306 char *fn_cipher_key_bytes_src;
294 307
295 if (!options) { 308 if (!options) {
296 rc = -EINVAL; 309 rc = -EINVAL;
@@ -322,10 +335,7 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
322 global_default_cipher_name; 335 global_default_cipher_name;
323 strncpy(cipher_name_dst, cipher_name_src, 336 strncpy(cipher_name_dst, cipher_name_src,
324 ECRYPTFS_MAX_CIPHER_NAME_SIZE); 337 ECRYPTFS_MAX_CIPHER_NAME_SIZE);
325 ecryptfs_printk(KERN_DEBUG, 338 cipher_name_dst[ECRYPTFS_MAX_CIPHER_NAME_SIZE] = '\0';
326 "The mount_crypt_stat "
327 "global_default_cipher_name set to: "
328 "[%s]\n", cipher_name_dst);
329 cipher_name_set = 1; 339 cipher_name_set = 1;
330 break; 340 break;
331 case ecryptfs_opt_ecryptfs_key_bytes: 341 case ecryptfs_opt_ecryptfs_key_bytes:
@@ -335,11 +345,6 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
335 &cipher_key_bytes_src, 0); 345 &cipher_key_bytes_src, 0);
336 mount_crypt_stat->global_default_cipher_key_size = 346 mount_crypt_stat->global_default_cipher_key_size =
337 cipher_key_bytes; 347 cipher_key_bytes;
338 ecryptfs_printk(KERN_DEBUG,
339 "The mount_crypt_stat "
340 "global_default_cipher_key_size "
341 "set to: [%d]\n", mount_crypt_stat->
342 global_default_cipher_key_size);
343 cipher_key_bytes_set = 1; 348 cipher_key_bytes_set = 1;
344 break; 349 break;
345 case ecryptfs_opt_passthrough: 350 case ecryptfs_opt_passthrough:
@@ -356,11 +361,51 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
356 mount_crypt_stat->flags |= 361 mount_crypt_stat->flags |=
357 ECRYPTFS_ENCRYPTED_VIEW_ENABLED; 362 ECRYPTFS_ENCRYPTED_VIEW_ENABLED;
358 break; 363 break;
364 case ecryptfs_opt_fnek_sig:
365 fnek_src = args[0].from;
366 fnek_dst =
367 mount_crypt_stat->global_default_fnek_sig;
368 strncpy(fnek_dst, fnek_src, ECRYPTFS_SIG_SIZE_HEX);
369 mount_crypt_stat->global_default_fnek_sig[
370 ECRYPTFS_SIG_SIZE_HEX] = '\0';
371 rc = ecryptfs_add_global_auth_tok(
372 mount_crypt_stat,
373 mount_crypt_stat->global_default_fnek_sig);
374 if (rc) {
375 printk(KERN_ERR "Error attempting to register "
376 "global fnek sig [%s]; rc = [%d]\n",
377 mount_crypt_stat->global_default_fnek_sig,
378 rc);
379 goto out;
380 }
381 mount_crypt_stat->flags |=
382 (ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES
383 | ECRYPTFS_GLOBAL_ENCFN_USE_MOUNT_FNEK);
384 break;
385 case ecryptfs_opt_fn_cipher:
386 fn_cipher_name_src = args[0].from;
387 fn_cipher_name_dst =
388 mount_crypt_stat->global_default_fn_cipher_name;
389 strncpy(fn_cipher_name_dst, fn_cipher_name_src,
390 ECRYPTFS_MAX_CIPHER_NAME_SIZE);
391 mount_crypt_stat->global_default_fn_cipher_name[
392 ECRYPTFS_MAX_CIPHER_NAME_SIZE] = '\0';
393 fn_cipher_name_set = 1;
394 break;
395 case ecryptfs_opt_fn_cipher_key_bytes:
396 fn_cipher_key_bytes_src = args[0].from;
397 fn_cipher_key_bytes =
398 (int)simple_strtol(fn_cipher_key_bytes_src,
399 &fn_cipher_key_bytes_src, 0);
400 mount_crypt_stat->global_default_fn_cipher_key_bytes =
401 fn_cipher_key_bytes;
402 fn_cipher_key_bytes_set = 1;
403 break;
359 case ecryptfs_opt_err: 404 case ecryptfs_opt_err:
360 default: 405 default:
361 ecryptfs_printk(KERN_WARNING, 406 printk(KERN_WARNING
362 "eCryptfs: unrecognized option '%s'\n", 407 "%s: eCryptfs: unrecognized option [%s]\n",
363 p); 408 __func__, p);
364 } 409 }
365 } 410 }
366 if (!sig_set) { 411 if (!sig_set) {
@@ -374,33 +419,60 @@ static int ecryptfs_parse_options(struct super_block *sb, char *options)
374 int cipher_name_len = strlen(ECRYPTFS_DEFAULT_CIPHER); 419 int cipher_name_len = strlen(ECRYPTFS_DEFAULT_CIPHER);
375 420
376 BUG_ON(cipher_name_len >= ECRYPTFS_MAX_CIPHER_NAME_SIZE); 421 BUG_ON(cipher_name_len >= ECRYPTFS_MAX_CIPHER_NAME_SIZE);
377
378 strcpy(mount_crypt_stat->global_default_cipher_name, 422 strcpy(mount_crypt_stat->global_default_cipher_name,
379 ECRYPTFS_DEFAULT_CIPHER); 423 ECRYPTFS_DEFAULT_CIPHER);
380 } 424 }
381 if (!cipher_key_bytes_set) { 425 if ((mount_crypt_stat->flags & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES)
426 && !fn_cipher_name_set)
427 strcpy(mount_crypt_stat->global_default_fn_cipher_name,
428 mount_crypt_stat->global_default_cipher_name);
429 if (!cipher_key_bytes_set)
382 mount_crypt_stat->global_default_cipher_key_size = 0; 430 mount_crypt_stat->global_default_cipher_key_size = 0;
383 } 431 if ((mount_crypt_stat->flags & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES)
432 && !fn_cipher_key_bytes_set)
433 mount_crypt_stat->global_default_fn_cipher_key_bytes =
434 mount_crypt_stat->global_default_cipher_key_size;
384 mutex_lock(&key_tfm_list_mutex); 435 mutex_lock(&key_tfm_list_mutex);
385 if (!ecryptfs_tfm_exists(mount_crypt_stat->global_default_cipher_name, 436 if (!ecryptfs_tfm_exists(mount_crypt_stat->global_default_cipher_name,
386 NULL)) 437 NULL)) {
387 rc = ecryptfs_add_new_key_tfm( 438 rc = ecryptfs_add_new_key_tfm(
388 NULL, mount_crypt_stat->global_default_cipher_name, 439 NULL, mount_crypt_stat->global_default_cipher_name,
389 mount_crypt_stat->global_default_cipher_key_size); 440 mount_crypt_stat->global_default_cipher_key_size);
390 mutex_unlock(&key_tfm_list_mutex); 441 if (rc) {
391 if (rc) { 442 printk(KERN_ERR "Error attempting to initialize "
392 printk(KERN_ERR "Error attempting to initialize cipher with " 443 "cipher with name = [%s] and key size = [%td]; "
393 "name = [%s] and key size = [%td]; rc = [%d]\n", 444 "rc = [%d]\n",
394 mount_crypt_stat->global_default_cipher_name, 445 mount_crypt_stat->global_default_cipher_name,
395 mount_crypt_stat->global_default_cipher_key_size, rc); 446 mount_crypt_stat->global_default_cipher_key_size,
396 rc = -EINVAL; 447 rc);
397 goto out; 448 rc = -EINVAL;
449 mutex_unlock(&key_tfm_list_mutex);
450 goto out;
451 }
398 } 452 }
453 if ((mount_crypt_stat->flags & ECRYPTFS_GLOBAL_ENCRYPT_FILENAMES)
454 && !ecryptfs_tfm_exists(
455 mount_crypt_stat->global_default_fn_cipher_name, NULL)) {
456 rc = ecryptfs_add_new_key_tfm(
457 NULL, mount_crypt_stat->global_default_fn_cipher_name,
458 mount_crypt_stat->global_default_fn_cipher_key_bytes);
459 if (rc) {
460 printk(KERN_ERR "Error attempting to initialize "
461 "cipher with name = [%s] and key size = [%td]; "
462 "rc = [%d]\n",
463 mount_crypt_stat->global_default_fn_cipher_name,
464 mount_crypt_stat->global_default_fn_cipher_key_bytes,
465 rc);
466 rc = -EINVAL;
467 mutex_unlock(&key_tfm_list_mutex);
468 goto out;
469 }
470 }
471 mutex_unlock(&key_tfm_list_mutex);
399 rc = ecryptfs_init_global_auth_toks(mount_crypt_stat); 472 rc = ecryptfs_init_global_auth_toks(mount_crypt_stat);
400 if (rc) { 473 if (rc)
401 printk(KERN_WARNING "One or more global auth toks could not " 474 printk(KERN_WARNING "One or more global auth toks could not "
402 "properly register; rc = [%d]\n", rc); 475 "properly register; rc = [%d]\n", rc);
403 }
404out: 476out:
405 return rc; 477 return rc;
406} 478}
diff --git a/fs/ecryptfs/messaging.c b/fs/ecryptfs/messaging.c
index 6913f727624d..96ef51489e01 100644
--- a/fs/ecryptfs/messaging.c
+++ b/fs/ecryptfs/messaging.c
@@ -193,7 +193,7 @@ ecryptfs_spawn_daemon(struct ecryptfs_daemon **daemon, uid_t euid,
193 (*daemon) = kzalloc(sizeof(**daemon), GFP_KERNEL); 193 (*daemon) = kzalloc(sizeof(**daemon), GFP_KERNEL);
194 if (!(*daemon)) { 194 if (!(*daemon)) {
195 rc = -ENOMEM; 195 rc = -ENOMEM;
196 printk(KERN_ERR "%s: Failed to allocate [%Zd] bytes of " 196 printk(KERN_ERR "%s: Failed to allocate [%zd] bytes of "
197 "GFP_KERNEL memory\n", __func__, sizeof(**daemon)); 197 "GFP_KERNEL memory\n", __func__, sizeof(**daemon));
198 goto out; 198 goto out;
199 } 199 }
@@ -435,7 +435,7 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
435 msg_ctx->msg = kmalloc(msg_size, GFP_KERNEL); 435 msg_ctx->msg = kmalloc(msg_size, GFP_KERNEL);
436 if (!msg_ctx->msg) { 436 if (!msg_ctx->msg) {
437 rc = -ENOMEM; 437 rc = -ENOMEM;
438 printk(KERN_ERR "%s: Failed to allocate [%Zd] bytes of " 438 printk(KERN_ERR "%s: Failed to allocate [%zd] bytes of "
439 "GFP_KERNEL memory\n", __func__, msg_size); 439 "GFP_KERNEL memory\n", __func__, msg_size);
440 goto unlock; 440 goto unlock;
441 } 441 }
diff --git a/fs/ecryptfs/miscdev.c b/fs/ecryptfs/miscdev.c
index efd95a0ed1ea..a67fea655f49 100644
--- a/fs/ecryptfs/miscdev.c
+++ b/fs/ecryptfs/miscdev.c
@@ -199,7 +199,7 @@ int ecryptfs_send_miscdev(char *data, size_t data_size,
199 if (!msg_ctx->msg) { 199 if (!msg_ctx->msg) {
200 rc = -ENOMEM; 200 rc = -ENOMEM;
201 printk(KERN_ERR "%s: Out of memory whilst attempting " 201 printk(KERN_ERR "%s: Out of memory whilst attempting "
202 "to kmalloc(%Zd, GFP_KERNEL)\n", __func__, 202 "to kmalloc(%zd, GFP_KERNEL)\n", __func__,
203 (sizeof(*msg_ctx->msg) + data_size)); 203 (sizeof(*msg_ctx->msg) + data_size));
204 goto out_unlock; 204 goto out_unlock;
205 } 205 }
@@ -322,7 +322,7 @@ check_list:
322 if (count < total_length) { 322 if (count < total_length) {
323 rc = 0; 323 rc = 0;
324 printk(KERN_WARNING "%s: Only given user buffer of " 324 printk(KERN_WARNING "%s: Only given user buffer of "
325 "size [%Zd], but we need [%Zd] to read the " 325 "size [%zd], but we need [%zd] to read the "
326 "pending message\n", __func__, count, total_length); 326 "pending message\n", __func__, count, total_length);
327 goto out_unlock_msg_ctx; 327 goto out_unlock_msg_ctx;
328 } 328 }
@@ -376,7 +376,7 @@ static int ecryptfs_miscdev_response(char *data, size_t data_size,
376 376
377 if ((sizeof(*msg) + msg->data_len) != data_size) { 377 if ((sizeof(*msg) + msg->data_len) != data_size) {
378 printk(KERN_WARNING "%s: (sizeof(*msg) + msg->data_len) = " 378 printk(KERN_WARNING "%s: (sizeof(*msg) + msg->data_len) = "
379 "[%Zd]; data_size = [%Zd]. Invalid packet.\n", __func__, 379 "[%zd]; data_size = [%zd]. Invalid packet.\n", __func__,
380 (sizeof(*msg) + msg->data_len), data_size); 380 (sizeof(*msg) + msg->data_len), data_size);
381 rc = -EINVAL; 381 rc = -EINVAL;
382 goto out; 382 goto out;
@@ -421,7 +421,7 @@ ecryptfs_miscdev_write(struct file *file, const char __user *buf,
421 data = kmalloc(count, GFP_KERNEL); 421 data = kmalloc(count, GFP_KERNEL);
422 if (!data) { 422 if (!data) {
423 printk(KERN_ERR "%s: Out of memory whilst attempting to " 423 printk(KERN_ERR "%s: Out of memory whilst attempting to "
424 "kmalloc([%Zd], GFP_KERNEL)\n", __func__, count); 424 "kmalloc([%zd], GFP_KERNEL)\n", __func__, count);
425 goto out; 425 goto out;
426 } 426 }
427 rc = copy_from_user(data, buf, count); 427 rc = copy_from_user(data, buf, count);
@@ -436,8 +436,8 @@ ecryptfs_miscdev_write(struct file *file, const char __user *buf,
436 case ECRYPTFS_MSG_RESPONSE: 436 case ECRYPTFS_MSG_RESPONSE:
437 if (count < (1 + 4 + 1 + sizeof(struct ecryptfs_message))) { 437 if (count < (1 + 4 + 1 + sizeof(struct ecryptfs_message))) {
438 printk(KERN_WARNING "%s: Minimum acceptable packet " 438 printk(KERN_WARNING "%s: Minimum acceptable packet "
439 "size is [%Zd], but amount of data written is " 439 "size is [%zd], but amount of data written is "
440 "only [%Zd]. Discarding response packet.\n", 440 "only [%zd]. Discarding response packet.\n",
441 __func__, 441 __func__,
442 (1 + 4 + 1 + sizeof(struct ecryptfs_message)), 442 (1 + 4 + 1 + sizeof(struct ecryptfs_message)),
443 count); 443 count);
@@ -455,9 +455,9 @@ ecryptfs_miscdev_write(struct file *file, const char __user *buf,
455 } 455 }
456 i += packet_size_length; 456 i += packet_size_length;
457 if ((1 + 4 + packet_size_length + packet_size) != count) { 457 if ((1 + 4 + packet_size_length + packet_size) != count) {
458 printk(KERN_WARNING "%s: (1 + packet_size_length([%Zd])" 458 printk(KERN_WARNING "%s: (1 + packet_size_length([%zd])"
459 " + packet_size([%Zd]))([%Zd]) != " 459 " + packet_size([%zd]))([%zd]) != "
460 "count([%Zd]). Invalid packet format.\n", 460 "count([%zd]). Invalid packet format.\n",
461 __func__, packet_size_length, packet_size, 461 __func__, packet_size_length, packet_size,
462 (1 + packet_size_length + packet_size), count); 462 (1 + packet_size_length + packet_size), count);
463 goto out_free; 463 goto out_free;
diff --git a/fs/exec.c b/fs/exec.c
index 9c33f542dc77..71a6efe5d8bd 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -232,13 +232,13 @@ static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
232 232
233static int __bprm_mm_init(struct linux_binprm *bprm) 233static int __bprm_mm_init(struct linux_binprm *bprm)
234{ 234{
235 int err = -ENOMEM; 235 int err;
236 struct vm_area_struct *vma = NULL; 236 struct vm_area_struct *vma = NULL;
237 struct mm_struct *mm = bprm->mm; 237 struct mm_struct *mm = bprm->mm;
238 238
239 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); 239 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
240 if (!vma) 240 if (!vma)
241 goto err; 241 return -ENOMEM;
242 242
243 down_write(&mm->mmap_sem); 243 down_write(&mm->mmap_sem);
244 vma->vm_mm = mm; 244 vma->vm_mm = mm;
@@ -251,28 +251,20 @@ static int __bprm_mm_init(struct linux_binprm *bprm)
251 */ 251 */
252 vma->vm_end = STACK_TOP_MAX; 252 vma->vm_end = STACK_TOP_MAX;
253 vma->vm_start = vma->vm_end - PAGE_SIZE; 253 vma->vm_start = vma->vm_end - PAGE_SIZE;
254
255 vma->vm_flags = VM_STACK_FLAGS; 254 vma->vm_flags = VM_STACK_FLAGS;
256 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); 255 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
257 err = insert_vm_struct(mm, vma); 256 err = insert_vm_struct(mm, vma);
258 if (err) { 257 if (err)
259 up_write(&mm->mmap_sem);
260 goto err; 258 goto err;
261 }
262 259
263 mm->stack_vm = mm->total_vm = 1; 260 mm->stack_vm = mm->total_vm = 1;
264 up_write(&mm->mmap_sem); 261 up_write(&mm->mmap_sem);
265
266 bprm->p = vma->vm_end - sizeof(void *); 262 bprm->p = vma->vm_end - sizeof(void *);
267
268 return 0; 263 return 0;
269
270err: 264err:
271 if (vma) { 265 up_write(&mm->mmap_sem);
272 bprm->vma = NULL; 266 bprm->vma = NULL;
273 kmem_cache_free(vm_area_cachep, vma); 267 kmem_cache_free(vm_area_cachep, vma);
274 }
275
276 return err; 268 return err;
277} 269}
278 270
@@ -1694,7 +1686,7 @@ int get_dumpable(struct mm_struct *mm)
1694 return (ret >= 2) ? 2 : ret; 1686 return (ret >= 2) ? 2 : ret;
1695} 1687}
1696 1688
1697int do_coredump(long signr, int exit_code, struct pt_regs * regs) 1689void do_coredump(long signr, int exit_code, struct pt_regs *regs)
1698{ 1690{
1699 struct core_state core_state; 1691 struct core_state core_state;
1700 char corename[CORENAME_MAX_SIZE + 1]; 1692 char corename[CORENAME_MAX_SIZE + 1];
@@ -1778,6 +1770,11 @@ int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1778 1770
1779 if (ispipe) { 1771 if (ispipe) {
1780 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc); 1772 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
1773 if (!helper_argv) {
1774 printk(KERN_WARNING "%s failed to allocate memory\n",
1775 __func__);
1776 goto fail_unlock;
1777 }
1781 /* Terminate the string before the first option */ 1778 /* Terminate the string before the first option */
1782 delimit = strchr(corename, ' '); 1779 delimit = strchr(corename, ' ');
1783 if (delimit) 1780 if (delimit)
@@ -1845,5 +1842,5 @@ fail_unlock:
1845 put_cred(cred); 1842 put_cred(cred);
1846 coredump_finish(mm); 1843 coredump_finish(mm);
1847fail: 1844fail:
1848 return retval; 1845 return;
1849} 1846}
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index b0537c827024..6c46c648430d 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1225,11 +1225,11 @@ do { \
1225} while (0) 1225} while (0)
1226 1226
1227#ifdef CONFIG_SMP 1227#ifdef CONFIG_SMP
1228/* Each CPU can accumulate FBC_BATCH blocks in their local 1228/* Each CPU can accumulate percpu_counter_batch blocks in their local
1229 * counters. So we need to make sure we have free blocks more 1229 * counters. So we need to make sure we have free blocks more
1230 * than FBC_BATCH * nr_cpu_ids. Also add a window of 4 times. 1230 * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
1231 */ 1231 */
1232#define EXT4_FREEBLOCKS_WATERMARK (4 * (FBC_BATCH * nr_cpu_ids)) 1232#define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
1233#else 1233#else
1234#define EXT4_FREEBLOCKS_WATERMARK 0 1234#define EXT4_FREEBLOCKS_WATERMARK 0
1235#endif 1235#endif
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 6702a49992a6..98d3fe7057ef 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2498,7 +2498,7 @@ static int ext4_nonda_switch(struct super_block *sb)
2498 /* 2498 /*
2499 * switch to non delalloc mode if we are running low 2499 * switch to non delalloc mode if we are running low
2500 * on free block. The free block accounting via percpu 2500 * on free block. The free block accounting via percpu
2501 * counters can get slightly wrong with FBC_BATCH getting 2501 * counters can get slightly wrong with percpu_counter_batch getting
2502 * accumulated on each CPU without updating global counters 2502 * accumulated on each CPU without updating global counters
2503 * Delalloc need an accurate free block accounting. So switch 2503 * Delalloc need an accurate free block accounting. So switch
2504 * to non delalloc when we are near to error range. 2504 * to non delalloc when we are near to error range.
diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index d0ff0b8cf309..e5eaa62fd17f 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -421,9 +421,6 @@ __writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
421 * If we're a pdlfush thread, then implement pdflush collision avoidance 421 * If we're a pdlfush thread, then implement pdflush collision avoidance
422 * against the entire list. 422 * against the entire list.
423 * 423 *
424 * WB_SYNC_HOLD is a hack for sys_sync(): reattach the inode to sb->s_dirty so
425 * that it can be located for waiting on in __writeback_single_inode().
426 *
427 * If `bdi' is non-zero then we're being asked to writeback a specific queue. 424 * If `bdi' is non-zero then we're being asked to writeback a specific queue.
428 * This function assumes that the blockdev superblock's inodes are backed by 425 * This function assumes that the blockdev superblock's inodes are backed by
429 * a variety of queues, so all inodes are searched. For other superblocks, 426 * a variety of queues, so all inodes are searched. For other superblocks,
@@ -443,6 +440,7 @@ void generic_sync_sb_inodes(struct super_block *sb,
443 struct writeback_control *wbc) 440 struct writeback_control *wbc)
444{ 441{
445 const unsigned long start = jiffies; /* livelock avoidance */ 442 const unsigned long start = jiffies; /* livelock avoidance */
443 int sync = wbc->sync_mode == WB_SYNC_ALL;
446 444
447 spin_lock(&inode_lock); 445 spin_lock(&inode_lock);
448 if (!wbc->for_kupdate || list_empty(&sb->s_io)) 446 if (!wbc->for_kupdate || list_empty(&sb->s_io))
@@ -499,10 +497,6 @@ void generic_sync_sb_inodes(struct super_block *sb,
499 __iget(inode); 497 __iget(inode);
500 pages_skipped = wbc->pages_skipped; 498 pages_skipped = wbc->pages_skipped;
501 __writeback_single_inode(inode, wbc); 499 __writeback_single_inode(inode, wbc);
502 if (wbc->sync_mode == WB_SYNC_HOLD) {
503 inode->dirtied_when = jiffies;
504 list_move(&inode->i_list, &sb->s_dirty);
505 }
506 if (current_is_pdflush()) 500 if (current_is_pdflush())
507 writeback_release(bdi); 501 writeback_release(bdi);
508 if (wbc->pages_skipped != pages_skipped) { 502 if (wbc->pages_skipped != pages_skipped) {
@@ -523,7 +517,49 @@ void generic_sync_sb_inodes(struct super_block *sb,
523 if (!list_empty(&sb->s_more_io)) 517 if (!list_empty(&sb->s_more_io))
524 wbc->more_io = 1; 518 wbc->more_io = 1;
525 } 519 }
526 spin_unlock(&inode_lock); 520
521 if (sync) {
522 struct inode *inode, *old_inode = NULL;
523
524 /*
525 * Data integrity sync. Must wait for all pages under writeback,
526 * because there may have been pages dirtied before our sync
527 * call, but which had writeout started before we write it out.
528 * In which case, the inode may not be on the dirty list, but
529 * we still have to wait for that writeout.
530 */
531 list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
532 struct address_space *mapping;
533
534 if (inode->i_state & (I_FREEING|I_WILL_FREE))
535 continue;
536 mapping = inode->i_mapping;
537 if (mapping->nrpages == 0)
538 continue;
539 __iget(inode);
540 spin_unlock(&inode_lock);
541 /*
542 * We hold a reference to 'inode' so it couldn't have
543 * been removed from s_inodes list while we dropped the
544 * inode_lock. We cannot iput the inode now as we can
545 * be holding the last reference and we cannot iput it
546 * under inode_lock. So we keep the reference and iput
547 * it later.
548 */
549 iput(old_inode);
550 old_inode = inode;
551
552 filemap_fdatawait(mapping);
553
554 cond_resched();
555
556 spin_lock(&inode_lock);
557 }
558 spin_unlock(&inode_lock);
559 iput(old_inode);
560 } else
561 spin_unlock(&inode_lock);
562
527 return; /* Leave any unwritten inodes on s_io */ 563 return; /* Leave any unwritten inodes on s_io */
528} 564}
529EXPORT_SYMBOL_GPL(generic_sync_sb_inodes); 565EXPORT_SYMBOL_GPL(generic_sync_sb_inodes);
@@ -588,8 +624,7 @@ restart:
588 624
589/* 625/*
590 * writeback and wait upon the filesystem's dirty inodes. The caller will 626 * writeback and wait upon the filesystem's dirty inodes. The caller will
591 * do this in two passes - one to write, and one to wait. WB_SYNC_HOLD is 627 * do this in two passes - one to write, and one to wait.
592 * used to park the written inodes on sb->s_dirty for the wait pass.
593 * 628 *
594 * A finite limit is set on the number of pages which will be written. 629 * A finite limit is set on the number of pages which will be written.
595 * To prevent infinite livelock of sys_sync(). 630 * To prevent infinite livelock of sys_sync().
@@ -600,30 +635,21 @@ restart:
600void sync_inodes_sb(struct super_block *sb, int wait) 635void sync_inodes_sb(struct super_block *sb, int wait)
601{ 636{
602 struct writeback_control wbc = { 637 struct writeback_control wbc = {
603 .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_HOLD, 638 .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_NONE,
604 .range_start = 0, 639 .range_start = 0,
605 .range_end = LLONG_MAX, 640 .range_end = LLONG_MAX,
606 }; 641 };
607 unsigned long nr_dirty = global_page_state(NR_FILE_DIRTY);
608 unsigned long nr_unstable = global_page_state(NR_UNSTABLE_NFS);
609 642
610 wbc.nr_to_write = nr_dirty + nr_unstable + 643 if (!wait) {
611 (inodes_stat.nr_inodes - inodes_stat.nr_unused) + 644 unsigned long nr_dirty = global_page_state(NR_FILE_DIRTY);
612 nr_dirty + nr_unstable; 645 unsigned long nr_unstable = global_page_state(NR_UNSTABLE_NFS);
613 wbc.nr_to_write += wbc.nr_to_write / 2; /* Bit more for luck */
614 sync_sb_inodes(sb, &wbc);
615}
616 646
617/* 647 wbc.nr_to_write = nr_dirty + nr_unstable +
618 * Rather lame livelock avoidance. 648 (inodes_stat.nr_inodes - inodes_stat.nr_unused);
619 */ 649 } else
620static void set_sb_syncing(int val) 650 wbc.nr_to_write = LONG_MAX; /* doesn't actually matter */
621{ 651
622 struct super_block *sb; 652 sync_sb_inodes(sb, &wbc);
623 spin_lock(&sb_lock);
624 list_for_each_entry_reverse(sb, &super_blocks, s_list)
625 sb->s_syncing = val;
626 spin_unlock(&sb_lock);
627} 653}
628 654
629/** 655/**
@@ -652,9 +678,6 @@ static void __sync_inodes(int wait)
652 spin_lock(&sb_lock); 678 spin_lock(&sb_lock);
653restart: 679restart:
654 list_for_each_entry(sb, &super_blocks, s_list) { 680 list_for_each_entry(sb, &super_blocks, s_list) {
655 if (sb->s_syncing)
656 continue;
657 sb->s_syncing = 1;
658 sb->s_count++; 681 sb->s_count++;
659 spin_unlock(&sb_lock); 682 spin_unlock(&sb_lock);
660 down_read(&sb->s_umount); 683 down_read(&sb->s_umount);
@@ -672,13 +695,10 @@ restart:
672 695
673void sync_inodes(int wait) 696void sync_inodes(int wait)
674{ 697{
675 set_sb_syncing(0);
676 __sync_inodes(0); 698 __sync_inodes(0);
677 699
678 if (wait) { 700 if (wait)
679 set_sb_syncing(0);
680 __sync_inodes(1); 701 __sync_inodes(1);
681 }
682} 702}
683 703
684/** 704/**
diff --git a/fs/fuse/control.c b/fs/fuse/control.c
index 4f3cab321415..99c99dfb0373 100644
--- a/fs/fuse/control.c
+++ b/fs/fuse/control.c
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -48,11 +48,13 @@ static ssize_t fuse_conn_waiting_read(struct file *file, char __user *buf,
48 size_t size; 48 size_t size;
49 49
50 if (!*ppos) { 50 if (!*ppos) {
51 long value;
51 struct fuse_conn *fc = fuse_ctl_file_conn_get(file); 52 struct fuse_conn *fc = fuse_ctl_file_conn_get(file);
52 if (!fc) 53 if (!fc)
53 return 0; 54 return 0;
54 55
55 file->private_data=(void *)(long)atomic_read(&fc->num_waiting); 56 value = atomic_read(&fc->num_waiting);
57 file->private_data = (void *)value;
56 fuse_conn_put(fc); 58 fuse_conn_put(fc);
57 } 59 }
58 size = sprintf(tmp, "%ld\n", (long)file->private_data); 60 size = sprintf(tmp, "%ld\n", (long)file->private_data);
diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
index fba571648a8e..e0c7ada08a1f 100644
--- a/fs/fuse/dev.c
+++ b/fs/fuse/dev.c
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -269,7 +269,7 @@ static void flush_bg_queue(struct fuse_conn *fc)
269 * Called with fc->lock, unlocks it 269 * Called with fc->lock, unlocks it
270 */ 270 */
271static void request_end(struct fuse_conn *fc, struct fuse_req *req) 271static void request_end(struct fuse_conn *fc, struct fuse_req *req)
272 __releases(fc->lock) 272__releases(&fc->lock)
273{ 273{
274 void (*end) (struct fuse_conn *, struct fuse_req *) = req->end; 274 void (*end) (struct fuse_conn *, struct fuse_req *) = req->end;
275 req->end = NULL; 275 req->end = NULL;
@@ -293,13 +293,13 @@ static void request_end(struct fuse_conn *fc, struct fuse_req *req)
293 wake_up(&req->waitq); 293 wake_up(&req->waitq);
294 if (end) 294 if (end)
295 end(fc, req); 295 end(fc, req);
296 else 296 fuse_put_request(fc, req);
297 fuse_put_request(fc, req);
298} 297}
299 298
300static void wait_answer_interruptible(struct fuse_conn *fc, 299static void wait_answer_interruptible(struct fuse_conn *fc,
301 struct fuse_req *req) 300 struct fuse_req *req)
302 __releases(fc->lock) __acquires(fc->lock) 301__releases(&fc->lock)
302__acquires(&fc->lock)
303{ 303{
304 if (signal_pending(current)) 304 if (signal_pending(current))
305 return; 305 return;
@@ -317,7 +317,8 @@ static void queue_interrupt(struct fuse_conn *fc, struct fuse_req *req)
317} 317}
318 318
319static void request_wait_answer(struct fuse_conn *fc, struct fuse_req *req) 319static void request_wait_answer(struct fuse_conn *fc, struct fuse_req *req)
320 __releases(fc->lock) __acquires(fc->lock) 320__releases(&fc->lock)
321__acquires(&fc->lock)
321{ 322{
322 if (!fc->no_interrupt) { 323 if (!fc->no_interrupt) {
323 /* Any signal may interrupt this */ 324 /* Any signal may interrupt this */
@@ -380,7 +381,7 @@ static void request_wait_answer(struct fuse_conn *fc, struct fuse_req *req)
380 } 381 }
381} 382}
382 383
383void request_send(struct fuse_conn *fc, struct fuse_req *req) 384void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req)
384{ 385{
385 req->isreply = 1; 386 req->isreply = 1;
386 spin_lock(&fc->lock); 387 spin_lock(&fc->lock);
@@ -399,8 +400,8 @@ void request_send(struct fuse_conn *fc, struct fuse_req *req)
399 spin_unlock(&fc->lock); 400 spin_unlock(&fc->lock);
400} 401}
401 402
402static void request_send_nowait_locked(struct fuse_conn *fc, 403static void fuse_request_send_nowait_locked(struct fuse_conn *fc,
403 struct fuse_req *req) 404 struct fuse_req *req)
404{ 405{
405 req->background = 1; 406 req->background = 1;
406 fc->num_background++; 407 fc->num_background++;
@@ -414,11 +415,11 @@ static void request_send_nowait_locked(struct fuse_conn *fc,
414 flush_bg_queue(fc); 415 flush_bg_queue(fc);
415} 416}
416 417
417static void request_send_nowait(struct fuse_conn *fc, struct fuse_req *req) 418static void fuse_request_send_nowait(struct fuse_conn *fc, struct fuse_req *req)
418{ 419{
419 spin_lock(&fc->lock); 420 spin_lock(&fc->lock);
420 if (fc->connected) { 421 if (fc->connected) {
421 request_send_nowait_locked(fc, req); 422 fuse_request_send_nowait_locked(fc, req);
422 spin_unlock(&fc->lock); 423 spin_unlock(&fc->lock);
423 } else { 424 } else {
424 req->out.h.error = -ENOTCONN; 425 req->out.h.error = -ENOTCONN;
@@ -426,16 +427,16 @@ static void request_send_nowait(struct fuse_conn *fc, struct fuse_req *req)
426 } 427 }
427} 428}
428 429
429void request_send_noreply(struct fuse_conn *fc, struct fuse_req *req) 430void fuse_request_send_noreply(struct fuse_conn *fc, struct fuse_req *req)
430{ 431{
431 req->isreply = 0; 432 req->isreply = 0;
432 request_send_nowait(fc, req); 433 fuse_request_send_nowait(fc, req);
433} 434}
434 435
435void request_send_background(struct fuse_conn *fc, struct fuse_req *req) 436void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req)
436{ 437{
437 req->isreply = 1; 438 req->isreply = 1;
438 request_send_nowait(fc, req); 439 fuse_request_send_nowait(fc, req);
439} 440}
440 441
441/* 442/*
@@ -443,10 +444,11 @@ void request_send_background(struct fuse_conn *fc, struct fuse_req *req)
443 * 444 *
444 * fc->connected must have been checked previously 445 * fc->connected must have been checked previously
445 */ 446 */
446void request_send_background_locked(struct fuse_conn *fc, struct fuse_req *req) 447void fuse_request_send_background_locked(struct fuse_conn *fc,
448 struct fuse_req *req)
447{ 449{
448 req->isreply = 1; 450 req->isreply = 1;
449 request_send_nowait_locked(fc, req); 451 fuse_request_send_nowait_locked(fc, req);
450} 452}
451 453
452/* 454/*
@@ -539,8 +541,8 @@ static int fuse_copy_fill(struct fuse_copy_state *cs)
539 BUG_ON(!cs->nr_segs); 541 BUG_ON(!cs->nr_segs);
540 cs->seglen = cs->iov[0].iov_len; 542 cs->seglen = cs->iov[0].iov_len;
541 cs->addr = (unsigned long) cs->iov[0].iov_base; 543 cs->addr = (unsigned long) cs->iov[0].iov_base;
542 cs->iov ++; 544 cs->iov++;
543 cs->nr_segs --; 545 cs->nr_segs--;
544 } 546 }
545 down_read(&current->mm->mmap_sem); 547 down_read(&current->mm->mmap_sem);
546 err = get_user_pages(current, current->mm, cs->addr, 1, cs->write, 0, 548 err = get_user_pages(current, current->mm, cs->addr, 1, cs->write, 0,
@@ -589,9 +591,11 @@ static int fuse_copy_page(struct fuse_copy_state *cs, struct page *page,
589 kunmap_atomic(mapaddr, KM_USER1); 591 kunmap_atomic(mapaddr, KM_USER1);
590 } 592 }
591 while (count) { 593 while (count) {
592 int err; 594 if (!cs->len) {
593 if (!cs->len && (err = fuse_copy_fill(cs))) 595 int err = fuse_copy_fill(cs);
594 return err; 596 if (err)
597 return err;
598 }
595 if (page) { 599 if (page) {
596 void *mapaddr = kmap_atomic(page, KM_USER1); 600 void *mapaddr = kmap_atomic(page, KM_USER1);
597 void *buf = mapaddr + offset; 601 void *buf = mapaddr + offset;
@@ -631,9 +635,11 @@ static int fuse_copy_pages(struct fuse_copy_state *cs, unsigned nbytes,
631static int fuse_copy_one(struct fuse_copy_state *cs, void *val, unsigned size) 635static int fuse_copy_one(struct fuse_copy_state *cs, void *val, unsigned size)
632{ 636{
633 while (size) { 637 while (size) {
634 int err; 638 if (!cs->len) {
635 if (!cs->len && (err = fuse_copy_fill(cs))) 639 int err = fuse_copy_fill(cs);
636 return err; 640 if (err)
641 return err;
642 }
637 fuse_copy_do(cs, &val, &size); 643 fuse_copy_do(cs, &val, &size);
638 } 644 }
639 return 0; 645 return 0;
@@ -664,6 +670,8 @@ static int request_pending(struct fuse_conn *fc)
664 670
665/* Wait until a request is available on the pending list */ 671/* Wait until a request is available on the pending list */
666static void request_wait(struct fuse_conn *fc) 672static void request_wait(struct fuse_conn *fc)
673__releases(&fc->lock)
674__acquires(&fc->lock)
667{ 675{
668 DECLARE_WAITQUEUE(wait, current); 676 DECLARE_WAITQUEUE(wait, current);
669 677
@@ -691,7 +699,7 @@ static void request_wait(struct fuse_conn *fc)
691 */ 699 */
692static int fuse_read_interrupt(struct fuse_conn *fc, struct fuse_req *req, 700static int fuse_read_interrupt(struct fuse_conn *fc, struct fuse_req *req,
693 const struct iovec *iov, unsigned long nr_segs) 701 const struct iovec *iov, unsigned long nr_segs)
694 __releases(fc->lock) 702__releases(&fc->lock)
695{ 703{
696 struct fuse_copy_state cs; 704 struct fuse_copy_state cs;
697 struct fuse_in_header ih; 705 struct fuse_in_header ih;
@@ -813,6 +821,34 @@ static ssize_t fuse_dev_read(struct kiocb *iocb, const struct iovec *iov,
813 return err; 821 return err;
814} 822}
815 823
824static int fuse_notify_poll(struct fuse_conn *fc, unsigned int size,
825 struct fuse_copy_state *cs)
826{
827 struct fuse_notify_poll_wakeup_out outarg;
828 int err;
829
830 if (size != sizeof(outarg))
831 return -EINVAL;
832
833 err = fuse_copy_one(cs, &outarg, sizeof(outarg));
834 if (err)
835 return err;
836
837 return fuse_notify_poll_wakeup(fc, &outarg);
838}
839
840static int fuse_notify(struct fuse_conn *fc, enum fuse_notify_code code,
841 unsigned int size, struct fuse_copy_state *cs)
842{
843 switch (code) {
844 case FUSE_NOTIFY_POLL:
845 return fuse_notify_poll(fc, size, cs);
846
847 default:
848 return -EINVAL;
849 }
850}
851
816/* Look up request on processing list by unique ID */ 852/* Look up request on processing list by unique ID */
817static struct fuse_req *request_find(struct fuse_conn *fc, u64 unique) 853static struct fuse_req *request_find(struct fuse_conn *fc, u64 unique)
818{ 854{
@@ -876,9 +912,23 @@ static ssize_t fuse_dev_write(struct kiocb *iocb, const struct iovec *iov,
876 err = fuse_copy_one(&cs, &oh, sizeof(oh)); 912 err = fuse_copy_one(&cs, &oh, sizeof(oh));
877 if (err) 913 if (err)
878 goto err_finish; 914 goto err_finish;
915
916 err = -EINVAL;
917 if (oh.len != nbytes)
918 goto err_finish;
919
920 /*
921 * Zero oh.unique indicates unsolicited notification message
922 * and error contains notification code.
923 */
924 if (!oh.unique) {
925 err = fuse_notify(fc, oh.error, nbytes - sizeof(oh), &cs);
926 fuse_copy_finish(&cs);
927 return err ? err : nbytes;
928 }
929
879 err = -EINVAL; 930 err = -EINVAL;
880 if (!oh.unique || oh.error <= -1000 || oh.error > 0 || 931 if (oh.error <= -1000 || oh.error > 0)
881 oh.len != nbytes)
882 goto err_finish; 932 goto err_finish;
883 933
884 spin_lock(&fc->lock); 934 spin_lock(&fc->lock);
@@ -966,6 +1016,8 @@ static unsigned fuse_dev_poll(struct file *file, poll_table *wait)
966 * This function releases and reacquires fc->lock 1016 * This function releases and reacquires fc->lock
967 */ 1017 */
968static void end_requests(struct fuse_conn *fc, struct list_head *head) 1018static void end_requests(struct fuse_conn *fc, struct list_head *head)
1019__releases(&fc->lock)
1020__acquires(&fc->lock)
969{ 1021{
970 while (!list_empty(head)) { 1022 while (!list_empty(head)) {
971 struct fuse_req *req; 1023 struct fuse_req *req;
@@ -988,7 +1040,8 @@ static void end_requests(struct fuse_conn *fc, struct list_head *head)
988 * locked). 1040 * locked).
989 */ 1041 */
990static void end_io_requests(struct fuse_conn *fc) 1042static void end_io_requests(struct fuse_conn *fc)
991 __releases(fc->lock) __acquires(fc->lock) 1043__releases(&fc->lock)
1044__acquires(&fc->lock)
992{ 1045{
993 while (!list_empty(&fc->io)) { 1046 while (!list_empty(&fc->io)) {
994 struct fuse_req *req = 1047 struct fuse_req *req =
@@ -1002,11 +1055,11 @@ static void end_io_requests(struct fuse_conn *fc)
1002 wake_up(&req->waitq); 1055 wake_up(&req->waitq);
1003 if (end) { 1056 if (end) {
1004 req->end = NULL; 1057 req->end = NULL;
1005 /* The end function will consume this reference */
1006 __fuse_get_request(req); 1058 __fuse_get_request(req);
1007 spin_unlock(&fc->lock); 1059 spin_unlock(&fc->lock);
1008 wait_event(req->waitq, !req->locked); 1060 wait_event(req->waitq, !req->locked);
1009 end(fc, req); 1061 end(fc, req);
1062 fuse_put_request(fc, req);
1010 spin_lock(&fc->lock); 1063 spin_lock(&fc->lock);
1011 } 1064 }
1012 } 1065 }
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index 95bc22bdd060..fdff346e96fd 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -189,7 +189,7 @@ static int fuse_dentry_revalidate(struct dentry *entry, struct nameidata *nd)
189 parent = dget_parent(entry); 189 parent = dget_parent(entry);
190 fuse_lookup_init(fc, req, get_node_id(parent->d_inode), 190 fuse_lookup_init(fc, req, get_node_id(parent->d_inode),
191 &entry->d_name, &outarg); 191 &entry->d_name, &outarg);
192 request_send(fc, req); 192 fuse_request_send(fc, req);
193 dput(parent); 193 dput(parent);
194 err = req->out.h.error; 194 err = req->out.h.error;
195 fuse_put_request(fc, req); 195 fuse_put_request(fc, req);
@@ -204,7 +204,7 @@ static int fuse_dentry_revalidate(struct dentry *entry, struct nameidata *nd)
204 return 0; 204 return 0;
205 } 205 }
206 spin_lock(&fc->lock); 206 spin_lock(&fc->lock);
207 fi->nlookup ++; 207 fi->nlookup++;
208 spin_unlock(&fc->lock); 208 spin_unlock(&fc->lock);
209 } 209 }
210 fuse_put_request(fc, forget_req); 210 fuse_put_request(fc, forget_req);
@@ -283,7 +283,7 @@ int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
283 attr_version = fuse_get_attr_version(fc); 283 attr_version = fuse_get_attr_version(fc);
284 284
285 fuse_lookup_init(fc, req, nodeid, name, outarg); 285 fuse_lookup_init(fc, req, nodeid, name, outarg);
286 request_send(fc, req); 286 fuse_request_send(fc, req);
287 err = req->out.h.error; 287 err = req->out.h.error;
288 fuse_put_request(fc, req); 288 fuse_put_request(fc, req);
289 /* Zero nodeid is same as -ENOENT, but with valid timeout */ 289 /* Zero nodeid is same as -ENOENT, but with valid timeout */
@@ -369,7 +369,7 @@ static void fuse_sync_release(struct fuse_conn *fc, struct fuse_file *ff,
369{ 369{
370 fuse_release_fill(ff, nodeid, flags, FUSE_RELEASE); 370 fuse_release_fill(ff, nodeid, flags, FUSE_RELEASE);
371 ff->reserved_req->force = 1; 371 ff->reserved_req->force = 1;
372 request_send(fc, ff->reserved_req); 372 fuse_request_send(fc, ff->reserved_req);
373 fuse_put_request(fc, ff->reserved_req); 373 fuse_put_request(fc, ff->reserved_req);
374 kfree(ff); 374 kfree(ff);
375} 375}
@@ -408,7 +408,7 @@ static int fuse_create_open(struct inode *dir, struct dentry *entry, int mode,
408 goto out_put_forget_req; 408 goto out_put_forget_req;
409 409
410 err = -ENOMEM; 410 err = -ENOMEM;
411 ff = fuse_file_alloc(); 411 ff = fuse_file_alloc(fc);
412 if (!ff) 412 if (!ff)
413 goto out_put_request; 413 goto out_put_request;
414 414
@@ -432,7 +432,7 @@ static int fuse_create_open(struct inode *dir, struct dentry *entry, int mode,
432 req->out.args[0].value = &outentry; 432 req->out.args[0].value = &outentry;
433 req->out.args[1].size = sizeof(outopen); 433 req->out.args[1].size = sizeof(outopen);
434 req->out.args[1].value = &outopen; 434 req->out.args[1].value = &outopen;
435 request_send(fc, req); 435 fuse_request_send(fc, req);
436 err = req->out.h.error; 436 err = req->out.h.error;
437 if (err) { 437 if (err) {
438 if (err == -ENOSYS) 438 if (err == -ENOSYS)
@@ -502,7 +502,7 @@ static int create_new_entry(struct fuse_conn *fc, struct fuse_req *req,
502 else 502 else
503 req->out.args[0].size = sizeof(outarg); 503 req->out.args[0].size = sizeof(outarg);
504 req->out.args[0].value = &outarg; 504 req->out.args[0].value = &outarg;
505 request_send(fc, req); 505 fuse_request_send(fc, req);
506 err = req->out.h.error; 506 err = req->out.h.error;
507 fuse_put_request(fc, req); 507 fuse_put_request(fc, req);
508 if (err) 508 if (err)
@@ -631,15 +631,17 @@ static int fuse_unlink(struct inode *dir, struct dentry *entry)
631 req->in.numargs = 1; 631 req->in.numargs = 1;
632 req->in.args[0].size = entry->d_name.len + 1; 632 req->in.args[0].size = entry->d_name.len + 1;
633 req->in.args[0].value = entry->d_name.name; 633 req->in.args[0].value = entry->d_name.name;
634 request_send(fc, req); 634 fuse_request_send(fc, req);
635 err = req->out.h.error; 635 err = req->out.h.error;
636 fuse_put_request(fc, req); 636 fuse_put_request(fc, req);
637 if (!err) { 637 if (!err) {
638 struct inode *inode = entry->d_inode; 638 struct inode *inode = entry->d_inode;
639 639
640 /* Set nlink to zero so the inode can be cleared, if 640 /*
641 the inode does have more links this will be 641 * Set nlink to zero so the inode can be cleared, if the inode
642 discovered at the next lookup/getattr */ 642 * does have more links this will be discovered at the next
643 * lookup/getattr.
644 */
643 clear_nlink(inode); 645 clear_nlink(inode);
644 fuse_invalidate_attr(inode); 646 fuse_invalidate_attr(inode);
645 fuse_invalidate_attr(dir); 647 fuse_invalidate_attr(dir);
@@ -662,7 +664,7 @@ static int fuse_rmdir(struct inode *dir, struct dentry *entry)
662 req->in.numargs = 1; 664 req->in.numargs = 1;
663 req->in.args[0].size = entry->d_name.len + 1; 665 req->in.args[0].size = entry->d_name.len + 1;
664 req->in.args[0].value = entry->d_name.name; 666 req->in.args[0].value = entry->d_name.name;
665 request_send(fc, req); 667 fuse_request_send(fc, req);
666 err = req->out.h.error; 668 err = req->out.h.error;
667 fuse_put_request(fc, req); 669 fuse_put_request(fc, req);
668 if (!err) { 670 if (!err) {
@@ -695,7 +697,7 @@ static int fuse_rename(struct inode *olddir, struct dentry *oldent,
695 req->in.args[1].value = oldent->d_name.name; 697 req->in.args[1].value = oldent->d_name.name;
696 req->in.args[2].size = newent->d_name.len + 1; 698 req->in.args[2].size = newent->d_name.len + 1;
697 req->in.args[2].value = newent->d_name.name; 699 req->in.args[2].value = newent->d_name.name;
698 request_send(fc, req); 700 fuse_request_send(fc, req);
699 err = req->out.h.error; 701 err = req->out.h.error;
700 fuse_put_request(fc, req); 702 fuse_put_request(fc, req);
701 if (!err) { 703 if (!err) {
@@ -811,7 +813,7 @@ static int fuse_do_getattr(struct inode *inode, struct kstat *stat,
811 else 813 else
812 req->out.args[0].size = sizeof(outarg); 814 req->out.args[0].size = sizeof(outarg);
813 req->out.args[0].value = &outarg; 815 req->out.args[0].value = &outarg;
814 request_send(fc, req); 816 fuse_request_send(fc, req);
815 err = req->out.h.error; 817 err = req->out.h.error;
816 fuse_put_request(fc, req); 818 fuse_put_request(fc, req);
817 if (!err) { 819 if (!err) {
@@ -911,7 +913,7 @@ static int fuse_access(struct inode *inode, int mask)
911 req->in.numargs = 1; 913 req->in.numargs = 1;
912 req->in.args[0].size = sizeof(inarg); 914 req->in.args[0].size = sizeof(inarg);
913 req->in.args[0].value = &inarg; 915 req->in.args[0].value = &inarg;
914 request_send(fc, req); 916 fuse_request_send(fc, req);
915 err = req->out.h.error; 917 err = req->out.h.error;
916 fuse_put_request(fc, req); 918 fuse_put_request(fc, req);
917 if (err == -ENOSYS) { 919 if (err == -ENOSYS) {
@@ -1033,7 +1035,7 @@ static int fuse_readdir(struct file *file, void *dstbuf, filldir_t filldir)
1033 req->num_pages = 1; 1035 req->num_pages = 1;
1034 req->pages[0] = page; 1036 req->pages[0] = page;
1035 fuse_read_fill(req, file, inode, file->f_pos, PAGE_SIZE, FUSE_READDIR); 1037 fuse_read_fill(req, file, inode, file->f_pos, PAGE_SIZE, FUSE_READDIR);
1036 request_send(fc, req); 1038 fuse_request_send(fc, req);
1037 nbytes = req->out.args[0].size; 1039 nbytes = req->out.args[0].size;
1038 err = req->out.h.error; 1040 err = req->out.h.error;
1039 fuse_put_request(fc, req); 1041 fuse_put_request(fc, req);
@@ -1067,7 +1069,7 @@ static char *read_link(struct dentry *dentry)
1067 req->out.numargs = 1; 1069 req->out.numargs = 1;
1068 req->out.args[0].size = PAGE_SIZE - 1; 1070 req->out.args[0].size = PAGE_SIZE - 1;
1069 req->out.args[0].value = link; 1071 req->out.args[0].value = link;
1070 request_send(fc, req); 1072 fuse_request_send(fc, req);
1071 if (req->out.h.error) { 1073 if (req->out.h.error) {
1072 free_page((unsigned long) link); 1074 free_page((unsigned long) link);
1073 link = ERR_PTR(req->out.h.error); 1075 link = ERR_PTR(req->out.h.error);
@@ -1273,7 +1275,7 @@ static int fuse_do_setattr(struct dentry *entry, struct iattr *attr,
1273 else 1275 else
1274 req->out.args[0].size = sizeof(outarg); 1276 req->out.args[0].size = sizeof(outarg);
1275 req->out.args[0].value = &outarg; 1277 req->out.args[0].value = &outarg;
1276 request_send(fc, req); 1278 fuse_request_send(fc, req);
1277 err = req->out.h.error; 1279 err = req->out.h.error;
1278 fuse_put_request(fc, req); 1280 fuse_put_request(fc, req);
1279 if (err) { 1281 if (err) {
@@ -1367,7 +1369,7 @@ static int fuse_setxattr(struct dentry *entry, const char *name,
1367 req->in.args[1].value = name; 1369 req->in.args[1].value = name;
1368 req->in.args[2].size = size; 1370 req->in.args[2].size = size;
1369 req->in.args[2].value = value; 1371 req->in.args[2].value = value;
1370 request_send(fc, req); 1372 fuse_request_send(fc, req);
1371 err = req->out.h.error; 1373 err = req->out.h.error;
1372 fuse_put_request(fc, req); 1374 fuse_put_request(fc, req);
1373 if (err == -ENOSYS) { 1375 if (err == -ENOSYS) {
@@ -1413,7 +1415,7 @@ static ssize_t fuse_getxattr(struct dentry *entry, const char *name,
1413 req->out.args[0].size = sizeof(outarg); 1415 req->out.args[0].size = sizeof(outarg);
1414 req->out.args[0].value = &outarg; 1416 req->out.args[0].value = &outarg;
1415 } 1417 }
1416 request_send(fc, req); 1418 fuse_request_send(fc, req);
1417 ret = req->out.h.error; 1419 ret = req->out.h.error;
1418 if (!ret) 1420 if (!ret)
1419 ret = size ? req->out.args[0].size : outarg.size; 1421 ret = size ? req->out.args[0].size : outarg.size;
@@ -1463,7 +1465,7 @@ static ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size)
1463 req->out.args[0].size = sizeof(outarg); 1465 req->out.args[0].size = sizeof(outarg);
1464 req->out.args[0].value = &outarg; 1466 req->out.args[0].value = &outarg;
1465 } 1467 }
1466 request_send(fc, req); 1468 fuse_request_send(fc, req);
1467 ret = req->out.h.error; 1469 ret = req->out.h.error;
1468 if (!ret) 1470 if (!ret)
1469 ret = size ? req->out.args[0].size : outarg.size; 1471 ret = size ? req->out.args[0].size : outarg.size;
@@ -1496,7 +1498,7 @@ static int fuse_removexattr(struct dentry *entry, const char *name)
1496 req->in.numargs = 1; 1498 req->in.numargs = 1;
1497 req->in.args[0].size = strlen(name) + 1; 1499 req->in.args[0].size = strlen(name) + 1;
1498 req->in.args[0].value = name; 1500 req->in.args[0].value = name;
1499 request_send(fc, req); 1501 fuse_request_send(fc, req);
1500 err = req->out.h.error; 1502 err = req->out.h.error;
1501 fuse_put_request(fc, req); 1503 fuse_put_request(fc, req);
1502 if (err == -ENOSYS) { 1504 if (err == -ENOSYS) {
diff --git a/fs/fuse/file.c b/fs/fuse/file.c
index 4c9ee7011265..e8162646a9b5 100644
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -39,14 +39,14 @@ static int fuse_send_open(struct inode *inode, struct file *file, int isdir,
39 req->out.numargs = 1; 39 req->out.numargs = 1;
40 req->out.args[0].size = sizeof(*outargp); 40 req->out.args[0].size = sizeof(*outargp);
41 req->out.args[0].value = outargp; 41 req->out.args[0].value = outargp;
42 request_send(fc, req); 42 fuse_request_send(fc, req);
43 err = req->out.h.error; 43 err = req->out.h.error;
44 fuse_put_request(fc, req); 44 fuse_put_request(fc, req);
45 45
46 return err; 46 return err;
47} 47}
48 48
49struct fuse_file *fuse_file_alloc(void) 49struct fuse_file *fuse_file_alloc(struct fuse_conn *fc)
50{ 50{
51 struct fuse_file *ff; 51 struct fuse_file *ff;
52 ff = kmalloc(sizeof(struct fuse_file), GFP_KERNEL); 52 ff = kmalloc(sizeof(struct fuse_file), GFP_KERNEL);
@@ -58,7 +58,12 @@ struct fuse_file *fuse_file_alloc(void)
58 } else { 58 } else {
59 INIT_LIST_HEAD(&ff->write_entry); 59 INIT_LIST_HEAD(&ff->write_entry);
60 atomic_set(&ff->count, 0); 60 atomic_set(&ff->count, 0);
61 spin_lock(&fc->lock);
62 ff->kh = ++fc->khctr;
63 spin_unlock(&fc->lock);
61 } 64 }
65 RB_CLEAR_NODE(&ff->polled_node);
66 init_waitqueue_head(&ff->poll_wait);
62 } 67 }
63 return ff; 68 return ff;
64} 69}
@@ -79,7 +84,6 @@ static void fuse_release_end(struct fuse_conn *fc, struct fuse_req *req)
79{ 84{
80 dput(req->misc.release.dentry); 85 dput(req->misc.release.dentry);
81 mntput(req->misc.release.vfsmount); 86 mntput(req->misc.release.vfsmount);
82 fuse_put_request(fc, req);
83} 87}
84 88
85static void fuse_file_put(struct fuse_file *ff) 89static void fuse_file_put(struct fuse_file *ff)
@@ -89,7 +93,7 @@ static void fuse_file_put(struct fuse_file *ff)
89 struct inode *inode = req->misc.release.dentry->d_inode; 93 struct inode *inode = req->misc.release.dentry->d_inode;
90 struct fuse_conn *fc = get_fuse_conn(inode); 94 struct fuse_conn *fc = get_fuse_conn(inode);
91 req->end = fuse_release_end; 95 req->end = fuse_release_end;
92 request_send_background(fc, req); 96 fuse_request_send_background(fc, req);
93 kfree(ff); 97 kfree(ff);
94 } 98 }
95} 99}
@@ -109,6 +113,7 @@ void fuse_finish_open(struct inode *inode, struct file *file,
109 113
110int fuse_open_common(struct inode *inode, struct file *file, int isdir) 114int fuse_open_common(struct inode *inode, struct file *file, int isdir)
111{ 115{
116 struct fuse_conn *fc = get_fuse_conn(inode);
112 struct fuse_open_out outarg; 117 struct fuse_open_out outarg;
113 struct fuse_file *ff; 118 struct fuse_file *ff;
114 int err; 119 int err;
@@ -121,7 +126,7 @@ int fuse_open_common(struct inode *inode, struct file *file, int isdir)
121 if (err) 126 if (err)
122 return err; 127 return err;
123 128
124 ff = fuse_file_alloc(); 129 ff = fuse_file_alloc(fc);
125 if (!ff) 130 if (!ff)
126 return -ENOMEM; 131 return -ENOMEM;
127 132
@@ -167,7 +172,11 @@ int fuse_release_common(struct inode *inode, struct file *file, int isdir)
167 172
168 spin_lock(&fc->lock); 173 spin_lock(&fc->lock);
169 list_del(&ff->write_entry); 174 list_del(&ff->write_entry);
175 if (!RB_EMPTY_NODE(&ff->polled_node))
176 rb_erase(&ff->polled_node, &fc->polled_files);
170 spin_unlock(&fc->lock); 177 spin_unlock(&fc->lock);
178
179 wake_up_interruptible_sync(&ff->poll_wait);
171 /* 180 /*
172 * Normally this will send the RELEASE request, 181 * Normally this will send the RELEASE request,
173 * however if some asynchronous READ or WRITE requests 182 * however if some asynchronous READ or WRITE requests
@@ -280,7 +289,7 @@ static int fuse_flush(struct file *file, fl_owner_t id)
280 req->in.args[0].size = sizeof(inarg); 289 req->in.args[0].size = sizeof(inarg);
281 req->in.args[0].value = &inarg; 290 req->in.args[0].value = &inarg;
282 req->force = 1; 291 req->force = 1;
283 request_send(fc, req); 292 fuse_request_send(fc, req);
284 err = req->out.h.error; 293 err = req->out.h.error;
285 fuse_put_request(fc, req); 294 fuse_put_request(fc, req);
286 if (err == -ENOSYS) { 295 if (err == -ENOSYS) {
@@ -344,7 +353,7 @@ int fuse_fsync_common(struct file *file, struct dentry *de, int datasync,
344 req->in.numargs = 1; 353 req->in.numargs = 1;
345 req->in.args[0].size = sizeof(inarg); 354 req->in.args[0].size = sizeof(inarg);
346 req->in.args[0].value = &inarg; 355 req->in.args[0].value = &inarg;
347 request_send(fc, req); 356 fuse_request_send(fc, req);
348 err = req->out.h.error; 357 err = req->out.h.error;
349 fuse_put_request(fc, req); 358 fuse_put_request(fc, req);
350 if (err == -ENOSYS) { 359 if (err == -ENOSYS) {
@@ -396,7 +405,7 @@ static size_t fuse_send_read(struct fuse_req *req, struct file *file,
396 inarg->read_flags |= FUSE_READ_LOCKOWNER; 405 inarg->read_flags |= FUSE_READ_LOCKOWNER;
397 inarg->lock_owner = fuse_lock_owner_id(fc, owner); 406 inarg->lock_owner = fuse_lock_owner_id(fc, owner);
398 } 407 }
399 request_send(fc, req); 408 fuse_request_send(fc, req);
400 return req->out.args[0].size; 409 return req->out.args[0].size;
401} 410}
402 411
@@ -493,7 +502,6 @@ static void fuse_readpages_end(struct fuse_conn *fc, struct fuse_req *req)
493 } 502 }
494 if (req->ff) 503 if (req->ff)
495 fuse_file_put(req->ff); 504 fuse_file_put(req->ff);
496 fuse_put_request(fc, req);
497} 505}
498 506
499static void fuse_send_readpages(struct fuse_req *req, struct file *file, 507static void fuse_send_readpages(struct fuse_req *req, struct file *file,
@@ -509,10 +517,11 @@ static void fuse_send_readpages(struct fuse_req *req, struct file *file,
509 struct fuse_file *ff = file->private_data; 517 struct fuse_file *ff = file->private_data;
510 req->ff = fuse_file_get(ff); 518 req->ff = fuse_file_get(ff);
511 req->end = fuse_readpages_end; 519 req->end = fuse_readpages_end;
512 request_send_background(fc, req); 520 fuse_request_send_background(fc, req);
513 } else { 521 } else {
514 request_send(fc, req); 522 fuse_request_send(fc, req);
515 fuse_readpages_end(fc, req); 523 fuse_readpages_end(fc, req);
524 fuse_put_request(fc, req);
516 } 525 }
517} 526}
518 527
@@ -543,7 +552,7 @@ static int fuse_readpages_fill(void *_data, struct page *page)
543 } 552 }
544 } 553 }
545 req->pages[req->num_pages] = page; 554 req->pages[req->num_pages] = page;
546 req->num_pages ++; 555 req->num_pages++;
547 return 0; 556 return 0;
548} 557}
549 558
@@ -636,7 +645,7 @@ static size_t fuse_send_write(struct fuse_req *req, struct file *file,
636 inarg->write_flags |= FUSE_WRITE_LOCKOWNER; 645 inarg->write_flags |= FUSE_WRITE_LOCKOWNER;
637 inarg->lock_owner = fuse_lock_owner_id(fc, owner); 646 inarg->lock_owner = fuse_lock_owner_id(fc, owner);
638 } 647 }
639 request_send(fc, req); 648 fuse_request_send(fc, req);
640 return req->misc.write.out.size; 649 return req->misc.write.out.size;
641} 650}
642 651
@@ -1042,7 +1051,6 @@ static void fuse_writepage_free(struct fuse_conn *fc, struct fuse_req *req)
1042{ 1051{
1043 __free_page(req->pages[0]); 1052 __free_page(req->pages[0]);
1044 fuse_file_put(req->ff); 1053 fuse_file_put(req->ff);
1045 fuse_put_request(fc, req);
1046} 1054}
1047 1055
1048static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req) 1056static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req)
@@ -1060,6 +1068,8 @@ static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req)
1060 1068
1061/* Called under fc->lock, may release and reacquire it */ 1069/* Called under fc->lock, may release and reacquire it */
1062static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req) 1070static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req)
1071__releases(&fc->lock)
1072__acquires(&fc->lock)
1063{ 1073{
1064 struct fuse_inode *fi = get_fuse_inode(req->inode); 1074 struct fuse_inode *fi = get_fuse_inode(req->inode);
1065 loff_t size = i_size_read(req->inode); 1075 loff_t size = i_size_read(req->inode);
@@ -1079,13 +1089,14 @@ static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req)
1079 1089
1080 req->in.args[1].size = inarg->size; 1090 req->in.args[1].size = inarg->size;
1081 fi->writectr++; 1091 fi->writectr++;
1082 request_send_background_locked(fc, req); 1092 fuse_request_send_background_locked(fc, req);
1083 return; 1093 return;
1084 1094
1085 out_free: 1095 out_free:
1086 fuse_writepage_finish(fc, req); 1096 fuse_writepage_finish(fc, req);
1087 spin_unlock(&fc->lock); 1097 spin_unlock(&fc->lock);
1088 fuse_writepage_free(fc, req); 1098 fuse_writepage_free(fc, req);
1099 fuse_put_request(fc, req);
1089 spin_lock(&fc->lock); 1100 spin_lock(&fc->lock);
1090} 1101}
1091 1102
@@ -1096,6 +1107,8 @@ static void fuse_send_writepage(struct fuse_conn *fc, struct fuse_req *req)
1096 * Called with fc->lock 1107 * Called with fc->lock
1097 */ 1108 */
1098void fuse_flush_writepages(struct inode *inode) 1109void fuse_flush_writepages(struct inode *inode)
1110__releases(&fc->lock)
1111__acquires(&fc->lock)
1099{ 1112{
1100 struct fuse_conn *fc = get_fuse_conn(inode); 1113 struct fuse_conn *fc = get_fuse_conn(inode);
1101 struct fuse_inode *fi = get_fuse_inode(inode); 1114 struct fuse_inode *fi = get_fuse_inode(inode);
@@ -1325,7 +1338,7 @@ static int fuse_getlk(struct file *file, struct file_lock *fl)
1325 req->out.numargs = 1; 1338 req->out.numargs = 1;
1326 req->out.args[0].size = sizeof(outarg); 1339 req->out.args[0].size = sizeof(outarg);
1327 req->out.args[0].value = &outarg; 1340 req->out.args[0].value = &outarg;
1328 request_send(fc, req); 1341 fuse_request_send(fc, req);
1329 err = req->out.h.error; 1342 err = req->out.h.error;
1330 fuse_put_request(fc, req); 1343 fuse_put_request(fc, req);
1331 if (!err) 1344 if (!err)
@@ -1357,7 +1370,7 @@ static int fuse_setlk(struct file *file, struct file_lock *fl, int flock)
1357 return PTR_ERR(req); 1370 return PTR_ERR(req);
1358 1371
1359 fuse_lk_fill(req, file, fl, opcode, pid, flock); 1372 fuse_lk_fill(req, file, fl, opcode, pid, flock);
1360 request_send(fc, req); 1373 fuse_request_send(fc, req);
1361 err = req->out.h.error; 1374 err = req->out.h.error;
1362 /* locking is restartable */ 1375 /* locking is restartable */
1363 if (err == -EINTR) 1376 if (err == -EINTR)
@@ -1433,7 +1446,7 @@ static sector_t fuse_bmap(struct address_space *mapping, sector_t block)
1433 req->out.numargs = 1; 1446 req->out.numargs = 1;
1434 req->out.args[0].size = sizeof(outarg); 1447 req->out.args[0].size = sizeof(outarg);
1435 req->out.args[0].value = &outarg; 1448 req->out.args[0].value = &outarg;
1436 request_send(fc, req); 1449 fuse_request_send(fc, req);
1437 err = req->out.h.error; 1450 err = req->out.h.error;
1438 fuse_put_request(fc, req); 1451 fuse_put_request(fc, req);
1439 if (err == -ENOSYS) 1452 if (err == -ENOSYS)
@@ -1470,6 +1483,406 @@ static loff_t fuse_file_llseek(struct file *file, loff_t offset, int origin)
1470 return retval; 1483 return retval;
1471} 1484}
1472 1485
1486static int fuse_ioctl_copy_user(struct page **pages, struct iovec *iov,
1487 unsigned int nr_segs, size_t bytes, bool to_user)
1488{
1489 struct iov_iter ii;
1490 int page_idx = 0;
1491
1492 if (!bytes)
1493 return 0;
1494
1495 iov_iter_init(&ii, iov, nr_segs, bytes, 0);
1496
1497 while (iov_iter_count(&ii)) {
1498 struct page *page = pages[page_idx++];
1499 size_t todo = min_t(size_t, PAGE_SIZE, iov_iter_count(&ii));
1500 void *kaddr, *map;
1501
1502 kaddr = map = kmap(page);
1503
1504 while (todo) {
1505 char __user *uaddr = ii.iov->iov_base + ii.iov_offset;
1506 size_t iov_len = ii.iov->iov_len - ii.iov_offset;
1507 size_t copy = min(todo, iov_len);
1508 size_t left;
1509
1510 if (!to_user)
1511 left = copy_from_user(kaddr, uaddr, copy);
1512 else
1513 left = copy_to_user(uaddr, kaddr, copy);
1514
1515 if (unlikely(left))
1516 return -EFAULT;
1517
1518 iov_iter_advance(&ii, copy);
1519 todo -= copy;
1520 kaddr += copy;
1521 }
1522
1523 kunmap(map);
1524 }
1525
1526 return 0;
1527}
1528
1529/*
1530 * For ioctls, there is no generic way to determine how much memory
1531 * needs to be read and/or written. Furthermore, ioctls are allowed
1532 * to dereference the passed pointer, so the parameter requires deep
1533 * copying but FUSE has no idea whatsoever about what to copy in or
1534 * out.
1535 *
1536 * This is solved by allowing FUSE server to retry ioctl with
1537 * necessary in/out iovecs. Let's assume the ioctl implementation
1538 * needs to read in the following structure.
1539 *
1540 * struct a {
1541 * char *buf;
1542 * size_t buflen;
1543 * }
1544 *
1545 * On the first callout to FUSE server, inarg->in_size and
1546 * inarg->out_size will be NULL; then, the server completes the ioctl
1547 * with FUSE_IOCTL_RETRY set in out->flags, out->in_iovs set to 1 and
1548 * the actual iov array to
1549 *
1550 * { { .iov_base = inarg.arg, .iov_len = sizeof(struct a) } }
1551 *
1552 * which tells FUSE to copy in the requested area and retry the ioctl.
1553 * On the second round, the server has access to the structure and
1554 * from that it can tell what to look for next, so on the invocation,
1555 * it sets FUSE_IOCTL_RETRY, out->in_iovs to 2 and iov array to
1556 *
1557 * { { .iov_base = inarg.arg, .iov_len = sizeof(struct a) },
1558 * { .iov_base = a.buf, .iov_len = a.buflen } }
1559 *
1560 * FUSE will copy both struct a and the pointed buffer from the
1561 * process doing the ioctl and retry ioctl with both struct a and the
1562 * buffer.
1563 *
1564 * This time, FUSE server has everything it needs and completes ioctl
1565 * without FUSE_IOCTL_RETRY which finishes the ioctl call.
1566 *
1567 * Copying data out works the same way.
1568 *
1569 * Note that if FUSE_IOCTL_UNRESTRICTED is clear, the kernel
1570 * automatically initializes in and out iovs by decoding @cmd with
1571 * _IOC_* macros and the server is not allowed to request RETRY. This
1572 * limits ioctl data transfers to well-formed ioctls and is the forced
1573 * behavior for all FUSE servers.
1574 */
1575static long fuse_file_do_ioctl(struct file *file, unsigned int cmd,
1576 unsigned long arg, unsigned int flags)
1577{
1578 struct inode *inode = file->f_dentry->d_inode;
1579 struct fuse_file *ff = file->private_data;
1580 struct fuse_conn *fc = get_fuse_conn(inode);
1581 struct fuse_ioctl_in inarg = {
1582 .fh = ff->fh,
1583 .cmd = cmd,
1584 .arg = arg,
1585 .flags = flags
1586 };
1587 struct fuse_ioctl_out outarg;
1588 struct fuse_req *req = NULL;
1589 struct page **pages = NULL;
1590 struct page *iov_page = NULL;
1591 struct iovec *in_iov = NULL, *out_iov = NULL;
1592 unsigned int in_iovs = 0, out_iovs = 0, num_pages = 0, max_pages;
1593 size_t in_size, out_size, transferred;
1594 int err;
1595
1596 /* assume all the iovs returned by client always fits in a page */
1597 BUILD_BUG_ON(sizeof(struct iovec) * FUSE_IOCTL_MAX_IOV > PAGE_SIZE);
1598
1599 if (!fuse_allow_task(fc, current))
1600 return -EACCES;
1601
1602 err = -EIO;
1603 if (is_bad_inode(inode))
1604 goto out;
1605
1606 err = -ENOMEM;
1607 pages = kzalloc(sizeof(pages[0]) * FUSE_MAX_PAGES_PER_REQ, GFP_KERNEL);
1608 iov_page = alloc_page(GFP_KERNEL);
1609 if (!pages || !iov_page)
1610 goto out;
1611
1612 /*
1613 * If restricted, initialize IO parameters as encoded in @cmd.
1614 * RETRY from server is not allowed.
1615 */
1616 if (!(flags & FUSE_IOCTL_UNRESTRICTED)) {
1617 struct iovec *iov = page_address(iov_page);
1618
1619 iov->iov_base = (void __user *)arg;
1620 iov->iov_len = _IOC_SIZE(cmd);
1621
1622 if (_IOC_DIR(cmd) & _IOC_WRITE) {
1623 in_iov = iov;
1624 in_iovs = 1;
1625 }
1626
1627 if (_IOC_DIR(cmd) & _IOC_READ) {
1628 out_iov = iov;
1629 out_iovs = 1;
1630 }
1631 }
1632
1633 retry:
1634 inarg.in_size = in_size = iov_length(in_iov, in_iovs);
1635 inarg.out_size = out_size = iov_length(out_iov, out_iovs);
1636
1637 /*
1638 * Out data can be used either for actual out data or iovs,
1639 * make sure there always is at least one page.
1640 */
1641 out_size = max_t(size_t, out_size, PAGE_SIZE);
1642 max_pages = DIV_ROUND_UP(max(in_size, out_size), PAGE_SIZE);
1643
1644 /* make sure there are enough buffer pages and init request with them */
1645 err = -ENOMEM;
1646 if (max_pages > FUSE_MAX_PAGES_PER_REQ)
1647 goto out;
1648 while (num_pages < max_pages) {
1649 pages[num_pages] = alloc_page(GFP_KERNEL | __GFP_HIGHMEM);
1650 if (!pages[num_pages])
1651 goto out;
1652 num_pages++;
1653 }
1654
1655 req = fuse_get_req(fc);
1656 if (IS_ERR(req)) {
1657 err = PTR_ERR(req);
1658 req = NULL;
1659 goto out;
1660 }
1661 memcpy(req->pages, pages, sizeof(req->pages[0]) * num_pages);
1662 req->num_pages = num_pages;
1663
1664 /* okay, let's send it to the client */
1665 req->in.h.opcode = FUSE_IOCTL;
1666 req->in.h.nodeid = get_node_id(inode);
1667 req->in.numargs = 1;
1668 req->in.args[0].size = sizeof(inarg);
1669 req->in.args[0].value = &inarg;
1670 if (in_size) {
1671 req->in.numargs++;
1672 req->in.args[1].size = in_size;
1673 req->in.argpages = 1;
1674
1675 err = fuse_ioctl_copy_user(pages, in_iov, in_iovs, in_size,
1676 false);
1677 if (err)
1678 goto out;
1679 }
1680
1681 req->out.numargs = 2;
1682 req->out.args[0].size = sizeof(outarg);
1683 req->out.args[0].value = &outarg;
1684 req->out.args[1].size = out_size;
1685 req->out.argpages = 1;
1686 req->out.argvar = 1;
1687
1688 fuse_request_send(fc, req);
1689 err = req->out.h.error;
1690 transferred = req->out.args[1].size;
1691 fuse_put_request(fc, req);
1692 req = NULL;
1693 if (err)
1694 goto out;
1695
1696 /* did it ask for retry? */
1697 if (outarg.flags & FUSE_IOCTL_RETRY) {
1698 char *vaddr;
1699
1700 /* no retry if in restricted mode */
1701 err = -EIO;
1702 if (!(flags & FUSE_IOCTL_UNRESTRICTED))
1703 goto out;
1704
1705 in_iovs = outarg.in_iovs;
1706 out_iovs = outarg.out_iovs;
1707
1708 /*
1709 * Make sure things are in boundary, separate checks
1710 * are to protect against overflow.
1711 */
1712 err = -ENOMEM;
1713 if (in_iovs > FUSE_IOCTL_MAX_IOV ||
1714 out_iovs > FUSE_IOCTL_MAX_IOV ||
1715 in_iovs + out_iovs > FUSE_IOCTL_MAX_IOV)
1716 goto out;
1717
1718 err = -EIO;
1719 if ((in_iovs + out_iovs) * sizeof(struct iovec) != transferred)
1720 goto out;
1721
1722 /* okay, copy in iovs and retry */
1723 vaddr = kmap_atomic(pages[0], KM_USER0);
1724 memcpy(page_address(iov_page), vaddr, transferred);
1725 kunmap_atomic(vaddr, KM_USER0);
1726
1727 in_iov = page_address(iov_page);
1728 out_iov = in_iov + in_iovs;
1729
1730 goto retry;
1731 }
1732
1733 err = -EIO;
1734 if (transferred > inarg.out_size)
1735 goto out;
1736
1737 err = fuse_ioctl_copy_user(pages, out_iov, out_iovs, transferred, true);
1738 out:
1739 if (req)
1740 fuse_put_request(fc, req);
1741 if (iov_page)
1742 __free_page(iov_page);
1743 while (num_pages)
1744 __free_page(pages[--num_pages]);
1745 kfree(pages);
1746
1747 return err ? err : outarg.result;
1748}
1749
1750static long fuse_file_ioctl(struct file *file, unsigned int cmd,
1751 unsigned long arg)
1752{
1753 return fuse_file_do_ioctl(file, cmd, arg, 0);
1754}
1755
1756static long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
1757 unsigned long arg)
1758{
1759 return fuse_file_do_ioctl(file, cmd, arg, FUSE_IOCTL_COMPAT);
1760}
1761
1762/*
1763 * All files which have been polled are linked to RB tree
1764 * fuse_conn->polled_files which is indexed by kh. Walk the tree and
1765 * find the matching one.
1766 */
1767static struct rb_node **fuse_find_polled_node(struct fuse_conn *fc, u64 kh,
1768 struct rb_node **parent_out)
1769{
1770 struct rb_node **link = &fc->polled_files.rb_node;
1771 struct rb_node *last = NULL;
1772
1773 while (*link) {
1774 struct fuse_file *ff;
1775
1776 last = *link;
1777 ff = rb_entry(last, struct fuse_file, polled_node);
1778
1779 if (kh < ff->kh)
1780 link = &last->rb_left;
1781 else if (kh > ff->kh)
1782 link = &last->rb_right;
1783 else
1784 return link;
1785 }
1786
1787 if (parent_out)
1788 *parent_out = last;
1789 return link;
1790}
1791
1792/*
1793 * The file is about to be polled. Make sure it's on the polled_files
1794 * RB tree. Note that files once added to the polled_files tree are
1795 * not removed before the file is released. This is because a file
1796 * polled once is likely to be polled again.
1797 */
1798static void fuse_register_polled_file(struct fuse_conn *fc,
1799 struct fuse_file *ff)
1800{
1801 spin_lock(&fc->lock);
1802 if (RB_EMPTY_NODE(&ff->polled_node)) {
1803 struct rb_node **link, *parent;
1804
1805 link = fuse_find_polled_node(fc, ff->kh, &parent);
1806 BUG_ON(*link);
1807 rb_link_node(&ff->polled_node, parent, link);
1808 rb_insert_color(&ff->polled_node, &fc->polled_files);
1809 }
1810 spin_unlock(&fc->lock);
1811}
1812
1813static unsigned fuse_file_poll(struct file *file, poll_table *wait)
1814{
1815 struct inode *inode = file->f_dentry->d_inode;
1816 struct fuse_file *ff = file->private_data;
1817 struct fuse_conn *fc = get_fuse_conn(inode);
1818 struct fuse_poll_in inarg = { .fh = ff->fh, .kh = ff->kh };
1819 struct fuse_poll_out outarg;
1820 struct fuse_req *req;
1821 int err;
1822
1823 if (fc->no_poll)
1824 return DEFAULT_POLLMASK;
1825
1826 poll_wait(file, &ff->poll_wait, wait);
1827
1828 /*
1829 * Ask for notification iff there's someone waiting for it.
1830 * The client may ignore the flag and always notify.
1831 */
1832 if (waitqueue_active(&ff->poll_wait)) {
1833 inarg.flags |= FUSE_POLL_SCHEDULE_NOTIFY;
1834 fuse_register_polled_file(fc, ff);
1835 }
1836
1837 req = fuse_get_req(fc);
1838 if (IS_ERR(req))
1839 return PTR_ERR(req);
1840
1841 req->in.h.opcode = FUSE_POLL;
1842 req->in.h.nodeid = get_node_id(inode);
1843 req->in.numargs = 1;
1844 req->in.args[0].size = sizeof(inarg);
1845 req->in.args[0].value = &inarg;
1846 req->out.numargs = 1;
1847 req->out.args[0].size = sizeof(outarg);
1848 req->out.args[0].value = &outarg;
1849 fuse_request_send(fc, req);
1850 err = req->out.h.error;
1851 fuse_put_request(fc, req);
1852
1853 if (!err)
1854 return outarg.revents;
1855 if (err == -ENOSYS) {
1856 fc->no_poll = 1;
1857 return DEFAULT_POLLMASK;
1858 }
1859 return POLLERR;
1860}
1861
1862/*
1863 * This is called from fuse_handle_notify() on FUSE_NOTIFY_POLL and
1864 * wakes up the poll waiters.
1865 */
1866int fuse_notify_poll_wakeup(struct fuse_conn *fc,
1867 struct fuse_notify_poll_wakeup_out *outarg)
1868{
1869 u64 kh = outarg->kh;
1870 struct rb_node **link;
1871
1872 spin_lock(&fc->lock);
1873
1874 link = fuse_find_polled_node(fc, kh, NULL);
1875 if (*link) {
1876 struct fuse_file *ff;
1877
1878 ff = rb_entry(*link, struct fuse_file, polled_node);
1879 wake_up_interruptible_sync(&ff->poll_wait);
1880 }
1881
1882 spin_unlock(&fc->lock);
1883 return 0;
1884}
1885
1473static const struct file_operations fuse_file_operations = { 1886static const struct file_operations fuse_file_operations = {
1474 .llseek = fuse_file_llseek, 1887 .llseek = fuse_file_llseek,
1475 .read = do_sync_read, 1888 .read = do_sync_read,
@@ -1484,6 +1897,9 @@ static const struct file_operations fuse_file_operations = {
1484 .lock = fuse_file_lock, 1897 .lock = fuse_file_lock,
1485 .flock = fuse_file_flock, 1898 .flock = fuse_file_flock,
1486 .splice_read = generic_file_splice_read, 1899 .splice_read = generic_file_splice_read,
1900 .unlocked_ioctl = fuse_file_ioctl,
1901 .compat_ioctl = fuse_file_compat_ioctl,
1902 .poll = fuse_file_poll,
1487}; 1903};
1488 1904
1489static const struct file_operations fuse_direct_io_file_operations = { 1905static const struct file_operations fuse_direct_io_file_operations = {
@@ -1496,6 +1912,9 @@ static const struct file_operations fuse_direct_io_file_operations = {
1496 .fsync = fuse_fsync, 1912 .fsync = fuse_fsync,
1497 .lock = fuse_file_lock, 1913 .lock = fuse_file_lock,
1498 .flock = fuse_file_flock, 1914 .flock = fuse_file_flock,
1915 .unlocked_ioctl = fuse_file_ioctl,
1916 .compat_ioctl = fuse_file_compat_ioctl,
1917 .poll = fuse_file_poll,
1499 /* no mmap and splice_read */ 1918 /* no mmap and splice_read */
1500}; 1919};
1501 1920
diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
index 35accfdd747f..5e64b815a5a1 100644
--- a/fs/fuse/fuse_i.h
+++ b/fs/fuse/fuse_i.h
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -19,6 +19,8 @@
19#include <linux/backing-dev.h> 19#include <linux/backing-dev.h>
20#include <linux/mutex.h> 20#include <linux/mutex.h>
21#include <linux/rwsem.h> 21#include <linux/rwsem.h>
22#include <linux/rbtree.h>
23#include <linux/poll.h>
22 24
23/** Max number of pages that can be used in a single read request */ 25/** Max number of pages that can be used in a single read request */
24#define FUSE_MAX_PAGES_PER_REQ 32 26#define FUSE_MAX_PAGES_PER_REQ 32
@@ -100,6 +102,9 @@ struct fuse_file {
100 /** Request reserved for flush and release */ 102 /** Request reserved for flush and release */
101 struct fuse_req *reserved_req; 103 struct fuse_req *reserved_req;
102 104
105 /** Kernel file handle guaranteed to be unique */
106 u64 kh;
107
103 /** File handle used by userspace */ 108 /** File handle used by userspace */
104 u64 fh; 109 u64 fh;
105 110
@@ -108,6 +113,12 @@ struct fuse_file {
108 113
109 /** Entry on inode's write_files list */ 114 /** Entry on inode's write_files list */
110 struct list_head write_entry; 115 struct list_head write_entry;
116
117 /** RB node to be linked on fuse_conn->polled_files */
118 struct rb_node polled_node;
119
120 /** Wait queue head for poll */
121 wait_queue_head_t poll_wait;
111}; 122};
112 123
113/** One input argument of a request */ 124/** One input argument of a request */
@@ -322,6 +333,12 @@ struct fuse_conn {
322 /** The list of requests under I/O */ 333 /** The list of requests under I/O */
323 struct list_head io; 334 struct list_head io;
324 335
336 /** The next unique kernel file handle */
337 u64 khctr;
338
339 /** rbtree of fuse_files waiting for poll events indexed by ph */
340 struct rb_root polled_files;
341
325 /** Number of requests currently in the background */ 342 /** Number of requests currently in the background */
326 unsigned num_background; 343 unsigned num_background;
327 344
@@ -355,19 +372,19 @@ struct fuse_conn {
355 /** Connection failed (version mismatch). Cannot race with 372 /** Connection failed (version mismatch). Cannot race with
356 setting other bitfields since it is only set once in INIT 373 setting other bitfields since it is only set once in INIT
357 reply, before any other request, and never cleared */ 374 reply, before any other request, and never cleared */
358 unsigned conn_error : 1; 375 unsigned conn_error:1;
359 376
360 /** Connection successful. Only set in INIT */ 377 /** Connection successful. Only set in INIT */
361 unsigned conn_init : 1; 378 unsigned conn_init:1;
362 379
363 /** Do readpages asynchronously? Only set in INIT */ 380 /** Do readpages asynchronously? Only set in INIT */
364 unsigned async_read : 1; 381 unsigned async_read:1;
365 382
366 /** Do not send separate SETATTR request before open(O_TRUNC) */ 383 /** Do not send separate SETATTR request before open(O_TRUNC) */
367 unsigned atomic_o_trunc : 1; 384 unsigned atomic_o_trunc:1;
368 385
369 /** Filesystem supports NFS exporting. Only set in INIT */ 386 /** Filesystem supports NFS exporting. Only set in INIT */
370 unsigned export_support : 1; 387 unsigned export_support:1;
371 388
372 /* 389 /*
373 * The following bitfields are only for optimization purposes 390 * The following bitfields are only for optimization purposes
@@ -375,43 +392,46 @@ struct fuse_conn {
375 */ 392 */
376 393
377 /** Is fsync not implemented by fs? */ 394 /** Is fsync not implemented by fs? */
378 unsigned no_fsync : 1; 395 unsigned no_fsync:1;
379 396
380 /** Is fsyncdir not implemented by fs? */ 397 /** Is fsyncdir not implemented by fs? */
381 unsigned no_fsyncdir : 1; 398 unsigned no_fsyncdir:1;
382 399
383 /** Is flush not implemented by fs? */ 400 /** Is flush not implemented by fs? */
384 unsigned no_flush : 1; 401 unsigned no_flush:1;
385 402
386 /** Is setxattr not implemented by fs? */ 403 /** Is setxattr not implemented by fs? */
387 unsigned no_setxattr : 1; 404 unsigned no_setxattr:1;
388 405
389 /** Is getxattr not implemented by fs? */ 406 /** Is getxattr not implemented by fs? */
390 unsigned no_getxattr : 1; 407 unsigned no_getxattr:1;
391 408
392 /** Is listxattr not implemented by fs? */ 409 /** Is listxattr not implemented by fs? */
393 unsigned no_listxattr : 1; 410 unsigned no_listxattr:1;
394 411
395 /** Is removexattr not implemented by fs? */ 412 /** Is removexattr not implemented by fs? */
396 unsigned no_removexattr : 1; 413 unsigned no_removexattr:1;
397 414
398 /** Are file locking primitives not implemented by fs? */ 415 /** Are file locking primitives not implemented by fs? */
399 unsigned no_lock : 1; 416 unsigned no_lock:1;
400 417
401 /** Is access not implemented by fs? */ 418 /** Is access not implemented by fs? */
402 unsigned no_access : 1; 419 unsigned no_access:1;
403 420
404 /** Is create not implemented by fs? */ 421 /** Is create not implemented by fs? */
405 unsigned no_create : 1; 422 unsigned no_create:1;
406 423
407 /** Is interrupt not implemented by fs? */ 424 /** Is interrupt not implemented by fs? */
408 unsigned no_interrupt : 1; 425 unsigned no_interrupt:1;
409 426
410 /** Is bmap not implemented by fs? */ 427 /** Is bmap not implemented by fs? */
411 unsigned no_bmap : 1; 428 unsigned no_bmap:1;
429
430 /** Is poll not implemented by fs? */
431 unsigned no_poll:1;
412 432
413 /** Do multi-page cached writes */ 433 /** Do multi-page cached writes */
414 unsigned big_writes : 1; 434 unsigned big_writes:1;
415 435
416 /** The number of requests waiting for completion */ 436 /** The number of requests waiting for completion */
417 atomic_t num_waiting; 437 atomic_t num_waiting;
@@ -445,6 +465,9 @@ struct fuse_conn {
445 465
446 /** Version counter for attribute changes */ 466 /** Version counter for attribute changes */
447 u64 attr_version; 467 u64 attr_version;
468
469 /** Called on final put */
470 void (*release)(struct fuse_conn *);
448}; 471};
449 472
450static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb) 473static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
@@ -499,7 +522,7 @@ void fuse_read_fill(struct fuse_req *req, struct file *file,
499 */ 522 */
500int fuse_open_common(struct inode *inode, struct file *file, int isdir); 523int fuse_open_common(struct inode *inode, struct file *file, int isdir);
501 524
502struct fuse_file *fuse_file_alloc(void); 525struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
503void fuse_file_free(struct fuse_file *ff); 526void fuse_file_free(struct fuse_file *ff);
504void fuse_finish_open(struct inode *inode, struct file *file, 527void fuse_finish_open(struct inode *inode, struct file *file,
505 struct fuse_file *ff, struct fuse_open_out *outarg); 528 struct fuse_file *ff, struct fuse_open_out *outarg);
@@ -519,6 +542,12 @@ int fuse_fsync_common(struct file *file, struct dentry *de, int datasync,
519 int isdir); 542 int isdir);
520 543
521/** 544/**
545 * Notify poll wakeup
546 */
547int fuse_notify_poll_wakeup(struct fuse_conn *fc,
548 struct fuse_notify_poll_wakeup_out *outarg);
549
550/**
522 * Initialize file operations on a regular file 551 * Initialize file operations on a regular file
523 */ 552 */
524void fuse_init_file_inode(struct inode *inode); 553void fuse_init_file_inode(struct inode *inode);
@@ -593,19 +622,20 @@ void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
593/** 622/**
594 * Send a request (synchronous) 623 * Send a request (synchronous)
595 */ 624 */
596void request_send(struct fuse_conn *fc, struct fuse_req *req); 625void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
597 626
598/** 627/**
599 * Send a request with no reply 628 * Send a request with no reply
600 */ 629 */
601void request_send_noreply(struct fuse_conn *fc, struct fuse_req *req); 630void fuse_request_send_noreply(struct fuse_conn *fc, struct fuse_req *req);
602 631
603/** 632/**
604 * Send a request in the background 633 * Send a request in the background
605 */ 634 */
606void request_send_background(struct fuse_conn *fc, struct fuse_req *req); 635void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
607 636
608void request_send_background_locked(struct fuse_conn *fc, struct fuse_req *req); 637void fuse_request_send_background_locked(struct fuse_conn *fc,
638 struct fuse_req *req);
609 639
610/* Abort all requests */ 640/* Abort all requests */
611void fuse_abort_conn(struct fuse_conn *fc); 641void fuse_abort_conn(struct fuse_conn *fc);
@@ -623,6 +653,11 @@ void fuse_invalidate_entry_cache(struct dentry *entry);
623struct fuse_conn *fuse_conn_get(struct fuse_conn *fc); 653struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
624 654
625/** 655/**
656 * Initialize fuse_conn
657 */
658int fuse_conn_init(struct fuse_conn *fc, struct super_block *sb);
659
660/**
626 * Release reference to fuse_conn 661 * Release reference to fuse_conn
627 */ 662 */
628void fuse_conn_put(struct fuse_conn *fc); 663void fuse_conn_put(struct fuse_conn *fc);
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 2e99f34b4435..47c96fdca1ac 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -1,6 +1,6 @@
1/* 1/*
2 FUSE: Filesystem in Userspace 2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2006 Miklos Szeredi <miklos@szeredi.hu> 3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4 4
5 This program can be distributed under the terms of the GNU GPL. 5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING. 6 See the file COPYING.
@@ -37,10 +37,10 @@ struct fuse_mount_data {
37 unsigned rootmode; 37 unsigned rootmode;
38 unsigned user_id; 38 unsigned user_id;
39 unsigned group_id; 39 unsigned group_id;
40 unsigned fd_present : 1; 40 unsigned fd_present:1;
41 unsigned rootmode_present : 1; 41 unsigned rootmode_present:1;
42 unsigned user_id_present : 1; 42 unsigned user_id_present:1;
43 unsigned group_id_present : 1; 43 unsigned group_id_present:1;
44 unsigned flags; 44 unsigned flags;
45 unsigned max_read; 45 unsigned max_read;
46 unsigned blksize; 46 unsigned blksize;
@@ -94,7 +94,7 @@ void fuse_send_forget(struct fuse_conn *fc, struct fuse_req *req,
94 req->in.numargs = 1; 94 req->in.numargs = 1;
95 req->in.args[0].size = sizeof(struct fuse_forget_in); 95 req->in.args[0].size = sizeof(struct fuse_forget_in);
96 req->in.args[0].value = inarg; 96 req->in.args[0].value = inarg;
97 request_send_noreply(fc, req); 97 fuse_request_send_noreply(fc, req);
98} 98}
99 99
100static void fuse_clear_inode(struct inode *inode) 100static void fuse_clear_inode(struct inode *inode)
@@ -250,7 +250,7 @@ struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
250 250
251 fi = get_fuse_inode(inode); 251 fi = get_fuse_inode(inode);
252 spin_lock(&fc->lock); 252 spin_lock(&fc->lock);
253 fi->nlookup ++; 253 fi->nlookup++;
254 spin_unlock(&fc->lock); 254 spin_unlock(&fc->lock);
255 fuse_change_attributes(inode, attr, attr_valid, attr_version); 255 fuse_change_attributes(inode, attr, attr_valid, attr_version);
256 256
@@ -269,7 +269,7 @@ static void fuse_send_destroy(struct fuse_conn *fc)
269 fc->destroy_req = NULL; 269 fc->destroy_req = NULL;
270 req->in.h.opcode = FUSE_DESTROY; 270 req->in.h.opcode = FUSE_DESTROY;
271 req->force = 1; 271 req->force = 1;
272 request_send(fc, req); 272 fuse_request_send(fc, req);
273 fuse_put_request(fc, req); 273 fuse_put_request(fc, req);
274 } 274 }
275} 275}
@@ -334,7 +334,7 @@ static int fuse_statfs(struct dentry *dentry, struct kstatfs *buf)
334 req->out.args[0].size = 334 req->out.args[0].size =
335 fc->minor < 4 ? FUSE_COMPAT_STATFS_SIZE : sizeof(outarg); 335 fc->minor < 4 ? FUSE_COMPAT_STATFS_SIZE : sizeof(outarg);
336 req->out.args[0].value = &outarg; 336 req->out.args[0].value = &outarg;
337 request_send(fc, req); 337 fuse_request_send(fc, req);
338 err = req->out.h.error; 338 err = req->out.h.error;
339 if (!err) 339 if (!err)
340 convert_fuse_statfs(buf, &outarg.st); 340 convert_fuse_statfs(buf, &outarg.st);
@@ -462,68 +462,69 @@ static int fuse_show_options(struct seq_file *m, struct vfsmount *mnt)
462 return 0; 462 return 0;
463} 463}
464 464
465static struct fuse_conn *new_conn(struct super_block *sb) 465int fuse_conn_init(struct fuse_conn *fc, struct super_block *sb)
466{ 466{
467 struct fuse_conn *fc;
468 int err; 467 int err;
469 468
470 fc = kzalloc(sizeof(*fc), GFP_KERNEL); 469 memset(fc, 0, sizeof(*fc));
471 if (fc) { 470 spin_lock_init(&fc->lock);
472 spin_lock_init(&fc->lock); 471 mutex_init(&fc->inst_mutex);
473 mutex_init(&fc->inst_mutex); 472 atomic_set(&fc->count, 1);
474 atomic_set(&fc->count, 1); 473 init_waitqueue_head(&fc->waitq);
475 init_waitqueue_head(&fc->waitq); 474 init_waitqueue_head(&fc->blocked_waitq);
476 init_waitqueue_head(&fc->blocked_waitq); 475 init_waitqueue_head(&fc->reserved_req_waitq);
477 init_waitqueue_head(&fc->reserved_req_waitq); 476 INIT_LIST_HEAD(&fc->pending);
478 INIT_LIST_HEAD(&fc->pending); 477 INIT_LIST_HEAD(&fc->processing);
479 INIT_LIST_HEAD(&fc->processing); 478 INIT_LIST_HEAD(&fc->io);
480 INIT_LIST_HEAD(&fc->io); 479 INIT_LIST_HEAD(&fc->interrupts);
481 INIT_LIST_HEAD(&fc->interrupts); 480 INIT_LIST_HEAD(&fc->bg_queue);
482 INIT_LIST_HEAD(&fc->bg_queue); 481 INIT_LIST_HEAD(&fc->entry);
483 atomic_set(&fc->num_waiting, 0); 482 atomic_set(&fc->num_waiting, 0);
484 fc->bdi.ra_pages = (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE; 483 fc->bdi.ra_pages = (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE;
485 fc->bdi.unplug_io_fn = default_unplug_io_fn; 484 fc->bdi.unplug_io_fn = default_unplug_io_fn;
486 /* fuse does it's own writeback accounting */ 485 /* fuse does it's own writeback accounting */
487 fc->bdi.capabilities = BDI_CAP_NO_ACCT_WB; 486 fc->bdi.capabilities = BDI_CAP_NO_ACCT_WB;
488 fc->dev = sb->s_dev; 487 fc->khctr = 0;
489 err = bdi_init(&fc->bdi); 488 fc->polled_files = RB_ROOT;
490 if (err) 489 fc->dev = sb->s_dev;
491 goto error_kfree; 490 err = bdi_init(&fc->bdi);
492 if (sb->s_bdev) { 491 if (err)
493 err = bdi_register(&fc->bdi, NULL, "%u:%u-fuseblk", 492 goto error_mutex_destroy;
494 MAJOR(fc->dev), MINOR(fc->dev)); 493 if (sb->s_bdev) {
495 } else { 494 err = bdi_register(&fc->bdi, NULL, "%u:%u-fuseblk",
496 err = bdi_register_dev(&fc->bdi, fc->dev); 495 MAJOR(fc->dev), MINOR(fc->dev));
497 } 496 } else {
498 if (err) 497 err = bdi_register_dev(&fc->bdi, fc->dev);
499 goto error_bdi_destroy;
500 /*
501 * For a single fuse filesystem use max 1% of dirty +
502 * writeback threshold.
503 *
504 * This gives about 1M of write buffer for memory maps on a
505 * machine with 1G and 10% dirty_ratio, which should be more
506 * than enough.
507 *
508 * Privileged users can raise it by writing to
509 *
510 * /sys/class/bdi/<bdi>/max_ratio
511 */
512 bdi_set_max_ratio(&fc->bdi, 1);
513 fc->reqctr = 0;
514 fc->blocked = 1;
515 fc->attr_version = 1;
516 get_random_bytes(&fc->scramble_key, sizeof(fc->scramble_key));
517 } 498 }
518 return fc; 499 if (err)
500 goto error_bdi_destroy;
501 /*
502 * For a single fuse filesystem use max 1% of dirty +
503 * writeback threshold.
504 *
505 * This gives about 1M of write buffer for memory maps on a
506 * machine with 1G and 10% dirty_ratio, which should be more
507 * than enough.
508 *
509 * Privileged users can raise it by writing to
510 *
511 * /sys/class/bdi/<bdi>/max_ratio
512 */
513 bdi_set_max_ratio(&fc->bdi, 1);
514 fc->reqctr = 0;
515 fc->blocked = 1;
516 fc->attr_version = 1;
517 get_random_bytes(&fc->scramble_key, sizeof(fc->scramble_key));
519 518
520error_bdi_destroy: 519 return 0;
520
521 error_bdi_destroy:
521 bdi_destroy(&fc->bdi); 522 bdi_destroy(&fc->bdi);
522error_kfree: 523 error_mutex_destroy:
523 mutex_destroy(&fc->inst_mutex); 524 mutex_destroy(&fc->inst_mutex);
524 kfree(fc); 525 return err;
525 return NULL;
526} 526}
527EXPORT_SYMBOL_GPL(fuse_conn_init);
527 528
528void fuse_conn_put(struct fuse_conn *fc) 529void fuse_conn_put(struct fuse_conn *fc)
529{ 530{
@@ -532,7 +533,7 @@ void fuse_conn_put(struct fuse_conn *fc)
532 fuse_request_free(fc->destroy_req); 533 fuse_request_free(fc->destroy_req);
533 mutex_destroy(&fc->inst_mutex); 534 mutex_destroy(&fc->inst_mutex);
534 bdi_destroy(&fc->bdi); 535 bdi_destroy(&fc->bdi);
535 kfree(fc); 536 fc->release(fc);
536 } 537 }
537} 538}
538 539
@@ -542,7 +543,7 @@ struct fuse_conn *fuse_conn_get(struct fuse_conn *fc)
542 return fc; 543 return fc;
543} 544}
544 545
545static struct inode *get_root_inode(struct super_block *sb, unsigned mode) 546static struct inode *fuse_get_root_inode(struct super_block *sb, unsigned mode)
546{ 547{
547 struct fuse_attr attr; 548 struct fuse_attr attr;
548 memset(&attr, 0, sizeof(attr)); 549 memset(&attr, 0, sizeof(attr));
@@ -553,8 +554,7 @@ static struct inode *get_root_inode(struct super_block *sb, unsigned mode)
553 return fuse_iget(sb, 1, 0, &attr, 0, 0); 554 return fuse_iget(sb, 1, 0, &attr, 0, 0);
554} 555}
555 556
556struct fuse_inode_handle 557struct fuse_inode_handle {
557{
558 u64 nodeid; 558 u64 nodeid;
559 u32 generation; 559 u32 generation;
560}; 560};
@@ -761,7 +761,6 @@ static void process_init_reply(struct fuse_conn *fc, struct fuse_req *req)
761 fc->max_write = max_t(unsigned, 4096, fc->max_write); 761 fc->max_write = max_t(unsigned, 4096, fc->max_write);
762 fc->conn_init = 1; 762 fc->conn_init = 1;
763 } 763 }
764 fuse_put_request(fc, req);
765 fc->blocked = 0; 764 fc->blocked = 0;
766 wake_up_all(&fc->blocked_waitq); 765 wake_up_all(&fc->blocked_waitq);
767} 766}
@@ -787,7 +786,12 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req)
787 req->out.args[0].size = sizeof(struct fuse_init_out); 786 req->out.args[0].size = sizeof(struct fuse_init_out);
788 req->out.args[0].value = &req->misc.init_out; 787 req->out.args[0].value = &req->misc.init_out;
789 req->end = process_init_reply; 788 req->end = process_init_reply;
790 request_send_background(fc, req); 789 fuse_request_send_background(fc, req);
790}
791
792static void fuse_free_conn(struct fuse_conn *fc)
793{
794 kfree(fc);
791} 795}
792 796
793static int fuse_fill_super(struct super_block *sb, void *data, int silent) 797static int fuse_fill_super(struct super_block *sb, void *data, int silent)
@@ -828,10 +832,17 @@ static int fuse_fill_super(struct super_block *sb, void *data, int silent)
828 if (file->f_op != &fuse_dev_operations) 832 if (file->f_op != &fuse_dev_operations)
829 return -EINVAL; 833 return -EINVAL;
830 834
831 fc = new_conn(sb); 835 fc = kmalloc(sizeof(*fc), GFP_KERNEL);
832 if (!fc) 836 if (!fc)
833 return -ENOMEM; 837 return -ENOMEM;
834 838
839 err = fuse_conn_init(fc, sb);
840 if (err) {
841 kfree(fc);
842 return err;
843 }
844
845 fc->release = fuse_free_conn;
835 fc->flags = d.flags; 846 fc->flags = d.flags;
836 fc->user_id = d.user_id; 847 fc->user_id = d.user_id;
837 fc->group_id = d.group_id; 848 fc->group_id = d.group_id;
@@ -841,7 +852,7 @@ static int fuse_fill_super(struct super_block *sb, void *data, int silent)
841 sb->s_fs_info = fc; 852 sb->s_fs_info = fc;
842 853
843 err = -ENOMEM; 854 err = -ENOMEM;
844 root = get_root_inode(sb, d.rootmode); 855 root = fuse_get_root_inode(sb, d.rootmode);
845 if (!root) 856 if (!root)
846 goto err; 857 goto err;
847 858
@@ -952,7 +963,7 @@ static inline void unregister_fuseblk(void)
952 963
953static void fuse_inode_init_once(void *foo) 964static void fuse_inode_init_once(void *foo)
954{ 965{
955 struct inode * inode = foo; 966 struct inode *inode = foo;
956 967
957 inode_init_once(inode); 968 inode_init_once(inode);
958} 969}
@@ -1031,7 +1042,7 @@ static int __init fuse_init(void)
1031{ 1042{
1032 int res; 1043 int res;
1033 1044
1034 printk("fuse init (API version %i.%i)\n", 1045 printk(KERN_INFO "fuse init (API version %i.%i)\n",
1035 FUSE_KERNEL_VERSION, FUSE_KERNEL_MINOR_VERSION); 1046 FUSE_KERNEL_VERSION, FUSE_KERNEL_MINOR_VERSION);
1036 1047
1037 INIT_LIST_HEAD(&fuse_conn_list); 1048 INIT_LIST_HEAD(&fuse_conn_list);
diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index 0ab0c6f5f438..6903d37af037 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -252,6 +252,7 @@ static ssize_t hugetlbfs_read(struct file *filp, char __user *buf,
252 for (;;) { 252 for (;;) {
253 struct page *page; 253 struct page *page;
254 unsigned long nr, ret; 254 unsigned long nr, ret;
255 int ra;
255 256
256 /* nr is the maximum number of bytes to copy from this page */ 257 /* nr is the maximum number of bytes to copy from this page */
257 nr = huge_page_size(h); 258 nr = huge_page_size(h);
@@ -274,16 +275,19 @@ static ssize_t hugetlbfs_read(struct file *filp, char __user *buf,
274 */ 275 */
275 ret = len < nr ? len : nr; 276 ret = len < nr ? len : nr;
276 if (clear_user(buf, ret)) 277 if (clear_user(buf, ret))
277 ret = -EFAULT; 278 ra = -EFAULT;
279 else
280 ra = 0;
278 } else { 281 } else {
279 /* 282 /*
280 * We have the page, copy it to user space buffer. 283 * We have the page, copy it to user space buffer.
281 */ 284 */
282 ret = hugetlbfs_read_actor(page, offset, buf, len, nr); 285 ra = hugetlbfs_read_actor(page, offset, buf, len, nr);
286 ret = ra;
283 } 287 }
284 if (ret < 0) { 288 if (ra < 0) {
285 if (retval == 0) 289 if (retval == 0)
286 retval = ret; 290 retval = ra;
287 if (page) 291 if (page)
288 page_cache_release(page); 292 page_cache_release(page);
289 goto out; 293 goto out;
diff --git a/fs/inode.c b/fs/inode.c
index bd48e5e6d3e8..7a6e8c2ff7b1 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -110,8 +110,8 @@ static void wake_up_inode(struct inode *inode)
110 110
111/** 111/**
112 * inode_init_always - perform inode structure intialisation 112 * inode_init_always - perform inode structure intialisation
113 * @sb - superblock inode belongs to. 113 * @sb: superblock inode belongs to
114 * @inode - inode to initialise 114 * @inode: inode to initialise
115 * 115 *
116 * These are initializations that need to be done on every inode 116 * These are initializations that need to be done on every inode
117 * allocation as the fields are not initialised by slab allocation. 117 * allocation as the fields are not initialised by slab allocation.
@@ -166,7 +166,7 @@ struct inode *inode_init_always(struct super_block *sb, struct inode *inode)
166 mapping->a_ops = &empty_aops; 166 mapping->a_ops = &empty_aops;
167 mapping->host = inode; 167 mapping->host = inode;
168 mapping->flags = 0; 168 mapping->flags = 0;
169 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_PAGECACHE); 169 mapping_set_gfp_mask(mapping, GFP_HIGHUSER_MOVABLE);
170 mapping->assoc_mapping = NULL; 170 mapping->assoc_mapping = NULL;
171 mapping->backing_dev_info = &default_backing_dev_info; 171 mapping->backing_dev_info = &default_backing_dev_info;
172 mapping->writeback_index = 0; 172 mapping->writeback_index = 0;
@@ -576,8 +576,8 @@ __inode_add_to_lists(struct super_block *sb, struct hlist_head *head,
576 576
577/** 577/**
578 * inode_add_to_lists - add a new inode to relevant lists 578 * inode_add_to_lists - add a new inode to relevant lists
579 * @sb - superblock inode belongs to. 579 * @sb: superblock inode belongs to
580 * @inode - inode to mark in use 580 * @inode: inode to mark in use
581 * 581 *
582 * When an inode is allocated it needs to be accounted for, added to the in use 582 * When an inode is allocated it needs to be accounted for, added to the in use
583 * list, the owning superblock and the inode hash. This needs to be done under 583 * list, the owning superblock and the inode hash. This needs to be done under
@@ -601,7 +601,7 @@ EXPORT_SYMBOL_GPL(inode_add_to_lists);
601 * @sb: superblock 601 * @sb: superblock
602 * 602 *
603 * Allocates a new inode for given superblock. The default gfp_mask 603 * Allocates a new inode for given superblock. The default gfp_mask
604 * for allocations related to inode->i_mapping is GFP_HIGHUSER_PAGECACHE. 604 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
605 * If HIGHMEM pages are unsuitable or it is known that pages allocated 605 * If HIGHMEM pages are unsuitable or it is known that pages allocated
606 * for the page cache are not reclaimable or migratable, 606 * for the page cache are not reclaimable or migratable,
607 * mapping_set_gfp_mask() must be called with suitable flags on the 607 * mapping_set_gfp_mask() must be called with suitable flags on the
diff --git a/fs/minix/dir.c b/fs/minix/dir.c
index f70433816a38..d4946c4c90e2 100644
--- a/fs/minix/dir.c
+++ b/fs/minix/dir.c
@@ -280,7 +280,7 @@ int minix_add_link(struct dentry *dentry, struct inode *inode)
280 return -EINVAL; 280 return -EINVAL;
281 281
282got_it: 282got_it:
283 pos = (page->index >> PAGE_CACHE_SHIFT) + p - (char*)page_address(page); 283 pos = page_offset(page) + p - (char *)page_address(page);
284 err = __minix_write_begin(NULL, page->mapping, pos, sbi->s_dirsize, 284 err = __minix_write_begin(NULL, page->mapping, pos, sbi->s_dirsize,
285 AOP_FLAG_UNINTERRUPTIBLE, &page, NULL); 285 AOP_FLAG_UNINTERRUPTIBLE, &page, NULL);
286 if (err) 286 if (err)
diff --git a/fs/mpage.c b/fs/mpage.c
index 552b80b3facc..16c3ef37eae3 100644
--- a/fs/mpage.c
+++ b/fs/mpage.c
@@ -241,7 +241,6 @@ do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
241 first_hole = page_block; 241 first_hole = page_block;
242 page_block++; 242 page_block++;
243 block_in_file++; 243 block_in_file++;
244 clear_buffer_mapped(map_bh);
245 continue; 244 continue;
246 } 245 }
247 246
@@ -308,7 +307,10 @@ alloc_new:
308 goto alloc_new; 307 goto alloc_new;
309 } 308 }
310 309
311 if (buffer_boundary(map_bh) || (first_hole != blocks_per_page)) 310 relative_block = block_in_file - *first_logical_block;
311 nblocks = map_bh->b_size >> blkbits;
312 if ((buffer_boundary(map_bh) && relative_block == nblocks) ||
313 (first_hole != blocks_per_page))
312 bio = mpage_bio_submit(READ, bio); 314 bio = mpage_bio_submit(READ, bio);
313 else 315 else
314 *last_block_in_bio = blocks[blocks_per_page - 1]; 316 *last_block_in_bio = blocks[blocks_per_page - 1];
diff --git a/fs/ncpfs/getopt.c b/fs/ncpfs/getopt.c
index 335b003dddf9..0af3349de851 100644
--- a/fs/ncpfs/getopt.c
+++ b/fs/ncpfs/getopt.c
@@ -16,7 +16,6 @@
16 * @opts: an array of &struct option entries controlling parser operations 16 * @opts: an array of &struct option entries controlling parser operations
17 * @optopt: output; will contain the current option 17 * @optopt: output; will contain the current option
18 * @optarg: output; will contain the value (if one exists) 18 * @optarg: output; will contain the value (if one exists)
19 * @flag: output; may be NULL; should point to a long for or'ing flags
20 * @value: output; may be NULL; will be overwritten with the integer value 19 * @value: output; may be NULL; will be overwritten with the integer value
21 * of the current argument. 20 * of the current argument.
22 * 21 *
diff --git a/fs/partitions/check.c b/fs/partitions/check.c
index 6d5b213b8a9b..5198ada67398 100644
--- a/fs/partitions/check.c
+++ b/fs/partitions/check.c
@@ -384,9 +384,9 @@ struct hd_struct *add_partition(struct gendisk *disk, int partno,
384 384
385 dname = dev_name(ddev); 385 dname = dev_name(ddev);
386 if (isdigit(dname[strlen(dname) - 1])) 386 if (isdigit(dname[strlen(dname) - 1]))
387 snprintf(pdev->bus_id, BUS_ID_SIZE, "%sp%d", dname, partno); 387 dev_set_name(pdev, "%sp%d", dname, partno);
388 else 388 else
389 snprintf(pdev->bus_id, BUS_ID_SIZE, "%s%d", dname, partno); 389 dev_set_name(pdev, "%s%d", dname, partno);
390 390
391 device_initialize(pdev); 391 device_initialize(pdev);
392 pdev->class = &block_class; 392 pdev->class = &block_class;
@@ -447,16 +447,11 @@ void register_disk(struct gendisk *disk)
447 struct block_device *bdev; 447 struct block_device *bdev;
448 struct disk_part_iter piter; 448 struct disk_part_iter piter;
449 struct hd_struct *part; 449 struct hd_struct *part;
450 char *s;
451 int err; 450 int err;
452 451
453 ddev->parent = disk->driverfs_dev; 452 ddev->parent = disk->driverfs_dev;
454 453
455 strlcpy(ddev->bus_id, disk->disk_name, BUS_ID_SIZE); 454 dev_set_name(ddev, disk->disk_name);
456 /* ewww... some of these buggers have / in the name... */
457 s = strchr(ddev->bus_id, '/');
458 if (s)
459 *s = '!';
460 455
461 /* delay uevents, until we scanned partition table */ 456 /* delay uevents, until we scanned partition table */
462 ddev->uevent_suppress = 1; 457 ddev->uevent_suppress = 1;
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 3a8bdd7f5756..94063840832a 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -396,7 +396,9 @@ static int show_smap(struct seq_file *m, void *v)
396 "Private_Clean: %8lu kB\n" 396 "Private_Clean: %8lu kB\n"
397 "Private_Dirty: %8lu kB\n" 397 "Private_Dirty: %8lu kB\n"
398 "Referenced: %8lu kB\n" 398 "Referenced: %8lu kB\n"
399 "Swap: %8lu kB\n", 399 "Swap: %8lu kB\n"
400 "KernelPageSize: %8lu kB\n"
401 "MMUPageSize: %8lu kB\n",
400 (vma->vm_end - vma->vm_start) >> 10, 402 (vma->vm_end - vma->vm_start) >> 10,
401 mss.resident >> 10, 403 mss.resident >> 10,
402 (unsigned long)(mss.pss >> (10 + PSS_SHIFT)), 404 (unsigned long)(mss.pss >> (10 + PSS_SHIFT)),
@@ -405,7 +407,9 @@ static int show_smap(struct seq_file *m, void *v)
405 mss.private_clean >> 10, 407 mss.private_clean >> 10,
406 mss.private_dirty >> 10, 408 mss.private_dirty >> 10,
407 mss.referenced >> 10, 409 mss.referenced >> 10,
408 mss.swap >> 10); 410 mss.swap >> 10,
411 vma_kernel_pagesize(vma) >> 10,
412 vma_mmu_pagesize(vma) >> 10);
409 413
410 if (m->count < m->size) /* vma is copied successfully */ 414 if (m->count < m->size) /* vma is copied successfully */
411 m->version = (vma != get_gate_vma(task)) ? vma->vm_start : 0; 415 m->version = (vma != get_gate_vma(task)) ? vma->vm_start : 0;
diff --git a/fs/select.c b/fs/select.c
index 87df51eadcf2..08b91beed806 100644
--- a/fs/select.c
+++ b/fs/select.c
@@ -109,11 +109,11 @@ static void __pollwait(struct file *filp, wait_queue_head_t *wait_address,
109void poll_initwait(struct poll_wqueues *pwq) 109void poll_initwait(struct poll_wqueues *pwq)
110{ 110{
111 init_poll_funcptr(&pwq->pt, __pollwait); 111 init_poll_funcptr(&pwq->pt, __pollwait);
112 pwq->polling_task = current;
112 pwq->error = 0; 113 pwq->error = 0;
113 pwq->table = NULL; 114 pwq->table = NULL;
114 pwq->inline_index = 0; 115 pwq->inline_index = 0;
115} 116}
116
117EXPORT_SYMBOL(poll_initwait); 117EXPORT_SYMBOL(poll_initwait);
118 118
119static void free_poll_entry(struct poll_table_entry *entry) 119static void free_poll_entry(struct poll_table_entry *entry)
@@ -142,12 +142,10 @@ void poll_freewait(struct poll_wqueues *pwq)
142 free_page((unsigned long) old); 142 free_page((unsigned long) old);
143 } 143 }
144} 144}
145
146EXPORT_SYMBOL(poll_freewait); 145EXPORT_SYMBOL(poll_freewait);
147 146
148static struct poll_table_entry *poll_get_entry(poll_table *_p) 147static struct poll_table_entry *poll_get_entry(struct poll_wqueues *p)
149{ 148{
150 struct poll_wqueues *p = container_of(_p, struct poll_wqueues, pt);
151 struct poll_table_page *table = p->table; 149 struct poll_table_page *table = p->table;
152 150
153 if (p->inline_index < N_INLINE_POLL_ENTRIES) 151 if (p->inline_index < N_INLINE_POLL_ENTRIES)
@@ -159,7 +157,6 @@ static struct poll_table_entry *poll_get_entry(poll_table *_p)
159 new_table = (struct poll_table_page *) __get_free_page(GFP_KERNEL); 157 new_table = (struct poll_table_page *) __get_free_page(GFP_KERNEL);
160 if (!new_table) { 158 if (!new_table) {
161 p->error = -ENOMEM; 159 p->error = -ENOMEM;
162 __set_current_state(TASK_RUNNING);
163 return NULL; 160 return NULL;
164 } 161 }
165 new_table->entry = new_table->entries; 162 new_table->entry = new_table->entries;
@@ -171,20 +168,75 @@ static struct poll_table_entry *poll_get_entry(poll_table *_p)
171 return table->entry++; 168 return table->entry++;
172} 169}
173 170
171static int pollwake(wait_queue_t *wait, unsigned mode, int sync, void *key)
172{
173 struct poll_wqueues *pwq = wait->private;
174 DECLARE_WAITQUEUE(dummy_wait, pwq->polling_task);
175
176 /*
177 * Although this function is called under waitqueue lock, LOCK
178 * doesn't imply write barrier and the users expect write
179 * barrier semantics on wakeup functions. The following
180 * smp_wmb() is equivalent to smp_wmb() in try_to_wake_up()
181 * and is paired with set_mb() in poll_schedule_timeout.
182 */
183 smp_wmb();
184 pwq->triggered = 1;
185
186 /*
187 * Perform the default wake up operation using a dummy
188 * waitqueue.
189 *
190 * TODO: This is hacky but there currently is no interface to
191 * pass in @sync. @sync is scheduled to be removed and once
192 * that happens, wake_up_process() can be used directly.
193 */
194 return default_wake_function(&dummy_wait, mode, sync, key);
195}
196
174/* Add a new entry */ 197/* Add a new entry */
175static void __pollwait(struct file *filp, wait_queue_head_t *wait_address, 198static void __pollwait(struct file *filp, wait_queue_head_t *wait_address,
176 poll_table *p) 199 poll_table *p)
177{ 200{
178 struct poll_table_entry *entry = poll_get_entry(p); 201 struct poll_wqueues *pwq = container_of(p, struct poll_wqueues, pt);
202 struct poll_table_entry *entry = poll_get_entry(pwq);
179 if (!entry) 203 if (!entry)
180 return; 204 return;
181 get_file(filp); 205 get_file(filp);
182 entry->filp = filp; 206 entry->filp = filp;
183 entry->wait_address = wait_address; 207 entry->wait_address = wait_address;
184 init_waitqueue_entry(&entry->wait, current); 208 init_waitqueue_func_entry(&entry->wait, pollwake);
209 entry->wait.private = pwq;
185 add_wait_queue(wait_address, &entry->wait); 210 add_wait_queue(wait_address, &entry->wait);
186} 211}
187 212
213int poll_schedule_timeout(struct poll_wqueues *pwq, int state,
214 ktime_t *expires, unsigned long slack)
215{
216 int rc = -EINTR;
217
218 set_current_state(state);
219 if (!pwq->triggered)
220 rc = schedule_hrtimeout_range(expires, slack, HRTIMER_MODE_ABS);
221 __set_current_state(TASK_RUNNING);
222
223 /*
224 * Prepare for the next iteration.
225 *
226 * The following set_mb() serves two purposes. First, it's
227 * the counterpart rmb of the wmb in pollwake() such that data
228 * written before wake up is always visible after wake up.
229 * Second, the full barrier guarantees that triggered clearing
230 * doesn't pass event check of the next iteration. Note that
231 * this problem doesn't exist for the first iteration as
232 * add_wait_queue() has full barrier semantics.
233 */
234 set_mb(pwq->triggered, 0);
235
236 return rc;
237}
238EXPORT_SYMBOL(poll_schedule_timeout);
239
188/** 240/**
189 * poll_select_set_timeout - helper function to setup the timeout value 241 * poll_select_set_timeout - helper function to setup the timeout value
190 * @to: pointer to timespec variable for the final timeout 242 * @to: pointer to timespec variable for the final timeout
@@ -340,8 +392,6 @@ int do_select(int n, fd_set_bits *fds, struct timespec *end_time)
340 for (;;) { 392 for (;;) {
341 unsigned long *rinp, *routp, *rexp, *inp, *outp, *exp; 393 unsigned long *rinp, *routp, *rexp, *inp, *outp, *exp;
342 394
343 set_current_state(TASK_INTERRUPTIBLE);
344
345 inp = fds->in; outp = fds->out; exp = fds->ex; 395 inp = fds->in; outp = fds->out; exp = fds->ex;
346 rinp = fds->res_in; routp = fds->res_out; rexp = fds->res_ex; 396 rinp = fds->res_in; routp = fds->res_out; rexp = fds->res_ex;
347 397
@@ -411,10 +461,10 @@ int do_select(int n, fd_set_bits *fds, struct timespec *end_time)
411 to = &expire; 461 to = &expire;
412 } 462 }
413 463
414 if (!schedule_hrtimeout_range(to, slack, HRTIMER_MODE_ABS)) 464 if (!poll_schedule_timeout(&table, TASK_INTERRUPTIBLE,
465 to, slack))
415 timed_out = 1; 466 timed_out = 1;
416 } 467 }
417 __set_current_state(TASK_RUNNING);
418 468
419 poll_freewait(&table); 469 poll_freewait(&table);
420 470
@@ -666,7 +716,6 @@ static int do_poll(unsigned int nfds, struct poll_list *list,
666 for (;;) { 716 for (;;) {
667 struct poll_list *walk; 717 struct poll_list *walk;
668 718
669 set_current_state(TASK_INTERRUPTIBLE);
670 for (walk = list; walk != NULL; walk = walk->next) { 719 for (walk = list; walk != NULL; walk = walk->next) {
671 struct pollfd * pfd, * pfd_end; 720 struct pollfd * pfd, * pfd_end;
672 721
@@ -709,10 +758,9 @@ static int do_poll(unsigned int nfds, struct poll_list *list,
709 to = &expire; 758 to = &expire;
710 } 759 }
711 760
712 if (!schedule_hrtimeout_range(to, slack, HRTIMER_MODE_ABS)) 761 if (!poll_schedule_timeout(wait, TASK_INTERRUPTIBLE, to, slack))
713 timed_out = 1; 762 timed_out = 1;
714 } 763 }
715 __set_current_state(TASK_RUNNING);
716 return count; 764 return count;
717} 765}
718 766
diff --git a/fs/sync.c b/fs/sync.c
index 0921d6d4b5e6..ac02b56548bc 100644
--- a/fs/sync.c
+++ b/fs/sync.c
@@ -295,7 +295,7 @@ int do_sync_mapping_range(struct address_space *mapping, loff_t offset,
295 295
296 if (flags & SYNC_FILE_RANGE_WRITE) { 296 if (flags & SYNC_FILE_RANGE_WRITE) {
297 ret = __filemap_fdatawrite_range(mapping, offset, endbyte, 297 ret = __filemap_fdatawrite_range(mapping, offset, endbyte,
298 WB_SYNC_NONE); 298 WB_SYNC_ALL);
299 if (ret < 0) 299 if (ret < 0)
300 goto out; 300 goto out;
301 } 301 }
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 0d7564b95f8e..89556ee72518 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -432,12 +432,19 @@ static int ubifs_sync_fs(struct super_block *sb, int wait)
432 int i, err; 432 int i, err;
433 struct ubifs_info *c = sb->s_fs_info; 433 struct ubifs_info *c = sb->s_fs_info;
434 struct writeback_control wbc = { 434 struct writeback_control wbc = {
435 .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_HOLD, 435 .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_NONE,
436 .range_start = 0, 436 .range_start = 0,
437 .range_end = LLONG_MAX, 437 .range_end = LLONG_MAX,
438 .nr_to_write = LONG_MAX, 438 .nr_to_write = LONG_MAX,
439 }; 439 };
440 440
441 /*
442 * Note by akpm about WB_SYNC_NONE used above: zero @wait is just an
443 * advisory thing to help the file system shove lots of data into the
444 * queues. If some gets missed then it'll be picked up on the second
445 * '->sync_fs()' call, with non-zero @wait.
446 */
447
441 if (sb->s_flags & MS_RDONLY) 448 if (sb->s_flags & MS_RDONLY)
442 return 0; 449 return 0;
443 450
generated by cgit v1.2.2 (git 2.25.0) at 2025-06-07 08:15:18 -0400