aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
Diffstat (limited to 'fs')
-rw-r--r--fs/bio-integrity.c7
-rw-r--r--fs/block_dev.c741
-rw-r--r--fs/btrfs/volumes.c28
-rw-r--r--fs/btrfs/volumes.h2
-rw-r--r--fs/char_dev.c2
-rw-r--r--fs/ext3/super.c12
-rw-r--r--fs/ext4/super.c12
-rw-r--r--fs/gfs2/ops_fstype.c8
-rw-r--r--fs/jfs/jfs_logmgr.c17
-rw-r--r--fs/logfs/dev_bdev.c7
-rw-r--r--fs/nfsd/vfs.c5
-rw-r--r--fs/nilfs2/super.c8
-rw-r--r--fs/ocfs2/cluster/heartbeat.c2
-rw-r--r--fs/partitions/check.c106
-rw-r--r--fs/reiserfs/journal.c21
-rw-r--r--fs/splice.c43
-rw-r--r--fs/super.c19
-rw-r--r--fs/xfs/linux-2.6/xfs_super.c5
18 files changed, 347 insertions, 698 deletions
diff --git a/fs/bio-integrity.c b/fs/bio-integrity.c
index 4d0ff5ee27b8..e49cce234c65 100644
--- a/fs/bio-integrity.c
+++ b/fs/bio-integrity.c
@@ -782,7 +782,12 @@ void __init bio_integrity_init(void)
782{ 782{
783 unsigned int i; 783 unsigned int i;
784 784
785 kintegrityd_wq = create_workqueue("kintegrityd"); 785 /*
786 * kintegrityd won't block much but may burn a lot of CPU cycles.
787 * Make it highpri CPU intensive wq with max concurrency of 1.
788 */
789 kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
790 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
786 if (!kintegrityd_wq) 791 if (!kintegrityd_wq)
787 panic("Failed to create kintegrityd\n"); 792 panic("Failed to create kintegrityd\n");
788 793
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 88da70355aa3..fe3f59c14a02 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -432,9 +432,6 @@ static void init_once(void *foo)
432 mutex_init(&bdev->bd_mutex); 432 mutex_init(&bdev->bd_mutex);
433 INIT_LIST_HEAD(&bdev->bd_inodes); 433 INIT_LIST_HEAD(&bdev->bd_inodes);
434 INIT_LIST_HEAD(&bdev->bd_list); 434 INIT_LIST_HEAD(&bdev->bd_list);
435#ifdef CONFIG_SYSFS
436 INIT_LIST_HEAD(&bdev->bd_holder_list);
437#endif
438 inode_init_once(&ei->vfs_inode); 435 inode_init_once(&ei->vfs_inode);
439 /* Initialize mutex for freeze. */ 436 /* Initialize mutex for freeze. */
440 mutex_init(&bdev->bd_fsfreeze_mutex); 437 mutex_init(&bdev->bd_fsfreeze_mutex);
@@ -669,7 +666,7 @@ static bool bd_may_claim(struct block_device *bdev, struct block_device *whole,
669 else if (bdev->bd_contains == bdev) 666 else if (bdev->bd_contains == bdev)
670 return true; /* is a whole device which isn't held */ 667 return true; /* is a whole device which isn't held */
671 668
672 else if (whole->bd_holder == bd_claim) 669 else if (whole->bd_holder == bd_may_claim)
673 return true; /* is a partition of a device that is being partitioned */ 670 return true; /* is a partition of a device that is being partitioned */
674 else if (whole->bd_holder != NULL) 671 else if (whole->bd_holder != NULL)
675 return false; /* is a partition of a held device */ 672 return false; /* is a partition of a held device */
@@ -781,439 +778,87 @@ static struct block_device *bd_start_claiming(struct block_device *bdev,
781 } 778 }
782} 779}
783 780
784/* releases bdev_lock */
785static void __bd_abort_claiming(struct block_device *whole, void *holder)
786{
787 BUG_ON(whole->bd_claiming != holder);
788 whole->bd_claiming = NULL;
789 wake_up_bit(&whole->bd_claiming, 0);
790
791 spin_unlock(&bdev_lock);
792 bdput(whole);
793}
794
795/**
796 * bd_abort_claiming - abort claiming a block device
797 * @whole: whole block device returned by bd_start_claiming()
798 * @holder: holder trying to claim @bdev
799 *
800 * Abort a claiming block started by bd_start_claiming(). Note that
801 * @whole is not the block device to be claimed but the whole device
802 * returned by bd_start_claiming().
803 *
804 * CONTEXT:
805 * Grabs and releases bdev_lock.
806 */
807static void bd_abort_claiming(struct block_device *whole, void *holder)
808{
809 spin_lock(&bdev_lock);
810 __bd_abort_claiming(whole, holder); /* releases bdev_lock */
811}
812
813/* increment holders when we have a legitimate claim. requires bdev_lock */
814static void __bd_claim(struct block_device *bdev, struct block_device *whole,
815 void *holder)
816{
817 /* note that for a whole device bd_holders
818 * will be incremented twice, and bd_holder will
819 * be set to bd_claim before being set to holder
820 */
821 whole->bd_holders++;
822 whole->bd_holder = bd_claim;
823 bdev->bd_holders++;
824 bdev->bd_holder = holder;
825}
826
827/**
828 * bd_finish_claiming - finish claiming a block device
829 * @bdev: block device of interest (passed to bd_start_claiming())
830 * @whole: whole block device returned by bd_start_claiming()
831 * @holder: holder trying to claim @bdev
832 *
833 * Finish a claiming block started by bd_start_claiming().
834 *
835 * CONTEXT:
836 * Grabs and releases bdev_lock.
837 */
838static void bd_finish_claiming(struct block_device *bdev,
839 struct block_device *whole, void *holder)
840{
841 spin_lock(&bdev_lock);
842 BUG_ON(!bd_may_claim(bdev, whole, holder));
843 __bd_claim(bdev, whole, holder);
844 __bd_abort_claiming(whole, holder); /* not actually an abort */
845}
846
847/**
848 * bd_claim - claim a block device
849 * @bdev: block device to claim
850 * @holder: holder trying to claim @bdev
851 *
852 * Try to claim @bdev which must have been opened successfully.
853 *
854 * CONTEXT:
855 * Might sleep.
856 *
857 * RETURNS:
858 * 0 if successful, -EBUSY if @bdev is already claimed.
859 */
860int bd_claim(struct block_device *bdev, void *holder)
861{
862 struct block_device *whole = bdev->bd_contains;
863 int res;
864
865 might_sleep();
866
867 spin_lock(&bdev_lock);
868 res = bd_prepare_to_claim(bdev, whole, holder);
869 if (res == 0)
870 __bd_claim(bdev, whole, holder);
871 spin_unlock(&bdev_lock);
872
873 return res;
874}
875EXPORT_SYMBOL(bd_claim);
876
877void bd_release(struct block_device *bdev)
878{
879 spin_lock(&bdev_lock);
880 if (!--bdev->bd_contains->bd_holders)
881 bdev->bd_contains->bd_holder = NULL;
882 if (!--bdev->bd_holders)
883 bdev->bd_holder = NULL;
884 spin_unlock(&bdev_lock);
885}
886
887EXPORT_SYMBOL(bd_release);
888
889#ifdef CONFIG_SYSFS 781#ifdef CONFIG_SYSFS
890/*
891 * Functions for bd_claim_by_kobject / bd_release_from_kobject
892 *
893 * If a kobject is passed to bd_claim_by_kobject()
894 * and the kobject has a parent directory,
895 * following symlinks are created:
896 * o from the kobject to the claimed bdev
897 * o from "holders" directory of the bdev to the parent of the kobject
898 * bd_release_from_kobject() removes these symlinks.
899 *
900 * Example:
901 * If /dev/dm-0 maps to /dev/sda, kobject corresponding to
902 * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then:
903 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
904 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
905 */
906
907static int add_symlink(struct kobject *from, struct kobject *to) 782static int add_symlink(struct kobject *from, struct kobject *to)
908{ 783{
909 if (!from || !to)
910 return 0;
911 return sysfs_create_link(from, to, kobject_name(to)); 784 return sysfs_create_link(from, to, kobject_name(to));
912} 785}
913 786
914static void del_symlink(struct kobject *from, struct kobject *to) 787static void del_symlink(struct kobject *from, struct kobject *to)
915{ 788{
916 if (!from || !to)
917 return;
918 sysfs_remove_link(from, kobject_name(to)); 789 sysfs_remove_link(from, kobject_name(to));
919} 790}
920 791
921/*
922 * 'struct bd_holder' contains pointers to kobjects symlinked by
923 * bd_claim_by_kobject.
924 * It's connected to bd_holder_list which is protected by bdev->bd_sem.
925 */
926struct bd_holder {
927 struct list_head list; /* chain of holders of the bdev */
928 int count; /* references from the holder */
929 struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */
930 struct kobject *hdev; /* e.g. "/block/dm-0" */
931 struct kobject *hdir; /* e.g. "/block/sda/holders" */
932 struct kobject *sdev; /* e.g. "/block/sda" */
933};
934
935/*
936 * Get references of related kobjects at once.
937 * Returns 1 on success. 0 on failure.
938 *
939 * Should call bd_holder_release_dirs() after successful use.
940 */
941static int bd_holder_grab_dirs(struct block_device *bdev,
942 struct bd_holder *bo)
943{
944 if (!bdev || !bo)
945 return 0;
946
947 bo->sdir = kobject_get(bo->sdir);
948 if (!bo->sdir)
949 return 0;
950
951 bo->hdev = kobject_get(bo->sdir->parent);
952 if (!bo->hdev)
953 goto fail_put_sdir;
954
955 bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj);
956 if (!bo->sdev)
957 goto fail_put_hdev;
958
959 bo->hdir = kobject_get(bdev->bd_part->holder_dir);
960 if (!bo->hdir)
961 goto fail_put_sdev;
962
963 return 1;
964
965fail_put_sdev:
966 kobject_put(bo->sdev);
967fail_put_hdev:
968 kobject_put(bo->hdev);
969fail_put_sdir:
970 kobject_put(bo->sdir);
971
972 return 0;
973}
974
975/* Put references of related kobjects at once. */
976static void bd_holder_release_dirs(struct bd_holder *bo)
977{
978 kobject_put(bo->hdir);
979 kobject_put(bo->sdev);
980 kobject_put(bo->hdev);
981 kobject_put(bo->sdir);
982}
983
984static struct bd_holder *alloc_bd_holder(struct kobject *kobj)
985{
986 struct bd_holder *bo;
987
988 bo = kzalloc(sizeof(*bo), GFP_KERNEL);
989 if (!bo)
990 return NULL;
991
992 bo->count = 1;
993 bo->sdir = kobj;
994
995 return bo;
996}
997
998static void free_bd_holder(struct bd_holder *bo)
999{
1000 kfree(bo);
1001}
1002
1003/** 792/**
1004 * find_bd_holder - find matching struct bd_holder from the block device 793 * bd_link_disk_holder - create symlinks between holding disk and slave bdev
794 * @bdev: the claimed slave bdev
795 * @disk: the holding disk
1005 * 796 *
1006 * @bdev: struct block device to be searched 797 * This functions creates the following sysfs symlinks.
1007 * @bo: target struct bd_holder
1008 *
1009 * Returns matching entry with @bo in @bdev->bd_holder_list.
1010 * If found, increment the reference count and return the pointer.
1011 * If not found, returns NULL.
1012 */
1013static struct bd_holder *find_bd_holder(struct block_device *bdev,
1014 struct bd_holder *bo)
1015{
1016 struct bd_holder *tmp;
1017
1018 list_for_each_entry(tmp, &bdev->bd_holder_list, list)
1019 if (tmp->sdir == bo->sdir) {
1020 tmp->count++;
1021 return tmp;
1022 }
1023
1024 return NULL;
1025}
1026
1027/**
1028 * add_bd_holder - create sysfs symlinks for bd_claim() relationship
1029 *
1030 * @bdev: block device to be bd_claimed
1031 * @bo: preallocated and initialized by alloc_bd_holder()
1032 *
1033 * Add @bo to @bdev->bd_holder_list, create symlinks.
1034 *
1035 * Returns 0 if symlinks are created.
1036 * Returns -ve if something fails.
1037 */
1038static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo)
1039{
1040 int err;
1041
1042 if (!bo)
1043 return -EINVAL;
1044
1045 if (!bd_holder_grab_dirs(bdev, bo))
1046 return -EBUSY;
1047
1048 err = add_symlink(bo->sdir, bo->sdev);
1049 if (err)
1050 return err;
1051
1052 err = add_symlink(bo->hdir, bo->hdev);
1053 if (err) {
1054 del_symlink(bo->sdir, bo->sdev);
1055 return err;
1056 }
1057
1058 list_add_tail(&bo->list, &bdev->bd_holder_list);
1059 return 0;
1060}
1061
1062/**
1063 * del_bd_holder - delete sysfs symlinks for bd_claim() relationship
1064 * 798 *
1065 * @bdev: block device to be bd_claimed 799 * - from "slaves" directory of the holder @disk to the claimed @bdev
1066 * @kobj: holder's kobject 800 * - from "holders" directory of the @bdev to the holder @disk
1067 * 801 *
1068 * If there is matching entry with @kobj in @bdev->bd_holder_list 802 * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
1069 * and no other bd_claim() from the same kobject, 803 * passed to bd_link_disk_holder(), then:
1070 * remove the struct bd_holder from the list, delete symlinks for it.
1071 * 804 *
1072 * Returns a pointer to the struct bd_holder when it's removed from the list 805 * /sys/block/dm-0/slaves/sda --> /sys/block/sda
1073 * and ready to be freed. 806 * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
1074 * Returns NULL if matching claim isn't found or there is other bd_claim()
1075 * by the same kobject.
1076 */
1077static struct bd_holder *del_bd_holder(struct block_device *bdev,
1078 struct kobject *kobj)
1079{
1080 struct bd_holder *bo;
1081
1082 list_for_each_entry(bo, &bdev->bd_holder_list, list) {
1083 if (bo->sdir == kobj) {
1084 bo->count--;
1085 BUG_ON(bo->count < 0);
1086 if (!bo->count) {
1087 list_del(&bo->list);
1088 del_symlink(bo->sdir, bo->sdev);
1089 del_symlink(bo->hdir, bo->hdev);
1090 bd_holder_release_dirs(bo);
1091 return bo;
1092 }
1093 break;
1094 }
1095 }
1096
1097 return NULL;
1098}
1099
1100/**
1101 * bd_claim_by_kobject - bd_claim() with additional kobject signature
1102 * 807 *
1103 * @bdev: block device to be claimed 808 * The caller must have claimed @bdev before calling this function and
1104 * @holder: holder's signature 809 * ensure that both @bdev and @disk are valid during the creation and
1105 * @kobj: holder's kobject 810 * lifetime of these symlinks.
1106 * 811 *
1107 * Do bd_claim() and if it succeeds, create sysfs symlinks between 812 * CONTEXT:
1108 * the bdev and the holder's kobject. 813 * Might sleep.
1109 * Use bd_release_from_kobject() when relesing the claimed bdev.
1110 * 814 *
1111 * Returns 0 on success. (same as bd_claim()) 815 * RETURNS:
1112 * Returns errno on failure. 816 * 0 on success, -errno on failure.
1113 */ 817 */
1114static int bd_claim_by_kobject(struct block_device *bdev, void *holder, 818int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk)
1115 struct kobject *kobj)
1116{ 819{
1117 int err; 820 int ret = 0;
1118 struct bd_holder *bo, *found;
1119
1120 if (!kobj)
1121 return -EINVAL;
1122
1123 bo = alloc_bd_holder(kobj);
1124 if (!bo)
1125 return -ENOMEM;
1126 821
1127 mutex_lock(&bdev->bd_mutex); 822 mutex_lock(&bdev->bd_mutex);
1128 823
1129 err = bd_claim(bdev, holder); 824 WARN_ON_ONCE(!bdev->bd_holder || bdev->bd_holder_disk);
1130 if (err)
1131 goto fail;
1132 825
1133 found = find_bd_holder(bdev, bo); 826 /* FIXME: remove the following once add_disk() handles errors */
1134 if (found) 827 if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir))
1135 goto fail; 828 goto out_unlock;
1136 829
1137 err = add_bd_holder(bdev, bo); 830 ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
1138 if (err) 831 if (ret)
1139 bd_release(bdev); 832 goto out_unlock;
1140 else
1141 bo = NULL;
1142fail:
1143 mutex_unlock(&bdev->bd_mutex);
1144 free_bd_holder(bo);
1145 return err;
1146}
1147 833
1148/** 834 ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj);
1149 * bd_release_from_kobject - bd_release() with additional kobject signature 835 if (ret) {
1150 * 836 del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
1151 * @bdev: block device to be released 837 goto out_unlock;
1152 * @kobj: holder's kobject 838 }
1153 *
1154 * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject().
1155 */
1156static void bd_release_from_kobject(struct block_device *bdev,
1157 struct kobject *kobj)
1158{
1159 if (!kobj)
1160 return;
1161 839
1162 mutex_lock(&bdev->bd_mutex); 840 bdev->bd_holder_disk = disk;
1163 bd_release(bdev); 841out_unlock:
1164 free_bd_holder(del_bd_holder(bdev, kobj));
1165 mutex_unlock(&bdev->bd_mutex); 842 mutex_unlock(&bdev->bd_mutex);
843 return ret;
1166} 844}
845EXPORT_SYMBOL_GPL(bd_link_disk_holder);
1167 846
1168/** 847static void bd_unlink_disk_holder(struct block_device *bdev)
1169 * bd_claim_by_disk - wrapper function for bd_claim_by_kobject()
1170 *
1171 * @bdev: block device to be claimed
1172 * @holder: holder's signature
1173 * @disk: holder's gendisk
1174 *
1175 * Call bd_claim_by_kobject() with getting @disk->slave_dir.
1176 */
1177int bd_claim_by_disk(struct block_device *bdev, void *holder,
1178 struct gendisk *disk)
1179{ 848{
1180 return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); 849 struct gendisk *disk = bdev->bd_holder_disk;
1181}
1182EXPORT_SYMBOL_GPL(bd_claim_by_disk);
1183 850
1184/** 851 bdev->bd_holder_disk = NULL;
1185 * bd_release_from_disk - wrapper function for bd_release_from_kobject() 852 if (!disk)
1186 * 853 return;
1187 * @bdev: block device to be claimed
1188 * @disk: holder's gendisk
1189 *
1190 * Call bd_release_from_kobject() and put @disk->slave_dir.
1191 */
1192void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk)
1193{
1194 bd_release_from_kobject(bdev, disk->slave_dir);
1195 kobject_put(disk->slave_dir);
1196}
1197EXPORT_SYMBOL_GPL(bd_release_from_disk);
1198#endif
1199 854
1200/* 855 del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj);
1201 * Tries to open block device by device number. Use it ONLY if you 856 del_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj);
1202 * really do not have anything better - i.e. when you are behind a
1203 * truly sucky interface and all you are given is a device number. _Never_
1204 * to be used for internal purposes. If you ever need it - reconsider
1205 * your API.
1206 */
1207struct block_device *open_by_devnum(dev_t dev, fmode_t mode)
1208{
1209 struct block_device *bdev = bdget(dev);
1210 int err = -ENOMEM;
1211 if (bdev)
1212 err = blkdev_get(bdev, mode);
1213 return err ? ERR_PTR(err) : bdev;
1214} 857}
1215 858#else
1216EXPORT_SYMBOL(open_by_devnum); 859static inline void bd_unlink_disk_holder(struct block_device *bdev)
860{ }
861#endif
1217 862
1218/** 863/**
1219 * flush_disk - invalidates all buffer-cache entries on a disk 864 * flush_disk - invalidates all buffer-cache entries on a disk
@@ -1309,10 +954,11 @@ int check_disk_change(struct block_device *bdev)
1309{ 954{
1310 struct gendisk *disk = bdev->bd_disk; 955 struct gendisk *disk = bdev->bd_disk;
1311 const struct block_device_operations *bdops = disk->fops; 956 const struct block_device_operations *bdops = disk->fops;
957 unsigned int events;
1312 958
1313 if (!bdops->media_changed) 959 events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE |
1314 return 0; 960 DISK_EVENT_EJECT_REQUEST);
1315 if (!bdops->media_changed(bdev->bd_disk)) 961 if (!(events & DISK_EVENT_MEDIA_CHANGE))
1316 return 0; 962 return 0;
1317 963
1318 flush_disk(bdev); 964 flush_disk(bdev);
@@ -1475,17 +1121,171 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part)
1475 return ret; 1121 return ret;
1476} 1122}
1477 1123
1478int blkdev_get(struct block_device *bdev, fmode_t mode) 1124/**
1125 * blkdev_get - open a block device
1126 * @bdev: block_device to open
1127 * @mode: FMODE_* mask
1128 * @holder: exclusive holder identifier
1129 *
1130 * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is
1131 * open with exclusive access. Specifying %FMODE_EXCL with %NULL
1132 * @holder is invalid. Exclusive opens may nest for the same @holder.
1133 *
1134 * On success, the reference count of @bdev is unchanged. On failure,
1135 * @bdev is put.
1136 *
1137 * CONTEXT:
1138 * Might sleep.
1139 *
1140 * RETURNS:
1141 * 0 on success, -errno on failure.
1142 */
1143int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder)
1479{ 1144{
1480 return __blkdev_get(bdev, mode, 0); 1145 struct block_device *whole = NULL;
1146 int res;
1147
1148 WARN_ON_ONCE((mode & FMODE_EXCL) && !holder);
1149
1150 if ((mode & FMODE_EXCL) && holder) {
1151 whole = bd_start_claiming(bdev, holder);
1152 if (IS_ERR(whole)) {
1153 bdput(bdev);
1154 return PTR_ERR(whole);
1155 }
1156 }
1157
1158 res = __blkdev_get(bdev, mode, 0);
1159
1160 /* __blkdev_get() may alter read only status, check it afterwards */
1161 if (!res && (mode & FMODE_WRITE) && bdev_read_only(bdev)) {
1162 __blkdev_put(bdev, mode, 0);
1163 res = -EACCES;
1164 }
1165
1166 if (whole) {
1167 /* finish claiming */
1168 mutex_lock(&bdev->bd_mutex);
1169 spin_lock(&bdev_lock);
1170
1171 if (!res) {
1172 BUG_ON(!bd_may_claim(bdev, whole, holder));
1173 /*
1174 * Note that for a whole device bd_holders
1175 * will be incremented twice, and bd_holder
1176 * will be set to bd_may_claim before being
1177 * set to holder
1178 */
1179 whole->bd_holders++;
1180 whole->bd_holder = bd_may_claim;
1181 bdev->bd_holders++;
1182 bdev->bd_holder = holder;
1183 }
1184
1185 /* tell others that we're done */
1186 BUG_ON(whole->bd_claiming != holder);
1187 whole->bd_claiming = NULL;
1188 wake_up_bit(&whole->bd_claiming, 0);
1189
1190 spin_unlock(&bdev_lock);
1191
1192 /*
1193 * Block event polling for write claims. Any write
1194 * holder makes the write_holder state stick until all
1195 * are released. This is good enough and tracking
1196 * individual writeable reference is too fragile given
1197 * the way @mode is used in blkdev_get/put().
1198 */
1199 if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder) {
1200 bdev->bd_write_holder = true;
1201 disk_block_events(bdev->bd_disk);
1202 }
1203
1204 mutex_unlock(&bdev->bd_mutex);
1205 bdput(whole);
1206 }
1207
1208 return res;
1481} 1209}
1482EXPORT_SYMBOL(blkdev_get); 1210EXPORT_SYMBOL(blkdev_get);
1483 1211
1212/**
1213 * blkdev_get_by_path - open a block device by name
1214 * @path: path to the block device to open
1215 * @mode: FMODE_* mask
1216 * @holder: exclusive holder identifier
1217 *
1218 * Open the blockdevice described by the device file at @path. @mode
1219 * and @holder are identical to blkdev_get().
1220 *
1221 * On success, the returned block_device has reference count of one.
1222 *
1223 * CONTEXT:
1224 * Might sleep.
1225 *
1226 * RETURNS:
1227 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1228 */
1229struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1230 void *holder)
1231{
1232 struct block_device *bdev;
1233 int err;
1234
1235 bdev = lookup_bdev(path);
1236 if (IS_ERR(bdev))
1237 return bdev;
1238
1239 err = blkdev_get(bdev, mode, holder);
1240 if (err)
1241 return ERR_PTR(err);
1242
1243 return bdev;
1244}
1245EXPORT_SYMBOL(blkdev_get_by_path);
1246
1247/**
1248 * blkdev_get_by_dev - open a block device by device number
1249 * @dev: device number of block device to open
1250 * @mode: FMODE_* mask
1251 * @holder: exclusive holder identifier
1252 *
1253 * Open the blockdevice described by device number @dev. @mode and
1254 * @holder are identical to blkdev_get().
1255 *
1256 * Use it ONLY if you really do not have anything better - i.e. when
1257 * you are behind a truly sucky interface and all you are given is a
1258 * device number. _Never_ to be used for internal purposes. If you
1259 * ever need it - reconsider your API.
1260 *
1261 * On success, the returned block_device has reference count of one.
1262 *
1263 * CONTEXT:
1264 * Might sleep.
1265 *
1266 * RETURNS:
1267 * Pointer to block_device on success, ERR_PTR(-errno) on failure.
1268 */
1269struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder)
1270{
1271 struct block_device *bdev;
1272 int err;
1273
1274 bdev = bdget(dev);
1275 if (!bdev)
1276 return ERR_PTR(-ENOMEM);
1277
1278 err = blkdev_get(bdev, mode, holder);
1279 if (err)
1280 return ERR_PTR(err);
1281
1282 return bdev;
1283}
1284EXPORT_SYMBOL(blkdev_get_by_dev);
1285
1484static int blkdev_open(struct inode * inode, struct file * filp) 1286static int blkdev_open(struct inode * inode, struct file * filp)
1485{ 1287{
1486 struct block_device *whole = NULL;
1487 struct block_device *bdev; 1288 struct block_device *bdev;
1488 int res;
1489 1289
1490 /* 1290 /*
1491 * Preserve backwards compatibility and allow large file access 1291 * Preserve backwards compatibility and allow large file access
@@ -1506,26 +1306,9 @@ static int blkdev_open(struct inode * inode, struct file * filp)
1506 if (bdev == NULL) 1306 if (bdev == NULL)
1507 return -ENOMEM; 1307 return -ENOMEM;
1508 1308
1509 if (filp->f_mode & FMODE_EXCL) {
1510 whole = bd_start_claiming(bdev, filp);
1511 if (IS_ERR(whole)) {
1512 bdput(bdev);
1513 return PTR_ERR(whole);
1514 }
1515 }
1516
1517 filp->f_mapping = bdev->bd_inode->i_mapping; 1309 filp->f_mapping = bdev->bd_inode->i_mapping;
1518 1310
1519 res = blkdev_get(bdev, filp->f_mode); 1311 return blkdev_get(bdev, filp->f_mode, filp);
1520
1521 if (whole) {
1522 if (res == 0)
1523 bd_finish_claiming(bdev, whole, filp);
1524 else
1525 bd_abort_claiming(whole, filp);
1526 }
1527
1528 return res;
1529} 1312}
1530 1313
1531static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) 1314static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
@@ -1539,6 +1322,7 @@ static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
1539 bdev->bd_part_count--; 1322 bdev->bd_part_count--;
1540 1323
1541 if (!--bdev->bd_openers) { 1324 if (!--bdev->bd_openers) {
1325 WARN_ON_ONCE(bdev->bd_holders);
1542 sync_blockdev(bdev); 1326 sync_blockdev(bdev);
1543 kill_bdev(bdev); 1327 kill_bdev(bdev);
1544 } 1328 }
@@ -1569,6 +1353,45 @@ static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part)
1569 1353
1570int blkdev_put(struct block_device *bdev, fmode_t mode) 1354int blkdev_put(struct block_device *bdev, fmode_t mode)
1571{ 1355{
1356 if (mode & FMODE_EXCL) {
1357 bool bdev_free;
1358
1359 /*
1360 * Release a claim on the device. The holder fields
1361 * are protected with bdev_lock. bd_mutex is to
1362 * synchronize disk_holder unlinking.
1363 */
1364 mutex_lock(&bdev->bd_mutex);
1365 spin_lock(&bdev_lock);
1366
1367 WARN_ON_ONCE(--bdev->bd_holders < 0);
1368 WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0);
1369
1370 /* bd_contains might point to self, check in a separate step */
1371 if ((bdev_free = !bdev->bd_holders))
1372 bdev->bd_holder = NULL;
1373 if (!bdev->bd_contains->bd_holders)
1374 bdev->bd_contains->bd_holder = NULL;
1375
1376 spin_unlock(&bdev_lock);
1377
1378 /*
1379 * If this was the last claim, remove holder link and
1380 * unblock evpoll if it was a write holder.
1381 */
1382 if (bdev_free) {
1383 bd_unlink_disk_holder(bdev);
1384 if (bdev->bd_write_holder) {
1385 disk_unblock_events(bdev->bd_disk);
1386 bdev->bd_write_holder = false;
1387 } else
1388 disk_check_events(bdev->bd_disk);
1389 }
1390
1391 mutex_unlock(&bdev->bd_mutex);
1392 } else
1393 disk_check_events(bdev->bd_disk);
1394
1572 return __blkdev_put(bdev, mode, 0); 1395 return __blkdev_put(bdev, mode, 0);
1573} 1396}
1574EXPORT_SYMBOL(blkdev_put); 1397EXPORT_SYMBOL(blkdev_put);
@@ -1576,8 +1399,7 @@ EXPORT_SYMBOL(blkdev_put);
1576static int blkdev_close(struct inode * inode, struct file * filp) 1399static int blkdev_close(struct inode * inode, struct file * filp)
1577{ 1400{
1578 struct block_device *bdev = I_BDEV(filp->f_mapping->host); 1401 struct block_device *bdev = I_BDEV(filp->f_mapping->host);
1579 if (bdev->bd_holder == filp) 1402
1580 bd_release(bdev);
1581 return blkdev_put(bdev, filp->f_mode); 1403 return blkdev_put(bdev, filp->f_mode);
1582} 1404}
1583 1405
@@ -1722,67 +1544,6 @@ fail:
1722} 1544}
1723EXPORT_SYMBOL(lookup_bdev); 1545EXPORT_SYMBOL(lookup_bdev);
1724 1546
1725/**
1726 * open_bdev_exclusive - open a block device by name and set it up for use
1727 *
1728 * @path: special file representing the block device
1729 * @mode: FMODE_... combination to pass be used
1730 * @holder: owner for exclusion
1731 *
1732 * Open the blockdevice described by the special file at @path, claim it
1733 * for the @holder.
1734 */
1735struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder)
1736{
1737 struct block_device *bdev, *whole;
1738 int error;
1739
1740 bdev = lookup_bdev(path);
1741 if (IS_ERR(bdev))
1742 return bdev;
1743
1744 whole = bd_start_claiming(bdev, holder);
1745 if (IS_ERR(whole)) {
1746 bdput(bdev);
1747 return whole;
1748 }
1749
1750 error = blkdev_get(bdev, mode);
1751 if (error)
1752 goto out_abort_claiming;
1753
1754 error = -EACCES;
1755 if ((mode & FMODE_WRITE) && bdev_read_only(bdev))
1756 goto out_blkdev_put;
1757
1758 bd_finish_claiming(bdev, whole, holder);
1759 return bdev;
1760
1761out_blkdev_put:
1762 blkdev_put(bdev, mode);
1763out_abort_claiming:
1764 bd_abort_claiming(whole, holder);
1765 return ERR_PTR(error);
1766}
1767
1768EXPORT_SYMBOL(open_bdev_exclusive);
1769
1770/**
1771 * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive()
1772 *
1773 * @bdev: blockdevice to close
1774 * @mode: mode, must match that used to open.
1775 *
1776 * This is the counterpart to open_bdev_exclusive().
1777 */
1778void close_bdev_exclusive(struct block_device *bdev, fmode_t mode)
1779{
1780 bd_release(bdev);
1781 blkdev_put(bdev, mode);
1782}
1783
1784EXPORT_SYMBOL(close_bdev_exclusive);
1785
1786int __invalidate_device(struct block_device *bdev) 1547int __invalidate_device(struct block_device *bdev)
1787{ 1548{
1788 struct super_block *sb = get_super(bdev); 1549 struct super_block *sb = get_super(bdev);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 6b9884507837..1718e1a5c320 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -493,7 +493,7 @@ again:
493 continue; 493 continue;
494 494
495 if (device->bdev) { 495 if (device->bdev) {
496 close_bdev_exclusive(device->bdev, device->mode); 496 blkdev_put(device->bdev, device->mode);
497 device->bdev = NULL; 497 device->bdev = NULL;
498 fs_devices->open_devices--; 498 fs_devices->open_devices--;
499 } 499 }
@@ -527,7 +527,7 @@ static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
527 527
528 list_for_each_entry(device, &fs_devices->devices, dev_list) { 528 list_for_each_entry(device, &fs_devices->devices, dev_list) {
529 if (device->bdev) { 529 if (device->bdev) {
530 close_bdev_exclusive(device->bdev, device->mode); 530 blkdev_put(device->bdev, device->mode);
531 fs_devices->open_devices--; 531 fs_devices->open_devices--;
532 } 532 }
533 if (device->writeable) { 533 if (device->writeable) {
@@ -584,13 +584,15 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
584 int seeding = 1; 584 int seeding = 1;
585 int ret = 0; 585 int ret = 0;
586 586
587 flags |= FMODE_EXCL;
588
587 list_for_each_entry(device, head, dev_list) { 589 list_for_each_entry(device, head, dev_list) {
588 if (device->bdev) 590 if (device->bdev)
589 continue; 591 continue;
590 if (!device->name) 592 if (!device->name)
591 continue; 593 continue;
592 594
593 bdev = open_bdev_exclusive(device->name, flags, holder); 595 bdev = blkdev_get_by_path(device->name, flags, holder);
594 if (IS_ERR(bdev)) { 596 if (IS_ERR(bdev)) {
595 printk(KERN_INFO "open %s failed\n", device->name); 597 printk(KERN_INFO "open %s failed\n", device->name);
596 goto error; 598 goto error;
@@ -642,7 +644,7 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
642error_brelse: 644error_brelse:
643 brelse(bh); 645 brelse(bh);
644error_close: 646error_close:
645 close_bdev_exclusive(bdev, FMODE_READ); 647 blkdev_put(bdev, flags);
646error: 648error:
647 continue; 649 continue;
648 } 650 }
@@ -688,7 +690,8 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
688 690
689 mutex_lock(&uuid_mutex); 691 mutex_lock(&uuid_mutex);
690 692
691 bdev = open_bdev_exclusive(path, flags, holder); 693 flags |= FMODE_EXCL;
694 bdev = blkdev_get_by_path(path, flags, holder);
692 695
693 if (IS_ERR(bdev)) { 696 if (IS_ERR(bdev)) {
694 ret = PTR_ERR(bdev); 697 ret = PTR_ERR(bdev);
@@ -720,7 +723,7 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
720 723
721 brelse(bh); 724 brelse(bh);
722error_close: 725error_close:
723 close_bdev_exclusive(bdev, flags); 726 blkdev_put(bdev, flags);
724error: 727error:
725 mutex_unlock(&uuid_mutex); 728 mutex_unlock(&uuid_mutex);
726 return ret; 729 return ret;
@@ -1183,8 +1186,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
1183 goto out; 1186 goto out;
1184 } 1187 }
1185 } else { 1188 } else {
1186 bdev = open_bdev_exclusive(device_path, FMODE_READ, 1189 bdev = blkdev_get_by_path(device_path, FMODE_READ | FMODE_EXCL,
1187 root->fs_info->bdev_holder); 1190 root->fs_info->bdev_holder);
1188 if (IS_ERR(bdev)) { 1191 if (IS_ERR(bdev)) {
1189 ret = PTR_ERR(bdev); 1192 ret = PTR_ERR(bdev);
1190 goto out; 1193 goto out;
@@ -1251,7 +1254,7 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
1251 root->fs_info->fs_devices->latest_bdev = next_device->bdev; 1254 root->fs_info->fs_devices->latest_bdev = next_device->bdev;
1252 1255
1253 if (device->bdev) { 1256 if (device->bdev) {
1254 close_bdev_exclusive(device->bdev, device->mode); 1257 blkdev_put(device->bdev, device->mode);
1255 device->bdev = NULL; 1258 device->bdev = NULL;
1256 device->fs_devices->open_devices--; 1259 device->fs_devices->open_devices--;
1257 } 1260 }
@@ -1294,7 +1297,7 @@ error_brelse:
1294 brelse(bh); 1297 brelse(bh);
1295error_close: 1298error_close:
1296 if (bdev) 1299 if (bdev)
1297 close_bdev_exclusive(bdev, FMODE_READ); 1300 blkdev_put(bdev, FMODE_READ | FMODE_EXCL);
1298out: 1301out:
1299 mutex_unlock(&root->fs_info->volume_mutex); 1302 mutex_unlock(&root->fs_info->volume_mutex);
1300 mutex_unlock(&uuid_mutex); 1303 mutex_unlock(&uuid_mutex);
@@ -1446,7 +1449,8 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
1446 if ((sb->s_flags & MS_RDONLY) && !root->fs_info->fs_devices->seeding) 1449 if ((sb->s_flags & MS_RDONLY) && !root->fs_info->fs_devices->seeding)
1447 return -EINVAL; 1450 return -EINVAL;
1448 1451
1449 bdev = open_bdev_exclusive(device_path, 0, root->fs_info->bdev_holder); 1452 bdev = blkdev_get_by_path(device_path, FMODE_EXCL,
1453 root->fs_info->bdev_holder);
1450 if (IS_ERR(bdev)) 1454 if (IS_ERR(bdev))
1451 return PTR_ERR(bdev); 1455 return PTR_ERR(bdev);
1452 1456
@@ -1572,7 +1576,7 @@ out:
1572 mutex_unlock(&root->fs_info->volume_mutex); 1576 mutex_unlock(&root->fs_info->volume_mutex);
1573 return ret; 1577 return ret;
1574error: 1578error:
1575 close_bdev_exclusive(bdev, 0); 1579 blkdev_put(bdev, FMODE_EXCL);
1576 if (seeding_dev) { 1580 if (seeding_dev) {
1577 mutex_unlock(&uuid_mutex); 1581 mutex_unlock(&uuid_mutex);
1578 up_write(&sb->s_umount); 1582 up_write(&sb->s_umount);
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 2740db49eb04..1be781079450 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -50,7 +50,7 @@ struct btrfs_device {
50 50
51 struct block_device *bdev; 51 struct block_device *bdev;
52 52
53 /* the mode sent to open_bdev_exclusive */ 53 /* the mode sent to blkdev_get */
54 fmode_t mode; 54 fmode_t mode;
55 55
56 char *name; 56 char *name;
diff --git a/fs/char_dev.c b/fs/char_dev.c
index 6e99b9ddd4e9..dca9e5e0f73b 100644
--- a/fs/char_dev.c
+++ b/fs/char_dev.c
@@ -59,7 +59,7 @@ static struct char_device_struct {
59} *chrdevs[CHRDEV_MAJOR_HASH_SIZE]; 59} *chrdevs[CHRDEV_MAJOR_HASH_SIZE];
60 60
61/* index in the above */ 61/* index in the above */
62static inline int major_to_index(int major) 62static inline int major_to_index(unsigned major)
63{ 63{
64 return major % CHRDEV_MAJOR_HASH_SIZE; 64 return major % CHRDEV_MAJOR_HASH_SIZE;
65} 65}
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index b7d0554631e4..7aa767d4f06f 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -364,7 +364,7 @@ static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
364 struct block_device *bdev; 364 struct block_device *bdev;
365 char b[BDEVNAME_SIZE]; 365 char b[BDEVNAME_SIZE];
366 366
367 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); 367 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
368 if (IS_ERR(bdev)) 368 if (IS_ERR(bdev))
369 goto fail; 369 goto fail;
370 return bdev; 370 return bdev;
@@ -381,8 +381,7 @@ fail:
381 */ 381 */
382static int ext3_blkdev_put(struct block_device *bdev) 382static int ext3_blkdev_put(struct block_device *bdev)
383{ 383{
384 bd_release(bdev); 384 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
385 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
386} 385}
387 386
388static int ext3_blkdev_remove(struct ext3_sb_info *sbi) 387static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
@@ -2162,13 +2161,6 @@ static journal_t *ext3_get_dev_journal(struct super_block *sb,
2162 if (bdev == NULL) 2161 if (bdev == NULL)
2163 return NULL; 2162 return NULL;
2164 2163
2165 if (bd_claim(bdev, sb)) {
2166 ext3_msg(sb, KERN_ERR,
2167 "error: failed to claim external journal device");
2168 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2169 return NULL;
2170 }
2171
2172 blocksize = sb->s_blocksize; 2164 blocksize = sb->s_blocksize;
2173 hblock = bdev_logical_block_size(bdev); 2165 hblock = bdev_logical_block_size(bdev);
2174 if (blocksize < hblock) { 2166 if (blocksize < hblock) {
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 29c80f6d8b27..cb10a06775e4 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -657,7 +657,7 @@ static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
657 struct block_device *bdev; 657 struct block_device *bdev;
658 char b[BDEVNAME_SIZE]; 658 char b[BDEVNAME_SIZE];
659 659
660 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE); 660 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
661 if (IS_ERR(bdev)) 661 if (IS_ERR(bdev))
662 goto fail; 662 goto fail;
663 return bdev; 663 return bdev;
@@ -673,8 +673,7 @@ fail:
673 */ 673 */
674static int ext4_blkdev_put(struct block_device *bdev) 674static int ext4_blkdev_put(struct block_device *bdev)
675{ 675{
676 bd_release(bdev); 676 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
678} 677}
679 678
680static int ext4_blkdev_remove(struct ext4_sb_info *sbi) 679static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
@@ -3778,13 +3777,6 @@ static journal_t *ext4_get_dev_journal(struct super_block *sb,
3778 if (bdev == NULL) 3777 if (bdev == NULL)
3779 return NULL; 3778 return NULL;
3780 3779
3781 if (bd_claim(bdev, sb)) {
3782 ext4_msg(sb, KERN_ERR,
3783 "failed to claim external journal device");
3784 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3785 return NULL;
3786 }
3787
3788 blocksize = sb->s_blocksize; 3780 blocksize = sb->s_blocksize;
3789 hblock = bdev_logical_block_size(bdev); 3781 hblock = bdev_logical_block_size(bdev);
3790 if (blocksize < hblock) { 3782 if (blocksize < hblock) {
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 693f4470a2df..777927ce6f79 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -1268,7 +1268,7 @@ static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1268{ 1268{
1269 struct block_device *bdev; 1269 struct block_device *bdev;
1270 struct super_block *s; 1270 struct super_block *s;
1271 fmode_t mode = FMODE_READ; 1271 fmode_t mode = FMODE_READ | FMODE_EXCL;
1272 int error; 1272 int error;
1273 struct gfs2_args args; 1273 struct gfs2_args args;
1274 struct gfs2_sbd *sdp; 1274 struct gfs2_sbd *sdp;
@@ -1276,7 +1276,7 @@ static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1276 if (!(flags & MS_RDONLY)) 1276 if (!(flags & MS_RDONLY))
1277 mode |= FMODE_WRITE; 1277 mode |= FMODE_WRITE;
1278 1278
1279 bdev = open_bdev_exclusive(dev_name, mode, fs_type); 1279 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
1280 if (IS_ERR(bdev)) 1280 if (IS_ERR(bdev))
1281 return ERR_CAST(bdev); 1281 return ERR_CAST(bdev);
1282 1282
@@ -1298,7 +1298,7 @@ static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1298 goto error_bdev; 1298 goto error_bdev;
1299 1299
1300 if (s->s_root) 1300 if (s->s_root)
1301 close_bdev_exclusive(bdev, mode); 1301 blkdev_put(bdev, mode);
1302 1302
1303 memset(&args, 0, sizeof(args)); 1303 memset(&args, 0, sizeof(args));
1304 args.ar_quota = GFS2_QUOTA_DEFAULT; 1304 args.ar_quota = GFS2_QUOTA_DEFAULT;
@@ -1342,7 +1342,7 @@ error_super:
1342 deactivate_locked_super(s); 1342 deactivate_locked_super(s);
1343 return ERR_PTR(error); 1343 return ERR_PTR(error);
1344error_bdev: 1344error_bdev:
1345 close_bdev_exclusive(bdev, mode); 1345 blkdev_put(bdev, mode);
1346 return ERR_PTR(error); 1346 return ERR_PTR(error);
1347} 1347}
1348 1348
diff --git a/fs/jfs/jfs_logmgr.c b/fs/jfs/jfs_logmgr.c
index e1b8493b9aaa..278e3fb40b71 100644
--- a/fs/jfs/jfs_logmgr.c
+++ b/fs/jfs/jfs_logmgr.c
@@ -1120,16 +1120,13 @@ int lmLogOpen(struct super_block *sb)
1120 * file systems to log may have n-to-1 relationship; 1120 * file systems to log may have n-to-1 relationship;
1121 */ 1121 */
1122 1122
1123 bdev = open_by_devnum(sbi->logdev, FMODE_READ|FMODE_WRITE); 1123 bdev = blkdev_get_by_dev(sbi->logdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
1124 log);
1124 if (IS_ERR(bdev)) { 1125 if (IS_ERR(bdev)) {
1125 rc = -PTR_ERR(bdev); 1126 rc = -PTR_ERR(bdev);
1126 goto free; 1127 goto free;
1127 } 1128 }
1128 1129
1129 if ((rc = bd_claim(bdev, log))) {
1130 goto close;
1131 }
1132
1133 log->bdev = bdev; 1130 log->bdev = bdev;
1134 memcpy(log->uuid, sbi->loguuid, sizeof(log->uuid)); 1131 memcpy(log->uuid, sbi->loguuid, sizeof(log->uuid));
1135 1132
@@ -1137,7 +1134,7 @@ int lmLogOpen(struct super_block *sb)
1137 * initialize log: 1134 * initialize log:
1138 */ 1135 */
1139 if ((rc = lmLogInit(log))) 1136 if ((rc = lmLogInit(log)))
1140 goto unclaim; 1137 goto close;
1141 1138
1142 list_add(&log->journal_list, &jfs_external_logs); 1139 list_add(&log->journal_list, &jfs_external_logs);
1143 1140
@@ -1163,11 +1160,8 @@ journal_found:
1163 list_del(&log->journal_list); 1160 list_del(&log->journal_list);
1164 lbmLogShutdown(log); 1161 lbmLogShutdown(log);
1165 1162
1166 unclaim:
1167 bd_release(bdev);
1168
1169 close: /* close external log device */ 1163 close: /* close external log device */
1170 blkdev_put(bdev, FMODE_READ|FMODE_WRITE); 1164 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1171 1165
1172 free: /* free log descriptor */ 1166 free: /* free log descriptor */
1173 mutex_unlock(&jfs_log_mutex); 1167 mutex_unlock(&jfs_log_mutex);
@@ -1512,8 +1506,7 @@ int lmLogClose(struct super_block *sb)
1512 bdev = log->bdev; 1506 bdev = log->bdev;
1513 rc = lmLogShutdown(log); 1507 rc = lmLogShutdown(log);
1514 1508
1515 bd_release(bdev); 1509 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1516 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
1517 1510
1518 kfree(log); 1511 kfree(log);
1519 1512
diff --git a/fs/logfs/dev_bdev.c b/fs/logfs/dev_bdev.c
index 92ca6fbe09bd..723bc5bca09a 100644
--- a/fs/logfs/dev_bdev.c
+++ b/fs/logfs/dev_bdev.c
@@ -300,7 +300,7 @@ static int bdev_write_sb(struct super_block *sb, struct page *page)
300 300
301static void bdev_put_device(struct logfs_super *s) 301static void bdev_put_device(struct logfs_super *s)
302{ 302{
303 close_bdev_exclusive(s->s_bdev, FMODE_READ|FMODE_WRITE); 303 blkdev_put(s->s_bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
304} 304}
305 305
306static int bdev_can_write_buf(struct super_block *sb, u64 ofs) 306static int bdev_can_write_buf(struct super_block *sb, u64 ofs)
@@ -325,13 +325,14 @@ int logfs_get_sb_bdev(struct logfs_super *p, struct file_system_type *type,
325{ 325{
326 struct block_device *bdev; 326 struct block_device *bdev;
327 327
328 bdev = open_bdev_exclusive(devname, FMODE_READ|FMODE_WRITE, type); 328 bdev = blkdev_get_by_path(devname, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
329 type);
329 if (IS_ERR(bdev)) 330 if (IS_ERR(bdev))
330 return PTR_ERR(bdev); 331 return PTR_ERR(bdev);
331 332
332 if (MAJOR(bdev->bd_dev) == MTD_BLOCK_MAJOR) { 333 if (MAJOR(bdev->bd_dev) == MTD_BLOCK_MAJOR) {
333 int mtdnr = MINOR(bdev->bd_dev); 334 int mtdnr = MINOR(bdev->bd_dev);
334 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE); 335 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
335 return logfs_get_sb_mtd(p, mtdnr); 336 return logfs_get_sb_mtd(p, mtdnr);
336 } 337 }
337 338
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 3a359023c9f7..230b79fbf005 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -845,11 +845,6 @@ nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
845 struct page **pp = rqstp->rq_respages + rqstp->rq_resused; 845 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
846 struct page *page = buf->page; 846 struct page *page = buf->page;
847 size_t size; 847 size_t size;
848 int ret;
849
850 ret = buf->ops->confirm(pipe, buf);
851 if (unlikely(ret))
852 return ret;
853 848
854 size = sd->len; 849 size = sd->len;
855 850
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index 70dfdd532b83..0994f6a76c07 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -1163,14 +1163,14 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
1163{ 1163{
1164 struct nilfs_super_data sd; 1164 struct nilfs_super_data sd;
1165 struct super_block *s; 1165 struct super_block *s;
1166 fmode_t mode = FMODE_READ; 1166 fmode_t mode = FMODE_READ | FMODE_EXCL;
1167 struct dentry *root_dentry; 1167 struct dentry *root_dentry;
1168 int err, s_new = false; 1168 int err, s_new = false;
1169 1169
1170 if (!(flags & MS_RDONLY)) 1170 if (!(flags & MS_RDONLY))
1171 mode |= FMODE_WRITE; 1171 mode |= FMODE_WRITE;
1172 1172
1173 sd.bdev = open_bdev_exclusive(dev_name, mode, fs_type); 1173 sd.bdev = blkdev_get_by_path(dev_name, mode, fs_type);
1174 if (IS_ERR(sd.bdev)) 1174 if (IS_ERR(sd.bdev))
1175 return ERR_CAST(sd.bdev); 1175 return ERR_CAST(sd.bdev);
1176 1176
@@ -1249,7 +1249,7 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
1249 } 1249 }
1250 1250
1251 if (!s_new) 1251 if (!s_new)
1252 close_bdev_exclusive(sd.bdev, mode); 1252 blkdev_put(sd.bdev, mode);
1253 1253
1254 return root_dentry; 1254 return root_dentry;
1255 1255
@@ -1258,7 +1258,7 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
1258 1258
1259 failed: 1259 failed:
1260 if (!s_new) 1260 if (!s_new)
1261 close_bdev_exclusive(sd.bdev, mode); 1261 blkdev_put(sd.bdev, mode);
1262 return ERR_PTR(err); 1262 return ERR_PTR(err);
1263} 1263}
1264 1264
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index a6cc05302e9f..b108e863d8f6 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -1729,7 +1729,7 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1729 goto out; 1729 goto out;
1730 1730
1731 reg->hr_bdev = I_BDEV(filp->f_mapping->host); 1731 reg->hr_bdev = I_BDEV(filp->f_mapping->host);
1732 ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ); 1732 ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL);
1733 if (ret) { 1733 if (ret) {
1734 reg->hr_bdev = NULL; 1734 reg->hr_bdev = NULL;
1735 goto out; 1735 goto out;
diff --git a/fs/partitions/check.c b/fs/partitions/check.c
index 0a8b0ad0c7e2..9c21119512b9 100644
--- a/fs/partitions/check.c
+++ b/fs/partitions/check.c
@@ -237,6 +237,13 @@ ssize_t part_size_show(struct device *dev,
237 return sprintf(buf, "%llu\n",(unsigned long long)p->nr_sects); 237 return sprintf(buf, "%llu\n",(unsigned long long)p->nr_sects);
238} 238}
239 239
240ssize_t part_ro_show(struct device *dev,
241 struct device_attribute *attr, char *buf)
242{
243 struct hd_struct *p = dev_to_part(dev);
244 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
245}
246
240ssize_t part_alignment_offset_show(struct device *dev, 247ssize_t part_alignment_offset_show(struct device *dev,
241 struct device_attribute *attr, char *buf) 248 struct device_attribute *attr, char *buf)
242{ 249{
@@ -312,6 +319,7 @@ ssize_t part_fail_store(struct device *dev,
312static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL); 319static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
313static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL); 320static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
314static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); 321static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
322static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
315static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL); 323static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
316static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show, 324static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
317 NULL); 325 NULL);
@@ -326,6 +334,7 @@ static struct attribute *part_attrs[] = {
326 &dev_attr_partition.attr, 334 &dev_attr_partition.attr,
327 &dev_attr_start.attr, 335 &dev_attr_start.attr,
328 &dev_attr_size.attr, 336 &dev_attr_size.attr,
337 &dev_attr_ro.attr,
329 &dev_attr_alignment_offset.attr, 338 &dev_attr_alignment_offset.attr,
330 &dev_attr_discard_alignment.attr, 339 &dev_attr_discard_alignment.attr,
331 &dev_attr_stat.attr, 340 &dev_attr_stat.attr,
@@ -372,6 +381,11 @@ static void delete_partition_rcu_cb(struct rcu_head *head)
372 put_device(part_to_dev(part)); 381 put_device(part_to_dev(part));
373} 382}
374 383
384void __delete_partition(struct hd_struct *part)
385{
386 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
387}
388
375void delete_partition(struct gendisk *disk, int partno) 389void delete_partition(struct gendisk *disk, int partno)
376{ 390{
377 struct disk_part_tbl *ptbl = disk->part_tbl; 391 struct disk_part_tbl *ptbl = disk->part_tbl;
@@ -390,7 +404,7 @@ void delete_partition(struct gendisk *disk, int partno)
390 kobject_put(part->holder_dir); 404 kobject_put(part->holder_dir);
391 device_del(part_to_dev(part)); 405 device_del(part_to_dev(part));
392 406
393 call_rcu(&part->rcu_head, delete_partition_rcu_cb); 407 hd_struct_put(part);
394} 408}
395 409
396static ssize_t whole_disk_show(struct device *dev, 410static ssize_t whole_disk_show(struct device *dev,
@@ -489,6 +503,7 @@ struct hd_struct *add_partition(struct gendisk *disk, int partno,
489 if (!dev_get_uevent_suppress(ddev)) 503 if (!dev_get_uevent_suppress(ddev))
490 kobject_uevent(&pdev->kobj, KOBJ_ADD); 504 kobject_uevent(&pdev->kobj, KOBJ_ADD);
491 505
506 hd_ref_init(p);
492 return p; 507 return p;
493 508
494out_free_info: 509out_free_info:
@@ -507,65 +522,6 @@ out_put:
507 return ERR_PTR(err); 522 return ERR_PTR(err);
508} 523}
509 524
510/* Not exported, helper to add_disk(). */
511void register_disk(struct gendisk *disk)
512{
513 struct device *ddev = disk_to_dev(disk);
514 struct block_device *bdev;
515 struct disk_part_iter piter;
516 struct hd_struct *part;
517 int err;
518
519 ddev->parent = disk->driverfs_dev;
520
521 dev_set_name(ddev, disk->disk_name);
522
523 /* delay uevents, until we scanned partition table */
524 dev_set_uevent_suppress(ddev, 1);
525
526 if (device_add(ddev))
527 return;
528 if (!sysfs_deprecated) {
529 err = sysfs_create_link(block_depr, &ddev->kobj,
530 kobject_name(&ddev->kobj));
531 if (err) {
532 device_del(ddev);
533 return;
534 }
535 }
536 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
537 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
538
539 /* No minors to use for partitions */
540 if (!disk_partitionable(disk))
541 goto exit;
542
543 /* No such device (e.g., media were just removed) */
544 if (!get_capacity(disk))
545 goto exit;
546
547 bdev = bdget_disk(disk, 0);
548 if (!bdev)
549 goto exit;
550
551 bdev->bd_invalidated = 1;
552 err = blkdev_get(bdev, FMODE_READ);
553 if (err < 0)
554 goto exit;
555 blkdev_put(bdev, FMODE_READ);
556
557exit:
558 /* announce disk after possible partitions are created */
559 dev_set_uevent_suppress(ddev, 0);
560 kobject_uevent(&ddev->kobj, KOBJ_ADD);
561
562 /* announce possible partitions */
563 disk_part_iter_init(&piter, disk, 0);
564 while ((part = disk_part_iter_next(&piter)))
565 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
566 disk_part_iter_exit(&piter);
567}
568
569static bool disk_unlock_native_capacity(struct gendisk *disk) 525static bool disk_unlock_native_capacity(struct gendisk *disk)
570{ 526{
571 const struct block_device_operations *bdops = disk->fops; 527 const struct block_device_operations *bdops = disk->fops;
@@ -728,33 +684,3 @@ fail:
728} 684}
729 685
730EXPORT_SYMBOL(read_dev_sector); 686EXPORT_SYMBOL(read_dev_sector);
731
732void del_gendisk(struct gendisk *disk)
733{
734 struct disk_part_iter piter;
735 struct hd_struct *part;
736
737 /* invalidate stuff */
738 disk_part_iter_init(&piter, disk,
739 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
740 while ((part = disk_part_iter_next(&piter))) {
741 invalidate_partition(disk, part->partno);
742 delete_partition(disk, part->partno);
743 }
744 disk_part_iter_exit(&piter);
745
746 invalidate_partition(disk, 0);
747 blk_free_devt(disk_to_dev(disk)->devt);
748 set_capacity(disk, 0);
749 disk->flags &= ~GENHD_FL_UP;
750 unlink_gendisk(disk);
751 part_stat_set_all(&disk->part0, 0);
752 disk->part0.stamp = 0;
753
754 kobject_put(disk->part0.holder_dir);
755 kobject_put(disk->slave_dir);
756 disk->driverfs_dev = NULL;
757 if (!sysfs_deprecated)
758 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
759 device_del(disk_to_dev(disk));
760}
diff --git a/fs/reiserfs/journal.c b/fs/reiserfs/journal.c
index d31bce1a9f90..3eea859e6990 100644
--- a/fs/reiserfs/journal.c
+++ b/fs/reiserfs/journal.c
@@ -2551,8 +2551,6 @@ static int release_journal_dev(struct super_block *super,
2551 result = 0; 2551 result = 0;
2552 2552
2553 if (journal->j_dev_bd != NULL) { 2553 if (journal->j_dev_bd != NULL) {
2554 if (journal->j_dev_bd->bd_dev != super->s_dev)
2555 bd_release(journal->j_dev_bd);
2556 result = blkdev_put(journal->j_dev_bd, journal->j_dev_mode); 2554 result = blkdev_put(journal->j_dev_bd, journal->j_dev_mode);
2557 journal->j_dev_bd = NULL; 2555 journal->j_dev_bd = NULL;
2558 } 2556 }
@@ -2570,7 +2568,7 @@ static int journal_init_dev(struct super_block *super,
2570{ 2568{
2571 int result; 2569 int result;
2572 dev_t jdev; 2570 dev_t jdev;
2573 fmode_t blkdev_mode = FMODE_READ | FMODE_WRITE; 2571 fmode_t blkdev_mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL;
2574 char b[BDEVNAME_SIZE]; 2572 char b[BDEVNAME_SIZE];
2575 2573
2576 result = 0; 2574 result = 0;
@@ -2584,7 +2582,10 @@ static int journal_init_dev(struct super_block *super,
2584 2582
2585 /* there is no "jdev" option and journal is on separate device */ 2583 /* there is no "jdev" option and journal is on separate device */
2586 if ((!jdev_name || !jdev_name[0])) { 2584 if ((!jdev_name || !jdev_name[0])) {
2587 journal->j_dev_bd = open_by_devnum(jdev, blkdev_mode); 2585 if (jdev == super->s_dev)
2586 blkdev_mode &= ~FMODE_EXCL;
2587 journal->j_dev_bd = blkdev_get_by_dev(jdev, blkdev_mode,
2588 journal);
2588 journal->j_dev_mode = blkdev_mode; 2589 journal->j_dev_mode = blkdev_mode;
2589 if (IS_ERR(journal->j_dev_bd)) { 2590 if (IS_ERR(journal->j_dev_bd)) {
2590 result = PTR_ERR(journal->j_dev_bd); 2591 result = PTR_ERR(journal->j_dev_bd);
@@ -2593,22 +2594,14 @@ static int journal_init_dev(struct super_block *super,
2593 "cannot init journal device '%s': %i", 2594 "cannot init journal device '%s': %i",
2594 __bdevname(jdev, b), result); 2595 __bdevname(jdev, b), result);
2595 return result; 2596 return result;
2596 } else if (jdev != super->s_dev) { 2597 } else if (jdev != super->s_dev)
2597 result = bd_claim(journal->j_dev_bd, journal);
2598 if (result) {
2599 blkdev_put(journal->j_dev_bd, blkdev_mode);
2600 return result;
2601 }
2602
2603 set_blocksize(journal->j_dev_bd, super->s_blocksize); 2598 set_blocksize(journal->j_dev_bd, super->s_blocksize);
2604 }
2605 2599
2606 return 0; 2600 return 0;
2607 } 2601 }
2608 2602
2609 journal->j_dev_mode = blkdev_mode; 2603 journal->j_dev_mode = blkdev_mode;
2610 journal->j_dev_bd = open_bdev_exclusive(jdev_name, 2604 journal->j_dev_bd = blkdev_get_by_path(jdev_name, blkdev_mode, journal);
2611 blkdev_mode, journal);
2612 if (IS_ERR(journal->j_dev_bd)) { 2605 if (IS_ERR(journal->j_dev_bd)) {
2613 result = PTR_ERR(journal->j_dev_bd); 2606 result = PTR_ERR(journal->j_dev_bd);
2614 journal->j_dev_bd = NULL; 2607 journal->j_dev_bd = NULL;
diff --git a/fs/splice.c b/fs/splice.c
index ce2f02579e35..50a5d978da16 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -682,19 +682,14 @@ static int pipe_to_sendpage(struct pipe_inode_info *pipe,
682{ 682{
683 struct file *file = sd->u.file; 683 struct file *file = sd->u.file;
684 loff_t pos = sd->pos; 684 loff_t pos = sd->pos;
685 int ret, more; 685 int more;
686
687 ret = buf->ops->confirm(pipe, buf);
688 if (!ret) {
689 more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
690 if (file->f_op && file->f_op->sendpage)
691 ret = file->f_op->sendpage(file, buf->page, buf->offset,
692 sd->len, &pos, more);
693 else
694 ret = -EINVAL;
695 }
696 686
697 return ret; 687 if (!likely(file->f_op && file->f_op->sendpage))
688 return -EINVAL;
689
690 more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
691 return file->f_op->sendpage(file, buf->page, buf->offset,
692 sd->len, &pos, more);
698} 693}
699 694
700/* 695/*
@@ -727,13 +722,6 @@ int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
727 void *fsdata; 722 void *fsdata;
728 int ret; 723 int ret;
729 724
730 /*
731 * make sure the data in this buffer is uptodate
732 */
733 ret = buf->ops->confirm(pipe, buf);
734 if (unlikely(ret))
735 return ret;
736
737 offset = sd->pos & ~PAGE_CACHE_MASK; 725 offset = sd->pos & ~PAGE_CACHE_MASK;
738 726
739 this_len = sd->len; 727 this_len = sd->len;
@@ -805,12 +793,17 @@ int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
805 if (sd->len > sd->total_len) 793 if (sd->len > sd->total_len)
806 sd->len = sd->total_len; 794 sd->len = sd->total_len;
807 795
808 ret = actor(pipe, buf, sd); 796 ret = buf->ops->confirm(pipe, buf);
809 if (ret <= 0) { 797 if (unlikely(ret)) {
810 if (ret == -ENODATA) 798 if (ret == -ENODATA)
811 ret = 0; 799 ret = 0;
812 return ret; 800 return ret;
813 } 801 }
802
803 ret = actor(pipe, buf, sd);
804 if (ret <= 0)
805 return ret;
806
814 buf->offset += ret; 807 buf->offset += ret;
815 buf->len -= ret; 808 buf->len -= ret;
816 809
@@ -1044,10 +1037,6 @@ static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
1044 int ret; 1037 int ret;
1045 void *data; 1038 void *data;
1046 1039
1047 ret = buf->ops->confirm(pipe, buf);
1048 if (ret)
1049 return ret;
1050
1051 data = buf->ops->map(pipe, buf, 0); 1040 data = buf->ops->map(pipe, buf, 0);
1052 ret = kernel_write(sd->u.file, data + buf->offset, sd->len, sd->pos); 1041 ret = kernel_write(sd->u.file, data + buf->offset, sd->len, sd->pos);
1053 buf->ops->unmap(pipe, buf, data); 1042 buf->ops->unmap(pipe, buf, data);
@@ -1495,10 +1484,6 @@ static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
1495 char *src; 1484 char *src;
1496 int ret; 1485 int ret;
1497 1486
1498 ret = buf->ops->confirm(pipe, buf);
1499 if (unlikely(ret))
1500 return ret;
1501
1502 /* 1487 /*
1503 * See if we can use the atomic maps, by prefaulting in the 1488 * See if we can use the atomic maps, by prefaulting in the
1504 * pages and doing an atomic copy 1489 * pages and doing an atomic copy
diff --git a/fs/super.c b/fs/super.c
index 823e061faa87..4f6a3571a634 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -767,13 +767,13 @@ struct dentry *mount_bdev(struct file_system_type *fs_type,
767{ 767{
768 struct block_device *bdev; 768 struct block_device *bdev;
769 struct super_block *s; 769 struct super_block *s;
770 fmode_t mode = FMODE_READ; 770 fmode_t mode = FMODE_READ | FMODE_EXCL;
771 int error = 0; 771 int error = 0;
772 772
773 if (!(flags & MS_RDONLY)) 773 if (!(flags & MS_RDONLY))
774 mode |= FMODE_WRITE; 774 mode |= FMODE_WRITE;
775 775
776 bdev = open_bdev_exclusive(dev_name, mode, fs_type); 776 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
777 if (IS_ERR(bdev)) 777 if (IS_ERR(bdev))
778 return ERR_CAST(bdev); 778 return ERR_CAST(bdev);
779 779
@@ -802,13 +802,13 @@ struct dentry *mount_bdev(struct file_system_type *fs_type,
802 802
803 /* 803 /*
804 * s_umount nests inside bd_mutex during 804 * s_umount nests inside bd_mutex during
805 * __invalidate_device(). close_bdev_exclusive() 805 * __invalidate_device(). blkdev_put() acquires
806 * acquires bd_mutex and can't be called under 806 * bd_mutex and can't be called under s_umount. Drop
807 * s_umount. Drop s_umount temporarily. This is safe 807 * s_umount temporarily. This is safe as we're
808 * as we're holding an active reference. 808 * holding an active reference.
809 */ 809 */
810 up_write(&s->s_umount); 810 up_write(&s->s_umount);
811 close_bdev_exclusive(bdev, mode); 811 blkdev_put(bdev, mode);
812 down_write(&s->s_umount); 812 down_write(&s->s_umount);
813 } else { 813 } else {
814 char b[BDEVNAME_SIZE]; 814 char b[BDEVNAME_SIZE];
@@ -832,7 +832,7 @@ struct dentry *mount_bdev(struct file_system_type *fs_type,
832error_s: 832error_s:
833 error = PTR_ERR(s); 833 error = PTR_ERR(s);
834error_bdev: 834error_bdev:
835 close_bdev_exclusive(bdev, mode); 835 blkdev_put(bdev, mode);
836error: 836error:
837 return ERR_PTR(error); 837 return ERR_PTR(error);
838} 838}
@@ -863,7 +863,8 @@ void kill_block_super(struct super_block *sb)
863 bdev->bd_super = NULL; 863 bdev->bd_super = NULL;
864 generic_shutdown_super(sb); 864 generic_shutdown_super(sb);
865 sync_blockdev(bdev); 865 sync_blockdev(bdev);
866 close_bdev_exclusive(bdev, mode); 866 WARN_ON_ONCE(!(mode & FMODE_EXCL));
867 blkdev_put(bdev, mode | FMODE_EXCL);
867} 868}
868 869
869EXPORT_SYMBOL(kill_block_super); 870EXPORT_SYMBOL(kill_block_super);
diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index a10f6416e563..bd07f7339366 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -606,7 +606,8 @@ xfs_blkdev_get(
606{ 606{
607 int error = 0; 607 int error = 0;
608 608
609 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp); 609 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
610 mp);
610 if (IS_ERR(*bdevp)) { 611 if (IS_ERR(*bdevp)) {
611 error = PTR_ERR(*bdevp); 612 error = PTR_ERR(*bdevp);
612 printk("XFS: Invalid device [%s], error=%d\n", name, error); 613 printk("XFS: Invalid device [%s], error=%d\n", name, error);
@@ -620,7 +621,7 @@ xfs_blkdev_put(
620 struct block_device *bdev) 621 struct block_device *bdev)
621{ 622{
622 if (bdev) 623 if (bdev)
623 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE); 624 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
624} 625}
625 626
626/* 627/*