From e09b457bdb7e8d23fc54dcef0930ac697d8de895 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:17 +0100 Subject: block: simplify holder symlink handling Code to manage symlinks in /sys/block/*/{holders|slaves} are overly complex with multiple holder considerations, redundant extra references to all involved kobjects, unused generic kobject holder support and unnecessary mixup with bd_claim/release functionalities. Strip it down to what's necessary (single gendisk holder) and make it use a separate interface. This is a step for cleaning up bd_claim/release. This patch makes dm-table slightly more complex but it will be simplified again with further changes. Signed-off-by: Tejun Heo Acked-by: Neil Brown Acked-by: Mike Snitzer Cc: dm-devel@redhat.com --- fs/block_dev.c | 322 ++++++++------------------------------------------------- 1 file changed, 44 insertions(+), 278 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 06e8ff12b97c..9329068684d2 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -426,9 +426,6 @@ static void init_once(void *foo) mutex_init(&bdev->bd_mutex); INIT_LIST_HEAD(&bdev->bd_inodes); INIT_LIST_HEAD(&bdev->bd_list); -#ifdef CONFIG_SYSFS - INIT_LIST_HEAD(&bdev->bd_holder_list); -#endif inode_init_once(&ei->vfs_inode); /* Initialize mutex for freeze. */ mutex_init(&bdev->bd_fsfreeze_mutex); @@ -881,314 +878,83 @@ void bd_release(struct block_device *bdev) EXPORT_SYMBOL(bd_release); #ifdef CONFIG_SYSFS -/* - * Functions for bd_claim_by_kobject / bd_release_from_kobject - * - * If a kobject is passed to bd_claim_by_kobject() - * and the kobject has a parent directory, - * following symlinks are created: - * o from the kobject to the claimed bdev - * o from "holders" directory of the bdev to the parent of the kobject - * bd_release_from_kobject() removes these symlinks. - * - * Example: - * If /dev/dm-0 maps to /dev/sda, kobject corresponding to - * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then: - * /sys/block/dm-0/slaves/sda --> /sys/block/sda - * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 - */ - static int add_symlink(struct kobject *from, struct kobject *to) { - if (!from || !to) - return 0; return sysfs_create_link(from, to, kobject_name(to)); } static void del_symlink(struct kobject *from, struct kobject *to) { - if (!from || !to) - return; sysfs_remove_link(from, kobject_name(to)); } -/* - * 'struct bd_holder' contains pointers to kobjects symlinked by - * bd_claim_by_kobject. - * It's connected to bd_holder_list which is protected by bdev->bd_sem. - */ -struct bd_holder { - struct list_head list; /* chain of holders of the bdev */ - int count; /* references from the holder */ - struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */ - struct kobject *hdev; /* e.g. "/block/dm-0" */ - struct kobject *hdir; /* e.g. "/block/sda/holders" */ - struct kobject *sdev; /* e.g. "/block/sda" */ -}; - -/* - * Get references of related kobjects at once. - * Returns 1 on success. 0 on failure. - * - * Should call bd_holder_release_dirs() after successful use. - */ -static int bd_holder_grab_dirs(struct block_device *bdev, - struct bd_holder *bo) -{ - if (!bdev || !bo) - return 0; - - bo->sdir = kobject_get(bo->sdir); - if (!bo->sdir) - return 0; - - bo->hdev = kobject_get(bo->sdir->parent); - if (!bo->hdev) - goto fail_put_sdir; - - bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj); - if (!bo->sdev) - goto fail_put_hdev; - - bo->hdir = kobject_get(bdev->bd_part->holder_dir); - if (!bo->hdir) - goto fail_put_sdev; - - return 1; - -fail_put_sdev: - kobject_put(bo->sdev); -fail_put_hdev: - kobject_put(bo->hdev); -fail_put_sdir: - kobject_put(bo->sdir); - - return 0; -} - -/* Put references of related kobjects at once. */ -static void bd_holder_release_dirs(struct bd_holder *bo) -{ - kobject_put(bo->hdir); - kobject_put(bo->sdev); - kobject_put(bo->hdev); - kobject_put(bo->sdir); -} - -static struct bd_holder *alloc_bd_holder(struct kobject *kobj) -{ - struct bd_holder *bo; - - bo = kzalloc(sizeof(*bo), GFP_KERNEL); - if (!bo) - return NULL; - - bo->count = 1; - bo->sdir = kobj; - - return bo; -} - -static void free_bd_holder(struct bd_holder *bo) -{ - kfree(bo); -} - /** - * find_bd_holder - find matching struct bd_holder from the block device + * bd_link_disk_holder - create symlinks between holding disk and slave bdev + * @bdev: the claimed slave bdev + * @disk: the holding disk * - * @bdev: struct block device to be searched - * @bo: target struct bd_holder + * This functions creates the following sysfs symlinks. * - * Returns matching entry with @bo in @bdev->bd_holder_list. - * If found, increment the reference count and return the pointer. - * If not found, returns NULL. - */ -static struct bd_holder *find_bd_holder(struct block_device *bdev, - struct bd_holder *bo) -{ - struct bd_holder *tmp; - - list_for_each_entry(tmp, &bdev->bd_holder_list, list) - if (tmp->sdir == bo->sdir) { - tmp->count++; - return tmp; - } - - return NULL; -} - -/** - * add_bd_holder - create sysfs symlinks for bd_claim() relationship + * - from "slaves" directory of the holder @disk to the claimed @bdev + * - from "holders" directory of the @bdev to the holder @disk * - * @bdev: block device to be bd_claimed - * @bo: preallocated and initialized by alloc_bd_holder() + * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is + * passed to bd_link_disk_holder(), then: * - * Add @bo to @bdev->bd_holder_list, create symlinks. + * /sys/block/dm-0/slaves/sda --> /sys/block/sda + * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 * - * Returns 0 if symlinks are created. - * Returns -ve if something fails. - */ -static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo) -{ - int err; - - if (!bo) - return -EINVAL; - - if (!bd_holder_grab_dirs(bdev, bo)) - return -EBUSY; - - err = add_symlink(bo->sdir, bo->sdev); - if (err) - return err; - - err = add_symlink(bo->hdir, bo->hdev); - if (err) { - del_symlink(bo->sdir, bo->sdev); - return err; - } - - list_add_tail(&bo->list, &bdev->bd_holder_list); - return 0; -} - -/** - * del_bd_holder - delete sysfs symlinks for bd_claim() relationship - * - * @bdev: block device to be bd_claimed - * @kobj: holder's kobject - * - * If there is matching entry with @kobj in @bdev->bd_holder_list - * and no other bd_claim() from the same kobject, - * remove the struct bd_holder from the list, delete symlinks for it. - * - * Returns a pointer to the struct bd_holder when it's removed from the list - * and ready to be freed. - * Returns NULL if matching claim isn't found or there is other bd_claim() - * by the same kobject. - */ -static struct bd_holder *del_bd_holder(struct block_device *bdev, - struct kobject *kobj) -{ - struct bd_holder *bo; - - list_for_each_entry(bo, &bdev->bd_holder_list, list) { - if (bo->sdir == kobj) { - bo->count--; - BUG_ON(bo->count < 0); - if (!bo->count) { - list_del(&bo->list); - del_symlink(bo->sdir, bo->sdev); - del_symlink(bo->hdir, bo->hdev); - bd_holder_release_dirs(bo); - return bo; - } - break; - } - } - - return NULL; -} - -/** - * bd_claim_by_kobject - bd_claim() with additional kobject signature - * - * @bdev: block device to be claimed - * @holder: holder's signature - * @kobj: holder's kobject + * The caller must have claimed @bdev before calling this function and + * ensure that both @bdev and @disk are valid during the creation and + * lifetime of these symlinks. * - * Do bd_claim() and if it succeeds, create sysfs symlinks between - * the bdev and the holder's kobject. - * Use bd_release_from_kobject() when relesing the claimed bdev. + * CONTEXT: + * Might sleep. * - * Returns 0 on success. (same as bd_claim()) - * Returns errno on failure. + * RETURNS: + * 0 on success, -errno on failure. */ -static int bd_claim_by_kobject(struct block_device *bdev, void *holder, - struct kobject *kobj) +int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk) { - int err; - struct bd_holder *bo, *found; - - if (!kobj) - return -EINVAL; - - bo = alloc_bd_holder(kobj); - if (!bo) - return -ENOMEM; + int ret = 0; mutex_lock(&bdev->bd_mutex); - err = bd_claim(bdev, holder); - if (err) - goto fail; + WARN_ON_ONCE(!bdev->bd_holder || bdev->bd_holder_disk); - found = find_bd_holder(bdev, bo); - if (found) - goto fail; + /* FIXME: remove the following once add_disk() handles errors */ + if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir)) + goto out_unlock; - err = add_bd_holder(bdev, bo); - if (err) - bd_release(bdev); - else - bo = NULL; -fail: - mutex_unlock(&bdev->bd_mutex); - free_bd_holder(bo); - return err; -} + ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); + if (ret) + goto out_unlock; -/** - * bd_release_from_kobject - bd_release() with additional kobject signature - * - * @bdev: block device to be released - * @kobj: holder's kobject - * - * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject(). - */ -static void bd_release_from_kobject(struct block_device *bdev, - struct kobject *kobj) -{ - if (!kobj) - return; + ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); + if (ret) { + del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); + goto out_unlock; + } - mutex_lock(&bdev->bd_mutex); - bd_release(bdev); - free_bd_holder(del_bd_holder(bdev, kobj)); + bdev->bd_holder_disk = disk; +out_unlock: mutex_unlock(&bdev->bd_mutex); + return ret; } +EXPORT_SYMBOL_GPL(bd_link_disk_holder); -/** - * bd_claim_by_disk - wrapper function for bd_claim_by_kobject() - * - * @bdev: block device to be claimed - * @holder: holder's signature - * @disk: holder's gendisk - * - * Call bd_claim_by_kobject() with getting @disk->slave_dir. - */ -int bd_claim_by_disk(struct block_device *bdev, void *holder, - struct gendisk *disk) +void bd_unlink_disk_holder(struct block_device *bdev) { - return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); -} -EXPORT_SYMBOL_GPL(bd_claim_by_disk); + struct gendisk *disk = bdev->bd_holder_disk; -/** - * bd_release_from_disk - wrapper function for bd_release_from_kobject() - * - * @bdev: block device to be claimed - * @disk: holder's gendisk - * - * Call bd_release_from_kobject() and put @disk->slave_dir. - */ -void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk) -{ - bd_release_from_kobject(bdev, disk->slave_dir); - kobject_put(disk->slave_dir); + bdev->bd_holder_disk = NULL; + if (!disk) + return; + + del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); + del_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); } -EXPORT_SYMBOL_GPL(bd_release_from_disk); +EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); #endif /* -- cgit v1.2.3 From e525fd89d380c4a94c0d63913a1dd1a593ed25e7 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:17 +0100 Subject: block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo Acked-by: Neil Brown Acked-by: Ryusuke Konishi Acked-by: Mike Snitzer Acked-by: Philipp Reisner Cc: Peter Osterlund Cc: Martin Schwidefsky Cc: Heiko Carstens Cc: Jan Kara Cc: Andrew Morton Cc: Andreas Dilger Cc: "Theodore Ts'o" Cc: Mark Fasheh Cc: Joel Becker Cc: Alex Elder Cc: Christoph Hellwig Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen Cc: Scott Branden Cc: Chris Mason Cc: Steven Whitehouse Cc: Dave Kleikamp Cc: Joern Engel Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro --- fs/block_dev.c | 149 +++++++++++++++++++-------------------------------------- 1 file changed, 50 insertions(+), 99 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 9329068684d2..fc48912354d1 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -660,7 +660,7 @@ static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, else if (bdev->bd_contains == bdev) return true; /* is a whole device which isn't held */ - else if (whole->bd_holder == bd_claim) + else if (whole->bd_holder == bd_may_claim) return true; /* is a partition of a device that is being partitioned */ else if (whole->bd_holder != NULL) return false; /* is a partition of a held device */ @@ -807,10 +807,10 @@ static void __bd_claim(struct block_device *bdev, struct block_device *whole, { /* note that for a whole device bd_holders * will be incremented twice, and bd_holder will - * be set to bd_claim before being set to holder + * be set to bd_may_claim before being set to holder */ whole->bd_holders++; - whole->bd_holder = bd_claim; + whole->bd_holder = bd_may_claim; bdev->bd_holders++; bdev->bd_holder = holder; } @@ -835,37 +835,7 @@ static void bd_finish_claiming(struct block_device *bdev, __bd_abort_claiming(whole, holder); /* not actually an abort */ } -/** - * bd_claim - claim a block device - * @bdev: block device to claim - * @holder: holder trying to claim @bdev - * - * Try to claim @bdev which must have been opened successfully. - * - * CONTEXT: - * Might sleep. - * - * RETURNS: - * 0 if successful, -EBUSY if @bdev is already claimed. - */ -int bd_claim(struct block_device *bdev, void *holder) -{ - struct block_device *whole = bdev->bd_contains; - int res; - - might_sleep(); - - spin_lock(&bdev_lock); - res = bd_prepare_to_claim(bdev, whole, holder); - if (res == 0) - __bd_claim(bdev, whole, holder); - spin_unlock(&bdev_lock); - - return res; -} -EXPORT_SYMBOL(bd_claim); - -void bd_release(struct block_device *bdev) +static void bd_release(struct block_device *bdev) { spin_lock(&bdev_lock); if (!--bdev->bd_contains->bd_holders) @@ -875,8 +845,6 @@ void bd_release(struct block_device *bdev) spin_unlock(&bdev_lock); } -EXPORT_SYMBOL(bd_release); - #ifdef CONFIG_SYSFS static int add_symlink(struct kobject *from, struct kobject *to) { @@ -943,7 +911,7 @@ out_unlock: } EXPORT_SYMBOL_GPL(bd_link_disk_holder); -void bd_unlink_disk_holder(struct block_device *bdev) +static void bd_unlink_disk_holder(struct block_device *bdev) { struct gendisk *disk = bdev->bd_holder_disk; @@ -954,7 +922,9 @@ void bd_unlink_disk_holder(struct block_device *bdev) del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); del_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); } -EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); +#else +static inline void bd_unlink_disk_holder(struct block_device *bdev) +{ } #endif /* @@ -964,12 +934,12 @@ EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); * to be used for internal purposes. If you ever need it - reconsider * your API. */ -struct block_device *open_by_devnum(dev_t dev, fmode_t mode) +struct block_device *open_by_devnum(dev_t dev, fmode_t mode, void *holder) { struct block_device *bdev = bdget(dev); int err = -ENOMEM; if (bdev) - err = blkdev_get(bdev, mode); + err = blkdev_get(bdev, mode, holder); return err ? ERR_PTR(err) : bdev; } @@ -1235,17 +1205,37 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) return ret; } -int blkdev_get(struct block_device *bdev, fmode_t mode) +int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) { - return __blkdev_get(bdev, mode, 0); + struct block_device *whole = NULL; + int res; + + WARN_ON_ONCE((mode & FMODE_EXCL) && !holder); + + if ((mode & FMODE_EXCL) && holder) { + whole = bd_start_claiming(bdev, holder); + if (IS_ERR(whole)) { + bdput(bdev); + return PTR_ERR(whole); + } + } + + res = __blkdev_get(bdev, mode, 0); + + if (whole) { + if (res == 0) + bd_finish_claiming(bdev, whole, holder); + else + bd_abort_claiming(whole, holder); + } + + return res; } EXPORT_SYMBOL(blkdev_get); static int blkdev_open(struct inode * inode, struct file * filp) { - struct block_device *whole = NULL; struct block_device *bdev; - int res; /* * Preserve backwards compatibility and allow large file access @@ -1266,26 +1256,9 @@ static int blkdev_open(struct inode * inode, struct file * filp) if (bdev == NULL) return -ENOMEM; - if (filp->f_mode & FMODE_EXCL) { - whole = bd_start_claiming(bdev, filp); - if (IS_ERR(whole)) { - bdput(bdev); - return PTR_ERR(whole); - } - } - filp->f_mapping = bdev->bd_inode->i_mapping; - res = blkdev_get(bdev, filp->f_mode); - - if (whole) { - if (res == 0) - bd_finish_claiming(bdev, whole, filp); - else - bd_abort_claiming(whole, filp); - } - - return res; + return blkdev_get(bdev, filp->f_mode, filp); } static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) @@ -1329,6 +1302,13 @@ static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) int blkdev_put(struct block_device *bdev, fmode_t mode) { + if (mode & FMODE_EXCL) { + mutex_lock(&bdev->bd_mutex); + bd_release(bdev); + if (!bdev->bd_holders) + bd_unlink_disk_holder(bdev); + mutex_unlock(&bdev->bd_mutex); + } return __blkdev_put(bdev, mode, 0); } EXPORT_SYMBOL(blkdev_put); @@ -1336,8 +1316,7 @@ EXPORT_SYMBOL(blkdev_put); static int blkdev_close(struct inode * inode, struct file * filp) { struct block_device *bdev = I_BDEV(filp->f_mapping->host); - if (bdev->bd_holder == filp) - bd_release(bdev); + return blkdev_put(bdev, filp->f_mode); } @@ -1494,55 +1473,27 @@ EXPORT_SYMBOL(lookup_bdev); */ struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder) { - struct block_device *bdev, *whole; + struct block_device *bdev; int error; bdev = lookup_bdev(path); if (IS_ERR(bdev)) return bdev; - whole = bd_start_claiming(bdev, holder); - if (IS_ERR(whole)) { - bdput(bdev); - return whole; - } - - error = blkdev_get(bdev, mode); + error = blkdev_get(bdev, mode | FMODE_EXCL, holder); if (error) - goto out_abort_claiming; + return ERR_PTR(error); - error = -EACCES; - if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) - goto out_blkdev_put; + if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) { + blkdev_put(bdev, mode); + return ERR_PTR(-EACCES); + } - bd_finish_claiming(bdev, whole, holder); return bdev; - -out_blkdev_put: - blkdev_put(bdev, mode); -out_abort_claiming: - bd_abort_claiming(whole, holder); - return ERR_PTR(error); } EXPORT_SYMBOL(open_bdev_exclusive); -/** - * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive() - * - * @bdev: blockdevice to close - * @mode: mode, must match that used to open. - * - * This is the counterpart to open_bdev_exclusive(). - */ -void close_bdev_exclusive(struct block_device *bdev, fmode_t mode) -{ - bd_release(bdev); - blkdev_put(bdev, mode); -} - -EXPORT_SYMBOL(close_bdev_exclusive); - int __invalidate_device(struct block_device *bdev) { struct super_block *sb = get_super(bdev); -- cgit v1.2.3 From 6a027eff62f6ae32d49f2ae5dadd6f4eee1ddae2 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:17 +0100 Subject: block: reorganize claim/release implementation With claim/release rolled into blkdev_get/put(), there's no reason to keep bd_abort/finish_claim(), __bd_claim() and bd_release() as separate functions. It only makes the code difficult to follow. Collapse them into blkdev_get/put(). This will ease future changes around claim/release. Signed-off-by: Tejun Heo --- fs/block_dev.c | 127 ++++++++++++++++++++++----------------------------------- 1 file changed, 48 insertions(+), 79 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index fc48912354d1..269bfbbd10fc 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -772,79 +772,6 @@ static struct block_device *bd_start_claiming(struct block_device *bdev, } } -/* releases bdev_lock */ -static void __bd_abort_claiming(struct block_device *whole, void *holder) -{ - BUG_ON(whole->bd_claiming != holder); - whole->bd_claiming = NULL; - wake_up_bit(&whole->bd_claiming, 0); - - spin_unlock(&bdev_lock); - bdput(whole); -} - -/** - * bd_abort_claiming - abort claiming a block device - * @whole: whole block device returned by bd_start_claiming() - * @holder: holder trying to claim @bdev - * - * Abort a claiming block started by bd_start_claiming(). Note that - * @whole is not the block device to be claimed but the whole device - * returned by bd_start_claiming(). - * - * CONTEXT: - * Grabs and releases bdev_lock. - */ -static void bd_abort_claiming(struct block_device *whole, void *holder) -{ - spin_lock(&bdev_lock); - __bd_abort_claiming(whole, holder); /* releases bdev_lock */ -} - -/* increment holders when we have a legitimate claim. requires bdev_lock */ -static void __bd_claim(struct block_device *bdev, struct block_device *whole, - void *holder) -{ - /* note that for a whole device bd_holders - * will be incremented twice, and bd_holder will - * be set to bd_may_claim before being set to holder - */ - whole->bd_holders++; - whole->bd_holder = bd_may_claim; - bdev->bd_holders++; - bdev->bd_holder = holder; -} - -/** - * bd_finish_claiming - finish claiming a block device - * @bdev: block device of interest (passed to bd_start_claiming()) - * @whole: whole block device returned by bd_start_claiming() - * @holder: holder trying to claim @bdev - * - * Finish a claiming block started by bd_start_claiming(). - * - * CONTEXT: - * Grabs and releases bdev_lock. - */ -static void bd_finish_claiming(struct block_device *bdev, - struct block_device *whole, void *holder) -{ - spin_lock(&bdev_lock); - BUG_ON(!bd_may_claim(bdev, whole, holder)); - __bd_claim(bdev, whole, holder); - __bd_abort_claiming(whole, holder); /* not actually an abort */ -} - -static void bd_release(struct block_device *bdev) -{ - spin_lock(&bdev_lock); - if (!--bdev->bd_contains->bd_holders) - bdev->bd_contains->bd_holder = NULL; - if (!--bdev->bd_holders) - bdev->bd_holder = NULL; - spin_unlock(&bdev_lock); -} - #ifdef CONFIG_SYSFS static int add_symlink(struct kobject *from, struct kobject *to) { @@ -1223,10 +1150,30 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) res = __blkdev_get(bdev, mode, 0); if (whole) { - if (res == 0) - bd_finish_claiming(bdev, whole, holder); - else - bd_abort_claiming(whole, holder); + /* finish claiming */ + spin_lock(&bdev_lock); + + if (res == 0) { + BUG_ON(!bd_may_claim(bdev, whole, holder)); + /* + * Note that for a whole device bd_holders + * will be incremented twice, and bd_holder + * will be set to bd_may_claim before being + * set to holder + */ + whole->bd_holders++; + whole->bd_holder = bd_may_claim; + bdev->bd_holders++; + bdev->bd_holder = holder; + } + + /* tell others that we're done */ + BUG_ON(whole->bd_claiming != holder); + whole->bd_claiming = NULL; + wake_up_bit(&whole->bd_claiming, 0); + + spin_unlock(&bdev_lock); + bdput(whole); } return res; @@ -1272,6 +1219,7 @@ static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) bdev->bd_part_count--; if (!--bdev->bd_openers) { + WARN_ON_ONCE(bdev->bd_holders); sync_blockdev(bdev); kill_bdev(bdev); } @@ -1303,10 +1251,31 @@ static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) int blkdev_put(struct block_device *bdev, fmode_t mode) { if (mode & FMODE_EXCL) { + bool bdev_free; + + /* + * Release a claim on the device. The holder fields + * are protected with bdev_lock. bd_mutex is to + * synchronize disk_holder unlinking. + */ mutex_lock(&bdev->bd_mutex); - bd_release(bdev); - if (!bdev->bd_holders) + spin_lock(&bdev_lock); + + WARN_ON_ONCE(--bdev->bd_holders < 0); + WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0); + + /* bd_contains might point to self, check in a separate step */ + if ((bdev_free = !bdev->bd_holders)) + bdev->bd_holder = NULL; + if (!bdev->bd_contains->bd_holders) + bdev->bd_contains->bd_holder = NULL; + + spin_unlock(&bdev_lock); + + /* if this was the last claim, holder link should go too */ + if (bdev_free) bd_unlink_disk_holder(bdev); + mutex_unlock(&bdev->bd_mutex); } return __blkdev_put(bdev, mode, 0); -- cgit v1.2.3 From 75f1dc0d076d1c1168f2115f1941ea627d38bd5a Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:17 +0100 Subject: block: check bdev_read_only() from blkdev_get() bdev read-only status can be queried using bdev_read_only() and may change while the device is being opened. Enforce it by checking it from blkdev_get() after open succeeds. This makes bdev_read_only() check in open_bdev_exclusive() and fsg_lun_open() unnecessary. Drop them. Signed-off-by: Tejun Heo Cc: David Brownell Cc: linux-usb@vger.kernel.org --- fs/block_dev.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 269bfbbd10fc..606a5259f87f 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -1149,6 +1149,12 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) res = __blkdev_get(bdev, mode, 0); + /* __blkdev_get() may alter read only status, check it afterwards */ + if (!res && (mode & FMODE_WRITE) && bdev_read_only(bdev)) { + __blkdev_put(bdev, mode, 0); + res = -EACCES; + } + if (whole) { /* finish claiming */ spin_lock(&bdev_lock); @@ -1453,11 +1459,6 @@ struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *h if (error) return ERR_PTR(error); - if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) { - blkdev_put(bdev, mode); - return ERR_PTR(-EACCES); - } - return bdev; } -- cgit v1.2.3 From d4d77629953eabd3c14f6fa5746f6b28babfc55f Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:18 +0100 Subject: block: clean up blkdev_get() wrappers and their users After recent blkdev_get() modifications, open_by_devnum() and open_bdev_exclusive() are simple wrappers around blkdev_get(). Replace them with blkdev_get_by_dev() and blkdev_get_by_path(). blkdev_get_by_dev() is identical to open_by_devnum(). blkdev_get_by_path() is slightly different in that it doesn't automatically add %FMODE_EXCL to @mode. All users are converted. Most conversions are mechanical and don't introduce any behavior difference. There are several exceptions. * btrfs now sets FMODE_EXCL in btrfs_device->mode, so there's no reason to OR it explicitly on blkdev_put(). * gfs2, nilfs2 and the generic mount_bdev() now set FMODE_EXCL in sb->s_mode. * With the above changes, sb->s_mode now always should contain FMODE_EXCL. WARN_ON_ONCE() added to kill_block_super() to detect errors. The new blkdev_get_*() functions are with proper docbook comments. While at it, add function description to blkdev_get() too. Signed-off-by: Tejun Heo Cc: Philipp Reisner Cc: Neil Brown Cc: Mike Snitzer Cc: Joern Engel Cc: Chris Mason Cc: Jan Kara Cc: "Theodore Ts'o" Cc: KONISHI Ryusuke Cc: reiserfs-devel@vger.kernel.org Cc: xfs-masters@oss.sgi.com Cc: Alexander Viro --- fs/block_dev.c | 139 ++++++++++++++++++++++++++++++++++++++------------------- 1 file changed, 93 insertions(+), 46 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 606a5259f87f..c1c1b8c3fb99 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -854,24 +854,6 @@ static inline void bd_unlink_disk_holder(struct block_device *bdev) { } #endif -/* - * Tries to open block device by device number. Use it ONLY if you - * really do not have anything better - i.e. when you are behind a - * truly sucky interface and all you are given is a device number. _Never_ - * to be used for internal purposes. If you ever need it - reconsider - * your API. - */ -struct block_device *open_by_devnum(dev_t dev, fmode_t mode, void *holder) -{ - struct block_device *bdev = bdget(dev); - int err = -ENOMEM; - if (bdev) - err = blkdev_get(bdev, mode, holder); - return err ? ERR_PTR(err) : bdev; -} - -EXPORT_SYMBOL(open_by_devnum); - /** * flush_disk - invalidates all buffer-cache entries on a disk * @@ -1132,6 +1114,25 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) return ret; } +/** + * blkdev_get - open a block device + * @bdev: block_device to open + * @mode: FMODE_* mask + * @holder: exclusive holder identifier + * + * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is + * open with exclusive access. Specifying %FMODE_EXCL with %NULL + * @holder is invalid. Exclusive opens may nest for the same @holder. + * + * On success, the reference count of @bdev is unchanged. On failure, + * @bdev is put. + * + * CONTEXT: + * Might sleep. + * + * RETURNS: + * 0 on success, -errno on failure. + */ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) { struct block_device *whole = NULL; @@ -1186,6 +1187,80 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) } EXPORT_SYMBOL(blkdev_get); +/** + * blkdev_get_by_path - open a block device by name + * @path: path to the block device to open + * @mode: FMODE_* mask + * @holder: exclusive holder identifier + * + * Open the blockdevice described by the device file at @path. @mode + * and @holder are identical to blkdev_get(). + * + * On success, the returned block_device has reference count of one. + * + * CONTEXT: + * Might sleep. + * + * RETURNS: + * Pointer to block_device on success, ERR_PTR(-errno) on failure. + */ +struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, + void *holder) +{ + struct block_device *bdev; + int err; + + bdev = lookup_bdev(path); + if (IS_ERR(bdev)) + return bdev; + + err = blkdev_get(bdev, mode, holder); + if (err) + return ERR_PTR(err); + + return bdev; +} +EXPORT_SYMBOL(blkdev_get_by_path); + +/** + * blkdev_get_by_dev - open a block device by device number + * @dev: device number of block device to open + * @mode: FMODE_* mask + * @holder: exclusive holder identifier + * + * Open the blockdevice described by device number @dev. @mode and + * @holder are identical to blkdev_get(). + * + * Use it ONLY if you really do not have anything better - i.e. when + * you are behind a truly sucky interface and all you are given is a + * device number. _Never_ to be used for internal purposes. If you + * ever need it - reconsider your API. + * + * On success, the returned block_device has reference count of one. + * + * CONTEXT: + * Might sleep. + * + * RETURNS: + * Pointer to block_device on success, ERR_PTR(-errno) on failure. + */ +struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder) +{ + struct block_device *bdev; + int err; + + bdev = bdget(dev); + if (!bdev) + return ERR_PTR(-ENOMEM); + + err = blkdev_get(bdev, mode, holder); + if (err) + return ERR_PTR(err); + + return bdev; +} +EXPORT_SYMBOL(blkdev_get_by_dev); + static int blkdev_open(struct inode * inode, struct file * filp) { struct block_device *bdev; @@ -1436,34 +1511,6 @@ fail: } EXPORT_SYMBOL(lookup_bdev); -/** - * open_bdev_exclusive - open a block device by name and set it up for use - * - * @path: special file representing the block device - * @mode: FMODE_... combination to pass be used - * @holder: owner for exclusion - * - * Open the blockdevice described by the special file at @path, claim it - * for the @holder. - */ -struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder) -{ - struct block_device *bdev; - int error; - - bdev = lookup_bdev(path); - if (IS_ERR(bdev)) - return bdev; - - error = blkdev_get(bdev, mode | FMODE_EXCL, holder); - if (error) - return ERR_PTR(error); - - return bdev; -} - -EXPORT_SYMBOL(open_bdev_exclusive); - int __invalidate_device(struct block_device *bdev) { struct super_block *sb = get_super(bdev); -- cgit v1.2.3 From 77ea887e433ad8389d416826936c110fa7910f80 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Dec 2010 20:57:37 +0100 Subject: implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo Cc: Kay Sievers Cc: Jan Kara Signed-off-by: Jens Axboe --- fs/block_dev.c | 41 ++++++++++++++++++++++++++++++++++------- 1 file changed, 34 insertions(+), 7 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index c1c1b8c3fb99..6017389711ee 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -948,10 +948,11 @@ int check_disk_change(struct block_device *bdev) { struct gendisk *disk = bdev->bd_disk; const struct block_device_operations *bdops = disk->fops; + unsigned int events; - if (!bdops->media_changed) - return 0; - if (!bdops->media_changed(bdev->bd_disk)) + events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE | + DISK_EVENT_EJECT_REQUEST); + if (!(events & DISK_EVENT_MEDIA_CHANGE)) return 0; flush_disk(bdev); @@ -1158,9 +1159,10 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) if (whole) { /* finish claiming */ + mutex_lock(&bdev->bd_mutex); spin_lock(&bdev_lock); - if (res == 0) { + if (!res) { BUG_ON(!bd_may_claim(bdev, whole, holder)); /* * Note that for a whole device bd_holders @@ -1180,6 +1182,20 @@ int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) wake_up_bit(&whole->bd_claiming, 0); spin_unlock(&bdev_lock); + + /* + * Block event polling for write claims. Any write + * holder makes the write_holder state stick until all + * are released. This is good enough and tracking + * individual writeable reference is too fragile given + * the way @mode is used in blkdev_get/put(). + */ + if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder) { + bdev->bd_write_holder = true; + disk_block_events(bdev->bd_disk); + } + + mutex_unlock(&bdev->bd_mutex); bdput(whole); } @@ -1353,12 +1369,23 @@ int blkdev_put(struct block_device *bdev, fmode_t mode) spin_unlock(&bdev_lock); - /* if this was the last claim, holder link should go too */ - if (bdev_free) + /* + * If this was the last claim, remove holder link and + * unblock evpoll if it was a write holder. + */ + if (bdev_free) { bd_unlink_disk_holder(bdev); + if (bdev->bd_write_holder) { + disk_unblock_events(bdev->bd_disk); + bdev->bd_write_holder = false; + } else + disk_check_events(bdev->bd_disk); + } mutex_unlock(&bdev->bd_mutex); - } + } else + disk_check_events(bdev->bd_disk); + return __blkdev_put(bdev, mode, 0); } EXPORT_SYMBOL(blkdev_put); -- cgit v1.2.3