From 87c0fded852ae20bddb7833da6ead082404de86a Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Thu, 29 Sep 2016 13:41:05 +0200 Subject: rbd: don't wait for the lock forever if blacklisted -EBLACKLISTED from __rbd_register_watch() means that our ceph_client got blacklisted - we won't be able to restore the watch and reacquire the lock. Wake up and fail all outstanding requests waiting for the lock and arrange for all new requests that require the lock to fail immediately. Signed-off-by: Ilya Dryomov Tested-by: Mike Christie --- drivers/block/rbd.c | 50 +++++++++++++++++++++++++++++++++----------------- 1 file changed, 33 insertions(+), 17 deletions(-) diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index abb71628ab61..633e8c2ea120 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -415,15 +415,15 @@ struct rbd_device { }; /* - * Flag bits for rbd_dev->flags. If atomicity is required, - * rbd_dev->lock is used to protect access. - * - * Currently, only the "removing" flag (which is coupled with the - * "open_count" field) requires atomic access. + * Flag bits for rbd_dev->flags: + * - REMOVING (which is coupled with rbd_dev->open_count) is protected + * by rbd_dev->lock + * - BLACKLISTED is protected by rbd_dev->lock_rwsem */ enum rbd_dev_flags { RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ + RBD_DEV_FLAG_BLACKLISTED, /* our ceph_client is blacklisted */ }; static DEFINE_MUTEX(client_mutex); /* Serialize client creation */ @@ -3926,6 +3926,7 @@ static void rbd_reregister_watch(struct work_struct *work) struct rbd_device *rbd_dev = container_of(to_delayed_work(work), struct rbd_device, watch_dwork); bool was_lock_owner = false; + bool need_to_wake = false; int ret; dout("%s rbd_dev %p\n", __func__, rbd_dev); @@ -3935,19 +3936,27 @@ static void rbd_reregister_watch(struct work_struct *work) was_lock_owner = rbd_release_lock(rbd_dev); mutex_lock(&rbd_dev->watch_mutex); - if (rbd_dev->watch_state != RBD_WATCH_STATE_ERROR) - goto fail_unlock; + if (rbd_dev->watch_state != RBD_WATCH_STATE_ERROR) { + mutex_unlock(&rbd_dev->watch_mutex); + goto out; + } ret = __rbd_register_watch(rbd_dev); if (ret) { rbd_warn(rbd_dev, "failed to reregister watch: %d", ret); - if (ret != -EBLACKLISTED) + if (ret == -EBLACKLISTED) { + set_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags); + need_to_wake = true; + } else { queue_delayed_work(rbd_dev->task_wq, &rbd_dev->watch_dwork, RBD_RETRY_DELAY); - goto fail_unlock; + } + mutex_unlock(&rbd_dev->watch_mutex); + goto out; } + need_to_wake = true; rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED; rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id; mutex_unlock(&rbd_dev->watch_mutex); @@ -3963,13 +3972,10 @@ static void rbd_reregister_watch(struct work_struct *work) ret); } +out: up_write(&rbd_dev->lock_rwsem); - wake_requests(rbd_dev, true); - return; - -fail_unlock: - mutex_unlock(&rbd_dev->watch_mutex); - up_write(&rbd_dev->lock_rwsem); + if (need_to_wake) + wake_requests(rbd_dev, true); } /* @@ -4074,7 +4080,9 @@ static void rbd_wait_state_locked(struct rbd_device *rbd_dev) up_read(&rbd_dev->lock_rwsem); schedule(); down_read(&rbd_dev->lock_rwsem); - } while (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED); + } while (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED && + !test_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags)); + finish_wait(&rbd_dev->lock_waitq, &wait); } @@ -4166,8 +4174,16 @@ static void rbd_queue_workfn(struct work_struct *work) if (must_be_locked) { down_read(&rbd_dev->lock_rwsem); - if (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED) + if (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED && + !test_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags)) rbd_wait_state_locked(rbd_dev); + + WARN_ON((rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED) ^ + !test_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags)); + if (test_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags)) { + result = -EBLACKLISTED; + goto err_unlock; + } } img_request = rbd_img_request_create(rbd_dev, offset, length, op_type, -- cgit v1.2.3 From 4d73644bc3d76dd161a84e3849c6f2c9c01c4ba7 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Thu, 29 Sep 2016 14:23:12 +0200 Subject: rbd: don't retry watch reregistration if header object is gone If the header object gets deleted (perhaps along with the entire pool), there is no point in attempting to reregister the watch. Treat this the same as blacklisting: fail all pending and new I/Os requiring the lock. Signed-off-by: Ilya Dryomov --- drivers/block/rbd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 633e8c2ea120..7b274ff4632c 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -3944,7 +3944,7 @@ static void rbd_reregister_watch(struct work_struct *work) ret = __rbd_register_watch(rbd_dev); if (ret) { rbd_warn(rbd_dev, "failed to reregister watch: %d", ret); - if (ret == -EBLACKLISTED) { + if (ret == -EBLACKLISTED || ret == -ENOENT) { set_bit(RBD_DEV_FLAG_BLACKLISTED, &rbd_dev->flags); need_to_wake = true; } else { -- cgit v1.2.3 From 0d7718f666be181fda1ba2d08f137d87c1419347 Mon Sep 17 00:00:00 2001 From: Nikolay Borisov Date: Mon, 10 Oct 2016 15:38:18 +0300 Subject: ceph: fix error handling in ceph_read_iter In case __ceph_do_getattr returns an error and the retry_op in ceph_read_iter is not READ_INLINE, then it's possible to invoke __free_page on a page which is NULL, this naturally leads to a crash. This can happen when, for example, a process waiting on a MDS reply receives sigterm. Fix this by explicitly checking whether the page is set or not. Cc: stable@vger.kernel.org # 3.19+ Signed-off-by: Nikolay Borisov Reviewed-by: Yan, Zheng Signed-off-by: Ilya Dryomov --- fs/ceph/file.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 7bf08825cc11..18630e800208 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -1272,7 +1272,8 @@ again: statret = __ceph_do_getattr(inode, page, CEPH_STAT_CAP_INLINE_DATA, !!page); if (statret < 0) { - __free_page(page); + if (page) + __free_page(page); if (statret == -ENODATA) { BUG_ON(retry_op != READ_INLINE); goto again; -- cgit v1.2.3 From f72f94555aa32b2c4374dde5ad9b960cc4ff32e2 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 12 Oct 2016 14:48:28 +0800 Subject: ceph: fix readdir vs fragmentation race following sequence of events tigger the race - client readdir frag 0* -> got item 'A' - MDS merges frag 0* and frag 1* - client send readdir request (frag 1*, offset 2, readdir_start 'A') - MDS reply items (that are after item 'A') in frag * Link: http://tracker.ceph.com/issues/17286 Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index bca1b49c1c4b..ef4d04647325 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1511,7 +1511,8 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req, ceph_fill_dirfrag(d_inode(parent), rinfo->dir_dir); } - if (ceph_frag_is_leftmost(frag) && req->r_readdir_offset == 2) { + if (ceph_frag_is_leftmost(frag) && req->r_readdir_offset == 2 && + !(rinfo->hash_order && req->r_path2)) { /* note dir version at start of readdir so we can tell * if any dentries get dropped */ req->r_dir_release_cnt = atomic64_read(&ci->i_release_count); -- cgit v1.2.3 From 31ca58781019de191c7f520f0626ca76a88c1f6e Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven Date: Thu, 13 Oct 2016 17:15:37 +0200 Subject: ceph: fix uninitialized dentry pointer in ceph_real_mount() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit fs/ceph/super.c: In function ‘ceph_real_mount’: fs/ceph/super.c:818: warning: ‘root’ may be used uninitialized in this function If s_root is already valid, dentry pointer root is never initialized, and returned by ceph_real_mount(). This will cause a crash later when the caller dereferences the pointer. Fixes: ce2728aaa82bbeba ("ceph: avoid accessing / when mounting a subpath") Signed-off-by: Geert Uytterhoeven Signed-off-by: Yan, Zheng --- fs/ceph/super.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/ceph/super.c b/fs/ceph/super.c index a29ffce98187..b382e5910eea 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -845,6 +845,8 @@ static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc) err = ceph_fs_debugfs_init(fsc); if (err < 0) goto fail; + } else { + root = dget(fsc->sb->s_root); } fsc->mount_state = CEPH_MOUNT_MOUNTED; -- cgit v1.2.3 From 5130ccea7cf4646a24c005be1309b7f86f1e91c9 Mon Sep 17 00:00:00 2001 From: Wei Yongjun Date: Mon, 17 Oct 2016 15:11:29 +0000 Subject: ceph: fix non static symbol warning Fixes the following sparse warning: fs/ceph/xattr.c:19:28: warning: symbol 'ceph_other_xattr_handler' was not declared. Should it be static? Signed-off-by: Wei Yongjun Signed-off-by: Ilya Dryomov --- fs/ceph/xattr.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c index 40b703217977..febc28f9e2c2 100644 --- a/fs/ceph/xattr.c +++ b/fs/ceph/xattr.c @@ -16,7 +16,7 @@ static int __remove_xattr(struct ceph_inode_info *ci, struct ceph_inode_xattr *xattr); -const struct xattr_handler ceph_other_xattr_handler; +static const struct xattr_handler ceph_other_xattr_handler; /* * List of handlers for synthetic system.* attributes. Other @@ -1086,7 +1086,7 @@ static int ceph_set_xattr_handler(const struct xattr_handler *handler, return __ceph_setxattr(inode, name, value, size, flags); } -const struct xattr_handler ceph_other_xattr_handler = { +static const struct xattr_handler ceph_other_xattr_handler = { .prefix = "", /* match any name => handlers called with full name */ .get = ceph_get_xattr_handler, .set = ceph_set_xattr_handler, -- cgit v1.2.3