kernel.git/fs/cifs/readdir.c, branch linux-4.12.y

cifs: initialize file_info_lock

2017-01-14T20:58:29Z

Reviewed-by: Jeff Layton CC: Stable file_info_lock is not initalized in initiate_cifs_search(), leading to the following splat after a simple "mount.cifs ... dir && ls dir/": BUG: spinlock bad magic on CPU#0, ls/486 lock: 0xffff880009301110, .magic: 00000000, .owner: /-1, .owner_cpu: 0 CPU: 0 PID: 486 Comm: ls Not tainted 4.9.0 #27 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996) ffffc900042f3db0 ffffffff81327533 0000000000000000 ffff880009301110 ffffc900042f3dd0 ffffffff810baf75 ffff880009301110 ffffffff817ae077 ffffc900042f3df0 ffffffff810baff6 ffff880009301110 ffff880008d69900 Call Trace: [] dump_stack+0x65/0x92 [] spin_dump+0x85/0xe0 [] spin_bug+0x26/0x30 [] do_raw_spin_lock+0xe9/0x130 [] _raw_spin_lock+0x1f/0x30 [] cifs_closedir+0x4d/0x100 [] __fput+0x5d/0x160 [] ____fput+0xe/0x10 [] task_work_run+0x7e/0xa0 [] exit_to_usermode_loop+0x92/0xa0 [] syscall_return_slowpath+0x49/0x50 [] entry_SYSCALL_64_fastpath+0xa7/0xa9 Fixes: 3afca265b5f53a0 ("Clarify locking of cifs file and tcon structures and make more granular") Signed-off-by: Rabin Vincent Signed-off-by: Steve French

Clarify locking of cifs file and tcon structures and make more granular

2016-10-12T17:08:32Z

Remove the global file_list_lock to simplify cifs/smb3 locking and have spinlocks that more closely match the information they are protecting. Add new tcon->open_file_lock and file->file_info_lock spinlocks. Locks continue to follow a heirachy, cifs_socket --> cifs_ses --> cifs_tcon --> cifs_file where global tcp_ses_lock still protects socket and cifs_ses, while the the newer locks protect the lower level structure's information (tcon and cifs_file respectively). CC: Stable Signed-off-by: Steve French Signed-off-by: Pavel Shilovsky Reviewed-by: Aurelien Aptel Reviewed-by: Germano Percossi

cifs: Use file_dentry()

2016-05-17T19:09:33Z

CIFS may be used as lower layer of overlayfs and accessing f_path.dentry can lead to a crash. Fix by replacing direct access of file->f_path.dentry with the file_dentry() accessor, which will always return a native object. Signed-off-by: Goldwyn Rodrigues Acked-by: Shirish Pargaonkar Signed-off-by: Steve French

cifs: switch to ->iterate_shared()

2016-05-02T23:49:31Z

Signed-off-by: Al Viro

don't bother with ->d_inode->i_sb - it's always equal to ->d_sb

2016-04-10T21:11:51Z

... and neither can ever be NULL Signed-off-by: Al Viro

cifs_dbg() outputs an uninitialized buffer in cifs_readdir()

2016-01-14T20:45:49Z

In some cases tmp_bug can be not filled in cifs_filldir and stay uninitialized, therefore its printk with "%s" modifier can leak content of kernelspace memory. If old content of this buffer does not contain '\0' access bejond end of allocated object can crash the host. Signed-off-by: Vasily Averin Signed-off-by: Steve French CC: Stable

fs: Drop unlikely before IS_ERR(_OR_NULL)

2015-09-29T13:13:58Z

IS_ERR(_OR_NULL) already contain an 'unlikely' compiler flag and there is no need to do that again from its callers. Drop it. Signed-off-by: Viresh Kumar Reviewed-by: Jeff Layton Reviewed-by: David Howells Reviewed-by: Steve French Signed-off-by: Jiri Kosina

cifs: Don't replace dentries for dfs mounts

2015-05-11T00:56:05Z

Doing a readdir on a dfs root can result in the dentries for directories with a dfs share mounted being replaced by new dentries for objects returned by the readdir call. These new dentries on shares mounted with unix extenstions show up as symlinks pointing to the dfs share. # mount -t cifs -o sec=none //vm140-31/dfsroot cifs # stat cifs/testlink/testfile; ls -l cifs File: ‘cifs/testlink/testfile’ Size: 0 Blocks: 0 IO Block: 16384 regular empty file Device: 27h/39d Inode: 130120 Links: 1 Access: (0644/-rw-r--r--) Uid: ( 0/ root) Gid: ( 0/ root) Access: 2015-03-31 13:55:50.106018200 +0100 Modify: 2015-03-31 13:55:50.106018200 +0100 Change: 2015-03-31 13:55:50.106018200 +0100 Birth: - total 0 drwxr-xr-x 2 root root 0 Mar 31 13:54 testdir lrwxrwxrwx 1 root root 19 Mar 24 14:25 testlink -> \vm140-31\test In the example above, the stat command mounts the dfs share at cifs/testlink. The subsequent ls on the dfsroot directory replaces the dentry for testlink with a symlink. In the earlier code, the d_invalidate command returned an -EBUSY error when attempting to invalidate directories. This stopped the code from replacing the directories with symlinks returned by the readdir call. Changes were recently made to the d_invalidate() command so that it no longer returns an error code. This results in the directory with the mounted dfs share being replaced by a symlink which denotes a dfs share. Signed-off-by: Sachin Prabhu Reviewed-by: Jeff Layton Signed-off-by: Steve French

VFS: normal filesystems (and lustre): d_inode() annotations

2015-04-15T19:06:57Z

that's the bulk of filesystem drivers dealing with inodes of their own Signed-off-by: David Howells Signed-off-by: Al Viro

cifs: make new inode cache when file type is different

2014-12-22T20:16:21Z

In spite of different file type, if file is same name and same inode number, old inode cache is used. This causes that you can not cd directory, can not cat SymbolicLink. So this patch is that if file type is different, return error. Reproducible sample : 1. create file 'a' at cifs client. 2. repeat rm and mkdir 'a' 4 times at server, then direcotry 'a' having same inode number is created. (Repeat 4 times, then same inode number is recycled.) (When server is under RHEL 6.6, 1 time is O.K. Always same inode number is recycled.) 3. ls -li at client, then you can not cd directory, can not remove directory. SymbolicLink has same problem. Bug link: https://bugzilla.kernel.org/show_bug.cgi?id=90011 Signed-off-by: Nakajima Akira Acked-by: Jeff Layton Signed-off-by: Steve French