HP CIFS File Locking Interoperation

10
An important detail is that Oplocks are invoked by the file system, not an application API.
Therefore, an application can close an oplocked file, but the file system does not relinquish
the Oplock. When the Oplock break is issued, the file system then simply closes the file in
preparation for the subsequent open by the second process.
4.2 UNIX
UNIX utilizes 2 kinds of file locking:
Advisory locking
Mandatory locking
Both lock types are implemented with the same function calls: fcntl and lockf. fcntl is the
POSIX interface (introduced at SVR4), and lockf is provided for backward compatibility. The
calling semantics are slightly different, but the functionality is basically the same.
Both functions specify read and/or write locking, and allow for byte-range locks or locking of
the entire file. A read lock allows concurrent access for reading, but not writing. A write
lock is exclusive.
ADVISORY LOCKING assumes that all processes attempting to access a file participate in
the locking protocol by calling the locking functions correctly and honoring the lock status of
the file. There is no requirement that a process must adhere to the advisory locking protocol.
A process that does not participate in the protocol can obliviously write or read to a file that
is locked with locking functions. Therefore there is a significant assumption of risk of data
corruption with advisory locking.
MANDATORY LOCKING is implemented with function calls identical to advisory locking.
Mandatory locking is then enabled per-file by manipulating the file’s permissions with
chmod. To enable a file for mandatory locking, the group-id bit must be set, and the execute
bit must be cleared (implying that executable files cannot be locked). Once set, any calls to
open, read, or write to a previously locked file are failed by the kernel, regardless of whether
the process participates in the file locking protocol. Most UNIX locking scenarios refer to
advisory locking protocol. The UNIX mandatory option seems to be rarely utilized, so further
references to mandatory locking refer to Windows only.
Note that UNIX does not employ the concept of Mandatory Share Mode (Open Mode) locking
like Windows does. UNIX locking consists of byte range locking and is strictly programmatic
on a previously opened file. Also, UNIX is incapable of sending an oplock break to a client,
and therefore cannot interoperate with Windows clients using Opportunistic Locking.
4.3 NFS
File locking for NFS is provided by two cooperating protocols:
Network Lock Manager
Network Status Monitor