Reliable over NFS? about filelock

[moi90]

SoftFileLock uses O_WRONLY | O_CREAT | O_EXCL | O_TRUNC.

According to man open:

O_CREAT: If pathname does not exist, create it as a regular file.
O_EXCL: Ensure that this call creates the file: if this flag is specified in conjunction with O_CREAT, and pathname already exists, then open() fails with the error EEXIST.
On NFS, O_EXCL is supported only when using NFSv3 or later on kernel 2.6 or later.

This is in compliance with the NFS FAQ D10:

It's worth noting that until early 2.6 kernels, O_EXCL creates were not atomic on Linux NFS clients. Don't use O_EXCL creates and expect atomic behavior among multiple NFS client unless you are running a kernel newer than 2.6.5.

So the answer should be: Yes, on modern systems.

What is the reason for O_TRUNC? It seems to be superfluous, because O_CREAT | O_EXCL prevent the file from being opened if it exists.

O_TRUNC: If the file already exists and is a regular file and the access mode allows writing (i.e., is O_RDWR or O_WRONLY) it will be truncated to length 0.

[andyneff]

I tested this on lusterfs and nfs both the lock mount flag enabled and disabled, and "unit tests passed". I do not know enough if this answers any of the reliability questions, but if the tests are good enough, this might be enough to verify it works on "RHEL 7" modern.

[moi90]

Great! Then I think this can be closed.