Re: struct file at VFS level (round 1.5)

[Christoph Badura <bad%bsd.de@localhost>]

On Tue, May 03, 2016 at 05:33:34AM +0200, Emmanuel Dreyfus wrote:
> Christoph Badura <bad%bsd.de@localhost> wrote:
> 
> > Mandatory file locking simply is a SMOP for the kernel to add a flag to a
> > vnode that indicates whether mandatory locking is in effect and adding the
> > necessary checks and interaction with file systems to open(2), read(2),
> > write(2) etc. at the syscall level.  VOP_ADVLOCK is the right place to
> > handle the necessary interaction with the file systems.[3]
> 
> Please tell me if I understood your point of view correctly: you mean
> that mandatory locks should be implemented above VFS, by turning for
> instance a write operation into VOP_ADVLOCK then on success VOP_WRITE?

Yes.  Mandatory locking is pretty much like:

 - for the read case
   off = current_file_offset
   VOP_ADVLOCK(F_RDLCK, off, size_of_read)
   VOP_READ(...)
   VOP_ADVLOCK(F_UNLCK, off, size_of_read)

 - for the write case
   off = current_file_offset
   VOP_ADVLOCK(F_WRLCK, off, size_of_write)
   VOP_WRITE(...)
   VOP_ADVLOCK(F_UNLCK, off, size_of_write)
   
Except that it is unclear to me how the F_UNLCK interacts with overlapping
read locks set explicitly by the process.  I.e. is the pre-existing read-lock
split or truncated?  But I guess it is OK to have the same lossage as with
advisory locks.

--chris