[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gfs-devel
Subject:    Re: New GFS Release
From:       "Matthew O'Keefe" <okeefe () lcse ! umn ! edu>
Date:       1999-05-08 13:08:08
[Download RAW message or body]


Many thanks to Ken and all the people who contributed 
(both testing and source code) to this
latest release of GFS.

Matt O'Keefe

> 
> There is a new GFS release on the ftp site (ftp://gfs.lcse.umn.edu/pub/GFS).
> 
> There are major changes in the way GFS does locking.  There are now locking
> modules that plug into GFS and let it use whatever locking protocol you wish
> to implement.  In this release, there are locking modules for the old Dlocks,
> the new Dlocks, and dummy locks (for local filesystems).  (Our IP based lock
> server is almost ready.  The server and a locking module for it will be
> included in the next release.)
> 
> Firmware that implements the Version 0.9.4 Dlock specification on Seagate 
> drives is on the ftp site.  For a list of the drive model it works on, see the
> file ftp://gfs.lcse.umn.edu/pub/GFS/firmware/README.
> 
> >From the ChangeLog:
> 
> Release Code Name:  Antimatter Woodchuck
> Requires Linux version 2.2.7+
> Major rewrite of the locking code.  GFS now has an abstract interface to 
> 	locking modules that can implement any locking protocol you want.  In
> 	this release, there are modules that allow GFS to work with the old
> 	Dlocks (Version 0.6), the new Dlocks (Version 0.9.4), and nolocks
> 	(for a local filesystem).  A description of the locking interface can
> 	be found in GFS2/src/fs/gfs_locking.h.
> When used with the dlock-0.9.4 and nolock modules, GFS now does full write
> 	caching.  Also supports reader/writer locks.
> Mkfs_gfs now takes the -p and -t options to specify locking module to be used.
> 	(See GFS2/INSTALLATION for instructions.)
> Directory entries are now variable length.  Directories are much more 
> 	efficient.
> Implemented lock sorting to avoid locking deadlock.  (Much better than the old
> 	backoff-and-retry.)
> Added new kernel thread, inoded, to avoid deadlock in clearing inodes.
> Improved the way bitmaps are read and written from disk.
> Fixed & instead of % typo in the block allocation routines.
> Improved method of handling structure packing.
> Added new locking code to Pool to handle the new Dlock command.  Pool now just
> 	maps SCSI CDBs and issues the commands.  It doesn't understand the
> 	Dlock protocol any more.
> Neither Pool nor GFS automatically allocate memory in proportion to the number
> 	of Dlocks in the system anymore.
> Pool will now work on IDE and loop devices (with no Dlocks).
> Mkfs_gfs was changed so GFS can be used without Pool.  (Pool is still needed
> 	if Dlocks are used, though.)
> Added code for a few new user-space tools that manipulate the old and new
> 	Dlocks.
> 
> 
> 
> 
> To unsubscribe from this list: send the line "unsubscribe gfs-announce" in
> the body of a message to majordomo@lcse.umn.edu
> 


To unsubscribe from this list: send the line "unsubscribe gfs-devel" in
the body of a message to majordomo@lcse.umn.edu

[prev in list] [next in list] [prev in thread] [next in thread]