freebsd-nq/sys/ufs/ufs
Konstantin Belousov 088ffd2086 Implement fine-grained locking for UFS quotas.
Each struct dquot gets dq_lock mutex to protect dq_flags and to interlock
with DQ_LOCK. qhash, dqfreelist and dq.dq_cnt are protected by global
dqhlock mutex.

i_dquot array for inode is protected by lockmgr' vnode lock, corresponding
assert added to the dqget(). Access to struct ufsmount quota-related fields
(um_quotas and um_qflags) is protected by um_lock.

Tested by:	Peter Holm
Reviewed by:	tegge
Approved by:	re (kensmith)

This work were not possible without enormous amount of help given by
Tor Egge and Peter Holm. Tor reviewed each version of patch, pointed out
numerous errors and provided invaluable suggestions. Peter did tireless
testing of the patch as it was developed.
2007-03-14 08:54:08 +00:00
..
acl.h
dinode.h o Fix a comment: ufs2_dinode.di_blocks counts blocks not bytes actually held. 2006-05-21 21:55:29 +00:00
dir.h /* -> /*- for license, minor formatting changes 2005-01-07 02:29:27 +00:00
dirhash.h /* -> /*- for license, minor formatting changes 2005-01-07 02:29:27 +00:00
extattr.h Move macros describing extended attributes in UFS from 2007-03-06 08:13:21 +00:00
gjournal.h Add gjournal specific code to the UFS file system: 2006-10-31 21:48:54 +00:00
inode.h Do not translate the IN_ACCESS inode flag into the IN_MODIFIED while filesystem 2006-10-10 09:20:54 +00:00
quota.h Implement fine-grained locking for UFS quotas. 2007-03-14 08:54:08 +00:00
README.acls
README.extattr
ufs_acl.c Canonicalize copyright: use a date range rather than comma-delimited 2007-01-08 17:55:32 +00:00
ufs_bmap.c Remove an unneeded "a" from comment. 2005-10-25 19:46:15 +00:00
ufs_dirhash.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
ufs_extattr.c Sweep kernel replacing suser(9) calls with priv(9) calls, assigning 2006-11-06 13:42:10 +00:00
ufs_extern.h Use vfs_hash instead of home-rolled. 2005-03-14 10:21:16 +00:00
ufs_gjournal.c Add gjournal specific code to the UFS file system: 2006-10-31 21:48:54 +00:00
ufs_inode.c Add gjournal specific code to the UFS file system: 2006-10-31 21:48:54 +00:00
ufs_lookup.c Call getinoquota() before allocating new block for the directory to properly 2007-03-14 08:50:27 +00:00
ufs_quota.c Implement fine-grained locking for UFS quotas. 2007-03-14 08:54:08 +00:00
ufs_vfsops.c Prevent quotactl calls that pass in an id of -1 from incorrectly 2007-02-01 02:13:53 +00:00
ufs_vnops.c Remove unneeded getinoquota() call in the ufs_access(). 2007-03-14 08:48:57 +00:00
ufsmount.h Eliminate softdep_flush() livelock by accounting for number of worklist items 2006-04-03 22:23:23 +00:00

$FreeBSD$

  UFS Extended Attributes Copyright

The UFS Extended Attributes implementation is copyright Robert Watson, and
is made available under a Berkeley-style license.

  About UFS Extended Attributes

Extended attributes allow the association of additional arbitrary
meta-data with files and directories.  Extended attributes are defined in
the form name=value, where name is an nul-terminated string in the style
of a filename, and value is a binary blob of zero or more bytes. The UFS
extended attribute service layers support for extended attributes onto a
backing file, in the style of the quota implementation, meaning that it
requires no underlying format changes in the filesystem.  This design
choice exchanges simplicity, usability and easy deployment for
performance.  When defined, extended attribute names exist in a series of
disjoint namespaces: currently, two namespaces are defined:
EXTATTR_NAMESPACE_SYSTEM and EXTATTR_NAMESPACE_USER.  The primary
distinction lies in the protection model: USER EAs are protected using the
normal inode protections, whereas SYSTEM EAs require privilege to access
or modify.

  Using UFS Extended Attributes

Support for UFS extended attributes is natively available in UFS2, and
requires no special configuration.  For reliability, administrative,
and performance reasons, if you plan to use extended attributes, it
is recommended that you use UFS2 in preference to UFS1.

Support for UFS extended attributes may be enabled for UFS1 by adding:

	options UFS_EXTATTR

to your kernel configuration file.  This allows UFS-based filesystems to
support extended attributes, but requires manual administration of EAs
using the extattrctl tool, including the starting of EA support for each
filesystem, and the enabling of individual attributes for the file
system.  The extattrctl utility may be used to initialize backing files
before first use, to start and stop EA service on a filesystem, and to
enable and disable named attributes.  The command lines for extattrctl
take the following forms:

  extattrctl start [path]
  extattrctl stop [path]
  extattrctl initattr [-f] [-p path] [attrsize] [attrfile]
  extattrctl enable [path] [attrnamespace] [attrname] [attrfile]
  extattrctl disable [path] [attrnamespace] [attrname]

In each case, [path] is used to indicate the mounted filesystem on which
to perform the operation.  [attrnamespace] refers to the namespace in
which the attribute is being manipulated, and may be "system" or "user".  
The [attrname] is the attribute name to use for the operation. The
[attrfile] argument specifies the attribute backing file to use. When
using the "initattr" function to initialize a backing file, the maximum
size of attribute data must be defined in bytes using the [attrsize]
field.  Optionally, the [-p path] argument may be used to indicate to
extattrctl that it should pre-allocate space for EA data, rather than
creating a sparse backing file.  This prevents attribute operations from
failing in low disk-space conditions (which can be important when EAs are
used for security purposes), but pre-allocation will consume space
proportional to the product of the defined maximum attribute size and
number of attributes on the specified filesystem.

Manual configuration increases administrative overhead, but also
introduces the possibility of race conditions during filesystem mount, if
EAs are used to support other features, as starting the EAs manually is
not atomic with the mount operation.  To address this problem, an
additional kernel option may be defined to auto-start EAs on a UFS file
system based on special directories at mount-time:

	options UFS_EXTATTR_AUTOSTART

If this option is defined, UFS will search for a ".attribute"
sub-directory of the filesystem root during the mount operation.  If it
is found, EA support will be started for the filesystem.  UFS will then
search for "system" and "user" sub-directories of the ".attribute"
directory for any potential backing files, and enable an EA for each valid
backing file with the name of the backing file as the attribute name.  
For example, by creating the following tree, the two EAs,
posix1e.acl_access and posix1e.acl_default will be enabled in the system
namespace of the root filesystem, reserving space for attribute data:

  mkdir -p /.attribute/system
  cd /.attribute/system
  extattrctl initattr -p / 388 posix1e.acl_access
  extattrctl initattr -p / 388 posix1e.acl_default

On the next mount of the root filesystem, the attributes will be
automatically started.