Commit Graph

790 Commits

Author SHA1 Message Date
Thomas Moestl
1fbcf0ac65 Call vn_close on the backing file vnode if ufs_extattr_enable failed to
avoid leaking it.

Reviewed by:	rwatson
2001-06-07 00:11:32 +00:00
Jonathan Lemon
3b6e32b01e Add a wrapper for the fifo kqfilter which falls through to the ufs routine.
This permits the fifo to inherit the ufs VNODE kqfilter.
2001-06-06 17:40:57 +00:00
Jonathan Lemon
4e92ec9dd0 Add a kqueue filter for writing to ufs filesystems which always returns
true.  This permits better interoperability with programs which register
filters on their stdin/stdout handles.

Submitted by: Niels Provos <provos@citi.umich.edu>
2001-06-05 13:52:37 +00:00
David E. O'Brien
1239674238 There seems to be a problem that the order of disk write operation being
incorrect due to a missing check for some dependency.  This change
avoids the freelist corruption (but not the temporarily inconsistent
state of the file system).

A message is printed as a reminder of the under lying problem when a
pagedep structure is not freed due to the NEWBLOCK flag being set.

Submitted by:	Tor.Egge@fast.no
2001-06-05 01:49:37 +00:00
John Baldwin
1c11b01562 Revert the previous commit in favor of the fix in rev 1.42 of
ufs/ffs/ffs_extern.h instead.

Requested by:	bde
2001-05-30 23:09:19 +00:00
John Baldwin
55d132317c Forward declare struct cg to quiet a warning.
Submitted by:	bde
2001-05-30 23:08:40 +00:00
John Baldwin
59718ee556 Include <ufs/ffs/fs.h> to get the definition of struct cg to quiet a
warning.
2001-05-29 23:53:16 +00:00
Poul-Henning Kamp
c7a3e2379c Remove last vestiges of MFS. 2001-05-29 21:21:53 +00:00
Poul-Henning Kamp
870b4959b7 Remove MFS from the kernel. 2001-05-29 18:50:30 +00:00
Thomas Moestl
3c436f07b7 Add a check to determine whether extended attributes have been
initialized on the file system before trying to grab the lock of the
per-mount extattr structure, as this lock is unitialized in that case.
This is needed because ufs_extattr_vnode_inactive is called from
ufs_inactive, which is also used by EA-unaware file systems such as
ext2fs.

Reviewed by:	rwatson
2001-05-25 18:24:52 +00:00
Robert Watson
b1fc0ec1a7 o Merge contents of struct pcred into struct ucred. Specifically, add the
real uid, saved uid, real gid, and saved gid to ucred, as well as the
  pcred->pc_uidinfo, which was associated with the real uid, only rename
  it to cr_ruidinfo so as not to conflict with cr_uidinfo, which
  corresponds to the effective uid.
o Remove p_cred from struct proc; add p_ucred to struct proc, replacing
  original macro that pointed.
  p->p_ucred to p->p_cred->pc_ucred.
o Universally update code so that it makes use of ucred instead of pcred,
  p->p_ucred instead of p->p_pcred, cr_ruidinfo instead of p_uidinfo,
  cr_{r,sv}{u,g}id instead of p_*, etc.
o Remove pcred0 and its initialization from init_main.c; initialize
  cr_ruidinfo there.
o Restruction many credential modification chunks to always crdup while
  we figure out locking and optimizations; generally speaking, this
  means moving to a structure like this:
        newcred = crdup(oldcred);
        ...
        p->p_ucred = newcred;
        crfree(oldcred);
  It's not race-free, but better than nothing.  There are also races
  in sys_process.c, all inter-process authorization, fork, exec, and
  exit.
o Remove sigio->sio_ruid since sigio->sio_ucred now contains the ruid;
  remove comments indicating that the old arrangement was a problem.
o Restructure exec1() a little to use newcred/oldcred arrangement, and
  use improved uid management primitives.
o Clean up exit1() so as to do less work in credential cleanup due to
  pcred removal.
o Clean up fork1() so as to do less work in credential cleanup and
  allocation.
o Clean up ktrcanset() to take into account changes, and move to using
  suser_xxx() instead of performing a direct uid==0 comparision.
o Improve commenting in various kern_prot.c credential modification
  calls to better document current behavior.  In a couple of places,
  current behavior is a little questionable and we need to check
  POSIX.1 to make sure it's "right".  More commenting work still
  remains to be done.
o Update credential management calls, such as crfree(), to take into
  account new ruidinfo reference.
o Modify or add the following uid and gid helper routines:
      change_euid()
      change_egid()
      change_ruid()
      change_rgid()
      change_svuid()
      change_svgid()
  In each case, the call now acts on a credential not a process, and as
  such no longer requires more complicated process locking/etc.  They
  now assume the caller will do any necessary allocation of an
  exclusive credential reference.  Each is commented to document its
  reference requirements.
o CANSIGIO() is simplified to require only credentials, not processes
  and pcreds.
o Remove lots of (p_pcred==NULL) checks.
o Add an XXX to authorization code in nfs_lock.c, since it's
  questionable, and needs to be considered carefully.
o Simplify posix4 authorization code to require only credentials, not
  processes and pcreds.  Note that this authorization, as well as
  CANSIGIO(), needs to be updated to use the p_cansignal() and
  p_cansched() centralized authorization routines, as they currently
  do not take into account some desirable restrictions that are handled
  by the centralized routines, as well as being inconsistent with other
  similar authorization instances.
o Update libkvm to take these changes into account.

Obtained from:	TrustedBSD Project
Reviewed by:	green, bde, jhb, freebsd-arch, freebsd-audit
2001-05-25 16:59:11 +00:00
Matthew Dillon
ac8f990bde This patch implements O_DIRECT about 80% of the way. It takes a patchset
Tor created a while ago, removes the raw I/O piece (that has cache coherency
problems), and adds a buffer cache / VM freeing piece.

Essentially this patch causes O_DIRECT I/O to not be left in the cache, but
does not prevent it from going through the cache, hence the 80%.  For
the last 20% we need a method by which the I/O can be issued directly to
buffer supplied by the user process and bypass the buffer cache entirely,
but still maintain cache coherency.

I also have the code working under -stable but the changes made to sys/file.h
may not be MFCable, so an MFC is not on the table yet.

Submitted by:	tegge, dillon
2001-05-24 07:22:27 +00:00
Alfred Perlstein
1752ee59ba ufs_bmaparray() may block on IO, drop vm mutex and aquire Giant when
calling it from the pager routine
2001-05-23 10:30:25 +00:00
Ruslan Ermilov
99d300a1ec - FDESC, FIFO, NULL, PORTAL, PROC, UMAP and UNION file
systems were repo-copied from sys/miscfs to sys/fs.

- Renamed the following file systems and their modules:
  fdesc -> fdescfs, portal -> portalfs, union -> unionfs.

- Renamed corresponding kernel options:
  FDESC -> FDESCFS, PORTAL -> PORTALFS, UNION -> UNIONFS.

- Install header files for the above file systems.

- Removed bogus -I${.CURDIR}/../../sys CFLAGS from userland
  Makefiles.
2001-05-23 09:42:29 +00:00
Kirk McKusick
57042c7f72 Update softdep_setup_directory_add prototype to reflect changes in
actual function.

Obtained from:	Jim Bloom <bloom@jbloom.jbloom.org>
2001-05-20 15:59:55 +00:00
Kirk McKusick
dc01275be9 Must ensure that all the entries on the pd_pendinghd list have been
committed to disk before clearing them. More specifically, when
free_newdirblk is called, we know that the inode claims the new
directory block. However, if the associated pagedep is still linked
onto the directory buffer dependency chain, then some of the entries
on the pd_pendinghd list may not be committed to disk yet. In this
case, we will simply note that the inode claims the block and let
the pd_pendinghd list be processed when the pagedep is next written.
If the pagedep is no longer on the buffer dependency chain, then
all the entries on the pd_pending list are committed to disk and
we can free them in free_newdirblk. This corrects a window of
vulnerability introduced in the code added in version 1.95.
2001-05-19 19:24:26 +00:00
Alfred Perlstein
2395531439 Introduce a global lock for the vm subsystem (vm_mtx).
vm_mtx does not recurse and is required for most low level
vm operations.

faults can not be taken without holding Giant.

Memory subsystems can now call the base page allocators safely.

Almost all atomic ops were removed as they are covered under the
vm mutex.

Alpha and ia64 now need to catch up to i386's trap handlers.

FFS and NFS have been tested, other filesystems will need minor
changes (grabbing the vm lock when twiddling page properties).

Reviewed (partially) by: jake, jhb
2001-05-19 01:28:09 +00:00
Kirk McKusick
9f5192ff71 Must be a bit less aggressive about freeing pagedep structures.
Obtained from:	Robert Watson <rwatson@FreeBSD.org> and
		Matthew Jacob <mjacob@feral.com>
2001-05-18 22:16:28 +00:00
Kirk McKusick
24a83a4b3f When a new block is allocated to a directory, an fsync of a file
whose name is within that block must ensure not only that the block
containing the file name has been written, but also that the on-disk
directory inode references that block. When a new directory block
is created, we allocate a newdirblk structure which is linked to
the associated allocdirect (on its ad_newdirblk list). When the
allocdirect has been satisfied, the newdirblk structure is moved
to the inodedep id_bufwait list of its directory to await the inode
being written.  When the inode is written, the directory entries
are fully committed and can be deleted from their pagedep->id_pendinghd
and inodedep->id_pendinghd lists.
2001-05-17 07:24:03 +00:00
Ian Dowse
0864ef1e8a Change the second argument of vflush() to an integer that specifies
the number of references on the filesystem root vnode to be both
expected and released. Many filesystems hold an extra reference on
the filesystem root vnode, which must be accounted for when
determining if the filesystem is busy and then released if it isn't
busy. The old `skipvp' approach required individual filesystem
xxx_unmount functions to re-implement much of vflush()'s logic to
deal with the root vnode.

All 9 filesystems that hold an extra reference on the root vnode
got the logic wrong in the case of forced unmounts, so `umount -f'
would always fail if there were any extra root vnode references.
Fix this issue centrally in vflush(), now that we can.

This commit also fixes a vnode reference leak in devfs, which could
result in idle devfs filesystems that refuse to unmount.

Reviewed by:	phk, bp
2001-05-16 18:04:37 +00:00
Kirk McKusick
7389126d9a Further fixes for deadlock in the presence of multiple snapshots.
There are still more to find, but this fix should cover the
common cases that folks are hitting.
2001-05-14 17:16:49 +00:00
Kirk McKusick
0b04113700 If the effective link count is zero when an NFS file handle request
comes in for it, the file is really gone, so return ESTALE.

The problem arises when the last reference to an FFS file is
released because soft-updates may delay the actual freeing of the
inode for some time. Since there are no filesystem links or open
file descriptors referencing the inode, from the point of view of
the system, the file is inaccessible. However, if the filesystem
is NFS exported, then the remote client can still access the inode
via ufs_fhtovp() until the inode really goes away. To prevent this
anomoly, it is necessary to begin returning ESTALE at the same time
that the file ceases to be accessible to the local filesystem.

Obtained from:	Ian Dowse <iedowse@maths.tcd.ie>
2001-05-13 23:30:45 +00:00
Kirk McKusick
9b35c30cf7 Remove yet another deadlock case. 2001-05-11 07:12:03 +00:00
Kirk McKusick
9ccb939ef0 When running with soft updates, track the number of blocks and files
that are committed to being freed and reflect these blocks in the
counts returned by statfs (and thus also by the `df' command). This
change allows programs such as those that do news expiration to
know when to stop if they are trying to create a certain percentage
of free space. Note that this change does not solve the much harder
problem of making this to-be-freed space available to applications
that want it (thus on a nearly full filesystem, you may still
encounter out-of-space conditions even though the free space will
show up eventually). Hopefully this harder problem will be the
subject of a future enhancement.
2001-05-08 07:42:20 +00:00
Kirk McKusick
27b047acf0 Several fixes for units errors:
1) Do not assume that the superblock will be of size fs->fs_bsize.
   This fixes a panic when taking a snapshot on a filesystem with
   a block size bigger than 8K.
2) Properly calculate the number of fragments that follow the
   superblock summary information. This fixes a bug with inconsistent
   snapshots.
3) When cleaning up a snapshot that is about to be removed, properly
   calculate the number of blocks that need to be checked. This fixes
   a bug that created partially allocated inodes.
4) When moving blocks from a snapshot that is about to be removed
   to another snapshot, properly account for the reduced number of
   blocks in the snapshot from which they are taken. This fixes a
   bug in which the number of blocks released from a snapshot did not
   match the number that it claimed to have.
2001-05-08 07:29:03 +00:00
Kirk McKusick
0c6fbff0a5 When syncing out snapshot metadata, we must temporarily allow recursive
buffer locking so as to avoid locking against ourselves if we need to
write filesystem metadata.
2001-05-08 07:13:00 +00:00
Kirk McKusick
23371b2f22 Refinement to revision 1.16 of ufs/ffs/ffs_snapshot.c to reduce
the amount of time that the filesystem must be suspended. The
current snapshot is elided as well as the earlier snapshots.
2001-05-04 05:49:28 +00:00
Poul-Henning Kamp
3858e5e797 Use ufs_bmaparray() rather than VOP_BMAP() on our own vnodes. 2001-05-01 09:12:39 +00:00
Poul-Henning Kamp
3c7a8027cb Remove blatantly pointless call to VOP_BMAP().
Use ufs_bmaparray() rather than VOP_BMAP() on our own vnodes.
2001-05-01 09:12:31 +00:00
Poul-Henning Kamp
a62615e59b Implement vop_std{get|put}pages() and add them to the default vop[].
Un-copy&paste all the VOP_{GET|PUT}PAGES() functions which do nothing but
the default.
2001-05-01 08:34:45 +00:00
Mark Murray
fb919e4d5a Undo part of the tangle of having sys/lock.h and sys/mutex.h included in
other "system" header files.

Also help the deprecation of lockmgr.h by making it a sub-include of
sys/lock.h and removing sys/lockmgr.h form kernel .c files.

Sort sys/*.h includes where possible in affected files.

OK'ed by:	bde (with reservations)
2001-05-01 08:13:21 +00:00
Poul-Henning Kamp
855aa097af VOP_BALLOC was never really a VOP in the first place, so convert it
to UFS_BALLOC like the other "between UFS and FFS function interfaces".
2001-04-29 12:36:52 +00:00
Poul-Henning Kamp
b7ebffbc08 Add a vop_stdbmap(), and make it part of the default vop vector.
Make 7 filesystems which don't really know about VOP_BMAP rely
on the default vector, rather than more or less complete local
vop_nopbmap() implementations.
2001-04-29 11:48:41 +00:00
Poul-Henning Kamp
f2ddd13ad2 Call ufs_bmaparray() directly instead of indirectly via VOP_BMAP(). 2001-04-29 10:25:30 +00:00
Poul-Henning Kamp
954a0e256e Remove two unused arguments from ufs_bmaparray(). 2001-04-29 10:24:58 +00:00
Poul-Henning Kamp
e955479077 Remove faint traces of blind copy&paste. 2001-04-29 10:23:50 +00:00
Poul-Henning Kamp
0c25dbeb17 Remove faint traces of non-existant ffs_bmap(). 2001-04-29 10:23:32 +00:00
Greg Lehey
60fb0ce365 Revert consequences of changes to mount.h, part 2.
Requested by:	bde
2001-04-29 02:45:39 +00:00
Kirk McKusick
c9509f5865 Rather than copying all the indirect blocks of the snapshot,
simply mark them as BLK_NOCOPY. This trick cuts the initial
size of the snapshot in half and cuts the time to take a
snapshot by a third.
2001-04-26 00:50:53 +00:00
Kirk McKusick
112f737245 When closing the last reference to an unlinked file, it is freed
by the inactive routine. Because the freeing causes the filesystem
to be modified, the close must be held up during periods when the
filesystem is suspended.

For snapshots to be consistent across crashes, they must write
blocks that they copy and claim those written blocks in their
on-disk block pointers before the old blocks that they referenced
can be allowed to be written.

Close a loophole that allowed unwritten blocks to be skipped when
doing ffs_sync with a request to wait for all I/O activity to be
completed.
2001-04-25 08:11:18 +00:00
Poul-Henning Kamp
a13234bb35 Move the netexport structure from the fs-specific mountstructure
to struct mount.

This makes the "struct netexport *" paramter to the vfs_export
and vfs_checkexport interface unneeded.

Consequently that all non-stacking filesystems can use
vfs_stdcheckexp().

At the same time, make it a pointer to a struct netexport
in struct mount, so that we can remove the bogus AF_MAX
and #include <net/radix.h> from <sys/mount.h>
2001-04-25 07:07:52 +00:00
Ian Dowse
5d69bac493 Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by:	mckusick
2001-04-24 00:37:16 +00:00
Greg Lehey
d98dc34f52 Correct #includes to work with fixed sys/mount.h. 2001-04-23 09:05:15 +00:00
Poul-Henning Kamp
f84e29a06c This patch removes the VOP_BWRITE() vector.
VOP_BWRITE() was a hack which made it possible for NFS client
side to use struct buf with non-bio backing.

This patch takes a more general approach and adds a bp->b_op
vector where more methods can be added.

The success of this patch depends on bp->b_op being initialized
all relevant places for some value of "relevant" which is not
easy to determine.  For now the buffers have grown a b_magic
element which will make such issues a tiny bit easier to debug.
2001-04-17 08:56:39 +00:00
Kirk McKusick
5819ab3f12 Add debugging option to always read/write cylinder groups as full
sized blocks. To enable this option, use: `sysctl -w debug.bigcgs=1'.
Add debugging option to disable background writes of cylinder
groups. To enable this option, use: `sysctl -w debug.dobkgrdwrite=0'.
These debugging options should be tried on systems that are panicing
with corrupted cylinder group maps to see if it makes the problem
go away. The set of panics in question are:

	ffs_clusteralloc: map mismatch
	ffs_nodealloccg: map corrupted
	ffs_nodealloccg: block not in map
	ffs_alloccg: map corrupted
	ffs_alloccg: block not in map
	ffs_alloccgblk: cyl groups corrupted
	ffs_alloccgblk: can't find blk in cyl
	ffs_checkblk: partially free fragment

The following panics are less likely to be related to this problem,
but might be helped by these debugging options:

	ffs_valloc: dup alloc
	ffs_blkfree: freeing free block
	ffs_blkfree: freeing free frag
	ffs_vfree: freeing free inode

If you try these options, please report whether they helped reduce your
bitmap corruption panics to Kirk McKusick at <mckusick@mckusick.com>
and to Matt Dillon <dillon@earth.backplane.com>.
2001-04-17 05:37:51 +00:00
Kirk McKusick
f0f3f19f05 Background fsck sysctl operations must use vn_start_write and
vn_finished_write so that they do not attempt to modify a
suspended filesystem.
2001-04-17 05:06:37 +00:00
Robert Watson
b114e127e6 In my first reading of POSIX.1e, I misinterpreted handling of the
ACL_USER_OBJ and ACL_GROUP_OBJ fields, believing that modification of the
access ACL could be used by privileged processes to change file/directory
ownership.  In fact, this is incorrect; ACL_*_OBJ (+ ACL_MASK and
ACL_OTHER) should have undefined ae_id fields; this commit attempts
to correct that misunderstanding.

o Modify arguments to vaccess_acl_posix1e() to accept the uid and gid
  associated with the vnode, as those can no longer be extracted from
  the ACL passed as an argument.  Perform all comparisons against
  the passed arguments.  This actually has the effect of simplifying
  a number of components of this call, as well as reducing the indent
  level, but now seperates handling of ACL_GROUP_OBJ from ACL_GROUP.

o Modify acl_posix1e_check() to return EINVAL if the ae_id field of
  any of the ACL_{USER_OBJ,GROUP_OBJ,MASK,OTHER} entries is a value
  other than ACL_UNDEFINED_ID.  As a temporary work-around to allow
  clean upgrades, set the ae_id field to ACL_UNDEFINED_ID before
  each check so that this cannot cause a failure in the short term
  (this work-around will be removed when the userland libraries and
  utilities are updated to take this change into account).

o Modify ufs_sync_acl_from_inode() so that it forces
  ACL_{USER_OBJ,GROUP_OBJ,MASK,OTHER} ae_id fields to ACL_UNDEFINED_ID
  when synchronizing the ACL from the inode.

o Modify ufs_sync_inode_from_acl to not propagate uid and gid
  information to the inode from the ACL during ACL update.  Also
  modify the masking of permission bits that may be set from
  ALLPERMS to (S_IRWXU|S_IRWXG|S_IRWXO), as ACLs currently do not
  carry none-ACCESSPERMS (S_ISUID, S_ISGID, S_ISTXT).

o Modify ufs_getacl() so that when it emulates an access ACL from
  the inode, it initializes the ae_id fields to ACL_UNDEFINED_ID.

o Clean up ufs_setacl() substantially since it is no longer possible
  to perform chown/chgrp operations using vop_setacl(), so all the
  access control for that can be eliminated.

o Modify ufs_access() so that it passes owner uid and gid information
  into vaccess_acl_posix1e().

Pointed out by:	jedger
Obtained from:	TrustedBSD Project
2001-04-17 04:33:34 +00:00
Kirk McKusick
74046077a7 Update to describe use of mdconfig instead of deprecated vnconfig.
Submitted by:	Steve Ames <steve@virtual-voodoo.com>
2001-04-14 18:32:09 +00:00
Kirk McKusick
1a6a661032 This checkin adds support in ufs/ffs for the FS_NEEDSFSCK flag.
It is described in ufs/ffs/fs.h as follows:

/*
 * Filesystem flags.
 *
 * Note that the FS_NEEDSFSCK flag is set and cleared only by the
 * fsck utility. It is set when background fsck finds an unexpected
 * inconsistency which requires a traditional foreground fsck to be
 * run. Such inconsistencies should only be found after an uncorrectable
 * disk error. A foreground fsck will clear the FS_NEEDSFSCK flag when
 * it has successfully cleaned up the filesystem. The kernel uses this
 * flag to enforce that inconsistent filesystems be mounted read-only.
 */
#define FS_UNCLEAN    0x01	/* filesystem not clean at mount */
#define FS_DOSOFTDEP  0x02	/* filesystem using soft dependencies */
#define FS_NEEDSFSCK  0x04	/* filesystem needs sync fsck before mount */
2001-04-14 05:26:28 +00:00
Kirk McKusick
a61ab64ac4 Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

  One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

  First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
   test is at wd1. Size of test file system is 8 Gb, number of cg=991,
   size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
   from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
   at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
   number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
   OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

                              Test Results

             tar -xzf ports.tar.gz               rm -rf ports
  mode  old dirpref new dirpref speedup old dirprefnew dirpref speedup
                             First system
 normal     667         472      1.41       477        331       1.44
 async      285         144      1.98       130         14       9.29
 sync       768         616      1.25       477        334       1.43
 softdep    413         252      1.64       241         38       6.34
                             Second system
 normal     329         81       4.06       263.5       93.5     2.81
 async      302         25.7    11.75       112          2.26   49.56
 sync       281         57.0     4.93       263         90.5     2.9
 softdep    341         40.6     8.4        284          4.76   59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
 * Find a cylinder to place a directory.
 *
 * The policy implemented by this algorithm is to select from
 * among those cylinder groups with above the average number of
 * free inodes, the one with the smallest number of directories.
 */

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

  1. A big filesystem is a filesystem which occupy 20-30 or more percent
     of total drive space, i.e. first and last cylinder are physically
     located relatively far from each other.
  2. It has a relatively large number of cylinder groups, for example
     more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
 * Find a cylinder group to place a directory.
 *
 * The policy implemented by this algorithm is to allocate a
 * directory inode in the same cylinder group as its parent
 * directory, but also to reserve space for its files inodes
 * and data. Restrict the number of directories which may be
 * allocated one after another in the same cylinder group
 * without intervening allocation of files.
 *
 * If we allocate a first level directory then force allocation
 * in another cylinder group.
 */

  My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

  My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

  The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

        int32_t  fs_avgfilesize;   /* expected average file size */
        int32_t  fs_avgfpdir;      /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from:	Grigoriy Orlov <gluk@ptci.ru>
2001-04-10 08:38:59 +00:00