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.
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.
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
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.
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.
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.
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>
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
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>.
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 */
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>
(as is done in unmount).
Remove a snapshot inode from the superblock list when its last
name goes away rather than when its last reference goes away.
That way it will be properly reclaimed by fsck after a crash
rather than reenabled when the filesystem is mounted.
options UFS_EXTATTR and UFS_EXTATTR_AUTOSTART respectively. This change
reflects the fact that our EA support is implemented entirely at the
UFS layer (modulo FFS start/stop/autostart hooks for mount and unmount
events). This also better reflects the fact that [shortly] MFS will also
support EAs, as well as possibly IFS.
o Consumers of the EA support in FFS are reminded that as a result, they
must change kernel config files to reflect the new option names.
Obtained from: TrustedBSD Project
"options FFS_EXTATTR". When extended attribute auto-starting
is enabled, FFS will scan the .attribute directory off of the
root of each file system, as it is mounted. If .attribute
exists, EA support will be started for the file system. If
there are files in the directory, FFS will attempt to start
them as attribute backing files for attributes baring the same
name. All attributes are started before access to the file
system is permitted, so this permits race-free enabling of
attributes. For attributes backing support for security
features, such as ACLs, MAC, Capabilities, this is vital, as
it prevents the file system attributes from getting out of
sync as a result of file system operations between mount-time
and the enabling of the extended attribute. The userland
extattrctl tool will still function exactly as previously.
Files must be placed directly in .attribute, which must be
directly off of the file system root: symbolic links are
not permitted. FFS_EXTATTR will continue to be able
to function without FFS_EXTATTR_AUTOSTART for sites that do not
want/require auto-starting. If you're using the UFS_ACL code
available from www.TrustedBSD.org, using FFS_EXTATTR_AUTOSTART
is recommended.
o This support is implemented by adding an invocation of
ufs_extattr_autostart() to ffs_mountfs(). In addition,
several new supporting calls are introduced in
ufs_extattr.c:
ufs_extattr_autostart(): start EAs on the specified mount
ufs_extattr_lookup(): given a directory and filename,
return the vnode for the file.
ufs_extattr_enable_with_open(): invoke ufs_extattr_enable()
after doing the equililent of vn_open()
on the passed file.
ufs_extattr_iterate_directory(): iterate over a directory,
invoking ufs_extattr_lookup() and
ufs_extattr_enable_with_open() on each
entry.
o This feature is not widely tested, and therefore may contain
bugs, caution is advised. Several changes are in the pipeline
for this feature, including breaking out of EA namespaces into
subdirectories of .attribute (this is waiting on the updated
EA API), as well as a per-filesystem flag indicating whether
or not EAs should be auto-started. This is required because
administrators may not want .attribute auto-started on all
file systems, especially if non-administrators have write access
to the root of a file system.
Obtained from: TrustedBSD Project
structure rather than assuming that the device vnode would reside
in the FFS filesystem (which is obviously a broken assumption with
the device filesystem).
An initial tidyup of the mount() syscall and VFS mount code.
This code replaces the earlier work done by jlemon in an attempt to
make linux_mount() work.
* the guts of the mount work has been moved into vfs_mount().
* move `type', `path' and `flags' from being userland variables into being
kernel variables in vfs_mount(). `data' remains a pointer into
userspace.
* Attempt to verify the `type' and `path' strings passed to vfs_mount()
aren't too long.
* rework mount() and linux_mount() to take the userland parameters
(besides data, as mentioned) and pass kernel variables to vfs_mount().
(linux_mount() already did this, I've just tidied it up a little more.)
* remove the copyin*() stuff for `path'. `data' still requires copyin*()
since its a pointer into userland.
* set `mount->mnt_statf_mntonname' in vfs_mount() rather than in each
filesystem. This variable is generally initialised with `path', and
each filesystem can override it if they want to.
* NOTE: f_mntonname is intiailised with "/" in the case of a root mount.
that was introduced in revision 1.80. The problem manifested
itself with a `locking against myself' panic and could also
result in soft updates inconsistences associated with inodedeps.
The two problems are:
1) One of the background operations could manipulate the bitmap
while holding it locked with intent to create. This held lock
results in a `locking against myself' panic, when the background
processing that we have been coopted to do tries to lock the bitmap
which we are already holding locked. To understand how to fix this
problem, first, observe that we can do the background cleanups in
inodedep_lookup only when allocating inodedeps (DEPALLOC is set in
the call to inodedep_lookup). Second observe that calls to
inodedep_lookup with DEPALLOC set can only happen from the following
calls into the softdep code:
softdep_setup_inomapdep
softdep_setup_allocdirect
softdep_setup_remove
softdep_setup_freeblocks
softdep_setup_directory_change
softdep_setup_directory_add
softdep_change_linkcnt
Only the first two of these can come from ffs_alloc.c while holding
a bitmap locked. Thus, inodedep_lookup must not go off to do
request_cleanups when being called from these functions. This change
adds a flag, NODELAY, that can be passed to inodedep_lookup to let
it know that it should not do background processing in those cases.
2) The return value from request_cleanup when helping out with the
cleanup was 0 instead of 1. This meant that despite the fact that
we may have slept while doing the cleanups, the code did not recheck
for the appearance of an inodedep (e.g., goto top in inodedep_lookup).
This lead to the softdep inconsistency in which we ended up with
two inodedep's for the same inode.
Reviewed by: Peter Wemm <peter@yahoo-inc.com>,
Matt Dillon <dillon@earth.backplane.com>
- All processes go into the same array of queues, with different
scheduling classes using different portions of the array. This
allows user processes to have their priorities propogated up into
interrupt thread range if need be.
- I chose 64 run queues as an arbitrary number that is greater than
32. We used to have 4 separate arrays of 32 queues each, so this
may not be optimal. The new run queue code was written with this
in mind; changing the number of run queues only requires changing
constants in runq.h and adjusting the priority levels.
- The new run queue code takes the run queue as a parameter. This
is intended to be used to create per-cpu run queues. Implement
wrappers for compatibility with the old interface which pass in
the global run queue structure.
- Group the priority level, user priority, native priority (before
propogation) and the scheduling class into a struct priority.
- Change any hard coded priority levels that I found to use
symbolic constants (TTIPRI and TTOPRI).
- Remove the curpriority global variable and use that of curproc.
This was used to detect when a process' priority had lowered and
it should yield. We now effectively yield on every interrupt.
- Activate propogate_priority(). It should now have the desired
effect without needing to also propogate the scheduling class.
- Temporarily comment out the call to vm_page_zero_idle() in the
idle loop. It interfered with propogate_priority() because
the idle process needed to do a non-blocking acquire of Giant
and then other processes would try to propogate their priority
onto it. The idle process should not do anything except idle.
vm_page_zero_idle() will return in the form of an idle priority
kernel thread which is woken up at apprioriate times by the vm
system.
- Update struct kinfo_proc to the new priority interface. Deliberately
change its size by adjusting the spare fields. It remained the same
size, but the layout has changed, so userland processes that use it
would parse the data incorrectly. The size constraint should really
be changed to an arbitrary version number. Also add a debug.sizeof
sysctl node for struct kinfo_proc.
mtx_enter(lock, type) becomes:
mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks)
mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized)
similarily, for releasing a lock, we now have:
mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN.
We change the caller interface for the two different types of locks
because the semantics are entirely different for each case, and this
makes it explicitly clear and, at the same time, it rids us of the
extra `type' argument.
The enter->lock and exit->unlock change has been made with the idea
that we're "locking data" and not "entering locked code" in mind.
Further, remove all additional "flags" previously passed to the
lock acquire/release routines with the exception of two:
MTX_QUIET and MTX_NOSWITCH
The functionality of these flags is preserved and they can be passed
to the lock/unlock routines by calling the corresponding wrappers:
mtx_{lock, unlock}_flags(lock, flag(s)) and
mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN
locks, respectively.
Re-inline some lock acq/rel code; in the sleep lock case, we only
inline the _obtain_lock()s in order to ensure that the inlined code
fits into a cache line. In the spin lock case, we inline recursion and
actually only perform a function call if we need to spin. This change
has been made with the idea that we generally tend to avoid spin locks
and that also the spin locks that we do have and are heavily used
(i.e. sched_lock) do recurse, and therefore in an effort to reduce
function call overhead for some architectures (such as alpha), we
inline recursion for this case.
Create a new malloc type for the witness code and retire from using
the M_DEV type. The new type is called M_WITNESS and is only declared
if WITNESS is enabled.
Begin cleaning up some machdep/mutex.h code - specifically updated the
"optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN
and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently
need those.
Finally, caught up to the interface changes in all sys code.
Contributors: jake, jhb, jasone (in no particular order)
filesystem softdep_process_worklist() is called in a loop until it indicates
that no dependancies remain, but the determination of that fact depends on
there only being one softdep_process_worklist() instance running. It was
possible for the syncer to also be running softdep_process_worklist()
and the pre-existing checks in the code to prevent this were not sufficient
to prevent the race. This patch solves the problem.
Approved-by: mckusick
in-core pointers to summary information. An array in this region
(fs_csp) could overflow on filesystems with a very large number of
cylinder groups (~16000 on i386 with 8k blocks). When this happens,
other fields in the superblock get corrupted, and fsck refuses to
check the filesystem.
Solve this problem by replacing the fs_csp array in 'struct fs'
with a single pointer, and add padding to keep the length of the
128-byte region fixed. Update the kernel and userland utilities
to use just this single pointer.
With this change, the kernel no longer makes use of the superblock
fields 'fs_csshift' and 'fs_csmask'. Add a comment to newfs/mkfs.c
to indicate that these fields must be calculated for compatibility
with older kernels.
Reviewed by: mckusick
