As Al Viro pointed out path resolution during Q_QUOTAON calls to quotactl
is prone to deadlocks. We hold s_umount semaphore for reading during the
path resolution and resolution itself may need to acquire the semaphore
for writing when e. g. autofs mountpoint is passed.
Solve the problem by performing the resolution before we get hold of the
superblock (and thus s_umount semaphore). The whole thing is complicated
by the fact that some filesystems (OCFS2) ignore the path argument. So to
distinguish between filesystem which want the path and which do not we
introduce new .quota_on_meta callback which does not get the path. OCFS2
then uses this callback instead of old .quota_on.
CC: Al Viro <viro@ZenIV.linux.org.uk>
CC: Christoph Hellwig <hch@lst.de>
CC: Ted Ts'o <tytso@mit.edu>
CC: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Jan Kara <jack@suse.cz>
RCU free the struct inode. This will allow:
- Subsequent store-free path walking patch. The inode must be consulted for
permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
to take i_lock no longer need to take sb_inode_list_lock to walk the list in
the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
page lock to follow page->mapping.
The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.
In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.
The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The big kernel lock has been removed from all these files at some point,
leaving only the #include.
Remove this too as a cleanup.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, the default behavior of O_DIRECT writes was allowing
concurrent writing among nodes to the same file, with no cluster
coherency guaranteed (no EX lock held). This can leave stale data in
the cache for buffered reads on other nodes.
The new mount option introduce a chance to choose two different
behaviors for O_DIRECT writes:
* coherency=full, as the default value, will disallow
concurrent O_DIRECT writes by taking
EX locks.
* coherency=buffered, allow concurrent O_DIRECT writes
without EX lock among nodes, which
gains high performance at risk of
getting stale data on other nodes.
Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Functions such as ocfs2_recovery_init() make use of osb->max_slots.
Initialize osb->max_slots early so the functions may use the correct
value.
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.de>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The BKL in ocfs2/dlmfs is used in put_super, fill_super and remount_fs
that are all three protected by the superblocks s_umount rw_semaphore.
The use in ocfs2_control_open is evidently unrelated and the function
is protected by ocfs2_control_lock.
Therefore it is safe to remove the BKL entirely.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
This patch is a preparation necessary to remove the BKL from do_new_mount().
It explicitly adds calls to lock_kernel()/unlock_kernel() around
get_sb/fill_super operations for filesystems that still uses the BKL.
I've read through all the code formerly covered by the BKL inside
do_kern_mount() and have satisfied myself that it doesn't need the BKL
any more.
do_kern_mount() is already called without the BKL when mounting the rootfs
and in nfsctl. do_kern_mount() calls vfs_kern_mount(), which is called
from various places without BKL: simple_pin_fs(), nfs_do_clone_mount()
through nfs_follow_mountpoint(), afs_mntpt_do_automount() through
afs_mntpt_follow_link(). Both later functions are actually the filesystems
follow_link inode operation. vfs_kern_mount() is calling the specified
get_sb function and lets the filesystem do its job by calling the given
fill_super function.
Therefore I think it is safe to push down the BKL from the VFS to the
low-level filesystems get_sb/fill_super operation.
[arnd: do not add the BKL to those file systems that already
don't use it elsewhere]
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Christoph Hellwig <hch@infradead.org>
Durring orphan scan, if we are slot 0, and we are replaying
orphan_dir:0001, the general process is that for every file
in this dir:
1. we will iget orphan_dir:0001, since there is no inode for it.
we will have to create an inode and read it from the disk.
2. do the normal work, such as delete_inode and remove it from
the dir if it is allowed.
3. call iput orphan_dir:0001 when we are done. In this case,
since we have no dcache for this inode, i_count will
reach 0, and VFS will have to call clear_inode and in
ocfs2_clear_inode we will checkpoint the inode which will let
ocfs2_cmt and journald begin to work.
4. We loop back to 1 for the next file.
So you see, actually for every deleted file, we have to read the
orphan dir from the disk and checkpoint the journal. It is very
time consuming and cause a lot of journal checkpoint I/O.
A better solution is that we can have another reference for these
inodes in ocfs2_super. So if there is no other race among
nodes(which will let dlmglue to checkpoint the inode), for step 3,
clear_inode won't be called and for step 1, we may only need to
read the inode for the 1st time. This is a big win for us.
So this patch will try to cache system inodes of other slots so
that we will have one more reference for these inodes and avoid
the extra inode read and journal checkpoint.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The OCFS2 developers have already done all of the hard work to allow
volumes larger than 16 TiB. But there is still a "sanity check" in
fs/ocfs2/super.c that prevents the mounting of such volumes, even when
the cluster size and journal options would allow it.
This patch replaces that sanity check with a more sophisticated one to
mount a huge volume provided that (a) it is addressable by the raw
word/address size of the system (borrowing a test from ext4); (b) the
volume is using JBD2; and (c) the JBD2_FEATURE_INCOMPAT_64BIT flag is
set on the journal.
I factored out the sanity check into its own function. I also moved it
from ocfs2_initialize_super() down to ocfs2_check_volume(); any earlier,
and the journal will not have been initialized yet.
This patch is one of a pair, and it depends on the other ("JBD2: Allow
feature checks before journal recovery").
I have tested this patch on small volumes, huge volumes, and huge
volumes without 64-bit block support in the journal. All of them appear
to work or to fail gracefully, as appropriate.
Signed-off-by: Patrick LoPresti <lopresti@gmail.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Follow the dquot_* style used elsewhere in dquot.c.
[Jan Kara: Fixed up missing conversion of ext2]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Remount handling has fully moved into the filesystem, so all this is
superflous now.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Instead of having wrappers in the VFS namespace export the dquot_suspend
and dquot_resume helpers directly. Also rename vfs_quota_disable to
dquot_disable while we're at it.
[Jan Kara: Moved dquot_suspend to quotaops.h and made it inline]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
We also have to cancel quota syncing thread on remount read only because
at that moment quota is being turned off. Otherwise quota syncing thread
will try to access already freed quota structures.
Signed-off-by: Jan Kara <jack@suse.cz>
We cannot cancel delayed work from ocfs2_local_free_info because that is called
with dqonoff_mutex held and the work it cancels requires dqonoff_mutex to
finish. Cancel the work before acquiring dqonoff_mutex.
Acked-by: Joel Becker <Joel.Becker@oracle.com>
Signed-off-by: Jan Kara <jack@suse.cz>
ocfs2 sometimes needs to block signals around dlm operations, but it
currently does it with sigprocmask(). Even worse, it's checking the
error code of sigprocmask(). The in-kernel sigprocmask() can only error
if you get the SIG_* argument wrong. We don't.
Wrap the sigprocmask() calls with ocfs2_[un]block_signals(). These
functions are void, but they will BUG() if somehow sigprocmask() returns
an error.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
OCFS2 has never really supported intr. This patch acknowledges this reality
and makes nointr the default mount option. In a later patch, we intend to
support intr.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The default behavior for directory reservations stays the same, but we add a
mount option so people can tweak the size of directory reservations
according to their workloads.
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
I have observed that the current size of 8M gives us pretty poor
fragmentation on multi-threaded workloads which do lots of writes.
Generally, I can increase the size of local alloc windows and observe a
marked decrease in fragmentation, even up and beyond window sizes of 512
megabytes. This makes sense for a couple reasons - larger local alloc means
more room for reservation windows. On multi-node workloads the larger local
alloc helps as well because we don't have to do window slides as often.
Also, I removed the OCFS2_DEFAULT_LOCAL_ALLOC_SIZE constant as it is no
longer used and the comment above it was out of date.
To test fragmentation, I used a workload which launched 4 threads that did
4k writes into a series of about 140 alternating files.
With resv_level=2, and a 4k/4k file system I observed the following average
fragmentation for various localalloc= parameters:
localalloc= avg. fragmentation
8 48
32 16
64 10
120 7
On larger cluster sizes, the difference is more dramatic.
The new default size top out at 256M, which we'll only get for cluster
sizes of 32K and above.
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
This patch pulls the local alloc sizing code into localalloc.c and provides
a callout to it from ocfs2_fill_super(). Behavior is essentially unchanged
except that I correctly calculate the maximum local alloc size. The old code
in ocfs2_parse_options() calculated the max size as:
ocfs2_local_alloc_size(sb) * 8
which is correct, in bits. Unfortunately though the option passed in is in
megabytes. Ultimately, this bug made no real difference - the shrink code
would catch a too-large size and bring it down to something reasonable.
Still, it's less than efficient as-is.
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
This patch improves Ocfs2 allocation policy by allowing an inode to
reserve a portion of the local alloc bitmap for itself. The reserved
portion (allocation window) is advisory in that other allocation
windows might steal it if the local alloc bitmap becomes
full. Otherwise, the reservations are honored and guaranteed to be
free. When the local alloc window is moved to a different portion of
the bitmap, existing reservations are discarded.
Reservation windows are represented internally by a red-black
tree. Within that tree, each node represents the reservation window of
one inode. An LRU of active reservations is also maintained. When new
data is written, we allocate it from the inodes window. When all bits
in a window are exhausted, we allocate a new one as close to the
previous one as possible. Should we not find free space, an existing
reservation is pulled off the LRU and cannibalized.
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
ocfs2_group_bitmap_size has to handle the case when the
volume don't have discontiguous block group support. So
pass the feature_incompat in and check it.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Defines the OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG feature bit and modifies
struct ocfs2_group_desc for the feature.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Tao Ma <tao.ma@oracle.com>
This patch add extent block (metadata) stealing mechanism for
extent allocation. This mechanism is same as the inode stealing.
if no room in slot specific extent_alloc, we will try to
allocate extent block from the next slot.
Signed-off-by: Tiger Yang <tiger.yang@oracle.com>
Acked-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Currently the f_fsid of struct kstatfs returned from ocfs2_statfs() is
undefined (vfs layer fills in 0 as default). Since in some conditions,
f_fsid value might be used in a (f_fsid, ino) pair to uniquely identify
a file, ocfs2 should return a unique defined f_fsid value from
ocfs2_statfs().
Because uuid_str is the same on big or litlle endian machine, it's
endian consistent to use osb->uuid_str to generate f_fsid value.
Signed-off-by: Coly Li <coly.li@suse.de>
Cc: Sunil Mushran <sunil.mushran@oracle.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We have to set MS_POSIXACL on remount as well. Otherwise VFS
would not know we started supporting ACLs after remount and
thus ACLs would not work.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Change acl mount options handling to match the one of XFS and BTRFS and
hopefully it is also easier to use now. When admin does not specify any
acl mount option, acls are enabled if and only if the filesystem has
xattr feature enabled. If admin specifies 'acl' mount option, we fail
the mount if the filesystem does not have xattr feature and thus acls
cannot be enabled.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
To become consistent with filesystems such as XFS or BTRFS, make posix
ACLs always available. This also reduces possibility of
misconfiguration on admin's side.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
In case of non-modular kernels the root filesystem is mounted by trying
several filesystems. If ocfs2 was tried before the actual filesystem
type, the mount would fail because ocfs2_sb_probe() returns -EAGAIN
instead of -EINVAL. ocfs2 will now return -EINVAL properly.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Reported-by: Laszlo Attila Toth <panther@balabit.hu>
* remove asm/atomic.h inclusion from linux/utsname.h --
not needed after kref conversion
* remove linux/utsname.h inclusion from files which do not need it
NOTE: it looks like fs/binfmt_elf.c do not need utsname.h, however
due to some personality stuff it _is_ needed -- cowardly leave ELF-related
headers and files alone.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In a clustered setup, we have to panic the box on journal abort. This is
because we don't have the facility to go hard readonly. With hard ro, another
node would detect node failure and initiate recovery.
Having said that, we shouldn't force panic if the volume is mounted locally.
This patch defers the handling to the mount option, errors.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Implement locking around struct ocfs2_refcount_tree. This protects
all read/write operations on refcount trees. ocfs2_refcount_tree
has its own lock and its own caching_info, protecting buffers among
multiple nodes.
User must call ocfs2_lock_refcount_tree before his operation on
the tree and unlock it after that.
ocfs2_refcount_trees are referenced by the block number of the
refcount tree root block, So we create an rb-tree on the ocfs2_super
to look them up.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Similar ip_last_trans, ip_created_trans tracks the creation of a journal
managed inode. This specifically tracks what transaction created the
inode. This is so the code can know if the inode has ever been written
to disk.
This behavior is desirable for any journal managed object. We move it
to struct ocfs2_caching_info as ci_created_trans so that any object
using ocfs2_caching_info can rely on this behavior.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We have the read side of metadata caching isolated to struct
ocfs2_caching_info, now we need the write side. This means the journal
functions. The journal only does a couple of things with struct inode.
This change moves the ip_last_trans field onto struct
ocfs2_caching_info as ci_last_trans. This field tells the journal
whether a pending journal flush is required.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We are really passing the inode into the ocfs2_read/write_blocks()
functions to get at the metadata cache. This commit passes the cache
directly into the metadata block functions, divorcing them from the
inode.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We don't really want to cart around too many new fields on the
ocfs2_caching_info structure. So let's wrap all our access of the
parent object in a set of operations. One pointer on caching_info, and
more flexibility to boot.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We want to use the ocfs2_caching_info structure in places that are not
inodes. To do that, it can no longer rely on referencing the inode
directly.
This patch moves the flags to ocfs2_caching_info->ci_flags, stores
pointers to the parent's locks on the ocfs2_caching_info, and renames
the constants and flags to reflect its independant state.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
If we fail to mount the filesystem, we have to be careful not to dereference
uninitialized structures in ocfs2_kill_sb.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We just set blockcheck stats to zeros but we should also
properly initialize the spinlock there.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
In commit ea455f8ab6, we moved the dentry lock
put process into ocfs2_wq. This causes problems during umount because ocfs2_wq
can drop references to inodes while they are being invalidated by
invalidate_inodes() causing all sorts of nasty things (invalidate_inodes()
ending in an infinite loop, "Busy inodes after umount" messages etc.).
We fix the problem by stopping ocfs2_wq from doing any further releasing of
inode references on the superblock being unmounted, wait until it finishes
the current round of releasing and finally cleaning up all the references in
dentry_lock_list from ocfs2_put_super().
The issue was tracked down by Tao Ma <tao.ma@oracle.com>.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
If the mount fails for any reason, ocfs2_dismount_volume calls
ocfs2_orphan_scan_stop. It requires that ocfs2_orphan_scan_init
be called to setup the mutex and work queues, but that doesn't
happen if the mount has failed and we oops accessing an uninitialized
work queue.
This patch splits the init and startup of the orphan scan, eliminating
the oops.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Al Viro