btrfs_new_inode doesn't call iput to free the inode
when it fails.
Signed-off-by: Shen Feng <shen@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Need to check kthread_should_stop after schedule_timeout() before calling
schedule(). This causes threads to sleep with potentially no one to wake them
up causing mount(2) to hang in btrfs_stop_workers waiting for threads to stop.
Signed-off-by: Amit Gud <gud@ksu.edu>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The 'flushoncommit' mount option forces any data dirtied by a write in a
prior transaction to commit as part of the current commit. This makes
the committed state a fully consistent view of the file system from the
application's perspective (i.e., it includes all completed file system
operations). This was previously the behavior only when a snapshot is
created.
This is used by Ceph to ensure that completed writes make it to the
platter along with the metadata operations they are bound to (by
BTRFS_IOC_TRANS_{START,END}).
Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Add a 'notreelog' mount option to disable the tree log (used by fsync,
O_SYNC writes). This is much slower, but the tree logging produces
inconsistent views into the FS for ceph.
Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs options can change at times other than mount, yet /proc/mounts shows the
options string used when the fs was mounted (an example would be when btrfs
determines that barriers aren't useful and turns them off.) This patch
instead outputs the actual options in use by btrfs.
Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Because btrfs is copy-on-write, we end up picking new locations for
blocks very often. This makes it fairly difficult to maintain perfect
read patterns over time, but we can at least do some optimizations
for writes.
This is done today by remembering the last place we allocated and
trying to find a free space hole big enough to hold more than just one
allocation. The end result is that we tend to write sequentially to
the drive.
This happens all the time for metadata and it happens for data
when mounted -o ssd. But, the way we record it is fairly racey
and it tends to fragment the free space over time because we are trying
to allocate fairly large areas at once.
This commit gets rid of the races by adding a free space cluster object
with dedicated locking to make sure that only one process at a time
is out replacing the cluster.
The free space fragmentation is somewhat solved by allowing a cluster
to be comprised of smaller free space extents. This part definitely
adds some CPU time to the cluster allocations, but it allows the allocator
to consume the small holes left behind by cow.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs_next_leaf was using blocking locks when it could have been using
faster spinning ones instead. This adds a few extra checks around
the pieces that block and switches over to spinning locks.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This patch removes the pinned_mutex. The extent io map has an internal tree
lock that protects the tree itself, and since we only copy the extent io map
when we are committing the transaction we don't need it there. We also don't
need it when caching the block group since searching through the tree is also
protected by the internal map spin lock.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
This patch removes the block group alloc mutex used to protect the free space
tree for allocations and replaces it with a spin lock which is used only to
protect the free space rb tree. This means we only take the lock when we are
directly manipulating the tree, which makes us a touch faster with
multi-threaded workloads.
This patch also gets rid of btrfs_find_free_space and replaces it with
btrfs_find_space_for_alloc, which takes the number of bytes you want to
allocate, and empty_size, which is used to indicate how much free space should
be at the end of the allocation.
It will return an offset for the allocator to use. If we don't end up using it
we _must_ call btrfs_add_free_space to put it back. This is the tradeoff to
kill the alloc_mutex, since we need to make sure nobody else comes along and
takes our space.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
I've replaced the strange looping constructs with a list_for_each_entry on
space_info->block_groups. If we have a hint we just jump into the loop with
the block group and start looking for space. If we don't find anything we
start at the beginning and start looking. We never come out of the loop with a
ref on the block_group _unless_ we found space to use, then we drop it after we
set the trans block_group.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
This patch cleans up the free space cache code a bit. It better documents the
idiosyncrasies of tree_search_offset and makes the code make a bit more sense.
I took out the info allocation at the start of __btrfs_add_free_space and put it
where it makes more sense. This was left over cruft from when alloc_mutex
existed. Also all of the re-searches we do to make sure we inserted properly.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Btrfs pages being written get set to writeback, and then may go through
a number of steps before they hit the block layer. This includes compression,
checksumming and async bio submission.
The end result is that someone who writes a page and then does
wait_on_page_writeback is likely to unplug the queue before the bio they
cared about got there.
We could fix this by marking bios sync, or by doing more frequent unplugs,
but this commit just changes the async bio submission code to unplug
after it has processed all the bios for a device. The async bio submission
does a fair job of collection bios, so this shouldn't be a huge problem
for reducing merging at the elevator.
For streaming O_DIRECT writes on a 5 drive array, it boosts performance
from 386MB/s to 460MB/s.
Thanks to Hisashi Hifumi for helping with this work.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Btrfs uses async helper threads to submit write bios so the checksumming
helper threads don't block on the disk.
The submit bio threads may process bios for more than one block device,
so when they find one device congested they try to move on to other
devices instead of blocking in get_request_wait for one device.
This does a pretty good job of keeping multiple devices busy, but the
congested flag has a number of problems. A congested device may still
give you a request, and other procs that aren't backing off the congested
device may starve you out.
This commit uses the io_context stored in current to decide if our process
has been made a batching process by the block layer. If so, it keeps
sending IO down for at least one batch. This helps make sure we do
a good amount of work each time we visit a bdev, and avoids large IO
stalls in multi-device workloads.
It's also very ugly. A better solution is in the works with Jens Axboe.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Allow MAP_PRIVATE mmaps of "direct_io" files. This is necessary for
execute support.
MAP_SHARED mappings require some sort of coherency between the
underlying file and the mapping. With "direct_io" it is difficult to
provide this, so for the moment just disallow shared (read-write and
read-only) mappings altogether.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Allow the kernel read and write on "direct_io" files. This is
necessary for nfs export and execute support.
The implementation is simple: if an access from the kernel is
detected, don't perform get_user_pages(), just use the kernel address
provided by the requester to copy from/to the userspace filesystem.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
We update information in logical volume integrity descriptor after each
allocation (as LVID contains free space, number of directories and files on
disk etc.). If the filesystem is on some phase change media, this leads to its
quick degradation as such media is able to handle only 10000 overwrites or so.
We solve the problem by writing new information into LVID only on umount,
remount-ro and sync. This solves the problem at the price of longer media
inconsistency (previously media became consistent after pdflush flushed dirty
LVID buffer) but that should be acceptable.
Report by and patch written in cooperation with
Rich Coe <Richard.Coe@med.ge.com>.
Signed-off-by: Jan Kara <jack@suse.cz>
Anchor block can be located at several places on the medium. Two of the
locations are relative to media end which is problematic to detect. Also
some drives report some block as last but are not able to read it or any
block nearby before it. So let's first try block 256 and if it is all fine,
don't look at other possible locations of anchor blocks to avoid IO errors.
This change required a larger reorganization of code but the new code is
hopefully more readable and definitely shorter.
Signed-off-by: Jan Kara <jack@suse.cz>
Make udf_check_valid() return 1 if the validity check passed and 0 otherwise.
So far it was the other way around which was a bit confusing. Also make
udf_vrs() return loff_t which is really the type it should return (not int).
Signed-off-by: Jan Kara <jack@suse.cz>
This patch makes the UDF FS driver use the hardware sector size as the
default logical block size, which is required by the UDF specifications.
While the previous default of 2048 bytes was correct for optical disks,
it was not for hard disks or USB storage devices, and made it impossible
to use such a device with the default mount options. (The Linux mkudffs
tool uses a default block size of 2048 bytes even on devices with
smaller hardware sectors, so this bug is unlikely to be noticed unless
UDF-formatted USB storage devices are exchanged with other OSs.)
To avoid regressions for people who use loopback optical disk images or
who used the (sometimes wrong) defaults of mkudffs, we also try with
a block size of 2048 bytes if no anchor was found with the hardware
sector size.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Functions udf_CS0toNLS() and udf_NLStoCS0() didn't count with the fact that
NLS can return negative length when invalid character is given to it for
conversion. Thus interesting things could happen (such as overwriting random
memory with the rest of filename). Add appropriate checks.
Signed-off-by: Jan Kara <jack@suse.cz>
This patch makes udf return f_fsid info for statfs(2).
Signed-off-by: Coly Li <coly.li@suse.de>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Jan Kara <jack@suse.cz>
On x86 (and several other archs) mode_t is defined as "unsigned short"
and comparing unsigned shorts to negative ints is broken (because short
is promoted to int and then compared). Fix it.
Reported-and-tested-by: Laurent Riffard <laurent.riffard@free.fr>
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
"dmode" allows overriding permissions of directories and
"mode" allows overriding permissions of files.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Jan Kara <jack@suse.cz>
Allocate strings with kmalloc.
Checkstack output:
Before: udf_get_filename: 600
After: udf_get_filename: 136
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Jan Kara <jack@suse.cz>
Allocate strings with kmalloc.
Checkstack output:
Before: udf_process_sequence: 712
After: udf_process_sequence: 200
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
* 'for-linus' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6: (58 commits)
SUNRPC: Ensure IPV6_V6ONLY is set on the socket before binding to a port
NSM: Fix unaligned accesses in nsm_init_private()
NFS: Simplify logic to compare socket addresses in client.c
NFS: Start PF_INET6 callback listener only if IPv6 support is available
lockd: Start PF_INET6 listener only if IPv6 support is available
SUNRPC: Remove CONFIG_SUNRPC_REGISTER_V4
SUNRPC: rpcb_register() should handle errors silently
SUNRPC: Simplify kernel RPC service registration
SUNRPC: Simplify svc_unregister()
SUNRPC: Allow callers to pass rpcb_v4_register a NULL address
SUNRPC: rpcbind actually interprets r_owner string
SUNRPC: Clean up address type casts in rpcb_v4_register()
SUNRPC: Don't return EPROTONOSUPPORT in svc_register()'s helpers
SUNRPC: Use IPv4 loopback for registering AF_INET6 kernel RPC services
SUNRPC: Set IPV6ONLY flag on PF_INET6 RPC listener sockets
NFS: Revert creation of IPv6 listeners for lockd and NFSv4 callbacks
SUNRPC: Remove @family argument from svc_create() and svc_create_pooled()
SUNRPC: Change svc_create_xprt() to take a @family argument
SUNRPC: svc_setup_socket() gets protocol family from socket
SUNRPC: Pass a family argument to svc_register()
...
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (33 commits)
ext4: Regularize mount options
ext4: fix locking typo in mballoc which could cause soft lockup hangs
ext4: fix typo which causes a memory leak on error path
jbd2: Update locking coments
ext4: Rename pa_linear to pa_type
ext4: add checks of block references for non-extent inodes
ext4: Check for an valid i_mode when reading the inode from disk
ext4: Use WRITE_SYNC for commits which are caused by fsync()
ext4: Add auto_da_alloc mount option
ext4: Use struct flex_groups to calculate get_orlov_stats()
ext4: Use atomic_t's in struct flex_groups
ext4: remove /proc tuning knobs
ext4: Add sysfs support
ext4: Track lifetime disk writes
ext4: Fix discard of inode prealloc space with delayed allocation.
ext4: Automatically allocate delay allocated blocks on rename
ext4: Automatically allocate delay allocated blocks on close
ext4: add EXT4_IOC_ALLOC_DA_BLKS ioctl
ext4: Simplify delalloc code by removing mpage_da_writepages()
ext4: Save stack space by removing fake buffer heads
...
This fixes unaligned accesses in nsm_init_private() when
creating nlm_reboot keys.
Signed-off-by: Mans Rullgard <mans@mansr.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
Btrfs: try to free metadata pages when we free btree blocks
Btrfs: add extra flushing for renames and truncates
Btrfs: make sure btrfs_update_delayed_ref doesn't increase ref_mod
Btrfs: optimize fsyncs on old files
Btrfs: tree logging unlink/rename fixes
Btrfs: Make sure i_nlink doesn't hit zero too soon during log replay
Btrfs: limit balancing work while flushing delayed refs
Btrfs: readahead checksums during btrfs_finish_ordered_io
Btrfs: leave btree locks spinning more often
Btrfs: Only let very young transactions grow during commit
Btrfs: Check for a blocking lock before taking the spin
Btrfs: reduce stack in cow_file_range
Btrfs: reduce stalls during transaction commit
Btrfs: process the delayed reference queue in clusters
Btrfs: try to cleanup delayed refs while freeing extents
Btrfs: reduce stack usage in some crucial tree balancing functions
Btrfs: do extent allocation and reference count updates in the background
Btrfs: don't preallocate metadata blocks during btrfs_search_slot
A deadlock can occur when user space uses a signal (autofs version 4 uses
SIGCHLD for this) to effect expire completion.
The order of events is:
Expire process completes, but before being able to send SIGCHLD to it's parent
...
Another process walks onto a different mount point and drops the directory
inode semaphore prior to sending the request to the daemon as it must ...
A third process does an lstat on on the expired mount point causing it to wait
on expire completion (unfortunately) holding the directory semaphore.
The mount request then arrives at the daemon which does an lstat and,
deadlock.
For some time I was concerned about releasing the directory semaphore around
the expire wait in autofs4_lookup as well as for the mount call back. I
finally realized that the last round of changes in this function made the
expiring dentry and the lookup dentry separate and distinct so the check and
possible wait can be done anywhere prior to the mount call back. This patch
moves the check to just before the mount call back and inside the directory
inode mutex release.
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A significant portion of the autofs_dev_ioctl_expire() and
autofs4_expire_multi() functions is duplicated code. This patch cleans that
up.
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=12843
"I use ramfs instead of tmpfs for /tmp because I don't use swap on my
laptop. Some apps need 1777 mode for /tmp directory, but ramfs does not
support 'mode=' mount option."
Reported-by: Avan Anishchuk <matimatik@gmail.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Introduce keyed event wakeups inside the eventfd code.
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the events hint now sent by some devices, to avoid unnecessary wakeups
for events that are of no interest for the caller. This code handles both
devices that are sending keyed events, and the ones that are not (and
event the ones that sometimes send events, and sometimes don't).
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
People started using eventfd in a semaphore-like way where before they
were using pipes.
That is, counter-based resource access. Where a "wait()" returns
immediately by decrementing the counter by one, if counter is greater than
zero. Otherwise will wait. And where a "post(count)" will add count to
the counter releasing the appropriate amount of waiters. If eventfd the
"post" (write) part is fine, while the "wait" (read) does not dequeue 1,
but the whole counter value.
The problem with eventfd is that a read() on the fd returns and wipes the
whole counter, making the use of it as semaphore a little bit more
cumbersome. You can do a read() followed by a write() of COUNTER-1, but
IMO it's pretty easy and cheap to make this work w/out extra steps. This
patch introduces a new eventfd flag that tells eventfd to only dequeue 1
from the counter, allowing simple read/write to make it behave like a
semaphore. Simple test here:
http://www.xmailserver.org/eventfd-sem.c
To be back-compatible with earlier kernels, userspace applications should
probe for the availability of this feature via
#ifdef EFD_SEMAPHORE
fd = eventfd2 (CNT, EFD_SEMAPHORE);
if (fd == -1 && errno == EINVAL)
<fallback>
#else
<fallback>
#endif
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: <linux-api@vger.kernel.org>
Tested-by: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
eventpoll.c uses void * in one place for no obvious reason; change it to
use the real type instead.
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ep_modify() doesn't need to set event.data from within the ep->lock
spinlock as the comment suggests. The only place event.data is used is
ep_send_events_proc(), and this is protected by ep->mtx instead of
ep->lock. Also update the comment for mutex_lock() at the top of
ep_scan_ready_list(), which mentions epoll_ctl(EPOLL_CTL_DEL) but not
epoll_ctl(EPOLL_CTL_MOD).
ep_modify() can also use spin_lock_irq() instead of spin_lock_irqsave().
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
xchg in ep_unregister_pollwait() is unnecessary because it is protected by
either epmutex or ep->mtx (the same protection as ep_remove()).
If xchg was necessary, it would be insufficient to protect against
problems: if multiple concurrent calls to ep_unregister_pollwait() were
possible then a second caller that returns without doing anything because
nwait == 0 could return before the waitqueues are removed by the first
caller, which looks like it could lead to problematic races with
ep_poll_callback().
So remove xchg and add comments about the locking.
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If epoll_wait returns -EFAULT, the event that was being returned when the
fault was encountered will be forgotten. This is not a big deal since
EFAULT will happen only if a buggy userspace program passes in a bad
address, in which case what happens later usually doesn't matter.
However, it is easy to remember the event for later, and this patch makes
a simple change to do that.
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
