We use the ocfs2_xattr_value_buf structure to manage external values.
It lets the value tree code do its work regardless of the containing
storage. ocfs2_xa_fill_value_buf() initializes a value buf from an
ocfs2_xa_loc entry.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Previously the xattr code would send in a fake value, containing a tree
root, to the function that installed name+value pairs. Instead, we pass
the real value to ocfs2_xa_set_inline_value(), and it notices that the
value cannot fit. Thus, it installs a tree root.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
We create two new functions on ocfs2_xa_loc, ocfs2_xa_prepare_entry()
and ocfs2_xa_store_inline_value().
ocfs2_xa_prepare_entry() makes sure that the xl_entry field of
ocfs2_xa_loc is ready to receive an xattr. The entry will point to an
appropriately sized name+value region in storage. If an existing entry
can be reused, it will be. If no entry already exists, it will be
allocated. If there isn't space to allocate it, -ENOSPC will be
returned.
ocfs2_xa_store_inline_value() stores the data that goes into the 'value'
part of the name+value pair. For values that don't fit directly, this
stores the value tree root.
A number of operations are added to ocfs2_xa_loc_operations to support
these functions. This reflects the disparate behaviors of xattr blocks
and buckets.
With these functions, the overlapping ocfs2_xattr_set_entry_local() and
ocfs2_xattr_set_entry_normal() can be replaced with a single call
scheme.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
An ocfs2 xattr entry stores the text name and value as a pair in the
storage area. Obviously names and values can be variable-sized. If a
value is too large for the entry storage, a tree root is stored instead.
The name+value pair is also padded.
Because of this, there are a million places in the code that do:
if (needs_external_tree(value_size)
namevalue_size = pad(name_size) + tree_root_size;
else
namevalue_size = pad(name_size) + pad(value_size);
Let's create some convenience functions to make the code more readable.
There are three forms. The first takes the raw sizes. The second takes
an ocfs2_xattr_info structure. The third takes an existing
ocfs2_xattr_entry.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Rather than calculating strlen all over the place, let's store the
name length directly on ocfs2_xattr_info.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
struct ocfs2_xattr_info is a useful structure describing an xattr
you'd like to set. Let's put prefixes on the member fields so it's
easier to read and use.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Add ocfs2_xa_remove_entry(), which will remove an xattr entry from its
storage via the ocfs2_xa_loc descriptor.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The ocfs2 extended attribute (xattr) code is very flexible. It can
store xattrs in the inode itself, in an external block, or in a tree of
data structures. This allows the number of xattrs to be bounded by the
filesystem size.
However, the code that manages each possible storage location is
different. Maintaining the ocfs2 xattr code requires changing each hunk
separately.
This patch is the start of a series introducing the ocfs2_xa_loc
structure. This structure wraps the on-disk details of an xattr
entry. The goal is that the generic xattr routines can use
ocfs2_xa_loc without knowing the underlying storage location.
This first pass merely implements the basic structure, initializing it,
and wiping the name+value pair of the entry.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Add current->comm to the standard mlog() output to help with debugging.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
When ocfs2 has to do CoW for refcounted extents, we disable direct I/O
and go through the buffered I/O path. This makes the combined check
easier to read.
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Joel Becker <joel.becker@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>
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2:
ocfs2/cluster: Make o2net connect messages KERN_NOTICE
ocfs2/dlm: Fix printing of lockname
ocfs2: Fix contiguousness check in ocfs2_try_to_merge_extent_map()
ocfs2/dlm: Remove BUG_ON in dlm recovery when freeing locks of a dead node
ocfs2: Plugs race between the dc thread and an unlock ast message
ocfs2: Remove overzealous BUG_ON during blocked lock processing
ocfs2: Do not downconvert if the lock level is already compatible
ocfs2: Prevent a livelock in dlmglue
ocfs2: Fix setting of OCFS2_LOCK_BLOCKED during bast
ocfs2: Use compat_ptr in reflink_arguments.
ocfs2/dlm: Handle EAGAIN for compatibility - v2
ocfs2: Add parenthesis to wrap the check for O_DIRECT.
ocfs2: Only bug out when page size is larger than cluster size.
ocfs2: Fix memory overflow in cow_by_page.
ocfs2/dlm: Print more messages during lock migration
ocfs2/dlm: Ignore LVBs of locks in the Blocked list
ocfs2/trivial: Remove trailing whitespaces
ocfs2: fix a misleading variable name
ocfs2: Sync max_inline_data_with_xattr from tools.
ocfs2: Fix refcnt leak on ocfs2_fast_follow_link() error path
Connect and disconnect messages are more than informational as they are required
during root cause analysis for failures. This patch changes them from KERN_INFO
to KERN_NOTICE.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Faseh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The debug call printing the name of the lock resource was chopping
off the last character. This patch fixes the problem.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
Take ima_file_free() to proper place.
ima: rename PATH_CHECK to FILE_CHECK
ima: rename ima_path_check to ima_file_check
ima: initialize ima before inodes can be allocated
fix ima breakage
Take ima_path_check() in nfsd past dentry_open() in nfsd_open()
freeze_bdev: don't deactivate successfully frozen MS_RDONLY sb
befs: fix leak
This reverts commit 7036251180 ("tty: fix race in tty_fasync") and
commit b04da8bfdf ("fnctl: f_modown should call write_lock_irqsave/
restore") that tried to fix up some of the fallout but was incomplete.
It turns out that we really cannot hold 'tty->ctrl_lock' over calling
__f_setown, because not only did that cause problems with interrupt
disables (which the second commit fixed), it also causes a potential
ABBA deadlock due to lock ordering.
Thanks to Tetsuo Handa for following up on the issue, and running
lockdep to show the problem. It goes roughly like this:
- f_getown gets filp->f_owner.lock for reading without interrupts
disabled, so an interrupt that happens while that lock is held can
cause a lockdep chain from f_owner.lock -> sighand->siglock.
- at the same time, the tty->ctrl_lock -> f_owner.lock chain that
commit 7036251180 introduced, together with the pre-existing
sighand->siglock -> tty->ctrl_lock chain means that we have a lock
dependency the other way too.
So instead of extending tty->ctrl_lock over the whole __f_setown() call,
we now just take a reference to the 'pid' structure while holding the
lock, and then release it after having done the __f_setown. That still
guarantees that 'struct pid' won't go away from under us, which is all
we really ever needed.
Reported-and-tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Acked-by: Américo Wang <xiyou.wangcong@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ima_path_check actually deals with files! call it ima_file_check instead.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The "Untangling ima mess, part 2 with counters" patch messed
up the counters. Based on conversations with Al Viro, this patch
streamlines ima_path_check() by removing the counter maintaince.
The counters are now updated independently, from measuring the file,
in __dentry_open() and alloc_file() by calling ima_counts_get().
ima_path_check() is called from nfsd and do_filp_open().
It also did not measure all files that should have been measured.
Reason: ima_path_check() got bogus value passed as mask.
[AV: mea culpa]
[AV: add missing nfsd bits]
Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Thanks Thomas and Christoph for testing and review.
I removed 'smp_wmb()' before up_write from the previous patch,
since up_write() should have necessary ordering constraints.
(I.e. the change of s_frozen is visible to others after up_write)
I'm quite sure the change is harmless but if you are uncomfortable
with Tested-by/Reviewed-by on the modified patch, please remove them.
If MS_RDONLY, freeze_bdev should just up_write(s_umount) instead of
deactivate_locked_super().
Also, keep sb->s_frozen consistent so that remount can check the frozen state.
Otherwise a crash reported here can happen:
http://lkml.org/lkml/2010/1/16/37http://lkml.org/lkml/2010/1/28/53
This patch should be applied for 2.6.32 stable series, too.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Thomas Backlund <tmb@mandriva.org>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The wrong member was compared in the continguousness check.
Acked-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
Btrfs: apply updated fallocate i_size fix
Btrfs: do not try and lookup the file extent when finishing ordered io
Btrfs: Fix oopsen when dropping empty tree.
Btrfs: remove BUG_ON() due to mounting bad filesystem
Btrfs: make error return negative in btrfs_sync_file()
Btrfs: fix race between allocate and release extent buffer.
* 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
NFS: Don't clobber the attribute type in nfs_update_inode()
NFS: Fix a umount race
NFS: Fix an Oops when truncating a file
NFS: Ensure that we handle NFS4ERR_STALE_STATEID correctly
NFSv4.1: Don't call nfs4_schedule_state_recovery() unnecessarily
NFSv4: Don't allow posix locking against servers that don't support it
NFSv4: Ensure that the NFSv4 locking can recover from stateid errors
NFS: Avoid warnings when CONFIG_NFS_V4=n
NFS: Make nfs_commitdata_release static
NFS: Try to commit unstable writes in nfs_release_page()
NFS: Fix a reference leak in nfs_wb_cancel_page()
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes:
GFS2: Extend umount wait coverage to full glock lifetime
GFS2: Wait for unlock completion on umount
This version of the i_size fix for fallocate makes sure we only update
the i_size when the current fallocate is really operating outside of
i_size.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
When running the following fio job
[torrent]
filename=torrent-test
rw=randwrite
size=4g
filesize=4g
bs=4k
ioengine=sync
you would see long stalls where no work was being done. That is because we were
doing all this extra work to read in the file extent outside of the transaction,
however in the random io case this ends up hurting us because the file extents
are not there to begin with. So axe this logic, since we end up reading in the
file extent when we go to update it anyway. This took the fio job from 11 mb/s
with several ~10 second stalls to 24 mb/s to a couple of 1-2 second stalls.
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
When dropping a empty tree, walk_down_tree() skips checking
extent information for the tree root. This will triggers a
BUG_ON in walk_up_proc().
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Mounting a bad filesystem caused a BUG_ON(). The following is steps to
reproduce it.
# mkfs.btrfs /dev/sda2
# mount /dev/sda2 /mnt
# mkfs.btrfs /dev/sda1 /dev/sda2
(the program says that /dev/sda2 was mounted, and then exits. )
# umount /mnt
# mount /dev/sda1 /mnt
At the third step, mkfs.btrfs exited in the way of make filesystem. So the
initialization of the filesystem didn't finish. So the filesystem was bad, and
it caused BUG_ON() when mounting it. But BUG_ON() should be called by the wrong
code, not user's operation, so I think it is a bug of btrfs.
This patch fixes it.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Increase extent buffer's reference count while holding the lock.
Otherwise it can race with try_release_extent_buffer.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
During recovery, the dlm frees the locks for the dead node. If it finds a
lock in a resource for the dead node, it expects that node to also have a
ref in that lock resource. If not, it BUGs.
ossbz#1175 was filed with the above BUG. Now, while it is correct that we
should be expecting the ref, I see no reason why we have to BUG. After all,
we are freeing up the lock and clearing the ref.
This patch replaces the BUG_ON with a printk(). Hopefully, that will give
us more clues next time this happens.
http://oss.oracle.com/bugzilla/show_bug.cgi?id=1175
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
This patch plugs a race between the downconvert thread and an unlock ast message.
Specifically, after the downconvert worker has done its task, the dc thread needs
to check whether an unlock ast made the downconvert moot.
Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Fasheh <mfasheh@sus.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
If the NFS_ATTR_FATTR_TYPE field isn't set in fattr->valid, then we should
not set the S_IFMT part of inode->i_mode.
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Ensure that we unregister the bdi before kill_anon_super() calls
ida_remove() on our device name.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
The VM/VFS does not allow mapping->a_ops->invalidatepage() to fail.
Unfortunately, nfs_wb_page_cancel() may fail if a fatal signal occurs.
Since the NFS code assumes that the page stays mapped for as long as the
writeback is active, we can end up Oopsing (among other things).
The only safe fix here is to convert nfs_wait_on_request(), so as to make
it uninterruptible (as is already the case with wait_on_page_writeback()).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
Although all glocks are, by the time of the umount glock wait,
scheduled for demotion, some of them haven't made it far
enough through the process for the original set of waiting
code to wait for them.
This extends the ref count to the whole glock lifetime in order
to ensure that the waiting does catch all glocks. It does make
it a bit more invasive, but it seems the only sensible solution
at the moment.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch adds a wait on umount between the point at which we
dispose of all glocks and the point at which we unmount the
lock protocol. This ensures that we've received all the replies
to our unlock requests before we stop the locking.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Reported-by: Fabio M. Di Nitto <fdinitto@redhat.com>
During blocked lock processing, we should consider the possibility that the
lock is no longer blocking.
Joel Becker <joel.becker@oracle.com> assisted in fixing this issue.
Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
During upconvert, if the master were to send a BAST, dlmglue will detect the
upconversion in process and send a cancel convert to the master. Upon receiving
the AST for the cancel convert, it will re-process the lock resource to determine
whether it needs downconverting. Say, the up was from PR to EX and the BAST was
for EX. After the cancel convert, it will need to downconvert to NL.
However, if the node was originally upconverting from NL to EX, then there would
be no reason to downconvert (assuming the same message sequence).
This patch makes dlmglue consider the possibility that the current lock level
is already compatible and that downconverting is not required.
Joel Becker <joel.becker@oracle.com> assisted in fixing this issue.
Fixes ossbz#1178
http://oss.oracle.com/bugzilla/show_bug.cgi?id=1178
Reported-by: Coly Li <coly.li@suse.de>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
There is possibility of a livelock in __ocfs2_cluster_lock(). If a node were
to get an ast for an upconvert request, followed immediately by a bast,
there is a small window where the fs may downconvert the lock before the
process requesting the upconvert is able to take the lock.
This patch adds a new flag to indicate that the upconvert is still in
progress and that the dc thread should not downconvert it right now.
Wengang Wang <wen.gang.wang@oracle.com> and Joel Becker
<joel.becker@oracle.com> contributed heavily to this patch.
Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
During bast, set the OCFS2_LOCK_BLOCKED flag only if the lock needs to
downconverted.
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Acked-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Although we use u64 to pass userspace pointers to the kernel
to avoid compat_ioctl, it doesn't work in some ppc platform.
So wrap them with compat_ptr and add compat_ioctl.
The detailed discussion about compat_ptr can be found in thread
http://lkml.org/lkml/2009/10/27/423.
We indeed met with a bug when testing on ppc(-EFAULT is returned
when using old_path). This patch try to fix this.
I have tested in ppc64(with 32 bit reflink) and x86_64(with i686
reflink), both works.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Mainline commit aad1b15310 made the
dlm_begin_reco_handler() return -EAGAIN instead of EAGAIN.
As this error is transmitted over the wire, we want the receiver,
dlm_send_begin_reco_message(), to understand both the older EAGAIN and
the newer -EAGAIN, to allow rolling upgrade of the cluster nodes.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
In CoW, we have to make sure that the page is already written
out to the disk. So we have a BUG_ON(PageDirty(page)).
In ppc platform we have pagesize=64K, so if the cs=4K, if the
file have fragmented clusters, we will map the page many times.
See this file as an example.
Tree Depth: 0 Count: 19 Next Free Rec: 14
## Offset Clusters Block# Flags
0 0 4 2164864 0x2 Refcounted
1 4 2 9302792 0x2 Refcounted
...
We have to replace the extent recs one by one, so the page with index 0
will be mapped and dirtied twice.
I'd like to leave the BUG_ON there while adding a check so that in
case we meet with an error in other platforms, we can find it easily.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
In ocfs2_duplicate_clusters_by_page, we calculate map_end
by shifting page_index. But actually in case we meet with
a large offset(say in a i686 box, poff_t is only 32 bits
and page_index=2056240), we will overflow. So change the
type of page_index to loff_t.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>
The cache alias problem will happen if the changes of user shared mapping
is not flushed before copying, then user and kernel mapping may be mapped
into two different cache line, it is impossible to guarantee the coherence
after iov_iter_copy_from_user_atomic. So the right steps should be:
flush_dcache_page(page);
kmap_atomic(page);
write to page;
kunmap_atomic(page);
flush_dcache_page(page);
More precisely, we might create two new APIs flush_dcache_user_page and
flush_dcache_kern_page to replace the two flush_dcache_page accordingly.
Here is a snippet tested on omap2430 with VIPT cache, and I think it is
not ARM-specific:
int val = 0x11111111;
fd = open("abc", O_RDWR);
addr = mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
*(addr+0) = 0x44444444;
tmp = *(addr+0);
*(addr+1) = 0x77777777;
write(fd, &val, sizeof(int));
close(fd);
The results are not always 0x11111111 0x77777777 at the beginning as expected. Sometimes we see 0x44444444 0x77777777.
Signed-off-by: Anfei <anfei.zhou@gmail.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: <linux-arch@vger.kernel.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>