Skip to content

L
linux
Switch branch/tag
  1. 04 Feb, 2009 2 commits
  3. 03 Feb, 2009 29 commits
  5. 02 Feb, 2009 9 commits
    • Mark Fasheh's avatar
      ocfs2: add quota call to ocfs2_remove_btree_range() · fd4ef231
      Mark Fasheh authored 
      We weren't reclaiming the clusters which get free'd from this function,
so any user punching holes in a file would still have those bytes accounted
against him/her. Add the call to vfs_dq_free_space_nodirty() to fix this.
Interestingly enough, the journal credits calculation already took this into
account.
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
Acked-by: default avatarJan Kara <jack@suse.cz>
      fd4ef231
    • Sunil Mushran's avatar
      ocfs2: Wakeup the downconvert thread after a successful cancel convert · a4b91965
      Sunil Mushran authored 
      When two nodes holding PR locks on a resource concurrently attempt to
upconvert the locks to EX, the master sends a BAST to one of the nodes. This
message tells that node to first cancel convert the upconvert request,
followed by downconvert to a NL. Only when this lock is downconverted to NL,
can the master upconvert the first node's lock to EX.

While the fs was doing the cancel convert, it was forgetting to wake up the
dc thread after a successful cancel, leading to a deadlock.
Reported-and-Tested-by: default avatarDavid Teigland <teigland@redhat.com>
Signed-off-by: default avatarSunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
      a4b91965
    • Tao Ma's avatar
      ocfs2: Access the xattr bucket only before modifying it. · 554e7f9e
      Tao Ma authored 
      In ocfs2_xattr_value_truncate, we may call b-tree codes which will
extend the journal transaction. It has a potential problem that it
may let the already-accessed-but-not-dirtied buffers gone. So we'd
better access the bucket after we call ocfs2_xattr_value_truncate.
And as for the root buffer for the xattr value, b-tree code will
acess and dirty it, so we don't need to worry about it.
Signed-off-by: default avatarTao Ma <tao.ma@oracle.com>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
      554e7f9e
    • Joel Becker's avatar
      configfs: Silence lockdep on mkdir(), rmdir() and configfs_depend_item() · 0e033342
      Joel Becker authored 
      When attaching default groups (subdirs) of a new group (in mkdir() or
in configfs_register()), configfs recursively takes inode's mutexes
along the path from the parent of the new group to the default
subdirs. This is needed to ensure that the VFS will not race with
operations on these sub-dirs. This is safe for the following reasons:

- the VFS allows one to lock first an inode and second one of its
  children (The lock subclasses for this pattern are respectively
  I_MUTEX_PARENT and I_MUTEX_CHILD);
- from this rule any inode path can be recursively locked in
  descending order as long as it stays under a single mountpoint and
  does not follow symlinks.

Unfortunately lockdep does not know (yet?) how to handle such
recursion.

I've tried to use Peter Zijlstra's lock_set_subclass() helper to
upgrade i_mutexes from I_MUTEX_CHILD to I_MUTEX_PARENT when we know
that we might recursively lock some of their descendant, but this
usage does not seem to fit the purpose of lock_set_subclass() because
it leads to several i_mutex locked with subclass I_MUTEX_PARENT by
the same task.

>From inside configfs it is not possible to serialize those recursive
locking with a top-level one, because mkdir() and rmdir() are already
called with inodes locked by the VFS. So using some
mutex_lock_nest_lock() is not an option.

I am proposing two solutions:
1) one that wraps recursive mutex_lock()s with
   lockdep_off()/lockdep_on().
2) (as suggested earlier by Peter Zijlstra) one that puts the
   i_mutexes recursively locked in different classes based on their
   depth from the top-level config_group created. This
   induces an arbitrary limit (MAX_LOCK_DEPTH - 2 == 46) on the
   nesting of configfs default groups whenever lockdep is activated
   but this limit looks reasonably high. Unfortunately, this alos
   isolates VFS operations on configfs default groups from the others
   and thus lowers the chances to detect locking issues.

This patch implements solution 1).

Solution 2) looks better from lockdep's point of view, but fails with
configfs_depend_item(). This needs to rework the locking
scheme of configfs_depend_item() by removing the variable lock recursion
depth, and I think that it's doable thanks to the configfs_dirent_lock.
For now, let's stick to solution 1).
Signed-off-by: default avatarLouis Rilling <louis.rilling@kerlabs.com>
Acked-by: default avatarJoel Becker <joel.becker@oracle.com>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
      0e033342
    • Jan Kara's avatar
      ocfs2: Fix possible deadlock in ocfs2_write_dquot() · f8afead7
      Jan Kara authored 
      It could happen that some limit has been set via quotactl() and in parallel
->mark_dirty() is called from another thread doing e.g. dquot_alloc_space(). In
such case ocfs2_write_dquot() must not try to sync the dquot because that needs
global quota lock but that ranks above transaction start.
Signed-off-by: default avatarJan Kara <jack@suse.cz>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
      f8afead7
    • Jan Kara's avatar
      ocfs2: Push out dropping of dentry lock to ocfs2_wq · ea455f8a
      Jan Kara authored 
      Dropping of last reference to dentry lock is a complicated operation involving
dropping of reference to inode. This can get complicated and quota code in
particular needs to obtain some quota locks which leads to potential deadlock.
Thus we defer dropping of inode reference to ocfs2_wq.
Signed-off-by: default avatarJan Kara <jack@suse.cz>
Signed-off-by: default avatarMark Fasheh <mfasheh@suse.com>
      ea455f8a
    • Cliff Brake's avatar
      pxamci: enable DMA for write ops after CMD/RESP · b6018958
      Cliff Brake authored 
      With the PXA270 MMC hardware, there seems to be an issue of
data corruption on writes where a 4KB data block is offset
by one byte.

If we delay enabling the DMA for writes until after the CMD/RESP
has finished, the problem seems to be fixed.

related to PXA270 Erratum #91
Tested-by: default avatarVernon Sauder <VernonInHand@gmail.com>
Signed-off-by: default avatarCliff Brake <cbrake@bec-systems.com>
Acked-by: default avatarEric Miao <eric.miao@marvell.com>
Signed-off-by: default avatarPierre Ossman <drzeus@drzeus.cx>
      b6018958
    • Cliff Brake's avatar
      pxamci: replace #ifdef CONFIG_PXA27x with if (cpu_is_pxa27x()) · e10a854c
      Cliff Brake authored 
      Signed-off-by: default avatarCliff Brake <cbrake@bec-systems.com>
Acked-by: default avatarEric Miao <eric.miao@marvell.com>
Signed-off-by: default avatarPierre Ossman <drzeus@drzeus.cx>
      e10a854c
    • philipl@overt.org's avatar
      ricoh_mmc: Use suspend_late/resume_early · 06cc1c88
      philipl@overt.org authored 
      If ricoh_mmc suspends before sdhci_pci, it will pull the card
out from under the controller, which could leave the system in
a very confused state.

Using suspend_late/resume_early ensures that sdhci_pci suspends first
and resumes second.
Signed-off-by: default avatarPhilip Langdale <philipl@overt.org>
Signed-off-by: default avatarPierre Ossman <drzeus@drzeus.cx>
      06cc1c88
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7