-----------------------------------------------------------------------------
>From 23c3010808de86f21436eb822aacfa551bfc17e4 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Fri, 14 Jan 2011 22:39:16 -0600
Subject: [PATCH 1/2] GFS2: remove iopen glocks from cache on failed deletes
When a file gets deleted on GFS2, if a node can't get an exclusive lock on the
file's iopen glock, it punts on actually freeing up the space, because another
node is using the file. When it does this, it needs to drop the iopen glock
from its cache so that the other node can get an exclusive lock on it. Now,
gfs2_delete_inode() sets GL_NOCACHE before dropping the shared lock on the
iopen glock in preparation for grabbing it in the exclusive state. Since the
node needs the glock in the exclusive state, dropping the shared lock from the
cache doesn't slow down the case where no other nodes are using the file.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
------------------------------------------------------------------------------
>From 24d9765fc18c7838ccdbb0d71fb706321d9b824c Mon Sep 17 00:00:00 2001
From: Steven Whitehouse <swhiteho@redhat.com>
Date: Tue, 18 Jan 2011 14:49:08 +0000
Subject: [PATCH 2/2] GFS2: Fix error path in gfs2_lookup_by_inum()
In the (impossible, except if there is fs corruption) error path
in gfs2_lookup_by_inum() if the call to gfs2_inode_refresh()
fails, it was leaving the function by calling iput() rather
than iget_failed(). This would cause future lookups of the same
inode to block forever.
This patch fixes the problem by moving the call to gfs2_inode_refresh()
into gfs2_inode_lookup() where iget_failed() is part of the error path
already. Also this cleans up some unreachable code and makes
gfs2_set_iop() static.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
/**
- * GFS2 lookup code fills in vfs inode contents based on info obtained
- * from directory entry inside gfs2_inode_lookup(). This has caused issues
- * with NFS code path since its get_dentry routine doesn't have the relevant
- * directory entry when gfs2_inode_lookup() is invoked. Part of the code
- * segment inside gfs2_inode_lookup code needs to get moved around.
+ * gfs2_set_iop - Sets inode operations
+ * @inode: The inode with correct i_mode filled in
*
- * Clears I_NEW as well.
- **/
+ * GFS2 lookup code fills in vfs inode contents based on info obtained
+ * from directory entry inside gfs2_inode_lookup().
+ */
- if ((type == DT_UNKNOWN) && (no_formal_ino == 0))
- goto gfs2_nfsbypass;
-
- inode->i_mode = DT2IF(type);
-
- /*
- * We must read the inode in order to work out its type in
- * this case. Note that this doesn't happen often as we normally
- * know the type beforehand. This code path only occurs during
- * unlinked inode recovery (where it is safe to do this glock,
- * which is not true in the general case).
- */
if (type == DT_UNKNOWN) {
- struct gfs2_holder gh;
- error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
- if (unlikely(error))
- goto fail_glock;
- /* Inode is now uptodate */
- gfs2_glock_dq_uninit(&gh);
+ /* Inode glock must be locked already */
+ error = gfs2_inode_refresh(GFS2_I(inode));
+ if (error)
+ goto fail_refresh;
+ } else {
+ inode->i_mode = DT2IF(type);
}
- error = gfs2_inode_refresh(GFS2_I(inode));
- if (error)
- goto fail_iput;
-
- /* Pick up the works we bypass in gfs2_inode_lookup */
- if (inode->i_state & I_NEW)
- gfs2_set_iop(inode);
-
/* Two extra checks for NFS only */
if (no_formal_ino) {
error = -ESTALE;
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index 732a183..3e00a66 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -96,7 +96,6 @@ err:
return -EIO;
}
These two changes just missed out on the last pull request. Please
consider pulling the following two fixes,
Steve.
-----------------------------------------------------------------------------
The following changes since commit c21fd1a8c68ce3f49b00caf10337169262cfb8ad:
Please consider pulling the following three fixes,
Steve.
--------------------------------------------------------------------------------
The following changes since commit 620917de59eeb934b9f8cf35cc2d95c1ac8ed0fc:
Linux 3.0-rc7 (2011-07-11 16:51:52 -0700)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes.git master
Benjamin Marzinski (1):
GFS2: force a log flush when invalidating the rindex glock
Steven Whitehouse (2):
GFS2: Fix race during filesystem mount
GFS2: Resolve inode eviction and ail list interaction bug
------------------------------------------------------------------------------
The following changes since commit dd775ae2549217d3ae09363e3edb305d0fa19928:
Linux 3.4-rc1 (2012-03-31 16:24:09 -0700)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes.git master
Benjamin Poirier (1):
GFS2: use depends instead of select in kconfig
Bob Peterson (3):
GFS2: put glock reference in error patch of read_rindex_entry
GFS2: Make sure rindex is uptodate before starting transactions
GFS2: Allow caching of rindex glock
[Apologies - I don't think the first message got signed as
it should have. Here is a second attempt...]
Hi,
Please consider pulling the following patches,
Steve.
------------------------------------------------------------------------------
The following changes since commit dd775ae2549217d3ae09363e3edb305d0fa19928:
Linux 3.4-rc1 (2012-03-31 16:24:09 -0700)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes.git master
Benjamin Poirier (1):
GFS2: use depends instead of select in kconfig
Bob Peterson (3):
GFS2: put glock reference in error patch of read_rindex_entry
GFS2: Make sure rindex is uptodate before starting transactions
GFS2: Allow caching of rindex glock
Here are three GFS2 fixes for the current kernel tree. These are all
related to the block reservation code which was added at the merge
window. That code will be getting an update at the forthcoming merge
window too. In the mean time though there are a few smaller issues
which should be fixed.
The first patch resolves an issue with write sizes of greater than
32 bits with the size hinting code. The second ensures that the
allocation data structure is initialised when using xattrs and the
third takes into account allocations which may have been made by
other nodes which affect a reservation on the local node.
The following changes since commit 0d7614f09c1ebdbaa1599a5aba7593f147bf96ee:
Linux 3.6-rc1 (2012-08-02 16:38:10 -0700)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes.git master
Steven Whitehouse (3):
GFS2: Make write size hinting code common
GFS2: Fix missing allocation data for set/remove xattr
GFS2: Take account of blockages when using reserved blocks
GFS2: Pull request (fixes)
