Commit Graph

582 Commits

Author SHA1 Message Date
Miklos Szeredi
68b47139ea [PATCH] namespace.c: fix bind mount from foreign namespace
I'm resending this patch, because I still believe it's the correct fix.

Tested before/after applying the patch with a test application
available from:

  http://www.inf.bme.hu/~mszeredi/nstest.c

Bind mount from a foreign namespace results in an un-removable mount.
The reason is that mnt->mnt_namespace is copied from the old mount in
clone_mnt().  Because of this check_mnt() in sys_umount() will fail.

The solution is to set mnt->mnt_namespace to current->namespace in
clone_mnt().  clone_mnt() is either called from do_loopback() or
copy_tree().  copy_tree() is called from do_loopback() or
copy_namespace().

When called (directly or indirectly) from do_loopback(), always
current->namspace is being modified: check_mnt(nd->mnt).  So setting
mnt->mnt_namespace to current->namspace is the right thing to do.

When called from copy_namespace(), the setting of mnt_namespace is
irrelevant, since mnt_namespace is reset later in that function for
all copied mounts.

Jamie said:

  This patch is correct.  The old code was buggy for more fundamental and
  serious reason: it broke the invariant that a tree of vfsmnts all have the
  same value of mnt_namespace (and the same for the mnt_list list).

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Acked-by: Jamie Lokier <jamie@shareable.org>
Cc: <viro@parcelfarce.linux.theplanet.co.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-07 10:00:38 -07:00
Andrew Morton
e525e153c7 [PATCH] __bio_clone() dead comment
Remove a very wrong comment.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-07 10:00:38 -07:00
Linus Torvalds
fab5a60a29 Check input buffer size in zisofs
This uses the new deflateBound() thing to sanity-check the input to the
zlib decompressor before we even bother to start reading in the blocks.

Problem noted by Tim Yamin <plasmaroo@gentoo.org>
2005-08-06 09:42:06 -07:00
John McCutchan
0c3dba1534 [PATCH] Clean up inotify delete race fix
This avoids the whole #ifdef mess by just getting a copy of
dentry->d_inode before d_delete is called - that makes the codepaths the
same for the INOTIFY/DNOTIFY cases as for the regular no-notify case.
I've been running this under a Gnome session for the last 10 minutes.
Inotify is being used extensively.

Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-04 21:37:39 -07:00
John McCutchan
e234f35c54 [PATCH] inotify delete race fix
The included patch fixes a problem where a inotify client would receive a
delete event before the file was actually deleted.  The bug affects both
dnotify & inotify.

Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-04 13:11:15 -07:00
Robert Love
3de11748c1 [PATCH] inotify: update help text
The inotify help text still refers to the character device.  Update it.

Fixes kernel bug #4993.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-04 13:11:15 -07:00
Roman Zippel
74f9c9c258 [PATCH] hfs: don't reference missing page
If there was a read error, the bnode might miss some pages, so skip them.

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-01 21:38:00 -07:00
Roman Zippel
f76d28d235 [PATCH] hfs: don't dirty unchanged inode
If inode size hasn't changed, don't do anything further in truncate, which
also prevents a dirty inode, what might upset some readonly devices quite
badly.

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-01 21:38:00 -07:00
John McCutchan
b9c55d29e9 [PATCH] inotify: fix race between the kernel and user space
When you rm a watch, an IN_IGNORED event is sent down the event queue
with the watch descriptor that you just rm'd.

If you then add a watch you could get the ignored watch's wd and if you
haven't read the entire event queue, user space will think that it's
newly created watch was just ignored.

To avoid this problem we just use idr_get_new_above instead of
idr_get_new.

Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-01 09:16:53 -07:00
John McCutchan
7544953685 [PATCH] inotify: fix file deletion by rename detection
When a file is moved over an existing file that you are watching,
inotify won't send you a DELETE_SELF event and it won't unref the inode
until the inotify instance is closed by the application.

Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-08-01 09:16:53 -07:00
Maneesh Soni
9ca1eb3282 [PATCH] sysfs: fix sysfs_setattr
o sysfs_dirent's s_mode field should also be updated in sysfs_setattr(), else
  there could be inconsistency in the two fields. s_mode is used while
  ->readdir so as not to bring in the inode to cache.

Signed-off-by: Maneesh Soni <maneesh@in.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-29 13:12:49 -07:00
Maneesh Soni
bc062b1b5c [PATCH] sysfs: fix sysfs_chmod_file
o sysfs_chmod_file() must update the new iattr field in sysfs_dirent else
  the mode change will not be persistent in case of inode evacuation from
  cache.

Signed-off-by: Maneesh Soni <maneesh@in.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-29 13:12:49 -07:00
Paolo 'Blaisorblade' Giarrusso
a2d76bd8fa [PATCH] uml: implement hostfs syncing
Actually implement the hostfs "sync" method.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-28 21:46:05 -07:00
Andrew Morton
a5453be48e [PATCH] bio_clone fix
Fix bug introduced in 2.6.11-rc2: when we clone a BIO we need to copy over the
current index into it as well.

It corrupts data with some MD setups.

See http://bugzilla.kernel.org/show_bug.cgi?id=4946

Huuuuuuuuge thanks to Matthew Stapleton <matthew4196@gmail.com> for doggedly
chasing this one down.

Acked-by: Jens Axboe <axboe@suse.de>
Cc: <linux-raid@vger.kernel.org>
Cc: <dm-devel@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-28 08:38:59 -07:00
Linus Torvalds
49302d0c42 Merge head 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/shaggy/jfs-2.6 2005-07-27 16:42:22 -07:00
Jesper Juhl
77933d7276 [PATCH] clean up inline static vs static inline
`gcc -W' likes to complain if the static keyword is not at the beginning of
the declaration.  This patch fixes all remaining occurrences of "inline
static" up with "static inline" in the entire kernel tree (140 occurrences in
47 files).

While making this change I came across a few lines with trailing whitespace
that I also fixed up, I have also added or removed a blank line or two here
and there, but there are no functional changes in the patch.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:26:20 -07:00
Olaf Hering
44456d37b5 [PATCH] turn many #if $undefined_string into #ifdef $undefined_string
turn many #if $undefined_string into #ifdef $undefined_string to fix some
warnings after -Wno-def was added to global CFLAGS

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:26:08 -07:00
Andreas Gruenbacher
02b775696f [PATCH] reiserfs doesn't use mbcache
reiserfs doesn't use the mbcache, so this can go.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:26:07 -07:00
Andreas Gruenbacher
8c52ab42c1 [PATCH] mbcache: Remove unused mb_cache_shrink parameter
The cache parameter to mb_cache_shrink isn't used.  We may as well remove
it.

Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:26:07 -07:00
Peter Staubach
c293621bbf [PATCH] stale POSIX lock handling
I believe that there is a problem with the handling of POSIX locks, which
the attached patch should address.

The problem appears to be a race between fcntl(2) and close(2).  A
multithreaded application could close a file descriptor at the same time as
it is trying to acquire a lock using the same file descriptor.  I would
suggest that that multithreaded application is not providing the proper
synchronization for itself, but the OS should still behave correctly.

SUS3 (Single UNIX Specification Version 3, read: POSIX) indicates that when
a file descriptor is closed, that all POSIX locks on the file, owned by the
process which closed the file descriptor, should be released.

The trick here is when those locks are released.  The current code releases
all locks which exist when close is processing, but any locks in progress
are handled when the last reference to the open file is released.

There are three cases to consider.

One is the simple case, a multithreaded (mt) process has a file open and
races to close it and acquire a lock on it.  In this case, the close will
release one reference to the open file and when the fcntl is done, it will
release the other reference.  For this situation, no locks should exist on
the file when both the close and fcntl operations are done.  The current
system will handle this case because the last reference to the open file is
being released.

The second case is when the mt process has dup(2)'d the file descriptor.
The close will release one reference to the file and the fcntl, when done,
will release another, but there will still be at least one more reference
to the open file.  One could argue that the existence of a lock on the file
after the close has completed is okay, because it was acquired after the
close operation and there is still a way for the application to release the
lock on the file, using an existing file descriptor.

The third case is when the mt process has forked, after opening the file
and either before or after becoming an mt process.  In this case, each
process would hold a reference to the open file.  For each process, this
degenerates to first case above.  However, the lock continues to exist
until both processes have released their references to the open file.  This
lock could block other lock requests.

The changes to release the lock when the last reference to the open file
aren't quite right because they would allow the lock to exist as long as
there was a reference to the open file.  This is too long.

The new proposed solution is to add support in the fcntl code path to
detect a race with close and then to release the lock which was just
acquired when such as race is detected.  This causes locks to be released
in a timely fashion and for the system to conform to the POSIX semantic
specification.

This was tested by instrumenting a kernel to detect the handling locks and
then running a program which generates case #3 above.  A dangling lock
could be reliably generated.  When the changes to detect the close/fcntl
race were added, a dangling lock could no longer be generated.

Cc: Matthew Wilcox <willy@debian.org>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:26:06 -07:00
Carsten Otte
0cfc11ed45 [PATCH] fix xip sparse file handling in ext2
Oliver Paukstadt from our test department is testing the xip patches in
Linus' git-tree.  He found a problem that shows when reading a file that
contains sparse blocks (holes) on a -o xip mounted ext2 filesystem: the
BUG_ON() in fs/ext2/xip.c:40 triggers where it should not.  The problem was
introduced by a cleanup in my previous patch, this patch fixes it.

Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:53 -07:00
Ian Kent
104e49fc1e [PATCH] autofs4: fix infamous "Busy inodes after umount ..." message
If the automount daemon receives a signal which causes it to sumarily
terminate the autofs4 module leaks dentries.  The same problem exists with
detached mount requests without the warning.

This patch cleans these dentries at umount.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:51 -07:00
Jan Kara
ab6862e6da [PATCH] ext3: drop quota references before releasing inode
We must drop references to quota structures before releasing the inode.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:50 -07:00
Jan Kara
c7e9a52ef0 [PATCH] ext2: drop quota reference before releasing inode
We must drop references to quota structures before releasing the inode.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:50 -07:00
Jeff Mahoney
b3bb8afd96 [PATCH] reiserfs: fix deadlock in inode creation failure path w/ default ACL
reiserfs_new_inode() can call iput() with the xattr lock held.  This will
cause a deadlock to occur when reiserfs_delete_xattrs() is called to clean
up.

The following patch releases the lock and reacquires it after the iput.
This is safe because interaction with xattrs is complete, and the relock is
just to balance out the release in the caller.

The locking needs some reworking to be more sane, but that's more intrusive
and I was just looking to fix this bug.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:50 -07:00
Nigel Cunningham
ef2a701d44 [PATCH] Fix missing refrigerator invocation in jffs2
Here's a patch to fix a missing refrigerator call in jffs2.

Signed-off-by: Nigel Cunningham <nigel@suspend2.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-27 16:25:49 -07:00
Andrew Morton
89373de7dd [PATCH] inotify: fix oops fix
Cc: Robert Love <rml@novell.com>
Cc: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 14:34:18 -07:00
Robert Love
e5ca844a9d [PATCH] inotify: check retval in init
Check for (unlikely) errors in the filesystem initialization stuff in
our module_init() function.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:37:22 -07:00
Robert Love
1b2ccf0cc1 [PATCH] inotify: change default limits
Change default inotify limits: Maximum instances per user to 128 and
maximum events per queue to 16k.  The max instances used to be 128; the
change to 8 was a mistake.  Memory consumption is fine.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:37:22 -07:00
Robert Love
5eb22cbcdb [PATCH] inotify: exit path cleanups
Handle error out paths better.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:37:22 -07:00
Robert Love
783bc29bbc [PATCH] inotify: oops fix
Bug fix: Ensure that the fd passed to inotify_add_watch() and
inotify_rm_watch() belongs to inotify.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:37:21 -07:00
Robert Love
33ea2f52b8 [PATCH] inotify: use fget_light
As an optimization, use fget_light() and fput_light() where possible.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:31:57 -07:00
Robert Love
b680716ed2 [PATCH] inotify: misc. cleanup
Miscellaneous invariant clean up, comment fixes, and so on.  Trivial
stuff.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-26 13:31:57 -07:00
Dave Kleikamp
18190cc08d JFS: Fix i_blocks accounting when allocation fails
A failure in dbAlloc caused a directory's i_blocks to be incorrectly
incremented, causing jfs_fsck to find the inode to be corrupt.

Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-26 09:29:13 -05:00
Dave Kleikamp
c2783f3a62 JFS: Don't set log_SYNCBARRIER when log->active == 0
If a metadata page is kept active, it is possible that the sync barrier logic
continues to trigger, even if all active transactions have been phyically
written to the journal.  This can cause a hang, since the completion of the
journal I/O is what unsets the sync barrier flag to allow new transactions
to be created.

Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-25 08:58:54 -05:00
Dave Kleikamp
c40c202493 JFS: Fix typo in last patch
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-22 11:08:44 -05:00
Dave Kleikamp
21d1ee8b37 Merge with /home/shaggy/git/linus-clean/
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-19 13:46:53 -05:00
Linus Torvalds
af6ea9ca23 Merge master.kernel.org:/pub/scm/linux/kernel/git/aia21/ntfs-2.6 2005-07-16 11:47:51 -07:00
Thomas Gleixner
2c4eec9802 Merge with rsync://fileserver/linux 2005-07-16 09:20:01 +02:00
Carsten Otte
afa597ba20 [PATCH] execute-in-place fixes
This patch includes feedback from Andrew and Christoph. Thanks for
taking time to review.

Use of empty_zero_page was eliminated to fix compilation for architectures
that don't have it.

This patch removes setting pages up-to-date in ext2_get_xip_page and all
bug checks to verify that the page is indeed up to date.  Setting the page
state on mapping to userland is bogus.  None of the code patchs involved
with these pages in mm cares about the page state.

still on my ToDo list: identify a place outside second extended where
__inode_direct_access should reside

Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-15 09:54:50 -07:00
Qu Fuping
3d9b1cdd24 JFS: fsync wrong behavior when I/O failure occurs
This is half of a patch that Qu Fuping submitted in April.  The first part
was applied to fs/mpage.c in 2.6.12-rc4.

jfs_fsync should return error, but it doesn't wait for the metadata page to
be uptodate, e.g.:
jfs_fsync->jfs_commit_inode->txCommit->diWrite->read_metapage->
__get_metapage->read_cache_page reads a page from disk. Because read is
async, when read_cache_page: err = filler(data, page), filler will not
return error, it just submits I/O request and returns. So, page is not
uptodate.  Checking only if(IS_ERROR(mp->page)) is not enough, we should
add "|| !PageUptodate(mp->page)"

Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-15 10:36:08 -05:00
Dave Kleikamp
56d1254917 JFS: Remove assert statement in dbJoin & return -EIO instead
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-15 09:43:36 -05:00
Thomas Gleixner
5d157885f3 [JFFS2] Fix node allocation leak
In the rare case of failing to write the cleanmarker
the allocated node was not freed.

Pointed out by Forrest Zhao
Initial cleanup by Joern Engel

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2005-07-15 08:14:44 +02:00
Dave Kleikamp
00be3e7e5c JFS: Remove bogus WARN_ON statement and some dead code
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
2005-07-14 15:15:39 -05:00
Paolo 'Blaisorblade' Giarrusso
a0d43df931 [PATCH] uml: hostfs: unuse ROOT_DEV
Minimal patch removing uses of ROOT_DEV; next patch unexports it.  I've
opposed this, but I've planned to reintroduce the functionality without using
ROOT_DEV.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-14 09:00:25 -07:00
Paolo 'Blaisorblade' Giarrusso
8e0a218124 [PATCH] uml: fix hppfs error path
Fix the error message to refer to the error code, i.e.  err, not count, plus
add some cosmetical fixes.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-14 09:00:25 -07:00
Anton Altaparmakov
c514720716 Automatic merge with /usr/src/ntfs-2.6.git. 2005-07-13 23:09:23 +01:00
Linus Torvalds
3720bd8b1e Merge master.kernel.org:/pub/scm/linux/kernel/git/tglx/mtd-2.6 2005-07-13 12:19:30 -07:00
Steve Dickson
7ee91ec14b [PATCH] NFS: procfs/sysctl interfaces for lockd do not work on x86_64
Allow the setting of NLM timeouts and grace periods through the proc and
sysclt interfaces on x86_64 architectures

Signed-off-by: Steve Dickson <steved@redhat.com>
Acked-by: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-13 11:25:24 -07:00
Anton Altaparmakov
88bd5121d6 [PATCH] Fix soft lockup due to NTFS: VFS part and explanation
Something has changed in the core kernel such that we now get concurrent
inode write outs, one e.g via pdflush and one via sys_sync or whatever.
This causes a nasty deadlock in ntfs.  The only clean solution
unfortunately requires a minor vfs api extension.

First the deadlock analysis:

Prerequisive knowledge: NTFS has a file $MFT (inode 0) loaded at mount
time.  The NTFS driver uses the page cache for storing the file contents as
usual.  More interestingly this file contains the table of on-disk inodes
as a sequence of MFT_RECORDs.  Thus NTFS driver accesses the on-disk inodes
by accessing the MFT_RECORDs in the page cache pages of the loaded inode
$MFT.

The situation: VFS inode X on a mounted ntfs volume is dirty.  For same
inode X, the ntfs_inode is dirty and thus corresponding on-disk inode,
which is as explained above in a dirty PAGE_CACHE_PAGE belonging to the
table of inodes ($MFT, inode 0).

What happens:

Process 1: sys_sync()/umount()/whatever...  calls __sync_single_inode() for
$MFT -> do_writepages() -> write_page for the dirty page containing the
on-disk inode X, the page is now locked -> ntfs_write_mst_block() which
clears PageUptodate() on the page to prevent anyone else getting hold of it
whilst it does the write out (this is necessary as the on-disk inode needs
"fixups" applied before the write to disk which are removed again after the
write and PageUptodate is then set again).  It then analyses the page
looking for dirty on-disk inodes and when it finds one it calls
ntfs_may_write_mft_record() to see if it is safe to write this on-disk
inode.  This then calls ilookup5() to check if the corresponding VFS inode
is in icache().  This in turn calls ifind() which waits on the inode lock
via wait_on_inode whilst holding the global inode_lock.

Process 2: pdflush results in a call to __sync_single_inode for the same
VFS inode X on the ntfs volume.  This locks the inode (I_LOCK) then calls
write-inode -> ntfs_write_inode -> map_mft_record() -> read_cache_page() of
the page (in page cache of table of inodes $MFT, inode 0) containing the
on-disk inode.  This page has PageUptodate() clear because of Process 1
(see above) so read_cache_page() blocks when tries to take the page lock
for the page so it can call ntfs_read_page().

Thus Process 1 is holding the page lock on the page containing the on-disk
inode X and it is waiting on the inode X to be unlocked in ifind() so it
can write the page out and then unlock the page.

And Process 2 is holding the inode lock on inode X and is waiting for the
page to be unlocked so it can call ntfs_readpage() or discover that
Process 1 set PageUptodate() again and use the page.

Thus we have a deadlock due to ifind() waiting on the inode lock.

The only sensible solution: NTFS does not care whether the VFS inode is
locked or not when it calls ilookup5() (it doesn't use the VFS inode at
all, it just uses it to find the corresponding ntfs_inode which is of
course attached to the VFS inode (both are one single struct); and it uses
the ntfs_inode which is subject to its own locking so I_LOCK is irrelevant)
hence we want a modified ilookup5_nowait() which is the same as ilookup5()
but it does not wait on the inode lock.

Without such functionality I would have to keep my own ntfs_inode cache in
the NTFS driver just so I can find ntfs_inodes independent of their VFS
inodes which would be slow, memory and cpu cycle wasting, and incredibly
stupid given the icache already exists in the VFS.

Below is a patch that does the ilookup5_nowait() implementation in
fs/inode.c and exports it.

ilookup5_nowait.diff:

Introduce ilookup5_nowait() which is basically the same as ilookup5() but
it does not wait on the inode's lock (i.e. it omits the wait_on_inode()
done in ifind()).

This is needed to avoid a nasty deadlock in NTFS.

Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-07-13 11:25:24 -07:00