Make NFSv4 return the fully initialized file pointer with the
stateid that it created in the lookup w/intent.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This is needed by NFSv4 for atomicity reasons: our open command is in
fact a lookup+open, so we need to be able to propagate open context
information from lookup() into the resulting struct file's
private_data field.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We no longer need to worry about collisions between close() and the state
recovery code, since the new close will automatically recheck the
file state once it is done waiting on its sequence slot.
Ditto for the nfs4_proc_locku() procedure.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Once the state_owner and lock_owner semaphores get removed, it will be
possible for other OPEN requests to reopen the same file if they have
lower sequence ids than our CLOSE call.
This patch ensures that we recheck the file state once
nfs_wait_on_sequence() has completed waiting.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
NFSv4 file state-changing functions such as OPEN, CLOSE, LOCK,... are all
labelled with "sequence identifiers" in order to prevent the server from
reordering RPC requests, as this could cause its file state to
become out of sync with the client.
Currently the NFS client code enforces this ordering locally using
semaphores to restrict access to structures until the RPC call is done.
This, of course, only works with synchronous RPC calls, since the
user process must first grab the semaphore.
By dropping semaphores, and instead teaching the RPC engine to hold
the RPC calls until they are ready to be sent, we can extend this
process to work nicely with asynchronous RPC calls too.
This patch adds a new list called "rpc_sequence" that defines the order
of the RPC calls to be sent. We add one such list for each state_owner.
When an RPC call is ready to be sent, it checks if it is top of the
rpc_sequence list. If so, it proceeds. If not, it goes back to sleep,
and loops until it hits top of the list.
Once the RPC call has completed, it can then bump the sequence id counter,
and remove itself from the rpc_sequence list, and then wake up the next
sleeper.
Note that the state_owner sequence ids and lock_owner sequence ids are
all indexed to the same rpc_sequence list, so OPEN, LOCK,... requests
are all ordered w.r.t. each other.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
lock_kiocb() was introduced to serialize retrying and cancellation. In the
process of doing so it tried to sleep waiting for KIF_LOCKED while holding
the ctx_lock spinlock. Recent fixes have ensured that multiple concurrent
retries won't be attempted for a given iocb. Cancel has other problems and
has no significant in-tree users that have been complaining about it. So
for the immediate future we'll revert sleeping with the lock held and will
address proper cancellation and retry serialization in the future.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Acked-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Currently you do not get all the map entries on nommu systems because the
start function doesn't index into the list using the value of "pos".
Signed-off-by: David McCullough <davidm@snapgear.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Oopsable since nfs_wait_on_inode() can get called as part of iput_final().
Unnecessary since the caller had better be damned sure that the inode won't
disappear from underneath it anyway.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If the data cache has been marked as potentially invalid by nfs_refresh_inode,
we should invalidate it rather than assume that changes are due to our own
activity.
Also ensure that we always start with a valid cache before declaring it
to be protected by a delegation.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
"proc_smaps_operations" is not defined in case of "CONFIG_MMU=n".
Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Nir Tzachar <tzachar@cs.bgu.ac.il> points out that if an ELF file specifies a
zero-length bss at a whacky address, we cannot load that binary because
padzero() tries to zero out the end of the page at the whacky address, and
that may not be writeable.
See also http://bugzilla.kernel.org/show_bug.cgi?id=5411
So teach load_elf_binary() to skip the bss settng altogether if the elf file
has a zero-length bss segment.
Cc: Roland McGrath <roland@redhat.com>
Cc: Daniel Jacobowitz <dan@debian.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Here is a compatibility fix between Linux and Solaris when used with VxFS
filesystems: Solaris usually accepts acl entries in any order, but with
VxFS it replies with NFSERR_INVAL when it sees a four-entry acl that is not
in canonical form. It may also fail with other non-canonical acls -- I
can't tell, because that case never triggers: We only send non-canonical
acls when we fake up an ACL_MASK entry.
Instead of adding fake ACL_MASK entries at the end, inserting them in the
correct position makes Solaris+VxFS happy. The Linux client and server
sides don't care about entry order. The three-entry-acl special case in
which we need a fake ACL_MASK entry was handled in xdr_nfsace_encode. The
patch moves this into nfsacl_encode.
Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Acked-by: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
v9fs_file_read and v9fs_file_write use kmalloc to allocate buffers as big
as the data buffer received as parameter. kmalloc cannot be used to
allocate buffers bigger than 128K, so reading/writing data in chunks bigger
than 128k fails.
This patch reorganizes v9fs_file_read and v9fs_file_write to allocate only
buffers as big as the maximum data that can be sent in one 9P message.
Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
file operations ->write(), ->aio_write(), and ->writev() for regular
files. This replaces the old use of generic_file_write(), et al and
the address space operations ->prepare_write and ->commit_write.
This means that both sparse and non-sparse (unencrypted and
uncompressed) files can now be extended using the normal write(2)
code path. There are two limitations at present and these are that
we never create sparse files and that we only have limited support
for highly fragmented files, i.e. ones whose data attribute is split
across multiple extents. When such a case is encountered,
EOPNOTSUPP is returned.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
and cond_resched() in the main loop as we could be dirtying a lot of
pages and this ensures we play nice with the VM and the system as a
whole.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
The third param in this call to vmap shouldn't be GFP_KERNEL, which
makes no sense, but rather VM_MAP. Thanks to Al Viro for spotting
this.
Signed-off-by: Tom Zanussi <zanussi@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- added typedef unsigned int __nocast gfp_t;
- replaced __nocast uses for gfp flags with gfp_t - it gives exactly
the same warnings as far as sparse is concerned, doesn't change
generated code (from gcc point of view we replaced unsigned int with
typedef) and documents what's going on far better.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The nameidata "last.name" is always allocated with "__getname()", and
should always be free'd with "__putname()".
Using "putname()" without the underscores will leak memory, because the
allocation will have been hidden from the AUDITSYSCALL code.
Arguably the real bug is that the AUDITSYSCALL code is really broken,
but in the meantime this fixes the problem people see.
Reported by Robert Derr, patch by Rick Lindsley.
Acked-by: Al Viro <viro@ftp.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
bfs_fill_super() walks the inode table to get the bitmap of free inodes
and collect stats. It has no business using iget() for that - it's a
lot of extra work, extra icache pollution and more complex code.
Switched to walking the damn thing directly.
Note: that also allows to kill ->i_dsk_ino in there - separate patch if
Tigran can confirm that this field can be zero only for deleted inodes
(i.e. something that could only be found during that scan and not by
normal lookups).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
the initial implementation of file truncation. Now both open(2)ing
a file with the O_TRUNC flag and the {,f}truncate(2) system calls
will resize a file appropriately. The limitations are that only
uncompressed and unencrypted files are supported. Also, there is
only very limited support for highly fragmented files (the ones whose
$DATA attribute is split into multiple attribute extents).
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
extend the allocation of an attributes. Optionally, the data size,
but not the initialized size can be extended, too.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
which is zero for a resident attribute but should no longer be zero
once the attribute is non-resident as it then has real clusters
allocated.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
as an extra parameter. This is needed since we need to know the size
before we can map the mft record and our callers always know it. The
reason we cannot simply read the size from the vfs inode i_size is
that this is not necessarily uptodate. This happens when
ntfs_attr_make_non_resident() is called in the ->truncate call path.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
specifying whether the cluster are being allocated to extend an
attribute or to fill a hole.
- Change ntfs_attr_make_non_resident() to call ntfs_cluster_alloc()
with @is_extension set to TRUE and remove the runlist terminator
fixup code as this is now done by ntfs_cluster_alloc().
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
search context as argument. This allows calling it with the mft
record mapped. Update all callers.
- Fix potential deadlock in ntfs_mft_data_extend_allocation_nolock()
error handling by passing in the active search context when calling
ntfs_cluster_free().
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
search context as argument. This allows calling it with the mft
record mapped. Update all callers.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
This patch fixes up a few problems with jfs's reserved inodes.
1. There is no need for the jfs code setting the I_DIRTY bits in i_state.
I am ashamed that the code ever did this, and surprised it hasn't been
noticed until now.
2. Make sure special inodes are on an inode hash list. If the inodes are
unhashed, __mark_inode_dirty will fail to put the inode on the
superblock's dirty list, and the data will not be flushed under memory
pressure.
3. Force writing journal data to disk when metapage_writepage is unable to
write a metadata page due to pending journal I/O.
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
Check O_DIRECT and return -EINVAL error in open. dentry_open() also checks
this but only after the open method is called. This patch optimizes away
the unnecessary upcalls in this case.
It could be a correctness issue too: if filesystem has open() with side
effect, then it should fail before doing the open, not after.
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Calling truncate() on hostfs spits a kernel warning "Something isn't
implemented here", but it still works fine.
Indeed, hostfs i_op->truncate doesn't do anything. But hostfs_setattr() ->
set_attr() correctly detects ATTR_SIZE and calls truncate() on the host. So
we should be safe (using ftruncate() may be better, in case the file is
unlinked on the host, but we aren't sure to have the file open for writing,
and reopening it would cause the same races; plus nobody should expect UML to
be so careful).
So, the warning is wrong, because the current implementation is working. Al,
am I correct, and can the warning be therefore dropped?
CC: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Recently aio_p{read,write} changed to perform retries internally rather
than returning -EIOCBRETRY. This inadvertantly resulted in always calling
aio_{read,write} with ki_left at 0 which would in turn immediately return
0. Harmless, but we can avoid this call by checking in the caller.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Benjamin LaHaise <bcrl@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Only one of the run or kick path is supposed to put an iocb on the run
list. If both of them do it than one of them can end up referencing a
freed iocb. The kick path could delete the task_list item from the wait
queue before getting the ctx_lock and putting the iocb on the run list.
The run path was testing the task_list item outside the lock so that it
could catch ki_retry methods that return -EIOCBRETRY *without* putting the
iocb on a wait queue and promising to call kick_iocb. This unlocked check
could then race with the kick path to cause both to try and put the iocb on
the run list.
The patch stops the run path from testing task_list by requring that any
ki_retry that returns -EIOCBRETRY *must* guarantee that kick_iocb() will be
called in the future. aio_p{read,write}, the only in-tree -EIOCBRETRY
users, are updated.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Benjamin LaHaise <bcrl@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Only one of the run or kick path is supposed to put an iocb on the run
list. If both of them do it than one of them can end up referencing a
freed iocb. The kick patch could set the Kicked bit before acquiring the
ctx_lock and putting the iocb on the run list. The run path, while holding
the ctx_lock, could see this partial kick and mistake it for a kick that
was deferred while it was doing work with the run_list NULLed out. It
would then race with the kick thread to add the iocb to the run list.
This patch moves the kick setting under the ctx_lock so that only one of
the kick or run path queues the iocb on the run list, as intended.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Benjamin LaHaise <bcrl@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
it seems that readv(2)/writev(2) syscalls do not call
file_permission callback. Looks like this is overlook.
I have filled the issue into redhat bugzilla as
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=169433
and got the recommendation to post this on lsm mailing list.
The following trivial patch solves the problem.
Signed-off-by: Kostik Belousov <kostikbel@gmail.com>
Signed-off-by: Chris Wright <chrisw@osdl.org>
