A file that is open for O_DIRECT is by definition not obeying
close-to-open cache consistency semantics, so let's not cache
the attributes too aggressively either.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Clean up...
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If we need to update the cached attributes, then we'd better make
sure that we also layoutcommit first. Otherwise, the server may have stale
attributes.

Prior to this patch, the revalidation code tried to "fix" this problem by
simply disabling attributes that would be affected by the layoutcommit.
That approach breaks nfs_writeback_check_extend(), leading to a file size
corruption.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Unless the user is using file locking, we must assume close-to-open
cache consistency when the file is open for writing. Adjust the
caching algorithm so that it does not clear the cache on out-of-order
writes and/or attribute revalidations.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If there were outstanding writes then chalk up the unexpected change
attribute on the server to them.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

use d_alloc_parallel() for sillyunlink/lookup exclusion and
explicit rwsem (nfs_rmdir() being a writer and nfs_call_unlink() -
a reader) for rmdir/sillyunlink one.

That ought to make lookup/readdir/!O_CREAT atomic_open really
parallel on NFS.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

NFS may be used as lower layer of overlayfs and accessing f_path.dentry can
lead to a crash.

Fix by replacing direct access of file->f_path.dentry with the
file_dentry() accessor, which will always return a native object.

Fixes: 4bacc9c9 ("overlayfs: Make f_path always point to the overlay and f_inode to the underlay")
Signed-off-by: NMiklos Szeredi <mszeredi@redhat.com>
Tested-by: NGoldwyn Rodrigues <rgoldwyn@suse.com>
Acked-by: NTrond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>
Cc: <stable@vger.kernel.org> # v4.2
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>

Just call inode_dio_wait directly instead of through a pointless wrapper.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

parallel to mutex_{lock,unlock,trylock,is_locked,lock_nested},
inode_foo(inode) being mutex_foo(&inode->i_mutex).

Please, use those for access to ->i_mutex; over the coming cycle
->i_mutex will become rwsem, with ->lookup() done with it held
only shared.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Mark those kmem allocations that are known to be easily triggered from
userspace as __GFP_ACCOUNT/SLAB_ACCOUNT, which makes them accounted to
memcg.  For the list, see below:

 - threadinfo
 - task_struct
 - task_delay_info
 - pid
 - cred
 - mm_struct
 - vm_area_struct and vm_region (nommu)
 - anon_vma and anon_vma_chain
 - signal_struct
 - sighand_struct
 - fs_struct
 - files_struct
 - fdtable and fdtable->full_fds_bits
 - dentry and external_name
 - inode for all filesystems. This is the most tedious part, because
   most filesystems overwrite the alloc_inode method.

The list is far from complete, so feel free to add more objects.
Nevertheless, it should be close to "account everything" approach and
keep most workloads within bounds.  Malevolent users will be able to
breach the limit, but this was possible even with the former "account
everything" approach (simply because it did not account everything in
fact).

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: NVladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NMichal Hocko <mhocko@suse.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The use of wait_on_atomic_t() for waiting on I/O to complete before
unlocking allows us to git rid of the NFS_IO_INPROGRESS flag, and thus the
nfs_iocounter's flags member, and finally the nfs_iocounter altogether.
The count of I/O is moved to the lock context, and the counter
increment/decrement functions become simple enough to open-code.
Signed-off-by: NBenjamin Coddington <bcodding@redhat.com>
[Trond: Fix up conflict with existing function nfs_wait_atomic_killable()]
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If a NFSv4 client uses the cache_consistency_bitmask in order to
request only information about the change attribute, timestamps and
size, then it has not revalidated all attributes, and hence the
attribute timeout timestamp should not be updated.
Reported-by: NDonald Buczek <buczek@molgen.mpg.de>
Cc: stable@vger.kernel.org
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

When we fail to queue a read page to IO descriptor,
we need to clean it up otherwise it is hanging around
preventing nfs module from being removed.

When we fail to queue a write page to IO descriptor,
we need to clean it up and also save the failure status
to open context. Then at file close, we can try to write
pages back again and drop the page if it fails to writeback
in .launder_page, which will be done in the next patch.
Signed-off-by: NPeng Tao <tao.peng@primarydata.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Jan Stancek reported that I wrecked things for him by fixing things for
Vladimir :/

His report was due to an UNINTERRUPTIBLE wait getting -EINTR, which
should not be possible, however my previous patch made this possible by
unconditionally checking signal_pending().

We cannot use current->state as was done previously, because the
instruction after the store to that variable it can be changed.  We must
instead pass the initial state along and use that.

Fixes: 68985633 ("sched/wait: Fix signal handling in bit wait helpers")
Reported-by: NJan Stancek <jstancek@redhat.com>
Reported-by: NChris Mason <clm@fb.com>
Tested-by: NJan Stancek <jstancek@redhat.com>
Tested-by: NVladimir Murzin <vladimir.murzin@arm.com>
Tested-by: NChris Mason <clm@fb.com>
Reviewed-by: NPaul Turner <pjt@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: tglx@linutronix.de
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: hpa@zytor.com
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

based upon the corresponding patch from Neil's March patchset,
again with kmap-related horrors removed.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

kmap() in page_follow_link_light() needed to go - allowing to hold
an arbitrary number of kmaps for long is a great way to deadlocking
the system.

new helper (inode_nohighmem(inode)) needs to be used for pagecache
symlinks inodes; done for all in-tree cases.  page_follow_link_light()
instrumented to yell about anything missed.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

If we pass in an empty nfs_fattr struct to nfs_update_inode, it will
(correctly) not update any of the attributes, but it then clears the
NFS_INO_INVALID_ATTR flag, which indicates that the attributes are
up to date. Don't clear the flag if the fattr struct has no valid
attrs to apply.
Reviewed-by: NSteve French <steve.french@primarydata.com>
Signed-off-by: NJeff Layton <jeff.layton@primarydata.com>
Cc: stable@vger.kernel.org
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If we get no post-op attributes back from a SETATTR operation, then no
attributes will of course be updated during the call to
nfs_update_inode.

We know however that the attributes are invalid at that point, since we
just changed some of them. At the very least, the ctime will be bogus.
If we get no post-op attributes back on the call, mark the attrcache
invalid to reflect that fact.
Reviewed-by: NSteve French <steve.french@primarydata.com>
Signed-off-by: NJeff Layton <jeff.layton@primarydata.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

This reverts commit f895c53f.

This commit causes a NFSv4 regression in that close()+unlink() can end
up failing. The reason is that we no longer have a guarantee that the
CLOSE has completed on the server, meaning that the subsequent call to
REMOVE may fail with NFS4ERR_FILE_OPEN if the server implements Windows
unlink() semantics.
Reported-by: N&lt;Olga Kornievskaia <aglo@umich.edu>
Cc: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If there is no cached data, then there is no need to track the file
change attribute on close.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Set rlimit for NFS's files is useless right now.
For local process's rlimit, it should be checked by nfs client.

The same, CIFS also call inode_change_ok checking rlimit at its client
in cifs_setattr_nounix() and cifs_setattr_unix().

v3, fix bad using of error
Signed-off-by: NKinglong Mee <kinglongmee@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If the ctime or mtime or change attribute have changed because
of an operation we initiated, we should make sure that we force
an attribute update. However we do not want to mark the page cache
for revalidation.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>
Cc: stable@vger.kernel.org # v4.0+

Chuck reports seeing cases where a GETATTR that happens to race
with an asynchronous WRITE is overriding the file size, despite
the attribute barrier being set by the writeback code.

The culprit turns out to be the check in nfs_ctime_need_update(),
which sees that the ctime is newer than the cached ctime, and
assumes that it is safe to override the attribute barrier.
This patch removes that override, and ensures that attribute
barriers are always respected.
Reported-by: NChuck Lever <chuck.lever@oracle.com>
Fixes: a08a8cd3 ("NFS: Add attribute update barriers to NFS writebacks")
Cc: stable@vger.kernel.org # v4.0+
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

And call nfs_file_clear_open_context() directly.  This makes it obvious
that nfs_file_release() will always return 0.
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Setting the change attribute has been mandatory for all NFS versions, since
commit 3a1556e8 ("NFSv2/v3: Simulate the change attribute"). We should
therefore not have anything be conditional on it being set/unset.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

We can't allow caching of data until the change attribute has been
initialised correctly.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

If we've ensured that the size and the change attribute are both correct,
then there is no point in marking those attributes as needing revalidation
again. Only do so if we know the size is incorrect and was not updated.

Fixes: f2467b6f ("NFS: Clear NFS_INO_REVAL_PAGECACHE when...")
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Have checking CONFIG_PROC_FS in include/linux/sunrpc/stats.h.
Signed-off-by: NKinglong Mee <kinglongmee@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

In glibc 2.21 (and several previous), a call to opendir() will
result in a 32K (BUFSIZ*4) buffer being allocated and passed to
getdents.

However a call to fdopendir() results in an 'fstat' request to
determine block size and a matching buffer allocated for subsequent
use with getdents.  This will typically be 1M.

The first getdents call on an NFS directory will always use
READDIR_PLUS (or NFSv4 equivalent) if available.  Subsequent getdents
calls only use this more expensive version if some 'stat' requests are
made between the getdents calls.

For this reason it is good to keep at least that first getdents call
relatively short.  When fdopendir() and readdir() is used on a large
directory, it takes approximately 32 times as long to complete as
using "opendir".  Current versions of 'find' use fdopendir() and
demonstrate this slowness.

'stat' on a directory currently returns the 'wsize'.  This number has
no meaning on directories.
Actual READDIR requests are limited to ->dtsize, which itself is
capped at 4 pages, coincidently the same as BUFSIZ*4.
So this is a meaningful number to use as the blocksize on directories,
and has the effect of making 'find' on large directories go a lot
faster.
Signed-off-by: NNeilBrown <neilb@suse.de>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Don't unnecessarily cast allocation return value in
fs/nfs/inode.c::nfs_alloc_inode().
Signed-off-by: NFiro Yang <firogm@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

2ef47eb1 (NFS: Fix use of nfs_attr_use_mounted_on_fileid()) was a good
start to fixing a circular directory structure warning for NFS v4
"junctioned" mountpoints.  Unfortunately, further testing continued to
generate this error.

My server is configured like this:

anna@nfsd ~ % df
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1       9.1G  2.0G  6.5G  24% /
/dev/vdc1      1014M   33M  982M   4% /exports
/dev/vdc2      1014M   33M  982M   4% /exports/vol1
/dev/vdc3      1014M   33M  982M   4% /exports/vol1/vol2

anna@nfsd ~ % cat /etc/exports
/exports/          *(rw,async,no_subtree_check,no_root_squash)
/exports/vol1/     *(rw,async,no_subtree_check,no_root_squash)
/exports/vol1/vol2 *(rw,async,no_subtree_check,no_root_squash)

I've been running chown across the entire mountpoint twice in a row to
hit this problem.  The first run succeeds, but the second one fails with
the circular directory warning along with:

anna@client ~ % dmesg
[Apr 3 14:28] NFS: server 192.168.100.204 error: fileid changed
              fsid 0:39: expected fileid 0x100080, got 0x80

WHere 0x80 is the mountpoint's fileid and 0x100080 is the mounted-on
fileid.

This patch fixes the issue by requesting an updated mounted-on fileid
from the server during nfs_update_inode(), and then checking that the
fileid stored in the nfs_inode matches either the fileid or mounted-on
fileid returned by the server.
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

This patch adds a GETATTR to the end of ALLOCATE and DEALLOCATE
operations so we can set the updated inode size and change attribute
directly.  DEALLOCATE will still need to release pagecache pages, so
nfs42_proc_deallocate() now calls truncate_pagecache_range() before
contacting the server.
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

that's the bulk of filesystem drivers dealing with inodes of their own
Signed-off-by: NDavid Howells <dhowells@redhat.com>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

We don't just want to sync out buffered writes, but also O_DIRECT ones.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Then apply it to nfs_setattr() and nfs_getattr().
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

nfs_vm_page_mkwrite() should wait until the page cache invalidation
is finished. This is the second patch in a 2 patch series to deprecate
the NFS client's reliance on nfs_release_page() in the context of
nfs_invalidate_mapping().
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

When invalidating the page cache for a regular file, we want to first
sync all dirty data to disk and then call invalidate_inode_pages2().
The latter relies on nfs_launder_page() and nfs_release_page() to deal
respectively with dirty pages, and unstable written pages.

When commit 95905446 ("NFS: avoid deadlocks with loop-back mounted
NFS filesystems.") changed the behaviour of nfs_release_page(), then it
made it possible for invalidate_inode_pages2() to fail with an EBUSY.
Unfortunately, that error is then propagated back to read().

Let's therefore work around the problem for now by protecting the call
to sync the data and invalidate_inode_pages2() so that they are atomic
w.r.t. the addition of new writes.
Later on, we can revisit whether or not we still need nfs_launder_page()
and nfs_release_page().
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>

Ensure that we don't regress the changes that were made to the
directory.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>
Tested-by: NChuck Lever <chuck.lever@oracle.com>

nfs_post_op_update_inode() is called after a self-induced attribute
update. Ensure that it also sets the barrier.
Signed-off-by: NTrond Myklebust <trond.myklebust@primarydata.com>
Tested-by: NChuck Lever <chuck.lever@oracle.com>