Merge remote-tracking branch 'hid/for-next'

Merge remote-tracking branch 'pci/next'

Merge remote-tracking branch 'vfs/for-next'

Conflicts:
fs/f2fs/file.c

Merge remote-tracking branch 'file-private-locks/linux-next'

Merge remote-tracking branch 'xfs/for-next'

Merge remote-tracking branch 'nfsd/nfsd-next'

Merge remote-tracking branch 'nfs/linux-next'

Merge remote-tracking branch 'logfs/master'

Merge remote-tracking branch 'jfs/jfs-next'

Merge remote-tracking branch 'gfs2/master'

Merge remote-tracking branch 'fscache/fscache'

Merge remote-tracking branch 'f2fs/dev'

Merge remote-tracking branch 'ext4/dev'

Merge remote-tracking branch 'ecryptfs/next'

Merge remote-tracking branch 'cifs/for-next'

Merge remote-tracking branch 'uml/next'

Conflicts:
arch/um/include/asm/Kbuild

Merge remote-tracking branch 'tile/master'

Merge remote-tracking branch 'sparc-next/master'

Merge remote-tracking branch 's390/features'

Merge remote-tracking branch 'powerpc/next'

Merge remote-tracking branch 'mips/mips-for-linux-next'

Merge remote-tracking branch 'm68knommu/for-next'

Merge remote-tracking branch 'tegra/for-next'

Merge remote-tracking branch 'samsung/for-next'

Merge remote-tracking branch 'renesas/next'

Merge remote-tracking branch 'mvebu/for-next'

Conflicts:
arch/arm/mach-mvebu/Kconfig

Merge remote-tracking branch 'keystone/next'

Merge remote-tracking branch 'imx-mxs/for-next'

Merge remote-tracking branch 'ep93xx/ep93xx-for-next'

Merge remote-tracking branch 'berlin/berlin/for-next'

Merge remote-tracking branch 'arm-soc/for-next'

Merge remote-tracking branch 'arm-perf/for-next/perf'

Merge remote-tracking branch 'arm-kvm-cpuresume/arm-kvm-cpuresume'

Merge remote-tracking branch 'arm/for-next'

Merge remote-tracking branch 'arc/for-next'

Merge remote-tracking branch 'drm-intel-fixes/for-linux-next-fixes'

Merge remote-tracking branch 'ide/master'

Merge remote-tracking branch 'crypto-current/master'

Merge remote-tracking branch 'input-current/for-linus'

Merge remote-tracking branch 'staging.current/staging-linus'

Merge remote-tracking branch 'wireless/master'

Merge remote-tracking branch 'sound-current/for-linus'

Merge remote-tracking branch 'ipsec/master'

Conflicts:
net/ipv4/ip_vti.c

NFSd: Mark nfs4_free_lockowner and nfs4_free_openowner as static functions

They do not need to be used outside fs/nfsd/nfs4state.c

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd: remove <linux/nfsd/debug.h>

There is almost nothing left it in, just merge it into the only file
that includes it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd: move <linux/nfsd/stats.h> to fs/nfsd

There are no legitimate users outside of fs/nfsd, so move it there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd: move <linux/nfsd/export.h> to fs/nfsd

There are no legitimate users outside of fs/nfsd, so move it there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

nfsd: remove <linux/nfsd/nfsfh.h>

The only real user of this header is fs/nfsd/nfsfh.h, so merge the
two.  Various lockѕ source files used it to indirectly get other
sunrpc or nfs headers, so fix those up.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

NFSd: Remove 'inline' designation for free_client()

It is large, it is used in more than one place, and it is not performance
critical. Let gcc figure out whether it should be inlined...

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

lockd: avoid warning when CONFIG_SYSCTL undefined

When building without CONFIG_SYSCTL, the compiler saw an unused
label. This moves the label into the #ifdef it is used under.

fs/lockd/svc.c: In function ‘init_nlm’:
fs/lockd/svc.c:626:1: warning: label ‘err_sysctl’ defined but not used [-Wunused-label]

Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

kill generic_file_splice_write()

no callers left

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ceph: switch to iter_file_splice_write()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

shmem: switch to iter_file_splice_write()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

nfs: switch to iter_splice_write_file()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fs/splice.c: remove unneeded exports

ocfs2 was using a bunch of splice.c guts...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ocfs2: switch to iter_file_splice_write()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

->splice_write() via ->write_iter()

iter_file_splice_write() - a ->splice_write() instance that gathers the
pipe buffers, builds a bio_vec-based iov_iter covering those and feeds
it to ->write_iter().  A bunch of simple cases coverted to that...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

bio_vec-backed iov_iter

New variant of iov_iter - ITER_BVEC in iter->type, backed with
bio_vec array instead of iovec one.  Primitives taught to deal
with such beasts, __swap_write() switched to using that kind
of iov_iter.

Note that bio_vec is just a <page, offset, length> triple - there's
nothing block-specific about it.  I've left the definition where it
was, but took it from under ifdef CONFIG_BLOCK.

Next target: ->splice_write()...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

optimize copy_page_{to,from}_iter()

if we'd ended up in the end of a segment, jump to the
beginning of the next one (iov_offset = 0, iov++),
rather than having the next primitive deal with that.

Ought to be folded back...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

bury generic_file_aio_{read,write}

no callers left

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

lustre: get rid of messing with iovecs

* switch to ->read_iter/->write_iter
* keep a pointer to iov_iter instead of iov/nr_segs
* do not modify iovecs; use iov_iter_truncate()/iov_iter_advance() and
a new primitive - iov_iter_reexpand() (expand previously truncated
iterator) istead.
* (racy) check for lustre VMAs intersecting with iovecs kept for now as
for_each_iov() loop.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ceph: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ceph_sync_direct_write: stop poking into iov_iter guts

all needed primitives are there...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ceph_sync_read: stop poking into iov_iter guts

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

new helper: copy_page_from_iter()

parallel to copy_page_to_iter().  pipe_write() switched to it (and became
->write_iter()).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fuse: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

btrfs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ocfs2: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

xfs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

afs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

gfs2: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

nfs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ubifs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

bury __generic_file_aio_write()

all users converted to __generic_file_write_iter() now

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

cifs: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

udf: switch to ->write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

convert ext4 to ->write_iter()

unfortunately, Ted's changes to ext4_file_write() are *still* an
incomplete fix - playing with rlimits can let you smuggle an
unaligned request past the checks.  So there almost certainly
will be more merge PITA around that place...

[fix from Peter Ujfalusi <peter.ujfalusi@ti.com> folded]

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Merge ext4 changes in ext4_file_write() into for-next

From ext4.git#dev, needed for switch of ext4 to ->write_iter() ;-/

blkdev_aio_write() - turn into blkdev_write_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

write_iter variants of {__,}generic_file_aio_write()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ceph: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

nfs: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fs/block_dev.c: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

shmem: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

pipe: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

cifs: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fuse_file_aio_read(): convert to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ocfs2: switch to ->read_iter()

tracepoints are evil, exhibit #6969...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ecryptfs: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

xfs: switch to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

switch simple generic_file_aio_read() users to ->read_iter()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

new methods: ->read_iter() and ->write_iter()

Beginning to introduce those.  Just the callers for now, and it's
clumsier than it'll eventually become; once we finish converting
aio_read and aio_write instances, the things will get nicer.

For now, these guys are in parallel to ->aio_read() and ->aio_write();
they take iocb and iov_iter, with everything in iov_iter already
validated.  File offset is passed in iocb->ki_pos, iov/nr_segs -
in iov_iter.

Main concerns in that series are stack footprint and ability to
split the damn thing cleanly.

[fix from Peter Ujfalusi <peter.ujfalusi@ti.com> folded]

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

replace checking for ->read/->aio_read presence with check in ->f_mode

Since we are about to introduce new methods (read_iter/write_iter), the
tests in a bunch of places would have to grow inconveniently.  Check
once (at open() time) and store results in ->f_mode as FMODE_CAN_READ
and FMODE_CAN_WRITE resp.  It might end up being a temporary measure -
once everything switches from ->aio_{read,write} to ->{read,write}_iter
it might make sense to return to open-coded checks.  We'll see...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

xfs: trim the argument lists of xfs_file_{dio,buffered}_aio_write()

pos is redundant (it's iocb->ki_pos), and iov/nr_segs/count are taken
care of by lifting iov_iter into the caller.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

blkdev_aio_read(): switch to generic_file_read_iter(), get rid of iov_shorten()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

iov_iter_truncate()

Now It Can Be Done(tm) - we don't need to do iov_shorten() in
generic_file_direct_write() anymore, now that all ->direct_IO()
instances are converted to proper iov_iter methods and honour
iter->count and iter->iov_offset properly.

Get rid of count/ocount arguments of generic_file_direct_write(),
while we are at it.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

btrfs: switch check_direct_IO() to iov_iter

... and don't open-code iov_iter_alignment() there

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

new helper: iov_iter_get_pages_alloc()

same as iov_iter_get_pages(), except that pages array is allocated
(kmalloc if possible, vmalloc if that fails) and left for caller to
free.  Lustre and NFS ->direct_IO() switched to it.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

new helper: iov_iter_npages()

counts the pages covered by iov_iter, up to given limit.
do_block_direct_io() and fuse_iter_npages() switched to
it.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

f2fs: switch to iov_iter_alignment()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>